diff options
Diffstat (limited to 'src/backend/optimizer')
59 files changed, 85072 insertions, 0 deletions
diff --git a/src/backend/optimizer/Makefile b/src/backend/optimizer/Makefile new file mode 100644 index 0000000..f523e5e --- /dev/null +++ b/src/backend/optimizer/Makefile @@ -0,0 +1,13 @@ +# +# Makefile for optimizer +# +# src/backend/optimizer/Makefile +# + +subdir = src/backend/optimizer +top_builddir = ../../.. +include $(top_builddir)/src/Makefile.global + +SUBDIRS = geqo path plan prep util + +include $(top_srcdir)/src/backend/common.mk diff --git a/src/backend/optimizer/README b/src/backend/optimizer/README new file mode 100644 index 0000000..41c120e --- /dev/null +++ b/src/backend/optimizer/README @@ -0,0 +1,1160 @@ +src/backend/optimizer/README + +Optimizer +========= + +These directories take the Query structure returned by the parser, and +generate a plan used by the executor. The /plan directory generates the +actual output plan, the /path code generates all possible ways to join the +tables, and /prep handles various preprocessing steps for special cases. +/util is utility stuff. /geqo is the separate "genetic optimization" planner +--- it does a semi-random search through the join tree space, rather than +exhaustively considering all possible join trees. (But each join considered +by /geqo is given to /path to create paths for, so we consider all possible +implementation paths for each specific join pair even in GEQO mode.) + + +Paths and Join Pairs +-------------------- + +During the planning/optimizing process, we build "Path" trees representing +the different ways of doing a query. We select the cheapest Path that +generates the desired relation and turn it into a Plan to pass to the +executor. (There is pretty nearly a one-to-one correspondence between the +Path and Plan trees, but Path nodes omit info that won't be needed during +planning, and include info needed for planning that won't be needed by the +executor.) + +The optimizer builds a RelOptInfo structure for each base relation used in +the query. Base rels are either primitive tables, or subquery subselects +that are planned via a separate recursive invocation of the planner. A +RelOptInfo is also built for each join relation that is considered during +planning. A join rel is simply a combination of base rels. There is only +one join RelOptInfo for any given set of baserels --- for example, the join +{A B C} is represented by the same RelOptInfo no matter whether we build it +by joining A and B first and then adding C, or joining B and C first and +then adding A, etc. These different means of building the joinrel are +represented as Paths. For each RelOptInfo we build a list of Paths that +represent plausible ways to implement the scan or join of that relation. +Once we've considered all the plausible Paths for a rel, we select the one +that is cheapest according to the planner's cost estimates. The final plan +is derived from the cheapest Path for the RelOptInfo that includes all the +base rels of the query. + +Possible Paths for a primitive table relation include plain old sequential +scan, plus index scans for any indexes that exist on the table, plus bitmap +index scans using one or more indexes. Specialized RTE types, such as +function RTEs, may have only one possible Path. + +Joins always occur using two RelOptInfos. One is outer, the other inner. +Outers drive lookups of values in the inner. In a nested loop, lookups of +values in the inner occur by scanning the inner path once per outer tuple +to find each matching inner row. In a mergejoin, inner and outer rows are +ordered, and are accessed in order, so only one scan is required to perform +the entire join: both inner and outer paths are scanned in-sync. (There's +not a lot of difference between inner and outer in a mergejoin...) In a +hashjoin, the inner is scanned first and all its rows are entered in a +hashtable, then the outer is scanned and for each row we lookup the join +key in the hashtable. + +A Path for a join relation is actually a tree structure, with the topmost +Path node representing the last-applied join method. It has left and right +subpaths that represent the scan or join methods used for the two input +relations. + + +Join Tree Construction +---------------------- + +The optimizer generates optimal query plans by doing a more-or-less +exhaustive search through the ways of executing the query. The best Path +tree is found by a recursive process: + +1) Take each base relation in the query, and make a RelOptInfo structure +for it. Find each potentially useful way of accessing the relation, +including sequential and index scans, and make Paths representing those +ways. All the Paths made for a given relation are placed in its +RelOptInfo.pathlist. (Actually, we discard Paths that are obviously +inferior alternatives before they ever get into the pathlist --- what +ends up in the pathlist is the cheapest way of generating each potentially +useful sort ordering and parameterization of the relation.) Also create a +RelOptInfo.joininfo list including all the join clauses that involve this +relation. For example, the WHERE clause "tab1.col1 = tab2.col1" generates +entries in both tab1 and tab2's joininfo lists. + +If we have only a single base relation in the query, we are done. +Otherwise we have to figure out how to join the base relations into a +single join relation. + +2) Normally, any explicit JOIN clauses are "flattened" so that we just +have a list of relations to join. However, FULL OUTER JOIN clauses are +never flattened, and other kinds of JOIN might not be either, if the +flattening process is stopped by join_collapse_limit or from_collapse_limit +restrictions. Therefore, we end up with a planning problem that contains +lists of relations to be joined in any order, where any individual item +might be a sub-list that has to be joined together before we can consider +joining it to its siblings. We process these sub-problems recursively, +bottom up. Note that the join list structure constrains the possible join +orders, but it doesn't constrain the join implementation method at each +join (nestloop, merge, hash), nor does it say which rel is considered outer +or inner at each join. We consider all these possibilities in building +Paths. We generate a Path for each feasible join method, and select the +cheapest Path. + +For each planning problem, therefore, we will have a list of relations +that are either base rels or joinrels constructed per sub-join-lists. +We can join these rels together in any order the planner sees fit. +The standard (non-GEQO) planner does this as follows: + +Consider joining each RelOptInfo to each other RelOptInfo for which there +is a usable joinclause, and generate a Path for each possible join method +for each such pair. (If we have a RelOptInfo with no join clauses, we have +no choice but to generate a clauseless Cartesian-product join; so we +consider joining that rel to each other available rel. But in the presence +of join clauses we will only consider joins that use available join +clauses. Note that join-order restrictions induced by outer joins and +IN/EXISTS clauses are also checked, to ensure that we find a workable join +order in cases where those restrictions force a clauseless join to be done.) + +If we only had two relations in the list, we are done: we just pick +the cheapest path for the join RelOptInfo. If we had more than two, we now +need to consider ways of joining join RelOptInfos to each other to make +join RelOptInfos that represent more than two list items. + +The join tree is constructed using a "dynamic programming" algorithm: +in the first pass (already described) we consider ways to create join rels +representing exactly two list items. The second pass considers ways +to make join rels that represent exactly three list items; the next pass, +four items, etc. The last pass considers how to make the final join +relation that includes all list items --- obviously there can be only one +join rel at this top level, whereas there can be more than one join rel +at lower levels. At each level we use joins that follow available join +clauses, if possible, just as described for the first level. + +For example: + + SELECT * + FROM tab1, tab2, tab3, tab4 + WHERE tab1.col = tab2.col AND + tab2.col = tab3.col AND + tab3.col = tab4.col + + Tables 1, 2, 3, and 4 are joined as: + {1 2},{2 3},{3 4} + {1 2 3},{2 3 4} + {1 2 3 4} + (other possibilities will be excluded for lack of join clauses) + + SELECT * + FROM tab1, tab2, tab3, tab4 + WHERE tab1.col = tab2.col AND + tab1.col = tab3.col AND + tab1.col = tab4.col + + Tables 1, 2, 3, and 4 are joined as: + {1 2},{1 3},{1 4} + {1 2 3},{1 3 4},{1 2 4} + {1 2 3 4} + +We consider left-handed plans (the outer rel of an upper join is a joinrel, +but the inner is always a single list item); right-handed plans (outer rel +is always a single item); and bushy plans (both inner and outer can be +joins themselves). For example, when building {1 2 3 4} we consider +joining {1 2 3} to {4} (left-handed), {4} to {1 2 3} (right-handed), and +{1 2} to {3 4} (bushy), among other choices. Although the jointree +scanning code produces these potential join combinations one at a time, +all the ways to produce the same set of joined base rels will share the +same RelOptInfo, so the paths produced from different join combinations +that produce equivalent joinrels will compete in add_path(). + +The dynamic-programming approach has an important property that's not +immediately obvious: we will finish constructing all paths for a given +relation before we construct any paths for relations containing that rel. +This means that we can reliably identify the "cheapest path" for each rel +before higher-level relations need to know that. Also, we can safely +discard a path when we find that another path for the same rel is better, +without worrying that maybe there is already a reference to that path in +some higher-level join path. Without this, memory management for paths +would be much more complicated. + +Once we have built the final join rel, we use either the cheapest path +for it or the cheapest path with the desired ordering (if that's cheaper +than applying a sort to the cheapest other path). + +If the query contains one-sided outer joins (LEFT or RIGHT joins), or +IN or EXISTS WHERE clauses that were converted to semijoins or antijoins, +then some of the possible join orders may be illegal. These are excluded +by having join_is_legal consult a side list of such "special" joins to see +whether a proposed join is illegal. (The same consultation allows it to +see which join style should be applied for a valid join, ie, JOIN_INNER, +JOIN_LEFT, etc.) + + +Valid OUTER JOIN Optimizations +------------------------------ + +The planner's treatment of outer join reordering is based on the following +identities: + +1. (A leftjoin B on (Pab)) innerjoin C on (Pac) + = (A innerjoin C on (Pac)) leftjoin B on (Pab) + +where Pac is a predicate referencing A and C, etc (in this case, clearly +Pac cannot reference B, or the transformation is nonsensical). + +2. (A leftjoin B on (Pab)) leftjoin C on (Pac) + = (A leftjoin C on (Pac)) leftjoin B on (Pab) + +3. (A leftjoin B on (Pab)) leftjoin C on (Pbc) + = A leftjoin (B leftjoin C on (Pbc)) on (Pab) + +Identity 3 only holds if predicate Pbc must fail for all-null B rows +(that is, Pbc is strict for at least one column of B). If Pbc is not +strict, the first form might produce some rows with nonnull C columns +where the second form would make those entries null. + +RIGHT JOIN is equivalent to LEFT JOIN after switching the two input +tables, so the same identities work for right joins. + +An example of a case that does *not* work is moving an innerjoin into or +out of the nullable side of an outer join: + + A leftjoin (B join C on (Pbc)) on (Pab) + != (A leftjoin B on (Pab)) join C on (Pbc) + +SEMI joins work a little bit differently. A semijoin can be reassociated +into or out of the lefthand side of another semijoin, left join, or +antijoin, but not into or out of the righthand side. Likewise, an inner +join, left join, or antijoin can be reassociated into or out of the +lefthand side of a semijoin, but not into or out of the righthand side. + +ANTI joins work approximately like LEFT joins, except that identity 3 +fails if the join to C is an antijoin (even if Pbc is strict, and in +both the cases where the other join is a leftjoin and where it is an +antijoin). So we can't reorder antijoins into or out of the RHS of a +leftjoin or antijoin, even if the relevant clause is strict. + +The current code does not attempt to re-order FULL JOINs at all. +FULL JOIN ordering is enforced by not collapsing FULL JOIN nodes when +translating the jointree to "joinlist" representation. Other types of +JOIN nodes are normally collapsed so that they participate fully in the +join order search. To avoid generating illegal join orders, the planner +creates a SpecialJoinInfo node for each non-inner join, and join_is_legal +checks this list to decide if a proposed join is legal. + +What we store in SpecialJoinInfo nodes are the minimum sets of Relids +required on each side of the join to form the outer join. Note that +these are minimums; there's no explicit maximum, since joining other +rels to the OJ's syntactic rels may be legal. Per identities 1 and 2, +non-FULL joins can be freely associated into the lefthand side of an +OJ, but in some cases they can't be associated into the righthand side. +So the restriction enforced by join_is_legal is that a proposed join +can't join a rel within or partly within an RHS boundary to one outside +the boundary, unless the proposed join is a LEFT join that can associate +into the SpecialJoinInfo's RHS using identity 3. + +The use of minimum Relid sets has some pitfalls; consider a query like + A leftjoin (B leftjoin (C innerjoin D) on (Pbcd)) on Pa +where Pa doesn't mention B/C/D at all. In this case a naive computation +would give the upper leftjoin's min LHS as {A} and min RHS as {C,D} (since +we know that the innerjoin can't associate out of the leftjoin's RHS, and +enforce that by including its relids in the leftjoin's min RHS). And the +lower leftjoin has min LHS of {B} and min RHS of {C,D}. Given such +information, join_is_legal would think it's okay to associate the upper +join into the lower join's RHS, transforming the query to + B leftjoin (A leftjoin (C innerjoin D) on Pa) on (Pbcd) +which yields totally wrong answers. We prevent that by forcing the min RHS +for the upper join to include B. This is perhaps overly restrictive, but +such cases don't arise often so it's not clear that it's worth developing a +more complicated system. + + +Pulling Up Subqueries +--------------------- + +As we described above, a subquery appearing in the range table is planned +independently and treated as a "black box" during planning of the outer +query. This is necessary when the subquery uses features such as +aggregates, GROUP, or DISTINCT. But if the subquery is just a simple +scan or join, treating the subquery as a black box may produce a poor plan +compared to considering it as part of the entire plan search space. +Therefore, at the start of the planning process the planner looks for +simple subqueries and pulls them up into the main query's jointree. + +Pulling up a subquery may result in FROM-list joins appearing below the top +of the join tree. Each FROM-list is planned using the dynamic-programming +search method described above. + +If pulling up a subquery produces a FROM-list as a direct child of another +FROM-list, then we can merge the two FROM-lists together. Once that's +done, the subquery is an absolutely integral part of the outer query and +will not constrain the join tree search space at all. However, that could +result in unpleasant growth of planning time, since the dynamic-programming +search has runtime exponential in the number of FROM-items considered. +Therefore, we don't merge FROM-lists if the result would have too many +FROM-items in one list. + + +Optimizer Functions +------------------- + +The primary entry point is planner(). + +planner() +set up for recursive handling of subqueries +-subquery_planner() + pull up sublinks and subqueries from rangetable, if possible + canonicalize qual + Attempt to simplify WHERE clause to the most useful form; this includes + flattening nested AND/ORs and detecting clauses that are duplicated in + different branches of an OR. + simplify constant expressions + process sublinks + convert Vars of outer query levels into Params +--grouping_planner() + preprocess target list for non-SELECT queries + handle UNION/INTERSECT/EXCEPT, GROUP BY, HAVING, aggregates, + ORDER BY, DISTINCT, LIMIT +---query_planner() + make list of base relations used in query + split up the qual into restrictions (a=1) and joins (b=c) + find qual clauses that enable merge and hash joins +----make_one_rel() + set_base_rel_pathlists() + find seqscan and all index paths for each base relation + find selectivity of columns used in joins + make_rel_from_joinlist() + hand off join subproblems to a plugin, GEQO, or standard_join_search() +------standard_join_search() + call join_search_one_level() for each level of join tree needed + join_search_one_level(): + For each joinrel of the prior level, do make_rels_by_clause_joins() + if it has join clauses, or make_rels_by_clauseless_joins() if not. + Also generate "bushy plan" joins between joinrels of lower levels. + Back at standard_join_search(), generate gather paths if needed for + each newly constructed joinrel, then apply set_cheapest() to extract + the cheapest path for it. + Loop back if this wasn't the top join level. + Back at grouping_planner: + do grouping (GROUP BY) and aggregation + do window functions + make unique (DISTINCT) + do sorting (ORDER BY) + do limit (LIMIT/OFFSET) +Back at planner(): +convert finished Path tree into a Plan tree +do final cleanup after planning + + +Optimizer Data Structures +------------------------- + +PlannerGlobal - global information for a single planner invocation + +PlannerInfo - information for planning a particular Query (we make + a separate PlannerInfo node for each sub-Query) + +RelOptInfo - a relation or joined relations + + RestrictInfo - WHERE clauses, like "x = 3" or "y = z" + (note the same structure is used for restriction and + join clauses) + + Path - every way to generate a RelOptInfo(sequential,index,joins) + A plain Path node can represent several simple plans, per its pathtype: + T_SeqScan - sequential scan + T_SampleScan - tablesample scan + T_FunctionScan - function-in-FROM scan + T_TableFuncScan - table function scan + T_ValuesScan - VALUES scan + T_CteScan - CTE (WITH) scan + T_NamedTuplestoreScan - ENR scan + T_WorkTableScan - scan worktable of a recursive CTE + T_Result - childless Result plan node (used for FROM-less SELECT) + IndexPath - index scan + BitmapHeapPath - top of a bitmapped index scan + TidPath - scan by CTID + TidRangePath - scan a contiguous range of CTIDs + SubqueryScanPath - scan a subquery-in-FROM + ForeignPath - scan a foreign table, foreign join or foreign upper-relation + CustomPath - for custom scan providers + AppendPath - append multiple subpaths together + MergeAppendPath - merge multiple subpaths, preserving their common sort order + GroupResultPath - childless Result plan node (used for degenerate grouping) + MaterialPath - a Material plan node + MemoizePath - a Memoize plan node for caching tuples from sub-paths + UniquePath - remove duplicate rows (either by hashing or sorting) + GatherPath - collect the results of parallel workers + GatherMergePath - collect parallel results, preserving their common sort order + ProjectionPath - a Result plan node with child (used for projection) + ProjectSetPath - a ProjectSet plan node applied to some sub-path + SortPath - a Sort plan node applied to some sub-path + IncrementalSortPath - an IncrementalSort plan node applied to some sub-path + GroupPath - a Group plan node applied to some sub-path + UpperUniquePath - a Unique plan node applied to some sub-path + AggPath - an Agg plan node applied to some sub-path + GroupingSetsPath - an Agg plan node used to implement GROUPING SETS + MinMaxAggPath - a Result plan node with subplans performing MIN/MAX + WindowAggPath - a WindowAgg plan node applied to some sub-path + SetOpPath - a SetOp plan node applied to some sub-path + RecursiveUnionPath - a RecursiveUnion plan node applied to two sub-paths + LockRowsPath - a LockRows plan node applied to some sub-path + ModifyTablePath - a ModifyTable plan node applied to some sub-path(s) + LimitPath - a Limit plan node applied to some sub-path + NestPath - nested-loop joins + MergePath - merge joins + HashPath - hash joins + + EquivalenceClass - a data structure representing a set of values known equal + + PathKey - a data structure representing the sort ordering of a path + +The optimizer spends a good deal of its time worrying about the ordering +of the tuples returned by a path. The reason this is useful is that by +knowing the sort ordering of a path, we may be able to use that path as +the left or right input of a mergejoin and avoid an explicit sort step. +Nestloops and hash joins don't really care what the order of their inputs +is, but mergejoin needs suitably ordered inputs. Therefore, all paths +generated during the optimization process are marked with their sort order +(to the extent that it is known) for possible use by a higher-level merge. + +It is also possible to avoid an explicit sort step to implement a user's +ORDER BY clause if the final path has the right ordering already, so the +sort ordering is of interest even at the top level. grouping_planner() will +look for the cheapest path with a sort order matching the desired order, +then compare its cost to the cost of using the cheapest-overall path and +doing an explicit sort on that. + +When we are generating paths for a particular RelOptInfo, we discard a path +if it is more expensive than another known path that has the same or better +sort order. We will never discard a path that is the only known way to +achieve a given sort order (without an explicit sort, that is). In this +way, the next level up will have the maximum freedom to build mergejoins +without sorting, since it can pick from any of the paths retained for its +inputs. + + +EquivalenceClasses +------------------ + +During the deconstruct_jointree() scan of the query's qual clauses, we look +for mergejoinable equality clauses A = B whose applicability is not delayed +by an outer join; these are called "equivalence clauses". When we find +one, we create an EquivalenceClass containing the expressions A and B to +record this knowledge. If we later find another equivalence clause B = C, +we add C to the existing EquivalenceClass for {A B}; this may require +merging two existing EquivalenceClasses. At the end of the scan, we have +sets of values that are known all transitively equal to each other. We can +therefore use a comparison of any pair of the values as a restriction or +join clause (when these values are available at the scan or join, of +course); furthermore, we need test only one such comparison, not all of +them. Therefore, equivalence clauses are removed from the standard qual +distribution process. Instead, when preparing a restriction or join clause +list, we examine each EquivalenceClass to see if it can contribute a +clause, and if so we select an appropriate pair of values to compare. For +example, if we are trying to join A's relation to C's, we can generate the +clause A = C, even though this appeared nowhere explicitly in the original +query. This may allow us to explore join paths that otherwise would have +been rejected as requiring Cartesian-product joins. + +Sometimes an EquivalenceClass may contain a pseudo-constant expression +(i.e., one not containing Vars or Aggs of the current query level, nor +volatile functions). In this case we do not follow the policy of +dynamically generating join clauses: instead, we dynamically generate +restriction clauses "var = const" wherever one of the variable members of +the class can first be computed. For example, if we have A = B and B = 42, +we effectively generate the restriction clauses A = 42 and B = 42, and then +we need not bother with explicitly testing the join clause A = B when the +relations are joined. In effect, all the class members can be tested at +relation-scan level and there's never a need for join tests. + +The precise technical interpretation of an EquivalenceClass is that it +asserts that at any plan node where more than one of its member values +can be computed, output rows in which the values are not all equal may +be discarded without affecting the query result. (We require all levels +of the plan to enforce EquivalenceClasses, hence a join need not recheck +equality of values that were computable by one of its children.) For an +ordinary EquivalenceClass that is "valid everywhere", we can further infer +that the values are all non-null, because all mergejoinable operators are +strict. However, we also allow equivalence clauses that appear below the +nullable side of an outer join to form EquivalenceClasses; for these +classes, the interpretation is that either all the values are equal, or +all (except pseudo-constants) have gone to null. (This requires a +limitation that non-constant members be strict, else they might not go +to null when the other members do.) Consider for example + + SELECT * + FROM a LEFT JOIN + (SELECT * FROM b JOIN c ON b.y = c.z WHERE b.y = 10) ss + ON a.x = ss.y + WHERE a.x = 42; + +We can form the below-outer-join EquivalenceClass {b.y c.z 10} and thereby +apply c.z = 10 while scanning c. (The reason we disallow outerjoin-delayed +clauses from forming EquivalenceClasses is exactly that we want to be able +to push any derived clauses as far down as possible.) But once above the +outer join it's no longer necessarily the case that b.y = 10, and thus we +cannot use such EquivalenceClasses to conclude that sorting is unnecessary +(see discussion of PathKeys below). + +In this example, notice also that a.x = ss.y (really a.x = b.y) is not an +equivalence clause because its applicability to b is delayed by the outer +join; thus we do not try to insert b.y into the equivalence class {a.x 42}. +But since we see that a.x has been equated to 42 above the outer join, we +are able to form a below-outer-join class {b.y 42}; this restriction can be +added because no b/c row not having b.y = 42 can contribute to the result +of the outer join, and so we need not compute such rows. Now this class +will get merged with {b.y c.z 10}, leading to the contradiction 10 = 42, +which lets the planner deduce that the b/c join need not be computed at all +because none of its rows can contribute to the outer join. (This gets +implemented as a gating Result filter, since more usually the potential +contradiction involves Param values rather than just Consts, and thus has +to be checked at runtime.) + +To aid in determining the sort ordering(s) that can work with a mergejoin, +we mark each mergejoinable clause with the EquivalenceClasses of its left +and right inputs. For an equivalence clause, these are of course the same +EquivalenceClass. For a non-equivalence mergejoinable clause (such as an +outer-join qualification), we generate two separate EquivalenceClasses for +the left and right inputs. This may result in creating single-item +equivalence "classes", though of course these are still subject to merging +if other equivalence clauses are later found to bear on the same +expressions. + +Another way that we may form a single-item EquivalenceClass is in creation +of a PathKey to represent a desired sort order (see below). This is a bit +different from the above cases because such an EquivalenceClass might +contain an aggregate function or volatile expression. (A clause containing +a volatile function will never be considered mergejoinable, even if its top +operator is mergejoinable, so there is no way for a volatile expression to +get into EquivalenceClasses otherwise. Aggregates are disallowed in WHERE +altogether, so will never be found in a mergejoinable clause.) This is just +a convenience to maintain a uniform PathKey representation: such an +EquivalenceClass will never be merged with any other. Note in particular +that a single-item EquivalenceClass {a.x} is *not* meant to imply an +assertion that a.x = a.x; the practical effect of this is that a.x could +be NULL. + +An EquivalenceClass also contains a list of btree opfamily OIDs, which +determines what the equalities it represents actually "mean". All the +equivalence clauses that contribute to an EquivalenceClass must have +equality operators that belong to the same set of opfamilies. (Note: most +of the time, a particular equality operator belongs to only one family, but +it's possible that it belongs to more than one. We keep track of all the +families to ensure that we can make use of an index belonging to any one of +the families for mergejoin purposes.) + +An EquivalenceClass can contain "em_is_child" members, which are copies +of members that contain appendrel parent relation Vars, transposed to +contain the equivalent child-relation variables or expressions. These +members are *not* full-fledged members of the EquivalenceClass and do not +affect the class's overall properties at all. They are kept only to +simplify matching of child-relation expressions to EquivalenceClasses. +Most operations on EquivalenceClasses should ignore child members. + + +PathKeys +-------- + +The PathKeys data structure represents what is known about the sort order +of the tuples generated by a particular Path. A path's pathkeys field is a +list of PathKey nodes, where the n'th item represents the n'th sort key of +the result. Each PathKey contains these fields: + + * a reference to an EquivalenceClass + * a btree opfamily OID (must match one of those in the EC) + * a sort direction (ascending or descending) + * a nulls-first-or-last flag + +The EquivalenceClass represents the value being sorted on. Since the +various members of an EquivalenceClass are known equal according to the +opfamily, we can consider a path sorted by any one of them to be sorted by +any other too; this is what justifies referencing the whole +EquivalenceClass rather than just one member of it. + +In single/base relation RelOptInfo's, the Paths represent various ways +of scanning the relation and the resulting ordering of the tuples. +Sequential scan Paths have NIL pathkeys, indicating no known ordering. +Index scans have Path.pathkeys that represent the chosen index's ordering, +if any. A single-key index would create a single-PathKey list, while a +multi-column index generates a list with one element per key index column. +Non-key columns specified in the INCLUDE clause of covering indexes don't +have corresponding PathKeys in the list, because the have no influence on +index ordering. (Actually, since an index can be scanned either forward or +backward, there are two possible sort orders and two possible PathKey lists +it can generate.) + +Note that a bitmap scan has NIL pathkeys since we can say nothing about +the overall order of its result. Also, an indexscan on an unordered type +of index generates NIL pathkeys. However, we can always create a pathkey +by doing an explicit sort. The pathkeys for a Sort plan's output just +represent the sort key fields and the ordering operators used. + +Things get more interesting when we consider joins. Suppose we do a +mergejoin between A and B using the mergeclause A.X = B.Y. The output +of the mergejoin is sorted by X --- but it is also sorted by Y. Again, +this can be represented by a PathKey referencing an EquivalenceClass +containing both X and Y. + +With a little further thought, it becomes apparent that nestloop joins +can also produce sorted output. For example, if we do a nestloop join +between outer relation A and inner relation B, then any pathkeys relevant +to A are still valid for the join result: we have not altered the order of +the tuples from A. Even more interesting, if there was an equivalence clause +A.X=B.Y, and A.X was a pathkey for the outer relation A, then we can assert +that B.Y is a pathkey for the join result; X was ordered before and still +is, and the joined values of Y are equal to the joined values of X, so Y +must now be ordered too. This is true even though we used neither an +explicit sort nor a mergejoin on Y. (Note: hash joins cannot be counted +on to preserve the order of their outer relation, because the executor +might decide to "batch" the join, so we always set pathkeys to NIL for +a hashjoin path.) Exception: a RIGHT or FULL join doesn't preserve the +ordering of its outer relation, because it might insert nulls at random +points in the ordering. + +In general, we can justify using EquivalenceClasses as the basis for +pathkeys because, whenever we scan a relation containing multiple +EquivalenceClass members or join two relations each containing +EquivalenceClass members, we apply restriction or join clauses derived from +the EquivalenceClass. This guarantees that any two values listed in the +EquivalenceClass are in fact equal in all tuples emitted by the scan or +join, and therefore that if the tuples are sorted by one of the values, +they can be considered sorted by any other as well. It does not matter +whether the test clause is used as a mergeclause, or merely enforced +after-the-fact as a qpqual filter. + +Note that there is no particular difficulty in labeling a path's sort +order with a PathKey referencing an EquivalenceClass that contains +variables not yet joined into the path's output. We can simply ignore +such entries as not being relevant (yet). This makes it possible to +use the same EquivalenceClasses throughout the join planning process. +In fact, by being careful not to generate multiple identical PathKey +objects, we can reduce comparison of EquivalenceClasses and PathKeys +to simple pointer comparison, which is a huge savings because add_path +has to make a large number of PathKey comparisons in deciding whether +competing Paths are equivalently sorted. + +Pathkeys are also useful to represent an ordering that we wish to achieve, +since they are easily compared to the pathkeys of a potential candidate +path. So, SortGroupClause lists are turned into pathkeys lists for use +inside the optimizer. + +An additional refinement we can make is to insist that canonical pathkey +lists (sort orderings) do not mention the same EquivalenceClass more than +once. For example, in all these cases the second sort column is redundant, +because it cannot distinguish values that are the same according to the +first sort column: + SELECT ... ORDER BY x, x + SELECT ... ORDER BY x, x DESC + SELECT ... WHERE x = y ORDER BY x, y +Although a user probably wouldn't write "ORDER BY x,x" directly, such +redundancies are more probable once equivalence classes have been +considered. Also, the system may generate redundant pathkey lists when +computing the sort ordering needed for a mergejoin. By eliminating the +redundancy, we save time and improve planning, since the planner will more +easily recognize equivalent orderings as being equivalent. + +Another interesting property is that if the underlying EquivalenceClass +contains a constant and is not below an outer join, then the pathkey is +completely redundant and need not be sorted by at all! Every row must +contain the same constant value, so there's no need to sort. (If the EC is +below an outer join, we still have to sort, since some of the rows might +have gone to null and others not. In this case we must be careful to pick +a non-const member to sort by. The assumption that all the non-const +members go to null at the same plan level is critical here, else they might +not produce the same sort order.) This might seem pointless because users +are unlikely to write "... WHERE x = 42 ORDER BY x", but it allows us to +recognize when particular index columns are irrelevant to the sort order: +if we have "... WHERE x = 42 ORDER BY y", scanning an index on (x,y) +produces correctly ordered data without a sort step. We used to have very +ugly ad-hoc code to recognize that in limited contexts, but discarding +constant ECs from pathkeys makes it happen cleanly and automatically. + +You might object that a below-outer-join EquivalenceClass doesn't always +represent the same values at every level of the join tree, and so using +it to uniquely identify a sort order is dubious. This is true, but we +can avoid dealing with the fact explicitly because we always consider that +an outer join destroys any ordering of its nullable inputs. Thus, even +if a path was sorted by {a.x} below an outer join, we'll re-sort if that +sort ordering was important; and so using the same PathKey for both sort +orderings doesn't create any real problem. + + +Order of processing for EquivalenceClasses and PathKeys +------------------------------------------------------- + +As alluded to above, there is a specific sequence of phases in the +processing of EquivalenceClasses and PathKeys during planning. During the +initial scanning of the query's quals (deconstruct_jointree followed by +reconsider_outer_join_clauses), we construct EquivalenceClasses based on +mergejoinable clauses found in the quals. At the end of this process, +we know all we can know about equivalence of different variables, so +subsequently there will be no further merging of EquivalenceClasses. +At that point it is possible to consider the EquivalenceClasses as +"canonical" and build canonical PathKeys that reference them. At this +time we construct PathKeys for the query's ORDER BY and related clauses. +(Any ordering expressions that do not appear elsewhere will result in +the creation of new EquivalenceClasses, but this cannot result in merging +existing classes, so canonical-ness is not lost.) + +Because all the EquivalenceClasses are known before we begin path +generation, we can use them as a guide to which indexes are of interest: +if an index's column is not mentioned in any EquivalenceClass then that +index's sort order cannot possibly be helpful for the query. This allows +short-circuiting of much of the processing of create_index_paths() for +irrelevant indexes. + +There are some cases where planner.c constructs additional +EquivalenceClasses and PathKeys after query_planner has completed. +In these cases, the extra ECs/PKs are needed to represent sort orders +that were not considered during query_planner. Such situations should be +minimized since it is impossible for query_planner to return a plan +producing such a sort order, meaning an explicit sort will always be needed. +Currently this happens only for queries involving multiple window functions +with different orderings, for which extra sorts are needed anyway. + + +Parameterized Paths +------------------- + +The naive way to join two relations using a clause like WHERE A.X = B.Y +is to generate a nestloop plan like this: + + NestLoop + Filter: A.X = B.Y + -> Seq Scan on A + -> Seq Scan on B + +We can make this better by using a merge or hash join, but it still +requires scanning all of both input relations. If A is very small and B is +very large, but there is an index on B.Y, it can be enormously better to do +something like this: + + NestLoop + -> Seq Scan on A + -> Index Scan using B_Y_IDX on B + Index Condition: B.Y = A.X + +Here, we are expecting that for each row scanned from A, the nestloop +plan node will pass down the current value of A.X into the scan of B. +That allows the indexscan to treat A.X as a constant for any one +invocation, and thereby use it as an index key. This is the only plan type +that can avoid fetching all of B, and for small numbers of rows coming from +A, that will dominate every other consideration. (As A gets larger, this +gets less attractive, and eventually a merge or hash join will win instead. +So we have to cost out all the alternatives to decide what to do.) + +It can be useful for the parameter value to be passed down through +intermediate layers of joins, for example: + + NestLoop + -> Seq Scan on A + Hash Join + Join Condition: B.Y = C.W + -> Seq Scan on B + -> Index Scan using C_Z_IDX on C + Index Condition: C.Z = A.X + +If all joins are plain inner joins then this is usually unnecessary, +because it's possible to reorder the joins so that a parameter is used +immediately below the nestloop node that provides it. But in the +presence of outer joins, such join reordering may not be possible. + +Also, the bottom-level scan might require parameters from more than one +other relation. In principle we could join the other relations first +so that all the parameters are supplied from a single nestloop level. +But if those other relations have no join clause in common (which is +common in star-schema queries for instance), the planner won't consider +joining them directly to each other. In such a case we need to be able +to create a plan like + + NestLoop + -> Seq Scan on SmallTable1 A + NestLoop + -> Seq Scan on SmallTable2 B + -> Index Scan using XYIndex on LargeTable C + Index Condition: C.X = A.AID and C.Y = B.BID + +so we should be willing to pass down A.AID through a join even though +there is no join order constraint forcing the plan to look like this. + +Before version 9.2, Postgres used ad-hoc methods for planning and +executing nestloop queries of this kind, and those methods could not +handle passing parameters down through multiple join levels. + +To plan such queries, we now use a notion of a "parameterized path", +which is a path that makes use of a join clause to a relation that's not +scanned by the path. In the example two above, we would construct a +path representing the possibility of doing this: + + -> Index Scan using C_Z_IDX on C + Index Condition: C.Z = A.X + +This path will be marked as being parameterized by relation A. (Note that +this is only one of the possible access paths for C; we'd still have a +plain unparameterized seqscan, and perhaps other possibilities.) The +parameterization marker does not prevent joining the path to B, so one of +the paths generated for the joinrel {B C} will represent + + Hash Join + Join Condition: B.Y = C.W + -> Seq Scan on B + -> Index Scan using C_Z_IDX on C + Index Condition: C.Z = A.X + +This path is still marked as being parameterized by A. When we attempt to +join {B C} to A to form the complete join tree, such a path can only be +used as the inner side of a nestloop join: it will be ignored for other +possible join types. So we will form a join path representing the query +plan shown above, and it will compete in the usual way with paths built +from non-parameterized scans. + +While all ordinary paths for a particular relation generate the same set +of rows (since they must all apply the same set of restriction clauses), +parameterized paths typically generate fewer rows than less-parameterized +paths, since they have additional clauses to work with. This means we +must consider the number of rows generated as an additional figure of +merit. A path that costs more than another, but generates fewer rows, +must be kept since the smaller number of rows might save work at some +intermediate join level. (It would not save anything if joined +immediately to the source of the parameters.) + +To keep cost estimation rules relatively simple, we make an implementation +restriction that all paths for a given relation of the same parameterization +(i.e., the same set of outer relations supplying parameters) must have the +same rowcount estimate. This is justified by insisting that each such path +apply *all* join clauses that are available with the named outer relations. +Different paths might, for instance, choose different join clauses to use +as index clauses; but they must then apply any other join clauses available +from the same outer relations as filter conditions, so that the set of rows +returned is held constant. This restriction doesn't degrade the quality of +the finished plan: it amounts to saying that we should always push down +movable join clauses to the lowest possible evaluation level, which is a +good thing anyway. The restriction is useful in particular to support +pre-filtering of join paths in add_path_precheck. Without this rule we +could never reject a parameterized path in advance of computing its rowcount +estimate, which would greatly reduce the value of the pre-filter mechanism. + +To limit planning time, we have to avoid generating an unreasonably large +number of parameterized paths. We do this by only generating parameterized +relation scan paths for index scans, and then only for indexes for which +suitable join clauses are available. There are also heuristics in join +planning that try to limit the number of parameterized paths considered. + +In particular, there's been a deliberate policy decision to favor hash +joins over merge joins for parameterized join steps (those occurring below +a nestloop that provides parameters to the lower join's inputs). While we +do not ignore merge joins entirely, joinpath.c does not fully explore the +space of potential merge joins with parameterized inputs. Also, add_path +treats parameterized paths as having no pathkeys, so that they compete +only on cost and rowcount; they don't get preference for producing a +special sort order. This creates additional bias against merge joins, +since we might discard a path that could have been useful for performing +a merge without an explicit sort step. Since a parameterized path must +ultimately be used on the inside of a nestloop, where its sort order is +uninteresting, these choices do not affect any requirement for the final +output order of a query --- they only make it harder to use a merge join +at a lower level. The savings in planning work justifies that. + +Similarly, parameterized paths do not normally get preference in add_path +for having cheap startup cost; that's seldom of much value when on the +inside of a nestloop, so it seems not worth keeping extra paths solely for +that. An exception occurs for parameterized paths for the RHS relation of +a SEMI or ANTI join: in those cases, we can stop the inner scan after the +first match, so it's primarily startup not total cost that we care about. + + +LATERAL subqueries +------------------ + +As of 9.3 we support SQL-standard LATERAL references from subqueries in +FROM (and also functions in FROM). The planner implements these by +generating parameterized paths for any RTE that contains lateral +references. In such cases, *all* paths for that relation will be +parameterized by at least the set of relations used in its lateral +references. (And in turn, join relations including such a subquery might +not have any unparameterized paths.) All the other comments made above for +parameterized paths still apply, though; in particular, each such path is +still expected to enforce any join clauses that can be pushed down to it, +so that all paths of the same parameterization have the same rowcount. + +We also allow LATERAL subqueries to be flattened (pulled up into the parent +query) by the optimizer, but only when this does not introduce lateral +references into JOIN/ON quals that would refer to relations outside the +lowest outer join at/above that qual. The semantics of such a qual would +be unclear. Note that even with this restriction, pullup of a LATERAL +subquery can result in creating PlaceHolderVars that contain lateral +references to relations outside their syntactic scope. We still evaluate +such PHVs at their syntactic location or lower, but the presence of such a +PHV in the quals or targetlist of a plan node requires that node to appear +on the inside of a nestloop join relative to the rel(s) supplying the +lateral reference. (Perhaps now that that stuff works, we could relax the +pullup restriction?) + + +Security-level constraints on qual clauses +------------------------------------------ + +To support row-level security and security-barrier views efficiently, +we mark qual clauses (RestrictInfo nodes) with a "security_level" field. +The basic concept is that a qual with a lower security_level must be +evaluated before one with a higher security_level. This ensures that +"leaky" quals that might expose sensitive data are not evaluated until +after the security barrier quals that are supposed to filter out +security-sensitive rows. However, many qual conditions are "leakproof", +that is we trust the functions they use to not expose data. To avoid +unnecessarily inefficient plans, a leakproof qual is not delayed by +security-level considerations, even if it has a higher syntactic +security_level than another qual. + +In a query that contains no use of RLS or security-barrier views, all +quals will have security_level zero, so that none of these restrictions +kick in; we don't even need to check leakproofness of qual conditions. + +If there are security-barrier quals, they get security_level zero (and +possibly higher, if there are multiple layers of barriers). Regular quals +coming from the query text get a security_level one more than the highest +level used for barrier quals. + +When new qual clauses are generated by EquivalenceClass processing, +they must be assigned a security_level. This is trickier than it seems. +One's first instinct is that it would be safe to use the largest level +found among the source quals for the EquivalenceClass, but that isn't +safe at all, because it allows unwanted delays of security-barrier quals. +Consider a barrier qual "t.x = t.y" plus a query qual "t.x = constant", +and suppose there is another query qual "leaky_function(t.z)" that +we mustn't evaluate before the barrier qual has been checked. +We will have an EC {t.x, t.y, constant} which will lead us to replace +the EC quals with "t.x = constant AND t.y = constant". (We do not want +to give up that behavior, either, since the latter condition could allow +use of an index on t.y, which we would never discover from the original +quals.) If these generated quals are assigned the same security_level as +the query quals, then it's possible for the leaky_function qual to be +evaluated first, allowing leaky_function to see data from rows that +possibly don't pass the barrier condition. + +Instead, our handling of security levels with ECs works like this: +* Quals are not accepted as source clauses for ECs in the first place +unless they are leakproof or have security_level zero. +* EC-derived quals are assigned the minimum (not maximum) security_level +found among the EC's source clauses. +* If the maximum security_level found among the EC's source clauses is +above zero, then the equality operators selected for derived quals must +be leakproof. When no such operator can be found, the EC is treated as +"broken" and we fall back to emitting its source clauses without any +additional derived quals. + +These rules together ensure that an untrusted qual clause (one with +security_level above zero) cannot cause an EC to generate a leaky derived +clause. This makes it safe to use the minimum not maximum security_level +for derived clauses. The rules could result in poor plans due to not +being able to generate derived clauses at all, but the risk of that is +small in practice because most btree equality operators are leakproof. +Also, by making exceptions for level-zero quals, we ensure that there is +no plan degradation when no barrier quals are present. + +Once we have security levels assigned to all clauses, enforcement +of barrier-qual ordering restrictions boils down to two rules: + +* Table scan plan nodes must not select quals for early execution +(for example, use them as index qualifiers in an indexscan) unless +they are leakproof or have security_level no higher than any other +qual that is due to be executed at the same plan node. (Use the +utility function restriction_is_securely_promotable() to check +whether it's okay to select a qual for early execution.) + +* Normal execution of a list of quals must execute them in an order +that satisfies the same security rule, ie higher security_levels must +be evaluated later unless leakproof. (This is handled in a single place +by order_qual_clauses() in createplan.c.) + +order_qual_clauses() uses a heuristic to decide exactly what to do with +leakproof clauses. Normally it sorts clauses by security_level then cost, +being careful that the sort is stable so that we don't reorder clauses +without a clear reason. But this could result in a very expensive qual +being done before a cheaper one that is of higher security_level. +If the cheaper qual is leaky we have no choice, but if it is leakproof +we could put it first. We choose to sort leakproof quals as if they +have security_level zero, but only when their cost is less than 10X +cpu_operator_cost; that restriction alleviates the opposite problem of +doing expensive quals first just because they're leakproof. + +Additional rules will be needed to support safe handling of join quals +when there is a mix of security levels among join quals; for example, it +will be necessary to prevent leaky higher-security-level quals from being +evaluated at a lower join level than other quals of lower security level. +Currently there is no need to consider that since security-prioritized +quals can only be single-table restriction quals coming from RLS policies +or security-barrier views, and security-barrier view subqueries are never +flattened into the parent query. Hence enforcement of security-prioritized +quals only happens at the table scan level. With extra rules for safe +handling of security levels among join quals, it should be possible to let +security-barrier views be flattened into the parent query, allowing more +flexibility of planning while still preserving required ordering of qual +evaluation. But that will come later. + + +Post scan/join planning +----------------------- + +So far we have discussed only scan/join planning, that is, implementation +of the FROM and WHERE clauses of a SQL query. But the planner must also +determine how to deal with GROUP BY, aggregation, and other higher-level +features of queries; and in many cases there are multiple ways to do these +steps and thus opportunities for optimization choices. These steps, like +scan/join planning, are handled by constructing Paths representing the +different ways to do a step, then choosing the cheapest Path. + +Since all Paths require a RelOptInfo as "parent", we create RelOptInfos +representing the outputs of these upper-level processing steps. These +RelOptInfos are mostly dummy, but their pathlist lists hold all the Paths +considered useful for each step. Currently, we may create these types of +additional RelOptInfos during upper-level planning: + +UPPERREL_SETOP result of UNION/INTERSECT/EXCEPT, if any +UPPERREL_PARTIAL_GROUP_AGG result of partial grouping/aggregation, if any +UPPERREL_GROUP_AGG result of grouping/aggregation, if any +UPPERREL_WINDOW result of window functions, if any +UPPERREL_PARTIAL_DISTINCT result of partial "SELECT DISTINCT", if any +UPPERREL_DISTINCT result of "SELECT DISTINCT", if any +UPPERREL_ORDERED result of ORDER BY, if any +UPPERREL_FINAL result of any remaining top-level actions + +UPPERREL_FINAL is used to represent any final processing steps, currently +LockRows (SELECT FOR UPDATE), LIMIT/OFFSET, and ModifyTable. There is no +flexibility about the order in which these steps are done, and thus no need +to subdivide this stage more finely. + +These "upper relations" are identified by the UPPERREL enum values shown +above, plus a relids set, which allows there to be more than one upperrel +of the same kind. We use NULL for the relids if there's no need for more +than one upperrel of the same kind. Currently, in fact, the relids set +is vestigial because it's always NULL, but that's expected to change in +the future. For example, in planning set operations, we might need the +relids to denote which subset of the leaf SELECTs has been combined in a +particular group of Paths that are competing with each other. + +The result of subquery_planner() is always returned as a set of Paths +stored in the UPPERREL_FINAL rel with NULL relids. The other types of +upperrels are created only if needed for the particular query. + + +Parallel Query and Partial Paths +-------------------------------- + +Parallel query involves dividing up the work that needs to be performed +either by an entire query or some portion of the query in such a way that +some of that work can be done by one or more worker processes, which are +called parallel workers. Parallel workers are a subtype of dynamic +background workers; see src/backend/access/transam/README.parallel for a +fuller description. The academic literature on parallel query suggests +that parallel execution strategies can be divided into essentially two +categories: pipelined parallelism, where the execution of the query is +divided into multiple stages and each stage is handled by a separate +process; and partitioning parallelism, where the data is split between +multiple processes and each process handles a subset of it. The +literature, however, suggests that gains from pipeline parallelism are +often very limited due to the difficulty of avoiding pipeline stalls. +Consequently, we do not currently attempt to generate query plans that +use this technique. + +Instead, we focus on partitioning parallelism, which does not require +that the underlying table be partitioned. It only requires that (1) +there is some method of dividing the data from at least one of the base +tables involved in the relation across multiple processes, (2) allowing +each process to handle its own portion of the data, and then (3) +collecting the results. Requirements (2) and (3) are satisfied by the +executor node Gather (or GatherMerge), which launches any number of worker +processes and executes its single child plan in all of them, and perhaps +in the leader also, if the children aren't generating enough data to keep +the leader busy. Requirement (1) is handled by the table scan node: when +invoked with parallel_aware = true, this node will, in effect, partition +the table on a block by block basis, returning a subset of the tuples from +the relation in each worker where that scan node is executed. + +Just as we do for non-parallel access methods, we build Paths to +represent access strategies that can be used in a parallel plan. These +are, in essence, the same strategies that are available in the +non-parallel plan, but there is an important difference: a path that +will run beneath a Gather node returns only a subset of the query +results in each worker, not all of them. To form a path that can +actually be executed, the (rather large) cost of the Gather node must be +accounted for. For this reason among others, paths intended to run +beneath a Gather node - which we call "partial" paths since they return +only a subset of the results in each worker - must be kept separate from +ordinary paths (see RelOptInfo's partial_pathlist and the function +add_partial_path). + +One of the keys to making parallel query effective is to run as much of +the query in parallel as possible. Therefore, we expect it to generally +be desirable to postpone the Gather stage until as near to the top of the +plan as possible. Expanding the range of cases in which more work can be +pushed below the Gather (and costing them accurately) is likely to keep us +busy for a long time to come. + +Partitionwise joins +------------------- + +A join between two similarly partitioned tables can be broken down into joins +between their matching partitions if there exists an equi-join condition +between the partition keys of the joining tables. The equi-join between +partition keys implies that all join partners for a given row in one +partitioned table must be in the corresponding partition of the other +partitioned table. Because of this the join between partitioned tables to be +broken into joins between the matching partitions. The resultant join is +partitioned in the same way as the joining relations, thus allowing an N-way +join between similarly partitioned tables having equi-join condition between +their partition keys to be broken down into N-way joins between their matching +partitions. This technique of breaking down a join between partitioned tables +into joins between their partitions is called partitionwise join. We will use +term "partitioned relation" for either a partitioned table or a join between +compatibly partitioned tables. + +Even if the joining relations don't have exactly the same partition bounds, +partitionwise join can still be applied by using an advanced +partition-matching algorithm. For both the joining relations, the algorithm +checks whether every partition of one joining relation only matches one +partition of the other joining relation at most. In such a case the join +between the joining relations can be broken down into joins between the +matching partitions. The join relation can then be considered partitioned. +The algorithm produces the pairs of the matching partitions, plus the +partition bounds for the join relation, to allow partitionwise join for +computing the join. The algorithm is implemented in partition_bounds_merge(). +For an N-way join relation considered partitioned this way, not every pair of +joining relations can use partitionwise join. For example: + + (A leftjoin B on (Pab)) innerjoin C on (Pac) + +where A, B, and C are partitioned tables, and A has an extra partition +compared to B and C. When considering partitionwise join for the join {A B}, +the extra partition of A doesn't have a matching partition on the nullable +side, which is the case that the current implementation of partitionwise join +can't handle. So {A B} is not considered partitioned, and the pair of {A B} +and C considered for the 3-way join can't use partitionwise join. On the +other hand, the pair of {A C} and B can use partitionwise join because {A C} +is considered partitioned by eliminating the extra partition (see identity 1 +on outer join reordering). Whether an N-way join can use partitionwise join +is determined based on the first pair of joining relations that are both +partitioned and can use partitionwise join. + +The partitioning properties of a partitioned relation are stored in its +RelOptInfo. The information about data types of partition keys are stored in +PartitionSchemeData structure. The planner maintains a list of canonical +partition schemes (distinct PartitionSchemeData objects) so that RelOptInfo of +any two partitioned relations with same partitioning scheme point to the same +PartitionSchemeData object. This reduces memory consumed by +PartitionSchemeData objects and makes it easy to compare the partition schemes +of joining relations. + +Partitionwise aggregates/grouping +--------------------------------- + +If the GROUP BY clause contains all of the partition keys, all the rows +that belong to a given group must come from a single partition; therefore, +aggregation can be done completely separately for each partition. Otherwise, +partial aggregates can be computed for each partition, and then finalized +after appending the results from the individual partitions. This technique of +breaking down aggregation or grouping over a partitioned relation into +aggregation or grouping over its partitions is called partitionwise +aggregation. Especially when the partition keys match the GROUP BY clause, +this can be significantly faster than the regular method. diff --git a/src/backend/optimizer/geqo/Makefile b/src/backend/optimizer/geqo/Makefile new file mode 100644 index 0000000..1978c14 --- /dev/null +++ b/src/backend/optimizer/geqo/Makefile @@ -0,0 +1,33 @@ +#------------------------------------------------------------------------- +# +# Makefile-- +# Makefile for the genetic query optimizer module +# +# Copyright (c) 1994, Regents of the University of California +# +# src/backend/optimizer/geqo/Makefile +# +#------------------------------------------------------------------------- + +subdir = src/backend/optimizer/geqo +top_builddir = ../../../.. +include $(top_builddir)/src/Makefile.global + +OBJS = \ + geqo_copy.o \ + geqo_cx.o \ + geqo_erx.o \ + geqo_eval.o \ + geqo_main.o \ + geqo_misc.o \ + geqo_mutation.o \ + geqo_ox1.o \ + geqo_ox2.o \ + geqo_pmx.o \ + geqo_pool.o \ + geqo_px.o \ + geqo_random.o \ + geqo_recombination.o \ + geqo_selection.o + +include $(top_srcdir)/src/backend/common.mk diff --git a/src/backend/optimizer/geqo/geqo_copy.c b/src/backend/optimizer/geqo/geqo_copy.c new file mode 100644 index 0000000..62302f4 --- /dev/null +++ b/src/backend/optimizer/geqo/geqo_copy.c @@ -0,0 +1,54 @@ +/*------------------------------------------------------------------------ + * + * geqo_copy.c + * + * Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group + * Portions Copyright (c) 1994, Regents of the University of California + * + * src/backend/optimizer/geqo/geqo_copy.c + * + *------------------------------------------------------------------------- + */ + +/* contributed by: + =*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*= + * Martin Utesch * Institute of Automatic Control * + = = University of Mining and Technology = + * utesch@aut.tu-freiberg.de * Freiberg, Germany * + =*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*= + */ + +/* this is adopted from D. Whitley's Genitor algorithm */ + +/*************************************************************/ +/* */ +/* Copyright (c) 1990 */ +/* Darrell L. Whitley */ +/* Computer Science Department */ +/* Colorado State University */ +/* */ +/* Permission is hereby granted to copy all or any part of */ +/* this program for free distribution. The author's name */ +/* and this copyright notice must be included in any copy. */ +/* */ +/*************************************************************/ + +#include "postgres.h" +#include "optimizer/geqo_copy.h" + +/* geqo_copy + * + * copies one gene to another + * + */ +void +geqo_copy(PlannerInfo *root, Chromosome *chromo1, Chromosome *chromo2, + int string_length) +{ + int i; + + for (i = 0; i < string_length; i++) + chromo1->string[i] = chromo2->string[i]; + + chromo1->worth = chromo2->worth; +} diff --git a/src/backend/optimizer/geqo/geqo_cx.c b/src/backend/optimizer/geqo/geqo_cx.c new file mode 100644 index 0000000..34cc53a --- /dev/null +++ b/src/backend/optimizer/geqo/geqo_cx.c @@ -0,0 +1,123 @@ +/*------------------------------------------------------------------------ +* +* geqo_cx.c +* +* cycle crossover [CX] routines; +* CX operator according to Oliver et al +* (Proc 2nd Int'l Conf on GA's) +* +* src/backend/optimizer/geqo/geqo_cx.c +* +*------------------------------------------------------------------------- +*/ + +/* contributed by: + =*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*= + * Martin Utesch * Institute of Automatic Control * + = = University of Mining and Technology = + * utesch@aut.tu-freiberg.de * Freiberg, Germany * + =*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*= + */ + +/* the cx algorithm is adopted from Genitor : */ +/*************************************************************/ +/* */ +/* Copyright (c) 1990 */ +/* Darrell L. Whitley */ +/* Computer Science Department */ +/* Colorado State University */ +/* */ +/* Permission is hereby granted to copy all or any part of */ +/* this program for free distribution. The author's name */ +/* and this copyright notice must be included in any copy. */ +/* */ +/*************************************************************/ + + +#include "postgres.h" +#include "optimizer/geqo_random.h" +#include "optimizer/geqo_recombination.h" + +#if defined(CX) + +/* cx + * + * cycle crossover + */ +int +cx(PlannerInfo *root, Gene *tour1, Gene *tour2, Gene *offspring, + int num_gene, City * city_table) +{ + int i, + start_pos, + curr_pos; + int count = 0; + int num_diffs = 0; + + /* initialize city table */ + for (i = 1; i <= num_gene; i++) + { + city_table[i].used = 0; + city_table[tour2[i - 1]].tour2_position = i - 1; + city_table[tour1[i - 1]].tour1_position = i - 1; + } + + /* choose random cycle starting position */ + start_pos = geqo_randint(root, num_gene - 1, 0); + + /* child inherits first city */ + offspring[start_pos] = tour1[start_pos]; + + /* begin cycle with tour1 */ + curr_pos = start_pos; + city_table[(int) tour1[start_pos]].used = 1; + + count++; + + /* cx main part */ + + +/* STEP 1 */ + + while (tour2[curr_pos] != tour1[start_pos]) + { + city_table[(int) tour2[curr_pos]].used = 1; + curr_pos = city_table[(int) tour2[curr_pos]].tour1_position; + offspring[curr_pos] = tour1[curr_pos]; + count++; + } + + +/* STEP 2 */ + + /* failed to create a complete tour */ + if (count < num_gene) + { + for (i = 1; i <= num_gene; i++) + { + if (!city_table[i].used) + { + offspring[city_table[i].tour2_position] = + tour2[(int) city_table[i].tour2_position]; + count++; + } + } + } + + +/* STEP 3 */ + + /* still failed to create a complete tour */ + if (count < num_gene) + { + + /* count the number of differences between mom and offspring */ + for (i = 0; i < num_gene; i++) + if (tour1[i] != offspring[i]) + num_diffs++; + } + + return num_diffs; +} + +#endif /* defined(CX) */ diff --git a/src/backend/optimizer/geqo/geqo_erx.c b/src/backend/optimizer/geqo/geqo_erx.c new file mode 100644 index 0000000..cc06613 --- /dev/null +++ b/src/backend/optimizer/geqo/geqo_erx.c @@ -0,0 +1,468 @@ +/*------------------------------------------------------------------------ +* +* geqo_erx.c +* edge recombination crossover [ER] +* +* src/backend/optimizer/geqo/geqo_erx.c +* +*------------------------------------------------------------------------- +*/ + +/* contributed by: + =*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*= + * Martin Utesch * Institute of Automatic Control * + = = University of Mining and Technology = + * utesch@aut.tu-freiberg.de * Freiberg, Germany * + =*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*= + */ + +/* the edge recombination algorithm is adopted from Genitor : */ +/*************************************************************/ +/* */ +/* Copyright (c) 1990 */ +/* Darrell L. Whitley */ +/* Computer Science Department */ +/* Colorado State University */ +/* */ +/* Permission is hereby granted to copy all or any part of */ +/* this program for free distribution. The author's name */ +/* and this copyright notice must be included in any copy. */ +/* */ +/*************************************************************/ + + +#include "postgres.h" +#include "optimizer/geqo_random.h" +#include "optimizer/geqo_recombination.h" + +#if defined(ERX) + +static int gimme_edge(PlannerInfo *root, Gene gene1, Gene gene2, Edge *edge_table); +static void remove_gene(PlannerInfo *root, Gene gene, Edge edge, Edge *edge_table); +static Gene gimme_gene(PlannerInfo *root, Edge edge, Edge *edge_table); + +static Gene edge_failure(PlannerInfo *root, Gene *gene, int index, Edge *edge_table, int num_gene); + + +/* alloc_edge_table + * + * allocate memory for edge table + * + */ + +Edge * +alloc_edge_table(PlannerInfo *root, int num_gene) +{ + Edge *edge_table; + + /* + * palloc one extra location so that nodes numbered 1..n can be indexed + * directly; 0 will not be used + */ + + edge_table = (Edge *) palloc((num_gene + 1) * sizeof(Edge)); + + return edge_table; +} + +/* free_edge_table + * + * deallocate memory of edge table + * + */ +void +free_edge_table(PlannerInfo *root, Edge *edge_table) +{ + pfree(edge_table); +} + +/* gimme_edge_table + * + * fills a data structure which represents the set of explicit + * edges between points in the (2) input genes + * + * assumes circular tours and bidirectional edges + * + * gimme_edge() will set "shared" edges to negative values + * + * returns average number edges/city in range 2.0 - 4.0 + * where 2.0=homogeneous; 4.0=diverse + * + */ +float +gimme_edge_table(PlannerInfo *root, Gene *tour1, Gene *tour2, + int num_gene, Edge *edge_table) +{ + int i, + index1, + index2; + int edge_total; /* total number of unique edges in two genes */ + + /* at first clear the edge table's old data */ + for (i = 1; i <= num_gene; i++) + { + edge_table[i].total_edges = 0; + edge_table[i].unused_edges = 0; + } + + /* fill edge table with new data */ + + edge_total = 0; + + for (index1 = 0; index1 < num_gene; index1++) + { + /* + * presume the tour is circular, i.e. 1->2, 2->3, 3->1 this operation + * maps n back to 1 + */ + + index2 = (index1 + 1) % num_gene; + + /* + * edges are bidirectional, i.e. 1->2 is same as 2->1 call gimme_edge + * twice per edge + */ + + edge_total += gimme_edge(root, tour1[index1], tour1[index2], edge_table); + gimme_edge(root, tour1[index2], tour1[index1], edge_table); + + edge_total += gimme_edge(root, tour2[index1], tour2[index2], edge_table); + gimme_edge(root, tour2[index2], tour2[index1], edge_table); + } + + /* return average number of edges per index */ + return ((float) (edge_total * 2) / (float) num_gene); +} + +/* gimme_edge + * + * registers edge from city1 to city2 in input edge table + * + * no assumptions about directionality are made; + * therefore it is up to the calling routine to + * call gimme_edge twice to make a bi-directional edge + * between city1 and city2; + * uni-directional edges are possible as well (just call gimme_edge + * once with the direction from city1 to city2) + * + * returns 1 if edge was not already registered and was just added; + * 0 if edge was already registered and edge_table is unchanged + */ +static int +gimme_edge(PlannerInfo *root, Gene gene1, Gene gene2, Edge *edge_table) +{ + int i; + int edges; + int city1 = (int) gene1; + int city2 = (int) gene2; + + + /* check whether edge city1->city2 already exists */ + edges = edge_table[city1].total_edges; + + for (i = 0; i < edges; i++) + { + if ((Gene) Abs(edge_table[city1].edge_list[i]) == city2) + { + + /* mark shared edges as negative */ + edge_table[city1].edge_list[i] = 0 - city2; + + return 0; + } + } + + /* add city1->city2; */ + edge_table[city1].edge_list[edges] = city2; + + /* increment the number of edges from city1 */ + edge_table[city1].total_edges++; + edge_table[city1].unused_edges++; + + return 1; +} + +/* gimme_tour + * + * creates a new tour using edges from the edge table. + * priority is given to "shared" edges (i.e. edges which + * all parent genes possess and are marked as negative + * in the edge table.) + * + */ +int +gimme_tour(PlannerInfo *root, Edge *edge_table, Gene *new_gene, int num_gene) +{ + int i; + int edge_failures = 0; + + /* choose int between 1 and num_gene */ + new_gene[0] = (Gene) geqo_randint(root, num_gene, 1); + + for (i = 1; i < num_gene; i++) + { + /* + * as each point is entered into the tour, remove it from the edge + * table + */ + + remove_gene(root, new_gene[i - 1], edge_table[(int) new_gene[i - 1]], edge_table); + + /* find destination for the newly entered point */ + + if (edge_table[new_gene[i - 1]].unused_edges > 0) + new_gene[i] = gimme_gene(root, edge_table[(int) new_gene[i - 1]], edge_table); + + else + { /* cope with fault */ + edge_failures++; + + new_gene[i] = edge_failure(root, new_gene, i - 1, edge_table, num_gene); + } + + /* mark this node as incorporated */ + edge_table[(int) new_gene[i - 1]].unused_edges = -1; + } /* for (i=1; i<num_gene; i++) */ + + return edge_failures; +} + +/* remove_gene + * + * removes input gene from edge_table. + * input edge is used + * to identify deletion locations within edge table. + * + */ +static void +remove_gene(PlannerInfo *root, Gene gene, Edge edge, Edge *edge_table) +{ + int i, + j; + int possess_edge; + int genes_remaining; + + /* + * do for every gene known to have an edge to input gene (i.e. in + * edge_list for input edge) + */ + + for (i = 0; i < edge.unused_edges; i++) + { + possess_edge = (int) Abs(edge.edge_list[i]); + genes_remaining = edge_table[possess_edge].unused_edges; + + /* find the input gene in all edge_lists and delete it */ + for (j = 0; j < genes_remaining; j++) + { + + if ((Gene) Abs(edge_table[possess_edge].edge_list[j]) == gene) + { + + edge_table[possess_edge].unused_edges--; + + edge_table[possess_edge].edge_list[j] = + edge_table[possess_edge].edge_list[genes_remaining - 1]; + + break; + } + } + } +} + +/* gimme_gene + * + * priority is given to "shared" edges + * (i.e. edges which both genes possess) + * + */ +static Gene +gimme_gene(PlannerInfo *root, Edge edge, Edge *edge_table) +{ + int i; + Gene friend; + int minimum_edges; + int minimum_count = -1; + int rand_decision; + + /* + * no point has edges to more than 4 other points thus, this contrived + * minimum will be replaced + */ + + minimum_edges = 5; + + /* consider candidate destination points in edge list */ + + for (i = 0; i < edge.unused_edges; i++) + { + friend = (Gene) edge.edge_list[i]; + + /* + * give priority to shared edges that are negative; so return 'em + */ + + /* + * negative values are caught here so we need not worry about + * converting to absolute values + */ + if (friend < 0) + return (Gene) Abs(friend); + + + /* + * give priority to candidates with fewest remaining unused edges; + * find out what the minimum number of unused edges is + * (minimum_edges); if there is more than one candidate with the + * minimum number of unused edges keep count of this number + * (minimum_count); + */ + + /* + * The test for minimum_count can probably be removed at some point + * but comments should probably indicate exactly why it is guaranteed + * that the test will always succeed the first time around. If it can + * fail then the code is in error + */ + + + if (edge_table[(int) friend].unused_edges < minimum_edges) + { + minimum_edges = edge_table[(int) friend].unused_edges; + minimum_count = 1; + } + else if (minimum_count == -1) + elog(ERROR, "minimum_count not set"); + else if (edge_table[(int) friend].unused_edges == minimum_edges) + minimum_count++; + } /* for (i=0; i<edge.unused_edges; i++) */ + + + /* random decision of the possible candidates to use */ + rand_decision = geqo_randint(root, minimum_count - 1, 0); + + + for (i = 0; i < edge.unused_edges; i++) + { + friend = (Gene) edge.edge_list[i]; + + /* return the chosen candidate point */ + if (edge_table[(int) friend].unused_edges == minimum_edges) + { + minimum_count--; + + if (minimum_count == rand_decision) + return friend; + } + } + + /* ... should never be reached */ + elog(ERROR, "neither shared nor minimum number nor random edge found"); + return 0; /* to keep the compiler quiet */ +} + +/* edge_failure + * + * routine for handling edge failure + * + */ +static Gene +edge_failure(PlannerInfo *root, Gene *gene, int index, Edge *edge_table, int num_gene) +{ + int i; + Gene fail_gene = gene[index]; + int remaining_edges = 0; + int four_count = 0; + int rand_decision; + + + /* + * how many edges remain? how many gene with four total (initial) edges + * remain? + */ + + for (i = 1; i <= num_gene; i++) + { + if ((edge_table[i].unused_edges != -1) && (i != (int) fail_gene)) + { + remaining_edges++; + + if (edge_table[i].total_edges == 4) + four_count++; + } + } + + /* + * random decision of the gene with remaining edges and whose total_edges + * == 4 + */ + + if (four_count != 0) + { + + rand_decision = geqo_randint(root, four_count - 1, 0); + + for (i = 1; i <= num_gene; i++) + { + + if ((Gene) i != fail_gene && + edge_table[i].unused_edges != -1 && + edge_table[i].total_edges == 4) + { + + four_count--; + + if (rand_decision == four_count) + return (Gene) i; + } + } + + elog(LOG, "no edge found via random decision and total_edges == 4"); + } + else if (remaining_edges != 0) + { + /* random decision of the gene with remaining edges */ + rand_decision = geqo_randint(root, remaining_edges - 1, 0); + + for (i = 1; i <= num_gene; i++) + { + + if ((Gene) i != fail_gene && + edge_table[i].unused_edges != -1) + { + + remaining_edges--; + + if (rand_decision == remaining_edges) + return i; + } + } + + elog(LOG, "no edge found via random decision with remaining edges"); + } + + /* + * edge table seems to be empty; this happens sometimes on the last point + * due to the fact that the first point is removed from the table even + * though only one of its edges has been determined + */ + + else + { /* occurs only at the last point in the tour; + * simply look for the point which is not yet + * used */ + + for (i = 1; i <= num_gene; i++) + if (edge_table[i].unused_edges >= 0) + return (Gene) i; + + elog(LOG, "no edge found via looking for the last unused point"); + } + + + /* ... should never be reached */ + elog(ERROR, "no edge found"); + return 0; /* to keep the compiler quiet */ +} + +#endif /* defined(ERX) */ diff --git a/src/backend/optimizer/geqo/geqo_eval.c b/src/backend/optimizer/geqo/geqo_eval.c new file mode 100644 index 0000000..004481d --- /dev/null +++ b/src/backend/optimizer/geqo/geqo_eval.c @@ -0,0 +1,338 @@ +/*------------------------------------------------------------------------ + * + * geqo_eval.c + * Routines to evaluate query trees + * + * Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group + * Portions Copyright (c) 1994, Regents of the University of California + * + * src/backend/optimizer/geqo/geqo_eval.c + * + *------------------------------------------------------------------------- + */ + +/* contributed by: + =*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*= + * Martin Utesch * Institute of Automatic Control * + = = University of Mining and Technology = + * utesch@aut.tu-freiberg.de * Freiberg, Germany * + =*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*= + */ + +#include "postgres.h" + +#include <float.h> +#include <limits.h> +#include <math.h> + +#include "optimizer/geqo.h" +#include "optimizer/joininfo.h" +#include "optimizer/pathnode.h" +#include "optimizer/paths.h" +#include "utils/memutils.h" + + +/* A "clump" of already-joined relations within gimme_tree */ +typedef struct +{ + RelOptInfo *joinrel; /* joinrel for the set of relations */ + int size; /* number of input relations in clump */ +} Clump; + +static List *merge_clump(PlannerInfo *root, List *clumps, Clump *new_clump, + int num_gene, bool force); +static bool desirable_join(PlannerInfo *root, + RelOptInfo *outer_rel, RelOptInfo *inner_rel); + + +/* + * geqo_eval + * + * Returns cost of a query tree as an individual of the population. + * + * If no legal join order can be extracted from the proposed tour, + * returns DBL_MAX. + */ +Cost +geqo_eval(PlannerInfo *root, Gene *tour, int num_gene) +{ + MemoryContext mycontext; + MemoryContext oldcxt; + RelOptInfo *joinrel; + Cost fitness; + int savelength; + struct HTAB *savehash; + + /* + * Create a private memory context that will hold all temp storage + * allocated inside gimme_tree(). + * + * Since geqo_eval() will be called many times, we can't afford to let all + * that memory go unreclaimed until end of statement. Note we make the + * temp context a child of the planner's normal context, so that it will + * be freed even if we abort via ereport(ERROR). + */ + mycontext = AllocSetContextCreate(CurrentMemoryContext, + "GEQO", + ALLOCSET_DEFAULT_SIZES); + oldcxt = MemoryContextSwitchTo(mycontext); + + /* + * gimme_tree will add entries to root->join_rel_list, which may or may + * not already contain some entries. The newly added entries will be + * recycled by the MemoryContextDelete below, so we must ensure that the + * list is restored to its former state before exiting. We can do this by + * truncating the list to its original length. NOTE this assumes that any + * added entries are appended at the end! + * + * We also must take care not to mess up the outer join_rel_hash, if there + * is one. We can do this by just temporarily setting the link to NULL. + * (If we are dealing with enough join rels, which we very likely are, a + * new hash table will get built and used locally.) + * + * join_rel_level[] shouldn't be in use, so just Assert it isn't. + */ + savelength = list_length(root->join_rel_list); + savehash = root->join_rel_hash; + Assert(root->join_rel_level == NULL); + + root->join_rel_hash = NULL; + + /* construct the best path for the given combination of relations */ + joinrel = gimme_tree(root, tour, num_gene); + + /* + * compute fitness, if we found a valid join + * + * XXX geqo does not currently support optimization for partial result + * retrieval, nor do we take any cognizance of possible use of + * parameterized paths --- how to fix? + */ + if (joinrel) + { + Path *best_path = joinrel->cheapest_total_path; + + fitness = best_path->total_cost; + } + else + fitness = DBL_MAX; + + /* + * Restore join_rel_list to its former state, and put back original + * hashtable if any. + */ + root->join_rel_list = list_truncate(root->join_rel_list, + savelength); + root->join_rel_hash = savehash; + + /* release all the memory acquired within gimme_tree */ + MemoryContextSwitchTo(oldcxt); + MemoryContextDelete(mycontext); + + return fitness; +} + +/* + * gimme_tree + * Form planner estimates for a join tree constructed in the specified + * order. + * + * 'tour' is the proposed join order, of length 'num_gene' + * + * Returns a new join relation whose cheapest path is the best plan for + * this join order. NB: will return NULL if join order is invalid and + * we can't modify it into a valid order. + * + * The original implementation of this routine always joined in the specified + * order, and so could only build left-sided plans (and right-sided and + * mixtures, as a byproduct of the fact that make_join_rel() is symmetric). + * It could never produce a "bushy" plan. This had a couple of big problems, + * of which the worst was that there are situations involving join order + * restrictions where the only valid plans are bushy. + * + * The present implementation takes the given tour as a guideline, but + * postpones joins that are illegal or seem unsuitable according to some + * heuristic rules. This allows correct bushy plans to be generated at need, + * and as a nice side-effect it seems to materially improve the quality of the + * generated plans. Note however that since it's just a heuristic, it can + * still fail in some cases. (In particular, we might clump together + * relations that actually mustn't be joined yet due to LATERAL restrictions; + * since there's no provision for un-clumping, this must lead to failure.) + */ +RelOptInfo * +gimme_tree(PlannerInfo *root, Gene *tour, int num_gene) +{ + GeqoPrivateData *private = (GeqoPrivateData *) root->join_search_private; + List *clumps; + int rel_count; + + /* + * Sometimes, a relation can't yet be joined to others due to heuristics + * or actual semantic restrictions. We maintain a list of "clumps" of + * successfully joined relations, with larger clumps at the front. Each + * new relation from the tour is added to the first clump it can be joined + * to; if there is none then it becomes a new clump of its own. When we + * enlarge an existing clump we check to see if it can now be merged with + * any other clumps. After the tour is all scanned, we forget about the + * heuristics and try to forcibly join any remaining clumps. If we are + * unable to merge all the clumps into one, fail. + */ + clumps = NIL; + + for (rel_count = 0; rel_count < num_gene; rel_count++) + { + int cur_rel_index; + RelOptInfo *cur_rel; + Clump *cur_clump; + + /* Get the next input relation */ + cur_rel_index = (int) tour[rel_count]; + cur_rel = (RelOptInfo *) list_nth(private->initial_rels, + cur_rel_index - 1); + + /* Make it into a single-rel clump */ + cur_clump = (Clump *) palloc(sizeof(Clump)); + cur_clump->joinrel = cur_rel; + cur_clump->size = 1; + + /* Merge it into the clumps list, using only desirable joins */ + clumps = merge_clump(root, clumps, cur_clump, num_gene, false); + } + + if (list_length(clumps) > 1) + { + /* Force-join the remaining clumps in some legal order */ + List *fclumps; + ListCell *lc; + + fclumps = NIL; + foreach(lc, clumps) + { + Clump *clump = (Clump *) lfirst(lc); + + fclumps = merge_clump(root, fclumps, clump, num_gene, true); + } + clumps = fclumps; + } + + /* Did we succeed in forming a single join relation? */ + if (list_length(clumps) != 1) + return NULL; + + return ((Clump *) linitial(clumps))->joinrel; +} + +/* + * Merge a "clump" into the list of existing clumps for gimme_tree. + * + * We try to merge the clump into some existing clump, and repeat if + * successful. When no more merging is possible, insert the clump + * into the list, preserving the list ordering rule (namely, that + * clumps of larger size appear earlier). + * + * If force is true, merge anywhere a join is legal, even if it causes + * a cartesian join to be performed. When force is false, do only + * "desirable" joins. + */ +static List * +merge_clump(PlannerInfo *root, List *clumps, Clump *new_clump, int num_gene, + bool force) +{ + ListCell *lc; + int pos; + + /* Look for a clump that new_clump can join to */ + foreach(lc, clumps) + { + Clump *old_clump = (Clump *) lfirst(lc); + + if (force || + desirable_join(root, old_clump->joinrel, new_clump->joinrel)) + { + RelOptInfo *joinrel; + + /* + * Construct a RelOptInfo representing the join of these two input + * relations. Note that we expect the joinrel not to exist in + * root->join_rel_list yet, and so the paths constructed for it + * will only include the ones we want. + */ + joinrel = make_join_rel(root, + old_clump->joinrel, + new_clump->joinrel); + + /* Keep searching if join order is not valid */ + if (joinrel) + { + /* Create paths for partitionwise joins. */ + generate_partitionwise_join_paths(root, joinrel); + + /* + * Except for the topmost scan/join rel, consider gathering + * partial paths. We'll do the same for the topmost scan/join + * rel once we know the final targetlist (see + * grouping_planner). + */ + if (!bms_equal(joinrel->relids, root->all_baserels)) + generate_useful_gather_paths(root, joinrel, false); + + /* Find and save the cheapest paths for this joinrel */ + set_cheapest(joinrel); + + /* Absorb new clump into old */ + old_clump->joinrel = joinrel; + old_clump->size += new_clump->size; + pfree(new_clump); + + /* Remove old_clump from list */ + clumps = foreach_delete_current(clumps, lc); + + /* + * Recursively try to merge the enlarged old_clump with + * others. When no further merge is possible, we'll reinsert + * it into the list. + */ + return merge_clump(root, clumps, old_clump, num_gene, force); + } + } + } + + /* + * No merging is possible, so add new_clump as an independent clump, in + * proper order according to size. We can be fast for the common case + * where it has size 1 --- it should always go at the end. + */ + if (clumps == NIL || new_clump->size == 1) + return lappend(clumps, new_clump); + + /* Else search for the place to insert it */ + for (pos = 0; pos < list_length(clumps); pos++) + { + Clump *old_clump = (Clump *) list_nth(clumps, pos); + + if (new_clump->size > old_clump->size) + break; /* new_clump belongs before old_clump */ + } + clumps = list_insert_nth(clumps, pos, new_clump); + + return clumps; +} + +/* + * Heuristics for gimme_tree: do we want to join these two relations? + */ +static bool +desirable_join(PlannerInfo *root, + RelOptInfo *outer_rel, RelOptInfo *inner_rel) +{ + /* + * Join if there is an applicable join clause, or if there is a join order + * restriction forcing these rels to be joined. + */ + if (have_relevant_joinclause(root, outer_rel, inner_rel) || + have_join_order_restriction(root, outer_rel, inner_rel)) + return true; + + /* Otherwise postpone the join till later. */ + return false; +} diff --git a/src/backend/optimizer/geqo/geqo_main.c b/src/backend/optimizer/geqo/geqo_main.c new file mode 100644 index 0000000..68ed743 --- /dev/null +++ b/src/backend/optimizer/geqo/geqo_main.c @@ -0,0 +1,353 @@ +/*------------------------------------------------------------------------ + * + * geqo_main.c + * solution to the query optimization problem + * by means of a Genetic Algorithm (GA) + * + * Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group + * Portions Copyright (c) 1994, Regents of the University of California + * + * src/backend/optimizer/geqo/geqo_main.c + * + *------------------------------------------------------------------------- + */ + +/* contributed by: + =*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*= + * Martin Utesch * Institute of Automatic Control * + = = University of Mining and Technology = + * utesch@aut.tu-freiberg.de * Freiberg, Germany * + =*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*= + */ + +/* -- parts of this are adapted from D. Whitley's Genitor algorithm -- */ + +#include "postgres.h" + +#include <math.h> + +#include "optimizer/geqo_misc.h" +#include "optimizer/geqo_mutation.h" +#include "optimizer/geqo_pool.h" +#include "optimizer/geqo_random.h" +#include "optimizer/geqo_selection.h" + + +/* + * Configuration options + */ +int Geqo_effort; +int Geqo_pool_size; +int Geqo_generations; +double Geqo_selection_bias; +double Geqo_seed; + + +static int gimme_pool_size(int nr_rel); +static int gimme_number_generations(int pool_size); + +/* complain if no recombination mechanism is #define'd */ +#if !defined(ERX) && \ + !defined(PMX) && \ + !defined(CX) && \ + !defined(PX) && \ + !defined(OX1) && \ + !defined(OX2) +#error "must choose one GEQO recombination mechanism in geqo.h" +#endif + + +/* + * geqo + * solution of the query optimization problem + * similar to a constrained Traveling Salesman Problem (TSP) + */ + +RelOptInfo * +geqo(PlannerInfo *root, int number_of_rels, List *initial_rels) +{ + GeqoPrivateData private; + int generation; + Chromosome *momma; + Chromosome *daddy; + Chromosome *kid; + Pool *pool; + int pool_size, + number_generations; + +#ifdef GEQO_DEBUG + int status_interval; +#endif + Gene *best_tour; + RelOptInfo *best_rel; + +#if defined(ERX) + Edge *edge_table; /* list of edges */ + int edge_failures = 0; +#endif +#if defined(CX) || defined(PX) || defined(OX1) || defined(OX2) + City *city_table; /* list of cities */ +#endif +#if defined(CX) + int cycle_diffs = 0; + int mutations = 0; +#endif + +/* set up private information */ + root->join_search_private = (void *) &private; + private.initial_rels = initial_rels; + +/* initialize private number generator */ + geqo_set_seed(root, Geqo_seed); + +/* set GA parameters */ + pool_size = gimme_pool_size(number_of_rels); + number_generations = gimme_number_generations(pool_size); +#ifdef GEQO_DEBUG + status_interval = 10; +#endif + +/* allocate genetic pool memory */ + pool = alloc_pool(root, pool_size, number_of_rels); + +/* random initialization of the pool */ + random_init_pool(root, pool); + +/* sort the pool according to cheapest path as fitness */ + sort_pool(root, pool); /* we have to do it only one time, since all + * kids replace the worst individuals in + * future (-> geqo_pool.c:spread_chromo ) */ + +#ifdef GEQO_DEBUG + elog(DEBUG1, "GEQO selected %d pool entries, best %.2f, worst %.2f", + pool_size, + pool->data[0].worth, + pool->data[pool_size - 1].worth); +#endif + +/* allocate chromosome momma and daddy memory */ + momma = alloc_chromo(root, pool->string_length); + daddy = alloc_chromo(root, pool->string_length); + +#if defined (ERX) +#ifdef GEQO_DEBUG + elog(DEBUG2, "using edge recombination crossover [ERX]"); +#endif +/* allocate edge table memory */ + edge_table = alloc_edge_table(root, pool->string_length); +#elif defined(PMX) +#ifdef GEQO_DEBUG + elog(DEBUG2, "using partially matched crossover [PMX]"); +#endif +/* allocate chromosome kid memory */ + kid = alloc_chromo(root, pool->string_length); +#elif defined(CX) +#ifdef GEQO_DEBUG + elog(DEBUG2, "using cycle crossover [CX]"); +#endif +/* allocate city table memory */ + kid = alloc_chromo(root, pool->string_length); + city_table = alloc_city_table(root, pool->string_length); +#elif defined(PX) +#ifdef GEQO_DEBUG + elog(DEBUG2, "using position crossover [PX]"); +#endif +/* allocate city table memory */ + kid = alloc_chromo(root, pool->string_length); + city_table = alloc_city_table(root, pool->string_length); +#elif defined(OX1) +#ifdef GEQO_DEBUG + elog(DEBUG2, "using order crossover [OX1]"); +#endif +/* allocate city table memory */ + kid = alloc_chromo(root, pool->string_length); + city_table = alloc_city_table(root, pool->string_length); +#elif defined(OX2) +#ifdef GEQO_DEBUG + elog(DEBUG2, "using order crossover [OX2]"); +#endif +/* allocate city table memory */ + kid = alloc_chromo(root, pool->string_length); + city_table = alloc_city_table(root, pool->string_length); +#endif + + +/* my pain main part: */ +/* iterative optimization */ + + for (generation = 0; generation < number_generations; generation++) + { + /* SELECTION: using linear bias function */ + geqo_selection(root, momma, daddy, pool, Geqo_selection_bias); + +#if defined (ERX) + /* EDGE RECOMBINATION CROSSOVER */ + gimme_edge_table(root, momma->string, daddy->string, pool->string_length, edge_table); + + kid = momma; + + /* are there any edge failures ? */ + edge_failures += gimme_tour(root, edge_table, kid->string, pool->string_length); +#elif defined(PMX) + /* PARTIALLY MATCHED CROSSOVER */ + pmx(root, momma->string, daddy->string, kid->string, pool->string_length); +#elif defined(CX) + /* CYCLE CROSSOVER */ + cycle_diffs = cx(root, momma->string, daddy->string, kid->string, pool->string_length, city_table); + /* mutate the child */ + if (cycle_diffs == 0) + { + mutations++; + geqo_mutation(root, kid->string, pool->string_length); + } +#elif defined(PX) + /* POSITION CROSSOVER */ + px(root, momma->string, daddy->string, kid->string, pool->string_length, city_table); +#elif defined(OX1) + /* ORDER CROSSOVER */ + ox1(root, momma->string, daddy->string, kid->string, pool->string_length, city_table); +#elif defined(OX2) + /* ORDER CROSSOVER */ + ox2(root, momma->string, daddy->string, kid->string, pool->string_length, city_table); +#endif + + + /* EVALUATE FITNESS */ + kid->worth = geqo_eval(root, kid->string, pool->string_length); + + /* push the kid into the wilderness of life according to its worth */ + spread_chromo(root, kid, pool); + + +#ifdef GEQO_DEBUG + if (status_interval && !(generation % status_interval)) + print_gen(stdout, pool, generation); +#endif + + } + + +#if defined(ERX) +#if defined(GEQO_DEBUG) + if (edge_failures != 0) + elog(LOG, "[GEQO] failures: %d, average: %d", + edge_failures, (int) number_generations / edge_failures); + else + elog(LOG, "[GEQO] no edge failures detected"); +#else + /* suppress variable-set-but-not-used warnings from some compilers */ + (void) edge_failures; +#endif +#endif + +#if defined(CX) && defined(GEQO_DEBUG) + if (mutations != 0) + elog(LOG, "[GEQO] mutations: %d, generations: %d", + mutations, number_generations); + else + elog(LOG, "[GEQO] no mutations processed"); +#endif + +#ifdef GEQO_DEBUG + print_pool(stdout, pool, 0, pool_size - 1); +#endif + +#ifdef GEQO_DEBUG + elog(DEBUG1, "GEQO best is %.2f after %d generations", + pool->data[0].worth, number_generations); +#endif + + + /* + * got the cheapest query tree processed by geqo; first element of the + * population indicates the best query tree + */ + best_tour = (Gene *) pool->data[0].string; + + best_rel = gimme_tree(root, best_tour, pool->string_length); + + if (best_rel == NULL) + elog(ERROR, "geqo failed to make a valid plan"); + + /* DBG: show the query plan */ +#ifdef NOT_USED + print_plan(best_plan, root); +#endif + + /* ... free memory stuff */ + free_chromo(root, momma); + free_chromo(root, daddy); + +#if defined (ERX) + free_edge_table(root, edge_table); +#elif defined(PMX) + free_chromo(root, kid); +#elif defined(CX) + free_chromo(root, kid); + free_city_table(root, city_table); +#elif defined(PX) + free_chromo(root, kid); + free_city_table(root, city_table); +#elif defined(OX1) + free_chromo(root, kid); + free_city_table(root, city_table); +#elif defined(OX2) + free_chromo(root, kid); + free_city_table(root, city_table); +#endif + + free_pool(root, pool); + + /* ... clear root pointer to our private storage */ + root->join_search_private = NULL; + + return best_rel; +} + + +/* + * Return either configured pool size or a good default + * + * The default is based on query size (no. of relations) = 2^(QS+1), + * but constrained to a range based on the effort value. + */ +static int +gimme_pool_size(int nr_rel) +{ + double size; + int minsize; + int maxsize; + + /* Legal pool size *must* be at least 2, so ignore attempt to select 1 */ + if (Geqo_pool_size >= 2) + return Geqo_pool_size; + + size = pow(2.0, nr_rel + 1.0); + + maxsize = 50 * Geqo_effort; /* 50 to 500 individuals */ + if (size > maxsize) + return maxsize; + + minsize = 10 * Geqo_effort; /* 10 to 100 individuals */ + if (size < minsize) + return minsize; + + return (int) ceil(size); +} + + +/* + * Return either configured number of generations or a good default + * + * The default is the same as the pool size, which allows us to be + * sure that less-fit individuals get pushed out of the breeding + * population before the run finishes. + */ +static int +gimme_number_generations(int pool_size) +{ + if (Geqo_generations > 0) + return Geqo_generations; + + return pool_size; +} diff --git a/src/backend/optimizer/geqo/geqo_misc.c b/src/backend/optimizer/geqo/geqo_misc.c new file mode 100644 index 0000000..890ac36 --- /dev/null +++ b/src/backend/optimizer/geqo/geqo_misc.c @@ -0,0 +1,132 @@ +/*------------------------------------------------------------------------ + * + * geqo_misc.c + * misc. printout and debug stuff + * + * Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group + * Portions Copyright (c) 1994, Regents of the University of California + * + * src/backend/optimizer/geqo/geqo_misc.c + * + *------------------------------------------------------------------------- + */ + +/* contributed by: + =*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*= + * Martin Utesch * Institute of Automatic Control * + = = University of Mining and Technology = + * utesch@aut.tu-freiberg.de * Freiberg, Germany * + =*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*= + */ + +#include "postgres.h" + +#include "optimizer/geqo_misc.h" + + +#ifdef GEQO_DEBUG + + +/* + * avg_pool + */ +static double +avg_pool(Pool *pool) +{ + int i; + double cumulative = 0.0; + + if (pool->size <= 0) + elog(ERROR, "pool_size is zero"); + + /* + * Since the pool may contain multiple occurrences of DBL_MAX, divide by + * pool->size before summing, not after, to avoid overflow. This loses a + * little in speed and accuracy, but this routine is only used for debug + * printouts, so we don't care that much. + */ + for (i = 0; i < pool->size; i++) + cumulative += pool->data[i].worth / pool->size; + + return cumulative; +} + +/* print_pool + */ +void +print_pool(FILE *fp, Pool *pool, int start, int stop) +{ + int i, + j; + + /* be extra careful that start and stop are valid inputs */ + + if (start < 0) + start = 0; + if (stop > pool->size) + stop = pool->size; + + if (start + stop > pool->size) + { + start = 0; + stop = pool->size; + } + + for (i = start; i < stop; i++) + { + fprintf(fp, "%d)\t", i); + for (j = 0; j < pool->string_length; j++) + fprintf(fp, "%d ", pool->data[i].string[j]); + fprintf(fp, "%g\n", pool->data[i].worth); + } + + fflush(fp); +} + +/* print_gen + * + * printout for chromosome: best, worst, mean, average + */ +void +print_gen(FILE *fp, Pool *pool, int generation) +{ + int lowest; + + /* Get index to lowest ranking gene in population. */ + /* Use 2nd to last since last is buffer. */ + lowest = pool->size > 1 ? pool->size - 2 : 0; + + fprintf(fp, + "%5d | Best: %g Worst: %g Mean: %g Avg: %g\n", + generation, + pool->data[0].worth, + pool->data[lowest].worth, + pool->data[pool->size / 2].worth, + avg_pool(pool)); + + fflush(fp); +} + + +void +print_edge_table(FILE *fp, Edge *edge_table, int num_gene) +{ + int i, + j; + + fprintf(fp, "\nEDGE TABLE\n"); + + for (i = 1; i <= num_gene; i++) + { + fprintf(fp, "%d :", i); + for (j = 0; j < edge_table[i].unused_edges; j++) + fprintf(fp, " %d", edge_table[i].edge_list[j]); + fprintf(fp, "\n"); + } + + fprintf(fp, "\n"); + + fflush(fp); +} + +#endif /* GEQO_DEBUG */ diff --git a/src/backend/optimizer/geqo/geqo_mutation.c b/src/backend/optimizer/geqo/geqo_mutation.c new file mode 100644 index 0000000..2af0295 --- /dev/null +++ b/src/backend/optimizer/geqo/geqo_mutation.c @@ -0,0 +1,66 @@ +/*------------------------------------------------------------------------ +* +* geqo_mutation.c +* +* TSP mutation routines +* +* src/backend/optimizer/geqo/geqo_mutation.c +* +*------------------------------------------------------------------------- +*/ + +/* contributed by: + =*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*= + * Martin Utesch * Institute of Automatic Control * + = = University of Mining and Technology = + * utesch@aut.tu-freiberg.de * Freiberg, Germany * + =*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*= + */ + +/* this is adopted from Genitor : */ +/*************************************************************/ +/* */ +/* Copyright (c) 1990 */ +/* Darrell L. Whitley */ +/* Computer Science Department */ +/* Colorado State University */ +/* */ +/* Permission is hereby granted to copy all or any part of */ +/* this program for free distribution. The author's name */ +/* and this copyright notice must be included in any copy. */ +/* */ +/*************************************************************/ + +#include "postgres.h" +#include "optimizer/geqo_mutation.h" +#include "optimizer/geqo_random.h" + +#if defined(CX) /* currently used only in CX mode */ + +void +geqo_mutation(PlannerInfo *root, Gene *tour, int num_gene) +{ + int swap1; + int swap2; + int num_swaps = geqo_randint(root, num_gene / 3, 0); + Gene temp; + + + while (num_swaps > 0) + { + swap1 = geqo_randint(root, num_gene - 1, 0); + swap2 = geqo_randint(root, num_gene - 1, 0); + + while (swap1 == swap2) + swap2 = geqo_randint(root, num_gene - 1, 0); + + temp = tour[swap1]; + tour[swap1] = tour[swap2]; + tour[swap2] = temp; + + + num_swaps -= 1; + } +} + +#endif /* defined(CX) */ diff --git a/src/backend/optimizer/geqo/geqo_ox1.c b/src/backend/optimizer/geqo/geqo_ox1.c new file mode 100644 index 0000000..ede7abb --- /dev/null +++ b/src/backend/optimizer/geqo/geqo_ox1.c @@ -0,0 +1,94 @@ +/*------------------------------------------------------------------------ +* +* geqo_ox1.c +* +* order crossover [OX] routines; +* OX1 operator according to Davis +* (Proc Int'l Joint Conf on AI) +* +* src/backend/optimizer/geqo/geqo_ox1.c +* +*------------------------------------------------------------------------- +*/ + +/* contributed by: + =*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*= + * Martin Utesch * Institute of Automatic Control * + = = University of Mining and Technology = + * utesch@aut.tu-freiberg.de * Freiberg, Germany * + =*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*= + */ + +/* the ox algorithm is adopted from Genitor : */ +/*************************************************************/ +/* */ +/* Copyright (c) 1990 */ +/* Darrell L. Whitley */ +/* Computer Science Department */ +/* Colorado State University */ +/* */ +/* Permission is hereby granted to copy all or any part of */ +/* this program for free distribution. The author's name */ +/* and this copyright notice must be included in any copy. */ +/* */ +/*************************************************************/ + +#include "postgres.h" +#include "optimizer/geqo_random.h" +#include "optimizer/geqo_recombination.h" + +#if defined(OX1) + +/* ox1 + * + * position crossover + */ +void +ox1(PlannerInfo *root, Gene *tour1, Gene *tour2, Gene *offspring, int num_gene, + City * city_table) +{ + int left, + right, + k, + p, + temp; + + /* initialize city table */ + for (k = 1; k <= num_gene; k++) + city_table[k].used = 0; + + /* select portion to copy from tour1 */ + left = geqo_randint(root, num_gene - 1, 0); + right = geqo_randint(root, num_gene - 1, 0); + + if (left > right) + { + temp = left; + left = right; + right = temp; + } + + /* copy portion from tour1 to offspring */ + for (k = left; k <= right; k++) + { + offspring[k] = tour1[k]; + city_table[(int) tour1[k]].used = 1; + } + + k = (right + 1) % num_gene; /* index into offspring */ + p = k; /* index into tour2 */ + + /* copy stuff from tour2 to offspring */ + while (k != left) + { + if (!city_table[(int) tour2[p]].used) + { + offspring[k] = tour2[p]; + k = (k + 1) % num_gene; + city_table[(int) tour2[p]].used = 1; + } + p = (p + 1) % num_gene; /* increment tour2-index */ + } +} + +#endif /* defined(OX1) */ diff --git a/src/backend/optimizer/geqo/geqo_ox2.c b/src/backend/optimizer/geqo/geqo_ox2.c new file mode 100644 index 0000000..080dbc0 --- /dev/null +++ b/src/backend/optimizer/geqo/geqo_ox2.c @@ -0,0 +1,111 @@ +/*------------------------------------------------------------------------ +* +* geqo_ox2.c +* +* order crossover [OX] routines; +* OX2 operator according to Syswerda +* (The Genetic Algorithms Handbook, ed L Davis) +* +* src/backend/optimizer/geqo/geqo_ox2.c +* +*------------------------------------------------------------------------- +*/ + +/* contributed by: + =*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*= + * Martin Utesch * Institute of Automatic Control * + = = University of Mining and Technology = + * utesch@aut.tu-freiberg.de * Freiberg, Germany * + =*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*= + */ + +/* the ox algorithm is adopted from Genitor : */ +/*************************************************************/ +/* */ +/* Copyright (c) 1990 */ +/* Darrell L. Whitley */ +/* Computer Science Department */ +/* Colorado State University */ +/* */ +/* Permission is hereby granted to copy all or any part of */ +/* this program for free distribution. The author's name */ +/* and this copyright notice must be included in any copy. */ +/* */ +/*************************************************************/ + +#include "postgres.h" +#include "optimizer/geqo_random.h" +#include "optimizer/geqo_recombination.h" + +#if defined(OX2) + +/* ox2 + * + * position crossover + */ +void +ox2(PlannerInfo *root, Gene *tour1, Gene *tour2, Gene *offspring, int num_gene, City * city_table) +{ + int k, + j, + count, + pos, + select, + num_positions; + + /* initialize city table */ + for (k = 1; k <= num_gene; k++) + { + city_table[k].used = 0; + city_table[k - 1].select_list = -1; + } + + /* determine the number of positions to be inherited from tour1 */ + num_positions = geqo_randint(root, 2 * num_gene / 3, num_gene / 3); + + /* make a list of selected cities */ + for (k = 0; k < num_positions; k++) + { + pos = geqo_randint(root, num_gene - 1, 0); + city_table[pos].select_list = (int) tour1[pos]; + city_table[(int) tour1[pos]].used = 1; /* mark used */ + } + + + count = 0; + k = 0; + + /* consolidate the select list to adjacent positions */ + while (count < num_positions) + { + if (city_table[k].select_list == -1) + { + j = k + 1; + while ((city_table[j].select_list == -1) && (j < num_gene)) + j++; + + city_table[k].select_list = city_table[j].select_list; + city_table[j].select_list = -1; + count++; + } + else + count++; + k++; + } + + select = 0; + + for (k = 0; k < num_gene; k++) + { + if (city_table[(int) tour2[k]].used) + { + offspring[k] = (Gene) city_table[select].select_list; + select++; /* next city in the select list */ + } + else + /* city isn't used yet, so inherit from tour2 */ + offspring[k] = tour2[k]; + } +} + +#endif /* defined(OX2) */ diff --git a/src/backend/optimizer/geqo/geqo_pmx.c b/src/backend/optimizer/geqo/geqo_pmx.c new file mode 100644 index 0000000..e44fd0b --- /dev/null +++ b/src/backend/optimizer/geqo/geqo_pmx.c @@ -0,0 +1,218 @@ +/*------------------------------------------------------------------------ +* +* geqo_pmx.c +* +* partially matched crossover [PMX] routines; +* PMX operator according to Goldberg & Lingle +* (Proc Int'l Conf on GA's) +* +* src/backend/optimizer/geqo/geqo_pmx.c +* +*------------------------------------------------------------------------- +*/ + +/* contributed by: + =*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*= + * Martin Utesch * Institute of Automatic Control * + = = University of Mining and Technology = + * utesch@aut.tu-freiberg.de * Freiberg, Germany * + =*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*= + */ + +/* the pmx algorithm is adopted from Genitor : */ +/*************************************************************/ +/* */ +/* Copyright (c) 1990 */ +/* Darrell L. Whitley */ +/* Computer Science Department */ +/* Colorado State University */ +/* */ +/* Permission is hereby granted to copy all or any part of */ +/* this program for free distribution. The author's name */ +/* and this copyright notice must be included in any copy. */ +/* */ +/*************************************************************/ + +#include "postgres.h" +#include "optimizer/geqo_random.h" +#include "optimizer/geqo_recombination.h" + +#if defined(PMX) + +/* pmx + * + * partially matched crossover + */ +void +pmx(PlannerInfo *root, Gene *tour1, Gene *tour2, Gene *offspring, int num_gene) +{ + int *failed = (int *) palloc((num_gene + 1) * sizeof(int)); + int *from = (int *) palloc((num_gene + 1) * sizeof(int)); + int *indx = (int *) palloc((num_gene + 1) * sizeof(int)); + int *check_list = (int *) palloc((num_gene + 1) * sizeof(int)); + + int left, + right, + temp, + i, + j, + k; + int mx_fail, + found, + mx_hold; + + +/* no mutation so start up the pmx replacement algorithm */ +/* initialize failed[], from[], check_list[] */ + for (k = 0; k < num_gene; k++) + { + failed[k] = -1; + from[k] = -1; + check_list[k + 1] = 0; + } + +/* locate crossover points */ + left = geqo_randint(root, num_gene - 1, 0); + right = geqo_randint(root, num_gene - 1, 0); + + if (left > right) + { + temp = left; + left = right; + right = temp; + } + + +/* copy tour2 into offspring */ + for (k = 0; k < num_gene; k++) + { + offspring[k] = tour2[k]; + from[k] = DAD; + check_list[tour2[k]]++; + } + +/* copy tour1 into offspring */ + for (k = left; k <= right; k++) + { + check_list[offspring[k]]--; + offspring[k] = tour1[k]; + from[k] = MOM; + check_list[tour1[k]]++; + } + + +/* pmx main part */ + + mx_fail = 0; + +/* STEP 1 */ + + for (k = left; k <= right; k++) + { /* for all elements in the tour1-2 */ + + if (tour1[k] == tour2[k]) + found = 1; /* find match in tour2 */ + + else + { + found = 0; /* substitute elements */ + + j = 0; + while (!(found) && (j < num_gene)) + { + if ((offspring[j] == tour1[k]) && (from[j] == DAD)) + { + + check_list[offspring[j]]--; + offspring[j] = tour2[k]; + found = 1; + check_list[tour2[k]]++; + } + + j++; + } + } + + if (!(found)) + { /* failed to replace gene */ + failed[mx_fail] = (int) tour1[k]; + indx[mx_fail] = k; + mx_fail++; + } + } /* ... for */ + + +/* STEP 2 */ + + /* see if any genes could not be replaced */ + if (mx_fail > 0) + { + mx_hold = mx_fail; + + for (k = 0; k < mx_hold; k++) + { + found = 0; + + j = 0; + while (!(found) && (j < num_gene)) + { + + if ((failed[k] == (int) offspring[j]) && (from[j] == DAD)) + { + check_list[offspring[j]]--; + offspring[j] = tour2[indx[k]]; + check_list[tour2[indx[k]]]++; + + found = 1; + failed[k] = -1; + mx_fail--; + } + + j++; + } + } /* ... for */ + } /* ... if */ + + +/* STEP 3 */ + + for (k = 1; k <= num_gene; k++) + { + + if (check_list[k] > 1) + { + i = 0; + + while (i < num_gene) + { + if ((offspring[i] == (Gene) k) && (from[i] == DAD)) + { + j = 1; + + while (j <= num_gene) + { + if (check_list[j] == 0) + { + offspring[i] = (Gene) j; + check_list[k]--; + check_list[j]++; + i = num_gene + 1; + j = i; + } + + j++; + } + } /* ... if */ + + i++; + } /* end while */ + } + } /* ... for */ + + pfree(failed); + pfree(from); + pfree(indx); + pfree(check_list); +} + +#endif /* defined(PMX) */ diff --git a/src/backend/optimizer/geqo/geqo_pool.c b/src/backend/optimizer/geqo/geqo_pool.c new file mode 100644 index 0000000..0dfef1f --- /dev/null +++ b/src/backend/optimizer/geqo/geqo_pool.c @@ -0,0 +1,265 @@ +/*------------------------------------------------------------------------ + * + * geqo_pool.c + * Genetic Algorithm (GA) pool stuff + * + * Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group + * Portions Copyright (c) 1994, Regents of the University of California + * + * src/backend/optimizer/geqo/geqo_pool.c + * + *------------------------------------------------------------------------- + */ + +/* contributed by: + =*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*= + * Martin Utesch * Institute of Automatic Control * + = = University of Mining and Technology = + * utesch@aut.tu-freiberg.de * Freiberg, Germany * + =*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*= + */ + +/* -- parts of this are adapted from D. Whitley's Genitor algorithm -- */ + +#include "postgres.h" + +#include <float.h> +#include <limits.h> +#include <math.h> + +#include "optimizer/geqo_copy.h" +#include "optimizer/geqo_pool.h" +#include "optimizer/geqo_recombination.h" + + +static int compare(const void *arg1, const void *arg2); + +/* + * alloc_pool + * allocates memory for GA pool + */ +Pool * +alloc_pool(PlannerInfo *root, int pool_size, int string_length) +{ + Pool *new_pool; + Chromosome *chromo; + int i; + + /* pool */ + new_pool = (Pool *) palloc(sizeof(Pool)); + new_pool->size = (int) pool_size; + new_pool->string_length = (int) string_length; + + /* all chromosome */ + new_pool->data = (Chromosome *) palloc(pool_size * sizeof(Chromosome)); + + /* all gene */ + chromo = (Chromosome *) new_pool->data; /* vector of all chromos */ + for (i = 0; i < pool_size; i++) + chromo[i].string = palloc((string_length + 1) * sizeof(Gene)); + + return new_pool; +} + +/* + * free_pool + * deallocates memory for GA pool + */ +void +free_pool(PlannerInfo *root, Pool *pool) +{ + Chromosome *chromo; + int i; + + /* all gene */ + chromo = (Chromosome *) pool->data; /* vector of all chromos */ + for (i = 0; i < pool->size; i++) + pfree(chromo[i].string); + + /* all chromosome */ + pfree(pool->data); + + /* pool */ + pfree(pool); +} + +/* + * random_init_pool + * initialize genetic pool + */ +void +random_init_pool(PlannerInfo *root, Pool *pool) +{ + Chromosome *chromo = (Chromosome *) pool->data; + int i; + int bad = 0; + + /* + * We immediately discard any invalid individuals (those that geqo_eval + * returns DBL_MAX for), thereby not wasting pool space on them. + * + * If we fail to make any valid individuals after 10000 tries, give up; + * this probably means something is broken, and we shouldn't just let + * ourselves get stuck in an infinite loop. + */ + i = 0; + while (i < pool->size) + { + init_tour(root, chromo[i].string, pool->string_length); + pool->data[i].worth = geqo_eval(root, chromo[i].string, + pool->string_length); + if (pool->data[i].worth < DBL_MAX) + i++; + else + { + bad++; + if (i == 0 && bad >= 10000) + elog(ERROR, "geqo failed to make a valid plan"); + } + } + +#ifdef GEQO_DEBUG + if (bad > 0) + elog(DEBUG1, "%d invalid tours found while selecting %d pool entries", + bad, pool->size); +#endif +} + +/* + * sort_pool + * sorts input pool according to worth, from smallest to largest + * + * maybe you have to change compare() for different ordering ... + */ +void +sort_pool(PlannerInfo *root, Pool *pool) +{ + qsort(pool->data, pool->size, sizeof(Chromosome), compare); +} + +/* + * compare + * qsort comparison function for sort_pool + */ +static int +compare(const void *arg1, const void *arg2) +{ + const Chromosome *chromo1 = (const Chromosome *) arg1; + const Chromosome *chromo2 = (const Chromosome *) arg2; + + if (chromo1->worth == chromo2->worth) + return 0; + else if (chromo1->worth > chromo2->worth) + return 1; + else + return -1; +} + +/* alloc_chromo + * allocates a chromosome and string space + */ +Chromosome * +alloc_chromo(PlannerInfo *root, int string_length) +{ + Chromosome *chromo; + + chromo = (Chromosome *) palloc(sizeof(Chromosome)); + chromo->string = (Gene *) palloc((string_length + 1) * sizeof(Gene)); + + return chromo; +} + +/* free_chromo + * deallocates a chromosome and string space + */ +void +free_chromo(PlannerInfo *root, Chromosome *chromo) +{ + pfree(chromo->string); + pfree(chromo); +} + +/* spread_chromo + * inserts a new chromosome into the pool, displacing worst gene in pool + * assumes best->worst = smallest->largest + */ +void +spread_chromo(PlannerInfo *root, Chromosome *chromo, Pool *pool) +{ + int top, + mid, + bot; + int i, + index; + Chromosome swap_chromo, + tmp_chromo; + + /* new chromo is so bad we can't use it */ + if (chromo->worth > pool->data[pool->size - 1].worth) + return; + + /* do a binary search to find the index of the new chromo */ + + top = 0; + mid = pool->size / 2; + bot = pool->size - 1; + index = -1; + + while (index == -1) + { + /* these 4 cases find a new location */ + + if (chromo->worth <= pool->data[top].worth) + index = top; + else if (chromo->worth == pool->data[mid].worth) + index = mid; + else if (chromo->worth == pool->data[bot].worth) + index = bot; + else if (bot - top <= 1) + index = bot; + + + /* + * these 2 cases move the search indices since a new location has not + * yet been found. + */ + + else if (chromo->worth < pool->data[mid].worth) + { + bot = mid; + mid = top + ((bot - top) / 2); + } + else + { /* (chromo->worth > pool->data[mid].worth) */ + top = mid; + mid = top + ((bot - top) / 2); + } + } /* ... while */ + + /* now we have index for chromo */ + + /* + * move every gene from index on down one position to make room for chromo + */ + + /* + * copy new gene into pool storage; always replace worst gene in pool + */ + + geqo_copy(root, &pool->data[pool->size - 1], chromo, pool->string_length); + + swap_chromo.string = pool->data[pool->size - 1].string; + swap_chromo.worth = pool->data[pool->size - 1].worth; + + for (i = index; i < pool->size; i++) + { + tmp_chromo.string = pool->data[i].string; + tmp_chromo.worth = pool->data[i].worth; + + pool->data[i].string = swap_chromo.string; + pool->data[i].worth = swap_chromo.worth; + + swap_chromo.string = tmp_chromo.string; + swap_chromo.worth = tmp_chromo.worth; + } +} diff --git a/src/backend/optimizer/geqo/geqo_px.c b/src/backend/optimizer/geqo/geqo_px.c new file mode 100644 index 0000000..914296b --- /dev/null +++ b/src/backend/optimizer/geqo/geqo_px.c @@ -0,0 +1,107 @@ +/*------------------------------------------------------------------------ +* +* geqo_px.c +* +* position crossover [PX] routines; +* PX operator according to Syswerda +* (The Genetic Algorithms Handbook, L Davis, ed) +* +* src/backend/optimizer/geqo/geqo_px.c +* +*------------------------------------------------------------------------- +*/ + +/* contributed by: + =*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*= + * Martin Utesch * Institute of Automatic Control * + = = University of Mining and Technology = + * utesch@aut.tu-freiberg.de * Freiberg, Germany * + =*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*= + */ + +/* the px algorithm is adopted from Genitor : */ +/*************************************************************/ +/* */ +/* Copyright (c) 1990 */ +/* Darrell L. Whitley */ +/* Computer Science Department */ +/* Colorado State University */ +/* */ +/* Permission is hereby granted to copy all or any part of */ +/* this program for free distribution. The author's name */ +/* and this copyright notice must be included in any copy. */ +/* */ +/*************************************************************/ + +#include "postgres.h" +#include "optimizer/geqo_random.h" +#include "optimizer/geqo_recombination.h" + +#if defined(PX) + +/* px + * + * position crossover + */ +void +px(PlannerInfo *root, Gene *tour1, Gene *tour2, Gene *offspring, int num_gene, + City * city_table) +{ + int num_positions; + int i, + pos, + tour2_index, + offspring_index; + + /* initialize city table */ + for (i = 1; i <= num_gene; i++) + city_table[i].used = 0; + + /* choose random positions that will be inherited directly from parent */ + num_positions = geqo_randint(root, 2 * num_gene / 3, num_gene / 3); + + /* choose random position */ + for (i = 0; i < num_positions; i++) + { + pos = geqo_randint(root, num_gene - 1, 0); + + offspring[pos] = tour1[pos]; /* transfer cities to child */ + city_table[(int) tour1[pos]].used = 1; /* mark city used */ + } + + tour2_index = 0; + offspring_index = 0; + + + /* px main part */ + + while (offspring_index < num_gene) + { + + /* next position in offspring filled */ + if (!city_table[(int) tour1[offspring_index]].used) + { + + /* next city in tour1 not used */ + if (!city_table[(int) tour2[tour2_index]].used) + { + + /* inherit from tour1 */ + offspring[offspring_index] = tour2[tour2_index]; + + tour2_index++; + offspring_index++; + } + else + { /* next city in tour2 has been used */ + tour2_index++; + } + } + else + { /* next position in offspring is filled */ + offspring_index++; + } + } +} + +#endif /* defined(PX) */ diff --git a/src/backend/optimizer/geqo/geqo_random.c b/src/backend/optimizer/geqo/geqo_random.c new file mode 100644 index 0000000..c6b7648 --- /dev/null +++ b/src/backend/optimizer/geqo/geqo_random.c @@ -0,0 +1,45 @@ +/*------------------------------------------------------------------------ + * + * geqo_random.c + * random number generator + * + * Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group + * Portions Copyright (c) 1994, Regents of the University of California + * + * src/backend/optimizer/geqo/geqo_random.c + * + *------------------------------------------------------------------------- + */ + +#include "postgres.h" + +#include "optimizer/geqo_random.h" + + +void +geqo_set_seed(PlannerInfo *root, double seed) +{ + GeqoPrivateData *private = (GeqoPrivateData *) root->join_search_private; + + pg_prng_fseed(&private->random_state, seed); +} + +double +geqo_rand(PlannerInfo *root) +{ + GeqoPrivateData *private = (GeqoPrivateData *) root->join_search_private; + + return pg_prng_double(&private->random_state); +} + +int +geqo_randint(PlannerInfo *root, int upper, int lower) +{ + GeqoPrivateData *private = (GeqoPrivateData *) root->join_search_private; + + /* + * In current usage, "lower" is never negative so we can just use + * pg_prng_uint64_range directly. + */ + return (int) pg_prng_uint64_range(&private->random_state, lower, upper); +} diff --git a/src/backend/optimizer/geqo/geqo_recombination.c b/src/backend/optimizer/geqo/geqo_recombination.c new file mode 100644 index 0000000..a5d3e47 --- /dev/null +++ b/src/backend/optimizer/geqo/geqo_recombination.c @@ -0,0 +1,92 @@ +/*------------------------------------------------------------------------ +* +* geqo_recombination.c +* misc recombination procedures +* +* src/backend/optimizer/geqo/geqo_recombination.c +* +*------------------------------------------------------------------------- +*/ + +/* contributed by: + =*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*= + * Martin Utesch * Institute of Automatic Control * + = = University of Mining and Technology = + * utesch@aut.tu-freiberg.de * Freiberg, Germany * + =*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*= + */ + +/* -- parts of this are adapted from D. Whitley's Genitor algorithm -- */ + +#include "postgres.h" + +#include "optimizer/geqo_random.h" +#include "optimizer/geqo_recombination.h" + + +/* + * init_tour + * + * Randomly generates a legal "traveling salesman" tour + * (i.e. where each point is visited only once.) + */ +void +init_tour(PlannerInfo *root, Gene *tour, int num_gene) +{ + int i, + j; + + /* + * We must fill the tour[] array with a random permutation of the numbers + * 1 .. num_gene. We can do that in one pass using the "inside-out" + * variant of the Fisher-Yates shuffle algorithm. Notionally, we append + * each new value to the array and then swap it with a randomly-chosen + * array element (possibly including itself, else we fail to generate + * permutations with the last city last). The swap step can be optimized + * by combining it with the insertion. + */ + if (num_gene > 0) + tour[0] = (Gene) 1; + + for (i = 1; i < num_gene; i++) + { + j = geqo_randint(root, i, 0); + /* i != j check avoids fetching uninitialized array element */ + if (i != j) + tour[i] = tour[j]; + tour[j] = (Gene) (i + 1); + } +} + +/* city table is used in these recombination methods: */ +#if defined(CX) || defined(PX) || defined(OX1) || defined(OX2) + +/* alloc_city_table + * + * allocate memory for city table + */ +City * +alloc_city_table(PlannerInfo *root, int num_gene) +{ + City *city_table; + + /* + * palloc one extra location so that nodes numbered 1..n can be indexed + * directly; 0 will not be used + */ + city_table = (City *) palloc((num_gene + 1) * sizeof(City)); + + return city_table; +} + +/* free_city_table + * + * deallocate memory of city table + */ +void +free_city_table(PlannerInfo *root, City * city_table) +{ + pfree(city_table); +} + +#endif /* CX || PX || OX1 || OX2 */ diff --git a/src/backend/optimizer/geqo/geqo_selection.c b/src/backend/optimizer/geqo/geqo_selection.c new file mode 100644 index 0000000..50f678a --- /dev/null +++ b/src/backend/optimizer/geqo/geqo_selection.c @@ -0,0 +1,111 @@ +/*------------------------------------------------------------------------- + * + * geqo_selection.c + * linear selection scheme for the genetic query optimizer + * + * Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group + * Portions Copyright (c) 1994, Regents of the University of California + * + * src/backend/optimizer/geqo/geqo_selection.c + * + *------------------------------------------------------------------------- + */ + +/* contributed by: + =*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*= + * Martin Utesch * Institute of Automatic Control * + = = University of Mining and Technology = + * utesch@aut.tu-freiberg.de * Freiberg, Germany * + =*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*= + */ + +/* this is adopted from D. Whitley's Genitor algorithm */ + +/*************************************************************/ +/* */ +/* Copyright (c) 1990 */ +/* Darrell L. Whitley */ +/* Computer Science Department */ +/* Colorado State University */ +/* */ +/* Permission is hereby granted to copy all or any part of */ +/* this program for free distribution. The author's name */ +/* and this copyright notice must be included in any copy. */ +/* */ +/*************************************************************/ + +#include "postgres.h" + +#include <math.h> + +#include "optimizer/geqo_copy.h" +#include "optimizer/geqo_random.h" +#include "optimizer/geqo_selection.h" + +static int linear_rand(PlannerInfo *root, int max, double bias); + + +/* + * geqo_selection + * according to bias described by input parameters, + * first and second genes are selected from the pool + */ +void +geqo_selection(PlannerInfo *root, Chromosome *momma, Chromosome *daddy, + Pool *pool, double bias) +{ + int first, + second; + + first = linear_rand(root, pool->size, bias); + second = linear_rand(root, pool->size, bias); + + /* + * Ensure we have selected different genes, except if pool size is only + * one, when we can't. + */ + if (pool->size > 1) + { + while (first == second) + second = linear_rand(root, pool->size, bias); + } + + geqo_copy(root, momma, &pool->data[first], pool->string_length); + geqo_copy(root, daddy, &pool->data[second], pool->string_length); +} + +/* + * linear_rand + * generates random integer between 0 and input max number + * using input linear bias + * + * bias is y-intercept of linear distribution + * + * probability distribution function is: f(x) = bias - 2(bias - 1)x + * bias = (prob of first rule) / (prob of middle rule) + */ +static int +linear_rand(PlannerInfo *root, int pool_size, double bias) +{ + double index; /* index between 0 and pool_size */ + double max = (double) pool_size; + + /* + * geqo_rand() is not supposed to return 1.0, but if it does then we will + * get exactly max from this equation, whereas we need 0 <= index < max. + * Also it seems possible that roundoff error might deliver values + * slightly outside the range; in particular avoid passing a value + * slightly less than 0 to sqrt(). If we get a bad value just try again. + */ + do + { + double sqrtval; + + sqrtval = (bias * bias) - 4.0 * (bias - 1.0) * geqo_rand(root); + if (sqrtval > 0.0) + sqrtval = sqrt(sqrtval); + index = max * (bias - sqrtval) / 2.0 / (bias - 1.0); + } while (index < 0.0 || index >= max); + + return (int) index; +} diff --git a/src/backend/optimizer/path/Makefile b/src/backend/optimizer/path/Makefile new file mode 100644 index 0000000..1e199ff --- /dev/null +++ b/src/backend/optimizer/path/Makefile @@ -0,0 +1,26 @@ +#------------------------------------------------------------------------- +# +# Makefile-- +# Makefile for optimizer/path +# +# IDENTIFICATION +# src/backend/optimizer/path/Makefile +# +#------------------------------------------------------------------------- + +subdir = src/backend/optimizer/path +top_builddir = ../../../.. +include $(top_builddir)/src/Makefile.global + +OBJS = \ + allpaths.o \ + clausesel.o \ + costsize.o \ + equivclass.o \ + indxpath.o \ + joinpath.o \ + joinrels.o \ + pathkeys.o \ + tidpath.o + +include $(top_srcdir)/src/backend/common.mk diff --git a/src/backend/optimizer/path/allpaths.c b/src/backend/optimizer/path/allpaths.c new file mode 100644 index 0000000..4e02439 --- /dev/null +++ b/src/backend/optimizer/path/allpaths.c @@ -0,0 +1,4668 @@ +/*------------------------------------------------------------------------- + * + * allpaths.c + * Routines to find possible search paths for processing a query + * + * Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group + * Portions Copyright (c) 1994, Regents of the University of California + * + * + * IDENTIFICATION + * src/backend/optimizer/path/allpaths.c + * + *------------------------------------------------------------------------- + */ + +#include "postgres.h" + +#include <limits.h> +#include <math.h> + +#include "access/sysattr.h" +#include "access/tsmapi.h" +#include "catalog/pg_class.h" +#include "catalog/pg_operator.h" +#include "catalog/pg_proc.h" +#include "foreign/fdwapi.h" +#include "miscadmin.h" +#include "nodes/makefuncs.h" +#include "nodes/nodeFuncs.h" +#include "nodes/supportnodes.h" +#ifdef OPTIMIZER_DEBUG +#include "nodes/print.h" +#endif +#include "optimizer/appendinfo.h" +#include "optimizer/clauses.h" +#include "optimizer/cost.h" +#include "optimizer/geqo.h" +#include "optimizer/inherit.h" +#include "optimizer/optimizer.h" +#include "optimizer/pathnode.h" +#include "optimizer/paths.h" +#include "optimizer/plancat.h" +#include "optimizer/planner.h" +#include "optimizer/restrictinfo.h" +#include "optimizer/tlist.h" +#include "parser/parse_clause.h" +#include "parser/parsetree.h" +#include "partitioning/partbounds.h" +#include "partitioning/partprune.h" +#include "rewrite/rewriteManip.h" +#include "utils/lsyscache.h" + + +/* Bitmask flags for pushdown_safety_info.unsafeFlags */ +#define UNSAFE_HAS_VOLATILE_FUNC (1 << 0) +#define UNSAFE_HAS_SET_FUNC (1 << 1) +#define UNSAFE_NOTIN_DISTINCTON_CLAUSE (1 << 2) +#define UNSAFE_NOTIN_PARTITIONBY_CLAUSE (1 << 3) +#define UNSAFE_TYPE_MISMATCH (1 << 4) + +/* results of subquery_is_pushdown_safe */ +typedef struct pushdown_safety_info +{ + unsigned char *unsafeFlags; /* bitmask of reasons why this target list + * column is unsafe for qual pushdown, or 0 if + * no reason. */ + bool unsafeVolatile; /* don't push down volatile quals */ + bool unsafeLeaky; /* don't push down leaky quals */ +} pushdown_safety_info; + +/* Return type for qual_is_pushdown_safe */ +typedef enum pushdown_safe_type +{ + PUSHDOWN_UNSAFE, /* unsafe to push qual into subquery */ + PUSHDOWN_SAFE, /* safe to push qual into subquery */ + PUSHDOWN_WINDOWCLAUSE_RUNCOND /* unsafe, but may work as WindowClause + * run condition */ +} pushdown_safe_type; + +/* These parameters are set by GUC */ +bool enable_geqo = false; /* just in case GUC doesn't set it */ +int geqo_threshold; +int min_parallel_table_scan_size; +int min_parallel_index_scan_size; + +/* Hook for plugins to get control in set_rel_pathlist() */ +set_rel_pathlist_hook_type set_rel_pathlist_hook = NULL; + +/* Hook for plugins to replace standard_join_search() */ +join_search_hook_type join_search_hook = NULL; + + +static void set_base_rel_consider_startup(PlannerInfo *root); +static void set_base_rel_sizes(PlannerInfo *root); +static void set_base_rel_pathlists(PlannerInfo *root); +static void set_rel_size(PlannerInfo *root, RelOptInfo *rel, + Index rti, RangeTblEntry *rte); +static void set_rel_pathlist(PlannerInfo *root, RelOptInfo *rel, + Index rti, RangeTblEntry *rte); +static void set_plain_rel_size(PlannerInfo *root, RelOptInfo *rel, + RangeTblEntry *rte); +static void create_plain_partial_paths(PlannerInfo *root, RelOptInfo *rel); +static void set_rel_consider_parallel(PlannerInfo *root, RelOptInfo *rel, + RangeTblEntry *rte); +static void set_plain_rel_pathlist(PlannerInfo *root, RelOptInfo *rel, + RangeTblEntry *rte); +static void set_tablesample_rel_size(PlannerInfo *root, RelOptInfo *rel, + RangeTblEntry *rte); +static void set_tablesample_rel_pathlist(PlannerInfo *root, RelOptInfo *rel, + RangeTblEntry *rte); +static void set_foreign_size(PlannerInfo *root, RelOptInfo *rel, + RangeTblEntry *rte); +static void set_foreign_pathlist(PlannerInfo *root, RelOptInfo *rel, + RangeTblEntry *rte); +static void set_append_rel_size(PlannerInfo *root, RelOptInfo *rel, + Index rti, RangeTblEntry *rte); +static void set_append_rel_pathlist(PlannerInfo *root, RelOptInfo *rel, + Index rti, RangeTblEntry *rte); +static void generate_orderedappend_paths(PlannerInfo *root, RelOptInfo *rel, + List *live_childrels, + List *all_child_pathkeys); +static Path *get_cheapest_parameterized_child_path(PlannerInfo *root, + RelOptInfo *rel, + Relids required_outer); +static void accumulate_append_subpath(Path *path, + List **subpaths, + List **special_subpaths); +static Path *get_singleton_append_subpath(Path *path); +static void set_dummy_rel_pathlist(RelOptInfo *rel); +static void set_subquery_pathlist(PlannerInfo *root, RelOptInfo *rel, + Index rti, RangeTblEntry *rte); +static void set_function_pathlist(PlannerInfo *root, RelOptInfo *rel, + RangeTblEntry *rte); +static void set_values_pathlist(PlannerInfo *root, RelOptInfo *rel, + RangeTblEntry *rte); +static void set_tablefunc_pathlist(PlannerInfo *root, RelOptInfo *rel, + RangeTblEntry *rte); +static void set_cte_pathlist(PlannerInfo *root, RelOptInfo *rel, + RangeTblEntry *rte); +static void set_namedtuplestore_pathlist(PlannerInfo *root, RelOptInfo *rel, + RangeTblEntry *rte); +static void set_result_pathlist(PlannerInfo *root, RelOptInfo *rel, + RangeTblEntry *rte); +static void set_worktable_pathlist(PlannerInfo *root, RelOptInfo *rel, + RangeTblEntry *rte); +static RelOptInfo *make_rel_from_joinlist(PlannerInfo *root, List *joinlist); +static bool subquery_is_pushdown_safe(Query *subquery, Query *topquery, + pushdown_safety_info *safetyInfo); +static bool recurse_pushdown_safe(Node *setOp, Query *topquery, + pushdown_safety_info *safetyInfo); +static void check_output_expressions(Query *subquery, + pushdown_safety_info *safetyInfo); +static void compare_tlist_datatypes(List *tlist, List *colTypes, + pushdown_safety_info *safetyInfo); +static bool targetIsInAllPartitionLists(TargetEntry *tle, Query *query); +static pushdown_safe_type qual_is_pushdown_safe(Query *subquery, Index rti, + RestrictInfo *rinfo, + pushdown_safety_info *safetyInfo); +static void subquery_push_qual(Query *subquery, + RangeTblEntry *rte, Index rti, Node *qual); +static void recurse_push_qual(Node *setOp, Query *topquery, + RangeTblEntry *rte, Index rti, Node *qual); +static void remove_unused_subquery_outputs(Query *subquery, RelOptInfo *rel, + Bitmapset *extra_used_attrs); + + +/* + * make_one_rel + * Finds all possible access paths for executing a query, returning a + * single rel that represents the join of all base rels in the query. + */ +RelOptInfo * +make_one_rel(PlannerInfo *root, List *joinlist) +{ + RelOptInfo *rel; + Index rti; + double total_pages; + + /* + * Construct the all_baserels Relids set. + */ + root->all_baserels = NULL; + for (rti = 1; rti < root->simple_rel_array_size; rti++) + { + RelOptInfo *brel = root->simple_rel_array[rti]; + + /* there may be empty slots corresponding to non-baserel RTEs */ + if (brel == NULL) + continue; + + Assert(brel->relid == rti); /* sanity check on array */ + + /* ignore RTEs that are "other rels" */ + if (brel->reloptkind != RELOPT_BASEREL) + continue; + + root->all_baserels = bms_add_member(root->all_baserels, brel->relid); + } + + /* Mark base rels as to whether we care about fast-start plans */ + set_base_rel_consider_startup(root); + + /* + * Compute size estimates and consider_parallel flags for each base rel. + */ + set_base_rel_sizes(root); + + /* + * We should now have size estimates for every actual table involved in + * the query, and we also know which if any have been deleted from the + * query by join removal, pruned by partition pruning, or eliminated by + * constraint exclusion. So we can now compute total_table_pages. + * + * Note that appendrels are not double-counted here, even though we don't + * bother to distinguish RelOptInfos for appendrel parents, because the + * parents will have pages = 0. + * + * XXX if a table is self-joined, we will count it once per appearance, + * which perhaps is the wrong thing ... but that's not completely clear, + * and detecting self-joins here is difficult, so ignore it for now. + */ + total_pages = 0; + for (rti = 1; rti < root->simple_rel_array_size; rti++) + { + RelOptInfo *brel = root->simple_rel_array[rti]; + + if (brel == NULL) + continue; + + Assert(brel->relid == rti); /* sanity check on array */ + + if (IS_DUMMY_REL(brel)) + continue; + + if (IS_SIMPLE_REL(brel)) + total_pages += (double) brel->pages; + } + root->total_table_pages = total_pages; + + /* + * Generate access paths for each base rel. + */ + set_base_rel_pathlists(root); + + /* + * Generate access paths for the entire join tree. + */ + rel = make_rel_from_joinlist(root, joinlist); + + /* + * The result should join all and only the query's base rels. + */ + Assert(bms_equal(rel->relids, root->all_baserels)); + + return rel; +} + +/* + * set_base_rel_consider_startup + * Set the consider_[param_]startup flags for each base-relation entry. + * + * For the moment, we only deal with consider_param_startup here; because the + * logic for consider_startup is pretty trivial and is the same for every base + * relation, we just let build_simple_rel() initialize that flag correctly to + * start with. If that logic ever gets more complicated it would probably + * be better to move it here. + */ +static void +set_base_rel_consider_startup(PlannerInfo *root) +{ + /* + * Since parameterized paths can only be used on the inside of a nestloop + * join plan, there is usually little value in considering fast-start + * plans for them. However, for relations that are on the RHS of a SEMI + * or ANTI join, a fast-start plan can be useful because we're only going + * to care about fetching one tuple anyway. + * + * To minimize growth of planning time, we currently restrict this to + * cases where the RHS is a single base relation, not a join; there is no + * provision for consider_param_startup to get set at all on joinrels. + * Also we don't worry about appendrels. costsize.c's costing rules for + * nestloop semi/antijoins don't consider such cases either. + */ + ListCell *lc; + + foreach(lc, root->join_info_list) + { + SpecialJoinInfo *sjinfo = (SpecialJoinInfo *) lfirst(lc); + int varno; + + if ((sjinfo->jointype == JOIN_SEMI || sjinfo->jointype == JOIN_ANTI) && + bms_get_singleton_member(sjinfo->syn_righthand, &varno)) + { + RelOptInfo *rel = find_base_rel(root, varno); + + rel->consider_param_startup = true; + } + } +} + +/* + * set_base_rel_sizes + * Set the size estimates (rows and widths) for each base-relation entry. + * Also determine whether to consider parallel paths for base relations. + * + * We do this in a separate pass over the base rels so that rowcount + * estimates are available for parameterized path generation, and also so + * that each rel's consider_parallel flag is set correctly before we begin to + * generate paths. + */ +static void +set_base_rel_sizes(PlannerInfo *root) +{ + Index rti; + + for (rti = 1; rti < root->simple_rel_array_size; rti++) + { + RelOptInfo *rel = root->simple_rel_array[rti]; + RangeTblEntry *rte; + + /* there may be empty slots corresponding to non-baserel RTEs */ + if (rel == NULL) + continue; + + Assert(rel->relid == rti); /* sanity check on array */ + + /* ignore RTEs that are "other rels" */ + if (rel->reloptkind != RELOPT_BASEREL) + continue; + + rte = root->simple_rte_array[rti]; + + /* + * If parallelism is allowable for this query in general, see whether + * it's allowable for this rel in particular. We have to do this + * before set_rel_size(), because (a) if this rel is an inheritance + * parent, set_append_rel_size() will use and perhaps change the rel's + * consider_parallel flag, and (b) for some RTE types, set_rel_size() + * goes ahead and makes paths immediately. + */ + if (root->glob->parallelModeOK) + set_rel_consider_parallel(root, rel, rte); + + set_rel_size(root, rel, rti, rte); + } +} + +/* + * set_base_rel_pathlists + * Finds all paths available for scanning each base-relation entry. + * Sequential scan and any available indices are considered. + * Each useful path is attached to its relation's 'pathlist' field. + */ +static void +set_base_rel_pathlists(PlannerInfo *root) +{ + Index rti; + + for (rti = 1; rti < root->simple_rel_array_size; rti++) + { + RelOptInfo *rel = root->simple_rel_array[rti]; + + /* there may be empty slots corresponding to non-baserel RTEs */ + if (rel == NULL) + continue; + + Assert(rel->relid == rti); /* sanity check on array */ + + /* ignore RTEs that are "other rels" */ + if (rel->reloptkind != RELOPT_BASEREL) + continue; + + set_rel_pathlist(root, rel, rti, root->simple_rte_array[rti]); + } +} + +/* + * set_rel_size + * Set size estimates for a base relation + */ +static void +set_rel_size(PlannerInfo *root, RelOptInfo *rel, + Index rti, RangeTblEntry *rte) +{ + if (rel->reloptkind == RELOPT_BASEREL && + relation_excluded_by_constraints(root, rel, rte)) + { + /* + * We proved we don't need to scan the rel via constraint exclusion, + * so set up a single dummy path for it. Here we only check this for + * regular baserels; if it's an otherrel, CE was already checked in + * set_append_rel_size(). + * + * In this case, we go ahead and set up the relation's path right away + * instead of leaving it for set_rel_pathlist to do. This is because + * we don't have a convention for marking a rel as dummy except by + * assigning a dummy path to it. + */ + set_dummy_rel_pathlist(rel); + } + else if (rte->inh) + { + /* It's an "append relation", process accordingly */ + set_append_rel_size(root, rel, rti, rte); + } + else + { + switch (rel->rtekind) + { + case RTE_RELATION: + if (rte->relkind == RELKIND_FOREIGN_TABLE) + { + /* Foreign table */ + set_foreign_size(root, rel, rte); + } + else if (rte->relkind == RELKIND_PARTITIONED_TABLE) + { + /* + * We could get here if asked to scan a partitioned table + * with ONLY. In that case we shouldn't scan any of the + * partitions, so mark it as a dummy rel. + */ + set_dummy_rel_pathlist(rel); + } + else if (rte->tablesample != NULL) + { + /* Sampled relation */ + set_tablesample_rel_size(root, rel, rte); + } + else + { + /* Plain relation */ + set_plain_rel_size(root, rel, rte); + } + break; + case RTE_SUBQUERY: + + /* + * Subqueries don't support making a choice between + * parameterized and unparameterized paths, so just go ahead + * and build their paths immediately. + */ + set_subquery_pathlist(root, rel, rti, rte); + break; + case RTE_FUNCTION: + set_function_size_estimates(root, rel); + break; + case RTE_TABLEFUNC: + set_tablefunc_size_estimates(root, rel); + break; + case RTE_VALUES: + set_values_size_estimates(root, rel); + break; + case RTE_CTE: + + /* + * CTEs don't support making a choice between parameterized + * and unparameterized paths, so just go ahead and build their + * paths immediately. + */ + if (rte->self_reference) + set_worktable_pathlist(root, rel, rte); + else + set_cte_pathlist(root, rel, rte); + break; + case RTE_NAMEDTUPLESTORE: + /* Might as well just build the path immediately */ + set_namedtuplestore_pathlist(root, rel, rte); + break; + case RTE_RESULT: + /* Might as well just build the path immediately */ + set_result_pathlist(root, rel, rte); + break; + default: + elog(ERROR, "unexpected rtekind: %d", (int) rel->rtekind); + break; + } + } + + /* + * We insist that all non-dummy rels have a nonzero rowcount estimate. + */ + Assert(rel->rows > 0 || IS_DUMMY_REL(rel)); +} + +/* + * set_rel_pathlist + * Build access paths for a base relation + */ +static void +set_rel_pathlist(PlannerInfo *root, RelOptInfo *rel, + Index rti, RangeTblEntry *rte) +{ + if (IS_DUMMY_REL(rel)) + { + /* We already proved the relation empty, so nothing more to do */ + } + else if (rte->inh) + { + /* It's an "append relation", process accordingly */ + set_append_rel_pathlist(root, rel, rti, rte); + } + else + { + switch (rel->rtekind) + { + case RTE_RELATION: + if (rte->relkind == RELKIND_FOREIGN_TABLE) + { + /* Foreign table */ + set_foreign_pathlist(root, rel, rte); + } + else if (rte->tablesample != NULL) + { + /* Sampled relation */ + set_tablesample_rel_pathlist(root, rel, rte); + } + else + { + /* Plain relation */ + set_plain_rel_pathlist(root, rel, rte); + } + break; + case RTE_SUBQUERY: + /* Subquery --- fully handled during set_rel_size */ + break; + case RTE_FUNCTION: + /* RangeFunction */ + set_function_pathlist(root, rel, rte); + break; + case RTE_TABLEFUNC: + /* Table Function */ + set_tablefunc_pathlist(root, rel, rte); + break; + case RTE_VALUES: + /* Values list */ + set_values_pathlist(root, rel, rte); + break; + case RTE_CTE: + /* CTE reference --- fully handled during set_rel_size */ + break; + case RTE_NAMEDTUPLESTORE: + /* tuplestore reference --- fully handled during set_rel_size */ + break; + case RTE_RESULT: + /* simple Result --- fully handled during set_rel_size */ + break; + default: + elog(ERROR, "unexpected rtekind: %d", (int) rel->rtekind); + break; + } + } + + /* + * Allow a plugin to editorialize on the set of Paths for this base + * relation. It could add new paths (such as CustomPaths) by calling + * add_path(), or add_partial_path() if parallel aware. It could also + * delete or modify paths added by the core code. + */ + if (set_rel_pathlist_hook) + (*set_rel_pathlist_hook) (root, rel, rti, rte); + + /* + * If this is a baserel, we should normally consider gathering any partial + * paths we may have created for it. We have to do this after calling the + * set_rel_pathlist_hook, else it cannot add partial paths to be included + * here. + * + * However, if this is an inheritance child, skip it. Otherwise, we could + * end up with a very large number of gather nodes, each trying to grab + * its own pool of workers. Instead, we'll consider gathering partial + * paths for the parent appendrel. + * + * Also, if this is the topmost scan/join rel, we postpone gathering until + * the final scan/join targetlist is available (see grouping_planner). + */ + if (rel->reloptkind == RELOPT_BASEREL && + !bms_equal(rel->relids, root->all_baserels)) + generate_useful_gather_paths(root, rel, false); + + /* Now find the cheapest of the paths for this rel */ + set_cheapest(rel); + +#ifdef OPTIMIZER_DEBUG + debug_print_rel(root, rel); +#endif +} + +/* + * set_plain_rel_size + * Set size estimates for a plain relation (no subquery, no inheritance) + */ +static void +set_plain_rel_size(PlannerInfo *root, RelOptInfo *rel, RangeTblEntry *rte) +{ + /* + * Test any partial indexes of rel for applicability. We must do this + * first since partial unique indexes can affect size estimates. + */ + check_index_predicates(root, rel); + + /* Mark rel with estimated output rows, width, etc */ + set_baserel_size_estimates(root, rel); +} + +/* + * If this relation could possibly be scanned from within a worker, then set + * its consider_parallel flag. + */ +static void +set_rel_consider_parallel(PlannerInfo *root, RelOptInfo *rel, + RangeTblEntry *rte) +{ + /* + * The flag has previously been initialized to false, so we can just + * return if it becomes clear that we can't safely set it. + */ + Assert(!rel->consider_parallel); + + /* Don't call this if parallelism is disallowed for the entire query. */ + Assert(root->glob->parallelModeOK); + + /* This should only be called for baserels and appendrel children. */ + Assert(IS_SIMPLE_REL(rel)); + + /* Assorted checks based on rtekind. */ + switch (rte->rtekind) + { + case RTE_RELATION: + + /* + * Currently, parallel workers can't access the leader's temporary + * tables. We could possibly relax this if we wrote all of its + * local buffers at the start of the query and made no changes + * thereafter (maybe we could allow hint bit changes), and if we + * taught the workers to read them. Writing a large number of + * temporary buffers could be expensive, though, and we don't have + * the rest of the necessary infrastructure right now anyway. So + * for now, bail out if we see a temporary table. + */ + if (get_rel_persistence(rte->relid) == RELPERSISTENCE_TEMP) + return; + + /* + * Table sampling can be pushed down to workers if the sample + * function and its arguments are safe. + */ + if (rte->tablesample != NULL) + { + char proparallel = func_parallel(rte->tablesample->tsmhandler); + + if (proparallel != PROPARALLEL_SAFE) + return; + if (!is_parallel_safe(root, (Node *) rte->tablesample->args)) + return; + } + + /* + * Ask FDWs whether they can support performing a ForeignScan + * within a worker. Most often, the answer will be no. For + * example, if the nature of the FDW is such that it opens a TCP + * connection with a remote server, each parallel worker would end + * up with a separate connection, and these connections might not + * be appropriately coordinated between workers and the leader. + */ + if (rte->relkind == RELKIND_FOREIGN_TABLE) + { + Assert(rel->fdwroutine); + if (!rel->fdwroutine->IsForeignScanParallelSafe) + return; + if (!rel->fdwroutine->IsForeignScanParallelSafe(root, rel, rte)) + return; + } + + /* + * There are additional considerations for appendrels, which we'll + * deal with in set_append_rel_size and set_append_rel_pathlist. + * For now, just set consider_parallel based on the rel's own + * quals and targetlist. + */ + break; + + case RTE_SUBQUERY: + + /* + * There's no intrinsic problem with scanning a subquery-in-FROM + * (as distinct from a SubPlan or InitPlan) in a parallel worker. + * If the subquery doesn't happen to have any parallel-safe paths, + * then flagging it as consider_parallel won't change anything, + * but that's true for plain tables, too. We must set + * consider_parallel based on the rel's own quals and targetlist, + * so that if a subquery path is parallel-safe but the quals and + * projection we're sticking onto it are not, we correctly mark + * the SubqueryScanPath as not parallel-safe. (Note that + * set_subquery_pathlist() might push some of these quals down + * into the subquery itself, but that doesn't change anything.) + * + * We can't push sub-select containing LIMIT/OFFSET to workers as + * there is no guarantee that the row order will be fully + * deterministic, and applying LIMIT/OFFSET will lead to + * inconsistent results at the top-level. (In some cases, where + * the result is ordered, we could relax this restriction. But it + * doesn't currently seem worth expending extra effort to do so.) + */ + { + Query *subquery = castNode(Query, rte->subquery); + + if (limit_needed(subquery)) + return; + } + break; + + case RTE_JOIN: + /* Shouldn't happen; we're only considering baserels here. */ + Assert(false); + return; + + case RTE_FUNCTION: + /* Check for parallel-restricted functions. */ + if (!is_parallel_safe(root, (Node *) rte->functions)) + return; + break; + + case RTE_TABLEFUNC: + /* not parallel safe */ + return; + + case RTE_VALUES: + /* Check for parallel-restricted functions. */ + if (!is_parallel_safe(root, (Node *) rte->values_lists)) + return; + break; + + case RTE_CTE: + + /* + * CTE tuplestores aren't shared among parallel workers, so we + * force all CTE scans to happen in the leader. Also, populating + * the CTE would require executing a subplan that's not available + * in the worker, might be parallel-restricted, and must get + * executed only once. + */ + return; + + case RTE_NAMEDTUPLESTORE: + + /* + * tuplestore cannot be shared, at least without more + * infrastructure to support that. + */ + return; + + case RTE_RESULT: + /* RESULT RTEs, in themselves, are no problem. */ + break; + } + + /* + * If there's anything in baserestrictinfo that's parallel-restricted, we + * give up on parallelizing access to this relation. We could consider + * instead postponing application of the restricted quals until we're + * above all the parallelism in the plan tree, but it's not clear that + * that would be a win in very many cases, and it might be tricky to make + * outer join clauses work correctly. It would likely break equivalence + * classes, too. + */ + if (!is_parallel_safe(root, (Node *) rel->baserestrictinfo)) + return; + + /* + * Likewise, if the relation's outputs are not parallel-safe, give up. + * (Usually, they're just Vars, but sometimes they're not.) + */ + if (!is_parallel_safe(root, (Node *) rel->reltarget->exprs)) + return; + + /* We have a winner. */ + rel->consider_parallel = true; +} + +/* + * set_plain_rel_pathlist + * Build access paths for a plain relation (no subquery, no inheritance) + */ +static void +set_plain_rel_pathlist(PlannerInfo *root, RelOptInfo *rel, RangeTblEntry *rte) +{ + Relids required_outer; + + /* + * We don't support pushing join clauses into the quals of a seqscan, but + * it could still have required parameterization due to LATERAL refs in + * its tlist. + */ + required_outer = rel->lateral_relids; + + /* Consider sequential scan */ + add_path(rel, create_seqscan_path(root, rel, required_outer, 0)); + + /* If appropriate, consider parallel sequential scan */ + if (rel->consider_parallel && required_outer == NULL) + create_plain_partial_paths(root, rel); + + /* Consider index scans */ + create_index_paths(root, rel); + + /* Consider TID scans */ + create_tidscan_paths(root, rel); +} + +/* + * create_plain_partial_paths + * Build partial access paths for parallel scan of a plain relation + */ +static void +create_plain_partial_paths(PlannerInfo *root, RelOptInfo *rel) +{ + int parallel_workers; + + parallel_workers = compute_parallel_worker(rel, rel->pages, -1, + max_parallel_workers_per_gather); + + /* If any limit was set to zero, the user doesn't want a parallel scan. */ + if (parallel_workers <= 0) + return; + + /* Add an unordered partial path based on a parallel sequential scan. */ + add_partial_path(rel, create_seqscan_path(root, rel, NULL, parallel_workers)); +} + +/* + * set_tablesample_rel_size + * Set size estimates for a sampled relation + */ +static void +set_tablesample_rel_size(PlannerInfo *root, RelOptInfo *rel, RangeTblEntry *rte) +{ + TableSampleClause *tsc = rte->tablesample; + TsmRoutine *tsm; + BlockNumber pages; + double tuples; + + /* + * Test any partial indexes of rel for applicability. We must do this + * first since partial unique indexes can affect size estimates. + */ + check_index_predicates(root, rel); + + /* + * Call the sampling method's estimation function to estimate the number + * of pages it will read and the number of tuples it will return. (Note: + * we assume the function returns sane values.) + */ + tsm = GetTsmRoutine(tsc->tsmhandler); + tsm->SampleScanGetSampleSize(root, rel, tsc->args, + &pages, &tuples); + + /* + * For the moment, because we will only consider a SampleScan path for the + * rel, it's okay to just overwrite the pages and tuples estimates for the + * whole relation. If we ever consider multiple path types for sampled + * rels, we'll need more complication. + */ + rel->pages = pages; + rel->tuples = tuples; + + /* Mark rel with estimated output rows, width, etc */ + set_baserel_size_estimates(root, rel); +} + +/* + * set_tablesample_rel_pathlist + * Build access paths for a sampled relation + */ +static void +set_tablesample_rel_pathlist(PlannerInfo *root, RelOptInfo *rel, RangeTblEntry *rte) +{ + Relids required_outer; + Path *path; + + /* + * We don't support pushing join clauses into the quals of a samplescan, + * but it could still have required parameterization due to LATERAL refs + * in its tlist or TABLESAMPLE arguments. + */ + required_outer = rel->lateral_relids; + + /* Consider sampled scan */ + path = create_samplescan_path(root, rel, required_outer); + + /* + * If the sampling method does not support repeatable scans, we must avoid + * plans that would scan the rel multiple times. Ideally, we'd simply + * avoid putting the rel on the inside of a nestloop join; but adding such + * a consideration to the planner seems like a great deal of complication + * to support an uncommon usage of second-rate sampling methods. Instead, + * if there is a risk that the query might perform an unsafe join, just + * wrap the SampleScan in a Materialize node. We can check for joins by + * counting the membership of all_baserels (note that this correctly + * counts inheritance trees as single rels). If we're inside a subquery, + * we can't easily check whether a join might occur in the outer query, so + * just assume one is possible. + * + * GetTsmRoutine is relatively expensive compared to the other tests here, + * so check repeatable_across_scans last, even though that's a bit odd. + */ + if ((root->query_level > 1 || + bms_membership(root->all_baserels) != BMS_SINGLETON) && + !(GetTsmRoutine(rte->tablesample->tsmhandler)->repeatable_across_scans)) + { + path = (Path *) create_material_path(rel, path); + } + + add_path(rel, path); + + /* For the moment, at least, there are no other paths to consider */ +} + +/* + * set_foreign_size + * Set size estimates for a foreign table RTE + */ +static void +set_foreign_size(PlannerInfo *root, RelOptInfo *rel, RangeTblEntry *rte) +{ + /* Mark rel with estimated output rows, width, etc */ + set_foreign_size_estimates(root, rel); + + /* Let FDW adjust the size estimates, if it can */ + rel->fdwroutine->GetForeignRelSize(root, rel, rte->relid); + + /* ... but do not let it set the rows estimate to zero */ + rel->rows = clamp_row_est(rel->rows); + + /* + * Also, make sure rel->tuples is not insane relative to rel->rows. + * Notably, this ensures sanity if pg_class.reltuples contains -1 and the + * FDW doesn't do anything to replace that. + */ + rel->tuples = Max(rel->tuples, rel->rows); +} + +/* + * set_foreign_pathlist + * Build access paths for a foreign table RTE + */ +static void +set_foreign_pathlist(PlannerInfo *root, RelOptInfo *rel, RangeTblEntry *rte) +{ + /* Call the FDW's GetForeignPaths function to generate path(s) */ + rel->fdwroutine->GetForeignPaths(root, rel, rte->relid); +} + +/* + * set_append_rel_size + * Set size estimates for a simple "append relation" + * + * The passed-in rel and RTE represent the entire append relation. The + * relation's contents are computed by appending together the output of the + * individual member relations. Note that in the non-partitioned inheritance + * case, the first member relation is actually the same table as is mentioned + * in the parent RTE ... but it has a different RTE and RelOptInfo. This is + * a good thing because their outputs are not the same size. + */ +static void +set_append_rel_size(PlannerInfo *root, RelOptInfo *rel, + Index rti, RangeTblEntry *rte) +{ + int parentRTindex = rti; + bool has_live_children; + double parent_rows; + double parent_size; + double *parent_attrsizes; + int nattrs; + ListCell *l; + + /* Guard against stack overflow due to overly deep inheritance tree. */ + check_stack_depth(); + + Assert(IS_SIMPLE_REL(rel)); + + /* + * If this is a partitioned baserel, set the consider_partitionwise_join + * flag; currently, we only consider partitionwise joins with the baserel + * if its targetlist doesn't contain a whole-row Var. + */ + if (enable_partitionwise_join && + rel->reloptkind == RELOPT_BASEREL && + rte->relkind == RELKIND_PARTITIONED_TABLE && + rel->attr_needed[InvalidAttrNumber - rel->min_attr] == NULL) + rel->consider_partitionwise_join = true; + + /* + * Initialize to compute size estimates for whole append relation. + * + * We handle width estimates by weighting the widths of different child + * rels proportionally to their number of rows. This is sensible because + * the use of width estimates is mainly to compute the total relation + * "footprint" if we have to sort or hash it. To do this, we sum the + * total equivalent size (in "double" arithmetic) and then divide by the + * total rowcount estimate. This is done separately for the total rel + * width and each attribute. + * + * Note: if you consider changing this logic, beware that child rels could + * have zero rows and/or width, if they were excluded by constraints. + */ + has_live_children = false; + parent_rows = 0; + parent_size = 0; + nattrs = rel->max_attr - rel->min_attr + 1; + parent_attrsizes = (double *) palloc0(nattrs * sizeof(double)); + + foreach(l, root->append_rel_list) + { + AppendRelInfo *appinfo = (AppendRelInfo *) lfirst(l); + int childRTindex; + RangeTblEntry *childRTE; + RelOptInfo *childrel; + ListCell *parentvars; + ListCell *childvars; + + /* append_rel_list contains all append rels; ignore others */ + if (appinfo->parent_relid != parentRTindex) + continue; + + childRTindex = appinfo->child_relid; + childRTE = root->simple_rte_array[childRTindex]; + + /* + * The child rel's RelOptInfo was already created during + * add_other_rels_to_query. + */ + childrel = find_base_rel(root, childRTindex); + Assert(childrel->reloptkind == RELOPT_OTHER_MEMBER_REL); + + /* We may have already proven the child to be dummy. */ + if (IS_DUMMY_REL(childrel)) + continue; + + /* + * We have to copy the parent's targetlist and quals to the child, + * with appropriate substitution of variables. However, the + * baserestrictinfo quals were already copied/substituted when the + * child RelOptInfo was built. So we don't need any additional setup + * before applying constraint exclusion. + */ + if (relation_excluded_by_constraints(root, childrel, childRTE)) + { + /* + * This child need not be scanned, so we can omit it from the + * appendrel. + */ + set_dummy_rel_pathlist(childrel); + continue; + } + + /* + * Constraint exclusion failed, so copy the parent's join quals and + * targetlist to the child, with appropriate variable substitutions. + * + * NB: the resulting childrel->reltarget->exprs may contain arbitrary + * expressions, which otherwise would not occur in a rel's targetlist. + * Code that might be looking at an appendrel child must cope with + * such. (Normally, a rel's targetlist would only include Vars and + * PlaceHolderVars.) XXX we do not bother to update the cost or width + * fields of childrel->reltarget; not clear if that would be useful. + */ + childrel->joininfo = (List *) + adjust_appendrel_attrs(root, + (Node *) rel->joininfo, + 1, &appinfo); + childrel->reltarget->exprs = (List *) + adjust_appendrel_attrs(root, + (Node *) rel->reltarget->exprs, + 1, &appinfo); + + /* + * We have to make child entries in the EquivalenceClass data + * structures as well. This is needed either if the parent + * participates in some eclass joins (because we will want to consider + * inner-indexscan joins on the individual children) or if the parent + * has useful pathkeys (because we should try to build MergeAppend + * paths that produce those sort orderings). + */ + if (rel->has_eclass_joins || has_useful_pathkeys(root, rel)) + add_child_rel_equivalences(root, appinfo, rel, childrel); + childrel->has_eclass_joins = rel->has_eclass_joins; + + /* + * Note: we could compute appropriate attr_needed data for the child's + * variables, by transforming the parent's attr_needed through the + * translated_vars mapping. However, currently there's no need + * because attr_needed is only examined for base relations not + * otherrels. So we just leave the child's attr_needed empty. + */ + + /* + * If we consider partitionwise joins with the parent rel, do the same + * for partitioned child rels. + * + * Note: here we abuse the consider_partitionwise_join flag by setting + * it for child rels that are not themselves partitioned. We do so to + * tell try_partitionwise_join() that the child rel is sufficiently + * valid to be used as a per-partition input, even if it later gets + * proven to be dummy. (It's not usable until we've set up the + * reltarget and EC entries, which we just did.) + */ + if (rel->consider_partitionwise_join) + childrel->consider_partitionwise_join = true; + + /* + * If parallelism is allowable for this query in general, see whether + * it's allowable for this childrel in particular. But if we've + * already decided the appendrel is not parallel-safe as a whole, + * there's no point in considering parallelism for this child. For + * consistency, do this before calling set_rel_size() for the child. + */ + if (root->glob->parallelModeOK && rel->consider_parallel) + set_rel_consider_parallel(root, childrel, childRTE); + + /* + * Compute the child's size. + */ + set_rel_size(root, childrel, childRTindex, childRTE); + + /* + * It is possible that constraint exclusion detected a contradiction + * within a child subquery, even though we didn't prove one above. If + * so, we can skip this child. + */ + if (IS_DUMMY_REL(childrel)) + continue; + + /* We have at least one live child. */ + has_live_children = true; + + /* + * If any live child is not parallel-safe, treat the whole appendrel + * as not parallel-safe. In future we might be able to generate plans + * in which some children are farmed out to workers while others are + * not; but we don't have that today, so it's a waste to consider + * partial paths anywhere in the appendrel unless it's all safe. + * (Child rels visited before this one will be unmarked in + * set_append_rel_pathlist().) + */ + if (!childrel->consider_parallel) + rel->consider_parallel = false; + + /* + * Accumulate size information from each live child. + */ + Assert(childrel->rows > 0); + + parent_rows += childrel->rows; + parent_size += childrel->reltarget->width * childrel->rows; + + /* + * Accumulate per-column estimates too. We need not do anything for + * PlaceHolderVars in the parent list. If child expression isn't a + * Var, or we didn't record a width estimate for it, we have to fall + * back on a datatype-based estimate. + * + * By construction, child's targetlist is 1-to-1 with parent's. + */ + forboth(parentvars, rel->reltarget->exprs, + childvars, childrel->reltarget->exprs) + { + Var *parentvar = (Var *) lfirst(parentvars); + Node *childvar = (Node *) lfirst(childvars); + + if (IsA(parentvar, Var) && parentvar->varno == parentRTindex) + { + int pndx = parentvar->varattno - rel->min_attr; + int32 child_width = 0; + + if (IsA(childvar, Var) && + ((Var *) childvar)->varno == childrel->relid) + { + int cndx = ((Var *) childvar)->varattno - childrel->min_attr; + + child_width = childrel->attr_widths[cndx]; + } + if (child_width <= 0) + child_width = get_typavgwidth(exprType(childvar), + exprTypmod(childvar)); + Assert(child_width > 0); + parent_attrsizes[pndx] += child_width * childrel->rows; + } + } + } + + if (has_live_children) + { + /* + * Save the finished size estimates. + */ + int i; + + Assert(parent_rows > 0); + rel->rows = parent_rows; + rel->reltarget->width = rint(parent_size / parent_rows); + for (i = 0; i < nattrs; i++) + rel->attr_widths[i] = rint(parent_attrsizes[i] / parent_rows); + + /* + * Set "raw tuples" count equal to "rows" for the appendrel; needed + * because some places assume rel->tuples is valid for any baserel. + */ + rel->tuples = parent_rows; + + /* + * Note that we leave rel->pages as zero; this is important to avoid + * double-counting the appendrel tree in total_table_pages. + */ + } + else + { + /* + * All children were excluded by constraints, so mark the whole + * appendrel dummy. We must do this in this phase so that the rel's + * dummy-ness is visible when we generate paths for other rels. + */ + set_dummy_rel_pathlist(rel); + } + + pfree(parent_attrsizes); +} + +/* + * set_append_rel_pathlist + * Build access paths for an "append relation" + */ +static void +set_append_rel_pathlist(PlannerInfo *root, RelOptInfo *rel, + Index rti, RangeTblEntry *rte) +{ + int parentRTindex = rti; + List *live_childrels = NIL; + ListCell *l; + + /* + * Generate access paths for each member relation, and remember the + * non-dummy children. + */ + foreach(l, root->append_rel_list) + { + AppendRelInfo *appinfo = (AppendRelInfo *) lfirst(l); + int childRTindex; + RangeTblEntry *childRTE; + RelOptInfo *childrel; + + /* append_rel_list contains all append rels; ignore others */ + if (appinfo->parent_relid != parentRTindex) + continue; + + /* Re-locate the child RTE and RelOptInfo */ + childRTindex = appinfo->child_relid; + childRTE = root->simple_rte_array[childRTindex]; + childrel = root->simple_rel_array[childRTindex]; + + /* + * If set_append_rel_size() decided the parent appendrel was + * parallel-unsafe at some point after visiting this child rel, we + * need to propagate the unsafety marking down to the child, so that + * we don't generate useless partial paths for it. + */ + if (!rel->consider_parallel) + childrel->consider_parallel = false; + + /* + * Compute the child's access paths. + */ + set_rel_pathlist(root, childrel, childRTindex, childRTE); + + /* + * If child is dummy, ignore it. + */ + if (IS_DUMMY_REL(childrel)) + continue; + + /* + * Child is live, so add it to the live_childrels list for use below. + */ + live_childrels = lappend(live_childrels, childrel); + } + + /* Add paths to the append relation. */ + add_paths_to_append_rel(root, rel, live_childrels); +} + + +/* + * add_paths_to_append_rel + * Generate paths for the given append relation given the set of non-dummy + * child rels. + * + * The function collects all parameterizations and orderings supported by the + * non-dummy children. For every such parameterization or ordering, it creates + * an append path collecting one path from each non-dummy child with given + * parameterization or ordering. Similarly it collects partial paths from + * non-dummy children to create partial append paths. + */ +void +add_paths_to_append_rel(PlannerInfo *root, RelOptInfo *rel, + List *live_childrels) +{ + List *subpaths = NIL; + bool subpaths_valid = true; + List *partial_subpaths = NIL; + List *pa_partial_subpaths = NIL; + List *pa_nonpartial_subpaths = NIL; + bool partial_subpaths_valid = true; + bool pa_subpaths_valid; + List *all_child_pathkeys = NIL; + List *all_child_outers = NIL; + ListCell *l; + double partial_rows = -1; + + /* If appropriate, consider parallel append */ + pa_subpaths_valid = enable_parallel_append && rel->consider_parallel; + + /* + * For every non-dummy child, remember the cheapest path. Also, identify + * all pathkeys (orderings) and parameterizations (required_outer sets) + * available for the non-dummy member relations. + */ + foreach(l, live_childrels) + { + RelOptInfo *childrel = lfirst(l); + ListCell *lcp; + Path *cheapest_partial_path = NULL; + + /* + * If child has an unparameterized cheapest-total path, add that to + * the unparameterized Append path we are constructing for the parent. + * If not, there's no workable unparameterized path. + * + * With partitionwise aggregates, the child rel's pathlist may be + * empty, so don't assume that a path exists here. + */ + if (childrel->pathlist != NIL && + childrel->cheapest_total_path->param_info == NULL) + accumulate_append_subpath(childrel->cheapest_total_path, + &subpaths, NULL); + else + subpaths_valid = false; + + /* Same idea, but for a partial plan. */ + if (childrel->partial_pathlist != NIL) + { + cheapest_partial_path = linitial(childrel->partial_pathlist); + accumulate_append_subpath(cheapest_partial_path, + &partial_subpaths, NULL); + } + else + partial_subpaths_valid = false; + + /* + * Same idea, but for a parallel append mixing partial and non-partial + * paths. + */ + if (pa_subpaths_valid) + { + Path *nppath = NULL; + + nppath = + get_cheapest_parallel_safe_total_inner(childrel->pathlist); + + if (cheapest_partial_path == NULL && nppath == NULL) + { + /* Neither a partial nor a parallel-safe path? Forget it. */ + pa_subpaths_valid = false; + } + else if (nppath == NULL || + (cheapest_partial_path != NULL && + cheapest_partial_path->total_cost < nppath->total_cost)) + { + /* Partial path is cheaper or the only option. */ + Assert(cheapest_partial_path != NULL); + accumulate_append_subpath(cheapest_partial_path, + &pa_partial_subpaths, + &pa_nonpartial_subpaths); + } + else + { + /* + * Either we've got only a non-partial path, or we think that + * a single backend can execute the best non-partial path + * faster than all the parallel backends working together can + * execute the best partial path. + * + * It might make sense to be more aggressive here. Even if + * the best non-partial path is more expensive than the best + * partial path, it could still be better to choose the + * non-partial path if there are several such paths that can + * be given to different workers. For now, we don't try to + * figure that out. + */ + accumulate_append_subpath(nppath, + &pa_nonpartial_subpaths, + NULL); + } + } + + /* + * Collect lists of all the available path orderings and + * parameterizations for all the children. We use these as a + * heuristic to indicate which sort orderings and parameterizations we + * should build Append and MergeAppend paths for. + */ + foreach(lcp, childrel->pathlist) + { + Path *childpath = (Path *) lfirst(lcp); + List *childkeys = childpath->pathkeys; + Relids childouter = PATH_REQ_OUTER(childpath); + + /* Unsorted paths don't contribute to pathkey list */ + if (childkeys != NIL) + { + ListCell *lpk; + bool found = false; + + /* Have we already seen this ordering? */ + foreach(lpk, all_child_pathkeys) + { + List *existing_pathkeys = (List *) lfirst(lpk); + + if (compare_pathkeys(existing_pathkeys, + childkeys) == PATHKEYS_EQUAL) + { + found = true; + break; + } + } + if (!found) + { + /* No, so add it to all_child_pathkeys */ + all_child_pathkeys = lappend(all_child_pathkeys, + childkeys); + } + } + + /* Unparameterized paths don't contribute to param-set list */ + if (childouter) + { + ListCell *lco; + bool found = false; + + /* Have we already seen this param set? */ + foreach(lco, all_child_outers) + { + Relids existing_outers = (Relids) lfirst(lco); + + if (bms_equal(existing_outers, childouter)) + { + found = true; + break; + } + } + if (!found) + { + /* No, so add it to all_child_outers */ + all_child_outers = lappend(all_child_outers, + childouter); + } + } + } + } + + /* + * If we found unparameterized paths for all children, build an unordered, + * unparameterized Append path for the rel. (Note: this is correct even + * if we have zero or one live subpath due to constraint exclusion.) + */ + if (subpaths_valid) + add_path(rel, (Path *) create_append_path(root, rel, subpaths, NIL, + NIL, NULL, 0, false, + -1)); + + /* + * Consider an append of unordered, unparameterized partial paths. Make + * it parallel-aware if possible. + */ + if (partial_subpaths_valid && partial_subpaths != NIL) + { + AppendPath *appendpath; + ListCell *lc; + int parallel_workers = 0; + + /* Find the highest number of workers requested for any subpath. */ + foreach(lc, partial_subpaths) + { + Path *path = lfirst(lc); + + parallel_workers = Max(parallel_workers, path->parallel_workers); + } + Assert(parallel_workers > 0); + + /* + * If the use of parallel append is permitted, always request at least + * log2(# of children) workers. We assume it can be useful to have + * extra workers in this case because they will be spread out across + * the children. The precise formula is just a guess, but we don't + * want to end up with a radically different answer for a table with N + * partitions vs. an unpartitioned table with the same data, so the + * use of some kind of log-scaling here seems to make some sense. + */ + if (enable_parallel_append) + { + parallel_workers = Max(parallel_workers, + fls(list_length(live_childrels))); + parallel_workers = Min(parallel_workers, + max_parallel_workers_per_gather); + } + Assert(parallel_workers > 0); + + /* Generate a partial append path. */ + appendpath = create_append_path(root, rel, NIL, partial_subpaths, + NIL, NULL, parallel_workers, + enable_parallel_append, + -1); + + /* + * Make sure any subsequent partial paths use the same row count + * estimate. + */ + partial_rows = appendpath->path.rows; + + /* Add the path. */ + add_partial_path(rel, (Path *) appendpath); + } + + /* + * Consider a parallel-aware append using a mix of partial and non-partial + * paths. (This only makes sense if there's at least one child which has + * a non-partial path that is substantially cheaper than any partial path; + * otherwise, we should use the append path added in the previous step.) + */ + if (pa_subpaths_valid && pa_nonpartial_subpaths != NIL) + { + AppendPath *appendpath; + ListCell *lc; + int parallel_workers = 0; + + /* + * Find the highest number of workers requested for any partial + * subpath. + */ + foreach(lc, pa_partial_subpaths) + { + Path *path = lfirst(lc); + + parallel_workers = Max(parallel_workers, path->parallel_workers); + } + + /* + * Same formula here as above. It's even more important in this + * instance because the non-partial paths won't contribute anything to + * the planned number of parallel workers. + */ + parallel_workers = Max(parallel_workers, + fls(list_length(live_childrels))); + parallel_workers = Min(parallel_workers, + max_parallel_workers_per_gather); + Assert(parallel_workers > 0); + + appendpath = create_append_path(root, rel, pa_nonpartial_subpaths, + pa_partial_subpaths, + NIL, NULL, parallel_workers, true, + partial_rows); + add_partial_path(rel, (Path *) appendpath); + } + + /* + * Also build unparameterized ordered append paths based on the collected + * list of child pathkeys. + */ + if (subpaths_valid) + generate_orderedappend_paths(root, rel, live_childrels, + all_child_pathkeys); + + /* + * Build Append paths for each parameterization seen among the child rels. + * (This may look pretty expensive, but in most cases of practical + * interest, the child rels will expose mostly the same parameterizations, + * so that not that many cases actually get considered here.) + * + * The Append node itself cannot enforce quals, so all qual checking must + * be done in the child paths. This means that to have a parameterized + * Append path, we must have the exact same parameterization for each + * child path; otherwise some children might be failing to check the + * moved-down quals. To make them match up, we can try to increase the + * parameterization of lesser-parameterized paths. + */ + foreach(l, all_child_outers) + { + Relids required_outer = (Relids) lfirst(l); + ListCell *lcr; + + /* Select the child paths for an Append with this parameterization */ + subpaths = NIL; + subpaths_valid = true; + foreach(lcr, live_childrels) + { + RelOptInfo *childrel = (RelOptInfo *) lfirst(lcr); + Path *subpath; + + if (childrel->pathlist == NIL) + { + /* failed to make a suitable path for this child */ + subpaths_valid = false; + break; + } + + subpath = get_cheapest_parameterized_child_path(root, + childrel, + required_outer); + if (subpath == NULL) + { + /* failed to make a suitable path for this child */ + subpaths_valid = false; + break; + } + accumulate_append_subpath(subpath, &subpaths, NULL); + } + + if (subpaths_valid) + add_path(rel, (Path *) + create_append_path(root, rel, subpaths, NIL, + NIL, required_outer, 0, false, + -1)); + } + + /* + * When there is only a single child relation, the Append path can inherit + * any ordering available for the child rel's path, so that it's useful to + * consider ordered partial paths. Above we only considered the cheapest + * partial path for each child, but let's also make paths using any + * partial paths that have pathkeys. + */ + if (list_length(live_childrels) == 1) + { + RelOptInfo *childrel = (RelOptInfo *) linitial(live_childrels); + + /* skip the cheapest partial path, since we already used that above */ + for_each_from(l, childrel->partial_pathlist, 1) + { + Path *path = (Path *) lfirst(l); + AppendPath *appendpath; + + /* skip paths with no pathkeys. */ + if (path->pathkeys == NIL) + continue; + + appendpath = create_append_path(root, rel, NIL, list_make1(path), + NIL, NULL, + path->parallel_workers, true, + partial_rows); + add_partial_path(rel, (Path *) appendpath); + } + } +} + +/* + * generate_orderedappend_paths + * Generate ordered append paths for an append relation + * + * Usually we generate MergeAppend paths here, but there are some special + * cases where we can generate simple Append paths, because the subpaths + * can provide tuples in the required order already. + * + * We generate a path for each ordering (pathkey list) appearing in + * all_child_pathkeys. + * + * We consider both cheapest-startup and cheapest-total cases, ie, for each + * interesting ordering, collect all the cheapest startup subpaths and all the + * cheapest total paths, and build a suitable path for each case. + * + * We don't currently generate any parameterized ordered paths here. While + * it would not take much more code here to do so, it's very unclear that it + * is worth the planning cycles to investigate such paths: there's little + * use for an ordered path on the inside of a nestloop. In fact, it's likely + * that the current coding of add_path would reject such paths out of hand, + * because add_path gives no credit for sort ordering of parameterized paths, + * and a parameterized MergeAppend is going to be more expensive than the + * corresponding parameterized Append path. If we ever try harder to support + * parameterized mergejoin plans, it might be worth adding support for + * parameterized paths here to feed such joins. (See notes in + * optimizer/README for why that might not ever happen, though.) + */ +static void +generate_orderedappend_paths(PlannerInfo *root, RelOptInfo *rel, + List *live_childrels, + List *all_child_pathkeys) +{ + ListCell *lcp; + List *partition_pathkeys = NIL; + List *partition_pathkeys_desc = NIL; + bool partition_pathkeys_partial = true; + bool partition_pathkeys_desc_partial = true; + + /* + * Some partitioned table setups may allow us to use an Append node + * instead of a MergeAppend. This is possible in cases such as RANGE + * partitioned tables where it's guaranteed that an earlier partition must + * contain rows which come earlier in the sort order. To detect whether + * this is relevant, build pathkey descriptions of the partition ordering, + * for both forward and reverse scans. + */ + if (rel->part_scheme != NULL && IS_SIMPLE_REL(rel) && + partitions_are_ordered(rel->boundinfo, rel->live_parts)) + { + partition_pathkeys = build_partition_pathkeys(root, rel, + ForwardScanDirection, + &partition_pathkeys_partial); + + partition_pathkeys_desc = build_partition_pathkeys(root, rel, + BackwardScanDirection, + &partition_pathkeys_desc_partial); + + /* + * You might think we should truncate_useless_pathkeys here, but + * allowing partition keys which are a subset of the query's pathkeys + * can often be useful. For example, consider a table partitioned by + * RANGE (a, b), and a query with ORDER BY a, b, c. If we have child + * paths that can produce the a, b, c ordering (perhaps via indexes on + * (a, b, c)) then it works to consider the appendrel output as + * ordered by a, b, c. + */ + } + + /* Now consider each interesting sort ordering */ + foreach(lcp, all_child_pathkeys) + { + List *pathkeys = (List *) lfirst(lcp); + List *startup_subpaths = NIL; + List *total_subpaths = NIL; + List *fractional_subpaths = NIL; + bool startup_neq_total = false; + ListCell *lcr; + bool match_partition_order; + bool match_partition_order_desc; + + /* + * Determine if this sort ordering matches any partition pathkeys we + * have, for both ascending and descending partition order. If the + * partition pathkeys happen to be contained in pathkeys then it still + * works, as described above, providing that the partition pathkeys + * are complete and not just a prefix of the partition keys. (In such + * cases we'll be relying on the child paths to have sorted the + * lower-order columns of the required pathkeys.) + */ + match_partition_order = + pathkeys_contained_in(pathkeys, partition_pathkeys) || + (!partition_pathkeys_partial && + pathkeys_contained_in(partition_pathkeys, pathkeys)); + + match_partition_order_desc = !match_partition_order && + (pathkeys_contained_in(pathkeys, partition_pathkeys_desc) || + (!partition_pathkeys_desc_partial && + pathkeys_contained_in(partition_pathkeys_desc, pathkeys))); + + /* Select the child paths for this ordering... */ + foreach(lcr, live_childrels) + { + RelOptInfo *childrel = (RelOptInfo *) lfirst(lcr); + Path *cheapest_startup, + *cheapest_total, + *cheapest_fractional = NULL; + + /* Locate the right paths, if they are available. */ + cheapest_startup = + get_cheapest_path_for_pathkeys(childrel->pathlist, + pathkeys, + NULL, + STARTUP_COST, + false); + cheapest_total = + get_cheapest_path_for_pathkeys(childrel->pathlist, + pathkeys, + NULL, + TOTAL_COST, + false); + + /* + * If we can't find any paths with the right order just use the + * cheapest-total path; we'll have to sort it later. + */ + if (cheapest_startup == NULL || cheapest_total == NULL) + { + cheapest_startup = cheapest_total = + childrel->cheapest_total_path; + /* Assert we do have an unparameterized path for this child */ + Assert(cheapest_total->param_info == NULL); + } + + /* + * When building a fractional path, determine a cheapest + * fractional path for each child relation too. Looking at startup + * and total costs is not enough, because the cheapest fractional + * path may be dominated by two separate paths (one for startup, + * one for total). + * + * When needed (building fractional path), determine the cheapest + * fractional path too. + */ + if (root->tuple_fraction > 0) + { + double path_fraction = (1.0 / root->tuple_fraction); + + cheapest_fractional = + get_cheapest_fractional_path_for_pathkeys(childrel->pathlist, + pathkeys, + NULL, + path_fraction); + + /* + * If we found no path with matching pathkeys, use the + * cheapest total path instead. + * + * XXX We might consider partially sorted paths too (with an + * incremental sort on top). But we'd have to build all the + * incremental paths, do the costing etc. + */ + if (!cheapest_fractional) + cheapest_fractional = cheapest_total; + } + + /* + * Notice whether we actually have different paths for the + * "cheapest" and "total" cases; frequently there will be no point + * in two create_merge_append_path() calls. + */ + if (cheapest_startup != cheapest_total) + startup_neq_total = true; + + /* + * Collect the appropriate child paths. The required logic varies + * for the Append and MergeAppend cases. + */ + if (match_partition_order) + { + /* + * We're going to make a plain Append path. We don't need + * most of what accumulate_append_subpath would do, but we do + * want to cut out child Appends or MergeAppends if they have + * just a single subpath (and hence aren't doing anything + * useful). + */ + cheapest_startup = get_singleton_append_subpath(cheapest_startup); + cheapest_total = get_singleton_append_subpath(cheapest_total); + + startup_subpaths = lappend(startup_subpaths, cheapest_startup); + total_subpaths = lappend(total_subpaths, cheapest_total); + + if (cheapest_fractional) + { + cheapest_fractional = get_singleton_append_subpath(cheapest_fractional); + fractional_subpaths = lappend(fractional_subpaths, cheapest_fractional); + } + } + else if (match_partition_order_desc) + { + /* + * As above, but we need to reverse the order of the children, + * because nodeAppend.c doesn't know anything about reverse + * ordering and will scan the children in the order presented. + */ + cheapest_startup = get_singleton_append_subpath(cheapest_startup); + cheapest_total = get_singleton_append_subpath(cheapest_total); + + startup_subpaths = lcons(cheapest_startup, startup_subpaths); + total_subpaths = lcons(cheapest_total, total_subpaths); + + if (cheapest_fractional) + { + cheapest_fractional = get_singleton_append_subpath(cheapest_fractional); + fractional_subpaths = lcons(cheapest_fractional, fractional_subpaths); + } + } + else + { + /* + * Otherwise, rely on accumulate_append_subpath to collect the + * child paths for the MergeAppend. + */ + accumulate_append_subpath(cheapest_startup, + &startup_subpaths, NULL); + accumulate_append_subpath(cheapest_total, + &total_subpaths, NULL); + + if (cheapest_fractional) + accumulate_append_subpath(cheapest_fractional, + &fractional_subpaths, NULL); + } + } + + /* ... and build the Append or MergeAppend paths */ + if (match_partition_order || match_partition_order_desc) + { + /* We only need Append */ + add_path(rel, (Path *) create_append_path(root, + rel, + startup_subpaths, + NIL, + pathkeys, + NULL, + 0, + false, + -1)); + if (startup_neq_total) + add_path(rel, (Path *) create_append_path(root, + rel, + total_subpaths, + NIL, + pathkeys, + NULL, + 0, + false, + -1)); + + if (fractional_subpaths) + add_path(rel, (Path *) create_append_path(root, + rel, + fractional_subpaths, + NIL, + pathkeys, + NULL, + 0, + false, + -1)); + } + else + { + /* We need MergeAppend */ + add_path(rel, (Path *) create_merge_append_path(root, + rel, + startup_subpaths, + pathkeys, + NULL)); + if (startup_neq_total) + add_path(rel, (Path *) create_merge_append_path(root, + rel, + total_subpaths, + pathkeys, + NULL)); + + if (fractional_subpaths) + add_path(rel, (Path *) create_merge_append_path(root, + rel, + fractional_subpaths, + pathkeys, + NULL)); + } + } +} + +/* + * get_cheapest_parameterized_child_path + * Get cheapest path for this relation that has exactly the requested + * parameterization. + * + * Returns NULL if unable to create such a path. + */ +static Path * +get_cheapest_parameterized_child_path(PlannerInfo *root, RelOptInfo *rel, + Relids required_outer) +{ + Path *cheapest; + ListCell *lc; + + /* + * Look up the cheapest existing path with no more than the needed + * parameterization. If it has exactly the needed parameterization, we're + * done. + */ + cheapest = get_cheapest_path_for_pathkeys(rel->pathlist, + NIL, + required_outer, + TOTAL_COST, + false); + Assert(cheapest != NULL); + if (bms_equal(PATH_REQ_OUTER(cheapest), required_outer)) + return cheapest; + + /* + * Otherwise, we can "reparameterize" an existing path to match the given + * parameterization, which effectively means pushing down additional + * joinquals to be checked within the path's scan. However, some existing + * paths might check the available joinquals already while others don't; + * therefore, it's not clear which existing path will be cheapest after + * reparameterization. We have to go through them all and find out. + */ + cheapest = NULL; + foreach(lc, rel->pathlist) + { + Path *path = (Path *) lfirst(lc); + + /* Can't use it if it needs more than requested parameterization */ + if (!bms_is_subset(PATH_REQ_OUTER(path), required_outer)) + continue; + + /* + * Reparameterization can only increase the path's cost, so if it's + * already more expensive than the current cheapest, forget it. + */ + if (cheapest != NULL && + compare_path_costs(cheapest, path, TOTAL_COST) <= 0) + continue; + + /* Reparameterize if needed, then recheck cost */ + if (!bms_equal(PATH_REQ_OUTER(path), required_outer)) + { + path = reparameterize_path(root, path, required_outer, 1.0); + if (path == NULL) + continue; /* failed to reparameterize this one */ + Assert(bms_equal(PATH_REQ_OUTER(path), required_outer)); + + if (cheapest != NULL && + compare_path_costs(cheapest, path, TOTAL_COST) <= 0) + continue; + } + + /* We have a new best path */ + cheapest = path; + } + + /* Return the best path, or NULL if we found no suitable candidate */ + return cheapest; +} + +/* + * accumulate_append_subpath + * Add a subpath to the list being built for an Append or MergeAppend. + * + * It's possible that the child is itself an Append or MergeAppend path, in + * which case we can "cut out the middleman" and just add its child paths to + * our own list. (We don't try to do this earlier because we need to apply + * both levels of transformation to the quals.) + * + * Note that if we omit a child MergeAppend in this way, we are effectively + * omitting a sort step, which seems fine: if the parent is to be an Append, + * its result would be unsorted anyway, while if the parent is to be a + * MergeAppend, there's no point in a separate sort on a child. + * + * Normally, either path is a partial path and subpaths is a list of partial + * paths, or else path is a non-partial plan and subpaths is a list of those. + * However, if path is a parallel-aware Append, then we add its partial path + * children to subpaths and the rest to special_subpaths. If the latter is + * NULL, we don't flatten the path at all (unless it contains only partial + * paths). + */ +static void +accumulate_append_subpath(Path *path, List **subpaths, List **special_subpaths) +{ + if (IsA(path, AppendPath)) + { + AppendPath *apath = (AppendPath *) path; + + if (!apath->path.parallel_aware || apath->first_partial_path == 0) + { + *subpaths = list_concat(*subpaths, apath->subpaths); + return; + } + else if (special_subpaths != NULL) + { + List *new_special_subpaths; + + /* Split Parallel Append into partial and non-partial subpaths */ + *subpaths = list_concat(*subpaths, + list_copy_tail(apath->subpaths, + apath->first_partial_path)); + new_special_subpaths = + list_truncate(list_copy(apath->subpaths), + apath->first_partial_path); + *special_subpaths = list_concat(*special_subpaths, + new_special_subpaths); + return; + } + } + else if (IsA(path, MergeAppendPath)) + { + MergeAppendPath *mpath = (MergeAppendPath *) path; + + *subpaths = list_concat(*subpaths, mpath->subpaths); + return; + } + + *subpaths = lappend(*subpaths, path); +} + +/* + * get_singleton_append_subpath + * Returns the single subpath of an Append/MergeAppend, or just + * return 'path' if it's not a single sub-path Append/MergeAppend. + * + * Note: 'path' must not be a parallel-aware path. + */ +static Path * +get_singleton_append_subpath(Path *path) +{ + Assert(!path->parallel_aware); + + if (IsA(path, AppendPath)) + { + AppendPath *apath = (AppendPath *) path; + + if (list_length(apath->subpaths) == 1) + return (Path *) linitial(apath->subpaths); + } + else if (IsA(path, MergeAppendPath)) + { + MergeAppendPath *mpath = (MergeAppendPath *) path; + + if (list_length(mpath->subpaths) == 1) + return (Path *) linitial(mpath->subpaths); + } + + return path; +} + +/* + * set_dummy_rel_pathlist + * Build a dummy path for a relation that's been excluded by constraints + * + * Rather than inventing a special "dummy" path type, we represent this as an + * AppendPath with no members (see also IS_DUMMY_APPEND/IS_DUMMY_REL macros). + * + * (See also mark_dummy_rel, which does basically the same thing, but is + * typically used to change a rel into dummy state after we already made + * paths for it.) + */ +static void +set_dummy_rel_pathlist(RelOptInfo *rel) +{ + /* Set dummy size estimates --- we leave attr_widths[] as zeroes */ + rel->rows = 0; + rel->reltarget->width = 0; + + /* Discard any pre-existing paths; no further need for them */ + rel->pathlist = NIL; + rel->partial_pathlist = NIL; + + /* Set up the dummy path */ + add_path(rel, (Path *) create_append_path(NULL, rel, NIL, NIL, + NIL, rel->lateral_relids, + 0, false, -1)); + + /* + * We set the cheapest-path fields immediately, just in case they were + * pointing at some discarded path. This is redundant when we're called + * from set_rel_size(), but not when called from elsewhere, and doing it + * twice is harmless anyway. + */ + set_cheapest(rel); +} + +/* quick-and-dirty test to see if any joining is needed */ +static bool +has_multiple_baserels(PlannerInfo *root) +{ + int num_base_rels = 0; + Index rti; + + for (rti = 1; rti < root->simple_rel_array_size; rti++) + { + RelOptInfo *brel = root->simple_rel_array[rti]; + + if (brel == NULL) + continue; + + /* ignore RTEs that are "other rels" */ + if (brel->reloptkind == RELOPT_BASEREL) + if (++num_base_rels > 1) + return true; + } + return false; +} + +/* + * find_window_run_conditions + * Determine if 'wfunc' is really a WindowFunc and call its prosupport + * function to determine the function's monotonic properties. We then + * see if 'opexpr' can be used to short-circuit execution. + * + * For example row_number() over (order by ...) always produces a value one + * higher than the previous. If someone has a window function in a subquery + * and has a WHERE clause in the outer query to filter rows <= 10, then we may + * as well stop processing the windowagg once the row number reaches 11. Here + * we check if 'opexpr' might help us to stop doing needless extra processing + * in WindowAgg nodes. + * + * '*keep_original' is set to true if the caller should also use 'opexpr' for + * its original purpose. This is set to false if the caller can assume that + * the run condition will handle all of the required filtering. + * + * Returns true if 'opexpr' was found to be useful and was added to the + * WindowClauses runCondition. We also set *keep_original accordingly and add + * 'attno' to *run_cond_attrs offset by FirstLowInvalidHeapAttributeNumber. + * If the 'opexpr' cannot be used then we set *keep_original to true and + * return false. + */ +static bool +find_window_run_conditions(Query *subquery, RangeTblEntry *rte, Index rti, + AttrNumber attno, WindowFunc *wfunc, OpExpr *opexpr, + bool wfunc_left, bool *keep_original, + Bitmapset **run_cond_attrs) +{ + Oid prosupport; + Expr *otherexpr; + SupportRequestWFuncMonotonic req; + SupportRequestWFuncMonotonic *res; + WindowClause *wclause; + List *opinfos; + OpExpr *runopexpr; + Oid runoperator; + ListCell *lc; + + *keep_original = true; + + while (IsA(wfunc, RelabelType)) + wfunc = (WindowFunc *) ((RelabelType *) wfunc)->arg; + + /* we can only work with window functions */ + if (!IsA(wfunc, WindowFunc)) + return false; + + /* can't use it if there are subplans in the WindowFunc */ + if (contain_subplans((Node *) wfunc)) + return false; + + prosupport = get_func_support(wfunc->winfnoid); + + /* Check if there's a support function for 'wfunc' */ + if (!OidIsValid(prosupport)) + return false; + + /* get the Expr from the other side of the OpExpr */ + if (wfunc_left) + otherexpr = lsecond(opexpr->args); + else + otherexpr = linitial(opexpr->args); + + /* + * The value being compared must not change during the evaluation of the + * window partition. + */ + if (!is_pseudo_constant_clause((Node *) otherexpr)) + return false; + + /* find the window clause belonging to the window function */ + wclause = (WindowClause *) list_nth(subquery->windowClause, + wfunc->winref - 1); + + req.type = T_SupportRequestWFuncMonotonic; + req.window_func = wfunc; + req.window_clause = wclause; + + /* call the support function */ + res = (SupportRequestWFuncMonotonic *) + DatumGetPointer(OidFunctionCall1(prosupport, + PointerGetDatum(&req))); + + /* + * Nothing to do if the function is neither monotonically increasing nor + * monotonically decreasing. + */ + if (res == NULL || res->monotonic == MONOTONICFUNC_NONE) + return false; + + runopexpr = NULL; + runoperator = InvalidOid; + opinfos = get_op_btree_interpretation(opexpr->opno); + + foreach(lc, opinfos) + { + OpBtreeInterpretation *opinfo = (OpBtreeInterpretation *) lfirst(lc); + int strategy = opinfo->strategy; + + /* handle < / <= */ + if (strategy == BTLessStrategyNumber || + strategy == BTLessEqualStrategyNumber) + { + /* + * < / <= is supported for monotonically increasing functions in + * the form <wfunc> op <pseudoconst> and <pseudoconst> op <wfunc> + * for monotonically decreasing functions. + */ + if ((wfunc_left && (res->monotonic & MONOTONICFUNC_INCREASING)) || + (!wfunc_left && (res->monotonic & MONOTONICFUNC_DECREASING))) + { + *keep_original = false; + runopexpr = opexpr; + runoperator = opexpr->opno; + } + break; + } + /* handle > / >= */ + else if (strategy == BTGreaterStrategyNumber || + strategy == BTGreaterEqualStrategyNumber) + { + /* + * > / >= is supported for monotonically decreasing functions in + * the form <wfunc> op <pseudoconst> and <pseudoconst> op <wfunc> + * for monotonically increasing functions. + */ + if ((wfunc_left && (res->monotonic & MONOTONICFUNC_DECREASING)) || + (!wfunc_left && (res->monotonic & MONOTONICFUNC_INCREASING))) + { + *keep_original = false; + runopexpr = opexpr; + runoperator = opexpr->opno; + } + break; + } + /* handle = */ + else if (strategy == BTEqualStrategyNumber) + { + int16 newstrategy; + + /* + * When both monotonically increasing and decreasing then the + * return value of the window function will be the same each time. + * We can simply use 'opexpr' as the run condition without + * modifying it. + */ + if ((res->monotonic & MONOTONICFUNC_BOTH) == MONOTONICFUNC_BOTH) + { + *keep_original = false; + runopexpr = opexpr; + runoperator = opexpr->opno; + break; + } + + /* + * When monotonically increasing we make a qual with <wfunc> <= + * <value> or <value> >= <wfunc> in order to filter out values + * which are above the value in the equality condition. For + * monotonically decreasing functions we want to filter values + * below the value in the equality condition. + */ + if (res->monotonic & MONOTONICFUNC_INCREASING) + newstrategy = wfunc_left ? BTLessEqualStrategyNumber : BTGreaterEqualStrategyNumber; + else + newstrategy = wfunc_left ? BTGreaterEqualStrategyNumber : BTLessEqualStrategyNumber; + + /* We must keep the original equality qual */ + *keep_original = true; + runopexpr = opexpr; + + /* determine the operator to use for the runCondition qual */ + runoperator = get_opfamily_member(opinfo->opfamily_id, + opinfo->oplefttype, + opinfo->oprighttype, + newstrategy); + break; + } + } + + if (runopexpr != NULL) + { + Expr *newexpr; + + /* + * Build the qual required for the run condition keeping the + * WindowFunc on the same side as it was originally. + */ + if (wfunc_left) + newexpr = make_opclause(runoperator, + runopexpr->opresulttype, + runopexpr->opretset, (Expr *) wfunc, + otherexpr, runopexpr->opcollid, + runopexpr->inputcollid); + else + newexpr = make_opclause(runoperator, + runopexpr->opresulttype, + runopexpr->opretset, + otherexpr, (Expr *) wfunc, + runopexpr->opcollid, + runopexpr->inputcollid); + + wclause->runCondition = lappend(wclause->runCondition, newexpr); + + /* record that this attno was used in a run condition */ + *run_cond_attrs = bms_add_member(*run_cond_attrs, + attno - FirstLowInvalidHeapAttributeNumber); + return true; + } + + /* unsupported OpExpr */ + return false; +} + +/* + * check_and_push_window_quals + * Check if 'clause' is a qual that can be pushed into a WindowFunc's + * WindowClause as a 'runCondition' qual. These, when present, allow + * some unnecessary work to be skipped during execution. + * + * 'run_cond_attrs' will be populated with all targetlist resnos of subquery + * targets (offset by FirstLowInvalidHeapAttributeNumber) that we pushed + * window quals for. + * + * Returns true if the caller still must keep the original qual or false if + * the caller can safely ignore the original qual because the WindowAgg node + * will use the runCondition to stop returning tuples. + */ +static bool +check_and_push_window_quals(Query *subquery, RangeTblEntry *rte, Index rti, + Node *clause, Bitmapset **run_cond_attrs) +{ + OpExpr *opexpr = (OpExpr *) clause; + bool keep_original = true; + Var *var1; + Var *var2; + + /* We're only able to use OpExprs with 2 operands */ + if (!IsA(opexpr, OpExpr)) + return true; + + if (list_length(opexpr->args) != 2) + return true; + + /* + * Currently, we restrict this optimization to strict OpExprs. The reason + * for this is that during execution, once the runcondition becomes false, + * we stop evaluating WindowFuncs. To avoid leaving around stale window + * function result values, we set them to NULL. Having only strict + * OpExprs here ensures that we properly filter out the tuples with NULLs + * in the top-level WindowAgg. + */ + set_opfuncid(opexpr); + if (!func_strict(opexpr->opfuncid)) + return true; + + /* + * Check for plain Vars that reference window functions in the subquery. + * If we find any, we'll ask find_window_run_conditions() if 'opexpr' can + * be used as part of the run condition. + */ + + /* Check the left side of the OpExpr */ + var1 = linitial(opexpr->args); + if (IsA(var1, Var) && var1->varattno > 0) + { + TargetEntry *tle = list_nth(subquery->targetList, var1->varattno - 1); + WindowFunc *wfunc = (WindowFunc *) tle->expr; + + if (find_window_run_conditions(subquery, rte, rti, tle->resno, wfunc, + opexpr, true, &keep_original, + run_cond_attrs)) + return keep_original; + } + + /* and check the right side */ + var2 = lsecond(opexpr->args); + if (IsA(var2, Var) && var2->varattno > 0) + { + TargetEntry *tle = list_nth(subquery->targetList, var2->varattno - 1); + WindowFunc *wfunc = (WindowFunc *) tle->expr; + + if (find_window_run_conditions(subquery, rte, rti, tle->resno, wfunc, + opexpr, false, &keep_original, + run_cond_attrs)) + return keep_original; + } + + return true; +} + +/* + * set_subquery_pathlist + * Generate SubqueryScan access paths for a subquery RTE + * + * We don't currently support generating parameterized paths for subqueries + * by pushing join clauses down into them; it seems too expensive to re-plan + * the subquery multiple times to consider different alternatives. + * (XXX that could stand to be reconsidered, now that we use Paths.) + * So the paths made here will be parameterized if the subquery contains + * LATERAL references, otherwise not. As long as that's true, there's no need + * for a separate set_subquery_size phase: just make the paths right away. + */ +static void +set_subquery_pathlist(PlannerInfo *root, RelOptInfo *rel, + Index rti, RangeTblEntry *rte) +{ + Query *parse = root->parse; + Query *subquery = rte->subquery; + Relids required_outer; + pushdown_safety_info safetyInfo; + double tuple_fraction; + RelOptInfo *sub_final_rel; + Bitmapset *run_cond_attrs = NULL; + ListCell *lc; + + /* + * Must copy the Query so that planning doesn't mess up the RTE contents + * (really really need to fix the planner to not scribble on its input, + * someday ... but see remove_unused_subquery_outputs to start with). + */ + subquery = copyObject(subquery); + + /* + * If it's a LATERAL subquery, it might contain some Vars of the current + * query level, requiring it to be treated as parameterized, even though + * we don't support pushing down join quals into subqueries. + */ + required_outer = rel->lateral_relids; + + /* + * Zero out result area for subquery_is_pushdown_safe, so that it can set + * flags as needed while recursing. In particular, we need a workspace + * for keeping track of the reasons why columns are unsafe to reference. + * These reasons are stored in the bits inside unsafeFlags[i] when we + * discover reasons that column i of the subquery is unsafe to be used in + * a pushed-down qual. + */ + memset(&safetyInfo, 0, sizeof(safetyInfo)); + safetyInfo.unsafeFlags = (unsigned char *) + palloc0((list_length(subquery->targetList) + 1) * sizeof(unsigned char)); + + /* + * If the subquery has the "security_barrier" flag, it means the subquery + * originated from a view that must enforce row-level security. Then we + * must not push down quals that contain leaky functions. (Ideally this + * would be checked inside subquery_is_pushdown_safe, but since we don't + * currently pass the RTE to that function, we must do it here.) + */ + safetyInfo.unsafeLeaky = rte->security_barrier; + + /* + * If there are any restriction clauses that have been attached to the + * subquery relation, consider pushing them down to become WHERE or HAVING + * quals of the subquery itself. This transformation is useful because it + * may allow us to generate a better plan for the subquery than evaluating + * all the subquery output rows and then filtering them. + * + * There are several cases where we cannot push down clauses. Restrictions + * involving the subquery are checked by subquery_is_pushdown_safe(). + * Restrictions on individual clauses are checked by + * qual_is_pushdown_safe(). Also, we don't want to push down + * pseudoconstant clauses; better to have the gating node above the + * subquery. + * + * Non-pushed-down clauses will get evaluated as qpquals of the + * SubqueryScan node. + * + * XXX Are there any cases where we want to make a policy decision not to + * push down a pushable qual, because it'd result in a worse plan? + */ + if (rel->baserestrictinfo != NIL && + subquery_is_pushdown_safe(subquery, subquery, &safetyInfo)) + { + /* OK to consider pushing down individual quals */ + List *upperrestrictlist = NIL; + ListCell *l; + + foreach(l, rel->baserestrictinfo) + { + RestrictInfo *rinfo = (RestrictInfo *) lfirst(l); + Node *clause = (Node *) rinfo->clause; + + if (rinfo->pseudoconstant) + { + upperrestrictlist = lappend(upperrestrictlist, rinfo); + continue; + } + + switch (qual_is_pushdown_safe(subquery, rti, rinfo, &safetyInfo)) + { + case PUSHDOWN_SAFE: + /* Push it down */ + subquery_push_qual(subquery, rte, rti, clause); + break; + + case PUSHDOWN_WINDOWCLAUSE_RUNCOND: + + /* + * Since we can't push the qual down into the subquery, + * check if it happens to reference a window function. If + * so then it might be useful to use for the WindowAgg's + * runCondition. + */ + if (!subquery->hasWindowFuncs || + check_and_push_window_quals(subquery, rte, rti, clause, + &run_cond_attrs)) + { + /* + * subquery has no window funcs or the clause is not a + * suitable window run condition qual or it is, but + * the original must also be kept in the upper query. + */ + upperrestrictlist = lappend(upperrestrictlist, rinfo); + } + break; + + case PUSHDOWN_UNSAFE: + upperrestrictlist = lappend(upperrestrictlist, rinfo); + break; + } + } + rel->baserestrictinfo = upperrestrictlist; + /* We don't bother recomputing baserestrict_min_security */ + } + + pfree(safetyInfo.unsafeFlags); + + /* + * The upper query might not use all the subquery's output columns; if + * not, we can simplify. Pass the attributes that were pushed down into + * WindowAgg run conditions to ensure we don't accidentally think those + * are unused. + */ + remove_unused_subquery_outputs(subquery, rel, run_cond_attrs); + + /* + * We can safely pass the outer tuple_fraction down to the subquery if the + * outer level has no joining, aggregation, or sorting to do. Otherwise + * we'd better tell the subquery to plan for full retrieval. (XXX This + * could probably be made more intelligent ...) + */ + if (parse->hasAggs || + parse->groupClause || + parse->groupingSets || + parse->havingQual || + parse->distinctClause || + parse->sortClause || + has_multiple_baserels(root)) + tuple_fraction = 0.0; /* default case */ + else + tuple_fraction = root->tuple_fraction; + + /* plan_params should not be in use in current query level */ + Assert(root->plan_params == NIL); + + /* Generate a subroot and Paths for the subquery */ + rel->subroot = subquery_planner(root->glob, subquery, + root, + false, tuple_fraction); + + /* Isolate the params needed by this specific subplan */ + rel->subplan_params = root->plan_params; + root->plan_params = NIL; + + /* + * It's possible that constraint exclusion proved the subquery empty. If + * so, it's desirable to produce an unadorned dummy path so that we will + * recognize appropriate optimizations at this query level. + */ + sub_final_rel = fetch_upper_rel(rel->subroot, UPPERREL_FINAL, NULL); + + if (IS_DUMMY_REL(sub_final_rel)) + { + set_dummy_rel_pathlist(rel); + return; + } + + /* + * Mark rel with estimated output rows, width, etc. Note that we have to + * do this before generating outer-query paths, else cost_subqueryscan is + * not happy. + */ + set_subquery_size_estimates(root, rel); + + /* + * For each Path that subquery_planner produced, make a SubqueryScanPath + * in the outer query. + */ + foreach(lc, sub_final_rel->pathlist) + { + Path *subpath = (Path *) lfirst(lc); + List *pathkeys; + + /* Convert subpath's pathkeys to outer representation */ + pathkeys = convert_subquery_pathkeys(root, + rel, + subpath->pathkeys, + make_tlist_from_pathtarget(subpath->pathtarget)); + + /* Generate outer path using this subpath */ + add_path(rel, (Path *) + create_subqueryscan_path(root, rel, subpath, + pathkeys, required_outer)); + } + + /* If outer rel allows parallelism, do same for partial paths. */ + if (rel->consider_parallel && bms_is_empty(required_outer)) + { + /* If consider_parallel is false, there should be no partial paths. */ + Assert(sub_final_rel->consider_parallel || + sub_final_rel->partial_pathlist == NIL); + + /* Same for partial paths. */ + foreach(lc, sub_final_rel->partial_pathlist) + { + Path *subpath = (Path *) lfirst(lc); + List *pathkeys; + + /* Convert subpath's pathkeys to outer representation */ + pathkeys = convert_subquery_pathkeys(root, + rel, + subpath->pathkeys, + make_tlist_from_pathtarget(subpath->pathtarget)); + + /* Generate outer path using this subpath */ + add_partial_path(rel, (Path *) + create_subqueryscan_path(root, rel, subpath, + pathkeys, + required_outer)); + } + } +} + +/* + * set_function_pathlist + * Build the (single) access path for a function RTE + */ +static void +set_function_pathlist(PlannerInfo *root, RelOptInfo *rel, RangeTblEntry *rte) +{ + Relids required_outer; + List *pathkeys = NIL; + + /* + * We don't support pushing join clauses into the quals of a function + * scan, but it could still have required parameterization due to LATERAL + * refs in the function expression. + */ + required_outer = rel->lateral_relids; + + /* + * The result is considered unordered unless ORDINALITY was used, in which + * case it is ordered by the ordinal column (the last one). See if we + * care, by checking for uses of that Var in equivalence classes. + */ + if (rte->funcordinality) + { + AttrNumber ordattno = rel->max_attr; + Var *var = NULL; + ListCell *lc; + + /* + * Is there a Var for it in rel's targetlist? If not, the query did + * not reference the ordinality column, or at least not in any way + * that would be interesting for sorting. + */ + foreach(lc, rel->reltarget->exprs) + { + Var *node = (Var *) lfirst(lc); + + /* checking varno/varlevelsup is just paranoia */ + if (IsA(node, Var) && + node->varattno == ordattno && + node->varno == rel->relid && + node->varlevelsup == 0) + { + var = node; + break; + } + } + + /* + * Try to build pathkeys for this Var with int8 sorting. We tell + * build_expression_pathkey not to build any new equivalence class; if + * the Var isn't already mentioned in some EC, it means that nothing + * cares about the ordering. + */ + if (var) + pathkeys = build_expression_pathkey(root, + (Expr *) var, + NULL, /* below outer joins */ + Int8LessOperator, + rel->relids, + false); + } + + /* Generate appropriate path */ + add_path(rel, create_functionscan_path(root, rel, + pathkeys, required_outer)); +} + +/* + * set_values_pathlist + * Build the (single) access path for a VALUES RTE + */ +static void +set_values_pathlist(PlannerInfo *root, RelOptInfo *rel, RangeTblEntry *rte) +{ + Relids required_outer; + + /* + * We don't support pushing join clauses into the quals of a values scan, + * but it could still have required parameterization due to LATERAL refs + * in the values expressions. + */ + required_outer = rel->lateral_relids; + + /* Generate appropriate path */ + add_path(rel, create_valuesscan_path(root, rel, required_outer)); +} + +/* + * set_tablefunc_pathlist + * Build the (single) access path for a table func RTE + */ +static void +set_tablefunc_pathlist(PlannerInfo *root, RelOptInfo *rel, RangeTblEntry *rte) +{ + Relids required_outer; + + /* + * We don't support pushing join clauses into the quals of a tablefunc + * scan, but it could still have required parameterization due to LATERAL + * refs in the function expression. + */ + required_outer = rel->lateral_relids; + + /* Generate appropriate path */ + add_path(rel, create_tablefuncscan_path(root, rel, + required_outer)); +} + +/* + * set_cte_pathlist + * Build the (single) access path for a non-self-reference CTE RTE + * + * There's no need for a separate set_cte_size phase, since we don't + * support join-qual-parameterized paths for CTEs. + */ +static void +set_cte_pathlist(PlannerInfo *root, RelOptInfo *rel, RangeTblEntry *rte) +{ + Plan *cteplan; + PlannerInfo *cteroot; + Index levelsup; + int ndx; + ListCell *lc; + int plan_id; + Relids required_outer; + + /* + * Find the referenced CTE, and locate the plan previously made for it. + */ + levelsup = rte->ctelevelsup; + cteroot = root; + while (levelsup-- > 0) + { + cteroot = cteroot->parent_root; + if (!cteroot) /* shouldn't happen */ + elog(ERROR, "bad levelsup for CTE \"%s\"", rte->ctename); + } + + /* + * Note: cte_plan_ids can be shorter than cteList, if we are still working + * on planning the CTEs (ie, this is a side-reference from another CTE). + * So we mustn't use forboth here. + */ + ndx = 0; + foreach(lc, cteroot->parse->cteList) + { + CommonTableExpr *cte = (CommonTableExpr *) lfirst(lc); + + if (strcmp(cte->ctename, rte->ctename) == 0) + break; + ndx++; + } + if (lc == NULL) /* shouldn't happen */ + elog(ERROR, "could not find CTE \"%s\"", rte->ctename); + if (ndx >= list_length(cteroot->cte_plan_ids)) + elog(ERROR, "could not find plan for CTE \"%s\"", rte->ctename); + plan_id = list_nth_int(cteroot->cte_plan_ids, ndx); + if (plan_id <= 0) + elog(ERROR, "no plan was made for CTE \"%s\"", rte->ctename); + cteplan = (Plan *) list_nth(root->glob->subplans, plan_id - 1); + + /* Mark rel with estimated output rows, width, etc */ + set_cte_size_estimates(root, rel, cteplan->plan_rows); + + /* + * We don't support pushing join clauses into the quals of a CTE scan, but + * it could still have required parameterization due to LATERAL refs in + * its tlist. + */ + required_outer = rel->lateral_relids; + + /* Generate appropriate path */ + add_path(rel, create_ctescan_path(root, rel, required_outer)); +} + +/* + * set_namedtuplestore_pathlist + * Build the (single) access path for a named tuplestore RTE + * + * There's no need for a separate set_namedtuplestore_size phase, since we + * don't support join-qual-parameterized paths for tuplestores. + */ +static void +set_namedtuplestore_pathlist(PlannerInfo *root, RelOptInfo *rel, + RangeTblEntry *rte) +{ + Relids required_outer; + + /* Mark rel with estimated output rows, width, etc */ + set_namedtuplestore_size_estimates(root, rel); + + /* + * We don't support pushing join clauses into the quals of a tuplestore + * scan, but it could still have required parameterization due to LATERAL + * refs in its tlist. + */ + required_outer = rel->lateral_relids; + + /* Generate appropriate path */ + add_path(rel, create_namedtuplestorescan_path(root, rel, required_outer)); + + /* Select cheapest path (pretty easy in this case...) */ + set_cheapest(rel); +} + +/* + * set_result_pathlist + * Build the (single) access path for an RTE_RESULT RTE + * + * There's no need for a separate set_result_size phase, since we + * don't support join-qual-parameterized paths for these RTEs. + */ +static void +set_result_pathlist(PlannerInfo *root, RelOptInfo *rel, + RangeTblEntry *rte) +{ + Relids required_outer; + + /* Mark rel with estimated output rows, width, etc */ + set_result_size_estimates(root, rel); + + /* + * We don't support pushing join clauses into the quals of a Result scan, + * but it could still have required parameterization due to LATERAL refs + * in its tlist. + */ + required_outer = rel->lateral_relids; + + /* Generate appropriate path */ + add_path(rel, create_resultscan_path(root, rel, required_outer)); + + /* Select cheapest path (pretty easy in this case...) */ + set_cheapest(rel); +} + +/* + * set_worktable_pathlist + * Build the (single) access path for a self-reference CTE RTE + * + * There's no need for a separate set_worktable_size phase, since we don't + * support join-qual-parameterized paths for CTEs. + */ +static void +set_worktable_pathlist(PlannerInfo *root, RelOptInfo *rel, RangeTblEntry *rte) +{ + Path *ctepath; + PlannerInfo *cteroot; + Index levelsup; + Relids required_outer; + + /* + * We need to find the non-recursive term's path, which is in the plan + * level that's processing the recursive UNION, which is one level *below* + * where the CTE comes from. + */ + levelsup = rte->ctelevelsup; + if (levelsup == 0) /* shouldn't happen */ + elog(ERROR, "bad levelsup for CTE \"%s\"", rte->ctename); + levelsup--; + cteroot = root; + while (levelsup-- > 0) + { + cteroot = cteroot->parent_root; + if (!cteroot) /* shouldn't happen */ + elog(ERROR, "bad levelsup for CTE \"%s\"", rte->ctename); + } + ctepath = cteroot->non_recursive_path; + if (!ctepath) /* shouldn't happen */ + elog(ERROR, "could not find path for CTE \"%s\"", rte->ctename); + + /* Mark rel with estimated output rows, width, etc */ + set_cte_size_estimates(root, rel, ctepath->rows); + + /* + * We don't support pushing join clauses into the quals of a worktable + * scan, but it could still have required parameterization due to LATERAL + * refs in its tlist. (I'm not sure this is actually possible given the + * restrictions on recursive references, but it's easy enough to support.) + */ + required_outer = rel->lateral_relids; + + /* Generate appropriate path */ + add_path(rel, create_worktablescan_path(root, rel, required_outer)); +} + +/* + * generate_gather_paths + * Generate parallel access paths for a relation by pushing a Gather or + * Gather Merge on top of a partial path. + * + * This must not be called until after we're done creating all partial paths + * for the specified relation. (Otherwise, add_partial_path might delete a + * path that some GatherPath or GatherMergePath has a reference to.) + * + * If we're generating paths for a scan or join relation, override_rows will + * be false, and we'll just use the relation's size estimate. When we're + * being called for a partially-grouped path, though, we need to override + * the rowcount estimate. (It's not clear that the particular value we're + * using here is actually best, but the underlying rel has no estimate so + * we must do something.) + */ +void +generate_gather_paths(PlannerInfo *root, RelOptInfo *rel, bool override_rows) +{ + Path *cheapest_partial_path; + Path *simple_gather_path; + ListCell *lc; + double rows; + double *rowsp = NULL; + + /* If there are no partial paths, there's nothing to do here. */ + if (rel->partial_pathlist == NIL) + return; + + /* Should we override the rel's rowcount estimate? */ + if (override_rows) + rowsp = &rows; + + /* + * The output of Gather is always unsorted, so there's only one partial + * path of interest: the cheapest one. That will be the one at the front + * of partial_pathlist because of the way add_partial_path works. + */ + cheapest_partial_path = linitial(rel->partial_pathlist); + rows = + cheapest_partial_path->rows * cheapest_partial_path->parallel_workers; + simple_gather_path = (Path *) + create_gather_path(root, rel, cheapest_partial_path, rel->reltarget, + NULL, rowsp); + add_path(rel, simple_gather_path); + + /* + * For each useful ordering, we can consider an order-preserving Gather + * Merge. + */ + foreach(lc, rel->partial_pathlist) + { + Path *subpath = (Path *) lfirst(lc); + GatherMergePath *path; + + if (subpath->pathkeys == NIL) + continue; + + rows = subpath->rows * subpath->parallel_workers; + path = create_gather_merge_path(root, rel, subpath, rel->reltarget, + subpath->pathkeys, NULL, rowsp); + add_path(rel, &path->path); + } +} + +/* + * get_useful_pathkeys_for_relation + * Determine which orderings of a relation might be useful. + * + * Getting data in sorted order can be useful either because the requested + * order matches the final output ordering for the overall query we're + * planning, or because it enables an efficient merge join. Here, we try + * to figure out which pathkeys to consider. + * + * This allows us to do incremental sort on top of an index scan under a gather + * merge node, i.e. parallelized. + * + * If the require_parallel_safe is true, we also require the expressions to + * be parallel safe (which allows pushing the sort below Gather Merge). + * + * XXX At the moment this can only ever return a list with a single element, + * because it looks at query_pathkeys only. So we might return the pathkeys + * directly, but it seems plausible we'll want to consider other orderings + * in the future. For example, we might want to consider pathkeys useful for + * merge joins. + */ +static List * +get_useful_pathkeys_for_relation(PlannerInfo *root, RelOptInfo *rel, + bool require_parallel_safe) +{ + List *useful_pathkeys_list = NIL; + + /* + * Considering query_pathkeys is always worth it, because it might allow + * us to avoid a total sort when we have a partially presorted path + * available or to push the total sort into the parallel portion of the + * query. + */ + if (root->query_pathkeys) + { + ListCell *lc; + int npathkeys = 0; /* useful pathkeys */ + + foreach(lc, root->query_pathkeys) + { + PathKey *pathkey = (PathKey *) lfirst(lc); + EquivalenceClass *pathkey_ec = pathkey->pk_eclass; + + /* + * We can only build a sort for pathkeys that contain a + * safe-to-compute-early EC member computable from the current + * relation's reltarget, so ignore the remainder of the list as + * soon as we find a pathkey without such a member. + * + * It's still worthwhile to return any prefix of the pathkeys list + * that meets this requirement, as we may be able to do an + * incremental sort. + * + * If requested, ensure the sort expression is parallel-safe too. + */ + if (!relation_can_be_sorted_early(root, rel, pathkey_ec, + require_parallel_safe)) + break; + + npathkeys++; + } + + /* + * The whole query_pathkeys list matches, so append it directly, to + * allow comparing pathkeys easily by comparing list pointer. If we + * have to truncate the pathkeys, we gotta do a copy though. + */ + if (npathkeys == list_length(root->query_pathkeys)) + useful_pathkeys_list = lappend(useful_pathkeys_list, + root->query_pathkeys); + else if (npathkeys > 0) + useful_pathkeys_list = lappend(useful_pathkeys_list, + list_truncate(list_copy(root->query_pathkeys), + npathkeys)); + } + + return useful_pathkeys_list; +} + +/* + * generate_useful_gather_paths + * Generate parallel access paths for a relation by pushing a Gather or + * Gather Merge on top of a partial path. + * + * Unlike plain generate_gather_paths, this looks both at pathkeys of input + * paths (aiming to preserve the ordering), but also considers ordering that + * might be useful for nodes above the gather merge node, and tries to add + * a sort (regular or incremental) to provide that. + */ +void +generate_useful_gather_paths(PlannerInfo *root, RelOptInfo *rel, bool override_rows) +{ + ListCell *lc; + double rows; + double *rowsp = NULL; + List *useful_pathkeys_list = NIL; + Path *cheapest_partial_path = NULL; + + /* If there are no partial paths, there's nothing to do here. */ + if (rel->partial_pathlist == NIL) + return; + + /* Should we override the rel's rowcount estimate? */ + if (override_rows) + rowsp = &rows; + + /* generate the regular gather (merge) paths */ + generate_gather_paths(root, rel, override_rows); + + /* consider incremental sort for interesting orderings */ + useful_pathkeys_list = get_useful_pathkeys_for_relation(root, rel, true); + + /* used for explicit (full) sort paths */ + cheapest_partial_path = linitial(rel->partial_pathlist); + + /* + * Consider sorted paths for each interesting ordering. We generate both + * incremental and full sort. + */ + foreach(lc, useful_pathkeys_list) + { + List *useful_pathkeys = lfirst(lc); + ListCell *lc2; + bool is_sorted; + int presorted_keys; + + foreach(lc2, rel->partial_pathlist) + { + Path *subpath = (Path *) lfirst(lc2); + GatherMergePath *path; + + is_sorted = pathkeys_count_contained_in(useful_pathkeys, + subpath->pathkeys, + &presorted_keys); + + /* + * We don't need to consider the case where a subpath is already + * fully sorted because generate_gather_paths already creates a + * gather merge path for every subpath that has pathkeys present. + * + * But since the subpath is already sorted, we know we don't need + * to consider adding a sort (full or incremental) on top of it, + * so we can continue here. + */ + if (is_sorted) + continue; + + /* + * Consider regular sort for the cheapest partial path (for each + * useful pathkeys). We know the path is not sorted, because we'd + * not get here otherwise. + * + * This is not redundant with the gather paths created in + * generate_gather_paths, because that doesn't generate ordered + * output. Here we add an explicit sort to match the useful + * ordering. + */ + if (cheapest_partial_path == subpath) + { + Path *tmp; + + tmp = (Path *) create_sort_path(root, + rel, + subpath, + useful_pathkeys, + -1.0); + + rows = tmp->rows * tmp->parallel_workers; + + path = create_gather_merge_path(root, rel, + tmp, + rel->reltarget, + tmp->pathkeys, + NULL, + rowsp); + + add_path(rel, &path->path); + + /* Fall through */ + } + + /* + * Consider incremental sort, but only when the subpath is already + * partially sorted on a pathkey prefix. + */ + if (enable_incremental_sort && presorted_keys > 0) + { + Path *tmp; + + /* + * We should have already excluded pathkeys of length 1 + * because then presorted_keys > 0 would imply is_sorted was + * true. + */ + Assert(list_length(useful_pathkeys) != 1); + + tmp = (Path *) create_incremental_sort_path(root, + rel, + subpath, + useful_pathkeys, + presorted_keys, + -1); + + path = create_gather_merge_path(root, rel, + tmp, + rel->reltarget, + tmp->pathkeys, + NULL, + rowsp); + + add_path(rel, &path->path); + } + } + } +} + +/* + * make_rel_from_joinlist + * Build access paths using a "joinlist" to guide the join path search. + * + * See comments for deconstruct_jointree() for definition of the joinlist + * data structure. + */ +static RelOptInfo * +make_rel_from_joinlist(PlannerInfo *root, List *joinlist) +{ + int levels_needed; + List *initial_rels; + ListCell *jl; + + /* + * Count the number of child joinlist nodes. This is the depth of the + * dynamic-programming algorithm we must employ to consider all ways of + * joining the child nodes. + */ + levels_needed = list_length(joinlist); + + if (levels_needed <= 0) + return NULL; /* nothing to do? */ + + /* + * Construct a list of rels corresponding to the child joinlist nodes. + * This may contain both base rels and rels constructed according to + * sub-joinlists. + */ + initial_rels = NIL; + foreach(jl, joinlist) + { + Node *jlnode = (Node *) lfirst(jl); + RelOptInfo *thisrel; + + if (IsA(jlnode, RangeTblRef)) + { + int varno = ((RangeTblRef *) jlnode)->rtindex; + + thisrel = find_base_rel(root, varno); + } + else if (IsA(jlnode, List)) + { + /* Recurse to handle subproblem */ + thisrel = make_rel_from_joinlist(root, (List *) jlnode); + } + else + { + elog(ERROR, "unrecognized joinlist node type: %d", + (int) nodeTag(jlnode)); + thisrel = NULL; /* keep compiler quiet */ + } + + initial_rels = lappend(initial_rels, thisrel); + } + + if (levels_needed == 1) + { + /* + * Single joinlist node, so we're done. + */ + return (RelOptInfo *) linitial(initial_rels); + } + else + { + /* + * Consider the different orders in which we could join the rels, + * using a plugin, GEQO, or the regular join search code. + * + * We put the initial_rels list into a PlannerInfo field because + * has_legal_joinclause() needs to look at it (ugly :-(). + */ + root->initial_rels = initial_rels; + + if (join_search_hook) + return (*join_search_hook) (root, levels_needed, initial_rels); + else if (enable_geqo && levels_needed >= geqo_threshold) + return geqo(root, levels_needed, initial_rels); + else + return standard_join_search(root, levels_needed, initial_rels); + } +} + +/* + * standard_join_search + * Find possible joinpaths for a query by successively finding ways + * to join component relations into join relations. + * + * 'levels_needed' is the number of iterations needed, ie, the number of + * independent jointree items in the query. This is > 1. + * + * 'initial_rels' is a list of RelOptInfo nodes for each independent + * jointree item. These are the components to be joined together. + * Note that levels_needed == list_length(initial_rels). + * + * Returns the final level of join relations, i.e., the relation that is + * the result of joining all the original relations together. + * At least one implementation path must be provided for this relation and + * all required sub-relations. + * + * To support loadable plugins that modify planner behavior by changing the + * join searching algorithm, we provide a hook variable that lets a plugin + * replace or supplement this function. Any such hook must return the same + * final join relation as the standard code would, but it might have a + * different set of implementation paths attached, and only the sub-joinrels + * needed for these paths need have been instantiated. + * + * Note to plugin authors: the functions invoked during standard_join_search() + * modify root->join_rel_list and root->join_rel_hash. If you want to do more + * than one join-order search, you'll probably need to save and restore the + * original states of those data structures. See geqo_eval() for an example. + */ +RelOptInfo * +standard_join_search(PlannerInfo *root, int levels_needed, List *initial_rels) +{ + int lev; + RelOptInfo *rel; + + /* + * This function cannot be invoked recursively within any one planning + * problem, so join_rel_level[] can't be in use already. + */ + Assert(root->join_rel_level == NULL); + + /* + * We employ a simple "dynamic programming" algorithm: we first find all + * ways to build joins of two jointree items, then all ways to build joins + * of three items (from two-item joins and single items), then four-item + * joins, and so on until we have considered all ways to join all the + * items into one rel. + * + * root->join_rel_level[j] is a list of all the j-item rels. Initially we + * set root->join_rel_level[1] to represent all the single-jointree-item + * relations. + */ + root->join_rel_level = (List **) palloc0((levels_needed + 1) * sizeof(List *)); + + root->join_rel_level[1] = initial_rels; + + for (lev = 2; lev <= levels_needed; lev++) + { + ListCell *lc; + + /* + * Determine all possible pairs of relations to be joined at this + * level, and build paths for making each one from every available + * pair of lower-level relations. + */ + join_search_one_level(root, lev); + + /* + * Run generate_partitionwise_join_paths() and + * generate_useful_gather_paths() for each just-processed joinrel. We + * could not do this earlier because both regular and partial paths + * can get added to a particular joinrel at multiple times within + * join_search_one_level. + * + * After that, we're done creating paths for the joinrel, so run + * set_cheapest(). + */ + foreach(lc, root->join_rel_level[lev]) + { + rel = (RelOptInfo *) lfirst(lc); + + /* Create paths for partitionwise joins. */ + generate_partitionwise_join_paths(root, rel); + + /* + * Except for the topmost scan/join rel, consider gathering + * partial paths. We'll do the same for the topmost scan/join rel + * once we know the final targetlist (see grouping_planner). + */ + if (!bms_equal(rel->relids, root->all_baserels)) + generate_useful_gather_paths(root, rel, false); + + /* Find and save the cheapest paths for this rel */ + set_cheapest(rel); + +#ifdef OPTIMIZER_DEBUG + debug_print_rel(root, rel); +#endif + } + } + + /* + * We should have a single rel at the final level. + */ + if (root->join_rel_level[levels_needed] == NIL) + elog(ERROR, "failed to build any %d-way joins", levels_needed); + Assert(list_length(root->join_rel_level[levels_needed]) == 1); + + rel = (RelOptInfo *) linitial(root->join_rel_level[levels_needed]); + + root->join_rel_level = NULL; + + return rel; +} + +/***************************************************************************** + * PUSHING QUALS DOWN INTO SUBQUERIES + *****************************************************************************/ + +/* + * subquery_is_pushdown_safe - is a subquery safe for pushing down quals? + * + * subquery is the particular component query being checked. topquery + * is the top component of a set-operations tree (the same Query if no + * set-op is involved). + * + * Conditions checked here: + * + * 1. If the subquery has a LIMIT clause, we must not push down any quals, + * since that could change the set of rows returned. + * + * 2. If the subquery contains EXCEPT or EXCEPT ALL set ops we cannot push + * quals into it, because that could change the results. + * + * 3. If the subquery uses DISTINCT, we cannot push volatile quals into it. + * This is because upper-level quals should semantically be evaluated only + * once per distinct row, not once per original row, and if the qual is + * volatile then extra evaluations could change the results. (This issue + * does not apply to other forms of aggregation such as GROUP BY, because + * when those are present we push into HAVING not WHERE, so that the quals + * are still applied after aggregation.) + * + * 4. If the subquery contains window functions, we cannot push volatile quals + * into it. The issue here is a bit different from DISTINCT: a volatile qual + * might succeed for some rows of a window partition and fail for others, + * thereby changing the partition contents and thus the window functions' + * results for rows that remain. + * + * 5. If the subquery contains any set-returning functions in its targetlist, + * we cannot push volatile quals into it. That would push them below the SRFs + * and thereby change the number of times they are evaluated. Also, a + * volatile qual could succeed for some SRF output rows and fail for others, + * a behavior that cannot occur if it's evaluated before SRF expansion. + * + * 6. If the subquery has nonempty grouping sets, we cannot push down any + * quals. The concern here is that a qual referencing a "constant" grouping + * column could get constant-folded, which would be improper because the value + * is potentially nullable by grouping-set expansion. This restriction could + * be removed if we had a parsetree representation that shows that such + * grouping columns are not really constant. (There are other ideas that + * could be used to relax this restriction, but that's the approach most + * likely to get taken in the future. Note that there's not much to be gained + * so long as subquery_planner can't move HAVING clauses to WHERE within such + * a subquery.) + * + * In addition, we make several checks on the subquery's output columns to see + * if it is safe to reference them in pushed-down quals. If output column k + * is found to be unsafe to reference, we set the reason for that inside + * safetyInfo->unsafeFlags[k], but we don't reject the subquery overall since + * column k might not be referenced by some/all quals. The unsafeFlags[] + * array will be consulted later by qual_is_pushdown_safe(). It's better to + * do it this way than to make the checks directly in qual_is_pushdown_safe(), + * because when the subquery involves set operations we have to check the + * output expressions in each arm of the set op. + * + * Note: pushing quals into a DISTINCT subquery is theoretically dubious: + * we're effectively assuming that the quals cannot distinguish values that + * the DISTINCT's equality operator sees as equal, yet there are many + * counterexamples to that assumption. However use of such a qual with a + * DISTINCT subquery would be unsafe anyway, since there's no guarantee which + * "equal" value will be chosen as the output value by the DISTINCT operation. + * So we don't worry too much about that. Another objection is that if the + * qual is expensive to evaluate, running it for each original row might cost + * more than we save by eliminating rows before the DISTINCT step. But it + * would be very hard to estimate that at this stage, and in practice pushdown + * seldom seems to make things worse, so we ignore that problem too. + * + * Note: likewise, pushing quals into a subquery with window functions is a + * bit dubious: the quals might remove some rows of a window partition while + * leaving others, causing changes in the window functions' results for the + * surviving rows. We insist that such a qual reference only partitioning + * columns, but again that only protects us if the qual does not distinguish + * values that the partitioning equality operator sees as equal. The risks + * here are perhaps larger than for DISTINCT, since no de-duplication of rows + * occurs and thus there is no theoretical problem with such a qual. But + * we'll do this anyway because the potential performance benefits are very + * large, and we've seen no field complaints about the longstanding comparable + * behavior with DISTINCT. + */ +static bool +subquery_is_pushdown_safe(Query *subquery, Query *topquery, + pushdown_safety_info *safetyInfo) +{ + SetOperationStmt *topop; + + /* Check point 1 */ + if (subquery->limitOffset != NULL || subquery->limitCount != NULL) + return false; + + /* Check point 6 */ + if (subquery->groupClause && subquery->groupingSets) + return false; + + /* Check points 3, 4, and 5 */ + if (subquery->distinctClause || + subquery->hasWindowFuncs || + subquery->hasTargetSRFs) + safetyInfo->unsafeVolatile = true; + + /* + * If we're at a leaf query, check for unsafe expressions in its target + * list, and mark any reasons why they're unsafe in unsafeFlags[]. + * (Non-leaf nodes in setop trees have only simple Vars in their tlists, + * so no need to check them.) + */ + if (subquery->setOperations == NULL) + check_output_expressions(subquery, safetyInfo); + + /* Are we at top level, or looking at a setop component? */ + if (subquery == topquery) + { + /* Top level, so check any component queries */ + if (subquery->setOperations != NULL) + if (!recurse_pushdown_safe(subquery->setOperations, topquery, + safetyInfo)) + return false; + } + else + { + /* Setop component must not have more components (too weird) */ + if (subquery->setOperations != NULL) + return false; + /* Check whether setop component output types match top level */ + topop = castNode(SetOperationStmt, topquery->setOperations); + Assert(topop); + compare_tlist_datatypes(subquery->targetList, + topop->colTypes, + safetyInfo); + } + return true; +} + +/* + * Helper routine to recurse through setOperations tree + */ +static bool +recurse_pushdown_safe(Node *setOp, Query *topquery, + pushdown_safety_info *safetyInfo) +{ + if (IsA(setOp, RangeTblRef)) + { + RangeTblRef *rtr = (RangeTblRef *) setOp; + RangeTblEntry *rte = rt_fetch(rtr->rtindex, topquery->rtable); + Query *subquery = rte->subquery; + + Assert(subquery != NULL); + return subquery_is_pushdown_safe(subquery, topquery, safetyInfo); + } + else if (IsA(setOp, SetOperationStmt)) + { + SetOperationStmt *op = (SetOperationStmt *) setOp; + + /* EXCEPT is no good (point 2 for subquery_is_pushdown_safe) */ + if (op->op == SETOP_EXCEPT) + return false; + /* Else recurse */ + if (!recurse_pushdown_safe(op->larg, topquery, safetyInfo)) + return false; + if (!recurse_pushdown_safe(op->rarg, topquery, safetyInfo)) + return false; + } + else + { + elog(ERROR, "unrecognized node type: %d", + (int) nodeTag(setOp)); + } + return true; +} + +/* + * check_output_expressions - check subquery's output expressions for safety + * + * There are several cases in which it's unsafe to push down an upper-level + * qual if it references a particular output column of a subquery. We check + * each output column of the subquery and set flags in unsafeFlags[k] when we + * see that column is unsafe for a pushed-down qual to reference. The + * conditions checked here are: + * + * 1. We must not push down any quals that refer to subselect outputs that + * return sets, else we'd introduce functions-returning-sets into the + * subquery's WHERE/HAVING quals. + * + * 2. We must not push down any quals that refer to subselect outputs that + * contain volatile functions, for fear of introducing strange results due + * to multiple evaluation of a volatile function. + * + * 3. If the subquery uses DISTINCT ON, we must not push down any quals that + * refer to non-DISTINCT output columns, because that could change the set + * of rows returned. (This condition is vacuous for DISTINCT, because then + * there are no non-DISTINCT output columns, so we needn't check. Note that + * subquery_is_pushdown_safe already reported that we can't use volatile + * quals if there's DISTINCT or DISTINCT ON.) + * + * 4. If the subquery has any window functions, we must not push down quals + * that reference any output columns that are not listed in all the subquery's + * window PARTITION BY clauses. We can push down quals that use only + * partitioning columns because they should succeed or fail identically for + * every row of any one window partition, and totally excluding some + * partitions will not change a window function's results for remaining + * partitions. (Again, this also requires nonvolatile quals, but + * subquery_is_pushdown_safe handles that.). Subquery columns marked as + * unsafe for this reason can still have WindowClause run conditions pushed + * down. + */ +static void +check_output_expressions(Query *subquery, pushdown_safety_info *safetyInfo) +{ + ListCell *lc; + + foreach(lc, subquery->targetList) + { + TargetEntry *tle = (TargetEntry *) lfirst(lc); + + if (tle->resjunk) + continue; /* ignore resjunk columns */ + + /* Functions returning sets are unsafe (point 1) */ + if (subquery->hasTargetSRFs && + (safetyInfo->unsafeFlags[tle->resno] & + UNSAFE_HAS_SET_FUNC) == 0 && + expression_returns_set((Node *) tle->expr)) + { + safetyInfo->unsafeFlags[tle->resno] |= UNSAFE_HAS_SET_FUNC; + continue; + } + + /* Volatile functions are unsafe (point 2) */ + if ((safetyInfo->unsafeFlags[tle->resno] & + UNSAFE_HAS_VOLATILE_FUNC) == 0 && + contain_volatile_functions((Node *) tle->expr)) + { + safetyInfo->unsafeFlags[tle->resno] |= UNSAFE_HAS_VOLATILE_FUNC; + continue; + } + + /* If subquery uses DISTINCT ON, check point 3 */ + if (subquery->hasDistinctOn && + (safetyInfo->unsafeFlags[tle->resno] & + UNSAFE_NOTIN_DISTINCTON_CLAUSE) == 0 && + !targetIsInSortList(tle, InvalidOid, subquery->distinctClause)) + { + /* non-DISTINCT column, so mark it unsafe */ + safetyInfo->unsafeFlags[tle->resno] |= UNSAFE_NOTIN_DISTINCTON_CLAUSE; + continue; + } + + /* If subquery uses window functions, check point 4 */ + if (subquery->hasWindowFuncs && + (safetyInfo->unsafeFlags[tle->resno] & + UNSAFE_NOTIN_DISTINCTON_CLAUSE) == 0 && + !targetIsInAllPartitionLists(tle, subquery)) + { + /* not present in all PARTITION BY clauses, so mark it unsafe */ + safetyInfo->unsafeFlags[tle->resno] |= UNSAFE_NOTIN_PARTITIONBY_CLAUSE; + continue; + } + } +} + +/* + * For subqueries using UNION/UNION ALL/INTERSECT/INTERSECT ALL, we can + * push quals into each component query, but the quals can only reference + * subquery columns that suffer no type coercions in the set operation. + * Otherwise there are possible semantic gotchas. So, we check the + * component queries to see if any of them have output types different from + * the top-level setop outputs. We set the UNSAFE_TYPE_MISMATCH bit in + * unsafeFlags[k] if column k has different type in any component. + * + * We don't have to care about typmods here: the only allowed difference + * between set-op input and output typmods is input is a specific typmod + * and output is -1, and that does not require a coercion. + * + * tlist is a subquery tlist. + * colTypes is an OID list of the top-level setop's output column types. + * safetyInfo is the pushdown_safety_info to set unsafeFlags[] for. + */ +static void +compare_tlist_datatypes(List *tlist, List *colTypes, + pushdown_safety_info *safetyInfo) +{ + ListCell *l; + ListCell *colType = list_head(colTypes); + + foreach(l, tlist) + { + TargetEntry *tle = (TargetEntry *) lfirst(l); + + if (tle->resjunk) + continue; /* ignore resjunk columns */ + if (colType == NULL) + elog(ERROR, "wrong number of tlist entries"); + if (exprType((Node *) tle->expr) != lfirst_oid(colType)) + safetyInfo->unsafeFlags[tle->resno] |= UNSAFE_TYPE_MISMATCH; + colType = lnext(colTypes, colType); + } + if (colType != NULL) + elog(ERROR, "wrong number of tlist entries"); +} + +/* + * targetIsInAllPartitionLists + * True if the TargetEntry is listed in the PARTITION BY clause + * of every window defined in the query. + * + * It would be safe to ignore windows not actually used by any window + * function, but it's not easy to get that info at this stage; and it's + * unlikely to be useful to spend any extra cycles getting it, since + * unreferenced window definitions are probably infrequent in practice. + */ +static bool +targetIsInAllPartitionLists(TargetEntry *tle, Query *query) +{ + ListCell *lc; + + foreach(lc, query->windowClause) + { + WindowClause *wc = (WindowClause *) lfirst(lc); + + if (!targetIsInSortList(tle, InvalidOid, wc->partitionClause)) + return false; + } + return true; +} + +/* + * qual_is_pushdown_safe - is a particular rinfo safe to push down? + * + * rinfo is a restriction clause applying to the given subquery (whose RTE + * has index rti in the parent query). + * + * Conditions checked here: + * + * 1. rinfo's clause must not contain any SubPlans (mainly because it's + * unclear that it will work correctly: SubLinks will already have been + * transformed into SubPlans in the qual, but not in the subquery). Note that + * SubLinks that transform to initplans are safe, and will be accepted here + * because what we'll see in the qual is just a Param referencing the initplan + * output. + * + * 2. If unsafeVolatile is set, rinfo's clause must not contain any volatile + * functions. + * + * 3. If unsafeLeaky is set, rinfo's clause must not contain any leaky + * functions that are passed Var nodes, and therefore might reveal values from + * the subquery as side effects. + * + * 4. rinfo's clause must not refer to the whole-row output of the subquery + * (since there is no easy way to name that within the subquery itself). + * + * 5. rinfo's clause must not refer to any subquery output columns that were + * found to be unsafe to reference by subquery_is_pushdown_safe(). + */ +static pushdown_safe_type +qual_is_pushdown_safe(Query *subquery, Index rti, RestrictInfo *rinfo, + pushdown_safety_info *safetyInfo) +{ + pushdown_safe_type safe = PUSHDOWN_SAFE; + Node *qual = (Node *) rinfo->clause; + List *vars; + ListCell *vl; + + /* Refuse subselects (point 1) */ + if (contain_subplans(qual)) + return PUSHDOWN_UNSAFE; + + /* Refuse volatile quals if we found they'd be unsafe (point 2) */ + if (safetyInfo->unsafeVolatile && + contain_volatile_functions((Node *) rinfo)) + return PUSHDOWN_UNSAFE; + + /* Refuse leaky quals if told to (point 3) */ + if (safetyInfo->unsafeLeaky && + contain_leaked_vars(qual)) + return PUSHDOWN_UNSAFE; + + /* + * It would be unsafe to push down window function calls, but at least for + * the moment we could never see any in a qual anyhow. (The same applies + * to aggregates, which we check for in pull_var_clause below.) + */ + Assert(!contain_window_function(qual)); + + /* + * Examine all Vars used in clause. Since it's a restriction clause, all + * such Vars must refer to subselect output columns ... unless this is + * part of a LATERAL subquery, in which case there could be lateral + * references. + */ + vars = pull_var_clause(qual, PVC_INCLUDE_PLACEHOLDERS); + foreach(vl, vars) + { + Var *var = (Var *) lfirst(vl); + + /* + * XXX Punt if we find any PlaceHolderVars in the restriction clause. + * It's not clear whether a PHV could safely be pushed down, and even + * less clear whether such a situation could arise in any cases of + * practical interest anyway. So for the moment, just refuse to push + * down. + */ + if (!IsA(var, Var)) + { + safe = PUSHDOWN_UNSAFE; + break; + } + + /* + * Punt if we find any lateral references. It would be safe to push + * these down, but we'd have to convert them into outer references, + * which subquery_push_qual lacks the infrastructure to do. The case + * arises so seldom that it doesn't seem worth working hard on. + */ + if (var->varno != rti) + { + safe = PUSHDOWN_UNSAFE; + break; + } + + /* Subqueries have no system columns */ + Assert(var->varattno >= 0); + + /* Check point 4 */ + if (var->varattno == 0) + { + safe = PUSHDOWN_UNSAFE; + break; + } + + /* Check point 5 */ + if (safetyInfo->unsafeFlags[var->varattno] != 0) + { + if (safetyInfo->unsafeFlags[var->varattno] & + (UNSAFE_HAS_VOLATILE_FUNC | UNSAFE_HAS_SET_FUNC | + UNSAFE_NOTIN_DISTINCTON_CLAUSE | UNSAFE_TYPE_MISMATCH)) + { + safe = PUSHDOWN_UNSAFE; + break; + } + else + { + /* UNSAFE_NOTIN_PARTITIONBY_CLAUSE is ok for run conditions */ + safe = PUSHDOWN_WINDOWCLAUSE_RUNCOND; + /* don't break, we might find another Var that's unsafe */ + } + } + } + + list_free(vars); + + return safe; +} + +/* + * subquery_push_qual - push down a qual that we have determined is safe + */ +static void +subquery_push_qual(Query *subquery, RangeTblEntry *rte, Index rti, Node *qual) +{ + if (subquery->setOperations != NULL) + { + /* Recurse to push it separately to each component query */ + recurse_push_qual(subquery->setOperations, subquery, + rte, rti, qual); + } + else + { + /* + * We need to replace Vars in the qual (which must refer to outputs of + * the subquery) with copies of the subquery's targetlist expressions. + * Note that at this point, any uplevel Vars in the qual should have + * been replaced with Params, so they need no work. + * + * This step also ensures that when we are pushing into a setop tree, + * each component query gets its own copy of the qual. + */ + qual = ReplaceVarsFromTargetList(qual, rti, 0, rte, + subquery->targetList, + REPLACEVARS_REPORT_ERROR, 0, + &subquery->hasSubLinks); + + /* + * Now attach the qual to the proper place: normally WHERE, but if the + * subquery uses grouping or aggregation, put it in HAVING (since the + * qual really refers to the group-result rows). + */ + if (subquery->hasAggs || subquery->groupClause || subquery->groupingSets || subquery->havingQual) + subquery->havingQual = make_and_qual(subquery->havingQual, qual); + else + subquery->jointree->quals = + make_and_qual(subquery->jointree->quals, qual); + + /* + * We need not change the subquery's hasAggs or hasSubLinks flags, + * since we can't be pushing down any aggregates that weren't there + * before, and we don't push down subselects at all. + */ + } +} + +/* + * Helper routine to recurse through setOperations tree + */ +static void +recurse_push_qual(Node *setOp, Query *topquery, + RangeTblEntry *rte, Index rti, Node *qual) +{ + if (IsA(setOp, RangeTblRef)) + { + RangeTblRef *rtr = (RangeTblRef *) setOp; + RangeTblEntry *subrte = rt_fetch(rtr->rtindex, topquery->rtable); + Query *subquery = subrte->subquery; + + Assert(subquery != NULL); + subquery_push_qual(subquery, rte, rti, qual); + } + else if (IsA(setOp, SetOperationStmt)) + { + SetOperationStmt *op = (SetOperationStmt *) setOp; + + recurse_push_qual(op->larg, topquery, rte, rti, qual); + recurse_push_qual(op->rarg, topquery, rte, rti, qual); + } + else + { + elog(ERROR, "unrecognized node type: %d", + (int) nodeTag(setOp)); + } +} + +/***************************************************************************** + * SIMPLIFYING SUBQUERY TARGETLISTS + *****************************************************************************/ + +/* + * remove_unused_subquery_outputs + * Remove subquery targetlist items we don't need + * + * It's possible, even likely, that the upper query does not read all the + * output columns of the subquery. We can remove any such outputs that are + * not needed by the subquery itself (e.g., as sort/group columns) and do not + * affect semantics otherwise (e.g., volatile functions can't be removed). + * This is useful not only because we might be able to remove expensive-to- + * compute expressions, but because deletion of output columns might allow + * optimizations such as join removal to occur within the subquery. + * + * extra_used_attrs can be passed as non-NULL to mark any columns (offset by + * FirstLowInvalidHeapAttributeNumber) that we should not remove. This + * parameter is modifed by the function, so callers must make a copy if they + * need to use the passed in Bitmapset after calling this function. + * + * To avoid affecting column numbering in the targetlist, we don't physically + * remove unused tlist entries, but rather replace their expressions with NULL + * constants. This is implemented by modifying subquery->targetList. + */ +static void +remove_unused_subquery_outputs(Query *subquery, RelOptInfo *rel, + Bitmapset *extra_used_attrs) +{ + Bitmapset *attrs_used; + ListCell *lc; + + /* + * Just point directly to extra_used_attrs. No need to bms_copy as none of + * the current callers use the Bitmapset after calling this function. + */ + attrs_used = extra_used_attrs; + + /* + * Do nothing if subquery has UNION/INTERSECT/EXCEPT: in principle we + * could update all the child SELECTs' tlists, but it seems not worth the + * trouble presently. + */ + if (subquery->setOperations) + return; + + /* + * If subquery has regular DISTINCT (not DISTINCT ON), we're wasting our + * time: all its output columns must be used in the distinctClause. + */ + if (subquery->distinctClause && !subquery->hasDistinctOn) + return; + + /* + * Collect a bitmap of all the output column numbers used by the upper + * query. + * + * Add all the attributes needed for joins or final output. Note: we must + * look at rel's targetlist, not the attr_needed data, because attr_needed + * isn't computed for inheritance child rels, cf set_append_rel_size(). + * (XXX might be worth changing that sometime.) + */ + pull_varattnos((Node *) rel->reltarget->exprs, rel->relid, &attrs_used); + + /* Add all the attributes used by un-pushed-down restriction clauses. */ + foreach(lc, rel->baserestrictinfo) + { + RestrictInfo *rinfo = (RestrictInfo *) lfirst(lc); + + pull_varattnos((Node *) rinfo->clause, rel->relid, &attrs_used); + } + + /* + * If there's a whole-row reference to the subquery, we can't remove + * anything. + */ + if (bms_is_member(0 - FirstLowInvalidHeapAttributeNumber, attrs_used)) + return; + + /* + * Run through the tlist and zap entries we don't need. It's okay to + * modify the tlist items in-place because set_subquery_pathlist made a + * copy of the subquery. + */ + foreach(lc, subquery->targetList) + { + TargetEntry *tle = (TargetEntry *) lfirst(lc); + Node *texpr = (Node *) tle->expr; + + /* + * If it has a sortgroupref number, it's used in some sort/group + * clause so we'd better not remove it. Also, don't remove any + * resjunk columns, since their reason for being has nothing to do + * with anybody reading the subquery's output. (It's likely that + * resjunk columns in a sub-SELECT would always have ressortgroupref + * set, but even if they don't, it seems imprudent to remove them.) + */ + if (tle->ressortgroupref || tle->resjunk) + continue; + + /* + * If it's used by the upper query, we can't remove it. + */ + if (bms_is_member(tle->resno - FirstLowInvalidHeapAttributeNumber, + attrs_used)) + continue; + + /* + * If it contains a set-returning function, we can't remove it since + * that could change the number of rows returned by the subquery. + */ + if (subquery->hasTargetSRFs && + expression_returns_set(texpr)) + continue; + + /* + * If it contains volatile functions, we daren't remove it for fear + * that the user is expecting their side-effects to happen. + */ + if (contain_volatile_functions(texpr)) + continue; + + /* + * OK, we don't need it. Replace the expression with a NULL constant. + * Preserve the exposed type of the expression, in case something + * looks at the rowtype of the subquery's result. + */ + tle->expr = (Expr *) makeNullConst(exprType(texpr), + exprTypmod(texpr), + exprCollation(texpr)); + } +} + +/* + * create_partial_bitmap_paths + * Build partial bitmap heap path for the relation + */ +void +create_partial_bitmap_paths(PlannerInfo *root, RelOptInfo *rel, + Path *bitmapqual) +{ + int parallel_workers; + double pages_fetched; + + /* Compute heap pages for bitmap heap scan */ + pages_fetched = compute_bitmap_pages(root, rel, bitmapqual, 1.0, + NULL, NULL); + + parallel_workers = compute_parallel_worker(rel, pages_fetched, -1, + max_parallel_workers_per_gather); + + if (parallel_workers <= 0) + return; + + add_partial_path(rel, (Path *) create_bitmap_heap_path(root, rel, + bitmapqual, rel->lateral_relids, 1.0, parallel_workers)); +} + +/* + * Compute the number of parallel workers that should be used to scan a + * relation. We compute the parallel workers based on the size of the heap to + * be scanned and the size of the index to be scanned, then choose a minimum + * of those. + * + * "heap_pages" is the number of pages from the table that we expect to scan, or + * -1 if we don't expect to scan any. + * + * "index_pages" is the number of pages from the index that we expect to scan, or + * -1 if we don't expect to scan any. + * + * "max_workers" is caller's limit on the number of workers. This typically + * comes from a GUC. + */ +int +compute_parallel_worker(RelOptInfo *rel, double heap_pages, double index_pages, + int max_workers) +{ + int parallel_workers = 0; + + /* + * If the user has set the parallel_workers reloption, use that; otherwise + * select a default number of workers. + */ + if (rel->rel_parallel_workers != -1) + parallel_workers = rel->rel_parallel_workers; + else + { + /* + * If the number of pages being scanned is insufficient to justify a + * parallel scan, just return zero ... unless it's an inheritance + * child. In that case, we want to generate a parallel path here + * anyway. It might not be worthwhile just for this relation, but + * when combined with all of its inheritance siblings it may well pay + * off. + */ + if (rel->reloptkind == RELOPT_BASEREL && + ((heap_pages >= 0 && heap_pages < min_parallel_table_scan_size) || + (index_pages >= 0 && index_pages < min_parallel_index_scan_size))) + return 0; + + if (heap_pages >= 0) + { + int heap_parallel_threshold; + int heap_parallel_workers = 1; + + /* + * Select the number of workers based on the log of the size of + * the relation. This probably needs to be a good deal more + * sophisticated, but we need something here for now. Note that + * the upper limit of the min_parallel_table_scan_size GUC is + * chosen to prevent overflow here. + */ + heap_parallel_threshold = Max(min_parallel_table_scan_size, 1); + while (heap_pages >= (BlockNumber) (heap_parallel_threshold * 3)) + { + heap_parallel_workers++; + heap_parallel_threshold *= 3; + if (heap_parallel_threshold > INT_MAX / 3) + break; /* avoid overflow */ + } + + parallel_workers = heap_parallel_workers; + } + + if (index_pages >= 0) + { + int index_parallel_workers = 1; + int index_parallel_threshold; + + /* same calculation as for heap_pages above */ + index_parallel_threshold = Max(min_parallel_index_scan_size, 1); + while (index_pages >= (BlockNumber) (index_parallel_threshold * 3)) + { + index_parallel_workers++; + index_parallel_threshold *= 3; + if (index_parallel_threshold > INT_MAX / 3) + break; /* avoid overflow */ + } + + if (parallel_workers > 0) + parallel_workers = Min(parallel_workers, index_parallel_workers); + else + parallel_workers = index_parallel_workers; + } + } + + /* In no case use more than caller supplied maximum number of workers */ + parallel_workers = Min(parallel_workers, max_workers); + + return parallel_workers; +} + +/* + * generate_partitionwise_join_paths + * Create paths representing partitionwise join for given partitioned + * join relation. + * + * This must not be called until after we are done adding paths for all + * child-joins. Otherwise, add_path might delete a path to which some path + * generated here has a reference. + */ +void +generate_partitionwise_join_paths(PlannerInfo *root, RelOptInfo *rel) +{ + List *live_children = NIL; + int cnt_parts; + int num_parts; + RelOptInfo **part_rels; + + /* Handle only join relations here. */ + if (!IS_JOIN_REL(rel)) + return; + + /* We've nothing to do if the relation is not partitioned. */ + if (!IS_PARTITIONED_REL(rel)) + return; + + /* The relation should have consider_partitionwise_join set. */ + Assert(rel->consider_partitionwise_join); + + /* Guard against stack overflow due to overly deep partition hierarchy. */ + check_stack_depth(); + + num_parts = rel->nparts; + part_rels = rel->part_rels; + + /* Collect non-dummy child-joins. */ + for (cnt_parts = 0; cnt_parts < num_parts; cnt_parts++) + { + RelOptInfo *child_rel = part_rels[cnt_parts]; + + /* If it's been pruned entirely, it's certainly dummy. */ + if (child_rel == NULL) + continue; + + /* Make partitionwise join paths for this partitioned child-join. */ + generate_partitionwise_join_paths(root, child_rel); + + /* If we failed to make any path for this child, we must give up. */ + if (child_rel->pathlist == NIL) + { + /* + * Mark the parent joinrel as unpartitioned so that later + * functions treat it correctly. + */ + rel->nparts = 0; + return; + } + + /* Else, identify the cheapest path for it. */ + set_cheapest(child_rel); + + /* Dummy children need not be scanned, so ignore those. */ + if (IS_DUMMY_REL(child_rel)) + continue; + +#ifdef OPTIMIZER_DEBUG + debug_print_rel(root, child_rel); +#endif + + live_children = lappend(live_children, child_rel); + } + + /* If all child-joins are dummy, parent join is also dummy. */ + if (!live_children) + { + mark_dummy_rel(rel); + return; + } + + /* Build additional paths for this rel from child-join paths. */ + add_paths_to_append_rel(root, rel, live_children); + list_free(live_children); +} + + +/***************************************************************************** + * DEBUG SUPPORT + *****************************************************************************/ + +#ifdef OPTIMIZER_DEBUG + +static void +print_relids(PlannerInfo *root, Relids relids) +{ + int x; + bool first = true; + + x = -1; + while ((x = bms_next_member(relids, x)) >= 0) + { + if (!first) + printf(" "); + if (x < root->simple_rel_array_size && + root->simple_rte_array[x]) + printf("%s", root->simple_rte_array[x]->eref->aliasname); + else + printf("%d", x); + first = false; + } +} + +static void +print_restrictclauses(PlannerInfo *root, List *clauses) +{ + ListCell *l; + + foreach(l, clauses) + { + RestrictInfo *c = lfirst(l); + + print_expr((Node *) c->clause, root->parse->rtable); + if (lnext(clauses, l)) + printf(", "); + } +} + +static void +print_path(PlannerInfo *root, Path *path, int indent) +{ + const char *ptype; + bool join = false; + Path *subpath = NULL; + int i; + + switch (nodeTag(path)) + { + case T_Path: + switch (path->pathtype) + { + case T_SeqScan: + ptype = "SeqScan"; + break; + case T_SampleScan: + ptype = "SampleScan"; + break; + case T_FunctionScan: + ptype = "FunctionScan"; + break; + case T_TableFuncScan: + ptype = "TableFuncScan"; + break; + case T_ValuesScan: + ptype = "ValuesScan"; + break; + case T_CteScan: + ptype = "CteScan"; + break; + case T_NamedTuplestoreScan: + ptype = "NamedTuplestoreScan"; + break; + case T_Result: + ptype = "Result"; + break; + case T_WorkTableScan: + ptype = "WorkTableScan"; + break; + default: + ptype = "???Path"; + break; + } + break; + case T_IndexPath: + ptype = "IdxScan"; + break; + case T_BitmapHeapPath: + ptype = "BitmapHeapScan"; + break; + case T_BitmapAndPath: + ptype = "BitmapAndPath"; + break; + case T_BitmapOrPath: + ptype = "BitmapOrPath"; + break; + case T_TidPath: + ptype = "TidScan"; + break; + case T_TidRangePath: + ptype = "TidRangePath"; + break; + case T_SubqueryScanPath: + ptype = "SubqueryScan"; + break; + case T_ForeignPath: + ptype = "ForeignScan"; + break; + case T_CustomPath: + ptype = "CustomScan"; + break; + case T_NestPath: + ptype = "NestLoop"; + join = true; + break; + case T_MergePath: + ptype = "MergeJoin"; + join = true; + break; + case T_HashPath: + ptype = "HashJoin"; + join = true; + break; + case T_AppendPath: + ptype = "Append"; + break; + case T_MergeAppendPath: + ptype = "MergeAppend"; + break; + case T_GroupResultPath: + ptype = "GroupResult"; + break; + case T_MaterialPath: + ptype = "Material"; + subpath = ((MaterialPath *) path)->subpath; + break; + case T_MemoizePath: + ptype = "Memoize"; + subpath = ((MemoizePath *) path)->subpath; + break; + case T_UniquePath: + ptype = "Unique"; + subpath = ((UniquePath *) path)->subpath; + break; + case T_GatherPath: + ptype = "Gather"; + subpath = ((GatherPath *) path)->subpath; + break; + case T_GatherMergePath: + ptype = "GatherMerge"; + subpath = ((GatherMergePath *) path)->subpath; + break; + case T_ProjectionPath: + ptype = "Projection"; + subpath = ((ProjectionPath *) path)->subpath; + break; + case T_ProjectSetPath: + ptype = "ProjectSet"; + subpath = ((ProjectSetPath *) path)->subpath; + break; + case T_SortPath: + ptype = "Sort"; + subpath = ((SortPath *) path)->subpath; + break; + case T_IncrementalSortPath: + ptype = "IncrementalSort"; + subpath = ((SortPath *) path)->subpath; + break; + case T_GroupPath: + ptype = "Group"; + subpath = ((GroupPath *) path)->subpath; + break; + case T_UpperUniquePath: + ptype = "UpperUnique"; + subpath = ((UpperUniquePath *) path)->subpath; + break; + case T_AggPath: + ptype = "Agg"; + subpath = ((AggPath *) path)->subpath; + break; + case T_GroupingSetsPath: + ptype = "GroupingSets"; + subpath = ((GroupingSetsPath *) path)->subpath; + break; + case T_MinMaxAggPath: + ptype = "MinMaxAgg"; + break; + case T_WindowAggPath: + ptype = "WindowAgg"; + subpath = ((WindowAggPath *) path)->subpath; + break; + case T_SetOpPath: + ptype = "SetOp"; + subpath = ((SetOpPath *) path)->subpath; + break; + case T_RecursiveUnionPath: + ptype = "RecursiveUnion"; + break; + case T_LockRowsPath: + ptype = "LockRows"; + subpath = ((LockRowsPath *) path)->subpath; + break; + case T_ModifyTablePath: + ptype = "ModifyTable"; + break; + case T_LimitPath: + ptype = "Limit"; + subpath = ((LimitPath *) path)->subpath; + break; + default: + ptype = "???Path"; + break; + } + + for (i = 0; i < indent; i++) + printf("\t"); + printf("%s", ptype); + + if (path->parent) + { + printf("("); + print_relids(root, path->parent->relids); + printf(")"); + } + if (path->param_info) + { + printf(" required_outer ("); + print_relids(root, path->param_info->ppi_req_outer); + printf(")"); + } + printf(" rows=%.0f cost=%.2f..%.2f\n", + path->rows, path->startup_cost, path->total_cost); + + if (path->pathkeys) + { + for (i = 0; i < indent; i++) + printf("\t"); + printf(" pathkeys: "); + print_pathkeys(path->pathkeys, root->parse->rtable); + } + + if (join) + { + JoinPath *jp = (JoinPath *) path; + + for (i = 0; i < indent; i++) + printf("\t"); + printf(" clauses: "); + print_restrictclauses(root, jp->joinrestrictinfo); + printf("\n"); + + if (IsA(path, MergePath)) + { + MergePath *mp = (MergePath *) path; + + for (i = 0; i < indent; i++) + printf("\t"); + printf(" sortouter=%d sortinner=%d materializeinner=%d\n", + ((mp->outersortkeys) ? 1 : 0), + ((mp->innersortkeys) ? 1 : 0), + ((mp->materialize_inner) ? 1 : 0)); + } + + print_path(root, jp->outerjoinpath, indent + 1); + print_path(root, jp->innerjoinpath, indent + 1); + } + + if (subpath) + print_path(root, subpath, indent + 1); +} + +void +debug_print_rel(PlannerInfo *root, RelOptInfo *rel) +{ + ListCell *l; + + printf("RELOPTINFO ("); + print_relids(root, rel->relids); + printf("): rows=%.0f width=%d\n", rel->rows, rel->reltarget->width); + + if (rel->baserestrictinfo) + { + printf("\tbaserestrictinfo: "); + print_restrictclauses(root, rel->baserestrictinfo); + printf("\n"); + } + + if (rel->joininfo) + { + printf("\tjoininfo: "); + print_restrictclauses(root, rel->joininfo); + printf("\n"); + } + + printf("\tpath list:\n"); + foreach(l, rel->pathlist) + print_path(root, lfirst(l), 1); + if (rel->cheapest_parameterized_paths) + { + printf("\n\tcheapest parameterized paths:\n"); + foreach(l, rel->cheapest_parameterized_paths) + print_path(root, lfirst(l), 1); + } + if (rel->cheapest_startup_path) + { + printf("\n\tcheapest startup path:\n"); + print_path(root, rel->cheapest_startup_path, 1); + } + if (rel->cheapest_total_path) + { + printf("\n\tcheapest total path:\n"); + print_path(root, rel->cheapest_total_path, 1); + } + printf("\n"); + fflush(stdout); +} + +#endif /* OPTIMIZER_DEBUG */ diff --git a/src/backend/optimizer/path/clausesel.c b/src/backend/optimizer/path/clausesel.c new file mode 100644 index 0000000..06f8363 --- /dev/null +++ b/src/backend/optimizer/path/clausesel.c @@ -0,0 +1,1000 @@ +/*------------------------------------------------------------------------- + * + * clausesel.c + * Routines to compute clause selectivities + * + * Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group + * Portions Copyright (c) 1994, Regents of the University of California + * + * + * IDENTIFICATION + * src/backend/optimizer/path/clausesel.c + * + *------------------------------------------------------------------------- + */ +#include "postgres.h" + +#include "nodes/makefuncs.h" +#include "nodes/nodeFuncs.h" +#include "optimizer/clauses.h" +#include "optimizer/cost.h" +#include "optimizer/optimizer.h" +#include "optimizer/pathnode.h" +#include "optimizer/plancat.h" +#include "statistics/statistics.h" +#include "utils/fmgroids.h" +#include "utils/lsyscache.h" +#include "utils/selfuncs.h" + +/* + * Data structure for accumulating info about possible range-query + * clause pairs in clauselist_selectivity. + */ +typedef struct RangeQueryClause +{ + struct RangeQueryClause *next; /* next in linked list */ + Node *var; /* The common variable of the clauses */ + bool have_lobound; /* found a low-bound clause yet? */ + bool have_hibound; /* found a high-bound clause yet? */ + Selectivity lobound; /* Selectivity of a var > something clause */ + Selectivity hibound; /* Selectivity of a var < something clause */ +} RangeQueryClause; + +static void addRangeClause(RangeQueryClause **rqlist, Node *clause, + bool varonleft, bool isLTsel, Selectivity s2); +static RelOptInfo *find_single_rel_for_clauses(PlannerInfo *root, + List *clauses); +static Selectivity clauselist_selectivity_or(PlannerInfo *root, + List *clauses, + int varRelid, + JoinType jointype, + SpecialJoinInfo *sjinfo, + bool use_extended_stats); + +/**************************************************************************** + * ROUTINES TO COMPUTE SELECTIVITIES + ****************************************************************************/ + +/* + * clauselist_selectivity - + * Compute the selectivity of an implicitly-ANDed list of boolean + * expression clauses. The list can be empty, in which case 1.0 + * must be returned. List elements may be either RestrictInfos + * or bare expression clauses --- the former is preferred since + * it allows caching of results. + * + * See clause_selectivity() for the meaning of the additional parameters. + * + * The basic approach is to apply extended statistics first, on as many + * clauses as possible, in order to capture cross-column dependencies etc. + * The remaining clauses are then estimated by taking the product of their + * selectivities, but that's only right if they have independent + * probabilities, and in reality they are often NOT independent even if they + * only refer to a single column. So, we want to be smarter where we can. + * + * We also recognize "range queries", such as "x > 34 AND x < 42". Clauses + * are recognized as possible range query components if they are restriction + * opclauses whose operators have scalarltsel or a related function as their + * restriction selectivity estimator. We pair up clauses of this form that + * refer to the same variable. An unpairable clause of this kind is simply + * multiplied into the selectivity product in the normal way. But when we + * find a pair, we know that the selectivities represent the relative + * positions of the low and high bounds within the column's range, so instead + * of figuring the selectivity as hisel * losel, we can figure it as hisel + + * losel - 1. (To visualize this, see that hisel is the fraction of the range + * below the high bound, while losel is the fraction above the low bound; so + * hisel can be interpreted directly as a 0..1 value but we need to convert + * losel to 1-losel before interpreting it as a value. Then the available + * range is 1-losel to hisel. However, this calculation double-excludes + * nulls, so really we need hisel + losel + null_frac - 1.) + * + * If either selectivity is exactly DEFAULT_INEQ_SEL, we forget this equation + * and instead use DEFAULT_RANGE_INEQ_SEL. The same applies if the equation + * yields an impossible (negative) result. + * + * A free side-effect is that we can recognize redundant inequalities such + * as "x < 4 AND x < 5"; only the tighter constraint will be counted. + * + * Of course this is all very dependent on the behavior of the inequality + * selectivity functions; perhaps some day we can generalize the approach. + */ +Selectivity +clauselist_selectivity(PlannerInfo *root, + List *clauses, + int varRelid, + JoinType jointype, + SpecialJoinInfo *sjinfo) +{ + return clauselist_selectivity_ext(root, clauses, varRelid, + jointype, sjinfo, true); +} + +/* + * clauselist_selectivity_ext - + * Extended version of clauselist_selectivity(). If "use_extended_stats" + * is false, all extended statistics will be ignored, and only per-column + * statistics will be used. + */ +Selectivity +clauselist_selectivity_ext(PlannerInfo *root, + List *clauses, + int varRelid, + JoinType jointype, + SpecialJoinInfo *sjinfo, + bool use_extended_stats) +{ + Selectivity s1 = 1.0; + RelOptInfo *rel; + Bitmapset *estimatedclauses = NULL; + RangeQueryClause *rqlist = NULL; + ListCell *l; + int listidx; + + /* + * If there's exactly one clause, just go directly to + * clause_selectivity_ext(). None of what we might do below is relevant. + */ + if (list_length(clauses) == 1) + return clause_selectivity_ext(root, (Node *) linitial(clauses), + varRelid, jointype, sjinfo, + use_extended_stats); + + /* + * Determine if these clauses reference a single relation. If so, and if + * it has extended statistics, try to apply those. + */ + rel = find_single_rel_for_clauses(root, clauses); + if (use_extended_stats && rel && rel->rtekind == RTE_RELATION && rel->statlist != NIL) + { + /* + * Estimate as many clauses as possible using extended statistics. + * + * 'estimatedclauses' is populated with the 0-based list position + * index of clauses estimated here, and that should be ignored below. + */ + s1 = statext_clauselist_selectivity(root, clauses, varRelid, + jointype, sjinfo, rel, + &estimatedclauses, false); + } + + /* + * Apply normal selectivity estimates for remaining clauses. We'll be + * careful to skip any clauses which were already estimated above. + * + * Anything that doesn't look like a potential rangequery clause gets + * multiplied into s1 and forgotten. Anything that does gets inserted into + * an rqlist entry. + */ + listidx = -1; + foreach(l, clauses) + { + Node *clause = (Node *) lfirst(l); + RestrictInfo *rinfo; + Selectivity s2; + + listidx++; + + /* + * Skip this clause if it's already been estimated by some other + * statistics above. + */ + if (bms_is_member(listidx, estimatedclauses)) + continue; + + /* Compute the selectivity of this clause in isolation */ + s2 = clause_selectivity_ext(root, clause, varRelid, jointype, sjinfo, + use_extended_stats); + + /* + * Check for being passed a RestrictInfo. + * + * If it's a pseudoconstant RestrictInfo, then s2 is either 1.0 or + * 0.0; just use that rather than looking for range pairs. + */ + if (IsA(clause, RestrictInfo)) + { + rinfo = (RestrictInfo *) clause; + if (rinfo->pseudoconstant) + { + s1 = s1 * s2; + continue; + } + clause = (Node *) rinfo->clause; + } + else + rinfo = NULL; + + /* + * See if it looks like a restriction clause with a pseudoconstant on + * one side. (Anything more complicated than that might not behave in + * the simple way we are expecting.) Most of the tests here can be + * done more efficiently with rinfo than without. + */ + if (is_opclause(clause) && list_length(((OpExpr *) clause)->args) == 2) + { + OpExpr *expr = (OpExpr *) clause; + bool varonleft = true; + bool ok; + + if (rinfo) + { + ok = (bms_membership(rinfo->clause_relids) == BMS_SINGLETON) && + (is_pseudo_constant_clause_relids(lsecond(expr->args), + rinfo->right_relids) || + (varonleft = false, + is_pseudo_constant_clause_relids(linitial(expr->args), + rinfo->left_relids))); + } + else + { + ok = (NumRelids(root, clause) == 1) && + (is_pseudo_constant_clause(lsecond(expr->args)) || + (varonleft = false, + is_pseudo_constant_clause(linitial(expr->args)))); + } + + if (ok) + { + /* + * If it's not a "<"/"<="/">"/">=" operator, just merge the + * selectivity in generically. But if it's the right oprrest, + * add the clause to rqlist for later processing. + */ + switch (get_oprrest(expr->opno)) + { + case F_SCALARLTSEL: + case F_SCALARLESEL: + addRangeClause(&rqlist, clause, + varonleft, true, s2); + break; + case F_SCALARGTSEL: + case F_SCALARGESEL: + addRangeClause(&rqlist, clause, + varonleft, false, s2); + break; + default: + /* Just merge the selectivity in generically */ + s1 = s1 * s2; + break; + } + continue; /* drop to loop bottom */ + } + } + + /* Not the right form, so treat it generically. */ + s1 = s1 * s2; + } + + /* + * Now scan the rangequery pair list. + */ + while (rqlist != NULL) + { + RangeQueryClause *rqnext; + + if (rqlist->have_lobound && rqlist->have_hibound) + { + /* Successfully matched a pair of range clauses */ + Selectivity s2; + + /* + * Exact equality to the default value probably means the + * selectivity function punted. This is not airtight but should + * be good enough. + */ + if (rqlist->hibound == DEFAULT_INEQ_SEL || + rqlist->lobound == DEFAULT_INEQ_SEL) + { + s2 = DEFAULT_RANGE_INEQ_SEL; + } + else + { + s2 = rqlist->hibound + rqlist->lobound - 1.0; + + /* Adjust for double-exclusion of NULLs */ + s2 += nulltestsel(root, IS_NULL, rqlist->var, + varRelid, jointype, sjinfo); + + /* + * A zero or slightly negative s2 should be converted into a + * small positive value; we probably are dealing with a very + * tight range and got a bogus result due to roundoff errors. + * However, if s2 is very negative, then we probably have + * default selectivity estimates on one or both sides of the + * range that we failed to recognize above for some reason. + */ + if (s2 <= 0.0) + { + if (s2 < -0.01) + { + /* + * No data available --- use a default estimate that + * is small, but not real small. + */ + s2 = DEFAULT_RANGE_INEQ_SEL; + } + else + { + /* + * It's just roundoff error; use a small positive + * value + */ + s2 = 1.0e-10; + } + } + } + /* Merge in the selectivity of the pair of clauses */ + s1 *= s2; + } + else + { + /* Only found one of a pair, merge it in generically */ + if (rqlist->have_lobound) + s1 *= rqlist->lobound; + else + s1 *= rqlist->hibound; + } + /* release storage and advance */ + rqnext = rqlist->next; + pfree(rqlist); + rqlist = rqnext; + } + + return s1; +} + +/* + * clauselist_selectivity_or - + * Compute the selectivity of an implicitly-ORed list of boolean + * expression clauses. The list can be empty, in which case 0.0 + * must be returned. List elements may be either RestrictInfos + * or bare expression clauses --- the former is preferred since + * it allows caching of results. + * + * See clause_selectivity() for the meaning of the additional parameters. + * + * The basic approach is to apply extended statistics first, on as many + * clauses as possible, in order to capture cross-column dependencies etc. + * The remaining clauses are then estimated as if they were independent. + */ +static Selectivity +clauselist_selectivity_or(PlannerInfo *root, + List *clauses, + int varRelid, + JoinType jointype, + SpecialJoinInfo *sjinfo, + bool use_extended_stats) +{ + Selectivity s1 = 0.0; + RelOptInfo *rel; + Bitmapset *estimatedclauses = NULL; + ListCell *lc; + int listidx; + + /* + * Determine if these clauses reference a single relation. If so, and if + * it has extended statistics, try to apply those. + */ + rel = find_single_rel_for_clauses(root, clauses); + if (use_extended_stats && rel && rel->rtekind == RTE_RELATION && rel->statlist != NIL) + { + /* + * Estimate as many clauses as possible using extended statistics. + * + * 'estimatedclauses' is populated with the 0-based list position + * index of clauses estimated here, and that should be ignored below. + */ + s1 = statext_clauselist_selectivity(root, clauses, varRelid, + jointype, sjinfo, rel, + &estimatedclauses, true); + } + + /* + * Estimate the remaining clauses as if they were independent. + * + * Selectivities for an OR clause are computed as s1+s2 - s1*s2 to account + * for the probable overlap of selected tuple sets. + * + * XXX is this too conservative? + */ + listidx = -1; + foreach(lc, clauses) + { + Selectivity s2; + + listidx++; + + /* + * Skip this clause if it's already been estimated by some other + * statistics above. + */ + if (bms_is_member(listidx, estimatedclauses)) + continue; + + s2 = clause_selectivity_ext(root, (Node *) lfirst(lc), varRelid, + jointype, sjinfo, use_extended_stats); + + s1 = s1 + s2 - s1 * s2; + } + + return s1; +} + +/* + * addRangeClause --- add a new range clause for clauselist_selectivity + * + * Here is where we try to match up pairs of range-query clauses + */ +static void +addRangeClause(RangeQueryClause **rqlist, Node *clause, + bool varonleft, bool isLTsel, Selectivity s2) +{ + RangeQueryClause *rqelem; + Node *var; + bool is_lobound; + + if (varonleft) + { + var = get_leftop((Expr *) clause); + is_lobound = !isLTsel; /* x < something is high bound */ + } + else + { + var = get_rightop((Expr *) clause); + is_lobound = isLTsel; /* something < x is low bound */ + } + + for (rqelem = *rqlist; rqelem; rqelem = rqelem->next) + { + /* + * We use full equal() here because the "var" might be a function of + * one or more attributes of the same relation... + */ + if (!equal(var, rqelem->var)) + continue; + /* Found the right group to put this clause in */ + if (is_lobound) + { + if (!rqelem->have_lobound) + { + rqelem->have_lobound = true; + rqelem->lobound = s2; + } + else + { + + /*------ + * We have found two similar clauses, such as + * x < y AND x <= z. + * Keep only the more restrictive one. + *------ + */ + if (rqelem->lobound > s2) + rqelem->lobound = s2; + } + } + else + { + if (!rqelem->have_hibound) + { + rqelem->have_hibound = true; + rqelem->hibound = s2; + } + else + { + + /*------ + * We have found two similar clauses, such as + * x > y AND x >= z. + * Keep only the more restrictive one. + *------ + */ + if (rqelem->hibound > s2) + rqelem->hibound = s2; + } + } + return; + } + + /* No matching var found, so make a new clause-pair data structure */ + rqelem = (RangeQueryClause *) palloc(sizeof(RangeQueryClause)); + rqelem->var = var; + if (is_lobound) + { + rqelem->have_lobound = true; + rqelem->have_hibound = false; + rqelem->lobound = s2; + } + else + { + rqelem->have_lobound = false; + rqelem->have_hibound = true; + rqelem->hibound = s2; + } + rqelem->next = *rqlist; + *rqlist = rqelem; +} + +/* + * find_single_rel_for_clauses + * Examine each clause in 'clauses' and determine if all clauses + * reference only a single relation. If so return that relation, + * otherwise return NULL. + */ +static RelOptInfo * +find_single_rel_for_clauses(PlannerInfo *root, List *clauses) +{ + int lastrelid = 0; + ListCell *l; + + foreach(l, clauses) + { + RestrictInfo *rinfo = (RestrictInfo *) lfirst(l); + int relid; + + /* + * If we have a list of bare clauses rather than RestrictInfos, we + * could pull out their relids the hard way with pull_varnos(). + * However, currently the extended-stats machinery won't do anything + * with non-RestrictInfo clauses anyway, so there's no point in + * spending extra cycles; just fail if that's what we have. + * + * An exception to that rule is if we have a bare BoolExpr AND clause. + * We treat this as a special case because the restrictinfo machinery + * doesn't build RestrictInfos on top of AND clauses. + */ + if (is_andclause(rinfo)) + { + RelOptInfo *rel; + + rel = find_single_rel_for_clauses(root, + ((BoolExpr *) rinfo)->args); + + if (rel == NULL) + return NULL; + if (lastrelid == 0) + lastrelid = rel->relid; + else if (rel->relid != lastrelid) + return NULL; + + continue; + } + + if (!IsA(rinfo, RestrictInfo)) + return NULL; + + if (bms_is_empty(rinfo->clause_relids)) + continue; /* we can ignore variable-free clauses */ + if (!bms_get_singleton_member(rinfo->clause_relids, &relid)) + return NULL; /* multiple relations in this clause */ + if (lastrelid == 0) + lastrelid = relid; /* first clause referencing a relation */ + else if (relid != lastrelid) + return NULL; /* relation not same as last one */ + } + + if (lastrelid != 0) + return find_base_rel(root, lastrelid); + + return NULL; /* no clauses */ +} + +/* + * bms_is_subset_singleton + * + * Same result as bms_is_subset(s, bms_make_singleton(x)), + * but a little faster and doesn't leak memory. + * + * Is this of use anywhere else? If so move to bitmapset.c ... + */ +static bool +bms_is_subset_singleton(const Bitmapset *s, int x) +{ + switch (bms_membership(s)) + { + case BMS_EMPTY_SET: + return true; + case BMS_SINGLETON: + return bms_is_member(x, s); + case BMS_MULTIPLE: + return false; + } + /* can't get here... */ + return false; +} + +/* + * treat_as_join_clause - + * Decide whether an operator clause is to be handled by the + * restriction or join estimator. Subroutine for clause_selectivity(). + */ +static inline bool +treat_as_join_clause(PlannerInfo *root, Node *clause, RestrictInfo *rinfo, + int varRelid, SpecialJoinInfo *sjinfo) +{ + if (varRelid != 0) + { + /* + * Caller is forcing restriction mode (eg, because we are examining an + * inner indexscan qual). + */ + return false; + } + else if (sjinfo == NULL) + { + /* + * It must be a restriction clause, since it's being evaluated at a + * scan node. + */ + return false; + } + else + { + /* + * Otherwise, it's a join if there's more than one relation used. We + * can optimize this calculation if an rinfo was passed. + * + * XXX Since we know the clause is being evaluated at a join, the + * only way it could be single-relation is if it was delayed by outer + * joins. Although we can make use of the restriction qual estimators + * anyway, it seems likely that we ought to account for the + * probability of injected nulls somehow. + */ + if (rinfo) + return (bms_membership(rinfo->clause_relids) == BMS_MULTIPLE); + else + return (NumRelids(root, clause) > 1); + } +} + + +/* + * clause_selectivity - + * Compute the selectivity of a general boolean expression clause. + * + * The clause can be either a RestrictInfo or a plain expression. If it's + * a RestrictInfo, we try to cache the selectivity for possible re-use, + * so passing RestrictInfos is preferred. + * + * varRelid is either 0 or a rangetable index. + * + * When varRelid is not 0, only variables belonging to that relation are + * considered in computing selectivity; other vars are treated as constants + * of unknown values. This is appropriate for estimating the selectivity of + * a join clause that is being used as a restriction clause in a scan of a + * nestloop join's inner relation --- varRelid should then be the ID of the + * inner relation. + * + * When varRelid is 0, all variables are treated as variables. This + * is appropriate for ordinary join clauses and restriction clauses. + * + * jointype is the join type, if the clause is a join clause. Pass JOIN_INNER + * if the clause isn't a join clause. + * + * sjinfo is NULL for a non-join clause, otherwise it provides additional + * context information about the join being performed. There are some + * special cases: + * 1. For a special (not INNER) join, sjinfo is always a member of + * root->join_info_list. + * 2. For an INNER join, sjinfo is just a transient struct, and only the + * relids and jointype fields in it can be trusted. + * It is possible for jointype to be different from sjinfo->jointype. + * This indicates we are considering a variant join: either with + * the LHS and RHS switched, or with one input unique-ified. + * + * Note: when passing nonzero varRelid, it's normally appropriate to set + * jointype == JOIN_INNER, sjinfo == NULL, even if the clause is really a + * join clause; because we aren't treating it as a join clause. + */ +Selectivity +clause_selectivity(PlannerInfo *root, + Node *clause, + int varRelid, + JoinType jointype, + SpecialJoinInfo *sjinfo) +{ + return clause_selectivity_ext(root, clause, varRelid, + jointype, sjinfo, true); +} + +/* + * clause_selectivity_ext - + * Extended version of clause_selectivity(). If "use_extended_stats" is + * false, all extended statistics will be ignored, and only per-column + * statistics will be used. + */ +Selectivity +clause_selectivity_ext(PlannerInfo *root, + Node *clause, + int varRelid, + JoinType jointype, + SpecialJoinInfo *sjinfo, + bool use_extended_stats) +{ + Selectivity s1 = 0.5; /* default for any unhandled clause type */ + RestrictInfo *rinfo = NULL; + bool cacheable = false; + + if (clause == NULL) /* can this still happen? */ + return s1; + + if (IsA(clause, RestrictInfo)) + { + rinfo = (RestrictInfo *) clause; + + /* + * If the clause is marked pseudoconstant, then it will be used as a + * gating qual and should not affect selectivity estimates; hence + * return 1.0. The only exception is that a constant FALSE may be + * taken as having selectivity 0.0, since it will surely mean no rows + * out of the plan. This case is simple enough that we need not + * bother caching the result. + */ + if (rinfo->pseudoconstant) + { + if (!IsA(rinfo->clause, Const)) + return (Selectivity) 1.0; + } + + /* + * If the clause is marked redundant, always return 1.0. + */ + if (rinfo->norm_selec > 1) + return (Selectivity) 1.0; + + /* + * If possible, cache the result of the selectivity calculation for + * the clause. We can cache if varRelid is zero or the clause + * contains only vars of that relid --- otherwise varRelid will affect + * the result, so mustn't cache. Outer join quals might be examined + * with either their join's actual jointype or JOIN_INNER, so we need + * two cache variables to remember both cases. Note: we assume the + * result won't change if we are switching the input relations or + * considering a unique-ified case, so we only need one cache variable + * for all non-JOIN_INNER cases. + */ + if (varRelid == 0 || + bms_is_subset_singleton(rinfo->clause_relids, varRelid)) + { + /* Cacheable --- do we already have the result? */ + if (jointype == JOIN_INNER) + { + if (rinfo->norm_selec >= 0) + return rinfo->norm_selec; + } + else + { + if (rinfo->outer_selec >= 0) + return rinfo->outer_selec; + } + cacheable = true; + } + + /* + * Proceed with examination of contained clause. If the clause is an + * OR-clause, we want to look at the variant with sub-RestrictInfos, + * so that per-subclause selectivities can be cached. + */ + if (rinfo->orclause) + clause = (Node *) rinfo->orclause; + else + clause = (Node *) rinfo->clause; + } + + if (IsA(clause, Var)) + { + Var *var = (Var *) clause; + + /* + * We probably shouldn't ever see an uplevel Var here, but if we do, + * return the default selectivity... + */ + if (var->varlevelsup == 0 && + (varRelid == 0 || varRelid == (int) var->varno)) + { + /* Use the restriction selectivity function for a bool Var */ + s1 = boolvarsel(root, (Node *) var, varRelid); + } + } + else if (IsA(clause, Const)) + { + /* bool constant is pretty easy... */ + Const *con = (Const *) clause; + + s1 = con->constisnull ? 0.0 : + DatumGetBool(con->constvalue) ? 1.0 : 0.0; + } + else if (IsA(clause, Param)) + { + /* see if we can replace the Param */ + Node *subst = estimate_expression_value(root, clause); + + if (IsA(subst, Const)) + { + /* bool constant is pretty easy... */ + Const *con = (Const *) subst; + + s1 = con->constisnull ? 0.0 : + DatumGetBool(con->constvalue) ? 1.0 : 0.0; + } + else + { + /* XXX any way to do better than default? */ + } + } + else if (is_notclause(clause)) + { + /* inverse of the selectivity of the underlying clause */ + s1 = 1.0 - clause_selectivity_ext(root, + (Node *) get_notclausearg((Expr *) clause), + varRelid, + jointype, + sjinfo, + use_extended_stats); + } + else if (is_andclause(clause)) + { + /* share code with clauselist_selectivity() */ + s1 = clauselist_selectivity_ext(root, + ((BoolExpr *) clause)->args, + varRelid, + jointype, + sjinfo, + use_extended_stats); + } + else if (is_orclause(clause)) + { + /* + * Almost the same thing as clauselist_selectivity, but with the + * clauses connected by OR. + */ + s1 = clauselist_selectivity_or(root, + ((BoolExpr *) clause)->args, + varRelid, + jointype, + sjinfo, + use_extended_stats); + } + else if (is_opclause(clause) || IsA(clause, DistinctExpr)) + { + OpExpr *opclause = (OpExpr *) clause; + Oid opno = opclause->opno; + + if (treat_as_join_clause(root, clause, rinfo, varRelid, sjinfo)) + { + /* Estimate selectivity for a join clause. */ + s1 = join_selectivity(root, opno, + opclause->args, + opclause->inputcollid, + jointype, + sjinfo); + } + else + { + /* Estimate selectivity for a restriction clause. */ + s1 = restriction_selectivity(root, opno, + opclause->args, + opclause->inputcollid, + varRelid); + } + + /* + * DistinctExpr has the same representation as OpExpr, but the + * contained operator is "=" not "<>", so we must negate the result. + * This estimation method doesn't give the right behavior for nulls, + * but it's better than doing nothing. + */ + if (IsA(clause, DistinctExpr)) + s1 = 1.0 - s1; + } + else if (is_funcclause(clause)) + { + FuncExpr *funcclause = (FuncExpr *) clause; + + /* Try to get an estimate from the support function, if any */ + s1 = function_selectivity(root, + funcclause->funcid, + funcclause->args, + funcclause->inputcollid, + treat_as_join_clause(root, clause, rinfo, + varRelid, sjinfo), + varRelid, + jointype, + sjinfo); + } + else if (IsA(clause, ScalarArrayOpExpr)) + { + /* Use node specific selectivity calculation function */ + s1 = scalararraysel(root, + (ScalarArrayOpExpr *) clause, + treat_as_join_clause(root, clause, rinfo, + varRelid, sjinfo), + varRelid, + jointype, + sjinfo); + } + else if (IsA(clause, RowCompareExpr)) + { + /* Use node specific selectivity calculation function */ + s1 = rowcomparesel(root, + (RowCompareExpr *) clause, + varRelid, + jointype, + sjinfo); + } + else if (IsA(clause, NullTest)) + { + /* Use node specific selectivity calculation function */ + s1 = nulltestsel(root, + ((NullTest *) clause)->nulltesttype, + (Node *) ((NullTest *) clause)->arg, + varRelid, + jointype, + sjinfo); + } + else if (IsA(clause, BooleanTest)) + { + /* Use node specific selectivity calculation function */ + s1 = booltestsel(root, + ((BooleanTest *) clause)->booltesttype, + (Node *) ((BooleanTest *) clause)->arg, + varRelid, + jointype, + sjinfo); + } + else if (IsA(clause, CurrentOfExpr)) + { + /* CURRENT OF selects at most one row of its table */ + CurrentOfExpr *cexpr = (CurrentOfExpr *) clause; + RelOptInfo *crel = find_base_rel(root, cexpr->cvarno); + + if (crel->tuples > 0) + s1 = 1.0 / crel->tuples; + } + else if (IsA(clause, RelabelType)) + { + /* Not sure this case is needed, but it can't hurt */ + s1 = clause_selectivity_ext(root, + (Node *) ((RelabelType *) clause)->arg, + varRelid, + jointype, + sjinfo, + use_extended_stats); + } + else if (IsA(clause, CoerceToDomain)) + { + /* Not sure this case is needed, but it can't hurt */ + s1 = clause_selectivity_ext(root, + (Node *) ((CoerceToDomain *) clause)->arg, + varRelid, + jointype, + sjinfo, + use_extended_stats); + } + else + { + /* + * For anything else, see if we can consider it as a boolean variable. + * This only works if it's an immutable expression in Vars of a single + * relation; but there's no point in us checking that here because + * boolvarsel() will do it internally, and return a suitable default + * selectivity if not. + */ + s1 = boolvarsel(root, clause, varRelid); + } + + /* Cache the result if possible */ + if (cacheable) + { + if (jointype == JOIN_INNER) + rinfo->norm_selec = s1; + else + rinfo->outer_selec = s1; + } + +#ifdef SELECTIVITY_DEBUG + elog(DEBUG4, "clause_selectivity: s1 %f", s1); +#endif /* SELECTIVITY_DEBUG */ + + return s1; +} diff --git a/src/backend/optimizer/path/costsize.c b/src/backend/optimizer/path/costsize.c new file mode 100644 index 0000000..0ba26b2 --- /dev/null +++ b/src/backend/optimizer/path/costsize.c @@ -0,0 +1,6221 @@ +/*------------------------------------------------------------------------- + * + * costsize.c + * Routines to compute (and set) relation sizes and path costs + * + * Path costs are measured in arbitrary units established by these basic + * parameters: + * + * seq_page_cost Cost of a sequential page fetch + * random_page_cost Cost of a non-sequential page fetch + * cpu_tuple_cost Cost of typical CPU time to process a tuple + * cpu_index_tuple_cost Cost of typical CPU time to process an index tuple + * cpu_operator_cost Cost of CPU time to execute an operator or function + * parallel_tuple_cost Cost of CPU time to pass a tuple from worker to leader backend + * parallel_setup_cost Cost of setting up shared memory for parallelism + * + * We expect that the kernel will typically do some amount of read-ahead + * optimization; this in conjunction with seek costs means that seq_page_cost + * is normally considerably less than random_page_cost. (However, if the + * database is fully cached in RAM, it is reasonable to set them equal.) + * + * We also use a rough estimate "effective_cache_size" of the number of + * disk pages in Postgres + OS-level disk cache. (We can't simply use + * NBuffers for this purpose because that would ignore the effects of + * the kernel's disk cache.) + * + * Obviously, taking constants for these values is an oversimplification, + * but it's tough enough to get any useful estimates even at this level of + * detail. Note that all of these parameters are user-settable, in case + * the default values are drastically off for a particular platform. + * + * seq_page_cost and random_page_cost can also be overridden for an individual + * tablespace, in case some data is on a fast disk and other data is on a slow + * disk. Per-tablespace overrides never apply to temporary work files such as + * an external sort or a materialize node that overflows work_mem. + * + * We compute two separate costs for each path: + * total_cost: total estimated cost to fetch all tuples + * startup_cost: cost that is expended before first tuple is fetched + * In some scenarios, such as when there is a LIMIT or we are implementing + * an EXISTS(...) sub-select, it is not necessary to fetch all tuples of the + * path's result. A caller can estimate the cost of fetching a partial + * result by interpolating between startup_cost and total_cost. In detail: + * actual_cost = startup_cost + + * (total_cost - startup_cost) * tuples_to_fetch / path->rows; + * Note that a base relation's rows count (and, by extension, plan_rows for + * plan nodes below the LIMIT node) are set without regard to any LIMIT, so + * that this equation works properly. (Note: while path->rows is never zero + * for ordinary relations, it is zero for paths for provably-empty relations, + * so beware of division-by-zero.) The LIMIT is applied as a top-level + * plan node. + * + * For largely historical reasons, most of the routines in this module use + * the passed result Path only to store their results (rows, startup_cost and + * total_cost) into. All the input data they need is passed as separate + * parameters, even though much of it could be extracted from the Path. + * An exception is made for the cost_XXXjoin() routines, which expect all + * the other fields of the passed XXXPath to be filled in, and similarly + * cost_index() assumes the passed IndexPath is valid except for its output + * values. + * + * + * Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group + * Portions Copyright (c) 1994, Regents of the University of California + * + * IDENTIFICATION + * src/backend/optimizer/path/costsize.c + * + *------------------------------------------------------------------------- + */ + +#include "postgres.h" + +#include <limits.h> +#include <math.h> + +#include "access/amapi.h" +#include "access/htup_details.h" +#include "access/tsmapi.h" +#include "executor/executor.h" +#include "executor/nodeAgg.h" +#include "executor/nodeHash.h" +#include "executor/nodeMemoize.h" +#include "miscadmin.h" +#include "nodes/makefuncs.h" +#include "nodes/nodeFuncs.h" +#include "optimizer/clauses.h" +#include "optimizer/cost.h" +#include "optimizer/optimizer.h" +#include "optimizer/pathnode.h" +#include "optimizer/paths.h" +#include "optimizer/placeholder.h" +#include "optimizer/plancat.h" +#include "optimizer/planmain.h" +#include "optimizer/restrictinfo.h" +#include "parser/parsetree.h" +#include "utils/lsyscache.h" +#include "utils/selfuncs.h" +#include "utils/spccache.h" +#include "utils/tuplesort.h" + + +#define LOG2(x) (log(x) / 0.693147180559945) + +/* + * Append and MergeAppend nodes are less expensive than some other operations + * which use cpu_tuple_cost; instead of adding a separate GUC, estimate the + * per-tuple cost as cpu_tuple_cost multiplied by this value. + */ +#define APPEND_CPU_COST_MULTIPLIER 0.5 + +/* + * Maximum value for row estimates. We cap row estimates to this to help + * ensure that costs based on these estimates remain within the range of what + * double can represent. add_path() wouldn't act sanely given infinite or NaN + * cost values. + */ +#define MAXIMUM_ROWCOUNT 1e100 + +double seq_page_cost = DEFAULT_SEQ_PAGE_COST; +double random_page_cost = DEFAULT_RANDOM_PAGE_COST; +double cpu_tuple_cost = DEFAULT_CPU_TUPLE_COST; +double cpu_index_tuple_cost = DEFAULT_CPU_INDEX_TUPLE_COST; +double cpu_operator_cost = DEFAULT_CPU_OPERATOR_COST; +double parallel_tuple_cost = DEFAULT_PARALLEL_TUPLE_COST; +double parallel_setup_cost = DEFAULT_PARALLEL_SETUP_COST; +double recursive_worktable_factor = DEFAULT_RECURSIVE_WORKTABLE_FACTOR; + +int effective_cache_size = DEFAULT_EFFECTIVE_CACHE_SIZE; + +Cost disable_cost = 1.0e10; + +int max_parallel_workers_per_gather = 2; + +bool enable_seqscan = true; +bool enable_indexscan = true; +bool enable_indexonlyscan = true; +bool enable_bitmapscan = true; +bool enable_tidscan = true; +bool enable_sort = true; +bool enable_incremental_sort = true; +bool enable_hashagg = true; +bool enable_nestloop = true; +bool enable_material = true; +bool enable_memoize = true; +bool enable_mergejoin = true; +bool enable_hashjoin = true; +bool enable_gathermerge = true; +bool enable_partitionwise_join = false; +bool enable_partitionwise_aggregate = false; +bool enable_parallel_append = true; +bool enable_parallel_hash = true; +bool enable_partition_pruning = true; +bool enable_async_append = true; + +typedef struct +{ + PlannerInfo *root; + QualCost total; +} cost_qual_eval_context; + +static List *extract_nonindex_conditions(List *qual_clauses, List *indexclauses); +static MergeScanSelCache *cached_scansel(PlannerInfo *root, + RestrictInfo *rinfo, + PathKey *pathkey); +static void cost_rescan(PlannerInfo *root, Path *path, + Cost *rescan_startup_cost, Cost *rescan_total_cost); +static bool cost_qual_eval_walker(Node *node, cost_qual_eval_context *context); +static void get_restriction_qual_cost(PlannerInfo *root, RelOptInfo *baserel, + ParamPathInfo *param_info, + QualCost *qpqual_cost); +static bool has_indexed_join_quals(NestPath *path); +static double approx_tuple_count(PlannerInfo *root, JoinPath *path, + List *quals); +static double calc_joinrel_size_estimate(PlannerInfo *root, + RelOptInfo *joinrel, + RelOptInfo *outer_rel, + RelOptInfo *inner_rel, + double outer_rows, + double inner_rows, + SpecialJoinInfo *sjinfo, + List *restrictlist); +static Selectivity get_foreign_key_join_selectivity(PlannerInfo *root, + Relids outer_relids, + Relids inner_relids, + SpecialJoinInfo *sjinfo, + List **restrictlist); +static Cost append_nonpartial_cost(List *subpaths, int numpaths, + int parallel_workers); +static void set_rel_width(PlannerInfo *root, RelOptInfo *rel); +static double relation_byte_size(double tuples, int width); +static double page_size(double tuples, int width); +static double get_parallel_divisor(Path *path); + + +/* + * clamp_row_est + * Force a row-count estimate to a sane value. + */ +double +clamp_row_est(double nrows) +{ + /* + * Avoid infinite and NaN row estimates. Costs derived from such values + * are going to be useless. Also force the estimate to be at least one + * row, to make explain output look better and to avoid possible + * divide-by-zero when interpolating costs. Make it an integer, too. + */ + if (nrows > MAXIMUM_ROWCOUNT || isnan(nrows)) + nrows = MAXIMUM_ROWCOUNT; + else if (nrows <= 1.0) + nrows = 1.0; + else + nrows = rint(nrows); + + return nrows; +} + +/* + * clamp_cardinality_to_long + * Cast a Cardinality value to a sane long value. + */ +long +clamp_cardinality_to_long(Cardinality x) +{ + /* + * Just for paranoia's sake, ensure we do something sane with negative or + * NaN values. + */ + if (isnan(x)) + return LONG_MAX; + if (x <= 0) + return 0; + + /* + * If "long" is 64 bits, then LONG_MAX cannot be represented exactly as a + * double. Casting it to double and back may well result in overflow due + * to rounding, so avoid doing that. We trust that any double value that + * compares strictly less than "(double) LONG_MAX" will cast to a + * representable "long" value. + */ + return (x < (double) LONG_MAX) ? (long) x : LONG_MAX; +} + + +/* + * cost_seqscan + * Determines and returns the cost of scanning a relation sequentially. + * + * 'baserel' is the relation to be scanned + * 'param_info' is the ParamPathInfo if this is a parameterized path, else NULL + */ +void +cost_seqscan(Path *path, PlannerInfo *root, + RelOptInfo *baserel, ParamPathInfo *param_info) +{ + Cost startup_cost = 0; + Cost cpu_run_cost; + Cost disk_run_cost; + double spc_seq_page_cost; + QualCost qpqual_cost; + Cost cpu_per_tuple; + + /* Should only be applied to base relations */ + Assert(baserel->relid > 0); + Assert(baserel->rtekind == RTE_RELATION); + + /* Mark the path with the correct row estimate */ + if (param_info) + path->rows = param_info->ppi_rows; + else + path->rows = baserel->rows; + + if (!enable_seqscan) + startup_cost += disable_cost; + + /* fetch estimated page cost for tablespace containing table */ + get_tablespace_page_costs(baserel->reltablespace, + NULL, + &spc_seq_page_cost); + + /* + * disk costs + */ + disk_run_cost = spc_seq_page_cost * baserel->pages; + + /* CPU costs */ + get_restriction_qual_cost(root, baserel, param_info, &qpqual_cost); + + startup_cost += qpqual_cost.startup; + cpu_per_tuple = cpu_tuple_cost + qpqual_cost.per_tuple; + cpu_run_cost = cpu_per_tuple * baserel->tuples; + /* tlist eval costs are paid per output row, not per tuple scanned */ + startup_cost += path->pathtarget->cost.startup; + cpu_run_cost += path->pathtarget->cost.per_tuple * path->rows; + + /* Adjust costing for parallelism, if used. */ + if (path->parallel_workers > 0) + { + double parallel_divisor = get_parallel_divisor(path); + + /* The CPU cost is divided among all the workers. */ + cpu_run_cost /= parallel_divisor; + + /* + * It may be possible to amortize some of the I/O cost, but probably + * not very much, because most operating systems already do aggressive + * prefetching. For now, we assume that the disk run cost can't be + * amortized at all. + */ + + /* + * In the case of a parallel plan, the row count needs to represent + * the number of tuples processed per worker. + */ + path->rows = clamp_row_est(path->rows / parallel_divisor); + } + + path->startup_cost = startup_cost; + path->total_cost = startup_cost + cpu_run_cost + disk_run_cost; +} + +/* + * cost_samplescan + * Determines and returns the cost of scanning a relation using sampling. + * + * 'baserel' is the relation to be scanned + * 'param_info' is the ParamPathInfo if this is a parameterized path, else NULL + */ +void +cost_samplescan(Path *path, PlannerInfo *root, + RelOptInfo *baserel, ParamPathInfo *param_info) +{ + Cost startup_cost = 0; + Cost run_cost = 0; + RangeTblEntry *rte; + TableSampleClause *tsc; + TsmRoutine *tsm; + double spc_seq_page_cost, + spc_random_page_cost, + spc_page_cost; + QualCost qpqual_cost; + Cost cpu_per_tuple; + + /* Should only be applied to base relations with tablesample clauses */ + Assert(baserel->relid > 0); + rte = planner_rt_fetch(baserel->relid, root); + Assert(rte->rtekind == RTE_RELATION); + tsc = rte->tablesample; + Assert(tsc != NULL); + tsm = GetTsmRoutine(tsc->tsmhandler); + + /* Mark the path with the correct row estimate */ + if (param_info) + path->rows = param_info->ppi_rows; + else + path->rows = baserel->rows; + + /* fetch estimated page cost for tablespace containing table */ + get_tablespace_page_costs(baserel->reltablespace, + &spc_random_page_cost, + &spc_seq_page_cost); + + /* if NextSampleBlock is used, assume random access, else sequential */ + spc_page_cost = (tsm->NextSampleBlock != NULL) ? + spc_random_page_cost : spc_seq_page_cost; + + /* + * disk costs (recall that baserel->pages has already been set to the + * number of pages the sampling method will visit) + */ + run_cost += spc_page_cost * baserel->pages; + + /* + * CPU costs (recall that baserel->tuples has already been set to the + * number of tuples the sampling method will select). Note that we ignore + * execution cost of the TABLESAMPLE parameter expressions; they will be + * evaluated only once per scan, and in most usages they'll likely be + * simple constants anyway. We also don't charge anything for the + * calculations the sampling method might do internally. + */ + get_restriction_qual_cost(root, baserel, param_info, &qpqual_cost); + + startup_cost += qpqual_cost.startup; + cpu_per_tuple = cpu_tuple_cost + qpqual_cost.per_tuple; + run_cost += cpu_per_tuple * baserel->tuples; + /* tlist eval costs are paid per output row, not per tuple scanned */ + startup_cost += path->pathtarget->cost.startup; + run_cost += path->pathtarget->cost.per_tuple * path->rows; + + path->startup_cost = startup_cost; + path->total_cost = startup_cost + run_cost; +} + +/* + * cost_gather + * Determines and returns the cost of gather path. + * + * 'rel' is the relation to be operated upon + * 'param_info' is the ParamPathInfo if this is a parameterized path, else NULL + * 'rows' may be used to point to a row estimate; if non-NULL, it overrides + * both 'rel' and 'param_info'. This is useful when the path doesn't exactly + * correspond to any particular RelOptInfo. + */ +void +cost_gather(GatherPath *path, PlannerInfo *root, + RelOptInfo *rel, ParamPathInfo *param_info, + double *rows) +{ + Cost startup_cost = 0; + Cost run_cost = 0; + + /* Mark the path with the correct row estimate */ + if (rows) + path->path.rows = *rows; + else if (param_info) + path->path.rows = param_info->ppi_rows; + else + path->path.rows = rel->rows; + + startup_cost = path->subpath->startup_cost; + + run_cost = path->subpath->total_cost - path->subpath->startup_cost; + + /* Parallel setup and communication cost. */ + startup_cost += parallel_setup_cost; + run_cost += parallel_tuple_cost * path->path.rows; + + path->path.startup_cost = startup_cost; + path->path.total_cost = (startup_cost + run_cost); +} + +/* + * cost_gather_merge + * Determines and returns the cost of gather merge path. + * + * GatherMerge merges several pre-sorted input streams, using a heap that at + * any given instant holds the next tuple from each stream. If there are N + * streams, we need about N*log2(N) tuple comparisons to construct the heap at + * startup, and then for each output tuple, about log2(N) comparisons to + * replace the top heap entry with the next tuple from the same stream. + */ +void +cost_gather_merge(GatherMergePath *path, PlannerInfo *root, + RelOptInfo *rel, ParamPathInfo *param_info, + Cost input_startup_cost, Cost input_total_cost, + double *rows) +{ + Cost startup_cost = 0; + Cost run_cost = 0; + Cost comparison_cost; + double N; + double logN; + + /* Mark the path with the correct row estimate */ + if (rows) + path->path.rows = *rows; + else if (param_info) + path->path.rows = param_info->ppi_rows; + else + path->path.rows = rel->rows; + + if (!enable_gathermerge) + startup_cost += disable_cost; + + /* + * Add one to the number of workers to account for the leader. This might + * be overgenerous since the leader will do less work than other workers + * in typical cases, but we'll go with it for now. + */ + Assert(path->num_workers > 0); + N = (double) path->num_workers + 1; + logN = LOG2(N); + + /* Assumed cost per tuple comparison */ + comparison_cost = 2.0 * cpu_operator_cost; + + /* Heap creation cost */ + startup_cost += comparison_cost * N * logN; + + /* Per-tuple heap maintenance cost */ + run_cost += path->path.rows * comparison_cost * logN; + + /* small cost for heap management, like cost_merge_append */ + run_cost += cpu_operator_cost * path->path.rows; + + /* + * Parallel setup and communication cost. Since Gather Merge, unlike + * Gather, requires us to block until a tuple is available from every + * worker, we bump the IPC cost up a little bit as compared with Gather. + * For lack of a better idea, charge an extra 5%. + */ + startup_cost += parallel_setup_cost; + run_cost += parallel_tuple_cost * path->path.rows * 1.05; + + path->path.startup_cost = startup_cost + input_startup_cost; + path->path.total_cost = (startup_cost + run_cost + input_total_cost); +} + +/* + * cost_index + * Determines and returns the cost of scanning a relation using an index. + * + * 'path' describes the indexscan under consideration, and is complete + * except for the fields to be set by this routine + * 'loop_count' is the number of repetitions of the indexscan to factor into + * estimates of caching behavior + * + * In addition to rows, startup_cost and total_cost, cost_index() sets the + * path's indextotalcost and indexselectivity fields. These values will be + * needed if the IndexPath is used in a BitmapIndexScan. + * + * NOTE: path->indexquals must contain only clauses usable as index + * restrictions. Any additional quals evaluated as qpquals may reduce the + * number of returned tuples, but they won't reduce the number of tuples + * we have to fetch from the table, so they don't reduce the scan cost. + */ +void +cost_index(IndexPath *path, PlannerInfo *root, double loop_count, + bool partial_path) +{ + IndexOptInfo *index = path->indexinfo; + RelOptInfo *baserel = index->rel; + bool indexonly = (path->path.pathtype == T_IndexOnlyScan); + amcostestimate_function amcostestimate; + List *qpquals; + Cost startup_cost = 0; + Cost run_cost = 0; + Cost cpu_run_cost = 0; + Cost indexStartupCost; + Cost indexTotalCost; + Selectivity indexSelectivity; + double indexCorrelation, + csquared; + double spc_seq_page_cost, + spc_random_page_cost; + Cost min_IO_cost, + max_IO_cost; + QualCost qpqual_cost; + Cost cpu_per_tuple; + double tuples_fetched; + double pages_fetched; + double rand_heap_pages; + double index_pages; + + /* Should only be applied to base relations */ + Assert(IsA(baserel, RelOptInfo) && + IsA(index, IndexOptInfo)); + Assert(baserel->relid > 0); + Assert(baserel->rtekind == RTE_RELATION); + + /* + * Mark the path with the correct row estimate, and identify which quals + * will need to be enforced as qpquals. We need not check any quals that + * are implied by the index's predicate, so we can use indrestrictinfo not + * baserestrictinfo as the list of relevant restriction clauses for the + * rel. + */ + if (path->path.param_info) + { + path->path.rows = path->path.param_info->ppi_rows; + /* qpquals come from the rel's restriction clauses and ppi_clauses */ + qpquals = list_concat(extract_nonindex_conditions(path->indexinfo->indrestrictinfo, + path->indexclauses), + extract_nonindex_conditions(path->path.param_info->ppi_clauses, + path->indexclauses)); + } + else + { + path->path.rows = baserel->rows; + /* qpquals come from just the rel's restriction clauses */ + qpquals = extract_nonindex_conditions(path->indexinfo->indrestrictinfo, + path->indexclauses); + } + + if (!enable_indexscan) + startup_cost += disable_cost; + /* we don't need to check enable_indexonlyscan; indxpath.c does that */ + + /* + * Call index-access-method-specific code to estimate the processing cost + * for scanning the index, as well as the selectivity of the index (ie, + * the fraction of main-table tuples we will have to retrieve) and its + * correlation to the main-table tuple order. We need a cast here because + * pathnodes.h uses a weak function type to avoid including amapi.h. + */ + amcostestimate = (amcostestimate_function) index->amcostestimate; + amcostestimate(root, path, loop_count, + &indexStartupCost, &indexTotalCost, + &indexSelectivity, &indexCorrelation, + &index_pages); + + /* + * Save amcostestimate's results for possible use in bitmap scan planning. + * We don't bother to save indexStartupCost or indexCorrelation, because a + * bitmap scan doesn't care about either. + */ + path->indextotalcost = indexTotalCost; + path->indexselectivity = indexSelectivity; + + /* all costs for touching index itself included here */ + startup_cost += indexStartupCost; + run_cost += indexTotalCost - indexStartupCost; + + /* estimate number of main-table tuples fetched */ + tuples_fetched = clamp_row_est(indexSelectivity * baserel->tuples); + + /* fetch estimated page costs for tablespace containing table */ + get_tablespace_page_costs(baserel->reltablespace, + &spc_random_page_cost, + &spc_seq_page_cost); + + /*---------- + * Estimate number of main-table pages fetched, and compute I/O cost. + * + * When the index ordering is uncorrelated with the table ordering, + * we use an approximation proposed by Mackert and Lohman (see + * index_pages_fetched() for details) to compute the number of pages + * fetched, and then charge spc_random_page_cost per page fetched. + * + * When the index ordering is exactly correlated with the table ordering + * (just after a CLUSTER, for example), the number of pages fetched should + * be exactly selectivity * table_size. What's more, all but the first + * will be sequential fetches, not the random fetches that occur in the + * uncorrelated case. So if the number of pages is more than 1, we + * ought to charge + * spc_random_page_cost + (pages_fetched - 1) * spc_seq_page_cost + * For partially-correlated indexes, we ought to charge somewhere between + * these two estimates. We currently interpolate linearly between the + * estimates based on the correlation squared (XXX is that appropriate?). + * + * If it's an index-only scan, then we will not need to fetch any heap + * pages for which the visibility map shows all tuples are visible. + * Hence, reduce the estimated number of heap fetches accordingly. + * We use the measured fraction of the entire heap that is all-visible, + * which might not be particularly relevant to the subset of the heap + * that this query will fetch; but it's not clear how to do better. + *---------- + */ + if (loop_count > 1) + { + /* + * For repeated indexscans, the appropriate estimate for the + * uncorrelated case is to scale up the number of tuples fetched in + * the Mackert and Lohman formula by the number of scans, so that we + * estimate the number of pages fetched by all the scans; then + * pro-rate the costs for one scan. In this case we assume all the + * fetches are random accesses. + */ + pages_fetched = index_pages_fetched(tuples_fetched * loop_count, + baserel->pages, + (double) index->pages, + root); + + if (indexonly) + pages_fetched = ceil(pages_fetched * (1.0 - baserel->allvisfrac)); + + rand_heap_pages = pages_fetched; + + max_IO_cost = (pages_fetched * spc_random_page_cost) / loop_count; + + /* + * In the perfectly correlated case, the number of pages touched by + * each scan is selectivity * table_size, and we can use the Mackert + * and Lohman formula at the page level to estimate how much work is + * saved by caching across scans. We still assume all the fetches are + * random, though, which is an overestimate that's hard to correct for + * without double-counting the cache effects. (But in most cases + * where such a plan is actually interesting, only one page would get + * fetched per scan anyway, so it shouldn't matter much.) + */ + pages_fetched = ceil(indexSelectivity * (double) baserel->pages); + + pages_fetched = index_pages_fetched(pages_fetched * loop_count, + baserel->pages, + (double) index->pages, + root); + + if (indexonly) + pages_fetched = ceil(pages_fetched * (1.0 - baserel->allvisfrac)); + + min_IO_cost = (pages_fetched * spc_random_page_cost) / loop_count; + } + else + { + /* + * Normal case: apply the Mackert and Lohman formula, and then + * interpolate between that and the correlation-derived result. + */ + pages_fetched = index_pages_fetched(tuples_fetched, + baserel->pages, + (double) index->pages, + root); + + if (indexonly) + pages_fetched = ceil(pages_fetched * (1.0 - baserel->allvisfrac)); + + rand_heap_pages = pages_fetched; + + /* max_IO_cost is for the perfectly uncorrelated case (csquared=0) */ + max_IO_cost = pages_fetched * spc_random_page_cost; + + /* min_IO_cost is for the perfectly correlated case (csquared=1) */ + pages_fetched = ceil(indexSelectivity * (double) baserel->pages); + + if (indexonly) + pages_fetched = ceil(pages_fetched * (1.0 - baserel->allvisfrac)); + + if (pages_fetched > 0) + { + min_IO_cost = spc_random_page_cost; + if (pages_fetched > 1) + min_IO_cost += (pages_fetched - 1) * spc_seq_page_cost; + } + else + min_IO_cost = 0; + } + + if (partial_path) + { + /* + * For index only scans compute workers based on number of index pages + * fetched; the number of heap pages we fetch might be so small as to + * effectively rule out parallelism, which we don't want to do. + */ + if (indexonly) + rand_heap_pages = -1; + + /* + * Estimate the number of parallel workers required to scan index. Use + * the number of heap pages computed considering heap fetches won't be + * sequential as for parallel scans the pages are accessed in random + * order. + */ + path->path.parallel_workers = compute_parallel_worker(baserel, + rand_heap_pages, + index_pages, + max_parallel_workers_per_gather); + + /* + * Fall out if workers can't be assigned for parallel scan, because in + * such a case this path will be rejected. So there is no benefit in + * doing extra computation. + */ + if (path->path.parallel_workers <= 0) + return; + + path->path.parallel_aware = true; + } + + /* + * Now interpolate based on estimated index order correlation to get total + * disk I/O cost for main table accesses. + */ + csquared = indexCorrelation * indexCorrelation; + + run_cost += max_IO_cost + csquared * (min_IO_cost - max_IO_cost); + + /* + * Estimate CPU costs per tuple. + * + * What we want here is cpu_tuple_cost plus the evaluation costs of any + * qual clauses that we have to evaluate as qpquals. + */ + cost_qual_eval(&qpqual_cost, qpquals, root); + + startup_cost += qpqual_cost.startup; + cpu_per_tuple = cpu_tuple_cost + qpqual_cost.per_tuple; + + cpu_run_cost += cpu_per_tuple * tuples_fetched; + + /* tlist eval costs are paid per output row, not per tuple scanned */ + startup_cost += path->path.pathtarget->cost.startup; + cpu_run_cost += path->path.pathtarget->cost.per_tuple * path->path.rows; + + /* Adjust costing for parallelism, if used. */ + if (path->path.parallel_workers > 0) + { + double parallel_divisor = get_parallel_divisor(&path->path); + + path->path.rows = clamp_row_est(path->path.rows / parallel_divisor); + + /* The CPU cost is divided among all the workers. */ + cpu_run_cost /= parallel_divisor; + } + + run_cost += cpu_run_cost; + + path->path.startup_cost = startup_cost; + path->path.total_cost = startup_cost + run_cost; +} + +/* + * extract_nonindex_conditions + * + * Given a list of quals to be enforced in an indexscan, extract the ones that + * will have to be applied as qpquals (ie, the index machinery won't handle + * them). Here we detect only whether a qual clause is directly redundant + * with some indexclause. If the index path is chosen for use, createplan.c + * will try a bit harder to get rid of redundant qual conditions; specifically + * it will see if quals can be proven to be implied by the indexquals. But + * it does not seem worth the cycles to try to factor that in at this stage, + * since we're only trying to estimate qual eval costs. Otherwise this must + * match the logic in create_indexscan_plan(). + * + * qual_clauses, and the result, are lists of RestrictInfos. + * indexclauses is a list of IndexClauses. + */ +static List * +extract_nonindex_conditions(List *qual_clauses, List *indexclauses) +{ + List *result = NIL; + ListCell *lc; + + foreach(lc, qual_clauses) + { + RestrictInfo *rinfo = lfirst_node(RestrictInfo, lc); + + if (rinfo->pseudoconstant) + continue; /* we may drop pseudoconstants here */ + if (is_redundant_with_indexclauses(rinfo, indexclauses)) + continue; /* dup or derived from same EquivalenceClass */ + /* ... skip the predicate proof attempt createplan.c will try ... */ + result = lappend(result, rinfo); + } + return result; +} + +/* + * index_pages_fetched + * Estimate the number of pages actually fetched after accounting for + * cache effects. + * + * We use an approximation proposed by Mackert and Lohman, "Index Scans + * Using a Finite LRU Buffer: A Validated I/O Model", ACM Transactions + * on Database Systems, Vol. 14, No. 3, September 1989, Pages 401-424. + * The Mackert and Lohman approximation is that the number of pages + * fetched is + * PF = + * min(2TNs/(2T+Ns), T) when T <= b + * 2TNs/(2T+Ns) when T > b and Ns <= 2Tb/(2T-b) + * b + (Ns - 2Tb/(2T-b))*(T-b)/T when T > b and Ns > 2Tb/(2T-b) + * where + * T = # pages in table + * N = # tuples in table + * s = selectivity = fraction of table to be scanned + * b = # buffer pages available (we include kernel space here) + * + * We assume that effective_cache_size is the total number of buffer pages + * available for the whole query, and pro-rate that space across all the + * tables in the query and the index currently under consideration. (This + * ignores space needed for other indexes used by the query, but since we + * don't know which indexes will get used, we can't estimate that very well; + * and in any case counting all the tables may well be an overestimate, since + * depending on the join plan not all the tables may be scanned concurrently.) + * + * The product Ns is the number of tuples fetched; we pass in that + * product rather than calculating it here. "pages" is the number of pages + * in the object under consideration (either an index or a table). + * "index_pages" is the amount to add to the total table space, which was + * computed for us by make_one_rel. + * + * Caller is expected to have ensured that tuples_fetched is greater than zero + * and rounded to integer (see clamp_row_est). The result will likewise be + * greater than zero and integral. + */ +double +index_pages_fetched(double tuples_fetched, BlockNumber pages, + double index_pages, PlannerInfo *root) +{ + double pages_fetched; + double total_pages; + double T, + b; + + /* T is # pages in table, but don't allow it to be zero */ + T = (pages > 1) ? (double) pages : 1.0; + + /* Compute number of pages assumed to be competing for cache space */ + total_pages = root->total_table_pages + index_pages; + total_pages = Max(total_pages, 1.0); + Assert(T <= total_pages); + + /* b is pro-rated share of effective_cache_size */ + b = (double) effective_cache_size * T / total_pages; + + /* force it positive and integral */ + if (b <= 1.0) + b = 1.0; + else + b = ceil(b); + + /* This part is the Mackert and Lohman formula */ + if (T <= b) + { + pages_fetched = + (2.0 * T * tuples_fetched) / (2.0 * T + tuples_fetched); + if (pages_fetched >= T) + pages_fetched = T; + else + pages_fetched = ceil(pages_fetched); + } + else + { + double lim; + + lim = (2.0 * T * b) / (2.0 * T - b); + if (tuples_fetched <= lim) + { + pages_fetched = + (2.0 * T * tuples_fetched) / (2.0 * T + tuples_fetched); + } + else + { + pages_fetched = + b + (tuples_fetched - lim) * (T - b) / T; + } + pages_fetched = ceil(pages_fetched); + } + return pages_fetched; +} + +/* + * get_indexpath_pages + * Determine the total size of the indexes used in a bitmap index path. + * + * Note: if the same index is used more than once in a bitmap tree, we will + * count it multiple times, which perhaps is the wrong thing ... but it's + * not completely clear, and detecting duplicates is difficult, so ignore it + * for now. + */ +static double +get_indexpath_pages(Path *bitmapqual) +{ + double result = 0; + ListCell *l; + + if (IsA(bitmapqual, BitmapAndPath)) + { + BitmapAndPath *apath = (BitmapAndPath *) bitmapqual; + + foreach(l, apath->bitmapquals) + { + result += get_indexpath_pages((Path *) lfirst(l)); + } + } + else if (IsA(bitmapqual, BitmapOrPath)) + { + BitmapOrPath *opath = (BitmapOrPath *) bitmapqual; + + foreach(l, opath->bitmapquals) + { + result += get_indexpath_pages((Path *) lfirst(l)); + } + } + else if (IsA(bitmapqual, IndexPath)) + { + IndexPath *ipath = (IndexPath *) bitmapqual; + + result = (double) ipath->indexinfo->pages; + } + else + elog(ERROR, "unrecognized node type: %d", nodeTag(bitmapqual)); + + return result; +} + +/* + * cost_bitmap_heap_scan + * Determines and returns the cost of scanning a relation using a bitmap + * index-then-heap plan. + * + * 'baserel' is the relation to be scanned + * 'param_info' is the ParamPathInfo if this is a parameterized path, else NULL + * 'bitmapqual' is a tree of IndexPaths, BitmapAndPaths, and BitmapOrPaths + * 'loop_count' is the number of repetitions of the indexscan to factor into + * estimates of caching behavior + * + * Note: the component IndexPaths in bitmapqual should have been costed + * using the same loop_count. + */ +void +cost_bitmap_heap_scan(Path *path, PlannerInfo *root, RelOptInfo *baserel, + ParamPathInfo *param_info, + Path *bitmapqual, double loop_count) +{ + Cost startup_cost = 0; + Cost run_cost = 0; + Cost indexTotalCost; + QualCost qpqual_cost; + Cost cpu_per_tuple; + Cost cost_per_page; + Cost cpu_run_cost; + double tuples_fetched; + double pages_fetched; + double spc_seq_page_cost, + spc_random_page_cost; + double T; + + /* Should only be applied to base relations */ + Assert(IsA(baserel, RelOptInfo)); + Assert(baserel->relid > 0); + Assert(baserel->rtekind == RTE_RELATION); + + /* Mark the path with the correct row estimate */ + if (param_info) + path->rows = param_info->ppi_rows; + else + path->rows = baserel->rows; + + if (!enable_bitmapscan) + startup_cost += disable_cost; + + pages_fetched = compute_bitmap_pages(root, baserel, bitmapqual, + loop_count, &indexTotalCost, + &tuples_fetched); + + startup_cost += indexTotalCost; + T = (baserel->pages > 1) ? (double) baserel->pages : 1.0; + + /* Fetch estimated page costs for tablespace containing table. */ + get_tablespace_page_costs(baserel->reltablespace, + &spc_random_page_cost, + &spc_seq_page_cost); + + /* + * For small numbers of pages we should charge spc_random_page_cost + * apiece, while if nearly all the table's pages are being read, it's more + * appropriate to charge spc_seq_page_cost apiece. The effect is + * nonlinear, too. For lack of a better idea, interpolate like this to + * determine the cost per page. + */ + if (pages_fetched >= 2.0) + cost_per_page = spc_random_page_cost - + (spc_random_page_cost - spc_seq_page_cost) + * sqrt(pages_fetched / T); + else + cost_per_page = spc_random_page_cost; + + run_cost += pages_fetched * cost_per_page; + + /* + * Estimate CPU costs per tuple. + * + * Often the indexquals don't need to be rechecked at each tuple ... but + * not always, especially not if there are enough tuples involved that the + * bitmaps become lossy. For the moment, just assume they will be + * rechecked always. This means we charge the full freight for all the + * scan clauses. + */ + get_restriction_qual_cost(root, baserel, param_info, &qpqual_cost); + + startup_cost += qpqual_cost.startup; + cpu_per_tuple = cpu_tuple_cost + qpqual_cost.per_tuple; + cpu_run_cost = cpu_per_tuple * tuples_fetched; + + /* Adjust costing for parallelism, if used. */ + if (path->parallel_workers > 0) + { + double parallel_divisor = get_parallel_divisor(path); + + /* The CPU cost is divided among all the workers. */ + cpu_run_cost /= parallel_divisor; + + path->rows = clamp_row_est(path->rows / parallel_divisor); + } + + + run_cost += cpu_run_cost; + + /* tlist eval costs are paid per output row, not per tuple scanned */ + startup_cost += path->pathtarget->cost.startup; + run_cost += path->pathtarget->cost.per_tuple * path->rows; + + path->startup_cost = startup_cost; + path->total_cost = startup_cost + run_cost; +} + +/* + * cost_bitmap_tree_node + * Extract cost and selectivity from a bitmap tree node (index/and/or) + */ +void +cost_bitmap_tree_node(Path *path, Cost *cost, Selectivity *selec) +{ + if (IsA(path, IndexPath)) + { + *cost = ((IndexPath *) path)->indextotalcost; + *selec = ((IndexPath *) path)->indexselectivity; + + /* + * Charge a small amount per retrieved tuple to reflect the costs of + * manipulating the bitmap. This is mostly to make sure that a bitmap + * scan doesn't look to be the same cost as an indexscan to retrieve a + * single tuple. + */ + *cost += 0.1 * cpu_operator_cost * path->rows; + } + else if (IsA(path, BitmapAndPath)) + { + *cost = path->total_cost; + *selec = ((BitmapAndPath *) path)->bitmapselectivity; + } + else if (IsA(path, BitmapOrPath)) + { + *cost = path->total_cost; + *selec = ((BitmapOrPath *) path)->bitmapselectivity; + } + else + { + elog(ERROR, "unrecognized node type: %d", nodeTag(path)); + *cost = *selec = 0; /* keep compiler quiet */ + } +} + +/* + * cost_bitmap_and_node + * Estimate the cost of a BitmapAnd node + * + * Note that this considers only the costs of index scanning and bitmap + * creation, not the eventual heap access. In that sense the object isn't + * truly a Path, but it has enough path-like properties (costs in particular) + * to warrant treating it as one. We don't bother to set the path rows field, + * however. + */ +void +cost_bitmap_and_node(BitmapAndPath *path, PlannerInfo *root) +{ + Cost totalCost; + Selectivity selec; + ListCell *l; + + /* + * We estimate AND selectivity on the assumption that the inputs are + * independent. This is probably often wrong, but we don't have the info + * to do better. + * + * The runtime cost of the BitmapAnd itself is estimated at 100x + * cpu_operator_cost for each tbm_intersect needed. Probably too small, + * definitely too simplistic? + */ + totalCost = 0.0; + selec = 1.0; + foreach(l, path->bitmapquals) + { + Path *subpath = (Path *) lfirst(l); + Cost subCost; + Selectivity subselec; + + cost_bitmap_tree_node(subpath, &subCost, &subselec); + + selec *= subselec; + + totalCost += subCost; + if (l != list_head(path->bitmapquals)) + totalCost += 100.0 * cpu_operator_cost; + } + path->bitmapselectivity = selec; + path->path.rows = 0; /* per above, not used */ + path->path.startup_cost = totalCost; + path->path.total_cost = totalCost; +} + +/* + * cost_bitmap_or_node + * Estimate the cost of a BitmapOr node + * + * See comments for cost_bitmap_and_node. + */ +void +cost_bitmap_or_node(BitmapOrPath *path, PlannerInfo *root) +{ + Cost totalCost; + Selectivity selec; + ListCell *l; + + /* + * We estimate OR selectivity on the assumption that the inputs are + * non-overlapping, since that's often the case in "x IN (list)" type + * situations. Of course, we clamp to 1.0 at the end. + * + * The runtime cost of the BitmapOr itself is estimated at 100x + * cpu_operator_cost for each tbm_union needed. Probably too small, + * definitely too simplistic? We are aware that the tbm_unions are + * optimized out when the inputs are BitmapIndexScans. + */ + totalCost = 0.0; + selec = 0.0; + foreach(l, path->bitmapquals) + { + Path *subpath = (Path *) lfirst(l); + Cost subCost; + Selectivity subselec; + + cost_bitmap_tree_node(subpath, &subCost, &subselec); + + selec += subselec; + + totalCost += subCost; + if (l != list_head(path->bitmapquals) && + !IsA(subpath, IndexPath)) + totalCost += 100.0 * cpu_operator_cost; + } + path->bitmapselectivity = Min(selec, 1.0); + path->path.rows = 0; /* per above, not used */ + path->path.startup_cost = totalCost; + path->path.total_cost = totalCost; +} + +/* + * cost_tidscan + * Determines and returns the cost of scanning a relation using TIDs. + * + * 'baserel' is the relation to be scanned + * 'tidquals' is the list of TID-checkable quals + * 'param_info' is the ParamPathInfo if this is a parameterized path, else NULL + */ +void +cost_tidscan(Path *path, PlannerInfo *root, + RelOptInfo *baserel, List *tidquals, ParamPathInfo *param_info) +{ + Cost startup_cost = 0; + Cost run_cost = 0; + bool isCurrentOf = false; + QualCost qpqual_cost; + Cost cpu_per_tuple; + QualCost tid_qual_cost; + int ntuples; + ListCell *l; + double spc_random_page_cost; + + /* Should only be applied to base relations */ + Assert(baserel->relid > 0); + Assert(baserel->rtekind == RTE_RELATION); + + /* Mark the path with the correct row estimate */ + if (param_info) + path->rows = param_info->ppi_rows; + else + path->rows = baserel->rows; + + /* Count how many tuples we expect to retrieve */ + ntuples = 0; + foreach(l, tidquals) + { + RestrictInfo *rinfo = lfirst_node(RestrictInfo, l); + Expr *qual = rinfo->clause; + + if (IsA(qual, ScalarArrayOpExpr)) + { + /* Each element of the array yields 1 tuple */ + ScalarArrayOpExpr *saop = (ScalarArrayOpExpr *) qual; + Node *arraynode = (Node *) lsecond(saop->args); + + ntuples += estimate_array_length(arraynode); + } + else if (IsA(qual, CurrentOfExpr)) + { + /* CURRENT OF yields 1 tuple */ + isCurrentOf = true; + ntuples++; + } + else + { + /* It's just CTID = something, count 1 tuple */ + ntuples++; + } + } + + /* + * We must force TID scan for WHERE CURRENT OF, because only nodeTidscan.c + * understands how to do it correctly. Therefore, honor enable_tidscan + * only when CURRENT OF isn't present. Also note that cost_qual_eval + * counts a CurrentOfExpr as having startup cost disable_cost, which we + * subtract off here; that's to prevent other plan types such as seqscan + * from winning. + */ + if (isCurrentOf) + { + Assert(baserel->baserestrictcost.startup >= disable_cost); + startup_cost -= disable_cost; + } + else if (!enable_tidscan) + startup_cost += disable_cost; + + /* + * The TID qual expressions will be computed once, any other baserestrict + * quals once per retrieved tuple. + */ + cost_qual_eval(&tid_qual_cost, tidquals, root); + + /* fetch estimated page cost for tablespace containing table */ + get_tablespace_page_costs(baserel->reltablespace, + &spc_random_page_cost, + NULL); + + /* disk costs --- assume each tuple on a different page */ + run_cost += spc_random_page_cost * ntuples; + + /* Add scanning CPU costs */ + get_restriction_qual_cost(root, baserel, param_info, &qpqual_cost); + + /* XXX currently we assume TID quals are a subset of qpquals */ + startup_cost += qpqual_cost.startup + tid_qual_cost.per_tuple; + cpu_per_tuple = cpu_tuple_cost + qpqual_cost.per_tuple - + tid_qual_cost.per_tuple; + run_cost += cpu_per_tuple * ntuples; + + /* tlist eval costs are paid per output row, not per tuple scanned */ + startup_cost += path->pathtarget->cost.startup; + run_cost += path->pathtarget->cost.per_tuple * path->rows; + + path->startup_cost = startup_cost; + path->total_cost = startup_cost + run_cost; +} + +/* + * cost_tidrangescan + * Determines and sets the costs of scanning a relation using a range of + * TIDs for 'path' + * + * 'baserel' is the relation to be scanned + * 'tidrangequals' is the list of TID-checkable range quals + * 'param_info' is the ParamPathInfo if this is a parameterized path, else NULL + */ +void +cost_tidrangescan(Path *path, PlannerInfo *root, + RelOptInfo *baserel, List *tidrangequals, + ParamPathInfo *param_info) +{ + Selectivity selectivity; + double pages; + Cost startup_cost = 0; + Cost run_cost = 0; + QualCost qpqual_cost; + Cost cpu_per_tuple; + QualCost tid_qual_cost; + double ntuples; + double nseqpages; + double spc_random_page_cost; + double spc_seq_page_cost; + + /* Should only be applied to base relations */ + Assert(baserel->relid > 0); + Assert(baserel->rtekind == RTE_RELATION); + + /* Mark the path with the correct row estimate */ + if (param_info) + path->rows = param_info->ppi_rows; + else + path->rows = baserel->rows; + + /* Count how many tuples and pages we expect to scan */ + selectivity = clauselist_selectivity(root, tidrangequals, baserel->relid, + JOIN_INNER, NULL); + pages = ceil(selectivity * baserel->pages); + + if (pages <= 0.0) + pages = 1.0; + + /* + * The first page in a range requires a random seek, but each subsequent + * page is just a normal sequential page read. NOTE: it's desirable for + * TID Range Scans to cost more than the equivalent Sequential Scans, + * because Seq Scans have some performance advantages such as scan + * synchronization and parallelizability, and we'd prefer one of them to + * be picked unless a TID Range Scan really is better. + */ + ntuples = selectivity * baserel->tuples; + nseqpages = pages - 1.0; + + if (!enable_tidscan) + startup_cost += disable_cost; + + /* + * The TID qual expressions will be computed once, any other baserestrict + * quals once per retrieved tuple. + */ + cost_qual_eval(&tid_qual_cost, tidrangequals, root); + + /* fetch estimated page cost for tablespace containing table */ + get_tablespace_page_costs(baserel->reltablespace, + &spc_random_page_cost, + &spc_seq_page_cost); + + /* disk costs; 1 random page and the remainder as seq pages */ + run_cost += spc_random_page_cost + spc_seq_page_cost * nseqpages; + + /* Add scanning CPU costs */ + get_restriction_qual_cost(root, baserel, param_info, &qpqual_cost); + + /* + * XXX currently we assume TID quals are a subset of qpquals at this + * point; they will be removed (if possible) when we create the plan, so + * we subtract their cost from the total qpqual cost. (If the TID quals + * can't be removed, this is a mistake and we're going to underestimate + * the CPU cost a bit.) + */ + startup_cost += qpqual_cost.startup + tid_qual_cost.per_tuple; + cpu_per_tuple = cpu_tuple_cost + qpqual_cost.per_tuple - + tid_qual_cost.per_tuple; + run_cost += cpu_per_tuple * ntuples; + + /* tlist eval costs are paid per output row, not per tuple scanned */ + startup_cost += path->pathtarget->cost.startup; + run_cost += path->pathtarget->cost.per_tuple * path->rows; + + path->startup_cost = startup_cost; + path->total_cost = startup_cost + run_cost; +} + +/* + * cost_subqueryscan + * Determines and returns the cost of scanning a subquery RTE. + * + * 'baserel' is the relation to be scanned + * 'param_info' is the ParamPathInfo if this is a parameterized path, else NULL + */ +void +cost_subqueryscan(SubqueryScanPath *path, PlannerInfo *root, + RelOptInfo *baserel, ParamPathInfo *param_info) +{ + Cost startup_cost; + Cost run_cost; + List *qpquals; + QualCost qpqual_cost; + Cost cpu_per_tuple; + + /* Should only be applied to base relations that are subqueries */ + Assert(baserel->relid > 0); + Assert(baserel->rtekind == RTE_SUBQUERY); + + /* + * We compute the rowcount estimate as the subplan's estimate times the + * selectivity of relevant restriction clauses. In simple cases this will + * come out the same as baserel->rows; but when dealing with parallelized + * paths we must do it like this to get the right answer. + */ + if (param_info) + qpquals = list_concat_copy(param_info->ppi_clauses, + baserel->baserestrictinfo); + else + qpquals = baserel->baserestrictinfo; + + path->path.rows = clamp_row_est(path->subpath->rows * + clauselist_selectivity(root, + qpquals, + 0, + JOIN_INNER, + NULL)); + + /* + * Cost of path is cost of evaluating the subplan, plus cost of evaluating + * any restriction clauses and tlist that will be attached to the + * SubqueryScan node, plus cpu_tuple_cost to account for selection and + * projection overhead. + */ + path->path.startup_cost = path->subpath->startup_cost; + path->path.total_cost = path->subpath->total_cost; + + get_restriction_qual_cost(root, baserel, param_info, &qpqual_cost); + + startup_cost = qpqual_cost.startup; + cpu_per_tuple = cpu_tuple_cost + qpqual_cost.per_tuple; + run_cost = cpu_per_tuple * path->subpath->rows; + + /* tlist eval costs are paid per output row, not per tuple scanned */ + startup_cost += path->path.pathtarget->cost.startup; + run_cost += path->path.pathtarget->cost.per_tuple * path->path.rows; + + path->path.startup_cost += startup_cost; + path->path.total_cost += startup_cost + run_cost; +} + +/* + * cost_functionscan + * Determines and returns the cost of scanning a function RTE. + * + * 'baserel' is the relation to be scanned + * 'param_info' is the ParamPathInfo if this is a parameterized path, else NULL + */ +void +cost_functionscan(Path *path, PlannerInfo *root, + RelOptInfo *baserel, ParamPathInfo *param_info) +{ + Cost startup_cost = 0; + Cost run_cost = 0; + QualCost qpqual_cost; + Cost cpu_per_tuple; + RangeTblEntry *rte; + QualCost exprcost; + + /* Should only be applied to base relations that are functions */ + Assert(baserel->relid > 0); + rte = planner_rt_fetch(baserel->relid, root); + Assert(rte->rtekind == RTE_FUNCTION); + + /* Mark the path with the correct row estimate */ + if (param_info) + path->rows = param_info->ppi_rows; + else + path->rows = baserel->rows; + + /* + * Estimate costs of executing the function expression(s). + * + * Currently, nodeFunctionscan.c always executes the functions to + * completion before returning any rows, and caches the results in a + * tuplestore. So the function eval cost is all startup cost, and per-row + * costs are minimal. + * + * XXX in principle we ought to charge tuplestore spill costs if the + * number of rows is large. However, given how phony our rowcount + * estimates for functions tend to be, there's not a lot of point in that + * refinement right now. + */ + cost_qual_eval_node(&exprcost, (Node *) rte->functions, root); + + startup_cost += exprcost.startup + exprcost.per_tuple; + + /* Add scanning CPU costs */ + get_restriction_qual_cost(root, baserel, param_info, &qpqual_cost); + + startup_cost += qpqual_cost.startup; + cpu_per_tuple = cpu_tuple_cost + qpqual_cost.per_tuple; + run_cost += cpu_per_tuple * baserel->tuples; + + /* tlist eval costs are paid per output row, not per tuple scanned */ + startup_cost += path->pathtarget->cost.startup; + run_cost += path->pathtarget->cost.per_tuple * path->rows; + + path->startup_cost = startup_cost; + path->total_cost = startup_cost + run_cost; +} + +/* + * cost_tablefuncscan + * Determines and returns the cost of scanning a table function. + * + * 'baserel' is the relation to be scanned + * 'param_info' is the ParamPathInfo if this is a parameterized path, else NULL + */ +void +cost_tablefuncscan(Path *path, PlannerInfo *root, + RelOptInfo *baserel, ParamPathInfo *param_info) +{ + Cost startup_cost = 0; + Cost run_cost = 0; + QualCost qpqual_cost; + Cost cpu_per_tuple; + RangeTblEntry *rte; + QualCost exprcost; + + /* Should only be applied to base relations that are functions */ + Assert(baserel->relid > 0); + rte = planner_rt_fetch(baserel->relid, root); + Assert(rte->rtekind == RTE_TABLEFUNC); + + /* Mark the path with the correct row estimate */ + if (param_info) + path->rows = param_info->ppi_rows; + else + path->rows = baserel->rows; + + /* + * Estimate costs of executing the table func expression(s). + * + * XXX in principle we ought to charge tuplestore spill costs if the + * number of rows is large. However, given how phony our rowcount + * estimates for tablefuncs tend to be, there's not a lot of point in that + * refinement right now. + */ + cost_qual_eval_node(&exprcost, (Node *) rte->tablefunc, root); + + startup_cost += exprcost.startup + exprcost.per_tuple; + + /* Add scanning CPU costs */ + get_restriction_qual_cost(root, baserel, param_info, &qpqual_cost); + + startup_cost += qpqual_cost.startup; + cpu_per_tuple = cpu_tuple_cost + qpqual_cost.per_tuple; + run_cost += cpu_per_tuple * baserel->tuples; + + /* tlist eval costs are paid per output row, not per tuple scanned */ + startup_cost += path->pathtarget->cost.startup; + run_cost += path->pathtarget->cost.per_tuple * path->rows; + + path->startup_cost = startup_cost; + path->total_cost = startup_cost + run_cost; +} + +/* + * cost_valuesscan + * Determines and returns the cost of scanning a VALUES RTE. + * + * 'baserel' is the relation to be scanned + * 'param_info' is the ParamPathInfo if this is a parameterized path, else NULL + */ +void +cost_valuesscan(Path *path, PlannerInfo *root, + RelOptInfo *baserel, ParamPathInfo *param_info) +{ + Cost startup_cost = 0; + Cost run_cost = 0; + QualCost qpqual_cost; + Cost cpu_per_tuple; + + /* Should only be applied to base relations that are values lists */ + Assert(baserel->relid > 0); + Assert(baserel->rtekind == RTE_VALUES); + + /* Mark the path with the correct row estimate */ + if (param_info) + path->rows = param_info->ppi_rows; + else + path->rows = baserel->rows; + + /* + * For now, estimate list evaluation cost at one operator eval per list + * (probably pretty bogus, but is it worth being smarter?) + */ + cpu_per_tuple = cpu_operator_cost; + + /* Add scanning CPU costs */ + get_restriction_qual_cost(root, baserel, param_info, &qpqual_cost); + + startup_cost += qpqual_cost.startup; + cpu_per_tuple += cpu_tuple_cost + qpqual_cost.per_tuple; + run_cost += cpu_per_tuple * baserel->tuples; + + /* tlist eval costs are paid per output row, not per tuple scanned */ + startup_cost += path->pathtarget->cost.startup; + run_cost += path->pathtarget->cost.per_tuple * path->rows; + + path->startup_cost = startup_cost; + path->total_cost = startup_cost + run_cost; +} + +/* + * cost_ctescan + * Determines and returns the cost of scanning a CTE RTE. + * + * Note: this is used for both self-reference and regular CTEs; the + * possible cost differences are below the threshold of what we could + * estimate accurately anyway. Note that the costs of evaluating the + * referenced CTE query are added into the final plan as initplan costs, + * and should NOT be counted here. + */ +void +cost_ctescan(Path *path, PlannerInfo *root, + RelOptInfo *baserel, ParamPathInfo *param_info) +{ + Cost startup_cost = 0; + Cost run_cost = 0; + QualCost qpqual_cost; + Cost cpu_per_tuple; + + /* Should only be applied to base relations that are CTEs */ + Assert(baserel->relid > 0); + Assert(baserel->rtekind == RTE_CTE); + + /* Mark the path with the correct row estimate */ + if (param_info) + path->rows = param_info->ppi_rows; + else + path->rows = baserel->rows; + + /* Charge one CPU tuple cost per row for tuplestore manipulation */ + cpu_per_tuple = cpu_tuple_cost; + + /* Add scanning CPU costs */ + get_restriction_qual_cost(root, baserel, param_info, &qpqual_cost); + + startup_cost += qpqual_cost.startup; + cpu_per_tuple += cpu_tuple_cost + qpqual_cost.per_tuple; + run_cost += cpu_per_tuple * baserel->tuples; + + /* tlist eval costs are paid per output row, not per tuple scanned */ + startup_cost += path->pathtarget->cost.startup; + run_cost += path->pathtarget->cost.per_tuple * path->rows; + + path->startup_cost = startup_cost; + path->total_cost = startup_cost + run_cost; +} + +/* + * cost_namedtuplestorescan + * Determines and returns the cost of scanning a named tuplestore. + */ +void +cost_namedtuplestorescan(Path *path, PlannerInfo *root, + RelOptInfo *baserel, ParamPathInfo *param_info) +{ + Cost startup_cost = 0; + Cost run_cost = 0; + QualCost qpqual_cost; + Cost cpu_per_tuple; + + /* Should only be applied to base relations that are Tuplestores */ + Assert(baserel->relid > 0); + Assert(baserel->rtekind == RTE_NAMEDTUPLESTORE); + + /* Mark the path with the correct row estimate */ + if (param_info) + path->rows = param_info->ppi_rows; + else + path->rows = baserel->rows; + + /* Charge one CPU tuple cost per row for tuplestore manipulation */ + cpu_per_tuple = cpu_tuple_cost; + + /* Add scanning CPU costs */ + get_restriction_qual_cost(root, baserel, param_info, &qpqual_cost); + + startup_cost += qpqual_cost.startup; + cpu_per_tuple += cpu_tuple_cost + qpqual_cost.per_tuple; + run_cost += cpu_per_tuple * baserel->tuples; + + path->startup_cost = startup_cost; + path->total_cost = startup_cost + run_cost; +} + +/* + * cost_resultscan + * Determines and returns the cost of scanning an RTE_RESULT relation. + */ +void +cost_resultscan(Path *path, PlannerInfo *root, + RelOptInfo *baserel, ParamPathInfo *param_info) +{ + Cost startup_cost = 0; + Cost run_cost = 0; + QualCost qpqual_cost; + Cost cpu_per_tuple; + + /* Should only be applied to RTE_RESULT base relations */ + Assert(baserel->relid > 0); + Assert(baserel->rtekind == RTE_RESULT); + + /* Mark the path with the correct row estimate */ + if (param_info) + path->rows = param_info->ppi_rows; + else + path->rows = baserel->rows; + + /* We charge qual cost plus cpu_tuple_cost */ + get_restriction_qual_cost(root, baserel, param_info, &qpqual_cost); + + startup_cost += qpqual_cost.startup; + cpu_per_tuple = cpu_tuple_cost + qpqual_cost.per_tuple; + run_cost += cpu_per_tuple * baserel->tuples; + + path->startup_cost = startup_cost; + path->total_cost = startup_cost + run_cost; +} + +/* + * cost_recursive_union + * Determines and returns the cost of performing a recursive union, + * and also the estimated output size. + * + * We are given Paths for the nonrecursive and recursive terms. + */ +void +cost_recursive_union(Path *runion, Path *nrterm, Path *rterm) +{ + Cost startup_cost; + Cost total_cost; + double total_rows; + + /* We probably have decent estimates for the non-recursive term */ + startup_cost = nrterm->startup_cost; + total_cost = nrterm->total_cost; + total_rows = nrterm->rows; + + /* + * We arbitrarily assume that about 10 recursive iterations will be + * needed, and that we've managed to get a good fix on the cost and output + * size of each one of them. These are mighty shaky assumptions but it's + * hard to see how to do better. + */ + total_cost += 10 * rterm->total_cost; + total_rows += 10 * rterm->rows; + + /* + * Also charge cpu_tuple_cost per row to account for the costs of + * manipulating the tuplestores. (We don't worry about possible + * spill-to-disk costs.) + */ + total_cost += cpu_tuple_cost * total_rows; + + runion->startup_cost = startup_cost; + runion->total_cost = total_cost; + runion->rows = total_rows; + runion->pathtarget->width = Max(nrterm->pathtarget->width, + rterm->pathtarget->width); +} + +/* + * cost_tuplesort + * Determines and returns the cost of sorting a relation using tuplesort, + * not including the cost of reading the input data. + * + * If the total volume of data to sort is less than sort_mem, we will do + * an in-memory sort, which requires no I/O and about t*log2(t) tuple + * comparisons for t tuples. + * + * If the total volume exceeds sort_mem, we switch to a tape-style merge + * algorithm. There will still be about t*log2(t) tuple comparisons in + * total, but we will also need to write and read each tuple once per + * merge pass. We expect about ceil(logM(r)) merge passes where r is the + * number of initial runs formed and M is the merge order used by tuplesort.c. + * Since the average initial run should be about sort_mem, we have + * disk traffic = 2 * relsize * ceil(logM(p / sort_mem)) + * cpu = comparison_cost * t * log2(t) + * + * If the sort is bounded (i.e., only the first k result tuples are needed) + * and k tuples can fit into sort_mem, we use a heap method that keeps only + * k tuples in the heap; this will require about t*log2(k) tuple comparisons. + * + * The disk traffic is assumed to be 3/4ths sequential and 1/4th random + * accesses (XXX can't we refine that guess?) + * + * By default, we charge two operator evals per tuple comparison, which should + * be in the right ballpark in most cases. The caller can tweak this by + * specifying nonzero comparison_cost; typically that's used for any extra + * work that has to be done to prepare the inputs to the comparison operators. + * + * 'tuples' is the number of tuples in the relation + * 'width' is the average tuple width in bytes + * 'comparison_cost' is the extra cost per comparison, if any + * 'sort_mem' is the number of kilobytes of work memory allowed for the sort + * 'limit_tuples' is the bound on the number of output tuples; -1 if no bound + */ +static void +cost_tuplesort(Cost *startup_cost, Cost *run_cost, + double tuples, int width, + Cost comparison_cost, int sort_mem, + double limit_tuples) +{ + double input_bytes = relation_byte_size(tuples, width); + double output_bytes; + double output_tuples; + long sort_mem_bytes = sort_mem * 1024L; + + /* + * We want to be sure the cost of a sort is never estimated as zero, even + * if passed-in tuple count is zero. Besides, mustn't do log(0)... + */ + if (tuples < 2.0) + tuples = 2.0; + + /* Include the default cost-per-comparison */ + comparison_cost += 2.0 * cpu_operator_cost; + + /* Do we have a useful LIMIT? */ + if (limit_tuples > 0 && limit_tuples < tuples) + { + output_tuples = limit_tuples; + output_bytes = relation_byte_size(output_tuples, width); + } + else + { + output_tuples = tuples; + output_bytes = input_bytes; + } + + if (output_bytes > sort_mem_bytes) + { + /* + * We'll have to use a disk-based sort of all the tuples + */ + double npages = ceil(input_bytes / BLCKSZ); + double nruns = input_bytes / sort_mem_bytes; + double mergeorder = tuplesort_merge_order(sort_mem_bytes); + double log_runs; + double npageaccesses; + + /* + * CPU costs + * + * Assume about N log2 N comparisons + */ + *startup_cost = comparison_cost * tuples * LOG2(tuples); + + /* Disk costs */ + + /* Compute logM(r) as log(r) / log(M) */ + if (nruns > mergeorder) + log_runs = ceil(log(nruns) / log(mergeorder)); + else + log_runs = 1.0; + npageaccesses = 2.0 * npages * log_runs; + /* Assume 3/4ths of accesses are sequential, 1/4th are not */ + *startup_cost += npageaccesses * + (seq_page_cost * 0.75 + random_page_cost * 0.25); + } + else if (tuples > 2 * output_tuples || input_bytes > sort_mem_bytes) + { + /* + * We'll use a bounded heap-sort keeping just K tuples in memory, for + * a total number of tuple comparisons of N log2 K; but the constant + * factor is a bit higher than for quicksort. Tweak it so that the + * cost curve is continuous at the crossover point. + */ + *startup_cost = comparison_cost * tuples * LOG2(2.0 * output_tuples); + } + else + { + /* We'll use plain quicksort on all the input tuples */ + *startup_cost = comparison_cost * tuples * LOG2(tuples); + } + + /* + * Also charge a small amount (arbitrarily set equal to operator cost) per + * extracted tuple. We don't charge cpu_tuple_cost because a Sort node + * doesn't do qual-checking or projection, so it has less overhead than + * most plan nodes. Note it's correct to use tuples not output_tuples + * here --- the upper LIMIT will pro-rate the run cost so we'd be double + * counting the LIMIT otherwise. + */ + *run_cost = cpu_operator_cost * tuples; +} + +/* + * cost_incremental_sort + * Determines and returns the cost of sorting a relation incrementally, when + * the input path is presorted by a prefix of the pathkeys. + * + * 'presorted_keys' is the number of leading pathkeys by which the input path + * is sorted. + * + * We estimate the number of groups into which the relation is divided by the + * leading pathkeys, and then calculate the cost of sorting a single group + * with tuplesort using cost_tuplesort(). + */ +void +cost_incremental_sort(Path *path, + PlannerInfo *root, List *pathkeys, int presorted_keys, + Cost input_startup_cost, Cost input_total_cost, + double input_tuples, int width, Cost comparison_cost, int sort_mem, + double limit_tuples) +{ + Cost startup_cost = 0, + run_cost = 0, + input_run_cost = input_total_cost - input_startup_cost; + double group_tuples, + input_groups; + Cost group_startup_cost, + group_run_cost, + group_input_run_cost; + List *presortedExprs = NIL; + ListCell *l; + int i = 0; + bool unknown_varno = false; + + Assert(presorted_keys != 0); + + /* + * We want to be sure the cost of a sort is never estimated as zero, even + * if passed-in tuple count is zero. Besides, mustn't do log(0)... + */ + if (input_tuples < 2.0) + input_tuples = 2.0; + + /* Default estimate of number of groups, capped to one group per row. */ + input_groups = Min(input_tuples, DEFAULT_NUM_DISTINCT); + + /* + * Extract presorted keys as list of expressions. + * + * We need to be careful about Vars containing "varno 0" which might have + * been introduced by generate_append_tlist, which would confuse + * estimate_num_groups (in fact it'd fail for such expressions). See + * recurse_set_operations which has to deal with the same issue. + * + * Unlike recurse_set_operations we can't access the original target list + * here, and even if we could it's not very clear how useful would that be + * for a set operation combining multiple tables. So we simply detect if + * there are any expressions with "varno 0" and use the default + * DEFAULT_NUM_DISTINCT in that case. + * + * We might also use either 1.0 (a single group) or input_tuples (each row + * being a separate group), pretty much the worst and best case for + * incremental sort. But those are extreme cases and using something in + * between seems reasonable. Furthermore, generate_append_tlist is used + * for set operations, which are likely to produce mostly unique output + * anyway - from that standpoint the DEFAULT_NUM_DISTINCT is defensive + * while maintaining lower startup cost. + */ + foreach(l, pathkeys) + { + PathKey *key = (PathKey *) lfirst(l); + EquivalenceMember *member = (EquivalenceMember *) + linitial(key->pk_eclass->ec_members); + + /* + * Check if the expression contains Var with "varno 0" so that we + * don't call estimate_num_groups in that case. + */ + if (bms_is_member(0, pull_varnos(root, (Node *) member->em_expr))) + { + unknown_varno = true; + break; + } + + /* expression not containing any Vars with "varno 0" */ + presortedExprs = lappend(presortedExprs, member->em_expr); + + i++; + if (i >= presorted_keys) + break; + } + + /* Estimate number of groups with equal presorted keys. */ + if (!unknown_varno) + input_groups = estimate_num_groups(root, presortedExprs, input_tuples, + NULL, NULL); + + group_tuples = input_tuples / input_groups; + group_input_run_cost = input_run_cost / input_groups; + + /* + * Estimate average cost of sorting of one group where presorted keys are + * equal. Incremental sort is sensitive to distribution of tuples to the + * groups, where we're relying on quite rough assumptions. Thus, we're + * pessimistic about incremental sort performance and increase its average + * group size by half. + */ + cost_tuplesort(&group_startup_cost, &group_run_cost, + 1.5 * group_tuples, width, comparison_cost, sort_mem, + limit_tuples); + + /* + * Startup cost of incremental sort is the startup cost of its first group + * plus the cost of its input. + */ + startup_cost += group_startup_cost + + input_startup_cost + group_input_run_cost; + + /* + * After we started producing tuples from the first group, the cost of + * producing all the tuples is given by the cost to finish processing this + * group, plus the total cost to process the remaining groups, plus the + * remaining cost of input. + */ + run_cost += group_run_cost + + (group_run_cost + group_startup_cost) * (input_groups - 1) + + group_input_run_cost * (input_groups - 1); + + /* + * Incremental sort adds some overhead by itself. Firstly, it has to + * detect the sort groups. This is roughly equal to one extra copy and + * comparison per tuple. Secondly, it has to reset the tuplesort context + * for every group. + */ + run_cost += (cpu_tuple_cost + comparison_cost) * input_tuples; + run_cost += 2.0 * cpu_tuple_cost * input_groups; + + path->rows = input_tuples; + path->startup_cost = startup_cost; + path->total_cost = startup_cost + run_cost; +} + +/* + * cost_sort + * Determines and returns the cost of sorting a relation, including + * the cost of reading the input data. + * + * NOTE: some callers currently pass NIL for pathkeys because they + * can't conveniently supply the sort keys. Since this routine doesn't + * currently do anything with pathkeys anyway, that doesn't matter... + * but if it ever does, it should react gracefully to lack of key data. + * (Actually, the thing we'd most likely be interested in is just the number + * of sort keys, which all callers *could* supply.) + */ +void +cost_sort(Path *path, PlannerInfo *root, + List *pathkeys, Cost input_cost, double tuples, int width, + Cost comparison_cost, int sort_mem, + double limit_tuples) + +{ + Cost startup_cost; + Cost run_cost; + + cost_tuplesort(&startup_cost, &run_cost, + tuples, width, + comparison_cost, sort_mem, + limit_tuples); + + if (!enable_sort) + startup_cost += disable_cost; + + startup_cost += input_cost; + + path->rows = tuples; + path->startup_cost = startup_cost; + path->total_cost = startup_cost + run_cost; +} + +/* + * append_nonpartial_cost + * Estimate the cost of the non-partial paths in a Parallel Append. + * The non-partial paths are assumed to be the first "numpaths" paths + * from the subpaths list, and to be in order of decreasing cost. + */ +static Cost +append_nonpartial_cost(List *subpaths, int numpaths, int parallel_workers) +{ + Cost *costarr; + int arrlen; + ListCell *l; + ListCell *cell; + int i; + int path_index; + int min_index; + int max_index; + + if (numpaths == 0) + return 0; + + /* + * Array length is number of workers or number of relevant paths, + * whichever is less. + */ + arrlen = Min(parallel_workers, numpaths); + costarr = (Cost *) palloc(sizeof(Cost) * arrlen); + + /* The first few paths will each be claimed by a different worker. */ + path_index = 0; + foreach(cell, subpaths) + { + Path *subpath = (Path *) lfirst(cell); + + if (path_index == arrlen) + break; + costarr[path_index++] = subpath->total_cost; + } + + /* + * Since subpaths are sorted by decreasing cost, the last one will have + * the minimum cost. + */ + min_index = arrlen - 1; + + /* + * For each of the remaining subpaths, add its cost to the array element + * with minimum cost. + */ + for_each_cell(l, subpaths, cell) + { + Path *subpath = (Path *) lfirst(l); + int i; + + /* Consider only the non-partial paths */ + if (path_index++ == numpaths) + break; + + costarr[min_index] += subpath->total_cost; + + /* Update the new min cost array index */ + for (min_index = i = 0; i < arrlen; i++) + { + if (costarr[i] < costarr[min_index]) + min_index = i; + } + } + + /* Return the highest cost from the array */ + for (max_index = i = 0; i < arrlen; i++) + { + if (costarr[i] > costarr[max_index]) + max_index = i; + } + + return costarr[max_index]; +} + +/* + * cost_append + * Determines and returns the cost of an Append node. + */ +void +cost_append(AppendPath *apath) +{ + ListCell *l; + + apath->path.startup_cost = 0; + apath->path.total_cost = 0; + apath->path.rows = 0; + + if (apath->subpaths == NIL) + return; + + if (!apath->path.parallel_aware) + { + List *pathkeys = apath->path.pathkeys; + + if (pathkeys == NIL) + { + Path *subpath = (Path *) linitial(apath->subpaths); + + /* + * For an unordered, non-parallel-aware Append we take the startup + * cost as the startup cost of the first subpath. + */ + apath->path.startup_cost = subpath->startup_cost; + + /* Compute rows and costs as sums of subplan rows and costs. */ + foreach(l, apath->subpaths) + { + Path *subpath = (Path *) lfirst(l); + + apath->path.rows += subpath->rows; + apath->path.total_cost += subpath->total_cost; + } + } + else + { + /* + * For an ordered, non-parallel-aware Append we take the startup + * cost as the sum of the subpath startup costs. This ensures + * that we don't underestimate the startup cost when a query's + * LIMIT is such that several of the children have to be run to + * satisfy it. This might be overkill --- another plausible hack + * would be to take the Append's startup cost as the maximum of + * the child startup costs. But we don't want to risk believing + * that an ORDER BY LIMIT query can be satisfied at small cost + * when the first child has small startup cost but later ones + * don't. (If we had the ability to deal with nonlinear cost + * interpolation for partial retrievals, we would not need to be + * so conservative about this.) + * + * This case is also different from the above in that we have to + * account for possibly injecting sorts into subpaths that aren't + * natively ordered. + */ + foreach(l, apath->subpaths) + { + Path *subpath = (Path *) lfirst(l); + Path sort_path; /* dummy for result of cost_sort */ + + if (!pathkeys_contained_in(pathkeys, subpath->pathkeys)) + { + /* + * We'll need to insert a Sort node, so include costs for + * that. We can use the parent's LIMIT if any, since we + * certainly won't pull more than that many tuples from + * any child. + */ + cost_sort(&sort_path, + NULL, /* doesn't currently need root */ + pathkeys, + subpath->total_cost, + subpath->rows, + subpath->pathtarget->width, + 0.0, + work_mem, + apath->limit_tuples); + subpath = &sort_path; + } + + apath->path.rows += subpath->rows; + apath->path.startup_cost += subpath->startup_cost; + apath->path.total_cost += subpath->total_cost; + } + } + } + else /* parallel-aware */ + { + int i = 0; + double parallel_divisor = get_parallel_divisor(&apath->path); + + /* Parallel-aware Append never produces ordered output. */ + Assert(apath->path.pathkeys == NIL); + + /* Calculate startup cost. */ + foreach(l, apath->subpaths) + { + Path *subpath = (Path *) lfirst(l); + + /* + * Append will start returning tuples when the child node having + * lowest startup cost is done setting up. We consider only the + * first few subplans that immediately get a worker assigned. + */ + if (i == 0) + apath->path.startup_cost = subpath->startup_cost; + else if (i < apath->path.parallel_workers) + apath->path.startup_cost = Min(apath->path.startup_cost, + subpath->startup_cost); + + /* + * Apply parallel divisor to subpaths. Scale the number of rows + * for each partial subpath based on the ratio of the parallel + * divisor originally used for the subpath to the one we adopted. + * Also add the cost of partial paths to the total cost, but + * ignore non-partial paths for now. + */ + if (i < apath->first_partial_path) + apath->path.rows += subpath->rows / parallel_divisor; + else + { + double subpath_parallel_divisor; + + subpath_parallel_divisor = get_parallel_divisor(subpath); + apath->path.rows += subpath->rows * (subpath_parallel_divisor / + parallel_divisor); + apath->path.total_cost += subpath->total_cost; + } + + apath->path.rows = clamp_row_est(apath->path.rows); + + i++; + } + + /* Add cost for non-partial subpaths. */ + apath->path.total_cost += + append_nonpartial_cost(apath->subpaths, + apath->first_partial_path, + apath->path.parallel_workers); + } + + /* + * Although Append does not do any selection or projection, it's not free; + * add a small per-tuple overhead. + */ + apath->path.total_cost += + cpu_tuple_cost * APPEND_CPU_COST_MULTIPLIER * apath->path.rows; +} + +/* + * cost_merge_append + * Determines and returns the cost of a MergeAppend node. + * + * MergeAppend merges several pre-sorted input streams, using a heap that + * at any given instant holds the next tuple from each stream. If there + * are N streams, we need about N*log2(N) tuple comparisons to construct + * the heap at startup, and then for each output tuple, about log2(N) + * comparisons to replace the top entry. + * + * (The effective value of N will drop once some of the input streams are + * exhausted, but it seems unlikely to be worth trying to account for that.) + * + * The heap is never spilled to disk, since we assume N is not very large. + * So this is much simpler than cost_sort. + * + * As in cost_sort, we charge two operator evals per tuple comparison. + * + * 'pathkeys' is a list of sort keys + * 'n_streams' is the number of input streams + * 'input_startup_cost' is the sum of the input streams' startup costs + * 'input_total_cost' is the sum of the input streams' total costs + * 'tuples' is the number of tuples in all the streams + */ +void +cost_merge_append(Path *path, PlannerInfo *root, + List *pathkeys, int n_streams, + Cost input_startup_cost, Cost input_total_cost, + double tuples) +{ + Cost startup_cost = 0; + Cost run_cost = 0; + Cost comparison_cost; + double N; + double logN; + + /* + * Avoid log(0)... + */ + N = (n_streams < 2) ? 2.0 : (double) n_streams; + logN = LOG2(N); + + /* Assumed cost per tuple comparison */ + comparison_cost = 2.0 * cpu_operator_cost; + + /* Heap creation cost */ + startup_cost += comparison_cost * N * logN; + + /* Per-tuple heap maintenance cost */ + run_cost += tuples * comparison_cost * logN; + + /* + * Although MergeAppend does not do any selection or projection, it's not + * free; add a small per-tuple overhead. + */ + run_cost += cpu_tuple_cost * APPEND_CPU_COST_MULTIPLIER * tuples; + + path->startup_cost = startup_cost + input_startup_cost; + path->total_cost = startup_cost + run_cost + input_total_cost; +} + +/* + * cost_material + * Determines and returns the cost of materializing a relation, including + * the cost of reading the input data. + * + * If the total volume of data to materialize exceeds work_mem, we will need + * to write it to disk, so the cost is much higher in that case. + * + * Note that here we are estimating the costs for the first scan of the + * relation, so the materialization is all overhead --- any savings will + * occur only on rescan, which is estimated in cost_rescan. + */ +void +cost_material(Path *path, + Cost input_startup_cost, Cost input_total_cost, + double tuples, int width) +{ + Cost startup_cost = input_startup_cost; + Cost run_cost = input_total_cost - input_startup_cost; + double nbytes = relation_byte_size(tuples, width); + long work_mem_bytes = work_mem * 1024L; + + path->rows = tuples; + + /* + * Whether spilling or not, charge 2x cpu_operator_cost per tuple to + * reflect bookkeeping overhead. (This rate must be more than what + * cost_rescan charges for materialize, ie, cpu_operator_cost per tuple; + * if it is exactly the same then there will be a cost tie between + * nestloop with A outer, materialized B inner and nestloop with B outer, + * materialized A inner. The extra cost ensures we'll prefer + * materializing the smaller rel.) Note that this is normally a good deal + * less than cpu_tuple_cost; which is OK because a Material plan node + * doesn't do qual-checking or projection, so it's got less overhead than + * most plan nodes. + */ + run_cost += 2 * cpu_operator_cost * tuples; + + /* + * If we will spill to disk, charge at the rate of seq_page_cost per page. + * This cost is assumed to be evenly spread through the plan run phase, + * which isn't exactly accurate but our cost model doesn't allow for + * nonuniform costs within the run phase. + */ + if (nbytes > work_mem_bytes) + { + double npages = ceil(nbytes / BLCKSZ); + + run_cost += seq_page_cost * npages; + } + + path->startup_cost = startup_cost; + path->total_cost = startup_cost + run_cost; +} + +/* + * cost_memoize_rescan + * Determines the estimated cost of rescanning a Memoize node. + * + * In order to estimate this, we must gain knowledge of how often we expect to + * be called and how many distinct sets of parameters we are likely to be + * called with. If we expect a good cache hit ratio, then we can set our + * costs to account for that hit ratio, plus a little bit of cost for the + * caching itself. Caching will not work out well if we expect to be called + * with too many distinct parameter values. The worst-case here is that we + * never see any parameter value twice, in which case we'd never get a cache + * hit and caching would be a complete waste of effort. + */ +static void +cost_memoize_rescan(PlannerInfo *root, MemoizePath *mpath, + Cost *rescan_startup_cost, Cost *rescan_total_cost) +{ + EstimationInfo estinfo; + Cost input_startup_cost = mpath->subpath->startup_cost; + Cost input_total_cost = mpath->subpath->total_cost; + double tuples = mpath->subpath->rows; + double calls = mpath->calls; + int width = mpath->subpath->pathtarget->width; + + double hash_mem_bytes; + double est_entry_bytes; + double est_cache_entries; + double ndistinct; + double evict_ratio; + double hit_ratio; + Cost startup_cost; + Cost total_cost; + + /* available cache space */ + hash_mem_bytes = get_hash_memory_limit(); + + /* + * Set the number of bytes each cache entry should consume in the cache. + * To provide us with better estimations on how many cache entries we can + * store at once, we make a call to the executor here to ask it what + * memory overheads there are for a single cache entry. + * + * XXX we also store the cache key, but that's not accounted for here. + */ + est_entry_bytes = relation_byte_size(tuples, width) + + ExecEstimateCacheEntryOverheadBytes(tuples); + + /* estimate on the upper limit of cache entries we can hold at once */ + est_cache_entries = floor(hash_mem_bytes / est_entry_bytes); + + /* estimate on the distinct number of parameter values */ + ndistinct = estimate_num_groups(root, mpath->param_exprs, calls, NULL, + &estinfo); + + /* + * When the estimation fell back on using a default value, it's a bit too + * risky to assume that it's ok to use a Memoize node. The use of a + * default could cause us to use a Memoize node when it's really + * inappropriate to do so. If we see that this has been done, then we'll + * assume that every call will have unique parameters, which will almost + * certainly mean a MemoizePath will never survive add_path(). + */ + if ((estinfo.flags & SELFLAG_USED_DEFAULT) != 0) + ndistinct = calls; + + /* + * Since we've already estimated the maximum number of entries we can + * store at once and know the estimated number of distinct values we'll be + * called with, we'll take this opportunity to set the path's est_entries. + * This will ultimately determine the hash table size that the executor + * will use. If we leave this at zero, the executor will just choose the + * size itself. Really this is not the right place to do this, but it's + * convenient since everything is already calculated. + */ + mpath->est_entries = Min(Min(ndistinct, est_cache_entries), + PG_UINT32_MAX); + + /* + * When the number of distinct parameter values is above the amount we can + * store in the cache, then we'll have to evict some entries from the + * cache. This is not free. Here we estimate how often we'll incur the + * cost of that eviction. + */ + evict_ratio = 1.0 - Min(est_cache_entries, ndistinct) / ndistinct; + + /* + * In order to estimate how costly a single scan will be, we need to + * attempt to estimate what the cache hit ratio will be. To do that we + * must look at how many scans are estimated in total for this node and + * how many of those scans we expect to get a cache hit. + */ + hit_ratio = 1.0 / ndistinct * Min(est_cache_entries, ndistinct) - + (ndistinct / calls); + + /* Ensure we don't go negative */ + hit_ratio = Max(hit_ratio, 0.0); + + /* + * Set the total_cost accounting for the expected cache hit ratio. We + * also add on a cpu_operator_cost to account for a cache lookup. This + * will happen regardless of whether it's a cache hit or not. + */ + total_cost = input_total_cost * (1.0 - hit_ratio) + cpu_operator_cost; + + /* Now adjust the total cost to account for cache evictions */ + + /* Charge a cpu_tuple_cost for evicting the actual cache entry */ + total_cost += cpu_tuple_cost * evict_ratio; + + /* + * Charge a 10th of cpu_operator_cost to evict every tuple in that entry. + * The per-tuple eviction is really just a pfree, so charging a whole + * cpu_operator_cost seems a little excessive. + */ + total_cost += cpu_operator_cost / 10.0 * evict_ratio * tuples; + + /* + * Now adjust for storing things in the cache, since that's not free + * either. Everything must go in the cache. We don't proportion this + * over any ratio, just apply it once for the scan. We charge a + * cpu_tuple_cost for the creation of the cache entry and also a + * cpu_operator_cost for each tuple we expect to cache. + */ + total_cost += cpu_tuple_cost + cpu_operator_cost * tuples; + + /* + * Getting the first row must be also be proportioned according to the + * expected cache hit ratio. + */ + startup_cost = input_startup_cost * (1.0 - hit_ratio); + + /* + * Additionally we charge a cpu_tuple_cost to account for cache lookups, + * which we'll do regardless of whether it was a cache hit or not. + */ + startup_cost += cpu_tuple_cost; + + *rescan_startup_cost = startup_cost; + *rescan_total_cost = total_cost; +} + +/* + * cost_agg + * Determines and returns the cost of performing an Agg plan node, + * including the cost of its input. + * + * aggcosts can be NULL when there are no actual aggregate functions (i.e., + * we are using a hashed Agg node just to do grouping). + * + * Note: when aggstrategy == AGG_SORTED, caller must ensure that input costs + * are for appropriately-sorted input. + */ +void +cost_agg(Path *path, PlannerInfo *root, + AggStrategy aggstrategy, const AggClauseCosts *aggcosts, + int numGroupCols, double numGroups, + List *quals, + Cost input_startup_cost, Cost input_total_cost, + double input_tuples, double input_width) +{ + double output_tuples; + Cost startup_cost; + Cost total_cost; + AggClauseCosts dummy_aggcosts; + + /* Use all-zero per-aggregate costs if NULL is passed */ + if (aggcosts == NULL) + { + Assert(aggstrategy == AGG_HASHED); + MemSet(&dummy_aggcosts, 0, sizeof(AggClauseCosts)); + aggcosts = &dummy_aggcosts; + } + + /* + * The transCost.per_tuple component of aggcosts should be charged once + * per input tuple, corresponding to the costs of evaluating the aggregate + * transfns and their input expressions. The finalCost.per_tuple component + * is charged once per output tuple, corresponding to the costs of + * evaluating the finalfns. Startup costs are of course charged but once. + * + * If we are grouping, we charge an additional cpu_operator_cost per + * grouping column per input tuple for grouping comparisons. + * + * We will produce a single output tuple if not grouping, and a tuple per + * group otherwise. We charge cpu_tuple_cost for each output tuple. + * + * Note: in this cost model, AGG_SORTED and AGG_HASHED have exactly the + * same total CPU cost, but AGG_SORTED has lower startup cost. If the + * input path is already sorted appropriately, AGG_SORTED should be + * preferred (since it has no risk of memory overflow). This will happen + * as long as the computed total costs are indeed exactly equal --- but if + * there's roundoff error we might do the wrong thing. So be sure that + * the computations below form the same intermediate values in the same + * order. + */ + if (aggstrategy == AGG_PLAIN) + { + startup_cost = input_total_cost; + startup_cost += aggcosts->transCost.startup; + startup_cost += aggcosts->transCost.per_tuple * input_tuples; + startup_cost += aggcosts->finalCost.startup; + startup_cost += aggcosts->finalCost.per_tuple; + /* we aren't grouping */ + total_cost = startup_cost + cpu_tuple_cost; + output_tuples = 1; + } + else if (aggstrategy == AGG_SORTED || aggstrategy == AGG_MIXED) + { + /* Here we are able to deliver output on-the-fly */ + startup_cost = input_startup_cost; + total_cost = input_total_cost; + if (aggstrategy == AGG_MIXED && !enable_hashagg) + { + startup_cost += disable_cost; + total_cost += disable_cost; + } + /* calcs phrased this way to match HASHED case, see note above */ + total_cost += aggcosts->transCost.startup; + total_cost += aggcosts->transCost.per_tuple * input_tuples; + total_cost += (cpu_operator_cost * numGroupCols) * input_tuples; + total_cost += aggcosts->finalCost.startup; + total_cost += aggcosts->finalCost.per_tuple * numGroups; + total_cost += cpu_tuple_cost * numGroups; + output_tuples = numGroups; + } + else + { + /* must be AGG_HASHED */ + startup_cost = input_total_cost; + if (!enable_hashagg) + startup_cost += disable_cost; + startup_cost += aggcosts->transCost.startup; + startup_cost += aggcosts->transCost.per_tuple * input_tuples; + /* cost of computing hash value */ + startup_cost += (cpu_operator_cost * numGroupCols) * input_tuples; + startup_cost += aggcosts->finalCost.startup; + + total_cost = startup_cost; + total_cost += aggcosts->finalCost.per_tuple * numGroups; + /* cost of retrieving from hash table */ + total_cost += cpu_tuple_cost * numGroups; + output_tuples = numGroups; + } + + /* + * Add the disk costs of hash aggregation that spills to disk. + * + * Groups that go into the hash table stay in memory until finalized, so + * spilling and reprocessing tuples doesn't incur additional invocations + * of transCost or finalCost. Furthermore, the computed hash value is + * stored with the spilled tuples, so we don't incur extra invocations of + * the hash function. + * + * Hash Agg begins returning tuples after the first batch is complete. + * Accrue writes (spilled tuples) to startup_cost and to total_cost; + * accrue reads only to total_cost. + */ + if (aggstrategy == AGG_HASHED || aggstrategy == AGG_MIXED) + { + double pages; + double pages_written = 0.0; + double pages_read = 0.0; + double spill_cost; + double hashentrysize; + double nbatches; + Size mem_limit; + uint64 ngroups_limit; + int num_partitions; + int depth; + + /* + * Estimate number of batches based on the computed limits. If less + * than or equal to one, all groups are expected to fit in memory; + * otherwise we expect to spill. + */ + hashentrysize = hash_agg_entry_size(list_length(root->aggtransinfos), + input_width, + aggcosts->transitionSpace); + hash_agg_set_limits(hashentrysize, numGroups, 0, &mem_limit, + &ngroups_limit, &num_partitions); + + nbatches = Max((numGroups * hashentrysize) / mem_limit, + numGroups / ngroups_limit); + + nbatches = Max(ceil(nbatches), 1.0); + num_partitions = Max(num_partitions, 2); + + /* + * The number of partitions can change at different levels of + * recursion; but for the purposes of this calculation assume it stays + * constant. + */ + depth = ceil(log(nbatches) / log(num_partitions)); + + /* + * Estimate number of pages read and written. For each level of + * recursion, a tuple must be written and then later read. + */ + pages = relation_byte_size(input_tuples, input_width) / BLCKSZ; + pages_written = pages_read = pages * depth; + + /* + * HashAgg has somewhat worse IO behavior than Sort on typical + * hardware/OS combinations. Account for this with a generic penalty. + */ + pages_read *= 2.0; + pages_written *= 2.0; + + startup_cost += pages_written * random_page_cost; + total_cost += pages_written * random_page_cost; + total_cost += pages_read * seq_page_cost; + + /* account for CPU cost of spilling a tuple and reading it back */ + spill_cost = depth * input_tuples * 2.0 * cpu_tuple_cost; + startup_cost += spill_cost; + total_cost += spill_cost; + } + + /* + * If there are quals (HAVING quals), account for their cost and + * selectivity. + */ + if (quals) + { + QualCost qual_cost; + + cost_qual_eval(&qual_cost, quals, root); + startup_cost += qual_cost.startup; + total_cost += qual_cost.startup + output_tuples * qual_cost.per_tuple; + + output_tuples = clamp_row_est(output_tuples * + clauselist_selectivity(root, + quals, + 0, + JOIN_INNER, + NULL)); + } + + path->rows = output_tuples; + path->startup_cost = startup_cost; + path->total_cost = total_cost; +} + +/* + * cost_windowagg + * Determines and returns the cost of performing a WindowAgg plan node, + * including the cost of its input. + * + * Input is assumed already properly sorted. + */ +void +cost_windowagg(Path *path, PlannerInfo *root, + List *windowFuncs, int numPartCols, int numOrderCols, + Cost input_startup_cost, Cost input_total_cost, + double input_tuples) +{ + Cost startup_cost; + Cost total_cost; + ListCell *lc; + + startup_cost = input_startup_cost; + total_cost = input_total_cost; + + /* + * Window functions are assumed to cost their stated execution cost, plus + * the cost of evaluating their input expressions, per tuple. Since they + * may in fact evaluate their inputs at multiple rows during each cycle, + * this could be a drastic underestimate; but without a way to know how + * many rows the window function will fetch, it's hard to do better. In + * any case, it's a good estimate for all the built-in window functions, + * so we'll just do this for now. + */ + foreach(lc, windowFuncs) + { + WindowFunc *wfunc = lfirst_node(WindowFunc, lc); + Cost wfunccost; + QualCost argcosts; + + argcosts.startup = argcosts.per_tuple = 0; + add_function_cost(root, wfunc->winfnoid, (Node *) wfunc, + &argcosts); + startup_cost += argcosts.startup; + wfunccost = argcosts.per_tuple; + + /* also add the input expressions' cost to per-input-row costs */ + cost_qual_eval_node(&argcosts, (Node *) wfunc->args, root); + startup_cost += argcosts.startup; + wfunccost += argcosts.per_tuple; + + /* + * Add the filter's cost to per-input-row costs. XXX We should reduce + * input expression costs according to filter selectivity. + */ + cost_qual_eval_node(&argcosts, (Node *) wfunc->aggfilter, root); + startup_cost += argcosts.startup; + wfunccost += argcosts.per_tuple; + + total_cost += wfunccost * input_tuples; + } + + /* + * We also charge cpu_operator_cost per grouping column per tuple for + * grouping comparisons, plus cpu_tuple_cost per tuple for general + * overhead. + * + * XXX this neglects costs of spooling the data to disk when it overflows + * work_mem. Sooner or later that should get accounted for. + */ + total_cost += cpu_operator_cost * (numPartCols + numOrderCols) * input_tuples; + total_cost += cpu_tuple_cost * input_tuples; + + path->rows = input_tuples; + path->startup_cost = startup_cost; + path->total_cost = total_cost; +} + +/* + * cost_group + * Determines and returns the cost of performing a Group plan node, + * including the cost of its input. + * + * Note: caller must ensure that input costs are for appropriately-sorted + * input. + */ +void +cost_group(Path *path, PlannerInfo *root, + int numGroupCols, double numGroups, + List *quals, + Cost input_startup_cost, Cost input_total_cost, + double input_tuples) +{ + double output_tuples; + Cost startup_cost; + Cost total_cost; + + output_tuples = numGroups; + startup_cost = input_startup_cost; + total_cost = input_total_cost; + + /* + * Charge one cpu_operator_cost per comparison per input tuple. We assume + * all columns get compared at most of the tuples. + */ + total_cost += cpu_operator_cost * input_tuples * numGroupCols; + + /* + * If there are quals (HAVING quals), account for their cost and + * selectivity. + */ + if (quals) + { + QualCost qual_cost; + + cost_qual_eval(&qual_cost, quals, root); + startup_cost += qual_cost.startup; + total_cost += qual_cost.startup + output_tuples * qual_cost.per_tuple; + + output_tuples = clamp_row_est(output_tuples * + clauselist_selectivity(root, + quals, + 0, + JOIN_INNER, + NULL)); + } + + path->rows = output_tuples; + path->startup_cost = startup_cost; + path->total_cost = total_cost; +} + +/* + * initial_cost_nestloop + * Preliminary estimate of the cost of a nestloop join path. + * + * This must quickly produce lower-bound estimates of the path's startup and + * total costs. If we are unable to eliminate the proposed path from + * consideration using the lower bounds, final_cost_nestloop will be called + * to obtain the final estimates. + * + * The exact division of labor between this function and final_cost_nestloop + * is private to them, and represents a tradeoff between speed of the initial + * estimate and getting a tight lower bound. We choose to not examine the + * join quals here, since that's by far the most expensive part of the + * calculations. The end result is that CPU-cost considerations must be + * left for the second phase; and for SEMI/ANTI joins, we must also postpone + * incorporation of the inner path's run cost. + * + * 'workspace' is to be filled with startup_cost, total_cost, and perhaps + * other data to be used by final_cost_nestloop + * 'jointype' is the type of join to be performed + * 'outer_path' is the outer input to the join + * 'inner_path' is the inner input to the join + * 'extra' contains miscellaneous information about the join + */ +void +initial_cost_nestloop(PlannerInfo *root, JoinCostWorkspace *workspace, + JoinType jointype, + Path *outer_path, Path *inner_path, + JoinPathExtraData *extra) +{ + Cost startup_cost = 0; + Cost run_cost = 0; + double outer_path_rows = outer_path->rows; + Cost inner_rescan_start_cost; + Cost inner_rescan_total_cost; + Cost inner_run_cost; + Cost inner_rescan_run_cost; + + /* estimate costs to rescan the inner relation */ + cost_rescan(root, inner_path, + &inner_rescan_start_cost, + &inner_rescan_total_cost); + + /* cost of source data */ + + /* + * NOTE: clearly, we must pay both outer and inner paths' startup_cost + * before we can start returning tuples, so the join's startup cost is + * their sum. We'll also pay the inner path's rescan startup cost + * multiple times. + */ + startup_cost += outer_path->startup_cost + inner_path->startup_cost; + run_cost += outer_path->total_cost - outer_path->startup_cost; + if (outer_path_rows > 1) + run_cost += (outer_path_rows - 1) * inner_rescan_start_cost; + + inner_run_cost = inner_path->total_cost - inner_path->startup_cost; + inner_rescan_run_cost = inner_rescan_total_cost - inner_rescan_start_cost; + + if (jointype == JOIN_SEMI || jointype == JOIN_ANTI || + extra->inner_unique) + { + /* + * With a SEMI or ANTI join, or if the innerrel is known unique, the + * executor will stop after the first match. + * + * Getting decent estimates requires inspection of the join quals, + * which we choose to postpone to final_cost_nestloop. + */ + + /* Save private data for final_cost_nestloop */ + workspace->inner_run_cost = inner_run_cost; + workspace->inner_rescan_run_cost = inner_rescan_run_cost; + } + else + { + /* Normal case; we'll scan whole input rel for each outer row */ + run_cost += inner_run_cost; + if (outer_path_rows > 1) + run_cost += (outer_path_rows - 1) * inner_rescan_run_cost; + } + + /* CPU costs left for later */ + + /* Public result fields */ + workspace->startup_cost = startup_cost; + workspace->total_cost = startup_cost + run_cost; + /* Save private data for final_cost_nestloop */ + workspace->run_cost = run_cost; +} + +/* + * final_cost_nestloop + * Final estimate of the cost and result size of a nestloop join path. + * + * 'path' is already filled in except for the rows and cost fields + * 'workspace' is the result from initial_cost_nestloop + * 'extra' contains miscellaneous information about the join + */ +void +final_cost_nestloop(PlannerInfo *root, NestPath *path, + JoinCostWorkspace *workspace, + JoinPathExtraData *extra) +{ + Path *outer_path = path->jpath.outerjoinpath; + Path *inner_path = path->jpath.innerjoinpath; + double outer_path_rows = outer_path->rows; + double inner_path_rows = inner_path->rows; + Cost startup_cost = workspace->startup_cost; + Cost run_cost = workspace->run_cost; + Cost cpu_per_tuple; + QualCost restrict_qual_cost; + double ntuples; + + /* Protect some assumptions below that rowcounts aren't zero */ + if (outer_path_rows <= 0) + outer_path_rows = 1; + if (inner_path_rows <= 0) + inner_path_rows = 1; + /* Mark the path with the correct row estimate */ + if (path->jpath.path.param_info) + path->jpath.path.rows = path->jpath.path.param_info->ppi_rows; + else + path->jpath.path.rows = path->jpath.path.parent->rows; + + /* For partial paths, scale row estimate. */ + if (path->jpath.path.parallel_workers > 0) + { + double parallel_divisor = get_parallel_divisor(&path->jpath.path); + + path->jpath.path.rows = + clamp_row_est(path->jpath.path.rows / parallel_divisor); + } + + /* + * We could include disable_cost in the preliminary estimate, but that + * would amount to optimizing for the case where the join method is + * disabled, which doesn't seem like the way to bet. + */ + if (!enable_nestloop) + startup_cost += disable_cost; + + /* cost of inner-relation source data (we already dealt with outer rel) */ + + if (path->jpath.jointype == JOIN_SEMI || path->jpath.jointype == JOIN_ANTI || + extra->inner_unique) + { + /* + * With a SEMI or ANTI join, or if the innerrel is known unique, the + * executor will stop after the first match. + */ + Cost inner_run_cost = workspace->inner_run_cost; + Cost inner_rescan_run_cost = workspace->inner_rescan_run_cost; + double outer_matched_rows; + double outer_unmatched_rows; + Selectivity inner_scan_frac; + + /* + * For an outer-rel row that has at least one match, we can expect the + * inner scan to stop after a fraction 1/(match_count+1) of the inner + * rows, if the matches are evenly distributed. Since they probably + * aren't quite evenly distributed, we apply a fuzz factor of 2.0 to + * that fraction. (If we used a larger fuzz factor, we'd have to + * clamp inner_scan_frac to at most 1.0; but since match_count is at + * least 1, no such clamp is needed now.) + */ + outer_matched_rows = rint(outer_path_rows * extra->semifactors.outer_match_frac); + outer_unmatched_rows = outer_path_rows - outer_matched_rows; + inner_scan_frac = 2.0 / (extra->semifactors.match_count + 1.0); + + /* + * Compute number of tuples processed (not number emitted!). First, + * account for successfully-matched outer rows. + */ + ntuples = outer_matched_rows * inner_path_rows * inner_scan_frac; + + /* + * Now we need to estimate the actual costs of scanning the inner + * relation, which may be quite a bit less than N times inner_run_cost + * due to early scan stops. We consider two cases. If the inner path + * is an indexscan using all the joinquals as indexquals, then an + * unmatched outer row results in an indexscan returning no rows, + * which is probably quite cheap. Otherwise, the executor will have + * to scan the whole inner rel for an unmatched row; not so cheap. + */ + if (has_indexed_join_quals(path)) + { + /* + * Successfully-matched outer rows will only require scanning + * inner_scan_frac of the inner relation. In this case, we don't + * need to charge the full inner_run_cost even when that's more + * than inner_rescan_run_cost, because we can assume that none of + * the inner scans ever scan the whole inner relation. So it's + * okay to assume that all the inner scan executions can be + * fractions of the full cost, even if materialization is reducing + * the rescan cost. At this writing, it's impossible to get here + * for a materialized inner scan, so inner_run_cost and + * inner_rescan_run_cost will be the same anyway; but just in + * case, use inner_run_cost for the first matched tuple and + * inner_rescan_run_cost for additional ones. + */ + run_cost += inner_run_cost * inner_scan_frac; + if (outer_matched_rows > 1) + run_cost += (outer_matched_rows - 1) * inner_rescan_run_cost * inner_scan_frac; + + /* + * Add the cost of inner-scan executions for unmatched outer rows. + * We estimate this as the same cost as returning the first tuple + * of a nonempty scan. We consider that these are all rescans, + * since we used inner_run_cost once already. + */ + run_cost += outer_unmatched_rows * + inner_rescan_run_cost / inner_path_rows; + + /* + * We won't be evaluating any quals at all for unmatched rows, so + * don't add them to ntuples. + */ + } + else + { + /* + * Here, a complicating factor is that rescans may be cheaper than + * first scans. If we never scan all the way to the end of the + * inner rel, it might be (depending on the plan type) that we'd + * never pay the whole inner first-scan run cost. However it is + * difficult to estimate whether that will happen (and it could + * not happen if there are any unmatched outer rows!), so be + * conservative and always charge the whole first-scan cost once. + * We consider this charge to correspond to the first unmatched + * outer row, unless there isn't one in our estimate, in which + * case blame it on the first matched row. + */ + + /* First, count all unmatched join tuples as being processed */ + ntuples += outer_unmatched_rows * inner_path_rows; + + /* Now add the forced full scan, and decrement appropriate count */ + run_cost += inner_run_cost; + if (outer_unmatched_rows >= 1) + outer_unmatched_rows -= 1; + else + outer_matched_rows -= 1; + + /* Add inner run cost for additional outer tuples having matches */ + if (outer_matched_rows > 0) + run_cost += outer_matched_rows * inner_rescan_run_cost * inner_scan_frac; + + /* Add inner run cost for additional unmatched outer tuples */ + if (outer_unmatched_rows > 0) + run_cost += outer_unmatched_rows * inner_rescan_run_cost; + } + } + else + { + /* Normal-case source costs were included in preliminary estimate */ + + /* Compute number of tuples processed (not number emitted!) */ + ntuples = outer_path_rows * inner_path_rows; + } + + /* CPU costs */ + cost_qual_eval(&restrict_qual_cost, path->jpath.joinrestrictinfo, root); + startup_cost += restrict_qual_cost.startup; + cpu_per_tuple = cpu_tuple_cost + restrict_qual_cost.per_tuple; + run_cost += cpu_per_tuple * ntuples; + + /* tlist eval costs are paid per output row, not per tuple scanned */ + startup_cost += path->jpath.path.pathtarget->cost.startup; + run_cost += path->jpath.path.pathtarget->cost.per_tuple * path->jpath.path.rows; + + path->jpath.path.startup_cost = startup_cost; + path->jpath.path.total_cost = startup_cost + run_cost; +} + +/* + * initial_cost_mergejoin + * Preliminary estimate of the cost of a mergejoin path. + * + * This must quickly produce lower-bound estimates of the path's startup and + * total costs. If we are unable to eliminate the proposed path from + * consideration using the lower bounds, final_cost_mergejoin will be called + * to obtain the final estimates. + * + * The exact division of labor between this function and final_cost_mergejoin + * is private to them, and represents a tradeoff between speed of the initial + * estimate and getting a tight lower bound. We choose to not examine the + * join quals here, except for obtaining the scan selectivity estimate which + * is really essential (but fortunately, use of caching keeps the cost of + * getting that down to something reasonable). + * We also assume that cost_sort is cheap enough to use here. + * + * 'workspace' is to be filled with startup_cost, total_cost, and perhaps + * other data to be used by final_cost_mergejoin + * 'jointype' is the type of join to be performed + * 'mergeclauses' is the list of joinclauses to be used as merge clauses + * 'outer_path' is the outer input to the join + * 'inner_path' is the inner input to the join + * 'outersortkeys' is the list of sort keys for the outer path + * 'innersortkeys' is the list of sort keys for the inner path + * 'extra' contains miscellaneous information about the join + * + * Note: outersortkeys and innersortkeys should be NIL if no explicit + * sort is needed because the respective source path is already ordered. + */ +void +initial_cost_mergejoin(PlannerInfo *root, JoinCostWorkspace *workspace, + JoinType jointype, + List *mergeclauses, + Path *outer_path, Path *inner_path, + List *outersortkeys, List *innersortkeys, + JoinPathExtraData *extra) +{ + Cost startup_cost = 0; + Cost run_cost = 0; + double outer_path_rows = outer_path->rows; + double inner_path_rows = inner_path->rows; + Cost inner_run_cost; + double outer_rows, + inner_rows, + outer_skip_rows, + inner_skip_rows; + Selectivity outerstartsel, + outerendsel, + innerstartsel, + innerendsel; + Path sort_path; /* dummy for result of cost_sort */ + + /* Protect some assumptions below that rowcounts aren't zero */ + if (outer_path_rows <= 0) + outer_path_rows = 1; + if (inner_path_rows <= 0) + inner_path_rows = 1; + + /* + * A merge join will stop as soon as it exhausts either input stream + * (unless it's an outer join, in which case the outer side has to be + * scanned all the way anyway). Estimate fraction of the left and right + * inputs that will actually need to be scanned. Likewise, we can + * estimate the number of rows that will be skipped before the first join + * pair is found, which should be factored into startup cost. We use only + * the first (most significant) merge clause for this purpose. Since + * mergejoinscansel() is a fairly expensive computation, we cache the + * results in the merge clause RestrictInfo. + */ + if (mergeclauses && jointype != JOIN_FULL) + { + RestrictInfo *firstclause = (RestrictInfo *) linitial(mergeclauses); + List *opathkeys; + List *ipathkeys; + PathKey *opathkey; + PathKey *ipathkey; + MergeScanSelCache *cache; + + /* Get the input pathkeys to determine the sort-order details */ + opathkeys = outersortkeys ? outersortkeys : outer_path->pathkeys; + ipathkeys = innersortkeys ? innersortkeys : inner_path->pathkeys; + Assert(opathkeys); + Assert(ipathkeys); + opathkey = (PathKey *) linitial(opathkeys); + ipathkey = (PathKey *) linitial(ipathkeys); + /* debugging check */ + if (opathkey->pk_opfamily != ipathkey->pk_opfamily || + opathkey->pk_eclass->ec_collation != ipathkey->pk_eclass->ec_collation || + opathkey->pk_strategy != ipathkey->pk_strategy || + opathkey->pk_nulls_first != ipathkey->pk_nulls_first) + elog(ERROR, "left and right pathkeys do not match in mergejoin"); + + /* Get the selectivity with caching */ + cache = cached_scansel(root, firstclause, opathkey); + + if (bms_is_subset(firstclause->left_relids, + outer_path->parent->relids)) + { + /* left side of clause is outer */ + outerstartsel = cache->leftstartsel; + outerendsel = cache->leftendsel; + innerstartsel = cache->rightstartsel; + innerendsel = cache->rightendsel; + } + else + { + /* left side of clause is inner */ + outerstartsel = cache->rightstartsel; + outerendsel = cache->rightendsel; + innerstartsel = cache->leftstartsel; + innerendsel = cache->leftendsel; + } + if (jointype == JOIN_LEFT || + jointype == JOIN_ANTI) + { + outerstartsel = 0.0; + outerendsel = 1.0; + } + else if (jointype == JOIN_RIGHT) + { + innerstartsel = 0.0; + innerendsel = 1.0; + } + } + else + { + /* cope with clauseless or full mergejoin */ + outerstartsel = innerstartsel = 0.0; + outerendsel = innerendsel = 1.0; + } + + /* + * Convert selectivities to row counts. We force outer_rows and + * inner_rows to be at least 1, but the skip_rows estimates can be zero. + */ + outer_skip_rows = rint(outer_path_rows * outerstartsel); + inner_skip_rows = rint(inner_path_rows * innerstartsel); + outer_rows = clamp_row_est(outer_path_rows * outerendsel); + inner_rows = clamp_row_est(inner_path_rows * innerendsel); + + Assert(outer_skip_rows <= outer_rows); + Assert(inner_skip_rows <= inner_rows); + + /* + * Readjust scan selectivities to account for above rounding. This is + * normally an insignificant effect, but when there are only a few rows in + * the inputs, failing to do this makes for a large percentage error. + */ + outerstartsel = outer_skip_rows / outer_path_rows; + innerstartsel = inner_skip_rows / inner_path_rows; + outerendsel = outer_rows / outer_path_rows; + innerendsel = inner_rows / inner_path_rows; + + Assert(outerstartsel <= outerendsel); + Assert(innerstartsel <= innerendsel); + + /* cost of source data */ + + if (outersortkeys) /* do we need to sort outer? */ + { + cost_sort(&sort_path, + root, + outersortkeys, + outer_path->total_cost, + outer_path_rows, + outer_path->pathtarget->width, + 0.0, + work_mem, + -1.0); + startup_cost += sort_path.startup_cost; + startup_cost += (sort_path.total_cost - sort_path.startup_cost) + * outerstartsel; + run_cost += (sort_path.total_cost - sort_path.startup_cost) + * (outerendsel - outerstartsel); + } + else + { + startup_cost += outer_path->startup_cost; + startup_cost += (outer_path->total_cost - outer_path->startup_cost) + * outerstartsel; + run_cost += (outer_path->total_cost - outer_path->startup_cost) + * (outerendsel - outerstartsel); + } + + if (innersortkeys) /* do we need to sort inner? */ + { + cost_sort(&sort_path, + root, + innersortkeys, + inner_path->total_cost, + inner_path_rows, + inner_path->pathtarget->width, + 0.0, + work_mem, + -1.0); + startup_cost += sort_path.startup_cost; + startup_cost += (sort_path.total_cost - sort_path.startup_cost) + * innerstartsel; + inner_run_cost = (sort_path.total_cost - sort_path.startup_cost) + * (innerendsel - innerstartsel); + } + else + { + startup_cost += inner_path->startup_cost; + startup_cost += (inner_path->total_cost - inner_path->startup_cost) + * innerstartsel; + inner_run_cost = (inner_path->total_cost - inner_path->startup_cost) + * (innerendsel - innerstartsel); + } + + /* + * We can't yet determine whether rescanning occurs, or whether + * materialization of the inner input should be done. The minimum + * possible inner input cost, regardless of rescan and materialization + * considerations, is inner_run_cost. We include that in + * workspace->total_cost, but not yet in run_cost. + */ + + /* CPU costs left for later */ + + /* Public result fields */ + workspace->startup_cost = startup_cost; + workspace->total_cost = startup_cost + run_cost + inner_run_cost; + /* Save private data for final_cost_mergejoin */ + workspace->run_cost = run_cost; + workspace->inner_run_cost = inner_run_cost; + workspace->outer_rows = outer_rows; + workspace->inner_rows = inner_rows; + workspace->outer_skip_rows = outer_skip_rows; + workspace->inner_skip_rows = inner_skip_rows; +} + +/* + * final_cost_mergejoin + * Final estimate of the cost and result size of a mergejoin path. + * + * Unlike other costsize functions, this routine makes two actual decisions: + * whether the executor will need to do mark/restore, and whether we should + * materialize the inner path. It would be logically cleaner to build + * separate paths testing these alternatives, but that would require repeating + * most of the cost calculations, which are not all that cheap. Since the + * choice will not affect output pathkeys or startup cost, only total cost, + * there is no possibility of wanting to keep more than one path. So it seems + * best to make the decisions here and record them in the path's + * skip_mark_restore and materialize_inner fields. + * + * Mark/restore overhead is usually required, but can be skipped if we know + * that the executor need find only one match per outer tuple, and that the + * mergeclauses are sufficient to identify a match. + * + * We materialize the inner path if we need mark/restore and either the inner + * path can't support mark/restore, or it's cheaper to use an interposed + * Material node to handle mark/restore. + * + * 'path' is already filled in except for the rows and cost fields and + * skip_mark_restore and materialize_inner + * 'workspace' is the result from initial_cost_mergejoin + * 'extra' contains miscellaneous information about the join + */ +void +final_cost_mergejoin(PlannerInfo *root, MergePath *path, + JoinCostWorkspace *workspace, + JoinPathExtraData *extra) +{ + Path *outer_path = path->jpath.outerjoinpath; + Path *inner_path = path->jpath.innerjoinpath; + double inner_path_rows = inner_path->rows; + List *mergeclauses = path->path_mergeclauses; + List *innersortkeys = path->innersortkeys; + Cost startup_cost = workspace->startup_cost; + Cost run_cost = workspace->run_cost; + Cost inner_run_cost = workspace->inner_run_cost; + double outer_rows = workspace->outer_rows; + double inner_rows = workspace->inner_rows; + double outer_skip_rows = workspace->outer_skip_rows; + double inner_skip_rows = workspace->inner_skip_rows; + Cost cpu_per_tuple, + bare_inner_cost, + mat_inner_cost; + QualCost merge_qual_cost; + QualCost qp_qual_cost; + double mergejointuples, + rescannedtuples; + double rescanratio; + + /* Protect some assumptions below that rowcounts aren't zero */ + if (inner_path_rows <= 0) + inner_path_rows = 1; + + /* Mark the path with the correct row estimate */ + if (path->jpath.path.param_info) + path->jpath.path.rows = path->jpath.path.param_info->ppi_rows; + else + path->jpath.path.rows = path->jpath.path.parent->rows; + + /* For partial paths, scale row estimate. */ + if (path->jpath.path.parallel_workers > 0) + { + double parallel_divisor = get_parallel_divisor(&path->jpath.path); + + path->jpath.path.rows = + clamp_row_est(path->jpath.path.rows / parallel_divisor); + } + + /* + * We could include disable_cost in the preliminary estimate, but that + * would amount to optimizing for the case where the join method is + * disabled, which doesn't seem like the way to bet. + */ + if (!enable_mergejoin) + startup_cost += disable_cost; + + /* + * Compute cost of the mergequals and qpquals (other restriction clauses) + * separately. + */ + cost_qual_eval(&merge_qual_cost, mergeclauses, root); + cost_qual_eval(&qp_qual_cost, path->jpath.joinrestrictinfo, root); + qp_qual_cost.startup -= merge_qual_cost.startup; + qp_qual_cost.per_tuple -= merge_qual_cost.per_tuple; + + /* + * With a SEMI or ANTI join, or if the innerrel is known unique, the + * executor will stop scanning for matches after the first match. When + * all the joinclauses are merge clauses, this means we don't ever need to + * back up the merge, and so we can skip mark/restore overhead. + */ + if ((path->jpath.jointype == JOIN_SEMI || + path->jpath.jointype == JOIN_ANTI || + extra->inner_unique) && + (list_length(path->jpath.joinrestrictinfo) == + list_length(path->path_mergeclauses))) + path->skip_mark_restore = true; + else + path->skip_mark_restore = false; + + /* + * Get approx # tuples passing the mergequals. We use approx_tuple_count + * here because we need an estimate done with JOIN_INNER semantics. + */ + mergejointuples = approx_tuple_count(root, &path->jpath, mergeclauses); + + /* + * When there are equal merge keys in the outer relation, the mergejoin + * must rescan any matching tuples in the inner relation. This means + * re-fetching inner tuples; we have to estimate how often that happens. + * + * For regular inner and outer joins, the number of re-fetches can be + * estimated approximately as size of merge join output minus size of + * inner relation. Assume that the distinct key values are 1, 2, ..., and + * denote the number of values of each key in the outer relation as m1, + * m2, ...; in the inner relation, n1, n2, ... Then we have + * + * size of join = m1 * n1 + m2 * n2 + ... + * + * number of rescanned tuples = (m1 - 1) * n1 + (m2 - 1) * n2 + ... = m1 * + * n1 + m2 * n2 + ... - (n1 + n2 + ...) = size of join - size of inner + * relation + * + * This equation works correctly for outer tuples having no inner match + * (nk = 0), but not for inner tuples having no outer match (mk = 0); we + * are effectively subtracting those from the number of rescanned tuples, + * when we should not. Can we do better without expensive selectivity + * computations? + * + * The whole issue is moot if we are working from a unique-ified outer + * input, or if we know we don't need to mark/restore at all. + */ + if (IsA(outer_path, UniquePath) || path->skip_mark_restore) + rescannedtuples = 0; + else + { + rescannedtuples = mergejointuples - inner_path_rows; + /* Must clamp because of possible underestimate */ + if (rescannedtuples < 0) + rescannedtuples = 0; + } + + /* + * We'll inflate various costs this much to account for rescanning. Note + * that this is to be multiplied by something involving inner_rows, or + * another number related to the portion of the inner rel we'll scan. + */ + rescanratio = 1.0 + (rescannedtuples / inner_rows); + + /* + * Decide whether we want to materialize the inner input to shield it from + * mark/restore and performing re-fetches. Our cost model for regular + * re-fetches is that a re-fetch costs the same as an original fetch, + * which is probably an overestimate; but on the other hand we ignore the + * bookkeeping costs of mark/restore. Not clear if it's worth developing + * a more refined model. So we just need to inflate the inner run cost by + * rescanratio. + */ + bare_inner_cost = inner_run_cost * rescanratio; + + /* + * When we interpose a Material node the re-fetch cost is assumed to be + * just cpu_operator_cost per tuple, independently of the underlying + * plan's cost; and we charge an extra cpu_operator_cost per original + * fetch as well. Note that we're assuming the materialize node will + * never spill to disk, since it only has to remember tuples back to the + * last mark. (If there are a huge number of duplicates, our other cost + * factors will make the path so expensive that it probably won't get + * chosen anyway.) So we don't use cost_rescan here. + * + * Note: keep this estimate in sync with create_mergejoin_plan's labeling + * of the generated Material node. + */ + mat_inner_cost = inner_run_cost + + cpu_operator_cost * inner_rows * rescanratio; + + /* + * If we don't need mark/restore at all, we don't need materialization. + */ + if (path->skip_mark_restore) + path->materialize_inner = false; + + /* + * Prefer materializing if it looks cheaper, unless the user has asked to + * suppress materialization. + */ + else if (enable_material && mat_inner_cost < bare_inner_cost) + path->materialize_inner = true; + + /* + * Even if materializing doesn't look cheaper, we *must* do it if the + * inner path is to be used directly (without sorting) and it doesn't + * support mark/restore. + * + * Since the inner side must be ordered, and only Sorts and IndexScans can + * create order to begin with, and they both support mark/restore, you + * might think there's no problem --- but you'd be wrong. Nestloop and + * merge joins can *preserve* the order of their inputs, so they can be + * selected as the input of a mergejoin, and they don't support + * mark/restore at present. + * + * We don't test the value of enable_material here, because + * materialization is required for correctness in this case, and turning + * it off does not entitle us to deliver an invalid plan. + */ + else if (innersortkeys == NIL && + !ExecSupportsMarkRestore(inner_path)) + path->materialize_inner = true; + + /* + * Also, force materializing if the inner path is to be sorted and the + * sort is expected to spill to disk. This is because the final merge + * pass can be done on-the-fly if it doesn't have to support mark/restore. + * We don't try to adjust the cost estimates for this consideration, + * though. + * + * Since materialization is a performance optimization in this case, + * rather than necessary for correctness, we skip it if enable_material is + * off. + */ + else if (enable_material && innersortkeys != NIL && + relation_byte_size(inner_path_rows, + inner_path->pathtarget->width) > + (work_mem * 1024L)) + path->materialize_inner = true; + else + path->materialize_inner = false; + + /* Charge the right incremental cost for the chosen case */ + if (path->materialize_inner) + run_cost += mat_inner_cost; + else + run_cost += bare_inner_cost; + + /* CPU costs */ + + /* + * The number of tuple comparisons needed is approximately number of outer + * rows plus number of inner rows plus number of rescanned tuples (can we + * refine this?). At each one, we need to evaluate the mergejoin quals. + */ + startup_cost += merge_qual_cost.startup; + startup_cost += merge_qual_cost.per_tuple * + (outer_skip_rows + inner_skip_rows * rescanratio); + run_cost += merge_qual_cost.per_tuple * + ((outer_rows - outer_skip_rows) + + (inner_rows - inner_skip_rows) * rescanratio); + + /* + * For each tuple that gets through the mergejoin proper, we charge + * cpu_tuple_cost plus the cost of evaluating additional restriction + * clauses that are to be applied at the join. (This is pessimistic since + * not all of the quals may get evaluated at each tuple.) + * + * Note: we could adjust for SEMI/ANTI joins skipping some qual + * evaluations here, but it's probably not worth the trouble. + */ + startup_cost += qp_qual_cost.startup; + cpu_per_tuple = cpu_tuple_cost + qp_qual_cost.per_tuple; + run_cost += cpu_per_tuple * mergejointuples; + + /* tlist eval costs are paid per output row, not per tuple scanned */ + startup_cost += path->jpath.path.pathtarget->cost.startup; + run_cost += path->jpath.path.pathtarget->cost.per_tuple * path->jpath.path.rows; + + path->jpath.path.startup_cost = startup_cost; + path->jpath.path.total_cost = startup_cost + run_cost; +} + +/* + * run mergejoinscansel() with caching + */ +static MergeScanSelCache * +cached_scansel(PlannerInfo *root, RestrictInfo *rinfo, PathKey *pathkey) +{ + MergeScanSelCache *cache; + ListCell *lc; + Selectivity leftstartsel, + leftendsel, + rightstartsel, + rightendsel; + MemoryContext oldcontext; + + /* Do we have this result already? */ + foreach(lc, rinfo->scansel_cache) + { + cache = (MergeScanSelCache *) lfirst(lc); + if (cache->opfamily == pathkey->pk_opfamily && + cache->collation == pathkey->pk_eclass->ec_collation && + cache->strategy == pathkey->pk_strategy && + cache->nulls_first == pathkey->pk_nulls_first) + return cache; + } + + /* Nope, do the computation */ + mergejoinscansel(root, + (Node *) rinfo->clause, + pathkey->pk_opfamily, + pathkey->pk_strategy, + pathkey->pk_nulls_first, + &leftstartsel, + &leftendsel, + &rightstartsel, + &rightendsel); + + /* Cache the result in suitably long-lived workspace */ + oldcontext = MemoryContextSwitchTo(root->planner_cxt); + + cache = (MergeScanSelCache *) palloc(sizeof(MergeScanSelCache)); + cache->opfamily = pathkey->pk_opfamily; + cache->collation = pathkey->pk_eclass->ec_collation; + cache->strategy = pathkey->pk_strategy; + cache->nulls_first = pathkey->pk_nulls_first; + cache->leftstartsel = leftstartsel; + cache->leftendsel = leftendsel; + cache->rightstartsel = rightstartsel; + cache->rightendsel = rightendsel; + + rinfo->scansel_cache = lappend(rinfo->scansel_cache, cache); + + MemoryContextSwitchTo(oldcontext); + + return cache; +} + +/* + * initial_cost_hashjoin + * Preliminary estimate of the cost of a hashjoin path. + * + * This must quickly produce lower-bound estimates of the path's startup and + * total costs. If we are unable to eliminate the proposed path from + * consideration using the lower bounds, final_cost_hashjoin will be called + * to obtain the final estimates. + * + * The exact division of labor between this function and final_cost_hashjoin + * is private to them, and represents a tradeoff between speed of the initial + * estimate and getting a tight lower bound. We choose to not examine the + * join quals here (other than by counting the number of hash clauses), + * so we can't do much with CPU costs. We do assume that + * ExecChooseHashTableSize is cheap enough to use here. + * + * 'workspace' is to be filled with startup_cost, total_cost, and perhaps + * other data to be used by final_cost_hashjoin + * 'jointype' is the type of join to be performed + * 'hashclauses' is the list of joinclauses to be used as hash clauses + * 'outer_path' is the outer input to the join + * 'inner_path' is the inner input to the join + * 'extra' contains miscellaneous information about the join + * 'parallel_hash' indicates that inner_path is partial and that a shared + * hash table will be built in parallel + */ +void +initial_cost_hashjoin(PlannerInfo *root, JoinCostWorkspace *workspace, + JoinType jointype, + List *hashclauses, + Path *outer_path, Path *inner_path, + JoinPathExtraData *extra, + bool parallel_hash) +{ + Cost startup_cost = 0; + Cost run_cost = 0; + double outer_path_rows = outer_path->rows; + double inner_path_rows = inner_path->rows; + double inner_path_rows_total = inner_path_rows; + int num_hashclauses = list_length(hashclauses); + int numbuckets; + int numbatches; + int num_skew_mcvs; + size_t space_allowed; /* unused */ + + /* cost of source data */ + startup_cost += outer_path->startup_cost; + run_cost += outer_path->total_cost - outer_path->startup_cost; + startup_cost += inner_path->total_cost; + + /* + * Cost of computing hash function: must do it once per input tuple. We + * charge one cpu_operator_cost for each column's hash function. Also, + * tack on one cpu_tuple_cost per inner row, to model the costs of + * inserting the row into the hashtable. + * + * XXX when a hashclause is more complex than a single operator, we really + * should charge the extra eval costs of the left or right side, as + * appropriate, here. This seems more work than it's worth at the moment. + */ + startup_cost += (cpu_operator_cost * num_hashclauses + cpu_tuple_cost) + * inner_path_rows; + run_cost += cpu_operator_cost * num_hashclauses * outer_path_rows; + + /* + * If this is a parallel hash build, then the value we have for + * inner_rows_total currently refers only to the rows returned by each + * participant. For shared hash table size estimation, we need the total + * number, so we need to undo the division. + */ + if (parallel_hash) + inner_path_rows_total *= get_parallel_divisor(inner_path); + + /* + * Get hash table size that executor would use for inner relation. + * + * XXX for the moment, always assume that skew optimization will be + * performed. As long as SKEW_HASH_MEM_PERCENT is small, it's not worth + * trying to determine that for sure. + * + * XXX at some point it might be interesting to try to account for skew + * optimization in the cost estimate, but for now, we don't. + */ + ExecChooseHashTableSize(inner_path_rows_total, + inner_path->pathtarget->width, + true, /* useskew */ + parallel_hash, /* try_combined_hash_mem */ + outer_path->parallel_workers, + &space_allowed, + &numbuckets, + &numbatches, + &num_skew_mcvs); + + /* + * If inner relation is too big then we will need to "batch" the join, + * which implies writing and reading most of the tuples to disk an extra + * time. Charge seq_page_cost per page, since the I/O should be nice and + * sequential. Writing the inner rel counts as startup cost, all the rest + * as run cost. + */ + if (numbatches > 1) + { + double outerpages = page_size(outer_path_rows, + outer_path->pathtarget->width); + double innerpages = page_size(inner_path_rows, + inner_path->pathtarget->width); + + startup_cost += seq_page_cost * innerpages; + run_cost += seq_page_cost * (innerpages + 2 * outerpages); + } + + /* CPU costs left for later */ + + /* Public result fields */ + workspace->startup_cost = startup_cost; + workspace->total_cost = startup_cost + run_cost; + /* Save private data for final_cost_hashjoin */ + workspace->run_cost = run_cost; + workspace->numbuckets = numbuckets; + workspace->numbatches = numbatches; + workspace->inner_rows_total = inner_path_rows_total; +} + +/* + * final_cost_hashjoin + * Final estimate of the cost and result size of a hashjoin path. + * + * Note: the numbatches estimate is also saved into 'path' for use later + * + * 'path' is already filled in except for the rows and cost fields and + * num_batches + * 'workspace' is the result from initial_cost_hashjoin + * 'extra' contains miscellaneous information about the join + */ +void +final_cost_hashjoin(PlannerInfo *root, HashPath *path, + JoinCostWorkspace *workspace, + JoinPathExtraData *extra) +{ + Path *outer_path = path->jpath.outerjoinpath; + Path *inner_path = path->jpath.innerjoinpath; + double outer_path_rows = outer_path->rows; + double inner_path_rows = inner_path->rows; + double inner_path_rows_total = workspace->inner_rows_total; + List *hashclauses = path->path_hashclauses; + Cost startup_cost = workspace->startup_cost; + Cost run_cost = workspace->run_cost; + int numbuckets = workspace->numbuckets; + int numbatches = workspace->numbatches; + Cost cpu_per_tuple; + QualCost hash_qual_cost; + QualCost qp_qual_cost; + double hashjointuples; + double virtualbuckets; + Selectivity innerbucketsize; + Selectivity innermcvfreq; + ListCell *hcl; + + /* Mark the path with the correct row estimate */ + if (path->jpath.path.param_info) + path->jpath.path.rows = path->jpath.path.param_info->ppi_rows; + else + path->jpath.path.rows = path->jpath.path.parent->rows; + + /* For partial paths, scale row estimate. */ + if (path->jpath.path.parallel_workers > 0) + { + double parallel_divisor = get_parallel_divisor(&path->jpath.path); + + path->jpath.path.rows = + clamp_row_est(path->jpath.path.rows / parallel_divisor); + } + + /* + * We could include disable_cost in the preliminary estimate, but that + * would amount to optimizing for the case where the join method is + * disabled, which doesn't seem like the way to bet. + */ + if (!enable_hashjoin) + startup_cost += disable_cost; + + /* mark the path with estimated # of batches */ + path->num_batches = numbatches; + + /* store the total number of tuples (sum of partial row estimates) */ + path->inner_rows_total = inner_path_rows_total; + + /* and compute the number of "virtual" buckets in the whole join */ + virtualbuckets = (double) numbuckets * (double) numbatches; + + /* + * Determine bucketsize fraction and MCV frequency for the inner relation. + * We use the smallest bucketsize or MCV frequency estimated for any + * individual hashclause; this is undoubtedly conservative. + * + * BUT: if inner relation has been unique-ified, we can assume it's good + * for hashing. This is important both because it's the right answer, and + * because we avoid contaminating the cache with a value that's wrong for + * non-unique-ified paths. + */ + if (IsA(inner_path, UniquePath)) + { + innerbucketsize = 1.0 / virtualbuckets; + innermcvfreq = 0.0; + } + else + { + innerbucketsize = 1.0; + innermcvfreq = 1.0; + foreach(hcl, hashclauses) + { + RestrictInfo *restrictinfo = lfirst_node(RestrictInfo, hcl); + Selectivity thisbucketsize; + Selectivity thismcvfreq; + + /* + * First we have to figure out which side of the hashjoin clause + * is the inner side. + * + * Since we tend to visit the same clauses over and over when + * planning a large query, we cache the bucket stats estimates in + * the RestrictInfo node to avoid repeated lookups of statistics. + */ + if (bms_is_subset(restrictinfo->right_relids, + inner_path->parent->relids)) + { + /* righthand side is inner */ + thisbucketsize = restrictinfo->right_bucketsize; + if (thisbucketsize < 0) + { + /* not cached yet */ + estimate_hash_bucket_stats(root, + get_rightop(restrictinfo->clause), + virtualbuckets, + &restrictinfo->right_mcvfreq, + &restrictinfo->right_bucketsize); + thisbucketsize = restrictinfo->right_bucketsize; + } + thismcvfreq = restrictinfo->right_mcvfreq; + } + else + { + Assert(bms_is_subset(restrictinfo->left_relids, + inner_path->parent->relids)); + /* lefthand side is inner */ + thisbucketsize = restrictinfo->left_bucketsize; + if (thisbucketsize < 0) + { + /* not cached yet */ + estimate_hash_bucket_stats(root, + get_leftop(restrictinfo->clause), + virtualbuckets, + &restrictinfo->left_mcvfreq, + &restrictinfo->left_bucketsize); + thisbucketsize = restrictinfo->left_bucketsize; + } + thismcvfreq = restrictinfo->left_mcvfreq; + } + + if (innerbucketsize > thisbucketsize) + innerbucketsize = thisbucketsize; + if (innermcvfreq > thismcvfreq) + innermcvfreq = thismcvfreq; + } + } + + /* + * If the bucket holding the inner MCV would exceed hash_mem, we don't + * want to hash unless there is really no other alternative, so apply + * disable_cost. (The executor normally copes with excessive memory usage + * by splitting batches, but obviously it cannot separate equal values + * that way, so it will be unable to drive the batch size below hash_mem + * when this is true.) + */ + if (relation_byte_size(clamp_row_est(inner_path_rows * innermcvfreq), + inner_path->pathtarget->width) > get_hash_memory_limit()) + startup_cost += disable_cost; + + /* + * Compute cost of the hashquals and qpquals (other restriction clauses) + * separately. + */ + cost_qual_eval(&hash_qual_cost, hashclauses, root); + cost_qual_eval(&qp_qual_cost, path->jpath.joinrestrictinfo, root); + qp_qual_cost.startup -= hash_qual_cost.startup; + qp_qual_cost.per_tuple -= hash_qual_cost.per_tuple; + + /* CPU costs */ + + if (path->jpath.jointype == JOIN_SEMI || + path->jpath.jointype == JOIN_ANTI || + extra->inner_unique) + { + double outer_matched_rows; + Selectivity inner_scan_frac; + + /* + * With a SEMI or ANTI join, or if the innerrel is known unique, the + * executor will stop after the first match. + * + * For an outer-rel row that has at least one match, we can expect the + * bucket scan to stop after a fraction 1/(match_count+1) of the + * bucket's rows, if the matches are evenly distributed. Since they + * probably aren't quite evenly distributed, we apply a fuzz factor of + * 2.0 to that fraction. (If we used a larger fuzz factor, we'd have + * to clamp inner_scan_frac to at most 1.0; but since match_count is + * at least 1, no such clamp is needed now.) + */ + outer_matched_rows = rint(outer_path_rows * extra->semifactors.outer_match_frac); + inner_scan_frac = 2.0 / (extra->semifactors.match_count + 1.0); + + startup_cost += hash_qual_cost.startup; + run_cost += hash_qual_cost.per_tuple * outer_matched_rows * + clamp_row_est(inner_path_rows * innerbucketsize * inner_scan_frac) * 0.5; + + /* + * For unmatched outer-rel rows, the picture is quite a lot different. + * In the first place, there is no reason to assume that these rows + * preferentially hit heavily-populated buckets; instead assume they + * are uncorrelated with the inner distribution and so they see an + * average bucket size of inner_path_rows / virtualbuckets. In the + * second place, it seems likely that they will have few if any exact + * hash-code matches and so very few of the tuples in the bucket will + * actually require eval of the hash quals. We don't have any good + * way to estimate how many will, but for the moment assume that the + * effective cost per bucket entry is one-tenth what it is for + * matchable tuples. + */ + run_cost += hash_qual_cost.per_tuple * + (outer_path_rows - outer_matched_rows) * + clamp_row_est(inner_path_rows / virtualbuckets) * 0.05; + + /* Get # of tuples that will pass the basic join */ + if (path->jpath.jointype == JOIN_ANTI) + hashjointuples = outer_path_rows - outer_matched_rows; + else + hashjointuples = outer_matched_rows; + } + else + { + /* + * The number of tuple comparisons needed is the number of outer + * tuples times the typical number of tuples in a hash bucket, which + * is the inner relation size times its bucketsize fraction. At each + * one, we need to evaluate the hashjoin quals. But actually, + * charging the full qual eval cost at each tuple is pessimistic, + * since we don't evaluate the quals unless the hash values match + * exactly. For lack of a better idea, halve the cost estimate to + * allow for that. + */ + startup_cost += hash_qual_cost.startup; + run_cost += hash_qual_cost.per_tuple * outer_path_rows * + clamp_row_est(inner_path_rows * innerbucketsize) * 0.5; + + /* + * Get approx # tuples passing the hashquals. We use + * approx_tuple_count here because we need an estimate done with + * JOIN_INNER semantics. + */ + hashjointuples = approx_tuple_count(root, &path->jpath, hashclauses); + } + + /* + * For each tuple that gets through the hashjoin proper, we charge + * cpu_tuple_cost plus the cost of evaluating additional restriction + * clauses that are to be applied at the join. (This is pessimistic since + * not all of the quals may get evaluated at each tuple.) + */ + startup_cost += qp_qual_cost.startup; + cpu_per_tuple = cpu_tuple_cost + qp_qual_cost.per_tuple; + run_cost += cpu_per_tuple * hashjointuples; + + /* tlist eval costs are paid per output row, not per tuple scanned */ + startup_cost += path->jpath.path.pathtarget->cost.startup; + run_cost += path->jpath.path.pathtarget->cost.per_tuple * path->jpath.path.rows; + + path->jpath.path.startup_cost = startup_cost; + path->jpath.path.total_cost = startup_cost + run_cost; +} + + +/* + * cost_subplan + * Figure the costs for a SubPlan (or initplan). + * + * Note: we could dig the subplan's Plan out of the root list, but in practice + * all callers have it handy already, so we make them pass it. + */ +void +cost_subplan(PlannerInfo *root, SubPlan *subplan, Plan *plan) +{ + QualCost sp_cost; + + /* Figure any cost for evaluating the testexpr */ + cost_qual_eval(&sp_cost, + make_ands_implicit((Expr *) subplan->testexpr), + root); + + if (subplan->useHashTable) + { + /* + * If we are using a hash table for the subquery outputs, then the + * cost of evaluating the query is a one-time cost. We charge one + * cpu_operator_cost per tuple for the work of loading the hashtable, + * too. + */ + sp_cost.startup += plan->total_cost + + cpu_operator_cost * plan->plan_rows; + + /* + * The per-tuple costs include the cost of evaluating the lefthand + * expressions, plus the cost of probing the hashtable. We already + * accounted for the lefthand expressions as part of the testexpr, and + * will also have counted one cpu_operator_cost for each comparison + * operator. That is probably too low for the probing cost, but it's + * hard to make a better estimate, so live with it for now. + */ + } + else + { + /* + * Otherwise we will be rescanning the subplan output on each + * evaluation. We need to estimate how much of the output we will + * actually need to scan. NOTE: this logic should agree with the + * tuple_fraction estimates used by make_subplan() in + * plan/subselect.c. + */ + Cost plan_run_cost = plan->total_cost - plan->startup_cost; + + if (subplan->subLinkType == EXISTS_SUBLINK) + { + /* we only need to fetch 1 tuple; clamp to avoid zero divide */ + sp_cost.per_tuple += plan_run_cost / clamp_row_est(plan->plan_rows); + } + else if (subplan->subLinkType == ALL_SUBLINK || + subplan->subLinkType == ANY_SUBLINK) + { + /* assume we need 50% of the tuples */ + sp_cost.per_tuple += 0.50 * plan_run_cost; + /* also charge a cpu_operator_cost per row examined */ + sp_cost.per_tuple += 0.50 * plan->plan_rows * cpu_operator_cost; + } + else + { + /* assume we need all tuples */ + sp_cost.per_tuple += plan_run_cost; + } + + /* + * Also account for subplan's startup cost. If the subplan is + * uncorrelated or undirect correlated, AND its topmost node is one + * that materializes its output, assume that we'll only need to pay + * its startup cost once; otherwise assume we pay the startup cost + * every time. + */ + if (subplan->parParam == NIL && + ExecMaterializesOutput(nodeTag(plan))) + sp_cost.startup += plan->startup_cost; + else + sp_cost.per_tuple += plan->startup_cost; + } + + subplan->startup_cost = sp_cost.startup; + subplan->per_call_cost = sp_cost.per_tuple; +} + + +/* + * cost_rescan + * Given a finished Path, estimate the costs of rescanning it after + * having done so the first time. For some Path types a rescan is + * cheaper than an original scan (if no parameters change), and this + * function embodies knowledge about that. The default is to return + * the same costs stored in the Path. (Note that the cost estimates + * actually stored in Paths are always for first scans.) + * + * This function is not currently intended to model effects such as rescans + * being cheaper due to disk block caching; what we are concerned with is + * plan types wherein the executor caches results explicitly, or doesn't + * redo startup calculations, etc. + */ +static void +cost_rescan(PlannerInfo *root, Path *path, + Cost *rescan_startup_cost, /* output parameters */ + Cost *rescan_total_cost) +{ + switch (path->pathtype) + { + case T_FunctionScan: + + /* + * Currently, nodeFunctionscan.c always executes the function to + * completion before returning any rows, and caches the results in + * a tuplestore. So the function eval cost is all startup cost + * and isn't paid over again on rescans. However, all run costs + * will be paid over again. + */ + *rescan_startup_cost = 0; + *rescan_total_cost = path->total_cost - path->startup_cost; + break; + case T_HashJoin: + + /* + * If it's a single-batch join, we don't need to rebuild the hash + * table during a rescan. + */ + if (((HashPath *) path)->num_batches == 1) + { + /* Startup cost is exactly the cost of hash table building */ + *rescan_startup_cost = 0; + *rescan_total_cost = path->total_cost - path->startup_cost; + } + else + { + /* Otherwise, no special treatment */ + *rescan_startup_cost = path->startup_cost; + *rescan_total_cost = path->total_cost; + } + break; + case T_CteScan: + case T_WorkTableScan: + { + /* + * These plan types materialize their final result in a + * tuplestore or tuplesort object. So the rescan cost is only + * cpu_tuple_cost per tuple, unless the result is large enough + * to spill to disk. + */ + Cost run_cost = cpu_tuple_cost * path->rows; + double nbytes = relation_byte_size(path->rows, + path->pathtarget->width); + long work_mem_bytes = work_mem * 1024L; + + if (nbytes > work_mem_bytes) + { + /* It will spill, so account for re-read cost */ + double npages = ceil(nbytes / BLCKSZ); + + run_cost += seq_page_cost * npages; + } + *rescan_startup_cost = 0; + *rescan_total_cost = run_cost; + } + break; + case T_Material: + case T_Sort: + { + /* + * These plan types not only materialize their results, but do + * not implement qual filtering or projection. So they are + * even cheaper to rescan than the ones above. We charge only + * cpu_operator_cost per tuple. (Note: keep that in sync with + * the run_cost charge in cost_sort, and also see comments in + * cost_material before you change it.) + */ + Cost run_cost = cpu_operator_cost * path->rows; + double nbytes = relation_byte_size(path->rows, + path->pathtarget->width); + long work_mem_bytes = work_mem * 1024L; + + if (nbytes > work_mem_bytes) + { + /* It will spill, so account for re-read cost */ + double npages = ceil(nbytes / BLCKSZ); + + run_cost += seq_page_cost * npages; + } + *rescan_startup_cost = 0; + *rescan_total_cost = run_cost; + } + break; + case T_Memoize: + /* All the hard work is done by cost_memoize_rescan */ + cost_memoize_rescan(root, (MemoizePath *) path, + rescan_startup_cost, rescan_total_cost); + break; + default: + *rescan_startup_cost = path->startup_cost; + *rescan_total_cost = path->total_cost; + break; + } +} + + +/* + * cost_qual_eval + * Estimate the CPU costs of evaluating a WHERE clause. + * The input can be either an implicitly-ANDed list of boolean + * expressions, or a list of RestrictInfo nodes. (The latter is + * preferred since it allows caching of the results.) + * The result includes both a one-time (startup) component, + * and a per-evaluation component. + */ +void +cost_qual_eval(QualCost *cost, List *quals, PlannerInfo *root) +{ + cost_qual_eval_context context; + ListCell *l; + + context.root = root; + context.total.startup = 0; + context.total.per_tuple = 0; + + /* We don't charge any cost for the implicit ANDing at top level ... */ + + foreach(l, quals) + { + Node *qual = (Node *) lfirst(l); + + cost_qual_eval_walker(qual, &context); + } + + *cost = context.total; +} + +/* + * cost_qual_eval_node + * As above, for a single RestrictInfo or expression. + */ +void +cost_qual_eval_node(QualCost *cost, Node *qual, PlannerInfo *root) +{ + cost_qual_eval_context context; + + context.root = root; + context.total.startup = 0; + context.total.per_tuple = 0; + + cost_qual_eval_walker(qual, &context); + + *cost = context.total; +} + +static bool +cost_qual_eval_walker(Node *node, cost_qual_eval_context *context) +{ + if (node == NULL) + return false; + + /* + * RestrictInfo nodes contain an eval_cost field reserved for this + * routine's use, so that it's not necessary to evaluate the qual clause's + * cost more than once. If the clause's cost hasn't been computed yet, + * the field's startup value will contain -1. + */ + if (IsA(node, RestrictInfo)) + { + RestrictInfo *rinfo = (RestrictInfo *) node; + + if (rinfo->eval_cost.startup < 0) + { + cost_qual_eval_context locContext; + + locContext.root = context->root; + locContext.total.startup = 0; + locContext.total.per_tuple = 0; + + /* + * For an OR clause, recurse into the marked-up tree so that we + * set the eval_cost for contained RestrictInfos too. + */ + if (rinfo->orclause) + cost_qual_eval_walker((Node *) rinfo->orclause, &locContext); + else + cost_qual_eval_walker((Node *) rinfo->clause, &locContext); + + /* + * If the RestrictInfo is marked pseudoconstant, it will be tested + * only once, so treat its cost as all startup cost. + */ + if (rinfo->pseudoconstant) + { + /* count one execution during startup */ + locContext.total.startup += locContext.total.per_tuple; + locContext.total.per_tuple = 0; + } + rinfo->eval_cost = locContext.total; + } + context->total.startup += rinfo->eval_cost.startup; + context->total.per_tuple += rinfo->eval_cost.per_tuple; + /* do NOT recurse into children */ + return false; + } + + /* + * For each operator or function node in the given tree, we charge the + * estimated execution cost given by pg_proc.procost (remember to multiply + * this by cpu_operator_cost). + * + * Vars and Consts are charged zero, and so are boolean operators (AND, + * OR, NOT). Simplistic, but a lot better than no model at all. + * + * Should we try to account for the possibility of short-circuit + * evaluation of AND/OR? Probably *not*, because that would make the + * results depend on the clause ordering, and we are not in any position + * to expect that the current ordering of the clauses is the one that's + * going to end up being used. The above per-RestrictInfo caching would + * not mix well with trying to re-order clauses anyway. + * + * Another issue that is entirely ignored here is that if a set-returning + * function is below top level in the tree, the functions/operators above + * it will need to be evaluated multiple times. In practical use, such + * cases arise so seldom as to not be worth the added complexity needed; + * moreover, since our rowcount estimates for functions tend to be pretty + * phony, the results would also be pretty phony. + */ + if (IsA(node, FuncExpr)) + { + add_function_cost(context->root, ((FuncExpr *) node)->funcid, node, + &context->total); + } + else if (IsA(node, OpExpr) || + IsA(node, DistinctExpr) || + IsA(node, NullIfExpr)) + { + /* rely on struct equivalence to treat these all alike */ + set_opfuncid((OpExpr *) node); + add_function_cost(context->root, ((OpExpr *) node)->opfuncid, node, + &context->total); + } + else if (IsA(node, ScalarArrayOpExpr)) + { + ScalarArrayOpExpr *saop = (ScalarArrayOpExpr *) node; + Node *arraynode = (Node *) lsecond(saop->args); + QualCost sacosts; + QualCost hcosts; + int estarraylen = estimate_array_length(arraynode); + + set_sa_opfuncid(saop); + sacosts.startup = sacosts.per_tuple = 0; + add_function_cost(context->root, saop->opfuncid, NULL, + &sacosts); + + if (OidIsValid(saop->hashfuncid)) + { + /* Handle costs for hashed ScalarArrayOpExpr */ + hcosts.startup = hcosts.per_tuple = 0; + + add_function_cost(context->root, saop->hashfuncid, NULL, &hcosts); + context->total.startup += sacosts.startup + hcosts.startup; + + /* Estimate the cost of building the hashtable. */ + context->total.startup += estarraylen * hcosts.per_tuple; + + /* + * XXX should we charge a little bit for sacosts.per_tuple when + * building the table, or is it ok to assume there will be zero + * hash collision? + */ + + /* + * Charge for hashtable lookups. Charge a single hash and a + * single comparison. + */ + context->total.per_tuple += hcosts.per_tuple + sacosts.per_tuple; + } + else + { + /* + * Estimate that the operator will be applied to about half of the + * array elements before the answer is determined. + */ + context->total.startup += sacosts.startup; + context->total.per_tuple += sacosts.per_tuple * + estimate_array_length(arraynode) * 0.5; + } + } + else if (IsA(node, Aggref) || + IsA(node, WindowFunc)) + { + /* + * Aggref and WindowFunc nodes are (and should be) treated like Vars, + * ie, zero execution cost in the current model, because they behave + * essentially like Vars at execution. We disregard the costs of + * their input expressions for the same reason. The actual execution + * costs of the aggregate/window functions and their arguments have to + * be factored into plan-node-specific costing of the Agg or WindowAgg + * plan node. + */ + return false; /* don't recurse into children */ + } + else if (IsA(node, GroupingFunc)) + { + /* Treat this as having cost 1 */ + context->total.per_tuple += cpu_operator_cost; + return false; /* don't recurse into children */ + } + else if (IsA(node, CoerceViaIO)) + { + CoerceViaIO *iocoerce = (CoerceViaIO *) node; + Oid iofunc; + Oid typioparam; + bool typisvarlena; + + /* check the result type's input function */ + getTypeInputInfo(iocoerce->resulttype, + &iofunc, &typioparam); + add_function_cost(context->root, iofunc, NULL, + &context->total); + /* check the input type's output function */ + getTypeOutputInfo(exprType((Node *) iocoerce->arg), + &iofunc, &typisvarlena); + add_function_cost(context->root, iofunc, NULL, + &context->total); + } + else if (IsA(node, ArrayCoerceExpr)) + { + ArrayCoerceExpr *acoerce = (ArrayCoerceExpr *) node; + QualCost perelemcost; + + cost_qual_eval_node(&perelemcost, (Node *) acoerce->elemexpr, + context->root); + context->total.startup += perelemcost.startup; + if (perelemcost.per_tuple > 0) + context->total.per_tuple += perelemcost.per_tuple * + estimate_array_length((Node *) acoerce->arg); + } + else if (IsA(node, RowCompareExpr)) + { + /* Conservatively assume we will check all the columns */ + RowCompareExpr *rcexpr = (RowCompareExpr *) node; + ListCell *lc; + + foreach(lc, rcexpr->opnos) + { + Oid opid = lfirst_oid(lc); + + add_function_cost(context->root, get_opcode(opid), NULL, + &context->total); + } + } + else if (IsA(node, MinMaxExpr) || + IsA(node, SQLValueFunction) || + IsA(node, XmlExpr) || + IsA(node, CoerceToDomain) || + IsA(node, NextValueExpr)) + { + /* Treat all these as having cost 1 */ + context->total.per_tuple += cpu_operator_cost; + } + else if (IsA(node, CurrentOfExpr)) + { + /* Report high cost to prevent selection of anything but TID scan */ + context->total.startup += disable_cost; + } + else if (IsA(node, SubLink)) + { + /* This routine should not be applied to un-planned expressions */ + elog(ERROR, "cannot handle unplanned sub-select"); + } + else if (IsA(node, SubPlan)) + { + /* + * A subplan node in an expression typically indicates that the + * subplan will be executed on each evaluation, so charge accordingly. + * (Sub-selects that can be executed as InitPlans have already been + * removed from the expression.) + */ + SubPlan *subplan = (SubPlan *) node; + + context->total.startup += subplan->startup_cost; + context->total.per_tuple += subplan->per_call_cost; + + /* + * We don't want to recurse into the testexpr, because it was already + * counted in the SubPlan node's costs. So we're done. + */ + return false; + } + else if (IsA(node, AlternativeSubPlan)) + { + /* + * Arbitrarily use the first alternative plan for costing. (We should + * certainly only include one alternative, and we don't yet have + * enough information to know which one the executor is most likely to + * use.) + */ + AlternativeSubPlan *asplan = (AlternativeSubPlan *) node; + + return cost_qual_eval_walker((Node *) linitial(asplan->subplans), + context); + } + else if (IsA(node, PlaceHolderVar)) + { + /* + * A PlaceHolderVar should be given cost zero when considering general + * expression evaluation costs. The expense of doing the contained + * expression is charged as part of the tlist eval costs of the scan + * or join where the PHV is first computed (see set_rel_width and + * add_placeholders_to_joinrel). If we charged it again here, we'd be + * double-counting the cost for each level of plan that the PHV + * bubbles up through. Hence, return without recursing into the + * phexpr. + */ + return false; + } + + /* recurse into children */ + return expression_tree_walker(node, cost_qual_eval_walker, + (void *) context); +} + +/* + * get_restriction_qual_cost + * Compute evaluation costs of a baserel's restriction quals, plus any + * movable join quals that have been pushed down to the scan. + * Results are returned into *qpqual_cost. + * + * This is a convenience subroutine that works for seqscans and other cases + * where all the given quals will be evaluated the hard way. It's not useful + * for cost_index(), for example, where the index machinery takes care of + * some of the quals. We assume baserestrictcost was previously set by + * set_baserel_size_estimates(). + */ +static void +get_restriction_qual_cost(PlannerInfo *root, RelOptInfo *baserel, + ParamPathInfo *param_info, + QualCost *qpqual_cost) +{ + if (param_info) + { + /* Include costs of pushed-down clauses */ + cost_qual_eval(qpqual_cost, param_info->ppi_clauses, root); + + qpqual_cost->startup += baserel->baserestrictcost.startup; + qpqual_cost->per_tuple += baserel->baserestrictcost.per_tuple; + } + else + *qpqual_cost = baserel->baserestrictcost; +} + + +/* + * compute_semi_anti_join_factors + * Estimate how much of the inner input a SEMI, ANTI, or inner_unique join + * can be expected to scan. + * + * In a hash or nestloop SEMI/ANTI join, the executor will stop scanning + * inner rows as soon as it finds a match to the current outer row. + * The same happens if we have detected the inner rel is unique. + * We should therefore adjust some of the cost components for this effect. + * This function computes some estimates needed for these adjustments. + * These estimates will be the same regardless of the particular paths used + * for the outer and inner relation, so we compute these once and then pass + * them to all the join cost estimation functions. + * + * Input parameters: + * joinrel: join relation under consideration + * outerrel: outer relation under consideration + * innerrel: inner relation under consideration + * jointype: if not JOIN_SEMI or JOIN_ANTI, we assume it's inner_unique + * sjinfo: SpecialJoinInfo relevant to this join + * restrictlist: join quals + * Output parameters: + * *semifactors is filled in (see pathnodes.h for field definitions) + */ +void +compute_semi_anti_join_factors(PlannerInfo *root, + RelOptInfo *joinrel, + RelOptInfo *outerrel, + RelOptInfo *innerrel, + JoinType jointype, + SpecialJoinInfo *sjinfo, + List *restrictlist, + SemiAntiJoinFactors *semifactors) +{ + Selectivity jselec; + Selectivity nselec; + Selectivity avgmatch; + SpecialJoinInfo norm_sjinfo; + List *joinquals; + ListCell *l; + + /* + * In an ANTI join, we must ignore clauses that are "pushed down", since + * those won't affect the match logic. In a SEMI join, we do not + * distinguish joinquals from "pushed down" quals, so just use the whole + * restrictinfo list. For other outer join types, we should consider only + * non-pushed-down quals, so that this devolves to an IS_OUTER_JOIN check. + */ + if (IS_OUTER_JOIN(jointype)) + { + joinquals = NIL; + foreach(l, restrictlist) + { + RestrictInfo *rinfo = lfirst_node(RestrictInfo, l); + + if (!RINFO_IS_PUSHED_DOWN(rinfo, joinrel->relids)) + joinquals = lappend(joinquals, rinfo); + } + } + else + joinquals = restrictlist; + + /* + * Get the JOIN_SEMI or JOIN_ANTI selectivity of the join clauses. + */ + jselec = clauselist_selectivity(root, + joinquals, + 0, + (jointype == JOIN_ANTI) ? JOIN_ANTI : JOIN_SEMI, + sjinfo); + + /* + * Also get the normal inner-join selectivity of the join clauses. + */ + norm_sjinfo.type = T_SpecialJoinInfo; + norm_sjinfo.min_lefthand = outerrel->relids; + norm_sjinfo.min_righthand = innerrel->relids; + norm_sjinfo.syn_lefthand = outerrel->relids; + norm_sjinfo.syn_righthand = innerrel->relids; + norm_sjinfo.jointype = JOIN_INNER; + /* we don't bother trying to make the remaining fields valid */ + norm_sjinfo.lhs_strict = false; + norm_sjinfo.delay_upper_joins = false; + norm_sjinfo.semi_can_btree = false; + norm_sjinfo.semi_can_hash = false; + norm_sjinfo.semi_operators = NIL; + norm_sjinfo.semi_rhs_exprs = NIL; + + nselec = clauselist_selectivity(root, + joinquals, + 0, + JOIN_INNER, + &norm_sjinfo); + + /* Avoid leaking a lot of ListCells */ + if (IS_OUTER_JOIN(jointype)) + list_free(joinquals); + + /* + * jselec can be interpreted as the fraction of outer-rel rows that have + * any matches (this is true for both SEMI and ANTI cases). And nselec is + * the fraction of the Cartesian product that matches. So, the average + * number of matches for each outer-rel row that has at least one match is + * nselec * inner_rows / jselec. + * + * Note: it is correct to use the inner rel's "rows" count here, even + * though we might later be considering a parameterized inner path with + * fewer rows. This is because we have included all the join clauses in + * the selectivity estimate. + */ + if (jselec > 0) /* protect against zero divide */ + { + avgmatch = nselec * innerrel->rows / jselec; + /* Clamp to sane range */ + avgmatch = Max(1.0, avgmatch); + } + else + avgmatch = 1.0; + + semifactors->outer_match_frac = jselec; + semifactors->match_count = avgmatch; +} + +/* + * has_indexed_join_quals + * Check whether all the joinquals of a nestloop join are used as + * inner index quals. + * + * If the inner path of a SEMI/ANTI join is an indexscan (including bitmap + * indexscan) that uses all the joinquals as indexquals, we can assume that an + * unmatched outer tuple is cheap to process, whereas otherwise it's probably + * expensive. + */ +static bool +has_indexed_join_quals(NestPath *path) +{ + JoinPath *joinpath = &path->jpath; + Relids joinrelids = joinpath->path.parent->relids; + Path *innerpath = joinpath->innerjoinpath; + List *indexclauses; + bool found_one; + ListCell *lc; + + /* If join still has quals to evaluate, it's not fast */ + if (joinpath->joinrestrictinfo != NIL) + return false; + /* Nor if the inner path isn't parameterized at all */ + if (innerpath->param_info == NULL) + return false; + + /* Find the indexclauses list for the inner scan */ + switch (innerpath->pathtype) + { + case T_IndexScan: + case T_IndexOnlyScan: + indexclauses = ((IndexPath *) innerpath)->indexclauses; + break; + case T_BitmapHeapScan: + { + /* Accept only a simple bitmap scan, not AND/OR cases */ + Path *bmqual = ((BitmapHeapPath *) innerpath)->bitmapqual; + + if (IsA(bmqual, IndexPath)) + indexclauses = ((IndexPath *) bmqual)->indexclauses; + else + return false; + break; + } + default: + + /* + * If it's not a simple indexscan, it probably doesn't run quickly + * for zero rows out, even if it's a parameterized path using all + * the joinquals. + */ + return false; + } + + /* + * Examine the inner path's param clauses. Any that are from the outer + * path must be found in the indexclauses list, either exactly or in an + * equivalent form generated by equivclass.c. Also, we must find at least + * one such clause, else it's a clauseless join which isn't fast. + */ + found_one = false; + foreach(lc, innerpath->param_info->ppi_clauses) + { + RestrictInfo *rinfo = (RestrictInfo *) lfirst(lc); + + if (join_clause_is_movable_into(rinfo, + innerpath->parent->relids, + joinrelids)) + { + if (!is_redundant_with_indexclauses(rinfo, indexclauses)) + return false; + found_one = true; + } + } + return found_one; +} + + +/* + * approx_tuple_count + * Quick-and-dirty estimation of the number of join rows passing + * a set of qual conditions. + * + * The quals can be either an implicitly-ANDed list of boolean expressions, + * or a list of RestrictInfo nodes (typically the latter). + * + * We intentionally compute the selectivity under JOIN_INNER rules, even + * if it's some type of outer join. This is appropriate because we are + * trying to figure out how many tuples pass the initial merge or hash + * join step. + * + * This is quick-and-dirty because we bypass clauselist_selectivity, and + * simply multiply the independent clause selectivities together. Now + * clauselist_selectivity often can't do any better than that anyhow, but + * for some situations (such as range constraints) it is smarter. However, + * we can't effectively cache the results of clauselist_selectivity, whereas + * the individual clause selectivities can be and are cached. + * + * Since we are only using the results to estimate how many potential + * output tuples are generated and passed through qpqual checking, it + * seems OK to live with the approximation. + */ +static double +approx_tuple_count(PlannerInfo *root, JoinPath *path, List *quals) +{ + double tuples; + double outer_tuples = path->outerjoinpath->rows; + double inner_tuples = path->innerjoinpath->rows; + SpecialJoinInfo sjinfo; + Selectivity selec = 1.0; + ListCell *l; + + /* + * Make up a SpecialJoinInfo for JOIN_INNER semantics. + */ + sjinfo.type = T_SpecialJoinInfo; + sjinfo.min_lefthand = path->outerjoinpath->parent->relids; + sjinfo.min_righthand = path->innerjoinpath->parent->relids; + sjinfo.syn_lefthand = path->outerjoinpath->parent->relids; + sjinfo.syn_righthand = path->innerjoinpath->parent->relids; + sjinfo.jointype = JOIN_INNER; + /* we don't bother trying to make the remaining fields valid */ + sjinfo.lhs_strict = false; + sjinfo.delay_upper_joins = false; + sjinfo.semi_can_btree = false; + sjinfo.semi_can_hash = false; + sjinfo.semi_operators = NIL; + sjinfo.semi_rhs_exprs = NIL; + + /* Get the approximate selectivity */ + foreach(l, quals) + { + Node *qual = (Node *) lfirst(l); + + /* Note that clause_selectivity will be able to cache its result */ + selec *= clause_selectivity(root, qual, 0, JOIN_INNER, &sjinfo); + } + + /* Apply it to the input relation sizes */ + tuples = selec * outer_tuples * inner_tuples; + + return clamp_row_est(tuples); +} + + +/* + * set_baserel_size_estimates + * Set the size estimates for the given base relation. + * + * The rel's targetlist and restrictinfo list must have been constructed + * already, and rel->tuples must be set. + * + * We set the following fields of the rel node: + * rows: the estimated number of output tuples (after applying + * restriction clauses). + * width: the estimated average output tuple width in bytes. + * baserestrictcost: estimated cost of evaluating baserestrictinfo clauses. + */ +void +set_baserel_size_estimates(PlannerInfo *root, RelOptInfo *rel) +{ + double nrows; + + /* Should only be applied to base relations */ + Assert(rel->relid > 0); + + nrows = rel->tuples * + clauselist_selectivity(root, + rel->baserestrictinfo, + 0, + JOIN_INNER, + NULL); + + rel->rows = clamp_row_est(nrows); + + cost_qual_eval(&rel->baserestrictcost, rel->baserestrictinfo, root); + + set_rel_width(root, rel); +} + +/* + * get_parameterized_baserel_size + * Make a size estimate for a parameterized scan of a base relation. + * + * 'param_clauses' lists the additional join clauses to be used. + * + * set_baserel_size_estimates must have been applied already. + */ +double +get_parameterized_baserel_size(PlannerInfo *root, RelOptInfo *rel, + List *param_clauses) +{ + List *allclauses; + double nrows; + + /* + * Estimate the number of rows returned by the parameterized scan, knowing + * that it will apply all the extra join clauses as well as the rel's own + * restriction clauses. Note that we force the clauses to be treated as + * non-join clauses during selectivity estimation. + */ + allclauses = list_concat_copy(param_clauses, rel->baserestrictinfo); + nrows = rel->tuples * + clauselist_selectivity(root, + allclauses, + rel->relid, /* do not use 0! */ + JOIN_INNER, + NULL); + nrows = clamp_row_est(nrows); + /* For safety, make sure result is not more than the base estimate */ + if (nrows > rel->rows) + nrows = rel->rows; + return nrows; +} + +/* + * set_joinrel_size_estimates + * Set the size estimates for the given join relation. + * + * The rel's targetlist must have been constructed already, and a + * restriction clause list that matches the given component rels must + * be provided. + * + * Since there is more than one way to make a joinrel for more than two + * base relations, the results we get here could depend on which component + * rel pair is provided. In theory we should get the same answers no matter + * which pair is provided; in practice, since the selectivity estimation + * routines don't handle all cases equally well, we might not. But there's + * not much to be done about it. (Would it make sense to repeat the + * calculations for each pair of input rels that's encountered, and somehow + * average the results? Probably way more trouble than it's worth, and + * anyway we must keep the rowcount estimate the same for all paths for the + * joinrel.) + * + * We set only the rows field here. The reltarget field was already set by + * build_joinrel_tlist, and baserestrictcost is not used for join rels. + */ +void +set_joinrel_size_estimates(PlannerInfo *root, RelOptInfo *rel, + RelOptInfo *outer_rel, + RelOptInfo *inner_rel, + SpecialJoinInfo *sjinfo, + List *restrictlist) +{ + rel->rows = calc_joinrel_size_estimate(root, + rel, + outer_rel, + inner_rel, + outer_rel->rows, + inner_rel->rows, + sjinfo, + restrictlist); +} + +/* + * get_parameterized_joinrel_size + * Make a size estimate for a parameterized scan of a join relation. + * + * 'rel' is the joinrel under consideration. + * 'outer_path', 'inner_path' are (probably also parameterized) Paths that + * produce the relations being joined. + * 'sjinfo' is any SpecialJoinInfo relevant to this join. + * 'restrict_clauses' lists the join clauses that need to be applied at the + * join node (including any movable clauses that were moved down to this join, + * and not including any movable clauses that were pushed down into the + * child paths). + * + * set_joinrel_size_estimates must have been applied already. + */ +double +get_parameterized_joinrel_size(PlannerInfo *root, RelOptInfo *rel, + Path *outer_path, + Path *inner_path, + SpecialJoinInfo *sjinfo, + List *restrict_clauses) +{ + double nrows; + + /* + * Estimate the number of rows returned by the parameterized join as the + * sizes of the input paths times the selectivity of the clauses that have + * ended up at this join node. + * + * As with set_joinrel_size_estimates, the rowcount estimate could depend + * on the pair of input paths provided, though ideally we'd get the same + * estimate for any pair with the same parameterization. + */ + nrows = calc_joinrel_size_estimate(root, + rel, + outer_path->parent, + inner_path->parent, + outer_path->rows, + inner_path->rows, + sjinfo, + restrict_clauses); + /* For safety, make sure result is not more than the base estimate */ + if (nrows > rel->rows) + nrows = rel->rows; + return nrows; +} + +/* + * calc_joinrel_size_estimate + * Workhorse for set_joinrel_size_estimates and + * get_parameterized_joinrel_size. + * + * outer_rel/inner_rel are the relations being joined, but they should be + * assumed to have sizes outer_rows/inner_rows; those numbers might be less + * than what rel->rows says, when we are considering parameterized paths. + */ +static double +calc_joinrel_size_estimate(PlannerInfo *root, + RelOptInfo *joinrel, + RelOptInfo *outer_rel, + RelOptInfo *inner_rel, + double outer_rows, + double inner_rows, + SpecialJoinInfo *sjinfo, + List *restrictlist_in) +{ + /* This apparently-useless variable dodges a compiler bug in VS2013: */ + List *restrictlist = restrictlist_in; + JoinType jointype = sjinfo->jointype; + Selectivity fkselec; + Selectivity jselec; + Selectivity pselec; + double nrows; + + /* + * Compute joinclause selectivity. Note that we are only considering + * clauses that become restriction clauses at this join level; we are not + * double-counting them because they were not considered in estimating the + * sizes of the component rels. + * + * First, see whether any of the joinclauses can be matched to known FK + * constraints. If so, drop those clauses from the restrictlist, and + * instead estimate their selectivity using FK semantics. (We do this + * without regard to whether said clauses are local or "pushed down". + * Probably, an FK-matching clause could never be seen as pushed down at + * an outer join, since it would be strict and hence would be grounds for + * join strength reduction.) fkselec gets the net selectivity for + * FK-matching clauses, or 1.0 if there are none. + */ + fkselec = get_foreign_key_join_selectivity(root, + outer_rel->relids, + inner_rel->relids, + sjinfo, + &restrictlist); + + /* + * For an outer join, we have to distinguish the selectivity of the join's + * own clauses (JOIN/ON conditions) from any clauses that were "pushed + * down". For inner joins we just count them all as joinclauses. + */ + if (IS_OUTER_JOIN(jointype)) + { + List *joinquals = NIL; + List *pushedquals = NIL; + ListCell *l; + + /* Grovel through the clauses to separate into two lists */ + foreach(l, restrictlist) + { + RestrictInfo *rinfo = lfirst_node(RestrictInfo, l); + + if (RINFO_IS_PUSHED_DOWN(rinfo, joinrel->relids)) + pushedquals = lappend(pushedquals, rinfo); + else + joinquals = lappend(joinquals, rinfo); + } + + /* Get the separate selectivities */ + jselec = clauselist_selectivity(root, + joinquals, + 0, + jointype, + sjinfo); + pselec = clauselist_selectivity(root, + pushedquals, + 0, + jointype, + sjinfo); + + /* Avoid leaking a lot of ListCells */ + list_free(joinquals); + list_free(pushedquals); + } + else + { + jselec = clauselist_selectivity(root, + restrictlist, + 0, + jointype, + sjinfo); + pselec = 0.0; /* not used, keep compiler quiet */ + } + + /* + * Basically, we multiply size of Cartesian product by selectivity. + * + * If we are doing an outer join, take that into account: the joinqual + * selectivity has to be clamped using the knowledge that the output must + * be at least as large as the non-nullable input. However, any + * pushed-down quals are applied after the outer join, so their + * selectivity applies fully. + * + * For JOIN_SEMI and JOIN_ANTI, the selectivity is defined as the fraction + * of LHS rows that have matches, and we apply that straightforwardly. + */ + switch (jointype) + { + case JOIN_INNER: + nrows = outer_rows * inner_rows * fkselec * jselec; + /* pselec not used */ + break; + case JOIN_LEFT: + nrows = outer_rows * inner_rows * fkselec * jselec; + if (nrows < outer_rows) + nrows = outer_rows; + nrows *= pselec; + break; + case JOIN_FULL: + nrows = outer_rows * inner_rows * fkselec * jselec; + if (nrows < outer_rows) + nrows = outer_rows; + if (nrows < inner_rows) + nrows = inner_rows; + nrows *= pselec; + break; + case JOIN_SEMI: + nrows = outer_rows * fkselec * jselec; + /* pselec not used */ + break; + case JOIN_ANTI: + nrows = outer_rows * (1.0 - fkselec * jselec); + nrows *= pselec; + break; + default: + /* other values not expected here */ + elog(ERROR, "unrecognized join type: %d", (int) jointype); + nrows = 0; /* keep compiler quiet */ + break; + } + + return clamp_row_est(nrows); +} + +/* + * get_foreign_key_join_selectivity + * Estimate join selectivity for foreign-key-related clauses. + * + * Remove any clauses that can be matched to FK constraints from *restrictlist, + * and return a substitute estimate of their selectivity. 1.0 is returned + * when there are no such clauses. + * + * The reason for treating such clauses specially is that we can get better + * estimates this way than by relying on clauselist_selectivity(), especially + * for multi-column FKs where that function's assumption that the clauses are + * independent falls down badly. But even with single-column FKs, we may be + * able to get a better answer when the pg_statistic stats are missing or out + * of date. + */ +static Selectivity +get_foreign_key_join_selectivity(PlannerInfo *root, + Relids outer_relids, + Relids inner_relids, + SpecialJoinInfo *sjinfo, + List **restrictlist) +{ + Selectivity fkselec = 1.0; + JoinType jointype = sjinfo->jointype; + List *worklist = *restrictlist; + ListCell *lc; + + /* Consider each FK constraint that is known to match the query */ + foreach(lc, root->fkey_list) + { + ForeignKeyOptInfo *fkinfo = (ForeignKeyOptInfo *) lfirst(lc); + bool ref_is_outer; + List *removedlist; + ListCell *cell; + + /* + * This FK is not relevant unless it connects a baserel on one side of + * this join to a baserel on the other side. + */ + if (bms_is_member(fkinfo->con_relid, outer_relids) && + bms_is_member(fkinfo->ref_relid, inner_relids)) + ref_is_outer = false; + else if (bms_is_member(fkinfo->ref_relid, outer_relids) && + bms_is_member(fkinfo->con_relid, inner_relids)) + ref_is_outer = true; + else + continue; + + /* + * If we're dealing with a semi/anti join, and the FK's referenced + * relation is on the outside, then knowledge of the FK doesn't help + * us figure out what we need to know (which is the fraction of outer + * rows that have matches). On the other hand, if the referenced rel + * is on the inside, then all outer rows must have matches in the + * referenced table (ignoring nulls). But any restriction or join + * clauses that filter that table will reduce the fraction of matches. + * We can account for restriction clauses, but it's too hard to guess + * how many table rows would get through a join that's inside the RHS. + * Hence, if either case applies, punt and ignore the FK. + */ + if ((jointype == JOIN_SEMI || jointype == JOIN_ANTI) && + (ref_is_outer || bms_membership(inner_relids) != BMS_SINGLETON)) + continue; + + /* + * Modify the restrictlist by removing clauses that match the FK (and + * putting them into removedlist instead). It seems unsafe to modify + * the originally-passed List structure, so we make a shallow copy the + * first time through. + */ + if (worklist == *restrictlist) + worklist = list_copy(worklist); + + removedlist = NIL; + foreach(cell, worklist) + { + RestrictInfo *rinfo = (RestrictInfo *) lfirst(cell); + bool remove_it = false; + int i; + + /* Drop this clause if it matches any column of the FK */ + for (i = 0; i < fkinfo->nkeys; i++) + { + if (rinfo->parent_ec) + { + /* + * EC-derived clauses can only match by EC. It is okay to + * consider any clause derived from the same EC as + * matching the FK: even if equivclass.c chose to generate + * a clause equating some other pair of Vars, it could + * have generated one equating the FK's Vars. So for + * purposes of estimation, we can act as though it did so. + * + * Note: checking parent_ec is a bit of a cheat because + * there are EC-derived clauses that don't have parent_ec + * set; but such clauses must compare expressions that + * aren't just Vars, so they cannot match the FK anyway. + */ + if (fkinfo->eclass[i] == rinfo->parent_ec) + { + remove_it = true; + break; + } + } + else + { + /* + * Otherwise, see if rinfo was previously matched to FK as + * a "loose" clause. + */ + if (list_member_ptr(fkinfo->rinfos[i], rinfo)) + { + remove_it = true; + break; + } + } + } + if (remove_it) + { + worklist = foreach_delete_current(worklist, cell); + removedlist = lappend(removedlist, rinfo); + } + } + + /* + * If we failed to remove all the matching clauses we expected to + * find, chicken out and ignore this FK; applying its selectivity + * might result in double-counting. Put any clauses we did manage to + * remove back into the worklist. + * + * Since the matching clauses are known not outerjoin-delayed, they + * would normally have appeared in the initial joinclause list. If we + * didn't find them, there are two possibilities: + * + * 1. If the FK match is based on an EC that is ec_has_const, it won't + * have generated any join clauses at all. We discount such ECs while + * checking to see if we have "all" the clauses. (Below, we'll adjust + * the selectivity estimate for this case.) + * + * 2. The clauses were matched to some other FK in a previous + * iteration of this loop, and thus removed from worklist. (A likely + * case is that two FKs are matched to the same EC; there will be only + * one EC-derived clause in the initial list, so the first FK will + * consume it.) Applying both FKs' selectivity independently risks + * underestimating the join size; in particular, this would undo one + * of the main things that ECs were invented for, namely to avoid + * double-counting the selectivity of redundant equality conditions. + * Later we might think of a reasonable way to combine the estimates, + * but for now, just punt, since this is a fairly uncommon situation. + */ + if (removedlist == NIL || + list_length(removedlist) != + (fkinfo->nmatched_ec - fkinfo->nconst_ec + fkinfo->nmatched_ri)) + { + worklist = list_concat(worklist, removedlist); + continue; + } + + /* + * Finally we get to the payoff: estimate selectivity using the + * knowledge that each referencing row will match exactly one row in + * the referenced table. + * + * XXX that's not true in the presence of nulls in the referencing + * column(s), so in principle we should derate the estimate for those. + * However (1) if there are any strict restriction clauses for the + * referencing column(s) elsewhere in the query, derating here would + * be double-counting the null fraction, and (2) it's not very clear + * how to combine null fractions for multiple referencing columns. So + * we do nothing for now about correcting for nulls. + * + * XXX another point here is that if either side of an FK constraint + * is an inheritance parent, we estimate as though the constraint + * covers all its children as well. This is not an unreasonable + * assumption for a referencing table, ie the user probably applied + * identical constraints to all child tables (though perhaps we ought + * to check that). But it's not possible to have done that for a + * referenced table. Fortunately, precisely because that doesn't + * work, it is uncommon in practice to have an FK referencing a parent + * table. So, at least for now, disregard inheritance here. + */ + if (jointype == JOIN_SEMI || jointype == JOIN_ANTI) + { + /* + * For JOIN_SEMI and JOIN_ANTI, we only get here when the FK's + * referenced table is exactly the inside of the join. The join + * selectivity is defined as the fraction of LHS rows that have + * matches. The FK implies that every LHS row has a match *in the + * referenced table*; but any restriction clauses on it will + * reduce the number of matches. Hence we take the join + * selectivity as equal to the selectivity of the table's + * restriction clauses, which is rows / tuples; but we must guard + * against tuples == 0. + */ + RelOptInfo *ref_rel = find_base_rel(root, fkinfo->ref_relid); + double ref_tuples = Max(ref_rel->tuples, 1.0); + + fkselec *= ref_rel->rows / ref_tuples; + } + else + { + /* + * Otherwise, selectivity is exactly 1/referenced-table-size; but + * guard against tuples == 0. Note we should use the raw table + * tuple count, not any estimate of its filtered or joined size. + */ + RelOptInfo *ref_rel = find_base_rel(root, fkinfo->ref_relid); + double ref_tuples = Max(ref_rel->tuples, 1.0); + + fkselec *= 1.0 / ref_tuples; + } + + /* + * If any of the FK columns participated in ec_has_const ECs, then + * equivclass.c will have generated "var = const" restrictions for + * each side of the join, thus reducing the sizes of both input + * relations. Taking the fkselec at face value would amount to + * double-counting the selectivity of the constant restriction for the + * referencing Var. Hence, look for the restriction clause(s) that + * were applied to the referencing Var(s), and divide out their + * selectivity to correct for this. + */ + if (fkinfo->nconst_ec > 0) + { + for (int i = 0; i < fkinfo->nkeys; i++) + { + EquivalenceClass *ec = fkinfo->eclass[i]; + + if (ec && ec->ec_has_const) + { + EquivalenceMember *em = fkinfo->fk_eclass_member[i]; + RestrictInfo *rinfo = find_derived_clause_for_ec_member(ec, + em); + + if (rinfo) + { + Selectivity s0; + + s0 = clause_selectivity(root, + (Node *) rinfo, + 0, + jointype, + sjinfo); + if (s0 > 0) + fkselec /= s0; + } + } + } + } + } + + *restrictlist = worklist; + CLAMP_PROBABILITY(fkselec); + return fkselec; +} + +/* + * set_subquery_size_estimates + * Set the size estimates for a base relation that is a subquery. + * + * The rel's targetlist and restrictinfo list must have been constructed + * already, and the Paths for the subquery must have been completed. + * We look at the subquery's PlannerInfo to extract data. + * + * We set the same fields as set_baserel_size_estimates. + */ +void +set_subquery_size_estimates(PlannerInfo *root, RelOptInfo *rel) +{ + PlannerInfo *subroot = rel->subroot; + RelOptInfo *sub_final_rel; + ListCell *lc; + + /* Should only be applied to base relations that are subqueries */ + Assert(rel->relid > 0); + Assert(planner_rt_fetch(rel->relid, root)->rtekind == RTE_SUBQUERY); + + /* + * Copy raw number of output rows from subquery. All of its paths should + * have the same output rowcount, so just look at cheapest-total. + */ + sub_final_rel = fetch_upper_rel(subroot, UPPERREL_FINAL, NULL); + rel->tuples = sub_final_rel->cheapest_total_path->rows; + + /* + * Compute per-output-column width estimates by examining the subquery's + * targetlist. For any output that is a plain Var, get the width estimate + * that was made while planning the subquery. Otherwise, we leave it to + * set_rel_width to fill in a datatype-based default estimate. + */ + foreach(lc, subroot->parse->targetList) + { + TargetEntry *te = lfirst_node(TargetEntry, lc); + Node *texpr = (Node *) te->expr; + int32 item_width = 0; + + /* junk columns aren't visible to upper query */ + if (te->resjunk) + continue; + + /* + * The subquery could be an expansion of a view that's had columns + * added to it since the current query was parsed, so that there are + * non-junk tlist columns in it that don't correspond to any column + * visible at our query level. Ignore such columns. + */ + if (te->resno < rel->min_attr || te->resno > rel->max_attr) + continue; + + /* + * XXX This currently doesn't work for subqueries containing set + * operations, because the Vars in their tlists are bogus references + * to the first leaf subquery, which wouldn't give the right answer + * even if we could still get to its PlannerInfo. + * + * Also, the subquery could be an appendrel for which all branches are + * known empty due to constraint exclusion, in which case + * set_append_rel_pathlist will have left the attr_widths set to zero. + * + * In either case, we just leave the width estimate zero until + * set_rel_width fixes it. + */ + if (IsA(texpr, Var) && + subroot->parse->setOperations == NULL) + { + Var *var = (Var *) texpr; + RelOptInfo *subrel = find_base_rel(subroot, var->varno); + + item_width = subrel->attr_widths[var->varattno - subrel->min_attr]; + } + rel->attr_widths[te->resno - rel->min_attr] = item_width; + } + + /* Now estimate number of output rows, etc */ + set_baserel_size_estimates(root, rel); +} + +/* + * set_function_size_estimates + * Set the size estimates for a base relation that is a function call. + * + * The rel's targetlist and restrictinfo list must have been constructed + * already. + * + * We set the same fields as set_baserel_size_estimates. + */ +void +set_function_size_estimates(PlannerInfo *root, RelOptInfo *rel) +{ + RangeTblEntry *rte; + ListCell *lc; + + /* Should only be applied to base relations that are functions */ + Assert(rel->relid > 0); + rte = planner_rt_fetch(rel->relid, root); + Assert(rte->rtekind == RTE_FUNCTION); + + /* + * Estimate number of rows the functions will return. The rowcount of the + * node is that of the largest function result. + */ + rel->tuples = 0; + foreach(lc, rte->functions) + { + RangeTblFunction *rtfunc = (RangeTblFunction *) lfirst(lc); + double ntup = expression_returns_set_rows(root, rtfunc->funcexpr); + + if (ntup > rel->tuples) + rel->tuples = ntup; + } + + /* Now estimate number of output rows, etc */ + set_baserel_size_estimates(root, rel); +} + +/* + * set_function_size_estimates + * Set the size estimates for a base relation that is a function call. + * + * The rel's targetlist and restrictinfo list must have been constructed + * already. + * + * We set the same fields as set_tablefunc_size_estimates. + */ +void +set_tablefunc_size_estimates(PlannerInfo *root, RelOptInfo *rel) +{ + /* Should only be applied to base relations that are functions */ + Assert(rel->relid > 0); + Assert(planner_rt_fetch(rel->relid, root)->rtekind == RTE_TABLEFUNC); + + rel->tuples = 100; + + /* Now estimate number of output rows, etc */ + set_baserel_size_estimates(root, rel); +} + +/* + * set_values_size_estimates + * Set the size estimates for a base relation that is a values list. + * + * The rel's targetlist and restrictinfo list must have been constructed + * already. + * + * We set the same fields as set_baserel_size_estimates. + */ +void +set_values_size_estimates(PlannerInfo *root, RelOptInfo *rel) +{ + RangeTblEntry *rte; + + /* Should only be applied to base relations that are values lists */ + Assert(rel->relid > 0); + rte = planner_rt_fetch(rel->relid, root); + Assert(rte->rtekind == RTE_VALUES); + + /* + * Estimate number of rows the values list will return. We know this + * precisely based on the list length (well, barring set-returning + * functions in list items, but that's a refinement not catered for + * anywhere else either). + */ + rel->tuples = list_length(rte->values_lists); + + /* Now estimate number of output rows, etc */ + set_baserel_size_estimates(root, rel); +} + +/* + * set_cte_size_estimates + * Set the size estimates for a base relation that is a CTE reference. + * + * The rel's targetlist and restrictinfo list must have been constructed + * already, and we need an estimate of the number of rows returned by the CTE + * (if a regular CTE) or the non-recursive term (if a self-reference). + * + * We set the same fields as set_baserel_size_estimates. + */ +void +set_cte_size_estimates(PlannerInfo *root, RelOptInfo *rel, double cte_rows) +{ + RangeTblEntry *rte; + + /* Should only be applied to base relations that are CTE references */ + Assert(rel->relid > 0); + rte = planner_rt_fetch(rel->relid, root); + Assert(rte->rtekind == RTE_CTE); + + if (rte->self_reference) + { + /* + * In a self-reference, we assume the average worktable size is a + * multiple of the nonrecursive term's size. The best multiplier will + * vary depending on query "fan-out", so make its value adjustable. + */ + rel->tuples = clamp_row_est(recursive_worktable_factor * cte_rows); + } + else + { + /* Otherwise just believe the CTE's rowcount estimate */ + rel->tuples = cte_rows; + } + + /* Now estimate number of output rows, etc */ + set_baserel_size_estimates(root, rel); +} + +/* + * set_namedtuplestore_size_estimates + * Set the size estimates for a base relation that is a tuplestore reference. + * + * The rel's targetlist and restrictinfo list must have been constructed + * already. + * + * We set the same fields as set_baserel_size_estimates. + */ +void +set_namedtuplestore_size_estimates(PlannerInfo *root, RelOptInfo *rel) +{ + RangeTblEntry *rte; + + /* Should only be applied to base relations that are tuplestore references */ + Assert(rel->relid > 0); + rte = planner_rt_fetch(rel->relid, root); + Assert(rte->rtekind == RTE_NAMEDTUPLESTORE); + + /* + * Use the estimate provided by the code which is generating the named + * tuplestore. In some cases, the actual number might be available; in + * others the same plan will be re-used, so a "typical" value might be + * estimated and used. + */ + rel->tuples = rte->enrtuples; + if (rel->tuples < 0) + rel->tuples = 1000; + + /* Now estimate number of output rows, etc */ + set_baserel_size_estimates(root, rel); +} + +/* + * set_result_size_estimates + * Set the size estimates for an RTE_RESULT base relation + * + * The rel's targetlist and restrictinfo list must have been constructed + * already. + * + * We set the same fields as set_baserel_size_estimates. + */ +void +set_result_size_estimates(PlannerInfo *root, RelOptInfo *rel) +{ + /* Should only be applied to RTE_RESULT base relations */ + Assert(rel->relid > 0); + Assert(planner_rt_fetch(rel->relid, root)->rtekind == RTE_RESULT); + + /* RTE_RESULT always generates a single row, natively */ + rel->tuples = 1; + + /* Now estimate number of output rows, etc */ + set_baserel_size_estimates(root, rel); +} + +/* + * set_foreign_size_estimates + * Set the size estimates for a base relation that is a foreign table. + * + * There is not a whole lot that we can do here; the foreign-data wrapper + * is responsible for producing useful estimates. We can do a decent job + * of estimating baserestrictcost, so we set that, and we also set up width + * using what will be purely datatype-driven estimates from the targetlist. + * There is no way to do anything sane with the rows value, so we just put + * a default estimate and hope that the wrapper can improve on it. The + * wrapper's GetForeignRelSize function will be called momentarily. + * + * The rel's targetlist and restrictinfo list must have been constructed + * already. + */ +void +set_foreign_size_estimates(PlannerInfo *root, RelOptInfo *rel) +{ + /* Should only be applied to base relations */ + Assert(rel->relid > 0); + + rel->rows = 1000; /* entirely bogus default estimate */ + + cost_qual_eval(&rel->baserestrictcost, rel->baserestrictinfo, root); + + set_rel_width(root, rel); +} + + +/* + * set_rel_width + * Set the estimated output width of a base relation. + * + * The estimated output width is the sum of the per-attribute width estimates + * for the actually-referenced columns, plus any PHVs or other expressions + * that have to be calculated at this relation. This is the amount of data + * we'd need to pass upwards in case of a sort, hash, etc. + * + * This function also sets reltarget->cost, so it's a bit misnamed now. + * + * NB: this works best on plain relations because it prefers to look at + * real Vars. For subqueries, set_subquery_size_estimates will already have + * copied up whatever per-column estimates were made within the subquery, + * and for other types of rels there isn't much we can do anyway. We fall + * back on (fairly stupid) datatype-based width estimates if we can't get + * any better number. + * + * The per-attribute width estimates are cached for possible re-use while + * building join relations or post-scan/join pathtargets. + */ +static void +set_rel_width(PlannerInfo *root, RelOptInfo *rel) +{ + Oid reloid = planner_rt_fetch(rel->relid, root)->relid; + int32 tuple_width = 0; + bool have_wholerow_var = false; + ListCell *lc; + + /* Vars are assumed to have cost zero, but other exprs do not */ + rel->reltarget->cost.startup = 0; + rel->reltarget->cost.per_tuple = 0; + + foreach(lc, rel->reltarget->exprs) + { + Node *node = (Node *) lfirst(lc); + + /* + * Ordinarily, a Var in a rel's targetlist must belong to that rel; + * but there are corner cases involving LATERAL references where that + * isn't so. If the Var has the wrong varno, fall through to the + * generic case (it doesn't seem worth the trouble to be any smarter). + */ + if (IsA(node, Var) && + ((Var *) node)->varno == rel->relid) + { + Var *var = (Var *) node; + int ndx; + int32 item_width; + + Assert(var->varattno >= rel->min_attr); + Assert(var->varattno <= rel->max_attr); + + ndx = var->varattno - rel->min_attr; + + /* + * If it's a whole-row Var, we'll deal with it below after we have + * already cached as many attr widths as possible. + */ + if (var->varattno == 0) + { + have_wholerow_var = true; + continue; + } + + /* + * The width may have been cached already (especially if it's a + * subquery), so don't duplicate effort. + */ + if (rel->attr_widths[ndx] > 0) + { + tuple_width += rel->attr_widths[ndx]; + continue; + } + + /* Try to get column width from statistics */ + if (reloid != InvalidOid && var->varattno > 0) + { + item_width = get_attavgwidth(reloid, var->varattno); + if (item_width > 0) + { + rel->attr_widths[ndx] = item_width; + tuple_width += item_width; + continue; + } + } + + /* + * Not a plain relation, or can't find statistics for it. Estimate + * using just the type info. + */ + item_width = get_typavgwidth(var->vartype, var->vartypmod); + Assert(item_width > 0); + rel->attr_widths[ndx] = item_width; + tuple_width += item_width; + } + else if (IsA(node, PlaceHolderVar)) + { + /* + * We will need to evaluate the PHV's contained expression while + * scanning this rel, so be sure to include it in reltarget->cost. + */ + PlaceHolderVar *phv = (PlaceHolderVar *) node; + PlaceHolderInfo *phinfo = find_placeholder_info(root, phv, false); + QualCost cost; + + tuple_width += phinfo->ph_width; + cost_qual_eval_node(&cost, (Node *) phv->phexpr, root); + rel->reltarget->cost.startup += cost.startup; + rel->reltarget->cost.per_tuple += cost.per_tuple; + } + else + { + /* + * We could be looking at an expression pulled up from a subquery, + * or a ROW() representing a whole-row child Var, etc. Do what we + * can using the expression type information. + */ + int32 item_width; + QualCost cost; + + item_width = get_typavgwidth(exprType(node), exprTypmod(node)); + Assert(item_width > 0); + tuple_width += item_width; + /* Not entirely clear if we need to account for cost, but do so */ + cost_qual_eval_node(&cost, node, root); + rel->reltarget->cost.startup += cost.startup; + rel->reltarget->cost.per_tuple += cost.per_tuple; + } + } + + /* + * If we have a whole-row reference, estimate its width as the sum of + * per-column widths plus heap tuple header overhead. + */ + if (have_wholerow_var) + { + int32 wholerow_width = MAXALIGN(SizeofHeapTupleHeader); + + if (reloid != InvalidOid) + { + /* Real relation, so estimate true tuple width */ + wholerow_width += get_relation_data_width(reloid, + rel->attr_widths - rel->min_attr); + } + else + { + /* Do what we can with info for a phony rel */ + AttrNumber i; + + for (i = 1; i <= rel->max_attr; i++) + wholerow_width += rel->attr_widths[i - rel->min_attr]; + } + + rel->attr_widths[0 - rel->min_attr] = wholerow_width; + + /* + * Include the whole-row Var as part of the output tuple. Yes, that + * really is what happens at runtime. + */ + tuple_width += wholerow_width; + } + + Assert(tuple_width >= 0); + rel->reltarget->width = tuple_width; +} + +/* + * set_pathtarget_cost_width + * Set the estimated eval cost and output width of a PathTarget tlist. + * + * As a notational convenience, returns the same PathTarget pointer passed in. + * + * Most, though not quite all, uses of this function occur after we've run + * set_rel_width() for base relations; so we can usually obtain cached width + * estimates for Vars. If we can't, fall back on datatype-based width + * estimates. Present early-planning uses of PathTargets don't need accurate + * widths badly enough to justify going to the catalogs for better data. + */ +PathTarget * +set_pathtarget_cost_width(PlannerInfo *root, PathTarget *target) +{ + int32 tuple_width = 0; + ListCell *lc; + + /* Vars are assumed to have cost zero, but other exprs do not */ + target->cost.startup = 0; + target->cost.per_tuple = 0; + + foreach(lc, target->exprs) + { + Node *node = (Node *) lfirst(lc); + + if (IsA(node, Var)) + { + Var *var = (Var *) node; + int32 item_width; + + /* We should not see any upper-level Vars here */ + Assert(var->varlevelsup == 0); + + /* Try to get data from RelOptInfo cache */ + if (!IS_SPECIAL_VARNO(var->varno) && + var->varno < root->simple_rel_array_size) + { + RelOptInfo *rel = root->simple_rel_array[var->varno]; + + if (rel != NULL && + var->varattno >= rel->min_attr && + var->varattno <= rel->max_attr) + { + int ndx = var->varattno - rel->min_attr; + + if (rel->attr_widths[ndx] > 0) + { + tuple_width += rel->attr_widths[ndx]; + continue; + } + } + } + + /* + * No cached data available, so estimate using just the type info. + */ + item_width = get_typavgwidth(var->vartype, var->vartypmod); + Assert(item_width > 0); + tuple_width += item_width; + } + else + { + /* + * Handle general expressions using type info. + */ + int32 item_width; + QualCost cost; + + item_width = get_typavgwidth(exprType(node), exprTypmod(node)); + Assert(item_width > 0); + tuple_width += item_width; + + /* Account for cost, too */ + cost_qual_eval_node(&cost, node, root); + target->cost.startup += cost.startup; + target->cost.per_tuple += cost.per_tuple; + } + } + + Assert(tuple_width >= 0); + target->width = tuple_width; + + return target; +} + +/* + * relation_byte_size + * Estimate the storage space in bytes for a given number of tuples + * of a given width (size in bytes). + */ +static double +relation_byte_size(double tuples, int width) +{ + return tuples * (MAXALIGN(width) + MAXALIGN(SizeofHeapTupleHeader)); +} + +/* + * page_size + * Returns an estimate of the number of pages covered by a given + * number of tuples of a given width (size in bytes). + */ +static double +page_size(double tuples, int width) +{ + return ceil(relation_byte_size(tuples, width) / BLCKSZ); +} + +/* + * Estimate the fraction of the work that each worker will do given the + * number of workers budgeted for the path. + */ +static double +get_parallel_divisor(Path *path) +{ + double parallel_divisor = path->parallel_workers; + + /* + * Early experience with parallel query suggests that when there is only + * one worker, the leader often makes a very substantial contribution to + * executing the parallel portion of the plan, but as more workers are + * added, it does less and less, because it's busy reading tuples from the + * workers and doing whatever non-parallel post-processing is needed. By + * the time we reach 4 workers, the leader no longer makes a meaningful + * contribution. Thus, for now, estimate that the leader spends 30% of + * its time servicing each worker, and the remainder executing the + * parallel plan. + */ + if (parallel_leader_participation) + { + double leader_contribution; + + leader_contribution = 1.0 - (0.3 * path->parallel_workers); + if (leader_contribution > 0) + parallel_divisor += leader_contribution; + } + + return parallel_divisor; +} + +/* + * compute_bitmap_pages + * + * compute number of pages fetched from heap in bitmap heap scan. + */ +double +compute_bitmap_pages(PlannerInfo *root, RelOptInfo *baserel, Path *bitmapqual, + int loop_count, Cost *cost, double *tuple) +{ + Cost indexTotalCost; + Selectivity indexSelectivity; + double T; + double pages_fetched; + double tuples_fetched; + double heap_pages; + long maxentries; + + /* + * Fetch total cost of obtaining the bitmap, as well as its total + * selectivity. + */ + cost_bitmap_tree_node(bitmapqual, &indexTotalCost, &indexSelectivity); + + /* + * Estimate number of main-table pages fetched. + */ + tuples_fetched = clamp_row_est(indexSelectivity * baserel->tuples); + + T = (baserel->pages > 1) ? (double) baserel->pages : 1.0; + + /* + * For a single scan, the number of heap pages that need to be fetched is + * the same as the Mackert and Lohman formula for the case T <= b (ie, no + * re-reads needed). + */ + pages_fetched = (2.0 * T * tuples_fetched) / (2.0 * T + tuples_fetched); + + /* + * Calculate the number of pages fetched from the heap. Then based on + * current work_mem estimate get the estimated maxentries in the bitmap. + * (Note that we always do this calculation based on the number of pages + * that would be fetched in a single iteration, even if loop_count > 1. + * That's correct, because only that number of entries will be stored in + * the bitmap at one time.) + */ + heap_pages = Min(pages_fetched, baserel->pages); + maxentries = tbm_calculate_entries(work_mem * 1024L); + + if (loop_count > 1) + { + /* + * For repeated bitmap scans, scale up the number of tuples fetched in + * the Mackert and Lohman formula by the number of scans, so that we + * estimate the number of pages fetched by all the scans. Then + * pro-rate for one scan. + */ + pages_fetched = index_pages_fetched(tuples_fetched * loop_count, + baserel->pages, + get_indexpath_pages(bitmapqual), + root); + pages_fetched /= loop_count; + } + + if (pages_fetched >= T) + pages_fetched = T; + else + pages_fetched = ceil(pages_fetched); + + if (maxentries < heap_pages) + { + double exact_pages; + double lossy_pages; + + /* + * Crude approximation of the number of lossy pages. Because of the + * way tbm_lossify() is coded, the number of lossy pages increases + * very sharply as soon as we run short of memory; this formula has + * that property and seems to perform adequately in testing, but it's + * possible we could do better somehow. + */ + lossy_pages = Max(0, heap_pages - maxentries / 2); + exact_pages = heap_pages - lossy_pages; + + /* + * If there are lossy pages then recompute the number of tuples + * processed by the bitmap heap node. We assume here that the chance + * of a given tuple coming from an exact page is the same as the + * chance that a given page is exact. This might not be true, but + * it's not clear how we can do any better. + */ + if (lossy_pages > 0) + tuples_fetched = + clamp_row_est(indexSelectivity * + (exact_pages / heap_pages) * baserel->tuples + + (lossy_pages / heap_pages) * baserel->tuples); + } + + if (cost) + *cost = indexTotalCost; + if (tuple) + *tuple = tuples_fetched; + + return pages_fetched; +} diff --git a/src/backend/optimizer/path/equivclass.c b/src/backend/optimizer/path/equivclass.c new file mode 100644 index 0000000..9f39f46 --- /dev/null +++ b/src/backend/optimizer/path/equivclass.c @@ -0,0 +1,3226 @@ +/*------------------------------------------------------------------------- + * + * equivclass.c + * Routines for managing EquivalenceClasses + * + * See src/backend/optimizer/README for discussion of EquivalenceClasses. + * + * + * Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group + * Portions Copyright (c) 1994, Regents of the University of California + * + * IDENTIFICATION + * src/backend/optimizer/path/equivclass.c + * + *------------------------------------------------------------------------- + */ +#include "postgres.h" + +#include <limits.h> + +#include "access/stratnum.h" +#include "catalog/pg_type.h" +#include "nodes/makefuncs.h" +#include "nodes/nodeFuncs.h" +#include "optimizer/appendinfo.h" +#include "optimizer/clauses.h" +#include "optimizer/optimizer.h" +#include "optimizer/pathnode.h" +#include "optimizer/paths.h" +#include "optimizer/planmain.h" +#include "optimizer/restrictinfo.h" +#include "utils/lsyscache.h" + + +static EquivalenceMember *add_eq_member(EquivalenceClass *ec, + Expr *expr, Relids relids, Relids nullable_relids, + bool is_child, Oid datatype); +static bool is_exprlist_member(Expr *node, List *exprs); +static void generate_base_implied_equalities_const(PlannerInfo *root, + EquivalenceClass *ec); +static void generate_base_implied_equalities_no_const(PlannerInfo *root, + EquivalenceClass *ec); +static void generate_base_implied_equalities_broken(PlannerInfo *root, + EquivalenceClass *ec); +static List *generate_join_implied_equalities_normal(PlannerInfo *root, + EquivalenceClass *ec, + Relids join_relids, + Relids outer_relids, + Relids inner_relids); +static List *generate_join_implied_equalities_broken(PlannerInfo *root, + EquivalenceClass *ec, + Relids nominal_join_relids, + Relids outer_relids, + Relids nominal_inner_relids, + RelOptInfo *inner_rel); +static Oid select_equality_operator(EquivalenceClass *ec, + Oid lefttype, Oid righttype); +static RestrictInfo *create_join_clause(PlannerInfo *root, + EquivalenceClass *ec, Oid opno, + EquivalenceMember *leftem, + EquivalenceMember *rightem, + EquivalenceClass *parent_ec); +static bool reconsider_outer_join_clause(PlannerInfo *root, + RestrictInfo *rinfo, + bool outer_on_left); +static bool reconsider_full_join_clause(PlannerInfo *root, + RestrictInfo *rinfo); +static Bitmapset *get_eclass_indexes_for_relids(PlannerInfo *root, + Relids relids); +static Bitmapset *get_common_eclass_indexes(PlannerInfo *root, Relids relids1, + Relids relids2); + + +/* + * process_equivalence + * The given clause has a mergejoinable operator and can be applied without + * any delay by an outer join, so its two sides can be considered equal + * anywhere they are both computable; moreover that equality can be + * extended transitively. Record this knowledge in the EquivalenceClass + * data structure, if applicable. Returns true if successful, false if not + * (in which case caller should treat the clause as ordinary, not an + * equivalence). + * + * In some cases, although we cannot convert a clause into EquivalenceClass + * knowledge, we can still modify it to a more useful form than the original. + * Then, *p_restrictinfo will be replaced by a new RestrictInfo, which is what + * the caller should use for further processing. + * + * If below_outer_join is true, then the clause was found below the nullable + * side of an outer join, so its sides might validly be both NULL rather than + * strictly equal. We can still deduce equalities in such cases, but we take + * care to mark an EquivalenceClass if it came from any such clauses. Also, + * we have to check that both sides are either pseudo-constants or strict + * functions of Vars, else they might not both go to NULL above the outer + * join. (This is the main reason why we need a failure return. It's more + * convenient to check this case here than at the call sites...) + * + * We also reject proposed equivalence clauses if they contain leaky functions + * and have security_level above zero. The EC evaluation rules require us to + * apply certain tests at certain joining levels, and we can't tolerate + * delaying any test on security_level grounds. By rejecting candidate clauses + * that might require security delays, we ensure it's safe to apply an EC + * clause as soon as it's supposed to be applied. + * + * On success return, we have also initialized the clause's left_ec/right_ec + * fields to point to the EquivalenceClass representing it. This saves lookup + * effort later. + * + * Note: constructing merged EquivalenceClasses is a standard UNION-FIND + * problem, for which there exist better data structures than simple lists. + * If this code ever proves to be a bottleneck then it could be sped up --- + * but for now, simple is beautiful. + * + * Note: this is only called during planner startup, not during GEQO + * exploration, so we need not worry about whether we're in the right + * memory context. + */ +bool +process_equivalence(PlannerInfo *root, + RestrictInfo **p_restrictinfo, + bool below_outer_join) +{ + RestrictInfo *restrictinfo = *p_restrictinfo; + Expr *clause = restrictinfo->clause; + Oid opno, + collation, + item1_type, + item2_type; + Expr *item1; + Expr *item2; + Relids item1_relids, + item2_relids, + item1_nullable_relids, + item2_nullable_relids; + List *opfamilies; + EquivalenceClass *ec1, + *ec2; + EquivalenceMember *em1, + *em2; + ListCell *lc1; + int ec2_idx; + + /* Should not already be marked as having generated an eclass */ + Assert(restrictinfo->left_ec == NULL); + Assert(restrictinfo->right_ec == NULL); + + /* Reject if it is potentially postponable by security considerations */ + if (restrictinfo->security_level > 0 && !restrictinfo->leakproof) + return false; + + /* Extract info from given clause */ + Assert(is_opclause(clause)); + opno = ((OpExpr *) clause)->opno; + collation = ((OpExpr *) clause)->inputcollid; + item1 = (Expr *) get_leftop(clause); + item2 = (Expr *) get_rightop(clause); + item1_relids = restrictinfo->left_relids; + item2_relids = restrictinfo->right_relids; + + /* + * Ensure both input expressions expose the desired collation (their types + * should be OK already); see comments for canonicalize_ec_expression. + */ + item1 = canonicalize_ec_expression(item1, + exprType((Node *) item1), + collation); + item2 = canonicalize_ec_expression(item2, + exprType((Node *) item2), + collation); + + /* + * Clauses of the form X=X cannot be translated into EquivalenceClasses. + * We'd either end up with a single-entry EC, losing the knowledge that + * the clause was present at all, or else make an EC with duplicate + * entries, causing other issues. + */ + if (equal(item1, item2)) + { + /* + * If the operator is strict, then the clause can be treated as just + * "X IS NOT NULL". (Since we know we are considering a top-level + * qual, we can ignore the difference between FALSE and NULL results.) + * It's worth making the conversion because we'll typically get a much + * better selectivity estimate than we would for X=X. + * + * If the operator is not strict, we can't be sure what it will do + * with NULLs, so don't attempt to optimize it. + */ + set_opfuncid((OpExpr *) clause); + if (func_strict(((OpExpr *) clause)->opfuncid)) + { + NullTest *ntest = makeNode(NullTest); + + ntest->arg = item1; + ntest->nulltesttype = IS_NOT_NULL; + ntest->argisrow = false; /* correct even if composite arg */ + ntest->location = -1; + + *p_restrictinfo = + make_restrictinfo(root, + (Expr *) ntest, + restrictinfo->is_pushed_down, + restrictinfo->outerjoin_delayed, + restrictinfo->pseudoconstant, + restrictinfo->security_level, + NULL, + restrictinfo->outer_relids, + restrictinfo->nullable_relids); + } + return false; + } + + /* + * If below outer join, check for strictness, else reject. + */ + if (below_outer_join) + { + if (!bms_is_empty(item1_relids) && + contain_nonstrict_functions((Node *) item1)) + return false; /* LHS is non-strict but not constant */ + if (!bms_is_empty(item2_relids) && + contain_nonstrict_functions((Node *) item2)) + return false; /* RHS is non-strict but not constant */ + } + + /* Calculate nullable-relid sets for each side of the clause */ + item1_nullable_relids = bms_intersect(item1_relids, + restrictinfo->nullable_relids); + item2_nullable_relids = bms_intersect(item2_relids, + restrictinfo->nullable_relids); + + /* + * We use the declared input types of the operator, not exprType() of the + * inputs, as the nominal datatypes for opfamily lookup. This presumes + * that btree operators are always registered with amoplefttype and + * amoprighttype equal to their declared input types. We will need this + * info anyway to build EquivalenceMember nodes, and by extracting it now + * we can use type comparisons to short-circuit some equal() tests. + */ + op_input_types(opno, &item1_type, &item2_type); + + opfamilies = restrictinfo->mergeopfamilies; + + /* + * Sweep through the existing EquivalenceClasses looking for matches to + * item1 and item2. These are the possible outcomes: + * + * 1. We find both in the same EC. The equivalence is already known, so + * there's nothing to do. + * + * 2. We find both in different ECs. Merge the two ECs together. + * + * 3. We find just one. Add the other to its EC. + * + * 4. We find neither. Make a new, two-entry EC. + * + * Note: since all ECs are built through this process or the similar + * search in get_eclass_for_sort_expr(), it's impossible that we'd match + * an item in more than one existing nonvolatile EC. So it's okay to stop + * at the first match. + */ + ec1 = ec2 = NULL; + em1 = em2 = NULL; + ec2_idx = -1; + foreach(lc1, root->eq_classes) + { + EquivalenceClass *cur_ec = (EquivalenceClass *) lfirst(lc1); + ListCell *lc2; + + /* Never match to a volatile EC */ + if (cur_ec->ec_has_volatile) + continue; + + /* + * The collation has to match; check this first since it's cheaper + * than the opfamily comparison. + */ + if (collation != cur_ec->ec_collation) + continue; + + /* + * A "match" requires matching sets of btree opfamilies. Use of + * equal() for this test has implications discussed in the comments + * for get_mergejoin_opfamilies(). + */ + if (!equal(opfamilies, cur_ec->ec_opfamilies)) + continue; + + foreach(lc2, cur_ec->ec_members) + { + EquivalenceMember *cur_em = (EquivalenceMember *) lfirst(lc2); + + Assert(!cur_em->em_is_child); /* no children yet */ + + /* + * If below an outer join, don't match constants: they're not as + * constant as they look. + */ + if ((below_outer_join || cur_ec->ec_below_outer_join) && + cur_em->em_is_const) + continue; + + if (!ec1 && + item1_type == cur_em->em_datatype && + equal(item1, cur_em->em_expr)) + { + ec1 = cur_ec; + em1 = cur_em; + if (ec2) + break; + } + + if (!ec2 && + item2_type == cur_em->em_datatype && + equal(item2, cur_em->em_expr)) + { + ec2 = cur_ec; + ec2_idx = foreach_current_index(lc1); + em2 = cur_em; + if (ec1) + break; + } + } + + if (ec1 && ec2) + break; + } + + /* Sweep finished, what did we find? */ + + if (ec1 && ec2) + { + /* If case 1, nothing to do, except add to sources */ + if (ec1 == ec2) + { + ec1->ec_sources = lappend(ec1->ec_sources, restrictinfo); + ec1->ec_below_outer_join |= below_outer_join; + ec1->ec_min_security = Min(ec1->ec_min_security, + restrictinfo->security_level); + ec1->ec_max_security = Max(ec1->ec_max_security, + restrictinfo->security_level); + /* mark the RI as associated with this eclass */ + restrictinfo->left_ec = ec1; + restrictinfo->right_ec = ec1; + /* mark the RI as usable with this pair of EMs */ + restrictinfo->left_em = em1; + restrictinfo->right_em = em2; + return true; + } + + /* + * Case 2: need to merge ec1 and ec2. This should never happen after + * the ECs have reached canonical state; otherwise, pathkeys could be + * rendered non-canonical by the merge, and relation eclass indexes + * would get broken by removal of an eq_classes list entry. + */ + if (root->ec_merging_done) + elog(ERROR, "too late to merge equivalence classes"); + + /* + * We add ec2's items to ec1, then set ec2's ec_merged link to point + * to ec1 and remove ec2 from the eq_classes list. We cannot simply + * delete ec2 because that could leave dangling pointers in existing + * PathKeys. We leave it behind with a link so that the merged EC can + * be found. + */ + ec1->ec_members = list_concat(ec1->ec_members, ec2->ec_members); + ec1->ec_sources = list_concat(ec1->ec_sources, ec2->ec_sources); + ec1->ec_derives = list_concat(ec1->ec_derives, ec2->ec_derives); + ec1->ec_relids = bms_join(ec1->ec_relids, ec2->ec_relids); + ec1->ec_has_const |= ec2->ec_has_const; + /* can't need to set has_volatile */ + ec1->ec_below_outer_join |= ec2->ec_below_outer_join; + ec1->ec_min_security = Min(ec1->ec_min_security, + ec2->ec_min_security); + ec1->ec_max_security = Max(ec1->ec_max_security, + ec2->ec_max_security); + ec2->ec_merged = ec1; + root->eq_classes = list_delete_nth_cell(root->eq_classes, ec2_idx); + /* just to avoid debugging confusion w/ dangling pointers: */ + ec2->ec_members = NIL; + ec2->ec_sources = NIL; + ec2->ec_derives = NIL; + ec2->ec_relids = NULL; + ec1->ec_sources = lappend(ec1->ec_sources, restrictinfo); + ec1->ec_below_outer_join |= below_outer_join; + ec1->ec_min_security = Min(ec1->ec_min_security, + restrictinfo->security_level); + ec1->ec_max_security = Max(ec1->ec_max_security, + restrictinfo->security_level); + /* mark the RI as associated with this eclass */ + restrictinfo->left_ec = ec1; + restrictinfo->right_ec = ec1; + /* mark the RI as usable with this pair of EMs */ + restrictinfo->left_em = em1; + restrictinfo->right_em = em2; + } + else if (ec1) + { + /* Case 3: add item2 to ec1 */ + em2 = add_eq_member(ec1, item2, item2_relids, item2_nullable_relids, + false, item2_type); + ec1->ec_sources = lappend(ec1->ec_sources, restrictinfo); + ec1->ec_below_outer_join |= below_outer_join; + ec1->ec_min_security = Min(ec1->ec_min_security, + restrictinfo->security_level); + ec1->ec_max_security = Max(ec1->ec_max_security, + restrictinfo->security_level); + /* mark the RI as associated with this eclass */ + restrictinfo->left_ec = ec1; + restrictinfo->right_ec = ec1; + /* mark the RI as usable with this pair of EMs */ + restrictinfo->left_em = em1; + restrictinfo->right_em = em2; + } + else if (ec2) + { + /* Case 3: add item1 to ec2 */ + em1 = add_eq_member(ec2, item1, item1_relids, item1_nullable_relids, + false, item1_type); + ec2->ec_sources = lappend(ec2->ec_sources, restrictinfo); + ec2->ec_below_outer_join |= below_outer_join; + ec2->ec_min_security = Min(ec2->ec_min_security, + restrictinfo->security_level); + ec2->ec_max_security = Max(ec2->ec_max_security, + restrictinfo->security_level); + /* mark the RI as associated with this eclass */ + restrictinfo->left_ec = ec2; + restrictinfo->right_ec = ec2; + /* mark the RI as usable with this pair of EMs */ + restrictinfo->left_em = em1; + restrictinfo->right_em = em2; + } + else + { + /* Case 4: make a new, two-entry EC */ + EquivalenceClass *ec = makeNode(EquivalenceClass); + + ec->ec_opfamilies = opfamilies; + ec->ec_collation = collation; + ec->ec_members = NIL; + ec->ec_sources = list_make1(restrictinfo); + ec->ec_derives = NIL; + ec->ec_relids = NULL; + ec->ec_has_const = false; + ec->ec_has_volatile = false; + ec->ec_below_outer_join = below_outer_join; + ec->ec_broken = false; + ec->ec_sortref = 0; + ec->ec_min_security = restrictinfo->security_level; + ec->ec_max_security = restrictinfo->security_level; + ec->ec_merged = NULL; + em1 = add_eq_member(ec, item1, item1_relids, item1_nullable_relids, + false, item1_type); + em2 = add_eq_member(ec, item2, item2_relids, item2_nullable_relids, + false, item2_type); + + root->eq_classes = lappend(root->eq_classes, ec); + + /* mark the RI as associated with this eclass */ + restrictinfo->left_ec = ec; + restrictinfo->right_ec = ec; + /* mark the RI as usable with this pair of EMs */ + restrictinfo->left_em = em1; + restrictinfo->right_em = em2; + } + + return true; +} + +/* + * canonicalize_ec_expression + * + * This function ensures that the expression exposes the expected type and + * collation, so that it will be equal() to other equivalence-class expressions + * that it ought to be equal() to. + * + * The rule for datatypes is that the exposed type should match what it would + * be for an input to an operator of the EC's opfamilies; which is usually + * the declared input type of the operator, but in the case of polymorphic + * operators no relabeling is wanted (compare the behavior of parse_coerce.c). + * Expressions coming in from quals will generally have the right type + * already, but expressions coming from indexkeys may not (because they are + * represented without any explicit relabel in pg_index), and the same problem + * occurs for sort expressions (because the parser is likewise cavalier about + * putting relabels on them). Such cases will be binary-compatible with the + * real operators, so adding a RelabelType is sufficient. + * + * Also, the expression's exposed collation must match the EC's collation. + * This is important because in comparisons like "foo < bar COLLATE baz", + * only one of the expressions has the correct exposed collation as we receive + * it from the parser. Forcing both of them to have it ensures that all + * variant spellings of such a construct behave the same. Again, we can + * stick on a RelabelType to force the right exposed collation. (It might + * work to not label the collation at all in EC members, but this is risky + * since some parts of the system expect exprCollation() to deliver the + * right answer for a sort key.) + */ +Expr * +canonicalize_ec_expression(Expr *expr, Oid req_type, Oid req_collation) +{ + Oid expr_type = exprType((Node *) expr); + + /* + * For a polymorphic-input-type opclass, just keep the same exposed type. + * RECORD opclasses work like polymorphic-type ones for this purpose. + */ + if (IsPolymorphicType(req_type) || req_type == RECORDOID) + req_type = expr_type; + + /* + * No work if the expression exposes the right type/collation already. + */ + if (expr_type != req_type || + exprCollation((Node *) expr) != req_collation) + { + /* + * If we have to change the type of the expression, set typmod to -1, + * since the new type may not have the same typmod interpretation. + * When we only have to change collation, preserve the exposed typmod. + */ + int32 req_typmod; + + if (expr_type != req_type) + req_typmod = -1; + else + req_typmod = exprTypmod((Node *) expr); + + /* + * Use applyRelabelType so that we preserve const-flatness. This is + * important since eval_const_expressions has already been applied. + */ + expr = (Expr *) applyRelabelType((Node *) expr, + req_type, req_typmod, req_collation, + COERCE_IMPLICIT_CAST, -1, false); + } + + return expr; +} + +/* + * add_eq_member - build a new EquivalenceMember and add it to an EC + */ +static EquivalenceMember * +add_eq_member(EquivalenceClass *ec, Expr *expr, Relids relids, + Relids nullable_relids, bool is_child, Oid datatype) +{ + EquivalenceMember *em = makeNode(EquivalenceMember); + + em->em_expr = expr; + em->em_relids = relids; + em->em_nullable_relids = nullable_relids; + em->em_is_const = false; + em->em_is_child = is_child; + em->em_datatype = datatype; + + if (bms_is_empty(relids)) + { + /* + * No Vars, assume it's a pseudoconstant. This is correct for entries + * generated from process_equivalence(), because a WHERE clause can't + * contain aggregates or SRFs, and non-volatility was checked before + * process_equivalence() ever got called. But + * get_eclass_for_sort_expr() has to work harder. We put the tests + * there not here to save cycles in the equivalence case. + */ + Assert(!is_child); + em->em_is_const = true; + ec->ec_has_const = true; + /* it can't affect ec_relids */ + } + else if (!is_child) /* child members don't add to ec_relids */ + { + ec->ec_relids = bms_add_members(ec->ec_relids, relids); + } + ec->ec_members = lappend(ec->ec_members, em); + + return em; +} + + +/* + * get_eclass_for_sort_expr + * Given an expression and opfamily/collation info, find an existing + * equivalence class it is a member of; if none, optionally build a new + * single-member EquivalenceClass for it. + * + * expr is the expression, and nullable_relids is the set of base relids + * that are potentially nullable below it. We actually only care about + * the set of such relids that are used in the expression; but for caller + * convenience, we perform that intersection step here. The caller need + * only be sure that nullable_relids doesn't omit any nullable rels that + * might appear in the expr. + * + * sortref is the SortGroupRef of the originating SortGroupClause, if any, + * or zero if not. (It should never be zero if the expression is volatile!) + * + * If rel is not NULL, it identifies a specific relation we're considering + * a path for, and indicates that child EC members for that relation can be + * considered. Otherwise child members are ignored. (Note: since child EC + * members aren't guaranteed unique, a non-NULL value means that there could + * be more than one EC that matches the expression; if so it's order-dependent + * which one you get. This is annoying but it only happens in corner cases, + * so for now we live with just reporting the first match. See also + * generate_implied_equalities_for_column and match_pathkeys_to_index.) + * + * If create_it is true, we'll build a new EquivalenceClass when there is no + * match. If create_it is false, we just return NULL when no match. + * + * This can be used safely both before and after EquivalenceClass merging; + * since it never causes merging it does not invalidate any existing ECs + * or PathKeys. However, ECs added after path generation has begun are + * of limited usefulness, so usually it's best to create them beforehand. + * + * Note: opfamilies must be chosen consistently with the way + * process_equivalence() would do; that is, generated from a mergejoinable + * equality operator. Else we might fail to detect valid equivalences, + * generating poor (but not incorrect) plans. + */ +EquivalenceClass * +get_eclass_for_sort_expr(PlannerInfo *root, + Expr *expr, + Relids nullable_relids, + List *opfamilies, + Oid opcintype, + Oid collation, + Index sortref, + Relids rel, + bool create_it) +{ + Relids expr_relids; + EquivalenceClass *newec; + EquivalenceMember *newem; + ListCell *lc1; + MemoryContext oldcontext; + + /* + * Ensure the expression exposes the correct type and collation. + */ + expr = canonicalize_ec_expression(expr, opcintype, collation); + + /* + * Scan through the existing EquivalenceClasses for a match + */ + foreach(lc1, root->eq_classes) + { + EquivalenceClass *cur_ec = (EquivalenceClass *) lfirst(lc1); + ListCell *lc2; + + /* + * Never match to a volatile EC, except when we are looking at another + * reference to the same volatile SortGroupClause. + */ + if (cur_ec->ec_has_volatile && + (sortref == 0 || sortref != cur_ec->ec_sortref)) + continue; + + if (collation != cur_ec->ec_collation) + continue; + if (!equal(opfamilies, cur_ec->ec_opfamilies)) + continue; + + foreach(lc2, cur_ec->ec_members) + { + EquivalenceMember *cur_em = (EquivalenceMember *) lfirst(lc2); + + /* + * Ignore child members unless they match the request. + */ + if (cur_em->em_is_child && + !bms_equal(cur_em->em_relids, rel)) + continue; + + /* + * If below an outer join, don't match constants: they're not as + * constant as they look. + */ + if (cur_ec->ec_below_outer_join && + cur_em->em_is_const) + continue; + + if (opcintype == cur_em->em_datatype && + equal(expr, cur_em->em_expr)) + return cur_ec; /* Match! */ + } + } + + /* No match; does caller want a NULL result? */ + if (!create_it) + return NULL; + + /* + * OK, build a new single-member EC + * + * Here, we must be sure that we construct the EC in the right context. + */ + oldcontext = MemoryContextSwitchTo(root->planner_cxt); + + newec = makeNode(EquivalenceClass); + newec->ec_opfamilies = list_copy(opfamilies); + newec->ec_collation = collation; + newec->ec_members = NIL; + newec->ec_sources = NIL; + newec->ec_derives = NIL; + newec->ec_relids = NULL; + newec->ec_has_const = false; + newec->ec_has_volatile = contain_volatile_functions((Node *) expr); + newec->ec_below_outer_join = false; + newec->ec_broken = false; + newec->ec_sortref = sortref; + newec->ec_min_security = UINT_MAX; + newec->ec_max_security = 0; + newec->ec_merged = NULL; + + if (newec->ec_has_volatile && sortref == 0) /* should not happen */ + elog(ERROR, "volatile EquivalenceClass has no sortref"); + + /* + * Get the precise set of nullable relids appearing in the expression. + */ + expr_relids = pull_varnos(root, (Node *) expr); + nullable_relids = bms_intersect(nullable_relids, expr_relids); + + newem = add_eq_member(newec, copyObject(expr), expr_relids, + nullable_relids, false, opcintype); + + /* + * add_eq_member doesn't check for volatile functions, set-returning + * functions, aggregates, or window functions, but such could appear in + * sort expressions; so we have to check whether its const-marking was + * correct. + */ + if (newec->ec_has_const) + { + if (newec->ec_has_volatile || + expression_returns_set((Node *) expr) || + contain_agg_clause((Node *) expr) || + contain_window_function((Node *) expr)) + { + newec->ec_has_const = false; + newem->em_is_const = false; + } + } + + root->eq_classes = lappend(root->eq_classes, newec); + + /* + * If EC merging is already complete, we have to mop up by adding the new + * EC to the eclass_indexes of the relation(s) mentioned in it. + */ + if (root->ec_merging_done) + { + int ec_index = list_length(root->eq_classes) - 1; + int i = -1; + + while ((i = bms_next_member(newec->ec_relids, i)) > 0) + { + RelOptInfo *rel = root->simple_rel_array[i]; + + Assert(rel->reloptkind == RELOPT_BASEREL || + rel->reloptkind == RELOPT_DEADREL); + + rel->eclass_indexes = bms_add_member(rel->eclass_indexes, + ec_index); + } + } + + MemoryContextSwitchTo(oldcontext); + + return newec; +} + +/* + * find_ec_member_matching_expr + * Locate an EquivalenceClass member matching the given expr, if any; + * return NULL if no match. + * + * "Matching" is defined as "equal after stripping RelabelTypes". + * This is used for identifying sort expressions, and we need to allow + * binary-compatible relabeling for some cases involving binary-compatible + * sort operators. + * + * Child EC members are ignored unless they belong to given 'relids'. + */ +EquivalenceMember * +find_ec_member_matching_expr(EquivalenceClass *ec, + Expr *expr, + Relids relids) +{ + ListCell *lc; + + /* We ignore binary-compatible relabeling on both ends */ + while (expr && IsA(expr, RelabelType)) + expr = ((RelabelType *) expr)->arg; + + foreach(lc, ec->ec_members) + { + EquivalenceMember *em = (EquivalenceMember *) lfirst(lc); + Expr *emexpr; + + /* + * We shouldn't be trying to sort by an equivalence class that + * contains a constant, so no need to consider such cases any further. + */ + if (em->em_is_const) + continue; + + /* + * Ignore child members unless they belong to the requested rel. + */ + if (em->em_is_child && + !bms_is_subset(em->em_relids, relids)) + continue; + + /* + * Match if same expression (after stripping relabel). + */ + emexpr = em->em_expr; + while (emexpr && IsA(emexpr, RelabelType)) + emexpr = ((RelabelType *) emexpr)->arg; + + if (equal(emexpr, expr)) + return em; + } + + return NULL; +} + +/* + * find_computable_ec_member + * Locate an EquivalenceClass member that can be computed from the + * expressions appearing in "exprs"; return NULL if no match. + * + * "exprs" can be either a list of bare expression trees, or a list of + * TargetEntry nodes. Either way, it should contain Vars and possibly + * Aggrefs and WindowFuncs, which are matched to the corresponding elements + * of the EquivalenceClass's expressions. + * + * Unlike find_ec_member_matching_expr, there's no special provision here + * for binary-compatible relabeling. This is intentional: if we have to + * compute an expression in this way, setrefs.c is going to insist on exact + * matches of Vars to the source tlist. + * + * Child EC members are ignored unless they belong to given 'relids'. + * Also, non-parallel-safe expressions are ignored if 'require_parallel_safe'. + * + * Note: some callers pass root == NULL for notational reasons. This is OK + * when require_parallel_safe is false. + */ +EquivalenceMember * +find_computable_ec_member(PlannerInfo *root, + EquivalenceClass *ec, + List *exprs, + Relids relids, + bool require_parallel_safe) +{ + ListCell *lc; + + foreach(lc, ec->ec_members) + { + EquivalenceMember *em = (EquivalenceMember *) lfirst(lc); + List *exprvars; + ListCell *lc2; + + /* + * We shouldn't be trying to sort by an equivalence class that + * contains a constant, so no need to consider such cases any further. + */ + if (em->em_is_const) + continue; + + /* + * Ignore child members unless they belong to the requested rel. + */ + if (em->em_is_child && + !bms_is_subset(em->em_relids, relids)) + continue; + + /* + * Match if all Vars and quasi-Vars are available in "exprs". + */ + exprvars = pull_var_clause((Node *) em->em_expr, + PVC_INCLUDE_AGGREGATES | + PVC_INCLUDE_WINDOWFUNCS | + PVC_INCLUDE_PLACEHOLDERS); + foreach(lc2, exprvars) + { + if (!is_exprlist_member(lfirst(lc2), exprs)) + break; + } + list_free(exprvars); + if (lc2) + continue; /* we hit a non-available Var */ + + /* + * If requested, reject expressions that are not parallel-safe. We + * check this last because it's a rather expensive test. + */ + if (require_parallel_safe && + !is_parallel_safe(root, (Node *) em->em_expr)) + continue; + + return em; /* found usable expression */ + } + + return NULL; +} + +/* + * is_exprlist_member + * Subroutine for find_computable_ec_member: is "node" in "exprs"? + * + * Per the requirements of that function, "exprs" might or might not have + * TargetEntry superstructure. + */ +static bool +is_exprlist_member(Expr *node, List *exprs) +{ + ListCell *lc; + + foreach(lc, exprs) + { + Expr *expr = (Expr *) lfirst(lc); + + if (expr && IsA(expr, TargetEntry)) + expr = ((TargetEntry *) expr)->expr; + + if (equal(node, expr)) + return true; + } + return false; +} + +/* + * relation_can_be_sorted_early + * Can this relation be sorted on this EC before the final output step? + * + * To succeed, we must find an EC member that prepare_sort_from_pathkeys knows + * how to sort on, given the rel's reltarget as input. There are also a few + * additional constraints based on the fact that the desired sort will be done + * "early", within the scan/join part of the plan. Also, non-parallel-safe + * expressions are ignored if 'require_parallel_safe'. + * + * At some point we might want to return the identified EquivalenceMember, + * but for now, callers only want to know if there is one. + */ +bool +relation_can_be_sorted_early(PlannerInfo *root, RelOptInfo *rel, + EquivalenceClass *ec, bool require_parallel_safe) +{ + PathTarget *target = rel->reltarget; + EquivalenceMember *em; + ListCell *lc; + + /* + * Reject volatile ECs immediately; such sorts must always be postponed. + */ + if (ec->ec_has_volatile) + return false; + + /* + * Try to find an EM directly matching some reltarget member. + */ + foreach(lc, target->exprs) + { + Expr *targetexpr = (Expr *) lfirst(lc); + + em = find_ec_member_matching_expr(ec, targetexpr, rel->relids); + if (!em) + continue; + + /* + * Reject expressions involving set-returning functions, as those + * can't be computed early either. (Note: this test and the following + * one are effectively checking properties of targetexpr, so there's + * no point in asking whether some other EC member would be better.) + */ + if (expression_returns_set((Node *) em->em_expr)) + continue; + + /* + * If requested, reject expressions that are not parallel-safe. We + * check this last because it's a rather expensive test. + */ + if (require_parallel_safe && + !is_parallel_safe(root, (Node *) em->em_expr)) + continue; + + return true; + } + + /* + * Try to find an expression computable from the reltarget. + */ + em = find_computable_ec_member(root, ec, target->exprs, rel->relids, + require_parallel_safe); + if (!em) + return false; + + /* + * Reject expressions involving set-returning functions, as those can't be + * computed early either. (There's no point in looking for another EC + * member in this case; since SRFs can't appear in WHERE, they cannot + * belong to multi-member ECs.) + */ + if (expression_returns_set((Node *) em->em_expr)) + return false; + + return true; +} + +/* + * generate_base_implied_equalities + * Generate any restriction clauses that we can deduce from equivalence + * classes. + * + * When an EC contains pseudoconstants, our strategy is to generate + * "member = const1" clauses where const1 is the first constant member, for + * every other member (including other constants). If we are able to do this + * then we don't need any "var = var" comparisons because we've successfully + * constrained all the vars at their points of creation. If we fail to + * generate any of these clauses due to lack of cross-type operators, we fall + * back to the "ec_broken" strategy described below. (XXX if there are + * multiple constants of different types, it's possible that we might succeed + * in forming all the required clauses if we started from a different const + * member; but this seems a sufficiently hokey corner case to not be worth + * spending lots of cycles on.) + * + * For ECs that contain no pseudoconstants, we generate derived clauses + * "member1 = member2" for each pair of members belonging to the same base + * relation (actually, if there are more than two for the same base relation, + * we only need enough clauses to link each to each other). This provides + * the base case for the recursion: each row emitted by a base relation scan + * will constrain all computable members of the EC to be equal. As each + * join path is formed, we'll add additional derived clauses on-the-fly + * to maintain this invariant (see generate_join_implied_equalities). + * + * If the opfamilies used by the EC do not provide complete sets of cross-type + * equality operators, it is possible that we will fail to generate a clause + * that must be generated to maintain the invariant. (An example: given + * "WHERE a.x = b.y AND b.y = a.z", the scheme breaks down if we cannot + * generate "a.x = a.z" as a restriction clause for A.) In this case we mark + * the EC "ec_broken" and fall back to regurgitating its original source + * RestrictInfos at appropriate times. We do not try to retract any derived + * clauses already generated from the broken EC, so the resulting plan could + * be poor due to bad selectivity estimates caused by redundant clauses. But + * the correct solution to that is to fix the opfamilies ... + * + * Equality clauses derived by this function are passed off to + * process_implied_equality (in plan/initsplan.c) to be inserted into the + * restrictinfo datastructures. Note that this must be called after initial + * scanning of the quals and before Path construction begins. + * + * We make no attempt to avoid generating duplicate RestrictInfos here: we + * don't search ec_sources or ec_derives for matches. It doesn't really + * seem worth the trouble to do so. + */ +void +generate_base_implied_equalities(PlannerInfo *root) +{ + int ec_index; + ListCell *lc; + + /* + * At this point, we're done absorbing knowledge of equivalences in the + * query, so no further EC merging should happen, and ECs remaining in the + * eq_classes list can be considered canonical. (But note that it's still + * possible for new single-member ECs to be added through + * get_eclass_for_sort_expr().) + */ + root->ec_merging_done = true; + + ec_index = 0; + foreach(lc, root->eq_classes) + { + EquivalenceClass *ec = (EquivalenceClass *) lfirst(lc); + bool can_generate_joinclause = false; + int i; + + Assert(ec->ec_merged == NULL); /* else shouldn't be in list */ + Assert(!ec->ec_broken); /* not yet anyway... */ + + /* + * Generate implied equalities that are restriction clauses. + * Single-member ECs won't generate any deductions, either here or at + * the join level. + */ + if (list_length(ec->ec_members) > 1) + { + if (ec->ec_has_const) + generate_base_implied_equalities_const(root, ec); + else + generate_base_implied_equalities_no_const(root, ec); + + /* Recover if we failed to generate required derived clauses */ + if (ec->ec_broken) + generate_base_implied_equalities_broken(root, ec); + + /* Detect whether this EC might generate join clauses */ + can_generate_joinclause = + (bms_membership(ec->ec_relids) == BMS_MULTIPLE); + } + + /* + * Mark the base rels cited in each eclass (which should all exist by + * now) with the eq_classes indexes of all eclasses mentioning them. + * This will let us avoid searching in subsequent lookups. While + * we're at it, we can mark base rels that have pending eclass joins; + * this is a cheap version of has_relevant_eclass_joinclause(). + */ + i = -1; + while ((i = bms_next_member(ec->ec_relids, i)) > 0) + { + RelOptInfo *rel = root->simple_rel_array[i]; + + Assert(rel->reloptkind == RELOPT_BASEREL); + + rel->eclass_indexes = bms_add_member(rel->eclass_indexes, + ec_index); + + if (can_generate_joinclause) + rel->has_eclass_joins = true; + } + + ec_index++; + } +} + +/* + * generate_base_implied_equalities when EC contains pseudoconstant(s) + */ +static void +generate_base_implied_equalities_const(PlannerInfo *root, + EquivalenceClass *ec) +{ + EquivalenceMember *const_em = NULL; + ListCell *lc; + + /* + * In the trivial case where we just had one "var = const" clause, push + * the original clause back into the main planner machinery. There is + * nothing to be gained by doing it differently, and we save the effort to + * re-build and re-analyze an equality clause that will be exactly + * equivalent to the old one. + */ + if (list_length(ec->ec_members) == 2 && + list_length(ec->ec_sources) == 1) + { + RestrictInfo *restrictinfo = (RestrictInfo *) linitial(ec->ec_sources); + + if (bms_membership(restrictinfo->required_relids) != BMS_MULTIPLE) + { + distribute_restrictinfo_to_rels(root, restrictinfo); + return; + } + } + + /* + * Find the constant member to use. We prefer an actual constant to + * pseudo-constants (such as Params), because the constraint exclusion + * machinery might be able to exclude relations on the basis of generated + * "var = const" equalities, but "var = param" won't work for that. + */ + foreach(lc, ec->ec_members) + { + EquivalenceMember *cur_em = (EquivalenceMember *) lfirst(lc); + + if (cur_em->em_is_const) + { + const_em = cur_em; + if (IsA(cur_em->em_expr, Const)) + break; + } + } + Assert(const_em != NULL); + + /* Generate a derived equality against each other member */ + foreach(lc, ec->ec_members) + { + EquivalenceMember *cur_em = (EquivalenceMember *) lfirst(lc); + Oid eq_op; + RestrictInfo *rinfo; + + Assert(!cur_em->em_is_child); /* no children yet */ + if (cur_em == const_em) + continue; + eq_op = select_equality_operator(ec, + cur_em->em_datatype, + const_em->em_datatype); + if (!OidIsValid(eq_op)) + { + /* failed... */ + ec->ec_broken = true; + break; + } + rinfo = process_implied_equality(root, eq_op, ec->ec_collation, + cur_em->em_expr, const_em->em_expr, + bms_copy(ec->ec_relids), + bms_union(cur_em->em_nullable_relids, + const_em->em_nullable_relids), + ec->ec_min_security, + ec->ec_below_outer_join, + cur_em->em_is_const); + + /* + * If the clause didn't degenerate to a constant, fill in the correct + * markings for a mergejoinable clause, and save it in ec_derives. (We + * will not re-use such clauses directly, but selectivity estimation + * may consult the list later. Note that this use of ec_derives does + * not overlap with its use for join clauses, since we never generate + * join clauses from an ec_has_const eclass.) + */ + if (rinfo && rinfo->mergeopfamilies) + { + /* it's not redundant, so don't set parent_ec */ + rinfo->left_ec = rinfo->right_ec = ec; + rinfo->left_em = cur_em; + rinfo->right_em = const_em; + ec->ec_derives = lappend(ec->ec_derives, rinfo); + } + } +} + +/* + * generate_base_implied_equalities when EC contains no pseudoconstants + */ +static void +generate_base_implied_equalities_no_const(PlannerInfo *root, + EquivalenceClass *ec) +{ + EquivalenceMember **prev_ems; + ListCell *lc; + + /* + * We scan the EC members once and track the last-seen member for each + * base relation. When we see another member of the same base relation, + * we generate "prev_em = cur_em". This results in the minimum number of + * derived clauses, but it's possible that it will fail when a different + * ordering would succeed. XXX FIXME: use a UNION-FIND algorithm similar + * to the way we build merged ECs. (Use a list-of-lists for each rel.) + */ + prev_ems = (EquivalenceMember **) + palloc0(root->simple_rel_array_size * sizeof(EquivalenceMember *)); + + foreach(lc, ec->ec_members) + { + EquivalenceMember *cur_em = (EquivalenceMember *) lfirst(lc); + int relid; + + Assert(!cur_em->em_is_child); /* no children yet */ + if (!bms_get_singleton_member(cur_em->em_relids, &relid)) + continue; + Assert(relid < root->simple_rel_array_size); + + if (prev_ems[relid] != NULL) + { + EquivalenceMember *prev_em = prev_ems[relid]; + Oid eq_op; + RestrictInfo *rinfo; + + eq_op = select_equality_operator(ec, + prev_em->em_datatype, + cur_em->em_datatype); + if (!OidIsValid(eq_op)) + { + /* failed... */ + ec->ec_broken = true; + break; + } + rinfo = process_implied_equality(root, eq_op, ec->ec_collation, + prev_em->em_expr, cur_em->em_expr, + bms_copy(ec->ec_relids), + bms_union(prev_em->em_nullable_relids, + cur_em->em_nullable_relids), + ec->ec_min_security, + ec->ec_below_outer_join, + false); + + /* + * If the clause didn't degenerate to a constant, fill in the + * correct markings for a mergejoinable clause. We don't put it + * in ec_derives however; we don't currently need to re-find such + * clauses, and we don't want to clutter that list with non-join + * clauses. + */ + if (rinfo && rinfo->mergeopfamilies) + { + /* it's not redundant, so don't set parent_ec */ + rinfo->left_ec = rinfo->right_ec = ec; + rinfo->left_em = prev_em; + rinfo->right_em = cur_em; + } + } + prev_ems[relid] = cur_em; + } + + pfree(prev_ems); + + /* + * We also have to make sure that all the Vars used in the member clauses + * will be available at any join node we might try to reference them at. + * For the moment we force all the Vars to be available at all join nodes + * for this eclass. Perhaps this could be improved by doing some + * pre-analysis of which members we prefer to join, but it's no worse than + * what happened in the pre-8.3 code. + */ + foreach(lc, ec->ec_members) + { + EquivalenceMember *cur_em = (EquivalenceMember *) lfirst(lc); + List *vars = pull_var_clause((Node *) cur_em->em_expr, + PVC_RECURSE_AGGREGATES | + PVC_RECURSE_WINDOWFUNCS | + PVC_INCLUDE_PLACEHOLDERS); + + add_vars_to_targetlist(root, vars, ec->ec_relids, false); + list_free(vars); + } +} + +/* + * generate_base_implied_equalities cleanup after failure + * + * What we must do here is push any zero- or one-relation source RestrictInfos + * of the EC back into the main restrictinfo datastructures. Multi-relation + * clauses will be regurgitated later by generate_join_implied_equalities(). + * (We do it this way to maintain continuity with the case that ec_broken + * becomes set only after we've gone up a join level or two.) However, for + * an EC that contains constants, we can adopt a simpler strategy and just + * throw back all the source RestrictInfos immediately; that works because + * we know that such an EC can't become broken later. (This rule justifies + * ignoring ec_has_const ECs in generate_join_implied_equalities, even when + * they are broken.) + */ +static void +generate_base_implied_equalities_broken(PlannerInfo *root, + EquivalenceClass *ec) +{ + ListCell *lc; + + foreach(lc, ec->ec_sources) + { + RestrictInfo *restrictinfo = (RestrictInfo *) lfirst(lc); + + if (ec->ec_has_const || + bms_membership(restrictinfo->required_relids) != BMS_MULTIPLE) + distribute_restrictinfo_to_rels(root, restrictinfo); + } +} + + +/* + * generate_join_implied_equalities + * Generate any join clauses that we can deduce from equivalence classes. + * + * At a join node, we must enforce restriction clauses sufficient to ensure + * that all equivalence-class members computable at that node are equal. + * Since the set of clauses to enforce can vary depending on which subset + * relations are the inputs, we have to compute this afresh for each join + * relation pair. Hence a fresh List of RestrictInfo nodes is built and + * passed back on each call. + * + * In addition to its use at join nodes, this can be applied to generate + * eclass-based join clauses for use in a parameterized scan of a base rel. + * The reason for the asymmetry of specifying the inner rel as a RelOptInfo + * and the outer rel by Relids is that this usage occurs before we have + * built any join RelOptInfos. + * + * An annoying special case for parameterized scans is that the inner rel can + * be an appendrel child (an "other rel"). In this case we must generate + * appropriate clauses using child EC members. add_child_rel_equivalences + * must already have been done for the child rel. + * + * The results are sufficient for use in merge, hash, and plain nestloop join + * methods. We do not worry here about selecting clauses that are optimal + * for use in a parameterized indexscan. indxpath.c makes its own selections + * of clauses to use, and if the ones we pick here are redundant with those, + * the extras will be eliminated at createplan time, using the parent_ec + * markers that we provide (see is_redundant_derived_clause()). + * + * Because the same join clauses are likely to be needed multiple times as + * we consider different join paths, we avoid generating multiple copies: + * whenever we select a particular pair of EquivalenceMembers to join, + * we check to see if the pair matches any original clause (in ec_sources) + * or previously-built clause (in ec_derives). This saves memory and allows + * re-use of information cached in RestrictInfos. + * + * join_relids should always equal bms_union(outer_relids, inner_rel->relids). + * We could simplify this function's API by computing it internally, but in + * most current uses, the caller has the value at hand anyway. + */ +List * +generate_join_implied_equalities(PlannerInfo *root, + Relids join_relids, + Relids outer_relids, + RelOptInfo *inner_rel) +{ + List *result = NIL; + Relids inner_relids = inner_rel->relids; + Relids nominal_inner_relids; + Relids nominal_join_relids; + Bitmapset *matching_ecs; + int i; + + /* If inner rel is a child, extra setup work is needed */ + if (IS_OTHER_REL(inner_rel)) + { + Assert(!bms_is_empty(inner_rel->top_parent_relids)); + + /* Fetch relid set for the topmost parent rel */ + nominal_inner_relids = inner_rel->top_parent_relids; + /* ECs will be marked with the parent's relid, not the child's */ + nominal_join_relids = bms_union(outer_relids, nominal_inner_relids); + } + else + { + nominal_inner_relids = inner_relids; + nominal_join_relids = join_relids; + } + + /* + * Get all eclasses that mention both inner and outer sides of the join + */ + matching_ecs = get_common_eclass_indexes(root, nominal_inner_relids, + outer_relids); + + i = -1; + while ((i = bms_next_member(matching_ecs, i)) >= 0) + { + EquivalenceClass *ec = (EquivalenceClass *) list_nth(root->eq_classes, i); + List *sublist = NIL; + + /* ECs containing consts do not need any further enforcement */ + if (ec->ec_has_const) + continue; + + /* Single-member ECs won't generate any deductions */ + if (list_length(ec->ec_members) <= 1) + continue; + + /* Sanity check that this eclass overlaps the join */ + Assert(bms_overlap(ec->ec_relids, nominal_join_relids)); + + if (!ec->ec_broken) + sublist = generate_join_implied_equalities_normal(root, + ec, + join_relids, + outer_relids, + inner_relids); + + /* Recover if we failed to generate required derived clauses */ + if (ec->ec_broken) + sublist = generate_join_implied_equalities_broken(root, + ec, + nominal_join_relids, + outer_relids, + nominal_inner_relids, + inner_rel); + + result = list_concat(result, sublist); + } + + return result; +} + +/* + * generate_join_implied_equalities_for_ecs + * As above, but consider only the listed ECs. + */ +List * +generate_join_implied_equalities_for_ecs(PlannerInfo *root, + List *eclasses, + Relids join_relids, + Relids outer_relids, + RelOptInfo *inner_rel) +{ + List *result = NIL; + Relids inner_relids = inner_rel->relids; + Relids nominal_inner_relids; + Relids nominal_join_relids; + ListCell *lc; + + /* If inner rel is a child, extra setup work is needed */ + if (IS_OTHER_REL(inner_rel)) + { + Assert(!bms_is_empty(inner_rel->top_parent_relids)); + + /* Fetch relid set for the topmost parent rel */ + nominal_inner_relids = inner_rel->top_parent_relids; + /* ECs will be marked with the parent's relid, not the child's */ + nominal_join_relids = bms_union(outer_relids, nominal_inner_relids); + } + else + { + nominal_inner_relids = inner_relids; + nominal_join_relids = join_relids; + } + + foreach(lc, eclasses) + { + EquivalenceClass *ec = (EquivalenceClass *) lfirst(lc); + List *sublist = NIL; + + /* ECs containing consts do not need any further enforcement */ + if (ec->ec_has_const) + continue; + + /* Single-member ECs won't generate any deductions */ + if (list_length(ec->ec_members) <= 1) + continue; + + /* We can quickly ignore any that don't overlap the join, too */ + if (!bms_overlap(ec->ec_relids, nominal_join_relids)) + continue; + + if (!ec->ec_broken) + sublist = generate_join_implied_equalities_normal(root, + ec, + join_relids, + outer_relids, + inner_relids); + + /* Recover if we failed to generate required derived clauses */ + if (ec->ec_broken) + sublist = generate_join_implied_equalities_broken(root, + ec, + nominal_join_relids, + outer_relids, + nominal_inner_relids, + inner_rel); + + result = list_concat(result, sublist); + } + + return result; +} + +/* + * generate_join_implied_equalities for a still-valid EC + */ +static List * +generate_join_implied_equalities_normal(PlannerInfo *root, + EquivalenceClass *ec, + Relids join_relids, + Relids outer_relids, + Relids inner_relids) +{ + List *result = NIL; + List *new_members = NIL; + List *outer_members = NIL; + List *inner_members = NIL; + ListCell *lc1; + + /* + * First, scan the EC to identify member values that are computable at the + * outer rel, at the inner rel, or at this relation but not in either + * input rel. The outer-rel members should already be enforced equal, + * likewise for the inner-rel members. We'll need to create clauses to + * enforce that any newly computable members are all equal to each other + * as well as to at least one input member, plus enforce at least one + * outer-rel member equal to at least one inner-rel member. + */ + foreach(lc1, ec->ec_members) + { + EquivalenceMember *cur_em = (EquivalenceMember *) lfirst(lc1); + + /* + * We don't need to check explicitly for child EC members. This test + * against join_relids will cause them to be ignored except when + * considering a child inner rel, which is what we want. + */ + if (!bms_is_subset(cur_em->em_relids, join_relids)) + continue; /* not computable yet, or wrong child */ + + if (bms_is_subset(cur_em->em_relids, outer_relids)) + outer_members = lappend(outer_members, cur_em); + else if (bms_is_subset(cur_em->em_relids, inner_relids)) + inner_members = lappend(inner_members, cur_em); + else + new_members = lappend(new_members, cur_em); + } + + /* + * First, select the joinclause if needed. We can equate any one outer + * member to any one inner member, but we have to find a datatype + * combination for which an opfamily member operator exists. If we have + * choices, we prefer simple Var members (possibly with RelabelType) since + * these are (a) cheapest to compute at runtime and (b) most likely to + * have useful statistics. Also, prefer operators that are also + * hashjoinable. + */ + if (outer_members && inner_members) + { + EquivalenceMember *best_outer_em = NULL; + EquivalenceMember *best_inner_em = NULL; + Oid best_eq_op = InvalidOid; + int best_score = -1; + RestrictInfo *rinfo; + + foreach(lc1, outer_members) + { + EquivalenceMember *outer_em = (EquivalenceMember *) lfirst(lc1); + ListCell *lc2; + + foreach(lc2, inner_members) + { + EquivalenceMember *inner_em = (EquivalenceMember *) lfirst(lc2); + Oid eq_op; + int score; + + eq_op = select_equality_operator(ec, + outer_em->em_datatype, + inner_em->em_datatype); + if (!OidIsValid(eq_op)) + continue; + score = 0; + if (IsA(outer_em->em_expr, Var) || + (IsA(outer_em->em_expr, RelabelType) && + IsA(((RelabelType *) outer_em->em_expr)->arg, Var))) + score++; + if (IsA(inner_em->em_expr, Var) || + (IsA(inner_em->em_expr, RelabelType) && + IsA(((RelabelType *) inner_em->em_expr)->arg, Var))) + score++; + if (op_hashjoinable(eq_op, + exprType((Node *) outer_em->em_expr))) + score++; + if (score > best_score) + { + best_outer_em = outer_em; + best_inner_em = inner_em; + best_eq_op = eq_op; + best_score = score; + if (best_score == 3) + break; /* no need to look further */ + } + } + if (best_score == 3) + break; /* no need to look further */ + } + if (best_score < 0) + { + /* failed... */ + ec->ec_broken = true; + return NIL; + } + + /* + * Create clause, setting parent_ec to mark it as redundant with other + * joinclauses + */ + rinfo = create_join_clause(root, ec, best_eq_op, + best_outer_em, best_inner_em, + ec); + + result = lappend(result, rinfo); + } + + /* + * Now deal with building restrictions for any expressions that involve + * Vars from both sides of the join. We have to equate all of these to + * each other as well as to at least one old member (if any). + * + * XXX as in generate_base_implied_equalities_no_const, we could be a lot + * smarter here to avoid unnecessary failures in cross-type situations. + * For now, use the same left-to-right method used there. + */ + if (new_members) + { + List *old_members = list_concat(outer_members, inner_members); + EquivalenceMember *prev_em = NULL; + RestrictInfo *rinfo; + + /* For now, arbitrarily take the first old_member as the one to use */ + if (old_members) + new_members = lappend(new_members, linitial(old_members)); + + foreach(lc1, new_members) + { + EquivalenceMember *cur_em = (EquivalenceMember *) lfirst(lc1); + + if (prev_em != NULL) + { + Oid eq_op; + + eq_op = select_equality_operator(ec, + prev_em->em_datatype, + cur_em->em_datatype); + if (!OidIsValid(eq_op)) + { + /* failed... */ + ec->ec_broken = true; + return NIL; + } + /* do NOT set parent_ec, this qual is not redundant! */ + rinfo = create_join_clause(root, ec, eq_op, + prev_em, cur_em, + NULL); + + result = lappend(result, rinfo); + } + prev_em = cur_em; + } + } + + return result; +} + +/* + * generate_join_implied_equalities cleanup after failure + * + * Return any original RestrictInfos that are enforceable at this join. + * + * In the case of a child inner relation, we have to translate the + * original RestrictInfos from parent to child Vars. + */ +static List * +generate_join_implied_equalities_broken(PlannerInfo *root, + EquivalenceClass *ec, + Relids nominal_join_relids, + Relids outer_relids, + Relids nominal_inner_relids, + RelOptInfo *inner_rel) +{ + List *result = NIL; + ListCell *lc; + + foreach(lc, ec->ec_sources) + { + RestrictInfo *restrictinfo = (RestrictInfo *) lfirst(lc); + Relids clause_relids = restrictinfo->required_relids; + + if (bms_is_subset(clause_relids, nominal_join_relids) && + !bms_is_subset(clause_relids, outer_relids) && + !bms_is_subset(clause_relids, nominal_inner_relids)) + result = lappend(result, restrictinfo); + } + + /* + * If we have to translate, just brute-force apply adjust_appendrel_attrs + * to all the RestrictInfos at once. This will result in returning + * RestrictInfos that are not listed in ec_derives, but there shouldn't be + * any duplication, and it's a sufficiently narrow corner case that we + * shouldn't sweat too much over it anyway. + * + * Since inner_rel might be an indirect descendant of the baserel + * mentioned in the ec_sources clauses, we have to be prepared to apply + * multiple levels of Var translation. + */ + if (IS_OTHER_REL(inner_rel) && result != NIL) + result = (List *) adjust_appendrel_attrs_multilevel(root, + (Node *) result, + inner_rel->relids, + inner_rel->top_parent_relids); + + return result; +} + + +/* + * select_equality_operator + * Select a suitable equality operator for comparing two EC members + * + * Returns InvalidOid if no operator can be found for this datatype combination + */ +static Oid +select_equality_operator(EquivalenceClass *ec, Oid lefttype, Oid righttype) +{ + ListCell *lc; + + foreach(lc, ec->ec_opfamilies) + { + Oid opfamily = lfirst_oid(lc); + Oid opno; + + opno = get_opfamily_member(opfamily, lefttype, righttype, + BTEqualStrategyNumber); + if (!OidIsValid(opno)) + continue; + /* If no barrier quals in query, don't worry about leaky operators */ + if (ec->ec_max_security == 0) + return opno; + /* Otherwise, insist that selected operators be leakproof */ + if (get_func_leakproof(get_opcode(opno))) + return opno; + } + return InvalidOid; +} + + +/* + * create_join_clause + * Find or make a RestrictInfo comparing the two given EC members + * with the given operator. + * + * parent_ec is either equal to ec (if the clause is a potentially-redundant + * join clause) or NULL (if not). We have to treat this as part of the + * match requirements --- it's possible that a clause comparing the same two + * EMs is a join clause in one join path and a restriction clause in another. + */ +static RestrictInfo * +create_join_clause(PlannerInfo *root, + EquivalenceClass *ec, Oid opno, + EquivalenceMember *leftem, + EquivalenceMember *rightem, + EquivalenceClass *parent_ec) +{ + RestrictInfo *rinfo; + ListCell *lc; + MemoryContext oldcontext; + + /* + * Search to see if we already built a RestrictInfo for this pair of + * EquivalenceMembers. We can use either original source clauses or + * previously-derived clauses. The check on opno is probably redundant, + * but be safe ... + */ + foreach(lc, ec->ec_sources) + { + rinfo = (RestrictInfo *) lfirst(lc); + if (rinfo->left_em == leftem && + rinfo->right_em == rightem && + rinfo->parent_ec == parent_ec && + opno == ((OpExpr *) rinfo->clause)->opno) + return rinfo; + } + + foreach(lc, ec->ec_derives) + { + rinfo = (RestrictInfo *) lfirst(lc); + if (rinfo->left_em == leftem && + rinfo->right_em == rightem && + rinfo->parent_ec == parent_ec && + opno == ((OpExpr *) rinfo->clause)->opno) + return rinfo; + } + + /* + * Not there, so build it, in planner context so we can re-use it. (Not + * important in normal planning, but definitely so in GEQO.) + */ + oldcontext = MemoryContextSwitchTo(root->planner_cxt); + + rinfo = build_implied_join_equality(root, + opno, + ec->ec_collation, + leftem->em_expr, + rightem->em_expr, + bms_union(leftem->em_relids, + rightem->em_relids), + bms_union(leftem->em_nullable_relids, + rightem->em_nullable_relids), + ec->ec_min_security); + + /* Mark the clause as redundant, or not */ + rinfo->parent_ec = parent_ec; + + /* + * We know the correct values for left_ec/right_ec, ie this particular EC, + * so we can just set them directly instead of forcing another lookup. + */ + rinfo->left_ec = ec; + rinfo->right_ec = ec; + + /* Mark it as usable with these EMs */ + rinfo->left_em = leftem; + rinfo->right_em = rightem; + /* and save it for possible re-use */ + ec->ec_derives = lappend(ec->ec_derives, rinfo); + + MemoryContextSwitchTo(oldcontext); + + return rinfo; +} + + +/* + * reconsider_outer_join_clauses + * Re-examine any outer-join clauses that were set aside by + * distribute_qual_to_rels(), and see if we can derive any + * EquivalenceClasses from them. Then, if they were not made + * redundant, push them out into the regular join-clause lists. + * + * When we have mergejoinable clauses A = B that are outer-join clauses, + * we can't blindly combine them with other clauses A = C to deduce B = C, + * since in fact the "equality" A = B won't necessarily hold above the + * outer join (one of the variables might be NULL instead). Nonetheless + * there are cases where we can add qual clauses using transitivity. + * + * One case that we look for here is an outer-join clause OUTERVAR = INNERVAR + * for which there is also an equivalence clause OUTERVAR = CONSTANT. + * It is safe and useful to push a clause INNERVAR = CONSTANT into the + * evaluation of the inner (nullable) relation, because any inner rows not + * meeting this condition will not contribute to the outer-join result anyway. + * (Any outer rows they could join to will be eliminated by the pushed-down + * equivalence clause.) + * + * Note that the above rule does not work for full outer joins; nor is it + * very interesting to consider cases where the generated equivalence clause + * would involve relations outside the outer join, since such clauses couldn't + * be pushed into the inner side's scan anyway. So the restriction to + * outervar = pseudoconstant is not really giving up anything. + * + * For full-join cases, we can only do something useful if it's a FULL JOIN + * USING and a merged column has an equivalence MERGEDVAR = CONSTANT. + * By the time it gets here, the merged column will look like + * COALESCE(LEFTVAR, RIGHTVAR) + * and we will have a full-join clause LEFTVAR = RIGHTVAR that we can match + * the COALESCE expression to. In this situation we can push LEFTVAR = CONSTANT + * and RIGHTVAR = CONSTANT into the input relations, since any rows not + * meeting these conditions cannot contribute to the join result. + * + * Again, there isn't any traction to be gained by trying to deal with + * clauses comparing a mergedvar to a non-pseudoconstant. So we can make + * use of the EquivalenceClasses to search for matching variables that were + * equivalenced to constants. The interesting outer-join clauses were + * accumulated for us by distribute_qual_to_rels. + * + * When we find one of these cases, we implement the changes we want by + * generating a new equivalence clause INNERVAR = CONSTANT (or LEFTVAR, etc) + * and pushing it into the EquivalenceClass structures. This is because we + * may already know that INNERVAR is equivalenced to some other var(s), and + * we'd like the constant to propagate to them too. Note that it would be + * unsafe to merge any existing EC for INNERVAR with the OUTERVAR's EC --- + * that could result in propagating constant restrictions from + * INNERVAR to OUTERVAR, which would be very wrong. + * + * It's possible that the INNERVAR is also an OUTERVAR for some other + * outer-join clause, in which case the process can be repeated. So we repeat + * looping over the lists of clauses until no further deductions can be made. + * Whenever we do make a deduction, we remove the generating clause from the + * lists, since we don't want to make the same deduction twice. + * + * If we don't find any match for a set-aside outer join clause, we must + * throw it back into the regular joinclause processing by passing it to + * distribute_restrictinfo_to_rels(). If we do generate a derived clause, + * however, the outer-join clause is redundant. We still throw it back, + * because otherwise the join will be seen as a clauseless join and avoided + * during join order searching; but we mark it as redundant to keep from + * messing up the joinrel's size estimate. (This behavior means that the + * API for this routine is uselessly complex: we could have just put all + * the clauses into the regular processing initially. We keep it because + * someday we might want to do something else, such as inserting "dummy" + * joinclauses instead of real ones.) + * + * Outer join clauses that are marked outerjoin_delayed are special: this + * condition means that one or both VARs might go to null due to a lower + * outer join. We can still push a constant through the clause, but only + * if its operator is strict; and we *have to* throw the clause back into + * regular joinclause processing. By keeping the strict join clause, + * we ensure that any null-extended rows that are mistakenly generated due + * to suppressing rows not matching the constant will be rejected at the + * upper outer join. (This doesn't work for full-join clauses.) + */ +void +reconsider_outer_join_clauses(PlannerInfo *root) +{ + bool found; + ListCell *cell; + + /* Outer loop repeats until we find no more deductions */ + do + { + found = false; + + /* Process the LEFT JOIN clauses */ + foreach(cell, root->left_join_clauses) + { + RestrictInfo *rinfo = (RestrictInfo *) lfirst(cell); + + if (reconsider_outer_join_clause(root, rinfo, true)) + { + found = true; + /* remove it from the list */ + root->left_join_clauses = + foreach_delete_current(root->left_join_clauses, cell); + /* we throw it back anyway (see notes above) */ + /* but the thrown-back clause has no extra selectivity */ + rinfo->norm_selec = 2.0; + rinfo->outer_selec = 1.0; + distribute_restrictinfo_to_rels(root, rinfo); + } + } + + /* Process the RIGHT JOIN clauses */ + foreach(cell, root->right_join_clauses) + { + RestrictInfo *rinfo = (RestrictInfo *) lfirst(cell); + + if (reconsider_outer_join_clause(root, rinfo, false)) + { + found = true; + /* remove it from the list */ + root->right_join_clauses = + foreach_delete_current(root->right_join_clauses, cell); + /* we throw it back anyway (see notes above) */ + /* but the thrown-back clause has no extra selectivity */ + rinfo->norm_selec = 2.0; + rinfo->outer_selec = 1.0; + distribute_restrictinfo_to_rels(root, rinfo); + } + } + + /* Process the FULL JOIN clauses */ + foreach(cell, root->full_join_clauses) + { + RestrictInfo *rinfo = (RestrictInfo *) lfirst(cell); + + if (reconsider_full_join_clause(root, rinfo)) + { + found = true; + /* remove it from the list */ + root->full_join_clauses = + foreach_delete_current(root->full_join_clauses, cell); + /* we throw it back anyway (see notes above) */ + /* but the thrown-back clause has no extra selectivity */ + rinfo->norm_selec = 2.0; + rinfo->outer_selec = 1.0; + distribute_restrictinfo_to_rels(root, rinfo); + } + } + } while (found); + + /* Now, any remaining clauses have to be thrown back */ + foreach(cell, root->left_join_clauses) + { + RestrictInfo *rinfo = (RestrictInfo *) lfirst(cell); + + distribute_restrictinfo_to_rels(root, rinfo); + } + foreach(cell, root->right_join_clauses) + { + RestrictInfo *rinfo = (RestrictInfo *) lfirst(cell); + + distribute_restrictinfo_to_rels(root, rinfo); + } + foreach(cell, root->full_join_clauses) + { + RestrictInfo *rinfo = (RestrictInfo *) lfirst(cell); + + distribute_restrictinfo_to_rels(root, rinfo); + } +} + +/* + * reconsider_outer_join_clauses for a single LEFT/RIGHT JOIN clause + * + * Returns true if we were able to propagate a constant through the clause. + */ +static bool +reconsider_outer_join_clause(PlannerInfo *root, RestrictInfo *rinfo, + bool outer_on_left) +{ + Expr *outervar, + *innervar; + Oid opno, + collation, + left_type, + right_type, + inner_datatype; + Relids inner_relids, + inner_nullable_relids; + ListCell *lc1; + + Assert(is_opclause(rinfo->clause)); + opno = ((OpExpr *) rinfo->clause)->opno; + collation = ((OpExpr *) rinfo->clause)->inputcollid; + + /* If clause is outerjoin_delayed, operator must be strict */ + if (rinfo->outerjoin_delayed && !op_strict(opno)) + return false; + + /* Extract needed info from the clause */ + op_input_types(opno, &left_type, &right_type); + if (outer_on_left) + { + outervar = (Expr *) get_leftop(rinfo->clause); + innervar = (Expr *) get_rightop(rinfo->clause); + inner_datatype = right_type; + inner_relids = rinfo->right_relids; + } + else + { + outervar = (Expr *) get_rightop(rinfo->clause); + innervar = (Expr *) get_leftop(rinfo->clause); + inner_datatype = left_type; + inner_relids = rinfo->left_relids; + } + inner_nullable_relids = bms_intersect(inner_relids, + rinfo->nullable_relids); + + /* Scan EquivalenceClasses for a match to outervar */ + foreach(lc1, root->eq_classes) + { + EquivalenceClass *cur_ec = (EquivalenceClass *) lfirst(lc1); + bool match; + ListCell *lc2; + + /* Ignore EC unless it contains pseudoconstants */ + if (!cur_ec->ec_has_const) + continue; + /* Never match to a volatile EC */ + if (cur_ec->ec_has_volatile) + continue; + /* It has to match the outer-join clause as to semantics, too */ + if (collation != cur_ec->ec_collation) + continue; + if (!equal(rinfo->mergeopfamilies, cur_ec->ec_opfamilies)) + continue; + /* Does it contain a match to outervar? */ + match = false; + foreach(lc2, cur_ec->ec_members) + { + EquivalenceMember *cur_em = (EquivalenceMember *) lfirst(lc2); + + Assert(!cur_em->em_is_child); /* no children yet */ + if (equal(outervar, cur_em->em_expr)) + { + match = true; + break; + } + } + if (!match) + continue; /* no match, so ignore this EC */ + + /* + * Yes it does! Try to generate a clause INNERVAR = CONSTANT for each + * CONSTANT in the EC. Note that we must succeed with at least one + * constant before we can decide to throw away the outer-join clause. + */ + match = false; + foreach(lc2, cur_ec->ec_members) + { + EquivalenceMember *cur_em = (EquivalenceMember *) lfirst(lc2); + Oid eq_op; + RestrictInfo *newrinfo; + + if (!cur_em->em_is_const) + continue; /* ignore non-const members */ + eq_op = select_equality_operator(cur_ec, + inner_datatype, + cur_em->em_datatype); + if (!OidIsValid(eq_op)) + continue; /* can't generate equality */ + newrinfo = build_implied_join_equality(root, + eq_op, + cur_ec->ec_collation, + innervar, + cur_em->em_expr, + bms_copy(inner_relids), + bms_copy(inner_nullable_relids), + cur_ec->ec_min_security); + if (process_equivalence(root, &newrinfo, true)) + match = true; + } + + /* + * If we were able to equate INNERVAR to any constant, report success. + * Otherwise, fall out of the search loop, since we know the OUTERVAR + * appears in at most one EC. + */ + if (match) + return true; + else + break; + } + + return false; /* failed to make any deduction */ +} + +/* + * reconsider_outer_join_clauses for a single FULL JOIN clause + * + * Returns true if we were able to propagate a constant through the clause. + */ +static bool +reconsider_full_join_clause(PlannerInfo *root, RestrictInfo *rinfo) +{ + Expr *leftvar; + Expr *rightvar; + Oid opno, + collation, + left_type, + right_type; + Relids left_relids, + right_relids, + left_nullable_relids, + right_nullable_relids; + ListCell *lc1; + + /* Can't use an outerjoin_delayed clause here */ + if (rinfo->outerjoin_delayed) + return false; + + /* Extract needed info from the clause */ + Assert(is_opclause(rinfo->clause)); + opno = ((OpExpr *) rinfo->clause)->opno; + collation = ((OpExpr *) rinfo->clause)->inputcollid; + op_input_types(opno, &left_type, &right_type); + leftvar = (Expr *) get_leftop(rinfo->clause); + rightvar = (Expr *) get_rightop(rinfo->clause); + left_relids = rinfo->left_relids; + right_relids = rinfo->right_relids; + left_nullable_relids = bms_intersect(left_relids, + rinfo->nullable_relids); + right_nullable_relids = bms_intersect(right_relids, + rinfo->nullable_relids); + + foreach(lc1, root->eq_classes) + { + EquivalenceClass *cur_ec = (EquivalenceClass *) lfirst(lc1); + EquivalenceMember *coal_em = NULL; + bool match; + bool matchleft; + bool matchright; + ListCell *lc2; + int coal_idx = -1; + + /* Ignore EC unless it contains pseudoconstants */ + if (!cur_ec->ec_has_const) + continue; + /* Never match to a volatile EC */ + if (cur_ec->ec_has_volatile) + continue; + /* It has to match the outer-join clause as to semantics, too */ + if (collation != cur_ec->ec_collation) + continue; + if (!equal(rinfo->mergeopfamilies, cur_ec->ec_opfamilies)) + continue; + + /* + * Does it contain a COALESCE(leftvar, rightvar) construct? + * + * We can assume the COALESCE() inputs are in the same order as the + * join clause, since both were automatically generated in the cases + * we care about. + * + * XXX currently this may fail to match in cross-type cases because + * the COALESCE will contain typecast operations while the join clause + * may not (if there is a cross-type mergejoin operator available for + * the two column types). Is it OK to strip implicit coercions from + * the COALESCE arguments? + */ + match = false; + foreach(lc2, cur_ec->ec_members) + { + coal_em = (EquivalenceMember *) lfirst(lc2); + Assert(!coal_em->em_is_child); /* no children yet */ + if (IsA(coal_em->em_expr, CoalesceExpr)) + { + CoalesceExpr *cexpr = (CoalesceExpr *) coal_em->em_expr; + Node *cfirst; + Node *csecond; + + if (list_length(cexpr->args) != 2) + continue; + cfirst = (Node *) linitial(cexpr->args); + csecond = (Node *) lsecond(cexpr->args); + + if (equal(leftvar, cfirst) && equal(rightvar, csecond)) + { + coal_idx = foreach_current_index(lc2); + match = true; + break; + } + } + } + if (!match) + continue; /* no match, so ignore this EC */ + + /* + * Yes it does! Try to generate clauses LEFTVAR = CONSTANT and + * RIGHTVAR = CONSTANT for each CONSTANT in the EC. Note that we must + * succeed with at least one constant for each var before we can + * decide to throw away the outer-join clause. + */ + matchleft = matchright = false; + foreach(lc2, cur_ec->ec_members) + { + EquivalenceMember *cur_em = (EquivalenceMember *) lfirst(lc2); + Oid eq_op; + RestrictInfo *newrinfo; + + if (!cur_em->em_is_const) + continue; /* ignore non-const members */ + eq_op = select_equality_operator(cur_ec, + left_type, + cur_em->em_datatype); + if (OidIsValid(eq_op)) + { + newrinfo = build_implied_join_equality(root, + eq_op, + cur_ec->ec_collation, + leftvar, + cur_em->em_expr, + bms_copy(left_relids), + bms_copy(left_nullable_relids), + cur_ec->ec_min_security); + if (process_equivalence(root, &newrinfo, true)) + matchleft = true; + } + eq_op = select_equality_operator(cur_ec, + right_type, + cur_em->em_datatype); + if (OidIsValid(eq_op)) + { + newrinfo = build_implied_join_equality(root, + eq_op, + cur_ec->ec_collation, + rightvar, + cur_em->em_expr, + bms_copy(right_relids), + bms_copy(right_nullable_relids), + cur_ec->ec_min_security); + if (process_equivalence(root, &newrinfo, true)) + matchright = true; + } + } + + /* + * If we were able to equate both vars to constants, we're done, and + * we can throw away the full-join clause as redundant. Moreover, we + * can remove the COALESCE entry from the EC, since the added + * restrictions ensure it will always have the expected value. (We + * don't bother trying to update ec_relids or ec_sources.) + */ + if (matchleft && matchright) + { + cur_ec->ec_members = list_delete_nth_cell(cur_ec->ec_members, coal_idx); + return true; + } + + /* + * Otherwise, fall out of the search loop, since we know the COALESCE + * appears in at most one EC (XXX might stop being true if we allow + * stripping of coercions above?) + */ + break; + } + + return false; /* failed to make any deduction */ +} + + +/* + * exprs_known_equal + * Detect whether two expressions are known equal due to equivalence + * relationships. + * + * Actually, this only shows that the expressions are equal according + * to some opfamily's notion of equality --- but we only use it for + * selectivity estimation, so a fuzzy idea of equality is OK. + * + * Note: does not bother to check for "equal(item1, item2)"; caller must + * check that case if it's possible to pass identical items. + */ +bool +exprs_known_equal(PlannerInfo *root, Node *item1, Node *item2) +{ + ListCell *lc1; + + foreach(lc1, root->eq_classes) + { + EquivalenceClass *ec = (EquivalenceClass *) lfirst(lc1); + bool item1member = false; + bool item2member = false; + ListCell *lc2; + + /* Never match to a volatile EC */ + if (ec->ec_has_volatile) + continue; + + foreach(lc2, ec->ec_members) + { + EquivalenceMember *em = (EquivalenceMember *) lfirst(lc2); + + if (em->em_is_child) + continue; /* ignore children here */ + if (equal(item1, em->em_expr)) + item1member = true; + else if (equal(item2, em->em_expr)) + item2member = true; + /* Exit as soon as equality is proven */ + if (item1member && item2member) + return true; + } + } + return false; +} + + +/* + * match_eclasses_to_foreign_key_col + * See whether a foreign key column match is proven by any eclass. + * + * If the referenced and referencing Vars of the fkey's colno'th column are + * known equal due to any eclass, return that eclass; otherwise return NULL. + * (In principle there might be more than one matching eclass if multiple + * collations are involved, but since collation doesn't matter for equality, + * we ignore that fine point here.) This is much like exprs_known_equal, + * except that we insist on the comparison operator matching the eclass, so + * that the result is definite not approximate. + * + * On success, we also set fkinfo->eclass[colno] to the matching eclass, + * and set fkinfo->fk_eclass_member[colno] to the eclass member for the + * referencing Var. + */ +EquivalenceClass * +match_eclasses_to_foreign_key_col(PlannerInfo *root, + ForeignKeyOptInfo *fkinfo, + int colno) +{ + Index var1varno = fkinfo->con_relid; + AttrNumber var1attno = fkinfo->conkey[colno]; + Index var2varno = fkinfo->ref_relid; + AttrNumber var2attno = fkinfo->confkey[colno]; + Oid eqop = fkinfo->conpfeqop[colno]; + RelOptInfo *rel1 = root->simple_rel_array[var1varno]; + RelOptInfo *rel2 = root->simple_rel_array[var2varno]; + List *opfamilies = NIL; /* compute only if needed */ + Bitmapset *matching_ecs; + int i; + + /* Consider only eclasses mentioning both relations */ + Assert(root->ec_merging_done); + Assert(IS_SIMPLE_REL(rel1)); + Assert(IS_SIMPLE_REL(rel2)); + matching_ecs = bms_intersect(rel1->eclass_indexes, + rel2->eclass_indexes); + + i = -1; + while ((i = bms_next_member(matching_ecs, i)) >= 0) + { + EquivalenceClass *ec = (EquivalenceClass *) list_nth(root->eq_classes, + i); + EquivalenceMember *item1_em = NULL; + EquivalenceMember *item2_em = NULL; + ListCell *lc2; + + /* Never match to a volatile EC */ + if (ec->ec_has_volatile) + continue; + /* Note: it seems okay to match to "broken" eclasses here */ + + foreach(lc2, ec->ec_members) + { + EquivalenceMember *em = (EquivalenceMember *) lfirst(lc2); + Var *var; + + if (em->em_is_child) + continue; /* ignore children here */ + + /* EM must be a Var, possibly with RelabelType */ + var = (Var *) em->em_expr; + while (var && IsA(var, RelabelType)) + var = (Var *) ((RelabelType *) var)->arg; + if (!(var && IsA(var, Var))) + continue; + + /* Match? */ + if (var->varno == var1varno && var->varattno == var1attno) + item1_em = em; + else if (var->varno == var2varno && var->varattno == var2attno) + item2_em = em; + + /* Have we found both PK and FK column in this EC? */ + if (item1_em && item2_em) + { + /* + * Succeed if eqop matches EC's opfamilies. We could test + * this before scanning the members, but it's probably cheaper + * to test for member matches first. + */ + if (opfamilies == NIL) /* compute if we didn't already */ + opfamilies = get_mergejoin_opfamilies(eqop); + if (equal(opfamilies, ec->ec_opfamilies)) + { + fkinfo->eclass[colno] = ec; + fkinfo->fk_eclass_member[colno] = item2_em; + return ec; + } + /* Otherwise, done with this EC, move on to the next */ + break; + } + } + } + return NULL; +} + +/* + * find_derived_clause_for_ec_member + * Search for a previously-derived clause mentioning the given EM. + * + * The eclass should be an ec_has_const EC, of which the EM is a non-const + * member. This should ensure there is just one derived clause mentioning + * the EM (and equating it to a constant). + * Returns NULL if no such clause can be found. + */ +RestrictInfo * +find_derived_clause_for_ec_member(EquivalenceClass *ec, + EquivalenceMember *em) +{ + ListCell *lc; + + Assert(ec->ec_has_const); + Assert(!em->em_is_const); + foreach(lc, ec->ec_derives) + { + RestrictInfo *rinfo = (RestrictInfo *) lfirst(lc); + + /* + * generate_base_implied_equalities_const will have put non-const + * members on the left side of derived clauses. + */ + if (rinfo->left_em == em) + return rinfo; + } + return NULL; +} + + +/* + * add_child_rel_equivalences + * Search for EC members that reference the root parent of child_rel, and + * add transformed members referencing the child_rel. + * + * Note that this function won't be called at all unless we have at least some + * reason to believe that the EC members it generates will be useful. + * + * parent_rel and child_rel could be derived from appinfo, but since the + * caller has already computed them, we might as well just pass them in. + * + * The passed-in AppendRelInfo is not used when the parent_rel is not a + * top-level baserel, since it shows the mapping from the parent_rel but + * we need to translate EC expressions that refer to the top-level parent. + * Using it is faster than using adjust_appendrel_attrs_multilevel(), though, + * so we prefer it when we can. + */ +void +add_child_rel_equivalences(PlannerInfo *root, + AppendRelInfo *appinfo, + RelOptInfo *parent_rel, + RelOptInfo *child_rel) +{ + Relids top_parent_relids = child_rel->top_parent_relids; + Relids child_relids = child_rel->relids; + int i; + + /* + * EC merging should be complete already, so we can use the parent rel's + * eclass_indexes to avoid searching all of root->eq_classes. + */ + Assert(root->ec_merging_done); + Assert(IS_SIMPLE_REL(parent_rel)); + + i = -1; + while ((i = bms_next_member(parent_rel->eclass_indexes, i)) >= 0) + { + EquivalenceClass *cur_ec = (EquivalenceClass *) list_nth(root->eq_classes, i); + int num_members; + + /* + * If this EC contains a volatile expression, then generating child + * EMs would be downright dangerous, so skip it. We rely on a + * volatile EC having only one EM. + */ + if (cur_ec->ec_has_volatile) + continue; + + /* Sanity check eclass_indexes only contain ECs for parent_rel */ + Assert(bms_is_subset(top_parent_relids, cur_ec->ec_relids)); + + /* + * We don't use foreach() here because there's no point in scanning + * newly-added child members, so we can stop after the last + * pre-existing EC member. + */ + num_members = list_length(cur_ec->ec_members); + for (int pos = 0; pos < num_members; pos++) + { + EquivalenceMember *cur_em = (EquivalenceMember *) list_nth(cur_ec->ec_members, pos); + + if (cur_em->em_is_const) + continue; /* ignore consts here */ + + /* + * We consider only original EC members here, not + * already-transformed child members. Otherwise, if some original + * member expression references more than one appendrel, we'd get + * an O(N^2) explosion of useless derived expressions for + * combinations of children. (But add_child_join_rel_equivalences + * may add targeted combinations for partitionwise-join purposes.) + */ + if (cur_em->em_is_child) + continue; /* ignore children here */ + + /* Does this member reference child's topmost parent rel? */ + if (bms_overlap(cur_em->em_relids, top_parent_relids)) + { + /* Yes, generate transformed child version */ + Expr *child_expr; + Relids new_relids; + Relids new_nullable_relids; + + if (parent_rel->reloptkind == RELOPT_BASEREL) + { + /* Simple single-level transformation */ + child_expr = (Expr *) + adjust_appendrel_attrs(root, + (Node *) cur_em->em_expr, + 1, &appinfo); + } + else + { + /* Must do multi-level transformation */ + child_expr = (Expr *) + adjust_appendrel_attrs_multilevel(root, + (Node *) cur_em->em_expr, + child_relids, + top_parent_relids); + } + + /* + * Transform em_relids to match. Note we do *not* do + * pull_varnos(child_expr) here, as for example the + * transformation might have substituted a constant, but we + * don't want the child member to be marked as constant. + */ + new_relids = bms_difference(cur_em->em_relids, + top_parent_relids); + new_relids = bms_add_members(new_relids, child_relids); + + /* + * And likewise for nullable_relids. Note this code assumes + * parent and child relids are singletons. + */ + new_nullable_relids = cur_em->em_nullable_relids; + if (bms_overlap(new_nullable_relids, top_parent_relids)) + { + new_nullable_relids = bms_difference(new_nullable_relids, + top_parent_relids); + new_nullable_relids = bms_add_members(new_nullable_relids, + child_relids); + } + + (void) add_eq_member(cur_ec, child_expr, + new_relids, new_nullable_relids, + true, cur_em->em_datatype); + + /* Record this EC index for the child rel */ + child_rel->eclass_indexes = bms_add_member(child_rel->eclass_indexes, i); + } + } + } +} + +/* + * add_child_join_rel_equivalences + * Like add_child_rel_equivalences(), but for joinrels + * + * Here we find the ECs relevant to the top parent joinrel and add transformed + * member expressions that refer to this child joinrel. + * + * Note that this function won't be called at all unless we have at least some + * reason to believe that the EC members it generates will be useful. + */ +void +add_child_join_rel_equivalences(PlannerInfo *root, + int nappinfos, AppendRelInfo **appinfos, + RelOptInfo *parent_joinrel, + RelOptInfo *child_joinrel) +{ + Relids top_parent_relids = child_joinrel->top_parent_relids; + Relids child_relids = child_joinrel->relids; + Bitmapset *matching_ecs; + MemoryContext oldcontext; + int i; + + Assert(IS_JOIN_REL(child_joinrel) && IS_JOIN_REL(parent_joinrel)); + + /* We need consider only ECs that mention the parent joinrel */ + matching_ecs = get_eclass_indexes_for_relids(root, top_parent_relids); + + /* + * If we're being called during GEQO join planning, we still have to + * create any new EC members in the main planner context, to avoid having + * a corrupt EC data structure after the GEQO context is reset. This is + * problematic since we'll leak memory across repeated GEQO cycles. For + * now, though, bloat is better than crash. If it becomes a real issue + * we'll have to do something to avoid generating duplicate EC members. + */ + oldcontext = MemoryContextSwitchTo(root->planner_cxt); + + i = -1; + while ((i = bms_next_member(matching_ecs, i)) >= 0) + { + EquivalenceClass *cur_ec = (EquivalenceClass *) list_nth(root->eq_classes, i); + int num_members; + + /* + * If this EC contains a volatile expression, then generating child + * EMs would be downright dangerous, so skip it. We rely on a + * volatile EC having only one EM. + */ + if (cur_ec->ec_has_volatile) + continue; + + /* Sanity check on get_eclass_indexes_for_relids result */ + Assert(bms_overlap(top_parent_relids, cur_ec->ec_relids)); + + /* + * We don't use foreach() here because there's no point in scanning + * newly-added child members, so we can stop after the last + * pre-existing EC member. + */ + num_members = list_length(cur_ec->ec_members); + for (int pos = 0; pos < num_members; pos++) + { + EquivalenceMember *cur_em = (EquivalenceMember *) list_nth(cur_ec->ec_members, pos); + + if (cur_em->em_is_const) + continue; /* ignore consts here */ + + /* + * We consider only original EC members here, not + * already-transformed child members. + */ + if (cur_em->em_is_child) + continue; /* ignore children here */ + + /* + * We may ignore expressions that reference a single baserel, + * because add_child_rel_equivalences should have handled them. + */ + if (bms_membership(cur_em->em_relids) != BMS_MULTIPLE) + continue; + + /* Does this member reference child's topmost parent rel? */ + if (bms_overlap(cur_em->em_relids, top_parent_relids)) + { + /* Yes, generate transformed child version */ + Expr *child_expr; + Relids new_relids; + Relids new_nullable_relids; + + if (parent_joinrel->reloptkind == RELOPT_JOINREL) + { + /* Simple single-level transformation */ + child_expr = (Expr *) + adjust_appendrel_attrs(root, + (Node *) cur_em->em_expr, + nappinfos, appinfos); + } + else + { + /* Must do multi-level transformation */ + Assert(parent_joinrel->reloptkind == RELOPT_OTHER_JOINREL); + child_expr = (Expr *) + adjust_appendrel_attrs_multilevel(root, + (Node *) cur_em->em_expr, + child_relids, + top_parent_relids); + } + + /* + * Transform em_relids to match. Note we do *not* do + * pull_varnos(child_expr) here, as for example the + * transformation might have substituted a constant, but we + * don't want the child member to be marked as constant. + */ + new_relids = bms_difference(cur_em->em_relids, + top_parent_relids); + new_relids = bms_add_members(new_relids, child_relids); + + /* + * For nullable_relids, we must selectively replace parent + * nullable relids with child ones. + */ + new_nullable_relids = cur_em->em_nullable_relids; + if (bms_overlap(new_nullable_relids, top_parent_relids)) + new_nullable_relids = + adjust_child_relids_multilevel(root, + new_nullable_relids, + child_relids, + top_parent_relids); + + (void) add_eq_member(cur_ec, child_expr, + new_relids, new_nullable_relids, + true, cur_em->em_datatype); + } + } + } + + MemoryContextSwitchTo(oldcontext); +} + + +/* + * generate_implied_equalities_for_column + * Create EC-derived joinclauses usable with a specific column. + * + * This is used by indxpath.c to extract potentially indexable joinclauses + * from ECs, and can be used by foreign data wrappers for similar purposes. + * We assume that only expressions in Vars of a single table are of interest, + * but the caller provides a callback function to identify exactly which + * such expressions it would like to know about. + * + * We assume that any given table/index column could appear in only one EC. + * (This should be true in all but the most pathological cases, and if it + * isn't, we stop on the first match anyway.) Therefore, what we return + * is a redundant list of clauses equating the table/index column to each of + * the other-relation values it is known to be equal to. Any one of + * these clauses can be used to create a parameterized path, and there + * is no value in using more than one. (But it *is* worthwhile to create + * a separate parameterized path for each one, since that leads to different + * join orders.) + * + * The caller can pass a Relids set of rels we aren't interested in joining + * to, so as to save the work of creating useless clauses. + */ +List * +generate_implied_equalities_for_column(PlannerInfo *root, + RelOptInfo *rel, + ec_matches_callback_type callback, + void *callback_arg, + Relids prohibited_rels) +{ + List *result = NIL; + bool is_child_rel = (rel->reloptkind == RELOPT_OTHER_MEMBER_REL); + Relids parent_relids; + int i; + + /* Should be OK to rely on eclass_indexes */ + Assert(root->ec_merging_done); + + /* Indexes are available only on base or "other" member relations. */ + Assert(IS_SIMPLE_REL(rel)); + + /* If it's a child rel, we'll need to know what its parent(s) are */ + if (is_child_rel) + parent_relids = find_childrel_parents(root, rel); + else + parent_relids = NULL; /* not used, but keep compiler quiet */ + + i = -1; + while ((i = bms_next_member(rel->eclass_indexes, i)) >= 0) + { + EquivalenceClass *cur_ec = (EquivalenceClass *) list_nth(root->eq_classes, i); + EquivalenceMember *cur_em; + ListCell *lc2; + + /* Sanity check eclass_indexes only contain ECs for rel */ + Assert(is_child_rel || bms_is_subset(rel->relids, cur_ec->ec_relids)); + + /* + * Won't generate joinclauses if const or single-member (the latter + * test covers the volatile case too) + */ + if (cur_ec->ec_has_const || list_length(cur_ec->ec_members) <= 1) + continue; + + /* + * Scan members, looking for a match to the target column. Note that + * child EC members are considered, but only when they belong to the + * target relation. (Unlike regular members, the same expression + * could be a child member of more than one EC. Therefore, it's + * potentially order-dependent which EC a child relation's target + * column gets matched to. This is annoying but it only happens in + * corner cases, so for now we live with just reporting the first + * match. See also get_eclass_for_sort_expr.) + */ + cur_em = NULL; + foreach(lc2, cur_ec->ec_members) + { + cur_em = (EquivalenceMember *) lfirst(lc2); + if (bms_equal(cur_em->em_relids, rel->relids) && + callback(root, rel, cur_ec, cur_em, callback_arg)) + break; + cur_em = NULL; + } + + if (!cur_em) + continue; + + /* + * Found our match. Scan the other EC members and attempt to generate + * joinclauses. + */ + foreach(lc2, cur_ec->ec_members) + { + EquivalenceMember *other_em = (EquivalenceMember *) lfirst(lc2); + Oid eq_op; + RestrictInfo *rinfo; + + if (other_em->em_is_child) + continue; /* ignore children here */ + + /* Make sure it'll be a join to a different rel */ + if (other_em == cur_em || + bms_overlap(other_em->em_relids, rel->relids)) + continue; + + /* Forget it if caller doesn't want joins to this rel */ + if (bms_overlap(other_em->em_relids, prohibited_rels)) + continue; + + /* + * Also, if this is a child rel, avoid generating a useless join + * to its parent rel(s). + */ + if (is_child_rel && + bms_overlap(parent_relids, other_em->em_relids)) + continue; + + eq_op = select_equality_operator(cur_ec, + cur_em->em_datatype, + other_em->em_datatype); + if (!OidIsValid(eq_op)) + continue; + + /* set parent_ec to mark as redundant with other joinclauses */ + rinfo = create_join_clause(root, cur_ec, eq_op, + cur_em, other_em, + cur_ec); + + result = lappend(result, rinfo); + } + + /* + * If somehow we failed to create any join clauses, we might as well + * keep scanning the ECs for another match. But if we did make any, + * we're done, because we don't want to return non-redundant clauses. + */ + if (result) + break; + } + + return result; +} + +/* + * have_relevant_eclass_joinclause + * Detect whether there is an EquivalenceClass that could produce + * a joinclause involving the two given relations. + * + * This is essentially a very cut-down version of + * generate_join_implied_equalities(). Note it's OK to occasionally say "yes" + * incorrectly. Hence we don't bother with details like whether the lack of a + * cross-type operator might prevent the clause from actually being generated. + */ +bool +have_relevant_eclass_joinclause(PlannerInfo *root, + RelOptInfo *rel1, RelOptInfo *rel2) +{ + Bitmapset *matching_ecs; + int i; + + /* Examine only eclasses mentioning both rel1 and rel2 */ + matching_ecs = get_common_eclass_indexes(root, rel1->relids, + rel2->relids); + + i = -1; + while ((i = bms_next_member(matching_ecs, i)) >= 0) + { + EquivalenceClass *ec = (EquivalenceClass *) list_nth(root->eq_classes, + i); + + /* + * Sanity check that get_common_eclass_indexes gave only ECs + * containing both rels. + */ + Assert(bms_overlap(rel1->relids, ec->ec_relids)); + Assert(bms_overlap(rel2->relids, ec->ec_relids)); + + /* + * Won't generate joinclauses if single-member (this test covers the + * volatile case too) + */ + if (list_length(ec->ec_members) <= 1) + continue; + + /* + * We do not need to examine the individual members of the EC, because + * all that we care about is whether each rel overlaps the relids of + * at least one member, and get_common_eclass_indexes() and the single + * member check above are sufficient to prove that. (As with + * have_relevant_joinclause(), it is not necessary that the EC be able + * to form a joinclause relating exactly the two given rels, only that + * it be able to form a joinclause mentioning both, and this will + * surely be true if both of them overlap ec_relids.) + * + * Note we don't test ec_broken; if we did, we'd need a separate code + * path to look through ec_sources. Checking the membership anyway is + * OK as a possibly-overoptimistic heuristic. + * + * We don't test ec_has_const either, even though a const eclass won't + * generate real join clauses. This is because if we had "WHERE a.x = + * b.y and a.x = 42", it is worth considering a join between a and b, + * since the join result is likely to be small even though it'll end + * up being an unqualified nestloop. + */ + + return true; + } + + return false; +} + + +/* + * has_relevant_eclass_joinclause + * Detect whether there is an EquivalenceClass that could produce + * a joinclause involving the given relation and anything else. + * + * This is the same as have_relevant_eclass_joinclause with the other rel + * implicitly defined as "everything else in the query". + */ +bool +has_relevant_eclass_joinclause(PlannerInfo *root, RelOptInfo *rel1) +{ + Bitmapset *matched_ecs; + int i; + + /* Examine only eclasses mentioning rel1 */ + matched_ecs = get_eclass_indexes_for_relids(root, rel1->relids); + + i = -1; + while ((i = bms_next_member(matched_ecs, i)) >= 0) + { + EquivalenceClass *ec = (EquivalenceClass *) list_nth(root->eq_classes, + i); + + /* + * Won't generate joinclauses if single-member (this test covers the + * volatile case too) + */ + if (list_length(ec->ec_members) <= 1) + continue; + + /* + * Per the comment in have_relevant_eclass_joinclause, it's sufficient + * to find an EC that mentions both this rel and some other rel. + */ + if (!bms_is_subset(ec->ec_relids, rel1->relids)) + return true; + } + + return false; +} + + +/* + * eclass_useful_for_merging + * Detect whether the EC could produce any mergejoinable join clauses + * against the specified relation. + * + * This is just a heuristic test and doesn't have to be exact; it's better + * to say "yes" incorrectly than "no". Hence we don't bother with details + * like whether the lack of a cross-type operator might prevent the clause + * from actually being generated. + */ +bool +eclass_useful_for_merging(PlannerInfo *root, + EquivalenceClass *eclass, + RelOptInfo *rel) +{ + Relids relids; + ListCell *lc; + + Assert(!eclass->ec_merged); + + /* + * Won't generate joinclauses if const or single-member (the latter test + * covers the volatile case too) + */ + if (eclass->ec_has_const || list_length(eclass->ec_members) <= 1) + return false; + + /* + * Note we don't test ec_broken; if we did, we'd need a separate code path + * to look through ec_sources. Checking the members anyway is OK as a + * possibly-overoptimistic heuristic. + */ + + /* If specified rel is a child, we must consider the topmost parent rel */ + if (IS_OTHER_REL(rel)) + { + Assert(!bms_is_empty(rel->top_parent_relids)); + relids = rel->top_parent_relids; + } + else + relids = rel->relids; + + /* If rel already includes all members of eclass, no point in searching */ + if (bms_is_subset(eclass->ec_relids, relids)) + return false; + + /* To join, we need a member not in the given rel */ + foreach(lc, eclass->ec_members) + { + EquivalenceMember *cur_em = (EquivalenceMember *) lfirst(lc); + + if (cur_em->em_is_child) + continue; /* ignore children here */ + + if (!bms_overlap(cur_em->em_relids, relids)) + return true; + } + + return false; +} + + +/* + * is_redundant_derived_clause + * Test whether rinfo is derived from same EC as any clause in clauselist; + * if so, it can be presumed to represent a condition that's redundant + * with that member of the list. + */ +bool +is_redundant_derived_clause(RestrictInfo *rinfo, List *clauselist) +{ + EquivalenceClass *parent_ec = rinfo->parent_ec; + ListCell *lc; + + /* Fail if it's not a potentially-redundant clause from some EC */ + if (parent_ec == NULL) + return false; + + foreach(lc, clauselist) + { + RestrictInfo *otherrinfo = (RestrictInfo *) lfirst(lc); + + if (otherrinfo->parent_ec == parent_ec) + return true; + } + + return false; +} + +/* + * is_redundant_with_indexclauses + * Test whether rinfo is redundant with any clause in the IndexClause + * list. Here, for convenience, we test both simple identity and + * whether it is derived from the same EC as any member of the list. + */ +bool +is_redundant_with_indexclauses(RestrictInfo *rinfo, List *indexclauses) +{ + EquivalenceClass *parent_ec = rinfo->parent_ec; + ListCell *lc; + + foreach(lc, indexclauses) + { + IndexClause *iclause = lfirst_node(IndexClause, lc); + RestrictInfo *otherrinfo = iclause->rinfo; + + /* If indexclause is lossy, it won't enforce the condition exactly */ + if (iclause->lossy) + continue; + + /* Match if it's same clause (pointer equality should be enough) */ + if (rinfo == otherrinfo) + return true; + /* Match if derived from same EC */ + if (parent_ec && otherrinfo->parent_ec == parent_ec) + return true; + + /* + * No need to look at the derived clauses in iclause->indexquals; they + * couldn't match if the parent clause didn't. + */ + } + + return false; +} + +/* + * get_eclass_indexes_for_relids + * Build and return a Bitmapset containing the indexes into root's + * eq_classes list for all eclasses that mention any of these relids + */ +static Bitmapset * +get_eclass_indexes_for_relids(PlannerInfo *root, Relids relids) +{ + Bitmapset *ec_indexes = NULL; + int i = -1; + + /* Should be OK to rely on eclass_indexes */ + Assert(root->ec_merging_done); + + while ((i = bms_next_member(relids, i)) > 0) + { + RelOptInfo *rel = root->simple_rel_array[i]; + + ec_indexes = bms_add_members(ec_indexes, rel->eclass_indexes); + } + return ec_indexes; +} + +/* + * get_common_eclass_indexes + * Build and return a Bitmapset containing the indexes into root's + * eq_classes list for all eclasses that mention rels in both + * relids1 and relids2. + */ +static Bitmapset * +get_common_eclass_indexes(PlannerInfo *root, Relids relids1, Relids relids2) +{ + Bitmapset *rel1ecs; + Bitmapset *rel2ecs; + int relid; + + rel1ecs = get_eclass_indexes_for_relids(root, relids1); + + /* + * We can get away with just using the relation's eclass_indexes directly + * when relids2 is a singleton set. + */ + if (bms_get_singleton_member(relids2, &relid)) + rel2ecs = root->simple_rel_array[relid]->eclass_indexes; + else + rel2ecs = get_eclass_indexes_for_relids(root, relids2); + + /* Calculate and return the common EC indexes, recycling the left input. */ + return bms_int_members(rel1ecs, rel2ecs); +} diff --git a/src/backend/optimizer/path/indxpath.c b/src/backend/optimizer/path/indxpath.c new file mode 100644 index 0000000..3800f0c --- /dev/null +++ b/src/backend/optimizer/path/indxpath.c @@ -0,0 +1,3817 @@ +/*------------------------------------------------------------------------- + * + * indxpath.c + * Routines to determine which indexes are usable for scanning a + * given relation, and create Paths accordingly. + * + * Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group + * Portions Copyright (c) 1994, Regents of the University of California + * + * + * IDENTIFICATION + * src/backend/optimizer/path/indxpath.c + * + *------------------------------------------------------------------------- + */ +#include "postgres.h" + +#include <math.h> + +#include "access/stratnum.h" +#include "access/sysattr.h" +#include "catalog/pg_am.h" +#include "catalog/pg_operator.h" +#include "catalog/pg_opfamily.h" +#include "catalog/pg_type.h" +#include "nodes/makefuncs.h" +#include "nodes/nodeFuncs.h" +#include "nodes/supportnodes.h" +#include "optimizer/cost.h" +#include "optimizer/optimizer.h" +#include "optimizer/pathnode.h" +#include "optimizer/paths.h" +#include "optimizer/prep.h" +#include "optimizer/restrictinfo.h" +#include "utils/lsyscache.h" +#include "utils/selfuncs.h" + + +/* XXX see PartCollMatchesExprColl */ +#define IndexCollMatchesExprColl(idxcollation, exprcollation) \ + ((idxcollation) == InvalidOid || (idxcollation) == (exprcollation)) + +/* Whether we are looking for plain indexscan, bitmap scan, or either */ +typedef enum +{ + ST_INDEXSCAN, /* must support amgettuple */ + ST_BITMAPSCAN, /* must support amgetbitmap */ + ST_ANYSCAN /* either is okay */ +} ScanTypeControl; + +/* Data structure for collecting qual clauses that match an index */ +typedef struct +{ + bool nonempty; /* True if lists are not all empty */ + /* Lists of IndexClause nodes, one list per index column */ + List *indexclauses[INDEX_MAX_KEYS]; +} IndexClauseSet; + +/* Per-path data used within choose_bitmap_and() */ +typedef struct +{ + Path *path; /* IndexPath, BitmapAndPath, or BitmapOrPath */ + List *quals; /* the WHERE clauses it uses */ + List *preds; /* predicates of its partial index(es) */ + Bitmapset *clauseids; /* quals+preds represented as a bitmapset */ + bool unclassifiable; /* has too many quals+preds to process? */ +} PathClauseUsage; + +/* Callback argument for ec_member_matches_indexcol */ +typedef struct +{ + IndexOptInfo *index; /* index we're considering */ + int indexcol; /* index column we want to match to */ +} ec_member_matches_arg; + + +static void consider_index_join_clauses(PlannerInfo *root, RelOptInfo *rel, + IndexOptInfo *index, + IndexClauseSet *rclauseset, + IndexClauseSet *jclauseset, + IndexClauseSet *eclauseset, + List **bitindexpaths); +static void consider_index_join_outer_rels(PlannerInfo *root, RelOptInfo *rel, + IndexOptInfo *index, + IndexClauseSet *rclauseset, + IndexClauseSet *jclauseset, + IndexClauseSet *eclauseset, + List **bitindexpaths, + List *indexjoinclauses, + int considered_clauses, + List **considered_relids); +static void get_join_index_paths(PlannerInfo *root, RelOptInfo *rel, + IndexOptInfo *index, + IndexClauseSet *rclauseset, + IndexClauseSet *jclauseset, + IndexClauseSet *eclauseset, + List **bitindexpaths, + Relids relids, + List **considered_relids); +static bool eclass_already_used(EquivalenceClass *parent_ec, Relids oldrelids, + List *indexjoinclauses); +static bool bms_equal_any(Relids relids, List *relids_list); +static void get_index_paths(PlannerInfo *root, RelOptInfo *rel, + IndexOptInfo *index, IndexClauseSet *clauses, + List **bitindexpaths); +static List *build_index_paths(PlannerInfo *root, RelOptInfo *rel, + IndexOptInfo *index, IndexClauseSet *clauses, + bool useful_predicate, + ScanTypeControl scantype, + bool *skip_nonnative_saop, + bool *skip_lower_saop); +static List *build_paths_for_OR(PlannerInfo *root, RelOptInfo *rel, + List *clauses, List *other_clauses); +static List *generate_bitmap_or_paths(PlannerInfo *root, RelOptInfo *rel, + List *clauses, List *other_clauses); +static Path *choose_bitmap_and(PlannerInfo *root, RelOptInfo *rel, + List *paths); +static int path_usage_comparator(const void *a, const void *b); +static Cost bitmap_scan_cost_est(PlannerInfo *root, RelOptInfo *rel, + Path *ipath); +static Cost bitmap_and_cost_est(PlannerInfo *root, RelOptInfo *rel, + List *paths); +static PathClauseUsage *classify_index_clause_usage(Path *path, + List **clauselist); +static void find_indexpath_quals(Path *bitmapqual, List **quals, List **preds); +static int find_list_position(Node *node, List **nodelist); +static bool check_index_only(RelOptInfo *rel, IndexOptInfo *index); +static double get_loop_count(PlannerInfo *root, Index cur_relid, Relids outer_relids); +static double adjust_rowcount_for_semijoins(PlannerInfo *root, + Index cur_relid, + Index outer_relid, + double rowcount); +static double approximate_joinrel_size(PlannerInfo *root, Relids relids); +static void match_restriction_clauses_to_index(PlannerInfo *root, + IndexOptInfo *index, + IndexClauseSet *clauseset); +static void match_join_clauses_to_index(PlannerInfo *root, + RelOptInfo *rel, IndexOptInfo *index, + IndexClauseSet *clauseset, + List **joinorclauses); +static void match_eclass_clauses_to_index(PlannerInfo *root, + IndexOptInfo *index, + IndexClauseSet *clauseset); +static void match_clauses_to_index(PlannerInfo *root, + List *clauses, + IndexOptInfo *index, + IndexClauseSet *clauseset); +static void match_clause_to_index(PlannerInfo *root, + RestrictInfo *rinfo, + IndexOptInfo *index, + IndexClauseSet *clauseset); +static IndexClause *match_clause_to_indexcol(PlannerInfo *root, + RestrictInfo *rinfo, + int indexcol, + IndexOptInfo *index); +static IndexClause *match_boolean_index_clause(PlannerInfo *root, + RestrictInfo *rinfo, + int indexcol, IndexOptInfo *index); +static IndexClause *match_opclause_to_indexcol(PlannerInfo *root, + RestrictInfo *rinfo, + int indexcol, + IndexOptInfo *index); +static IndexClause *match_funcclause_to_indexcol(PlannerInfo *root, + RestrictInfo *rinfo, + int indexcol, + IndexOptInfo *index); +static IndexClause *get_index_clause_from_support(PlannerInfo *root, + RestrictInfo *rinfo, + Oid funcid, + int indexarg, + int indexcol, + IndexOptInfo *index); +static IndexClause *match_saopclause_to_indexcol(PlannerInfo *root, + RestrictInfo *rinfo, + int indexcol, + IndexOptInfo *index); +static IndexClause *match_rowcompare_to_indexcol(PlannerInfo *root, + RestrictInfo *rinfo, + int indexcol, + IndexOptInfo *index); +static IndexClause *expand_indexqual_rowcompare(PlannerInfo *root, + RestrictInfo *rinfo, + int indexcol, + IndexOptInfo *index, + Oid expr_op, + bool var_on_left); +static void match_pathkeys_to_index(IndexOptInfo *index, List *pathkeys, + List **orderby_clauses_p, + List **clause_columns_p); +static Expr *match_clause_to_ordering_op(IndexOptInfo *index, + int indexcol, Expr *clause, Oid pk_opfamily); +static bool ec_member_matches_indexcol(PlannerInfo *root, RelOptInfo *rel, + EquivalenceClass *ec, EquivalenceMember *em, + void *arg); + + +/* + * create_index_paths() + * Generate all interesting index paths for the given relation. + * Candidate paths are added to the rel's pathlist (using add_path). + * + * To be considered for an index scan, an index must match one or more + * restriction clauses or join clauses from the query's qual condition, + * or match the query's ORDER BY condition, or have a predicate that + * matches the query's qual condition. + * + * There are two basic kinds of index scans. A "plain" index scan uses + * only restriction clauses (possibly none at all) in its indexqual, + * so it can be applied in any context. A "parameterized" index scan uses + * join clauses (plus restriction clauses, if available) in its indexqual. + * When joining such a scan to one of the relations supplying the other + * variables used in its indexqual, the parameterized scan must appear as + * the inner relation of a nestloop join; it can't be used on the outer side, + * nor in a merge or hash join. In that context, values for the other rels' + * attributes are available and fixed during any one scan of the indexpath. + * + * An IndexPath is generated and submitted to add_path() for each plain or + * parameterized index scan this routine deems potentially interesting for + * the current query. + * + * 'rel' is the relation for which we want to generate index paths + * + * Note: check_index_predicates() must have been run previously for this rel. + * + * Note: in cases involving LATERAL references in the relation's tlist, it's + * possible that rel->lateral_relids is nonempty. Currently, we include + * lateral_relids into the parameterization reported for each path, but don't + * take it into account otherwise. The fact that any such rels *must* be + * available as parameter sources perhaps should influence our choices of + * index quals ... but for now, it doesn't seem worth troubling over. + * In particular, comments below about "unparameterized" paths should be read + * as meaning "unparameterized so far as the indexquals are concerned". + */ +void +create_index_paths(PlannerInfo *root, RelOptInfo *rel) +{ + List *indexpaths; + List *bitindexpaths; + List *bitjoinpaths; + List *joinorclauses; + IndexClauseSet rclauseset; + IndexClauseSet jclauseset; + IndexClauseSet eclauseset; + ListCell *lc; + + /* Skip the whole mess if no indexes */ + if (rel->indexlist == NIL) + return; + + /* Bitmap paths are collected and then dealt with at the end */ + bitindexpaths = bitjoinpaths = joinorclauses = NIL; + + /* Examine each index in turn */ + foreach(lc, rel->indexlist) + { + IndexOptInfo *index = (IndexOptInfo *) lfirst(lc); + + /* Protect limited-size array in IndexClauseSets */ + Assert(index->nkeycolumns <= INDEX_MAX_KEYS); + + /* + * Ignore partial indexes that do not match the query. + * (generate_bitmap_or_paths() might be able to do something with + * them, but that's of no concern here.) + */ + if (index->indpred != NIL && !index->predOK) + continue; + + /* + * Identify the restriction clauses that can match the index. + */ + MemSet(&rclauseset, 0, sizeof(rclauseset)); + match_restriction_clauses_to_index(root, index, &rclauseset); + + /* + * Build index paths from the restriction clauses. These will be + * non-parameterized paths. Plain paths go directly to add_path(), + * bitmap paths are added to bitindexpaths to be handled below. + */ + get_index_paths(root, rel, index, &rclauseset, + &bitindexpaths); + + /* + * Identify the join clauses that can match the index. For the moment + * we keep them separate from the restriction clauses. Note that this + * step finds only "loose" join clauses that have not been merged into + * EquivalenceClasses. Also, collect join OR clauses for later. + */ + MemSet(&jclauseset, 0, sizeof(jclauseset)); + match_join_clauses_to_index(root, rel, index, + &jclauseset, &joinorclauses); + + /* + * Look for EquivalenceClasses that can generate joinclauses matching + * the index. + */ + MemSet(&eclauseset, 0, sizeof(eclauseset)); + match_eclass_clauses_to_index(root, index, + &eclauseset); + + /* + * If we found any plain or eclass join clauses, build parameterized + * index paths using them. + */ + if (jclauseset.nonempty || eclauseset.nonempty) + consider_index_join_clauses(root, rel, index, + &rclauseset, + &jclauseset, + &eclauseset, + &bitjoinpaths); + } + + /* + * Generate BitmapOrPaths for any suitable OR-clauses present in the + * restriction list. Add these to bitindexpaths. + */ + indexpaths = generate_bitmap_or_paths(root, rel, + rel->baserestrictinfo, NIL); + bitindexpaths = list_concat(bitindexpaths, indexpaths); + + /* + * Likewise, generate BitmapOrPaths for any suitable OR-clauses present in + * the joinclause list. Add these to bitjoinpaths. + */ + indexpaths = generate_bitmap_or_paths(root, rel, + joinorclauses, rel->baserestrictinfo); + bitjoinpaths = list_concat(bitjoinpaths, indexpaths); + + /* + * If we found anything usable, generate a BitmapHeapPath for the most + * promising combination of restriction bitmap index paths. Note there + * will be only one such path no matter how many indexes exist. This + * should be sufficient since there's basically only one figure of merit + * (total cost) for such a path. + */ + if (bitindexpaths != NIL) + { + Path *bitmapqual; + BitmapHeapPath *bpath; + + bitmapqual = choose_bitmap_and(root, rel, bitindexpaths); + bpath = create_bitmap_heap_path(root, rel, bitmapqual, + rel->lateral_relids, 1.0, 0); + add_path(rel, (Path *) bpath); + + /* create a partial bitmap heap path */ + if (rel->consider_parallel && rel->lateral_relids == NULL) + create_partial_bitmap_paths(root, rel, bitmapqual); + } + + /* + * Likewise, if we found anything usable, generate BitmapHeapPaths for the + * most promising combinations of join bitmap index paths. Our strategy + * is to generate one such path for each distinct parameterization seen + * among the available bitmap index paths. This may look pretty + * expensive, but usually there won't be very many distinct + * parameterizations. (This logic is quite similar to that in + * consider_index_join_clauses, but we're working with whole paths not + * individual clauses.) + */ + if (bitjoinpaths != NIL) + { + List *all_path_outers; + ListCell *lc; + + /* Identify each distinct parameterization seen in bitjoinpaths */ + all_path_outers = NIL; + foreach(lc, bitjoinpaths) + { + Path *path = (Path *) lfirst(lc); + Relids required_outer = PATH_REQ_OUTER(path); + + if (!bms_equal_any(required_outer, all_path_outers)) + all_path_outers = lappend(all_path_outers, required_outer); + } + + /* Now, for each distinct parameterization set ... */ + foreach(lc, all_path_outers) + { + Relids max_outers = (Relids) lfirst(lc); + List *this_path_set; + Path *bitmapqual; + Relids required_outer; + double loop_count; + BitmapHeapPath *bpath; + ListCell *lcp; + + /* Identify all the bitmap join paths needing no more than that */ + this_path_set = NIL; + foreach(lcp, bitjoinpaths) + { + Path *path = (Path *) lfirst(lcp); + + if (bms_is_subset(PATH_REQ_OUTER(path), max_outers)) + this_path_set = lappend(this_path_set, path); + } + + /* + * Add in restriction bitmap paths, since they can be used + * together with any join paths. + */ + this_path_set = list_concat(this_path_set, bitindexpaths); + + /* Select best AND combination for this parameterization */ + bitmapqual = choose_bitmap_and(root, rel, this_path_set); + + /* And push that path into the mix */ + required_outer = PATH_REQ_OUTER(bitmapqual); + loop_count = get_loop_count(root, rel->relid, required_outer); + bpath = create_bitmap_heap_path(root, rel, bitmapqual, + required_outer, loop_count, 0); + add_path(rel, (Path *) bpath); + } + } +} + +/* + * consider_index_join_clauses + * Given sets of join clauses for an index, decide which parameterized + * index paths to build. + * + * Plain indexpaths are sent directly to add_path, while potential + * bitmap indexpaths are added to *bitindexpaths for later processing. + * + * 'rel' is the index's heap relation + * 'index' is the index for which we want to generate paths + * 'rclauseset' is the collection of indexable restriction clauses + * 'jclauseset' is the collection of indexable simple join clauses + * 'eclauseset' is the collection of indexable clauses from EquivalenceClasses + * '*bitindexpaths' is the list to add bitmap paths to + */ +static void +consider_index_join_clauses(PlannerInfo *root, RelOptInfo *rel, + IndexOptInfo *index, + IndexClauseSet *rclauseset, + IndexClauseSet *jclauseset, + IndexClauseSet *eclauseset, + List **bitindexpaths) +{ + int considered_clauses = 0; + List *considered_relids = NIL; + int indexcol; + + /* + * The strategy here is to identify every potentially useful set of outer + * rels that can provide indexable join clauses. For each such set, + * select all the join clauses available from those outer rels, add on all + * the indexable restriction clauses, and generate plain and/or bitmap + * index paths for that set of clauses. This is based on the assumption + * that it's always better to apply a clause as an indexqual than as a + * filter (qpqual); which is where an available clause would end up being + * applied if we omit it from the indexquals. + * + * This looks expensive, but in most practical cases there won't be very + * many distinct sets of outer rels to consider. As a safety valve when + * that's not true, we use a heuristic: limit the number of outer rel sets + * considered to a multiple of the number of clauses considered. (We'll + * always consider using each individual join clause, though.) + * + * For simplicity in selecting relevant clauses, we represent each set of + * outer rels as a maximum set of clause_relids --- that is, the indexed + * relation itself is also included in the relids set. considered_relids + * lists all relids sets we've already tried. + */ + for (indexcol = 0; indexcol < index->nkeycolumns; indexcol++) + { + /* Consider each applicable simple join clause */ + considered_clauses += list_length(jclauseset->indexclauses[indexcol]); + consider_index_join_outer_rels(root, rel, index, + rclauseset, jclauseset, eclauseset, + bitindexpaths, + jclauseset->indexclauses[indexcol], + considered_clauses, + &considered_relids); + /* Consider each applicable eclass join clause */ + considered_clauses += list_length(eclauseset->indexclauses[indexcol]); + consider_index_join_outer_rels(root, rel, index, + rclauseset, jclauseset, eclauseset, + bitindexpaths, + eclauseset->indexclauses[indexcol], + considered_clauses, + &considered_relids); + } +} + +/* + * consider_index_join_outer_rels + * Generate parameterized paths based on clause relids in the clause list. + * + * Workhorse for consider_index_join_clauses; see notes therein for rationale. + * + * 'rel', 'index', 'rclauseset', 'jclauseset', 'eclauseset', and + * 'bitindexpaths' as above + * 'indexjoinclauses' is a list of IndexClauses for join clauses + * 'considered_clauses' is the total number of clauses considered (so far) + * '*considered_relids' is a list of all relids sets already considered + */ +static void +consider_index_join_outer_rels(PlannerInfo *root, RelOptInfo *rel, + IndexOptInfo *index, + IndexClauseSet *rclauseset, + IndexClauseSet *jclauseset, + IndexClauseSet *eclauseset, + List **bitindexpaths, + List *indexjoinclauses, + int considered_clauses, + List **considered_relids) +{ + ListCell *lc; + + /* Examine relids of each joinclause in the given list */ + foreach(lc, indexjoinclauses) + { + IndexClause *iclause = (IndexClause *) lfirst(lc); + Relids clause_relids = iclause->rinfo->clause_relids; + EquivalenceClass *parent_ec = iclause->rinfo->parent_ec; + int num_considered_relids; + + /* If we already tried its relids set, no need to do so again */ + if (bms_equal_any(clause_relids, *considered_relids)) + continue; + + /* + * Generate the union of this clause's relids set with each + * previously-tried set. This ensures we try this clause along with + * every interesting subset of previous clauses. However, to avoid + * exponential growth of planning time when there are many clauses, + * limit the number of relid sets accepted to 10 * considered_clauses. + * + * Note: get_join_index_paths appends entries to *considered_relids, + * but we do not need to visit such newly-added entries within this + * loop, so we don't use foreach() here. No real harm would be done + * if we did visit them, since the subset check would reject them; but + * it would waste some cycles. + */ + num_considered_relids = list_length(*considered_relids); + for (int pos = 0; pos < num_considered_relids; pos++) + { + Relids oldrelids = (Relids) list_nth(*considered_relids, pos); + + /* + * If either is a subset of the other, no new set is possible. + * This isn't a complete test for redundancy, but it's easy and + * cheap. get_join_index_paths will check more carefully if we + * already generated the same relids set. + */ + if (bms_subset_compare(clause_relids, oldrelids) != BMS_DIFFERENT) + continue; + + /* + * If this clause was derived from an equivalence class, the + * clause list may contain other clauses derived from the same + * eclass. We should not consider that combining this clause with + * one of those clauses generates a usefully different + * parameterization; so skip if any clause derived from the same + * eclass would already have been included when using oldrelids. + */ + if (parent_ec && + eclass_already_used(parent_ec, oldrelids, + indexjoinclauses)) + continue; + + /* + * If the number of relid sets considered exceeds our heuristic + * limit, stop considering combinations of clauses. We'll still + * consider the current clause alone, though (below this loop). + */ + if (list_length(*considered_relids) >= 10 * considered_clauses) + break; + + /* OK, try the union set */ + get_join_index_paths(root, rel, index, + rclauseset, jclauseset, eclauseset, + bitindexpaths, + bms_union(clause_relids, oldrelids), + considered_relids); + } + + /* Also try this set of relids by itself */ + get_join_index_paths(root, rel, index, + rclauseset, jclauseset, eclauseset, + bitindexpaths, + clause_relids, + considered_relids); + } +} + +/* + * get_join_index_paths + * Generate index paths using clauses from the specified outer relations. + * In addition to generating paths, relids is added to *considered_relids + * if not already present. + * + * Workhorse for consider_index_join_clauses; see notes therein for rationale. + * + * 'rel', 'index', 'rclauseset', 'jclauseset', 'eclauseset', + * 'bitindexpaths', 'considered_relids' as above + * 'relids' is the current set of relids to consider (the target rel plus + * one or more outer rels) + */ +static void +get_join_index_paths(PlannerInfo *root, RelOptInfo *rel, + IndexOptInfo *index, + IndexClauseSet *rclauseset, + IndexClauseSet *jclauseset, + IndexClauseSet *eclauseset, + List **bitindexpaths, + Relids relids, + List **considered_relids) +{ + IndexClauseSet clauseset; + int indexcol; + + /* If we already considered this relids set, don't repeat the work */ + if (bms_equal_any(relids, *considered_relids)) + return; + + /* Identify indexclauses usable with this relids set */ + MemSet(&clauseset, 0, sizeof(clauseset)); + + for (indexcol = 0; indexcol < index->nkeycolumns; indexcol++) + { + ListCell *lc; + + /* First find applicable simple join clauses */ + foreach(lc, jclauseset->indexclauses[indexcol]) + { + IndexClause *iclause = (IndexClause *) lfirst(lc); + + if (bms_is_subset(iclause->rinfo->clause_relids, relids)) + clauseset.indexclauses[indexcol] = + lappend(clauseset.indexclauses[indexcol], iclause); + } + + /* + * Add applicable eclass join clauses. The clauses generated for each + * column are redundant (cf generate_implied_equalities_for_column), + * so we need at most one. This is the only exception to the general + * rule of using all available index clauses. + */ + foreach(lc, eclauseset->indexclauses[indexcol]) + { + IndexClause *iclause = (IndexClause *) lfirst(lc); + + if (bms_is_subset(iclause->rinfo->clause_relids, relids)) + { + clauseset.indexclauses[indexcol] = + lappend(clauseset.indexclauses[indexcol], iclause); + break; + } + } + + /* Add restriction clauses */ + clauseset.indexclauses[indexcol] = + list_concat(clauseset.indexclauses[indexcol], + rclauseset->indexclauses[indexcol]); + + if (clauseset.indexclauses[indexcol] != NIL) + clauseset.nonempty = true; + } + + /* We should have found something, else caller passed silly relids */ + Assert(clauseset.nonempty); + + /* Build index path(s) using the collected set of clauses */ + get_index_paths(root, rel, index, &clauseset, bitindexpaths); + + /* + * Remember we considered paths for this set of relids. + */ + *considered_relids = lappend(*considered_relids, relids); +} + +/* + * eclass_already_used + * True if any join clause usable with oldrelids was generated from + * the specified equivalence class. + */ +static bool +eclass_already_used(EquivalenceClass *parent_ec, Relids oldrelids, + List *indexjoinclauses) +{ + ListCell *lc; + + foreach(lc, indexjoinclauses) + { + IndexClause *iclause = (IndexClause *) lfirst(lc); + RestrictInfo *rinfo = iclause->rinfo; + + if (rinfo->parent_ec == parent_ec && + bms_is_subset(rinfo->clause_relids, oldrelids)) + return true; + } + return false; +} + +/* + * bms_equal_any + * True if relids is bms_equal to any member of relids_list + * + * Perhaps this should be in bitmapset.c someday. + */ +static bool +bms_equal_any(Relids relids, List *relids_list) +{ + ListCell *lc; + + foreach(lc, relids_list) + { + if (bms_equal(relids, (Relids) lfirst(lc))) + return true; + } + return false; +} + + +/* + * get_index_paths + * Given an index and a set of index clauses for it, construct IndexPaths. + * + * Plain indexpaths are sent directly to add_path, while potential + * bitmap indexpaths are added to *bitindexpaths for later processing. + * + * This is a fairly simple frontend to build_index_paths(). Its reason for + * existence is mainly to handle ScalarArrayOpExpr quals properly. If the + * index AM supports them natively, we should just include them in simple + * index paths. If not, we should exclude them while building simple index + * paths, and then make a separate attempt to include them in bitmap paths. + * Furthermore, we should consider excluding lower-order ScalarArrayOpExpr + * quals so as to create ordered paths. + */ +static void +get_index_paths(PlannerInfo *root, RelOptInfo *rel, + IndexOptInfo *index, IndexClauseSet *clauses, + List **bitindexpaths) +{ + List *indexpaths; + bool skip_nonnative_saop = false; + bool skip_lower_saop = false; + ListCell *lc; + + /* + * Build simple index paths using the clauses. Allow ScalarArrayOpExpr + * clauses only if the index AM supports them natively, and skip any such + * clauses for index columns after the first (so that we produce ordered + * paths if possible). + */ + indexpaths = build_index_paths(root, rel, + index, clauses, + index->predOK, + ST_ANYSCAN, + &skip_nonnative_saop, + &skip_lower_saop); + + /* + * If we skipped any lower-order ScalarArrayOpExprs on an index with an AM + * that supports them, then try again including those clauses. This will + * produce paths with more selectivity but no ordering. + */ + if (skip_lower_saop) + { + indexpaths = list_concat(indexpaths, + build_index_paths(root, rel, + index, clauses, + index->predOK, + ST_ANYSCAN, + &skip_nonnative_saop, + NULL)); + } + + /* + * Submit all the ones that can form plain IndexScan plans to add_path. (A + * plain IndexPath can represent either a plain IndexScan or an + * IndexOnlyScan, but for our purposes here that distinction does not + * matter. However, some of the indexes might support only bitmap scans, + * and those we mustn't submit to add_path here.) + * + * Also, pick out the ones that are usable as bitmap scans. For that, we + * must discard indexes that don't support bitmap scans, and we also are + * only interested in paths that have some selectivity; we should discard + * anything that was generated solely for ordering purposes. + */ + foreach(lc, indexpaths) + { + IndexPath *ipath = (IndexPath *) lfirst(lc); + + if (index->amhasgettuple) + add_path(rel, (Path *) ipath); + + if (index->amhasgetbitmap && + (ipath->path.pathkeys == NIL || + ipath->indexselectivity < 1.0)) + *bitindexpaths = lappend(*bitindexpaths, ipath); + } + + /* + * If there were ScalarArrayOpExpr clauses that the index can't handle + * natively, generate bitmap scan paths relying on executor-managed + * ScalarArrayOpExpr. + */ + if (skip_nonnative_saop) + { + indexpaths = build_index_paths(root, rel, + index, clauses, + false, + ST_BITMAPSCAN, + NULL, + NULL); + *bitindexpaths = list_concat(*bitindexpaths, indexpaths); + } +} + +/* + * build_index_paths + * Given an index and a set of index clauses for it, construct zero + * or more IndexPaths. It also constructs zero or more partial IndexPaths. + * + * We return a list of paths because (1) this routine checks some cases + * that should cause us to not generate any IndexPath, and (2) in some + * cases we want to consider both a forward and a backward scan, so as + * to obtain both sort orders. Note that the paths are just returned + * to the caller and not immediately fed to add_path(). + * + * At top level, useful_predicate should be exactly the index's predOK flag + * (ie, true if it has a predicate that was proven from the restriction + * clauses). When working on an arm of an OR clause, useful_predicate + * should be true if the predicate required the current OR list to be proven. + * Note that this routine should never be called at all if the index has an + * unprovable predicate. + * + * scantype indicates whether we want to create plain indexscans, bitmap + * indexscans, or both. When it's ST_BITMAPSCAN, we will not consider + * index ordering while deciding if a Path is worth generating. + * + * If skip_nonnative_saop is non-NULL, we ignore ScalarArrayOpExpr clauses + * unless the index AM supports them directly, and we set *skip_nonnative_saop + * to true if we found any such clauses (caller must initialize the variable + * to false). If it's NULL, we do not ignore ScalarArrayOpExpr clauses. + * + * If skip_lower_saop is non-NULL, we ignore ScalarArrayOpExpr clauses for + * non-first index columns, and we set *skip_lower_saop to true if we found + * any such clauses (caller must initialize the variable to false). If it's + * NULL, we do not ignore non-first ScalarArrayOpExpr clauses, but they will + * result in considering the scan's output to be unordered. + * + * 'rel' is the index's heap relation + * 'index' is the index for which we want to generate paths + * 'clauses' is the collection of indexable clauses (IndexClause nodes) + * 'useful_predicate' indicates whether the index has a useful predicate + * 'scantype' indicates whether we need plain or bitmap scan support + * 'skip_nonnative_saop' indicates whether to accept SAOP if index AM doesn't + * 'skip_lower_saop' indicates whether to accept non-first-column SAOP + */ +static List * +build_index_paths(PlannerInfo *root, RelOptInfo *rel, + IndexOptInfo *index, IndexClauseSet *clauses, + bool useful_predicate, + ScanTypeControl scantype, + bool *skip_nonnative_saop, + bool *skip_lower_saop) +{ + List *result = NIL; + IndexPath *ipath; + List *index_clauses; + Relids outer_relids; + double loop_count; + List *orderbyclauses; + List *orderbyclausecols; + List *index_pathkeys; + List *useful_pathkeys; + bool found_lower_saop_clause; + bool pathkeys_possibly_useful; + bool index_is_ordered; + bool index_only_scan; + int indexcol; + + /* + * Check that index supports the desired scan type(s) + */ + switch (scantype) + { + case ST_INDEXSCAN: + if (!index->amhasgettuple) + return NIL; + break; + case ST_BITMAPSCAN: + if (!index->amhasgetbitmap) + return NIL; + break; + case ST_ANYSCAN: + /* either or both are OK */ + break; + } + + /* + * 1. Combine the per-column IndexClause lists into an overall list. + * + * In the resulting list, clauses are ordered by index key, so that the + * column numbers form a nondecreasing sequence. (This order is depended + * on by btree and possibly other places.) The list can be empty, if the + * index AM allows that. + * + * found_lower_saop_clause is set true if we accept a ScalarArrayOpExpr + * index clause for a non-first index column. This prevents us from + * assuming that the scan result is ordered. (Actually, the result is + * still ordered if there are equality constraints for all earlier + * columns, but it seems too expensive and non-modular for this code to be + * aware of that refinement.) + * + * We also build a Relids set showing which outer rels are required by the + * selected clauses. Any lateral_relids are included in that, but not + * otherwise accounted for. + */ + index_clauses = NIL; + found_lower_saop_clause = false; + outer_relids = bms_copy(rel->lateral_relids); + for (indexcol = 0; indexcol < index->nkeycolumns; indexcol++) + { + ListCell *lc; + + foreach(lc, clauses->indexclauses[indexcol]) + { + IndexClause *iclause = (IndexClause *) lfirst(lc); + RestrictInfo *rinfo = iclause->rinfo; + + /* We might need to omit ScalarArrayOpExpr clauses */ + if (IsA(rinfo->clause, ScalarArrayOpExpr)) + { + if (!index->amsearcharray) + { + if (skip_nonnative_saop) + { + /* Ignore because not supported by index */ + *skip_nonnative_saop = true; + continue; + } + /* Caller had better intend this only for bitmap scan */ + Assert(scantype == ST_BITMAPSCAN); + } + if (indexcol > 0) + { + if (skip_lower_saop) + { + /* Caller doesn't want to lose index ordering */ + *skip_lower_saop = true; + continue; + } + found_lower_saop_clause = true; + } + } + + /* OK to include this clause */ + index_clauses = lappend(index_clauses, iclause); + outer_relids = bms_add_members(outer_relids, + rinfo->clause_relids); + } + + /* + * If no clauses match the first index column, check for amoptionalkey + * restriction. We can't generate a scan over an index with + * amoptionalkey = false unless there's at least one index clause. + * (When working on columns after the first, this test cannot fail. It + * is always okay for columns after the first to not have any + * clauses.) + */ + if (index_clauses == NIL && !index->amoptionalkey) + return NIL; + } + + /* We do not want the index's rel itself listed in outer_relids */ + outer_relids = bms_del_member(outer_relids, rel->relid); + /* Enforce convention that outer_relids is exactly NULL if empty */ + if (bms_is_empty(outer_relids)) + outer_relids = NULL; + + /* Compute loop_count for cost estimation purposes */ + loop_count = get_loop_count(root, rel->relid, outer_relids); + + /* + * 2. Compute pathkeys describing index's ordering, if any, then see how + * many of them are actually useful for this query. This is not relevant + * if we are only trying to build bitmap indexscans, nor if we have to + * assume the scan is unordered. + */ + pathkeys_possibly_useful = (scantype != ST_BITMAPSCAN && + !found_lower_saop_clause && + has_useful_pathkeys(root, rel)); + index_is_ordered = (index->sortopfamily != NULL); + if (index_is_ordered && pathkeys_possibly_useful) + { + index_pathkeys = build_index_pathkeys(root, index, + ForwardScanDirection); + useful_pathkeys = truncate_useless_pathkeys(root, rel, + index_pathkeys); + orderbyclauses = NIL; + orderbyclausecols = NIL; + } + else if (index->amcanorderbyop && pathkeys_possibly_useful) + { + /* see if we can generate ordering operators for query_pathkeys */ + match_pathkeys_to_index(index, root->query_pathkeys, + &orderbyclauses, + &orderbyclausecols); + if (orderbyclauses) + useful_pathkeys = root->query_pathkeys; + else + useful_pathkeys = NIL; + } + else + { + useful_pathkeys = NIL; + orderbyclauses = NIL; + orderbyclausecols = NIL; + } + + /* + * 3. Check if an index-only scan is possible. If we're not building + * plain indexscans, this isn't relevant since bitmap scans don't support + * index data retrieval anyway. + */ + index_only_scan = (scantype != ST_BITMAPSCAN && + check_index_only(rel, index)); + + /* + * 4. Generate an indexscan path if there are relevant restriction clauses + * in the current clauses, OR the index ordering is potentially useful for + * later merging or final output ordering, OR the index has a useful + * predicate, OR an index-only scan is possible. + */ + if (index_clauses != NIL || useful_pathkeys != NIL || useful_predicate || + index_only_scan) + { + ipath = create_index_path(root, index, + index_clauses, + orderbyclauses, + orderbyclausecols, + useful_pathkeys, + index_is_ordered ? + ForwardScanDirection : + NoMovementScanDirection, + index_only_scan, + outer_relids, + loop_count, + false); + result = lappend(result, ipath); + + /* + * If appropriate, consider parallel index scan. We don't allow + * parallel index scan for bitmap index scans. + */ + if (index->amcanparallel && + rel->consider_parallel && outer_relids == NULL && + scantype != ST_BITMAPSCAN) + { + ipath = create_index_path(root, index, + index_clauses, + orderbyclauses, + orderbyclausecols, + useful_pathkeys, + index_is_ordered ? + ForwardScanDirection : + NoMovementScanDirection, + index_only_scan, + outer_relids, + loop_count, + true); + + /* + * if, after costing the path, we find that it's not worth using + * parallel workers, just free it. + */ + if (ipath->path.parallel_workers > 0) + add_partial_path(rel, (Path *) ipath); + else + pfree(ipath); + } + } + + /* + * 5. If the index is ordered, a backwards scan might be interesting. + */ + if (index_is_ordered && pathkeys_possibly_useful) + { + index_pathkeys = build_index_pathkeys(root, index, + BackwardScanDirection); + useful_pathkeys = truncate_useless_pathkeys(root, rel, + index_pathkeys); + if (useful_pathkeys != NIL) + { + ipath = create_index_path(root, index, + index_clauses, + NIL, + NIL, + useful_pathkeys, + BackwardScanDirection, + index_only_scan, + outer_relids, + loop_count, + false); + result = lappend(result, ipath); + + /* If appropriate, consider parallel index scan */ + if (index->amcanparallel && + rel->consider_parallel && outer_relids == NULL && + scantype != ST_BITMAPSCAN) + { + ipath = create_index_path(root, index, + index_clauses, + NIL, + NIL, + useful_pathkeys, + BackwardScanDirection, + index_only_scan, + outer_relids, + loop_count, + true); + + /* + * if, after costing the path, we find that it's not worth + * using parallel workers, just free it. + */ + if (ipath->path.parallel_workers > 0) + add_partial_path(rel, (Path *) ipath); + else + pfree(ipath); + } + } + } + + return result; +} + +/* + * build_paths_for_OR + * Given a list of restriction clauses from one arm of an OR clause, + * construct all matching IndexPaths for the relation. + * + * Here we must scan all indexes of the relation, since a bitmap OR tree + * can use multiple indexes. + * + * The caller actually supplies two lists of restriction clauses: some + * "current" ones and some "other" ones. Both lists can be used freely + * to match keys of the index, but an index must use at least one of the + * "current" clauses to be considered usable. The motivation for this is + * examples like + * WHERE (x = 42) AND (... OR (y = 52 AND z = 77) OR ....) + * While we are considering the y/z subclause of the OR, we can use "x = 42" + * as one of the available index conditions; but we shouldn't match the + * subclause to any index on x alone, because such a Path would already have + * been generated at the upper level. So we could use an index on x,y,z + * or an index on x,y for the OR subclause, but not an index on just x. + * When dealing with a partial index, a match of the index predicate to + * one of the "current" clauses also makes the index usable. + * + * 'rel' is the relation for which we want to generate index paths + * 'clauses' is the current list of clauses (RestrictInfo nodes) + * 'other_clauses' is the list of additional upper-level clauses + */ +static List * +build_paths_for_OR(PlannerInfo *root, RelOptInfo *rel, + List *clauses, List *other_clauses) +{ + List *result = NIL; + List *all_clauses = NIL; /* not computed till needed */ + ListCell *lc; + + foreach(lc, rel->indexlist) + { + IndexOptInfo *index = (IndexOptInfo *) lfirst(lc); + IndexClauseSet clauseset; + List *indexpaths; + bool useful_predicate; + + /* Ignore index if it doesn't support bitmap scans */ + if (!index->amhasgetbitmap) + continue; + + /* + * Ignore partial indexes that do not match the query. If a partial + * index is marked predOK then we know it's OK. Otherwise, we have to + * test whether the added clauses are sufficient to imply the + * predicate. If so, we can use the index in the current context. + * + * We set useful_predicate to true iff the predicate was proven using + * the current set of clauses. This is needed to prevent matching a + * predOK index to an arm of an OR, which would be a legal but + * pointlessly inefficient plan. (A better plan will be generated by + * just scanning the predOK index alone, no OR.) + */ + useful_predicate = false; + if (index->indpred != NIL) + { + if (index->predOK) + { + /* Usable, but don't set useful_predicate */ + } + else + { + /* Form all_clauses if not done already */ + if (all_clauses == NIL) + all_clauses = list_concat_copy(clauses, other_clauses); + + if (!predicate_implied_by(index->indpred, all_clauses, false)) + continue; /* can't use it at all */ + + if (!predicate_implied_by(index->indpred, other_clauses, false)) + useful_predicate = true; + } + } + + /* + * Identify the restriction clauses that can match the index. + */ + MemSet(&clauseset, 0, sizeof(clauseset)); + match_clauses_to_index(root, clauses, index, &clauseset); + + /* + * If no matches so far, and the index predicate isn't useful, we + * don't want it. + */ + if (!clauseset.nonempty && !useful_predicate) + continue; + + /* + * Add "other" restriction clauses to the clauseset. + */ + match_clauses_to_index(root, other_clauses, index, &clauseset); + + /* + * Construct paths if possible. + */ + indexpaths = build_index_paths(root, rel, + index, &clauseset, + useful_predicate, + ST_BITMAPSCAN, + NULL, + NULL); + result = list_concat(result, indexpaths); + } + + return result; +} + +/* + * generate_bitmap_or_paths + * Look through the list of clauses to find OR clauses, and generate + * a BitmapOrPath for each one we can handle that way. Return a list + * of the generated BitmapOrPaths. + * + * other_clauses is a list of additional clauses that can be assumed true + * for the purpose of generating indexquals, but are not to be searched for + * ORs. (See build_paths_for_OR() for motivation.) + */ +static List * +generate_bitmap_or_paths(PlannerInfo *root, RelOptInfo *rel, + List *clauses, List *other_clauses) +{ + List *result = NIL; + List *all_clauses; + ListCell *lc; + + /* + * We can use both the current and other clauses as context for + * build_paths_for_OR; no need to remove ORs from the lists. + */ + all_clauses = list_concat_copy(clauses, other_clauses); + + foreach(lc, clauses) + { + RestrictInfo *rinfo = lfirst_node(RestrictInfo, lc); + List *pathlist; + Path *bitmapqual; + ListCell *j; + + /* Ignore RestrictInfos that aren't ORs */ + if (!restriction_is_or_clause(rinfo)) + continue; + + /* + * We must be able to match at least one index to each of the arms of + * the OR, else we can't use it. + */ + pathlist = NIL; + foreach(j, ((BoolExpr *) rinfo->orclause)->args) + { + Node *orarg = (Node *) lfirst(j); + List *indlist; + + /* OR arguments should be ANDs or sub-RestrictInfos */ + if (is_andclause(orarg)) + { + List *andargs = ((BoolExpr *) orarg)->args; + + indlist = build_paths_for_OR(root, rel, + andargs, + all_clauses); + + /* Recurse in case there are sub-ORs */ + indlist = list_concat(indlist, + generate_bitmap_or_paths(root, rel, + andargs, + all_clauses)); + } + else + { + RestrictInfo *rinfo = castNode(RestrictInfo, orarg); + List *orargs; + + Assert(!restriction_is_or_clause(rinfo)); + orargs = list_make1(rinfo); + + indlist = build_paths_for_OR(root, rel, + orargs, + all_clauses); + } + + /* + * If nothing matched this arm, we can't do anything with this OR + * clause. + */ + if (indlist == NIL) + { + pathlist = NIL; + break; + } + + /* + * OK, pick the most promising AND combination, and add it to + * pathlist. + */ + bitmapqual = choose_bitmap_and(root, rel, indlist); + pathlist = lappend(pathlist, bitmapqual); + } + + /* + * If we have a match for every arm, then turn them into a + * BitmapOrPath, and add to result list. + */ + if (pathlist != NIL) + { + bitmapqual = (Path *) create_bitmap_or_path(root, rel, pathlist); + result = lappend(result, bitmapqual); + } + } + + return result; +} + + +/* + * choose_bitmap_and + * Given a nonempty list of bitmap paths, AND them into one path. + * + * This is a nontrivial decision since we can legally use any subset of the + * given path set. We want to choose a good tradeoff between selectivity + * and cost of computing the bitmap. + * + * The result is either a single one of the inputs, or a BitmapAndPath + * combining multiple inputs. + */ +static Path * +choose_bitmap_and(PlannerInfo *root, RelOptInfo *rel, List *paths) +{ + int npaths = list_length(paths); + PathClauseUsage **pathinfoarray; + PathClauseUsage *pathinfo; + List *clauselist; + List *bestpaths = NIL; + Cost bestcost = 0; + int i, + j; + ListCell *l; + + Assert(npaths > 0); /* else caller error */ + if (npaths == 1) + return (Path *) linitial(paths); /* easy case */ + + /* + * In theory we should consider every nonempty subset of the given paths. + * In practice that seems like overkill, given the crude nature of the + * estimates, not to mention the possible effects of higher-level AND and + * OR clauses. Moreover, it's completely impractical if there are a large + * number of paths, since the work would grow as O(2^N). + * + * As a heuristic, we first check for paths using exactly the same sets of + * WHERE clauses + index predicate conditions, and reject all but the + * cheapest-to-scan in any such group. This primarily gets rid of indexes + * that include the interesting columns but also irrelevant columns. (In + * situations where the DBA has gone overboard on creating variant + * indexes, this can make for a very large reduction in the number of + * paths considered further.) + * + * We then sort the surviving paths with the cheapest-to-scan first, and + * for each path, consider using that path alone as the basis for a bitmap + * scan. Then we consider bitmap AND scans formed from that path plus + * each subsequent (higher-cost) path, adding on a subsequent path if it + * results in a reduction in the estimated total scan cost. This means we + * consider about O(N^2) rather than O(2^N) path combinations, which is + * quite tolerable, especially given than N is usually reasonably small + * because of the prefiltering step. The cheapest of these is returned. + * + * We will only consider AND combinations in which no two indexes use the + * same WHERE clause. This is a bit of a kluge: it's needed because + * costsize.c and clausesel.c aren't very smart about redundant clauses. + * They will usually double-count the redundant clauses, producing a + * too-small selectivity that makes a redundant AND step look like it + * reduces the total cost. Perhaps someday that code will be smarter and + * we can remove this limitation. (But note that this also defends + * against flat-out duplicate input paths, which can happen because + * match_join_clauses_to_index will find the same OR join clauses that + * extract_restriction_or_clauses has pulled OR restriction clauses out + * of.) + * + * For the same reason, we reject AND combinations in which an index + * predicate clause duplicates another clause. Here we find it necessary + * to be even stricter: we'll reject a partial index if any of its + * predicate clauses are implied by the set of WHERE clauses and predicate + * clauses used so far. This covers cases such as a condition "x = 42" + * used with a plain index, followed by a clauseless scan of a partial + * index "WHERE x >= 40 AND x < 50". The partial index has been accepted + * only because "x = 42" was present, and so allowing it would partially + * double-count selectivity. (We could use predicate_implied_by on + * regular qual clauses too, to have a more intelligent, but much more + * expensive, check for redundancy --- but in most cases simple equality + * seems to suffice.) + */ + + /* + * Extract clause usage info and detect any paths that use exactly the + * same set of clauses; keep only the cheapest-to-scan of any such groups. + * The surviving paths are put into an array for qsort'ing. + */ + pathinfoarray = (PathClauseUsage **) + palloc(npaths * sizeof(PathClauseUsage *)); + clauselist = NIL; + npaths = 0; + foreach(l, paths) + { + Path *ipath = (Path *) lfirst(l); + + pathinfo = classify_index_clause_usage(ipath, &clauselist); + + /* If it's unclassifiable, treat it as distinct from all others */ + if (pathinfo->unclassifiable) + { + pathinfoarray[npaths++] = pathinfo; + continue; + } + + for (i = 0; i < npaths; i++) + { + if (!pathinfoarray[i]->unclassifiable && + bms_equal(pathinfo->clauseids, pathinfoarray[i]->clauseids)) + break; + } + if (i < npaths) + { + /* duplicate clauseids, keep the cheaper one */ + Cost ncost; + Cost ocost; + Selectivity nselec; + Selectivity oselec; + + cost_bitmap_tree_node(pathinfo->path, &ncost, &nselec); + cost_bitmap_tree_node(pathinfoarray[i]->path, &ocost, &oselec); + if (ncost < ocost) + pathinfoarray[i] = pathinfo; + } + else + { + /* not duplicate clauseids, add to array */ + pathinfoarray[npaths++] = pathinfo; + } + } + + /* If only one surviving path, we're done */ + if (npaths == 1) + return pathinfoarray[0]->path; + + /* Sort the surviving paths by index access cost */ + qsort(pathinfoarray, npaths, sizeof(PathClauseUsage *), + path_usage_comparator); + + /* + * For each surviving index, consider it as an "AND group leader", and see + * whether adding on any of the later indexes results in an AND path with + * cheaper total cost than before. Then take the cheapest AND group. + * + * Note: paths that are either clauseless or unclassifiable will have + * empty clauseids, so that they will not be rejected by the clauseids + * filter here, nor will they cause later paths to be rejected by it. + */ + for (i = 0; i < npaths; i++) + { + Cost costsofar; + List *qualsofar; + Bitmapset *clauseidsofar; + + pathinfo = pathinfoarray[i]; + paths = list_make1(pathinfo->path); + costsofar = bitmap_scan_cost_est(root, rel, pathinfo->path); + qualsofar = list_concat_copy(pathinfo->quals, pathinfo->preds); + clauseidsofar = bms_copy(pathinfo->clauseids); + + for (j = i + 1; j < npaths; j++) + { + Cost newcost; + + pathinfo = pathinfoarray[j]; + /* Check for redundancy */ + if (bms_overlap(pathinfo->clauseids, clauseidsofar)) + continue; /* consider it redundant */ + if (pathinfo->preds) + { + bool redundant = false; + + /* we check each predicate clause separately */ + foreach(l, pathinfo->preds) + { + Node *np = (Node *) lfirst(l); + + if (predicate_implied_by(list_make1(np), qualsofar, false)) + { + redundant = true; + break; /* out of inner foreach loop */ + } + } + if (redundant) + continue; + } + /* tentatively add new path to paths, so we can estimate cost */ + paths = lappend(paths, pathinfo->path); + newcost = bitmap_and_cost_est(root, rel, paths); + if (newcost < costsofar) + { + /* keep new path in paths, update subsidiary variables */ + costsofar = newcost; + qualsofar = list_concat(qualsofar, pathinfo->quals); + qualsofar = list_concat(qualsofar, pathinfo->preds); + clauseidsofar = bms_add_members(clauseidsofar, + pathinfo->clauseids); + } + else + { + /* reject new path, remove it from paths list */ + paths = list_truncate(paths, list_length(paths) - 1); + } + } + + /* Keep the cheapest AND-group (or singleton) */ + if (i == 0 || costsofar < bestcost) + { + bestpaths = paths; + bestcost = costsofar; + } + + /* some easy cleanup (we don't try real hard though) */ + list_free(qualsofar); + } + + if (list_length(bestpaths) == 1) + return (Path *) linitial(bestpaths); /* no need for AND */ + return (Path *) create_bitmap_and_path(root, rel, bestpaths); +} + +/* qsort comparator to sort in increasing index access cost order */ +static int +path_usage_comparator(const void *a, const void *b) +{ + PathClauseUsage *pa = *(PathClauseUsage *const *) a; + PathClauseUsage *pb = *(PathClauseUsage *const *) b; + Cost acost; + Cost bcost; + Selectivity aselec; + Selectivity bselec; + + cost_bitmap_tree_node(pa->path, &acost, &aselec); + cost_bitmap_tree_node(pb->path, &bcost, &bselec); + + /* + * If costs are the same, sort by selectivity. + */ + if (acost < bcost) + return -1; + if (acost > bcost) + return 1; + + if (aselec < bselec) + return -1; + if (aselec > bselec) + return 1; + + return 0; +} + +/* + * Estimate the cost of actually executing a bitmap scan with a single + * index path (which could be a BitmapAnd or BitmapOr node). + */ +static Cost +bitmap_scan_cost_est(PlannerInfo *root, RelOptInfo *rel, Path *ipath) +{ + BitmapHeapPath bpath; + + /* Set up a dummy BitmapHeapPath */ + bpath.path.type = T_BitmapHeapPath; + bpath.path.pathtype = T_BitmapHeapScan; + bpath.path.parent = rel; + bpath.path.pathtarget = rel->reltarget; + bpath.path.param_info = ipath->param_info; + bpath.path.pathkeys = NIL; + bpath.bitmapqual = ipath; + + /* + * Check the cost of temporary path without considering parallelism. + * Parallel bitmap heap path will be considered at later stage. + */ + bpath.path.parallel_workers = 0; + + /* Now we can do cost_bitmap_heap_scan */ + cost_bitmap_heap_scan(&bpath.path, root, rel, + bpath.path.param_info, + ipath, + get_loop_count(root, rel->relid, + PATH_REQ_OUTER(ipath))); + + return bpath.path.total_cost; +} + +/* + * Estimate the cost of actually executing a BitmapAnd scan with the given + * inputs. + */ +static Cost +bitmap_and_cost_est(PlannerInfo *root, RelOptInfo *rel, List *paths) +{ + BitmapAndPath *apath; + + /* + * Might as well build a real BitmapAndPath here, as the work is slightly + * too complicated to be worth repeating just to save one palloc. + */ + apath = create_bitmap_and_path(root, rel, paths); + + return bitmap_scan_cost_est(root, rel, (Path *) apath); +} + + +/* + * classify_index_clause_usage + * Construct a PathClauseUsage struct describing the WHERE clauses and + * index predicate clauses used by the given indexscan path. + * We consider two clauses the same if they are equal(). + * + * At some point we might want to migrate this info into the Path data + * structure proper, but for the moment it's only needed within + * choose_bitmap_and(). + * + * *clauselist is used and expanded as needed to identify all the distinct + * clauses seen across successive calls. Caller must initialize it to NIL + * before first call of a set. + */ +static PathClauseUsage * +classify_index_clause_usage(Path *path, List **clauselist) +{ + PathClauseUsage *result; + Bitmapset *clauseids; + ListCell *lc; + + result = (PathClauseUsage *) palloc(sizeof(PathClauseUsage)); + result->path = path; + + /* Recursively find the quals and preds used by the path */ + result->quals = NIL; + result->preds = NIL; + find_indexpath_quals(path, &result->quals, &result->preds); + + /* + * Some machine-generated queries have outlandish numbers of qual clauses. + * To avoid getting into O(N^2) behavior even in this preliminary + * classification step, we want to limit the number of entries we can + * accumulate in *clauselist. Treat any path with more than 100 quals + + * preds as unclassifiable, which will cause calling code to consider it + * distinct from all other paths. + */ + if (list_length(result->quals) + list_length(result->preds) > 100) + { + result->clauseids = NULL; + result->unclassifiable = true; + return result; + } + + /* Build up a bitmapset representing the quals and preds */ + clauseids = NULL; + foreach(lc, result->quals) + { + Node *node = (Node *) lfirst(lc); + + clauseids = bms_add_member(clauseids, + find_list_position(node, clauselist)); + } + foreach(lc, result->preds) + { + Node *node = (Node *) lfirst(lc); + + clauseids = bms_add_member(clauseids, + find_list_position(node, clauselist)); + } + result->clauseids = clauseids; + result->unclassifiable = false; + + return result; +} + + +/* + * find_indexpath_quals + * + * Given the Path structure for a plain or bitmap indexscan, extract lists + * of all the index clauses and index predicate conditions used in the Path. + * These are appended to the initial contents of *quals and *preds (hence + * caller should initialize those to NIL). + * + * Note we are not trying to produce an accurate representation of the AND/OR + * semantics of the Path, but just find out all the base conditions used. + * + * The result lists contain pointers to the expressions used in the Path, + * but all the list cells are freshly built, so it's safe to destructively + * modify the lists (eg, by concat'ing with other lists). + */ +static void +find_indexpath_quals(Path *bitmapqual, List **quals, List **preds) +{ + if (IsA(bitmapqual, BitmapAndPath)) + { + BitmapAndPath *apath = (BitmapAndPath *) bitmapqual; + ListCell *l; + + foreach(l, apath->bitmapquals) + { + find_indexpath_quals((Path *) lfirst(l), quals, preds); + } + } + else if (IsA(bitmapqual, BitmapOrPath)) + { + BitmapOrPath *opath = (BitmapOrPath *) bitmapqual; + ListCell *l; + + foreach(l, opath->bitmapquals) + { + find_indexpath_quals((Path *) lfirst(l), quals, preds); + } + } + else if (IsA(bitmapqual, IndexPath)) + { + IndexPath *ipath = (IndexPath *) bitmapqual; + ListCell *l; + + foreach(l, ipath->indexclauses) + { + IndexClause *iclause = (IndexClause *) lfirst(l); + + *quals = lappend(*quals, iclause->rinfo->clause); + } + *preds = list_concat(*preds, ipath->indexinfo->indpred); + } + else + elog(ERROR, "unrecognized node type: %d", nodeTag(bitmapqual)); +} + + +/* + * find_list_position + * Return the given node's position (counting from 0) in the given + * list of nodes. If it's not equal() to any existing list member, + * add it at the end, and return that position. + */ +static int +find_list_position(Node *node, List **nodelist) +{ + int i; + ListCell *lc; + + i = 0; + foreach(lc, *nodelist) + { + Node *oldnode = (Node *) lfirst(lc); + + if (equal(node, oldnode)) + return i; + i++; + } + + *nodelist = lappend(*nodelist, node); + + return i; +} + + +/* + * check_index_only + * Determine whether an index-only scan is possible for this index. + */ +static bool +check_index_only(RelOptInfo *rel, IndexOptInfo *index) +{ + bool result; + Bitmapset *attrs_used = NULL; + Bitmapset *index_canreturn_attrs = NULL; + ListCell *lc; + int i; + + /* Index-only scans must be enabled */ + if (!enable_indexonlyscan) + return false; + + /* + * Check that all needed attributes of the relation are available from the + * index. + */ + + /* + * First, identify all the attributes needed for joins or final output. + * Note: we must look at rel's targetlist, not the attr_needed data, + * because attr_needed isn't computed for inheritance child rels. + */ + pull_varattnos((Node *) rel->reltarget->exprs, rel->relid, &attrs_used); + + /* + * Add all the attributes used by restriction clauses; but consider only + * those clauses not implied by the index predicate, since ones that are + * so implied don't need to be checked explicitly in the plan. + * + * Note: attributes used only in index quals would not be needed at + * runtime either, if we are certain that the index is not lossy. However + * it'd be complicated to account for that accurately, and it doesn't + * matter in most cases, since we'd conclude that such attributes are + * available from the index anyway. + */ + foreach(lc, index->indrestrictinfo) + { + RestrictInfo *rinfo = (RestrictInfo *) lfirst(lc); + + pull_varattnos((Node *) rinfo->clause, rel->relid, &attrs_used); + } + + /* + * Construct a bitmapset of columns that the index can return back in an + * index-only scan. + */ + for (i = 0; i < index->ncolumns; i++) + { + int attno = index->indexkeys[i]; + + /* + * For the moment, we just ignore index expressions. It might be nice + * to do something with them, later. + */ + if (attno == 0) + continue; + + if (index->canreturn[i]) + index_canreturn_attrs = + bms_add_member(index_canreturn_attrs, + attno - FirstLowInvalidHeapAttributeNumber); + } + + /* Do we have all the necessary attributes? */ + result = bms_is_subset(attrs_used, index_canreturn_attrs); + + bms_free(attrs_used); + bms_free(index_canreturn_attrs); + + return result; +} + +/* + * get_loop_count + * Choose the loop count estimate to use for costing a parameterized path + * with the given set of outer relids. + * + * Since we produce parameterized paths before we've begun to generate join + * relations, it's impossible to predict exactly how many times a parameterized + * path will be iterated; we don't know the size of the relation that will be + * on the outside of the nestloop. However, we should try to account for + * multiple iterations somehow in costing the path. The heuristic embodied + * here is to use the rowcount of the smallest other base relation needed in + * the join clauses used by the path. (We could alternatively consider the + * largest one, but that seems too optimistic.) This is of course the right + * answer for single-other-relation cases, and it seems like a reasonable + * zero-order approximation for multiway-join cases. + * + * In addition, we check to see if the other side of each join clause is on + * the inside of some semijoin that the current relation is on the outside of. + * If so, the only way that a parameterized path could be used is if the + * semijoin RHS has been unique-ified, so we should use the number of unique + * RHS rows rather than using the relation's raw rowcount. + * + * Note: for this to work, allpaths.c must establish all baserel size + * estimates before it begins to compute paths, or at least before it + * calls create_index_paths(). + */ +static double +get_loop_count(PlannerInfo *root, Index cur_relid, Relids outer_relids) +{ + double result; + int outer_relid; + + /* For a non-parameterized path, just return 1.0 quickly */ + if (outer_relids == NULL) + return 1.0; + + result = 0.0; + outer_relid = -1; + while ((outer_relid = bms_next_member(outer_relids, outer_relid)) >= 0) + { + RelOptInfo *outer_rel; + double rowcount; + + /* Paranoia: ignore bogus relid indexes */ + if (outer_relid >= root->simple_rel_array_size) + continue; + outer_rel = root->simple_rel_array[outer_relid]; + if (outer_rel == NULL) + continue; + Assert(outer_rel->relid == outer_relid); /* sanity check on array */ + + /* Other relation could be proven empty, if so ignore */ + if (IS_DUMMY_REL(outer_rel)) + continue; + + /* Otherwise, rel's rows estimate should be valid by now */ + Assert(outer_rel->rows > 0); + + /* Check to see if rel is on the inside of any semijoins */ + rowcount = adjust_rowcount_for_semijoins(root, + cur_relid, + outer_relid, + outer_rel->rows); + + /* Remember smallest row count estimate among the outer rels */ + if (result == 0.0 || result > rowcount) + result = rowcount; + } + /* Return 1.0 if we found no valid relations (shouldn't happen) */ + return (result > 0.0) ? result : 1.0; +} + +/* + * Check to see if outer_relid is on the inside of any semijoin that cur_relid + * is on the outside of. If so, replace rowcount with the estimated number of + * unique rows from the semijoin RHS (assuming that's smaller, which it might + * not be). The estimate is crude but it's the best we can do at this stage + * of the proceedings. + */ +static double +adjust_rowcount_for_semijoins(PlannerInfo *root, + Index cur_relid, + Index outer_relid, + double rowcount) +{ + ListCell *lc; + + foreach(lc, root->join_info_list) + { + SpecialJoinInfo *sjinfo = (SpecialJoinInfo *) lfirst(lc); + + if (sjinfo->jointype == JOIN_SEMI && + bms_is_member(cur_relid, sjinfo->syn_lefthand) && + bms_is_member(outer_relid, sjinfo->syn_righthand)) + { + /* Estimate number of unique-ified rows */ + double nraw; + double nunique; + + nraw = approximate_joinrel_size(root, sjinfo->syn_righthand); + nunique = estimate_num_groups(root, + sjinfo->semi_rhs_exprs, + nraw, + NULL, + NULL); + if (rowcount > nunique) + rowcount = nunique; + } + } + return rowcount; +} + +/* + * Make an approximate estimate of the size of a joinrel. + * + * We don't have enough info at this point to get a good estimate, so we + * just multiply the base relation sizes together. Fortunately, this is + * the right answer anyway for the most common case with a single relation + * on the RHS of a semijoin. Also, estimate_num_groups() has only a weak + * dependency on its input_rows argument (it basically uses it as a clamp). + * So we might be able to get a fairly decent end result even with a severe + * overestimate of the RHS's raw size. + */ +static double +approximate_joinrel_size(PlannerInfo *root, Relids relids) +{ + double rowcount = 1.0; + int relid; + + relid = -1; + while ((relid = bms_next_member(relids, relid)) >= 0) + { + RelOptInfo *rel; + + /* Paranoia: ignore bogus relid indexes */ + if (relid >= root->simple_rel_array_size) + continue; + rel = root->simple_rel_array[relid]; + if (rel == NULL) + continue; + Assert(rel->relid == relid); /* sanity check on array */ + + /* Relation could be proven empty, if so ignore */ + if (IS_DUMMY_REL(rel)) + continue; + + /* Otherwise, rel's rows estimate should be valid by now */ + Assert(rel->rows > 0); + + /* Accumulate product */ + rowcount *= rel->rows; + } + return rowcount; +} + + +/**************************************************************************** + * ---- ROUTINES TO CHECK QUERY CLAUSES ---- + ****************************************************************************/ + +/* + * match_restriction_clauses_to_index + * Identify restriction clauses for the rel that match the index. + * Matching clauses are added to *clauseset. + */ +static void +match_restriction_clauses_to_index(PlannerInfo *root, + IndexOptInfo *index, + IndexClauseSet *clauseset) +{ + /* We can ignore clauses that are implied by the index predicate */ + match_clauses_to_index(root, index->indrestrictinfo, index, clauseset); +} + +/* + * match_join_clauses_to_index + * Identify join clauses for the rel that match the index. + * Matching clauses are added to *clauseset. + * Also, add any potentially usable join OR clauses to *joinorclauses. + */ +static void +match_join_clauses_to_index(PlannerInfo *root, + RelOptInfo *rel, IndexOptInfo *index, + IndexClauseSet *clauseset, + List **joinorclauses) +{ + ListCell *lc; + + /* Scan the rel's join clauses */ + foreach(lc, rel->joininfo) + { + RestrictInfo *rinfo = (RestrictInfo *) lfirst(lc); + + /* Check if clause can be moved to this rel */ + if (!join_clause_is_movable_to(rinfo, rel)) + continue; + + /* Potentially usable, so see if it matches the index or is an OR */ + if (restriction_is_or_clause(rinfo)) + *joinorclauses = lappend(*joinorclauses, rinfo); + else + match_clause_to_index(root, rinfo, index, clauseset); + } +} + +/* + * match_eclass_clauses_to_index + * Identify EquivalenceClass join clauses for the rel that match the index. + * Matching clauses are added to *clauseset. + */ +static void +match_eclass_clauses_to_index(PlannerInfo *root, IndexOptInfo *index, + IndexClauseSet *clauseset) +{ + int indexcol; + + /* No work if rel is not in any such ECs */ + if (!index->rel->has_eclass_joins) + return; + + for (indexcol = 0; indexcol < index->nkeycolumns; indexcol++) + { + ec_member_matches_arg arg; + List *clauses; + + /* Generate clauses, skipping any that join to lateral_referencers */ + arg.index = index; + arg.indexcol = indexcol; + clauses = generate_implied_equalities_for_column(root, + index->rel, + ec_member_matches_indexcol, + (void *) &arg, + index->rel->lateral_referencers); + + /* + * We have to check whether the results actually do match the index, + * since for non-btree indexes the EC's equality operators might not + * be in the index opclass (cf ec_member_matches_indexcol). + */ + match_clauses_to_index(root, clauses, index, clauseset); + } +} + +/* + * match_clauses_to_index + * Perform match_clause_to_index() for each clause in a list. + * Matching clauses are added to *clauseset. + */ +static void +match_clauses_to_index(PlannerInfo *root, + List *clauses, + IndexOptInfo *index, + IndexClauseSet *clauseset) +{ + ListCell *lc; + + foreach(lc, clauses) + { + RestrictInfo *rinfo = lfirst_node(RestrictInfo, lc); + + match_clause_to_index(root, rinfo, index, clauseset); + } +} + +/* + * match_clause_to_index + * Test whether a qual clause can be used with an index. + * + * If the clause is usable, add an IndexClause entry for it to the appropriate + * list in *clauseset. (*clauseset must be initialized to zeroes before first + * call.) + * + * Note: in some circumstances we may find the same RestrictInfos coming from + * multiple places. Defend against redundant outputs by refusing to add a + * clause twice (pointer equality should be a good enough check for this). + * + * Note: it's possible that a badly-defined index could have multiple matching + * columns. We always select the first match if so; this avoids scenarios + * wherein we get an inflated idea of the index's selectivity by using the + * same clause multiple times with different index columns. + */ +static void +match_clause_to_index(PlannerInfo *root, + RestrictInfo *rinfo, + IndexOptInfo *index, + IndexClauseSet *clauseset) +{ + int indexcol; + + /* + * Never match pseudoconstants to indexes. (Normally a match could not + * happen anyway, since a pseudoconstant clause couldn't contain a Var, + * but what if someone builds an expression index on a constant? It's not + * totally unreasonable to do so with a partial index, either.) + */ + if (rinfo->pseudoconstant) + return; + + /* + * If clause can't be used as an indexqual because it must wait till after + * some lower-security-level restriction clause, reject it. + */ + if (!restriction_is_securely_promotable(rinfo, index->rel)) + return; + + /* OK, check each index key column for a match */ + for (indexcol = 0; indexcol < index->nkeycolumns; indexcol++) + { + IndexClause *iclause; + ListCell *lc; + + /* Ignore duplicates */ + foreach(lc, clauseset->indexclauses[indexcol]) + { + IndexClause *iclause = (IndexClause *) lfirst(lc); + + if (iclause->rinfo == rinfo) + return; + } + + /* OK, try to match the clause to the index column */ + iclause = match_clause_to_indexcol(root, + rinfo, + indexcol, + index); + if (iclause) + { + /* Success, so record it */ + clauseset->indexclauses[indexcol] = + lappend(clauseset->indexclauses[indexcol], iclause); + clauseset->nonempty = true; + return; + } + } +} + +/* + * match_clause_to_indexcol() + * Determine whether a restriction clause matches a column of an index, + * and if so, build an IndexClause node describing the details. + * + * To match an index normally, an operator clause: + * + * (1) must be in the form (indexkey op const) or (const op indexkey); + * and + * (2) must contain an operator which is in the index's operator family + * for this column; and + * (3) must match the collation of the index, if collation is relevant. + * + * Our definition of "const" is exceedingly liberal: we allow anything that + * doesn't involve a volatile function or a Var of the index's relation. + * In particular, Vars belonging to other relations of the query are + * accepted here, since a clause of that form can be used in a + * parameterized indexscan. It's the responsibility of higher code levels + * to manage restriction and join clauses appropriately. + * + * Note: we do need to check for Vars of the index's relation on the + * "const" side of the clause, since clauses like (a.f1 OP (b.f2 OP a.f3)) + * are not processable by a parameterized indexscan on a.f1, whereas + * something like (a.f1 OP (b.f2 OP c.f3)) is. + * + * Presently, the executor can only deal with indexquals that have the + * indexkey on the left, so we can only use clauses that have the indexkey + * on the right if we can commute the clause to put the key on the left. + * We handle that by generating an IndexClause with the correctly-commuted + * opclause as a derived indexqual. + * + * If the index has a collation, the clause must have the same collation. + * For collation-less indexes, we assume it doesn't matter; this is + * necessary for cases like "hstore ? text", wherein hstore's operators + * don't care about collation but the clause will get marked with a + * collation anyway because of the text argument. (This logic is + * embodied in the macro IndexCollMatchesExprColl.) + * + * It is also possible to match RowCompareExpr clauses to indexes (but + * currently, only btree indexes handle this). + * + * It is also possible to match ScalarArrayOpExpr clauses to indexes, when + * the clause is of the form "indexkey op ANY (arrayconst)". + * + * For boolean indexes, it is also possible to match the clause directly + * to the indexkey; or perhaps the clause is (NOT indexkey). + * + * And, last but not least, some operators and functions can be processed + * to derive (typically lossy) indexquals from a clause that isn't in + * itself indexable. If we see that any operand of an OpExpr or FuncExpr + * matches the index key, and the function has a planner support function + * attached to it, we'll invoke the support function to see if such an + * indexqual can be built. + * + * 'rinfo' is the clause to be tested (as a RestrictInfo node). + * 'indexcol' is a column number of 'index' (counting from 0). + * 'index' is the index of interest. + * + * Returns an IndexClause if the clause can be used with this index key, + * or NULL if not. + * + * NOTE: returns NULL if clause is an OR or AND clause; it is the + * responsibility of higher-level routines to cope with those. + */ +static IndexClause * +match_clause_to_indexcol(PlannerInfo *root, + RestrictInfo *rinfo, + int indexcol, + IndexOptInfo *index) +{ + IndexClause *iclause; + Expr *clause = rinfo->clause; + Oid opfamily; + + Assert(indexcol < index->nkeycolumns); + + /* + * Historically this code has coped with NULL clauses. That's probably + * not possible anymore, but we might as well continue to cope. + */ + if (clause == NULL) + return NULL; + + /* First check for boolean-index cases. */ + opfamily = index->opfamily[indexcol]; + if (IsBooleanOpfamily(opfamily)) + { + iclause = match_boolean_index_clause(root, rinfo, indexcol, index); + if (iclause) + return iclause; + } + + /* + * Clause must be an opclause, funcclause, ScalarArrayOpExpr, or + * RowCompareExpr. Or, if the index supports it, we can handle IS + * NULL/NOT NULL clauses. + */ + if (IsA(clause, OpExpr)) + { + return match_opclause_to_indexcol(root, rinfo, indexcol, index); + } + else if (IsA(clause, FuncExpr)) + { + return match_funcclause_to_indexcol(root, rinfo, indexcol, index); + } + else if (IsA(clause, ScalarArrayOpExpr)) + { + return match_saopclause_to_indexcol(root, rinfo, indexcol, index); + } + else if (IsA(clause, RowCompareExpr)) + { + return match_rowcompare_to_indexcol(root, rinfo, indexcol, index); + } + else if (index->amsearchnulls && IsA(clause, NullTest)) + { + NullTest *nt = (NullTest *) clause; + + if (!nt->argisrow && + match_index_to_operand((Node *) nt->arg, indexcol, index)) + { + iclause = makeNode(IndexClause); + iclause->rinfo = rinfo; + iclause->indexquals = list_make1(rinfo); + iclause->lossy = false; + iclause->indexcol = indexcol; + iclause->indexcols = NIL; + return iclause; + } + } + + return NULL; +} + +/* + * match_boolean_index_clause + * Recognize restriction clauses that can be matched to a boolean index. + * + * The idea here is that, for an index on a boolean column that supports the + * BooleanEqualOperator, we can transform a plain reference to the indexkey + * into "indexkey = true", or "NOT indexkey" into "indexkey = false", etc, + * so as to make the expression indexable using the index's "=" operator. + * Since Postgres 8.1, we must do this because constant simplification does + * the reverse transformation; without this code there'd be no way to use + * such an index at all. + * + * This should be called only when IsBooleanOpfamily() recognizes the + * index's operator family. We check to see if the clause matches the + * index's key, and if so, build a suitable IndexClause. + */ +static IndexClause * +match_boolean_index_clause(PlannerInfo *root, + RestrictInfo *rinfo, + int indexcol, + IndexOptInfo *index) +{ + Node *clause = (Node *) rinfo->clause; + Expr *op = NULL; + + /* Direct match? */ + if (match_index_to_operand(clause, indexcol, index)) + { + /* convert to indexkey = TRUE */ + op = make_opclause(BooleanEqualOperator, BOOLOID, false, + (Expr *) clause, + (Expr *) makeBoolConst(true, false), + InvalidOid, InvalidOid); + } + /* NOT clause? */ + else if (is_notclause(clause)) + { + Node *arg = (Node *) get_notclausearg((Expr *) clause); + + if (match_index_to_operand(arg, indexcol, index)) + { + /* convert to indexkey = FALSE */ + op = make_opclause(BooleanEqualOperator, BOOLOID, false, + (Expr *) arg, + (Expr *) makeBoolConst(false, false), + InvalidOid, InvalidOid); + } + } + + /* + * Since we only consider clauses at top level of WHERE, we can convert + * indexkey IS TRUE and indexkey IS FALSE to index searches as well. The + * different meaning for NULL isn't important. + */ + else if (clause && IsA(clause, BooleanTest)) + { + BooleanTest *btest = (BooleanTest *) clause; + Node *arg = (Node *) btest->arg; + + if (btest->booltesttype == IS_TRUE && + match_index_to_operand(arg, indexcol, index)) + { + /* convert to indexkey = TRUE */ + op = make_opclause(BooleanEqualOperator, BOOLOID, false, + (Expr *) arg, + (Expr *) makeBoolConst(true, false), + InvalidOid, InvalidOid); + } + else if (btest->booltesttype == IS_FALSE && + match_index_to_operand(arg, indexcol, index)) + { + /* convert to indexkey = FALSE */ + op = make_opclause(BooleanEqualOperator, BOOLOID, false, + (Expr *) arg, + (Expr *) makeBoolConst(false, false), + InvalidOid, InvalidOid); + } + } + + /* + * If we successfully made an operator clause from the given qual, we must + * wrap it in an IndexClause. It's not lossy. + */ + if (op) + { + IndexClause *iclause = makeNode(IndexClause); + + iclause->rinfo = rinfo; + iclause->indexquals = list_make1(make_simple_restrictinfo(root, op)); + iclause->lossy = false; + iclause->indexcol = indexcol; + iclause->indexcols = NIL; + return iclause; + } + + return NULL; +} + +/* + * match_opclause_to_indexcol() + * Handles the OpExpr case for match_clause_to_indexcol(), + * which see for comments. + */ +static IndexClause * +match_opclause_to_indexcol(PlannerInfo *root, + RestrictInfo *rinfo, + int indexcol, + IndexOptInfo *index) +{ + IndexClause *iclause; + OpExpr *clause = (OpExpr *) rinfo->clause; + Node *leftop, + *rightop; + Oid expr_op; + Oid expr_coll; + Index index_relid; + Oid opfamily; + Oid idxcollation; + + /* + * Only binary operators need apply. (In theory, a planner support + * function could do something with a unary operator, but it seems + * unlikely to be worth the cycles to check.) + */ + if (list_length(clause->args) != 2) + return NULL; + + leftop = (Node *) linitial(clause->args); + rightop = (Node *) lsecond(clause->args); + expr_op = clause->opno; + expr_coll = clause->inputcollid; + + index_relid = index->rel->relid; + opfamily = index->opfamily[indexcol]; + idxcollation = index->indexcollations[indexcol]; + + /* + * Check for clauses of the form: (indexkey operator constant) or + * (constant operator indexkey). See match_clause_to_indexcol's notes + * about const-ness. + * + * Note that we don't ask the support function about clauses that don't + * have one of these forms. Again, in principle it might be possible to + * do something, but it seems unlikely to be worth the cycles to check. + */ + if (match_index_to_operand(leftop, indexcol, index) && + !bms_is_member(index_relid, rinfo->right_relids) && + !contain_volatile_functions(rightop)) + { + if (IndexCollMatchesExprColl(idxcollation, expr_coll) && + op_in_opfamily(expr_op, opfamily)) + { + iclause = makeNode(IndexClause); + iclause->rinfo = rinfo; + iclause->indexquals = list_make1(rinfo); + iclause->lossy = false; + iclause->indexcol = indexcol; + iclause->indexcols = NIL; + return iclause; + } + + /* + * If we didn't find a member of the index's opfamily, try the support + * function for the operator's underlying function. + */ + set_opfuncid(clause); /* make sure we have opfuncid */ + return get_index_clause_from_support(root, + rinfo, + clause->opfuncid, + 0, /* indexarg on left */ + indexcol, + index); + } + + if (match_index_to_operand(rightop, indexcol, index) && + !bms_is_member(index_relid, rinfo->left_relids) && + !contain_volatile_functions(leftop)) + { + if (IndexCollMatchesExprColl(idxcollation, expr_coll)) + { + Oid comm_op = get_commutator(expr_op); + + if (OidIsValid(comm_op) && + op_in_opfamily(comm_op, opfamily)) + { + RestrictInfo *commrinfo; + + /* Build a commuted OpExpr and RestrictInfo */ + commrinfo = commute_restrictinfo(rinfo, comm_op); + + /* Make an IndexClause showing that as a derived qual */ + iclause = makeNode(IndexClause); + iclause->rinfo = rinfo; + iclause->indexquals = list_make1(commrinfo); + iclause->lossy = false; + iclause->indexcol = indexcol; + iclause->indexcols = NIL; + return iclause; + } + } + + /* + * If we didn't find a member of the index's opfamily, try the support + * function for the operator's underlying function. + */ + set_opfuncid(clause); /* make sure we have opfuncid */ + return get_index_clause_from_support(root, + rinfo, + clause->opfuncid, + 1, /* indexarg on right */ + indexcol, + index); + } + + return NULL; +} + +/* + * match_funcclause_to_indexcol() + * Handles the FuncExpr case for match_clause_to_indexcol(), + * which see for comments. + */ +static IndexClause * +match_funcclause_to_indexcol(PlannerInfo *root, + RestrictInfo *rinfo, + int indexcol, + IndexOptInfo *index) +{ + FuncExpr *clause = (FuncExpr *) rinfo->clause; + int indexarg; + ListCell *lc; + + /* + * We have no built-in intelligence about function clauses, but if there's + * a planner support function, it might be able to do something. But, to + * cut down on wasted planning cycles, only call the support function if + * at least one argument matches the target index column. + * + * Note that we don't insist on the other arguments being pseudoconstants; + * the support function has to check that. This is to allow cases where + * only some of the other arguments need to be included in the indexqual. + */ + indexarg = 0; + foreach(lc, clause->args) + { + Node *op = (Node *) lfirst(lc); + + if (match_index_to_operand(op, indexcol, index)) + { + return get_index_clause_from_support(root, + rinfo, + clause->funcid, + indexarg, + indexcol, + index); + } + + indexarg++; + } + + return NULL; +} + +/* + * get_index_clause_from_support() + * If the function has a planner support function, try to construct + * an IndexClause using indexquals created by the support function. + */ +static IndexClause * +get_index_clause_from_support(PlannerInfo *root, + RestrictInfo *rinfo, + Oid funcid, + int indexarg, + int indexcol, + IndexOptInfo *index) +{ + Oid prosupport = get_func_support(funcid); + SupportRequestIndexCondition req; + List *sresult; + + if (!OidIsValid(prosupport)) + return NULL; + + req.type = T_SupportRequestIndexCondition; + req.root = root; + req.funcid = funcid; + req.node = (Node *) rinfo->clause; + req.indexarg = indexarg; + req.index = index; + req.indexcol = indexcol; + req.opfamily = index->opfamily[indexcol]; + req.indexcollation = index->indexcollations[indexcol]; + + req.lossy = true; /* default assumption */ + + sresult = (List *) + DatumGetPointer(OidFunctionCall1(prosupport, + PointerGetDatum(&req))); + + if (sresult != NIL) + { + IndexClause *iclause = makeNode(IndexClause); + List *indexquals = NIL; + ListCell *lc; + + /* + * The support function API says it should just give back bare + * clauses, so here we must wrap each one in a RestrictInfo. + */ + foreach(lc, sresult) + { + Expr *clause = (Expr *) lfirst(lc); + + indexquals = lappend(indexquals, + make_simple_restrictinfo(root, clause)); + } + + iclause->rinfo = rinfo; + iclause->indexquals = indexquals; + iclause->lossy = req.lossy; + iclause->indexcol = indexcol; + iclause->indexcols = NIL; + + return iclause; + } + + return NULL; +} + +/* + * match_saopclause_to_indexcol() + * Handles the ScalarArrayOpExpr case for match_clause_to_indexcol(), + * which see for comments. + */ +static IndexClause * +match_saopclause_to_indexcol(PlannerInfo *root, + RestrictInfo *rinfo, + int indexcol, + IndexOptInfo *index) +{ + ScalarArrayOpExpr *saop = (ScalarArrayOpExpr *) rinfo->clause; + Node *leftop, + *rightop; + Relids right_relids; + Oid expr_op; + Oid expr_coll; + Index index_relid; + Oid opfamily; + Oid idxcollation; + + /* We only accept ANY clauses, not ALL */ + if (!saop->useOr) + return NULL; + leftop = (Node *) linitial(saop->args); + rightop = (Node *) lsecond(saop->args); + right_relids = pull_varnos(root, rightop); + expr_op = saop->opno; + expr_coll = saop->inputcollid; + + index_relid = index->rel->relid; + opfamily = index->opfamily[indexcol]; + idxcollation = index->indexcollations[indexcol]; + + /* + * We must have indexkey on the left and a pseudo-constant array argument. + */ + if (match_index_to_operand(leftop, indexcol, index) && + !bms_is_member(index_relid, right_relids) && + !contain_volatile_functions(rightop)) + { + if (IndexCollMatchesExprColl(idxcollation, expr_coll) && + op_in_opfamily(expr_op, opfamily)) + { + IndexClause *iclause = makeNode(IndexClause); + + iclause->rinfo = rinfo; + iclause->indexquals = list_make1(rinfo); + iclause->lossy = false; + iclause->indexcol = indexcol; + iclause->indexcols = NIL; + return iclause; + } + + /* + * We do not currently ask support functions about ScalarArrayOpExprs, + * though in principle we could. + */ + } + + return NULL; +} + +/* + * match_rowcompare_to_indexcol() + * Handles the RowCompareExpr case for match_clause_to_indexcol(), + * which see for comments. + * + * In this routine we check whether the first column of the row comparison + * matches the target index column. This is sufficient to guarantee that some + * index condition can be constructed from the RowCompareExpr --- the rest + * is handled by expand_indexqual_rowcompare(). + */ +static IndexClause * +match_rowcompare_to_indexcol(PlannerInfo *root, + RestrictInfo *rinfo, + int indexcol, + IndexOptInfo *index) +{ + RowCompareExpr *clause = (RowCompareExpr *) rinfo->clause; + Index index_relid; + Oid opfamily; + Oid idxcollation; + Node *leftop, + *rightop; + bool var_on_left; + Oid expr_op; + Oid expr_coll; + + /* Forget it if we're not dealing with a btree index */ + if (index->relam != BTREE_AM_OID) + return NULL; + + index_relid = index->rel->relid; + opfamily = index->opfamily[indexcol]; + idxcollation = index->indexcollations[indexcol]; + + /* + * We could do the matching on the basis of insisting that the opfamily + * shown in the RowCompareExpr be the same as the index column's opfamily, + * but that could fail in the presence of reverse-sort opfamilies: it'd be + * a matter of chance whether RowCompareExpr had picked the forward or + * reverse-sort family. So look only at the operator, and match if it is + * a member of the index's opfamily (after commutation, if the indexkey is + * on the right). We'll worry later about whether any additional + * operators are matchable to the index. + */ + leftop = (Node *) linitial(clause->largs); + rightop = (Node *) linitial(clause->rargs); + expr_op = linitial_oid(clause->opnos); + expr_coll = linitial_oid(clause->inputcollids); + + /* Collations must match, if relevant */ + if (!IndexCollMatchesExprColl(idxcollation, expr_coll)) + return NULL; + + /* + * These syntactic tests are the same as in match_opclause_to_indexcol() + */ + if (match_index_to_operand(leftop, indexcol, index) && + !bms_is_member(index_relid, pull_varnos(root, rightop)) && + !contain_volatile_functions(rightop)) + { + /* OK, indexkey is on left */ + var_on_left = true; + } + else if (match_index_to_operand(rightop, indexcol, index) && + !bms_is_member(index_relid, pull_varnos(root, leftop)) && + !contain_volatile_functions(leftop)) + { + /* indexkey is on right, so commute the operator */ + expr_op = get_commutator(expr_op); + if (expr_op == InvalidOid) + return NULL; + var_on_left = false; + } + else + return NULL; + + /* We're good if the operator is the right type of opfamily member */ + switch (get_op_opfamily_strategy(expr_op, opfamily)) + { + case BTLessStrategyNumber: + case BTLessEqualStrategyNumber: + case BTGreaterEqualStrategyNumber: + case BTGreaterStrategyNumber: + return expand_indexqual_rowcompare(root, + rinfo, + indexcol, + index, + expr_op, + var_on_left); + } + + return NULL; +} + +/* + * expand_indexqual_rowcompare --- expand a single indexqual condition + * that is a RowCompareExpr + * + * It's already known that the first column of the row comparison matches + * the specified column of the index. We can use additional columns of the + * row comparison as index qualifications, so long as they match the index + * in the "same direction", ie, the indexkeys are all on the same side of the + * clause and the operators are all the same-type members of the opfamilies. + * + * If all the columns of the RowCompareExpr match in this way, we just use it + * as-is, except for possibly commuting it to put the indexkeys on the left. + * + * Otherwise, we build a shortened RowCompareExpr (if more than one + * column matches) or a simple OpExpr (if the first-column match is all + * there is). In these cases the modified clause is always "<=" or ">=" + * even when the original was "<" or ">" --- this is necessary to match all + * the rows that could match the original. (We are building a lossy version + * of the row comparison when we do this, so we set lossy = true.) + * + * Note: this is really just the last half of match_rowcompare_to_indexcol, + * but we split it out for comprehensibility. + */ +static IndexClause * +expand_indexqual_rowcompare(PlannerInfo *root, + RestrictInfo *rinfo, + int indexcol, + IndexOptInfo *index, + Oid expr_op, + bool var_on_left) +{ + IndexClause *iclause = makeNode(IndexClause); + RowCompareExpr *clause = (RowCompareExpr *) rinfo->clause; + int op_strategy; + Oid op_lefttype; + Oid op_righttype; + int matching_cols; + List *expr_ops; + List *opfamilies; + List *lefttypes; + List *righttypes; + List *new_ops; + List *var_args; + List *non_var_args; + + iclause->rinfo = rinfo; + iclause->indexcol = indexcol; + + if (var_on_left) + { + var_args = clause->largs; + non_var_args = clause->rargs; + } + else + { + var_args = clause->rargs; + non_var_args = clause->largs; + } + + get_op_opfamily_properties(expr_op, index->opfamily[indexcol], false, + &op_strategy, + &op_lefttype, + &op_righttype); + + /* Initialize returned list of which index columns are used */ + iclause->indexcols = list_make1_int(indexcol); + + /* Build lists of ops, opfamilies and operator datatypes in case needed */ + expr_ops = list_make1_oid(expr_op); + opfamilies = list_make1_oid(index->opfamily[indexcol]); + lefttypes = list_make1_oid(op_lefttype); + righttypes = list_make1_oid(op_righttype); + + /* + * See how many of the remaining columns match some index column in the + * same way. As in match_clause_to_indexcol(), the "other" side of any + * potential index condition is OK as long as it doesn't use Vars from the + * indexed relation. + */ + matching_cols = 1; + + while (matching_cols < list_length(var_args)) + { + Node *varop = (Node *) list_nth(var_args, matching_cols); + Node *constop = (Node *) list_nth(non_var_args, matching_cols); + int i; + + expr_op = list_nth_oid(clause->opnos, matching_cols); + if (!var_on_left) + { + /* indexkey is on right, so commute the operator */ + expr_op = get_commutator(expr_op); + if (expr_op == InvalidOid) + break; /* operator is not usable */ + } + if (bms_is_member(index->rel->relid, pull_varnos(root, constop))) + break; /* no good, Var on wrong side */ + if (contain_volatile_functions(constop)) + break; /* no good, volatile comparison value */ + + /* + * The Var side can match any key column of the index. + */ + for (i = 0; i < index->nkeycolumns; i++) + { + if (match_index_to_operand(varop, i, index) && + get_op_opfamily_strategy(expr_op, + index->opfamily[i]) == op_strategy && + IndexCollMatchesExprColl(index->indexcollations[i], + list_nth_oid(clause->inputcollids, + matching_cols))) + break; + } + if (i >= index->nkeycolumns) + break; /* no match found */ + + /* Add column number to returned list */ + iclause->indexcols = lappend_int(iclause->indexcols, i); + + /* Add operator info to lists */ + get_op_opfamily_properties(expr_op, index->opfamily[i], false, + &op_strategy, + &op_lefttype, + &op_righttype); + expr_ops = lappend_oid(expr_ops, expr_op); + opfamilies = lappend_oid(opfamilies, index->opfamily[i]); + lefttypes = lappend_oid(lefttypes, op_lefttype); + righttypes = lappend_oid(righttypes, op_righttype); + + /* This column matches, keep scanning */ + matching_cols++; + } + + /* Result is non-lossy if all columns are usable as index quals */ + iclause->lossy = (matching_cols != list_length(clause->opnos)); + + /* + * We can use rinfo->clause as-is if we have var on left and it's all + * usable as index quals. + */ + if (var_on_left && !iclause->lossy) + iclause->indexquals = list_make1(rinfo); + else + { + /* + * We have to generate a modified rowcompare (possibly just one + * OpExpr). The painful part of this is changing < to <= or > to >=, + * so deal with that first. + */ + if (!iclause->lossy) + { + /* very easy, just use the commuted operators */ + new_ops = expr_ops; + } + else if (op_strategy == BTLessEqualStrategyNumber || + op_strategy == BTGreaterEqualStrategyNumber) + { + /* easy, just use the same (possibly commuted) operators */ + new_ops = list_truncate(expr_ops, matching_cols); + } + else + { + ListCell *opfamilies_cell; + ListCell *lefttypes_cell; + ListCell *righttypes_cell; + + if (op_strategy == BTLessStrategyNumber) + op_strategy = BTLessEqualStrategyNumber; + else if (op_strategy == BTGreaterStrategyNumber) + op_strategy = BTGreaterEqualStrategyNumber; + else + elog(ERROR, "unexpected strategy number %d", op_strategy); + new_ops = NIL; + forthree(opfamilies_cell, opfamilies, + lefttypes_cell, lefttypes, + righttypes_cell, righttypes) + { + Oid opfam = lfirst_oid(opfamilies_cell); + Oid lefttype = lfirst_oid(lefttypes_cell); + Oid righttype = lfirst_oid(righttypes_cell); + + expr_op = get_opfamily_member(opfam, lefttype, righttype, + op_strategy); + if (!OidIsValid(expr_op)) /* should not happen */ + elog(ERROR, "missing operator %d(%u,%u) in opfamily %u", + op_strategy, lefttype, righttype, opfam); + new_ops = lappend_oid(new_ops, expr_op); + } + } + + /* If we have more than one matching col, create a subset rowcompare */ + if (matching_cols > 1) + { + RowCompareExpr *rc = makeNode(RowCompareExpr); + + rc->rctype = (RowCompareType) op_strategy; + rc->opnos = new_ops; + rc->opfamilies = list_truncate(list_copy(clause->opfamilies), + matching_cols); + rc->inputcollids = list_truncate(list_copy(clause->inputcollids), + matching_cols); + rc->largs = list_truncate(copyObject(var_args), + matching_cols); + rc->rargs = list_truncate(copyObject(non_var_args), + matching_cols); + iclause->indexquals = list_make1(make_simple_restrictinfo(root, + (Expr *) rc)); + } + else + { + Expr *op; + + /* We don't report an index column list in this case */ + iclause->indexcols = NIL; + + op = make_opclause(linitial_oid(new_ops), BOOLOID, false, + copyObject(linitial(var_args)), + copyObject(linitial(non_var_args)), + InvalidOid, + linitial_oid(clause->inputcollids)); + iclause->indexquals = list_make1(make_simple_restrictinfo(root, op)); + } + } + + return iclause; +} + + +/**************************************************************************** + * ---- ROUTINES TO CHECK ORDERING OPERATORS ---- + ****************************************************************************/ + +/* + * match_pathkeys_to_index + * Test whether an index can produce output ordered according to the + * given pathkeys using "ordering operators". + * + * If it can, return a list of suitable ORDER BY expressions, each of the form + * "indexedcol operator pseudoconstant", along with an integer list of the + * index column numbers (zero based) that each clause would be used with. + * NIL lists are returned if the ordering is not achievable this way. + * + * On success, the result list is ordered by pathkeys, and in fact is + * one-to-one with the requested pathkeys. + */ +static void +match_pathkeys_to_index(IndexOptInfo *index, List *pathkeys, + List **orderby_clauses_p, + List **clause_columns_p) +{ + List *orderby_clauses = NIL; + List *clause_columns = NIL; + ListCell *lc1; + + *orderby_clauses_p = NIL; /* set default results */ + *clause_columns_p = NIL; + + /* Only indexes with the amcanorderbyop property are interesting here */ + if (!index->amcanorderbyop) + return; + + foreach(lc1, pathkeys) + { + PathKey *pathkey = (PathKey *) lfirst(lc1); + bool found = false; + ListCell *lc2; + + /* + * Note: for any failure to match, we just return NIL immediately. + * There is no value in matching just some of the pathkeys. + */ + + /* Pathkey must request default sort order for the target opfamily */ + if (pathkey->pk_strategy != BTLessStrategyNumber || + pathkey->pk_nulls_first) + return; + + /* If eclass is volatile, no hope of using an indexscan */ + if (pathkey->pk_eclass->ec_has_volatile) + return; + + /* + * Try to match eclass member expression(s) to index. Note that child + * EC members are considered, but only when they belong to the target + * relation. (Unlike regular members, the same expression could be a + * child member of more than one EC. Therefore, the same index could + * be considered to match more than one pathkey list, which is OK + * here. See also get_eclass_for_sort_expr.) + */ + foreach(lc2, pathkey->pk_eclass->ec_members) + { + EquivalenceMember *member = (EquivalenceMember *) lfirst(lc2); + int indexcol; + + /* No possibility of match if it references other relations */ + if (!bms_equal(member->em_relids, index->rel->relids)) + continue; + + /* + * We allow any column of the index to match each pathkey; they + * don't have to match left-to-right as you might expect. This is + * correct for GiST, and it doesn't matter for SP-GiST because + * that doesn't handle multiple columns anyway, and no other + * existing AMs support amcanorderbyop. We might need different + * logic in future for other implementations. + */ + for (indexcol = 0; indexcol < index->nkeycolumns; indexcol++) + { + Expr *expr; + + expr = match_clause_to_ordering_op(index, + indexcol, + member->em_expr, + pathkey->pk_opfamily); + if (expr) + { + orderby_clauses = lappend(orderby_clauses, expr); + clause_columns = lappend_int(clause_columns, indexcol); + found = true; + break; + } + } + + if (found) /* don't want to look at remaining members */ + break; + } + + if (!found) /* fail if no match for this pathkey */ + return; + } + + *orderby_clauses_p = orderby_clauses; /* success! */ + *clause_columns_p = clause_columns; +} + +/* + * match_clause_to_ordering_op + * Determines whether an ordering operator expression matches an + * index column. + * + * This is similar to, but simpler than, match_clause_to_indexcol. + * We only care about simple OpExpr cases. The input is a bare + * expression that is being ordered by, which must be of the form + * (indexkey op const) or (const op indexkey) where op is an ordering + * operator for the column's opfamily. + * + * 'index' is the index of interest. + * 'indexcol' is a column number of 'index' (counting from 0). + * 'clause' is the ordering expression to be tested. + * 'pk_opfamily' is the btree opfamily describing the required sort order. + * + * Note that we currently do not consider the collation of the ordering + * operator's result. In practical cases the result type will be numeric + * and thus have no collation, and it's not very clear what to match to + * if it did have a collation. The index's collation should match the + * ordering operator's input collation, not its result. + * + * If successful, return 'clause' as-is if the indexkey is on the left, + * otherwise a commuted copy of 'clause'. If no match, return NULL. + */ +static Expr * +match_clause_to_ordering_op(IndexOptInfo *index, + int indexcol, + Expr *clause, + Oid pk_opfamily) +{ + Oid opfamily; + Oid idxcollation; + Node *leftop, + *rightop; + Oid expr_op; + Oid expr_coll; + Oid sortfamily; + bool commuted; + + Assert(indexcol < index->nkeycolumns); + + opfamily = index->opfamily[indexcol]; + idxcollation = index->indexcollations[indexcol]; + + /* + * Clause must be a binary opclause. + */ + if (!is_opclause(clause)) + return NULL; + leftop = get_leftop(clause); + rightop = get_rightop(clause); + if (!leftop || !rightop) + return NULL; + expr_op = ((OpExpr *) clause)->opno; + expr_coll = ((OpExpr *) clause)->inputcollid; + + /* + * We can forget the whole thing right away if wrong collation. + */ + if (!IndexCollMatchesExprColl(idxcollation, expr_coll)) + return NULL; + + /* + * Check for clauses of the form: (indexkey operator constant) or + * (constant operator indexkey). + */ + if (match_index_to_operand(leftop, indexcol, index) && + !contain_var_clause(rightop) && + !contain_volatile_functions(rightop)) + { + commuted = false; + } + else if (match_index_to_operand(rightop, indexcol, index) && + !contain_var_clause(leftop) && + !contain_volatile_functions(leftop)) + { + /* Might match, but we need a commuted operator */ + expr_op = get_commutator(expr_op); + if (expr_op == InvalidOid) + return NULL; + commuted = true; + } + else + return NULL; + + /* + * Is the (commuted) operator an ordering operator for the opfamily? And + * if so, does it yield the right sorting semantics? + */ + sortfamily = get_op_opfamily_sortfamily(expr_op, opfamily); + if (sortfamily != pk_opfamily) + return NULL; + + /* We have a match. Return clause or a commuted version thereof. */ + if (commuted) + { + OpExpr *newclause = makeNode(OpExpr); + + /* flat-copy all the fields of clause */ + memcpy(newclause, clause, sizeof(OpExpr)); + + /* commute it */ + newclause->opno = expr_op; + newclause->opfuncid = InvalidOid; + newclause->args = list_make2(rightop, leftop); + + clause = (Expr *) newclause; + } + + return clause; +} + + +/**************************************************************************** + * ---- ROUTINES TO DO PARTIAL INDEX PREDICATE TESTS ---- + ****************************************************************************/ + +/* + * check_index_predicates + * Set the predicate-derived IndexOptInfo fields for each index + * of the specified relation. + * + * predOK is set true if the index is partial and its predicate is satisfied + * for this query, ie the query's WHERE clauses imply the predicate. + * + * indrestrictinfo is set to the relation's baserestrictinfo list less any + * conditions that are implied by the index's predicate. (Obviously, for a + * non-partial index, this is the same as baserestrictinfo.) Such conditions + * can be dropped from the plan when using the index, in certain cases. + * + * At one time it was possible for this to get re-run after adding more + * restrictions to the rel, thus possibly letting us prove more indexes OK. + * That doesn't happen any more (at least not in the core code's usage), + * but this code still supports it in case extensions want to mess with the + * baserestrictinfo list. We assume that adding more restrictions can't make + * an index not predOK. We must recompute indrestrictinfo each time, though, + * to make sure any newly-added restrictions get into it if needed. + */ +void +check_index_predicates(PlannerInfo *root, RelOptInfo *rel) +{ + List *clauselist; + bool have_partial; + bool is_target_rel; + Relids otherrels; + ListCell *lc; + + /* Indexes are available only on base or "other" member relations. */ + Assert(IS_SIMPLE_REL(rel)); + + /* + * Initialize the indrestrictinfo lists to be identical to + * baserestrictinfo, and check whether there are any partial indexes. If + * not, this is all we need to do. + */ + have_partial = false; + foreach(lc, rel->indexlist) + { + IndexOptInfo *index = (IndexOptInfo *) lfirst(lc); + + index->indrestrictinfo = rel->baserestrictinfo; + if (index->indpred) + have_partial = true; + } + if (!have_partial) + return; + + /* + * Construct a list of clauses that we can assume true for the purpose of + * proving the index(es) usable. Restriction clauses for the rel are + * always usable, and so are any join clauses that are "movable to" this + * rel. Also, we can consider any EC-derivable join clauses (which must + * be "movable to" this rel, by definition). + */ + clauselist = list_copy(rel->baserestrictinfo); + + /* Scan the rel's join clauses */ + foreach(lc, rel->joininfo) + { + RestrictInfo *rinfo = (RestrictInfo *) lfirst(lc); + + /* Check if clause can be moved to this rel */ + if (!join_clause_is_movable_to(rinfo, rel)) + continue; + + clauselist = lappend(clauselist, rinfo); + } + + /* + * Add on any equivalence-derivable join clauses. Computing the correct + * relid sets for generate_join_implied_equalities is slightly tricky + * because the rel could be a child rel rather than a true baserel, and in + * that case we must remove its parents' relid(s) from all_baserels. + */ + if (rel->reloptkind == RELOPT_OTHER_MEMBER_REL) + otherrels = bms_difference(root->all_baserels, + find_childrel_parents(root, rel)); + else + otherrels = bms_difference(root->all_baserels, rel->relids); + + if (!bms_is_empty(otherrels)) + clauselist = + list_concat(clauselist, + generate_join_implied_equalities(root, + bms_union(rel->relids, + otherrels), + otherrels, + rel)); + + /* + * Normally we remove quals that are implied by a partial index's + * predicate from indrestrictinfo, indicating that they need not be + * checked explicitly by an indexscan plan using this index. However, if + * the rel is a target relation of UPDATE/DELETE/MERGE/SELECT FOR UPDATE, + * we cannot remove such quals from the plan, because they need to be in + * the plan so that they will be properly rechecked by EvalPlanQual + * testing. Some day we might want to remove such quals from the main + * plan anyway and pass them through to EvalPlanQual via a side channel; + * but for now, we just don't remove implied quals at all for target + * relations. + */ + is_target_rel = (bms_is_member(rel->relid, root->all_result_relids) || + get_plan_rowmark(root->rowMarks, rel->relid) != NULL); + + /* + * Now try to prove each index predicate true, and compute the + * indrestrictinfo lists for partial indexes. Note that we compute the + * indrestrictinfo list even for non-predOK indexes; this might seem + * wasteful, but we may be able to use such indexes in OR clauses, cf + * generate_bitmap_or_paths(). + */ + foreach(lc, rel->indexlist) + { + IndexOptInfo *index = (IndexOptInfo *) lfirst(lc); + ListCell *lcr; + + if (index->indpred == NIL) + continue; /* ignore non-partial indexes here */ + + if (!index->predOK) /* don't repeat work if already proven OK */ + index->predOK = predicate_implied_by(index->indpred, clauselist, + false); + + /* If rel is an update target, leave indrestrictinfo as set above */ + if (is_target_rel) + continue; + + /* Else compute indrestrictinfo as the non-implied quals */ + index->indrestrictinfo = NIL; + foreach(lcr, rel->baserestrictinfo) + { + RestrictInfo *rinfo = (RestrictInfo *) lfirst(lcr); + + /* predicate_implied_by() assumes first arg is immutable */ + if (contain_mutable_functions((Node *) rinfo->clause) || + !predicate_implied_by(list_make1(rinfo->clause), + index->indpred, false)) + index->indrestrictinfo = lappend(index->indrestrictinfo, rinfo); + } + } +} + +/**************************************************************************** + * ---- ROUTINES TO CHECK EXTERNALLY-VISIBLE CONDITIONS ---- + ****************************************************************************/ + +/* + * ec_member_matches_indexcol + * Test whether an EquivalenceClass member matches an index column. + * + * This is a callback for use by generate_implied_equalities_for_column. + */ +static bool +ec_member_matches_indexcol(PlannerInfo *root, RelOptInfo *rel, + EquivalenceClass *ec, EquivalenceMember *em, + void *arg) +{ + IndexOptInfo *index = ((ec_member_matches_arg *) arg)->index; + int indexcol = ((ec_member_matches_arg *) arg)->indexcol; + Oid curFamily; + Oid curCollation; + + Assert(indexcol < index->nkeycolumns); + + curFamily = index->opfamily[indexcol]; + curCollation = index->indexcollations[indexcol]; + + /* + * If it's a btree index, we can reject it if its opfamily isn't + * compatible with the EC, since no clause generated from the EC could be + * used with the index. For non-btree indexes, we can't easily tell + * whether clauses generated from the EC could be used with the index, so + * don't check the opfamily. This might mean we return "true" for a + * useless EC, so we have to recheck the results of + * generate_implied_equalities_for_column; see + * match_eclass_clauses_to_index. + */ + if (index->relam == BTREE_AM_OID && + !list_member_oid(ec->ec_opfamilies, curFamily)) + return false; + + /* We insist on collation match for all index types, though */ + if (!IndexCollMatchesExprColl(curCollation, ec->ec_collation)) + return false; + + return match_index_to_operand((Node *) em->em_expr, indexcol, index); +} + +/* + * relation_has_unique_index_for + * Determine whether the relation provably has at most one row satisfying + * a set of equality conditions, because the conditions constrain all + * columns of some unique index. + * + * The conditions can be represented in either or both of two ways: + * 1. A list of RestrictInfo nodes, where the caller has already determined + * that each condition is a mergejoinable equality with an expression in + * this relation on one side, and an expression not involving this relation + * on the other. The transient outer_is_left flag is used to identify which + * side we should look at: left side if outer_is_left is false, right side + * if it is true. + * 2. A list of expressions in this relation, and a corresponding list of + * equality operators. The caller must have already checked that the operators + * represent equality. (Note: the operators could be cross-type; the + * expressions should correspond to their RHS inputs.) + * + * The caller need only supply equality conditions arising from joins; + * this routine automatically adds in any usable baserestrictinfo clauses. + * (Note that the passed-in restrictlist will be destructively modified!) + */ +bool +relation_has_unique_index_for(PlannerInfo *root, RelOptInfo *rel, + List *restrictlist, + List *exprlist, List *oprlist) +{ + ListCell *ic; + + Assert(list_length(exprlist) == list_length(oprlist)); + + /* Short-circuit if no indexes... */ + if (rel->indexlist == NIL) + return false; + + /* + * Examine the rel's restriction clauses for usable var = const clauses + * that we can add to the restrictlist. + */ + foreach(ic, rel->baserestrictinfo) + { + RestrictInfo *restrictinfo = (RestrictInfo *) lfirst(ic); + + /* + * Note: can_join won't be set for a restriction clause, but + * mergeopfamilies will be if it has a mergejoinable operator and + * doesn't contain volatile functions. + */ + if (restrictinfo->mergeopfamilies == NIL) + continue; /* not mergejoinable */ + + /* + * The clause certainly doesn't refer to anything but the given rel. + * If either side is pseudoconstant then we can use it. + */ + if (bms_is_empty(restrictinfo->left_relids)) + { + /* righthand side is inner */ + restrictinfo->outer_is_left = true; + } + else if (bms_is_empty(restrictinfo->right_relids)) + { + /* lefthand side is inner */ + restrictinfo->outer_is_left = false; + } + else + continue; + + /* OK, add to list */ + restrictlist = lappend(restrictlist, restrictinfo); + } + + /* Short-circuit the easy case */ + if (restrictlist == NIL && exprlist == NIL) + return false; + + /* Examine each index of the relation ... */ + foreach(ic, rel->indexlist) + { + IndexOptInfo *ind = (IndexOptInfo *) lfirst(ic); + int c; + + /* + * If the index is not unique, or not immediately enforced, or if it's + * a partial index, it's useless here. We're unable to make use of + * predOK partial unique indexes due to the fact that + * check_index_predicates() also makes use of join predicates to + * determine if the partial index is usable. Here we need proofs that + * hold true before any joins are evaluated. + */ + if (!ind->unique || !ind->immediate || ind->indpred != NIL) + continue; + + /* + * Try to find each index column in the lists of conditions. This is + * O(N^2) or worse, but we expect all the lists to be short. + */ + for (c = 0; c < ind->nkeycolumns; c++) + { + bool matched = false; + ListCell *lc; + ListCell *lc2; + + foreach(lc, restrictlist) + { + RestrictInfo *rinfo = (RestrictInfo *) lfirst(lc); + Node *rexpr; + + /* + * The condition's equality operator must be a member of the + * index opfamily, else it is not asserting the right kind of + * equality behavior for this index. We check this first + * since it's probably cheaper than match_index_to_operand(). + */ + if (!list_member_oid(rinfo->mergeopfamilies, ind->opfamily[c])) + continue; + + /* + * XXX at some point we may need to check collations here too. + * For the moment we assume all collations reduce to the same + * notion of equality. + */ + + /* OK, see if the condition operand matches the index key */ + if (rinfo->outer_is_left) + rexpr = get_rightop(rinfo->clause); + else + rexpr = get_leftop(rinfo->clause); + + if (match_index_to_operand(rexpr, c, ind)) + { + matched = true; /* column is unique */ + break; + } + } + + if (matched) + continue; + + forboth(lc, exprlist, lc2, oprlist) + { + Node *expr = (Node *) lfirst(lc); + Oid opr = lfirst_oid(lc2); + + /* See if the expression matches the index key */ + if (!match_index_to_operand(expr, c, ind)) + continue; + + /* + * The equality operator must be a member of the index + * opfamily, else it is not asserting the right kind of + * equality behavior for this index. We assume the caller + * determined it is an equality operator, so we don't need to + * check any more tightly than this. + */ + if (!op_in_opfamily(opr, ind->opfamily[c])) + continue; + + /* + * XXX at some point we may need to check collations here too. + * For the moment we assume all collations reduce to the same + * notion of equality. + */ + + matched = true; /* column is unique */ + break; + } + + if (!matched) + break; /* no match; this index doesn't help us */ + } + + /* Matched all key columns of this index? */ + if (c == ind->nkeycolumns) + return true; + } + + return false; +} + +/* + * indexcol_is_bool_constant_for_query + * + * If an index column is constrained to have a constant value by the query's + * WHERE conditions, then it's irrelevant for sort-order considerations. + * Usually that means we have a restriction clause WHERE indexcol = constant, + * which gets turned into an EquivalenceClass containing a constant, which + * is recognized as redundant by build_index_pathkeys(). But if the index + * column is a boolean variable (or expression), then we are not going to + * see WHERE indexcol = constant, because expression preprocessing will have + * simplified that to "WHERE indexcol" or "WHERE NOT indexcol". So we are not + * going to have a matching EquivalenceClass (unless the query also contains + * "ORDER BY indexcol"). To allow such cases to work the same as they would + * for non-boolean values, this function is provided to detect whether the + * specified index column matches a boolean restriction clause. + */ +bool +indexcol_is_bool_constant_for_query(PlannerInfo *root, + IndexOptInfo *index, + int indexcol) +{ + ListCell *lc; + + /* If the index isn't boolean, we can't possibly get a match */ + if (!IsBooleanOpfamily(index->opfamily[indexcol])) + return false; + + /* Check each restriction clause for the index's rel */ + foreach(lc, index->rel->baserestrictinfo) + { + RestrictInfo *rinfo = (RestrictInfo *) lfirst(lc); + + /* + * As in match_clause_to_indexcol, never match pseudoconstants to + * indexes. (It might be semantically okay to do so here, but the + * odds of getting a match are negligible, so don't waste the cycles.) + */ + if (rinfo->pseudoconstant) + continue; + + /* See if we can match the clause's expression to the index column */ + if (match_boolean_index_clause(root, rinfo, indexcol, index)) + return true; + } + + return false; +} + + +/**************************************************************************** + * ---- ROUTINES TO CHECK OPERANDS ---- + ****************************************************************************/ + +/* + * match_index_to_operand() + * Generalized test for a match between an index's key + * and the operand on one side of a restriction or join clause. + * + * operand: the nodetree to be compared to the index + * indexcol: the column number of the index (counting from 0) + * index: the index of interest + * + * Note that we aren't interested in collations here; the caller must check + * for a collation match, if it's dealing with an operator where that matters. + * + * This is exported for use in selfuncs.c. + */ +bool +match_index_to_operand(Node *operand, + int indexcol, + IndexOptInfo *index) +{ + int indkey; + + /* + * Ignore any RelabelType node above the operand. This is needed to be + * able to apply indexscanning in binary-compatible-operator cases. Note: + * we can assume there is at most one RelabelType node; + * eval_const_expressions() will have simplified if more than one. + */ + if (operand && IsA(operand, RelabelType)) + operand = (Node *) ((RelabelType *) operand)->arg; + + indkey = index->indexkeys[indexcol]; + if (indkey != 0) + { + /* + * Simple index column; operand must be a matching Var. + */ + if (operand && IsA(operand, Var) && + index->rel->relid == ((Var *) operand)->varno && + indkey == ((Var *) operand)->varattno) + return true; + } + else + { + /* + * Index expression; find the correct expression. (This search could + * be avoided, at the cost of complicating all the callers of this + * routine; doesn't seem worth it.) + */ + ListCell *indexpr_item; + int i; + Node *indexkey; + + indexpr_item = list_head(index->indexprs); + for (i = 0; i < indexcol; i++) + { + if (index->indexkeys[i] == 0) + { + if (indexpr_item == NULL) + elog(ERROR, "wrong number of index expressions"); + indexpr_item = lnext(index->indexprs, indexpr_item); + } + } + if (indexpr_item == NULL) + elog(ERROR, "wrong number of index expressions"); + indexkey = (Node *) lfirst(indexpr_item); + + /* + * Does it match the operand? Again, strip any relabeling. + */ + if (indexkey && IsA(indexkey, RelabelType)) + indexkey = (Node *) ((RelabelType *) indexkey)->arg; + + if (equal(indexkey, operand)) + return true; + } + + return false; +} + +/* + * is_pseudo_constant_for_index() + * Test whether the given expression can be used as an indexscan + * comparison value. + * + * An indexscan comparison value must not contain any volatile functions, + * and it can't contain any Vars of the index's own table. Vars of + * other tables are okay, though; in that case we'd be producing an + * indexqual usable in a parameterized indexscan. This is, therefore, + * a weaker condition than is_pseudo_constant_clause(). + * + * This function is exported for use by planner support functions, + * which will have available the IndexOptInfo, but not any RestrictInfo + * infrastructure. It is making the same test made by functions above + * such as match_opclause_to_indexcol(), but those rely where possible + * on RestrictInfo information about variable membership. + * + * expr: the nodetree to be checked + * index: the index of interest + */ +bool +is_pseudo_constant_for_index(PlannerInfo *root, Node *expr, IndexOptInfo *index) +{ + /* pull_varnos is cheaper than volatility check, so do that first */ + if (bms_is_member(index->rel->relid, pull_varnos(root, expr))) + return false; /* no good, contains Var of table */ + if (contain_volatile_functions(expr)) + return false; /* no good, volatile comparison value */ + return true; +} diff --git a/src/backend/optimizer/path/joinpath.c b/src/backend/optimizer/path/joinpath.c new file mode 100644 index 0000000..bad3dd1 --- /dev/null +++ b/src/backend/optimizer/path/joinpath.c @@ -0,0 +1,2367 @@ +/*------------------------------------------------------------------------- + * + * joinpath.c + * Routines to find all possible paths for processing a set of joins + * + * Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group + * Portions Copyright (c) 1994, Regents of the University of California + * + * + * IDENTIFICATION + * src/backend/optimizer/path/joinpath.c + * + *------------------------------------------------------------------------- + */ +#include "postgres.h" + +#include <math.h> + +#include "executor/executor.h" +#include "foreign/fdwapi.h" +#include "nodes/nodeFuncs.h" +#include "optimizer/cost.h" +#include "optimizer/optimizer.h" +#include "optimizer/pathnode.h" +#include "optimizer/paths.h" +#include "optimizer/planmain.h" +#include "optimizer/restrictinfo.h" +#include "utils/typcache.h" + +/* Hook for plugins to get control in add_paths_to_joinrel() */ +set_join_pathlist_hook_type set_join_pathlist_hook = NULL; + +/* + * Paths parameterized by the parent can be considered to be parameterized by + * any of its child. + */ +#define PATH_PARAM_BY_PARENT(path, rel) \ + ((path)->param_info && bms_overlap(PATH_REQ_OUTER(path), \ + (rel)->top_parent_relids)) +#define PATH_PARAM_BY_REL_SELF(path, rel) \ + ((path)->param_info && bms_overlap(PATH_REQ_OUTER(path), (rel)->relids)) + +#define PATH_PARAM_BY_REL(path, rel) \ + (PATH_PARAM_BY_REL_SELF(path, rel) || PATH_PARAM_BY_PARENT(path, rel)) + +static void try_partial_mergejoin_path(PlannerInfo *root, + RelOptInfo *joinrel, + Path *outer_path, + Path *inner_path, + List *pathkeys, + List *mergeclauses, + List *outersortkeys, + List *innersortkeys, + JoinType jointype, + JoinPathExtraData *extra); +static void sort_inner_and_outer(PlannerInfo *root, RelOptInfo *joinrel, + RelOptInfo *outerrel, RelOptInfo *innerrel, + JoinType jointype, JoinPathExtraData *extra); +static inline bool clause_sides_match_join(RestrictInfo *rinfo, + RelOptInfo *outerrel, + RelOptInfo *innerrel); +static void match_unsorted_outer(PlannerInfo *root, RelOptInfo *joinrel, + RelOptInfo *outerrel, RelOptInfo *innerrel, + JoinType jointype, JoinPathExtraData *extra); +static void consider_parallel_nestloop(PlannerInfo *root, + RelOptInfo *joinrel, + RelOptInfo *outerrel, + RelOptInfo *innerrel, + JoinType jointype, + JoinPathExtraData *extra); +static void consider_parallel_mergejoin(PlannerInfo *root, + RelOptInfo *joinrel, + RelOptInfo *outerrel, + RelOptInfo *innerrel, + JoinType jointype, + JoinPathExtraData *extra, + Path *inner_cheapest_total); +static void hash_inner_and_outer(PlannerInfo *root, RelOptInfo *joinrel, + RelOptInfo *outerrel, RelOptInfo *innerrel, + JoinType jointype, JoinPathExtraData *extra); +static List *select_mergejoin_clauses(PlannerInfo *root, + RelOptInfo *joinrel, + RelOptInfo *outerrel, + RelOptInfo *innerrel, + List *restrictlist, + JoinType jointype, + bool *mergejoin_allowed); +static void generate_mergejoin_paths(PlannerInfo *root, + RelOptInfo *joinrel, + RelOptInfo *innerrel, + Path *outerpath, + JoinType jointype, + JoinPathExtraData *extra, + bool useallclauses, + Path *inner_cheapest_total, + List *merge_pathkeys, + bool is_partial); + + +/* + * add_paths_to_joinrel + * Given a join relation and two component rels from which it can be made, + * consider all possible paths that use the two component rels as outer + * and inner rel respectively. Add these paths to the join rel's pathlist + * if they survive comparison with other paths (and remove any existing + * paths that are dominated by these paths). + * + * Modifies the pathlist field of the joinrel node to contain the best + * paths found so far. + * + * jointype is not necessarily the same as sjinfo->jointype; it might be + * "flipped around" if we are considering joining the rels in the opposite + * direction from what's indicated in sjinfo. + * + * Also, this routine and others in this module accept the special JoinTypes + * JOIN_UNIQUE_OUTER and JOIN_UNIQUE_INNER to indicate that we should + * unique-ify the outer or inner relation and then apply a regular inner + * join. These values are not allowed to propagate outside this module, + * however. Path cost estimation code may need to recognize that it's + * dealing with such a case --- the combination of nominal jointype INNER + * with sjinfo->jointype == JOIN_SEMI indicates that. + */ +void +add_paths_to_joinrel(PlannerInfo *root, + RelOptInfo *joinrel, + RelOptInfo *outerrel, + RelOptInfo *innerrel, + JoinType jointype, + SpecialJoinInfo *sjinfo, + List *restrictlist) +{ + JoinPathExtraData extra; + bool mergejoin_allowed = true; + bool consider_join_pushdown = false; + ListCell *lc; + Relids joinrelids; + + /* + * PlannerInfo doesn't contain the SpecialJoinInfos created for joins + * between child relations, even if there is a SpecialJoinInfo node for + * the join between the topmost parents. So, while calculating Relids set + * representing the restriction, consider relids of topmost parent of + * partitions. + */ + if (joinrel->reloptkind == RELOPT_OTHER_JOINREL) + joinrelids = joinrel->top_parent_relids; + else + joinrelids = joinrel->relids; + + extra.restrictlist = restrictlist; + extra.mergeclause_list = NIL; + extra.sjinfo = sjinfo; + extra.param_source_rels = NULL; + + /* + * See if the inner relation is provably unique for this outer rel. + * + * We have some special cases: for JOIN_SEMI and JOIN_ANTI, it doesn't + * matter since the executor can make the equivalent optimization anyway; + * we need not expend planner cycles on proofs. For JOIN_UNIQUE_INNER, we + * must be considering a semijoin whose inner side is not provably unique + * (else reduce_unique_semijoins would've simplified it), so there's no + * point in calling innerrel_is_unique. However, if the LHS covers all of + * the semijoin's min_lefthand, then it's appropriate to set inner_unique + * because the path produced by create_unique_path will be unique relative + * to the LHS. (If we have an LHS that's only part of the min_lefthand, + * that is *not* true.) For JOIN_UNIQUE_OUTER, pass JOIN_INNER to avoid + * letting that value escape this module. + */ + switch (jointype) + { + case JOIN_SEMI: + case JOIN_ANTI: + + /* + * XXX it may be worth proving this to allow a Memoize to be + * considered for Nested Loop Semi/Anti Joins. + */ + extra.inner_unique = false; /* well, unproven */ + break; + case JOIN_UNIQUE_INNER: + extra.inner_unique = bms_is_subset(sjinfo->min_lefthand, + outerrel->relids); + break; + case JOIN_UNIQUE_OUTER: + extra.inner_unique = innerrel_is_unique(root, + joinrel->relids, + outerrel->relids, + innerrel, + JOIN_INNER, + restrictlist, + false); + break; + default: + extra.inner_unique = innerrel_is_unique(root, + joinrel->relids, + outerrel->relids, + innerrel, + jointype, + restrictlist, + false); + break; + } + + /* + * Find potential mergejoin clauses. We can skip this if we are not + * interested in doing a mergejoin. However, mergejoin may be our only + * way of implementing a full outer join, so override enable_mergejoin if + * it's a full join. + */ + if (enable_mergejoin || jointype == JOIN_FULL) + extra.mergeclause_list = select_mergejoin_clauses(root, + joinrel, + outerrel, + innerrel, + restrictlist, + jointype, + &mergejoin_allowed); + + /* + * If it's SEMI, ANTI, or inner_unique join, compute correction factors + * for cost estimation. These will be the same for all paths. + */ + if (jointype == JOIN_SEMI || jointype == JOIN_ANTI || extra.inner_unique) + compute_semi_anti_join_factors(root, joinrel, outerrel, innerrel, + jointype, sjinfo, restrictlist, + &extra.semifactors); + + /* + * Decide whether it's sensible to generate parameterized paths for this + * joinrel, and if so, which relations such paths should require. There + * is usually no need to create a parameterized result path unless there + * is a join order restriction that prevents joining one of our input rels + * directly to the parameter source rel instead of joining to the other + * input rel. (But see allow_star_schema_join().) This restriction + * reduces the number of parameterized paths we have to deal with at + * higher join levels, without compromising the quality of the resulting + * plan. We express the restriction as a Relids set that must overlap the + * parameterization of any proposed join path. + */ + foreach(lc, root->join_info_list) + { + SpecialJoinInfo *sjinfo2 = (SpecialJoinInfo *) lfirst(lc); + + /* + * SJ is relevant to this join if we have some part of its RHS + * (possibly not all of it), and haven't yet joined to its LHS. (This + * test is pretty simplistic, but should be sufficient considering the + * join has already been proven legal.) If the SJ is relevant, it + * presents constraints for joining to anything not in its RHS. + */ + if (bms_overlap(joinrelids, sjinfo2->min_righthand) && + !bms_overlap(joinrelids, sjinfo2->min_lefthand)) + extra.param_source_rels = bms_join(extra.param_source_rels, + bms_difference(root->all_baserels, + sjinfo2->min_righthand)); + + /* full joins constrain both sides symmetrically */ + if (sjinfo2->jointype == JOIN_FULL && + bms_overlap(joinrelids, sjinfo2->min_lefthand) && + !bms_overlap(joinrelids, sjinfo2->min_righthand)) + extra.param_source_rels = bms_join(extra.param_source_rels, + bms_difference(root->all_baserels, + sjinfo2->min_lefthand)); + } + + /* + * However, when a LATERAL subquery is involved, there will simply not be + * any paths for the joinrel that aren't parameterized by whatever the + * subquery is parameterized by, unless its parameterization is resolved + * within the joinrel. So we might as well allow additional dependencies + * on whatever residual lateral dependencies the joinrel will have. + */ + extra.param_source_rels = bms_add_members(extra.param_source_rels, + joinrel->lateral_relids); + + /* + * 1. Consider mergejoin paths where both relations must be explicitly + * sorted. Skip this if we can't mergejoin. + */ + if (mergejoin_allowed) + sort_inner_and_outer(root, joinrel, outerrel, innerrel, + jointype, &extra); + + /* + * 2. Consider paths where the outer relation need not be explicitly + * sorted. This includes both nestloops and mergejoins where the outer + * path is already ordered. Again, skip this if we can't mergejoin. + * (That's okay because we know that nestloop can't handle right/full + * joins at all, so it wouldn't work in the prohibited cases either.) + */ + if (mergejoin_allowed) + match_unsorted_outer(root, joinrel, outerrel, innerrel, + jointype, &extra); + +#ifdef NOT_USED + + /* + * 3. Consider paths where the inner relation need not be explicitly + * sorted. This includes mergejoins only (nestloops were already built in + * match_unsorted_outer). + * + * Diked out as redundant 2/13/2000 -- tgl. There isn't any really + * significant difference between the inner and outer side of a mergejoin, + * so match_unsorted_inner creates no paths that aren't equivalent to + * those made by match_unsorted_outer when add_paths_to_joinrel() is + * invoked with the two rels given in the other order. + */ + if (mergejoin_allowed) + match_unsorted_inner(root, joinrel, outerrel, innerrel, + jointype, &extra); +#endif + + /* + * 4. Consider paths where both outer and inner relations must be hashed + * before being joined. As above, disregard enable_hashjoin for full + * joins, because there may be no other alternative. + */ + if (enable_hashjoin || jointype == JOIN_FULL) + hash_inner_and_outer(root, joinrel, outerrel, innerrel, + jointype, &extra); + + /* + * createplan.c does not currently support handling of pseudoconstant + * clauses assigned to joins pushed down by extensions; check if the + * restrictlist has such clauses, and if not, allow them to consider + * pushing down joins. + */ + if ((joinrel->fdwroutine && + joinrel->fdwroutine->GetForeignJoinPaths) || + set_join_pathlist_hook) + consider_join_pushdown = !has_pseudoconstant_clauses(root, + restrictlist); + + /* + * 5. If inner and outer relations are foreign tables (or joins) belonging + * to the same server and assigned to the same user to check access + * permissions as, give the FDW a chance to push down joins. + */ + if (joinrel->fdwroutine && + joinrel->fdwroutine->GetForeignJoinPaths && + consider_join_pushdown) + joinrel->fdwroutine->GetForeignJoinPaths(root, joinrel, + outerrel, innerrel, + jointype, &extra); + + /* + * 6. Finally, give extensions a chance to manipulate the path list. They + * could add new paths (such as CustomPaths) by calling add_path(), or + * add_partial_path() if parallel aware. They could also delete or modify + * paths added by the core code. + */ + if (set_join_pathlist_hook && + consider_join_pushdown) + set_join_pathlist_hook(root, joinrel, outerrel, innerrel, + jointype, &extra); +} + +/* + * We override the param_source_rels heuristic to accept nestloop paths in + * which the outer rel satisfies some but not all of the inner path's + * parameterization. This is necessary to get good plans for star-schema + * scenarios, in which a parameterized path for a large table may require + * parameters from multiple small tables that will not get joined directly to + * each other. We can handle that by stacking nestloops that have the small + * tables on the outside; but this breaks the rule the param_source_rels + * heuristic is based on, namely that parameters should not be passed down + * across joins unless there's a join-order-constraint-based reason to do so. + * So we ignore the param_source_rels restriction when this case applies. + * + * allow_star_schema_join() returns true if the param_source_rels restriction + * should be overridden, ie, it's okay to perform this join. + */ +static inline bool +allow_star_schema_join(PlannerInfo *root, + Relids outerrelids, + Relids inner_paramrels) +{ + /* + * It's a star-schema case if the outer rel provides some but not all of + * the inner rel's parameterization. + */ + return (bms_overlap(inner_paramrels, outerrelids) && + bms_nonempty_difference(inner_paramrels, outerrelids)); +} + +/* + * paraminfo_get_equal_hashops + * Determine if param_info and innerrel's lateral_vars can be hashed. + * Returns true the hashing is possible, otherwise return false. + * + * Additionally we also collect the outer exprs and the hash operators for + * each parameter to innerrel. These set in 'param_exprs', 'operators' and + * 'binary_mode' when we return true. + */ +static bool +paraminfo_get_equal_hashops(PlannerInfo *root, ParamPathInfo *param_info, + RelOptInfo *outerrel, RelOptInfo *innerrel, + List **param_exprs, List **operators, + bool *binary_mode) + +{ + ListCell *lc; + + *param_exprs = NIL; + *operators = NIL; + *binary_mode = false; + + if (param_info != NULL) + { + List *clauses = param_info->ppi_clauses; + + foreach(lc, clauses) + { + RestrictInfo *rinfo = (RestrictInfo *) lfirst(lc); + OpExpr *opexpr; + Node *expr; + Oid hasheqoperator; + + opexpr = (OpExpr *) rinfo->clause; + + /* + * Bail if the rinfo is not compatible. We need a join OpExpr + * with 2 args. + */ + if (!IsA(opexpr, OpExpr) || list_length(opexpr->args) != 2 || + !clause_sides_match_join(rinfo, outerrel, innerrel)) + { + list_free(*operators); + list_free(*param_exprs); + return false; + } + + if (rinfo->outer_is_left) + { + expr = (Node *) linitial(opexpr->args); + hasheqoperator = rinfo->left_hasheqoperator; + } + else + { + expr = (Node *) lsecond(opexpr->args); + hasheqoperator = rinfo->right_hasheqoperator; + } + + /* can't do memoize if we can't hash the outer type */ + if (!OidIsValid(hasheqoperator)) + { + list_free(*operators); + list_free(*param_exprs); + return false; + } + + *operators = lappend_oid(*operators, hasheqoperator); + *param_exprs = lappend(*param_exprs, expr); + + /* + * When the join operator is not hashable then it's possible that + * the operator will be able to distinguish something that the + * hash equality operator could not. For example with floating + * point types -0.0 and +0.0 are classed as equal by the hash + * function and equality function, but some other operator may be + * able to tell those values apart. This means that we must put + * memoize into binary comparison mode so that it does bit-by-bit + * comparisons rather than a "logical" comparison as it would + * using the hash equality operator. + */ + if (!OidIsValid(rinfo->hashjoinoperator)) + *binary_mode = true; + } + } + + /* Now add any lateral vars to the cache key too */ + foreach(lc, innerrel->lateral_vars) + { + Node *expr = (Node *) lfirst(lc); + TypeCacheEntry *typentry; + + /* Reject if there are any volatile functions */ + if (contain_volatile_functions(expr)) + { + list_free(*operators); + list_free(*param_exprs); + return false; + } + + typentry = lookup_type_cache(exprType(expr), + TYPECACHE_HASH_PROC | TYPECACHE_EQ_OPR); + + /* can't use a memoize node without a valid hash equals operator */ + if (!OidIsValid(typentry->hash_proc) || !OidIsValid(typentry->eq_opr)) + { + list_free(*operators); + list_free(*param_exprs); + return false; + } + + *operators = lappend_oid(*operators, typentry->eq_opr); + *param_exprs = lappend(*param_exprs, expr); + + /* + * We must go into binary mode as we don't have too much of an idea of + * how these lateral Vars are being used. See comment above when we + * set *binary_mode for the non-lateral Var case. This could be + * relaxed a bit if we had the RestrictInfos and knew the operators + * being used, however for cases like Vars that are arguments to + * functions we must operate in binary mode as we don't have + * visibility into what the function is doing with the Vars. + */ + *binary_mode = true; + } + + /* We're okay to use memoize */ + return true; +} + +/* + * get_memoize_path + * If possible, make and return a Memoize path atop of 'inner_path'. + * Otherwise return NULL. + */ +static Path * +get_memoize_path(PlannerInfo *root, RelOptInfo *innerrel, + RelOptInfo *outerrel, Path *inner_path, + Path *outer_path, JoinType jointype, + JoinPathExtraData *extra) +{ + RelOptInfo *top_outerrel; + List *param_exprs; + List *hash_operators; + ListCell *lc; + bool binary_mode; + + /* Obviously not if it's disabled */ + if (!enable_memoize) + return NULL; + + /* + * We can safely not bother with all this unless we expect to perform more + * than one inner scan. The first scan is always going to be a cache + * miss. This would likely fail later anyway based on costs, so this is + * really just to save some wasted effort. + */ + if (outer_path->parent->rows < 2) + return NULL; + + /* + * We can only have a memoize node when there's some kind of cache key, + * either parameterized path clauses or lateral Vars. No cache key sounds + * more like something a Materialize node might be more useful for. + */ + if ((inner_path->param_info == NULL || + inner_path->param_info->ppi_clauses == NIL) && + innerrel->lateral_vars == NIL) + return NULL; + + /* + * Currently we don't do this for SEMI and ANTI joins unless they're + * marked as inner_unique. This is because nested loop SEMI/ANTI joins + * don't scan the inner node to completion, which will mean memoize cannot + * mark the cache entry as complete. + * + * XXX Currently we don't attempt to mark SEMI/ANTI joins as inner_unique + * = true. Should we? See add_paths_to_joinrel() + */ + if (!extra->inner_unique && (jointype == JOIN_SEMI || + jointype == JOIN_ANTI)) + return NULL; + + /* + * Memoize normally marks cache entries as complete when it runs out of + * tuples to read from its subplan. However, with unique joins, Nested + * Loop will skip to the next outer tuple after finding the first matching + * inner tuple. This means that we may not read the inner side of the + * join to completion which leaves no opportunity to mark the cache entry + * as complete. To work around that, when the join is unique we + * automatically mark cache entries as complete after fetching the first + * tuple. This works when the entire join condition is parameterized. + * Otherwise, when the parameterization is only a subset of the join + * condition, we can't be sure which part of it causes the join to be + * unique. This means there are no guarantees that only 1 tuple will be + * read. We cannot mark the cache entry as complete after reading the + * first tuple without that guarantee. This means the scope of Memoize + * node's usefulness is limited to only outer rows that have no join + * partner as this is the only case where Nested Loop would exhaust the + * inner scan of a unique join. Since the scope is limited to that, we + * just don't bother making a memoize path in this case. + * + * Lateral vars needn't be considered here as they're not considered when + * determining if the join is unique. + * + * XXX this could be enabled if the remaining join quals were made part of + * the inner scan's filter instead of the join filter. Maybe it's worth + * considering doing that? + */ + if (extra->inner_unique && + (inner_path->param_info == NULL || + list_length(inner_path->param_info->ppi_clauses) < + list_length(extra->restrictlist))) + return NULL; + + /* + * We can't use a memoize node if there are volatile functions in the + * inner rel's target list or restrict list. A cache hit could reduce the + * number of calls to these functions. + */ + if (contain_volatile_functions((Node *) innerrel->reltarget)) + return NULL; + + foreach(lc, innerrel->baserestrictinfo) + { + RestrictInfo *rinfo = (RestrictInfo *) lfirst(lc); + + if (contain_volatile_functions((Node *) rinfo)) + return NULL; + } + + /* + * Also check the parameterized path restrictinfos for volatile functions. + * Indexed functions must be immutable so shouldn't have any volatile + * functions, however, with a lateral join the inner scan may not be an + * index scan. + */ + if (inner_path->param_info != NULL) + { + foreach(lc, inner_path->param_info->ppi_clauses) + { + RestrictInfo *rinfo = (RestrictInfo *) lfirst(lc); + + if (contain_volatile_functions((Node *) rinfo)) + return NULL; + } + } + + /* + * When considering a partitionwise join, we have clauses that reference + * the outerrel's top parent not outerrel itself. + */ + if (outerrel->reloptkind == RELOPT_OTHER_MEMBER_REL) + top_outerrel = find_base_rel(root, bms_singleton_member(outerrel->top_parent_relids)); + else if (outerrel->reloptkind == RELOPT_OTHER_JOINREL) + top_outerrel = find_join_rel(root, outerrel->top_parent_relids); + else + top_outerrel = outerrel; + + /* Check if we have hash ops for each parameter to the path */ + if (paraminfo_get_equal_hashops(root, + inner_path->param_info, + top_outerrel, + innerrel, + ¶m_exprs, + &hash_operators, + &binary_mode)) + { + return (Path *) create_memoize_path(root, + innerrel, + inner_path, + param_exprs, + hash_operators, + extra->inner_unique, + binary_mode, + outer_path->rows); + } + + return NULL; +} + +/* + * try_nestloop_path + * Consider a nestloop join path; if it appears useful, push it into + * the joinrel's pathlist via add_path(). + */ +static void +try_nestloop_path(PlannerInfo *root, + RelOptInfo *joinrel, + Path *outer_path, + Path *inner_path, + List *pathkeys, + JoinType jointype, + JoinPathExtraData *extra) +{ + Relids required_outer; + JoinCostWorkspace workspace; + RelOptInfo *innerrel = inner_path->parent; + RelOptInfo *outerrel = outer_path->parent; + Relids innerrelids; + Relids outerrelids; + Relids inner_paramrels = PATH_REQ_OUTER(inner_path); + Relids outer_paramrels = PATH_REQ_OUTER(outer_path); + + /* + * Paths are parameterized by top-level parents, so run parameterization + * tests on the parent relids. + */ + if (innerrel->top_parent_relids) + innerrelids = innerrel->top_parent_relids; + else + innerrelids = innerrel->relids; + + if (outerrel->top_parent_relids) + outerrelids = outerrel->top_parent_relids; + else + outerrelids = outerrel->relids; + + /* + * Check to see if proposed path is still parameterized, and reject if the + * parameterization wouldn't be sensible --- unless allow_star_schema_join + * says to allow it anyway. Also, we must reject if have_dangerous_phv + * doesn't like the look of it, which could only happen if the nestloop is + * still parameterized. + */ + required_outer = calc_nestloop_required_outer(outerrelids, outer_paramrels, + innerrelids, inner_paramrels); + if (required_outer && + ((!bms_overlap(required_outer, extra->param_source_rels) && + !allow_star_schema_join(root, outerrelids, inner_paramrels)) || + have_dangerous_phv(root, outerrelids, inner_paramrels))) + { + /* Waste no memory when we reject a path here */ + bms_free(required_outer); + return; + } + + /* + * Do a precheck to quickly eliminate obviously-inferior paths. We + * calculate a cheap lower bound on the path's cost and then use + * add_path_precheck() to see if the path is clearly going to be dominated + * by some existing path for the joinrel. If not, do the full pushup with + * creating a fully valid path structure and submitting it to add_path(). + * The latter two steps are expensive enough to make this two-phase + * methodology worthwhile. + */ + initial_cost_nestloop(root, &workspace, jointype, + outer_path, inner_path, extra); + + if (add_path_precheck(joinrel, + workspace.startup_cost, workspace.total_cost, + pathkeys, required_outer)) + { + /* + * If the inner path is parameterized, it is parameterized by the + * topmost parent of the outer rel, not the outer rel itself. Fix + * that. + */ + if (PATH_PARAM_BY_PARENT(inner_path, outer_path->parent)) + { + inner_path = reparameterize_path_by_child(root, inner_path, + outer_path->parent); + + /* + * If we could not translate the path, we can't create nest loop + * path. + */ + if (!inner_path) + { + bms_free(required_outer); + return; + } + } + + add_path(joinrel, (Path *) + create_nestloop_path(root, + joinrel, + jointype, + &workspace, + extra, + outer_path, + inner_path, + extra->restrictlist, + pathkeys, + required_outer)); + } + else + { + /* Waste no memory when we reject a path here */ + bms_free(required_outer); + } +} + +/* + * try_partial_nestloop_path + * Consider a partial nestloop join path; if it appears useful, push it into + * the joinrel's partial_pathlist via add_partial_path(). + */ +static void +try_partial_nestloop_path(PlannerInfo *root, + RelOptInfo *joinrel, + Path *outer_path, + Path *inner_path, + List *pathkeys, + JoinType jointype, + JoinPathExtraData *extra) +{ + JoinCostWorkspace workspace; + + /* + * If the inner path is parameterized, the parameterization must be fully + * satisfied by the proposed outer path. Parameterized partial paths are + * not supported. The caller should already have verified that no lateral + * rels are required here. + */ + Assert(bms_is_empty(joinrel->lateral_relids)); + if (inner_path->param_info != NULL) + { + Relids inner_paramrels = inner_path->param_info->ppi_req_outer; + RelOptInfo *outerrel = outer_path->parent; + Relids outerrelids; + + /* + * The inner and outer paths are parameterized, if at all, by the top + * level parents, not the child relations, so we must use those relids + * for our parameterization tests. + */ + if (outerrel->top_parent_relids) + outerrelids = outerrel->top_parent_relids; + else + outerrelids = outerrel->relids; + + if (!bms_is_subset(inner_paramrels, outerrelids)) + return; + } + + /* + * Before creating a path, get a quick lower bound on what it is likely to + * cost. Bail out right away if it looks terrible. + */ + initial_cost_nestloop(root, &workspace, jointype, + outer_path, inner_path, extra); + if (!add_partial_path_precheck(joinrel, workspace.total_cost, pathkeys)) + return; + + /* + * If the inner path is parameterized, it is parameterized by the topmost + * parent of the outer rel, not the outer rel itself. Fix that. + */ + if (PATH_PARAM_BY_PARENT(inner_path, outer_path->parent)) + { + inner_path = reparameterize_path_by_child(root, inner_path, + outer_path->parent); + + /* + * If we could not translate the path, we can't create nest loop path. + */ + if (!inner_path) + return; + } + + /* Might be good enough to be worth trying, so let's try it. */ + add_partial_path(joinrel, (Path *) + create_nestloop_path(root, + joinrel, + jointype, + &workspace, + extra, + outer_path, + inner_path, + extra->restrictlist, + pathkeys, + NULL)); +} + +/* + * try_mergejoin_path + * Consider a merge join path; if it appears useful, push it into + * the joinrel's pathlist via add_path(). + */ +static void +try_mergejoin_path(PlannerInfo *root, + RelOptInfo *joinrel, + Path *outer_path, + Path *inner_path, + List *pathkeys, + List *mergeclauses, + List *outersortkeys, + List *innersortkeys, + JoinType jointype, + JoinPathExtraData *extra, + bool is_partial) +{ + Relids required_outer; + JoinCostWorkspace workspace; + + if (is_partial) + { + try_partial_mergejoin_path(root, + joinrel, + outer_path, + inner_path, + pathkeys, + mergeclauses, + outersortkeys, + innersortkeys, + jointype, + extra); + return; + } + + /* + * Check to see if proposed path is still parameterized, and reject if the + * parameterization wouldn't be sensible. + */ + required_outer = calc_non_nestloop_required_outer(outer_path, + inner_path); + if (required_outer && + !bms_overlap(required_outer, extra->param_source_rels)) + { + /* Waste no memory when we reject a path here */ + bms_free(required_outer); + return; + } + + /* + * If the given paths are already well enough ordered, we can skip doing + * an explicit sort. + */ + if (outersortkeys && + pathkeys_contained_in(outersortkeys, outer_path->pathkeys)) + outersortkeys = NIL; + if (innersortkeys && + pathkeys_contained_in(innersortkeys, inner_path->pathkeys)) + innersortkeys = NIL; + + /* + * See comments in try_nestloop_path(). + */ + initial_cost_mergejoin(root, &workspace, jointype, mergeclauses, + outer_path, inner_path, + outersortkeys, innersortkeys, + extra); + + if (add_path_precheck(joinrel, + workspace.startup_cost, workspace.total_cost, + pathkeys, required_outer)) + { + add_path(joinrel, (Path *) + create_mergejoin_path(root, + joinrel, + jointype, + &workspace, + extra, + outer_path, + inner_path, + extra->restrictlist, + pathkeys, + required_outer, + mergeclauses, + outersortkeys, + innersortkeys)); + } + else + { + /* Waste no memory when we reject a path here */ + bms_free(required_outer); + } +} + +/* + * try_partial_mergejoin_path + * Consider a partial merge join path; if it appears useful, push it into + * the joinrel's pathlist via add_partial_path(). + */ +static void +try_partial_mergejoin_path(PlannerInfo *root, + RelOptInfo *joinrel, + Path *outer_path, + Path *inner_path, + List *pathkeys, + List *mergeclauses, + List *outersortkeys, + List *innersortkeys, + JoinType jointype, + JoinPathExtraData *extra) +{ + JoinCostWorkspace workspace; + + /* + * See comments in try_partial_hashjoin_path(). + */ + Assert(bms_is_empty(joinrel->lateral_relids)); + if (inner_path->param_info != NULL) + { + Relids inner_paramrels = inner_path->param_info->ppi_req_outer; + + if (!bms_is_empty(inner_paramrels)) + return; + } + + /* + * If the given paths are already well enough ordered, we can skip doing + * an explicit sort. + */ + if (outersortkeys && + pathkeys_contained_in(outersortkeys, outer_path->pathkeys)) + outersortkeys = NIL; + if (innersortkeys && + pathkeys_contained_in(innersortkeys, inner_path->pathkeys)) + innersortkeys = NIL; + + /* + * See comments in try_partial_nestloop_path(). + */ + initial_cost_mergejoin(root, &workspace, jointype, mergeclauses, + outer_path, inner_path, + outersortkeys, innersortkeys, + extra); + + if (!add_partial_path_precheck(joinrel, workspace.total_cost, pathkeys)) + return; + + /* Might be good enough to be worth trying, so let's try it. */ + add_partial_path(joinrel, (Path *) + create_mergejoin_path(root, + joinrel, + jointype, + &workspace, + extra, + outer_path, + inner_path, + extra->restrictlist, + pathkeys, + NULL, + mergeclauses, + outersortkeys, + innersortkeys)); +} + +/* + * try_hashjoin_path + * Consider a hash join path; if it appears useful, push it into + * the joinrel's pathlist via add_path(). + */ +static void +try_hashjoin_path(PlannerInfo *root, + RelOptInfo *joinrel, + Path *outer_path, + Path *inner_path, + List *hashclauses, + JoinType jointype, + JoinPathExtraData *extra) +{ + Relids required_outer; + JoinCostWorkspace workspace; + + /* + * Check to see if proposed path is still parameterized, and reject if the + * parameterization wouldn't be sensible. + */ + required_outer = calc_non_nestloop_required_outer(outer_path, + inner_path); + if (required_outer && + !bms_overlap(required_outer, extra->param_source_rels)) + { + /* Waste no memory when we reject a path here */ + bms_free(required_outer); + return; + } + + /* + * See comments in try_nestloop_path(). Also note that hashjoin paths + * never have any output pathkeys, per comments in create_hashjoin_path. + */ + initial_cost_hashjoin(root, &workspace, jointype, hashclauses, + outer_path, inner_path, extra, false); + + if (add_path_precheck(joinrel, + workspace.startup_cost, workspace.total_cost, + NIL, required_outer)) + { + add_path(joinrel, (Path *) + create_hashjoin_path(root, + joinrel, + jointype, + &workspace, + extra, + outer_path, + inner_path, + false, /* parallel_hash */ + extra->restrictlist, + required_outer, + hashclauses)); + } + else + { + /* Waste no memory when we reject a path here */ + bms_free(required_outer); + } +} + +/* + * try_partial_hashjoin_path + * Consider a partial hashjoin join path; if it appears useful, push it into + * the joinrel's partial_pathlist via add_partial_path(). + * The outer side is partial. If parallel_hash is true, then the inner path + * must be partial and will be run in parallel to create one or more shared + * hash tables; otherwise the inner path must be complete and a copy of it + * is run in every process to create separate identical private hash tables. + */ +static void +try_partial_hashjoin_path(PlannerInfo *root, + RelOptInfo *joinrel, + Path *outer_path, + Path *inner_path, + List *hashclauses, + JoinType jointype, + JoinPathExtraData *extra, + bool parallel_hash) +{ + JoinCostWorkspace workspace; + + /* + * If the inner path is parameterized, the parameterization must be fully + * satisfied by the proposed outer path. Parameterized partial paths are + * not supported. The caller should already have verified that no lateral + * rels are required here. + */ + Assert(bms_is_empty(joinrel->lateral_relids)); + if (inner_path->param_info != NULL) + { + Relids inner_paramrels = inner_path->param_info->ppi_req_outer; + + if (!bms_is_empty(inner_paramrels)) + return; + } + + /* + * Before creating a path, get a quick lower bound on what it is likely to + * cost. Bail out right away if it looks terrible. + */ + initial_cost_hashjoin(root, &workspace, jointype, hashclauses, + outer_path, inner_path, extra, parallel_hash); + if (!add_partial_path_precheck(joinrel, workspace.total_cost, NIL)) + return; + + /* Might be good enough to be worth trying, so let's try it. */ + add_partial_path(joinrel, (Path *) + create_hashjoin_path(root, + joinrel, + jointype, + &workspace, + extra, + outer_path, + inner_path, + parallel_hash, + extra->restrictlist, + NULL, + hashclauses)); +} + +/* + * clause_sides_match_join + * Determine whether a join clause is of the right form to use in this join. + * + * We already know that the clause is a binary opclause referencing only the + * rels in the current join. The point here is to check whether it has the + * form "outerrel_expr op innerrel_expr" or "innerrel_expr op outerrel_expr", + * rather than mixing outer and inner vars on either side. If it matches, + * we set the transient flag outer_is_left to identify which side is which. + */ +static inline bool +clause_sides_match_join(RestrictInfo *rinfo, RelOptInfo *outerrel, + RelOptInfo *innerrel) +{ + if (bms_is_subset(rinfo->left_relids, outerrel->relids) && + bms_is_subset(rinfo->right_relids, innerrel->relids)) + { + /* lefthand side is outer */ + rinfo->outer_is_left = true; + return true; + } + else if (bms_is_subset(rinfo->left_relids, innerrel->relids) && + bms_is_subset(rinfo->right_relids, outerrel->relids)) + { + /* righthand side is outer */ + rinfo->outer_is_left = false; + return true; + } + return false; /* no good for these input relations */ +} + +/* + * sort_inner_and_outer + * Create mergejoin join paths by explicitly sorting both the outer and + * inner join relations on each available merge ordering. + * + * 'joinrel' is the join relation + * 'outerrel' is the outer join relation + * 'innerrel' is the inner join relation + * 'jointype' is the type of join to do + * 'extra' contains additional input values + */ +static void +sort_inner_and_outer(PlannerInfo *root, + RelOptInfo *joinrel, + RelOptInfo *outerrel, + RelOptInfo *innerrel, + JoinType jointype, + JoinPathExtraData *extra) +{ + JoinType save_jointype = jointype; + Path *outer_path; + Path *inner_path; + Path *cheapest_partial_outer = NULL; + Path *cheapest_safe_inner = NULL; + List *all_pathkeys; + ListCell *l; + + /* + * We only consider the cheapest-total-cost input paths, since we are + * assuming here that a sort is required. We will consider + * cheapest-startup-cost input paths later, and only if they don't need a + * sort. + * + * This function intentionally does not consider parameterized input + * paths, except when the cheapest-total is parameterized. If we did so, + * we'd have a combinatorial explosion of mergejoin paths of dubious + * value. This interacts with decisions elsewhere that also discriminate + * against mergejoins with parameterized inputs; see comments in + * src/backend/optimizer/README. + */ + outer_path = outerrel->cheapest_total_path; + inner_path = innerrel->cheapest_total_path; + + /* + * If either cheapest-total path is parameterized by the other rel, we + * can't use a mergejoin. (There's no use looking for alternative input + * paths, since these should already be the least-parameterized available + * paths.) + */ + if (PATH_PARAM_BY_REL(outer_path, innerrel) || + PATH_PARAM_BY_REL(inner_path, outerrel)) + return; + + /* + * If unique-ification is requested, do it and then handle as a plain + * inner join. + */ + if (jointype == JOIN_UNIQUE_OUTER) + { + outer_path = (Path *) create_unique_path(root, outerrel, + outer_path, extra->sjinfo); + Assert(outer_path); + jointype = JOIN_INNER; + } + else if (jointype == JOIN_UNIQUE_INNER) + { + inner_path = (Path *) create_unique_path(root, innerrel, + inner_path, extra->sjinfo); + Assert(inner_path); + jointype = JOIN_INNER; + } + + /* + * If the joinrel is parallel-safe, we may be able to consider a partial + * merge join. However, we can't handle JOIN_UNIQUE_OUTER, because the + * outer path will be partial, and therefore we won't be able to properly + * guarantee uniqueness. Similarly, we can't handle JOIN_FULL and + * JOIN_RIGHT, because they can produce false null extended rows. Also, + * the resulting path must not be parameterized. + */ + if (joinrel->consider_parallel && + save_jointype != JOIN_UNIQUE_OUTER && + save_jointype != JOIN_FULL && + save_jointype != JOIN_RIGHT && + outerrel->partial_pathlist != NIL && + bms_is_empty(joinrel->lateral_relids)) + { + cheapest_partial_outer = (Path *) linitial(outerrel->partial_pathlist); + + if (inner_path->parallel_safe) + cheapest_safe_inner = inner_path; + else if (save_jointype != JOIN_UNIQUE_INNER) + cheapest_safe_inner = + get_cheapest_parallel_safe_total_inner(innerrel->pathlist); + } + + /* + * Each possible ordering of the available mergejoin clauses will generate + * a differently-sorted result path at essentially the same cost. We have + * no basis for choosing one over another at this level of joining, but + * some sort orders may be more useful than others for higher-level + * mergejoins, so it's worth considering multiple orderings. + * + * Actually, it's not quite true that every mergeclause ordering will + * generate a different path order, because some of the clauses may be + * partially redundant (refer to the same EquivalenceClasses). Therefore, + * what we do is convert the mergeclause list to a list of canonical + * pathkeys, and then consider different orderings of the pathkeys. + * + * Generating a path for *every* permutation of the pathkeys doesn't seem + * like a winning strategy; the cost in planning time is too high. For + * now, we generate one path for each pathkey, listing that pathkey first + * and the rest in random order. This should allow at least a one-clause + * mergejoin without re-sorting against any other possible mergejoin + * partner path. But if we've not guessed the right ordering of secondary + * keys, we may end up evaluating clauses as qpquals when they could have + * been done as mergeclauses. (In practice, it's rare that there's more + * than two or three mergeclauses, so expending a huge amount of thought + * on that is probably not worth it.) + * + * The pathkey order returned by select_outer_pathkeys_for_merge() has + * some heuristics behind it (see that function), so be sure to try it + * exactly as-is as well as making variants. + */ + all_pathkeys = select_outer_pathkeys_for_merge(root, + extra->mergeclause_list, + joinrel); + + foreach(l, all_pathkeys) + { + PathKey *front_pathkey = (PathKey *) lfirst(l); + List *cur_mergeclauses; + List *outerkeys; + List *innerkeys; + List *merge_pathkeys; + + /* Make a pathkey list with this guy first */ + if (l != list_head(all_pathkeys)) + outerkeys = lcons(front_pathkey, + list_delete_nth_cell(list_copy(all_pathkeys), + foreach_current_index(l))); + else + outerkeys = all_pathkeys; /* no work at first one... */ + + /* Sort the mergeclauses into the corresponding ordering */ + cur_mergeclauses = + find_mergeclauses_for_outer_pathkeys(root, + outerkeys, + extra->mergeclause_list); + + /* Should have used them all... */ + Assert(list_length(cur_mergeclauses) == list_length(extra->mergeclause_list)); + + /* Build sort pathkeys for the inner side */ + innerkeys = make_inner_pathkeys_for_merge(root, + cur_mergeclauses, + outerkeys); + + /* Build pathkeys representing output sort order */ + merge_pathkeys = build_join_pathkeys(root, joinrel, jointype, + outerkeys); + + /* + * And now we can make the path. + * + * Note: it's possible that the cheapest paths will already be sorted + * properly. try_mergejoin_path will detect that case and suppress an + * explicit sort step, so we needn't do so here. + */ + try_mergejoin_path(root, + joinrel, + outer_path, + inner_path, + merge_pathkeys, + cur_mergeclauses, + outerkeys, + innerkeys, + jointype, + extra, + false); + + /* + * If we have partial outer and parallel safe inner path then try + * partial mergejoin path. + */ + if (cheapest_partial_outer && cheapest_safe_inner) + try_partial_mergejoin_path(root, + joinrel, + cheapest_partial_outer, + cheapest_safe_inner, + merge_pathkeys, + cur_mergeclauses, + outerkeys, + innerkeys, + jointype, + extra); + } +} + +/* + * generate_mergejoin_paths + * Creates possible mergejoin paths for input outerpath. + * + * We generate mergejoins if mergejoin clauses are available. We have + * two ways to generate the inner path for a mergejoin: sort the cheapest + * inner path, or use an inner path that is already suitably ordered for the + * merge. If we have several mergeclauses, it could be that there is no inner + * path (or only a very expensive one) for the full list of mergeclauses, but + * better paths exist if we truncate the mergeclause list (thereby discarding + * some sort key requirements). So, we consider truncations of the + * mergeclause list as well as the full list. (Ideally we'd consider all + * subsets of the mergeclause list, but that seems way too expensive.) + */ +static void +generate_mergejoin_paths(PlannerInfo *root, + RelOptInfo *joinrel, + RelOptInfo *innerrel, + Path *outerpath, + JoinType jointype, + JoinPathExtraData *extra, + bool useallclauses, + Path *inner_cheapest_total, + List *merge_pathkeys, + bool is_partial) +{ + List *mergeclauses; + List *innersortkeys; + List *trialsortkeys; + Path *cheapest_startup_inner; + Path *cheapest_total_inner; + JoinType save_jointype = jointype; + int num_sortkeys; + int sortkeycnt; + + if (jointype == JOIN_UNIQUE_OUTER || jointype == JOIN_UNIQUE_INNER) + jointype = JOIN_INNER; + + /* Look for useful mergeclauses (if any) */ + mergeclauses = + find_mergeclauses_for_outer_pathkeys(root, + outerpath->pathkeys, + extra->mergeclause_list); + + /* + * Done with this outer path if no chance for a mergejoin. + * + * Special corner case: for "x FULL JOIN y ON true", there will be no join + * clauses at all. Ordinarily we'd generate a clauseless nestloop path, + * but since mergejoin is our only join type that supports FULL JOIN + * without any join clauses, it's necessary to generate a clauseless + * mergejoin path instead. + */ + if (mergeclauses == NIL) + { + if (jointype == JOIN_FULL) + /* okay to try for mergejoin */ ; + else + return; + } + if (useallclauses && + list_length(mergeclauses) != list_length(extra->mergeclause_list)) + return; + + /* Compute the required ordering of the inner path */ + innersortkeys = make_inner_pathkeys_for_merge(root, + mergeclauses, + outerpath->pathkeys); + + /* + * Generate a mergejoin on the basis of sorting the cheapest inner. Since + * a sort will be needed, only cheapest total cost matters. (But + * try_mergejoin_path will do the right thing if inner_cheapest_total is + * already correctly sorted.) + */ + try_mergejoin_path(root, + joinrel, + outerpath, + inner_cheapest_total, + merge_pathkeys, + mergeclauses, + NIL, + innersortkeys, + jointype, + extra, + is_partial); + + /* Can't do anything else if inner path needs to be unique'd */ + if (save_jointype == JOIN_UNIQUE_INNER) + return; + + /* + * Look for presorted inner paths that satisfy the innersortkey list --- + * or any truncation thereof, if we are allowed to build a mergejoin using + * a subset of the merge clauses. Here, we consider both cheap startup + * cost and cheap total cost. + * + * Currently we do not consider parameterized inner paths here. This + * interacts with decisions elsewhere that also discriminate against + * mergejoins with parameterized inputs; see comments in + * src/backend/optimizer/README. + * + * As we shorten the sortkey list, we should consider only paths that are + * strictly cheaper than (in particular, not the same as) any path found + * in an earlier iteration. Otherwise we'd be intentionally using fewer + * merge keys than a given path allows (treating the rest as plain + * joinquals), which is unlikely to be a good idea. Also, eliminating + * paths here on the basis of compare_path_costs is a lot cheaper than + * building the mergejoin path only to throw it away. + * + * If inner_cheapest_total is well enough sorted to have not required a + * sort in the path made above, we shouldn't make a duplicate path with + * it, either. We handle that case with the same logic that handles the + * previous consideration, by initializing the variables that track + * cheapest-so-far properly. Note that we do NOT reject + * inner_cheapest_total if we find it matches some shorter set of + * pathkeys. That case corresponds to using fewer mergekeys to avoid + * sorting inner_cheapest_total, whereas we did sort it above, so the + * plans being considered are different. + */ + if (pathkeys_contained_in(innersortkeys, + inner_cheapest_total->pathkeys)) + { + /* inner_cheapest_total didn't require a sort */ + cheapest_startup_inner = inner_cheapest_total; + cheapest_total_inner = inner_cheapest_total; + } + else + { + /* it did require a sort, at least for the full set of keys */ + cheapest_startup_inner = NULL; + cheapest_total_inner = NULL; + } + num_sortkeys = list_length(innersortkeys); + if (num_sortkeys > 1 && !useallclauses) + trialsortkeys = list_copy(innersortkeys); /* need modifiable copy */ + else + trialsortkeys = innersortkeys; /* won't really truncate */ + + for (sortkeycnt = num_sortkeys; sortkeycnt > 0; sortkeycnt--) + { + Path *innerpath; + List *newclauses = NIL; + + /* + * Look for an inner path ordered well enough for the first + * 'sortkeycnt' innersortkeys. NB: trialsortkeys list is modified + * destructively, which is why we made a copy... + */ + trialsortkeys = list_truncate(trialsortkeys, sortkeycnt); + innerpath = get_cheapest_path_for_pathkeys(innerrel->pathlist, + trialsortkeys, + NULL, + TOTAL_COST, + is_partial); + if (innerpath != NULL && + (cheapest_total_inner == NULL || + compare_path_costs(innerpath, cheapest_total_inner, + TOTAL_COST) < 0)) + { + /* Found a cheap (or even-cheaper) sorted path */ + /* Select the right mergeclauses, if we didn't already */ + if (sortkeycnt < num_sortkeys) + { + newclauses = + trim_mergeclauses_for_inner_pathkeys(root, + mergeclauses, + trialsortkeys); + Assert(newclauses != NIL); + } + else + newclauses = mergeclauses; + try_mergejoin_path(root, + joinrel, + outerpath, + innerpath, + merge_pathkeys, + newclauses, + NIL, + NIL, + jointype, + extra, + is_partial); + cheapest_total_inner = innerpath; + } + /* Same on the basis of cheapest startup cost ... */ + innerpath = get_cheapest_path_for_pathkeys(innerrel->pathlist, + trialsortkeys, + NULL, + STARTUP_COST, + is_partial); + if (innerpath != NULL && + (cheapest_startup_inner == NULL || + compare_path_costs(innerpath, cheapest_startup_inner, + STARTUP_COST) < 0)) + { + /* Found a cheap (or even-cheaper) sorted path */ + if (innerpath != cheapest_total_inner) + { + /* + * Avoid rebuilding clause list if we already made one; saves + * memory in big join trees... + */ + if (newclauses == NIL) + { + if (sortkeycnt < num_sortkeys) + { + newclauses = + trim_mergeclauses_for_inner_pathkeys(root, + mergeclauses, + trialsortkeys); + Assert(newclauses != NIL); + } + else + newclauses = mergeclauses; + } + try_mergejoin_path(root, + joinrel, + outerpath, + innerpath, + merge_pathkeys, + newclauses, + NIL, + NIL, + jointype, + extra, + is_partial); + } + cheapest_startup_inner = innerpath; + } + + /* + * Don't consider truncated sortkeys if we need all clauses. + */ + if (useallclauses) + break; + } +} + +/* + * match_unsorted_outer + * Creates possible join paths for processing a single join relation + * 'joinrel' by employing either iterative substitution or + * mergejoining on each of its possible outer paths (considering + * only outer paths that are already ordered well enough for merging). + * + * We always generate a nestloop path for each available outer path. + * In fact we may generate as many as five: one on the cheapest-total-cost + * inner path, one on the same with materialization, one on the + * cheapest-startup-cost inner path (if different), one on the + * cheapest-total inner-indexscan path (if any), and one on the + * cheapest-startup inner-indexscan path (if different). + * + * We also consider mergejoins if mergejoin clauses are available. See + * detailed comments in generate_mergejoin_paths. + * + * 'joinrel' is the join relation + * 'outerrel' is the outer join relation + * 'innerrel' is the inner join relation + * 'jointype' is the type of join to do + * 'extra' contains additional input values + */ +static void +match_unsorted_outer(PlannerInfo *root, + RelOptInfo *joinrel, + RelOptInfo *outerrel, + RelOptInfo *innerrel, + JoinType jointype, + JoinPathExtraData *extra) +{ + JoinType save_jointype = jointype; + bool nestjoinOK; + bool useallclauses; + Path *inner_cheapest_total = innerrel->cheapest_total_path; + Path *matpath = NULL; + ListCell *lc1; + + /* + * Nestloop only supports inner, left, semi, and anti joins. Also, if we + * are doing a right or full mergejoin, we must use *all* the mergeclauses + * as join clauses, else we will not have a valid plan. (Although these + * two flags are currently inverses, keep them separate for clarity and + * possible future changes.) + */ + switch (jointype) + { + case JOIN_INNER: + case JOIN_LEFT: + case JOIN_SEMI: + case JOIN_ANTI: + nestjoinOK = true; + useallclauses = false; + break; + case JOIN_RIGHT: + case JOIN_FULL: + nestjoinOK = false; + useallclauses = true; + break; + case JOIN_UNIQUE_OUTER: + case JOIN_UNIQUE_INNER: + jointype = JOIN_INNER; + nestjoinOK = true; + useallclauses = false; + break; + default: + elog(ERROR, "unrecognized join type: %d", + (int) jointype); + nestjoinOK = false; /* keep compiler quiet */ + useallclauses = false; + break; + } + + /* + * If inner_cheapest_total is parameterized by the outer rel, ignore it; + * we will consider it below as a member of cheapest_parameterized_paths, + * but the other possibilities considered in this routine aren't usable. + */ + if (PATH_PARAM_BY_REL(inner_cheapest_total, outerrel)) + inner_cheapest_total = NULL; + + /* + * If we need to unique-ify the inner path, we will consider only the + * cheapest-total inner. + */ + if (save_jointype == JOIN_UNIQUE_INNER) + { + /* No way to do this with an inner path parameterized by outer rel */ + if (inner_cheapest_total == NULL) + return; + inner_cheapest_total = (Path *) + create_unique_path(root, innerrel, inner_cheapest_total, extra->sjinfo); + Assert(inner_cheapest_total); + } + else if (nestjoinOK) + { + /* + * Consider materializing the cheapest inner path, unless + * enable_material is off or the path in question materializes its + * output anyway. + */ + if (enable_material && inner_cheapest_total != NULL && + !ExecMaterializesOutput(inner_cheapest_total->pathtype)) + matpath = (Path *) + create_material_path(innerrel, inner_cheapest_total); + } + + foreach(lc1, outerrel->pathlist) + { + Path *outerpath = (Path *) lfirst(lc1); + List *merge_pathkeys; + + /* + * We cannot use an outer path that is parameterized by the inner rel. + */ + if (PATH_PARAM_BY_REL(outerpath, innerrel)) + continue; + + /* + * If we need to unique-ify the outer path, it's pointless to consider + * any but the cheapest outer. (XXX we don't consider parameterized + * outers, nor inners, for unique-ified cases. Should we?) + */ + if (save_jointype == JOIN_UNIQUE_OUTER) + { + if (outerpath != outerrel->cheapest_total_path) + continue; + outerpath = (Path *) create_unique_path(root, outerrel, + outerpath, extra->sjinfo); + Assert(outerpath); + } + + /* + * The result will have this sort order (even if it is implemented as + * a nestloop, and even if some of the mergeclauses are implemented by + * qpquals rather than as true mergeclauses): + */ + merge_pathkeys = build_join_pathkeys(root, joinrel, jointype, + outerpath->pathkeys); + + if (save_jointype == JOIN_UNIQUE_INNER) + { + /* + * Consider nestloop join, but only with the unique-ified cheapest + * inner path + */ + try_nestloop_path(root, + joinrel, + outerpath, + inner_cheapest_total, + merge_pathkeys, + jointype, + extra); + } + else if (nestjoinOK) + { + /* + * Consider nestloop joins using this outer path and various + * available paths for the inner relation. We consider the + * cheapest-total paths for each available parameterization of the + * inner relation, including the unparameterized case. + */ + ListCell *lc2; + + foreach(lc2, innerrel->cheapest_parameterized_paths) + { + Path *innerpath = (Path *) lfirst(lc2); + Path *mpath; + + try_nestloop_path(root, + joinrel, + outerpath, + innerpath, + merge_pathkeys, + jointype, + extra); + + /* + * Try generating a memoize path and see if that makes the + * nested loop any cheaper. + */ + mpath = get_memoize_path(root, innerrel, outerrel, + innerpath, outerpath, jointype, + extra); + if (mpath != NULL) + try_nestloop_path(root, + joinrel, + outerpath, + mpath, + merge_pathkeys, + jointype, + extra); + } + + /* Also consider materialized form of the cheapest inner path */ + if (matpath != NULL) + try_nestloop_path(root, + joinrel, + outerpath, + matpath, + merge_pathkeys, + jointype, + extra); + } + + /* Can't do anything else if outer path needs to be unique'd */ + if (save_jointype == JOIN_UNIQUE_OUTER) + continue; + + /* Can't do anything else if inner rel is parameterized by outer */ + if (inner_cheapest_total == NULL) + continue; + + /* Generate merge join paths */ + generate_mergejoin_paths(root, joinrel, innerrel, outerpath, + save_jointype, extra, useallclauses, + inner_cheapest_total, merge_pathkeys, + false); + } + + /* + * Consider partial nestloop and mergejoin plan if outerrel has any + * partial path and the joinrel is parallel-safe. However, we can't + * handle JOIN_UNIQUE_OUTER, because the outer path will be partial, and + * therefore we won't be able to properly guarantee uniqueness. Nor can + * we handle joins needing lateral rels, since partial paths must not be + * parameterized. Similarly, we can't handle JOIN_FULL and JOIN_RIGHT, + * because they can produce false null extended rows. + */ + if (joinrel->consider_parallel && + save_jointype != JOIN_UNIQUE_OUTER && + save_jointype != JOIN_FULL && + save_jointype != JOIN_RIGHT && + outerrel->partial_pathlist != NIL && + bms_is_empty(joinrel->lateral_relids)) + { + if (nestjoinOK) + consider_parallel_nestloop(root, joinrel, outerrel, innerrel, + save_jointype, extra); + + /* + * If inner_cheapest_total is NULL or non parallel-safe then find the + * cheapest total parallel safe path. If doing JOIN_UNIQUE_INNER, we + * can't use any alternative inner path. + */ + if (inner_cheapest_total == NULL || + !inner_cheapest_total->parallel_safe) + { + if (save_jointype == JOIN_UNIQUE_INNER) + return; + + inner_cheapest_total = get_cheapest_parallel_safe_total_inner(innerrel->pathlist); + } + + if (inner_cheapest_total) + consider_parallel_mergejoin(root, joinrel, outerrel, innerrel, + save_jointype, extra, + inner_cheapest_total); + } +} + +/* + * consider_parallel_mergejoin + * Try to build partial paths for a joinrel by joining a partial path + * for the outer relation to a complete path for the inner relation. + * + * 'joinrel' is the join relation + * 'outerrel' is the outer join relation + * 'innerrel' is the inner join relation + * 'jointype' is the type of join to do + * 'extra' contains additional input values + * 'inner_cheapest_total' cheapest total path for innerrel + */ +static void +consider_parallel_mergejoin(PlannerInfo *root, + RelOptInfo *joinrel, + RelOptInfo *outerrel, + RelOptInfo *innerrel, + JoinType jointype, + JoinPathExtraData *extra, + Path *inner_cheapest_total) +{ + ListCell *lc1; + + /* generate merge join path for each partial outer path */ + foreach(lc1, outerrel->partial_pathlist) + { + Path *outerpath = (Path *) lfirst(lc1); + List *merge_pathkeys; + + /* + * Figure out what useful ordering any paths we create will have. + */ + merge_pathkeys = build_join_pathkeys(root, joinrel, jointype, + outerpath->pathkeys); + + generate_mergejoin_paths(root, joinrel, innerrel, outerpath, jointype, + extra, false, inner_cheapest_total, + merge_pathkeys, true); + } +} + +/* + * consider_parallel_nestloop + * Try to build partial paths for a joinrel by joining a partial path for the + * outer relation to a complete path for the inner relation. + * + * 'joinrel' is the join relation + * 'outerrel' is the outer join relation + * 'innerrel' is the inner join relation + * 'jointype' is the type of join to do + * 'extra' contains additional input values + */ +static void +consider_parallel_nestloop(PlannerInfo *root, + RelOptInfo *joinrel, + RelOptInfo *outerrel, + RelOptInfo *innerrel, + JoinType jointype, + JoinPathExtraData *extra) +{ + JoinType save_jointype = jointype; + ListCell *lc1; + + if (jointype == JOIN_UNIQUE_INNER) + jointype = JOIN_INNER; + + foreach(lc1, outerrel->partial_pathlist) + { + Path *outerpath = (Path *) lfirst(lc1); + List *pathkeys; + ListCell *lc2; + + /* Figure out what useful ordering any paths we create will have. */ + pathkeys = build_join_pathkeys(root, joinrel, jointype, + outerpath->pathkeys); + + /* + * Try the cheapest parameterized paths; only those which will produce + * an unparameterized path when joined to this outerrel will survive + * try_partial_nestloop_path. The cheapest unparameterized path is + * also in this list. + */ + foreach(lc2, innerrel->cheapest_parameterized_paths) + { + Path *innerpath = (Path *) lfirst(lc2); + Path *mpath; + + /* Can't join to an inner path that is not parallel-safe */ + if (!innerpath->parallel_safe) + continue; + + /* + * If we're doing JOIN_UNIQUE_INNER, we can only use the inner's + * cheapest_total_path, and we have to unique-ify it. (We might + * be able to relax this to allow other safe, unparameterized + * inner paths, but right now create_unique_path is not on board + * with that.) + */ + if (save_jointype == JOIN_UNIQUE_INNER) + { + if (innerpath != innerrel->cheapest_total_path) + continue; + innerpath = (Path *) create_unique_path(root, innerrel, + innerpath, + extra->sjinfo); + Assert(innerpath); + } + + try_partial_nestloop_path(root, joinrel, outerpath, innerpath, + pathkeys, jointype, extra); + + /* + * Try generating a memoize path and see if that makes the nested + * loop any cheaper. + */ + mpath = get_memoize_path(root, innerrel, outerrel, + innerpath, outerpath, jointype, + extra); + if (mpath != NULL) + try_partial_nestloop_path(root, joinrel, outerpath, mpath, + pathkeys, jointype, extra); + } + } +} + +/* + * hash_inner_and_outer + * Create hashjoin join paths by explicitly hashing both the outer and + * inner keys of each available hash clause. + * + * 'joinrel' is the join relation + * 'outerrel' is the outer join relation + * 'innerrel' is the inner join relation + * 'jointype' is the type of join to do + * 'extra' contains additional input values + */ +static void +hash_inner_and_outer(PlannerInfo *root, + RelOptInfo *joinrel, + RelOptInfo *outerrel, + RelOptInfo *innerrel, + JoinType jointype, + JoinPathExtraData *extra) +{ + JoinType save_jointype = jointype; + bool isouterjoin = IS_OUTER_JOIN(jointype); + List *hashclauses; + ListCell *l; + + /* + * We need to build only one hashclauses list for any given pair of outer + * and inner relations; all of the hashable clauses will be used as keys. + * + * Scan the join's restrictinfo list to find hashjoinable clauses that are + * usable with this pair of sub-relations. + */ + hashclauses = NIL; + foreach(l, extra->restrictlist) + { + RestrictInfo *restrictinfo = (RestrictInfo *) lfirst(l); + + /* + * If processing an outer join, only use its own join clauses for + * hashing. For inner joins we need not be so picky. + */ + if (isouterjoin && RINFO_IS_PUSHED_DOWN(restrictinfo, joinrel->relids)) + continue; + + if (!restrictinfo->can_join || + restrictinfo->hashjoinoperator == InvalidOid) + continue; /* not hashjoinable */ + + /* + * Check if clause has the form "outer op inner" or "inner op outer". + */ + if (!clause_sides_match_join(restrictinfo, outerrel, innerrel)) + continue; /* no good for these input relations */ + + hashclauses = lappend(hashclauses, restrictinfo); + } + + /* If we found any usable hashclauses, make paths */ + if (hashclauses) + { + /* + * We consider both the cheapest-total-cost and cheapest-startup-cost + * outer paths. There's no need to consider any but the + * cheapest-total-cost inner path, however. + */ + Path *cheapest_startup_outer = outerrel->cheapest_startup_path; + Path *cheapest_total_outer = outerrel->cheapest_total_path; + Path *cheapest_total_inner = innerrel->cheapest_total_path; + + /* + * If either cheapest-total path is parameterized by the other rel, we + * can't use a hashjoin. (There's no use looking for alternative + * input paths, since these should already be the least-parameterized + * available paths.) + */ + if (PATH_PARAM_BY_REL(cheapest_total_outer, innerrel) || + PATH_PARAM_BY_REL(cheapest_total_inner, outerrel)) + return; + + /* Unique-ify if need be; we ignore parameterized possibilities */ + if (jointype == JOIN_UNIQUE_OUTER) + { + cheapest_total_outer = (Path *) + create_unique_path(root, outerrel, + cheapest_total_outer, extra->sjinfo); + Assert(cheapest_total_outer); + jointype = JOIN_INNER; + try_hashjoin_path(root, + joinrel, + cheapest_total_outer, + cheapest_total_inner, + hashclauses, + jointype, + extra); + /* no possibility of cheap startup here */ + } + else if (jointype == JOIN_UNIQUE_INNER) + { + cheapest_total_inner = (Path *) + create_unique_path(root, innerrel, + cheapest_total_inner, extra->sjinfo); + Assert(cheapest_total_inner); + jointype = JOIN_INNER; + try_hashjoin_path(root, + joinrel, + cheapest_total_outer, + cheapest_total_inner, + hashclauses, + jointype, + extra); + if (cheapest_startup_outer != NULL && + cheapest_startup_outer != cheapest_total_outer) + try_hashjoin_path(root, + joinrel, + cheapest_startup_outer, + cheapest_total_inner, + hashclauses, + jointype, + extra); + } + else + { + /* + * For other jointypes, we consider the cheapest startup outer + * together with the cheapest total inner, and then consider + * pairings of cheapest-total paths including parameterized ones. + * There is no use in generating parameterized paths on the basis + * of possibly cheap startup cost, so this is sufficient. + */ + ListCell *lc1; + ListCell *lc2; + + if (cheapest_startup_outer != NULL) + try_hashjoin_path(root, + joinrel, + cheapest_startup_outer, + cheapest_total_inner, + hashclauses, + jointype, + extra); + + foreach(lc1, outerrel->cheapest_parameterized_paths) + { + Path *outerpath = (Path *) lfirst(lc1); + + /* + * We cannot use an outer path that is parameterized by the + * inner rel. + */ + if (PATH_PARAM_BY_REL(outerpath, innerrel)) + continue; + + foreach(lc2, innerrel->cheapest_parameterized_paths) + { + Path *innerpath = (Path *) lfirst(lc2); + + /* + * We cannot use an inner path that is parameterized by + * the outer rel, either. + */ + if (PATH_PARAM_BY_REL(innerpath, outerrel)) + continue; + + if (outerpath == cheapest_startup_outer && + innerpath == cheapest_total_inner) + continue; /* already tried it */ + + try_hashjoin_path(root, + joinrel, + outerpath, + innerpath, + hashclauses, + jointype, + extra); + } + } + } + + /* + * If the joinrel is parallel-safe, we may be able to consider a + * partial hash join. However, we can't handle JOIN_UNIQUE_OUTER, + * because the outer path will be partial, and therefore we won't be + * able to properly guarantee uniqueness. Similarly, we can't handle + * JOIN_FULL and JOIN_RIGHT, because they can produce false null + * extended rows. Also, the resulting path must not be parameterized. + * We would be able to support JOIN_FULL and JOIN_RIGHT for Parallel + * Hash, since in that case we're back to a single hash table with a + * single set of match bits for each batch, but that will require + * figuring out a deadlock-free way to wait for the probe to finish. + */ + if (joinrel->consider_parallel && + save_jointype != JOIN_UNIQUE_OUTER && + save_jointype != JOIN_FULL && + save_jointype != JOIN_RIGHT && + outerrel->partial_pathlist != NIL && + bms_is_empty(joinrel->lateral_relids)) + { + Path *cheapest_partial_outer; + Path *cheapest_partial_inner = NULL; + Path *cheapest_safe_inner = NULL; + + cheapest_partial_outer = + (Path *) linitial(outerrel->partial_pathlist); + + /* + * Can we use a partial inner plan too, so that we can build a + * shared hash table in parallel? We can't handle + * JOIN_UNIQUE_INNER because we can't guarantee uniqueness. + */ + if (innerrel->partial_pathlist != NIL && + save_jointype != JOIN_UNIQUE_INNER && + enable_parallel_hash) + { + cheapest_partial_inner = + (Path *) linitial(innerrel->partial_pathlist); + try_partial_hashjoin_path(root, joinrel, + cheapest_partial_outer, + cheapest_partial_inner, + hashclauses, jointype, extra, + true /* parallel_hash */ ); + } + + /* + * Normally, given that the joinrel is parallel-safe, the cheapest + * total inner path will also be parallel-safe, but if not, we'll + * have to search for the cheapest safe, unparameterized inner + * path. If doing JOIN_UNIQUE_INNER, we can't use any alternative + * inner path. + */ + if (cheapest_total_inner->parallel_safe) + cheapest_safe_inner = cheapest_total_inner; + else if (save_jointype != JOIN_UNIQUE_INNER) + cheapest_safe_inner = + get_cheapest_parallel_safe_total_inner(innerrel->pathlist); + + if (cheapest_safe_inner != NULL) + try_partial_hashjoin_path(root, joinrel, + cheapest_partial_outer, + cheapest_safe_inner, + hashclauses, jointype, extra, + false /* parallel_hash */ ); + } + } +} + +/* + * select_mergejoin_clauses + * Select mergejoin clauses that are usable for a particular join. + * Returns a list of RestrictInfo nodes for those clauses. + * + * *mergejoin_allowed is normally set to true, but it is set to false if + * this is a right/full join and there are nonmergejoinable join clauses. + * The executor's mergejoin machinery cannot handle such cases, so we have + * to avoid generating a mergejoin plan. (Note that this flag does NOT + * consider whether there are actually any mergejoinable clauses. This is + * correct because in some cases we need to build a clauseless mergejoin. + * Simply returning NIL is therefore not enough to distinguish safe from + * unsafe cases.) + * + * We also mark each selected RestrictInfo to show which side is currently + * being considered as outer. These are transient markings that are only + * good for the duration of the current add_paths_to_joinrel() call! + * + * We examine each restrictinfo clause known for the join to see + * if it is mergejoinable and involves vars from the two sub-relations + * currently of interest. + */ +static List * +select_mergejoin_clauses(PlannerInfo *root, + RelOptInfo *joinrel, + RelOptInfo *outerrel, + RelOptInfo *innerrel, + List *restrictlist, + JoinType jointype, + bool *mergejoin_allowed) +{ + List *result_list = NIL; + bool isouterjoin = IS_OUTER_JOIN(jointype); + bool have_nonmergeable_joinclause = false; + ListCell *l; + + foreach(l, restrictlist) + { + RestrictInfo *restrictinfo = (RestrictInfo *) lfirst(l); + + /* + * If processing an outer join, only use its own join clauses in the + * merge. For inner joins we can use pushed-down clauses too. (Note: + * we don't set have_nonmergeable_joinclause here because pushed-down + * clauses will become otherquals not joinquals.) + */ + if (isouterjoin && RINFO_IS_PUSHED_DOWN(restrictinfo, joinrel->relids)) + continue; + + /* Check that clause is a mergeable operator clause */ + if (!restrictinfo->can_join || + restrictinfo->mergeopfamilies == NIL) + { + /* + * The executor can handle extra joinquals that are constants, but + * not anything else, when doing right/full merge join. (The + * reason to support constants is so we can do FULL JOIN ON + * FALSE.) + */ + if (!restrictinfo->clause || !IsA(restrictinfo->clause, Const)) + have_nonmergeable_joinclause = true; + continue; /* not mergejoinable */ + } + + /* + * Check if clause has the form "outer op inner" or "inner op outer". + */ + if (!clause_sides_match_join(restrictinfo, outerrel, innerrel)) + { + have_nonmergeable_joinclause = true; + continue; /* no good for these input relations */ + } + + /* + * Insist that each side have a non-redundant eclass. This + * restriction is needed because various bits of the planner expect + * that each clause in a merge be associable with some pathkey in a + * canonical pathkey list, but redundant eclasses can't appear in + * canonical sort orderings. (XXX it might be worth relaxing this, + * but not enough time to address it for 8.3.) + * + * Note: it would be bad if this condition failed for an otherwise + * mergejoinable FULL JOIN clause, since that would result in + * undesirable planner failure. I believe that is not possible + * however; a variable involved in a full join could only appear in + * below_outer_join eclasses, which aren't considered redundant. + * + * This case *can* happen for left/right join clauses: the outer-side + * variable could be equated to a constant. Because we will propagate + * that constant across the join clause, the loss of ability to do a + * mergejoin is not really all that big a deal, and so it's not clear + * that improving this is important. + */ + update_mergeclause_eclasses(root, restrictinfo); + + if (EC_MUST_BE_REDUNDANT(restrictinfo->left_ec) || + EC_MUST_BE_REDUNDANT(restrictinfo->right_ec)) + { + have_nonmergeable_joinclause = true; + continue; /* can't handle redundant eclasses */ + } + + result_list = lappend(result_list, restrictinfo); + } + + /* + * Report whether mergejoin is allowed (see comment at top of function). + */ + switch (jointype) + { + case JOIN_RIGHT: + case JOIN_FULL: + *mergejoin_allowed = !have_nonmergeable_joinclause; + break; + default: + *mergejoin_allowed = true; + break; + } + + return result_list; +} diff --git a/src/backend/optimizer/path/joinrels.c b/src/backend/optimizer/path/joinrels.c new file mode 100644 index 0000000..9da3ff2 --- /dev/null +++ b/src/backend/optimizer/path/joinrels.c @@ -0,0 +1,1783 @@ +/*------------------------------------------------------------------------- + * + * joinrels.c + * Routines to determine which relations should be joined + * + * Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group + * Portions Copyright (c) 1994, Regents of the University of California + * + * + * IDENTIFICATION + * src/backend/optimizer/path/joinrels.c + * + *------------------------------------------------------------------------- + */ +#include "postgres.h" + +#include "miscadmin.h" +#include "optimizer/appendinfo.h" +#include "optimizer/joininfo.h" +#include "optimizer/pathnode.h" +#include "optimizer/paths.h" +#include "partitioning/partbounds.h" +#include "utils/memutils.h" + + +static void make_rels_by_clause_joins(PlannerInfo *root, + RelOptInfo *old_rel, + List *other_rels_list, + ListCell *other_rels); +static void make_rels_by_clauseless_joins(PlannerInfo *root, + RelOptInfo *old_rel, + List *other_rels); +static bool has_join_restriction(PlannerInfo *root, RelOptInfo *rel); +static bool has_legal_joinclause(PlannerInfo *root, RelOptInfo *rel); +static bool restriction_is_constant_false(List *restrictlist, + RelOptInfo *joinrel, + bool only_pushed_down); +static void populate_joinrel_with_paths(PlannerInfo *root, RelOptInfo *rel1, + RelOptInfo *rel2, RelOptInfo *joinrel, + SpecialJoinInfo *sjinfo, List *restrictlist); +static void try_partitionwise_join(PlannerInfo *root, RelOptInfo *rel1, + RelOptInfo *rel2, RelOptInfo *joinrel, + SpecialJoinInfo *parent_sjinfo, + List *parent_restrictlist); +static SpecialJoinInfo *build_child_join_sjinfo(PlannerInfo *root, + SpecialJoinInfo *parent_sjinfo, + Relids left_relids, Relids right_relids); +static void compute_partition_bounds(PlannerInfo *root, RelOptInfo *rel1, + RelOptInfo *rel2, RelOptInfo *joinrel, + SpecialJoinInfo *parent_sjinfo, + List **parts1, List **parts2); +static void get_matching_part_pairs(PlannerInfo *root, RelOptInfo *joinrel, + RelOptInfo *rel1, RelOptInfo *rel2, + List **parts1, List **parts2); + + +/* + * join_search_one_level + * Consider ways to produce join relations containing exactly 'level' + * jointree items. (This is one step of the dynamic-programming method + * embodied in standard_join_search.) Join rel nodes for each feasible + * combination of lower-level rels are created and returned in a list. + * Implementation paths are created for each such joinrel, too. + * + * level: level of rels we want to make this time + * root->join_rel_level[j], 1 <= j < level, is a list of rels containing j items + * + * The result is returned in root->join_rel_level[level]. + */ +void +join_search_one_level(PlannerInfo *root, int level) +{ + List **joinrels = root->join_rel_level; + ListCell *r; + int k; + + Assert(joinrels[level] == NIL); + + /* Set join_cur_level so that new joinrels are added to proper list */ + root->join_cur_level = level; + + /* + * First, consider left-sided and right-sided plans, in which rels of + * exactly level-1 member relations are joined against initial relations. + * We prefer to join using join clauses, but if we find a rel of level-1 + * members that has no join clauses, we will generate Cartesian-product + * joins against all initial rels not already contained in it. + */ + foreach(r, joinrels[level - 1]) + { + RelOptInfo *old_rel = (RelOptInfo *) lfirst(r); + + if (old_rel->joininfo != NIL || old_rel->has_eclass_joins || + has_join_restriction(root, old_rel)) + { + /* + * There are join clauses or join order restrictions relevant to + * this rel, so consider joins between this rel and (only) those + * initial rels it is linked to by a clause or restriction. + * + * At level 2 this condition is symmetric, so there is no need to + * look at initial rels before this one in the list; we already + * considered such joins when we were at the earlier rel. (The + * mirror-image joins are handled automatically by make_join_rel.) + * In later passes (level > 2), we join rels of the previous level + * to each initial rel they don't already include but have a join + * clause or restriction with. + */ + List *other_rels_list; + ListCell *other_rels; + + if (level == 2) /* consider remaining initial rels */ + { + other_rels_list = joinrels[level - 1]; + other_rels = lnext(other_rels_list, r); + } + else /* consider all initial rels */ + { + other_rels_list = joinrels[1]; + other_rels = list_head(other_rels_list); + } + + make_rels_by_clause_joins(root, + old_rel, + other_rels_list, + other_rels); + } + else + { + /* + * Oops, we have a relation that is not joined to any other + * relation, either directly or by join-order restrictions. + * Cartesian product time. + * + * We consider a cartesian product with each not-already-included + * initial rel, whether it has other join clauses or not. At + * level 2, if there are two or more clauseless initial rels, we + * will redundantly consider joining them in both directions; but + * such cases aren't common enough to justify adding complexity to + * avoid the duplicated effort. + */ + make_rels_by_clauseless_joins(root, + old_rel, + joinrels[1]); + } + } + + /* + * Now, consider "bushy plans" in which relations of k initial rels are + * joined to relations of level-k initial rels, for 2 <= k <= level-2. + * + * We only consider bushy-plan joins for pairs of rels where there is a + * suitable join clause (or join order restriction), in order to avoid + * unreasonable growth of planning time. + */ + for (k = 2;; k++) + { + int other_level = level - k; + + /* + * Since make_join_rel(x, y) handles both x,y and y,x cases, we only + * need to go as far as the halfway point. + */ + if (k > other_level) + break; + + foreach(r, joinrels[k]) + { + RelOptInfo *old_rel = (RelOptInfo *) lfirst(r); + List *other_rels_list; + ListCell *other_rels; + ListCell *r2; + + /* + * We can ignore relations without join clauses here, unless they + * participate in join-order restrictions --- then we might have + * to force a bushy join plan. + */ + if (old_rel->joininfo == NIL && !old_rel->has_eclass_joins && + !has_join_restriction(root, old_rel)) + continue; + + if (k == other_level) + { + /* only consider remaining rels */ + other_rels_list = joinrels[k]; + other_rels = lnext(other_rels_list, r); + } + else + { + other_rels_list = joinrels[other_level]; + other_rels = list_head(other_rels_list); + } + + for_each_cell(r2, other_rels_list, other_rels) + { + RelOptInfo *new_rel = (RelOptInfo *) lfirst(r2); + + if (!bms_overlap(old_rel->relids, new_rel->relids)) + { + /* + * OK, we can build a rel of the right level from this + * pair of rels. Do so if there is at least one relevant + * join clause or join order restriction. + */ + if (have_relevant_joinclause(root, old_rel, new_rel) || + have_join_order_restriction(root, old_rel, new_rel)) + { + (void) make_join_rel(root, old_rel, new_rel); + } + } + } + } + } + + /*---------- + * Last-ditch effort: if we failed to find any usable joins so far, force + * a set of cartesian-product joins to be generated. This handles the + * special case where all the available rels have join clauses but we + * cannot use any of those clauses yet. This can only happen when we are + * considering a join sub-problem (a sub-joinlist) and all the rels in the + * sub-problem have only join clauses with rels outside the sub-problem. + * An example is + * + * SELECT ... FROM a INNER JOIN b ON TRUE, c, d, ... + * WHERE a.w = c.x and b.y = d.z; + * + * If the "a INNER JOIN b" sub-problem does not get flattened into the + * upper level, we must be willing to make a cartesian join of a and b; + * but the code above will not have done so, because it thought that both + * a and b have joinclauses. We consider only left-sided and right-sided + * cartesian joins in this case (no bushy). + *---------- + */ + if (joinrels[level] == NIL) + { + /* + * This loop is just like the first one, except we always call + * make_rels_by_clauseless_joins(). + */ + foreach(r, joinrels[level - 1]) + { + RelOptInfo *old_rel = (RelOptInfo *) lfirst(r); + + make_rels_by_clauseless_joins(root, + old_rel, + joinrels[1]); + } + + /*---------- + * When special joins are involved, there may be no legal way + * to make an N-way join for some values of N. For example consider + * + * SELECT ... FROM t1 WHERE + * x IN (SELECT ... FROM t2,t3 WHERE ...) AND + * y IN (SELECT ... FROM t4,t5 WHERE ...) + * + * We will flatten this query to a 5-way join problem, but there are + * no 4-way joins that join_is_legal() will consider legal. We have + * to accept failure at level 4 and go on to discover a workable + * bushy plan at level 5. + * + * However, if there are no special joins and no lateral references + * then join_is_legal() should never fail, and so the following sanity + * check is useful. + *---------- + */ + if (joinrels[level] == NIL && + root->join_info_list == NIL && + !root->hasLateralRTEs) + elog(ERROR, "failed to build any %d-way joins", level); + } +} + +/* + * make_rels_by_clause_joins + * Build joins between the given relation 'old_rel' and other relations + * that participate in join clauses that 'old_rel' also participates in + * (or participate in join-order restrictions with it). + * The join rels are returned in root->join_rel_level[join_cur_level]. + * + * Note: at levels above 2 we will generate the same joined relation in + * multiple ways --- for example (a join b) join c is the same RelOptInfo as + * (b join c) join a, though the second case will add a different set of Paths + * to it. This is the reason for using the join_rel_level mechanism, which + * automatically ensures that each new joinrel is only added to the list once. + * + * 'old_rel' is the relation entry for the relation to be joined + * 'other_rels_list': a list containing the other + * rels to be considered for joining + * 'other_rels': the first cell to be considered + * + * Currently, this is only used with initial rels in other_rels, but it + * will work for joining to joinrels too. + */ +static void +make_rels_by_clause_joins(PlannerInfo *root, + RelOptInfo *old_rel, + List *other_rels_list, + ListCell *other_rels) +{ + ListCell *l; + + for_each_cell(l, other_rels_list, other_rels) + { + RelOptInfo *other_rel = (RelOptInfo *) lfirst(l); + + if (!bms_overlap(old_rel->relids, other_rel->relids) && + (have_relevant_joinclause(root, old_rel, other_rel) || + have_join_order_restriction(root, old_rel, other_rel))) + { + (void) make_join_rel(root, old_rel, other_rel); + } + } +} + +/* + * make_rels_by_clauseless_joins + * Given a relation 'old_rel' and a list of other relations + * 'other_rels', create a join relation between 'old_rel' and each + * member of 'other_rels' that isn't already included in 'old_rel'. + * The join rels are returned in root->join_rel_level[join_cur_level]. + * + * 'old_rel' is the relation entry for the relation to be joined + * 'other_rels': a list containing the other rels to be considered for joining + * + * Currently, this is only used with initial rels in other_rels, but it would + * work for joining to joinrels too. + */ +static void +make_rels_by_clauseless_joins(PlannerInfo *root, + RelOptInfo *old_rel, + List *other_rels) +{ + ListCell *l; + + foreach(l, other_rels) + { + RelOptInfo *other_rel = (RelOptInfo *) lfirst(l); + + if (!bms_overlap(other_rel->relids, old_rel->relids)) + { + (void) make_join_rel(root, old_rel, other_rel); + } + } +} + + +/* + * join_is_legal + * Determine whether a proposed join is legal given the query's + * join order constraints; and if it is, determine the join type. + * + * Caller must supply not only the two rels, but the union of their relids. + * (We could simplify the API by computing joinrelids locally, but this + * would be redundant work in the normal path through make_join_rel.) + * + * On success, *sjinfo_p is set to NULL if this is to be a plain inner join, + * else it's set to point to the associated SpecialJoinInfo node. Also, + * *reversed_p is set true if the given relations need to be swapped to + * match the SpecialJoinInfo node. + */ +static bool +join_is_legal(PlannerInfo *root, RelOptInfo *rel1, RelOptInfo *rel2, + Relids joinrelids, + SpecialJoinInfo **sjinfo_p, bool *reversed_p) +{ + SpecialJoinInfo *match_sjinfo; + bool reversed; + bool unique_ified; + bool must_be_leftjoin; + ListCell *l; + + /* + * Ensure output params are set on failure return. This is just to + * suppress uninitialized-variable warnings from overly anal compilers. + */ + *sjinfo_p = NULL; + *reversed_p = false; + + /* + * If we have any special joins, the proposed join might be illegal; and + * in any case we have to determine its join type. Scan the join info + * list for matches and conflicts. + */ + match_sjinfo = NULL; + reversed = false; + unique_ified = false; + must_be_leftjoin = false; + + foreach(l, root->join_info_list) + { + SpecialJoinInfo *sjinfo = (SpecialJoinInfo *) lfirst(l); + + /* + * This special join is not relevant unless its RHS overlaps the + * proposed join. (Check this first as a fast path for dismissing + * most irrelevant SJs quickly.) + */ + if (!bms_overlap(sjinfo->min_righthand, joinrelids)) + continue; + + /* + * Also, not relevant if proposed join is fully contained within RHS + * (ie, we're still building up the RHS). + */ + if (bms_is_subset(joinrelids, sjinfo->min_righthand)) + continue; + + /* + * Also, not relevant if SJ is already done within either input. + */ + if (bms_is_subset(sjinfo->min_lefthand, rel1->relids) && + bms_is_subset(sjinfo->min_righthand, rel1->relids)) + continue; + if (bms_is_subset(sjinfo->min_lefthand, rel2->relids) && + bms_is_subset(sjinfo->min_righthand, rel2->relids)) + continue; + + /* + * If it's a semijoin and we already joined the RHS to any other rels + * within either input, then we must have unique-ified the RHS at that + * point (see below). Therefore the semijoin is no longer relevant in + * this join path. + */ + if (sjinfo->jointype == JOIN_SEMI) + { + if (bms_is_subset(sjinfo->syn_righthand, rel1->relids) && + !bms_equal(sjinfo->syn_righthand, rel1->relids)) + continue; + if (bms_is_subset(sjinfo->syn_righthand, rel2->relids) && + !bms_equal(sjinfo->syn_righthand, rel2->relids)) + continue; + } + + /* + * If one input contains min_lefthand and the other contains + * min_righthand, then we can perform the SJ at this join. + * + * Reject if we get matches to more than one SJ; that implies we're + * considering something that's not really valid. + */ + if (bms_is_subset(sjinfo->min_lefthand, rel1->relids) && + bms_is_subset(sjinfo->min_righthand, rel2->relids)) + { + if (match_sjinfo) + return false; /* invalid join path */ + match_sjinfo = sjinfo; + reversed = false; + } + else if (bms_is_subset(sjinfo->min_lefthand, rel2->relids) && + bms_is_subset(sjinfo->min_righthand, rel1->relids)) + { + if (match_sjinfo) + return false; /* invalid join path */ + match_sjinfo = sjinfo; + reversed = true; + } + else if (sjinfo->jointype == JOIN_SEMI && + bms_equal(sjinfo->syn_righthand, rel2->relids) && + create_unique_path(root, rel2, rel2->cheapest_total_path, + sjinfo) != NULL) + { + /*---------- + * For a semijoin, we can join the RHS to anything else by + * unique-ifying the RHS (if the RHS can be unique-ified). + * We will only get here if we have the full RHS but less + * than min_lefthand on the LHS. + * + * The reason to consider such a join path is exemplified by + * SELECT ... FROM a,b WHERE (a.x,b.y) IN (SELECT c1,c2 FROM c) + * If we insist on doing this as a semijoin we will first have + * to form the cartesian product of A*B. But if we unique-ify + * C then the semijoin becomes a plain innerjoin and we can join + * in any order, eg C to A and then to B. When C is much smaller + * than A and B this can be a huge win. So we allow C to be + * joined to just A or just B here, and then make_join_rel has + * to handle the case properly. + * + * Note that actually we'll allow unique-ified C to be joined to + * some other relation D here, too. That is legal, if usually not + * very sane, and this routine is only concerned with legality not + * with whether the join is good strategy. + *---------- + */ + if (match_sjinfo) + return false; /* invalid join path */ + match_sjinfo = sjinfo; + reversed = false; + unique_ified = true; + } + else if (sjinfo->jointype == JOIN_SEMI && + bms_equal(sjinfo->syn_righthand, rel1->relids) && + create_unique_path(root, rel1, rel1->cheapest_total_path, + sjinfo) != NULL) + { + /* Reversed semijoin case */ + if (match_sjinfo) + return false; /* invalid join path */ + match_sjinfo = sjinfo; + reversed = true; + unique_ified = true; + } + else + { + /* + * Otherwise, the proposed join overlaps the RHS but isn't a valid + * implementation of this SJ. But don't panic quite yet: the RHS + * violation might have occurred previously, in one or both input + * relations, in which case we must have previously decided that + * it was OK to commute some other SJ with this one. If we need + * to perform this join to finish building up the RHS, rejecting + * it could lead to not finding any plan at all. (This can occur + * because of the heuristics elsewhere in this file that postpone + * clauseless joins: we might not consider doing a clauseless join + * within the RHS until after we've performed other, validly + * commutable SJs with one or both sides of the clauseless join.) + * This consideration boils down to the rule that if both inputs + * overlap the RHS, we can allow the join --- they are either + * fully within the RHS, or represent previously-allowed joins to + * rels outside it. + */ + if (bms_overlap(rel1->relids, sjinfo->min_righthand) && + bms_overlap(rel2->relids, sjinfo->min_righthand)) + continue; /* assume valid previous violation of RHS */ + + /* + * The proposed join could still be legal, but only if we're + * allowed to associate it into the RHS of this SJ. That means + * this SJ must be a LEFT join (not SEMI or ANTI, and certainly + * not FULL) and the proposed join must not overlap the LHS. + */ + if (sjinfo->jointype != JOIN_LEFT || + bms_overlap(joinrelids, sjinfo->min_lefthand)) + return false; /* invalid join path */ + + /* + * To be valid, the proposed join must be a LEFT join; otherwise + * it can't associate into this SJ's RHS. But we may not yet have + * found the SpecialJoinInfo matching the proposed join, so we + * can't test that yet. Remember the requirement for later. + */ + must_be_leftjoin = true; + } + } + + /* + * Fail if violated any SJ's RHS and didn't match to a LEFT SJ: the + * proposed join can't associate into an SJ's RHS. + * + * Also, fail if the proposed join's predicate isn't strict; we're + * essentially checking to see if we can apply outer-join identity 3, and + * that's a requirement. (This check may be redundant with checks in + * make_outerjoininfo, but I'm not quite sure, and it's cheap to test.) + */ + if (must_be_leftjoin && + (match_sjinfo == NULL || + match_sjinfo->jointype != JOIN_LEFT || + !match_sjinfo->lhs_strict)) + return false; /* invalid join path */ + + /* + * We also have to check for constraints imposed by LATERAL references. + */ + if (root->hasLateralRTEs) + { + bool lateral_fwd; + bool lateral_rev; + Relids join_lateral_rels; + + /* + * The proposed rels could each contain lateral references to the + * other, in which case the join is impossible. If there are lateral + * references in just one direction, then the join has to be done with + * a nestloop with the lateral referencer on the inside. If the join + * matches an SJ that cannot be implemented by such a nestloop, the + * join is impossible. + * + * Also, if the lateral reference is only indirect, we should reject + * the join; whatever rel(s) the reference chain goes through must be + * joined to first. + * + * Another case that might keep us from building a valid plan is the + * implementation restriction described by have_dangerous_phv(). + */ + lateral_fwd = bms_overlap(rel1->relids, rel2->lateral_relids); + lateral_rev = bms_overlap(rel2->relids, rel1->lateral_relids); + if (lateral_fwd && lateral_rev) + return false; /* have lateral refs in both directions */ + if (lateral_fwd) + { + /* has to be implemented as nestloop with rel1 on left */ + if (match_sjinfo && + (reversed || + unique_ified || + match_sjinfo->jointype == JOIN_FULL)) + return false; /* not implementable as nestloop */ + /* check there is a direct reference from rel2 to rel1 */ + if (!bms_overlap(rel1->relids, rel2->direct_lateral_relids)) + return false; /* only indirect refs, so reject */ + /* check we won't have a dangerous PHV */ + if (have_dangerous_phv(root, rel1->relids, rel2->lateral_relids)) + return false; /* might be unable to handle required PHV */ + } + else if (lateral_rev) + { + /* has to be implemented as nestloop with rel2 on left */ + if (match_sjinfo && + (!reversed || + unique_ified || + match_sjinfo->jointype == JOIN_FULL)) + return false; /* not implementable as nestloop */ + /* check there is a direct reference from rel1 to rel2 */ + if (!bms_overlap(rel2->relids, rel1->direct_lateral_relids)) + return false; /* only indirect refs, so reject */ + /* check we won't have a dangerous PHV */ + if (have_dangerous_phv(root, rel2->relids, rel1->lateral_relids)) + return false; /* might be unable to handle required PHV */ + } + + /* + * LATERAL references could also cause problems later on if we accept + * this join: if the join's minimum parameterization includes any rels + * that would have to be on the inside of an outer join with this join + * rel, then it's never going to be possible to build the complete + * query using this join. We should reject this join not only because + * it'll save work, but because if we don't, the clauseless-join + * heuristics might think that legality of this join means that some + * other join rel need not be formed, and that could lead to failure + * to find any plan at all. We have to consider not only rels that + * are directly on the inner side of an OJ with the joinrel, but also + * ones that are indirectly so, so search to find all such rels. + */ + join_lateral_rels = min_join_parameterization(root, joinrelids, + rel1, rel2); + if (join_lateral_rels) + { + Relids join_plus_rhs = bms_copy(joinrelids); + bool more; + + do + { + more = false; + foreach(l, root->join_info_list) + { + SpecialJoinInfo *sjinfo = (SpecialJoinInfo *) lfirst(l); + + /* ignore full joins --- their ordering is predetermined */ + if (sjinfo->jointype == JOIN_FULL) + continue; + + if (bms_overlap(sjinfo->min_lefthand, join_plus_rhs) && + !bms_is_subset(sjinfo->min_righthand, join_plus_rhs)) + { + join_plus_rhs = bms_add_members(join_plus_rhs, + sjinfo->min_righthand); + more = true; + } + } + } while (more); + if (bms_overlap(join_plus_rhs, join_lateral_rels)) + return false; /* will not be able to join to some RHS rel */ + } + } + + /* Otherwise, it's a valid join */ + *sjinfo_p = match_sjinfo; + *reversed_p = reversed; + return true; +} + + +/* + * make_join_rel + * Find or create a join RelOptInfo that represents the join of + * the two given rels, and add to it path information for paths + * created with the two rels as outer and inner rel. + * (The join rel may already contain paths generated from other + * pairs of rels that add up to the same set of base rels.) + * + * NB: will return NULL if attempted join is not valid. This can happen + * when working with outer joins, or with IN or EXISTS clauses that have been + * turned into joins. + */ +RelOptInfo * +make_join_rel(PlannerInfo *root, RelOptInfo *rel1, RelOptInfo *rel2) +{ + Relids joinrelids; + SpecialJoinInfo *sjinfo; + bool reversed; + SpecialJoinInfo sjinfo_data; + RelOptInfo *joinrel; + List *restrictlist; + + /* We should never try to join two overlapping sets of rels. */ + Assert(!bms_overlap(rel1->relids, rel2->relids)); + + /* Construct Relids set that identifies the joinrel. */ + joinrelids = bms_union(rel1->relids, rel2->relids); + + /* Check validity and determine join type. */ + if (!join_is_legal(root, rel1, rel2, joinrelids, + &sjinfo, &reversed)) + { + /* invalid join path */ + bms_free(joinrelids); + return NULL; + } + + /* Swap rels if needed to match the join info. */ + if (reversed) + { + RelOptInfo *trel = rel1; + + rel1 = rel2; + rel2 = trel; + } + + /* + * If it's a plain inner join, then we won't have found anything in + * join_info_list. Make up a SpecialJoinInfo so that selectivity + * estimation functions will know what's being joined. + */ + if (sjinfo == NULL) + { + sjinfo = &sjinfo_data; + sjinfo->type = T_SpecialJoinInfo; + sjinfo->min_lefthand = rel1->relids; + sjinfo->min_righthand = rel2->relids; + sjinfo->syn_lefthand = rel1->relids; + sjinfo->syn_righthand = rel2->relids; + sjinfo->jointype = JOIN_INNER; + /* we don't bother trying to make the remaining fields valid */ + sjinfo->lhs_strict = false; + sjinfo->delay_upper_joins = false; + sjinfo->semi_can_btree = false; + sjinfo->semi_can_hash = false; + sjinfo->semi_operators = NIL; + sjinfo->semi_rhs_exprs = NIL; + } + + /* + * Find or build the join RelOptInfo, and compute the restrictlist that + * goes with this particular joining. + */ + joinrel = build_join_rel(root, joinrelids, rel1, rel2, sjinfo, + &restrictlist); + + /* + * If we've already proven this join is empty, we needn't consider any + * more paths for it. + */ + if (is_dummy_rel(joinrel)) + { + bms_free(joinrelids); + return joinrel; + } + + /* Add paths to the join relation. */ + populate_joinrel_with_paths(root, rel1, rel2, joinrel, sjinfo, + restrictlist); + + bms_free(joinrelids); + + return joinrel; +} + +/* + * populate_joinrel_with_paths + * Add paths to the given joinrel for given pair of joining relations. The + * SpecialJoinInfo provides details about the join and the restrictlist + * contains the join clauses and the other clauses applicable for given pair + * of the joining relations. + */ +static void +populate_joinrel_with_paths(PlannerInfo *root, RelOptInfo *rel1, + RelOptInfo *rel2, RelOptInfo *joinrel, + SpecialJoinInfo *sjinfo, List *restrictlist) +{ + /* + * Consider paths using each rel as both outer and inner. Depending on + * the join type, a provably empty outer or inner rel might mean the join + * is provably empty too; in which case throw away any previously computed + * paths and mark the join as dummy. (We do it this way since it's + * conceivable that dummy-ness of a multi-element join might only be + * noticeable for certain construction paths.) + * + * Also, a provably constant-false join restriction typically means that + * we can skip evaluating one or both sides of the join. We do this by + * marking the appropriate rel as dummy. For outer joins, a + * constant-false restriction that is pushed down still means the whole + * join is dummy, while a non-pushed-down one means that no inner rows + * will join so we can treat the inner rel as dummy. + * + * We need only consider the jointypes that appear in join_info_list, plus + * JOIN_INNER. + */ + switch (sjinfo->jointype) + { + case JOIN_INNER: + if (is_dummy_rel(rel1) || is_dummy_rel(rel2) || + restriction_is_constant_false(restrictlist, joinrel, false)) + { + mark_dummy_rel(joinrel); + break; + } + add_paths_to_joinrel(root, joinrel, rel1, rel2, + JOIN_INNER, sjinfo, + restrictlist); + add_paths_to_joinrel(root, joinrel, rel2, rel1, + JOIN_INNER, sjinfo, + restrictlist); + break; + case JOIN_LEFT: + if (is_dummy_rel(rel1) || + restriction_is_constant_false(restrictlist, joinrel, true)) + { + mark_dummy_rel(joinrel); + break; + } + if (restriction_is_constant_false(restrictlist, joinrel, false) && + bms_is_subset(rel2->relids, sjinfo->syn_righthand)) + mark_dummy_rel(rel2); + add_paths_to_joinrel(root, joinrel, rel1, rel2, + JOIN_LEFT, sjinfo, + restrictlist); + add_paths_to_joinrel(root, joinrel, rel2, rel1, + JOIN_RIGHT, sjinfo, + restrictlist); + break; + case JOIN_FULL: + if ((is_dummy_rel(rel1) && is_dummy_rel(rel2)) || + restriction_is_constant_false(restrictlist, joinrel, true)) + { + mark_dummy_rel(joinrel); + break; + } + add_paths_to_joinrel(root, joinrel, rel1, rel2, + JOIN_FULL, sjinfo, + restrictlist); + add_paths_to_joinrel(root, joinrel, rel2, rel1, + JOIN_FULL, sjinfo, + restrictlist); + + /* + * If there are join quals that aren't mergeable or hashable, we + * may not be able to build any valid plan. Complain here so that + * we can give a somewhat-useful error message. (Since we have no + * flexibility of planning for a full join, there's no chance of + * succeeding later with another pair of input rels.) + */ + if (joinrel->pathlist == NIL) + ereport(ERROR, + (errcode(ERRCODE_FEATURE_NOT_SUPPORTED), + errmsg("FULL JOIN is only supported with merge-joinable or hash-joinable join conditions"))); + break; + case JOIN_SEMI: + + /* + * We might have a normal semijoin, or a case where we don't have + * enough rels to do the semijoin but can unique-ify the RHS and + * then do an innerjoin (see comments in join_is_legal). In the + * latter case we can't apply JOIN_SEMI joining. + */ + if (bms_is_subset(sjinfo->min_lefthand, rel1->relids) && + bms_is_subset(sjinfo->min_righthand, rel2->relids)) + { + if (is_dummy_rel(rel1) || is_dummy_rel(rel2) || + restriction_is_constant_false(restrictlist, joinrel, false)) + { + mark_dummy_rel(joinrel); + break; + } + add_paths_to_joinrel(root, joinrel, rel1, rel2, + JOIN_SEMI, sjinfo, + restrictlist); + } + + /* + * If we know how to unique-ify the RHS and one input rel is + * exactly the RHS (not a superset) we can consider unique-ifying + * it and then doing a regular join. (The create_unique_path + * check here is probably redundant with what join_is_legal did, + * but if so the check is cheap because it's cached. So test + * anyway to be sure.) + */ + if (bms_equal(sjinfo->syn_righthand, rel2->relids) && + create_unique_path(root, rel2, rel2->cheapest_total_path, + sjinfo) != NULL) + { + if (is_dummy_rel(rel1) || is_dummy_rel(rel2) || + restriction_is_constant_false(restrictlist, joinrel, false)) + { + mark_dummy_rel(joinrel); + break; + } + add_paths_to_joinrel(root, joinrel, rel1, rel2, + JOIN_UNIQUE_INNER, sjinfo, + restrictlist); + add_paths_to_joinrel(root, joinrel, rel2, rel1, + JOIN_UNIQUE_OUTER, sjinfo, + restrictlist); + } + break; + case JOIN_ANTI: + if (is_dummy_rel(rel1) || + restriction_is_constant_false(restrictlist, joinrel, true)) + { + mark_dummy_rel(joinrel); + break; + } + if (restriction_is_constant_false(restrictlist, joinrel, false) && + bms_is_subset(rel2->relids, sjinfo->syn_righthand)) + mark_dummy_rel(rel2); + add_paths_to_joinrel(root, joinrel, rel1, rel2, + JOIN_ANTI, sjinfo, + restrictlist); + break; + default: + /* other values not expected here */ + elog(ERROR, "unrecognized join type: %d", (int) sjinfo->jointype); + break; + } + + /* Apply partitionwise join technique, if possible. */ + try_partitionwise_join(root, rel1, rel2, joinrel, sjinfo, restrictlist); +} + + +/* + * have_join_order_restriction + * Detect whether the two relations should be joined to satisfy + * a join-order restriction arising from special or lateral joins. + * + * In practice this is always used with have_relevant_joinclause(), and so + * could be merged with that function, but it seems clearer to separate the + * two concerns. We need this test because there are degenerate cases where + * a clauseless join must be performed to satisfy join-order restrictions. + * Also, if one rel has a lateral reference to the other, or both are needed + * to compute some PHV, we should consider joining them even if the join would + * be clauseless. + * + * Note: this is only a problem if one side of a degenerate outer join + * contains multiple rels, or a clauseless join is required within an + * IN/EXISTS RHS; else we will find a join path via the "last ditch" case in + * join_search_one_level(). We could dispense with this test if we were + * willing to try bushy plans in the "last ditch" case, but that seems much + * less efficient. + */ +bool +have_join_order_restriction(PlannerInfo *root, + RelOptInfo *rel1, RelOptInfo *rel2) +{ + bool result = false; + ListCell *l; + + /* + * If either side has a direct lateral reference to the other, attempt the + * join regardless of outer-join considerations. + */ + if (bms_overlap(rel1->relids, rel2->direct_lateral_relids) || + bms_overlap(rel2->relids, rel1->direct_lateral_relids)) + return true; + + /* + * Likewise, if both rels are needed to compute some PlaceHolderVar, + * attempt the join regardless of outer-join considerations. (This is not + * very desirable, because a PHV with a large eval_at set will cause a lot + * of probably-useless joins to be considered, but failing to do this can + * cause us to fail to construct a plan at all.) + */ + foreach(l, root->placeholder_list) + { + PlaceHolderInfo *phinfo = (PlaceHolderInfo *) lfirst(l); + + if (bms_is_subset(rel1->relids, phinfo->ph_eval_at) && + bms_is_subset(rel2->relids, phinfo->ph_eval_at)) + return true; + } + + /* + * It's possible that the rels correspond to the left and right sides of a + * degenerate outer join, that is, one with no joinclause mentioning the + * non-nullable side; in which case we should force the join to occur. + * + * Also, the two rels could represent a clauseless join that has to be + * completed to build up the LHS or RHS of an outer join. + */ + foreach(l, root->join_info_list) + { + SpecialJoinInfo *sjinfo = (SpecialJoinInfo *) lfirst(l); + + /* ignore full joins --- other mechanisms handle them */ + if (sjinfo->jointype == JOIN_FULL) + continue; + + /* Can we perform the SJ with these rels? */ + if (bms_is_subset(sjinfo->min_lefthand, rel1->relids) && + bms_is_subset(sjinfo->min_righthand, rel2->relids)) + { + result = true; + break; + } + if (bms_is_subset(sjinfo->min_lefthand, rel2->relids) && + bms_is_subset(sjinfo->min_righthand, rel1->relids)) + { + result = true; + break; + } + + /* + * Might we need to join these rels to complete the RHS? We have to + * use "overlap" tests since either rel might include a lower SJ that + * has been proven to commute with this one. + */ + if (bms_overlap(sjinfo->min_righthand, rel1->relids) && + bms_overlap(sjinfo->min_righthand, rel2->relids)) + { + result = true; + break; + } + + /* Likewise for the LHS. */ + if (bms_overlap(sjinfo->min_lefthand, rel1->relids) && + bms_overlap(sjinfo->min_lefthand, rel2->relids)) + { + result = true; + break; + } + } + + /* + * We do not force the join to occur if either input rel can legally be + * joined to anything else using joinclauses. This essentially means that + * clauseless bushy joins are put off as long as possible. The reason is + * that when there is a join order restriction high up in the join tree + * (that is, with many rels inside the LHS or RHS), we would otherwise + * expend lots of effort considering very stupid join combinations within + * its LHS or RHS. + */ + if (result) + { + if (has_legal_joinclause(root, rel1) || + has_legal_joinclause(root, rel2)) + result = false; + } + + return result; +} + + +/* + * has_join_restriction + * Detect whether the specified relation has join-order restrictions, + * due to being inside an outer join or an IN (sub-SELECT), + * or participating in any LATERAL references or multi-rel PHVs. + * + * Essentially, this tests whether have_join_order_restriction() could + * succeed with this rel and some other one. It's OK if we sometimes + * say "true" incorrectly. (Therefore, we don't bother with the relatively + * expensive has_legal_joinclause test.) + */ +static bool +has_join_restriction(PlannerInfo *root, RelOptInfo *rel) +{ + ListCell *l; + + if (rel->lateral_relids != NULL || rel->lateral_referencers != NULL) + return true; + + foreach(l, root->placeholder_list) + { + PlaceHolderInfo *phinfo = (PlaceHolderInfo *) lfirst(l); + + if (bms_is_subset(rel->relids, phinfo->ph_eval_at) && + !bms_equal(rel->relids, phinfo->ph_eval_at)) + return true; + } + + foreach(l, root->join_info_list) + { + SpecialJoinInfo *sjinfo = (SpecialJoinInfo *) lfirst(l); + + /* ignore full joins --- other mechanisms preserve their ordering */ + if (sjinfo->jointype == JOIN_FULL) + continue; + + /* ignore if SJ is already contained in rel */ + if (bms_is_subset(sjinfo->min_lefthand, rel->relids) && + bms_is_subset(sjinfo->min_righthand, rel->relids)) + continue; + + /* restricted if it overlaps LHS or RHS, but doesn't contain SJ */ + if (bms_overlap(sjinfo->min_lefthand, rel->relids) || + bms_overlap(sjinfo->min_righthand, rel->relids)) + return true; + } + + return false; +} + + +/* + * has_legal_joinclause + * Detect whether the specified relation can legally be joined + * to any other rels using join clauses. + * + * We consider only joins to single other relations in the current + * initial_rels list. This is sufficient to get a "true" result in most real + * queries, and an occasional erroneous "false" will only cost a bit more + * planning time. The reason for this limitation is that considering joins to + * other joins would require proving that the other join rel can legally be + * formed, which seems like too much trouble for something that's only a + * heuristic to save planning time. (Note: we must look at initial_rels + * and not all of the query, since when we are planning a sub-joinlist we + * may be forced to make clauseless joins within initial_rels even though + * there are join clauses linking to other parts of the query.) + */ +static bool +has_legal_joinclause(PlannerInfo *root, RelOptInfo *rel) +{ + ListCell *lc; + + foreach(lc, root->initial_rels) + { + RelOptInfo *rel2 = (RelOptInfo *) lfirst(lc); + + /* ignore rels that are already in "rel" */ + if (bms_overlap(rel->relids, rel2->relids)) + continue; + + if (have_relevant_joinclause(root, rel, rel2)) + { + Relids joinrelids; + SpecialJoinInfo *sjinfo; + bool reversed; + + /* join_is_legal needs relids of the union */ + joinrelids = bms_union(rel->relids, rel2->relids); + + if (join_is_legal(root, rel, rel2, joinrelids, + &sjinfo, &reversed)) + { + /* Yes, this will work */ + bms_free(joinrelids); + return true; + } + + bms_free(joinrelids); + } + } + + return false; +} + + +/* + * There's a pitfall for creating parameterized nestloops: suppose the inner + * rel (call it A) has a parameter that is a PlaceHolderVar, and that PHV's + * minimum eval_at set includes the outer rel (B) and some third rel (C). + * We might think we could create a B/A nestloop join that's parameterized by + * C. But we would end up with a plan in which the PHV's expression has to be + * evaluated as a nestloop parameter at the B/A join; and the executor is only + * set up to handle simple Vars as NestLoopParams. Rather than add complexity + * and overhead to the executor for such corner cases, it seems better to + * forbid the join. (Note that we can still make use of A's parameterized + * path with pre-joined B+C as the outer rel. have_join_order_restriction() + * ensures that we will consider making such a join even if there are not + * other reasons to do so.) + * + * So we check whether any PHVs used in the query could pose such a hazard. + * We don't have any simple way of checking whether a risky PHV would actually + * be used in the inner plan, and the case is so unusual that it doesn't seem + * worth working very hard on it. + * + * This needs to be checked in two places. If the inner rel's minimum + * parameterization would trigger the restriction, then join_is_legal() should + * reject the join altogether, because there will be no workable paths for it. + * But joinpath.c has to check again for every proposed nestloop path, because + * the inner path might have more than the minimum parameterization, causing + * some PHV to be dangerous for it that otherwise wouldn't be. + */ +bool +have_dangerous_phv(PlannerInfo *root, + Relids outer_relids, Relids inner_params) +{ + ListCell *lc; + + foreach(lc, root->placeholder_list) + { + PlaceHolderInfo *phinfo = (PlaceHolderInfo *) lfirst(lc); + + if (!bms_is_subset(phinfo->ph_eval_at, inner_params)) + continue; /* ignore, could not be a nestloop param */ + if (!bms_overlap(phinfo->ph_eval_at, outer_relids)) + continue; /* ignore, not relevant to this join */ + if (bms_is_subset(phinfo->ph_eval_at, outer_relids)) + continue; /* safe, it can be eval'd within outerrel */ + /* Otherwise, it's potentially unsafe, so reject the join */ + return true; + } + + /* OK to perform the join */ + return false; +} + + +/* + * is_dummy_rel --- has relation been proven empty? + */ +bool +is_dummy_rel(RelOptInfo *rel) +{ + Path *path; + + /* + * A rel that is known dummy will have just one path that is a childless + * Append. (Even if somehow it has more paths, a childless Append will + * have cost zero and hence should be at the front of the pathlist.) + */ + if (rel->pathlist == NIL) + return false; + path = (Path *) linitial(rel->pathlist); + + /* + * Initially, a dummy path will just be a childless Append. But in later + * planning stages we might stick a ProjectSetPath and/or ProjectionPath + * on top, since Append can't project. Rather than make assumptions about + * which combinations can occur, just descend through whatever we find. + */ + for (;;) + { + if (IsA(path, ProjectionPath)) + path = ((ProjectionPath *) path)->subpath; + else if (IsA(path, ProjectSetPath)) + path = ((ProjectSetPath *) path)->subpath; + else + break; + } + if (IS_DUMMY_APPEND(path)) + return true; + return false; +} + +/* + * Mark a relation as proven empty. + * + * During GEQO planning, this can get invoked more than once on the same + * baserel struct, so it's worth checking to see if the rel is already marked + * dummy. + * + * Also, when called during GEQO join planning, we are in a short-lived + * memory context. We must make sure that the dummy path attached to a + * baserel survives the GEQO cycle, else the baserel is trashed for future + * GEQO cycles. On the other hand, when we are marking a joinrel during GEQO, + * we don't want the dummy path to clutter the main planning context. Upshot + * is that the best solution is to explicitly make the dummy path in the same + * context the given RelOptInfo is in. + */ +void +mark_dummy_rel(RelOptInfo *rel) +{ + MemoryContext oldcontext; + + /* Already marked? */ + if (is_dummy_rel(rel)) + return; + + /* No, so choose correct context to make the dummy path in */ + oldcontext = MemoryContextSwitchTo(GetMemoryChunkContext(rel)); + + /* Set dummy size estimate */ + rel->rows = 0; + + /* Evict any previously chosen paths */ + rel->pathlist = NIL; + rel->partial_pathlist = NIL; + + /* Set up the dummy path */ + add_path(rel, (Path *) create_append_path(NULL, rel, NIL, NIL, + NIL, rel->lateral_relids, + 0, false, -1)); + + /* Set or update cheapest_total_path and related fields */ + set_cheapest(rel); + + MemoryContextSwitchTo(oldcontext); +} + + +/* + * restriction_is_constant_false --- is a restrictlist just FALSE? + * + * In cases where a qual is provably constant FALSE, eval_const_expressions + * will generally have thrown away anything that's ANDed with it. In outer + * join situations this will leave us computing cartesian products only to + * decide there's no match for an outer row, which is pretty stupid. So, + * we need to detect the case. + * + * If only_pushed_down is true, then consider only quals that are pushed-down + * from the point of view of the joinrel. + */ +static bool +restriction_is_constant_false(List *restrictlist, + RelOptInfo *joinrel, + bool only_pushed_down) +{ + ListCell *lc; + + /* + * Despite the above comment, the restriction list we see here might + * possibly have other members besides the FALSE constant, since other + * quals could get "pushed down" to the outer join level. So we check + * each member of the list. + */ + foreach(lc, restrictlist) + { + RestrictInfo *rinfo = lfirst_node(RestrictInfo, lc); + + if (only_pushed_down && !RINFO_IS_PUSHED_DOWN(rinfo, joinrel->relids)) + continue; + + if (rinfo->clause && IsA(rinfo->clause, Const)) + { + Const *con = (Const *) rinfo->clause; + + /* constant NULL is as good as constant FALSE for our purposes */ + if (con->constisnull) + return true; + if (!DatumGetBool(con->constvalue)) + return true; + } + } + return false; +} + +/* + * Assess whether join between given two partitioned relations can be broken + * down into joins between matching partitions; a technique called + * "partitionwise join" + * + * Partitionwise join is possible when a. Joining relations have same + * partitioning scheme b. There exists an equi-join between the partition keys + * of the two relations. + * + * Partitionwise join is planned as follows (details: optimizer/README.) + * + * 1. Create the RelOptInfos for joins between matching partitions i.e + * child-joins and add paths to them. + * + * 2. Construct Append or MergeAppend paths across the set of child joins. + * This second phase is implemented by generate_partitionwise_join_paths(). + * + * The RelOptInfo, SpecialJoinInfo and restrictlist for each child join are + * obtained by translating the respective parent join structures. + */ +static void +try_partitionwise_join(PlannerInfo *root, RelOptInfo *rel1, RelOptInfo *rel2, + RelOptInfo *joinrel, SpecialJoinInfo *parent_sjinfo, + List *parent_restrictlist) +{ + bool rel1_is_simple = IS_SIMPLE_REL(rel1); + bool rel2_is_simple = IS_SIMPLE_REL(rel2); + List *parts1 = NIL; + List *parts2 = NIL; + ListCell *lcr1 = NULL; + ListCell *lcr2 = NULL; + int cnt_parts; + + /* Guard against stack overflow due to overly deep partition hierarchy. */ + check_stack_depth(); + + /* Nothing to do, if the join relation is not partitioned. */ + if (joinrel->part_scheme == NULL || joinrel->nparts == 0) + return; + + /* The join relation should have consider_partitionwise_join set. */ + Assert(joinrel->consider_partitionwise_join); + + /* + * We can not perform partitionwise join if either of the joining + * relations is not partitioned. + */ + if (!IS_PARTITIONED_REL(rel1) || !IS_PARTITIONED_REL(rel2)) + return; + + Assert(REL_HAS_ALL_PART_PROPS(rel1) && REL_HAS_ALL_PART_PROPS(rel2)); + + /* The joining relations should have consider_partitionwise_join set. */ + Assert(rel1->consider_partitionwise_join && + rel2->consider_partitionwise_join); + + /* + * The partition scheme of the join relation should match that of the + * joining relations. + */ + Assert(joinrel->part_scheme == rel1->part_scheme && + joinrel->part_scheme == rel2->part_scheme); + + Assert(!(joinrel->partbounds_merged && (joinrel->nparts <= 0))); + + compute_partition_bounds(root, rel1, rel2, joinrel, parent_sjinfo, + &parts1, &parts2); + + if (joinrel->partbounds_merged) + { + lcr1 = list_head(parts1); + lcr2 = list_head(parts2); + } + + /* + * Create child-join relations for this partitioned join, if those don't + * exist. Add paths to child-joins for a pair of child relations + * corresponding to the given pair of parent relations. + */ + for (cnt_parts = 0; cnt_parts < joinrel->nparts; cnt_parts++) + { + RelOptInfo *child_rel1; + RelOptInfo *child_rel2; + bool rel1_empty; + bool rel2_empty; + SpecialJoinInfo *child_sjinfo; + List *child_restrictlist; + RelOptInfo *child_joinrel; + Relids child_joinrelids; + AppendRelInfo **appinfos; + int nappinfos; + + if (joinrel->partbounds_merged) + { + child_rel1 = lfirst_node(RelOptInfo, lcr1); + child_rel2 = lfirst_node(RelOptInfo, lcr2); + lcr1 = lnext(parts1, lcr1); + lcr2 = lnext(parts2, lcr2); + } + else + { + child_rel1 = rel1->part_rels[cnt_parts]; + child_rel2 = rel2->part_rels[cnt_parts]; + } + + rel1_empty = (child_rel1 == NULL || IS_DUMMY_REL(child_rel1)); + rel2_empty = (child_rel2 == NULL || IS_DUMMY_REL(child_rel2)); + + /* + * Check for cases where we can prove that this segment of the join + * returns no rows, due to one or both inputs being empty (including + * inputs that have been pruned away entirely). If so just ignore it. + * These rules are equivalent to populate_joinrel_with_paths's rules + * for dummy input relations. + */ + switch (parent_sjinfo->jointype) + { + case JOIN_INNER: + case JOIN_SEMI: + if (rel1_empty || rel2_empty) + continue; /* ignore this join segment */ + break; + case JOIN_LEFT: + case JOIN_ANTI: + if (rel1_empty) + continue; /* ignore this join segment */ + break; + case JOIN_FULL: + if (rel1_empty && rel2_empty) + continue; /* ignore this join segment */ + break; + default: + /* other values not expected here */ + elog(ERROR, "unrecognized join type: %d", + (int) parent_sjinfo->jointype); + break; + } + + /* + * If a child has been pruned entirely then we can't generate paths + * for it, so we have to reject partitionwise joining unless we were + * able to eliminate this partition above. + */ + if (child_rel1 == NULL || child_rel2 == NULL) + { + /* + * Mark the joinrel as unpartitioned so that later functions treat + * it correctly. + */ + joinrel->nparts = 0; + return; + } + + /* + * If a leaf relation has consider_partitionwise_join=false, it means + * that it's a dummy relation for which we skipped setting up tlist + * expressions and adding EC members in set_append_rel_size(), so + * again we have to fail here. + */ + if (rel1_is_simple && !child_rel1->consider_partitionwise_join) + { + Assert(child_rel1->reloptkind == RELOPT_OTHER_MEMBER_REL); + Assert(IS_DUMMY_REL(child_rel1)); + joinrel->nparts = 0; + return; + } + if (rel2_is_simple && !child_rel2->consider_partitionwise_join) + { + Assert(child_rel2->reloptkind == RELOPT_OTHER_MEMBER_REL); + Assert(IS_DUMMY_REL(child_rel2)); + joinrel->nparts = 0; + return; + } + + /* We should never try to join two overlapping sets of rels. */ + Assert(!bms_overlap(child_rel1->relids, child_rel2->relids)); + child_joinrelids = bms_union(child_rel1->relids, child_rel2->relids); + appinfos = find_appinfos_by_relids(root, child_joinrelids, &nappinfos); + + /* + * Construct SpecialJoinInfo from parent join relations's + * SpecialJoinInfo. + */ + child_sjinfo = build_child_join_sjinfo(root, parent_sjinfo, + child_rel1->relids, + child_rel2->relids); + + /* + * Construct restrictions applicable to the child join from those + * applicable to the parent join. + */ + child_restrictlist = + (List *) adjust_appendrel_attrs(root, + (Node *) parent_restrictlist, + nappinfos, appinfos); + pfree(appinfos); + + child_joinrel = joinrel->part_rels[cnt_parts]; + if (!child_joinrel) + { + child_joinrel = build_child_join_rel(root, child_rel1, child_rel2, + joinrel, child_restrictlist, + child_sjinfo, + child_sjinfo->jointype); + joinrel->part_rels[cnt_parts] = child_joinrel; + joinrel->live_parts = bms_add_member(joinrel->live_parts, cnt_parts); + joinrel->all_partrels = bms_add_members(joinrel->all_partrels, + child_joinrel->relids); + } + + Assert(bms_equal(child_joinrel->relids, child_joinrelids)); + + populate_joinrel_with_paths(root, child_rel1, child_rel2, + child_joinrel, child_sjinfo, + child_restrictlist); + } +} + +/* + * Construct the SpecialJoinInfo for a child-join by translating + * SpecialJoinInfo for the join between parents. left_relids and right_relids + * are the relids of left and right side of the join respectively. + */ +static SpecialJoinInfo * +build_child_join_sjinfo(PlannerInfo *root, SpecialJoinInfo *parent_sjinfo, + Relids left_relids, Relids right_relids) +{ + SpecialJoinInfo *sjinfo = makeNode(SpecialJoinInfo); + AppendRelInfo **left_appinfos; + int left_nappinfos; + AppendRelInfo **right_appinfos; + int right_nappinfos; + + memcpy(sjinfo, parent_sjinfo, sizeof(SpecialJoinInfo)); + left_appinfos = find_appinfos_by_relids(root, left_relids, + &left_nappinfos); + right_appinfos = find_appinfos_by_relids(root, right_relids, + &right_nappinfos); + + sjinfo->min_lefthand = adjust_child_relids(sjinfo->min_lefthand, + left_nappinfos, left_appinfos); + sjinfo->min_righthand = adjust_child_relids(sjinfo->min_righthand, + right_nappinfos, + right_appinfos); + sjinfo->syn_lefthand = adjust_child_relids(sjinfo->syn_lefthand, + left_nappinfos, left_appinfos); + sjinfo->syn_righthand = adjust_child_relids(sjinfo->syn_righthand, + right_nappinfos, + right_appinfos); + sjinfo->semi_rhs_exprs = (List *) adjust_appendrel_attrs(root, + (Node *) sjinfo->semi_rhs_exprs, + right_nappinfos, + right_appinfos); + + pfree(left_appinfos); + pfree(right_appinfos); + + return sjinfo; +} + +/* + * compute_partition_bounds + * Compute the partition bounds for a join rel from those for inputs + */ +static void +compute_partition_bounds(PlannerInfo *root, RelOptInfo *rel1, + RelOptInfo *rel2, RelOptInfo *joinrel, + SpecialJoinInfo *parent_sjinfo, + List **parts1, List **parts2) +{ + /* + * If we don't have the partition bounds for the join rel yet, try to + * compute those along with pairs of partitions to be joined. + */ + if (joinrel->nparts == -1) + { + PartitionScheme part_scheme = joinrel->part_scheme; + PartitionBoundInfo boundinfo = NULL; + int nparts = 0; + + Assert(joinrel->boundinfo == NULL); + Assert(joinrel->part_rels == NULL); + + /* + * See if the partition bounds for inputs are exactly the same, in + * which case we don't need to work hard: the join rel will have the + * same partition bounds as inputs, and the partitions with the same + * cardinal positions will form the pairs. + * + * Note: even in cases where one or both inputs have merged bounds, it + * would be possible for both the bounds to be exactly the same, but + * it seems unlikely to be worth the cycles to check. + */ + if (!rel1->partbounds_merged && + !rel2->partbounds_merged && + rel1->nparts == rel2->nparts && + partition_bounds_equal(part_scheme->partnatts, + part_scheme->parttyplen, + part_scheme->parttypbyval, + rel1->boundinfo, rel2->boundinfo)) + { + boundinfo = rel1->boundinfo; + nparts = rel1->nparts; + } + else + { + /* Try merging the partition bounds for inputs. */ + boundinfo = partition_bounds_merge(part_scheme->partnatts, + part_scheme->partsupfunc, + part_scheme->partcollation, + rel1, rel2, + parent_sjinfo->jointype, + parts1, parts2); + if (boundinfo == NULL) + { + joinrel->nparts = 0; + return; + } + nparts = list_length(*parts1); + joinrel->partbounds_merged = true; + } + + Assert(nparts > 0); + joinrel->boundinfo = boundinfo; + joinrel->nparts = nparts; + joinrel->part_rels = + (RelOptInfo **) palloc0(sizeof(RelOptInfo *) * nparts); + } + else + { + Assert(joinrel->nparts > 0); + Assert(joinrel->boundinfo); + Assert(joinrel->part_rels); + + /* + * If the join rel's partbounds_merged flag is true, it means inputs + * are not guaranteed to have the same partition bounds, therefore we + * can't assume that the partitions at the same cardinal positions + * form the pairs; let get_matching_part_pairs() generate the pairs. + * Otherwise, nothing to do since we can assume that. + */ + if (joinrel->partbounds_merged) + { + get_matching_part_pairs(root, joinrel, rel1, rel2, + parts1, parts2); + Assert(list_length(*parts1) == joinrel->nparts); + Assert(list_length(*parts2) == joinrel->nparts); + } + } +} + +/* + * get_matching_part_pairs + * Generate pairs of partitions to be joined from inputs + */ +static void +get_matching_part_pairs(PlannerInfo *root, RelOptInfo *joinrel, + RelOptInfo *rel1, RelOptInfo *rel2, + List **parts1, List **parts2) +{ + bool rel1_is_simple = IS_SIMPLE_REL(rel1); + bool rel2_is_simple = IS_SIMPLE_REL(rel2); + int cnt_parts; + + *parts1 = NIL; + *parts2 = NIL; + + for (cnt_parts = 0; cnt_parts < joinrel->nparts; cnt_parts++) + { + RelOptInfo *child_joinrel = joinrel->part_rels[cnt_parts]; + RelOptInfo *child_rel1; + RelOptInfo *child_rel2; + Relids child_relids1; + Relids child_relids2; + + /* + * If this segment of the join is empty, it means that this segment + * was ignored when previously creating child-join paths for it in + * try_partitionwise_join() as it would not contribute to the join + * result, due to one or both inputs being empty; add NULL to each of + * the given lists so that this segment will be ignored again in that + * function. + */ + if (!child_joinrel) + { + *parts1 = lappend(*parts1, NULL); + *parts2 = lappend(*parts2, NULL); + continue; + } + + /* + * Get a relids set of partition(s) involved in this join segment that + * are from the rel1 side. + */ + child_relids1 = bms_intersect(child_joinrel->relids, + rel1->all_partrels); + Assert(bms_num_members(child_relids1) == bms_num_members(rel1->relids)); + + /* + * Get a child rel for rel1 with the relids. Note that we should have + * the child rel even if rel1 is a join rel, because in that case the + * partitions specified in the relids would have matching/overlapping + * boundaries, so the specified partitions should be considered as + * ones to be joined when planning partitionwise joins of rel1, + * meaning that the child rel would have been built by the time we get + * here. + */ + if (rel1_is_simple) + { + int varno = bms_singleton_member(child_relids1); + + child_rel1 = find_base_rel(root, varno); + } + else + child_rel1 = find_join_rel(root, child_relids1); + Assert(child_rel1); + + /* + * Get a relids set of partition(s) involved in this join segment that + * are from the rel2 side. + */ + child_relids2 = bms_intersect(child_joinrel->relids, + rel2->all_partrels); + Assert(bms_num_members(child_relids2) == bms_num_members(rel2->relids)); + + /* + * Get a child rel for rel2 with the relids. See above comments. + */ + if (rel2_is_simple) + { + int varno = bms_singleton_member(child_relids2); + + child_rel2 = find_base_rel(root, varno); + } + else + child_rel2 = find_join_rel(root, child_relids2); + Assert(child_rel2); + + /* + * The join of rel1 and rel2 is legal, so is the join of the child + * rels obtained above; add them to the given lists as a join pair + * producing this join segment. + */ + *parts1 = lappend(*parts1, child_rel1); + *parts2 = lappend(*parts2, child_rel2); + } +} diff --git a/src/backend/optimizer/path/pathkeys.c b/src/backend/optimizer/path/pathkeys.c new file mode 100644 index 0000000..86a35cd --- /dev/null +++ b/src/backend/optimizer/path/pathkeys.c @@ -0,0 +1,1917 @@ +/*------------------------------------------------------------------------- + * + * pathkeys.c + * Utilities for matching and building path keys + * + * See src/backend/optimizer/README for a great deal of information about + * the nature and use of path keys. + * + * + * Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group + * Portions Copyright (c) 1994, Regents of the University of California + * + * IDENTIFICATION + * src/backend/optimizer/path/pathkeys.c + * + *------------------------------------------------------------------------- + */ +#include "postgres.h" + +#include "access/stratnum.h" +#include "catalog/pg_opfamily.h" +#include "nodes/makefuncs.h" +#include "nodes/nodeFuncs.h" +#include "nodes/plannodes.h" +#include "optimizer/optimizer.h" +#include "optimizer/pathnode.h" +#include "optimizer/paths.h" +#include "partitioning/partbounds.h" +#include "utils/lsyscache.h" + + +static bool pathkey_is_redundant(PathKey *new_pathkey, List *pathkeys); +static bool matches_boolean_partition_clause(RestrictInfo *rinfo, + RelOptInfo *partrel, + int partkeycol); +static Var *find_var_for_subquery_tle(RelOptInfo *rel, TargetEntry *tle); +static bool right_merge_direction(PlannerInfo *root, PathKey *pathkey); + + +/**************************************************************************** + * PATHKEY CONSTRUCTION AND REDUNDANCY TESTING + ****************************************************************************/ + +/* + * make_canonical_pathkey + * Given the parameters for a PathKey, find any pre-existing matching + * pathkey in the query's list of "canonical" pathkeys. Make a new + * entry if there's not one already. + * + * Note that this function must not be used until after we have completed + * merging EquivalenceClasses. + */ +PathKey * +make_canonical_pathkey(PlannerInfo *root, + EquivalenceClass *eclass, Oid opfamily, + int strategy, bool nulls_first) +{ + PathKey *pk; + ListCell *lc; + MemoryContext oldcontext; + + /* Can't make canonical pathkeys if the set of ECs might still change */ + if (!root->ec_merging_done) + elog(ERROR, "too soon to build canonical pathkeys"); + + /* The passed eclass might be non-canonical, so chase up to the top */ + while (eclass->ec_merged) + eclass = eclass->ec_merged; + + foreach(lc, root->canon_pathkeys) + { + pk = (PathKey *) lfirst(lc); + if (eclass == pk->pk_eclass && + opfamily == pk->pk_opfamily && + strategy == pk->pk_strategy && + nulls_first == pk->pk_nulls_first) + return pk; + } + + /* + * Be sure canonical pathkeys are allocated in the main planning context. + * Not an issue in normal planning, but it is for GEQO. + */ + oldcontext = MemoryContextSwitchTo(root->planner_cxt); + + pk = makeNode(PathKey); + pk->pk_eclass = eclass; + pk->pk_opfamily = opfamily; + pk->pk_strategy = strategy; + pk->pk_nulls_first = nulls_first; + + root->canon_pathkeys = lappend(root->canon_pathkeys, pk); + + MemoryContextSwitchTo(oldcontext); + + return pk; +} + +/* + * pathkey_is_redundant + * Is a pathkey redundant with one already in the given list? + * + * We detect two cases: + * + * 1. If the new pathkey's equivalence class contains a constant, and isn't + * below an outer join, then we can disregard it as a sort key. An example: + * SELECT ... WHERE x = 42 ORDER BY x, y; + * We may as well just sort by y. Note that because of opfamily matching, + * this is semantically correct: we know that the equality constraint is one + * that actually binds the variable to a single value in the terms of any + * ordering operator that might go with the eclass. This rule not only lets + * us simplify (or even skip) explicit sorts, but also allows matching index + * sort orders to a query when there are don't-care index columns. + * + * 2. If the new pathkey's equivalence class is the same as that of any + * existing member of the pathkey list, then it is redundant. Some examples: + * SELECT ... ORDER BY x, x; + * SELECT ... ORDER BY x, x DESC; + * SELECT ... WHERE x = y ORDER BY x, y; + * In all these cases the second sort key cannot distinguish values that are + * considered equal by the first, and so there's no point in using it. + * Note in particular that we need not compare opfamily (all the opfamilies + * of the EC have the same notion of equality) nor sort direction. + * + * Both the given pathkey and the list members must be canonical for this + * to work properly, but that's okay since we no longer ever construct any + * non-canonical pathkeys. (Note: the notion of a pathkey *list* being + * canonical includes the additional requirement of no redundant entries, + * which is exactly what we are checking for here.) + * + * Because the equivclass.c machinery forms only one copy of any EC per query, + * pointer comparison is enough to decide whether canonical ECs are the same. + */ +static bool +pathkey_is_redundant(PathKey *new_pathkey, List *pathkeys) +{ + EquivalenceClass *new_ec = new_pathkey->pk_eclass; + ListCell *lc; + + /* Check for EC containing a constant --- unconditionally redundant */ + if (EC_MUST_BE_REDUNDANT(new_ec)) + return true; + + /* If same EC already used in list, then redundant */ + foreach(lc, pathkeys) + { + PathKey *old_pathkey = (PathKey *) lfirst(lc); + + if (new_ec == old_pathkey->pk_eclass) + return true; + } + + return false; +} + +/* + * make_pathkey_from_sortinfo + * Given an expression and sort-order information, create a PathKey. + * The result is always a "canonical" PathKey, but it might be redundant. + * + * expr is the expression, and nullable_relids is the set of base relids + * that are potentially nullable below it. + * + * If the PathKey is being generated from a SortGroupClause, sortref should be + * the SortGroupClause's SortGroupRef; otherwise zero. + * + * If rel is not NULL, it identifies a specific relation we're considering + * a path for, and indicates that child EC members for that relation can be + * considered. Otherwise child members are ignored. (See the comments for + * get_eclass_for_sort_expr.) + * + * create_it is true if we should create any missing EquivalenceClass + * needed to represent the sort key. If it's false, we return NULL if the + * sort key isn't already present in any EquivalenceClass. + */ +static PathKey * +make_pathkey_from_sortinfo(PlannerInfo *root, + Expr *expr, + Relids nullable_relids, + Oid opfamily, + Oid opcintype, + Oid collation, + bool reverse_sort, + bool nulls_first, + Index sortref, + Relids rel, + bool create_it) +{ + int16 strategy; + Oid equality_op; + List *opfamilies; + EquivalenceClass *eclass; + + strategy = reverse_sort ? BTGreaterStrategyNumber : BTLessStrategyNumber; + + /* + * EquivalenceClasses need to contain opfamily lists based on the family + * membership of mergejoinable equality operators, which could belong to + * more than one opfamily. So we have to look up the opfamily's equality + * operator and get its membership. + */ + equality_op = get_opfamily_member(opfamily, + opcintype, + opcintype, + BTEqualStrategyNumber); + if (!OidIsValid(equality_op)) /* shouldn't happen */ + elog(ERROR, "missing operator %d(%u,%u) in opfamily %u", + BTEqualStrategyNumber, opcintype, opcintype, opfamily); + opfamilies = get_mergejoin_opfamilies(equality_op); + if (!opfamilies) /* certainly should find some */ + elog(ERROR, "could not find opfamilies for equality operator %u", + equality_op); + + /* Now find or (optionally) create a matching EquivalenceClass */ + eclass = get_eclass_for_sort_expr(root, expr, nullable_relids, + opfamilies, opcintype, collation, + sortref, rel, create_it); + + /* Fail if no EC and !create_it */ + if (!eclass) + return NULL; + + /* And finally we can find or create a PathKey node */ + return make_canonical_pathkey(root, eclass, opfamily, + strategy, nulls_first); +} + +/* + * make_pathkey_from_sortop + * Like make_pathkey_from_sortinfo, but work from a sort operator. + * + * This should eventually go away, but we need to restructure SortGroupClause + * first. + */ +static PathKey * +make_pathkey_from_sortop(PlannerInfo *root, + Expr *expr, + Relids nullable_relids, + Oid ordering_op, + bool nulls_first, + Index sortref, + bool create_it) +{ + Oid opfamily, + opcintype, + collation; + int16 strategy; + + /* Find the operator in pg_amop --- failure shouldn't happen */ + if (!get_ordering_op_properties(ordering_op, + &opfamily, &opcintype, &strategy)) + elog(ERROR, "operator %u is not a valid ordering operator", + ordering_op); + + /* Because SortGroupClause doesn't carry collation, consult the expr */ + collation = exprCollation((Node *) expr); + + return make_pathkey_from_sortinfo(root, + expr, + nullable_relids, + opfamily, + opcintype, + collation, + (strategy == BTGreaterStrategyNumber), + nulls_first, + sortref, + NULL, + create_it); +} + + +/**************************************************************************** + * PATHKEY COMPARISONS + ****************************************************************************/ + +/* + * compare_pathkeys + * Compare two pathkeys to see if they are equivalent, and if not whether + * one is "better" than the other. + * + * We assume the pathkeys are canonical, and so they can be checked for + * equality by simple pointer comparison. + */ +PathKeysComparison +compare_pathkeys(List *keys1, List *keys2) +{ + ListCell *key1, + *key2; + + /* + * Fall out quickly if we are passed two identical lists. This mostly + * catches the case where both are NIL, but that's common enough to + * warrant the test. + */ + if (keys1 == keys2) + return PATHKEYS_EQUAL; + + forboth(key1, keys1, key2, keys2) + { + PathKey *pathkey1 = (PathKey *) lfirst(key1); + PathKey *pathkey2 = (PathKey *) lfirst(key2); + + if (pathkey1 != pathkey2) + return PATHKEYS_DIFFERENT; /* no need to keep looking */ + } + + /* + * If we reached the end of only one list, the other is longer and + * therefore not a subset. + */ + if (key1 != NULL) + return PATHKEYS_BETTER1; /* key1 is longer */ + if (key2 != NULL) + return PATHKEYS_BETTER2; /* key2 is longer */ + return PATHKEYS_EQUAL; +} + +/* + * pathkeys_contained_in + * Common special case of compare_pathkeys: we just want to know + * if keys2 are at least as well sorted as keys1. + */ +bool +pathkeys_contained_in(List *keys1, List *keys2) +{ + switch (compare_pathkeys(keys1, keys2)) + { + case PATHKEYS_EQUAL: + case PATHKEYS_BETTER2: + return true; + default: + break; + } + return false; +} + +/* + * pathkeys_count_contained_in + * Same as pathkeys_contained_in, but also sets length of longest + * common prefix of keys1 and keys2. + */ +bool +pathkeys_count_contained_in(List *keys1, List *keys2, int *n_common) +{ + int n = 0; + ListCell *key1, + *key2; + + /* + * See if we can avoiding looping through both lists. This optimization + * gains us several percent in planning time in a worst-case test. + */ + if (keys1 == keys2) + { + *n_common = list_length(keys1); + return true; + } + else if (keys1 == NIL) + { + *n_common = 0; + return true; + } + else if (keys2 == NIL) + { + *n_common = 0; + return false; + } + + /* + * If both lists are non-empty, iterate through both to find out how many + * items are shared. + */ + forboth(key1, keys1, key2, keys2) + { + PathKey *pathkey1 = (PathKey *) lfirst(key1); + PathKey *pathkey2 = (PathKey *) lfirst(key2); + + if (pathkey1 != pathkey2) + { + *n_common = n; + return false; + } + n++; + } + + /* If we ended with a null value, then we've processed the whole list. */ + *n_common = n; + return (key1 == NULL); +} + +/* + * get_cheapest_path_for_pathkeys + * Find the cheapest path (according to the specified criterion) that + * satisfies the given pathkeys and parameterization. + * Return NULL if no such path. + * + * 'paths' is a list of possible paths that all generate the same relation + * 'pathkeys' represents a required ordering (in canonical form!) + * 'required_outer' denotes allowable outer relations for parameterized paths + * 'cost_criterion' is STARTUP_COST or TOTAL_COST + * 'require_parallel_safe' causes us to consider only parallel-safe paths + */ +Path * +get_cheapest_path_for_pathkeys(List *paths, List *pathkeys, + Relids required_outer, + CostSelector cost_criterion, + bool require_parallel_safe) +{ + Path *matched_path = NULL; + ListCell *l; + + foreach(l, paths) + { + Path *path = (Path *) lfirst(l); + + /* + * Since cost comparison is a lot cheaper than pathkey comparison, do + * that first. (XXX is that still true?) + */ + if (matched_path != NULL && + compare_path_costs(matched_path, path, cost_criterion) <= 0) + continue; + + if (require_parallel_safe && !path->parallel_safe) + continue; + + if (pathkeys_contained_in(pathkeys, path->pathkeys) && + bms_is_subset(PATH_REQ_OUTER(path), required_outer)) + matched_path = path; + } + return matched_path; +} + +/* + * get_cheapest_fractional_path_for_pathkeys + * Find the cheapest path (for retrieving a specified fraction of all + * the tuples) that satisfies the given pathkeys and parameterization. + * Return NULL if no such path. + * + * See compare_fractional_path_costs() for the interpretation of the fraction + * parameter. + * + * 'paths' is a list of possible paths that all generate the same relation + * 'pathkeys' represents a required ordering (in canonical form!) + * 'required_outer' denotes allowable outer relations for parameterized paths + * 'fraction' is the fraction of the total tuples expected to be retrieved + */ +Path * +get_cheapest_fractional_path_for_pathkeys(List *paths, + List *pathkeys, + Relids required_outer, + double fraction) +{ + Path *matched_path = NULL; + ListCell *l; + + foreach(l, paths) + { + Path *path = (Path *) lfirst(l); + + /* + * Since cost comparison is a lot cheaper than pathkey comparison, do + * that first. (XXX is that still true?) + */ + if (matched_path != NULL && + compare_fractional_path_costs(matched_path, path, fraction) <= 0) + continue; + + if (pathkeys_contained_in(pathkeys, path->pathkeys) && + bms_is_subset(PATH_REQ_OUTER(path), required_outer)) + matched_path = path; + } + return matched_path; +} + + +/* + * get_cheapest_parallel_safe_total_inner + * Find the unparameterized parallel-safe path with the least total cost. + */ +Path * +get_cheapest_parallel_safe_total_inner(List *paths) +{ + ListCell *l; + + foreach(l, paths) + { + Path *innerpath = (Path *) lfirst(l); + + if (innerpath->parallel_safe && + bms_is_empty(PATH_REQ_OUTER(innerpath))) + return innerpath; + } + + return NULL; +} + +/**************************************************************************** + * NEW PATHKEY FORMATION + ****************************************************************************/ + +/* + * build_index_pathkeys + * Build a pathkeys list that describes the ordering induced by an index + * scan using the given index. (Note that an unordered index doesn't + * induce any ordering, so we return NIL.) + * + * If 'scandir' is BackwardScanDirection, build pathkeys representing a + * backwards scan of the index. + * + * We iterate only key columns of covering indexes, since non-key columns + * don't influence index ordering. The result is canonical, meaning that + * redundant pathkeys are removed; it may therefore have fewer entries than + * there are key columns in the index. + * + * Another reason for stopping early is that we may be able to tell that + * an index column's sort order is uninteresting for this query. However, + * that test is just based on the existence of an EquivalenceClass and not + * on position in pathkey lists, so it's not complete. Caller should call + * truncate_useless_pathkeys() to possibly remove more pathkeys. + */ +List * +build_index_pathkeys(PlannerInfo *root, + IndexOptInfo *index, + ScanDirection scandir) +{ + List *retval = NIL; + ListCell *lc; + int i; + + if (index->sortopfamily == NULL) + return NIL; /* non-orderable index */ + + i = 0; + foreach(lc, index->indextlist) + { + TargetEntry *indextle = (TargetEntry *) lfirst(lc); + Expr *indexkey; + bool reverse_sort; + bool nulls_first; + PathKey *cpathkey; + + /* + * INCLUDE columns are stored in index unordered, so they don't + * support ordered index scan. + */ + if (i >= index->nkeycolumns) + break; + + /* We assume we don't need to make a copy of the tlist item */ + indexkey = indextle->expr; + + if (ScanDirectionIsBackward(scandir)) + { + reverse_sort = !index->reverse_sort[i]; + nulls_first = !index->nulls_first[i]; + } + else + { + reverse_sort = index->reverse_sort[i]; + nulls_first = index->nulls_first[i]; + } + + /* + * OK, try to make a canonical pathkey for this sort key. Note we're + * underneath any outer joins, so nullable_relids should be NULL. + */ + cpathkey = make_pathkey_from_sortinfo(root, + indexkey, + NULL, + index->sortopfamily[i], + index->opcintype[i], + index->indexcollations[i], + reverse_sort, + nulls_first, + 0, + index->rel->relids, + false); + + if (cpathkey) + { + /* + * We found the sort key in an EquivalenceClass, so it's relevant + * for this query. Add it to list, unless it's redundant. + */ + if (!pathkey_is_redundant(cpathkey, retval)) + retval = lappend(retval, cpathkey); + } + else + { + /* + * Boolean index keys might be redundant even if they do not + * appear in an EquivalenceClass, because of our special treatment + * of boolean equality conditions --- see the comment for + * indexcol_is_bool_constant_for_query(). If that applies, we can + * continue to examine lower-order index columns. Otherwise, the + * sort key is not an interesting sort order for this query, so we + * should stop considering index columns; any lower-order sort + * keys won't be useful either. + */ + if (!indexcol_is_bool_constant_for_query(root, index, i)) + break; + } + + i++; + } + + return retval; +} + +/* + * partkey_is_bool_constant_for_query + * + * If a partition key column is constrained to have a constant value by the + * query's WHERE conditions, then it's irrelevant for sort-order + * considerations. Usually that means we have a restriction clause + * WHERE partkeycol = constant, which gets turned into an EquivalenceClass + * containing a constant, which is recognized as redundant by + * build_partition_pathkeys(). But if the partition key column is a + * boolean variable (or expression), then we are not going to see such a + * WHERE clause, because expression preprocessing will have simplified it + * to "WHERE partkeycol" or "WHERE NOT partkeycol". So we are not going + * to have a matching EquivalenceClass (unless the query also contains + * "ORDER BY partkeycol"). To allow such cases to work the same as they would + * for non-boolean values, this function is provided to detect whether the + * specified partition key column matches a boolean restriction clause. + */ +static bool +partkey_is_bool_constant_for_query(RelOptInfo *partrel, int partkeycol) +{ + PartitionScheme partscheme = partrel->part_scheme; + ListCell *lc; + + /* If the partkey isn't boolean, we can't possibly get a match */ + if (!IsBooleanOpfamily(partscheme->partopfamily[partkeycol])) + return false; + + /* Check each restriction clause for the partitioned rel */ + foreach(lc, partrel->baserestrictinfo) + { + RestrictInfo *rinfo = (RestrictInfo *) lfirst(lc); + + /* Ignore pseudoconstant quals, they won't match */ + if (rinfo->pseudoconstant) + continue; + + /* See if we can match the clause's expression to the partkey column */ + if (matches_boolean_partition_clause(rinfo, partrel, partkeycol)) + return true; + } + + return false; +} + +/* + * matches_boolean_partition_clause + * Determine if the boolean clause described by rinfo matches + * partrel's partkeycol-th partition key column. + * + * "Matches" can be either an exact match (equivalent to partkey = true), + * or a NOT above an exact match (equivalent to partkey = false). + */ +static bool +matches_boolean_partition_clause(RestrictInfo *rinfo, + RelOptInfo *partrel, int partkeycol) +{ + Node *clause = (Node *) rinfo->clause; + Node *partexpr = (Node *) linitial(partrel->partexprs[partkeycol]); + + /* Direct match? */ + if (equal(partexpr, clause)) + return true; + /* NOT clause? */ + else if (is_notclause(clause)) + { + Node *arg = (Node *) get_notclausearg((Expr *) clause); + + if (equal(partexpr, arg)) + return true; + } + + return false; +} + +/* + * build_partition_pathkeys + * Build a pathkeys list that describes the ordering induced by the + * partitions of partrel, under either forward or backward scan + * as per scandir. + * + * Caller must have checked that the partitions are properly ordered, + * as detected by partitions_are_ordered(). + * + * Sets *partialkeys to true if pathkeys were only built for a prefix of the + * partition key, or false if the pathkeys include all columns of the + * partition key. + */ +List * +build_partition_pathkeys(PlannerInfo *root, RelOptInfo *partrel, + ScanDirection scandir, bool *partialkeys) +{ + List *retval = NIL; + PartitionScheme partscheme = partrel->part_scheme; + int i; + + Assert(partscheme != NULL); + Assert(partitions_are_ordered(partrel->boundinfo, partrel->live_parts)); + /* For now, we can only cope with baserels */ + Assert(IS_SIMPLE_REL(partrel)); + + for (i = 0; i < partscheme->partnatts; i++) + { + PathKey *cpathkey; + Expr *keyCol = (Expr *) linitial(partrel->partexprs[i]); + + /* + * Try to make a canonical pathkey for this partkey. + * + * We're considering a baserel scan, so nullable_relids should be + * NULL. Also, we assume the PartitionDesc lists any NULL partition + * last, so we treat the scan like a NULLS LAST index: we have + * nulls_first for backwards scan only. + */ + cpathkey = make_pathkey_from_sortinfo(root, + keyCol, + NULL, + partscheme->partopfamily[i], + partscheme->partopcintype[i], + partscheme->partcollation[i], + ScanDirectionIsBackward(scandir), + ScanDirectionIsBackward(scandir), + 0, + partrel->relids, + false); + + + if (cpathkey) + { + /* + * We found the sort key in an EquivalenceClass, so it's relevant + * for this query. Add it to list, unless it's redundant. + */ + if (!pathkey_is_redundant(cpathkey, retval)) + retval = lappend(retval, cpathkey); + } + else + { + /* + * Boolean partition keys might be redundant even if they do not + * appear in an EquivalenceClass, because of our special treatment + * of boolean equality conditions --- see the comment for + * partkey_is_bool_constant_for_query(). If that applies, we can + * continue to examine lower-order partition keys. Otherwise, the + * sort key is not an interesting sort order for this query, so we + * should stop considering partition columns; any lower-order sort + * keys won't be useful either. + */ + if (!partkey_is_bool_constant_for_query(partrel, i)) + { + *partialkeys = true; + return retval; + } + } + } + + *partialkeys = false; + return retval; +} + +/* + * build_expression_pathkey + * Build a pathkeys list that describes an ordering by a single expression + * using the given sort operator. + * + * expr, nullable_relids, and rel are as for make_pathkey_from_sortinfo. + * We induce the other arguments assuming default sort order for the operator. + * + * Similarly to make_pathkey_from_sortinfo, the result is NIL if create_it + * is false and the expression isn't already in some EquivalenceClass. + */ +List * +build_expression_pathkey(PlannerInfo *root, + Expr *expr, + Relids nullable_relids, + Oid opno, + Relids rel, + bool create_it) +{ + List *pathkeys; + Oid opfamily, + opcintype; + int16 strategy; + PathKey *cpathkey; + + /* Find the operator in pg_amop --- failure shouldn't happen */ + if (!get_ordering_op_properties(opno, + &opfamily, &opcintype, &strategy)) + elog(ERROR, "operator %u is not a valid ordering operator", + opno); + + cpathkey = make_pathkey_from_sortinfo(root, + expr, + nullable_relids, + opfamily, + opcintype, + exprCollation((Node *) expr), + (strategy == BTGreaterStrategyNumber), + (strategy == BTGreaterStrategyNumber), + 0, + rel, + create_it); + + if (cpathkey) + pathkeys = list_make1(cpathkey); + else + pathkeys = NIL; + + return pathkeys; +} + +/* + * convert_subquery_pathkeys + * Build a pathkeys list that describes the ordering of a subquery's + * result, in the terms of the outer query. This is essentially a + * task of conversion. + * + * 'rel': outer query's RelOptInfo for the subquery relation. + * 'subquery_pathkeys': the subquery's output pathkeys, in its terms. + * 'subquery_tlist': the subquery's output targetlist, in its terms. + * + * We intentionally don't do truncate_useless_pathkeys() here, because there + * are situations where seeing the raw ordering of the subquery is helpful. + * For example, if it returns ORDER BY x DESC, that may prompt us to + * construct a mergejoin using DESC order rather than ASC order; but the + * right_merge_direction heuristic would have us throw the knowledge away. + */ +List * +convert_subquery_pathkeys(PlannerInfo *root, RelOptInfo *rel, + List *subquery_pathkeys, + List *subquery_tlist) +{ + List *retval = NIL; + int retvallen = 0; + int outer_query_keys = list_length(root->query_pathkeys); + ListCell *i; + + foreach(i, subquery_pathkeys) + { + PathKey *sub_pathkey = (PathKey *) lfirst(i); + EquivalenceClass *sub_eclass = sub_pathkey->pk_eclass; + PathKey *best_pathkey = NULL; + + if (sub_eclass->ec_has_volatile) + { + /* + * If the sub_pathkey's EquivalenceClass is volatile, then it must + * have come from an ORDER BY clause, and we have to match it to + * that same targetlist entry. + */ + TargetEntry *tle; + Var *outer_var; + + if (sub_eclass->ec_sortref == 0) /* can't happen */ + elog(ERROR, "volatile EquivalenceClass has no sortref"); + tle = get_sortgroupref_tle(sub_eclass->ec_sortref, subquery_tlist); + Assert(tle); + /* Is TLE actually available to the outer query? */ + outer_var = find_var_for_subquery_tle(rel, tle); + if (outer_var) + { + /* We can represent this sub_pathkey */ + EquivalenceMember *sub_member; + EquivalenceClass *outer_ec; + + Assert(list_length(sub_eclass->ec_members) == 1); + sub_member = (EquivalenceMember *) linitial(sub_eclass->ec_members); + + /* + * Note: it might look funny to be setting sortref = 0 for a + * reference to a volatile sub_eclass. However, the + * expression is *not* volatile in the outer query: it's just + * a Var referencing whatever the subquery emitted. (IOW, the + * outer query isn't going to re-execute the volatile + * expression itself.) So this is okay. Likewise, it's + * correct to pass nullable_relids = NULL, because we're + * underneath any outer joins appearing in the outer query. + */ + outer_ec = + get_eclass_for_sort_expr(root, + (Expr *) outer_var, + NULL, + sub_eclass->ec_opfamilies, + sub_member->em_datatype, + sub_eclass->ec_collation, + 0, + rel->relids, + false); + + /* + * If we don't find a matching EC, sub-pathkey isn't + * interesting to the outer query + */ + if (outer_ec) + best_pathkey = + make_canonical_pathkey(root, + outer_ec, + sub_pathkey->pk_opfamily, + sub_pathkey->pk_strategy, + sub_pathkey->pk_nulls_first); + } + } + else + { + /* + * Otherwise, the sub_pathkey's EquivalenceClass could contain + * multiple elements (representing knowledge that multiple items + * are effectively equal). Each element might match none, one, or + * more of the output columns that are visible to the outer query. + * This means we may have multiple possible representations of the + * sub_pathkey in the context of the outer query. Ideally we + * would generate them all and put them all into an EC of the + * outer query, thereby propagating equality knowledge up to the + * outer query. Right now we cannot do so, because the outer + * query's EquivalenceClasses are already frozen when this is + * called. Instead we prefer the one that has the highest "score" + * (number of EC peers, plus one if it matches the outer + * query_pathkeys). This is the most likely to be useful in the + * outer query. + */ + int best_score = -1; + ListCell *j; + + foreach(j, sub_eclass->ec_members) + { + EquivalenceMember *sub_member = (EquivalenceMember *) lfirst(j); + Expr *sub_expr = sub_member->em_expr; + Oid sub_expr_type = sub_member->em_datatype; + Oid sub_expr_coll = sub_eclass->ec_collation; + ListCell *k; + + if (sub_member->em_is_child) + continue; /* ignore children here */ + + foreach(k, subquery_tlist) + { + TargetEntry *tle = (TargetEntry *) lfirst(k); + Var *outer_var; + Expr *tle_expr; + EquivalenceClass *outer_ec; + PathKey *outer_pk; + int score; + + /* Is TLE actually available to the outer query? */ + outer_var = find_var_for_subquery_tle(rel, tle); + if (!outer_var) + continue; + + /* + * The targetlist entry is considered to match if it + * matches after sort-key canonicalization. That is + * needed since the sub_expr has been through the same + * process. + */ + tle_expr = canonicalize_ec_expression(tle->expr, + sub_expr_type, + sub_expr_coll); + if (!equal(tle_expr, sub_expr)) + continue; + + /* See if we have a matching EC for the TLE */ + outer_ec = get_eclass_for_sort_expr(root, + (Expr *) outer_var, + NULL, + sub_eclass->ec_opfamilies, + sub_expr_type, + sub_expr_coll, + 0, + rel->relids, + false); + + /* + * If we don't find a matching EC, this sub-pathkey isn't + * interesting to the outer query + */ + if (!outer_ec) + continue; + + outer_pk = make_canonical_pathkey(root, + outer_ec, + sub_pathkey->pk_opfamily, + sub_pathkey->pk_strategy, + sub_pathkey->pk_nulls_first); + /* score = # of equivalence peers */ + score = list_length(outer_ec->ec_members) - 1; + /* +1 if it matches the proper query_pathkeys item */ + if (retvallen < outer_query_keys && + list_nth(root->query_pathkeys, retvallen) == outer_pk) + score++; + if (score > best_score) + { + best_pathkey = outer_pk; + best_score = score; + } + } + } + } + + /* + * If we couldn't find a representation of this sub_pathkey, we're + * done (we can't use the ones to its right, either). + */ + if (!best_pathkey) + break; + + /* + * Eliminate redundant ordering info; could happen if outer query + * equivalences subquery keys... + */ + if (!pathkey_is_redundant(best_pathkey, retval)) + { + retval = lappend(retval, best_pathkey); + retvallen++; + } + } + + return retval; +} + +/* + * find_var_for_subquery_tle + * + * If the given subquery tlist entry is due to be emitted by the subquery's + * scan node, return a Var for it, else return NULL. + * + * We need this to ensure that we don't return pathkeys describing values + * that are unavailable above the level of the subquery scan. + */ +static Var * +find_var_for_subquery_tle(RelOptInfo *rel, TargetEntry *tle) +{ + ListCell *lc; + + /* If the TLE is resjunk, it's certainly not visible to the outer query */ + if (tle->resjunk) + return NULL; + + /* Search the rel's targetlist to see what it will return */ + foreach(lc, rel->reltarget->exprs) + { + Var *var = (Var *) lfirst(lc); + + /* Ignore placeholders */ + if (!IsA(var, Var)) + continue; + Assert(var->varno == rel->relid); + + /* If we find a Var referencing this TLE, we're good */ + if (var->varattno == tle->resno) + return copyObject(var); /* Make a copy for safety */ + } + return NULL; +} + +/* + * build_join_pathkeys + * Build the path keys for a join relation constructed by mergejoin or + * nestloop join. This is normally the same as the outer path's keys. + * + * EXCEPTION: in a FULL or RIGHT join, we cannot treat the result as + * having the outer path's path keys, because null lefthand rows may be + * inserted at random points. It must be treated as unsorted. + * + * We truncate away any pathkeys that are uninteresting for higher joins. + * + * 'joinrel' is the join relation that paths are being formed for + * 'jointype' is the join type (inner, left, full, etc) + * 'outer_pathkeys' is the list of the current outer path's path keys + * + * Returns the list of new path keys. + */ +List * +build_join_pathkeys(PlannerInfo *root, + RelOptInfo *joinrel, + JoinType jointype, + List *outer_pathkeys) +{ + if (jointype == JOIN_FULL || jointype == JOIN_RIGHT) + return NIL; + + /* + * This used to be quite a complex bit of code, but now that all pathkey + * sublists start out life canonicalized, we don't have to do a darn thing + * here! + * + * We do, however, need to truncate the pathkeys list, since it may + * contain pathkeys that were useful for forming this joinrel but are + * uninteresting to higher levels. + */ + return truncate_useless_pathkeys(root, joinrel, outer_pathkeys); +} + +/**************************************************************************** + * PATHKEYS AND SORT CLAUSES + ****************************************************************************/ + +/* + * make_pathkeys_for_sortclauses + * Generate a pathkeys list that represents the sort order specified + * by a list of SortGroupClauses + * + * The resulting PathKeys are always in canonical form. (Actually, there + * is no longer any code anywhere that creates non-canonical PathKeys.) + * + * We assume that root->nullable_baserels is the set of base relids that could + * have gone to NULL below the SortGroupClause expressions. This is okay if + * the expressions came from the query's top level (ORDER BY, DISTINCT, etc) + * and if this function is only invoked after deconstruct_jointree. In the + * future we might have to make callers pass in the appropriate + * nullable-relids set, but for now it seems unnecessary. + * + * 'sortclauses' is a list of SortGroupClause nodes + * 'tlist' is the targetlist to find the referenced tlist entries in + */ +List * +make_pathkeys_for_sortclauses(PlannerInfo *root, + List *sortclauses, + List *tlist) +{ + List *pathkeys = NIL; + ListCell *l; + + foreach(l, sortclauses) + { + SortGroupClause *sortcl = (SortGroupClause *) lfirst(l); + Expr *sortkey; + PathKey *pathkey; + + sortkey = (Expr *) get_sortgroupclause_expr(sortcl, tlist); + Assert(OidIsValid(sortcl->sortop)); + pathkey = make_pathkey_from_sortop(root, + sortkey, + root->nullable_baserels, + sortcl->sortop, + sortcl->nulls_first, + sortcl->tleSortGroupRef, + true); + + /* Canonical form eliminates redundant ordering keys */ + if (!pathkey_is_redundant(pathkey, pathkeys)) + pathkeys = lappend(pathkeys, pathkey); + } + return pathkeys; +} + +/**************************************************************************** + * PATHKEYS AND MERGECLAUSES + ****************************************************************************/ + +/* + * initialize_mergeclause_eclasses + * Set the EquivalenceClass links in a mergeclause restrictinfo. + * + * RestrictInfo contains fields in which we may cache pointers to + * EquivalenceClasses for the left and right inputs of the mergeclause. + * (If the mergeclause is a true equivalence clause these will be the + * same EquivalenceClass, otherwise not.) If the mergeclause is either + * used to generate an EquivalenceClass, or derived from an EquivalenceClass, + * then it's easy to set up the left_ec and right_ec members --- otherwise, + * this function should be called to set them up. We will generate new + * EquivalenceClauses if necessary to represent the mergeclause's left and + * right sides. + * + * Note this is called before EC merging is complete, so the links won't + * necessarily point to canonical ECs. Before they are actually used for + * anything, update_mergeclause_eclasses must be called to ensure that + * they've been updated to point to canonical ECs. + */ +void +initialize_mergeclause_eclasses(PlannerInfo *root, RestrictInfo *restrictinfo) +{ + Expr *clause = restrictinfo->clause; + Oid lefttype, + righttype; + + /* Should be a mergeclause ... */ + Assert(restrictinfo->mergeopfamilies != NIL); + /* ... with links not yet set */ + Assert(restrictinfo->left_ec == NULL); + Assert(restrictinfo->right_ec == NULL); + + /* Need the declared input types of the operator */ + op_input_types(((OpExpr *) clause)->opno, &lefttype, &righttype); + + /* Find or create a matching EquivalenceClass for each side */ + restrictinfo->left_ec = + get_eclass_for_sort_expr(root, + (Expr *) get_leftop(clause), + restrictinfo->nullable_relids, + restrictinfo->mergeopfamilies, + lefttype, + ((OpExpr *) clause)->inputcollid, + 0, + NULL, + true); + restrictinfo->right_ec = + get_eclass_for_sort_expr(root, + (Expr *) get_rightop(clause), + restrictinfo->nullable_relids, + restrictinfo->mergeopfamilies, + righttype, + ((OpExpr *) clause)->inputcollid, + 0, + NULL, + true); +} + +/* + * update_mergeclause_eclasses + * Make the cached EquivalenceClass links valid in a mergeclause + * restrictinfo. + * + * These pointers should have been set by process_equivalence or + * initialize_mergeclause_eclasses, but they might have been set to + * non-canonical ECs that got merged later. Chase up to the canonical + * merged parent if so. + */ +void +update_mergeclause_eclasses(PlannerInfo *root, RestrictInfo *restrictinfo) +{ + /* Should be a merge clause ... */ + Assert(restrictinfo->mergeopfamilies != NIL); + /* ... with pointers already set */ + Assert(restrictinfo->left_ec != NULL); + Assert(restrictinfo->right_ec != NULL); + + /* Chase up to the top as needed */ + while (restrictinfo->left_ec->ec_merged) + restrictinfo->left_ec = restrictinfo->left_ec->ec_merged; + while (restrictinfo->right_ec->ec_merged) + restrictinfo->right_ec = restrictinfo->right_ec->ec_merged; +} + +/* + * find_mergeclauses_for_outer_pathkeys + * This routine attempts to find a list of mergeclauses that can be + * used with a specified ordering for the join's outer relation. + * If successful, it returns a list of mergeclauses. + * + * 'pathkeys' is a pathkeys list showing the ordering of an outer-rel path. + * 'restrictinfos' is a list of mergejoinable restriction clauses for the + * join relation being formed, in no particular order. + * + * The restrictinfos must be marked (via outer_is_left) to show which side + * of each clause is associated with the current outer path. (See + * select_mergejoin_clauses()) + * + * The result is NIL if no merge can be done, else a maximal list of + * usable mergeclauses (represented as a list of their restrictinfo nodes). + * The list is ordered to match the pathkeys, as required for execution. + */ +List * +find_mergeclauses_for_outer_pathkeys(PlannerInfo *root, + List *pathkeys, + List *restrictinfos) +{ + List *mergeclauses = NIL; + ListCell *i; + + /* make sure we have eclasses cached in the clauses */ + foreach(i, restrictinfos) + { + RestrictInfo *rinfo = (RestrictInfo *) lfirst(i); + + update_mergeclause_eclasses(root, rinfo); + } + + foreach(i, pathkeys) + { + PathKey *pathkey = (PathKey *) lfirst(i); + EquivalenceClass *pathkey_ec = pathkey->pk_eclass; + List *matched_restrictinfos = NIL; + ListCell *j; + + /*---------- + * A mergejoin clause matches a pathkey if it has the same EC. + * If there are multiple matching clauses, take them all. In plain + * inner-join scenarios we expect only one match, because + * equivalence-class processing will have removed any redundant + * mergeclauses. However, in outer-join scenarios there might be + * multiple matches. An example is + * + * select * from a full join b + * on a.v1 = b.v1 and a.v2 = b.v2 and a.v1 = b.v2; + * + * Given the pathkeys ({a.v1}, {a.v2}) it is okay to return all three + * clauses (in the order a.v1=b.v1, a.v1=b.v2, a.v2=b.v2) and indeed + * we *must* do so or we will be unable to form a valid plan. + * + * We expect that the given pathkeys list is canonical, which means + * no two members have the same EC, so it's not possible for this + * code to enter the same mergeclause into the result list twice. + * + * It's possible that multiple matching clauses might have different + * ECs on the other side, in which case the order we put them into our + * result makes a difference in the pathkeys required for the inner + * input rel. However this routine hasn't got any info about which + * order would be best, so we don't worry about that. + * + * It's also possible that the selected mergejoin clauses produce + * a noncanonical ordering of pathkeys for the inner side, ie, we + * might select clauses that reference b.v1, b.v2, b.v1 in that + * order. This is not harmful in itself, though it suggests that + * the clauses are partially redundant. Since the alternative is + * to omit mergejoin clauses and thereby possibly fail to generate a + * plan altogether, we live with it. make_inner_pathkeys_for_merge() + * has to delete duplicates when it constructs the inner pathkeys + * list, and we also have to deal with such cases specially in + * create_mergejoin_plan(). + *---------- + */ + foreach(j, restrictinfos) + { + RestrictInfo *rinfo = (RestrictInfo *) lfirst(j); + EquivalenceClass *clause_ec; + + clause_ec = rinfo->outer_is_left ? + rinfo->left_ec : rinfo->right_ec; + if (clause_ec == pathkey_ec) + matched_restrictinfos = lappend(matched_restrictinfos, rinfo); + } + + /* + * If we didn't find a mergeclause, we're done --- any additional + * sort-key positions in the pathkeys are useless. (But we can still + * mergejoin if we found at least one mergeclause.) + */ + if (matched_restrictinfos == NIL) + break; + + /* + * If we did find usable mergeclause(s) for this sort-key position, + * add them to result list. + */ + mergeclauses = list_concat(mergeclauses, matched_restrictinfos); + } + + return mergeclauses; +} + +/* + * select_outer_pathkeys_for_merge + * Builds a pathkey list representing a possible sort ordering + * that can be used with the given mergeclauses. + * + * 'mergeclauses' is a list of RestrictInfos for mergejoin clauses + * that will be used in a merge join. + * 'joinrel' is the join relation we are trying to construct. + * + * The restrictinfos must be marked (via outer_is_left) to show which side + * of each clause is associated with the current outer path. (See + * select_mergejoin_clauses()) + * + * Returns a pathkeys list that can be applied to the outer relation. + * + * Since we assume here that a sort is required, there is no particular use + * in matching any available ordering of the outerrel. (joinpath.c has an + * entirely separate code path for considering sort-free mergejoins.) Rather, + * it's interesting to try to match the requested query_pathkeys so that a + * second output sort may be avoided; and failing that, we try to list "more + * popular" keys (those with the most unmatched EquivalenceClass peers) + * earlier, in hopes of making the resulting ordering useful for as many + * higher-level mergejoins as possible. + */ +List * +select_outer_pathkeys_for_merge(PlannerInfo *root, + List *mergeclauses, + RelOptInfo *joinrel) +{ + List *pathkeys = NIL; + int nClauses = list_length(mergeclauses); + EquivalenceClass **ecs; + int *scores; + int necs; + ListCell *lc; + int j; + + /* Might have no mergeclauses */ + if (nClauses == 0) + return NIL; + + /* + * Make arrays of the ECs used by the mergeclauses (dropping any + * duplicates) and their "popularity" scores. + */ + ecs = (EquivalenceClass **) palloc(nClauses * sizeof(EquivalenceClass *)); + scores = (int *) palloc(nClauses * sizeof(int)); + necs = 0; + + foreach(lc, mergeclauses) + { + RestrictInfo *rinfo = (RestrictInfo *) lfirst(lc); + EquivalenceClass *oeclass; + int score; + ListCell *lc2; + + /* get the outer eclass */ + update_mergeclause_eclasses(root, rinfo); + + if (rinfo->outer_is_left) + oeclass = rinfo->left_ec; + else + oeclass = rinfo->right_ec; + + /* reject duplicates */ + for (j = 0; j < necs; j++) + { + if (ecs[j] == oeclass) + break; + } + if (j < necs) + continue; + + /* compute score */ + score = 0; + foreach(lc2, oeclass->ec_members) + { + EquivalenceMember *em = (EquivalenceMember *) lfirst(lc2); + + /* Potential future join partner? */ + if (!em->em_is_const && !em->em_is_child && + !bms_overlap(em->em_relids, joinrel->relids)) + score++; + } + + ecs[necs] = oeclass; + scores[necs] = score; + necs++; + } + + /* + * Find out if we have all the ECs mentioned in query_pathkeys; if so we + * can generate a sort order that's also useful for final output. There is + * no percentage in a partial match, though, so we have to have 'em all. + */ + if (root->query_pathkeys) + { + foreach(lc, root->query_pathkeys) + { + PathKey *query_pathkey = (PathKey *) lfirst(lc); + EquivalenceClass *query_ec = query_pathkey->pk_eclass; + + for (j = 0; j < necs; j++) + { + if (ecs[j] == query_ec) + break; /* found match */ + } + if (j >= necs) + break; /* didn't find match */ + } + /* if we got to the end of the list, we have them all */ + if (lc == NULL) + { + /* copy query_pathkeys as starting point for our output */ + pathkeys = list_copy(root->query_pathkeys); + /* mark their ECs as already-emitted */ + foreach(lc, root->query_pathkeys) + { + PathKey *query_pathkey = (PathKey *) lfirst(lc); + EquivalenceClass *query_ec = query_pathkey->pk_eclass; + + for (j = 0; j < necs; j++) + { + if (ecs[j] == query_ec) + { + scores[j] = -1; + break; + } + } + } + } + } + + /* + * Add remaining ECs to the list in popularity order, using a default sort + * ordering. (We could use qsort() here, but the list length is usually + * so small it's not worth it.) + */ + for (;;) + { + int best_j; + int best_score; + EquivalenceClass *ec; + PathKey *pathkey; + + best_j = 0; + best_score = scores[0]; + for (j = 1; j < necs; j++) + { + if (scores[j] > best_score) + { + best_j = j; + best_score = scores[j]; + } + } + if (best_score < 0) + break; /* all done */ + ec = ecs[best_j]; + scores[best_j] = -1; + pathkey = make_canonical_pathkey(root, + ec, + linitial_oid(ec->ec_opfamilies), + BTLessStrategyNumber, + false); + /* can't be redundant because no duplicate ECs */ + Assert(!pathkey_is_redundant(pathkey, pathkeys)); + pathkeys = lappend(pathkeys, pathkey); + } + + pfree(ecs); + pfree(scores); + + return pathkeys; +} + +/* + * make_inner_pathkeys_for_merge + * Builds a pathkey list representing the explicit sort order that + * must be applied to an inner path to make it usable with the + * given mergeclauses. + * + * 'mergeclauses' is a list of RestrictInfos for the mergejoin clauses + * that will be used in a merge join, in order. + * 'outer_pathkeys' are the already-known canonical pathkeys for the outer + * side of the join. + * + * The restrictinfos must be marked (via outer_is_left) to show which side + * of each clause is associated with the current outer path. (See + * select_mergejoin_clauses()) + * + * Returns a pathkeys list that can be applied to the inner relation. + * + * Note that it is not this routine's job to decide whether sorting is + * actually needed for a particular input path. Assume a sort is necessary; + * just make the keys, eh? + */ +List * +make_inner_pathkeys_for_merge(PlannerInfo *root, + List *mergeclauses, + List *outer_pathkeys) +{ + List *pathkeys = NIL; + EquivalenceClass *lastoeclass; + PathKey *opathkey; + ListCell *lc; + ListCell *lop; + + lastoeclass = NULL; + opathkey = NULL; + lop = list_head(outer_pathkeys); + + foreach(lc, mergeclauses) + { + RestrictInfo *rinfo = (RestrictInfo *) lfirst(lc); + EquivalenceClass *oeclass; + EquivalenceClass *ieclass; + PathKey *pathkey; + + update_mergeclause_eclasses(root, rinfo); + + if (rinfo->outer_is_left) + { + oeclass = rinfo->left_ec; + ieclass = rinfo->right_ec; + } + else + { + oeclass = rinfo->right_ec; + ieclass = rinfo->left_ec; + } + + /* outer eclass should match current or next pathkeys */ + /* we check this carefully for debugging reasons */ + if (oeclass != lastoeclass) + { + if (!lop) + elog(ERROR, "too few pathkeys for mergeclauses"); + opathkey = (PathKey *) lfirst(lop); + lop = lnext(outer_pathkeys, lop); + lastoeclass = opathkey->pk_eclass; + if (oeclass != lastoeclass) + elog(ERROR, "outer pathkeys do not match mergeclause"); + } + + /* + * Often, we'll have same EC on both sides, in which case the outer + * pathkey is also canonical for the inner side, and we can skip a + * useless search. + */ + if (ieclass == oeclass) + pathkey = opathkey; + else + pathkey = make_canonical_pathkey(root, + ieclass, + opathkey->pk_opfamily, + opathkey->pk_strategy, + opathkey->pk_nulls_first); + + /* + * Don't generate redundant pathkeys (which can happen if multiple + * mergeclauses refer to the same EC). Because we do this, the output + * pathkey list isn't necessarily ordered like the mergeclauses, which + * complicates life for create_mergejoin_plan(). But if we didn't, + * we'd have a noncanonical sort key list, which would be bad; for one + * reason, it certainly wouldn't match any available sort order for + * the input relation. + */ + if (!pathkey_is_redundant(pathkey, pathkeys)) + pathkeys = lappend(pathkeys, pathkey); + } + + return pathkeys; +} + +/* + * trim_mergeclauses_for_inner_pathkeys + * This routine trims a list of mergeclauses to include just those that + * work with a specified ordering for the join's inner relation. + * + * 'mergeclauses' is a list of RestrictInfos for mergejoin clauses for the + * join relation being formed, in an order known to work for the + * currently-considered sort ordering of the join's outer rel. + * 'pathkeys' is a pathkeys list showing the ordering of an inner-rel path; + * it should be equal to, or a truncation of, the result of + * make_inner_pathkeys_for_merge for these mergeclauses. + * + * What we return will be a prefix of the given mergeclauses list. + * + * We need this logic because make_inner_pathkeys_for_merge's result isn't + * necessarily in the same order as the mergeclauses. That means that if we + * consider an inner-rel pathkey list that is a truncation of that result, + * we might need to drop mergeclauses even though they match a surviving inner + * pathkey. This happens when they are to the right of a mergeclause that + * matches a removed inner pathkey. + * + * The mergeclauses must be marked (via outer_is_left) to show which side + * of each clause is associated with the current outer path. (See + * select_mergejoin_clauses()) + */ +List * +trim_mergeclauses_for_inner_pathkeys(PlannerInfo *root, + List *mergeclauses, + List *pathkeys) +{ + List *new_mergeclauses = NIL; + PathKey *pathkey; + EquivalenceClass *pathkey_ec; + bool matched_pathkey; + ListCell *lip; + ListCell *i; + + /* No pathkeys => no mergeclauses (though we don't expect this case) */ + if (pathkeys == NIL) + return NIL; + /* Initialize to consider first pathkey */ + lip = list_head(pathkeys); + pathkey = (PathKey *) lfirst(lip); + pathkey_ec = pathkey->pk_eclass; + lip = lnext(pathkeys, lip); + matched_pathkey = false; + + /* Scan mergeclauses to see how many we can use */ + foreach(i, mergeclauses) + { + RestrictInfo *rinfo = (RestrictInfo *) lfirst(i); + EquivalenceClass *clause_ec; + + /* Assume we needn't do update_mergeclause_eclasses again here */ + + /* Check clause's inner-rel EC against current pathkey */ + clause_ec = rinfo->outer_is_left ? + rinfo->right_ec : rinfo->left_ec; + + /* If we don't have a match, attempt to advance to next pathkey */ + if (clause_ec != pathkey_ec) + { + /* If we had no clauses matching this inner pathkey, must stop */ + if (!matched_pathkey) + break; + + /* Advance to next inner pathkey, if any */ + if (lip == NULL) + break; + pathkey = (PathKey *) lfirst(lip); + pathkey_ec = pathkey->pk_eclass; + lip = lnext(pathkeys, lip); + matched_pathkey = false; + } + + /* If mergeclause matches current inner pathkey, we can use it */ + if (clause_ec == pathkey_ec) + { + new_mergeclauses = lappend(new_mergeclauses, rinfo); + matched_pathkey = true; + } + else + { + /* Else, no hope of adding any more mergeclauses */ + break; + } + } + + return new_mergeclauses; +} + + +/**************************************************************************** + * PATHKEY USEFULNESS CHECKS + * + * We only want to remember as many of the pathkeys of a path as have some + * potential use, either for subsequent mergejoins or for meeting the query's + * requested output ordering. This ensures that add_path() won't consider + * a path to have a usefully different ordering unless it really is useful. + * These routines check for usefulness of given pathkeys. + ****************************************************************************/ + +/* + * pathkeys_useful_for_merging + * Count the number of pathkeys that may be useful for mergejoins + * above the given relation. + * + * We consider a pathkey potentially useful if it corresponds to the merge + * ordering of either side of any joinclause for the rel. This might be + * overoptimistic, since joinclauses that require different other relations + * might never be usable at the same time, but trying to be exact is likely + * to be more trouble than it's worth. + * + * To avoid doubling the number of mergejoin paths considered, we would like + * to consider only one of the two scan directions (ASC or DESC) as useful + * for merging for any given target column. The choice is arbitrary unless + * one of the directions happens to match an ORDER BY key, in which case + * that direction should be preferred, in hopes of avoiding a final sort step. + * right_merge_direction() implements this heuristic. + */ +static int +pathkeys_useful_for_merging(PlannerInfo *root, RelOptInfo *rel, List *pathkeys) +{ + int useful = 0; + ListCell *i; + + foreach(i, pathkeys) + { + PathKey *pathkey = (PathKey *) lfirst(i); + bool matched = false; + ListCell *j; + + /* If "wrong" direction, not useful for merging */ + if (!right_merge_direction(root, pathkey)) + break; + + /* + * First look into the EquivalenceClass of the pathkey, to see if + * there are any members not yet joined to the rel. If so, it's + * surely possible to generate a mergejoin clause using them. + */ + if (rel->has_eclass_joins && + eclass_useful_for_merging(root, pathkey->pk_eclass, rel)) + matched = true; + else + { + /* + * Otherwise search the rel's joininfo list, which contains + * non-EquivalenceClass-derivable join clauses that might + * nonetheless be mergejoinable. + */ + foreach(j, rel->joininfo) + { + RestrictInfo *restrictinfo = (RestrictInfo *) lfirst(j); + + if (restrictinfo->mergeopfamilies == NIL) + continue; + update_mergeclause_eclasses(root, restrictinfo); + + if (pathkey->pk_eclass == restrictinfo->left_ec || + pathkey->pk_eclass == restrictinfo->right_ec) + { + matched = true; + break; + } + } + } + + /* + * If we didn't find a mergeclause, we're done --- any additional + * sort-key positions in the pathkeys are useless. (But we can still + * mergejoin if we found at least one mergeclause.) + */ + if (matched) + useful++; + else + break; + } + + return useful; +} + +/* + * right_merge_direction + * Check whether the pathkey embodies the preferred sort direction + * for merging its target column. + */ +static bool +right_merge_direction(PlannerInfo *root, PathKey *pathkey) +{ + ListCell *l; + + foreach(l, root->query_pathkeys) + { + PathKey *query_pathkey = (PathKey *) lfirst(l); + + if (pathkey->pk_eclass == query_pathkey->pk_eclass && + pathkey->pk_opfamily == query_pathkey->pk_opfamily) + { + /* + * Found a matching query sort column. Prefer this pathkey's + * direction iff it matches. Note that we ignore pk_nulls_first, + * which means that a sort might be needed anyway ... but we still + * want to prefer only one of the two possible directions, and we + * might as well use this one. + */ + return (pathkey->pk_strategy == query_pathkey->pk_strategy); + } + } + + /* If no matching ORDER BY request, prefer the ASC direction */ + return (pathkey->pk_strategy == BTLessStrategyNumber); +} + +/* + * pathkeys_useful_for_ordering + * Count the number of pathkeys that are useful for meeting the + * query's requested output ordering. + * + * Because we the have the possibility of incremental sort, a prefix list of + * keys is potentially useful for improving the performance of the requested + * ordering. Thus we return 0, if no valuable keys are found, or the number + * of leading keys shared by the list and the requested ordering.. + */ +static int +pathkeys_useful_for_ordering(PlannerInfo *root, List *pathkeys) +{ + int n_common_pathkeys; + + if (root->query_pathkeys == NIL) + return 0; /* no special ordering requested */ + + if (pathkeys == NIL) + return 0; /* unordered path */ + + (void) pathkeys_count_contained_in(root->query_pathkeys, pathkeys, + &n_common_pathkeys); + + return n_common_pathkeys; +} + +/* + * truncate_useless_pathkeys + * Shorten the given pathkey list to just the useful pathkeys. + */ +List * +truncate_useless_pathkeys(PlannerInfo *root, + RelOptInfo *rel, + List *pathkeys) +{ + int nuseful; + int nuseful2; + + nuseful = pathkeys_useful_for_merging(root, rel, pathkeys); + nuseful2 = pathkeys_useful_for_ordering(root, pathkeys); + if (nuseful2 > nuseful) + nuseful = nuseful2; + + /* + * Note: not safe to modify input list destructively, but we can avoid + * copying the list if we're not actually going to change it + */ + if (nuseful == 0) + return NIL; + else if (nuseful == list_length(pathkeys)) + return pathkeys; + else + return list_truncate(list_copy(pathkeys), nuseful); +} + +/* + * has_useful_pathkeys + * Detect whether the specified rel could have any pathkeys that are + * useful according to truncate_useless_pathkeys(). + * + * This is a cheap test that lets us skip building pathkeys at all in very + * simple queries. It's OK to err in the direction of returning "true" when + * there really aren't any usable pathkeys, but erring in the other direction + * is bad --- so keep this in sync with the routines above! + * + * We could make the test more complex, for example checking to see if any of + * the joinclauses are really mergejoinable, but that likely wouldn't win + * often enough to repay the extra cycles. Queries with neither a join nor + * a sort are reasonably common, though, so this much work seems worthwhile. + */ +bool +has_useful_pathkeys(PlannerInfo *root, RelOptInfo *rel) +{ + if (rel->joininfo != NIL || rel->has_eclass_joins) + return true; /* might be able to use pathkeys for merging */ + if (root->query_pathkeys != NIL) + return true; /* might be able to use them for ordering */ + return false; /* definitely useless */ +} diff --git a/src/backend/optimizer/path/tidpath.c b/src/backend/optimizer/path/tidpath.c new file mode 100644 index 0000000..279ca1f --- /dev/null +++ b/src/backend/optimizer/path/tidpath.c @@ -0,0 +1,528 @@ +/*------------------------------------------------------------------------- + * + * tidpath.c + * Routines to determine which TID conditions are usable for scanning + * a given relation, and create TidPaths and TidRangePaths accordingly. + * + * For TidPaths, we look for WHERE conditions of the form + * "CTID = pseudoconstant", which can be implemented by just fetching + * the tuple directly via heap_fetch(). We can also handle OR'd conditions + * such as (CTID = const1) OR (CTID = const2), as well as ScalarArrayOpExpr + * conditions of the form CTID = ANY(pseudoconstant_array). In particular + * this allows + * WHERE ctid IN (tid1, tid2, ...) + * + * As with indexscans, our definition of "pseudoconstant" is pretty liberal: + * we allow anything that doesn't involve a volatile function or a Var of + * the relation under consideration. Vars belonging to other relations of + * the query are allowed, giving rise to parameterized TID scans. + * + * We also support "WHERE CURRENT OF cursor" conditions (CurrentOfExpr), + * which amount to "CTID = run-time-determined-TID". These could in + * theory be translated to a simple comparison of CTID to the result of + * a function, but in practice it works better to keep the special node + * representation all the way through to execution. + * + * Additionally, TidRangePaths may be created for conditions of the form + * "CTID relop pseudoconstant", where relop is one of >,>=,<,<=, and + * AND-clauses composed of such conditions. + * + * Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group + * Portions Copyright (c) 1994, Regents of the University of California + * + * + * IDENTIFICATION + * src/backend/optimizer/path/tidpath.c + * + *------------------------------------------------------------------------- + */ +#include "postgres.h" + +#include "access/sysattr.h" +#include "catalog/pg_operator.h" +#include "catalog/pg_type.h" +#include "nodes/nodeFuncs.h" +#include "optimizer/clauses.h" +#include "optimizer/optimizer.h" +#include "optimizer/pathnode.h" +#include "optimizer/paths.h" +#include "optimizer/restrictinfo.h" + + +/* + * Does this Var represent the CTID column of the specified baserel? + */ +static inline bool +IsCTIDVar(Var *var, RelOptInfo *rel) +{ + /* The vartype check is strictly paranoia */ + if (var->varattno == SelfItemPointerAttributeNumber && + var->vartype == TIDOID && + var->varno == rel->relid && + var->varlevelsup == 0) + return true; + return false; +} + +/* + * Check to see if a RestrictInfo is of the form + * CTID OP pseudoconstant + * or + * pseudoconstant OP CTID + * where OP is a binary operation, the CTID Var belongs to relation "rel", + * and nothing on the other side of the clause does. + */ +static bool +IsBinaryTidClause(RestrictInfo *rinfo, RelOptInfo *rel) +{ + OpExpr *node; + Node *arg1, + *arg2, + *other; + Relids other_relids; + + /* Must be an OpExpr */ + if (!is_opclause(rinfo->clause)) + return false; + node = (OpExpr *) rinfo->clause; + + /* OpExpr must have two arguments */ + if (list_length(node->args) != 2) + return false; + arg1 = linitial(node->args); + arg2 = lsecond(node->args); + + /* Look for CTID as either argument */ + other = NULL; + other_relids = NULL; + if (arg1 && IsA(arg1, Var) && + IsCTIDVar((Var *) arg1, rel)) + { + other = arg2; + other_relids = rinfo->right_relids; + } + if (!other && arg2 && IsA(arg2, Var) && + IsCTIDVar((Var *) arg2, rel)) + { + other = arg1; + other_relids = rinfo->left_relids; + } + if (!other) + return false; + + /* The other argument must be a pseudoconstant */ + if (bms_is_member(rel->relid, other_relids) || + contain_volatile_functions(other)) + return false; + + return true; /* success */ +} + +/* + * Check to see if a RestrictInfo is of the form + * CTID = pseudoconstant + * or + * pseudoconstant = CTID + * where the CTID Var belongs to relation "rel", and nothing on the + * other side of the clause does. + */ +static bool +IsTidEqualClause(RestrictInfo *rinfo, RelOptInfo *rel) +{ + if (!IsBinaryTidClause(rinfo, rel)) + return false; + + if (((OpExpr *) rinfo->clause)->opno == TIDEqualOperator) + return true; + + return false; +} + +/* + * Check to see if a RestrictInfo is of the form + * CTID OP pseudoconstant + * or + * pseudoconstant OP CTID + * where OP is a range operator such as <, <=, >, or >=, the CTID Var belongs + * to relation "rel", and nothing on the other side of the clause does. + */ +static bool +IsTidRangeClause(RestrictInfo *rinfo, RelOptInfo *rel) +{ + Oid opno; + + if (!IsBinaryTidClause(rinfo, rel)) + return false; + opno = ((OpExpr *) rinfo->clause)->opno; + + if (opno == TIDLessOperator || opno == TIDLessEqOperator || + opno == TIDGreaterOperator || opno == TIDGreaterEqOperator) + return true; + + return false; +} + +/* + * Check to see if a RestrictInfo is of the form + * CTID = ANY (pseudoconstant_array) + * where the CTID Var belongs to relation "rel", and nothing on the + * other side of the clause does. + */ +static bool +IsTidEqualAnyClause(PlannerInfo *root, RestrictInfo *rinfo, RelOptInfo *rel) +{ + ScalarArrayOpExpr *node; + Node *arg1, + *arg2; + + /* Must be a ScalarArrayOpExpr */ + if (!(rinfo->clause && IsA(rinfo->clause, ScalarArrayOpExpr))) + return false; + node = (ScalarArrayOpExpr *) rinfo->clause; + + /* Operator must be tideq */ + if (node->opno != TIDEqualOperator) + return false; + if (!node->useOr) + return false; + Assert(list_length(node->args) == 2); + arg1 = linitial(node->args); + arg2 = lsecond(node->args); + + /* CTID must be first argument */ + if (arg1 && IsA(arg1, Var) && + IsCTIDVar((Var *) arg1, rel)) + { + /* The other argument must be a pseudoconstant */ + if (bms_is_member(rel->relid, pull_varnos(root, arg2)) || + contain_volatile_functions(arg2)) + return false; + + return true; /* success */ + } + + return false; +} + +/* + * Check to see if a RestrictInfo is a CurrentOfExpr referencing "rel". + */ +static bool +IsCurrentOfClause(RestrictInfo *rinfo, RelOptInfo *rel) +{ + CurrentOfExpr *node; + + /* Must be a CurrentOfExpr */ + if (!(rinfo->clause && IsA(rinfo->clause, CurrentOfExpr))) + return false; + node = (CurrentOfExpr *) rinfo->clause; + + /* If it references this rel, we're good */ + if (node->cvarno == rel->relid) + return true; + + return false; +} + +/* + * Extract a set of CTID conditions from the given RestrictInfo + * + * Returns a List of CTID qual RestrictInfos for the specified rel (with + * implicit OR semantics across the list), or NIL if there are no usable + * conditions. + * + * This function considers only base cases; AND/OR combination is handled + * below. Therefore the returned List never has more than one element. + * (Using a List may seem a bit weird, but it simplifies the caller.) + */ +static List * +TidQualFromRestrictInfo(PlannerInfo *root, RestrictInfo *rinfo, RelOptInfo *rel) +{ + /* + * We may ignore pseudoconstant clauses (they can't contain Vars, so could + * not match anyway). + */ + if (rinfo->pseudoconstant) + return NIL; + + /* + * If clause must wait till after some lower-security-level restriction + * clause, reject it. + */ + if (!restriction_is_securely_promotable(rinfo, rel)) + return NIL; + + /* + * Check all base cases. If we get a match, return the clause. + */ + if (IsTidEqualClause(rinfo, rel) || + IsTidEqualAnyClause(root, rinfo, rel) || + IsCurrentOfClause(rinfo, rel)) + return list_make1(rinfo); + + return NIL; +} + +/* + * Extract a set of CTID conditions from implicit-AND List of RestrictInfos + * + * Returns a List of CTID qual RestrictInfos for the specified rel (with + * implicit OR semantics across the list), or NIL if there are no usable + * equality conditions. + * + * This function is just concerned with handling AND/OR recursion. + */ +static List * +TidQualFromRestrictInfoList(PlannerInfo *root, List *rlist, RelOptInfo *rel) +{ + List *rlst = NIL; + ListCell *l; + + foreach(l, rlist) + { + RestrictInfo *rinfo = lfirst_node(RestrictInfo, l); + + if (restriction_is_or_clause(rinfo)) + { + ListCell *j; + + /* + * We must be able to extract a CTID condition from every + * sub-clause of an OR, or we can't use it. + */ + foreach(j, ((BoolExpr *) rinfo->orclause)->args) + { + Node *orarg = (Node *) lfirst(j); + List *sublist; + + /* OR arguments should be ANDs or sub-RestrictInfos */ + if (is_andclause(orarg)) + { + List *andargs = ((BoolExpr *) orarg)->args; + + /* Recurse in case there are sub-ORs */ + sublist = TidQualFromRestrictInfoList(root, andargs, rel); + } + else + { + RestrictInfo *rinfo = castNode(RestrictInfo, orarg); + + Assert(!restriction_is_or_clause(rinfo)); + sublist = TidQualFromRestrictInfo(root, rinfo, rel); + } + + /* + * If nothing found in this arm, we can't do anything with + * this OR clause. + */ + if (sublist == NIL) + { + rlst = NIL; /* forget anything we had */ + break; /* out of loop over OR args */ + } + + /* + * OK, continue constructing implicitly-OR'ed result list. + */ + rlst = list_concat(rlst, sublist); + } + } + else + { + /* Not an OR clause, so handle base cases */ + rlst = TidQualFromRestrictInfo(root, rinfo, rel); + } + + /* + * Stop as soon as we find any usable CTID condition. In theory we + * could get CTID equality conditions from different AND'ed clauses, + * in which case we could try to pick the most efficient one. In + * practice, such usage seems very unlikely, so we don't bother; we + * just exit as soon as we find the first candidate. + */ + if (rlst) + break; + } + + return rlst; +} + +/* + * Extract a set of CTID range conditions from implicit-AND List of RestrictInfos + * + * Returns a List of CTID range qual RestrictInfos for the specified rel + * (with implicit AND semantics across the list), or NIL if there are no + * usable range conditions or if the rel's table AM does not support TID range + * scans. + */ +static List * +TidRangeQualFromRestrictInfoList(List *rlist, RelOptInfo *rel) +{ + List *rlst = NIL; + ListCell *l; + + if ((rel->amflags & AMFLAG_HAS_TID_RANGE) == 0) + return NIL; + + foreach(l, rlist) + { + RestrictInfo *rinfo = lfirst_node(RestrictInfo, l); + + if (IsTidRangeClause(rinfo, rel)) + rlst = lappend(rlst, rinfo); + } + + return rlst; +} + +/* + * Given a list of join clauses involving our rel, create a parameterized + * TidPath for each one that is a suitable TidEqual clause. + * + * In principle we could combine clauses that reference the same outer rels, + * but it doesn't seem like such cases would arise often enough to be worth + * troubling over. + */ +static void +BuildParameterizedTidPaths(PlannerInfo *root, RelOptInfo *rel, List *clauses) +{ + ListCell *l; + + foreach(l, clauses) + { + RestrictInfo *rinfo = lfirst_node(RestrictInfo, l); + List *tidquals; + Relids required_outer; + + /* + * Validate whether each clause is actually usable; we must check this + * even when examining clauses generated from an EquivalenceClass, + * since they might not satisfy the restriction on not having Vars of + * our rel on the other side, or somebody might've built an operator + * class that accepts type "tid" but has other operators in it. + * + * We currently consider only TidEqual join clauses. In principle we + * might find a suitable ScalarArrayOpExpr in the rel's joininfo list, + * but it seems unlikely to be worth expending the cycles to check. + * And we definitely won't find a CurrentOfExpr here. Hence, we don't + * use TidQualFromRestrictInfo; but this must match that function + * otherwise. + */ + if (rinfo->pseudoconstant || + !restriction_is_securely_promotable(rinfo, rel) || + !IsTidEqualClause(rinfo, rel)) + continue; + + /* + * Check if clause can be moved to this rel; this is probably + * redundant when considering EC-derived clauses, but we must check it + * for "loose" join clauses. + */ + if (!join_clause_is_movable_to(rinfo, rel)) + continue; + + /* OK, make list of clauses for this path */ + tidquals = list_make1(rinfo); + + /* Compute required outer rels for this path */ + required_outer = bms_union(rinfo->required_relids, rel->lateral_relids); + required_outer = bms_del_member(required_outer, rel->relid); + + add_path(rel, (Path *) create_tidscan_path(root, rel, tidquals, + required_outer)); + } +} + +/* + * Test whether an EquivalenceClass member matches our rel's CTID Var. + * + * This is a callback for use by generate_implied_equalities_for_column. + */ +static bool +ec_member_matches_ctid(PlannerInfo *root, RelOptInfo *rel, + EquivalenceClass *ec, EquivalenceMember *em, + void *arg) +{ + if (em->em_expr && IsA(em->em_expr, Var) && + IsCTIDVar((Var *) em->em_expr, rel)) + return true; + return false; +} + +/* + * create_tidscan_paths + * Create paths corresponding to direct TID scans of the given rel. + * + * Candidate paths are added to the rel's pathlist (using add_path). + */ +void +create_tidscan_paths(PlannerInfo *root, RelOptInfo *rel) +{ + List *tidquals; + List *tidrangequals; + + /* + * If any suitable quals exist in the rel's baserestrict list, generate a + * plain (unparameterized) TidPath with them. + */ + tidquals = TidQualFromRestrictInfoList(root, rel->baserestrictinfo, rel); + + if (tidquals != NIL) + { + /* + * This path uses no join clauses, but it could still have required + * parameterization due to LATERAL refs in its tlist. + */ + Relids required_outer = rel->lateral_relids; + + add_path(rel, (Path *) create_tidscan_path(root, rel, tidquals, + required_outer)); + } + + /* + * If there are range quals in the baserestrict list, generate a + * TidRangePath. + */ + tidrangequals = TidRangeQualFromRestrictInfoList(rel->baserestrictinfo, + rel); + + if (tidrangequals != NIL) + { + /* + * This path uses no join clauses, but it could still have required + * parameterization due to LATERAL refs in its tlist. + */ + Relids required_outer = rel->lateral_relids; + + add_path(rel, (Path *) create_tidrangescan_path(root, rel, + tidrangequals, + required_outer)); + } + + /* + * Try to generate parameterized TidPaths using equality clauses extracted + * from EquivalenceClasses. (This is important since simple "t1.ctid = + * t2.ctid" clauses will turn into ECs.) + */ + if (rel->has_eclass_joins) + { + List *clauses; + + /* Generate clauses, skipping any that join to lateral_referencers */ + clauses = generate_implied_equalities_for_column(root, + rel, + ec_member_matches_ctid, + NULL, + rel->lateral_referencers); + + /* Generate a path for each usable join clause */ + BuildParameterizedTidPaths(root, rel, clauses); + } + + /* + * Also consider parameterized TidPaths using "loose" join quals. Quals + * of the form "t1.ctid = t2.ctid" would turn into these if they are outer + * join quals, for example. + */ + BuildParameterizedTidPaths(root, rel, rel->joininfo); +} diff --git a/src/backend/optimizer/plan/Makefile b/src/backend/optimizer/plan/Makefile new file mode 100644 index 0000000..80ef162 --- /dev/null +++ b/src/backend/optimizer/plan/Makefile @@ -0,0 +1,25 @@ +#------------------------------------------------------------------------- +# +# Makefile-- +# Makefile for optimizer/plan +# +# IDENTIFICATION +# src/backend/optimizer/plan/Makefile +# +#------------------------------------------------------------------------- + +subdir = src/backend/optimizer/plan +top_builddir = ../../../.. +include $(top_builddir)/src/Makefile.global + +OBJS = \ + analyzejoins.o \ + createplan.o \ + initsplan.o \ + planagg.o \ + planmain.o \ + planner.o \ + setrefs.o \ + subselect.o + +include $(top_srcdir)/src/backend/common.mk diff --git a/src/backend/optimizer/plan/README b/src/backend/optimizer/plan/README new file mode 100644 index 0000000..013c0f9 --- /dev/null +++ b/src/backend/optimizer/plan/README @@ -0,0 +1,158 @@ +src/backend/optimizer/plan/README + +Subselects +========== + +Vadim B. Mikheev + + +From owner-pgsql-hackers@hub.org Fri Feb 13 09:01:19 1998 +Received: from renoir.op.net (root@renoir.op.net [209.152.193.4]) + by candle.pha.pa.us (8.8.5/8.8.5) with ESMTP id JAA11576 + for <maillist@candle.pha.pa.us>; Fri, 13 Feb 1998 09:01:17 -0500 (EST) +Received: from hub.org (hub.org [209.47.148.200]) by renoir.op.net (o1/$Revision: 1.14 $) with ESMTP id IAA09761 for <maillist@candle.pha.pa.us>; Fri, 13 Feb 1998 08:41:22 -0500 (EST) +Received: from localhost (majordom@localhost) by hub.org (8.8.8/8.7.5) with SMTP id IAA08135; Fri, 13 Feb 1998 08:40:17 -0500 (EST) +Received: by hub.org (TLB v0.10a (1.23 tibbs 1997/01/09 00:29:32)); Fri, 13 Feb 1998 08:38:42 -0500 (EST) +Received: (from majordom@localhost) by hub.org (8.8.8/8.7.5) id IAA06646 for pgsql-hackers-outgoing; Fri, 13 Feb 1998 08:38:35 -0500 (EST) +Received: from dune.krasnet.ru (dune.krasnet.ru [193.125.44.86]) by hub.org (8.8.8/8.7.5) with ESMTP id IAA04568 for <hackers@postgreSQL.org>; Fri, 13 Feb 1998 08:37:16 -0500 (EST) +Received: from sable.krasnoyarsk.su (dune.krasnet.ru [193.125.44.86]) + by dune.krasnet.ru (8.8.7/8.8.7) with ESMTP id UAA13717 + for <hackers@postgreSQL.org>; Fri, 13 Feb 1998 20:51:03 +0700 (KRS) + (envelope-from vadim@sable.krasnoyarsk.su) +Message-ID: <34E44FBA.D64E7997@sable.krasnoyarsk.su> +Date: Fri, 13 Feb 1998 20:50:50 +0700 +From: "Vadim B. Mikheev" <vadim@sable.krasnoyarsk.su> +Organization: ITTS (Krasnoyarsk) +X-Mailer: Mozilla 4.04 [en] (X11; I; FreeBSD 2.2.5-RELEASE i386) +MIME-Version: 1.0 +To: PostgreSQL Developers List <hackers@postgreSQL.org> +Subject: [HACKERS] Subselects are in CVS... +Content-Type: text/plain; charset=us-ascii +Content-Transfer-Encoding: 7bit +Sender: owner-pgsql-hackers@hub.org +Precedence: bulk +Status: OR + +This is some implementation notes and opened issues... + +First, implementation uses new type of parameters - PARAM_EXEC - to deal +with correlation Vars. When query_planner() is called, it first tries to +replace all upper queries Var referenced in current query with Param of +this type. Some global variables are used to keep mapping of Vars to +Params and Params to Vars. + +After this, all current query' SubLinks are processed: for each SubLink +found in query' qual union_planner() (old planner() function) will be +called to plan corresponding subselect (union_planner() calls +query_planner() for "simple" query and supports UNIONs). After subselect +are planned, optimizer knows about is this correlated, un-correlated or +_undirect_ correlated (references some grand-parent Vars but no parent +ones: uncorrelated from the parent' point of view) query. + +For uncorrelated and undirect correlated subqueries of EXPRession or +EXISTS type SubLinks will be replaced with "normal" clauses from +SubLink->Oper list (I changed this list to be list of EXPR nodes, +not just Oper ones). Right sides of these nodes are replaced with +PARAM_EXEC parameters. This is second use of new parameter type. +At run-time these parameters get value from result of subquery +evaluation (i.e. - from target list of subquery). Execution plan of +subquery itself becomes init plan of parent query. InitPlan knows +what parameters are to get values from subquery' results and will be +executed "on-demand" (for query select * from table where x > 0 and +y > (select max(a) from table_a) subquery will not be executed at all +if there are no tuples with x > 0 _and_ y is not used in index scan). + +SubLinks for subqueries of all other types are transformed into +new type of Expr node - SUBPLAN_EXPR. Expr->args are just correlation +variables from _parent_ query. Expr->oper is new SubPlan node. + +This node is used for InitPlan too. It keeps subquery range table, +indices of Params which are to get value from _parent_ query Vars +(i.e. - from Expr->args), indices of Params into which subquery' +results are to be substituted (this is for InitPlans), SubLink +and subquery' execution plan. + +Plan node was changed to know about dependencies on Params from +parent queries and InitPlans, to keep list of changed Params +(from the above) and so be re-scanned if this list is not NULL. +Also, added list of InitPlans (actually, all of them for current +query are in topmost plan node now) and other SubPlans (from +plan->qual) - to initialize them and let them know about changed +Params (from the list of their "interests"). + +After all SubLinks are processed, query_planner() calls qual' +canonificator and does "normal" work. By using Params optimizer +is mostly unchanged. + +Well, Executor. To get subplans re-evaluated without ExecutorStart() +and ExecutorEnd() (without opening and closing relations and indices +and without many palloc() and pfree() - this is what SQL-funcs does +on each call) ExecReScan() now supports most of Plan types... + +Explanation of EXPLAIN. + +vac=> explain select * from tmp where x >= (select max(x2) from test2 +where y2 = y and exists (select * from tempx where tx = x)); +NOTICE: QUERY PLAN: + +Seq Scan on tmp (cost=40.03 size=101 width=8) + SubPlan + ^^^^^^^ subquery is in Seq Scan' qual, its plan is below + -> Aggregate (cost=2.05 size=0 width=0) + InitPlan + ^^^^^^^^ EXISTS subsubquery is InitPlan of subquery + -> Seq Scan on tempx (cost=4.33 size=1 width=4) + -> Result (cost=2.05 size=0 width=0) + ^^^^^^ EXISTS subsubquery was transformed into Param + and so we have Result node here + -> Index Scan on test2 (cost=2.05 size=1 width=4) + + +Opened issues. + +1. No read permissions checking (easy, just not done yet). +2. readfuncs.c can't read subplan-s (easy, not critical, because of + we currently nowhere use ascii representation of execution plans). +3. ExecReScan() doesn't support all plan types. At least support for + MergeJoin has to be implemented. +4. Memory leaks in ExecReScan(). +5. I need in advice: if subquery introduced with NOT IN doesn't return + any tuples then qualification is failed, yes ? +6. Regression tests !!!!!!!!!!!!!!!!!!!! + (Could we use data/queries from MySQL' crash.me ? + Copyright-ed ? Could they give us rights ?) +7. Performance. + - Should be good when subquery is transformed into InitPlan. + - Something should be done for uncorrelated subqueries introduced + with ANY/ALL - keep thinking. Currently, subplan will be re-scanned + for each parent tuple - very slow... + +Results of some test. TMP is table with x,y (int4-s), x in 0-9, +y = 100 - x, 1000 tuples (10 duplicates of each tuple). TEST2 is table +with x2, y2 (int4-s), x2 in 1-99, y2 = 100 -x2, 10000 tuples (100 dups). + + Trying + +select * from tmp where x >= (select max(x2) from test2 where y2 = y); + + and + +begin; +select y as ty, max(x2) as mx into table tsub from test2, tmp +where y2 = y group by ty; +vacuum tsub; +select x, y from tmp, tsub where x >= mx and y = ty; +drop table tsub; +end; + + Without index on test2(y2): + +SubSelect -> 320 sec +Using temp table -> 32 sec + + Having index + +SubSelect -> 17 sec (2M of memory) +Using temp table -> 32 sec (12M of memory: -S 8192) + +Vadim diff --git a/src/backend/optimizer/plan/analyzejoins.c b/src/backend/optimizer/plan/analyzejoins.c new file mode 100644 index 0000000..34efeee --- /dev/null +++ b/src/backend/optimizer/plan/analyzejoins.c @@ -0,0 +1,1127 @@ +/*------------------------------------------------------------------------- + * + * analyzejoins.c + * Routines for simplifying joins after initial query analysis + * + * While we do a great deal of join simplification in prep/prepjointree.c, + * certain optimizations cannot be performed at that stage for lack of + * detailed information about the query. The routines here are invoked + * after initsplan.c has done its work, and can do additional join removal + * and simplification steps based on the information extracted. The penalty + * is that we have to work harder to clean up after ourselves when we modify + * the query, since the derived data structures have to be updated too. + * + * Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group + * Portions Copyright (c) 1994, Regents of the University of California + * + * + * IDENTIFICATION + * src/backend/optimizer/plan/analyzejoins.c + * + *------------------------------------------------------------------------- + */ +#include "postgres.h" + +#include "nodes/nodeFuncs.h" +#include "optimizer/clauses.h" +#include "optimizer/joininfo.h" +#include "optimizer/optimizer.h" +#include "optimizer/pathnode.h" +#include "optimizer/paths.h" +#include "optimizer/planmain.h" +#include "optimizer/tlist.h" +#include "utils/lsyscache.h" + +/* local functions */ +static bool join_is_removable(PlannerInfo *root, SpecialJoinInfo *sjinfo); +static void remove_rel_from_query(PlannerInfo *root, int relid, + Relids joinrelids); +static List *remove_rel_from_joinlist(List *joinlist, int relid, int *nremoved); +static bool rel_supports_distinctness(PlannerInfo *root, RelOptInfo *rel); +static bool rel_is_distinct_for(PlannerInfo *root, RelOptInfo *rel, + List *clause_list); +static Oid distinct_col_search(int colno, List *colnos, List *opids); +static bool is_innerrel_unique_for(PlannerInfo *root, + Relids joinrelids, + Relids outerrelids, + RelOptInfo *innerrel, + JoinType jointype, + List *restrictlist); + + +/* + * remove_useless_joins + * Check for relations that don't actually need to be joined at all, + * and remove them from the query. + * + * We are passed the current joinlist and return the updated list. Other + * data structures that have to be updated are accessible via "root". + */ +List * +remove_useless_joins(PlannerInfo *root, List *joinlist) +{ + ListCell *lc; + + /* + * We are only interested in relations that are left-joined to, so we can + * scan the join_info_list to find them easily. + */ +restart: + foreach(lc, root->join_info_list) + { + SpecialJoinInfo *sjinfo = (SpecialJoinInfo *) lfirst(lc); + int innerrelid; + int nremoved; + + /* Skip if not removable */ + if (!join_is_removable(root, sjinfo)) + continue; + + /* + * Currently, join_is_removable can only succeed when the sjinfo's + * righthand is a single baserel. Remove that rel from the query and + * joinlist. + */ + innerrelid = bms_singleton_member(sjinfo->min_righthand); + + remove_rel_from_query(root, innerrelid, + bms_union(sjinfo->min_lefthand, + sjinfo->min_righthand)); + + /* We verify that exactly one reference gets removed from joinlist */ + nremoved = 0; + joinlist = remove_rel_from_joinlist(joinlist, innerrelid, &nremoved); + if (nremoved != 1) + elog(ERROR, "failed to find relation %d in joinlist", innerrelid); + + /* + * We can delete this SpecialJoinInfo from the list too, since it's no + * longer of interest. (Since we'll restart the foreach loop + * immediately, we don't bother with foreach_delete_current.) + */ + root->join_info_list = list_delete_cell(root->join_info_list, lc); + + /* + * Restart the scan. This is necessary to ensure we find all + * removable joins independently of ordering of the join_info_list + * (note that removal of attr_needed bits may make a join appear + * removable that did not before). + */ + goto restart; + } + + return joinlist; +} + +/* + * clause_sides_match_join + * Determine whether a join clause is of the right form to use in this join. + * + * We already know that the clause is a binary opclause referencing only the + * rels in the current join. The point here is to check whether it has the + * form "outerrel_expr op innerrel_expr" or "innerrel_expr op outerrel_expr", + * rather than mixing outer and inner vars on either side. If it matches, + * we set the transient flag outer_is_left to identify which side is which. + */ +static inline bool +clause_sides_match_join(RestrictInfo *rinfo, Relids outerrelids, + Relids innerrelids) +{ + if (bms_is_subset(rinfo->left_relids, outerrelids) && + bms_is_subset(rinfo->right_relids, innerrelids)) + { + /* lefthand side is outer */ + rinfo->outer_is_left = true; + return true; + } + else if (bms_is_subset(rinfo->left_relids, innerrelids) && + bms_is_subset(rinfo->right_relids, outerrelids)) + { + /* righthand side is outer */ + rinfo->outer_is_left = false; + return true; + } + return false; /* no good for these input relations */ +} + +/* + * join_is_removable + * Check whether we need not perform this special join at all, because + * it will just duplicate its left input. + * + * This is true for a left join for which the join condition cannot match + * more than one inner-side row. (There are other possibly interesting + * cases, but we don't have the infrastructure to prove them.) We also + * have to check that the inner side doesn't generate any variables needed + * above the join. + */ +static bool +join_is_removable(PlannerInfo *root, SpecialJoinInfo *sjinfo) +{ + int innerrelid; + RelOptInfo *innerrel; + Relids joinrelids; + List *clause_list = NIL; + ListCell *l; + int attroff; + + /* + * Must be a non-delaying left join to a single baserel, else we aren't + * going to be able to do anything with it. + */ + if (sjinfo->jointype != JOIN_LEFT || + sjinfo->delay_upper_joins) + return false; + + if (!bms_get_singleton_member(sjinfo->min_righthand, &innerrelid)) + return false; + + /* + * Never try to eliminate a left join to the query result rel. Although + * the case is syntactically impossible in standard SQL, MERGE will build + * a join tree that looks exactly like that. + */ + if (innerrelid == root->parse->resultRelation) + return false; + + innerrel = find_base_rel(root, innerrelid); + + /* + * Before we go to the effort of checking whether any innerrel variables + * are needed above the join, make a quick check to eliminate cases in + * which we will surely be unable to prove uniqueness of the innerrel. + */ + if (!rel_supports_distinctness(root, innerrel)) + return false; + + /* Compute the relid set for the join we are considering */ + joinrelids = bms_union(sjinfo->min_lefthand, sjinfo->min_righthand); + + /* + * We can't remove the join if any inner-rel attributes are used above the + * join. + * + * Note that this test only detects use of inner-rel attributes in higher + * join conditions and the target list. There might be such attributes in + * pushed-down conditions at this join, too. We check that case below. + * + * As a micro-optimization, it seems better to start with max_attr and + * count down rather than starting with min_attr and counting up, on the + * theory that the system attributes are somewhat less likely to be wanted + * and should be tested last. + */ + for (attroff = innerrel->max_attr - innerrel->min_attr; + attroff >= 0; + attroff--) + { + if (!bms_is_subset(innerrel->attr_needed[attroff], joinrelids)) + return false; + } + + /* + * Similarly check that the inner rel isn't needed by any PlaceHolderVars + * that will be used above the join. We only need to fail if such a PHV + * actually references some inner-rel attributes; but the correct check + * for that is relatively expensive, so we first check against ph_eval_at, + * which must mention the inner rel if the PHV uses any inner-rel attrs as + * non-lateral references. Note that if the PHV's syntactic scope is just + * the inner rel, we can't drop the rel even if the PHV is variable-free. + */ + foreach(l, root->placeholder_list) + { + PlaceHolderInfo *phinfo = (PlaceHolderInfo *) lfirst(l); + + if (bms_overlap(phinfo->ph_lateral, innerrel->relids)) + return false; /* it references innerrel laterally */ + if (bms_is_subset(phinfo->ph_needed, joinrelids)) + continue; /* PHV is not used above the join */ + if (!bms_overlap(phinfo->ph_eval_at, innerrel->relids)) + continue; /* it definitely doesn't reference innerrel */ + if (bms_is_subset(phinfo->ph_eval_at, innerrel->relids)) + return false; /* there isn't any other place to eval PHV */ + if (bms_overlap(pull_varnos(root, (Node *) phinfo->ph_var->phexpr), + innerrel->relids)) + return false; /* it does reference innerrel */ + } + + /* + * Search for mergejoinable clauses that constrain the inner rel against + * either the outer rel or a pseudoconstant. If an operator is + * mergejoinable then it behaves like equality for some btree opclass, so + * it's what we want. The mergejoinability test also eliminates clauses + * containing volatile functions, which we couldn't depend on. + */ + foreach(l, innerrel->joininfo) + { + RestrictInfo *restrictinfo = (RestrictInfo *) lfirst(l); + + /* + * If it's not a join clause for this outer join, we can't use it. + * Note that if the clause is pushed-down, then it is logically from + * above the outer join, even if it references no other rels (it might + * be from WHERE, for example). + */ + if (RINFO_IS_PUSHED_DOWN(restrictinfo, joinrelids)) + { + /* + * If such a clause actually references the inner rel then join + * removal has to be disallowed. We have to check this despite + * the previous attr_needed checks because of the possibility of + * pushed-down clauses referencing the rel. + */ + if (bms_is_member(innerrelid, restrictinfo->clause_relids)) + return false; + continue; /* else, ignore; not useful here */ + } + + /* Ignore if it's not a mergejoinable clause */ + if (!restrictinfo->can_join || + restrictinfo->mergeopfamilies == NIL) + continue; /* not mergejoinable */ + + /* + * Check if clause has the form "outer op inner" or "inner op outer", + * and if so mark which side is inner. + */ + if (!clause_sides_match_join(restrictinfo, sjinfo->min_lefthand, + innerrel->relids)) + continue; /* no good for these input relations */ + + /* OK, add to list */ + clause_list = lappend(clause_list, restrictinfo); + } + + /* + * Now that we have the relevant equality join clauses, try to prove the + * innerrel distinct. + */ + if (rel_is_distinct_for(root, innerrel, clause_list)) + return true; + + /* + * Some day it would be nice to check for other methods of establishing + * distinctness. + */ + return false; +} + + +/* + * Remove the target relid from the planner's data structures, having + * determined that there is no need to include it in the query. + * + * We are not terribly thorough here. We must make sure that the rel is + * no longer treated as a baserel, and that attributes of other baserels + * are no longer marked as being needed at joins involving this rel. + * Also, join quals involving the rel have to be removed from the joininfo + * lists, but only if they belong to the outer join identified by joinrelids. + */ +static void +remove_rel_from_query(PlannerInfo *root, int relid, Relids joinrelids) +{ + RelOptInfo *rel = find_base_rel(root, relid); + List *joininfos; + Index rti; + ListCell *l; + + /* + * Mark the rel as "dead" to show it is no longer part of the join tree. + * (Removing it from the baserel array altogether seems too risky.) + */ + rel->reloptkind = RELOPT_DEADREL; + + /* + * Remove references to the rel from other baserels' attr_needed arrays. + */ + for (rti = 1; rti < root->simple_rel_array_size; rti++) + { + RelOptInfo *otherrel = root->simple_rel_array[rti]; + int attroff; + + /* there may be empty slots corresponding to non-baserel RTEs */ + if (otherrel == NULL) + continue; + + Assert(otherrel->relid == rti); /* sanity check on array */ + + /* no point in processing target rel itself */ + if (otherrel == rel) + continue; + + for (attroff = otherrel->max_attr - otherrel->min_attr; + attroff >= 0; + attroff--) + { + otherrel->attr_needed[attroff] = + bms_del_member(otherrel->attr_needed[attroff], relid); + } + } + + /* + * Likewise remove references from SpecialJoinInfo data structures. + * + * This is relevant in case the outer join we're deleting is nested inside + * other outer joins: the upper joins' relid sets have to be adjusted. The + * RHS of the target outer join will be made empty here, but that's OK + * since caller will delete that SpecialJoinInfo entirely. + */ + foreach(l, root->join_info_list) + { + SpecialJoinInfo *sjinfo = (SpecialJoinInfo *) lfirst(l); + + sjinfo->min_lefthand = bms_del_member(sjinfo->min_lefthand, relid); + sjinfo->min_righthand = bms_del_member(sjinfo->min_righthand, relid); + sjinfo->syn_lefthand = bms_del_member(sjinfo->syn_lefthand, relid); + sjinfo->syn_righthand = bms_del_member(sjinfo->syn_righthand, relid); + } + + /* + * Likewise remove references from PlaceHolderVar data structures, + * removing any no-longer-needed placeholders entirely. + * + * Removal is a bit trickier than it might seem: we can remove PHVs that + * are used at the target rel and/or in the join qual, but not those that + * are used at join partner rels or above the join. It's not that easy to + * distinguish PHVs used at partner rels from those used in the join qual, + * since they will both have ph_needed sets that are subsets of + * joinrelids. However, a PHV used at a partner rel could not have the + * target rel in ph_eval_at, so we check that while deciding whether to + * remove or just update the PHV. There is no corresponding test in + * join_is_removable because it doesn't need to distinguish those cases. + */ + foreach(l, root->placeholder_list) + { + PlaceHolderInfo *phinfo = (PlaceHolderInfo *) lfirst(l); + + Assert(!bms_is_member(relid, phinfo->ph_lateral)); + if (bms_is_subset(phinfo->ph_needed, joinrelids) && + bms_is_member(relid, phinfo->ph_eval_at)) + root->placeholder_list = foreach_delete_current(root->placeholder_list, + l); + else + { + phinfo->ph_eval_at = bms_del_member(phinfo->ph_eval_at, relid); + Assert(!bms_is_empty(phinfo->ph_eval_at)); + phinfo->ph_needed = bms_del_member(phinfo->ph_needed, relid); + } + } + + /* + * Remove any joinquals referencing the rel from the joininfo lists. + * + * In some cases, a joinqual has to be put back after deleting its + * reference to the target rel. This can occur for pseudoconstant and + * outerjoin-delayed quals, which can get marked as requiring the rel in + * order to force them to be evaluated at or above the join. We can't + * just discard them, though. Only quals that logically belonged to the + * outer join being discarded should be removed from the query. + * + * We must make a copy of the rel's old joininfo list before starting the + * loop, because otherwise remove_join_clause_from_rels would destroy the + * list while we're scanning it. + */ + joininfos = list_copy(rel->joininfo); + foreach(l, joininfos) + { + RestrictInfo *rinfo = (RestrictInfo *) lfirst(l); + + remove_join_clause_from_rels(root, rinfo, rinfo->required_relids); + + if (RINFO_IS_PUSHED_DOWN(rinfo, joinrelids)) + { + /* Recheck that qual doesn't actually reference the target rel */ + Assert(!bms_is_member(relid, rinfo->clause_relids)); + + /* + * The required_relids probably aren't shared with anything else, + * but let's copy them just to be sure. + */ + rinfo->required_relids = bms_copy(rinfo->required_relids); + rinfo->required_relids = bms_del_member(rinfo->required_relids, + relid); + distribute_restrictinfo_to_rels(root, rinfo); + } + } + + /* + * There may be references to the rel in root->fkey_list, but if so, + * match_foreign_keys_to_quals() will get rid of them. + */ +} + +/* + * Remove any occurrences of the target relid from a joinlist structure. + * + * It's easiest to build a whole new list structure, so we handle it that + * way. Efficiency is not a big deal here. + * + * *nremoved is incremented by the number of occurrences removed (there + * should be exactly one, but the caller checks that). + */ +static List * +remove_rel_from_joinlist(List *joinlist, int relid, int *nremoved) +{ + List *result = NIL; + ListCell *jl; + + foreach(jl, joinlist) + { + Node *jlnode = (Node *) lfirst(jl); + + if (IsA(jlnode, RangeTblRef)) + { + int varno = ((RangeTblRef *) jlnode)->rtindex; + + if (varno == relid) + (*nremoved)++; + else + result = lappend(result, jlnode); + } + else if (IsA(jlnode, List)) + { + /* Recurse to handle subproblem */ + List *sublist; + + sublist = remove_rel_from_joinlist((List *) jlnode, + relid, nremoved); + /* Avoid including empty sub-lists in the result */ + if (sublist) + result = lappend(result, sublist); + } + else + { + elog(ERROR, "unrecognized joinlist node type: %d", + (int) nodeTag(jlnode)); + } + } + + return result; +} + + +/* + * reduce_unique_semijoins + * Check for semijoins that can be simplified to plain inner joins + * because the inner relation is provably unique for the join clauses. + * + * Ideally this would happen during reduce_outer_joins, but we don't have + * enough information at that point. + * + * To perform the strength reduction when applicable, we need only delete + * the semijoin's SpecialJoinInfo from root->join_info_list. (We don't + * bother fixing the join type attributed to it in the query jointree, + * since that won't be consulted again.) + */ +void +reduce_unique_semijoins(PlannerInfo *root) +{ + ListCell *lc; + + /* + * Scan the join_info_list to find semijoins. + */ + foreach(lc, root->join_info_list) + { + SpecialJoinInfo *sjinfo = (SpecialJoinInfo *) lfirst(lc); + int innerrelid; + RelOptInfo *innerrel; + Relids joinrelids; + List *restrictlist; + + /* + * Must be a non-delaying semijoin to a single baserel, else we aren't + * going to be able to do anything with it. (It's probably not + * possible for delay_upper_joins to be set on a semijoin, but we + * might as well check.) + */ + if (sjinfo->jointype != JOIN_SEMI || + sjinfo->delay_upper_joins) + continue; + + if (!bms_get_singleton_member(sjinfo->min_righthand, &innerrelid)) + continue; + + innerrel = find_base_rel(root, innerrelid); + + /* + * Before we trouble to run generate_join_implied_equalities, make a + * quick check to eliminate cases in which we will surely be unable to + * prove uniqueness of the innerrel. + */ + if (!rel_supports_distinctness(root, innerrel)) + continue; + + /* Compute the relid set for the join we are considering */ + joinrelids = bms_union(sjinfo->min_lefthand, sjinfo->min_righthand); + + /* + * Since we're only considering a single-rel RHS, any join clauses it + * has must be clauses linking it to the semijoin's min_lefthand. We + * can also consider EC-derived join clauses. + */ + restrictlist = + list_concat(generate_join_implied_equalities(root, + joinrelids, + sjinfo->min_lefthand, + innerrel), + innerrel->joininfo); + + /* Test whether the innerrel is unique for those clauses. */ + if (!innerrel_is_unique(root, + joinrelids, sjinfo->min_lefthand, innerrel, + JOIN_SEMI, restrictlist, true)) + continue; + + /* OK, remove the SpecialJoinInfo from the list. */ + root->join_info_list = foreach_delete_current(root->join_info_list, lc); + } +} + + +/* + * rel_supports_distinctness + * Could the relation possibly be proven distinct on some set of columns? + * + * This is effectively a pre-checking function for rel_is_distinct_for(). + * It must return true if rel_is_distinct_for() could possibly return true + * with this rel, but it should not expend a lot of cycles. The idea is + * that callers can avoid doing possibly-expensive processing to compute + * rel_is_distinct_for()'s argument lists if the call could not possibly + * succeed. + */ +static bool +rel_supports_distinctness(PlannerInfo *root, RelOptInfo *rel) +{ + /* We only know about baserels ... */ + if (rel->reloptkind != RELOPT_BASEREL) + return false; + if (rel->rtekind == RTE_RELATION) + { + /* + * For a plain relation, we only know how to prove uniqueness by + * reference to unique indexes. Make sure there's at least one + * suitable unique index. It must be immediately enforced, and not a + * partial index. (Keep these conditions in sync with + * relation_has_unique_index_for!) + */ + ListCell *lc; + + foreach(lc, rel->indexlist) + { + IndexOptInfo *ind = (IndexOptInfo *) lfirst(lc); + + if (ind->unique && ind->immediate && ind->indpred == NIL) + return true; + } + } + else if (rel->rtekind == RTE_SUBQUERY) + { + Query *subquery = root->simple_rte_array[rel->relid]->subquery; + + /* Check if the subquery has any qualities that support distinctness */ + if (query_supports_distinctness(subquery)) + return true; + } + /* We have no proof rules for any other rtekinds. */ + return false; +} + +/* + * rel_is_distinct_for + * Does the relation return only distinct rows according to clause_list? + * + * clause_list is a list of join restriction clauses involving this rel and + * some other one. Return true if no two rows emitted by this rel could + * possibly join to the same row of the other rel. + * + * The caller must have already determined that each condition is a + * mergejoinable equality with an expression in this relation on one side, and + * an expression not involving this relation on the other. The transient + * outer_is_left flag is used to identify which side references this relation: + * left side if outer_is_left is false, right side if it is true. + * + * Note that the passed-in clause_list may be destructively modified! This + * is OK for current uses, because the clause_list is built by the caller for + * the sole purpose of passing to this function. + */ +static bool +rel_is_distinct_for(PlannerInfo *root, RelOptInfo *rel, List *clause_list) +{ + /* + * We could skip a couple of tests here if we assume all callers checked + * rel_supports_distinctness first, but it doesn't seem worth taking any + * risk for. + */ + if (rel->reloptkind != RELOPT_BASEREL) + return false; + if (rel->rtekind == RTE_RELATION) + { + /* + * Examine the indexes to see if we have a matching unique index. + * relation_has_unique_index_for automatically adds any usable + * restriction clauses for the rel, so we needn't do that here. + */ + if (relation_has_unique_index_for(root, rel, clause_list, NIL, NIL)) + return true; + } + else if (rel->rtekind == RTE_SUBQUERY) + { + Index relid = rel->relid; + Query *subquery = root->simple_rte_array[relid]->subquery; + List *colnos = NIL; + List *opids = NIL; + ListCell *l; + + /* + * Build the argument lists for query_is_distinct_for: a list of + * output column numbers that the query needs to be distinct over, and + * a list of equality operators that the output columns need to be + * distinct according to. + * + * (XXX we are not considering restriction clauses attached to the + * subquery; is that worth doing?) + */ + foreach(l, clause_list) + { + RestrictInfo *rinfo = lfirst_node(RestrictInfo, l); + Oid op; + Var *var; + + /* + * Get the equality operator we need uniqueness according to. + * (This might be a cross-type operator and thus not exactly the + * same operator the subquery would consider; that's all right + * since query_is_distinct_for can resolve such cases.) The + * caller's mergejoinability test should have selected only + * OpExprs. + */ + op = castNode(OpExpr, rinfo->clause)->opno; + + /* caller identified the inner side for us */ + if (rinfo->outer_is_left) + var = (Var *) get_rightop(rinfo->clause); + else + var = (Var *) get_leftop(rinfo->clause); + + /* + * We may ignore any RelabelType node above the operand. (There + * won't be more than one, since eval_const_expressions() has been + * applied already.) + */ + if (var && IsA(var, RelabelType)) + var = (Var *) ((RelabelType *) var)->arg; + + /* + * If inner side isn't a Var referencing a subquery output column, + * this clause doesn't help us. + */ + if (!var || !IsA(var, Var) || + var->varno != relid || var->varlevelsup != 0) + continue; + + colnos = lappend_int(colnos, var->varattno); + opids = lappend_oid(opids, op); + } + + if (query_is_distinct_for(subquery, colnos, opids)) + return true; + } + return false; +} + + +/* + * query_supports_distinctness - could the query possibly be proven distinct + * on some set of output columns? + * + * This is effectively a pre-checking function for query_is_distinct_for(). + * It must return true if query_is_distinct_for() could possibly return true + * with this query, but it should not expend a lot of cycles. The idea is + * that callers can avoid doing possibly-expensive processing to compute + * query_is_distinct_for()'s argument lists if the call could not possibly + * succeed. + */ +bool +query_supports_distinctness(Query *query) +{ + /* SRFs break distinctness except with DISTINCT, see below */ + if (query->hasTargetSRFs && query->distinctClause == NIL) + return false; + + /* check for features we can prove distinctness with */ + if (query->distinctClause != NIL || + query->groupClause != NIL || + query->groupingSets != NIL || + query->hasAggs || + query->havingQual || + query->setOperations) + return true; + + return false; +} + +/* + * query_is_distinct_for - does query never return duplicates of the + * specified columns? + * + * query is a not-yet-planned subquery (in current usage, it's always from + * a subquery RTE, which the planner avoids scribbling on). + * + * colnos is an integer list of output column numbers (resno's). We are + * interested in whether rows consisting of just these columns are certain + * to be distinct. "Distinctness" is defined according to whether the + * corresponding upper-level equality operators listed in opids would think + * the values are distinct. (Note: the opids entries could be cross-type + * operators, and thus not exactly the equality operators that the subquery + * would use itself. We use equality_ops_are_compatible() to check + * compatibility. That looks at btree or hash opfamily membership, and so + * should give trustworthy answers for all operators that we might need + * to deal with here.) + */ +bool +query_is_distinct_for(Query *query, List *colnos, List *opids) +{ + ListCell *l; + Oid opid; + + Assert(list_length(colnos) == list_length(opids)); + + /* + * DISTINCT (including DISTINCT ON) guarantees uniqueness if all the + * columns in the DISTINCT clause appear in colnos and operator semantics + * match. This is true even if there are SRFs in the DISTINCT columns or + * elsewhere in the tlist. + */ + if (query->distinctClause) + { + foreach(l, query->distinctClause) + { + SortGroupClause *sgc = (SortGroupClause *) lfirst(l); + TargetEntry *tle = get_sortgroupclause_tle(sgc, + query->targetList); + + opid = distinct_col_search(tle->resno, colnos, opids); + if (!OidIsValid(opid) || + !equality_ops_are_compatible(opid, sgc->eqop)) + break; /* exit early if no match */ + } + if (l == NULL) /* had matches for all? */ + return true; + } + + /* + * Otherwise, a set-returning function in the query's targetlist can + * result in returning duplicate rows, despite any grouping that might + * occur before tlist evaluation. (If all tlist SRFs are within GROUP BY + * columns, it would be safe because they'd be expanded before grouping. + * But it doesn't currently seem worth the effort to check for that.) + */ + if (query->hasTargetSRFs) + return false; + + /* + * Similarly, GROUP BY without GROUPING SETS guarantees uniqueness if all + * the grouped columns appear in colnos and operator semantics match. + */ + if (query->groupClause && !query->groupingSets) + { + foreach(l, query->groupClause) + { + SortGroupClause *sgc = (SortGroupClause *) lfirst(l); + TargetEntry *tle = get_sortgroupclause_tle(sgc, + query->targetList); + + opid = distinct_col_search(tle->resno, colnos, opids); + if (!OidIsValid(opid) || + !equality_ops_are_compatible(opid, sgc->eqop)) + break; /* exit early if no match */ + } + if (l == NULL) /* had matches for all? */ + return true; + } + else if (query->groupingSets) + { + /* + * If we have grouping sets with expressions, we probably don't have + * uniqueness and analysis would be hard. Punt. + */ + if (query->groupClause) + return false; + + /* + * If we have no groupClause (therefore no grouping expressions), we + * might have one or many empty grouping sets. If there's just one, + * then we're returning only one row and are certainly unique. But + * otherwise, we know we're certainly not unique. + */ + if (list_length(query->groupingSets) == 1 && + ((GroupingSet *) linitial(query->groupingSets))->kind == GROUPING_SET_EMPTY) + return true; + else + return false; + } + else + { + /* + * If we have no GROUP BY, but do have aggregates or HAVING, then the + * result is at most one row so it's surely unique, for any operators. + */ + if (query->hasAggs || query->havingQual) + return true; + } + + /* + * UNION, INTERSECT, EXCEPT guarantee uniqueness of the whole output row, + * except with ALL. + */ + if (query->setOperations) + { + SetOperationStmt *topop = castNode(SetOperationStmt, query->setOperations); + + Assert(topop->op != SETOP_NONE); + + if (!topop->all) + { + ListCell *lg; + + /* We're good if all the nonjunk output columns are in colnos */ + lg = list_head(topop->groupClauses); + foreach(l, query->targetList) + { + TargetEntry *tle = (TargetEntry *) lfirst(l); + SortGroupClause *sgc; + + if (tle->resjunk) + continue; /* ignore resjunk columns */ + + /* non-resjunk columns should have grouping clauses */ + Assert(lg != NULL); + sgc = (SortGroupClause *) lfirst(lg); + lg = lnext(topop->groupClauses, lg); + + opid = distinct_col_search(tle->resno, colnos, opids); + if (!OidIsValid(opid) || + !equality_ops_are_compatible(opid, sgc->eqop)) + break; /* exit early if no match */ + } + if (l == NULL) /* had matches for all? */ + return true; + } + } + + /* + * XXX Are there any other cases in which we can easily see the result + * must be distinct? + * + * If you do add more smarts to this function, be sure to update + * query_supports_distinctness() to match. + */ + + return false; +} + +/* + * distinct_col_search - subroutine for query_is_distinct_for + * + * If colno is in colnos, return the corresponding element of opids, + * else return InvalidOid. (Ordinarily colnos would not contain duplicates, + * but if it does, we arbitrarily select the first match.) + */ +static Oid +distinct_col_search(int colno, List *colnos, List *opids) +{ + ListCell *lc1, + *lc2; + + forboth(lc1, colnos, lc2, opids) + { + if (colno == lfirst_int(lc1)) + return lfirst_oid(lc2); + } + return InvalidOid; +} + + +/* + * innerrel_is_unique + * Check if the innerrel provably contains at most one tuple matching any + * tuple from the outerrel, based on join clauses in the 'restrictlist'. + * + * We need an actual RelOptInfo for the innerrel, but it's sufficient to + * identify the outerrel by its Relids. This asymmetry supports use of this + * function before joinrels have been built. (The caller is expected to + * also supply the joinrelids, just to save recalculating that.) + * + * The proof must be made based only on clauses that will be "joinquals" + * rather than "otherquals" at execution. For an inner join there's no + * difference; but if the join is outer, we must ignore pushed-down quals, + * as those will become "otherquals". Note that this means the answer might + * vary depending on whether IS_OUTER_JOIN(jointype); since we cache the + * answer without regard to that, callers must take care not to call this + * with jointypes that would be classified differently by IS_OUTER_JOIN(). + * + * The actual proof is undertaken by is_innerrel_unique_for(); this function + * is a frontend that is mainly concerned with caching the answers. + * In particular, the force_cache argument allows overriding the internal + * heuristic about whether to cache negative answers; it should be "true" + * if making an inquiry that is not part of the normal bottom-up join search + * sequence. + */ +bool +innerrel_is_unique(PlannerInfo *root, + Relids joinrelids, + Relids outerrelids, + RelOptInfo *innerrel, + JoinType jointype, + List *restrictlist, + bool force_cache) +{ + MemoryContext old_context; + ListCell *lc; + + /* Certainly can't prove uniqueness when there are no joinclauses */ + if (restrictlist == NIL) + return false; + + /* + * Make a quick check to eliminate cases in which we will surely be unable + * to prove uniqueness of the innerrel. + */ + if (!rel_supports_distinctness(root, innerrel)) + return false; + + /* + * Query the cache to see if we've managed to prove that innerrel is + * unique for any subset of this outerrel. We don't need an exact match, + * as extra outerrels can't make the innerrel any less unique (or more + * formally, the restrictlist for a join to a superset outerrel must be a + * superset of the conditions we successfully used before). + */ + foreach(lc, innerrel->unique_for_rels) + { + Relids unique_for_rels = (Relids) lfirst(lc); + + if (bms_is_subset(unique_for_rels, outerrelids)) + return true; /* Success! */ + } + + /* + * Conversely, we may have already determined that this outerrel, or some + * superset thereof, cannot prove this innerrel to be unique. + */ + foreach(lc, innerrel->non_unique_for_rels) + { + Relids unique_for_rels = (Relids) lfirst(lc); + + if (bms_is_subset(outerrelids, unique_for_rels)) + return false; + } + + /* No cached information, so try to make the proof. */ + if (is_innerrel_unique_for(root, joinrelids, outerrelids, innerrel, + jointype, restrictlist)) + { + /* + * Cache the positive result for future probes, being sure to keep it + * in the planner_cxt even if we are working in GEQO. + * + * Note: one might consider trying to isolate the minimal subset of + * the outerrels that proved the innerrel unique. But it's not worth + * the trouble, because the planner builds up joinrels incrementally + * and so we'll see the minimally sufficient outerrels before any + * supersets of them anyway. + */ + old_context = MemoryContextSwitchTo(root->planner_cxt); + innerrel->unique_for_rels = lappend(innerrel->unique_for_rels, + bms_copy(outerrelids)); + MemoryContextSwitchTo(old_context); + + return true; /* Success! */ + } + else + { + /* + * None of the join conditions for outerrel proved innerrel unique, so + * we can safely reject this outerrel or any subset of it in future + * checks. + * + * However, in normal planning mode, caching this knowledge is totally + * pointless; it won't be queried again, because we build up joinrels + * from smaller to larger. It is useful in GEQO mode, where the + * knowledge can be carried across successive planning attempts; and + * it's likely to be useful when using join-search plugins, too. Hence + * cache when join_search_private is non-NULL. (Yeah, that's a hack, + * but it seems reasonable.) + * + * Also, allow callers to override that heuristic and force caching; + * that's useful for reduce_unique_semijoins, which calls here before + * the normal join search starts. + */ + if (force_cache || root->join_search_private) + { + old_context = MemoryContextSwitchTo(root->planner_cxt); + innerrel->non_unique_for_rels = + lappend(innerrel->non_unique_for_rels, + bms_copy(outerrelids)); + MemoryContextSwitchTo(old_context); + } + + return false; + } +} + +/* + * is_innerrel_unique_for + * Check if the innerrel provably contains at most one tuple matching any + * tuple from the outerrel, based on join clauses in the 'restrictlist'. + */ +static bool +is_innerrel_unique_for(PlannerInfo *root, + Relids joinrelids, + Relids outerrelids, + RelOptInfo *innerrel, + JoinType jointype, + List *restrictlist) +{ + List *clause_list = NIL; + ListCell *lc; + + /* + * Search for mergejoinable clauses that constrain the inner rel against + * the outer rel. If an operator is mergejoinable then it behaves like + * equality for some btree opclass, so it's what we want. The + * mergejoinability test also eliminates clauses containing volatile + * functions, which we couldn't depend on. + */ + foreach(lc, restrictlist) + { + RestrictInfo *restrictinfo = (RestrictInfo *) lfirst(lc); + + /* + * As noted above, if it's a pushed-down clause and we're at an outer + * join, we can't use it. + */ + if (IS_OUTER_JOIN(jointype) && + RINFO_IS_PUSHED_DOWN(restrictinfo, joinrelids)) + continue; + + /* Ignore if it's not a mergejoinable clause */ + if (!restrictinfo->can_join || + restrictinfo->mergeopfamilies == NIL) + continue; /* not mergejoinable */ + + /* + * Check if clause has the form "outer op inner" or "inner op outer", + * and if so mark which side is inner. + */ + if (!clause_sides_match_join(restrictinfo, outerrelids, + innerrel->relids)) + continue; /* no good for these input relations */ + + /* OK, add to list */ + clause_list = lappend(clause_list, restrictinfo); + } + + /* Let rel_is_distinct_for() do the hard work */ + return rel_is_distinct_for(root, innerrel, clause_list); +} diff --git a/src/backend/optimizer/plan/createplan.c b/src/backend/optimizer/plan/createplan.c new file mode 100644 index 0000000..1bc59c9 --- /dev/null +++ b/src/backend/optimizer/plan/createplan.c @@ -0,0 +1,7244 @@ +/*------------------------------------------------------------------------- + * + * createplan.c + * Routines to create the desired plan for processing a query. + * Planning is complete, we just need to convert the selected + * Path into a Plan. + * + * Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group + * Portions Copyright (c) 1994, Regents of the University of California + * + * + * IDENTIFICATION + * src/backend/optimizer/plan/createplan.c + * + *------------------------------------------------------------------------- + */ +#include "postgres.h" + +#include <math.h> + +#include "access/sysattr.h" +#include "catalog/pg_class.h" +#include "foreign/fdwapi.h" +#include "miscadmin.h" +#include "nodes/extensible.h" +#include "nodes/makefuncs.h" +#include "nodes/nodeFuncs.h" +#include "optimizer/clauses.h" +#include "optimizer/cost.h" +#include "optimizer/optimizer.h" +#include "optimizer/paramassign.h" +#include "optimizer/paths.h" +#include "optimizer/placeholder.h" +#include "optimizer/plancat.h" +#include "optimizer/planmain.h" +#include "optimizer/prep.h" +#include "optimizer/restrictinfo.h" +#include "optimizer/subselect.h" +#include "optimizer/tlist.h" +#include "parser/parse_clause.h" +#include "parser/parsetree.h" +#include "partitioning/partprune.h" +#include "utils/lsyscache.h" + + +/* + * Flag bits that can appear in the flags argument of create_plan_recurse(). + * These can be OR-ed together. + * + * CP_EXACT_TLIST specifies that the generated plan node must return exactly + * the tlist specified by the path's pathtarget (this overrides both + * CP_SMALL_TLIST and CP_LABEL_TLIST, if those are set). Otherwise, the + * plan node is allowed to return just the Vars and PlaceHolderVars needed + * to evaluate the pathtarget. + * + * CP_SMALL_TLIST specifies that a narrower tlist is preferred. This is + * passed down by parent nodes such as Sort and Hash, which will have to + * store the returned tuples. + * + * CP_LABEL_TLIST specifies that the plan node must return columns matching + * any sortgrouprefs specified in its pathtarget, with appropriate + * ressortgroupref labels. This is passed down by parent nodes such as Sort + * and Group, which need these values to be available in their inputs. + * + * CP_IGNORE_TLIST specifies that the caller plans to replace the targetlist, + * and therefore it doesn't matter a bit what target list gets generated. + */ +#define CP_EXACT_TLIST 0x0001 /* Plan must return specified tlist */ +#define CP_SMALL_TLIST 0x0002 /* Prefer narrower tlists */ +#define CP_LABEL_TLIST 0x0004 /* tlist must contain sortgrouprefs */ +#define CP_IGNORE_TLIST 0x0008 /* caller will replace tlist */ + + +static Plan *create_plan_recurse(PlannerInfo *root, Path *best_path, + int flags); +static Plan *create_scan_plan(PlannerInfo *root, Path *best_path, + int flags); +static List *build_path_tlist(PlannerInfo *root, Path *path); +static bool use_physical_tlist(PlannerInfo *root, Path *path, int flags); +static List *get_gating_quals(PlannerInfo *root, List *quals); +static Plan *create_gating_plan(PlannerInfo *root, Path *path, Plan *plan, + List *gating_quals); +static Plan *create_join_plan(PlannerInfo *root, JoinPath *best_path); +static bool mark_async_capable_plan(Plan *plan, Path *path); +static Plan *create_append_plan(PlannerInfo *root, AppendPath *best_path, + int flags); +static Plan *create_merge_append_plan(PlannerInfo *root, MergeAppendPath *best_path, + int flags); +static Result *create_group_result_plan(PlannerInfo *root, + GroupResultPath *best_path); +static ProjectSet *create_project_set_plan(PlannerInfo *root, ProjectSetPath *best_path); +static Material *create_material_plan(PlannerInfo *root, MaterialPath *best_path, + int flags); +static Memoize *create_memoize_plan(PlannerInfo *root, MemoizePath *best_path, + int flags); +static Plan *create_unique_plan(PlannerInfo *root, UniquePath *best_path, + int flags); +static Gather *create_gather_plan(PlannerInfo *root, GatherPath *best_path); +static Plan *create_projection_plan(PlannerInfo *root, + ProjectionPath *best_path, + int flags); +static Plan *inject_projection_plan(Plan *subplan, List *tlist, bool parallel_safe); +static Sort *create_sort_plan(PlannerInfo *root, SortPath *best_path, int flags); +static IncrementalSort *create_incrementalsort_plan(PlannerInfo *root, + IncrementalSortPath *best_path, int flags); +static Group *create_group_plan(PlannerInfo *root, GroupPath *best_path); +static Unique *create_upper_unique_plan(PlannerInfo *root, UpperUniquePath *best_path, + int flags); +static Agg *create_agg_plan(PlannerInfo *root, AggPath *best_path); +static Plan *create_groupingsets_plan(PlannerInfo *root, GroupingSetsPath *best_path); +static Result *create_minmaxagg_plan(PlannerInfo *root, MinMaxAggPath *best_path); +static WindowAgg *create_windowagg_plan(PlannerInfo *root, WindowAggPath *best_path); +static SetOp *create_setop_plan(PlannerInfo *root, SetOpPath *best_path, + int flags); +static RecursiveUnion *create_recursiveunion_plan(PlannerInfo *root, RecursiveUnionPath *best_path); +static LockRows *create_lockrows_plan(PlannerInfo *root, LockRowsPath *best_path, + int flags); +static ModifyTable *create_modifytable_plan(PlannerInfo *root, ModifyTablePath *best_path); +static Limit *create_limit_plan(PlannerInfo *root, LimitPath *best_path, + int flags); +static SeqScan *create_seqscan_plan(PlannerInfo *root, Path *best_path, + List *tlist, List *scan_clauses); +static SampleScan *create_samplescan_plan(PlannerInfo *root, Path *best_path, + List *tlist, List *scan_clauses); +static Scan *create_indexscan_plan(PlannerInfo *root, IndexPath *best_path, + List *tlist, List *scan_clauses, bool indexonly); +static BitmapHeapScan *create_bitmap_scan_plan(PlannerInfo *root, + BitmapHeapPath *best_path, + List *tlist, List *scan_clauses); +static Plan *create_bitmap_subplan(PlannerInfo *root, Path *bitmapqual, + List **qual, List **indexqual, List **indexECs); +static void bitmap_subplan_mark_shared(Plan *plan); +static TidScan *create_tidscan_plan(PlannerInfo *root, TidPath *best_path, + List *tlist, List *scan_clauses); +static TidRangeScan *create_tidrangescan_plan(PlannerInfo *root, + TidRangePath *best_path, + List *tlist, + List *scan_clauses); +static SubqueryScan *create_subqueryscan_plan(PlannerInfo *root, + SubqueryScanPath *best_path, + List *tlist, List *scan_clauses); +static FunctionScan *create_functionscan_plan(PlannerInfo *root, Path *best_path, + List *tlist, List *scan_clauses); +static ValuesScan *create_valuesscan_plan(PlannerInfo *root, Path *best_path, + List *tlist, List *scan_clauses); +static TableFuncScan *create_tablefuncscan_plan(PlannerInfo *root, Path *best_path, + List *tlist, List *scan_clauses); +static CteScan *create_ctescan_plan(PlannerInfo *root, Path *best_path, + List *tlist, List *scan_clauses); +static NamedTuplestoreScan *create_namedtuplestorescan_plan(PlannerInfo *root, + Path *best_path, List *tlist, List *scan_clauses); +static Result *create_resultscan_plan(PlannerInfo *root, Path *best_path, + List *tlist, List *scan_clauses); +static WorkTableScan *create_worktablescan_plan(PlannerInfo *root, Path *best_path, + List *tlist, List *scan_clauses); +static ForeignScan *create_foreignscan_plan(PlannerInfo *root, ForeignPath *best_path, + List *tlist, List *scan_clauses); +static CustomScan *create_customscan_plan(PlannerInfo *root, + CustomPath *best_path, + List *tlist, List *scan_clauses); +static NestLoop *create_nestloop_plan(PlannerInfo *root, NestPath *best_path); +static MergeJoin *create_mergejoin_plan(PlannerInfo *root, MergePath *best_path); +static HashJoin *create_hashjoin_plan(PlannerInfo *root, HashPath *best_path); +static Node *replace_nestloop_params(PlannerInfo *root, Node *expr); +static Node *replace_nestloop_params_mutator(Node *node, PlannerInfo *root); +static void fix_indexqual_references(PlannerInfo *root, IndexPath *index_path, + List **stripped_indexquals_p, + List **fixed_indexquals_p); +static List *fix_indexorderby_references(PlannerInfo *root, IndexPath *index_path); +static Node *fix_indexqual_clause(PlannerInfo *root, + IndexOptInfo *index, int indexcol, + Node *clause, List *indexcolnos); +static Node *fix_indexqual_operand(Node *node, IndexOptInfo *index, int indexcol); +static List *get_switched_clauses(List *clauses, Relids outerrelids); +static List *order_qual_clauses(PlannerInfo *root, List *clauses); +static void copy_generic_path_info(Plan *dest, Path *src); +static void copy_plan_costsize(Plan *dest, Plan *src); +static void label_sort_with_costsize(PlannerInfo *root, Sort *plan, + double limit_tuples); +static SeqScan *make_seqscan(List *qptlist, List *qpqual, Index scanrelid); +static SampleScan *make_samplescan(List *qptlist, List *qpqual, Index scanrelid, + TableSampleClause *tsc); +static IndexScan *make_indexscan(List *qptlist, List *qpqual, Index scanrelid, + Oid indexid, List *indexqual, List *indexqualorig, + List *indexorderby, List *indexorderbyorig, + List *indexorderbyops, + ScanDirection indexscandir); +static IndexOnlyScan *make_indexonlyscan(List *qptlist, List *qpqual, + Index scanrelid, Oid indexid, + List *indexqual, List *recheckqual, + List *indexorderby, + List *indextlist, + ScanDirection indexscandir); +static BitmapIndexScan *make_bitmap_indexscan(Index scanrelid, Oid indexid, + List *indexqual, + List *indexqualorig); +static BitmapHeapScan *make_bitmap_heapscan(List *qptlist, + List *qpqual, + Plan *lefttree, + List *bitmapqualorig, + Index scanrelid); +static TidScan *make_tidscan(List *qptlist, List *qpqual, Index scanrelid, + List *tidquals); +static TidRangeScan *make_tidrangescan(List *qptlist, List *qpqual, + Index scanrelid, List *tidrangequals); +static SubqueryScan *make_subqueryscan(List *qptlist, + List *qpqual, + Index scanrelid, + Plan *subplan); +static FunctionScan *make_functionscan(List *qptlist, List *qpqual, + Index scanrelid, List *functions, bool funcordinality); +static ValuesScan *make_valuesscan(List *qptlist, List *qpqual, + Index scanrelid, List *values_lists); +static TableFuncScan *make_tablefuncscan(List *qptlist, List *qpqual, + Index scanrelid, TableFunc *tablefunc); +static CteScan *make_ctescan(List *qptlist, List *qpqual, + Index scanrelid, int ctePlanId, int cteParam); +static NamedTuplestoreScan *make_namedtuplestorescan(List *qptlist, List *qpqual, + Index scanrelid, char *enrname); +static WorkTableScan *make_worktablescan(List *qptlist, List *qpqual, + Index scanrelid, int wtParam); +static RecursiveUnion *make_recursive_union(List *tlist, + Plan *lefttree, + Plan *righttree, + int wtParam, + List *distinctList, + long numGroups); +static BitmapAnd *make_bitmap_and(List *bitmapplans); +static BitmapOr *make_bitmap_or(List *bitmapplans); +static NestLoop *make_nestloop(List *tlist, + List *joinclauses, List *otherclauses, List *nestParams, + Plan *lefttree, Plan *righttree, + JoinType jointype, bool inner_unique); +static HashJoin *make_hashjoin(List *tlist, + List *joinclauses, List *otherclauses, + List *hashclauses, + List *hashoperators, List *hashcollations, + List *hashkeys, + Plan *lefttree, Plan *righttree, + JoinType jointype, bool inner_unique); +static Hash *make_hash(Plan *lefttree, + List *hashkeys, + Oid skewTable, + AttrNumber skewColumn, + bool skewInherit); +static MergeJoin *make_mergejoin(List *tlist, + List *joinclauses, List *otherclauses, + List *mergeclauses, + Oid *mergefamilies, + Oid *mergecollations, + int *mergestrategies, + bool *mergenullsfirst, + Plan *lefttree, Plan *righttree, + JoinType jointype, bool inner_unique, + bool skip_mark_restore); +static Sort *make_sort(Plan *lefttree, int numCols, + AttrNumber *sortColIdx, Oid *sortOperators, + Oid *collations, bool *nullsFirst); +static IncrementalSort *make_incrementalsort(Plan *lefttree, + int numCols, int nPresortedCols, + AttrNumber *sortColIdx, Oid *sortOperators, + Oid *collations, bool *nullsFirst); +static Plan *prepare_sort_from_pathkeys(Plan *lefttree, List *pathkeys, + Relids relids, + const AttrNumber *reqColIdx, + bool adjust_tlist_in_place, + int *p_numsortkeys, + AttrNumber **p_sortColIdx, + Oid **p_sortOperators, + Oid **p_collations, + bool **p_nullsFirst); +static Sort *make_sort_from_pathkeys(Plan *lefttree, List *pathkeys, + Relids relids); +static IncrementalSort *make_incrementalsort_from_pathkeys(Plan *lefttree, + List *pathkeys, Relids relids, int nPresortedCols); +static Sort *make_sort_from_groupcols(List *groupcls, + AttrNumber *grpColIdx, + Plan *lefttree); +static Material *make_material(Plan *lefttree); +static Memoize *make_memoize(Plan *lefttree, Oid *hashoperators, + Oid *collations, List *param_exprs, + bool singlerow, bool binary_mode, + uint32 est_entries, Bitmapset *keyparamids); +static WindowAgg *make_windowagg(List *tlist, Index winref, + int partNumCols, AttrNumber *partColIdx, Oid *partOperators, Oid *partCollations, + int ordNumCols, AttrNumber *ordColIdx, Oid *ordOperators, Oid *ordCollations, + int frameOptions, Node *startOffset, Node *endOffset, + Oid startInRangeFunc, Oid endInRangeFunc, + Oid inRangeColl, bool inRangeAsc, bool inRangeNullsFirst, + List *runCondition, List *qual, bool topWindow, + Plan *lefttree); +static Group *make_group(List *tlist, List *qual, int numGroupCols, + AttrNumber *grpColIdx, Oid *grpOperators, Oid *grpCollations, + Plan *lefttree); +static Unique *make_unique_from_sortclauses(Plan *lefttree, List *distinctList); +static Unique *make_unique_from_pathkeys(Plan *lefttree, + List *pathkeys, int numCols); +static Gather *make_gather(List *qptlist, List *qpqual, + int nworkers, int rescan_param, bool single_copy, Plan *subplan); +static SetOp *make_setop(SetOpCmd cmd, SetOpStrategy strategy, Plan *lefttree, + List *distinctList, AttrNumber flagColIdx, int firstFlag, + long numGroups); +static LockRows *make_lockrows(Plan *lefttree, List *rowMarks, int epqParam); +static Result *make_result(List *tlist, Node *resconstantqual, Plan *subplan); +static ProjectSet *make_project_set(List *tlist, Plan *subplan); +static ModifyTable *make_modifytable(PlannerInfo *root, Plan *subplan, + CmdType operation, bool canSetTag, + Index nominalRelation, Index rootRelation, + bool partColsUpdated, + List *resultRelations, + List *updateColnosLists, + List *withCheckOptionLists, List *returningLists, + List *rowMarks, OnConflictExpr *onconflict, + List *mergeActionList, int epqParam); +static GatherMerge *create_gather_merge_plan(PlannerInfo *root, + GatherMergePath *best_path); + + +/* + * create_plan + * Creates the access plan for a query by recursively processing the + * desired tree of pathnodes, starting at the node 'best_path'. For + * every pathnode found, we create a corresponding plan node containing + * appropriate id, target list, and qualification information. + * + * The tlists and quals in the plan tree are still in planner format, + * ie, Vars still correspond to the parser's numbering. This will be + * fixed later by setrefs.c. + * + * best_path is the best access path + * + * Returns a Plan tree. + */ +Plan * +create_plan(PlannerInfo *root, Path *best_path) +{ + Plan *plan; + + /* plan_params should not be in use in current query level */ + Assert(root->plan_params == NIL); + + /* Initialize this module's workspace in PlannerInfo */ + root->curOuterRels = NULL; + root->curOuterParams = NIL; + + /* Recursively process the path tree, demanding the correct tlist result */ + plan = create_plan_recurse(root, best_path, CP_EXACT_TLIST); + + /* + * Make sure the topmost plan node's targetlist exposes the original + * column names and other decorative info. Targetlists generated within + * the planner don't bother with that stuff, but we must have it on the + * top-level tlist seen at execution time. However, ModifyTable plan + * nodes don't have a tlist matching the querytree targetlist. + */ + if (!IsA(plan, ModifyTable)) + apply_tlist_labeling(plan->targetlist, root->processed_tlist); + + /* + * Attach any initPlans created in this query level to the topmost plan + * node. (In principle the initplans could go in any plan node at or + * above where they're referenced, but there seems no reason to put them + * any lower than the topmost node for the query level. Also, see + * comments for SS_finalize_plan before you try to change this.) + */ + SS_attach_initplans(root, plan); + + /* Check we successfully assigned all NestLoopParams to plan nodes */ + if (root->curOuterParams != NIL) + elog(ERROR, "failed to assign all NestLoopParams to plan nodes"); + + /* + * Reset plan_params to ensure param IDs used for nestloop params are not + * re-used later + */ + root->plan_params = NIL; + + return plan; +} + +/* + * create_plan_recurse + * Recursive guts of create_plan(). + */ +static Plan * +create_plan_recurse(PlannerInfo *root, Path *best_path, int flags) +{ + Plan *plan; + + /* Guard against stack overflow due to overly complex plans */ + check_stack_depth(); + + switch (best_path->pathtype) + { + case T_SeqScan: + case T_SampleScan: + case T_IndexScan: + case T_IndexOnlyScan: + case T_BitmapHeapScan: + case T_TidScan: + case T_TidRangeScan: + case T_SubqueryScan: + case T_FunctionScan: + case T_TableFuncScan: + case T_ValuesScan: + case T_CteScan: + case T_WorkTableScan: + case T_NamedTuplestoreScan: + case T_ForeignScan: + case T_CustomScan: + plan = create_scan_plan(root, best_path, flags); + break; + case T_HashJoin: + case T_MergeJoin: + case T_NestLoop: + plan = create_join_plan(root, + (JoinPath *) best_path); + break; + case T_Append: + plan = create_append_plan(root, + (AppendPath *) best_path, + flags); + break; + case T_MergeAppend: + plan = create_merge_append_plan(root, + (MergeAppendPath *) best_path, + flags); + break; + case T_Result: + if (IsA(best_path, ProjectionPath)) + { + plan = create_projection_plan(root, + (ProjectionPath *) best_path, + flags); + } + else if (IsA(best_path, MinMaxAggPath)) + { + plan = (Plan *) create_minmaxagg_plan(root, + (MinMaxAggPath *) best_path); + } + else if (IsA(best_path, GroupResultPath)) + { + plan = (Plan *) create_group_result_plan(root, + (GroupResultPath *) best_path); + } + else + { + /* Simple RTE_RESULT base relation */ + Assert(IsA(best_path, Path)); + plan = create_scan_plan(root, best_path, flags); + } + break; + case T_ProjectSet: + plan = (Plan *) create_project_set_plan(root, + (ProjectSetPath *) best_path); + break; + case T_Material: + plan = (Plan *) create_material_plan(root, + (MaterialPath *) best_path, + flags); + break; + case T_Memoize: + plan = (Plan *) create_memoize_plan(root, + (MemoizePath *) best_path, + flags); + break; + case T_Unique: + if (IsA(best_path, UpperUniquePath)) + { + plan = (Plan *) create_upper_unique_plan(root, + (UpperUniquePath *) best_path, + flags); + } + else + { + Assert(IsA(best_path, UniquePath)); + plan = create_unique_plan(root, + (UniquePath *) best_path, + flags); + } + break; + case T_Gather: + plan = (Plan *) create_gather_plan(root, + (GatherPath *) best_path); + break; + case T_Sort: + plan = (Plan *) create_sort_plan(root, + (SortPath *) best_path, + flags); + break; + case T_IncrementalSort: + plan = (Plan *) create_incrementalsort_plan(root, + (IncrementalSortPath *) best_path, + flags); + break; + case T_Group: + plan = (Plan *) create_group_plan(root, + (GroupPath *) best_path); + break; + case T_Agg: + if (IsA(best_path, GroupingSetsPath)) + plan = create_groupingsets_plan(root, + (GroupingSetsPath *) best_path); + else + { + Assert(IsA(best_path, AggPath)); + plan = (Plan *) create_agg_plan(root, + (AggPath *) best_path); + } + break; + case T_WindowAgg: + plan = (Plan *) create_windowagg_plan(root, + (WindowAggPath *) best_path); + break; + case T_SetOp: + plan = (Plan *) create_setop_plan(root, + (SetOpPath *) best_path, + flags); + break; + case T_RecursiveUnion: + plan = (Plan *) create_recursiveunion_plan(root, + (RecursiveUnionPath *) best_path); + break; + case T_LockRows: + plan = (Plan *) create_lockrows_plan(root, + (LockRowsPath *) best_path, + flags); + break; + case T_ModifyTable: + plan = (Plan *) create_modifytable_plan(root, + (ModifyTablePath *) best_path); + break; + case T_Limit: + plan = (Plan *) create_limit_plan(root, + (LimitPath *) best_path, + flags); + break; + case T_GatherMerge: + plan = (Plan *) create_gather_merge_plan(root, + (GatherMergePath *) best_path); + break; + default: + elog(ERROR, "unrecognized node type: %d", + (int) best_path->pathtype); + plan = NULL; /* keep compiler quiet */ + break; + } + + return plan; +} + +/* + * create_scan_plan + * Create a scan plan for the parent relation of 'best_path'. + */ +static Plan * +create_scan_plan(PlannerInfo *root, Path *best_path, int flags) +{ + RelOptInfo *rel = best_path->parent; + List *scan_clauses; + List *gating_clauses; + List *tlist; + Plan *plan; + + /* + * Extract the relevant restriction clauses from the parent relation. The + * executor must apply all these restrictions during the scan, except for + * pseudoconstants which we'll take care of below. + * + * If this is a plain indexscan or index-only scan, we need not consider + * restriction clauses that are implied by the index's predicate, so use + * indrestrictinfo not baserestrictinfo. Note that we can't do that for + * bitmap indexscans, since there's not necessarily a single index + * involved; but it doesn't matter since create_bitmap_scan_plan() will be + * able to get rid of such clauses anyway via predicate proof. + */ + switch (best_path->pathtype) + { + case T_IndexScan: + case T_IndexOnlyScan: + scan_clauses = castNode(IndexPath, best_path)->indexinfo->indrestrictinfo; + break; + default: + scan_clauses = rel->baserestrictinfo; + break; + } + + /* + * If this is a parameterized scan, we also need to enforce all the join + * clauses available from the outer relation(s). + * + * For paranoia's sake, don't modify the stored baserestrictinfo list. + */ + if (best_path->param_info) + scan_clauses = list_concat_copy(scan_clauses, + best_path->param_info->ppi_clauses); + + /* + * Detect whether we have any pseudoconstant quals to deal with. Then, if + * we'll need a gating Result node, it will be able to project, so there + * are no requirements on the child's tlist. + */ + gating_clauses = get_gating_quals(root, scan_clauses); + if (gating_clauses) + flags = 0; + + /* + * For table scans, rather than using the relation targetlist (which is + * only those Vars actually needed by the query), we prefer to generate a + * tlist containing all Vars in order. This will allow the executor to + * optimize away projection of the table tuples, if possible. + * + * But if the caller is going to ignore our tlist anyway, then don't + * bother generating one at all. We use an exact equality test here, so + * that this only applies when CP_IGNORE_TLIST is the only flag set. + */ + if (flags == CP_IGNORE_TLIST) + { + tlist = NULL; + } + else if (use_physical_tlist(root, best_path, flags)) + { + if (best_path->pathtype == T_IndexOnlyScan) + { + /* For index-only scan, the preferred tlist is the index's */ + tlist = copyObject(((IndexPath *) best_path)->indexinfo->indextlist); + + /* + * Transfer sortgroupref data to the replacement tlist, if + * requested (use_physical_tlist checked that this will work). + */ + if (flags & CP_LABEL_TLIST) + apply_pathtarget_labeling_to_tlist(tlist, best_path->pathtarget); + } + else + { + tlist = build_physical_tlist(root, rel); + if (tlist == NIL) + { + /* Failed because of dropped cols, so use regular method */ + tlist = build_path_tlist(root, best_path); + } + else + { + /* As above, transfer sortgroupref data to replacement tlist */ + if (flags & CP_LABEL_TLIST) + apply_pathtarget_labeling_to_tlist(tlist, best_path->pathtarget); + } + } + } + else + { + tlist = build_path_tlist(root, best_path); + } + + switch (best_path->pathtype) + { + case T_SeqScan: + plan = (Plan *) create_seqscan_plan(root, + best_path, + tlist, + scan_clauses); + break; + + case T_SampleScan: + plan = (Plan *) create_samplescan_plan(root, + best_path, + tlist, + scan_clauses); + break; + + case T_IndexScan: + plan = (Plan *) create_indexscan_plan(root, + (IndexPath *) best_path, + tlist, + scan_clauses, + false); + break; + + case T_IndexOnlyScan: + plan = (Plan *) create_indexscan_plan(root, + (IndexPath *) best_path, + tlist, + scan_clauses, + true); + break; + + case T_BitmapHeapScan: + plan = (Plan *) create_bitmap_scan_plan(root, + (BitmapHeapPath *) best_path, + tlist, + scan_clauses); + break; + + case T_TidScan: + plan = (Plan *) create_tidscan_plan(root, + (TidPath *) best_path, + tlist, + scan_clauses); + break; + + case T_TidRangeScan: + plan = (Plan *) create_tidrangescan_plan(root, + (TidRangePath *) best_path, + tlist, + scan_clauses); + break; + + case T_SubqueryScan: + plan = (Plan *) create_subqueryscan_plan(root, + (SubqueryScanPath *) best_path, + tlist, + scan_clauses); + break; + + case T_FunctionScan: + plan = (Plan *) create_functionscan_plan(root, + best_path, + tlist, + scan_clauses); + break; + + case T_TableFuncScan: + plan = (Plan *) create_tablefuncscan_plan(root, + best_path, + tlist, + scan_clauses); + break; + + case T_ValuesScan: + plan = (Plan *) create_valuesscan_plan(root, + best_path, + tlist, + scan_clauses); + break; + + case T_CteScan: + plan = (Plan *) create_ctescan_plan(root, + best_path, + tlist, + scan_clauses); + break; + + case T_NamedTuplestoreScan: + plan = (Plan *) create_namedtuplestorescan_plan(root, + best_path, + tlist, + scan_clauses); + break; + + case T_Result: + plan = (Plan *) create_resultscan_plan(root, + best_path, + tlist, + scan_clauses); + break; + + case T_WorkTableScan: + plan = (Plan *) create_worktablescan_plan(root, + best_path, + tlist, + scan_clauses); + break; + + case T_ForeignScan: + plan = (Plan *) create_foreignscan_plan(root, + (ForeignPath *) best_path, + tlist, + scan_clauses); + break; + + case T_CustomScan: + plan = (Plan *) create_customscan_plan(root, + (CustomPath *) best_path, + tlist, + scan_clauses); + break; + + default: + elog(ERROR, "unrecognized node type: %d", + (int) best_path->pathtype); + plan = NULL; /* keep compiler quiet */ + break; + } + + /* + * If there are any pseudoconstant clauses attached to this node, insert a + * gating Result node that evaluates the pseudoconstants as one-time + * quals. + */ + if (gating_clauses) + plan = create_gating_plan(root, best_path, plan, gating_clauses); + + return plan; +} + +/* + * Build a target list (ie, a list of TargetEntry) for the Path's output. + * + * This is almost just make_tlist_from_pathtarget(), but we also have to + * deal with replacing nestloop params. + */ +static List * +build_path_tlist(PlannerInfo *root, Path *path) +{ + List *tlist = NIL; + Index *sortgrouprefs = path->pathtarget->sortgrouprefs; + int resno = 1; + ListCell *v; + + foreach(v, path->pathtarget->exprs) + { + Node *node = (Node *) lfirst(v); + TargetEntry *tle; + + /* + * If it's a parameterized path, there might be lateral references in + * the tlist, which need to be replaced with Params. There's no need + * to remake the TargetEntry nodes, so apply this to each list item + * separately. + */ + if (path->param_info) + node = replace_nestloop_params(root, node); + + tle = makeTargetEntry((Expr *) node, + resno, + NULL, + false); + if (sortgrouprefs) + tle->ressortgroupref = sortgrouprefs[resno - 1]; + + tlist = lappend(tlist, tle); + resno++; + } + return tlist; +} + +/* + * use_physical_tlist + * Decide whether to use a tlist matching relation structure, + * rather than only those Vars actually referenced. + */ +static bool +use_physical_tlist(PlannerInfo *root, Path *path, int flags) +{ + RelOptInfo *rel = path->parent; + int i; + ListCell *lc; + + /* + * Forget it if either exact tlist or small tlist is demanded. + */ + if (flags & (CP_EXACT_TLIST | CP_SMALL_TLIST)) + return false; + + /* + * We can do this for real relation scans, subquery scans, function scans, + * tablefunc scans, values scans, and CTE scans (but not for, eg, joins). + */ + if (rel->rtekind != RTE_RELATION && + rel->rtekind != RTE_SUBQUERY && + rel->rtekind != RTE_FUNCTION && + rel->rtekind != RTE_TABLEFUNC && + rel->rtekind != RTE_VALUES && + rel->rtekind != RTE_CTE) + return false; + + /* + * Can't do it with inheritance cases either (mainly because Append + * doesn't project; this test may be unnecessary now that + * create_append_plan instructs its children to return an exact tlist). + */ + if (rel->reloptkind != RELOPT_BASEREL) + return false; + + /* + * Also, don't do it to a CustomPath; the premise that we're extracting + * columns from a simple physical tuple is unlikely to hold for those. + * (When it does make sense, the custom path creator can set up the path's + * pathtarget that way.) + */ + if (IsA(path, CustomPath)) + return false; + + /* + * If a bitmap scan's tlist is empty, keep it as-is. This may allow the + * executor to skip heap page fetches, and in any case, the benefit of + * using a physical tlist instead would be minimal. + */ + if (IsA(path, BitmapHeapPath) && + path->pathtarget->exprs == NIL) + return false; + + /* + * Can't do it if any system columns or whole-row Vars are requested. + * (This could possibly be fixed but would take some fragile assumptions + * in setrefs.c, I think.) + */ + for (i = rel->min_attr; i <= 0; i++) + { + if (!bms_is_empty(rel->attr_needed[i - rel->min_attr])) + return false; + } + + /* + * Can't do it if the rel is required to emit any placeholder expressions, + * either. + */ + foreach(lc, root->placeholder_list) + { + PlaceHolderInfo *phinfo = (PlaceHolderInfo *) lfirst(lc); + + if (bms_nonempty_difference(phinfo->ph_needed, rel->relids) && + bms_is_subset(phinfo->ph_eval_at, rel->relids)) + return false; + } + + /* + * For an index-only scan, the "physical tlist" is the index's indextlist. + * We can only return that without a projection if all the index's columns + * are returnable. + */ + if (path->pathtype == T_IndexOnlyScan) + { + IndexOptInfo *indexinfo = ((IndexPath *) path)->indexinfo; + + for (i = 0; i < indexinfo->ncolumns; i++) + { + if (!indexinfo->canreturn[i]) + return false; + } + } + + /* + * Also, can't do it if CP_LABEL_TLIST is specified and path is requested + * to emit any sort/group columns that are not simple Vars. (If they are + * simple Vars, they should appear in the physical tlist, and + * apply_pathtarget_labeling_to_tlist will take care of getting them + * labeled again.) We also have to check that no two sort/group columns + * are the same Var, else that element of the physical tlist would need + * conflicting ressortgroupref labels. + */ + if ((flags & CP_LABEL_TLIST) && path->pathtarget->sortgrouprefs) + { + Bitmapset *sortgroupatts = NULL; + + i = 0; + foreach(lc, path->pathtarget->exprs) + { + Expr *expr = (Expr *) lfirst(lc); + + if (path->pathtarget->sortgrouprefs[i]) + { + if (expr && IsA(expr, Var)) + { + int attno = ((Var *) expr)->varattno; + + attno -= FirstLowInvalidHeapAttributeNumber; + if (bms_is_member(attno, sortgroupatts)) + return false; + sortgroupatts = bms_add_member(sortgroupatts, attno); + } + else + return false; + } + i++; + } + } + + return true; +} + +/* + * get_gating_quals + * See if there are pseudoconstant quals in a node's quals list + * + * If the node's quals list includes any pseudoconstant quals, + * return just those quals. + */ +static List * +get_gating_quals(PlannerInfo *root, List *quals) +{ + /* No need to look if we know there are no pseudoconstants */ + if (!root->hasPseudoConstantQuals) + return NIL; + + /* Sort into desirable execution order while still in RestrictInfo form */ + quals = order_qual_clauses(root, quals); + + /* Pull out any pseudoconstant quals from the RestrictInfo list */ + return extract_actual_clauses(quals, true); +} + +/* + * create_gating_plan + * Deal with pseudoconstant qual clauses + * + * Add a gating Result node atop the already-built plan. + */ +static Plan * +create_gating_plan(PlannerInfo *root, Path *path, Plan *plan, + List *gating_quals) +{ + Plan *gplan; + Plan *splan; + + Assert(gating_quals); + + /* + * We might have a trivial Result plan already. Stacking one Result atop + * another is silly, so if that applies, just discard the input plan. + * (We're assuming its targetlist is uninteresting; it should be either + * the same as the result of build_path_tlist, or a simplified version.) + */ + splan = plan; + if (IsA(plan, Result)) + { + Result *rplan = (Result *) plan; + + if (rplan->plan.lefttree == NULL && + rplan->resconstantqual == NULL) + splan = NULL; + } + + /* + * Since we need a Result node anyway, always return the path's requested + * tlist; that's never a wrong choice, even if the parent node didn't ask + * for CP_EXACT_TLIST. + */ + gplan = (Plan *) make_result(build_path_tlist(root, path), + (Node *) gating_quals, + splan); + + /* + * Notice that we don't change cost or size estimates when doing gating. + * The costs of qual eval were already included in the subplan's cost. + * Leaving the size alone amounts to assuming that the gating qual will + * succeed, which is the conservative estimate for planning upper queries. + * We certainly don't want to assume the output size is zero (unless the + * gating qual is actually constant FALSE, and that case is dealt with in + * clausesel.c). Interpolating between the two cases is silly, because it + * doesn't reflect what will really happen at runtime, and besides which + * in most cases we have only a very bad idea of the probability of the + * gating qual being true. + */ + copy_plan_costsize(gplan, plan); + + /* Gating quals could be unsafe, so better use the Path's safety flag */ + gplan->parallel_safe = path->parallel_safe; + + return gplan; +} + +/* + * create_join_plan + * Create a join plan for 'best_path' and (recursively) plans for its + * inner and outer paths. + */ +static Plan * +create_join_plan(PlannerInfo *root, JoinPath *best_path) +{ + Plan *plan; + List *gating_clauses; + + switch (best_path->path.pathtype) + { + case T_MergeJoin: + plan = (Plan *) create_mergejoin_plan(root, + (MergePath *) best_path); + break; + case T_HashJoin: + plan = (Plan *) create_hashjoin_plan(root, + (HashPath *) best_path); + break; + case T_NestLoop: + plan = (Plan *) create_nestloop_plan(root, + (NestPath *) best_path); + break; + default: + elog(ERROR, "unrecognized node type: %d", + (int) best_path->path.pathtype); + plan = NULL; /* keep compiler quiet */ + break; + } + + /* + * If there are any pseudoconstant clauses attached to this node, insert a + * gating Result node that evaluates the pseudoconstants as one-time + * quals. + */ + gating_clauses = get_gating_quals(root, best_path->joinrestrictinfo); + if (gating_clauses) + plan = create_gating_plan(root, (Path *) best_path, plan, + gating_clauses); + +#ifdef NOT_USED + + /* + * * Expensive function pullups may have pulled local predicates * into + * this path node. Put them in the qpqual of the plan node. * JMH, + * 6/15/92 + */ + if (get_loc_restrictinfo(best_path) != NIL) + set_qpqual((Plan) plan, + list_concat(get_qpqual((Plan) plan), + get_actual_clauses(get_loc_restrictinfo(best_path)))); +#endif + + return plan; +} + +/* + * mark_async_capable_plan + * Check whether the Plan node created from a Path node is async-capable, + * and if so, mark the Plan node as such and return true, otherwise + * return false. + */ +static bool +mark_async_capable_plan(Plan *plan, Path *path) +{ + switch (nodeTag(path)) + { + case T_SubqueryScanPath: + { + SubqueryScan *scan_plan = (SubqueryScan *) plan; + + /* + * If the generated plan node includes a gating Result node, + * we can't execute it asynchronously. + */ + if (IsA(plan, Result)) + return false; + + /* + * If a SubqueryScan node atop of an async-capable plan node + * is deletable, consider it as async-capable. + */ + if (trivial_subqueryscan(scan_plan) && + mark_async_capable_plan(scan_plan->subplan, + ((SubqueryScanPath *) path)->subpath)) + break; + return false; + } + case T_ForeignPath: + { + FdwRoutine *fdwroutine = path->parent->fdwroutine; + + /* + * If the generated plan node includes a gating Result node, + * we can't execute it asynchronously. + */ + if (IsA(plan, Result)) + return false; + + Assert(fdwroutine != NULL); + if (fdwroutine->IsForeignPathAsyncCapable != NULL && + fdwroutine->IsForeignPathAsyncCapable((ForeignPath *) path)) + break; + return false; + } + case T_ProjectionPath: + + /* + * If the generated plan node includes a Result node for the + * projection, we can't execute it asynchronously. + */ + if (IsA(plan, Result)) + return false; + + /* + * create_projection_plan() would have pulled up the subplan, so + * check the capability using the subpath. + */ + if (mark_async_capable_plan(plan, + ((ProjectionPath *) path)->subpath)) + return true; + return false; + default: + return false; + } + + plan->async_capable = true; + + return true; +} + +/* + * create_append_plan + * Create an Append plan for 'best_path' and (recursively) plans + * for its subpaths. + * + * Returns a Plan node. + */ +static Plan * +create_append_plan(PlannerInfo *root, AppendPath *best_path, int flags) +{ + Append *plan; + List *tlist = build_path_tlist(root, &best_path->path); + int orig_tlist_length = list_length(tlist); + bool tlist_was_changed = false; + List *pathkeys = best_path->path.pathkeys; + List *subplans = NIL; + ListCell *subpaths; + int nasyncplans = 0; + RelOptInfo *rel = best_path->path.parent; + PartitionPruneInfo *partpruneinfo = NULL; + int nodenumsortkeys = 0; + AttrNumber *nodeSortColIdx = NULL; + Oid *nodeSortOperators = NULL; + Oid *nodeCollations = NULL; + bool *nodeNullsFirst = NULL; + bool consider_async = false; + + /* + * The subpaths list could be empty, if every child was proven empty by + * constraint exclusion. In that case generate a dummy plan that returns + * no rows. + * + * Note that an AppendPath with no members is also generated in certain + * cases where there was no appending construct at all, but we know the + * relation is empty (see set_dummy_rel_pathlist and mark_dummy_rel). + */ + if (best_path->subpaths == NIL) + { + /* Generate a Result plan with constant-FALSE gating qual */ + Plan *plan; + + plan = (Plan *) make_result(tlist, + (Node *) list_make1(makeBoolConst(false, + false)), + NULL); + + copy_generic_path_info(plan, (Path *) best_path); + + return plan; + } + + /* + * Otherwise build an Append plan. Note that if there's just one child, + * the Append is pretty useless; but we wait till setrefs.c to get rid of + * it. Doing so here doesn't work because the varno of the child scan + * plan won't match the parent-rel Vars it'll be asked to emit. + * + * We don't have the actual creation of the Append node split out into a + * separate make_xxx function. This is because we want to run + * prepare_sort_from_pathkeys on it before we do so on the individual + * child plans, to make cross-checking the sort info easier. + */ + plan = makeNode(Append); + plan->plan.targetlist = tlist; + plan->plan.qual = NIL; + plan->plan.lefttree = NULL; + plan->plan.righttree = NULL; + plan->apprelids = rel->relids; + + if (pathkeys != NIL) + { + /* + * Compute sort column info, and adjust the Append's tlist as needed. + * Because we pass adjust_tlist_in_place = true, we may ignore the + * function result; it must be the same plan node. However, we then + * need to detect whether any tlist entries were added. + */ + (void) prepare_sort_from_pathkeys((Plan *) plan, pathkeys, + best_path->path.parent->relids, + NULL, + true, + &nodenumsortkeys, + &nodeSortColIdx, + &nodeSortOperators, + &nodeCollations, + &nodeNullsFirst); + tlist_was_changed = (orig_tlist_length != list_length(plan->plan.targetlist)); + } + + /* If appropriate, consider async append */ + consider_async = (enable_async_append && pathkeys == NIL && + !best_path->path.parallel_safe && + list_length(best_path->subpaths) > 1); + + /* Build the plan for each child */ + foreach(subpaths, best_path->subpaths) + { + Path *subpath = (Path *) lfirst(subpaths); + Plan *subplan; + + /* Must insist that all children return the same tlist */ + subplan = create_plan_recurse(root, subpath, CP_EXACT_TLIST); + + /* + * For ordered Appends, we must insert a Sort node if subplan isn't + * sufficiently ordered. + */ + if (pathkeys != NIL) + { + int numsortkeys; + AttrNumber *sortColIdx; + Oid *sortOperators; + Oid *collations; + bool *nullsFirst; + + /* + * Compute sort column info, and adjust subplan's tlist as needed. + * We must apply prepare_sort_from_pathkeys even to subplans that + * don't need an explicit sort, to make sure they are returning + * the same sort key columns the Append expects. + */ + subplan = prepare_sort_from_pathkeys(subplan, pathkeys, + subpath->parent->relids, + nodeSortColIdx, + false, + &numsortkeys, + &sortColIdx, + &sortOperators, + &collations, + &nullsFirst); + + /* + * Check that we got the same sort key information. We just + * Assert that the sortops match, since those depend only on the + * pathkeys; but it seems like a good idea to check the sort + * column numbers explicitly, to ensure the tlists match up. + */ + Assert(numsortkeys == nodenumsortkeys); + if (memcmp(sortColIdx, nodeSortColIdx, + numsortkeys * sizeof(AttrNumber)) != 0) + elog(ERROR, "Append child's targetlist doesn't match Append"); + Assert(memcmp(sortOperators, nodeSortOperators, + numsortkeys * sizeof(Oid)) == 0); + Assert(memcmp(collations, nodeCollations, + numsortkeys * sizeof(Oid)) == 0); + Assert(memcmp(nullsFirst, nodeNullsFirst, + numsortkeys * sizeof(bool)) == 0); + + /* Now, insert a Sort node if subplan isn't sufficiently ordered */ + if (!pathkeys_contained_in(pathkeys, subpath->pathkeys)) + { + Sort *sort = make_sort(subplan, numsortkeys, + sortColIdx, sortOperators, + collations, nullsFirst); + + label_sort_with_costsize(root, sort, best_path->limit_tuples); + subplan = (Plan *) sort; + } + } + + /* If needed, check to see if subplan can be executed asynchronously */ + if (consider_async && mark_async_capable_plan(subplan, subpath)) + { + Assert(subplan->async_capable); + ++nasyncplans; + } + + subplans = lappend(subplans, subplan); + } + + /* + * If any quals exist, they may be useful to perform further partition + * pruning during execution. Gather information needed by the executor to + * do partition pruning. + */ + if (enable_partition_pruning) + { + List *prunequal; + + prunequal = extract_actual_clauses(rel->baserestrictinfo, false); + + if (best_path->path.param_info) + { + List *prmquals = best_path->path.param_info->ppi_clauses; + + prmquals = extract_actual_clauses(prmquals, false); + prmquals = (List *) replace_nestloop_params(root, + (Node *) prmquals); + + prunequal = list_concat(prunequal, prmquals); + } + + if (prunequal != NIL) + partpruneinfo = + make_partition_pruneinfo(root, rel, + best_path->subpaths, + prunequal); + } + + plan->appendplans = subplans; + plan->nasyncplans = nasyncplans; + plan->first_partial_plan = best_path->first_partial_path; + plan->part_prune_info = partpruneinfo; + + copy_generic_path_info(&plan->plan, (Path *) best_path); + + /* + * If prepare_sort_from_pathkeys added sort columns, but we were told to + * produce either the exact tlist or a narrow tlist, we should get rid of + * the sort columns again. We must inject a projection node to do so. + */ + if (tlist_was_changed && (flags & (CP_EXACT_TLIST | CP_SMALL_TLIST))) + { + tlist = list_truncate(list_copy(plan->plan.targetlist), + orig_tlist_length); + return inject_projection_plan((Plan *) plan, tlist, + plan->plan.parallel_safe); + } + else + return (Plan *) plan; +} + +/* + * create_merge_append_plan + * Create a MergeAppend plan for 'best_path' and (recursively) plans + * for its subpaths. + * + * Returns a Plan node. + */ +static Plan * +create_merge_append_plan(PlannerInfo *root, MergeAppendPath *best_path, + int flags) +{ + MergeAppend *node = makeNode(MergeAppend); + Plan *plan = &node->plan; + List *tlist = build_path_tlist(root, &best_path->path); + int orig_tlist_length = list_length(tlist); + bool tlist_was_changed; + List *pathkeys = best_path->path.pathkeys; + List *subplans = NIL; + ListCell *subpaths; + RelOptInfo *rel = best_path->path.parent; + PartitionPruneInfo *partpruneinfo = NULL; + + /* + * We don't have the actual creation of the MergeAppend node split out + * into a separate make_xxx function. This is because we want to run + * prepare_sort_from_pathkeys on it before we do so on the individual + * child plans, to make cross-checking the sort info easier. + */ + copy_generic_path_info(plan, (Path *) best_path); + plan->targetlist = tlist; + plan->qual = NIL; + plan->lefttree = NULL; + plan->righttree = NULL; + node->apprelids = rel->relids; + + /* + * Compute sort column info, and adjust MergeAppend's tlist as needed. + * Because we pass adjust_tlist_in_place = true, we may ignore the + * function result; it must be the same plan node. However, we then need + * to detect whether any tlist entries were added. + */ + (void) prepare_sort_from_pathkeys(plan, pathkeys, + best_path->path.parent->relids, + NULL, + true, + &node->numCols, + &node->sortColIdx, + &node->sortOperators, + &node->collations, + &node->nullsFirst); + tlist_was_changed = (orig_tlist_length != list_length(plan->targetlist)); + + /* + * Now prepare the child plans. We must apply prepare_sort_from_pathkeys + * even to subplans that don't need an explicit sort, to make sure they + * are returning the same sort key columns the MergeAppend expects. + */ + foreach(subpaths, best_path->subpaths) + { + Path *subpath = (Path *) lfirst(subpaths); + Plan *subplan; + int numsortkeys; + AttrNumber *sortColIdx; + Oid *sortOperators; + Oid *collations; + bool *nullsFirst; + + /* Build the child plan */ + /* Must insist that all children return the same tlist */ + subplan = create_plan_recurse(root, subpath, CP_EXACT_TLIST); + + /* Compute sort column info, and adjust subplan's tlist as needed */ + subplan = prepare_sort_from_pathkeys(subplan, pathkeys, + subpath->parent->relids, + node->sortColIdx, + false, + &numsortkeys, + &sortColIdx, + &sortOperators, + &collations, + &nullsFirst); + + /* + * Check that we got the same sort key information. We just Assert + * that the sortops match, since those depend only on the pathkeys; + * but it seems like a good idea to check the sort column numbers + * explicitly, to ensure the tlists really do match up. + */ + Assert(numsortkeys == node->numCols); + if (memcmp(sortColIdx, node->sortColIdx, + numsortkeys * sizeof(AttrNumber)) != 0) + elog(ERROR, "MergeAppend child's targetlist doesn't match MergeAppend"); + Assert(memcmp(sortOperators, node->sortOperators, + numsortkeys * sizeof(Oid)) == 0); + Assert(memcmp(collations, node->collations, + numsortkeys * sizeof(Oid)) == 0); + Assert(memcmp(nullsFirst, node->nullsFirst, + numsortkeys * sizeof(bool)) == 0); + + /* Now, insert a Sort node if subplan isn't sufficiently ordered */ + if (!pathkeys_contained_in(pathkeys, subpath->pathkeys)) + { + Sort *sort = make_sort(subplan, numsortkeys, + sortColIdx, sortOperators, + collations, nullsFirst); + + label_sort_with_costsize(root, sort, best_path->limit_tuples); + subplan = (Plan *) sort; + } + + subplans = lappend(subplans, subplan); + } + + /* + * If any quals exist, they may be useful to perform further partition + * pruning during execution. Gather information needed by the executor to + * do partition pruning. + */ + if (enable_partition_pruning) + { + List *prunequal; + + prunequal = extract_actual_clauses(rel->baserestrictinfo, false); + + if (best_path->path.param_info) + { + List *prmquals = best_path->path.param_info->ppi_clauses; + + prmquals = extract_actual_clauses(prmquals, false); + prmquals = (List *) replace_nestloop_params(root, + (Node *) prmquals); + + prunequal = list_concat(prunequal, prmquals); + } + + if (prunequal != NIL) + partpruneinfo = make_partition_pruneinfo(root, rel, + best_path->subpaths, + prunequal); + } + + node->mergeplans = subplans; + node->part_prune_info = partpruneinfo; + + /* + * If prepare_sort_from_pathkeys added sort columns, but we were told to + * produce either the exact tlist or a narrow tlist, we should get rid of + * the sort columns again. We must inject a projection node to do so. + */ + if (tlist_was_changed && (flags & (CP_EXACT_TLIST | CP_SMALL_TLIST))) + { + tlist = list_truncate(list_copy(plan->targetlist), orig_tlist_length); + return inject_projection_plan(plan, tlist, plan->parallel_safe); + } + else + return plan; +} + +/* + * create_group_result_plan + * Create a Result plan for 'best_path'. + * This is only used for degenerate grouping cases. + * + * Returns a Plan node. + */ +static Result * +create_group_result_plan(PlannerInfo *root, GroupResultPath *best_path) +{ + Result *plan; + List *tlist; + List *quals; + + tlist = build_path_tlist(root, &best_path->path); + + /* best_path->quals is just bare clauses */ + quals = order_qual_clauses(root, best_path->quals); + + plan = make_result(tlist, (Node *) quals, NULL); + + copy_generic_path_info(&plan->plan, (Path *) best_path); + + return plan; +} + +/* + * create_project_set_plan + * Create a ProjectSet plan for 'best_path'. + * + * Returns a Plan node. + */ +static ProjectSet * +create_project_set_plan(PlannerInfo *root, ProjectSetPath *best_path) +{ + ProjectSet *plan; + Plan *subplan; + List *tlist; + + /* Since we intend to project, we don't need to constrain child tlist */ + subplan = create_plan_recurse(root, best_path->subpath, 0); + + tlist = build_path_tlist(root, &best_path->path); + + plan = make_project_set(tlist, subplan); + + copy_generic_path_info(&plan->plan, (Path *) best_path); + + return plan; +} + +/* + * create_material_plan + * Create a Material plan for 'best_path' and (recursively) plans + * for its subpaths. + * + * Returns a Plan node. + */ +static Material * +create_material_plan(PlannerInfo *root, MaterialPath *best_path, int flags) +{ + Material *plan; + Plan *subplan; + + /* + * We don't want any excess columns in the materialized tuples, so request + * a smaller tlist. Otherwise, since Material doesn't project, tlist + * requirements pass through. + */ + subplan = create_plan_recurse(root, best_path->subpath, + flags | CP_SMALL_TLIST); + + plan = make_material(subplan); + + copy_generic_path_info(&plan->plan, (Path *) best_path); + + return plan; +} + +/* + * create_memoize_plan + * Create a Memoize plan for 'best_path' and (recursively) plans for its + * subpaths. + * + * Returns a Plan node. + */ +static Memoize * +create_memoize_plan(PlannerInfo *root, MemoizePath *best_path, int flags) +{ + Memoize *plan; + Bitmapset *keyparamids; + Plan *subplan; + Oid *operators; + Oid *collations; + List *param_exprs = NIL; + ListCell *lc; + ListCell *lc2; + int nkeys; + int i; + + subplan = create_plan_recurse(root, best_path->subpath, + flags | CP_SMALL_TLIST); + + param_exprs = (List *) replace_nestloop_params(root, (Node *) + best_path->param_exprs); + + nkeys = list_length(param_exprs); + Assert(nkeys > 0); + operators = palloc(nkeys * sizeof(Oid)); + collations = palloc(nkeys * sizeof(Oid)); + + i = 0; + forboth(lc, param_exprs, lc2, best_path->hash_operators) + { + Expr *param_expr = (Expr *) lfirst(lc); + Oid opno = lfirst_oid(lc2); + + operators[i] = opno; + collations[i] = exprCollation((Node *) param_expr); + i++; + } + + keyparamids = pull_paramids((Expr *) param_exprs); + + plan = make_memoize(subplan, operators, collations, param_exprs, + best_path->singlerow, best_path->binary_mode, + best_path->est_entries, keyparamids); + + copy_generic_path_info(&plan->plan, (Path *) best_path); + + return plan; +} + +/* + * create_unique_plan + * Create a Unique plan for 'best_path' and (recursively) plans + * for its subpaths. + * + * Returns a Plan node. + */ +static Plan * +create_unique_plan(PlannerInfo *root, UniquePath *best_path, int flags) +{ + Plan *plan; + Plan *subplan; + List *in_operators; + List *uniq_exprs; + List *newtlist; + int nextresno; + bool newitems; + int numGroupCols; + AttrNumber *groupColIdx; + Oid *groupCollations; + int groupColPos; + ListCell *l; + + /* Unique doesn't project, so tlist requirements pass through */ + subplan = create_plan_recurse(root, best_path->subpath, flags); + + /* Done if we don't need to do any actual unique-ifying */ + if (best_path->umethod == UNIQUE_PATH_NOOP) + return subplan; + + /* + * As constructed, the subplan has a "flat" tlist containing just the Vars + * needed here and at upper levels. The values we are supposed to + * unique-ify may be expressions in these variables. We have to add any + * such expressions to the subplan's tlist. + * + * The subplan may have a "physical" tlist if it is a simple scan plan. If + * we're going to sort, this should be reduced to the regular tlist, so + * that we don't sort more data than we need to. For hashing, the tlist + * should be left as-is if we don't need to add any expressions; but if we + * do have to add expressions, then a projection step will be needed at + * runtime anyway, so we may as well remove unneeded items. Therefore + * newtlist starts from build_path_tlist() not just a copy of the + * subplan's tlist; and we don't install it into the subplan unless we are + * sorting or stuff has to be added. + */ + in_operators = best_path->in_operators; + uniq_exprs = best_path->uniq_exprs; + + /* initialize modified subplan tlist as just the "required" vars */ + newtlist = build_path_tlist(root, &best_path->path); + nextresno = list_length(newtlist) + 1; + newitems = false; + + foreach(l, uniq_exprs) + { + Expr *uniqexpr = lfirst(l); + TargetEntry *tle; + + tle = tlist_member(uniqexpr, newtlist); + if (!tle) + { + tle = makeTargetEntry((Expr *) uniqexpr, + nextresno, + NULL, + false); + newtlist = lappend(newtlist, tle); + nextresno++; + newitems = true; + } + } + + /* Use change_plan_targetlist in case we need to insert a Result node */ + if (newitems || best_path->umethod == UNIQUE_PATH_SORT) + subplan = change_plan_targetlist(subplan, newtlist, + best_path->path.parallel_safe); + + /* + * Build control information showing which subplan output columns are to + * be examined by the grouping step. Unfortunately we can't merge this + * with the previous loop, since we didn't then know which version of the + * subplan tlist we'd end up using. + */ + newtlist = subplan->targetlist; + numGroupCols = list_length(uniq_exprs); + groupColIdx = (AttrNumber *) palloc(numGroupCols * sizeof(AttrNumber)); + groupCollations = (Oid *) palloc(numGroupCols * sizeof(Oid)); + + groupColPos = 0; + foreach(l, uniq_exprs) + { + Expr *uniqexpr = lfirst(l); + TargetEntry *tle; + + tle = tlist_member(uniqexpr, newtlist); + if (!tle) /* shouldn't happen */ + elog(ERROR, "failed to find unique expression in subplan tlist"); + groupColIdx[groupColPos] = tle->resno; + groupCollations[groupColPos] = exprCollation((Node *) tle->expr); + groupColPos++; + } + + if (best_path->umethod == UNIQUE_PATH_HASH) + { + Oid *groupOperators; + + /* + * Get the hashable equality operators for the Agg node to use. + * Normally these are the same as the IN clause operators, but if + * those are cross-type operators then the equality operators are the + * ones for the IN clause operators' RHS datatype. + */ + groupOperators = (Oid *) palloc(numGroupCols * sizeof(Oid)); + groupColPos = 0; + foreach(l, in_operators) + { + Oid in_oper = lfirst_oid(l); + Oid eq_oper; + + if (!get_compatible_hash_operators(in_oper, NULL, &eq_oper)) + elog(ERROR, "could not find compatible hash operator for operator %u", + in_oper); + groupOperators[groupColPos++] = eq_oper; + } + + /* + * Since the Agg node is going to project anyway, we can give it the + * minimum output tlist, without any stuff we might have added to the + * subplan tlist. + */ + plan = (Plan *) make_agg(build_path_tlist(root, &best_path->path), + NIL, + AGG_HASHED, + AGGSPLIT_SIMPLE, + numGroupCols, + groupColIdx, + groupOperators, + groupCollations, + NIL, + NIL, + best_path->path.rows, + 0, + subplan); + } + else + { + List *sortList = NIL; + Sort *sort; + + /* Create an ORDER BY list to sort the input compatibly */ + groupColPos = 0; + foreach(l, in_operators) + { + Oid in_oper = lfirst_oid(l); + Oid sortop; + Oid eqop; + TargetEntry *tle; + SortGroupClause *sortcl; + + sortop = get_ordering_op_for_equality_op(in_oper, false); + if (!OidIsValid(sortop)) /* shouldn't happen */ + elog(ERROR, "could not find ordering operator for equality operator %u", + in_oper); + + /* + * The Unique node will need equality operators. Normally these + * are the same as the IN clause operators, but if those are + * cross-type operators then the equality operators are the ones + * for the IN clause operators' RHS datatype. + */ + eqop = get_equality_op_for_ordering_op(sortop, NULL); + if (!OidIsValid(eqop)) /* shouldn't happen */ + elog(ERROR, "could not find equality operator for ordering operator %u", + sortop); + + tle = get_tle_by_resno(subplan->targetlist, + groupColIdx[groupColPos]); + Assert(tle != NULL); + + sortcl = makeNode(SortGroupClause); + sortcl->tleSortGroupRef = assignSortGroupRef(tle, + subplan->targetlist); + sortcl->eqop = eqop; + sortcl->sortop = sortop; + sortcl->nulls_first = false; + sortcl->hashable = false; /* no need to make this accurate */ + sortList = lappend(sortList, sortcl); + groupColPos++; + } + sort = make_sort_from_sortclauses(sortList, subplan); + label_sort_with_costsize(root, sort, -1.0); + plan = (Plan *) make_unique_from_sortclauses((Plan *) sort, sortList); + } + + /* Copy cost data from Path to Plan */ + copy_generic_path_info(plan, &best_path->path); + + return plan; +} + +/* + * create_gather_plan + * + * Create a Gather plan for 'best_path' and (recursively) plans + * for its subpaths. + */ +static Gather * +create_gather_plan(PlannerInfo *root, GatherPath *best_path) +{ + Gather *gather_plan; + Plan *subplan; + List *tlist; + + /* + * Push projection down to the child node. That way, the projection work + * is parallelized, and there can be no system columns in the result (they + * can't travel through a tuple queue because it uses MinimalTuple + * representation). + */ + subplan = create_plan_recurse(root, best_path->subpath, CP_EXACT_TLIST); + + tlist = build_path_tlist(root, &best_path->path); + + gather_plan = make_gather(tlist, + NIL, + best_path->num_workers, + assign_special_exec_param(root), + best_path->single_copy, + subplan); + + copy_generic_path_info(&gather_plan->plan, &best_path->path); + + /* use parallel mode for parallel plans. */ + root->glob->parallelModeNeeded = true; + + return gather_plan; +} + +/* + * create_gather_merge_plan + * + * Create a Gather Merge plan for 'best_path' and (recursively) + * plans for its subpaths. + */ +static GatherMerge * +create_gather_merge_plan(PlannerInfo *root, GatherMergePath *best_path) +{ + GatherMerge *gm_plan; + Plan *subplan; + List *pathkeys = best_path->path.pathkeys; + List *tlist = build_path_tlist(root, &best_path->path); + + /* As with Gather, project away columns in the workers. */ + subplan = create_plan_recurse(root, best_path->subpath, CP_EXACT_TLIST); + + /* Create a shell for a GatherMerge plan. */ + gm_plan = makeNode(GatherMerge); + gm_plan->plan.targetlist = tlist; + gm_plan->num_workers = best_path->num_workers; + copy_generic_path_info(&gm_plan->plan, &best_path->path); + + /* Assign the rescan Param. */ + gm_plan->rescan_param = assign_special_exec_param(root); + + /* Gather Merge is pointless with no pathkeys; use Gather instead. */ + Assert(pathkeys != NIL); + + /* Compute sort column info, and adjust subplan's tlist as needed */ + subplan = prepare_sort_from_pathkeys(subplan, pathkeys, + best_path->subpath->parent->relids, + gm_plan->sortColIdx, + false, + &gm_plan->numCols, + &gm_plan->sortColIdx, + &gm_plan->sortOperators, + &gm_plan->collations, + &gm_plan->nullsFirst); + + + /* + * All gather merge paths should have already guaranteed the necessary + * sort order either by adding an explicit sort node or by using presorted + * input. We can't simply add a sort here on additional pathkeys, because + * we can't guarantee the sort would be safe. For example, expressions may + * be volatile or otherwise parallel unsafe. + */ + if (!pathkeys_contained_in(pathkeys, best_path->subpath->pathkeys)) + elog(ERROR, "gather merge input not sufficiently sorted"); + + /* Now insert the subplan under GatherMerge. */ + gm_plan->plan.lefttree = subplan; + + /* use parallel mode for parallel plans. */ + root->glob->parallelModeNeeded = true; + + return gm_plan; +} + +/* + * create_projection_plan + * + * Create a plan tree to do a projection step and (recursively) plans + * for its subpaths. We may need a Result node for the projection, + * but sometimes we can just let the subplan do the work. + */ +static Plan * +create_projection_plan(PlannerInfo *root, ProjectionPath *best_path, int flags) +{ + Plan *plan; + Plan *subplan; + List *tlist; + bool needs_result_node = false; + + /* + * Convert our subpath to a Plan and determine whether we need a Result + * node. + * + * In most cases where we don't need to project, creation_projection_path + * will have set dummypp, but not always. First, some createplan.c + * routines change the tlists of their nodes. (An example is that + * create_merge_append_plan might add resjunk sort columns to a + * MergeAppend.) Second, create_projection_path has no way of knowing + * what path node will be placed on top of the projection path and + * therefore can't predict whether it will require an exact tlist. For + * both of these reasons, we have to recheck here. + */ + if (use_physical_tlist(root, &best_path->path, flags)) + { + /* + * Our caller doesn't really care what tlist we return, so we don't + * actually need to project. However, we may still need to ensure + * proper sortgroupref labels, if the caller cares about those. + */ + subplan = create_plan_recurse(root, best_path->subpath, 0); + tlist = subplan->targetlist; + if (flags & CP_LABEL_TLIST) + apply_pathtarget_labeling_to_tlist(tlist, + best_path->path.pathtarget); + } + else if (is_projection_capable_path(best_path->subpath)) + { + /* + * Our caller requires that we return the exact tlist, but no separate + * result node is needed because the subpath is projection-capable. + * Tell create_plan_recurse that we're going to ignore the tlist it + * produces. + */ + subplan = create_plan_recurse(root, best_path->subpath, + CP_IGNORE_TLIST); + Assert(is_projection_capable_plan(subplan)); + tlist = build_path_tlist(root, &best_path->path); + } + else + { + /* + * It looks like we need a result node, unless by good fortune the + * requested tlist is exactly the one the child wants to produce. + */ + subplan = create_plan_recurse(root, best_path->subpath, 0); + tlist = build_path_tlist(root, &best_path->path); + needs_result_node = !tlist_same_exprs(tlist, subplan->targetlist); + } + + /* + * If we make a different decision about whether to include a Result node + * than create_projection_path did, we'll have made slightly wrong cost + * estimates; but label the plan with the cost estimates we actually used, + * not "corrected" ones. (XXX this could be cleaned up if we moved more + * of the sortcolumn setup logic into Path creation, but that would add + * expense to creating Paths we might end up not using.) + */ + if (!needs_result_node) + { + /* Don't need a separate Result, just assign tlist to subplan */ + plan = subplan; + plan->targetlist = tlist; + + /* Label plan with the estimated costs we actually used */ + plan->startup_cost = best_path->path.startup_cost; + plan->total_cost = best_path->path.total_cost; + plan->plan_rows = best_path->path.rows; + plan->plan_width = best_path->path.pathtarget->width; + plan->parallel_safe = best_path->path.parallel_safe; + /* ... but don't change subplan's parallel_aware flag */ + } + else + { + /* We need a Result node */ + plan = (Plan *) make_result(tlist, NULL, subplan); + + copy_generic_path_info(plan, (Path *) best_path); + } + + return plan; +} + +/* + * inject_projection_plan + * Insert a Result node to do a projection step. + * + * This is used in a few places where we decide on-the-fly that we need a + * projection step as part of the tree generated for some Path node. + * We should try to get rid of this in favor of doing it more honestly. + * + * One reason it's ugly is we have to be told the right parallel_safe marking + * to apply (since the tlist might be unsafe even if the child plan is safe). + */ +static Plan * +inject_projection_plan(Plan *subplan, List *tlist, bool parallel_safe) +{ + Plan *plan; + + plan = (Plan *) make_result(tlist, NULL, subplan); + + /* + * In principle, we should charge tlist eval cost plus cpu_per_tuple per + * row for the Result node. But the former has probably been factored in + * already and the latter was not accounted for during Path construction, + * so being formally correct might just make the EXPLAIN output look less + * consistent not more so. Hence, just copy the subplan's cost. + */ + copy_plan_costsize(plan, subplan); + plan->parallel_safe = parallel_safe; + + return plan; +} + +/* + * change_plan_targetlist + * Externally available wrapper for inject_projection_plan. + * + * This is meant for use by FDW plan-generation functions, which might + * want to adjust the tlist computed by some subplan tree. In general, + * a Result node is needed to compute the new tlist, but we can optimize + * some cases. + * + * In most cases, tlist_parallel_safe can just be passed as the parallel_safe + * flag of the FDW's own Path node. + */ +Plan * +change_plan_targetlist(Plan *subplan, List *tlist, bool tlist_parallel_safe) +{ + /* + * If the top plan node can't do projections and its existing target list + * isn't already what we need, we need to add a Result node to help it + * along. + */ + if (!is_projection_capable_plan(subplan) && + !tlist_same_exprs(tlist, subplan->targetlist)) + subplan = inject_projection_plan(subplan, tlist, + subplan->parallel_safe && + tlist_parallel_safe); + else + { + /* Else we can just replace the plan node's tlist */ + subplan->targetlist = tlist; + subplan->parallel_safe &= tlist_parallel_safe; + } + return subplan; +} + +/* + * create_sort_plan + * + * Create a Sort plan for 'best_path' and (recursively) plans + * for its subpaths. + */ +static Sort * +create_sort_plan(PlannerInfo *root, SortPath *best_path, int flags) +{ + Sort *plan; + Plan *subplan; + + /* + * We don't want any excess columns in the sorted tuples, so request a + * smaller tlist. Otherwise, since Sort doesn't project, tlist + * requirements pass through. + */ + subplan = create_plan_recurse(root, best_path->subpath, + flags | CP_SMALL_TLIST); + + /* + * make_sort_from_pathkeys indirectly calls find_ec_member_matching_expr, + * which will ignore any child EC members that don't belong to the given + * relids. Thus, if this sort path is based on a child relation, we must + * pass its relids. + */ + plan = make_sort_from_pathkeys(subplan, best_path->path.pathkeys, + IS_OTHER_REL(best_path->subpath->parent) ? + best_path->path.parent->relids : NULL); + + copy_generic_path_info(&plan->plan, (Path *) best_path); + + return plan; +} + +/* + * create_incrementalsort_plan + * + * Do the same as create_sort_plan, but create IncrementalSort plan. + */ +static IncrementalSort * +create_incrementalsort_plan(PlannerInfo *root, IncrementalSortPath *best_path, + int flags) +{ + IncrementalSort *plan; + Plan *subplan; + + /* See comments in create_sort_plan() above */ + subplan = create_plan_recurse(root, best_path->spath.subpath, + flags | CP_SMALL_TLIST); + plan = make_incrementalsort_from_pathkeys(subplan, + best_path->spath.path.pathkeys, + IS_OTHER_REL(best_path->spath.subpath->parent) ? + best_path->spath.path.parent->relids : NULL, + best_path->nPresortedCols); + + copy_generic_path_info(&plan->sort.plan, (Path *) best_path); + + return plan; +} + +/* + * create_group_plan + * + * Create a Group plan for 'best_path' and (recursively) plans + * for its subpaths. + */ +static Group * +create_group_plan(PlannerInfo *root, GroupPath *best_path) +{ + Group *plan; + Plan *subplan; + List *tlist; + List *quals; + + /* + * Group can project, so no need to be terribly picky about child tlist, + * but we do need grouping columns to be available + */ + subplan = create_plan_recurse(root, best_path->subpath, CP_LABEL_TLIST); + + tlist = build_path_tlist(root, &best_path->path); + + quals = order_qual_clauses(root, best_path->qual); + + plan = make_group(tlist, + quals, + list_length(best_path->groupClause), + extract_grouping_cols(best_path->groupClause, + subplan->targetlist), + extract_grouping_ops(best_path->groupClause), + extract_grouping_collations(best_path->groupClause, + subplan->targetlist), + subplan); + + copy_generic_path_info(&plan->plan, (Path *) best_path); + + return plan; +} + +/* + * create_upper_unique_plan + * + * Create a Unique plan for 'best_path' and (recursively) plans + * for its subpaths. + */ +static Unique * +create_upper_unique_plan(PlannerInfo *root, UpperUniquePath *best_path, int flags) +{ + Unique *plan; + Plan *subplan; + + /* + * Unique doesn't project, so tlist requirements pass through; moreover we + * need grouping columns to be labeled. + */ + subplan = create_plan_recurse(root, best_path->subpath, + flags | CP_LABEL_TLIST); + + plan = make_unique_from_pathkeys(subplan, + best_path->path.pathkeys, + best_path->numkeys); + + copy_generic_path_info(&plan->plan, (Path *) best_path); + + return plan; +} + +/* + * create_agg_plan + * + * Create an Agg plan for 'best_path' and (recursively) plans + * for its subpaths. + */ +static Agg * +create_agg_plan(PlannerInfo *root, AggPath *best_path) +{ + Agg *plan; + Plan *subplan; + List *tlist; + List *quals; + + /* + * Agg can project, so no need to be terribly picky about child tlist, but + * we do need grouping columns to be available + */ + subplan = create_plan_recurse(root, best_path->subpath, CP_LABEL_TLIST); + + tlist = build_path_tlist(root, &best_path->path); + + quals = order_qual_clauses(root, best_path->qual); + + plan = make_agg(tlist, quals, + best_path->aggstrategy, + best_path->aggsplit, + list_length(best_path->groupClause), + extract_grouping_cols(best_path->groupClause, + subplan->targetlist), + extract_grouping_ops(best_path->groupClause), + extract_grouping_collations(best_path->groupClause, + subplan->targetlist), + NIL, + NIL, + best_path->numGroups, + best_path->transitionSpace, + subplan); + + copy_generic_path_info(&plan->plan, (Path *) best_path); + + return plan; +} + +/* + * Given a groupclause for a collection of grouping sets, produce the + * corresponding groupColIdx. + * + * root->grouping_map maps the tleSortGroupRef to the actual column position in + * the input tuple. So we get the ref from the entries in the groupclause and + * look them up there. + */ +static AttrNumber * +remap_groupColIdx(PlannerInfo *root, List *groupClause) +{ + AttrNumber *grouping_map = root->grouping_map; + AttrNumber *new_grpColIdx; + ListCell *lc; + int i; + + Assert(grouping_map); + + new_grpColIdx = palloc0(sizeof(AttrNumber) * list_length(groupClause)); + + i = 0; + foreach(lc, groupClause) + { + SortGroupClause *clause = lfirst(lc); + + new_grpColIdx[i++] = grouping_map[clause->tleSortGroupRef]; + } + + return new_grpColIdx; +} + +/* + * create_groupingsets_plan + * Create a plan for 'best_path' and (recursively) plans + * for its subpaths. + * + * What we emit is an Agg plan with some vestigial Agg and Sort nodes + * hanging off the side. The top Agg implements the last grouping set + * specified in the GroupingSetsPath, and any additional grouping sets + * each give rise to a subsidiary Agg and Sort node in the top Agg's + * "chain" list. These nodes don't participate in the plan directly, + * but they are a convenient way to represent the required data for + * the extra steps. + * + * Returns a Plan node. + */ +static Plan * +create_groupingsets_plan(PlannerInfo *root, GroupingSetsPath *best_path) +{ + Agg *plan; + Plan *subplan; + List *rollups = best_path->rollups; + AttrNumber *grouping_map; + int maxref; + List *chain; + ListCell *lc; + + /* Shouldn't get here without grouping sets */ + Assert(root->parse->groupingSets); + Assert(rollups != NIL); + + /* + * Agg can project, so no need to be terribly picky about child tlist, but + * we do need grouping columns to be available + */ + subplan = create_plan_recurse(root, best_path->subpath, CP_LABEL_TLIST); + + /* + * Compute the mapping from tleSortGroupRef to column index in the child's + * tlist. First, identify max SortGroupRef in groupClause, for array + * sizing. + */ + maxref = 0; + foreach(lc, root->parse->groupClause) + { + SortGroupClause *gc = (SortGroupClause *) lfirst(lc); + + if (gc->tleSortGroupRef > maxref) + maxref = gc->tleSortGroupRef; + } + + grouping_map = (AttrNumber *) palloc0((maxref + 1) * sizeof(AttrNumber)); + + /* Now look up the column numbers in the child's tlist */ + foreach(lc, root->parse->groupClause) + { + SortGroupClause *gc = (SortGroupClause *) lfirst(lc); + TargetEntry *tle = get_sortgroupclause_tle(gc, subplan->targetlist); + + grouping_map[gc->tleSortGroupRef] = tle->resno; + } + + /* + * During setrefs.c, we'll need the grouping_map to fix up the cols lists + * in GroupingFunc nodes. Save it for setrefs.c to use. + */ + Assert(root->grouping_map == NULL); + root->grouping_map = grouping_map; + + /* + * Generate the side nodes that describe the other sort and group + * operations besides the top one. Note that we don't worry about putting + * accurate cost estimates in the side nodes; only the topmost Agg node's + * costs will be shown by EXPLAIN. + */ + chain = NIL; + if (list_length(rollups) > 1) + { + bool is_first_sort = ((RollupData *) linitial(rollups))->is_hashed; + + for_each_from(lc, rollups, 1) + { + RollupData *rollup = lfirst(lc); + AttrNumber *new_grpColIdx; + Plan *sort_plan = NULL; + Plan *agg_plan; + AggStrategy strat; + + new_grpColIdx = remap_groupColIdx(root, rollup->groupClause); + + if (!rollup->is_hashed && !is_first_sort) + { + sort_plan = (Plan *) + make_sort_from_groupcols(rollup->groupClause, + new_grpColIdx, + subplan); + } + + if (!rollup->is_hashed) + is_first_sort = false; + + if (rollup->is_hashed) + strat = AGG_HASHED; + else if (list_length(linitial(rollup->gsets)) == 0) + strat = AGG_PLAIN; + else + strat = AGG_SORTED; + + agg_plan = (Plan *) make_agg(NIL, + NIL, + strat, + AGGSPLIT_SIMPLE, + list_length((List *) linitial(rollup->gsets)), + new_grpColIdx, + extract_grouping_ops(rollup->groupClause), + extract_grouping_collations(rollup->groupClause, subplan->targetlist), + rollup->gsets, + NIL, + rollup->numGroups, + best_path->transitionSpace, + sort_plan); + + /* + * Remove stuff we don't need to avoid bloating debug output. + */ + if (sort_plan) + { + sort_plan->targetlist = NIL; + sort_plan->lefttree = NULL; + } + + chain = lappend(chain, agg_plan); + } + } + + /* + * Now make the real Agg node + */ + { + RollupData *rollup = linitial(rollups); + AttrNumber *top_grpColIdx; + int numGroupCols; + + top_grpColIdx = remap_groupColIdx(root, rollup->groupClause); + + numGroupCols = list_length((List *) linitial(rollup->gsets)); + + plan = make_agg(build_path_tlist(root, &best_path->path), + best_path->qual, + best_path->aggstrategy, + AGGSPLIT_SIMPLE, + numGroupCols, + top_grpColIdx, + extract_grouping_ops(rollup->groupClause), + extract_grouping_collations(rollup->groupClause, subplan->targetlist), + rollup->gsets, + chain, + rollup->numGroups, + best_path->transitionSpace, + subplan); + + /* Copy cost data from Path to Plan */ + copy_generic_path_info(&plan->plan, &best_path->path); + } + + return (Plan *) plan; +} + +/* + * create_minmaxagg_plan + * + * Create a Result plan for 'best_path' and (recursively) plans + * for its subpaths. + */ +static Result * +create_minmaxagg_plan(PlannerInfo *root, MinMaxAggPath *best_path) +{ + Result *plan; + List *tlist; + ListCell *lc; + + /* Prepare an InitPlan for each aggregate's subquery. */ + foreach(lc, best_path->mmaggregates) + { + MinMaxAggInfo *mminfo = (MinMaxAggInfo *) lfirst(lc); + PlannerInfo *subroot = mminfo->subroot; + Query *subparse = subroot->parse; + Plan *plan; + + /* + * Generate the plan for the subquery. We already have a Path, but we + * have to convert it to a Plan and attach a LIMIT node above it. + * Since we are entering a different planner context (subroot), + * recurse to create_plan not create_plan_recurse. + */ + plan = create_plan(subroot, mminfo->path); + + plan = (Plan *) make_limit(plan, + subparse->limitOffset, + subparse->limitCount, + subparse->limitOption, + 0, NULL, NULL, NULL); + + /* Must apply correct cost/width data to Limit node */ + plan->startup_cost = mminfo->path->startup_cost; + plan->total_cost = mminfo->pathcost; + plan->plan_rows = 1; + plan->plan_width = mminfo->path->pathtarget->width; + plan->parallel_aware = false; + plan->parallel_safe = mminfo->path->parallel_safe; + + /* Convert the plan into an InitPlan in the outer query. */ + SS_make_initplan_from_plan(root, subroot, plan, mminfo->param); + } + + /* Generate the output plan --- basically just a Result */ + tlist = build_path_tlist(root, &best_path->path); + + plan = make_result(tlist, (Node *) best_path->quals, NULL); + + copy_generic_path_info(&plan->plan, (Path *) best_path); + + /* + * During setrefs.c, we'll need to replace references to the Agg nodes + * with InitPlan output params. (We can't just do that locally in the + * MinMaxAgg node, because path nodes above here may have Agg references + * as well.) Save the mmaggregates list to tell setrefs.c to do that. + */ + Assert(root->minmax_aggs == NIL); + root->minmax_aggs = best_path->mmaggregates; + + return plan; +} + +/* + * create_windowagg_plan + * + * Create a WindowAgg plan for 'best_path' and (recursively) plans + * for its subpaths. + */ +static WindowAgg * +create_windowagg_plan(PlannerInfo *root, WindowAggPath *best_path) +{ + WindowAgg *plan; + WindowClause *wc = best_path->winclause; + int numPart = list_length(wc->partitionClause); + int numOrder = list_length(wc->orderClause); + Plan *subplan; + List *tlist; + int partNumCols; + AttrNumber *partColIdx; + Oid *partOperators; + Oid *partCollations; + int ordNumCols; + AttrNumber *ordColIdx; + Oid *ordOperators; + Oid *ordCollations; + ListCell *lc; + + /* + * Choice of tlist here is motivated by the fact that WindowAgg will be + * storing the input rows of window frames in a tuplestore; it therefore + * behooves us to request a small tlist to avoid wasting space. We do of + * course need grouping columns to be available. + */ + subplan = create_plan_recurse(root, best_path->subpath, + CP_LABEL_TLIST | CP_SMALL_TLIST); + + tlist = build_path_tlist(root, &best_path->path); + + /* + * Convert SortGroupClause lists into arrays of attr indexes and equality + * operators, as wanted by executor. (Note: in principle, it's possible + * to drop some of the sort columns, if they were proved redundant by + * pathkey logic. However, it doesn't seem worth going out of our way to + * optimize such cases. In any case, we must *not* remove the ordering + * column for RANGE OFFSET cases, as the executor needs that for in_range + * tests even if it's known to be equal to some partitioning column.) + */ + partColIdx = (AttrNumber *) palloc(sizeof(AttrNumber) * numPart); + partOperators = (Oid *) palloc(sizeof(Oid) * numPart); + partCollations = (Oid *) palloc(sizeof(Oid) * numPart); + + partNumCols = 0; + foreach(lc, wc->partitionClause) + { + SortGroupClause *sgc = (SortGroupClause *) lfirst(lc); + TargetEntry *tle = get_sortgroupclause_tle(sgc, subplan->targetlist); + + Assert(OidIsValid(sgc->eqop)); + partColIdx[partNumCols] = tle->resno; + partOperators[partNumCols] = sgc->eqop; + partCollations[partNumCols] = exprCollation((Node *) tle->expr); + partNumCols++; + } + + ordColIdx = (AttrNumber *) palloc(sizeof(AttrNumber) * numOrder); + ordOperators = (Oid *) palloc(sizeof(Oid) * numOrder); + ordCollations = (Oid *) palloc(sizeof(Oid) * numOrder); + + ordNumCols = 0; + foreach(lc, wc->orderClause) + { + SortGroupClause *sgc = (SortGroupClause *) lfirst(lc); + TargetEntry *tle = get_sortgroupclause_tle(sgc, subplan->targetlist); + + Assert(OidIsValid(sgc->eqop)); + ordColIdx[ordNumCols] = tle->resno; + ordOperators[ordNumCols] = sgc->eqop; + ordCollations[ordNumCols] = exprCollation((Node *) tle->expr); + ordNumCols++; + } + + /* And finally we can make the WindowAgg node */ + plan = make_windowagg(tlist, + wc->winref, + partNumCols, + partColIdx, + partOperators, + partCollations, + ordNumCols, + ordColIdx, + ordOperators, + ordCollations, + wc->frameOptions, + wc->startOffset, + wc->endOffset, + wc->startInRangeFunc, + wc->endInRangeFunc, + wc->inRangeColl, + wc->inRangeAsc, + wc->inRangeNullsFirst, + wc->runCondition, + best_path->qual, + best_path->topwindow, + subplan); + + copy_generic_path_info(&plan->plan, (Path *) best_path); + + return plan; +} + +/* + * create_setop_plan + * + * Create a SetOp plan for 'best_path' and (recursively) plans + * for its subpaths. + */ +static SetOp * +create_setop_plan(PlannerInfo *root, SetOpPath *best_path, int flags) +{ + SetOp *plan; + Plan *subplan; + long numGroups; + + /* + * SetOp doesn't project, so tlist requirements pass through; moreover we + * need grouping columns to be labeled. + */ + subplan = create_plan_recurse(root, best_path->subpath, + flags | CP_LABEL_TLIST); + + /* Convert numGroups to long int --- but 'ware overflow! */ + numGroups = clamp_cardinality_to_long(best_path->numGroups); + + plan = make_setop(best_path->cmd, + best_path->strategy, + subplan, + best_path->distinctList, + best_path->flagColIdx, + best_path->firstFlag, + numGroups); + + copy_generic_path_info(&plan->plan, (Path *) best_path); + + return plan; +} + +/* + * create_recursiveunion_plan + * + * Create a RecursiveUnion plan for 'best_path' and (recursively) plans + * for its subpaths. + */ +static RecursiveUnion * +create_recursiveunion_plan(PlannerInfo *root, RecursiveUnionPath *best_path) +{ + RecursiveUnion *plan; + Plan *leftplan; + Plan *rightplan; + List *tlist; + long numGroups; + + /* Need both children to produce same tlist, so force it */ + leftplan = create_plan_recurse(root, best_path->leftpath, CP_EXACT_TLIST); + rightplan = create_plan_recurse(root, best_path->rightpath, CP_EXACT_TLIST); + + tlist = build_path_tlist(root, &best_path->path); + + /* Convert numGroups to long int --- but 'ware overflow! */ + numGroups = clamp_cardinality_to_long(best_path->numGroups); + + plan = make_recursive_union(tlist, + leftplan, + rightplan, + best_path->wtParam, + best_path->distinctList, + numGroups); + + copy_generic_path_info(&plan->plan, (Path *) best_path); + + return plan; +} + +/* + * create_lockrows_plan + * + * Create a LockRows plan for 'best_path' and (recursively) plans + * for its subpaths. + */ +static LockRows * +create_lockrows_plan(PlannerInfo *root, LockRowsPath *best_path, + int flags) +{ + LockRows *plan; + Plan *subplan; + + /* LockRows doesn't project, so tlist requirements pass through */ + subplan = create_plan_recurse(root, best_path->subpath, flags); + + plan = make_lockrows(subplan, best_path->rowMarks, best_path->epqParam); + + copy_generic_path_info(&plan->plan, (Path *) best_path); + + return plan; +} + +/* + * create_modifytable_plan + * Create a ModifyTable plan for 'best_path'. + * + * Returns a Plan node. + */ +static ModifyTable * +create_modifytable_plan(PlannerInfo *root, ModifyTablePath *best_path) +{ + ModifyTable *plan; + Path *subpath = best_path->subpath; + Plan *subplan; + + /* Subplan must produce exactly the specified tlist */ + subplan = create_plan_recurse(root, subpath, CP_EXACT_TLIST); + + /* Transfer resname/resjunk labeling, too, to keep executor happy */ + apply_tlist_labeling(subplan->targetlist, root->processed_tlist); + + plan = make_modifytable(root, + subplan, + best_path->operation, + best_path->canSetTag, + best_path->nominalRelation, + best_path->rootRelation, + best_path->partColsUpdated, + best_path->resultRelations, + best_path->updateColnosLists, + best_path->withCheckOptionLists, + best_path->returningLists, + best_path->rowMarks, + best_path->onconflict, + best_path->mergeActionLists, + best_path->epqParam); + + copy_generic_path_info(&plan->plan, &best_path->path); + + return plan; +} + +/* + * create_limit_plan + * + * Create a Limit plan for 'best_path' and (recursively) plans + * for its subpaths. + */ +static Limit * +create_limit_plan(PlannerInfo *root, LimitPath *best_path, int flags) +{ + Limit *plan; + Plan *subplan; + int numUniqkeys = 0; + AttrNumber *uniqColIdx = NULL; + Oid *uniqOperators = NULL; + Oid *uniqCollations = NULL; + + /* Limit doesn't project, so tlist requirements pass through */ + subplan = create_plan_recurse(root, best_path->subpath, flags); + + /* Extract information necessary for comparing rows for WITH TIES. */ + if (best_path->limitOption == LIMIT_OPTION_WITH_TIES) + { + Query *parse = root->parse; + ListCell *l; + + numUniqkeys = list_length(parse->sortClause); + uniqColIdx = (AttrNumber *) palloc(numUniqkeys * sizeof(AttrNumber)); + uniqOperators = (Oid *) palloc(numUniqkeys * sizeof(Oid)); + uniqCollations = (Oid *) palloc(numUniqkeys * sizeof(Oid)); + + numUniqkeys = 0; + foreach(l, parse->sortClause) + { + SortGroupClause *sortcl = (SortGroupClause *) lfirst(l); + TargetEntry *tle = get_sortgroupclause_tle(sortcl, parse->targetList); + + uniqColIdx[numUniqkeys] = tle->resno; + uniqOperators[numUniqkeys] = sortcl->eqop; + uniqCollations[numUniqkeys] = exprCollation((Node *) tle->expr); + numUniqkeys++; + } + } + + plan = make_limit(subplan, + best_path->limitOffset, + best_path->limitCount, + best_path->limitOption, + numUniqkeys, uniqColIdx, uniqOperators, uniqCollations); + + copy_generic_path_info(&plan->plan, (Path *) best_path); + + return plan; +} + + +/***************************************************************************** + * + * BASE-RELATION SCAN METHODS + * + *****************************************************************************/ + + +/* + * create_seqscan_plan + * Returns a seqscan plan for the base relation scanned by 'best_path' + * with restriction clauses 'scan_clauses' and targetlist 'tlist'. + */ +static SeqScan * +create_seqscan_plan(PlannerInfo *root, Path *best_path, + List *tlist, List *scan_clauses) +{ + SeqScan *scan_plan; + Index scan_relid = best_path->parent->relid; + + /* it should be a base rel... */ + Assert(scan_relid > 0); + Assert(best_path->parent->rtekind == RTE_RELATION); + + /* Sort clauses into best execution order */ + scan_clauses = order_qual_clauses(root, scan_clauses); + + /* Reduce RestrictInfo list to bare expressions; ignore pseudoconstants */ + scan_clauses = extract_actual_clauses(scan_clauses, false); + + /* Replace any outer-relation variables with nestloop params */ + if (best_path->param_info) + { + scan_clauses = (List *) + replace_nestloop_params(root, (Node *) scan_clauses); + } + + scan_plan = make_seqscan(tlist, + scan_clauses, + scan_relid); + + copy_generic_path_info(&scan_plan->scan.plan, best_path); + + return scan_plan; +} + +/* + * create_samplescan_plan + * Returns a samplescan plan for the base relation scanned by 'best_path' + * with restriction clauses 'scan_clauses' and targetlist 'tlist'. + */ +static SampleScan * +create_samplescan_plan(PlannerInfo *root, Path *best_path, + List *tlist, List *scan_clauses) +{ + SampleScan *scan_plan; + Index scan_relid = best_path->parent->relid; + RangeTblEntry *rte; + TableSampleClause *tsc; + + /* it should be a base rel with a tablesample clause... */ + Assert(scan_relid > 0); + rte = planner_rt_fetch(scan_relid, root); + Assert(rte->rtekind == RTE_RELATION); + tsc = rte->tablesample; + Assert(tsc != NULL); + + /* Sort clauses into best execution order */ + scan_clauses = order_qual_clauses(root, scan_clauses); + + /* Reduce RestrictInfo list to bare expressions; ignore pseudoconstants */ + scan_clauses = extract_actual_clauses(scan_clauses, false); + + /* Replace any outer-relation variables with nestloop params */ + if (best_path->param_info) + { + scan_clauses = (List *) + replace_nestloop_params(root, (Node *) scan_clauses); + tsc = (TableSampleClause *) + replace_nestloop_params(root, (Node *) tsc); + } + + scan_plan = make_samplescan(tlist, + scan_clauses, + scan_relid, + tsc); + + copy_generic_path_info(&scan_plan->scan.plan, best_path); + + return scan_plan; +} + +/* + * create_indexscan_plan + * Returns an indexscan plan for the base relation scanned by 'best_path' + * with restriction clauses 'scan_clauses' and targetlist 'tlist'. + * + * We use this for both plain IndexScans and IndexOnlyScans, because the + * qual preprocessing work is the same for both. Note that the caller tells + * us which to build --- we don't look at best_path->path.pathtype, because + * create_bitmap_subplan needs to be able to override the prior decision. + */ +static Scan * +create_indexscan_plan(PlannerInfo *root, + IndexPath *best_path, + List *tlist, + List *scan_clauses, + bool indexonly) +{ + Scan *scan_plan; + List *indexclauses = best_path->indexclauses; + List *indexorderbys = best_path->indexorderbys; + Index baserelid = best_path->path.parent->relid; + IndexOptInfo *indexinfo = best_path->indexinfo; + Oid indexoid = indexinfo->indexoid; + List *qpqual; + List *stripped_indexquals; + List *fixed_indexquals; + List *fixed_indexorderbys; + List *indexorderbyops = NIL; + ListCell *l; + + /* it should be a base rel... */ + Assert(baserelid > 0); + Assert(best_path->path.parent->rtekind == RTE_RELATION); + + /* + * Extract the index qual expressions (stripped of RestrictInfos) from the + * IndexClauses list, and prepare a copy with index Vars substituted for + * table Vars. (This step also does replace_nestloop_params on the + * fixed_indexquals.) + */ + fix_indexqual_references(root, best_path, + &stripped_indexquals, + &fixed_indexquals); + + /* + * Likewise fix up index attr references in the ORDER BY expressions. + */ + fixed_indexorderbys = fix_indexorderby_references(root, best_path); + + /* + * The qpqual list must contain all restrictions not automatically handled + * by the index, other than pseudoconstant clauses which will be handled + * by a separate gating plan node. All the predicates in the indexquals + * will be checked (either by the index itself, or by nodeIndexscan.c), + * but if there are any "special" operators involved then they must be + * included in qpqual. The upshot is that qpqual must contain + * scan_clauses minus whatever appears in indexquals. + * + * is_redundant_with_indexclauses() detects cases where a scan clause is + * present in the indexclauses list or is generated from the same + * EquivalenceClass as some indexclause, and is therefore redundant with + * it, though not equal. (The latter happens when indxpath.c prefers a + * different derived equality than what generate_join_implied_equalities + * picked for a parameterized scan's ppi_clauses.) Note that it will not + * match to lossy index clauses, which is critical because we have to + * include the original clause in qpqual in that case. + * + * In some situations (particularly with OR'd index conditions) we may + * have scan_clauses that are not equal to, but are logically implied by, + * the index quals; so we also try a predicate_implied_by() check to see + * if we can discard quals that way. (predicate_implied_by assumes its + * first input contains only immutable functions, so we have to check + * that.) + * + * Note: if you change this bit of code you should also look at + * extract_nonindex_conditions() in costsize.c. + */ + qpqual = NIL; + foreach(l, scan_clauses) + { + RestrictInfo *rinfo = lfirst_node(RestrictInfo, l); + + if (rinfo->pseudoconstant) + continue; /* we may drop pseudoconstants here */ + if (is_redundant_with_indexclauses(rinfo, indexclauses)) + continue; /* dup or derived from same EquivalenceClass */ + if (!contain_mutable_functions((Node *) rinfo->clause) && + predicate_implied_by(list_make1(rinfo->clause), stripped_indexquals, + false)) + continue; /* provably implied by indexquals */ + qpqual = lappend(qpqual, rinfo); + } + + /* Sort clauses into best execution order */ + qpqual = order_qual_clauses(root, qpqual); + + /* Reduce RestrictInfo list to bare expressions; ignore pseudoconstants */ + qpqual = extract_actual_clauses(qpqual, false); + + /* + * We have to replace any outer-relation variables with nestloop params in + * the indexqualorig, qpqual, and indexorderbyorig expressions. A bit + * annoying to have to do this separately from the processing in + * fix_indexqual_references --- rethink this when generalizing the inner + * indexscan support. But note we can't really do this earlier because + * it'd break the comparisons to predicates above ... (or would it? Those + * wouldn't have outer refs) + */ + if (best_path->path.param_info) + { + stripped_indexquals = (List *) + replace_nestloop_params(root, (Node *) stripped_indexquals); + qpqual = (List *) + replace_nestloop_params(root, (Node *) qpqual); + indexorderbys = (List *) + replace_nestloop_params(root, (Node *) indexorderbys); + } + + /* + * If there are ORDER BY expressions, look up the sort operators for their + * result datatypes. + */ + if (indexorderbys) + { + ListCell *pathkeyCell, + *exprCell; + + /* + * PathKey contains OID of the btree opfamily we're sorting by, but + * that's not quite enough because we need the expression's datatype + * to look up the sort operator in the operator family. + */ + Assert(list_length(best_path->path.pathkeys) == list_length(indexorderbys)); + forboth(pathkeyCell, best_path->path.pathkeys, exprCell, indexorderbys) + { + PathKey *pathkey = (PathKey *) lfirst(pathkeyCell); + Node *expr = (Node *) lfirst(exprCell); + Oid exprtype = exprType(expr); + Oid sortop; + + /* Get sort operator from opfamily */ + sortop = get_opfamily_member(pathkey->pk_opfamily, + exprtype, + exprtype, + pathkey->pk_strategy); + if (!OidIsValid(sortop)) + elog(ERROR, "missing operator %d(%u,%u) in opfamily %u", + pathkey->pk_strategy, exprtype, exprtype, pathkey->pk_opfamily); + indexorderbyops = lappend_oid(indexorderbyops, sortop); + } + } + + /* + * For an index-only scan, we must mark indextlist entries as resjunk if + * they are columns that the index AM can't return; this cues setrefs.c to + * not generate references to those columns. + */ + if (indexonly) + { + int i = 0; + + foreach(l, indexinfo->indextlist) + { + TargetEntry *indextle = (TargetEntry *) lfirst(l); + + indextle->resjunk = !indexinfo->canreturn[i]; + i++; + } + } + + /* Finally ready to build the plan node */ + if (indexonly) + scan_plan = (Scan *) make_indexonlyscan(tlist, + qpqual, + baserelid, + indexoid, + fixed_indexquals, + stripped_indexquals, + fixed_indexorderbys, + indexinfo->indextlist, + best_path->indexscandir); + else + scan_plan = (Scan *) make_indexscan(tlist, + qpqual, + baserelid, + indexoid, + fixed_indexquals, + stripped_indexquals, + fixed_indexorderbys, + indexorderbys, + indexorderbyops, + best_path->indexscandir); + + copy_generic_path_info(&scan_plan->plan, &best_path->path); + + return scan_plan; +} + +/* + * create_bitmap_scan_plan + * Returns a bitmap scan plan for the base relation scanned by 'best_path' + * with restriction clauses 'scan_clauses' and targetlist 'tlist'. + */ +static BitmapHeapScan * +create_bitmap_scan_plan(PlannerInfo *root, + BitmapHeapPath *best_path, + List *tlist, + List *scan_clauses) +{ + Index baserelid = best_path->path.parent->relid; + Plan *bitmapqualplan; + List *bitmapqualorig; + List *indexquals; + List *indexECs; + List *qpqual; + ListCell *l; + BitmapHeapScan *scan_plan; + + /* it should be a base rel... */ + Assert(baserelid > 0); + Assert(best_path->path.parent->rtekind == RTE_RELATION); + + /* Process the bitmapqual tree into a Plan tree and qual lists */ + bitmapqualplan = create_bitmap_subplan(root, best_path->bitmapqual, + &bitmapqualorig, &indexquals, + &indexECs); + + if (best_path->path.parallel_aware) + bitmap_subplan_mark_shared(bitmapqualplan); + + /* + * The qpqual list must contain all restrictions not automatically handled + * by the index, other than pseudoconstant clauses which will be handled + * by a separate gating plan node. All the predicates in the indexquals + * will be checked (either by the index itself, or by + * nodeBitmapHeapscan.c), but if there are any "special" operators + * involved then they must be added to qpqual. The upshot is that qpqual + * must contain scan_clauses minus whatever appears in indexquals. + * + * This loop is similar to the comparable code in create_indexscan_plan(), + * but with some differences because it has to compare the scan clauses to + * stripped (no RestrictInfos) indexquals. See comments there for more + * info. + * + * In normal cases simple equal() checks will be enough to spot duplicate + * clauses, so we try that first. We next see if the scan clause is + * redundant with any top-level indexqual by virtue of being generated + * from the same EC. After that, try predicate_implied_by(). + * + * Unlike create_indexscan_plan(), the predicate_implied_by() test here is + * useful for getting rid of qpquals that are implied by index predicates, + * because the predicate conditions are included in the "indexquals" + * returned by create_bitmap_subplan(). Bitmap scans have to do it that + * way because predicate conditions need to be rechecked if the scan + * becomes lossy, so they have to be included in bitmapqualorig. + */ + qpqual = NIL; + foreach(l, scan_clauses) + { + RestrictInfo *rinfo = lfirst_node(RestrictInfo, l); + Node *clause = (Node *) rinfo->clause; + + if (rinfo->pseudoconstant) + continue; /* we may drop pseudoconstants here */ + if (list_member(indexquals, clause)) + continue; /* simple duplicate */ + if (rinfo->parent_ec && list_member_ptr(indexECs, rinfo->parent_ec)) + continue; /* derived from same EquivalenceClass */ + if (!contain_mutable_functions(clause) && + predicate_implied_by(list_make1(clause), indexquals, false)) + continue; /* provably implied by indexquals */ + qpqual = lappend(qpqual, rinfo); + } + + /* Sort clauses into best execution order */ + qpqual = order_qual_clauses(root, qpqual); + + /* Reduce RestrictInfo list to bare expressions; ignore pseudoconstants */ + qpqual = extract_actual_clauses(qpqual, false); + + /* + * When dealing with special operators, we will at this point have + * duplicate clauses in qpqual and bitmapqualorig. We may as well drop + * 'em from bitmapqualorig, since there's no point in making the tests + * twice. + */ + bitmapqualorig = list_difference_ptr(bitmapqualorig, qpqual); + + /* + * We have to replace any outer-relation variables with nestloop params in + * the qpqual and bitmapqualorig expressions. (This was already done for + * expressions attached to plan nodes in the bitmapqualplan tree.) + */ + if (best_path->path.param_info) + { + qpqual = (List *) + replace_nestloop_params(root, (Node *) qpqual); + bitmapqualorig = (List *) + replace_nestloop_params(root, (Node *) bitmapqualorig); + } + + /* Finally ready to build the plan node */ + scan_plan = make_bitmap_heapscan(tlist, + qpqual, + bitmapqualplan, + bitmapqualorig, + baserelid); + + copy_generic_path_info(&scan_plan->scan.plan, &best_path->path); + + return scan_plan; +} + +/* + * Given a bitmapqual tree, generate the Plan tree that implements it + * + * As byproducts, we also return in *qual and *indexqual the qual lists + * (in implicit-AND form, without RestrictInfos) describing the original index + * conditions and the generated indexqual conditions. (These are the same in + * simple cases, but when special index operators are involved, the former + * list includes the special conditions while the latter includes the actual + * indexable conditions derived from them.) Both lists include partial-index + * predicates, because we have to recheck predicates as well as index + * conditions if the bitmap scan becomes lossy. + * + * In addition, we return a list of EquivalenceClass pointers for all the + * top-level indexquals that were possibly-redundantly derived from ECs. + * This allows removal of scan_clauses that are redundant with such quals. + * (We do not attempt to detect such redundancies for quals that are within + * OR subtrees. This could be done in a less hacky way if we returned the + * indexquals in RestrictInfo form, but that would be slower and still pretty + * messy, since we'd have to build new RestrictInfos in many cases.) + */ +static Plan * +create_bitmap_subplan(PlannerInfo *root, Path *bitmapqual, + List **qual, List **indexqual, List **indexECs) +{ + Plan *plan; + + if (IsA(bitmapqual, BitmapAndPath)) + { + BitmapAndPath *apath = (BitmapAndPath *) bitmapqual; + List *subplans = NIL; + List *subquals = NIL; + List *subindexquals = NIL; + List *subindexECs = NIL; + ListCell *l; + + /* + * There may well be redundant quals among the subplans, since a + * top-level WHERE qual might have gotten used to form several + * different index quals. We don't try exceedingly hard to eliminate + * redundancies, but we do eliminate obvious duplicates by using + * list_concat_unique. + */ + foreach(l, apath->bitmapquals) + { + Plan *subplan; + List *subqual; + List *subindexqual; + List *subindexEC; + + subplan = create_bitmap_subplan(root, (Path *) lfirst(l), + &subqual, &subindexqual, + &subindexEC); + subplans = lappend(subplans, subplan); + subquals = list_concat_unique(subquals, subqual); + subindexquals = list_concat_unique(subindexquals, subindexqual); + /* Duplicates in indexECs aren't worth getting rid of */ + subindexECs = list_concat(subindexECs, subindexEC); + } + plan = (Plan *) make_bitmap_and(subplans); + plan->startup_cost = apath->path.startup_cost; + plan->total_cost = apath->path.total_cost; + plan->plan_rows = + clamp_row_est(apath->bitmapselectivity * apath->path.parent->tuples); + plan->plan_width = 0; /* meaningless */ + plan->parallel_aware = false; + plan->parallel_safe = apath->path.parallel_safe; + *qual = subquals; + *indexqual = subindexquals; + *indexECs = subindexECs; + } + else if (IsA(bitmapqual, BitmapOrPath)) + { + BitmapOrPath *opath = (BitmapOrPath *) bitmapqual; + List *subplans = NIL; + List *subquals = NIL; + List *subindexquals = NIL; + bool const_true_subqual = false; + bool const_true_subindexqual = false; + ListCell *l; + + /* + * Here, we only detect qual-free subplans. A qual-free subplan would + * cause us to generate "... OR true ..." which we may as well reduce + * to just "true". We do not try to eliminate redundant subclauses + * because (a) it's not as likely as in the AND case, and (b) we might + * well be working with hundreds or even thousands of OR conditions, + * perhaps from a long IN list. The performance of list_append_unique + * would be unacceptable. + */ + foreach(l, opath->bitmapquals) + { + Plan *subplan; + List *subqual; + List *subindexqual; + List *subindexEC; + + subplan = create_bitmap_subplan(root, (Path *) lfirst(l), + &subqual, &subindexqual, + &subindexEC); + subplans = lappend(subplans, subplan); + if (subqual == NIL) + const_true_subqual = true; + else if (!const_true_subqual) + subquals = lappend(subquals, + make_ands_explicit(subqual)); + if (subindexqual == NIL) + const_true_subindexqual = true; + else if (!const_true_subindexqual) + subindexquals = lappend(subindexquals, + make_ands_explicit(subindexqual)); + } + + /* + * In the presence of ScalarArrayOpExpr quals, we might have built + * BitmapOrPaths with just one subpath; don't add an OR step. + */ + if (list_length(subplans) == 1) + { + plan = (Plan *) linitial(subplans); + } + else + { + plan = (Plan *) make_bitmap_or(subplans); + plan->startup_cost = opath->path.startup_cost; + plan->total_cost = opath->path.total_cost; + plan->plan_rows = + clamp_row_est(opath->bitmapselectivity * opath->path.parent->tuples); + plan->plan_width = 0; /* meaningless */ + plan->parallel_aware = false; + plan->parallel_safe = opath->path.parallel_safe; + } + + /* + * If there were constant-TRUE subquals, the OR reduces to constant + * TRUE. Also, avoid generating one-element ORs, which could happen + * due to redundancy elimination or ScalarArrayOpExpr quals. + */ + if (const_true_subqual) + *qual = NIL; + else if (list_length(subquals) <= 1) + *qual = subquals; + else + *qual = list_make1(make_orclause(subquals)); + if (const_true_subindexqual) + *indexqual = NIL; + else if (list_length(subindexquals) <= 1) + *indexqual = subindexquals; + else + *indexqual = list_make1(make_orclause(subindexquals)); + *indexECs = NIL; + } + else if (IsA(bitmapqual, IndexPath)) + { + IndexPath *ipath = (IndexPath *) bitmapqual; + IndexScan *iscan; + List *subquals; + List *subindexquals; + List *subindexECs; + ListCell *l; + + /* Use the regular indexscan plan build machinery... */ + iscan = castNode(IndexScan, + create_indexscan_plan(root, ipath, + NIL, NIL, false)); + /* then convert to a bitmap indexscan */ + plan = (Plan *) make_bitmap_indexscan(iscan->scan.scanrelid, + iscan->indexid, + iscan->indexqual, + iscan->indexqualorig); + /* and set its cost/width fields appropriately */ + plan->startup_cost = 0.0; + plan->total_cost = ipath->indextotalcost; + plan->plan_rows = + clamp_row_est(ipath->indexselectivity * ipath->path.parent->tuples); + plan->plan_width = 0; /* meaningless */ + plan->parallel_aware = false; + plan->parallel_safe = ipath->path.parallel_safe; + /* Extract original index clauses, actual index quals, relevant ECs */ + subquals = NIL; + subindexquals = NIL; + subindexECs = NIL; + foreach(l, ipath->indexclauses) + { + IndexClause *iclause = (IndexClause *) lfirst(l); + RestrictInfo *rinfo = iclause->rinfo; + + Assert(!rinfo->pseudoconstant); + subquals = lappend(subquals, rinfo->clause); + subindexquals = list_concat(subindexquals, + get_actual_clauses(iclause->indexquals)); + if (rinfo->parent_ec) + subindexECs = lappend(subindexECs, rinfo->parent_ec); + } + /* We can add any index predicate conditions, too */ + foreach(l, ipath->indexinfo->indpred) + { + Expr *pred = (Expr *) lfirst(l); + + /* + * We know that the index predicate must have been implied by the + * query condition as a whole, but it may or may not be implied by + * the conditions that got pushed into the bitmapqual. Avoid + * generating redundant conditions. + */ + if (!predicate_implied_by(list_make1(pred), subquals, false)) + { + subquals = lappend(subquals, pred); + subindexquals = lappend(subindexquals, pred); + } + } + *qual = subquals; + *indexqual = subindexquals; + *indexECs = subindexECs; + } + else + { + elog(ERROR, "unrecognized node type: %d", nodeTag(bitmapqual)); + plan = NULL; /* keep compiler quiet */ + } + + return plan; +} + +/* + * create_tidscan_plan + * Returns a tidscan plan for the base relation scanned by 'best_path' + * with restriction clauses 'scan_clauses' and targetlist 'tlist'. + */ +static TidScan * +create_tidscan_plan(PlannerInfo *root, TidPath *best_path, + List *tlist, List *scan_clauses) +{ + TidScan *scan_plan; + Index scan_relid = best_path->path.parent->relid; + List *tidquals = best_path->tidquals; + + /* it should be a base rel... */ + Assert(scan_relid > 0); + Assert(best_path->path.parent->rtekind == RTE_RELATION); + + /* + * The qpqual list must contain all restrictions not enforced by the + * tidquals list. Since tidquals has OR semantics, we have to be careful + * about matching it up to scan_clauses. It's convenient to handle the + * single-tidqual case separately from the multiple-tidqual case. In the + * single-tidqual case, we look through the scan_clauses while they are + * still in RestrictInfo form, and drop any that are redundant with the + * tidqual. + * + * In normal cases simple pointer equality checks will be enough to spot + * duplicate RestrictInfos, so we try that first. + * + * Another common case is that a scan_clauses entry is generated from the + * same EquivalenceClass as some tidqual, and is therefore redundant with + * it, though not equal. + * + * Unlike indexpaths, we don't bother with predicate_implied_by(); the + * number of cases where it could win are pretty small. + */ + if (list_length(tidquals) == 1) + { + List *qpqual = NIL; + ListCell *l; + + foreach(l, scan_clauses) + { + RestrictInfo *rinfo = lfirst_node(RestrictInfo, l); + + if (rinfo->pseudoconstant) + continue; /* we may drop pseudoconstants here */ + if (list_member_ptr(tidquals, rinfo)) + continue; /* simple duplicate */ + if (is_redundant_derived_clause(rinfo, tidquals)) + continue; /* derived from same EquivalenceClass */ + qpqual = lappend(qpqual, rinfo); + } + scan_clauses = qpqual; + } + + /* Sort clauses into best execution order */ + scan_clauses = order_qual_clauses(root, scan_clauses); + + /* Reduce RestrictInfo lists to bare expressions; ignore pseudoconstants */ + tidquals = extract_actual_clauses(tidquals, false); + scan_clauses = extract_actual_clauses(scan_clauses, false); + + /* + * If we have multiple tidquals, it's more convenient to remove duplicate + * scan_clauses after stripping the RestrictInfos. In this situation, + * because the tidquals represent OR sub-clauses, they could not have come + * from EquivalenceClasses so we don't have to worry about matching up + * non-identical clauses. On the other hand, because tidpath.c will have + * extracted those sub-clauses from some OR clause and built its own list, + * we will certainly not have pointer equality to any scan clause. So + * convert the tidquals list to an explicit OR clause and see if we can + * match it via equal() to any scan clause. + */ + if (list_length(tidquals) > 1) + scan_clauses = list_difference(scan_clauses, + list_make1(make_orclause(tidquals))); + + /* Replace any outer-relation variables with nestloop params */ + if (best_path->path.param_info) + { + tidquals = (List *) + replace_nestloop_params(root, (Node *) tidquals); + scan_clauses = (List *) + replace_nestloop_params(root, (Node *) scan_clauses); + } + + scan_plan = make_tidscan(tlist, + scan_clauses, + scan_relid, + tidquals); + + copy_generic_path_info(&scan_plan->scan.plan, &best_path->path); + + return scan_plan; +} + +/* + * create_tidrangescan_plan + * Returns a tidrangescan plan for the base relation scanned by 'best_path' + * with restriction clauses 'scan_clauses' and targetlist 'tlist'. + */ +static TidRangeScan * +create_tidrangescan_plan(PlannerInfo *root, TidRangePath *best_path, + List *tlist, List *scan_clauses) +{ + TidRangeScan *scan_plan; + Index scan_relid = best_path->path.parent->relid; + List *tidrangequals = best_path->tidrangequals; + + /* it should be a base rel... */ + Assert(scan_relid > 0); + Assert(best_path->path.parent->rtekind == RTE_RELATION); + + /* + * The qpqual list must contain all restrictions not enforced by the + * tidrangequals list. tidrangequals has AND semantics, so we can simply + * remove any qual that appears in it. + */ + { + List *qpqual = NIL; + ListCell *l; + + foreach(l, scan_clauses) + { + RestrictInfo *rinfo = lfirst_node(RestrictInfo, l); + + if (rinfo->pseudoconstant) + continue; /* we may drop pseudoconstants here */ + if (list_member_ptr(tidrangequals, rinfo)) + continue; /* simple duplicate */ + qpqual = lappend(qpqual, rinfo); + } + scan_clauses = qpqual; + } + + /* Sort clauses into best execution order */ + scan_clauses = order_qual_clauses(root, scan_clauses); + + /* Reduce RestrictInfo lists to bare expressions; ignore pseudoconstants */ + tidrangequals = extract_actual_clauses(tidrangequals, false); + scan_clauses = extract_actual_clauses(scan_clauses, false); + + /* Replace any outer-relation variables with nestloop params */ + if (best_path->path.param_info) + { + tidrangequals = (List *) + replace_nestloop_params(root, (Node *) tidrangequals); + scan_clauses = (List *) + replace_nestloop_params(root, (Node *) scan_clauses); + } + + scan_plan = make_tidrangescan(tlist, + scan_clauses, + scan_relid, + tidrangequals); + + copy_generic_path_info(&scan_plan->scan.plan, &best_path->path); + + return scan_plan; +} + +/* + * create_subqueryscan_plan + * Returns a subqueryscan plan for the base relation scanned by 'best_path' + * with restriction clauses 'scan_clauses' and targetlist 'tlist'. + */ +static SubqueryScan * +create_subqueryscan_plan(PlannerInfo *root, SubqueryScanPath *best_path, + List *tlist, List *scan_clauses) +{ + SubqueryScan *scan_plan; + RelOptInfo *rel = best_path->path.parent; + Index scan_relid = rel->relid; + Plan *subplan; + + /* it should be a subquery base rel... */ + Assert(scan_relid > 0); + Assert(rel->rtekind == RTE_SUBQUERY); + + /* + * Recursively create Plan from Path for subquery. Since we are entering + * a different planner context (subroot), recurse to create_plan not + * create_plan_recurse. + */ + subplan = create_plan(rel->subroot, best_path->subpath); + + /* Sort clauses into best execution order */ + scan_clauses = order_qual_clauses(root, scan_clauses); + + /* Reduce RestrictInfo list to bare expressions; ignore pseudoconstants */ + scan_clauses = extract_actual_clauses(scan_clauses, false); + + /* Replace any outer-relation variables with nestloop params */ + if (best_path->path.param_info) + { + scan_clauses = (List *) + replace_nestloop_params(root, (Node *) scan_clauses); + process_subquery_nestloop_params(root, + rel->subplan_params); + } + + scan_plan = make_subqueryscan(tlist, + scan_clauses, + scan_relid, + subplan); + + copy_generic_path_info(&scan_plan->scan.plan, &best_path->path); + + return scan_plan; +} + +/* + * create_functionscan_plan + * Returns a functionscan plan for the base relation scanned by 'best_path' + * with restriction clauses 'scan_clauses' and targetlist 'tlist'. + */ +static FunctionScan * +create_functionscan_plan(PlannerInfo *root, Path *best_path, + List *tlist, List *scan_clauses) +{ + FunctionScan *scan_plan; + Index scan_relid = best_path->parent->relid; + RangeTblEntry *rte; + List *functions; + + /* it should be a function base rel... */ + Assert(scan_relid > 0); + rte = planner_rt_fetch(scan_relid, root); + Assert(rte->rtekind == RTE_FUNCTION); + functions = rte->functions; + + /* Sort clauses into best execution order */ + scan_clauses = order_qual_clauses(root, scan_clauses); + + /* Reduce RestrictInfo list to bare expressions; ignore pseudoconstants */ + scan_clauses = extract_actual_clauses(scan_clauses, false); + + /* Replace any outer-relation variables with nestloop params */ + if (best_path->param_info) + { + scan_clauses = (List *) + replace_nestloop_params(root, (Node *) scan_clauses); + /* The function expressions could contain nestloop params, too */ + functions = (List *) replace_nestloop_params(root, (Node *) functions); + } + + scan_plan = make_functionscan(tlist, scan_clauses, scan_relid, + functions, rte->funcordinality); + + copy_generic_path_info(&scan_plan->scan.plan, best_path); + + return scan_plan; +} + +/* + * create_tablefuncscan_plan + * Returns a tablefuncscan plan for the base relation scanned by 'best_path' + * with restriction clauses 'scan_clauses' and targetlist 'tlist'. + */ +static TableFuncScan * +create_tablefuncscan_plan(PlannerInfo *root, Path *best_path, + List *tlist, List *scan_clauses) +{ + TableFuncScan *scan_plan; + Index scan_relid = best_path->parent->relid; + RangeTblEntry *rte; + TableFunc *tablefunc; + + /* it should be a function base rel... */ + Assert(scan_relid > 0); + rte = planner_rt_fetch(scan_relid, root); + Assert(rte->rtekind == RTE_TABLEFUNC); + tablefunc = rte->tablefunc; + + /* Sort clauses into best execution order */ + scan_clauses = order_qual_clauses(root, scan_clauses); + + /* Reduce RestrictInfo list to bare expressions; ignore pseudoconstants */ + scan_clauses = extract_actual_clauses(scan_clauses, false); + + /* Replace any outer-relation variables with nestloop params */ + if (best_path->param_info) + { + scan_clauses = (List *) + replace_nestloop_params(root, (Node *) scan_clauses); + /* The function expressions could contain nestloop params, too */ + tablefunc = (TableFunc *) replace_nestloop_params(root, (Node *) tablefunc); + } + + scan_plan = make_tablefuncscan(tlist, scan_clauses, scan_relid, + tablefunc); + + copy_generic_path_info(&scan_plan->scan.plan, best_path); + + return scan_plan; +} + +/* + * create_valuesscan_plan + * Returns a valuesscan plan for the base relation scanned by 'best_path' + * with restriction clauses 'scan_clauses' and targetlist 'tlist'. + */ +static ValuesScan * +create_valuesscan_plan(PlannerInfo *root, Path *best_path, + List *tlist, List *scan_clauses) +{ + ValuesScan *scan_plan; + Index scan_relid = best_path->parent->relid; + RangeTblEntry *rte; + List *values_lists; + + /* it should be a values base rel... */ + Assert(scan_relid > 0); + rte = planner_rt_fetch(scan_relid, root); + Assert(rte->rtekind == RTE_VALUES); + values_lists = rte->values_lists; + + /* Sort clauses into best execution order */ + scan_clauses = order_qual_clauses(root, scan_clauses); + + /* Reduce RestrictInfo list to bare expressions; ignore pseudoconstants */ + scan_clauses = extract_actual_clauses(scan_clauses, false); + + /* Replace any outer-relation variables with nestloop params */ + if (best_path->param_info) + { + scan_clauses = (List *) + replace_nestloop_params(root, (Node *) scan_clauses); + /* The values lists could contain nestloop params, too */ + values_lists = (List *) + replace_nestloop_params(root, (Node *) values_lists); + } + + scan_plan = make_valuesscan(tlist, scan_clauses, scan_relid, + values_lists); + + copy_generic_path_info(&scan_plan->scan.plan, best_path); + + return scan_plan; +} + +/* + * create_ctescan_plan + * Returns a ctescan plan for the base relation scanned by 'best_path' + * with restriction clauses 'scan_clauses' and targetlist 'tlist'. + */ +static CteScan * +create_ctescan_plan(PlannerInfo *root, Path *best_path, + List *tlist, List *scan_clauses) +{ + CteScan *scan_plan; + Index scan_relid = best_path->parent->relid; + RangeTblEntry *rte; + SubPlan *ctesplan = NULL; + int plan_id; + int cte_param_id; + PlannerInfo *cteroot; + Index levelsup; + int ndx; + ListCell *lc; + + Assert(scan_relid > 0); + rte = planner_rt_fetch(scan_relid, root); + Assert(rte->rtekind == RTE_CTE); + Assert(!rte->self_reference); + + /* + * Find the referenced CTE, and locate the SubPlan previously made for it. + */ + levelsup = rte->ctelevelsup; + cteroot = root; + while (levelsup-- > 0) + { + cteroot = cteroot->parent_root; + if (!cteroot) /* shouldn't happen */ + elog(ERROR, "bad levelsup for CTE \"%s\"", rte->ctename); + } + + /* + * Note: cte_plan_ids can be shorter than cteList, if we are still working + * on planning the CTEs (ie, this is a side-reference from another CTE). + * So we mustn't use forboth here. + */ + ndx = 0; + foreach(lc, cteroot->parse->cteList) + { + CommonTableExpr *cte = (CommonTableExpr *) lfirst(lc); + + if (strcmp(cte->ctename, rte->ctename) == 0) + break; + ndx++; + } + if (lc == NULL) /* shouldn't happen */ + elog(ERROR, "could not find CTE \"%s\"", rte->ctename); + if (ndx >= list_length(cteroot->cte_plan_ids)) + elog(ERROR, "could not find plan for CTE \"%s\"", rte->ctename); + plan_id = list_nth_int(cteroot->cte_plan_ids, ndx); + if (plan_id <= 0) + elog(ERROR, "no plan was made for CTE \"%s\"", rte->ctename); + foreach(lc, cteroot->init_plans) + { + ctesplan = (SubPlan *) lfirst(lc); + if (ctesplan->plan_id == plan_id) + break; + } + if (lc == NULL) /* shouldn't happen */ + elog(ERROR, "could not find plan for CTE \"%s\"", rte->ctename); + + /* + * We need the CTE param ID, which is the sole member of the SubPlan's + * setParam list. + */ + cte_param_id = linitial_int(ctesplan->setParam); + + /* Sort clauses into best execution order */ + scan_clauses = order_qual_clauses(root, scan_clauses); + + /* Reduce RestrictInfo list to bare expressions; ignore pseudoconstants */ + scan_clauses = extract_actual_clauses(scan_clauses, false); + + /* Replace any outer-relation variables with nestloop params */ + if (best_path->param_info) + { + scan_clauses = (List *) + replace_nestloop_params(root, (Node *) scan_clauses); + } + + scan_plan = make_ctescan(tlist, scan_clauses, scan_relid, + plan_id, cte_param_id); + + copy_generic_path_info(&scan_plan->scan.plan, best_path); + + return scan_plan; +} + +/* + * create_namedtuplestorescan_plan + * Returns a tuplestorescan plan for the base relation scanned by + * 'best_path' with restriction clauses 'scan_clauses' and targetlist + * 'tlist'. + */ +static NamedTuplestoreScan * +create_namedtuplestorescan_plan(PlannerInfo *root, Path *best_path, + List *tlist, List *scan_clauses) +{ + NamedTuplestoreScan *scan_plan; + Index scan_relid = best_path->parent->relid; + RangeTblEntry *rte; + + Assert(scan_relid > 0); + rte = planner_rt_fetch(scan_relid, root); + Assert(rte->rtekind == RTE_NAMEDTUPLESTORE); + + /* Sort clauses into best execution order */ + scan_clauses = order_qual_clauses(root, scan_clauses); + + /* Reduce RestrictInfo list to bare expressions; ignore pseudoconstants */ + scan_clauses = extract_actual_clauses(scan_clauses, false); + + /* Replace any outer-relation variables with nestloop params */ + if (best_path->param_info) + { + scan_clauses = (List *) + replace_nestloop_params(root, (Node *) scan_clauses); + } + + scan_plan = make_namedtuplestorescan(tlist, scan_clauses, scan_relid, + rte->enrname); + + copy_generic_path_info(&scan_plan->scan.plan, best_path); + + return scan_plan; +} + +/* + * create_resultscan_plan + * Returns a Result plan for the RTE_RESULT base relation scanned by + * 'best_path' with restriction clauses 'scan_clauses' and targetlist + * 'tlist'. + */ +static Result * +create_resultscan_plan(PlannerInfo *root, Path *best_path, + List *tlist, List *scan_clauses) +{ + Result *scan_plan; + Index scan_relid = best_path->parent->relid; + RangeTblEntry *rte PG_USED_FOR_ASSERTS_ONLY; + + Assert(scan_relid > 0); + rte = planner_rt_fetch(scan_relid, root); + Assert(rte->rtekind == RTE_RESULT); + + /* Sort clauses into best execution order */ + scan_clauses = order_qual_clauses(root, scan_clauses); + + /* Reduce RestrictInfo list to bare expressions; ignore pseudoconstants */ + scan_clauses = extract_actual_clauses(scan_clauses, false); + + /* Replace any outer-relation variables with nestloop params */ + if (best_path->param_info) + { + scan_clauses = (List *) + replace_nestloop_params(root, (Node *) scan_clauses); + } + + scan_plan = make_result(tlist, (Node *) scan_clauses, NULL); + + copy_generic_path_info(&scan_plan->plan, best_path); + + return scan_plan; +} + +/* + * create_worktablescan_plan + * Returns a worktablescan plan for the base relation scanned by 'best_path' + * with restriction clauses 'scan_clauses' and targetlist 'tlist'. + */ +static WorkTableScan * +create_worktablescan_plan(PlannerInfo *root, Path *best_path, + List *tlist, List *scan_clauses) +{ + WorkTableScan *scan_plan; + Index scan_relid = best_path->parent->relid; + RangeTblEntry *rte; + Index levelsup; + PlannerInfo *cteroot; + + Assert(scan_relid > 0); + rte = planner_rt_fetch(scan_relid, root); + Assert(rte->rtekind == RTE_CTE); + Assert(rte->self_reference); + + /* + * We need to find the worktable param ID, which is in the plan level + * that's processing the recursive UNION, which is one level *below* where + * the CTE comes from. + */ + levelsup = rte->ctelevelsup; + if (levelsup == 0) /* shouldn't happen */ + elog(ERROR, "bad levelsup for CTE \"%s\"", rte->ctename); + levelsup--; + cteroot = root; + while (levelsup-- > 0) + { + cteroot = cteroot->parent_root; + if (!cteroot) /* shouldn't happen */ + elog(ERROR, "bad levelsup for CTE \"%s\"", rte->ctename); + } + if (cteroot->wt_param_id < 0) /* shouldn't happen */ + elog(ERROR, "could not find param ID for CTE \"%s\"", rte->ctename); + + /* Sort clauses into best execution order */ + scan_clauses = order_qual_clauses(root, scan_clauses); + + /* Reduce RestrictInfo list to bare expressions; ignore pseudoconstants */ + scan_clauses = extract_actual_clauses(scan_clauses, false); + + /* Replace any outer-relation variables with nestloop params */ + if (best_path->param_info) + { + scan_clauses = (List *) + replace_nestloop_params(root, (Node *) scan_clauses); + } + + scan_plan = make_worktablescan(tlist, scan_clauses, scan_relid, + cteroot->wt_param_id); + + copy_generic_path_info(&scan_plan->scan.plan, best_path); + + return scan_plan; +} + +/* + * create_foreignscan_plan + * Returns a foreignscan plan for the relation scanned by 'best_path' + * with restriction clauses 'scan_clauses' and targetlist 'tlist'. + */ +static ForeignScan * +create_foreignscan_plan(PlannerInfo *root, ForeignPath *best_path, + List *tlist, List *scan_clauses) +{ + ForeignScan *scan_plan; + RelOptInfo *rel = best_path->path.parent; + Index scan_relid = rel->relid; + Oid rel_oid = InvalidOid; + Plan *outer_plan = NULL; + + Assert(rel->fdwroutine != NULL); + + /* transform the child path if any */ + if (best_path->fdw_outerpath) + outer_plan = create_plan_recurse(root, best_path->fdw_outerpath, + CP_EXACT_TLIST); + + /* + * If we're scanning a base relation, fetch its OID. (Irrelevant if + * scanning a join relation.) + */ + if (scan_relid > 0) + { + RangeTblEntry *rte; + + Assert(rel->rtekind == RTE_RELATION); + rte = planner_rt_fetch(scan_relid, root); + Assert(rte->rtekind == RTE_RELATION); + rel_oid = rte->relid; + } + + /* + * Sort clauses into best execution order. We do this first since the FDW + * might have more info than we do and wish to adjust the ordering. + */ + scan_clauses = order_qual_clauses(root, scan_clauses); + + /* + * Let the FDW perform its processing on the restriction clauses and + * generate the plan node. Note that the FDW might remove restriction + * clauses that it intends to execute remotely, or even add more (if it + * has selected some join clauses for remote use but also wants them + * rechecked locally). + */ + scan_plan = rel->fdwroutine->GetForeignPlan(root, rel, rel_oid, + best_path, + tlist, scan_clauses, + outer_plan); + + /* Copy cost data from Path to Plan; no need to make FDW do this */ + copy_generic_path_info(&scan_plan->scan.plan, &best_path->path); + + /* Copy foreign server OID; likewise, no need to make FDW do this */ + scan_plan->fs_server = rel->serverid; + + /* + * Likewise, copy the relids that are represented by this foreign scan. An + * upper rel doesn't have relids set, but it covers all the base relations + * participating in the underlying scan, so use root's all_baserels. + */ + if (rel->reloptkind == RELOPT_UPPER_REL) + scan_plan->fs_relids = root->all_baserels; + else + scan_plan->fs_relids = best_path->path.parent->relids; + + /* + * If this is a foreign join, and to make it valid to push down we had to + * assume that the current user is the same as some user explicitly named + * in the query, mark the finished plan as depending on the current user. + */ + if (rel->useridiscurrent) + root->glob->dependsOnRole = true; + + /* + * Replace any outer-relation variables with nestloop params in the qual, + * fdw_exprs and fdw_recheck_quals expressions. We do this last so that + * the FDW doesn't have to be involved. (Note that parts of fdw_exprs or + * fdw_recheck_quals could have come from join clauses, so doing this + * beforehand on the scan_clauses wouldn't work.) We assume + * fdw_scan_tlist contains no such variables. + */ + if (best_path->path.param_info) + { + scan_plan->scan.plan.qual = (List *) + replace_nestloop_params(root, (Node *) scan_plan->scan.plan.qual); + scan_plan->fdw_exprs = (List *) + replace_nestloop_params(root, (Node *) scan_plan->fdw_exprs); + scan_plan->fdw_recheck_quals = (List *) + replace_nestloop_params(root, + (Node *) scan_plan->fdw_recheck_quals); + } + + /* + * If rel is a base relation, detect whether any system columns are + * requested from the rel. (If rel is a join relation, rel->relid will be + * 0, but there can be no Var with relid 0 in the rel's targetlist or the + * restriction clauses, so we skip this in that case. Note that any such + * columns in base relations that were joined are assumed to be contained + * in fdw_scan_tlist.) This is a bit of a kluge and might go away + * someday, so we intentionally leave it out of the API presented to FDWs. + */ + scan_plan->fsSystemCol = false; + if (scan_relid > 0) + { + Bitmapset *attrs_used = NULL; + ListCell *lc; + int i; + + /* + * First, examine all the attributes needed for joins or final output. + * Note: we must look at rel's targetlist, not the attr_needed data, + * because attr_needed isn't computed for inheritance child rels. + */ + pull_varattnos((Node *) rel->reltarget->exprs, scan_relid, &attrs_used); + + /* Add all the attributes used by restriction clauses. */ + foreach(lc, rel->baserestrictinfo) + { + RestrictInfo *rinfo = (RestrictInfo *) lfirst(lc); + + pull_varattnos((Node *) rinfo->clause, scan_relid, &attrs_used); + } + + /* Now, are any system columns requested from rel? */ + for (i = FirstLowInvalidHeapAttributeNumber + 1; i < 0; i++) + { + if (bms_is_member(i - FirstLowInvalidHeapAttributeNumber, attrs_used)) + { + scan_plan->fsSystemCol = true; + break; + } + } + + bms_free(attrs_used); + } + + return scan_plan; +} + +/* + * create_customscan_plan + * + * Transform a CustomPath into a Plan. + */ +static CustomScan * +create_customscan_plan(PlannerInfo *root, CustomPath *best_path, + List *tlist, List *scan_clauses) +{ + CustomScan *cplan; + RelOptInfo *rel = best_path->path.parent; + List *custom_plans = NIL; + ListCell *lc; + + /* Recursively transform child paths. */ + foreach(lc, best_path->custom_paths) + { + Plan *plan = create_plan_recurse(root, (Path *) lfirst(lc), + CP_EXACT_TLIST); + + custom_plans = lappend(custom_plans, plan); + } + + /* + * Sort clauses into the best execution order, although custom-scan + * provider can reorder them again. + */ + scan_clauses = order_qual_clauses(root, scan_clauses); + + /* + * Invoke custom plan provider to create the Plan node represented by the + * CustomPath. + */ + cplan = castNode(CustomScan, + best_path->methods->PlanCustomPath(root, + rel, + best_path, + tlist, + scan_clauses, + custom_plans)); + + /* + * Copy cost data from Path to Plan; no need to make custom-plan providers + * do this + */ + copy_generic_path_info(&cplan->scan.plan, &best_path->path); + + /* Likewise, copy the relids that are represented by this custom scan */ + cplan->custom_relids = best_path->path.parent->relids; + + /* + * Replace any outer-relation variables with nestloop params in the qual + * and custom_exprs expressions. We do this last so that the custom-plan + * provider doesn't have to be involved. (Note that parts of custom_exprs + * could have come from join clauses, so doing this beforehand on the + * scan_clauses wouldn't work.) We assume custom_scan_tlist contains no + * such variables. + */ + if (best_path->path.param_info) + { + cplan->scan.plan.qual = (List *) + replace_nestloop_params(root, (Node *) cplan->scan.plan.qual); + cplan->custom_exprs = (List *) + replace_nestloop_params(root, (Node *) cplan->custom_exprs); + } + + return cplan; +} + + +/***************************************************************************** + * + * JOIN METHODS + * + *****************************************************************************/ + +static NestLoop * +create_nestloop_plan(PlannerInfo *root, + NestPath *best_path) +{ + NestLoop *join_plan; + Plan *outer_plan; + Plan *inner_plan; + List *tlist = build_path_tlist(root, &best_path->jpath.path); + List *joinrestrictclauses = best_path->jpath.joinrestrictinfo; + List *joinclauses; + List *otherclauses; + Relids outerrelids; + List *nestParams; + Relids saveOuterRels = root->curOuterRels; + + /* NestLoop can project, so no need to be picky about child tlists */ + outer_plan = create_plan_recurse(root, best_path->jpath.outerjoinpath, 0); + + /* For a nestloop, include outer relids in curOuterRels for inner side */ + root->curOuterRels = bms_union(root->curOuterRels, + best_path->jpath.outerjoinpath->parent->relids); + + inner_plan = create_plan_recurse(root, best_path->jpath.innerjoinpath, 0); + + /* Restore curOuterRels */ + bms_free(root->curOuterRels); + root->curOuterRels = saveOuterRels; + + /* Sort join qual clauses into best execution order */ + joinrestrictclauses = order_qual_clauses(root, joinrestrictclauses); + + /* Get the join qual clauses (in plain expression form) */ + /* Any pseudoconstant clauses are ignored here */ + if (IS_OUTER_JOIN(best_path->jpath.jointype)) + { + extract_actual_join_clauses(joinrestrictclauses, + best_path->jpath.path.parent->relids, + &joinclauses, &otherclauses); + } + else + { + /* We can treat all clauses alike for an inner join */ + joinclauses = extract_actual_clauses(joinrestrictclauses, false); + otherclauses = NIL; + } + + /* Replace any outer-relation variables with nestloop params */ + if (best_path->jpath.path.param_info) + { + joinclauses = (List *) + replace_nestloop_params(root, (Node *) joinclauses); + otherclauses = (List *) + replace_nestloop_params(root, (Node *) otherclauses); + } + + /* + * Identify any nestloop parameters that should be supplied by this join + * node, and remove them from root->curOuterParams. + */ + outerrelids = best_path->jpath.outerjoinpath->parent->relids; + nestParams = identify_current_nestloop_params(root, outerrelids); + + join_plan = make_nestloop(tlist, + joinclauses, + otherclauses, + nestParams, + outer_plan, + inner_plan, + best_path->jpath.jointype, + best_path->jpath.inner_unique); + + copy_generic_path_info(&join_plan->join.plan, &best_path->jpath.path); + + return join_plan; +} + +static MergeJoin * +create_mergejoin_plan(PlannerInfo *root, + MergePath *best_path) +{ + MergeJoin *join_plan; + Plan *outer_plan; + Plan *inner_plan; + List *tlist = build_path_tlist(root, &best_path->jpath.path); + List *joinclauses; + List *otherclauses; + List *mergeclauses; + List *outerpathkeys; + List *innerpathkeys; + int nClauses; + Oid *mergefamilies; + Oid *mergecollations; + int *mergestrategies; + bool *mergenullsfirst; + PathKey *opathkey; + EquivalenceClass *opeclass; + int i; + ListCell *lc; + ListCell *lop; + ListCell *lip; + Path *outer_path = best_path->jpath.outerjoinpath; + Path *inner_path = best_path->jpath.innerjoinpath; + + /* + * MergeJoin can project, so we don't have to demand exact tlists from the + * inputs. However, if we're intending to sort an input's result, it's + * best to request a small tlist so we aren't sorting more data than + * necessary. + */ + outer_plan = create_plan_recurse(root, best_path->jpath.outerjoinpath, + (best_path->outersortkeys != NIL) ? CP_SMALL_TLIST : 0); + + inner_plan = create_plan_recurse(root, best_path->jpath.innerjoinpath, + (best_path->innersortkeys != NIL) ? CP_SMALL_TLIST : 0); + + /* Sort join qual clauses into best execution order */ + /* NB: do NOT reorder the mergeclauses */ + joinclauses = order_qual_clauses(root, best_path->jpath.joinrestrictinfo); + + /* Get the join qual clauses (in plain expression form) */ + /* Any pseudoconstant clauses are ignored here */ + if (IS_OUTER_JOIN(best_path->jpath.jointype)) + { + extract_actual_join_clauses(joinclauses, + best_path->jpath.path.parent->relids, + &joinclauses, &otherclauses); + } + else + { + /* We can treat all clauses alike for an inner join */ + joinclauses = extract_actual_clauses(joinclauses, false); + otherclauses = NIL; + } + + /* + * Remove the mergeclauses from the list of join qual clauses, leaving the + * list of quals that must be checked as qpquals. + */ + mergeclauses = get_actual_clauses(best_path->path_mergeclauses); + joinclauses = list_difference(joinclauses, mergeclauses); + + /* + * Replace any outer-relation variables with nestloop params. There + * should not be any in the mergeclauses. + */ + if (best_path->jpath.path.param_info) + { + joinclauses = (List *) + replace_nestloop_params(root, (Node *) joinclauses); + otherclauses = (List *) + replace_nestloop_params(root, (Node *) otherclauses); + } + + /* + * Rearrange mergeclauses, if needed, so that the outer variable is always + * on the left; mark the mergeclause restrictinfos with correct + * outer_is_left status. + */ + mergeclauses = get_switched_clauses(best_path->path_mergeclauses, + best_path->jpath.outerjoinpath->parent->relids); + + /* + * Create explicit sort nodes for the outer and inner paths if necessary. + */ + if (best_path->outersortkeys) + { + Relids outer_relids = outer_path->parent->relids; + Sort *sort = make_sort_from_pathkeys(outer_plan, + best_path->outersortkeys, + outer_relids); + + label_sort_with_costsize(root, sort, -1.0); + outer_plan = (Plan *) sort; + outerpathkeys = best_path->outersortkeys; + } + else + outerpathkeys = best_path->jpath.outerjoinpath->pathkeys; + + if (best_path->innersortkeys) + { + Relids inner_relids = inner_path->parent->relids; + Sort *sort = make_sort_from_pathkeys(inner_plan, + best_path->innersortkeys, + inner_relids); + + label_sort_with_costsize(root, sort, -1.0); + inner_plan = (Plan *) sort; + innerpathkeys = best_path->innersortkeys; + } + else + innerpathkeys = best_path->jpath.innerjoinpath->pathkeys; + + /* + * If specified, add a materialize node to shield the inner plan from the + * need to handle mark/restore. + */ + if (best_path->materialize_inner) + { + Plan *matplan = (Plan *) make_material(inner_plan); + + /* + * We assume the materialize will not spill to disk, and therefore + * charge just cpu_operator_cost per tuple. (Keep this estimate in + * sync with final_cost_mergejoin.) + */ + copy_plan_costsize(matplan, inner_plan); + matplan->total_cost += cpu_operator_cost * matplan->plan_rows; + + inner_plan = matplan; + } + + /* + * Compute the opfamily/collation/strategy/nullsfirst arrays needed by the + * executor. The information is in the pathkeys for the two inputs, but + * we need to be careful about the possibility of mergeclauses sharing a + * pathkey, as well as the possibility that the inner pathkeys are not in + * an order matching the mergeclauses. + */ + nClauses = list_length(mergeclauses); + Assert(nClauses == list_length(best_path->path_mergeclauses)); + mergefamilies = (Oid *) palloc(nClauses * sizeof(Oid)); + mergecollations = (Oid *) palloc(nClauses * sizeof(Oid)); + mergestrategies = (int *) palloc(nClauses * sizeof(int)); + mergenullsfirst = (bool *) palloc(nClauses * sizeof(bool)); + + opathkey = NULL; + opeclass = NULL; + lop = list_head(outerpathkeys); + lip = list_head(innerpathkeys); + i = 0; + foreach(lc, best_path->path_mergeclauses) + { + RestrictInfo *rinfo = lfirst_node(RestrictInfo, lc); + EquivalenceClass *oeclass; + EquivalenceClass *ieclass; + PathKey *ipathkey = NULL; + EquivalenceClass *ipeclass = NULL; + bool first_inner_match = false; + + /* fetch outer/inner eclass from mergeclause */ + if (rinfo->outer_is_left) + { + oeclass = rinfo->left_ec; + ieclass = rinfo->right_ec; + } + else + { + oeclass = rinfo->right_ec; + ieclass = rinfo->left_ec; + } + Assert(oeclass != NULL); + Assert(ieclass != NULL); + + /* + * We must identify the pathkey elements associated with this clause + * by matching the eclasses (which should give a unique match, since + * the pathkey lists should be canonical). In typical cases the merge + * clauses are one-to-one with the pathkeys, but when dealing with + * partially redundant query conditions, things are more complicated. + * + * lop and lip reference the first as-yet-unmatched pathkey elements. + * If they're NULL then all pathkey elements have been matched. + * + * The ordering of the outer pathkeys should match the mergeclauses, + * by construction (see find_mergeclauses_for_outer_pathkeys()). There + * could be more than one mergeclause for the same outer pathkey, but + * no pathkey may be entirely skipped over. + */ + if (oeclass != opeclass) /* multiple matches are not interesting */ + { + /* doesn't match the current opathkey, so must match the next */ + if (lop == NULL) + elog(ERROR, "outer pathkeys do not match mergeclauses"); + opathkey = (PathKey *) lfirst(lop); + opeclass = opathkey->pk_eclass; + lop = lnext(outerpathkeys, lop); + if (oeclass != opeclass) + elog(ERROR, "outer pathkeys do not match mergeclauses"); + } + + /* + * The inner pathkeys likewise should not have skipped-over keys, but + * it's possible for a mergeclause to reference some earlier inner + * pathkey if we had redundant pathkeys. For example we might have + * mergeclauses like "o.a = i.x AND o.b = i.y AND o.c = i.x". The + * implied inner ordering is then "ORDER BY x, y, x", but the pathkey + * mechanism drops the second sort by x as redundant, and this code + * must cope. + * + * It's also possible for the implied inner-rel ordering to be like + * "ORDER BY x, y, x DESC". We still drop the second instance of x as + * redundant; but this means that the sort ordering of a redundant + * inner pathkey should not be considered significant. So we must + * detect whether this is the first clause matching an inner pathkey. + */ + if (lip) + { + ipathkey = (PathKey *) lfirst(lip); + ipeclass = ipathkey->pk_eclass; + if (ieclass == ipeclass) + { + /* successful first match to this inner pathkey */ + lip = lnext(innerpathkeys, lip); + first_inner_match = true; + } + } + if (!first_inner_match) + { + /* redundant clause ... must match something before lip */ + ListCell *l2; + + foreach(l2, innerpathkeys) + { + if (l2 == lip) + break; + ipathkey = (PathKey *) lfirst(l2); + ipeclass = ipathkey->pk_eclass; + if (ieclass == ipeclass) + break; + } + if (ieclass != ipeclass) + elog(ERROR, "inner pathkeys do not match mergeclauses"); + } + + /* + * The pathkeys should always match each other as to opfamily and + * collation (which affect equality), but if we're considering a + * redundant inner pathkey, its sort ordering might not match. In + * such cases we may ignore the inner pathkey's sort ordering and use + * the outer's. (In effect, we're lying to the executor about the + * sort direction of this inner column, but it does not matter since + * the run-time row comparisons would only reach this column when + * there's equality for the earlier column containing the same eclass. + * There could be only one value in this column for the range of inner + * rows having a given value in the earlier column, so it does not + * matter which way we imagine this column to be ordered.) But a + * non-redundant inner pathkey had better match outer's ordering too. + */ + if (opathkey->pk_opfamily != ipathkey->pk_opfamily || + opathkey->pk_eclass->ec_collation != ipathkey->pk_eclass->ec_collation) + elog(ERROR, "left and right pathkeys do not match in mergejoin"); + if (first_inner_match && + (opathkey->pk_strategy != ipathkey->pk_strategy || + opathkey->pk_nulls_first != ipathkey->pk_nulls_first)) + elog(ERROR, "left and right pathkeys do not match in mergejoin"); + + /* OK, save info for executor */ + mergefamilies[i] = opathkey->pk_opfamily; + mergecollations[i] = opathkey->pk_eclass->ec_collation; + mergestrategies[i] = opathkey->pk_strategy; + mergenullsfirst[i] = opathkey->pk_nulls_first; + i++; + } + + /* + * Note: it is not an error if we have additional pathkey elements (i.e., + * lop or lip isn't NULL here). The input paths might be better-sorted + * than we need for the current mergejoin. + */ + + /* + * Now we can build the mergejoin node. + */ + join_plan = make_mergejoin(tlist, + joinclauses, + otherclauses, + mergeclauses, + mergefamilies, + mergecollations, + mergestrategies, + mergenullsfirst, + outer_plan, + inner_plan, + best_path->jpath.jointype, + best_path->jpath.inner_unique, + best_path->skip_mark_restore); + + /* Costs of sort and material steps are included in path cost already */ + copy_generic_path_info(&join_plan->join.plan, &best_path->jpath.path); + + return join_plan; +} + +static HashJoin * +create_hashjoin_plan(PlannerInfo *root, + HashPath *best_path) +{ + HashJoin *join_plan; + Hash *hash_plan; + Plan *outer_plan; + Plan *inner_plan; + List *tlist = build_path_tlist(root, &best_path->jpath.path); + List *joinclauses; + List *otherclauses; + List *hashclauses; + List *hashoperators = NIL; + List *hashcollations = NIL; + List *inner_hashkeys = NIL; + List *outer_hashkeys = NIL; + Oid skewTable = InvalidOid; + AttrNumber skewColumn = InvalidAttrNumber; + bool skewInherit = false; + ListCell *lc; + + /* + * HashJoin can project, so we don't have to demand exact tlists from the + * inputs. However, it's best to request a small tlist from the inner + * side, so that we aren't storing more data than necessary. Likewise, if + * we anticipate batching, request a small tlist from the outer side so + * that we don't put extra data in the outer batch files. + */ + outer_plan = create_plan_recurse(root, best_path->jpath.outerjoinpath, + (best_path->num_batches > 1) ? CP_SMALL_TLIST : 0); + + inner_plan = create_plan_recurse(root, best_path->jpath.innerjoinpath, + CP_SMALL_TLIST); + + /* Sort join qual clauses into best execution order */ + joinclauses = order_qual_clauses(root, best_path->jpath.joinrestrictinfo); + /* There's no point in sorting the hash clauses ... */ + + /* Get the join qual clauses (in plain expression form) */ + /* Any pseudoconstant clauses are ignored here */ + if (IS_OUTER_JOIN(best_path->jpath.jointype)) + { + extract_actual_join_clauses(joinclauses, + best_path->jpath.path.parent->relids, + &joinclauses, &otherclauses); + } + else + { + /* We can treat all clauses alike for an inner join */ + joinclauses = extract_actual_clauses(joinclauses, false); + otherclauses = NIL; + } + + /* + * Remove the hashclauses from the list of join qual clauses, leaving the + * list of quals that must be checked as qpquals. + */ + hashclauses = get_actual_clauses(best_path->path_hashclauses); + joinclauses = list_difference(joinclauses, hashclauses); + + /* + * Replace any outer-relation variables with nestloop params. There + * should not be any in the hashclauses. + */ + if (best_path->jpath.path.param_info) + { + joinclauses = (List *) + replace_nestloop_params(root, (Node *) joinclauses); + otherclauses = (List *) + replace_nestloop_params(root, (Node *) otherclauses); + } + + /* + * Rearrange hashclauses, if needed, so that the outer variable is always + * on the left. + */ + hashclauses = get_switched_clauses(best_path->path_hashclauses, + best_path->jpath.outerjoinpath->parent->relids); + + /* + * If there is a single join clause and we can identify the outer variable + * as a simple column reference, supply its identity for possible use in + * skew optimization. (Note: in principle we could do skew optimization + * with multiple join clauses, but we'd have to be able to determine the + * most common combinations of outer values, which we don't currently have + * enough stats for.) + */ + if (list_length(hashclauses) == 1) + { + OpExpr *clause = (OpExpr *) linitial(hashclauses); + Node *node; + + Assert(is_opclause(clause)); + node = (Node *) linitial(clause->args); + if (IsA(node, RelabelType)) + node = (Node *) ((RelabelType *) node)->arg; + if (IsA(node, Var)) + { + Var *var = (Var *) node; + RangeTblEntry *rte; + + rte = root->simple_rte_array[var->varno]; + if (rte->rtekind == RTE_RELATION) + { + skewTable = rte->relid; + skewColumn = var->varattno; + skewInherit = rte->inh; + } + } + } + + /* + * Collect hash related information. The hashed expressions are + * deconstructed into outer/inner expressions, so they can be computed + * separately (inner expressions are used to build the hashtable via Hash, + * outer expressions to perform lookups of tuples from HashJoin's outer + * plan in the hashtable). Also collect operator information necessary to + * build the hashtable. + */ + foreach(lc, hashclauses) + { + OpExpr *hclause = lfirst_node(OpExpr, lc); + + hashoperators = lappend_oid(hashoperators, hclause->opno); + hashcollations = lappend_oid(hashcollations, hclause->inputcollid); + outer_hashkeys = lappend(outer_hashkeys, linitial(hclause->args)); + inner_hashkeys = lappend(inner_hashkeys, lsecond(hclause->args)); + } + + /* + * Build the hash node and hash join node. + */ + hash_plan = make_hash(inner_plan, + inner_hashkeys, + skewTable, + skewColumn, + skewInherit); + + /* + * Set Hash node's startup & total costs equal to total cost of input + * plan; this only affects EXPLAIN display not decisions. + */ + copy_plan_costsize(&hash_plan->plan, inner_plan); + hash_plan->plan.startup_cost = hash_plan->plan.total_cost; + + /* + * If parallel-aware, the executor will also need an estimate of the total + * number of rows expected from all participants so that it can size the + * shared hash table. + */ + if (best_path->jpath.path.parallel_aware) + { + hash_plan->plan.parallel_aware = true; + hash_plan->rows_total = best_path->inner_rows_total; + } + + join_plan = make_hashjoin(tlist, + joinclauses, + otherclauses, + hashclauses, + hashoperators, + hashcollations, + outer_hashkeys, + outer_plan, + (Plan *) hash_plan, + best_path->jpath.jointype, + best_path->jpath.inner_unique); + + copy_generic_path_info(&join_plan->join.plan, &best_path->jpath.path); + + return join_plan; +} + + +/***************************************************************************** + * + * SUPPORTING ROUTINES + * + *****************************************************************************/ + +/* + * replace_nestloop_params + * Replace outer-relation Vars and PlaceHolderVars in the given expression + * with nestloop Params + * + * All Vars and PlaceHolderVars belonging to the relation(s) identified by + * root->curOuterRels are replaced by Params, and entries are added to + * root->curOuterParams if not already present. + */ +static Node * +replace_nestloop_params(PlannerInfo *root, Node *expr) +{ + /* No setup needed for tree walk, so away we go */ + return replace_nestloop_params_mutator(expr, root); +} + +static Node * +replace_nestloop_params_mutator(Node *node, PlannerInfo *root) +{ + if (node == NULL) + return NULL; + if (IsA(node, Var)) + { + Var *var = (Var *) node; + + /* Upper-level Vars should be long gone at this point */ + Assert(var->varlevelsup == 0); + /* If not to be replaced, we can just return the Var unmodified */ + if (IS_SPECIAL_VARNO(var->varno) || + !bms_is_member(var->varno, root->curOuterRels)) + return node; + /* Replace the Var with a nestloop Param */ + return (Node *) replace_nestloop_param_var(root, var); + } + if (IsA(node, PlaceHolderVar)) + { + PlaceHolderVar *phv = (PlaceHolderVar *) node; + + /* Upper-level PlaceHolderVars should be long gone at this point */ + Assert(phv->phlevelsup == 0); + + /* + * Check whether we need to replace the PHV. We use bms_overlap as a + * cheap/quick test to see if the PHV might be evaluated in the outer + * rels, and then grab its PlaceHolderInfo to tell for sure. + */ + if (!bms_overlap(phv->phrels, root->curOuterRels) || + !bms_is_subset(find_placeholder_info(root, phv, false)->ph_eval_at, + root->curOuterRels)) + { + /* + * We can't replace the whole PHV, but we might still need to + * replace Vars or PHVs within its expression, in case it ends up + * actually getting evaluated here. (It might get evaluated in + * this plan node, or some child node; in the latter case we don't + * really need to process the expression here, but we haven't got + * enough info to tell if that's the case.) Flat-copy the PHV + * node and then recurse on its expression. + * + * Note that after doing this, we might have different + * representations of the contents of the same PHV in different + * parts of the plan tree. This is OK because equal() will just + * match on phid/phlevelsup, so setrefs.c will still recognize an + * upper-level reference to a lower-level copy of the same PHV. + */ + PlaceHolderVar *newphv = makeNode(PlaceHolderVar); + + memcpy(newphv, phv, sizeof(PlaceHolderVar)); + newphv->phexpr = (Expr *) + replace_nestloop_params_mutator((Node *) phv->phexpr, + root); + return (Node *) newphv; + } + /* Replace the PlaceHolderVar with a nestloop Param */ + return (Node *) replace_nestloop_param_placeholdervar(root, phv); + } + return expression_tree_mutator(node, + replace_nestloop_params_mutator, + (void *) root); +} + +/* + * fix_indexqual_references + * Adjust indexqual clauses to the form the executor's indexqual + * machinery needs. + * + * We have three tasks here: + * * Select the actual qual clauses out of the input IndexClause list, + * and remove RestrictInfo nodes from the qual clauses. + * * Replace any outer-relation Var or PHV nodes with nestloop Params. + * (XXX eventually, that responsibility should go elsewhere?) + * * Index keys must be represented by Var nodes with varattno set to the + * index's attribute number, not the attribute number in the original rel. + * + * *stripped_indexquals_p receives a list of the actual qual clauses. + * + * *fixed_indexquals_p receives a list of the adjusted quals. This is a copy + * that shares no substructure with the original; this is needed in case there + * are subplans in it (we need two separate copies of the subplan tree, or + * things will go awry). + */ +static void +fix_indexqual_references(PlannerInfo *root, IndexPath *index_path, + List **stripped_indexquals_p, List **fixed_indexquals_p) +{ + IndexOptInfo *index = index_path->indexinfo; + List *stripped_indexquals; + List *fixed_indexquals; + ListCell *lc; + + stripped_indexquals = fixed_indexquals = NIL; + + foreach(lc, index_path->indexclauses) + { + IndexClause *iclause = lfirst_node(IndexClause, lc); + int indexcol = iclause->indexcol; + ListCell *lc2; + + foreach(lc2, iclause->indexquals) + { + RestrictInfo *rinfo = lfirst_node(RestrictInfo, lc2); + Node *clause = (Node *) rinfo->clause; + + stripped_indexquals = lappend(stripped_indexquals, clause); + clause = fix_indexqual_clause(root, index, indexcol, + clause, iclause->indexcols); + fixed_indexquals = lappend(fixed_indexquals, clause); + } + } + + *stripped_indexquals_p = stripped_indexquals; + *fixed_indexquals_p = fixed_indexquals; +} + +/* + * fix_indexorderby_references + * Adjust indexorderby clauses to the form the executor's index + * machinery needs. + * + * This is a simplified version of fix_indexqual_references. The input is + * bare clauses and a separate indexcol list, instead of IndexClauses. + */ +static List * +fix_indexorderby_references(PlannerInfo *root, IndexPath *index_path) +{ + IndexOptInfo *index = index_path->indexinfo; + List *fixed_indexorderbys; + ListCell *lcc, + *lci; + + fixed_indexorderbys = NIL; + + forboth(lcc, index_path->indexorderbys, lci, index_path->indexorderbycols) + { + Node *clause = (Node *) lfirst(lcc); + int indexcol = lfirst_int(lci); + + clause = fix_indexqual_clause(root, index, indexcol, clause, NIL); + fixed_indexorderbys = lappend(fixed_indexorderbys, clause); + } + + return fixed_indexorderbys; +} + +/* + * fix_indexqual_clause + * Convert a single indexqual clause to the form needed by the executor. + * + * We replace nestloop params here, and replace the index key variables + * or expressions by index Var nodes. + */ +static Node * +fix_indexqual_clause(PlannerInfo *root, IndexOptInfo *index, int indexcol, + Node *clause, List *indexcolnos) +{ + /* + * Replace any outer-relation variables with nestloop params. + * + * This also makes a copy of the clause, so it's safe to modify it + * in-place below. + */ + clause = replace_nestloop_params(root, clause); + + if (IsA(clause, OpExpr)) + { + OpExpr *op = (OpExpr *) clause; + + /* Replace the indexkey expression with an index Var. */ + linitial(op->args) = fix_indexqual_operand(linitial(op->args), + index, + indexcol); + } + else if (IsA(clause, RowCompareExpr)) + { + RowCompareExpr *rc = (RowCompareExpr *) clause; + ListCell *lca, + *lcai; + + /* Replace the indexkey expressions with index Vars. */ + Assert(list_length(rc->largs) == list_length(indexcolnos)); + forboth(lca, rc->largs, lcai, indexcolnos) + { + lfirst(lca) = fix_indexqual_operand(lfirst(lca), + index, + lfirst_int(lcai)); + } + } + else if (IsA(clause, ScalarArrayOpExpr)) + { + ScalarArrayOpExpr *saop = (ScalarArrayOpExpr *) clause; + + /* Replace the indexkey expression with an index Var. */ + linitial(saop->args) = fix_indexqual_operand(linitial(saop->args), + index, + indexcol); + } + else if (IsA(clause, NullTest)) + { + NullTest *nt = (NullTest *) clause; + + /* Replace the indexkey expression with an index Var. */ + nt->arg = (Expr *) fix_indexqual_operand((Node *) nt->arg, + index, + indexcol); + } + else + elog(ERROR, "unsupported indexqual type: %d", + (int) nodeTag(clause)); + + return clause; +} + +/* + * fix_indexqual_operand + * Convert an indexqual expression to a Var referencing the index column. + * + * We represent index keys by Var nodes having varno == INDEX_VAR and varattno + * equal to the index's attribute number (index column position). + * + * Most of the code here is just for sanity cross-checking that the given + * expression actually matches the index column it's claimed to. + */ +static Node * +fix_indexqual_operand(Node *node, IndexOptInfo *index, int indexcol) +{ + Var *result; + int pos; + ListCell *indexpr_item; + + /* + * Remove any binary-compatible relabeling of the indexkey + */ + if (IsA(node, RelabelType)) + node = (Node *) ((RelabelType *) node)->arg; + + Assert(indexcol >= 0 && indexcol < index->ncolumns); + + if (index->indexkeys[indexcol] != 0) + { + /* It's a simple index column */ + if (IsA(node, Var) && + ((Var *) node)->varno == index->rel->relid && + ((Var *) node)->varattno == index->indexkeys[indexcol]) + { + result = (Var *) copyObject(node); + result->varno = INDEX_VAR; + result->varattno = indexcol + 1; + return (Node *) result; + } + else + elog(ERROR, "index key does not match expected index column"); + } + + /* It's an index expression, so find and cross-check the expression */ + indexpr_item = list_head(index->indexprs); + for (pos = 0; pos < index->ncolumns; pos++) + { + if (index->indexkeys[pos] == 0) + { + if (indexpr_item == NULL) + elog(ERROR, "too few entries in indexprs list"); + if (pos == indexcol) + { + Node *indexkey; + + indexkey = (Node *) lfirst(indexpr_item); + if (indexkey && IsA(indexkey, RelabelType)) + indexkey = (Node *) ((RelabelType *) indexkey)->arg; + if (equal(node, indexkey)) + { + result = makeVar(INDEX_VAR, indexcol + 1, + exprType(lfirst(indexpr_item)), -1, + exprCollation(lfirst(indexpr_item)), + 0); + return (Node *) result; + } + else + elog(ERROR, "index key does not match expected index column"); + } + indexpr_item = lnext(index->indexprs, indexpr_item); + } + } + + /* Oops... */ + elog(ERROR, "index key does not match expected index column"); + return NULL; /* keep compiler quiet */ +} + +/* + * get_switched_clauses + * Given a list of merge or hash joinclauses (as RestrictInfo nodes), + * extract the bare clauses, and rearrange the elements within the + * clauses, if needed, so the outer join variable is on the left and + * the inner is on the right. The original clause data structure is not + * touched; a modified list is returned. We do, however, set the transient + * outer_is_left field in each RestrictInfo to show which side was which. + */ +static List * +get_switched_clauses(List *clauses, Relids outerrelids) +{ + List *t_list = NIL; + ListCell *l; + + foreach(l, clauses) + { + RestrictInfo *restrictinfo = (RestrictInfo *) lfirst(l); + OpExpr *clause = (OpExpr *) restrictinfo->clause; + + Assert(is_opclause(clause)); + if (bms_is_subset(restrictinfo->right_relids, outerrelids)) + { + /* + * Duplicate just enough of the structure to allow commuting the + * clause without changing the original list. Could use + * copyObject, but a complete deep copy is overkill. + */ + OpExpr *temp = makeNode(OpExpr); + + temp->opno = clause->opno; + temp->opfuncid = InvalidOid; + temp->opresulttype = clause->opresulttype; + temp->opretset = clause->opretset; + temp->opcollid = clause->opcollid; + temp->inputcollid = clause->inputcollid; + temp->args = list_copy(clause->args); + temp->location = clause->location; + /* Commute it --- note this modifies the temp node in-place. */ + CommuteOpExpr(temp); + t_list = lappend(t_list, temp); + restrictinfo->outer_is_left = false; + } + else + { + Assert(bms_is_subset(restrictinfo->left_relids, outerrelids)); + t_list = lappend(t_list, clause); + restrictinfo->outer_is_left = true; + } + } + return t_list; +} + +/* + * order_qual_clauses + * Given a list of qual clauses that will all be evaluated at the same + * plan node, sort the list into the order we want to check the quals + * in at runtime. + * + * When security barrier quals are used in the query, we may have quals with + * different security levels in the list. Quals of lower security_level + * must go before quals of higher security_level, except that we can grant + * exceptions to move up quals that are leakproof. When security level + * doesn't force the decision, we prefer to order clauses by estimated + * execution cost, cheapest first. + * + * Ideally the order should be driven by a combination of execution cost and + * selectivity, but it's not immediately clear how to account for both, + * and given the uncertainty of the estimates the reliability of the decisions + * would be doubtful anyway. So we just order by security level then + * estimated per-tuple cost, being careful not to change the order when + * (as is often the case) the estimates are identical. + * + * Although this will work on either bare clauses or RestrictInfos, it's + * much faster to apply it to RestrictInfos, since it can re-use cost + * information that is cached in RestrictInfos. XXX in the bare-clause + * case, we are also not able to apply security considerations. That is + * all right for the moment, because the bare-clause case doesn't occur + * anywhere that barrier quals could be present, but it would be better to + * get rid of it. + * + * Note: some callers pass lists that contain entries that will later be + * removed; this is the easiest way to let this routine see RestrictInfos + * instead of bare clauses. This is another reason why trying to consider + * selectivity in the ordering would likely do the wrong thing. + */ +static List * +order_qual_clauses(PlannerInfo *root, List *clauses) +{ + typedef struct + { + Node *clause; + Cost cost; + Index security_level; + } QualItem; + int nitems = list_length(clauses); + QualItem *items; + ListCell *lc; + int i; + List *result; + + /* No need to work hard for 0 or 1 clause */ + if (nitems <= 1) + return clauses; + + /* + * Collect the items and costs into an array. This is to avoid repeated + * cost_qual_eval work if the inputs aren't RestrictInfos. + */ + items = (QualItem *) palloc(nitems * sizeof(QualItem)); + i = 0; + foreach(lc, clauses) + { + Node *clause = (Node *) lfirst(lc); + QualCost qcost; + + cost_qual_eval_node(&qcost, clause, root); + items[i].clause = clause; + items[i].cost = qcost.per_tuple; + if (IsA(clause, RestrictInfo)) + { + RestrictInfo *rinfo = (RestrictInfo *) clause; + + /* + * If a clause is leakproof, it doesn't have to be constrained by + * its nominal security level. If it's also reasonably cheap + * (here defined as 10X cpu_operator_cost), pretend it has + * security_level 0, which will allow it to go in front of + * more-expensive quals of lower security levels. Of course, that + * will also force it to go in front of cheaper quals of its own + * security level, which is not so great, but we can alleviate + * that risk by applying the cost limit cutoff. + */ + if (rinfo->leakproof && items[i].cost < 10 * cpu_operator_cost) + items[i].security_level = 0; + else + items[i].security_level = rinfo->security_level; + } + else + items[i].security_level = 0; + i++; + } + + /* + * Sort. We don't use qsort() because it's not guaranteed stable for + * equal keys. The expected number of entries is small enough that a + * simple insertion sort should be good enough. + */ + for (i = 1; i < nitems; i++) + { + QualItem newitem = items[i]; + int j; + + /* insert newitem into the already-sorted subarray */ + for (j = i; j > 0; j--) + { + QualItem *olditem = &items[j - 1]; + + if (newitem.security_level > olditem->security_level || + (newitem.security_level == olditem->security_level && + newitem.cost >= olditem->cost)) + break; + items[j] = *olditem; + } + items[j] = newitem; + } + + /* Convert back to a list */ + result = NIL; + for (i = 0; i < nitems; i++) + result = lappend(result, items[i].clause); + + return result; +} + +/* + * Copy cost and size info from a Path node to the Plan node created from it. + * The executor usually won't use this info, but it's needed by EXPLAIN. + * Also copy the parallel-related flags, which the executor *will* use. + */ +static void +copy_generic_path_info(Plan *dest, Path *src) +{ + dest->startup_cost = src->startup_cost; + dest->total_cost = src->total_cost; + dest->plan_rows = src->rows; + dest->plan_width = src->pathtarget->width; + dest->parallel_aware = src->parallel_aware; + dest->parallel_safe = src->parallel_safe; +} + +/* + * Copy cost and size info from a lower plan node to an inserted node. + * (Most callers alter the info after copying it.) + */ +static void +copy_plan_costsize(Plan *dest, Plan *src) +{ + dest->startup_cost = src->startup_cost; + dest->total_cost = src->total_cost; + dest->plan_rows = src->plan_rows; + dest->plan_width = src->plan_width; + /* Assume the inserted node is not parallel-aware. */ + dest->parallel_aware = false; + /* Assume the inserted node is parallel-safe, if child plan is. */ + dest->parallel_safe = src->parallel_safe; +} + +/* + * Some places in this file build Sort nodes that don't have a directly + * corresponding Path node. The cost of the sort is, or should have been, + * included in the cost of the Path node we're working from, but since it's + * not split out, we have to re-figure it using cost_sort(). This is just + * to label the Sort node nicely for EXPLAIN. + * + * limit_tuples is as for cost_sort (in particular, pass -1 if no limit) + */ +static void +label_sort_with_costsize(PlannerInfo *root, Sort *plan, double limit_tuples) +{ + Plan *lefttree = plan->plan.lefttree; + Path sort_path; /* dummy for result of cost_sort */ + + /* + * This function shouldn't have to deal with IncrementalSort plans because + * they are only created from corresponding Path nodes. + */ + Assert(IsA(plan, Sort)); + + cost_sort(&sort_path, root, NIL, + lefttree->total_cost, + lefttree->plan_rows, + lefttree->plan_width, + 0.0, + work_mem, + limit_tuples); + plan->plan.startup_cost = sort_path.startup_cost; + plan->plan.total_cost = sort_path.total_cost; + plan->plan.plan_rows = lefttree->plan_rows; + plan->plan.plan_width = lefttree->plan_width; + plan->plan.parallel_aware = false; + plan->plan.parallel_safe = lefttree->parallel_safe; +} + +/* + * bitmap_subplan_mark_shared + * Set isshared flag in bitmap subplan so that it will be created in + * shared memory. + */ +static void +bitmap_subplan_mark_shared(Plan *plan) +{ + if (IsA(plan, BitmapAnd)) + bitmap_subplan_mark_shared(linitial(((BitmapAnd *) plan)->bitmapplans)); + else if (IsA(plan, BitmapOr)) + { + ((BitmapOr *) plan)->isshared = true; + bitmap_subplan_mark_shared(linitial(((BitmapOr *) plan)->bitmapplans)); + } + else if (IsA(plan, BitmapIndexScan)) + ((BitmapIndexScan *) plan)->isshared = true; + else + elog(ERROR, "unrecognized node type: %d", nodeTag(plan)); +} + +/***************************************************************************** + * + * PLAN NODE BUILDING ROUTINES + * + * In general, these functions are not passed the original Path and therefore + * leave it to the caller to fill in the cost/width fields from the Path, + * typically by calling copy_generic_path_info(). This convention is + * somewhat historical, but it does support a few places above where we build + * a plan node without having an exactly corresponding Path node. Under no + * circumstances should one of these functions do its own cost calculations, + * as that would be redundant with calculations done while building Paths. + * + *****************************************************************************/ + +static SeqScan * +make_seqscan(List *qptlist, + List *qpqual, + Index scanrelid) +{ + SeqScan *node = makeNode(SeqScan); + Plan *plan = &node->scan.plan; + + plan->targetlist = qptlist; + plan->qual = qpqual; + plan->lefttree = NULL; + plan->righttree = NULL; + node->scan.scanrelid = scanrelid; + + return node; +} + +static SampleScan * +make_samplescan(List *qptlist, + List *qpqual, + Index scanrelid, + TableSampleClause *tsc) +{ + SampleScan *node = makeNode(SampleScan); + Plan *plan = &node->scan.plan; + + plan->targetlist = qptlist; + plan->qual = qpqual; + plan->lefttree = NULL; + plan->righttree = NULL; + node->scan.scanrelid = scanrelid; + node->tablesample = tsc; + + return node; +} + +static IndexScan * +make_indexscan(List *qptlist, + List *qpqual, + Index scanrelid, + Oid indexid, + List *indexqual, + List *indexqualorig, + List *indexorderby, + List *indexorderbyorig, + List *indexorderbyops, + ScanDirection indexscandir) +{ + IndexScan *node = makeNode(IndexScan); + Plan *plan = &node->scan.plan; + + plan->targetlist = qptlist; + plan->qual = qpqual; + plan->lefttree = NULL; + plan->righttree = NULL; + node->scan.scanrelid = scanrelid; + node->indexid = indexid; + node->indexqual = indexqual; + node->indexqualorig = indexqualorig; + node->indexorderby = indexorderby; + node->indexorderbyorig = indexorderbyorig; + node->indexorderbyops = indexorderbyops; + node->indexorderdir = indexscandir; + + return node; +} + +static IndexOnlyScan * +make_indexonlyscan(List *qptlist, + List *qpqual, + Index scanrelid, + Oid indexid, + List *indexqual, + List *recheckqual, + List *indexorderby, + List *indextlist, + ScanDirection indexscandir) +{ + IndexOnlyScan *node = makeNode(IndexOnlyScan); + Plan *plan = &node->scan.plan; + + plan->targetlist = qptlist; + plan->qual = qpqual; + plan->lefttree = NULL; + plan->righttree = NULL; + node->scan.scanrelid = scanrelid; + node->indexid = indexid; + node->indexqual = indexqual; + node->recheckqual = recheckqual; + node->indexorderby = indexorderby; + node->indextlist = indextlist; + node->indexorderdir = indexscandir; + + return node; +} + +static BitmapIndexScan * +make_bitmap_indexscan(Index scanrelid, + Oid indexid, + List *indexqual, + List *indexqualorig) +{ + BitmapIndexScan *node = makeNode(BitmapIndexScan); + Plan *plan = &node->scan.plan; + + plan->targetlist = NIL; /* not used */ + plan->qual = NIL; /* not used */ + plan->lefttree = NULL; + plan->righttree = NULL; + node->scan.scanrelid = scanrelid; + node->indexid = indexid; + node->indexqual = indexqual; + node->indexqualorig = indexqualorig; + + return node; +} + +static BitmapHeapScan * +make_bitmap_heapscan(List *qptlist, + List *qpqual, + Plan *lefttree, + List *bitmapqualorig, + Index scanrelid) +{ + BitmapHeapScan *node = makeNode(BitmapHeapScan); + Plan *plan = &node->scan.plan; + + plan->targetlist = qptlist; + plan->qual = qpqual; + plan->lefttree = lefttree; + plan->righttree = NULL; + node->scan.scanrelid = scanrelid; + node->bitmapqualorig = bitmapqualorig; + + return node; +} + +static TidScan * +make_tidscan(List *qptlist, + List *qpqual, + Index scanrelid, + List *tidquals) +{ + TidScan *node = makeNode(TidScan); + Plan *plan = &node->scan.plan; + + plan->targetlist = qptlist; + plan->qual = qpqual; + plan->lefttree = NULL; + plan->righttree = NULL; + node->scan.scanrelid = scanrelid; + node->tidquals = tidquals; + + return node; +} + +static TidRangeScan * +make_tidrangescan(List *qptlist, + List *qpqual, + Index scanrelid, + List *tidrangequals) +{ + TidRangeScan *node = makeNode(TidRangeScan); + Plan *plan = &node->scan.plan; + + plan->targetlist = qptlist; + plan->qual = qpqual; + plan->lefttree = NULL; + plan->righttree = NULL; + node->scan.scanrelid = scanrelid; + node->tidrangequals = tidrangequals; + + return node; +} + +static SubqueryScan * +make_subqueryscan(List *qptlist, + List *qpqual, + Index scanrelid, + Plan *subplan) +{ + SubqueryScan *node = makeNode(SubqueryScan); + Plan *plan = &node->scan.plan; + + plan->targetlist = qptlist; + plan->qual = qpqual; + plan->lefttree = NULL; + plan->righttree = NULL; + node->scan.scanrelid = scanrelid; + node->subplan = subplan; + node->scanstatus = SUBQUERY_SCAN_UNKNOWN; + + return node; +} + +static FunctionScan * +make_functionscan(List *qptlist, + List *qpqual, + Index scanrelid, + List *functions, + bool funcordinality) +{ + FunctionScan *node = makeNode(FunctionScan); + Plan *plan = &node->scan.plan; + + plan->targetlist = qptlist; + plan->qual = qpqual; + plan->lefttree = NULL; + plan->righttree = NULL; + node->scan.scanrelid = scanrelid; + node->functions = functions; + node->funcordinality = funcordinality; + + return node; +} + +static TableFuncScan * +make_tablefuncscan(List *qptlist, + List *qpqual, + Index scanrelid, + TableFunc *tablefunc) +{ + TableFuncScan *node = makeNode(TableFuncScan); + Plan *plan = &node->scan.plan; + + plan->targetlist = qptlist; + plan->qual = qpqual; + plan->lefttree = NULL; + plan->righttree = NULL; + node->scan.scanrelid = scanrelid; + node->tablefunc = tablefunc; + + return node; +} + +static ValuesScan * +make_valuesscan(List *qptlist, + List *qpqual, + Index scanrelid, + List *values_lists) +{ + ValuesScan *node = makeNode(ValuesScan); + Plan *plan = &node->scan.plan; + + plan->targetlist = qptlist; + plan->qual = qpqual; + plan->lefttree = NULL; + plan->righttree = NULL; + node->scan.scanrelid = scanrelid; + node->values_lists = values_lists; + + return node; +} + +static CteScan * +make_ctescan(List *qptlist, + List *qpqual, + Index scanrelid, + int ctePlanId, + int cteParam) +{ + CteScan *node = makeNode(CteScan); + Plan *plan = &node->scan.plan; + + plan->targetlist = qptlist; + plan->qual = qpqual; + plan->lefttree = NULL; + plan->righttree = NULL; + node->scan.scanrelid = scanrelid; + node->ctePlanId = ctePlanId; + node->cteParam = cteParam; + + return node; +} + +static NamedTuplestoreScan * +make_namedtuplestorescan(List *qptlist, + List *qpqual, + Index scanrelid, + char *enrname) +{ + NamedTuplestoreScan *node = makeNode(NamedTuplestoreScan); + Plan *plan = &node->scan.plan; + + /* cost should be inserted by caller */ + plan->targetlist = qptlist; + plan->qual = qpqual; + plan->lefttree = NULL; + plan->righttree = NULL; + node->scan.scanrelid = scanrelid; + node->enrname = enrname; + + return node; +} + +static WorkTableScan * +make_worktablescan(List *qptlist, + List *qpqual, + Index scanrelid, + int wtParam) +{ + WorkTableScan *node = makeNode(WorkTableScan); + Plan *plan = &node->scan.plan; + + plan->targetlist = qptlist; + plan->qual = qpqual; + plan->lefttree = NULL; + plan->righttree = NULL; + node->scan.scanrelid = scanrelid; + node->wtParam = wtParam; + + return node; +} + +ForeignScan * +make_foreignscan(List *qptlist, + List *qpqual, + Index scanrelid, + List *fdw_exprs, + List *fdw_private, + List *fdw_scan_tlist, + List *fdw_recheck_quals, + Plan *outer_plan) +{ + ForeignScan *node = makeNode(ForeignScan); + Plan *plan = &node->scan.plan; + + /* cost will be filled in by create_foreignscan_plan */ + plan->targetlist = qptlist; + plan->qual = qpqual; + plan->lefttree = outer_plan; + plan->righttree = NULL; + node->scan.scanrelid = scanrelid; + + /* these may be overridden by the FDW's PlanDirectModify callback. */ + node->operation = CMD_SELECT; + node->resultRelation = 0; + + /* fs_server will be filled in by create_foreignscan_plan */ + node->fs_server = InvalidOid; + node->fdw_exprs = fdw_exprs; + node->fdw_private = fdw_private; + node->fdw_scan_tlist = fdw_scan_tlist; + node->fdw_recheck_quals = fdw_recheck_quals; + /* fs_relids will be filled in by create_foreignscan_plan */ + node->fs_relids = NULL; + /* fsSystemCol will be filled in by create_foreignscan_plan */ + node->fsSystemCol = false; + + return node; +} + +static RecursiveUnion * +make_recursive_union(List *tlist, + Plan *lefttree, + Plan *righttree, + int wtParam, + List *distinctList, + long numGroups) +{ + RecursiveUnion *node = makeNode(RecursiveUnion); + Plan *plan = &node->plan; + int numCols = list_length(distinctList); + + plan->targetlist = tlist; + plan->qual = NIL; + plan->lefttree = lefttree; + plan->righttree = righttree; + node->wtParam = wtParam; + + /* + * convert SortGroupClause list into arrays of attr indexes and equality + * operators, as wanted by executor + */ + node->numCols = numCols; + if (numCols > 0) + { + int keyno = 0; + AttrNumber *dupColIdx; + Oid *dupOperators; + Oid *dupCollations; + ListCell *slitem; + + dupColIdx = (AttrNumber *) palloc(sizeof(AttrNumber) * numCols); + dupOperators = (Oid *) palloc(sizeof(Oid) * numCols); + dupCollations = (Oid *) palloc(sizeof(Oid) * numCols); + + foreach(slitem, distinctList) + { + SortGroupClause *sortcl = (SortGroupClause *) lfirst(slitem); + TargetEntry *tle = get_sortgroupclause_tle(sortcl, + plan->targetlist); + + dupColIdx[keyno] = tle->resno; + dupOperators[keyno] = sortcl->eqop; + dupCollations[keyno] = exprCollation((Node *) tle->expr); + Assert(OidIsValid(dupOperators[keyno])); + keyno++; + } + node->dupColIdx = dupColIdx; + node->dupOperators = dupOperators; + node->dupCollations = dupCollations; + } + node->numGroups = numGroups; + + return node; +} + +static BitmapAnd * +make_bitmap_and(List *bitmapplans) +{ + BitmapAnd *node = makeNode(BitmapAnd); + Plan *plan = &node->plan; + + plan->targetlist = NIL; + plan->qual = NIL; + plan->lefttree = NULL; + plan->righttree = NULL; + node->bitmapplans = bitmapplans; + + return node; +} + +static BitmapOr * +make_bitmap_or(List *bitmapplans) +{ + BitmapOr *node = makeNode(BitmapOr); + Plan *plan = &node->plan; + + plan->targetlist = NIL; + plan->qual = NIL; + plan->lefttree = NULL; + plan->righttree = NULL; + node->bitmapplans = bitmapplans; + + return node; +} + +static NestLoop * +make_nestloop(List *tlist, + List *joinclauses, + List *otherclauses, + List *nestParams, + Plan *lefttree, + Plan *righttree, + JoinType jointype, + bool inner_unique) +{ + NestLoop *node = makeNode(NestLoop); + Plan *plan = &node->join.plan; + + plan->targetlist = tlist; + plan->qual = otherclauses; + plan->lefttree = lefttree; + plan->righttree = righttree; + node->join.jointype = jointype; + node->join.inner_unique = inner_unique; + node->join.joinqual = joinclauses; + node->nestParams = nestParams; + + return node; +} + +static HashJoin * +make_hashjoin(List *tlist, + List *joinclauses, + List *otherclauses, + List *hashclauses, + List *hashoperators, + List *hashcollations, + List *hashkeys, + Plan *lefttree, + Plan *righttree, + JoinType jointype, + bool inner_unique) +{ + HashJoin *node = makeNode(HashJoin); + Plan *plan = &node->join.plan; + + plan->targetlist = tlist; + plan->qual = otherclauses; + plan->lefttree = lefttree; + plan->righttree = righttree; + node->hashclauses = hashclauses; + node->hashoperators = hashoperators; + node->hashcollations = hashcollations; + node->hashkeys = hashkeys; + node->join.jointype = jointype; + node->join.inner_unique = inner_unique; + node->join.joinqual = joinclauses; + + return node; +} + +static Hash * +make_hash(Plan *lefttree, + List *hashkeys, + Oid skewTable, + AttrNumber skewColumn, + bool skewInherit) +{ + Hash *node = makeNode(Hash); + Plan *plan = &node->plan; + + plan->targetlist = lefttree->targetlist; + plan->qual = NIL; + plan->lefttree = lefttree; + plan->righttree = NULL; + + node->hashkeys = hashkeys; + node->skewTable = skewTable; + node->skewColumn = skewColumn; + node->skewInherit = skewInherit; + + return node; +} + +static MergeJoin * +make_mergejoin(List *tlist, + List *joinclauses, + List *otherclauses, + List *mergeclauses, + Oid *mergefamilies, + Oid *mergecollations, + int *mergestrategies, + bool *mergenullsfirst, + Plan *lefttree, + Plan *righttree, + JoinType jointype, + bool inner_unique, + bool skip_mark_restore) +{ + MergeJoin *node = makeNode(MergeJoin); + Plan *plan = &node->join.plan; + + plan->targetlist = tlist; + plan->qual = otherclauses; + plan->lefttree = lefttree; + plan->righttree = righttree; + node->skip_mark_restore = skip_mark_restore; + node->mergeclauses = mergeclauses; + node->mergeFamilies = mergefamilies; + node->mergeCollations = mergecollations; + node->mergeStrategies = mergestrategies; + node->mergeNullsFirst = mergenullsfirst; + node->join.jointype = jointype; + node->join.inner_unique = inner_unique; + node->join.joinqual = joinclauses; + + return node; +} + +/* + * make_sort --- basic routine to build a Sort plan node + * + * Caller must have built the sortColIdx, sortOperators, collations, and + * nullsFirst arrays already. + */ +static Sort * +make_sort(Plan *lefttree, int numCols, + AttrNumber *sortColIdx, Oid *sortOperators, + Oid *collations, bool *nullsFirst) +{ + Sort *node; + Plan *plan; + + node = makeNode(Sort); + + plan = &node->plan; + plan->targetlist = lefttree->targetlist; + plan->qual = NIL; + plan->lefttree = lefttree; + plan->righttree = NULL; + node->numCols = numCols; + node->sortColIdx = sortColIdx; + node->sortOperators = sortOperators; + node->collations = collations; + node->nullsFirst = nullsFirst; + + return node; +} + +/* + * make_incrementalsort --- basic routine to build an IncrementalSort plan node + * + * Caller must have built the sortColIdx, sortOperators, collations, and + * nullsFirst arrays already. + */ +static IncrementalSort * +make_incrementalsort(Plan *lefttree, int numCols, int nPresortedCols, + AttrNumber *sortColIdx, Oid *sortOperators, + Oid *collations, bool *nullsFirst) +{ + IncrementalSort *node; + Plan *plan; + + node = makeNode(IncrementalSort); + + plan = &node->sort.plan; + plan->targetlist = lefttree->targetlist; + plan->qual = NIL; + plan->lefttree = lefttree; + plan->righttree = NULL; + node->nPresortedCols = nPresortedCols; + node->sort.numCols = numCols; + node->sort.sortColIdx = sortColIdx; + node->sort.sortOperators = sortOperators; + node->sort.collations = collations; + node->sort.nullsFirst = nullsFirst; + + return node; +} + +/* + * prepare_sort_from_pathkeys + * Prepare to sort according to given pathkeys + * + * This is used to set up for Sort, MergeAppend, and Gather Merge nodes. It + * calculates the executor's representation of the sort key information, and + * adjusts the plan targetlist if needed to add resjunk sort columns. + * + * Input parameters: + * 'lefttree' is the plan node which yields input tuples + * 'pathkeys' is the list of pathkeys by which the result is to be sorted + * 'relids' identifies the child relation being sorted, if any + * 'reqColIdx' is NULL or an array of required sort key column numbers + * 'adjust_tlist_in_place' is true if lefttree must be modified in-place + * + * We must convert the pathkey information into arrays of sort key column + * numbers, sort operator OIDs, collation OIDs, and nulls-first flags, + * which is the representation the executor wants. These are returned into + * the output parameters *p_numsortkeys etc. + * + * When looking for matches to an EquivalenceClass's members, we will only + * consider child EC members if they belong to given 'relids'. This protects + * against possible incorrect matches to child expressions that contain no + * Vars. + * + * If reqColIdx isn't NULL then it contains sort key column numbers that + * we should match. This is used when making child plans for a MergeAppend; + * it's an error if we can't match the columns. + * + * If the pathkeys include expressions that aren't simple Vars, we will + * usually need to add resjunk items to the input plan's targetlist to + * compute these expressions, since a Sort or MergeAppend node itself won't + * do any such calculations. If the input plan type isn't one that can do + * projections, this means adding a Result node just to do the projection. + * However, the caller can pass adjust_tlist_in_place = true to force the + * lefttree tlist to be modified in-place regardless of whether the node type + * can project --- we use this for fixing the tlist of MergeAppend itself. + * + * Returns the node which is to be the input to the Sort (either lefttree, + * or a Result stacked atop lefttree). + */ +static Plan * +prepare_sort_from_pathkeys(Plan *lefttree, List *pathkeys, + Relids relids, + const AttrNumber *reqColIdx, + bool adjust_tlist_in_place, + int *p_numsortkeys, + AttrNumber **p_sortColIdx, + Oid **p_sortOperators, + Oid **p_collations, + bool **p_nullsFirst) +{ + List *tlist = lefttree->targetlist; + ListCell *i; + int numsortkeys; + AttrNumber *sortColIdx; + Oid *sortOperators; + Oid *collations; + bool *nullsFirst; + + /* + * We will need at most list_length(pathkeys) sort columns; possibly less + */ + numsortkeys = list_length(pathkeys); + sortColIdx = (AttrNumber *) palloc(numsortkeys * sizeof(AttrNumber)); + sortOperators = (Oid *) palloc(numsortkeys * sizeof(Oid)); + collations = (Oid *) palloc(numsortkeys * sizeof(Oid)); + nullsFirst = (bool *) palloc(numsortkeys * sizeof(bool)); + + numsortkeys = 0; + + foreach(i, pathkeys) + { + PathKey *pathkey = (PathKey *) lfirst(i); + EquivalenceClass *ec = pathkey->pk_eclass; + EquivalenceMember *em; + TargetEntry *tle = NULL; + Oid pk_datatype = InvalidOid; + Oid sortop; + ListCell *j; + + if (ec->ec_has_volatile) + { + /* + * If the pathkey's EquivalenceClass is volatile, then it must + * have come from an ORDER BY clause, and we have to match it to + * that same targetlist entry. + */ + if (ec->ec_sortref == 0) /* can't happen */ + elog(ERROR, "volatile EquivalenceClass has no sortref"); + tle = get_sortgroupref_tle(ec->ec_sortref, tlist); + Assert(tle); + Assert(list_length(ec->ec_members) == 1); + pk_datatype = ((EquivalenceMember *) linitial(ec->ec_members))->em_datatype; + } + else if (reqColIdx != NULL) + { + /* + * If we are given a sort column number to match, only consider + * the single TLE at that position. It's possible that there is + * no such TLE, in which case fall through and generate a resjunk + * targetentry (we assume this must have happened in the parent + * plan as well). If there is a TLE but it doesn't match the + * pathkey's EC, we do the same, which is probably the wrong thing + * but we'll leave it to caller to complain about the mismatch. + */ + tle = get_tle_by_resno(tlist, reqColIdx[numsortkeys]); + if (tle) + { + em = find_ec_member_matching_expr(ec, tle->expr, relids); + if (em) + { + /* found expr at right place in tlist */ + pk_datatype = em->em_datatype; + } + else + tle = NULL; + } + } + else + { + /* + * Otherwise, we can sort by any non-constant expression listed in + * the pathkey's EquivalenceClass. For now, we take the first + * tlist item found in the EC. If there's no match, we'll generate + * a resjunk entry using the first EC member that is an expression + * in the input's vars. (The non-const restriction only matters + * if the EC is below_outer_join; but if it isn't, it won't + * contain consts anyway, else we'd have discarded the pathkey as + * redundant.) + * + * XXX if we have a choice, is there any way of figuring out which + * might be cheapest to execute? (For example, int4lt is likely + * much cheaper to execute than numericlt, but both might appear + * in the same equivalence class...) Not clear that we ever will + * have an interesting choice in practice, so it may not matter. + */ + foreach(j, tlist) + { + tle = (TargetEntry *) lfirst(j); + em = find_ec_member_matching_expr(ec, tle->expr, relids); + if (em) + { + /* found expr already in tlist */ + pk_datatype = em->em_datatype; + break; + } + tle = NULL; + } + } + + if (!tle) + { + /* + * No matching tlist item; look for a computable expression. + */ + em = find_computable_ec_member(NULL, ec, tlist, relids, false); + if (!em) + elog(ERROR, "could not find pathkey item to sort"); + pk_datatype = em->em_datatype; + + /* + * Do we need to insert a Result node? + */ + if (!adjust_tlist_in_place && + !is_projection_capable_plan(lefttree)) + { + /* copy needed so we don't modify input's tlist below */ + tlist = copyObject(tlist); + lefttree = inject_projection_plan(lefttree, tlist, + lefttree->parallel_safe); + } + + /* Don't bother testing is_projection_capable_plan again */ + adjust_tlist_in_place = true; + + /* + * Add resjunk entry to input's tlist + */ + tle = makeTargetEntry(copyObject(em->em_expr), + list_length(tlist) + 1, + NULL, + true); + tlist = lappend(tlist, tle); + lefttree->targetlist = tlist; /* just in case NIL before */ + } + + /* + * Look up the correct sort operator from the PathKey's slightly + * abstracted representation. + */ + sortop = get_opfamily_member(pathkey->pk_opfamily, + pk_datatype, + pk_datatype, + pathkey->pk_strategy); + if (!OidIsValid(sortop)) /* should not happen */ + elog(ERROR, "missing operator %d(%u,%u) in opfamily %u", + pathkey->pk_strategy, pk_datatype, pk_datatype, + pathkey->pk_opfamily); + + /* Add the column to the sort arrays */ + sortColIdx[numsortkeys] = tle->resno; + sortOperators[numsortkeys] = sortop; + collations[numsortkeys] = ec->ec_collation; + nullsFirst[numsortkeys] = pathkey->pk_nulls_first; + numsortkeys++; + } + + /* Return results */ + *p_numsortkeys = numsortkeys; + *p_sortColIdx = sortColIdx; + *p_sortOperators = sortOperators; + *p_collations = collations; + *p_nullsFirst = nullsFirst; + + return lefttree; +} + +/* + * make_sort_from_pathkeys + * Create sort plan to sort according to given pathkeys + * + * 'lefttree' is the node which yields input tuples + * 'pathkeys' is the list of pathkeys by which the result is to be sorted + * 'relids' is the set of relations required by prepare_sort_from_pathkeys() + */ +static Sort * +make_sort_from_pathkeys(Plan *lefttree, List *pathkeys, Relids relids) +{ + int numsortkeys; + AttrNumber *sortColIdx; + Oid *sortOperators; + Oid *collations; + bool *nullsFirst; + + /* Compute sort column info, and adjust lefttree as needed */ + lefttree = prepare_sort_from_pathkeys(lefttree, pathkeys, + relids, + NULL, + false, + &numsortkeys, + &sortColIdx, + &sortOperators, + &collations, + &nullsFirst); + + /* Now build the Sort node */ + return make_sort(lefttree, numsortkeys, + sortColIdx, sortOperators, + collations, nullsFirst); +} + +/* + * make_incrementalsort_from_pathkeys + * Create sort plan to sort according to given pathkeys + * + * 'lefttree' is the node which yields input tuples + * 'pathkeys' is the list of pathkeys by which the result is to be sorted + * 'relids' is the set of relations required by prepare_sort_from_pathkeys() + * 'nPresortedCols' is the number of presorted columns in input tuples + */ +static IncrementalSort * +make_incrementalsort_from_pathkeys(Plan *lefttree, List *pathkeys, + Relids relids, int nPresortedCols) +{ + int numsortkeys; + AttrNumber *sortColIdx; + Oid *sortOperators; + Oid *collations; + bool *nullsFirst; + + /* Compute sort column info, and adjust lefttree as needed */ + lefttree = prepare_sort_from_pathkeys(lefttree, pathkeys, + relids, + NULL, + false, + &numsortkeys, + &sortColIdx, + &sortOperators, + &collations, + &nullsFirst); + + /* Now build the Sort node */ + return make_incrementalsort(lefttree, numsortkeys, nPresortedCols, + sortColIdx, sortOperators, + collations, nullsFirst); +} + +/* + * make_sort_from_sortclauses + * Create sort plan to sort according to given sortclauses + * + * 'sortcls' is a list of SortGroupClauses + * 'lefttree' is the node which yields input tuples + */ +Sort * +make_sort_from_sortclauses(List *sortcls, Plan *lefttree) +{ + List *sub_tlist = lefttree->targetlist; + ListCell *l; + int numsortkeys; + AttrNumber *sortColIdx; + Oid *sortOperators; + Oid *collations; + bool *nullsFirst; + + /* Convert list-ish representation to arrays wanted by executor */ + numsortkeys = list_length(sortcls); + sortColIdx = (AttrNumber *) palloc(numsortkeys * sizeof(AttrNumber)); + sortOperators = (Oid *) palloc(numsortkeys * sizeof(Oid)); + collations = (Oid *) palloc(numsortkeys * sizeof(Oid)); + nullsFirst = (bool *) palloc(numsortkeys * sizeof(bool)); + + numsortkeys = 0; + foreach(l, sortcls) + { + SortGroupClause *sortcl = (SortGroupClause *) lfirst(l); + TargetEntry *tle = get_sortgroupclause_tle(sortcl, sub_tlist); + + sortColIdx[numsortkeys] = tle->resno; + sortOperators[numsortkeys] = sortcl->sortop; + collations[numsortkeys] = exprCollation((Node *) tle->expr); + nullsFirst[numsortkeys] = sortcl->nulls_first; + numsortkeys++; + } + + return make_sort(lefttree, numsortkeys, + sortColIdx, sortOperators, + collations, nullsFirst); +} + +/* + * make_sort_from_groupcols + * Create sort plan to sort based on grouping columns + * + * 'groupcls' is the list of SortGroupClauses + * 'grpColIdx' gives the column numbers to use + * + * This might look like it could be merged with make_sort_from_sortclauses, + * but presently we *must* use the grpColIdx[] array to locate sort columns, + * because the child plan's tlist is not marked with ressortgroupref info + * appropriate to the grouping node. So, only the sort ordering info + * is used from the SortGroupClause entries. + */ +static Sort * +make_sort_from_groupcols(List *groupcls, + AttrNumber *grpColIdx, + Plan *lefttree) +{ + List *sub_tlist = lefttree->targetlist; + ListCell *l; + int numsortkeys; + AttrNumber *sortColIdx; + Oid *sortOperators; + Oid *collations; + bool *nullsFirst; + + /* Convert list-ish representation to arrays wanted by executor */ + numsortkeys = list_length(groupcls); + sortColIdx = (AttrNumber *) palloc(numsortkeys * sizeof(AttrNumber)); + sortOperators = (Oid *) palloc(numsortkeys * sizeof(Oid)); + collations = (Oid *) palloc(numsortkeys * sizeof(Oid)); + nullsFirst = (bool *) palloc(numsortkeys * sizeof(bool)); + + numsortkeys = 0; + foreach(l, groupcls) + { + SortGroupClause *grpcl = (SortGroupClause *) lfirst(l); + TargetEntry *tle = get_tle_by_resno(sub_tlist, grpColIdx[numsortkeys]); + + if (!tle) + elog(ERROR, "could not retrieve tle for sort-from-groupcols"); + + sortColIdx[numsortkeys] = tle->resno; + sortOperators[numsortkeys] = grpcl->sortop; + collations[numsortkeys] = exprCollation((Node *) tle->expr); + nullsFirst[numsortkeys] = grpcl->nulls_first; + numsortkeys++; + } + + return make_sort(lefttree, numsortkeys, + sortColIdx, sortOperators, + collations, nullsFirst); +} + +static Material * +make_material(Plan *lefttree) +{ + Material *node = makeNode(Material); + Plan *plan = &node->plan; + + plan->targetlist = lefttree->targetlist; + plan->qual = NIL; + plan->lefttree = lefttree; + plan->righttree = NULL; + + return node; +} + +/* + * materialize_finished_plan: stick a Material node atop a completed plan + * + * There are a couple of places where we want to attach a Material node + * after completion of create_plan(), without any MaterialPath path. + * Those places should probably be refactored someday to do this on the + * Path representation, but it's not worth the trouble yet. + */ +Plan * +materialize_finished_plan(Plan *subplan) +{ + Plan *matplan; + Path matpath; /* dummy for result of cost_material */ + + matplan = (Plan *) make_material(subplan); + + /* + * XXX horrid kluge: if there are any initPlans attached to the subplan, + * move them up to the Material node, which is now effectively the top + * plan node in its query level. This prevents failure in + * SS_finalize_plan(), which see for comments. We don't bother adjusting + * the subplan's cost estimate for this. + */ + matplan->initPlan = subplan->initPlan; + subplan->initPlan = NIL; + + /* Set cost data */ + cost_material(&matpath, + subplan->startup_cost, + subplan->total_cost, + subplan->plan_rows, + subplan->plan_width); + matplan->startup_cost = matpath.startup_cost; + matplan->total_cost = matpath.total_cost; + matplan->plan_rows = subplan->plan_rows; + matplan->plan_width = subplan->plan_width; + matplan->parallel_aware = false; + matplan->parallel_safe = subplan->parallel_safe; + + return matplan; +} + +static Memoize * +make_memoize(Plan *lefttree, Oid *hashoperators, Oid *collations, + List *param_exprs, bool singlerow, bool binary_mode, + uint32 est_entries, Bitmapset *keyparamids) +{ + Memoize *node = makeNode(Memoize); + Plan *plan = &node->plan; + + plan->targetlist = lefttree->targetlist; + plan->qual = NIL; + plan->lefttree = lefttree; + plan->righttree = NULL; + + node->numKeys = list_length(param_exprs); + node->hashOperators = hashoperators; + node->collations = collations; + node->param_exprs = param_exprs; + node->singlerow = singlerow; + node->binary_mode = binary_mode; + node->est_entries = est_entries; + node->keyparamids = keyparamids; + + return node; +} + +Agg * +make_agg(List *tlist, List *qual, + AggStrategy aggstrategy, AggSplit aggsplit, + int numGroupCols, AttrNumber *grpColIdx, Oid *grpOperators, Oid *grpCollations, + List *groupingSets, List *chain, double dNumGroups, + Size transitionSpace, Plan *lefttree) +{ + Agg *node = makeNode(Agg); + Plan *plan = &node->plan; + long numGroups; + + /* Reduce to long, but 'ware overflow! */ + numGroups = clamp_cardinality_to_long(dNumGroups); + + node->aggstrategy = aggstrategy; + node->aggsplit = aggsplit; + node->numCols = numGroupCols; + node->grpColIdx = grpColIdx; + node->grpOperators = grpOperators; + node->grpCollations = grpCollations; + node->numGroups = numGroups; + node->transitionSpace = transitionSpace; + node->aggParams = NULL; /* SS_finalize_plan() will fill this */ + node->groupingSets = groupingSets; + node->chain = chain; + + plan->qual = qual; + plan->targetlist = tlist; + plan->lefttree = lefttree; + plan->righttree = NULL; + + return node; +} + +static WindowAgg * +make_windowagg(List *tlist, Index winref, + int partNumCols, AttrNumber *partColIdx, Oid *partOperators, Oid *partCollations, + int ordNumCols, AttrNumber *ordColIdx, Oid *ordOperators, Oid *ordCollations, + int frameOptions, Node *startOffset, Node *endOffset, + Oid startInRangeFunc, Oid endInRangeFunc, + Oid inRangeColl, bool inRangeAsc, bool inRangeNullsFirst, + List *runCondition, List *qual, bool topWindow, Plan *lefttree) +{ + WindowAgg *node = makeNode(WindowAgg); + Plan *plan = &node->plan; + + node->winref = winref; + node->partNumCols = partNumCols; + node->partColIdx = partColIdx; + node->partOperators = partOperators; + node->partCollations = partCollations; + node->ordNumCols = ordNumCols; + node->ordColIdx = ordColIdx; + node->ordOperators = ordOperators; + node->ordCollations = ordCollations; + node->frameOptions = frameOptions; + node->startOffset = startOffset; + node->endOffset = endOffset; + node->runCondition = runCondition; + /* a duplicate of the above for EXPLAIN */ + node->runConditionOrig = runCondition; + node->startInRangeFunc = startInRangeFunc; + node->endInRangeFunc = endInRangeFunc; + node->inRangeColl = inRangeColl; + node->inRangeAsc = inRangeAsc; + node->inRangeNullsFirst = inRangeNullsFirst; + node->topWindow = topWindow; + + plan->targetlist = tlist; + plan->lefttree = lefttree; + plan->righttree = NULL; + plan->qual = qual; + + return node; +} + +static Group * +make_group(List *tlist, + List *qual, + int numGroupCols, + AttrNumber *grpColIdx, + Oid *grpOperators, + Oid *grpCollations, + Plan *lefttree) +{ + Group *node = makeNode(Group); + Plan *plan = &node->plan; + + node->numCols = numGroupCols; + node->grpColIdx = grpColIdx; + node->grpOperators = grpOperators; + node->grpCollations = grpCollations; + + plan->qual = qual; + plan->targetlist = tlist; + plan->lefttree = lefttree; + plan->righttree = NULL; + + return node; +} + +/* + * distinctList is a list of SortGroupClauses, identifying the targetlist items + * that should be considered by the Unique filter. The input path must + * already be sorted accordingly. + */ +static Unique * +make_unique_from_sortclauses(Plan *lefttree, List *distinctList) +{ + Unique *node = makeNode(Unique); + Plan *plan = &node->plan; + int numCols = list_length(distinctList); + int keyno = 0; + AttrNumber *uniqColIdx; + Oid *uniqOperators; + Oid *uniqCollations; + ListCell *slitem; + + plan->targetlist = lefttree->targetlist; + plan->qual = NIL; + plan->lefttree = lefttree; + plan->righttree = NULL; + + /* + * convert SortGroupClause list into arrays of attr indexes and equality + * operators, as wanted by executor + */ + Assert(numCols > 0); + uniqColIdx = (AttrNumber *) palloc(sizeof(AttrNumber) * numCols); + uniqOperators = (Oid *) palloc(sizeof(Oid) * numCols); + uniqCollations = (Oid *) palloc(sizeof(Oid) * numCols); + + foreach(slitem, distinctList) + { + SortGroupClause *sortcl = (SortGroupClause *) lfirst(slitem); + TargetEntry *tle = get_sortgroupclause_tle(sortcl, plan->targetlist); + + uniqColIdx[keyno] = tle->resno; + uniqOperators[keyno] = sortcl->eqop; + uniqCollations[keyno] = exprCollation((Node *) tle->expr); + Assert(OidIsValid(uniqOperators[keyno])); + keyno++; + } + + node->numCols = numCols; + node->uniqColIdx = uniqColIdx; + node->uniqOperators = uniqOperators; + node->uniqCollations = uniqCollations; + + return node; +} + +/* + * as above, but use pathkeys to identify the sort columns and semantics + */ +static Unique * +make_unique_from_pathkeys(Plan *lefttree, List *pathkeys, int numCols) +{ + Unique *node = makeNode(Unique); + Plan *plan = &node->plan; + int keyno = 0; + AttrNumber *uniqColIdx; + Oid *uniqOperators; + Oid *uniqCollations; + ListCell *lc; + + plan->targetlist = lefttree->targetlist; + plan->qual = NIL; + plan->lefttree = lefttree; + plan->righttree = NULL; + + /* + * Convert pathkeys list into arrays of attr indexes and equality + * operators, as wanted by executor. This has a lot in common with + * prepare_sort_from_pathkeys ... maybe unify sometime? + */ + Assert(numCols >= 0 && numCols <= list_length(pathkeys)); + uniqColIdx = (AttrNumber *) palloc(sizeof(AttrNumber) * numCols); + uniqOperators = (Oid *) palloc(sizeof(Oid) * numCols); + uniqCollations = (Oid *) palloc(sizeof(Oid) * numCols); + + foreach(lc, pathkeys) + { + PathKey *pathkey = (PathKey *) lfirst(lc); + EquivalenceClass *ec = pathkey->pk_eclass; + EquivalenceMember *em; + TargetEntry *tle = NULL; + Oid pk_datatype = InvalidOid; + Oid eqop; + ListCell *j; + + /* Ignore pathkeys beyond the specified number of columns */ + if (keyno >= numCols) + break; + + if (ec->ec_has_volatile) + { + /* + * If the pathkey's EquivalenceClass is volatile, then it must + * have come from an ORDER BY clause, and we have to match it to + * that same targetlist entry. + */ + if (ec->ec_sortref == 0) /* can't happen */ + elog(ERROR, "volatile EquivalenceClass has no sortref"); + tle = get_sortgroupref_tle(ec->ec_sortref, plan->targetlist); + Assert(tle); + Assert(list_length(ec->ec_members) == 1); + pk_datatype = ((EquivalenceMember *) linitial(ec->ec_members))->em_datatype; + } + else + { + /* + * Otherwise, we can use any non-constant expression listed in the + * pathkey's EquivalenceClass. For now, we take the first tlist + * item found in the EC. + */ + foreach(j, plan->targetlist) + { + tle = (TargetEntry *) lfirst(j); + em = find_ec_member_matching_expr(ec, tle->expr, NULL); + if (em) + { + /* found expr already in tlist */ + pk_datatype = em->em_datatype; + break; + } + tle = NULL; + } + } + + if (!tle) + elog(ERROR, "could not find pathkey item to sort"); + + /* + * Look up the correct equality operator from the PathKey's slightly + * abstracted representation. + */ + eqop = get_opfamily_member(pathkey->pk_opfamily, + pk_datatype, + pk_datatype, + BTEqualStrategyNumber); + if (!OidIsValid(eqop)) /* should not happen */ + elog(ERROR, "missing operator %d(%u,%u) in opfamily %u", + BTEqualStrategyNumber, pk_datatype, pk_datatype, + pathkey->pk_opfamily); + + uniqColIdx[keyno] = tle->resno; + uniqOperators[keyno] = eqop; + uniqCollations[keyno] = ec->ec_collation; + + keyno++; + } + + node->numCols = numCols; + node->uniqColIdx = uniqColIdx; + node->uniqOperators = uniqOperators; + node->uniqCollations = uniqCollations; + + return node; +} + +static Gather * +make_gather(List *qptlist, + List *qpqual, + int nworkers, + int rescan_param, + bool single_copy, + Plan *subplan) +{ + Gather *node = makeNode(Gather); + Plan *plan = &node->plan; + + plan->targetlist = qptlist; + plan->qual = qpqual; + plan->lefttree = subplan; + plan->righttree = NULL; + node->num_workers = nworkers; + node->rescan_param = rescan_param; + node->single_copy = single_copy; + node->invisible = false; + node->initParam = NULL; + + return node; +} + +/* + * distinctList is a list of SortGroupClauses, identifying the targetlist + * items that should be considered by the SetOp filter. The input path must + * already be sorted accordingly. + */ +static SetOp * +make_setop(SetOpCmd cmd, SetOpStrategy strategy, Plan *lefttree, + List *distinctList, AttrNumber flagColIdx, int firstFlag, + long numGroups) +{ + SetOp *node = makeNode(SetOp); + Plan *plan = &node->plan; + int numCols = list_length(distinctList); + int keyno = 0; + AttrNumber *dupColIdx; + Oid *dupOperators; + Oid *dupCollations; + ListCell *slitem; + + plan->targetlist = lefttree->targetlist; + plan->qual = NIL; + plan->lefttree = lefttree; + plan->righttree = NULL; + + /* + * convert SortGroupClause list into arrays of attr indexes and equality + * operators, as wanted by executor + */ + dupColIdx = (AttrNumber *) palloc(sizeof(AttrNumber) * numCols); + dupOperators = (Oid *) palloc(sizeof(Oid) * numCols); + dupCollations = (Oid *) palloc(sizeof(Oid) * numCols); + + foreach(slitem, distinctList) + { + SortGroupClause *sortcl = (SortGroupClause *) lfirst(slitem); + TargetEntry *tle = get_sortgroupclause_tle(sortcl, plan->targetlist); + + dupColIdx[keyno] = tle->resno; + dupOperators[keyno] = sortcl->eqop; + dupCollations[keyno] = exprCollation((Node *) tle->expr); + Assert(OidIsValid(dupOperators[keyno])); + keyno++; + } + + node->cmd = cmd; + node->strategy = strategy; + node->numCols = numCols; + node->dupColIdx = dupColIdx; + node->dupOperators = dupOperators; + node->dupCollations = dupCollations; + node->flagColIdx = flagColIdx; + node->firstFlag = firstFlag; + node->numGroups = numGroups; + + return node; +} + +/* + * make_lockrows + * Build a LockRows plan node + */ +static LockRows * +make_lockrows(Plan *lefttree, List *rowMarks, int epqParam) +{ + LockRows *node = makeNode(LockRows); + Plan *plan = &node->plan; + + plan->targetlist = lefttree->targetlist; + plan->qual = NIL; + plan->lefttree = lefttree; + plan->righttree = NULL; + + node->rowMarks = rowMarks; + node->epqParam = epqParam; + + return node; +} + +/* + * make_limit + * Build a Limit plan node + */ +Limit * +make_limit(Plan *lefttree, Node *limitOffset, Node *limitCount, + LimitOption limitOption, int uniqNumCols, AttrNumber *uniqColIdx, + Oid *uniqOperators, Oid *uniqCollations) +{ + Limit *node = makeNode(Limit); + Plan *plan = &node->plan; + + plan->targetlist = lefttree->targetlist; + plan->qual = NIL; + plan->lefttree = lefttree; + plan->righttree = NULL; + + node->limitOffset = limitOffset; + node->limitCount = limitCount; + node->limitOption = limitOption; + node->uniqNumCols = uniqNumCols; + node->uniqColIdx = uniqColIdx; + node->uniqOperators = uniqOperators; + node->uniqCollations = uniqCollations; + + return node; +} + +/* + * make_result + * Build a Result plan node + */ +static Result * +make_result(List *tlist, + Node *resconstantqual, + Plan *subplan) +{ + Result *node = makeNode(Result); + Plan *plan = &node->plan; + + plan->targetlist = tlist; + plan->qual = NIL; + plan->lefttree = subplan; + plan->righttree = NULL; + node->resconstantqual = resconstantqual; + + return node; +} + +/* + * make_project_set + * Build a ProjectSet plan node + */ +static ProjectSet * +make_project_set(List *tlist, + Plan *subplan) +{ + ProjectSet *node = makeNode(ProjectSet); + Plan *plan = &node->plan; + + plan->targetlist = tlist; + plan->qual = NIL; + plan->lefttree = subplan; + plan->righttree = NULL; + + return node; +} + +/* + * make_modifytable + * Build a ModifyTable plan node + */ +static ModifyTable * +make_modifytable(PlannerInfo *root, Plan *subplan, + CmdType operation, bool canSetTag, + Index nominalRelation, Index rootRelation, + bool partColsUpdated, + List *resultRelations, + List *updateColnosLists, + List *withCheckOptionLists, List *returningLists, + List *rowMarks, OnConflictExpr *onconflict, + List *mergeActionLists, int epqParam) +{ + ModifyTable *node = makeNode(ModifyTable); + List *fdw_private_list; + Bitmapset *direct_modify_plans; + ListCell *lc; + int i; + + Assert(operation == CMD_MERGE || + (operation == CMD_UPDATE ? + list_length(resultRelations) == list_length(updateColnosLists) : + updateColnosLists == NIL)); + Assert(withCheckOptionLists == NIL || + list_length(resultRelations) == list_length(withCheckOptionLists)); + Assert(returningLists == NIL || + list_length(resultRelations) == list_length(returningLists)); + + node->plan.lefttree = subplan; + node->plan.righttree = NULL; + node->plan.qual = NIL; + /* setrefs.c will fill in the targetlist, if needed */ + node->plan.targetlist = NIL; + + node->operation = operation; + node->canSetTag = canSetTag; + node->nominalRelation = nominalRelation; + node->rootRelation = rootRelation; + node->partColsUpdated = partColsUpdated; + node->resultRelations = resultRelations; + if (!onconflict) + { + node->onConflictAction = ONCONFLICT_NONE; + node->onConflictSet = NIL; + node->onConflictCols = NIL; + node->onConflictWhere = NULL; + node->arbiterIndexes = NIL; + node->exclRelRTI = 0; + node->exclRelTlist = NIL; + } + else + { + node->onConflictAction = onconflict->action; + + /* + * Here we convert the ON CONFLICT UPDATE tlist, if any, to the + * executor's convention of having consecutive resno's. The actual + * target column numbers are saved in node->onConflictCols. (This + * could be done earlier, but there seems no need to.) + */ + node->onConflictSet = onconflict->onConflictSet; + node->onConflictCols = + extract_update_targetlist_colnos(node->onConflictSet); + node->onConflictWhere = onconflict->onConflictWhere; + + /* + * If a set of unique index inference elements was provided (an + * INSERT...ON CONFLICT "inference specification"), then infer + * appropriate unique indexes (or throw an error if none are + * available). + */ + node->arbiterIndexes = infer_arbiter_indexes(root); + + node->exclRelRTI = onconflict->exclRelIndex; + node->exclRelTlist = onconflict->exclRelTlist; + } + node->updateColnosLists = updateColnosLists; + node->withCheckOptionLists = withCheckOptionLists; + node->returningLists = returningLists; + node->rowMarks = rowMarks; + node->mergeActionLists = mergeActionLists; + node->epqParam = epqParam; + + /* + * For each result relation that is a foreign table, allow the FDW to + * construct private plan data, and accumulate it all into a list. + */ + fdw_private_list = NIL; + direct_modify_plans = NULL; + i = 0; + foreach(lc, resultRelations) + { + Index rti = lfirst_int(lc); + FdwRoutine *fdwroutine; + List *fdw_private; + bool direct_modify; + + /* + * If possible, we want to get the FdwRoutine from our RelOptInfo for + * the table. But sometimes we don't have a RelOptInfo and must get + * it the hard way. (In INSERT, the target relation is not scanned, + * so it's not a baserel; and there are also corner cases for + * updatable views where the target rel isn't a baserel.) + */ + if (rti < root->simple_rel_array_size && + root->simple_rel_array[rti] != NULL) + { + RelOptInfo *resultRel = root->simple_rel_array[rti]; + + fdwroutine = resultRel->fdwroutine; + + /* + * MERGE is not currently supported for foreign tables and we + * already checked when the table mentioned in the query is + * foreign; but we can still get here if a partitioned table has a + * foreign table as partition. Disallow that now, to avoid an + * uglier error message later. + */ + if (operation == CMD_MERGE && fdwroutine != NULL) + { + RangeTblEntry *rte = root->simple_rte_array[rti]; + + ereport(ERROR, + errcode(ERRCODE_FEATURE_NOT_SUPPORTED), + errmsg("cannot execute MERGE on relation \"%s\"", + get_rel_name(rte->relid)), + errdetail_relkind_not_supported(rte->relkind)); + } + + } + else + { + RangeTblEntry *rte = planner_rt_fetch(rti, root); + + Assert(rte->rtekind == RTE_RELATION); + Assert(operation != CMD_MERGE); + if (rte->relkind == RELKIND_FOREIGN_TABLE) + fdwroutine = GetFdwRoutineByRelId(rte->relid); + else + fdwroutine = NULL; + } + + /* + * Try to modify the foreign table directly if (1) the FDW provides + * callback functions needed for that and (2) there are no local + * structures that need to be run for each modified row: row-level + * triggers on the foreign table, stored generated columns, WITH CHECK + * OPTIONs from parent views. + */ + direct_modify = false; + if (fdwroutine != NULL && + fdwroutine->PlanDirectModify != NULL && + fdwroutine->BeginDirectModify != NULL && + fdwroutine->IterateDirectModify != NULL && + fdwroutine->EndDirectModify != NULL && + withCheckOptionLists == NIL && + !has_row_triggers(root, rti, operation) && + !has_stored_generated_columns(root, rti)) + direct_modify = fdwroutine->PlanDirectModify(root, node, rti, i); + if (direct_modify) + direct_modify_plans = bms_add_member(direct_modify_plans, i); + + if (!direct_modify && + fdwroutine != NULL && + fdwroutine->PlanForeignModify != NULL) + fdw_private = fdwroutine->PlanForeignModify(root, node, rti, i); + else + fdw_private = NIL; + fdw_private_list = lappend(fdw_private_list, fdw_private); + i++; + } + node->fdwPrivLists = fdw_private_list; + node->fdwDirectModifyPlans = direct_modify_plans; + + return node; +} + +/* + * is_projection_capable_path + * Check whether a given Path node is able to do projection. + */ +bool +is_projection_capable_path(Path *path) +{ + /* Most plan types can project, so just list the ones that can't */ + switch (path->pathtype) + { + case T_Hash: + case T_Material: + case T_Memoize: + case T_Sort: + case T_IncrementalSort: + case T_Unique: + case T_SetOp: + case T_LockRows: + case T_Limit: + case T_ModifyTable: + case T_MergeAppend: + case T_RecursiveUnion: + return false; + case T_CustomScan: + if (castNode(CustomPath, path)->flags & CUSTOMPATH_SUPPORT_PROJECTION) + return true; + return false; + case T_Append: + + /* + * Append can't project, but if an AppendPath is being used to + * represent a dummy path, what will actually be generated is a + * Result which can project. + */ + return IS_DUMMY_APPEND(path); + case T_ProjectSet: + + /* + * Although ProjectSet certainly projects, say "no" because we + * don't want the planner to randomly replace its tlist with + * something else; the SRFs have to stay at top level. This might + * get relaxed later. + */ + return false; + default: + break; + } + return true; +} + +/* + * is_projection_capable_plan + * Check whether a given Plan node is able to do projection. + */ +bool +is_projection_capable_plan(Plan *plan) +{ + /* Most plan types can project, so just list the ones that can't */ + switch (nodeTag(plan)) + { + case T_Hash: + case T_Material: + case T_Memoize: + case T_Sort: + case T_Unique: + case T_SetOp: + case T_LockRows: + case T_Limit: + case T_ModifyTable: + case T_Append: + case T_MergeAppend: + case T_RecursiveUnion: + return false; + case T_CustomScan: + if (((CustomScan *) plan)->flags & CUSTOMPATH_SUPPORT_PROJECTION) + return true; + return false; + case T_ProjectSet: + + /* + * Although ProjectSet certainly projects, say "no" because we + * don't want the planner to randomly replace its tlist with + * something else; the SRFs have to stay at top level. This might + * get relaxed later. + */ + return false; + default: + break; + } + return true; +} diff --git a/src/backend/optimizer/plan/initsplan.c b/src/backend/optimizer/plan/initsplan.c new file mode 100644 index 0000000..023efba --- /dev/null +++ b/src/backend/optimizer/plan/initsplan.c @@ -0,0 +1,2752 @@ +/*------------------------------------------------------------------------- + * + * initsplan.c + * Target list, qualification, joininfo initialization routines + * + * Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group + * Portions Copyright (c) 1994, Regents of the University of California + * + * + * IDENTIFICATION + * src/backend/optimizer/plan/initsplan.c + * + *------------------------------------------------------------------------- + */ +#include "postgres.h" + +#include "catalog/pg_class.h" +#include "catalog/pg_type.h" +#include "nodes/makefuncs.h" +#include "nodes/nodeFuncs.h" +#include "optimizer/clauses.h" +#include "optimizer/cost.h" +#include "optimizer/inherit.h" +#include "optimizer/joininfo.h" +#include "optimizer/optimizer.h" +#include "optimizer/pathnode.h" +#include "optimizer/paths.h" +#include "optimizer/placeholder.h" +#include "optimizer/planmain.h" +#include "optimizer/planner.h" +#include "optimizer/prep.h" +#include "optimizer/restrictinfo.h" +#include "parser/analyze.h" +#include "rewrite/rewriteManip.h" +#include "utils/lsyscache.h" +#include "utils/typcache.h" + +/* These parameters are set by GUC */ +int from_collapse_limit; +int join_collapse_limit; + + +/* Elements of the postponed_qual_list used during deconstruct_recurse */ +typedef struct PostponedQual +{ + Node *qual; /* a qual clause waiting to be processed */ + Relids relids; /* the set of baserels it references */ +} PostponedQual; + + +static void extract_lateral_references(PlannerInfo *root, RelOptInfo *brel, + Index rtindex); +static List *deconstruct_recurse(PlannerInfo *root, Node *jtnode, + bool below_outer_join, + Relids *qualscope, Relids *inner_join_rels, + List **postponed_qual_list); +static void process_security_barrier_quals(PlannerInfo *root, + int rti, Relids qualscope, + bool below_outer_join); +static SpecialJoinInfo *make_outerjoininfo(PlannerInfo *root, + Relids left_rels, Relids right_rels, + Relids inner_join_rels, + JoinType jointype, List *clause); +static void compute_semijoin_info(PlannerInfo *root, SpecialJoinInfo *sjinfo, + List *clause); +static void distribute_qual_to_rels(PlannerInfo *root, Node *clause, + bool below_outer_join, + JoinType jointype, + Index security_level, + Relids qualscope, + Relids ojscope, + Relids outerjoin_nonnullable, + List **postponed_qual_list); +static bool check_outerjoin_delay(PlannerInfo *root, Relids *relids_p, + Relids *nullable_relids_p, bool is_pushed_down); +static bool check_equivalence_delay(PlannerInfo *root, + RestrictInfo *restrictinfo); +static bool check_redundant_nullability_qual(PlannerInfo *root, Node *clause); +static void check_mergejoinable(RestrictInfo *restrictinfo); +static void check_hashjoinable(RestrictInfo *restrictinfo); +static void check_memoizable(RestrictInfo *restrictinfo); + + +/***************************************************************************** + * + * JOIN TREES + * + *****************************************************************************/ + +/* + * add_base_rels_to_query + * + * Scan the query's jointree and create baserel RelOptInfos for all + * the base relations (e.g., table, subquery, and function RTEs) + * appearing in the jointree. + * + * The initial invocation must pass root->parse->jointree as the value of + * jtnode. Internally, the function recurses through the jointree. + * + * At the end of this process, there should be one baserel RelOptInfo for + * every non-join RTE that is used in the query. Some of the baserels + * may be appendrel parents, which will require additional "otherrel" + * RelOptInfos for their member rels, but those are added later. + */ +void +add_base_rels_to_query(PlannerInfo *root, Node *jtnode) +{ + if (jtnode == NULL) + return; + if (IsA(jtnode, RangeTblRef)) + { + int varno = ((RangeTblRef *) jtnode)->rtindex; + + (void) build_simple_rel(root, varno, NULL); + } + else if (IsA(jtnode, FromExpr)) + { + FromExpr *f = (FromExpr *) jtnode; + ListCell *l; + + foreach(l, f->fromlist) + add_base_rels_to_query(root, lfirst(l)); + } + else if (IsA(jtnode, JoinExpr)) + { + JoinExpr *j = (JoinExpr *) jtnode; + + add_base_rels_to_query(root, j->larg); + add_base_rels_to_query(root, j->rarg); + } + else + elog(ERROR, "unrecognized node type: %d", + (int) nodeTag(jtnode)); +} + +/* + * add_other_rels_to_query + * create "otherrel" RelOptInfos for the children of appendrel baserels + * + * At the end of this process, there should be RelOptInfos for all relations + * that will be scanned by the query. + */ +void +add_other_rels_to_query(PlannerInfo *root) +{ + int rti; + + for (rti = 1; rti < root->simple_rel_array_size; rti++) + { + RelOptInfo *rel = root->simple_rel_array[rti]; + RangeTblEntry *rte = root->simple_rte_array[rti]; + + /* there may be empty slots corresponding to non-baserel RTEs */ + if (rel == NULL) + continue; + + /* Ignore any "otherrels" that were already added. */ + if (rel->reloptkind != RELOPT_BASEREL) + continue; + + /* If it's marked as inheritable, look for children. */ + if (rte->inh) + expand_inherited_rtentry(root, rel, rte, rti); + } +} + + +/***************************************************************************** + * + * TARGET LISTS + * + *****************************************************************************/ + +/* + * build_base_rel_tlists + * Add targetlist entries for each var needed in the query's final tlist + * (and HAVING clause, if any) to the appropriate base relations. + * + * We mark such vars as needed by "relation 0" to ensure that they will + * propagate up through all join plan steps. + */ +void +build_base_rel_tlists(PlannerInfo *root, List *final_tlist) +{ + List *tlist_vars = pull_var_clause((Node *) final_tlist, + PVC_RECURSE_AGGREGATES | + PVC_RECURSE_WINDOWFUNCS | + PVC_INCLUDE_PLACEHOLDERS); + + if (tlist_vars != NIL) + { + add_vars_to_targetlist(root, tlist_vars, bms_make_singleton(0), true); + list_free(tlist_vars); + } + + /* + * If there's a HAVING clause, we'll need the Vars it uses, too. Note + * that HAVING can contain Aggrefs but not WindowFuncs. + */ + if (root->parse->havingQual) + { + List *having_vars = pull_var_clause(root->parse->havingQual, + PVC_RECURSE_AGGREGATES | + PVC_INCLUDE_PLACEHOLDERS); + + if (having_vars != NIL) + { + add_vars_to_targetlist(root, having_vars, + bms_make_singleton(0), true); + list_free(having_vars); + } + } +} + +/* + * add_vars_to_targetlist + * For each variable appearing in the list, add it to the owning + * relation's targetlist if not already present, and mark the variable + * as being needed for the indicated join (or for final output if + * where_needed includes "relation 0"). + * + * The list may also contain PlaceHolderVars. These don't necessarily + * have a single owning relation; we keep their attr_needed info in + * root->placeholder_list instead. If create_new_ph is true, it's OK + * to create new PlaceHolderInfos; otherwise, the PlaceHolderInfos must + * already exist, and we should only update their ph_needed. (This should + * be true before deconstruct_jointree begins, and false after that.) + */ +void +add_vars_to_targetlist(PlannerInfo *root, List *vars, + Relids where_needed, bool create_new_ph) +{ + ListCell *temp; + + Assert(!bms_is_empty(where_needed)); + + foreach(temp, vars) + { + Node *node = (Node *) lfirst(temp); + + if (IsA(node, Var)) + { + Var *var = (Var *) node; + RelOptInfo *rel = find_base_rel(root, var->varno); + int attno = var->varattno; + + if (bms_is_subset(where_needed, rel->relids)) + continue; + Assert(attno >= rel->min_attr && attno <= rel->max_attr); + attno -= rel->min_attr; + if (rel->attr_needed[attno] == NULL) + { + /* Variable not yet requested, so add to rel's targetlist */ + /* XXX is copyObject necessary here? */ + rel->reltarget->exprs = lappend(rel->reltarget->exprs, + copyObject(var)); + /* reltarget cost and width will be computed later */ + } + rel->attr_needed[attno] = bms_add_members(rel->attr_needed[attno], + where_needed); + } + else if (IsA(node, PlaceHolderVar)) + { + PlaceHolderVar *phv = (PlaceHolderVar *) node; + PlaceHolderInfo *phinfo = find_placeholder_info(root, phv, + create_new_ph); + + phinfo->ph_needed = bms_add_members(phinfo->ph_needed, + where_needed); + } + else + elog(ERROR, "unrecognized node type: %d", (int) nodeTag(node)); + } +} + + +/***************************************************************************** + * + * LATERAL REFERENCES + * + *****************************************************************************/ + +/* + * find_lateral_references + * For each LATERAL subquery, extract all its references to Vars and + * PlaceHolderVars of the current query level, and make sure those values + * will be available for evaluation of the subquery. + * + * While later planning steps ensure that the Var/PHV source rels are on the + * outside of nestloops relative to the LATERAL subquery, we also need to + * ensure that the Vars/PHVs propagate up to the nestloop join level; this + * means setting suitable where_needed values for them. + * + * Note that this only deals with lateral references in unflattened LATERAL + * subqueries. When we flatten a LATERAL subquery, its lateral references + * become plain Vars in the parent query, but they may have to be wrapped in + * PlaceHolderVars if they need to be forced NULL by outer joins that don't + * also null the LATERAL subquery. That's all handled elsewhere. + * + * This has to run before deconstruct_jointree, since it might result in + * creation of PlaceHolderInfos. + */ +void +find_lateral_references(PlannerInfo *root) +{ + Index rti; + + /* We need do nothing if the query contains no LATERAL RTEs */ + if (!root->hasLateralRTEs) + return; + + /* + * Examine all baserels (the rel array has been set up by now). + */ + for (rti = 1; rti < root->simple_rel_array_size; rti++) + { + RelOptInfo *brel = root->simple_rel_array[rti]; + + /* there may be empty slots corresponding to non-baserel RTEs */ + if (brel == NULL) + continue; + + Assert(brel->relid == rti); /* sanity check on array */ + + /* + * This bit is less obvious than it might look. We ignore appendrel + * otherrels and consider only their parent baserels. In a case where + * a LATERAL-containing UNION ALL subquery was pulled up, it is the + * otherrel that is actually going to be in the plan. However, we + * want to mark all its lateral references as needed by the parent, + * because it is the parent's relid that will be used for join + * planning purposes. And the parent's RTE will contain all the + * lateral references we need to know, since the pulled-up member is + * nothing but a copy of parts of the original RTE's subquery. We + * could visit the parent's children instead and transform their + * references back to the parent's relid, but it would be much more + * complicated for no real gain. (Important here is that the child + * members have not yet received any processing beyond being pulled + * up.) Similarly, in appendrels created by inheritance expansion, + * it's sufficient to look at the parent relation. + */ + + /* ignore RTEs that are "other rels" */ + if (brel->reloptkind != RELOPT_BASEREL) + continue; + + extract_lateral_references(root, brel, rti); + } +} + +static void +extract_lateral_references(PlannerInfo *root, RelOptInfo *brel, Index rtindex) +{ + RangeTblEntry *rte = root->simple_rte_array[rtindex]; + List *vars; + List *newvars; + Relids where_needed; + ListCell *lc; + + /* No cross-references are possible if it's not LATERAL */ + if (!rte->lateral) + return; + + /* Fetch the appropriate variables */ + if (rte->rtekind == RTE_RELATION) + vars = pull_vars_of_level((Node *) rte->tablesample, 0); + else if (rte->rtekind == RTE_SUBQUERY) + vars = pull_vars_of_level((Node *) rte->subquery, 1); + else if (rte->rtekind == RTE_FUNCTION) + vars = pull_vars_of_level((Node *) rte->functions, 0); + else if (rte->rtekind == RTE_TABLEFUNC) + vars = pull_vars_of_level((Node *) rte->tablefunc, 0); + else if (rte->rtekind == RTE_VALUES) + vars = pull_vars_of_level((Node *) rte->values_lists, 0); + else + { + Assert(false); + return; /* keep compiler quiet */ + } + + if (vars == NIL) + return; /* nothing to do */ + + /* Copy each Var (or PlaceHolderVar) and adjust it to match our level */ + newvars = NIL; + foreach(lc, vars) + { + Node *node = (Node *) lfirst(lc); + + node = copyObject(node); + if (IsA(node, Var)) + { + Var *var = (Var *) node; + + /* Adjustment is easy since it's just one node */ + var->varlevelsup = 0; + } + else if (IsA(node, PlaceHolderVar)) + { + PlaceHolderVar *phv = (PlaceHolderVar *) node; + int levelsup = phv->phlevelsup; + + /* Have to work harder to adjust the contained expression too */ + if (levelsup != 0) + IncrementVarSublevelsUp(node, -levelsup, 0); + + /* + * If we pulled the PHV out of a subquery RTE, its expression + * needs to be preprocessed. subquery_planner() already did this + * for level-zero PHVs in function and values RTEs, though. + */ + if (levelsup > 0) + phv->phexpr = preprocess_phv_expression(root, phv->phexpr); + } + else + Assert(false); + newvars = lappend(newvars, node); + } + + list_free(vars); + + /* + * We mark the Vars as being "needed" at the LATERAL RTE. This is a bit + * of a cheat: a more formal approach would be to mark each one as needed + * at the join of the LATERAL RTE with its source RTE. But it will work, + * and it's much less tedious than computing a separate where_needed for + * each Var. + */ + where_needed = bms_make_singleton(rtindex); + + /* + * Push Vars into their source relations' targetlists, and PHVs into + * root->placeholder_list. + */ + add_vars_to_targetlist(root, newvars, where_needed, true); + + /* Remember the lateral references for create_lateral_join_info */ + brel->lateral_vars = newvars; +} + +/* + * create_lateral_join_info + * Fill in the per-base-relation direct_lateral_relids, lateral_relids + * and lateral_referencers sets. + * + * This has to run after deconstruct_jointree, because we need to know the + * final ph_eval_at values for PlaceHolderVars. + */ +void +create_lateral_join_info(PlannerInfo *root) +{ + bool found_laterals = false; + Index rti; + ListCell *lc; + + /* We need do nothing if the query contains no LATERAL RTEs */ + if (!root->hasLateralRTEs) + return; + + /* + * Examine all baserels (the rel array has been set up by now). + */ + for (rti = 1; rti < root->simple_rel_array_size; rti++) + { + RelOptInfo *brel = root->simple_rel_array[rti]; + Relids lateral_relids; + + /* there may be empty slots corresponding to non-baserel RTEs */ + if (brel == NULL) + continue; + + Assert(brel->relid == rti); /* sanity check on array */ + + /* ignore RTEs that are "other rels" */ + if (brel->reloptkind != RELOPT_BASEREL) + continue; + + lateral_relids = NULL; + + /* consider each laterally-referenced Var or PHV */ + foreach(lc, brel->lateral_vars) + { + Node *node = (Node *) lfirst(lc); + + if (IsA(node, Var)) + { + Var *var = (Var *) node; + + found_laterals = true; + lateral_relids = bms_add_member(lateral_relids, + var->varno); + } + else if (IsA(node, PlaceHolderVar)) + { + PlaceHolderVar *phv = (PlaceHolderVar *) node; + PlaceHolderInfo *phinfo = find_placeholder_info(root, phv, + false); + + found_laterals = true; + lateral_relids = bms_add_members(lateral_relids, + phinfo->ph_eval_at); + } + else + Assert(false); + } + + /* We now have all the simple lateral refs from this rel */ + brel->direct_lateral_relids = lateral_relids; + brel->lateral_relids = bms_copy(lateral_relids); + } + + /* + * Now check for lateral references within PlaceHolderVars, and mark their + * eval_at rels as having lateral references to the source rels. + * + * For a PHV that is due to be evaluated at a baserel, mark its source(s) + * as direct lateral dependencies of the baserel (adding onto the ones + * recorded above). If it's due to be evaluated at a join, mark its + * source(s) as indirect lateral dependencies of each baserel in the join, + * ie put them into lateral_relids but not direct_lateral_relids. This is + * appropriate because we can't put any such baserel on the outside of a + * join to one of the PHV's lateral dependencies, but on the other hand we + * also can't yet join it directly to the dependency. + */ + foreach(lc, root->placeholder_list) + { + PlaceHolderInfo *phinfo = (PlaceHolderInfo *) lfirst(lc); + Relids eval_at = phinfo->ph_eval_at; + int varno; + + if (phinfo->ph_lateral == NULL) + continue; /* PHV is uninteresting if no lateral refs */ + + found_laterals = true; + + if (bms_get_singleton_member(eval_at, &varno)) + { + /* Evaluation site is a baserel */ + RelOptInfo *brel = find_base_rel(root, varno); + + brel->direct_lateral_relids = + bms_add_members(brel->direct_lateral_relids, + phinfo->ph_lateral); + brel->lateral_relids = + bms_add_members(brel->lateral_relids, + phinfo->ph_lateral); + } + else + { + /* Evaluation site is a join */ + varno = -1; + while ((varno = bms_next_member(eval_at, varno)) >= 0) + { + RelOptInfo *brel = find_base_rel(root, varno); + + brel->lateral_relids = bms_add_members(brel->lateral_relids, + phinfo->ph_lateral); + } + } + } + + /* + * If we found no actual lateral references, we're done; but reset the + * hasLateralRTEs flag to avoid useless work later. + */ + if (!found_laterals) + { + root->hasLateralRTEs = false; + return; + } + + /* + * Calculate the transitive closure of the lateral_relids sets, so that + * they describe both direct and indirect lateral references. If relation + * X references Y laterally, and Y references Z laterally, then we will + * have to scan X on the inside of a nestloop with Z, so for all intents + * and purposes X is laterally dependent on Z too. + * + * This code is essentially Warshall's algorithm for transitive closure. + * The outer loop considers each baserel, and propagates its lateral + * dependencies to those baserels that have a lateral dependency on it. + */ + for (rti = 1; rti < root->simple_rel_array_size; rti++) + { + RelOptInfo *brel = root->simple_rel_array[rti]; + Relids outer_lateral_relids; + Index rti2; + + if (brel == NULL || brel->reloptkind != RELOPT_BASEREL) + continue; + + /* need not consider baserel further if it has no lateral refs */ + outer_lateral_relids = brel->lateral_relids; + if (outer_lateral_relids == NULL) + continue; + + /* else scan all baserels */ + for (rti2 = 1; rti2 < root->simple_rel_array_size; rti2++) + { + RelOptInfo *brel2 = root->simple_rel_array[rti2]; + + if (brel2 == NULL || brel2->reloptkind != RELOPT_BASEREL) + continue; + + /* if brel2 has lateral ref to brel, propagate brel's refs */ + if (bms_is_member(rti, brel2->lateral_relids)) + brel2->lateral_relids = bms_add_members(brel2->lateral_relids, + outer_lateral_relids); + } + } + + /* + * Now that we've identified all lateral references, mark each baserel + * with the set of relids of rels that reference it laterally (possibly + * indirectly) --- that is, the inverse mapping of lateral_relids. + */ + for (rti = 1; rti < root->simple_rel_array_size; rti++) + { + RelOptInfo *brel = root->simple_rel_array[rti]; + Relids lateral_relids; + int rti2; + + if (brel == NULL || brel->reloptkind != RELOPT_BASEREL) + continue; + + /* Nothing to do at rels with no lateral refs */ + lateral_relids = brel->lateral_relids; + if (lateral_relids == NULL) + continue; + + /* + * We should not have broken the invariant that lateral_relids is + * exactly NULL if empty. + */ + Assert(!bms_is_empty(lateral_relids)); + + /* Also, no rel should have a lateral dependency on itself */ + Assert(!bms_is_member(rti, lateral_relids)); + + /* Mark this rel's referencees */ + rti2 = -1; + while ((rti2 = bms_next_member(lateral_relids, rti2)) >= 0) + { + RelOptInfo *brel2 = root->simple_rel_array[rti2]; + + Assert(brel2 != NULL && brel2->reloptkind == RELOPT_BASEREL); + brel2->lateral_referencers = + bms_add_member(brel2->lateral_referencers, rti); + } + } +} + + +/***************************************************************************** + * + * JOIN TREE PROCESSING + * + *****************************************************************************/ + +/* + * deconstruct_jointree + * Recursively scan the query's join tree for WHERE and JOIN/ON qual + * clauses, and add these to the appropriate restrictinfo and joininfo + * lists belonging to base RelOptInfos. Also, add SpecialJoinInfo nodes + * to root->join_info_list for any outer joins appearing in the query tree. + * Return a "joinlist" data structure showing the join order decisions + * that need to be made by make_one_rel(). + * + * The "joinlist" result is a list of items that are either RangeTblRef + * jointree nodes or sub-joinlists. All the items at the same level of + * joinlist must be joined in an order to be determined by make_one_rel() + * (note that legal orders may be constrained by SpecialJoinInfo nodes). + * A sub-joinlist represents a subproblem to be planned separately. Currently + * sub-joinlists arise only from FULL OUTER JOIN or when collapsing of + * subproblems is stopped by join_collapse_limit or from_collapse_limit. + * + * NOTE: when dealing with inner joins, it is appropriate to let a qual clause + * be evaluated at the lowest level where all the variables it mentions are + * available. However, we cannot push a qual down into the nullable side(s) + * of an outer join since the qual might eliminate matching rows and cause a + * NULL row to be incorrectly emitted by the join. Therefore, we artificially + * OR the minimum-relids of such an outer join into the required_relids of + * clauses appearing above it. This forces those clauses to be delayed until + * application of the outer join (or maybe even higher in the join tree). + */ +List * +deconstruct_jointree(PlannerInfo *root) +{ + List *result; + Relids qualscope; + Relids inner_join_rels; + List *postponed_qual_list = NIL; + + /* Start recursion at top of jointree */ + Assert(root->parse->jointree != NULL && + IsA(root->parse->jointree, FromExpr)); + + /* this is filled as we scan the jointree */ + root->nullable_baserels = NULL; + + result = deconstruct_recurse(root, (Node *) root->parse->jointree, false, + &qualscope, &inner_join_rels, + &postponed_qual_list); + + /* Shouldn't be any leftover quals */ + Assert(postponed_qual_list == NIL); + + return result; +} + +/* + * deconstruct_recurse + * One recursion level of deconstruct_jointree processing. + * + * Inputs: + * jtnode is the jointree node to examine + * below_outer_join is true if this node is within the nullable side of a + * higher-level outer join + * Outputs: + * *qualscope gets the set of base Relids syntactically included in this + * jointree node (do not modify or free this, as it may also be pointed + * to by RestrictInfo and SpecialJoinInfo nodes) + * *inner_join_rels gets the set of base Relids syntactically included in + * inner joins appearing at or below this jointree node (do not modify + * or free this, either) + * *postponed_qual_list is a list of PostponedQual structs, which we can + * add quals to if they turn out to belong to a higher join level + * Return value is the appropriate joinlist for this jointree node + * + * In addition, entries will be added to root->join_info_list for outer joins. + */ +static List * +deconstruct_recurse(PlannerInfo *root, Node *jtnode, bool below_outer_join, + Relids *qualscope, Relids *inner_join_rels, + List **postponed_qual_list) +{ + List *joinlist; + + if (jtnode == NULL) + { + *qualscope = NULL; + *inner_join_rels = NULL; + return NIL; + } + if (IsA(jtnode, RangeTblRef)) + { + int varno = ((RangeTblRef *) jtnode)->rtindex; + + /* qualscope is just the one RTE */ + *qualscope = bms_make_singleton(varno); + /* Deal with any securityQuals attached to the RTE */ + if (root->qual_security_level > 0) + process_security_barrier_quals(root, + varno, + *qualscope, + below_outer_join); + /* A single baserel does not create an inner join */ + *inner_join_rels = NULL; + joinlist = list_make1(jtnode); + } + else if (IsA(jtnode, FromExpr)) + { + FromExpr *f = (FromExpr *) jtnode; + List *child_postponed_quals = NIL; + int remaining; + ListCell *l; + + /* + * First, recurse to handle child joins. We collapse subproblems into + * a single joinlist whenever the resulting joinlist wouldn't exceed + * from_collapse_limit members. Also, always collapse one-element + * subproblems, since that won't lengthen the joinlist anyway. + */ + *qualscope = NULL; + *inner_join_rels = NULL; + joinlist = NIL; + remaining = list_length(f->fromlist); + foreach(l, f->fromlist) + { + Relids sub_qualscope; + List *sub_joinlist; + int sub_members; + + sub_joinlist = deconstruct_recurse(root, lfirst(l), + below_outer_join, + &sub_qualscope, + inner_join_rels, + &child_postponed_quals); + *qualscope = bms_add_members(*qualscope, sub_qualscope); + sub_members = list_length(sub_joinlist); + remaining--; + if (sub_members <= 1 || + list_length(joinlist) + sub_members + remaining <= from_collapse_limit) + joinlist = list_concat(joinlist, sub_joinlist); + else + joinlist = lappend(joinlist, sub_joinlist); + } + + /* + * A FROM with more than one list element is an inner join subsuming + * all below it, so we should report inner_join_rels = qualscope. If + * there was exactly one element, we should (and already did) report + * whatever its inner_join_rels were. If there were no elements (is + * that still possible?) the initialization before the loop fixed it. + */ + if (list_length(f->fromlist) > 1) + *inner_join_rels = *qualscope; + + /* + * Try to process any quals postponed by children. If they need + * further postponement, add them to my output postponed_qual_list. + */ + foreach(l, child_postponed_quals) + { + PostponedQual *pq = (PostponedQual *) lfirst(l); + + if (bms_is_subset(pq->relids, *qualscope)) + distribute_qual_to_rels(root, pq->qual, + below_outer_join, JOIN_INNER, + root->qual_security_level, + *qualscope, NULL, NULL, + NULL); + else + *postponed_qual_list = lappend(*postponed_qual_list, pq); + } + + /* + * Now process the top-level quals. + */ + foreach(l, (List *) f->quals) + { + Node *qual = (Node *) lfirst(l); + + distribute_qual_to_rels(root, qual, + below_outer_join, JOIN_INNER, + root->qual_security_level, + *qualscope, NULL, NULL, + postponed_qual_list); + } + } + else if (IsA(jtnode, JoinExpr)) + { + JoinExpr *j = (JoinExpr *) jtnode; + List *child_postponed_quals = NIL; + Relids leftids, + rightids, + left_inners, + right_inners, + nonnullable_rels, + nullable_rels, + ojscope; + List *leftjoinlist, + *rightjoinlist; + List *my_quals; + SpecialJoinInfo *sjinfo; + ListCell *l; + + /* + * Order of operations here is subtle and critical. First we recurse + * to handle sub-JOINs. Their join quals will be placed without + * regard for whether this level is an outer join, which is correct. + * Then we place our own join quals, which are restricted by lower + * outer joins in any case, and are forced to this level if this is an + * outer join and they mention the outer side. Finally, if this is an + * outer join, we create a join_info_list entry for the join. This + * will prevent quals above us in the join tree that use those rels + * from being pushed down below this level. (It's okay for upper + * quals to be pushed down to the outer side, however.) + */ + switch (j->jointype) + { + case JOIN_INNER: + leftjoinlist = deconstruct_recurse(root, j->larg, + below_outer_join, + &leftids, &left_inners, + &child_postponed_quals); + rightjoinlist = deconstruct_recurse(root, j->rarg, + below_outer_join, + &rightids, &right_inners, + &child_postponed_quals); + *qualscope = bms_union(leftids, rightids); + *inner_join_rels = *qualscope; + /* Inner join adds no restrictions for quals */ + nonnullable_rels = NULL; + /* and it doesn't force anything to null, either */ + nullable_rels = NULL; + break; + case JOIN_LEFT: + case JOIN_ANTI: + leftjoinlist = deconstruct_recurse(root, j->larg, + below_outer_join, + &leftids, &left_inners, + &child_postponed_quals); + rightjoinlist = deconstruct_recurse(root, j->rarg, + true, + &rightids, &right_inners, + &child_postponed_quals); + *qualscope = bms_union(leftids, rightids); + *inner_join_rels = bms_union(left_inners, right_inners); + nonnullable_rels = leftids; + nullable_rels = rightids; + break; + case JOIN_SEMI: + leftjoinlist = deconstruct_recurse(root, j->larg, + below_outer_join, + &leftids, &left_inners, + &child_postponed_quals); + rightjoinlist = deconstruct_recurse(root, j->rarg, + below_outer_join, + &rightids, &right_inners, + &child_postponed_quals); + *qualscope = bms_union(leftids, rightids); + *inner_join_rels = bms_union(left_inners, right_inners); + /* Semi join adds no restrictions for quals */ + nonnullable_rels = NULL; + + /* + * Theoretically, a semijoin would null the RHS; but since the + * RHS can't be accessed above the join, this is immaterial + * and we needn't account for it. + */ + nullable_rels = NULL; + break; + case JOIN_FULL: + leftjoinlist = deconstruct_recurse(root, j->larg, + true, + &leftids, &left_inners, + &child_postponed_quals); + rightjoinlist = deconstruct_recurse(root, j->rarg, + true, + &rightids, &right_inners, + &child_postponed_quals); + *qualscope = bms_union(leftids, rightids); + *inner_join_rels = bms_union(left_inners, right_inners); + /* each side is both outer and inner */ + nonnullable_rels = *qualscope; + nullable_rels = *qualscope; + break; + default: + /* JOIN_RIGHT was eliminated during reduce_outer_joins() */ + elog(ERROR, "unrecognized join type: %d", + (int) j->jointype); + nonnullable_rels = NULL; /* keep compiler quiet */ + nullable_rels = NULL; + leftjoinlist = rightjoinlist = NIL; + break; + } + + /* Report all rels that will be nulled anywhere in the jointree */ + root->nullable_baserels = bms_add_members(root->nullable_baserels, + nullable_rels); + + /* + * Try to process any quals postponed by children. If they need + * further postponement, add them to my output postponed_qual_list. + * Quals that can be processed now must be included in my_quals, so + * that they'll be handled properly in make_outerjoininfo. + */ + my_quals = NIL; + foreach(l, child_postponed_quals) + { + PostponedQual *pq = (PostponedQual *) lfirst(l); + + if (bms_is_subset(pq->relids, *qualscope)) + my_quals = lappend(my_quals, pq->qual); + else + { + /* + * We should not be postponing any quals past an outer join. + * If this Assert fires, pull_up_subqueries() messed up. + */ + Assert(j->jointype == JOIN_INNER); + *postponed_qual_list = lappend(*postponed_qual_list, pq); + } + } + my_quals = list_concat(my_quals, (List *) j->quals); + + /* + * For an OJ, form the SpecialJoinInfo now, because we need the OJ's + * semantic scope (ojscope) to pass to distribute_qual_to_rels. But + * we mustn't add it to join_info_list just yet, because we don't want + * distribute_qual_to_rels to think it is an outer join below us. + * + * Semijoins are a bit of a hybrid: we build a SpecialJoinInfo, but we + * want ojscope = NULL for distribute_qual_to_rels. + */ + if (j->jointype != JOIN_INNER) + { + sjinfo = make_outerjoininfo(root, + leftids, rightids, + *inner_join_rels, + j->jointype, + my_quals); + if (j->jointype == JOIN_SEMI) + ojscope = NULL; + else + ojscope = bms_union(sjinfo->min_lefthand, + sjinfo->min_righthand); + } + else + { + sjinfo = NULL; + ojscope = NULL; + } + + /* Process the JOIN's qual clauses */ + foreach(l, my_quals) + { + Node *qual = (Node *) lfirst(l); + + distribute_qual_to_rels(root, qual, + below_outer_join, j->jointype, + root->qual_security_level, + *qualscope, + ojscope, nonnullable_rels, + postponed_qual_list); + } + + /* Now we can add the SpecialJoinInfo to join_info_list */ + if (sjinfo) + { + root->join_info_list = lappend(root->join_info_list, sjinfo); + /* Each time we do that, recheck placeholder eval levels */ + update_placeholder_eval_levels(root, sjinfo); + } + + /* + * Finally, compute the output joinlist. We fold subproblems together + * except at a FULL JOIN or where join_collapse_limit would be + * exceeded. + */ + if (j->jointype == JOIN_FULL) + { + /* force the join order exactly at this node */ + joinlist = list_make1(list_make2(leftjoinlist, rightjoinlist)); + } + else if (list_length(leftjoinlist) + list_length(rightjoinlist) <= + join_collapse_limit) + { + /* OK to combine subproblems */ + joinlist = list_concat(leftjoinlist, rightjoinlist); + } + else + { + /* can't combine, but needn't force join order above here */ + Node *leftpart, + *rightpart; + + /* avoid creating useless 1-element sublists */ + if (list_length(leftjoinlist) == 1) + leftpart = (Node *) linitial(leftjoinlist); + else + leftpart = (Node *) leftjoinlist; + if (list_length(rightjoinlist) == 1) + rightpart = (Node *) linitial(rightjoinlist); + else + rightpart = (Node *) rightjoinlist; + joinlist = list_make2(leftpart, rightpart); + } + } + else + { + elog(ERROR, "unrecognized node type: %d", + (int) nodeTag(jtnode)); + joinlist = NIL; /* keep compiler quiet */ + } + return joinlist; +} + +/* + * process_security_barrier_quals + * Transfer security-barrier quals into relation's baserestrictinfo list. + * + * The rewriter put any relevant security-barrier conditions into the RTE's + * securityQuals field, but it's now time to copy them into the rel's + * baserestrictinfo. + * + * In inheritance cases, we only consider quals attached to the parent rel + * here; they will be valid for all children too, so it's okay to consider + * them for purposes like equivalence class creation. Quals attached to + * individual child rels will be dealt with during path creation. + */ +static void +process_security_barrier_quals(PlannerInfo *root, + int rti, Relids qualscope, + bool below_outer_join) +{ + RangeTblEntry *rte = root->simple_rte_array[rti]; + Index security_level = 0; + ListCell *lc; + + /* + * Each element of the securityQuals list has been preprocessed into an + * implicitly-ANDed list of clauses. All the clauses in a given sublist + * should get the same security level, but successive sublists get higher + * levels. + */ + foreach(lc, rte->securityQuals) + { + List *qualset = (List *) lfirst(lc); + ListCell *lc2; + + foreach(lc2, qualset) + { + Node *qual = (Node *) lfirst(lc2); + + /* + * We cheat to the extent of passing ojscope = qualscope rather + * than its more logical value of NULL. The only effect this has + * is to force a Var-free qual to be evaluated at the rel rather + * than being pushed up to top of tree, which we don't want. + */ + distribute_qual_to_rels(root, qual, + below_outer_join, + JOIN_INNER, + security_level, + qualscope, + qualscope, + NULL, + NULL); + } + security_level++; + } + + /* Assert that qual_security_level is higher than anything we just used */ + Assert(security_level <= root->qual_security_level); +} + +/* + * make_outerjoininfo + * Build a SpecialJoinInfo for the current outer join + * + * Inputs: + * left_rels: the base Relids syntactically on outer side of join + * right_rels: the base Relids syntactically on inner side of join + * inner_join_rels: base Relids participating in inner joins below this one + * jointype: what it says (must always be LEFT, FULL, SEMI, or ANTI) + * clause: the outer join's join condition (in implicit-AND format) + * + * The node should eventually be appended to root->join_info_list, but we + * do not do that here. + * + * Note: we assume that this function is invoked bottom-up, so that + * root->join_info_list already contains entries for all outer joins that are + * syntactically below this one. + */ +static SpecialJoinInfo * +make_outerjoininfo(PlannerInfo *root, + Relids left_rels, Relids right_rels, + Relids inner_join_rels, + JoinType jointype, List *clause) +{ + SpecialJoinInfo *sjinfo = makeNode(SpecialJoinInfo); + Relids clause_relids; + Relids strict_relids; + Relids min_lefthand; + Relids min_righthand; + ListCell *l; + + /* + * We should not see RIGHT JOIN here because left/right were switched + * earlier + */ + Assert(jointype != JOIN_INNER); + Assert(jointype != JOIN_RIGHT); + + /* + * Presently the executor cannot support FOR [KEY] UPDATE/SHARE marking of + * rels appearing on the nullable side of an outer join. (It's somewhat + * unclear what that would mean, anyway: what should we mark when a result + * row is generated from no element of the nullable relation?) So, + * complain if any nullable rel is FOR [KEY] UPDATE/SHARE. + * + * You might be wondering why this test isn't made far upstream in the + * parser. It's because the parser hasn't got enough info --- consider + * FOR UPDATE applied to a view. Only after rewriting and flattening do + * we know whether the view contains an outer join. + * + * We use the original RowMarkClause list here; the PlanRowMark list would + * list everything. + */ + foreach(l, root->parse->rowMarks) + { + RowMarkClause *rc = (RowMarkClause *) lfirst(l); + + if (bms_is_member(rc->rti, right_rels) || + (jointype == JOIN_FULL && bms_is_member(rc->rti, left_rels))) + ereport(ERROR, + (errcode(ERRCODE_FEATURE_NOT_SUPPORTED), + /*------ + translator: %s is a SQL row locking clause such as FOR UPDATE */ + errmsg("%s cannot be applied to the nullable side of an outer join", + LCS_asString(rc->strength)))); + } + + sjinfo->syn_lefthand = left_rels; + sjinfo->syn_righthand = right_rels; + sjinfo->jointype = jointype; + /* this always starts out false */ + sjinfo->delay_upper_joins = false; + + compute_semijoin_info(root, sjinfo, clause); + + /* If it's a full join, no need to be very smart */ + if (jointype == JOIN_FULL) + { + sjinfo->min_lefthand = bms_copy(left_rels); + sjinfo->min_righthand = bms_copy(right_rels); + sjinfo->lhs_strict = false; /* don't care about this */ + return sjinfo; + } + + /* + * Retrieve all relids mentioned within the join clause. + */ + clause_relids = pull_varnos(root, (Node *) clause); + + /* + * For which relids is the clause strict, ie, it cannot succeed if the + * rel's columns are all NULL? + */ + strict_relids = find_nonnullable_rels((Node *) clause); + + /* Remember whether the clause is strict for any LHS relations */ + sjinfo->lhs_strict = bms_overlap(strict_relids, left_rels); + + /* + * Required LHS always includes the LHS rels mentioned in the clause. We + * may have to add more rels based on lower outer joins; see below. + */ + min_lefthand = bms_intersect(clause_relids, left_rels); + + /* + * Similarly for required RHS. But here, we must also include any lower + * inner joins, to ensure we don't try to commute with any of them. + */ + min_righthand = bms_int_members(bms_union(clause_relids, inner_join_rels), + right_rels); + + /* + * Now check previous outer joins for ordering restrictions. + */ + foreach(l, root->join_info_list) + { + SpecialJoinInfo *otherinfo = (SpecialJoinInfo *) lfirst(l); + + /* + * A full join is an optimization barrier: we can't associate into or + * out of it. Hence, if it overlaps either LHS or RHS of the current + * rel, expand that side's min relset to cover the whole full join. + */ + if (otherinfo->jointype == JOIN_FULL) + { + if (bms_overlap(left_rels, otherinfo->syn_lefthand) || + bms_overlap(left_rels, otherinfo->syn_righthand)) + { + min_lefthand = bms_add_members(min_lefthand, + otherinfo->syn_lefthand); + min_lefthand = bms_add_members(min_lefthand, + otherinfo->syn_righthand); + } + if (bms_overlap(right_rels, otherinfo->syn_lefthand) || + bms_overlap(right_rels, otherinfo->syn_righthand)) + { + min_righthand = bms_add_members(min_righthand, + otherinfo->syn_lefthand); + min_righthand = bms_add_members(min_righthand, + otherinfo->syn_righthand); + } + /* Needn't do anything else with the full join */ + continue; + } + + /* + * For a lower OJ in our LHS, if our join condition uses the lower + * join's RHS and is not strict for that rel, we must preserve the + * ordering of the two OJs, so add lower OJ's full syntactic relset to + * min_lefthand. (We must use its full syntactic relset, not just its + * min_lefthand + min_righthand. This is because there might be other + * OJs below this one that this one can commute with, but we cannot + * commute with them if we don't with this one.) Also, if the current + * join is a semijoin or antijoin, we must preserve ordering + * regardless of strictness. + * + * Note: I believe we have to insist on being strict for at least one + * rel in the lower OJ's min_righthand, not its whole syn_righthand. + */ + if (bms_overlap(left_rels, otherinfo->syn_righthand)) + { + if (bms_overlap(clause_relids, otherinfo->syn_righthand) && + (jointype == JOIN_SEMI || jointype == JOIN_ANTI || + !bms_overlap(strict_relids, otherinfo->min_righthand))) + { + min_lefthand = bms_add_members(min_lefthand, + otherinfo->syn_lefthand); + min_lefthand = bms_add_members(min_lefthand, + otherinfo->syn_righthand); + } + } + + /* + * For a lower OJ in our RHS, if our join condition does not use the + * lower join's RHS and the lower OJ's join condition is strict, we + * can interchange the ordering of the two OJs; otherwise we must add + * the lower OJ's full syntactic relset to min_righthand. + * + * Also, if our join condition does not use the lower join's LHS + * either, force the ordering to be preserved. Otherwise we can end + * up with SpecialJoinInfos with identical min_righthands, which can + * confuse join_is_legal (see discussion in backend/optimizer/README). + * + * Also, we must preserve ordering anyway if either the current join + * or the lower OJ is either a semijoin or an antijoin. + * + * Here, we have to consider that "our join condition" includes any + * clauses that syntactically appeared above the lower OJ and below + * ours; those are equivalent to degenerate clauses in our OJ and must + * be treated as such. Such clauses obviously can't reference our + * LHS, and they must be non-strict for the lower OJ's RHS (else + * reduce_outer_joins would have reduced the lower OJ to a plain + * join). Hence the other ways in which we handle clauses within our + * join condition are not affected by them. The net effect is + * therefore sufficiently represented by the delay_upper_joins flag + * saved for us by check_outerjoin_delay. + */ + if (bms_overlap(right_rels, otherinfo->syn_righthand)) + { + if (bms_overlap(clause_relids, otherinfo->syn_righthand) || + !bms_overlap(clause_relids, otherinfo->min_lefthand) || + jointype == JOIN_SEMI || + jointype == JOIN_ANTI || + otherinfo->jointype == JOIN_SEMI || + otherinfo->jointype == JOIN_ANTI || + !otherinfo->lhs_strict || otherinfo->delay_upper_joins) + { + min_righthand = bms_add_members(min_righthand, + otherinfo->syn_lefthand); + min_righthand = bms_add_members(min_righthand, + otherinfo->syn_righthand); + } + } + } + + /* + * Examine PlaceHolderVars. If a PHV is supposed to be evaluated within + * this join's nullable side, then ensure that min_righthand contains the + * full eval_at set of the PHV. This ensures that the PHV actually can be + * evaluated within the RHS. Note that this works only because we should + * already have determined the final eval_at level for any PHV + * syntactically within this join. + */ + foreach(l, root->placeholder_list) + { + PlaceHolderInfo *phinfo = (PlaceHolderInfo *) lfirst(l); + Relids ph_syn_level = phinfo->ph_var->phrels; + + /* Ignore placeholder if it didn't syntactically come from RHS */ + if (!bms_is_subset(ph_syn_level, right_rels)) + continue; + + /* Else, prevent join from being formed before we eval the PHV */ + min_righthand = bms_add_members(min_righthand, phinfo->ph_eval_at); + } + + /* + * If we found nothing to put in min_lefthand, punt and make it the full + * LHS, to avoid having an empty min_lefthand which will confuse later + * processing. (We don't try to be smart about such cases, just correct.) + * Likewise for min_righthand. + */ + if (bms_is_empty(min_lefthand)) + min_lefthand = bms_copy(left_rels); + if (bms_is_empty(min_righthand)) + min_righthand = bms_copy(right_rels); + + /* Now they'd better be nonempty */ + Assert(!bms_is_empty(min_lefthand)); + Assert(!bms_is_empty(min_righthand)); + /* Shouldn't overlap either */ + Assert(!bms_overlap(min_lefthand, min_righthand)); + + sjinfo->min_lefthand = min_lefthand; + sjinfo->min_righthand = min_righthand; + + return sjinfo; +} + +/* + * compute_semijoin_info + * Fill semijoin-related fields of a new SpecialJoinInfo + * + * Note: this relies on only the jointype and syn_righthand fields of the + * SpecialJoinInfo; the rest may not be set yet. + */ +static void +compute_semijoin_info(PlannerInfo *root, SpecialJoinInfo *sjinfo, List *clause) +{ + List *semi_operators; + List *semi_rhs_exprs; + bool all_btree; + bool all_hash; + ListCell *lc; + + /* Initialize semijoin-related fields in case we can't unique-ify */ + sjinfo->semi_can_btree = false; + sjinfo->semi_can_hash = false; + sjinfo->semi_operators = NIL; + sjinfo->semi_rhs_exprs = NIL; + + /* Nothing more to do if it's not a semijoin */ + if (sjinfo->jointype != JOIN_SEMI) + return; + + /* + * Look to see whether the semijoin's join quals consist of AND'ed + * equality operators, with (only) RHS variables on only one side of each + * one. If so, we can figure out how to enforce uniqueness for the RHS. + * + * Note that the input clause list is the list of quals that are + * *syntactically* associated with the semijoin, which in practice means + * the synthesized comparison list for an IN or the WHERE of an EXISTS. + * Particularly in the latter case, it might contain clauses that aren't + * *semantically* associated with the join, but refer to just one side or + * the other. We can ignore such clauses here, as they will just drop + * down to be processed within one side or the other. (It is okay to + * consider only the syntactically-associated clauses here because for a + * semijoin, no higher-level quals could refer to the RHS, and so there + * can be no other quals that are semantically associated with this join. + * We do things this way because it is useful to have the set of potential + * unique-ification expressions before we can extract the list of quals + * that are actually semantically associated with the particular join.) + * + * Note that the semi_operators list consists of the joinqual operators + * themselves (but commuted if needed to put the RHS value on the right). + * These could be cross-type operators, in which case the operator + * actually needed for uniqueness is a related single-type operator. We + * assume here that that operator will be available from the btree or hash + * opclass when the time comes ... if not, create_unique_plan() will fail. + */ + semi_operators = NIL; + semi_rhs_exprs = NIL; + all_btree = true; + all_hash = enable_hashagg; /* don't consider hash if not enabled */ + foreach(lc, clause) + { + OpExpr *op = (OpExpr *) lfirst(lc); + Oid opno; + Node *left_expr; + Node *right_expr; + Relids left_varnos; + Relids right_varnos; + Relids all_varnos; + Oid opinputtype; + + /* Is it a binary opclause? */ + if (!IsA(op, OpExpr) || + list_length(op->args) != 2) + { + /* No, but does it reference both sides? */ + all_varnos = pull_varnos(root, (Node *) op); + if (!bms_overlap(all_varnos, sjinfo->syn_righthand) || + bms_is_subset(all_varnos, sjinfo->syn_righthand)) + { + /* + * Clause refers to only one rel, so ignore it --- unless it + * contains volatile functions, in which case we'd better + * punt. + */ + if (contain_volatile_functions((Node *) op)) + return; + continue; + } + /* Non-operator clause referencing both sides, must punt */ + return; + } + + /* Extract data from binary opclause */ + opno = op->opno; + left_expr = linitial(op->args); + right_expr = lsecond(op->args); + left_varnos = pull_varnos(root, left_expr); + right_varnos = pull_varnos(root, right_expr); + all_varnos = bms_union(left_varnos, right_varnos); + opinputtype = exprType(left_expr); + + /* Does it reference both sides? */ + if (!bms_overlap(all_varnos, sjinfo->syn_righthand) || + bms_is_subset(all_varnos, sjinfo->syn_righthand)) + { + /* + * Clause refers to only one rel, so ignore it --- unless it + * contains volatile functions, in which case we'd better punt. + */ + if (contain_volatile_functions((Node *) op)) + return; + continue; + } + + /* check rel membership of arguments */ + if (!bms_is_empty(right_varnos) && + bms_is_subset(right_varnos, sjinfo->syn_righthand) && + !bms_overlap(left_varnos, sjinfo->syn_righthand)) + { + /* typical case, right_expr is RHS variable */ + } + else if (!bms_is_empty(left_varnos) && + bms_is_subset(left_varnos, sjinfo->syn_righthand) && + !bms_overlap(right_varnos, sjinfo->syn_righthand)) + { + /* flipped case, left_expr is RHS variable */ + opno = get_commutator(opno); + if (!OidIsValid(opno)) + return; + right_expr = left_expr; + } + else + { + /* mixed membership of args, punt */ + return; + } + + /* all operators must be btree equality or hash equality */ + if (all_btree) + { + /* oprcanmerge is considered a hint... */ + if (!op_mergejoinable(opno, opinputtype) || + get_mergejoin_opfamilies(opno) == NIL) + all_btree = false; + } + if (all_hash) + { + /* ... but oprcanhash had better be correct */ + if (!op_hashjoinable(opno, opinputtype)) + all_hash = false; + } + if (!(all_btree || all_hash)) + return; + + /* so far so good, keep building lists */ + semi_operators = lappend_oid(semi_operators, opno); + semi_rhs_exprs = lappend(semi_rhs_exprs, copyObject(right_expr)); + } + + /* Punt if we didn't find at least one column to unique-ify */ + if (semi_rhs_exprs == NIL) + return; + + /* + * The expressions we'd need to unique-ify mustn't be volatile. + */ + if (contain_volatile_functions((Node *) semi_rhs_exprs)) + return; + + /* + * If we get here, we can unique-ify the semijoin's RHS using at least one + * of sorting and hashing. Save the information about how to do that. + */ + sjinfo->semi_can_btree = all_btree; + sjinfo->semi_can_hash = all_hash; + sjinfo->semi_operators = semi_operators; + sjinfo->semi_rhs_exprs = semi_rhs_exprs; +} + + +/***************************************************************************** + * + * QUALIFICATIONS + * + *****************************************************************************/ + +/* + * distribute_qual_to_rels + * Add clause information to either the baserestrictinfo or joininfo list + * (depending on whether the clause is a join) of each base relation + * mentioned in the clause. A RestrictInfo node is created and added to + * the appropriate list for each rel. Alternatively, if the clause uses a + * mergejoinable operator and is not delayed by outer-join rules, enter + * the left- and right-side expressions into the query's list of + * EquivalenceClasses. Alternatively, if the clause needs to be treated + * as belonging to a higher join level, just add it to postponed_qual_list. + * + * 'clause': the qual clause to be distributed + * 'below_outer_join': true if the qual is from a JOIN/ON that is below the + * nullable side of a higher-level outer join + * 'jointype': type of join the qual is from (JOIN_INNER for a WHERE clause) + * 'security_level': security_level to assign to the qual + * 'qualscope': set of baserels the qual's syntactic scope covers + * 'ojscope': NULL if not an outer-join qual, else the minimum set of baserels + * needed to form this join + * 'outerjoin_nonnullable': NULL if not an outer-join qual, else the set of + * baserels appearing on the outer (nonnullable) side of the join + * (for FULL JOIN this includes both sides of the join, and must in fact + * equal qualscope) + * 'postponed_qual_list': list of PostponedQual structs, which we can add + * this qual to if it turns out to belong to a higher join level. + * Can be NULL if caller knows postponement is impossible. + * + * 'qualscope' identifies what level of JOIN the qual came from syntactically. + * 'ojscope' is needed if we decide to force the qual up to the outer-join + * level, which will be ojscope not necessarily qualscope. + * + * At the time this is called, root->join_info_list must contain entries for + * all and only those special joins that are syntactically below this qual. + */ +static void +distribute_qual_to_rels(PlannerInfo *root, Node *clause, + bool below_outer_join, + JoinType jointype, + Index security_level, + Relids qualscope, + Relids ojscope, + Relids outerjoin_nonnullable, + List **postponed_qual_list) +{ + Relids relids; + bool is_pushed_down; + bool outerjoin_delayed; + bool pseudoconstant = false; + bool maybe_equivalence; + bool maybe_outer_join; + Relids nullable_relids; + RestrictInfo *restrictinfo; + + /* + * Retrieve all relids mentioned within the clause. + */ + relids = pull_varnos(root, clause); + + /* + * In ordinary SQL, a WHERE or JOIN/ON clause can't reference any rels + * that aren't within its syntactic scope; however, if we pulled up a + * LATERAL subquery then we might find such references in quals that have + * been pulled up. We need to treat such quals as belonging to the join + * level that includes every rel they reference. Although we could make + * pull_up_subqueries() place such quals correctly to begin with, it's + * easier to handle it here. When we find a clause that contains Vars + * outside its syntactic scope, we add it to the postponed-quals list, and + * process it once we've recursed back up to the appropriate join level. + */ + if (!bms_is_subset(relids, qualscope)) + { + PostponedQual *pq = (PostponedQual *) palloc(sizeof(PostponedQual)); + + Assert(root->hasLateralRTEs); /* shouldn't happen otherwise */ + Assert(jointype == JOIN_INNER); /* mustn't postpone past outer join */ + pq->qual = clause; + pq->relids = relids; + *postponed_qual_list = lappend(*postponed_qual_list, pq); + return; + } + + /* + * If it's an outer-join clause, also check that relids is a subset of + * ojscope. (This should not fail if the syntactic scope check passed.) + */ + if (ojscope && !bms_is_subset(relids, ojscope)) + elog(ERROR, "JOIN qualification cannot refer to other relations"); + + /* + * If the clause is variable-free, our normal heuristic for pushing it + * down to just the mentioned rels doesn't work, because there are none. + * + * If the clause is an outer-join clause, we must force it to the OJ's + * semantic level to preserve semantics. + * + * Otherwise, when the clause contains volatile functions, we force it to + * be evaluated at its original syntactic level. This preserves the + * expected semantics. + * + * When the clause contains no volatile functions either, it is actually a + * pseudoconstant clause that will not change value during any one + * execution of the plan, and hence can be used as a one-time qual in a + * gating Result plan node. We put such a clause into the regular + * RestrictInfo lists for the moment, but eventually createplan.c will + * pull it out and make a gating Result node immediately above whatever + * plan node the pseudoconstant clause is assigned to. It's usually best + * to put a gating node as high in the plan tree as possible. If we are + * not below an outer join, we can actually push the pseudoconstant qual + * all the way to the top of the tree. If we are below an outer join, we + * leave the qual at its original syntactic level (we could push it up to + * just below the outer join, but that seems more complex than it's + * worth). + */ + if (bms_is_empty(relids)) + { + if (ojscope) + { + /* clause is attached to outer join, eval it there */ + relids = bms_copy(ojscope); + /* mustn't use as gating qual, so don't mark pseudoconstant */ + } + else + { + /* eval at original syntactic level */ + relids = bms_copy(qualscope); + if (!contain_volatile_functions(clause)) + { + /* mark as gating qual */ + pseudoconstant = true; + /* tell createplan.c to check for gating quals */ + root->hasPseudoConstantQuals = true; + /* if not below outer join, push it to top of tree */ + if (!below_outer_join) + { + relids = + get_relids_in_jointree((Node *) root->parse->jointree, + false); + qualscope = bms_copy(relids); + } + } + } + } + + /*---------- + * Check to see if clause application must be delayed by outer-join + * considerations. + * + * A word about is_pushed_down: we mark the qual as "pushed down" if + * it is (potentially) applicable at a level different from its original + * syntactic level. This flag is used to distinguish OUTER JOIN ON quals + * from other quals pushed down to the same joinrel. The rules are: + * WHERE quals and INNER JOIN quals: is_pushed_down = true. + * Non-degenerate OUTER JOIN quals: is_pushed_down = false. + * Degenerate OUTER JOIN quals: is_pushed_down = true. + * A "degenerate" OUTER JOIN qual is one that doesn't mention the + * non-nullable side, and hence can be pushed down into the nullable side + * without changing the join result. It is correct to treat it as a + * regular filter condition at the level where it is evaluated. + * + * Note: it is not immediately obvious that a simple boolean is enough + * for this: if for some reason we were to attach a degenerate qual to + * its original join level, it would need to be treated as an outer join + * qual there. However, this cannot happen, because all the rels the + * clause mentions must be in the outer join's min_righthand, therefore + * the join it needs must be formed before the outer join; and we always + * attach quals to the lowest level where they can be evaluated. But + * if we were ever to re-introduce a mechanism for delaying evaluation + * of "expensive" quals, this area would need work. + * + * Note: generally, use of is_pushed_down has to go through the macro + * RINFO_IS_PUSHED_DOWN, because that flag alone is not always sufficient + * to tell whether a clause must be treated as pushed-down in context. + * This seems like another reason why it should perhaps be rethought. + *---------- + */ + if (bms_overlap(relids, outerjoin_nonnullable)) + { + /* + * The qual is attached to an outer join and mentions (some of the) + * rels on the nonnullable side, so it's not degenerate. + * + * We can't use such a clause to deduce equivalence (the left and + * right sides might be unequal above the join because one of them has + * gone to NULL) ... but we might be able to use it for more limited + * deductions, if it is mergejoinable. So consider adding it to the + * lists of set-aside outer-join clauses. + */ + is_pushed_down = false; + maybe_equivalence = false; + maybe_outer_join = true; + + /* Check to see if must be delayed by lower outer join */ + outerjoin_delayed = check_outerjoin_delay(root, + &relids, + &nullable_relids, + false); + + /* + * Now force the qual to be evaluated exactly at the level of joining + * corresponding to the outer join. We cannot let it get pushed down + * into the nonnullable side, since then we'd produce no output rows, + * rather than the intended single null-extended row, for any + * nonnullable-side rows failing the qual. + * + * (Do this step after calling check_outerjoin_delay, because that + * trashes relids.) + */ + Assert(ojscope); + relids = ojscope; + Assert(!pseudoconstant); + } + else + { + /* + * Normal qual clause or degenerate outer-join clause. Either way, we + * can mark it as pushed-down. + */ + is_pushed_down = true; + + /* Check to see if must be delayed by lower outer join */ + outerjoin_delayed = check_outerjoin_delay(root, + &relids, + &nullable_relids, + true); + + if (outerjoin_delayed) + { + /* Should still be a subset of current scope ... */ + Assert(root->hasLateralRTEs || bms_is_subset(relids, qualscope)); + Assert(ojscope == NULL || bms_is_subset(relids, ojscope)); + + /* + * Because application of the qual will be delayed by outer join, + * we mustn't assume its vars are equal everywhere. + */ + maybe_equivalence = false; + + /* + * It's possible that this is an IS NULL clause that's redundant + * with a lower antijoin; if so we can just discard it. We need + * not test in any of the other cases, because this will only be + * possible for pushed-down, delayed clauses. + */ + if (check_redundant_nullability_qual(root, clause)) + return; + } + else + { + /* + * Qual is not delayed by any lower outer-join restriction, so we + * can consider feeding it to the equivalence machinery. However, + * if it's itself within an outer-join clause, treat it as though + * it appeared below that outer join (note that we can only get + * here when the clause references only nullable-side rels). + */ + maybe_equivalence = true; + if (outerjoin_nonnullable != NULL) + below_outer_join = true; + } + + /* + * Since it doesn't mention the LHS, it's certainly not useful as a + * set-aside OJ clause, even if it's in an OJ. + */ + maybe_outer_join = false; + } + + /* + * Build the RestrictInfo node itself. + */ + restrictinfo = make_restrictinfo(root, + (Expr *) clause, + is_pushed_down, + outerjoin_delayed, + pseudoconstant, + security_level, + relids, + outerjoin_nonnullable, + nullable_relids); + + /* + * If it's a join clause (either naturally, or because delayed by + * outer-join rules), add vars used in the clause to targetlists of their + * relations, so that they will be emitted by the plan nodes that scan + * those relations (else they won't be available at the join node!). + * + * Note: if the clause gets absorbed into an EquivalenceClass then this + * may be unnecessary, but for now we have to do it to cover the case + * where the EC becomes ec_broken and we end up reinserting the original + * clauses into the plan. + */ + if (bms_membership(relids) == BMS_MULTIPLE) + { + List *vars = pull_var_clause(clause, + PVC_RECURSE_AGGREGATES | + PVC_RECURSE_WINDOWFUNCS | + PVC_INCLUDE_PLACEHOLDERS); + + add_vars_to_targetlist(root, vars, relids, false); + list_free(vars); + } + + /* + * We check "mergejoinability" of every clause, not only join clauses, + * because we want to know about equivalences between vars of the same + * relation, or between vars and consts. + */ + check_mergejoinable(restrictinfo); + + /* + * If it is a true equivalence clause, send it to the EquivalenceClass + * machinery. We do *not* attach it directly to any restriction or join + * lists. The EC code will propagate it to the appropriate places later. + * + * If the clause has a mergejoinable operator and is not + * outerjoin-delayed, yet isn't an equivalence because it is an outer-join + * clause, the EC code may yet be able to do something with it. We add it + * to appropriate lists for further consideration later. Specifically: + * + * If it is a left or right outer-join qualification that relates the two + * sides of the outer join (no funny business like leftvar1 = leftvar2 + + * rightvar), we add it to root->left_join_clauses or + * root->right_join_clauses according to which side the nonnullable + * variable appears on. + * + * If it is a full outer-join qualification, we add it to + * root->full_join_clauses. (Ideally we'd discard cases that aren't + * leftvar = rightvar, as we do for left/right joins, but this routine + * doesn't have the info needed to do that; and the current usage of the + * full_join_clauses list doesn't require that, so it's not currently + * worth complicating this routine's API to make it possible.) + * + * If none of the above hold, pass it off to + * distribute_restrictinfo_to_rels(). + * + * In all cases, it's important to initialize the left_ec and right_ec + * fields of a mergejoinable clause, so that all possibly mergejoinable + * expressions have representations in EquivalenceClasses. If + * process_equivalence is successful, it will take care of that; + * otherwise, we have to call initialize_mergeclause_eclasses to do it. + */ + if (restrictinfo->mergeopfamilies) + { + if (maybe_equivalence) + { + if (check_equivalence_delay(root, restrictinfo) && + process_equivalence(root, &restrictinfo, below_outer_join)) + return; + /* EC rejected it, so set left_ec/right_ec the hard way ... */ + if (restrictinfo->mergeopfamilies) /* EC might have changed this */ + initialize_mergeclause_eclasses(root, restrictinfo); + /* ... and fall through to distribute_restrictinfo_to_rels */ + } + else if (maybe_outer_join && restrictinfo->can_join) + { + /* we need to set up left_ec/right_ec the hard way */ + initialize_mergeclause_eclasses(root, restrictinfo); + /* now see if it should go to any outer-join lists */ + if (bms_is_subset(restrictinfo->left_relids, + outerjoin_nonnullable) && + !bms_overlap(restrictinfo->right_relids, + outerjoin_nonnullable)) + { + /* we have outervar = innervar */ + root->left_join_clauses = lappend(root->left_join_clauses, + restrictinfo); + return; + } + if (bms_is_subset(restrictinfo->right_relids, + outerjoin_nonnullable) && + !bms_overlap(restrictinfo->left_relids, + outerjoin_nonnullable)) + { + /* we have innervar = outervar */ + root->right_join_clauses = lappend(root->right_join_clauses, + restrictinfo); + return; + } + if (jointype == JOIN_FULL) + { + /* FULL JOIN (above tests cannot match in this case) */ + root->full_join_clauses = lappend(root->full_join_clauses, + restrictinfo); + return; + } + /* nope, so fall through to distribute_restrictinfo_to_rels */ + } + else + { + /* we still need to set up left_ec/right_ec */ + initialize_mergeclause_eclasses(root, restrictinfo); + } + } + + /* No EC special case applies, so push it into the clause lists */ + distribute_restrictinfo_to_rels(root, restrictinfo); +} + +/* + * check_outerjoin_delay + * Detect whether a qual referencing the given relids must be delayed + * in application due to the presence of a lower outer join, and/or + * may force extra delay of higher-level outer joins. + * + * If the qual must be delayed, add relids to *relids_p to reflect the lowest + * safe level for evaluating the qual, and return true. Any extra delay for + * higher-level joins is reflected by setting delay_upper_joins to true in + * SpecialJoinInfo structs. We also compute nullable_relids, the set of + * referenced relids that are nullable by lower outer joins (note that this + * can be nonempty even for a non-delayed qual). + * + * For an is_pushed_down qual, we can evaluate the qual as soon as (1) we have + * all the rels it mentions, and (2) we are at or above any outer joins that + * can null any of these rels and are below the syntactic location of the + * given qual. We must enforce (2) because pushing down such a clause below + * the OJ might cause the OJ to emit null-extended rows that should not have + * been formed, or that should have been rejected by the clause. (This is + * only an issue for non-strict quals, since if we can prove a qual mentioning + * only nullable rels is strict, we'd have reduced the outer join to an inner + * join in reduce_outer_joins().) + * + * To enforce (2), scan the join_info_list and merge the required-relid sets of + * any such OJs into the clause's own reference list. At the time we are + * called, the join_info_list contains only outer joins below this qual. We + * have to repeat the scan until no new relids get added; this ensures that + * the qual is suitably delayed regardless of the order in which OJs get + * executed. As an example, if we have one OJ with LHS=A, RHS=B, and one with + * LHS=B, RHS=C, it is implied that these can be done in either order; if the + * B/C join is done first then the join to A can null C, so a qual actually + * mentioning only C cannot be applied below the join to A. + * + * For a non-pushed-down qual, this isn't going to determine where we place the + * qual, but we need to determine outerjoin_delayed and nullable_relids anyway + * for use later in the planning process. + * + * Lastly, a pushed-down qual that references the nullable side of any current + * join_info_list member and has to be evaluated above that OJ (because its + * required relids overlap the LHS too) causes that OJ's delay_upper_joins + * flag to be set true. This will prevent any higher-level OJs from + * being interchanged with that OJ, which would result in not having any + * correct place to evaluate the qual. (The case we care about here is a + * sub-select WHERE clause within the RHS of some outer join. The WHERE + * clause must effectively be treated as a degenerate clause of that outer + * join's condition. Rather than trying to match such clauses with joins + * directly, we set delay_upper_joins here, and when the upper outer join + * is processed by make_outerjoininfo, it will refrain from allowing the + * two OJs to commute.) + */ +static bool +check_outerjoin_delay(PlannerInfo *root, + Relids *relids_p, /* in/out parameter */ + Relids *nullable_relids_p, /* output parameter */ + bool is_pushed_down) +{ + Relids relids; + Relids nullable_relids; + bool outerjoin_delayed; + bool found_some; + + /* fast path if no special joins */ + if (root->join_info_list == NIL) + { + *nullable_relids_p = NULL; + return false; + } + + /* must copy relids because we need the original value at the end */ + relids = bms_copy(*relids_p); + nullable_relids = NULL; + outerjoin_delayed = false; + do + { + ListCell *l; + + found_some = false; + foreach(l, root->join_info_list) + { + SpecialJoinInfo *sjinfo = (SpecialJoinInfo *) lfirst(l); + + /* do we reference any nullable rels of this OJ? */ + if (bms_overlap(relids, sjinfo->min_righthand) || + (sjinfo->jointype == JOIN_FULL && + bms_overlap(relids, sjinfo->min_lefthand))) + { + /* yes; have we included all its rels in relids? */ + if (!bms_is_subset(sjinfo->min_lefthand, relids) || + !bms_is_subset(sjinfo->min_righthand, relids)) + { + /* no, so add them in */ + relids = bms_add_members(relids, sjinfo->min_lefthand); + relids = bms_add_members(relids, sjinfo->min_righthand); + outerjoin_delayed = true; + /* we'll need another iteration */ + found_some = true; + } + /* track all the nullable rels of relevant OJs */ + nullable_relids = bms_add_members(nullable_relids, + sjinfo->min_righthand); + if (sjinfo->jointype == JOIN_FULL) + nullable_relids = bms_add_members(nullable_relids, + sjinfo->min_lefthand); + /* set delay_upper_joins if needed */ + if (is_pushed_down && sjinfo->jointype != JOIN_FULL && + bms_overlap(relids, sjinfo->min_lefthand)) + sjinfo->delay_upper_joins = true; + } + } + } while (found_some); + + /* identify just the actually-referenced nullable rels */ + nullable_relids = bms_int_members(nullable_relids, *relids_p); + + /* replace *relids_p, and return nullable_relids */ + bms_free(*relids_p); + *relids_p = relids; + *nullable_relids_p = nullable_relids; + return outerjoin_delayed; +} + +/* + * check_equivalence_delay + * Detect whether a potential equivalence clause is rendered unsafe + * by outer-join-delay considerations. Return true if it's safe. + * + * The initial tests in distribute_qual_to_rels will consider a mergejoinable + * clause to be a potential equivalence clause if it is not outerjoin_delayed. + * But since the point of equivalence processing is that we will recombine the + * two sides of the clause with others, we have to check that each side + * satisfies the not-outerjoin_delayed condition on its own; otherwise it might + * not be safe to evaluate everywhere we could place a derived equivalence + * condition. + */ +static bool +check_equivalence_delay(PlannerInfo *root, + RestrictInfo *restrictinfo) +{ + Relids relids; + Relids nullable_relids; + + /* fast path if no special joins */ + if (root->join_info_list == NIL) + return true; + + /* must copy restrictinfo's relids to avoid changing it */ + relids = bms_copy(restrictinfo->left_relids); + /* check left side does not need delay */ + if (check_outerjoin_delay(root, &relids, &nullable_relids, true)) + return false; + + /* and similarly for the right side */ + relids = bms_copy(restrictinfo->right_relids); + if (check_outerjoin_delay(root, &relids, &nullable_relids, true)) + return false; + + return true; +} + +/* + * check_redundant_nullability_qual + * Check to see if the qual is an IS NULL qual that is redundant with + * a lower JOIN_ANTI join. + * + * We want to suppress redundant IS NULL quals, not so much to save cycles + * as to avoid generating bogus selectivity estimates for them. So if + * redundancy is detected here, distribute_qual_to_rels() just throws away + * the qual. + */ +static bool +check_redundant_nullability_qual(PlannerInfo *root, Node *clause) +{ + Var *forced_null_var; + Index forced_null_rel; + ListCell *lc; + + /* Check for IS NULL, and identify the Var forced to NULL */ + forced_null_var = find_forced_null_var(clause); + if (forced_null_var == NULL) + return false; + forced_null_rel = forced_null_var->varno; + + /* + * If the Var comes from the nullable side of a lower antijoin, the IS + * NULL condition is necessarily true. + */ + foreach(lc, root->join_info_list) + { + SpecialJoinInfo *sjinfo = (SpecialJoinInfo *) lfirst(lc); + + if (sjinfo->jointype == JOIN_ANTI && + bms_is_member(forced_null_rel, sjinfo->syn_righthand)) + return true; + } + + return false; +} + +/* + * distribute_restrictinfo_to_rels + * Push a completed RestrictInfo into the proper restriction or join + * clause list(s). + * + * This is the last step of distribute_qual_to_rels() for ordinary qual + * clauses. Clauses that are interesting for equivalence-class processing + * are diverted to the EC machinery, but may ultimately get fed back here. + */ +void +distribute_restrictinfo_to_rels(PlannerInfo *root, + RestrictInfo *restrictinfo) +{ + Relids relids = restrictinfo->required_relids; + RelOptInfo *rel; + + switch (bms_membership(relids)) + { + case BMS_SINGLETON: + + /* + * There is only one relation participating in the clause, so it + * is a restriction clause for that relation. + */ + rel = find_base_rel(root, bms_singleton_member(relids)); + + /* Add clause to rel's restriction list */ + rel->baserestrictinfo = lappend(rel->baserestrictinfo, + restrictinfo); + /* Update security level info */ + rel->baserestrict_min_security = Min(rel->baserestrict_min_security, + restrictinfo->security_level); + break; + case BMS_MULTIPLE: + + /* + * The clause is a join clause, since there is more than one rel + * in its relid set. + */ + + /* + * Check for hashjoinable operators. (We don't bother setting the + * hashjoin info except in true join clauses.) + */ + check_hashjoinable(restrictinfo); + + /* + * Likewise, check if the clause is suitable to be used with a + * Memoize node to cache inner tuples during a parameterized + * nested loop. + */ + check_memoizable(restrictinfo); + + /* + * Add clause to the join lists of all the relevant relations. + */ + add_join_clause_to_rels(root, restrictinfo, relids); + break; + default: + + /* + * clause references no rels, and therefore we have no place to + * attach it. Shouldn't get here if callers are working properly. + */ + elog(ERROR, "cannot cope with variable-free clause"); + break; + } +} + +/* + * process_implied_equality + * Create a restrictinfo item that says "item1 op item2", and push it + * into the appropriate lists. (In practice opno is always a btree + * equality operator.) + * + * "qualscope" is the nominal syntactic level to impute to the restrictinfo. + * This must contain at least all the rels used in the expressions, but it + * is used only to set the qual application level when both exprs are + * variable-free. Otherwise the qual is applied at the lowest join level + * that provides all its variables. + * + * "nullable_relids" is the set of relids used in the expressions that are + * potentially nullable below the expressions. (This has to be supplied by + * caller because this function is used after deconstruct_jointree, so we + * don't have knowledge of where the clause items came from.) + * + * "security_level" is the security level to assign to the new restrictinfo. + * + * "both_const" indicates whether both items are known pseudo-constant; + * in this case it is worth applying eval_const_expressions() in case we + * can produce constant TRUE or constant FALSE. (Otherwise it's not, + * because the expressions went through eval_const_expressions already.) + * + * Returns the generated RestrictInfo, if any. The result will be NULL + * if both_const is true and we successfully reduced the clause to + * constant TRUE. + * + * Note: this function will copy item1 and item2, but it is caller's + * responsibility to make sure that the Relids parameters are fresh copies + * not shared with other uses. + * + * Note: we do not do initialize_mergeclause_eclasses() here. It is + * caller's responsibility that left_ec/right_ec be set as necessary. + */ +RestrictInfo * +process_implied_equality(PlannerInfo *root, + Oid opno, + Oid collation, + Expr *item1, + Expr *item2, + Relids qualscope, + Relids nullable_relids, + Index security_level, + bool below_outer_join, + bool both_const) +{ + RestrictInfo *restrictinfo; + Node *clause; + Relids relids; + bool pseudoconstant = false; + + /* + * Build the new clause. Copy to ensure it shares no substructure with + * original (this is necessary in case there are subselects in there...) + */ + clause = (Node *) make_opclause(opno, + BOOLOID, /* opresulttype */ + false, /* opretset */ + copyObject(item1), + copyObject(item2), + InvalidOid, + collation); + + /* If both constant, try to reduce to a boolean constant. */ + if (both_const) + { + clause = eval_const_expressions(root, clause); + + /* If we produced const TRUE, just drop the clause */ + if (clause && IsA(clause, Const)) + { + Const *cclause = (Const *) clause; + + Assert(cclause->consttype == BOOLOID); + if (!cclause->constisnull && DatumGetBool(cclause->constvalue)) + return NULL; + } + } + + /* + * The rest of this is a very cut-down version of distribute_qual_to_rels. + * We can skip most of the work therein, but there are a couple of special + * cases we still have to handle. + * + * Retrieve all relids mentioned within the possibly-simplified clause. + */ + relids = pull_varnos(root, clause); + Assert(bms_is_subset(relids, qualscope)); + + /* + * If the clause is variable-free, our normal heuristic for pushing it + * down to just the mentioned rels doesn't work, because there are none. + * Apply at the given qualscope, or at the top of tree if it's nonvolatile + * (which it very likely is, but we'll check, just to be sure). + */ + if (bms_is_empty(relids)) + { + /* eval at original syntactic level */ + relids = bms_copy(qualscope); + if (!contain_volatile_functions(clause)) + { + /* mark as gating qual */ + pseudoconstant = true; + /* tell createplan.c to check for gating quals */ + root->hasPseudoConstantQuals = true; + /* if not below outer join, push it to top of tree */ + if (!below_outer_join) + { + relids = + get_relids_in_jointree((Node *) root->parse->jointree, + false); + } + } + } + + /* + * Build the RestrictInfo node itself. + */ + restrictinfo = make_restrictinfo(root, + (Expr *) clause, + true, /* is_pushed_down */ + false, /* outerjoin_delayed */ + pseudoconstant, + security_level, + relids, + NULL, /* outer_relids */ + nullable_relids); + + /* + * If it's a join clause, add vars used in the clause to targetlists of + * their relations, so that they will be emitted by the plan nodes that + * scan those relations (else they won't be available at the join node!). + * + * Typically, we'd have already done this when the component expressions + * were first seen by distribute_qual_to_rels; but it is possible that + * some of the Vars could have missed having that done because they only + * appeared in single-relation clauses originally. So do it here for + * safety. + */ + if (bms_membership(relids) == BMS_MULTIPLE) + { + List *vars = pull_var_clause(clause, + PVC_RECURSE_AGGREGATES | + PVC_RECURSE_WINDOWFUNCS | + PVC_INCLUDE_PLACEHOLDERS); + + add_vars_to_targetlist(root, vars, relids, false); + list_free(vars); + } + + /* + * Check mergejoinability. This will usually succeed, since the op came + * from an EquivalenceClass; but we could have reduced the original clause + * to a constant. + */ + check_mergejoinable(restrictinfo); + + /* + * Note we don't do initialize_mergeclause_eclasses(); the caller can + * handle that much more cheaply than we can. It's okay to call + * distribute_restrictinfo_to_rels() before that happens. + */ + + /* + * Push the new clause into all the appropriate restrictinfo lists. + */ + distribute_restrictinfo_to_rels(root, restrictinfo); + + return restrictinfo; +} + +/* + * build_implied_join_equality --- build a RestrictInfo for a derived equality + * + * This overlaps the functionality of process_implied_equality(), but we + * must not push the RestrictInfo into the joininfo tree. + * + * Note: this function will copy item1 and item2, but it is caller's + * responsibility to make sure that the Relids parameters are fresh copies + * not shared with other uses. + * + * Note: we do not do initialize_mergeclause_eclasses() here. It is + * caller's responsibility that left_ec/right_ec be set as necessary. + */ +RestrictInfo * +build_implied_join_equality(PlannerInfo *root, + Oid opno, + Oid collation, + Expr *item1, + Expr *item2, + Relids qualscope, + Relids nullable_relids, + Index security_level) +{ + RestrictInfo *restrictinfo; + Expr *clause; + + /* + * Build the new clause. Copy to ensure it shares no substructure with + * original (this is necessary in case there are subselects in there...) + */ + clause = make_opclause(opno, + BOOLOID, /* opresulttype */ + false, /* opretset */ + copyObject(item1), + copyObject(item2), + InvalidOid, + collation); + + /* + * Build the RestrictInfo node itself. + */ + restrictinfo = make_restrictinfo(root, + clause, + true, /* is_pushed_down */ + false, /* outerjoin_delayed */ + false, /* pseudoconstant */ + security_level, /* security_level */ + qualscope, /* required_relids */ + NULL, /* outer_relids */ + nullable_relids); /* nullable_relids */ + + /* Set mergejoinability/hashjoinability flags */ + check_mergejoinable(restrictinfo); + check_hashjoinable(restrictinfo); + check_memoizable(restrictinfo); + + return restrictinfo; +} + + +/* + * match_foreign_keys_to_quals + * Match foreign-key constraints to equivalence classes and join quals + * + * The idea here is to see which query join conditions match equality + * constraints of a foreign-key relationship. For such join conditions, + * we can use the FK semantics to make selectivity estimates that are more + * reliable than estimating from statistics, especially for multiple-column + * FKs, where the normal assumption of independent conditions tends to fail. + * + * In this function we annotate the ForeignKeyOptInfos in root->fkey_list + * with info about which eclasses and join qual clauses they match, and + * discard any ForeignKeyOptInfos that are irrelevant for the query. + */ +void +match_foreign_keys_to_quals(PlannerInfo *root) +{ + List *newlist = NIL; + ListCell *lc; + + foreach(lc, root->fkey_list) + { + ForeignKeyOptInfo *fkinfo = (ForeignKeyOptInfo *) lfirst(lc); + RelOptInfo *con_rel; + RelOptInfo *ref_rel; + int colno; + + /* + * Either relid might identify a rel that is in the query's rtable but + * isn't referenced by the jointree so won't have a RelOptInfo. Hence + * don't use find_base_rel() here. We can ignore such FKs. + */ + if (fkinfo->con_relid >= root->simple_rel_array_size || + fkinfo->ref_relid >= root->simple_rel_array_size) + continue; /* just paranoia */ + con_rel = root->simple_rel_array[fkinfo->con_relid]; + if (con_rel == NULL) + continue; + ref_rel = root->simple_rel_array[fkinfo->ref_relid]; + if (ref_rel == NULL) + continue; + + /* + * Ignore FK unless both rels are baserels. This gets rid of FKs that + * link to inheritance child rels (otherrels) and those that link to + * rels removed by join removal (dead rels). + */ + if (con_rel->reloptkind != RELOPT_BASEREL || + ref_rel->reloptkind != RELOPT_BASEREL) + continue; + + /* + * Scan the columns and try to match them to eclasses and quals. + * + * Note: for simple inner joins, any match should be in an eclass. + * "Loose" quals that syntactically match an FK equality must have + * been rejected for EC status because they are outer-join quals or + * similar. We can still consider them to match the FK if they are + * not outerjoin_delayed. + */ + for (colno = 0; colno < fkinfo->nkeys; colno++) + { + EquivalenceClass *ec; + AttrNumber con_attno, + ref_attno; + Oid fpeqop; + ListCell *lc2; + + ec = match_eclasses_to_foreign_key_col(root, fkinfo, colno); + /* Don't bother looking for loose quals if we got an EC match */ + if (ec != NULL) + { + fkinfo->nmatched_ec++; + if (ec->ec_has_const) + fkinfo->nconst_ec++; + continue; + } + + /* + * Scan joininfo list for relevant clauses. Either rel's joininfo + * list would do equally well; we use con_rel's. + */ + con_attno = fkinfo->conkey[colno]; + ref_attno = fkinfo->confkey[colno]; + fpeqop = InvalidOid; /* we'll look this up only if needed */ + + foreach(lc2, con_rel->joininfo) + { + RestrictInfo *rinfo = (RestrictInfo *) lfirst(lc2); + OpExpr *clause = (OpExpr *) rinfo->clause; + Var *leftvar; + Var *rightvar; + + /* Ignore outerjoin-delayed clauses */ + if (rinfo->outerjoin_delayed) + continue; + + /* Only binary OpExprs are useful for consideration */ + if (!IsA(clause, OpExpr) || + list_length(clause->args) != 2) + continue; + leftvar = (Var *) get_leftop((Expr *) clause); + rightvar = (Var *) get_rightop((Expr *) clause); + + /* Operands must be Vars, possibly with RelabelType */ + while (leftvar && IsA(leftvar, RelabelType)) + leftvar = (Var *) ((RelabelType *) leftvar)->arg; + if (!(leftvar && IsA(leftvar, Var))) + continue; + while (rightvar && IsA(rightvar, RelabelType)) + rightvar = (Var *) ((RelabelType *) rightvar)->arg; + if (!(rightvar && IsA(rightvar, Var))) + continue; + + /* Now try to match the vars to the current foreign key cols */ + if (fkinfo->ref_relid == leftvar->varno && + ref_attno == leftvar->varattno && + fkinfo->con_relid == rightvar->varno && + con_attno == rightvar->varattno) + { + /* Vars match, but is it the right operator? */ + if (clause->opno == fkinfo->conpfeqop[colno]) + { + fkinfo->rinfos[colno] = lappend(fkinfo->rinfos[colno], + rinfo); + fkinfo->nmatched_ri++; + } + } + else if (fkinfo->ref_relid == rightvar->varno && + ref_attno == rightvar->varattno && + fkinfo->con_relid == leftvar->varno && + con_attno == leftvar->varattno) + { + /* + * Reverse match, must check commutator operator. Look it + * up if we didn't already. (In the worst case we might + * do multiple lookups here, but that would require an FK + * equality operator without commutator, which is + * unlikely.) + */ + if (!OidIsValid(fpeqop)) + fpeqop = get_commutator(fkinfo->conpfeqop[colno]); + if (clause->opno == fpeqop) + { + fkinfo->rinfos[colno] = lappend(fkinfo->rinfos[colno], + rinfo); + fkinfo->nmatched_ri++; + } + } + } + /* If we found any matching loose quals, count col as matched */ + if (fkinfo->rinfos[colno]) + fkinfo->nmatched_rcols++; + } + + /* + * Currently, we drop multicolumn FKs that aren't fully matched to the + * query. Later we might figure out how to derive some sort of + * estimate from them, in which case this test should be weakened to + * "if ((fkinfo->nmatched_ec + fkinfo->nmatched_rcols) > 0)". + */ + if ((fkinfo->nmatched_ec + fkinfo->nmatched_rcols) == fkinfo->nkeys) + newlist = lappend(newlist, fkinfo); + } + /* Replace fkey_list, thereby discarding any useless entries */ + root->fkey_list = newlist; +} + + +/***************************************************************************** + * + * CHECKS FOR MERGEJOINABLE AND HASHJOINABLE CLAUSES + * + *****************************************************************************/ + +/* + * check_mergejoinable + * If the restrictinfo's clause is mergejoinable, set the mergejoin + * info fields in the restrictinfo. + * + * Currently, we support mergejoin for binary opclauses where + * the operator is a mergejoinable operator. The arguments can be + * anything --- as long as there are no volatile functions in them. + */ +static void +check_mergejoinable(RestrictInfo *restrictinfo) +{ + Expr *clause = restrictinfo->clause; + Oid opno; + Node *leftarg; + + if (restrictinfo->pseudoconstant) + return; + if (!is_opclause(clause)) + return; + if (list_length(((OpExpr *) clause)->args) != 2) + return; + + opno = ((OpExpr *) clause)->opno; + leftarg = linitial(((OpExpr *) clause)->args); + + if (op_mergejoinable(opno, exprType(leftarg)) && + !contain_volatile_functions((Node *) restrictinfo)) + restrictinfo->mergeopfamilies = get_mergejoin_opfamilies(opno); + + /* + * Note: op_mergejoinable is just a hint; if we fail to find the operator + * in any btree opfamilies, mergeopfamilies remains NIL and so the clause + * is not treated as mergejoinable. + */ +} + +/* + * check_hashjoinable + * If the restrictinfo's clause is hashjoinable, set the hashjoin + * info fields in the restrictinfo. + * + * Currently, we support hashjoin for binary opclauses where + * the operator is a hashjoinable operator. The arguments can be + * anything --- as long as there are no volatile functions in them. + */ +static void +check_hashjoinable(RestrictInfo *restrictinfo) +{ + Expr *clause = restrictinfo->clause; + Oid opno; + Node *leftarg; + + if (restrictinfo->pseudoconstant) + return; + if (!is_opclause(clause)) + return; + if (list_length(((OpExpr *) clause)->args) != 2) + return; + + opno = ((OpExpr *) clause)->opno; + leftarg = linitial(((OpExpr *) clause)->args); + + if (op_hashjoinable(opno, exprType(leftarg)) && + !contain_volatile_functions((Node *) restrictinfo)) + restrictinfo->hashjoinoperator = opno; +} + +/* + * check_memoizable + * If the restrictinfo's clause is suitable to be used for a Memoize node, + * set the lefthasheqoperator and righthasheqoperator to the hash equality + * operator that will be needed during caching. + */ +static void +check_memoizable(RestrictInfo *restrictinfo) +{ + TypeCacheEntry *typentry; + Expr *clause = restrictinfo->clause; + Oid lefttype; + Oid righttype; + + if (restrictinfo->pseudoconstant) + return; + if (!is_opclause(clause)) + return; + if (list_length(((OpExpr *) clause)->args) != 2) + return; + + lefttype = exprType(linitial(((OpExpr *) clause)->args)); + + typentry = lookup_type_cache(lefttype, TYPECACHE_HASH_PROC | + TYPECACHE_EQ_OPR); + + if (OidIsValid(typentry->hash_proc) && OidIsValid(typentry->eq_opr)) + restrictinfo->left_hasheqoperator = typentry->eq_opr; + + righttype = exprType(lsecond(((OpExpr *) clause)->args)); + + /* + * Lookup the right type, unless it's the same as the left type, in which + * case typentry is already pointing to the required TypeCacheEntry. + */ + if (lefttype != righttype) + typentry = lookup_type_cache(righttype, TYPECACHE_HASH_PROC | + TYPECACHE_EQ_OPR); + + if (OidIsValid(typentry->hash_proc) && OidIsValid(typentry->eq_opr)) + restrictinfo->right_hasheqoperator = typentry->eq_opr; +} diff --git a/src/backend/optimizer/plan/planagg.c b/src/backend/optimizer/plan/planagg.c new file mode 100644 index 0000000..9330908 --- /dev/null +++ b/src/backend/optimizer/plan/planagg.c @@ -0,0 +1,513 @@ +/*------------------------------------------------------------------------- + * + * planagg.c + * Special planning for aggregate queries. + * + * This module tries to replace MIN/MAX aggregate functions by subqueries + * of the form + * (SELECT col FROM tab + * WHERE col IS NOT NULL AND existing-quals + * ORDER BY col ASC/DESC + * LIMIT 1) + * Given a suitable index on tab.col, this can be much faster than the + * generic scan-all-the-rows aggregation plan. We can handle multiple + * MIN/MAX aggregates by generating multiple subqueries, and their + * orderings can be different. However, if the query contains any + * non-optimizable aggregates, there's no point since we'll have to + * scan all the rows anyway. + * + * + * Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group + * Portions Copyright (c) 1994, Regents of the University of California + * + * + * IDENTIFICATION + * src/backend/optimizer/plan/planagg.c + * + *------------------------------------------------------------------------- + */ +#include "postgres.h" + +#include "access/htup_details.h" +#include "catalog/pg_aggregate.h" +#include "catalog/pg_type.h" +#include "nodes/makefuncs.h" +#include "nodes/nodeFuncs.h" +#include "optimizer/clauses.h" +#include "optimizer/cost.h" +#include "optimizer/optimizer.h" +#include "optimizer/pathnode.h" +#include "optimizer/paths.h" +#include "optimizer/planmain.h" +#include "optimizer/subselect.h" +#include "optimizer/tlist.h" +#include "parser/parse_clause.h" +#include "parser/parsetree.h" +#include "rewrite/rewriteManip.h" +#include "utils/lsyscache.h" +#include "utils/syscache.h" + +static bool can_minmax_aggs(PlannerInfo *root, List **context); +static bool build_minmax_path(PlannerInfo *root, MinMaxAggInfo *mminfo, + Oid eqop, Oid sortop, bool nulls_first); +static void minmax_qp_callback(PlannerInfo *root, void *extra); +static Oid fetch_agg_sort_op(Oid aggfnoid); + + +/* + * preprocess_minmax_aggregates - preprocess MIN/MAX aggregates + * + * Check to see whether the query contains MIN/MAX aggregate functions that + * might be optimizable via indexscans. If it does, and all the aggregates + * are potentially optimizable, then create a MinMaxAggPath and add it to + * the (UPPERREL_GROUP_AGG, NULL) upperrel. + * + * This should be called by grouping_planner() just before it's ready to call + * query_planner(), because we generate indexscan paths by cloning the + * planner's state and invoking query_planner() on a modified version of + * the query parsetree. Thus, all preprocessing needed before query_planner() + * must already be done. This relies on the list of aggregates in + * root->agginfos, so preprocess_aggrefs() must have been called already, too. + */ +void +preprocess_minmax_aggregates(PlannerInfo *root) +{ + Query *parse = root->parse; + FromExpr *jtnode; + RangeTblRef *rtr; + RangeTblEntry *rte; + List *aggs_list; + RelOptInfo *grouped_rel; + ListCell *lc; + + /* minmax_aggs list should be empty at this point */ + Assert(root->minmax_aggs == NIL); + + /* Nothing to do if query has no aggregates */ + if (!parse->hasAggs) + return; + + Assert(!parse->setOperations); /* shouldn't get here if a setop */ + Assert(parse->rowMarks == NIL); /* nor if FOR UPDATE */ + + /* + * Reject unoptimizable cases. + * + * We don't handle GROUP BY or windowing, because our current + * implementations of grouping require looking at all the rows anyway, and + * so there's not much point in optimizing MIN/MAX. + */ + if (parse->groupClause || list_length(parse->groupingSets) > 1 || + parse->hasWindowFuncs) + return; + + /* + * Reject if query contains any CTEs; there's no way to build an indexscan + * on one so we couldn't succeed here. (If the CTEs are unreferenced, + * that's not true, but it doesn't seem worth expending cycles to check.) + */ + if (parse->cteList) + return; + + /* + * We also restrict the query to reference exactly one table, since join + * conditions can't be handled reasonably. (We could perhaps handle a + * query containing cartesian-product joins, but it hardly seems worth the + * trouble.) However, the single table could be buried in several levels + * of FromExpr due to subqueries. Note the "single" table could be an + * inheritance parent, too, including the case of a UNION ALL subquery + * that's been flattened to an appendrel. + */ + jtnode = parse->jointree; + while (IsA(jtnode, FromExpr)) + { + if (list_length(jtnode->fromlist) != 1) + return; + jtnode = linitial(jtnode->fromlist); + } + if (!IsA(jtnode, RangeTblRef)) + return; + rtr = (RangeTblRef *) jtnode; + rte = planner_rt_fetch(rtr->rtindex, root); + if (rte->rtekind == RTE_RELATION) + /* ordinary relation, ok */ ; + else if (rte->rtekind == RTE_SUBQUERY && rte->inh) + /* flattened UNION ALL subquery, ok */ ; + else + return; + + /* + * Scan the tlist and HAVING qual to find all the aggregates and verify + * all are MIN/MAX aggregates. Stop as soon as we find one that isn't. + */ + aggs_list = NIL; + if (!can_minmax_aggs(root, &aggs_list)) + return; + + /* + * OK, there is at least the possibility of performing the optimization. + * Build an access path for each aggregate. If any of the aggregates + * prove to be non-indexable, give up; there is no point in optimizing + * just some of them. + */ + foreach(lc, aggs_list) + { + MinMaxAggInfo *mminfo = (MinMaxAggInfo *) lfirst(lc); + Oid eqop; + bool reverse; + + /* + * We'll need the equality operator that goes with the aggregate's + * ordering operator. + */ + eqop = get_equality_op_for_ordering_op(mminfo->aggsortop, &reverse); + if (!OidIsValid(eqop)) /* shouldn't happen */ + elog(ERROR, "could not find equality operator for ordering operator %u", + mminfo->aggsortop); + + /* + * We can use either an ordering that gives NULLS FIRST or one that + * gives NULLS LAST; furthermore there's unlikely to be much + * performance difference between them, so it doesn't seem worth + * costing out both ways if we get a hit on the first one. NULLS + * FIRST is more likely to be available if the operator is a + * reverse-sort operator, so try that first if reverse. + */ + if (build_minmax_path(root, mminfo, eqop, mminfo->aggsortop, reverse)) + continue; + if (build_minmax_path(root, mminfo, eqop, mminfo->aggsortop, !reverse)) + continue; + + /* No indexable path for this aggregate, so fail */ + return; + } + + /* + * OK, we can do the query this way. Prepare to create a MinMaxAggPath + * node. + * + * First, create an output Param node for each agg. (If we end up not + * using the MinMaxAggPath, we'll waste a PARAM_EXEC slot for each agg, + * which is not worth worrying about. We can't wait till create_plan time + * to decide whether to make the Param, unfortunately.) + */ + foreach(lc, aggs_list) + { + MinMaxAggInfo *mminfo = (MinMaxAggInfo *) lfirst(lc); + + mminfo->param = + SS_make_initplan_output_param(root, + exprType((Node *) mminfo->target), + -1, + exprCollation((Node *) mminfo->target)); + } + + /* + * Create a MinMaxAggPath node with the appropriate estimated costs and + * other needed data, and add it to the UPPERREL_GROUP_AGG upperrel, where + * it will compete against the standard aggregate implementation. (It + * will likely always win, but we need not assume that here.) + * + * Note: grouping_planner won't have created this upperrel yet, but it's + * fine for us to create it first. We will not have inserted the correct + * consider_parallel value in it, but MinMaxAggPath paths are currently + * never parallel-safe anyway, so that doesn't matter. Likewise, it + * doesn't matter that we haven't filled FDW-related fields in the rel. + * Also, because there are no rowmarks, we know that the processed_tlist + * doesn't need to change anymore, so making the pathtarget now is safe. + */ + grouped_rel = fetch_upper_rel(root, UPPERREL_GROUP_AGG, NULL); + add_path(grouped_rel, (Path *) + create_minmaxagg_path(root, grouped_rel, + create_pathtarget(root, + root->processed_tlist), + aggs_list, + (List *) parse->havingQual)); +} + +/* + * can_minmax_aggs + * Walk through all the aggregates in the query, and check + * if they are all MIN/MAX aggregates. If so, build a list of the + * distinct aggregate calls in the tree. + * + * Returns false if a non-MIN/MAX aggregate is found, true otherwise. + * + * This does not descend into subqueries, and so should be used only after + * reduction of sublinks to subplans. There mustn't be outer-aggregate + * references either. + */ +static bool +can_minmax_aggs(PlannerInfo *root, List **context) +{ + ListCell *lc; + + foreach(lc, root->agginfos) + { + AggInfo *agginfo = (AggInfo *) lfirst(lc); + Aggref *aggref = agginfo->representative_aggref; + Oid aggsortop; + TargetEntry *curTarget; + MinMaxAggInfo *mminfo; + + Assert(aggref->agglevelsup == 0); + if (list_length(aggref->args) != 1) + return false; /* it couldn't be MIN/MAX */ + + /* + * ORDER BY is usually irrelevant for MIN/MAX, but it can change the + * outcome if the aggsortop's operator class recognizes non-identical + * values as equal. For example, 4.0 and 4.00 are equal according to + * numeric_ops, yet distinguishable. If MIN() receives more than one + * value equal to 4.0 and no value less than 4.0, it is unspecified + * which of those equal values MIN() returns. An ORDER BY expression + * that differs for each of those equal values of the argument + * expression makes the result predictable once again. This is a + * niche requirement, and we do not implement it with subquery paths. + * In any case, this test lets us reject ordered-set aggregates + * quickly. + */ + if (aggref->aggorder != NIL) + return false; + /* note: we do not care if DISTINCT is mentioned ... */ + + /* + * We might implement the optimization when a FILTER clause is present + * by adding the filter to the quals of the generated subquery. For + * now, just punt. + */ + if (aggref->aggfilter != NULL) + return false; + + aggsortop = fetch_agg_sort_op(aggref->aggfnoid); + if (!OidIsValid(aggsortop)) + return false; /* not a MIN/MAX aggregate */ + + curTarget = (TargetEntry *) linitial(aggref->args); + + if (contain_mutable_functions((Node *) curTarget->expr)) + return false; /* not potentially indexable */ + + if (type_is_rowtype(exprType((Node *) curTarget->expr))) + return false; /* IS NOT NULL would have weird semantics */ + + mminfo = makeNode(MinMaxAggInfo); + mminfo->aggfnoid = aggref->aggfnoid; + mminfo->aggsortop = aggsortop; + mminfo->target = curTarget->expr; + mminfo->subroot = NULL; /* don't compute path yet */ + mminfo->path = NULL; + mminfo->pathcost = 0; + mminfo->param = NULL; + + *context = lappend(*context, mminfo); + } + return true; +} + +/* + * build_minmax_path + * Given a MIN/MAX aggregate, try to build an indexscan Path it can be + * optimized with. + * + * If successful, stash the best path in *mminfo and return true. + * Otherwise, return false. + */ +static bool +build_minmax_path(PlannerInfo *root, MinMaxAggInfo *mminfo, + Oid eqop, Oid sortop, bool nulls_first) +{ + PlannerInfo *subroot; + Query *parse; + TargetEntry *tle; + List *tlist; + NullTest *ntest; + SortGroupClause *sortcl; + RelOptInfo *final_rel; + Path *sorted_path; + Cost path_cost; + double path_fraction; + + /* + * We are going to construct what is effectively a sub-SELECT query, so + * clone the current query level's state and adjust it to make it look + * like a subquery. Any outer references will now be one level higher + * than before. (This means that when we are done, there will be no Vars + * of level 1, which is why the subquery can become an initplan.) + */ + subroot = (PlannerInfo *) palloc(sizeof(PlannerInfo)); + memcpy(subroot, root, sizeof(PlannerInfo)); + subroot->query_level++; + subroot->parent_root = root; + /* reset subplan-related stuff */ + subroot->plan_params = NIL; + subroot->outer_params = NULL; + subroot->init_plans = NIL; + subroot->agginfos = NIL; + subroot->aggtransinfos = NIL; + + subroot->parse = parse = copyObject(root->parse); + IncrementVarSublevelsUp((Node *) parse, 1, 1); + + /* append_rel_list might contain outer Vars? */ + subroot->append_rel_list = copyObject(root->append_rel_list); + IncrementVarSublevelsUp((Node *) subroot->append_rel_list, 1, 1); + /* There shouldn't be any OJ info to translate, as yet */ + Assert(subroot->join_info_list == NIL); + /* and we haven't made equivalence classes, either */ + Assert(subroot->eq_classes == NIL); + /* and we haven't created PlaceHolderInfos, either */ + Assert(subroot->placeholder_list == NIL); + + /*---------- + * Generate modified query of the form + * (SELECT col FROM tab + * WHERE col IS NOT NULL AND existing-quals + * ORDER BY col ASC/DESC + * LIMIT 1) + *---------- + */ + /* single tlist entry that is the aggregate target */ + tle = makeTargetEntry(copyObject(mminfo->target), + (AttrNumber) 1, + pstrdup("agg_target"), + false); + tlist = list_make1(tle); + subroot->processed_tlist = parse->targetList = tlist; + + /* No HAVING, no DISTINCT, no aggregates anymore */ + parse->havingQual = NULL; + subroot->hasHavingQual = false; + parse->distinctClause = NIL; + parse->hasDistinctOn = false; + parse->hasAggs = false; + + /* Build "target IS NOT NULL" expression */ + ntest = makeNode(NullTest); + ntest->nulltesttype = IS_NOT_NULL; + ntest->arg = copyObject(mminfo->target); + /* we checked it wasn't a rowtype in find_minmax_aggs_walker */ + ntest->argisrow = false; + ntest->location = -1; + + /* User might have had that in WHERE already */ + if (!list_member((List *) parse->jointree->quals, ntest)) + parse->jointree->quals = (Node *) + lcons(ntest, (List *) parse->jointree->quals); + + /* Build suitable ORDER BY clause */ + sortcl = makeNode(SortGroupClause); + sortcl->tleSortGroupRef = assignSortGroupRef(tle, subroot->processed_tlist); + sortcl->eqop = eqop; + sortcl->sortop = sortop; + sortcl->nulls_first = nulls_first; + sortcl->hashable = false; /* no need to make this accurate */ + parse->sortClause = list_make1(sortcl); + + /* set up expressions for LIMIT 1 */ + parse->limitOffset = NULL; + parse->limitCount = (Node *) makeConst(INT8OID, -1, InvalidOid, + sizeof(int64), + Int64GetDatum(1), false, + FLOAT8PASSBYVAL); + + /* + * Generate the best paths for this query, telling query_planner that we + * have LIMIT 1. + */ + subroot->tuple_fraction = 1.0; + subroot->limit_tuples = 1.0; + + final_rel = query_planner(subroot, minmax_qp_callback, NULL); + + /* + * Since we didn't go through subquery_planner() to handle the subquery, + * we have to do some of the same cleanup it would do, in particular cope + * with params and initplans used within this subquery. (This won't + * matter if we end up not using the subplan.) + */ + SS_identify_outer_params(subroot); + SS_charge_for_initplans(subroot, final_rel); + + /* + * Get the best presorted path, that being the one that's cheapest for + * fetching just one row. If there's no such path, fail. + */ + if (final_rel->rows > 1.0) + path_fraction = 1.0 / final_rel->rows; + else + path_fraction = 1.0; + + sorted_path = + get_cheapest_fractional_path_for_pathkeys(final_rel->pathlist, + subroot->query_pathkeys, + NULL, + path_fraction); + if (!sorted_path) + return false; + + /* + * The path might not return exactly what we want, so fix that. (We + * assume that this won't change any conclusions about which was the + * cheapest path.) + */ + sorted_path = apply_projection_to_path(subroot, final_rel, sorted_path, + create_pathtarget(subroot, + subroot->processed_tlist)); + + /* + * Determine cost to get just the first row of the presorted path. + * + * Note: cost calculation here should match + * compare_fractional_path_costs(). + */ + path_cost = sorted_path->startup_cost + + path_fraction * (sorted_path->total_cost - sorted_path->startup_cost); + + /* Save state for further processing */ + mminfo->subroot = subroot; + mminfo->path = sorted_path; + mminfo->pathcost = path_cost; + + return true; +} + +/* + * Compute query_pathkeys and other pathkeys during query_planner() + */ +static void +minmax_qp_callback(PlannerInfo *root, void *extra) +{ + root->group_pathkeys = NIL; + root->window_pathkeys = NIL; + root->distinct_pathkeys = NIL; + + root->sort_pathkeys = + make_pathkeys_for_sortclauses(root, + root->parse->sortClause, + root->parse->targetList); + + root->query_pathkeys = root->sort_pathkeys; +} + +/* + * Get the OID of the sort operator, if any, associated with an aggregate. + * Returns InvalidOid if there is no such operator. + */ +static Oid +fetch_agg_sort_op(Oid aggfnoid) +{ + HeapTuple aggTuple; + Form_pg_aggregate aggform; + Oid aggsortop; + + /* fetch aggregate entry from pg_aggregate */ + aggTuple = SearchSysCache1(AGGFNOID, ObjectIdGetDatum(aggfnoid)); + if (!HeapTupleIsValid(aggTuple)) + return InvalidOid; + aggform = (Form_pg_aggregate) GETSTRUCT(aggTuple); + aggsortop = aggform->aggsortop; + ReleaseSysCache(aggTuple); + + return aggsortop; +} diff --git a/src/backend/optimizer/plan/planmain.c b/src/backend/optimizer/plan/planmain.c new file mode 100644 index 0000000..5a1d006 --- /dev/null +++ b/src/backend/optimizer/plan/planmain.c @@ -0,0 +1,284 @@ +/*------------------------------------------------------------------------- + * + * planmain.c + * Routines to plan a single query + * + * What's in a name, anyway? The top-level entry point of the planner/ + * optimizer is over in planner.c, not here as you might think from the + * file name. But this is the main code for planning a basic join operation, + * shorn of features like subselects, inheritance, aggregates, grouping, + * and so on. (Those are the things planner.c deals with.) + * + * Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group + * Portions Copyright (c) 1994, Regents of the University of California + * + * + * IDENTIFICATION + * src/backend/optimizer/plan/planmain.c + * + *------------------------------------------------------------------------- + */ +#include "postgres.h" + +#include "optimizer/appendinfo.h" +#include "optimizer/clauses.h" +#include "optimizer/inherit.h" +#include "optimizer/optimizer.h" +#include "optimizer/orclauses.h" +#include "optimizer/pathnode.h" +#include "optimizer/paths.h" +#include "optimizer/placeholder.h" +#include "optimizer/planmain.h" + + +/* + * query_planner + * Generate a path (that is, a simplified plan) for a basic query, + * which may involve joins but not any fancier features. + * + * Since query_planner does not handle the toplevel processing (grouping, + * sorting, etc) it cannot select the best path by itself. Instead, it + * returns the RelOptInfo for the top level of joining, and the caller + * (grouping_planner) can choose among the surviving paths for the rel. + * + * root describes the query to plan + * qp_callback is a function to compute query_pathkeys once it's safe to do so + * qp_extra is optional extra data to pass to qp_callback + * + * Note: the PlannerInfo node also includes a query_pathkeys field, which + * tells query_planner the sort order that is desired in the final output + * plan. This value is *not* available at call time, but is computed by + * qp_callback once we have completed merging the query's equivalence classes. + * (We cannot construct canonical pathkeys until that's done.) + */ +RelOptInfo * +query_planner(PlannerInfo *root, + query_pathkeys_callback qp_callback, void *qp_extra) +{ + Query *parse = root->parse; + List *joinlist; + RelOptInfo *final_rel; + + /* + * Init planner lists to empty. + * + * NOTE: append_rel_list was set up by subquery_planner, so do not touch + * here. + */ + root->join_rel_list = NIL; + root->join_rel_hash = NULL; + root->join_rel_level = NULL; + root->join_cur_level = 0; + root->canon_pathkeys = NIL; + root->left_join_clauses = NIL; + root->right_join_clauses = NIL; + root->full_join_clauses = NIL; + root->join_info_list = NIL; + root->placeholder_list = NIL; + root->fkey_list = NIL; + root->initial_rels = NIL; + + /* + * Set up arrays for accessing base relations and AppendRelInfos. + */ + setup_simple_rel_arrays(root); + + /* + * In the trivial case where the jointree is a single RTE_RESULT relation, + * bypass all the rest of this function and just make a RelOptInfo and its + * one access path. This is worth optimizing because it applies for + * common cases like "SELECT expression" and "INSERT ... VALUES()". + */ + Assert(parse->jointree->fromlist != NIL); + if (list_length(parse->jointree->fromlist) == 1) + { + Node *jtnode = (Node *) linitial(parse->jointree->fromlist); + + if (IsA(jtnode, RangeTblRef)) + { + int varno = ((RangeTblRef *) jtnode)->rtindex; + RangeTblEntry *rte = root->simple_rte_array[varno]; + + Assert(rte != NULL); + if (rte->rtekind == RTE_RESULT) + { + /* Make the RelOptInfo for it directly */ + final_rel = build_simple_rel(root, varno, NULL); + + /* + * If query allows parallelism in general, check whether the + * quals are parallel-restricted. (We need not check + * final_rel->reltarget because it's empty at this point. + * Anything parallel-restricted in the query tlist will be + * dealt with later.) This is normally pretty silly, because + * a Result-only plan would never be interesting to + * parallelize. However, if force_parallel_mode is on, then + * we want to execute the Result in a parallel worker if + * possible, so we must do this. + */ + if (root->glob->parallelModeOK && + force_parallel_mode != FORCE_PARALLEL_OFF) + final_rel->consider_parallel = + is_parallel_safe(root, parse->jointree->quals); + + /* + * The only path for it is a trivial Result path. We cheat a + * bit here by using a GroupResultPath, because that way we + * can just jam the quals into it without preprocessing them. + * (But, if you hold your head at the right angle, a FROM-less + * SELECT is a kind of degenerate-grouping case, so it's not + * that much of a cheat.) + */ + add_path(final_rel, (Path *) + create_group_result_path(root, final_rel, + final_rel->reltarget, + (List *) parse->jointree->quals)); + + /* Select cheapest path (pretty easy in this case...) */ + set_cheapest(final_rel); + + /* + * We don't need to run generate_base_implied_equalities, but + * we do need to pretend that EC merging is complete. + */ + root->ec_merging_done = true; + + /* + * We still are required to call qp_callback, in case it's + * something like "SELECT 2+2 ORDER BY 1". + */ + (*qp_callback) (root, qp_extra); + + return final_rel; + } + } + } + + /* + * Construct RelOptInfo nodes for all base relations used in the query. + * Appendrel member relations ("other rels") will be added later. + * + * Note: the reason we find the baserels by searching the jointree, rather + * than scanning the rangetable, is that the rangetable may contain RTEs + * for rels not actively part of the query, for example views. We don't + * want to make RelOptInfos for them. + */ + add_base_rels_to_query(root, (Node *) parse->jointree); + + /* + * Examine the targetlist and join tree, adding entries to baserel + * targetlists for all referenced Vars, and generating PlaceHolderInfo + * entries for all referenced PlaceHolderVars. Restrict and join clauses + * are added to appropriate lists belonging to the mentioned relations. We + * also build EquivalenceClasses for provably equivalent expressions. The + * SpecialJoinInfo list is also built to hold information about join order + * restrictions. Finally, we form a target joinlist for make_one_rel() to + * work from. + */ + build_base_rel_tlists(root, root->processed_tlist); + + find_placeholders_in_jointree(root); + + find_lateral_references(root); + + joinlist = deconstruct_jointree(root); + + /* + * Reconsider any postponed outer-join quals now that we have built up + * equivalence classes. (This could result in further additions or + * mergings of classes.) + */ + reconsider_outer_join_clauses(root); + + /* + * If we formed any equivalence classes, generate additional restriction + * clauses as appropriate. (Implied join clauses are formed on-the-fly + * later.) + */ + generate_base_implied_equalities(root); + + /* + * We have completed merging equivalence sets, so it's now possible to + * generate pathkeys in canonical form; so compute query_pathkeys and + * other pathkeys fields in PlannerInfo. + */ + (*qp_callback) (root, qp_extra); + + /* + * Examine any "placeholder" expressions generated during subquery pullup. + * Make sure that the Vars they need are marked as needed at the relevant + * join level. This must be done before join removal because it might + * cause Vars or placeholders to be needed above a join when they weren't + * so marked before. + */ + fix_placeholder_input_needed_levels(root); + + /* + * Remove any useless outer joins. Ideally this would be done during + * jointree preprocessing, but the necessary information isn't available + * until we've built baserel data structures and classified qual clauses. + */ + joinlist = remove_useless_joins(root, joinlist); + + /* + * Also, reduce any semijoins with unique inner rels to plain inner joins. + * Likewise, this can't be done until now for lack of needed info. + */ + reduce_unique_semijoins(root); + + /* + * Now distribute "placeholders" to base rels as needed. This has to be + * done after join removal because removal could change whether a + * placeholder is evaluable at a base rel. + */ + add_placeholders_to_base_rels(root); + + /* + * Construct the lateral reference sets now that we have finalized + * PlaceHolderVar eval levels. + */ + create_lateral_join_info(root); + + /* + * Match foreign keys to equivalence classes and join quals. This must be + * done after finalizing equivalence classes, and it's useful to wait till + * after join removal so that we can skip processing foreign keys + * involving removed relations. + */ + match_foreign_keys_to_quals(root); + + /* + * Look for join OR clauses that we can extract single-relation + * restriction OR clauses from. + */ + extract_restriction_or_clauses(root); + + /* + * Now expand appendrels by adding "otherrels" for their children. We + * delay this to the end so that we have as much information as possible + * available for each baserel, including all restriction clauses. That + * let us prune away partitions that don't satisfy a restriction clause. + * Also note that some information such as lateral_relids is propagated + * from baserels to otherrels here, so we must have computed it already. + */ + add_other_rels_to_query(root); + + /* + * Distribute any UPDATE/DELETE/MERGE row identity variables to the target + * relations. This can't be done till we've finished expansion of + * appendrels. + */ + distribute_row_identity_vars(root); + + /* + * Ready to do the primary planning. + */ + final_rel = make_one_rel(root, joinlist); + + /* Check that we got at least one usable path */ + if (!final_rel || !final_rel->cheapest_total_path || + final_rel->cheapest_total_path->param_info != NULL) + elog(ERROR, "failed to construct the join relation"); + + return final_rel; +} diff --git a/src/backend/optimizer/plan/planner.c b/src/backend/optimizer/plan/planner.c new file mode 100644 index 0000000..bd4e4ce --- /dev/null +++ b/src/backend/optimizer/plan/planner.c @@ -0,0 +1,7492 @@ +/*------------------------------------------------------------------------- + * + * planner.c + * The query optimizer external interface. + * + * Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group + * Portions Copyright (c) 1994, Regents of the University of California + * + * + * IDENTIFICATION + * src/backend/optimizer/plan/planner.c + * + *------------------------------------------------------------------------- + */ + +#include "postgres.h" + +#include <limits.h> +#include <math.h> + +#include "access/genam.h" +#include "access/htup_details.h" +#include "access/parallel.h" +#include "access/sysattr.h" +#include "access/table.h" +#include "access/xact.h" +#include "catalog/pg_constraint.h" +#include "catalog/pg_inherits.h" +#include "catalog/pg_proc.h" +#include "catalog/pg_type.h" +#include "executor/executor.h" +#include "executor/nodeAgg.h" +#include "foreign/fdwapi.h" +#include "jit/jit.h" +#include "lib/bipartite_match.h" +#include "lib/knapsack.h" +#include "miscadmin.h" +#include "nodes/makefuncs.h" +#include "nodes/nodeFuncs.h" +#ifdef OPTIMIZER_DEBUG +#include "nodes/print.h" +#endif +#include "optimizer/appendinfo.h" +#include "optimizer/clauses.h" +#include "optimizer/cost.h" +#include "optimizer/inherit.h" +#include "optimizer/optimizer.h" +#include "optimizer/paramassign.h" +#include "optimizer/pathnode.h" +#include "optimizer/paths.h" +#include "optimizer/plancat.h" +#include "optimizer/planmain.h" +#include "optimizer/planner.h" +#include "optimizer/prep.h" +#include "optimizer/subselect.h" +#include "optimizer/tlist.h" +#include "parser/analyze.h" +#include "parser/parse_agg.h" +#include "parser/parsetree.h" +#include "partitioning/partdesc.h" +#include "rewrite/rewriteManip.h" +#include "storage/dsm_impl.h" +#include "utils/lsyscache.h" +#include "utils/rel.h" +#include "utils/selfuncs.h" +#include "utils/syscache.h" + +/* GUC parameters */ +double cursor_tuple_fraction = DEFAULT_CURSOR_TUPLE_FRACTION; +int force_parallel_mode = FORCE_PARALLEL_OFF; +bool parallel_leader_participation = true; + +/* Hook for plugins to get control in planner() */ +planner_hook_type planner_hook = NULL; + +/* Hook for plugins to get control when grouping_planner() plans upper rels */ +create_upper_paths_hook_type create_upper_paths_hook = NULL; + + +/* Expression kind codes for preprocess_expression */ +#define EXPRKIND_QUAL 0 +#define EXPRKIND_TARGET 1 +#define EXPRKIND_RTFUNC 2 +#define EXPRKIND_RTFUNC_LATERAL 3 +#define EXPRKIND_VALUES 4 +#define EXPRKIND_VALUES_LATERAL 5 +#define EXPRKIND_LIMIT 6 +#define EXPRKIND_APPINFO 7 +#define EXPRKIND_PHV 8 +#define EXPRKIND_TABLESAMPLE 9 +#define EXPRKIND_ARBITER_ELEM 10 +#define EXPRKIND_TABLEFUNC 11 +#define EXPRKIND_TABLEFUNC_LATERAL 12 + +/* Passthrough data for standard_qp_callback */ +typedef struct +{ + List *activeWindows; /* active windows, if any */ + List *groupClause; /* overrides parse->groupClause */ +} standard_qp_extra; + +/* + * Data specific to grouping sets + */ + +typedef struct +{ + List *rollups; + List *hash_sets_idx; + double dNumHashGroups; + bool any_hashable; + Bitmapset *unsortable_refs; + Bitmapset *unhashable_refs; + List *unsortable_sets; + int *tleref_to_colnum_map; +} grouping_sets_data; + +/* + * Temporary structure for use during WindowClause reordering in order to be + * able to sort WindowClauses on partitioning/ordering prefix. + */ +typedef struct +{ + WindowClause *wc; + List *uniqueOrder; /* A List of unique ordering/partitioning + * clauses per Window */ +} WindowClauseSortData; + +/* Local functions */ +static Node *preprocess_expression(PlannerInfo *root, Node *expr, int kind); +static void preprocess_qual_conditions(PlannerInfo *root, Node *jtnode); +static void grouping_planner(PlannerInfo *root, double tuple_fraction); +static grouping_sets_data *preprocess_grouping_sets(PlannerInfo *root); +static List *remap_to_groupclause_idx(List *groupClause, List *gsets, + int *tleref_to_colnum_map); +static void preprocess_rowmarks(PlannerInfo *root); +static double preprocess_limit(PlannerInfo *root, + double tuple_fraction, + int64 *offset_est, int64 *count_est); +static void remove_useless_groupby_columns(PlannerInfo *root); +static List *preprocess_groupclause(PlannerInfo *root, List *force); +static List *extract_rollup_sets(List *groupingSets); +static List *reorder_grouping_sets(List *groupingSets, List *sortclause); +static void standard_qp_callback(PlannerInfo *root, void *extra); +static double get_number_of_groups(PlannerInfo *root, + double path_rows, + grouping_sets_data *gd, + List *target_list); +static RelOptInfo *create_grouping_paths(PlannerInfo *root, + RelOptInfo *input_rel, + PathTarget *target, + bool target_parallel_safe, + grouping_sets_data *gd); +static bool is_degenerate_grouping(PlannerInfo *root); +static void create_degenerate_grouping_paths(PlannerInfo *root, + RelOptInfo *input_rel, + RelOptInfo *grouped_rel); +static RelOptInfo *make_grouping_rel(PlannerInfo *root, RelOptInfo *input_rel, + PathTarget *target, bool target_parallel_safe, + Node *havingQual); +static void create_ordinary_grouping_paths(PlannerInfo *root, + RelOptInfo *input_rel, + RelOptInfo *grouped_rel, + const AggClauseCosts *agg_costs, + grouping_sets_data *gd, + GroupPathExtraData *extra, + RelOptInfo **partially_grouped_rel_p); +static void consider_groupingsets_paths(PlannerInfo *root, + RelOptInfo *grouped_rel, + Path *path, + bool is_sorted, + bool can_hash, + grouping_sets_data *gd, + const AggClauseCosts *agg_costs, + double dNumGroups); +static RelOptInfo *create_window_paths(PlannerInfo *root, + RelOptInfo *input_rel, + PathTarget *input_target, + PathTarget *output_target, + bool output_target_parallel_safe, + WindowFuncLists *wflists, + List *activeWindows); +static void create_one_window_path(PlannerInfo *root, + RelOptInfo *window_rel, + Path *path, + PathTarget *input_target, + PathTarget *output_target, + WindowFuncLists *wflists, + List *activeWindows); +static RelOptInfo *create_distinct_paths(PlannerInfo *root, + RelOptInfo *input_rel); +static void create_partial_distinct_paths(PlannerInfo *root, + RelOptInfo *input_rel, + RelOptInfo *final_distinct_rel); +static RelOptInfo *create_final_distinct_paths(PlannerInfo *root, + RelOptInfo *input_rel, + RelOptInfo *distinct_rel); +static RelOptInfo *create_ordered_paths(PlannerInfo *root, + RelOptInfo *input_rel, + PathTarget *target, + bool target_parallel_safe, + double limit_tuples); +static PathTarget *make_group_input_target(PlannerInfo *root, + PathTarget *final_target); +static PathTarget *make_partial_grouping_target(PlannerInfo *root, + PathTarget *grouping_target, + Node *havingQual); +static List *postprocess_setop_tlist(List *new_tlist, List *orig_tlist); +static List *select_active_windows(PlannerInfo *root, WindowFuncLists *wflists); +static PathTarget *make_window_input_target(PlannerInfo *root, + PathTarget *final_target, + List *activeWindows); +static List *make_pathkeys_for_window(PlannerInfo *root, WindowClause *wc, + List *tlist); +static PathTarget *make_sort_input_target(PlannerInfo *root, + PathTarget *final_target, + bool *have_postponed_srfs); +static void adjust_paths_for_srfs(PlannerInfo *root, RelOptInfo *rel, + List *targets, List *targets_contain_srfs); +static void add_paths_to_grouping_rel(PlannerInfo *root, RelOptInfo *input_rel, + RelOptInfo *grouped_rel, + RelOptInfo *partially_grouped_rel, + const AggClauseCosts *agg_costs, + grouping_sets_data *gd, + double dNumGroups, + GroupPathExtraData *extra); +static RelOptInfo *create_partial_grouping_paths(PlannerInfo *root, + RelOptInfo *grouped_rel, + RelOptInfo *input_rel, + grouping_sets_data *gd, + GroupPathExtraData *extra, + bool force_rel_creation); +static void gather_grouping_paths(PlannerInfo *root, RelOptInfo *rel); +static bool can_partial_agg(PlannerInfo *root); +static void apply_scanjoin_target_to_paths(PlannerInfo *root, + RelOptInfo *rel, + List *scanjoin_targets, + List *scanjoin_targets_contain_srfs, + bool scanjoin_target_parallel_safe, + bool tlist_same_exprs); +static void create_partitionwise_grouping_paths(PlannerInfo *root, + RelOptInfo *input_rel, + RelOptInfo *grouped_rel, + RelOptInfo *partially_grouped_rel, + const AggClauseCosts *agg_costs, + grouping_sets_data *gd, + PartitionwiseAggregateType patype, + GroupPathExtraData *extra); +static bool group_by_has_partkey(RelOptInfo *input_rel, + List *targetList, + List *groupClause); +static int common_prefix_cmp(const void *a, const void *b); + + +/***************************************************************************** + * + * Query optimizer entry point + * + * To support loadable plugins that monitor or modify planner behavior, + * we provide a hook variable that lets a plugin get control before and + * after the standard planning process. The plugin would normally call + * standard_planner(). + * + * Note to plugin authors: standard_planner() scribbles on its Query input, + * so you'd better copy that data structure if you want to plan more than once. + * + *****************************************************************************/ +PlannedStmt * +planner(Query *parse, const char *query_string, int cursorOptions, + ParamListInfo boundParams) +{ + PlannedStmt *result; + + if (planner_hook) + result = (*planner_hook) (parse, query_string, cursorOptions, boundParams); + else + result = standard_planner(parse, query_string, cursorOptions, boundParams); + return result; +} + +PlannedStmt * +standard_planner(Query *parse, const char *query_string, int cursorOptions, + ParamListInfo boundParams) +{ + PlannedStmt *result; + PlannerGlobal *glob; + double tuple_fraction; + PlannerInfo *root; + RelOptInfo *final_rel; + Path *best_path; + Plan *top_plan; + ListCell *lp, + *lr; + + /* + * Set up global state for this planner invocation. This data is needed + * across all levels of sub-Query that might exist in the given command, + * so we keep it in a separate struct that's linked to by each per-Query + * PlannerInfo. + */ + glob = makeNode(PlannerGlobal); + + glob->boundParams = boundParams; + glob->subplans = NIL; + glob->subroots = NIL; + glob->rewindPlanIDs = NULL; + glob->finalrtable = NIL; + glob->finalrowmarks = NIL; + glob->resultRelations = NIL; + glob->appendRelations = NIL; + glob->relationOids = NIL; + glob->invalItems = NIL; + glob->paramExecTypes = NIL; + glob->lastPHId = 0; + glob->lastRowMarkId = 0; + glob->lastPlanNodeId = 0; + glob->transientPlan = false; + glob->dependsOnRole = false; + + /* + * Assess whether it's feasible to use parallel mode for this query. We + * can't do this in a standalone backend, or if the command will try to + * modify any data, or if this is a cursor operation, or if GUCs are set + * to values that don't permit parallelism, or if parallel-unsafe + * functions are present in the query tree. + * + * (Note that we do allow CREATE TABLE AS, SELECT INTO, and CREATE + * MATERIALIZED VIEW to use parallel plans, but this is safe only because + * the command is writing into a completely new table which workers won't + * be able to see. If the workers could see the table, the fact that + * group locking would cause them to ignore the leader's heavyweight + * GIN page locks would make this unsafe. We'll have to fix that somehow + * if we want to allow parallel inserts in general; updates and deletes + * have additional problems especially around combo CIDs.) + * + * For now, we don't try to use parallel mode if we're running inside a + * parallel worker. We might eventually be able to relax this + * restriction, but for now it seems best not to have parallel workers + * trying to create their own parallel workers. + */ + if ((cursorOptions & CURSOR_OPT_PARALLEL_OK) != 0 && + IsUnderPostmaster && + parse->commandType == CMD_SELECT && + !parse->hasModifyingCTE && + max_parallel_workers_per_gather > 0 && + !IsParallelWorker()) + { + /* all the cheap tests pass, so scan the query tree */ + glob->maxParallelHazard = max_parallel_hazard(parse); + glob->parallelModeOK = (glob->maxParallelHazard != PROPARALLEL_UNSAFE); + } + else + { + /* skip the query tree scan, just assume it's unsafe */ + glob->maxParallelHazard = PROPARALLEL_UNSAFE; + glob->parallelModeOK = false; + } + + /* + * glob->parallelModeNeeded is normally set to false here and changed to + * true during plan creation if a Gather or Gather Merge plan is actually + * created (cf. create_gather_plan, create_gather_merge_plan). + * + * However, if force_parallel_mode = on or force_parallel_mode = regress, + * then we impose parallel mode whenever it's safe to do so, even if the + * final plan doesn't use parallelism. It's not safe to do so if the + * query contains anything parallel-unsafe; parallelModeOK will be false + * in that case. Note that parallelModeOK can't change after this point. + * Otherwise, everything in the query is either parallel-safe or + * parallel-restricted, and in either case it should be OK to impose + * parallel-mode restrictions. If that ends up breaking something, then + * either some function the user included in the query is incorrectly + * labeled as parallel-safe or parallel-restricted when in reality it's + * parallel-unsafe, or else the query planner itself has a bug. + */ + glob->parallelModeNeeded = glob->parallelModeOK && + (force_parallel_mode != FORCE_PARALLEL_OFF); + + /* Determine what fraction of the plan is likely to be scanned */ + if (cursorOptions & CURSOR_OPT_FAST_PLAN) + { + /* + * We have no real idea how many tuples the user will ultimately FETCH + * from a cursor, but it is often the case that he doesn't want 'em + * all, or would prefer a fast-start plan anyway so that he can + * process some of the tuples sooner. Use a GUC parameter to decide + * what fraction to optimize for. + */ + tuple_fraction = cursor_tuple_fraction; + + /* + * We document cursor_tuple_fraction as simply being a fraction, which + * means the edge cases 0 and 1 have to be treated specially here. We + * convert 1 to 0 ("all the tuples") and 0 to a very small fraction. + */ + if (tuple_fraction >= 1.0) + tuple_fraction = 0.0; + else if (tuple_fraction <= 0.0) + tuple_fraction = 1e-10; + } + else + { + /* Default assumption is we need all the tuples */ + tuple_fraction = 0.0; + } + + /* primary planning entry point (may recurse for subqueries) */ + root = subquery_planner(glob, parse, NULL, + false, tuple_fraction); + + /* Select best Path and turn it into a Plan */ + final_rel = fetch_upper_rel(root, UPPERREL_FINAL, NULL); + best_path = get_cheapest_fractional_path(final_rel, tuple_fraction); + + top_plan = create_plan(root, best_path); + + /* + * If creating a plan for a scrollable cursor, make sure it can run + * backwards on demand. Add a Material node at the top at need. + */ + if (cursorOptions & CURSOR_OPT_SCROLL) + { + if (!ExecSupportsBackwardScan(top_plan)) + top_plan = materialize_finished_plan(top_plan); + } + + /* + * Optionally add a Gather node for testing purposes, provided this is + * actually a safe thing to do. + */ + if (force_parallel_mode != FORCE_PARALLEL_OFF && top_plan->parallel_safe) + { + Gather *gather = makeNode(Gather); + + /* + * Top plan must not have any initPlans, else it shouldn't have been + * marked parallel-safe. + */ + Assert(top_plan->initPlan == NIL); + + gather->plan.targetlist = top_plan->targetlist; + gather->plan.qual = NIL; + gather->plan.lefttree = top_plan; + gather->plan.righttree = NULL; + gather->num_workers = 1; + gather->single_copy = true; + gather->invisible = (force_parallel_mode == FORCE_PARALLEL_REGRESS); + + /* + * Since this Gather has no parallel-aware descendants to signal to, + * we don't need a rescan Param. + */ + gather->rescan_param = -1; + + /* + * Ideally we'd use cost_gather here, but setting up dummy path data + * to satisfy it doesn't seem much cleaner than knowing what it does. + */ + gather->plan.startup_cost = top_plan->startup_cost + + parallel_setup_cost; + gather->plan.total_cost = top_plan->total_cost + + parallel_setup_cost + parallel_tuple_cost * top_plan->plan_rows; + gather->plan.plan_rows = top_plan->plan_rows; + gather->plan.plan_width = top_plan->plan_width; + gather->plan.parallel_aware = false; + gather->plan.parallel_safe = false; + + /* use parallel mode for parallel plans. */ + root->glob->parallelModeNeeded = true; + + top_plan = &gather->plan; + } + + /* + * If any Params were generated, run through the plan tree and compute + * each plan node's extParam/allParam sets. Ideally we'd merge this into + * set_plan_references' tree traversal, but for now it has to be separate + * because we need to visit subplans before not after main plan. + */ + if (glob->paramExecTypes != NIL) + { + Assert(list_length(glob->subplans) == list_length(glob->subroots)); + forboth(lp, glob->subplans, lr, glob->subroots) + { + Plan *subplan = (Plan *) lfirst(lp); + PlannerInfo *subroot = lfirst_node(PlannerInfo, lr); + + SS_finalize_plan(subroot, subplan); + } + SS_finalize_plan(root, top_plan); + } + + /* final cleanup of the plan */ + Assert(glob->finalrtable == NIL); + Assert(glob->finalrowmarks == NIL); + Assert(glob->resultRelations == NIL); + Assert(glob->appendRelations == NIL); + top_plan = set_plan_references(root, top_plan); + /* ... and the subplans (both regular subplans and initplans) */ + Assert(list_length(glob->subplans) == list_length(glob->subroots)); + forboth(lp, glob->subplans, lr, glob->subroots) + { + Plan *subplan = (Plan *) lfirst(lp); + PlannerInfo *subroot = lfirst_node(PlannerInfo, lr); + + lfirst(lp) = set_plan_references(subroot, subplan); + } + + /* build the PlannedStmt result */ + result = makeNode(PlannedStmt); + + result->commandType = parse->commandType; + result->queryId = parse->queryId; + result->hasReturning = (parse->returningList != NIL); + result->hasModifyingCTE = parse->hasModifyingCTE; + result->canSetTag = parse->canSetTag; + result->transientPlan = glob->transientPlan; + result->dependsOnRole = glob->dependsOnRole; + result->parallelModeNeeded = glob->parallelModeNeeded; + result->planTree = top_plan; + result->rtable = glob->finalrtable; + result->resultRelations = glob->resultRelations; + result->appendRelations = glob->appendRelations; + result->subplans = glob->subplans; + result->rewindPlanIDs = glob->rewindPlanIDs; + result->rowMarks = glob->finalrowmarks; + result->relationOids = glob->relationOids; + result->invalItems = glob->invalItems; + result->paramExecTypes = glob->paramExecTypes; + /* utilityStmt should be null, but we might as well copy it */ + result->utilityStmt = parse->utilityStmt; + result->stmt_location = parse->stmt_location; + result->stmt_len = parse->stmt_len; + + result->jitFlags = PGJIT_NONE; + if (jit_enabled && jit_above_cost >= 0 && + top_plan->total_cost > jit_above_cost) + { + result->jitFlags |= PGJIT_PERFORM; + + /* + * Decide how much effort should be put into generating better code. + */ + if (jit_optimize_above_cost >= 0 && + top_plan->total_cost > jit_optimize_above_cost) + result->jitFlags |= PGJIT_OPT3; + if (jit_inline_above_cost >= 0 && + top_plan->total_cost > jit_inline_above_cost) + result->jitFlags |= PGJIT_INLINE; + + /* + * Decide which operations should be JITed. + */ + if (jit_expressions) + result->jitFlags |= PGJIT_EXPR; + if (jit_tuple_deforming) + result->jitFlags |= PGJIT_DEFORM; + } + + if (glob->partition_directory != NULL) + DestroyPartitionDirectory(glob->partition_directory); + + return result; +} + + +/*-------------------- + * subquery_planner + * Invokes the planner on a subquery. We recurse to here for each + * sub-SELECT found in the query tree. + * + * glob is the global state for the current planner run. + * parse is the querytree produced by the parser & rewriter. + * parent_root is the immediate parent Query's info (NULL at the top level). + * hasRecursion is true if this is a recursive WITH query. + * tuple_fraction is the fraction of tuples we expect will be retrieved. + * tuple_fraction is interpreted as explained for grouping_planner, below. + * + * Basically, this routine does the stuff that should only be done once + * per Query object. It then calls grouping_planner. At one time, + * grouping_planner could be invoked recursively on the same Query object; + * that's not currently true, but we keep the separation between the two + * routines anyway, in case we need it again someday. + * + * subquery_planner will be called recursively to handle sub-Query nodes + * found within the query's expressions and rangetable. + * + * Returns the PlannerInfo struct ("root") that contains all data generated + * while planning the subquery. In particular, the Path(s) attached to + * the (UPPERREL_FINAL, NULL) upperrel represent our conclusions about the + * cheapest way(s) to implement the query. The top level will select the + * best Path and pass it through createplan.c to produce a finished Plan. + *-------------------- + */ +PlannerInfo * +subquery_planner(PlannerGlobal *glob, Query *parse, + PlannerInfo *parent_root, + bool hasRecursion, double tuple_fraction) +{ + PlannerInfo *root; + List *newWithCheckOptions; + List *newHaving; + bool hasOuterJoins; + bool hasResultRTEs; + RelOptInfo *final_rel; + ListCell *l; + + /* Create a PlannerInfo data structure for this subquery */ + root = makeNode(PlannerInfo); + root->parse = parse; + root->glob = glob; + root->query_level = parent_root ? parent_root->query_level + 1 : 1; + root->parent_root = parent_root; + root->plan_params = NIL; + root->outer_params = NULL; + root->planner_cxt = CurrentMemoryContext; + root->init_plans = NIL; + root->cte_plan_ids = NIL; + root->multiexpr_params = NIL; + root->eq_classes = NIL; + root->ec_merging_done = false; + root->all_result_relids = + parse->resultRelation ? bms_make_singleton(parse->resultRelation) : NULL; + root->leaf_result_relids = NULL; /* we'll find out leaf-ness later */ + root->append_rel_list = NIL; + root->row_identity_vars = NIL; + root->rowMarks = NIL; + memset(root->upper_rels, 0, sizeof(root->upper_rels)); + memset(root->upper_targets, 0, sizeof(root->upper_targets)); + root->processed_tlist = NIL; + root->update_colnos = NIL; + root->grouping_map = NULL; + root->minmax_aggs = NIL; + root->qual_security_level = 0; + root->hasPseudoConstantQuals = false; + root->hasAlternativeSubPlans = false; + root->hasRecursion = hasRecursion; + if (hasRecursion) + root->wt_param_id = assign_special_exec_param(root); + else + root->wt_param_id = -1; + root->non_recursive_path = NULL; + root->partColsUpdated = false; + + /* + * If there is a WITH list, process each WITH query and either convert it + * to RTE_SUBQUERY RTE(s) or build an initplan SubPlan structure for it. + */ + if (parse->cteList) + SS_process_ctes(root); + + /* + * If it's a MERGE command, transform the joinlist as appropriate. + */ + transform_MERGE_to_join(parse); + + /* + * If the FROM clause is empty, replace it with a dummy RTE_RESULT RTE, so + * that we don't need so many special cases to deal with that situation. + */ + replace_empty_jointree(parse); + + /* + * Look for ANY and EXISTS SubLinks in WHERE and JOIN/ON clauses, and try + * to transform them into joins. Note that this step does not descend + * into subqueries; if we pull up any subqueries below, their SubLinks are + * processed just before pulling them up. + */ + if (parse->hasSubLinks) + pull_up_sublinks(root); + + /* + * Scan the rangetable for function RTEs, do const-simplification on them, + * and then inline them if possible (producing subqueries that might get + * pulled up next). Recursion issues here are handled in the same way as + * for SubLinks. + */ + preprocess_function_rtes(root); + + /* + * Check to see if any subqueries in the jointree can be merged into this + * query. + */ + pull_up_subqueries(root); + + /* + * If this is a simple UNION ALL query, flatten it into an appendrel. We + * do this now because it requires applying pull_up_subqueries to the leaf + * queries of the UNION ALL, which weren't touched above because they + * weren't referenced by the jointree (they will be after we do this). + */ + if (parse->setOperations) + flatten_simple_union_all(root); + + /* + * Survey the rangetable to see what kinds of entries are present. We can + * skip some later processing if relevant SQL features are not used; for + * example if there are no JOIN RTEs we can avoid the expense of doing + * flatten_join_alias_vars(). This must be done after we have finished + * adding rangetable entries, of course. (Note: actually, processing of + * inherited or partitioned rels can cause RTEs for their child tables to + * get added later; but those must all be RTE_RELATION entries, so they + * don't invalidate the conclusions drawn here.) + */ + root->hasJoinRTEs = false; + root->hasLateralRTEs = false; + hasOuterJoins = false; + hasResultRTEs = false; + foreach(l, parse->rtable) + { + RangeTblEntry *rte = lfirst_node(RangeTblEntry, l); + + switch (rte->rtekind) + { + case RTE_RELATION: + if (rte->inh) + { + /* + * Check to see if the relation actually has any children; + * if not, clear the inh flag so we can treat it as a + * plain base relation. + * + * Note: this could give a false-positive result, if the + * rel once had children but no longer does. We used to + * be able to clear rte->inh later on when we discovered + * that, but no more; we have to handle such cases as + * full-fledged inheritance. + */ + rte->inh = has_subclass(rte->relid); + } + break; + case RTE_JOIN: + root->hasJoinRTEs = true; + if (IS_OUTER_JOIN(rte->jointype)) + hasOuterJoins = true; + break; + case RTE_RESULT: + hasResultRTEs = true; + break; + default: + /* No work here for other RTE types */ + break; + } + + if (rte->lateral) + root->hasLateralRTEs = true; + + /* + * We can also determine the maximum security level required for any + * securityQuals now. Addition of inheritance-child RTEs won't affect + * this, because child tables don't have their own securityQuals; see + * expand_single_inheritance_child(). + */ + if (rte->securityQuals) + root->qual_security_level = Max(root->qual_security_level, + list_length(rte->securityQuals)); + } + + /* + * If we have now verified that the query target relation is + * non-inheriting, mark it as a leaf target. + */ + if (parse->resultRelation) + { + RangeTblEntry *rte = rt_fetch(parse->resultRelation, parse->rtable); + + if (!rte->inh) + root->leaf_result_relids = + bms_make_singleton(parse->resultRelation); + } + + /* + * Preprocess RowMark information. We need to do this after subquery + * pullup, so that all base relations are present. + */ + preprocess_rowmarks(root); + + /* + * Set hasHavingQual to remember if HAVING clause is present. Needed + * because preprocess_expression will reduce a constant-true condition to + * an empty qual list ... but "HAVING TRUE" is not a semantic no-op. + */ + root->hasHavingQual = (parse->havingQual != NULL); + + /* + * Do expression preprocessing on targetlist and quals, as well as other + * random expressions in the querytree. Note that we do not need to + * handle sort/group expressions explicitly, because they are actually + * part of the targetlist. + */ + parse->targetList = (List *) + preprocess_expression(root, (Node *) parse->targetList, + EXPRKIND_TARGET); + + /* Constant-folding might have removed all set-returning functions */ + if (parse->hasTargetSRFs) + parse->hasTargetSRFs = expression_returns_set((Node *) parse->targetList); + + newWithCheckOptions = NIL; + foreach(l, parse->withCheckOptions) + { + WithCheckOption *wco = lfirst_node(WithCheckOption, l); + + wco->qual = preprocess_expression(root, wco->qual, + EXPRKIND_QUAL); + if (wco->qual != NULL) + newWithCheckOptions = lappend(newWithCheckOptions, wco); + } + parse->withCheckOptions = newWithCheckOptions; + + parse->returningList = (List *) + preprocess_expression(root, (Node *) parse->returningList, + EXPRKIND_TARGET); + + preprocess_qual_conditions(root, (Node *) parse->jointree); + + parse->havingQual = preprocess_expression(root, parse->havingQual, + EXPRKIND_QUAL); + + foreach(l, parse->windowClause) + { + WindowClause *wc = lfirst_node(WindowClause, l); + + /* partitionClause/orderClause are sort/group expressions */ + wc->startOffset = preprocess_expression(root, wc->startOffset, + EXPRKIND_LIMIT); + wc->endOffset = preprocess_expression(root, wc->endOffset, + EXPRKIND_LIMIT); + wc->runCondition = (List *) preprocess_expression(root, + (Node *) wc->runCondition, + EXPRKIND_TARGET); + } + + parse->limitOffset = preprocess_expression(root, parse->limitOffset, + EXPRKIND_LIMIT); + parse->limitCount = preprocess_expression(root, parse->limitCount, + EXPRKIND_LIMIT); + + if (parse->onConflict) + { + parse->onConflict->arbiterElems = (List *) + preprocess_expression(root, + (Node *) parse->onConflict->arbiterElems, + EXPRKIND_ARBITER_ELEM); + parse->onConflict->arbiterWhere = + preprocess_expression(root, + parse->onConflict->arbiterWhere, + EXPRKIND_QUAL); + parse->onConflict->onConflictSet = (List *) + preprocess_expression(root, + (Node *) parse->onConflict->onConflictSet, + EXPRKIND_TARGET); + parse->onConflict->onConflictWhere = + preprocess_expression(root, + parse->onConflict->onConflictWhere, + EXPRKIND_QUAL); + /* exclRelTlist contains only Vars, so no preprocessing needed */ + } + + foreach(l, parse->mergeActionList) + { + MergeAction *action = (MergeAction *) lfirst(l); + + action->targetList = (List *) + preprocess_expression(root, + (Node *) action->targetList, + EXPRKIND_TARGET); + action->qual = + preprocess_expression(root, + (Node *) action->qual, + EXPRKIND_QUAL); + } + + root->append_rel_list = (List *) + preprocess_expression(root, (Node *) root->append_rel_list, + EXPRKIND_APPINFO); + + /* Also need to preprocess expressions within RTEs */ + foreach(l, parse->rtable) + { + RangeTblEntry *rte = lfirst_node(RangeTblEntry, l); + int kind; + ListCell *lcsq; + + if (rte->rtekind == RTE_RELATION) + { + if (rte->tablesample) + rte->tablesample = (TableSampleClause *) + preprocess_expression(root, + (Node *) rte->tablesample, + EXPRKIND_TABLESAMPLE); + } + else if (rte->rtekind == RTE_SUBQUERY) + { + /* + * We don't want to do all preprocessing yet on the subquery's + * expressions, since that will happen when we plan it. But if it + * contains any join aliases of our level, those have to get + * expanded now, because planning of the subquery won't do it. + * That's only possible if the subquery is LATERAL. + */ + if (rte->lateral && root->hasJoinRTEs) + rte->subquery = (Query *) + flatten_join_alias_vars(root->parse, + (Node *) rte->subquery); + } + else if (rte->rtekind == RTE_FUNCTION) + { + /* Preprocess the function expression(s) fully */ + kind = rte->lateral ? EXPRKIND_RTFUNC_LATERAL : EXPRKIND_RTFUNC; + rte->functions = (List *) + preprocess_expression(root, (Node *) rte->functions, kind); + } + else if (rte->rtekind == RTE_TABLEFUNC) + { + /* Preprocess the function expression(s) fully */ + kind = rte->lateral ? EXPRKIND_TABLEFUNC_LATERAL : EXPRKIND_TABLEFUNC; + rte->tablefunc = (TableFunc *) + preprocess_expression(root, (Node *) rte->tablefunc, kind); + } + else if (rte->rtekind == RTE_VALUES) + { + /* Preprocess the values lists fully */ + kind = rte->lateral ? EXPRKIND_VALUES_LATERAL : EXPRKIND_VALUES; + rte->values_lists = (List *) + preprocess_expression(root, (Node *) rte->values_lists, kind); + } + + /* + * Process each element of the securityQuals list as if it were a + * separate qual expression (as indeed it is). We need to do it this + * way to get proper canonicalization of AND/OR structure. Note that + * this converts each element into an implicit-AND sublist. + */ + foreach(lcsq, rte->securityQuals) + { + lfirst(lcsq) = preprocess_expression(root, + (Node *) lfirst(lcsq), + EXPRKIND_QUAL); + } + } + + /* + * Now that we are done preprocessing expressions, and in particular done + * flattening join alias variables, get rid of the joinaliasvars lists. + * They no longer match what expressions in the rest of the tree look + * like, because we have not preprocessed expressions in those lists (and + * do not want to; for example, expanding a SubLink there would result in + * a useless unreferenced subplan). Leaving them in place simply creates + * a hazard for later scans of the tree. We could try to prevent that by + * using QTW_IGNORE_JOINALIASES in every tree scan done after this point, + * but that doesn't sound very reliable. + */ + if (root->hasJoinRTEs) + { + foreach(l, parse->rtable) + { + RangeTblEntry *rte = lfirst_node(RangeTblEntry, l); + + rte->joinaliasvars = NIL; + } + } + + /* + * In some cases we may want to transfer a HAVING clause into WHERE. We + * cannot do so if the HAVING clause contains aggregates (obviously) or + * volatile functions (since a HAVING clause is supposed to be executed + * only once per group). We also can't do this if there are any nonempty + * grouping sets; moving such a clause into WHERE would potentially change + * the results, if any referenced column isn't present in all the grouping + * sets. (If there are only empty grouping sets, then the HAVING clause + * must be degenerate as discussed below.) + * + * Also, it may be that the clause is so expensive to execute that we're + * better off doing it only once per group, despite the loss of + * selectivity. This is hard to estimate short of doing the entire + * planning process twice, so we use a heuristic: clauses containing + * subplans are left in HAVING. Otherwise, we move or copy the HAVING + * clause into WHERE, in hopes of eliminating tuples before aggregation + * instead of after. + * + * If the query has explicit grouping then we can simply move such a + * clause into WHERE; any group that fails the clause will not be in the + * output because none of its tuples will reach the grouping or + * aggregation stage. Otherwise we must have a degenerate (variable-free) + * HAVING clause, which we put in WHERE so that query_planner() can use it + * in a gating Result node, but also keep in HAVING to ensure that we + * don't emit a bogus aggregated row. (This could be done better, but it + * seems not worth optimizing.) + * + * Note that both havingQual and parse->jointree->quals are in + * implicitly-ANDed-list form at this point, even though they are declared + * as Node *. + */ + newHaving = NIL; + foreach(l, (List *) parse->havingQual) + { + Node *havingclause = (Node *) lfirst(l); + + if ((parse->groupClause && parse->groupingSets) || + contain_agg_clause(havingclause) || + contain_volatile_functions(havingclause) || + contain_subplans(havingclause)) + { + /* keep it in HAVING */ + newHaving = lappend(newHaving, havingclause); + } + else if (parse->groupClause && !parse->groupingSets) + { + /* move it to WHERE */ + parse->jointree->quals = (Node *) + lappend((List *) parse->jointree->quals, havingclause); + } + else + { + /* put a copy in WHERE, keep it in HAVING */ + parse->jointree->quals = (Node *) + lappend((List *) parse->jointree->quals, + copyObject(havingclause)); + newHaving = lappend(newHaving, havingclause); + } + } + parse->havingQual = (Node *) newHaving; + + /* Remove any redundant GROUP BY columns */ + remove_useless_groupby_columns(root); + + /* + * If we have any outer joins, try to reduce them to plain inner joins. + * This step is most easily done after we've done expression + * preprocessing. + */ + if (hasOuterJoins) + reduce_outer_joins(root); + + /* + * If we have any RTE_RESULT relations, see if they can be deleted from + * the jointree. This step is most effectively done after we've done + * expression preprocessing and outer join reduction. + */ + if (hasResultRTEs) + remove_useless_result_rtes(root); + + /* + * Do the main planning. + */ + grouping_planner(root, tuple_fraction); + + /* + * Capture the set of outer-level param IDs we have access to, for use in + * extParam/allParam calculations later. + */ + SS_identify_outer_params(root); + + /* + * If any initPlans were created in this query level, adjust the surviving + * Paths' costs and parallel-safety flags to account for them. The + * initPlans won't actually get attached to the plan tree till + * create_plan() runs, but we must include their effects now. + */ + final_rel = fetch_upper_rel(root, UPPERREL_FINAL, NULL); + SS_charge_for_initplans(root, final_rel); + + /* + * Make sure we've identified the cheapest Path for the final rel. (By + * doing this here not in grouping_planner, we include initPlan costs in + * the decision, though it's unlikely that will change anything.) + */ + set_cheapest(final_rel); + + return root; +} + +/* + * preprocess_expression + * Do subquery_planner's preprocessing work for an expression, + * which can be a targetlist, a WHERE clause (including JOIN/ON + * conditions), a HAVING clause, or a few other things. + */ +static Node * +preprocess_expression(PlannerInfo *root, Node *expr, int kind) +{ + /* + * Fall out quickly if expression is empty. This occurs often enough to + * be worth checking. Note that null->null is the correct conversion for + * implicit-AND result format, too. + */ + if (expr == NULL) + return NULL; + + /* + * If the query has any join RTEs, replace join alias variables with + * base-relation variables. We must do this first, since any expressions + * we may extract from the joinaliasvars lists have not been preprocessed. + * For example, if we did this after sublink processing, sublinks expanded + * out from join aliases would not get processed. But we can skip this in + * non-lateral RTE functions, VALUES lists, and TABLESAMPLE clauses, since + * they can't contain any Vars of the current query level. + */ + if (root->hasJoinRTEs && + !(kind == EXPRKIND_RTFUNC || + kind == EXPRKIND_VALUES || + kind == EXPRKIND_TABLESAMPLE || + kind == EXPRKIND_TABLEFUNC)) + expr = flatten_join_alias_vars(root->parse, expr); + + /* + * Simplify constant expressions. For function RTEs, this was already + * done by preprocess_function_rtes. (But note we must do it again for + * EXPRKIND_RTFUNC_LATERAL, because those might by now contain + * un-simplified subexpressions inserted by flattening of subqueries or + * join alias variables.) + * + * Note: an essential effect of this is to convert named-argument function + * calls to positional notation and insert the current actual values of + * any default arguments for functions. To ensure that happens, we *must* + * process all expressions here. Previous PG versions sometimes skipped + * const-simplification if it didn't seem worth the trouble, but we can't + * do that anymore. + * + * Note: this also flattens nested AND and OR expressions into N-argument + * form. All processing of a qual expression after this point must be + * careful to maintain AND/OR flatness --- that is, do not generate a tree + * with AND directly under AND, nor OR directly under OR. + */ + if (kind != EXPRKIND_RTFUNC) + expr = eval_const_expressions(root, expr); + + /* + * If it's a qual or havingQual, canonicalize it. + */ + if (kind == EXPRKIND_QUAL) + { + expr = (Node *) canonicalize_qual((Expr *) expr, false); + +#ifdef OPTIMIZER_DEBUG + printf("After canonicalize_qual()\n"); + pprint(expr); +#endif + } + + /* + * Check for ANY ScalarArrayOpExpr with Const arrays and set the + * hashfuncid of any that might execute more quickly by using hash lookups + * instead of a linear search. + */ + if (kind == EXPRKIND_QUAL || kind == EXPRKIND_TARGET) + { + convert_saop_to_hashed_saop(expr); + } + + /* Expand SubLinks to SubPlans */ + if (root->parse->hasSubLinks) + expr = SS_process_sublinks(root, expr, (kind == EXPRKIND_QUAL)); + + /* + * XXX do not insert anything here unless you have grokked the comments in + * SS_replace_correlation_vars ... + */ + + /* Replace uplevel vars with Param nodes (this IS possible in VALUES) */ + if (root->query_level > 1) + expr = SS_replace_correlation_vars(root, expr); + + /* + * If it's a qual or havingQual, convert it to implicit-AND format. (We + * don't want to do this before eval_const_expressions, since the latter + * would be unable to simplify a top-level AND correctly. Also, + * SS_process_sublinks expects explicit-AND format.) + */ + if (kind == EXPRKIND_QUAL) + expr = (Node *) make_ands_implicit((Expr *) expr); + + return expr; +} + +/* + * preprocess_qual_conditions + * Recursively scan the query's jointree and do subquery_planner's + * preprocessing work on each qual condition found therein. + */ +static void +preprocess_qual_conditions(PlannerInfo *root, Node *jtnode) +{ + if (jtnode == NULL) + return; + if (IsA(jtnode, RangeTblRef)) + { + /* nothing to do here */ + } + else if (IsA(jtnode, FromExpr)) + { + FromExpr *f = (FromExpr *) jtnode; + ListCell *l; + + foreach(l, f->fromlist) + preprocess_qual_conditions(root, lfirst(l)); + + f->quals = preprocess_expression(root, f->quals, EXPRKIND_QUAL); + } + else if (IsA(jtnode, JoinExpr)) + { + JoinExpr *j = (JoinExpr *) jtnode; + + preprocess_qual_conditions(root, j->larg); + preprocess_qual_conditions(root, j->rarg); + + j->quals = preprocess_expression(root, j->quals, EXPRKIND_QUAL); + } + else + elog(ERROR, "unrecognized node type: %d", + (int) nodeTag(jtnode)); +} + +/* + * preprocess_phv_expression + * Do preprocessing on a PlaceHolderVar expression that's been pulled up. + * + * If a LATERAL subquery references an output of another subquery, and that + * output must be wrapped in a PlaceHolderVar because of an intermediate outer + * join, then we'll push the PlaceHolderVar expression down into the subquery + * and later pull it back up during find_lateral_references, which runs after + * subquery_planner has preprocessed all the expressions that were in the + * current query level to start with. So we need to preprocess it then. + */ +Expr * +preprocess_phv_expression(PlannerInfo *root, Expr *expr) +{ + return (Expr *) preprocess_expression(root, (Node *) expr, EXPRKIND_PHV); +} + +/*-------------------- + * grouping_planner + * Perform planning steps related to grouping, aggregation, etc. + * + * This function adds all required top-level processing to the scan/join + * Path(s) produced by query_planner. + * + * tuple_fraction is the fraction of tuples we expect will be retrieved. + * tuple_fraction is interpreted as follows: + * 0: expect all tuples to be retrieved (normal case) + * 0 < tuple_fraction < 1: expect the given fraction of tuples available + * from the plan to be retrieved + * tuple_fraction >= 1: tuple_fraction is the absolute number of tuples + * expected to be retrieved (ie, a LIMIT specification) + * + * Returns nothing; the useful output is in the Paths we attach to the + * (UPPERREL_FINAL, NULL) upperrel in *root. In addition, + * root->processed_tlist contains the final processed targetlist. + * + * Note that we have not done set_cheapest() on the final rel; it's convenient + * to leave this to the caller. + *-------------------- + */ +static void +grouping_planner(PlannerInfo *root, double tuple_fraction) +{ + Query *parse = root->parse; + int64 offset_est = 0; + int64 count_est = 0; + double limit_tuples = -1.0; + bool have_postponed_srfs = false; + PathTarget *final_target; + List *final_targets; + List *final_targets_contain_srfs; + bool final_target_parallel_safe; + RelOptInfo *current_rel; + RelOptInfo *final_rel; + FinalPathExtraData extra; + ListCell *lc; + + /* Tweak caller-supplied tuple_fraction if have LIMIT/OFFSET */ + if (parse->limitCount || parse->limitOffset) + { + tuple_fraction = preprocess_limit(root, tuple_fraction, + &offset_est, &count_est); + + /* + * If we have a known LIMIT, and don't have an unknown OFFSET, we can + * estimate the effects of using a bounded sort. + */ + if (count_est > 0 && offset_est >= 0) + limit_tuples = (double) count_est + (double) offset_est; + } + + /* Make tuple_fraction accessible to lower-level routines */ + root->tuple_fraction = tuple_fraction; + + if (parse->setOperations) + { + /* + * If there's a top-level ORDER BY, assume we have to fetch all the + * tuples. This might be too simplistic given all the hackery below + * to possibly avoid the sort; but the odds of accurate estimates here + * are pretty low anyway. XXX try to get rid of this in favor of + * letting plan_set_operations generate both fast-start and + * cheapest-total paths. + */ + if (parse->sortClause) + root->tuple_fraction = 0.0; + + /* + * Construct Paths for set operations. The results will not need any + * work except perhaps a top-level sort and/or LIMIT. Note that any + * special work for recursive unions is the responsibility of + * plan_set_operations. + */ + current_rel = plan_set_operations(root); + + /* + * We should not need to call preprocess_targetlist, since we must be + * in a SELECT query node. Instead, use the processed_tlist returned + * by plan_set_operations (since this tells whether it returned any + * resjunk columns!), and transfer any sort key information from the + * original tlist. + */ + Assert(parse->commandType == CMD_SELECT); + + /* for safety, copy processed_tlist instead of modifying in-place */ + root->processed_tlist = + postprocess_setop_tlist(copyObject(root->processed_tlist), + parse->targetList); + + /* Also extract the PathTarget form of the setop result tlist */ + final_target = current_rel->cheapest_total_path->pathtarget; + + /* And check whether it's parallel safe */ + final_target_parallel_safe = + is_parallel_safe(root, (Node *) final_target->exprs); + + /* The setop result tlist couldn't contain any SRFs */ + Assert(!parse->hasTargetSRFs); + final_targets = final_targets_contain_srfs = NIL; + + /* + * Can't handle FOR [KEY] UPDATE/SHARE here (parser should have + * checked already, but let's make sure). + */ + if (parse->rowMarks) + ereport(ERROR, + (errcode(ERRCODE_FEATURE_NOT_SUPPORTED), + /*------ + translator: %s is a SQL row locking clause such as FOR UPDATE */ + errmsg("%s is not allowed with UNION/INTERSECT/EXCEPT", + LCS_asString(linitial_node(RowMarkClause, + parse->rowMarks)->strength)))); + + /* + * Calculate pathkeys that represent result ordering requirements + */ + Assert(parse->distinctClause == NIL); + root->sort_pathkeys = make_pathkeys_for_sortclauses(root, + parse->sortClause, + root->processed_tlist); + } + else + { + /* No set operations, do regular planning */ + PathTarget *sort_input_target; + List *sort_input_targets; + List *sort_input_targets_contain_srfs; + bool sort_input_target_parallel_safe; + PathTarget *grouping_target; + List *grouping_targets; + List *grouping_targets_contain_srfs; + bool grouping_target_parallel_safe; + PathTarget *scanjoin_target; + List *scanjoin_targets; + List *scanjoin_targets_contain_srfs; + bool scanjoin_target_parallel_safe; + bool scanjoin_target_same_exprs; + bool have_grouping; + WindowFuncLists *wflists = NULL; + List *activeWindows = NIL; + grouping_sets_data *gset_data = NULL; + standard_qp_extra qp_extra; + + /* A recursive query should always have setOperations */ + Assert(!root->hasRecursion); + + /* Preprocess grouping sets and GROUP BY clause, if any */ + if (parse->groupingSets) + { + gset_data = preprocess_grouping_sets(root); + } + else + { + /* Preprocess regular GROUP BY clause, if any */ + if (parse->groupClause) + parse->groupClause = preprocess_groupclause(root, NIL); + } + + /* + * Preprocess targetlist. Note that much of the remaining planning + * work will be done with the PathTarget representation of tlists, but + * we must also maintain the full representation of the final tlist so + * that we can transfer its decoration (resnames etc) to the topmost + * tlist of the finished Plan. This is kept in processed_tlist. + */ + preprocess_targetlist(root); + + /* + * Mark all the aggregates with resolved aggtranstypes, and detect + * aggregates that are duplicates or can share transition state. We + * must do this before slicing and dicing the tlist into various + * pathtargets, else some copies of the Aggref nodes might escape + * being marked. + */ + if (parse->hasAggs) + { + preprocess_aggrefs(root, (Node *) root->processed_tlist); + preprocess_aggrefs(root, (Node *) parse->havingQual); + } + + /* + * Locate any window functions in the tlist. (We don't need to look + * anywhere else, since expressions used in ORDER BY will be in there + * too.) Note that they could all have been eliminated by constant + * folding, in which case we don't need to do any more work. + */ + if (parse->hasWindowFuncs) + { + wflists = find_window_functions((Node *) root->processed_tlist, + list_length(parse->windowClause)); + if (wflists->numWindowFuncs > 0) + activeWindows = select_active_windows(root, wflists); + else + parse->hasWindowFuncs = false; + } + + /* + * Preprocess MIN/MAX aggregates, if any. Note: be careful about + * adding logic between here and the query_planner() call. Anything + * that is needed in MIN/MAX-optimizable cases will have to be + * duplicated in planagg.c. + */ + if (parse->hasAggs) + preprocess_minmax_aggregates(root); + + /* + * Figure out whether there's a hard limit on the number of rows that + * query_planner's result subplan needs to return. Even if we know a + * hard limit overall, it doesn't apply if the query has any + * grouping/aggregation operations, or SRFs in the tlist. + */ + if (parse->groupClause || + parse->groupingSets || + parse->distinctClause || + parse->hasAggs || + parse->hasWindowFuncs || + parse->hasTargetSRFs || + root->hasHavingQual) + root->limit_tuples = -1.0; + else + root->limit_tuples = limit_tuples; + + /* Set up data needed by standard_qp_callback */ + qp_extra.activeWindows = activeWindows; + qp_extra.groupClause = (gset_data + ? (gset_data->rollups ? linitial_node(RollupData, gset_data->rollups)->groupClause : NIL) + : parse->groupClause); + + /* + * Generate the best unsorted and presorted paths for the scan/join + * portion of this Query, ie the processing represented by the + * FROM/WHERE clauses. (Note there may not be any presorted paths.) + * We also generate (in standard_qp_callback) pathkey representations + * of the query's sort clause, distinct clause, etc. + */ + current_rel = query_planner(root, standard_qp_callback, &qp_extra); + + /* + * Convert the query's result tlist into PathTarget format. + * + * Note: this cannot be done before query_planner() has performed + * appendrel expansion, because that might add resjunk entries to + * root->processed_tlist. Waiting till afterwards is also helpful + * because the target width estimates can use per-Var width numbers + * that were obtained within query_planner(). + */ + final_target = create_pathtarget(root, root->processed_tlist); + final_target_parallel_safe = + is_parallel_safe(root, (Node *) final_target->exprs); + + /* + * If ORDER BY was given, consider whether we should use a post-sort + * projection, and compute the adjusted target for preceding steps if + * so. + */ + if (parse->sortClause) + { + sort_input_target = make_sort_input_target(root, + final_target, + &have_postponed_srfs); + sort_input_target_parallel_safe = + is_parallel_safe(root, (Node *) sort_input_target->exprs); + } + else + { + sort_input_target = final_target; + sort_input_target_parallel_safe = final_target_parallel_safe; + } + + /* + * If we have window functions to deal with, the output from any + * grouping step needs to be what the window functions want; + * otherwise, it should be sort_input_target. + */ + if (activeWindows) + { + grouping_target = make_window_input_target(root, + final_target, + activeWindows); + grouping_target_parallel_safe = + is_parallel_safe(root, (Node *) grouping_target->exprs); + } + else + { + grouping_target = sort_input_target; + grouping_target_parallel_safe = sort_input_target_parallel_safe; + } + + /* + * If we have grouping or aggregation to do, the topmost scan/join + * plan node must emit what the grouping step wants; otherwise, it + * should emit grouping_target. + */ + have_grouping = (parse->groupClause || parse->groupingSets || + parse->hasAggs || root->hasHavingQual); + if (have_grouping) + { + scanjoin_target = make_group_input_target(root, final_target); + scanjoin_target_parallel_safe = + is_parallel_safe(root, (Node *) scanjoin_target->exprs); + } + else + { + scanjoin_target = grouping_target; + scanjoin_target_parallel_safe = grouping_target_parallel_safe; + } + + /* + * If there are any SRFs in the targetlist, we must separate each of + * these PathTargets into SRF-computing and SRF-free targets. Replace + * each of the named targets with a SRF-free version, and remember the + * list of additional projection steps we need to add afterwards. + */ + if (parse->hasTargetSRFs) + { + /* final_target doesn't recompute any SRFs in sort_input_target */ + split_pathtarget_at_srfs(root, final_target, sort_input_target, + &final_targets, + &final_targets_contain_srfs); + final_target = linitial_node(PathTarget, final_targets); + Assert(!linitial_int(final_targets_contain_srfs)); + /* likewise for sort_input_target vs. grouping_target */ + split_pathtarget_at_srfs(root, sort_input_target, grouping_target, + &sort_input_targets, + &sort_input_targets_contain_srfs); + sort_input_target = linitial_node(PathTarget, sort_input_targets); + Assert(!linitial_int(sort_input_targets_contain_srfs)); + /* likewise for grouping_target vs. scanjoin_target */ + split_pathtarget_at_srfs(root, grouping_target, scanjoin_target, + &grouping_targets, + &grouping_targets_contain_srfs); + grouping_target = linitial_node(PathTarget, grouping_targets); + Assert(!linitial_int(grouping_targets_contain_srfs)); + /* scanjoin_target will not have any SRFs precomputed for it */ + split_pathtarget_at_srfs(root, scanjoin_target, NULL, + &scanjoin_targets, + &scanjoin_targets_contain_srfs); + scanjoin_target = linitial_node(PathTarget, scanjoin_targets); + Assert(!linitial_int(scanjoin_targets_contain_srfs)); + } + else + { + /* initialize lists; for most of these, dummy values are OK */ + final_targets = final_targets_contain_srfs = NIL; + sort_input_targets = sort_input_targets_contain_srfs = NIL; + grouping_targets = grouping_targets_contain_srfs = NIL; + scanjoin_targets = list_make1(scanjoin_target); + scanjoin_targets_contain_srfs = NIL; + } + + /* Apply scan/join target. */ + scanjoin_target_same_exprs = list_length(scanjoin_targets) == 1 + && equal(scanjoin_target->exprs, current_rel->reltarget->exprs); + apply_scanjoin_target_to_paths(root, current_rel, scanjoin_targets, + scanjoin_targets_contain_srfs, + scanjoin_target_parallel_safe, + scanjoin_target_same_exprs); + + /* + * Save the various upper-rel PathTargets we just computed into + * root->upper_targets[]. The core code doesn't use this, but it + * provides a convenient place for extensions to get at the info. For + * consistency, we save all the intermediate targets, even though some + * of the corresponding upperrels might not be needed for this query. + */ + root->upper_targets[UPPERREL_FINAL] = final_target; + root->upper_targets[UPPERREL_ORDERED] = final_target; + root->upper_targets[UPPERREL_PARTIAL_DISTINCT] = sort_input_target; + root->upper_targets[UPPERREL_DISTINCT] = sort_input_target; + root->upper_targets[UPPERREL_WINDOW] = sort_input_target; + root->upper_targets[UPPERREL_GROUP_AGG] = grouping_target; + + /* + * If we have grouping and/or aggregation, consider ways to implement + * that. We build a new upperrel representing the output of this + * phase. + */ + if (have_grouping) + { + current_rel = create_grouping_paths(root, + current_rel, + grouping_target, + grouping_target_parallel_safe, + gset_data); + /* Fix things up if grouping_target contains SRFs */ + if (parse->hasTargetSRFs) + adjust_paths_for_srfs(root, current_rel, + grouping_targets, + grouping_targets_contain_srfs); + } + + /* + * If we have window functions, consider ways to implement those. We + * build a new upperrel representing the output of this phase. + */ + if (activeWindows) + { + current_rel = create_window_paths(root, + current_rel, + grouping_target, + sort_input_target, + sort_input_target_parallel_safe, + wflists, + activeWindows); + /* Fix things up if sort_input_target contains SRFs */ + if (parse->hasTargetSRFs) + adjust_paths_for_srfs(root, current_rel, + sort_input_targets, + sort_input_targets_contain_srfs); + } + + /* + * If there is a DISTINCT clause, consider ways to implement that. We + * build a new upperrel representing the output of this phase. + */ + if (parse->distinctClause) + { + current_rel = create_distinct_paths(root, + current_rel); + } + } /* end of if (setOperations) */ + + /* + * If ORDER BY was given, consider ways to implement that, and generate a + * new upperrel containing only paths that emit the correct ordering and + * project the correct final_target. We can apply the original + * limit_tuples limit in sort costing here, but only if there are no + * postponed SRFs. + */ + if (parse->sortClause) + { + current_rel = create_ordered_paths(root, + current_rel, + final_target, + final_target_parallel_safe, + have_postponed_srfs ? -1.0 : + limit_tuples); + /* Fix things up if final_target contains SRFs */ + if (parse->hasTargetSRFs) + adjust_paths_for_srfs(root, current_rel, + final_targets, + final_targets_contain_srfs); + } + + /* + * Now we are prepared to build the final-output upperrel. + */ + final_rel = fetch_upper_rel(root, UPPERREL_FINAL, NULL); + + /* + * If the input rel is marked consider_parallel and there's nothing that's + * not parallel-safe in the LIMIT clause, then the final_rel can be marked + * consider_parallel as well. Note that if the query has rowMarks or is + * not a SELECT, consider_parallel will be false for every relation in the + * query. + */ + if (current_rel->consider_parallel && + is_parallel_safe(root, parse->limitOffset) && + is_parallel_safe(root, parse->limitCount)) + final_rel->consider_parallel = true; + + /* + * If the current_rel belongs to a single FDW, so does the final_rel. + */ + final_rel->serverid = current_rel->serverid; + final_rel->userid = current_rel->userid; + final_rel->useridiscurrent = current_rel->useridiscurrent; + final_rel->fdwroutine = current_rel->fdwroutine; + + /* + * Generate paths for the final_rel. Insert all surviving paths, with + * LockRows, Limit, and/or ModifyTable steps added if needed. + */ + foreach(lc, current_rel->pathlist) + { + Path *path = (Path *) lfirst(lc); + + /* + * If there is a FOR [KEY] UPDATE/SHARE clause, add the LockRows node. + * (Note: we intentionally test parse->rowMarks not root->rowMarks + * here. If there are only non-locking rowmarks, they should be + * handled by the ModifyTable node instead. However, root->rowMarks + * is what goes into the LockRows node.) + */ + if (parse->rowMarks) + { + path = (Path *) create_lockrows_path(root, final_rel, path, + root->rowMarks, + assign_special_exec_param(root)); + } + + /* + * If there is a LIMIT/OFFSET clause, add the LIMIT node. + */ + if (limit_needed(parse)) + { + path = (Path *) create_limit_path(root, final_rel, path, + parse->limitOffset, + parse->limitCount, + parse->limitOption, + offset_est, count_est); + } + + /* + * If this is an INSERT/UPDATE/DELETE/MERGE, add the ModifyTable node. + */ + if (parse->commandType != CMD_SELECT) + { + Index rootRelation; + List *resultRelations = NIL; + List *updateColnosLists = NIL; + List *withCheckOptionLists = NIL; + List *returningLists = NIL; + List *mergeActionLists = NIL; + List *rowMarks; + + if (bms_membership(root->all_result_relids) == BMS_MULTIPLE) + { + /* Inherited UPDATE/DELETE/MERGE */ + RelOptInfo *top_result_rel = find_base_rel(root, + parse->resultRelation); + int resultRelation = -1; + + /* Pass the root result rel forward to the executor. */ + rootRelation = parse->resultRelation; + + /* Add only leaf children to ModifyTable. */ + while ((resultRelation = bms_next_member(root->leaf_result_relids, + resultRelation)) >= 0) + { + RelOptInfo *this_result_rel = find_base_rel(root, + resultRelation); + + /* + * Also exclude any leaf rels that have turned dummy since + * being added to the list, for example, by being excluded + * by constraint exclusion. + */ + if (IS_DUMMY_REL(this_result_rel)) + continue; + + /* Build per-target-rel lists needed by ModifyTable */ + resultRelations = lappend_int(resultRelations, + resultRelation); + if (parse->commandType == CMD_UPDATE) + { + List *update_colnos = root->update_colnos; + + if (this_result_rel != top_result_rel) + update_colnos = + adjust_inherited_attnums_multilevel(root, + update_colnos, + this_result_rel->relid, + top_result_rel->relid); + updateColnosLists = lappend(updateColnosLists, + update_colnos); + } + if (parse->withCheckOptions) + { + List *withCheckOptions = parse->withCheckOptions; + + if (this_result_rel != top_result_rel) + withCheckOptions = (List *) + adjust_appendrel_attrs_multilevel(root, + (Node *) withCheckOptions, + this_result_rel->relids, + top_result_rel->relids); + withCheckOptionLists = lappend(withCheckOptionLists, + withCheckOptions); + } + if (parse->returningList) + { + List *returningList = parse->returningList; + + if (this_result_rel != top_result_rel) + returningList = (List *) + adjust_appendrel_attrs_multilevel(root, + (Node *) returningList, + this_result_rel->relids, + top_result_rel->relids); + returningLists = lappend(returningLists, + returningList); + } + if (parse->mergeActionList) + { + ListCell *l; + List *mergeActionList = NIL; + + /* + * Copy MergeActions and translate stuff that + * references attribute numbers. + */ + foreach(l, parse->mergeActionList) + { + MergeAction *action = lfirst(l), + *leaf_action = copyObject(action); + + leaf_action->qual = + adjust_appendrel_attrs_multilevel(root, + (Node *) action->qual, + this_result_rel->relids, + top_result_rel->relids); + leaf_action->targetList = (List *) + adjust_appendrel_attrs_multilevel(root, + (Node *) action->targetList, + this_result_rel->relids, + top_result_rel->relids); + if (leaf_action->commandType == CMD_UPDATE) + leaf_action->updateColnos = + adjust_inherited_attnums_multilevel(root, + action->updateColnos, + this_result_rel->relid, + top_result_rel->relid); + mergeActionList = lappend(mergeActionList, + leaf_action); + } + + mergeActionLists = lappend(mergeActionLists, + mergeActionList); + } + } + + if (resultRelations == NIL) + { + /* + * We managed to exclude every child rel, so generate a + * dummy one-relation plan using info for the top target + * rel (even though that may not be a leaf target). + * Although it's clear that no data will be updated or + * deleted, we still need to have a ModifyTable node so + * that any statement triggers will be executed. (This + * could be cleaner if we fixed nodeModifyTable.c to allow + * zero target relations, but that probably wouldn't be a + * net win.) + */ + resultRelations = list_make1_int(parse->resultRelation); + if (parse->commandType == CMD_UPDATE) + updateColnosLists = list_make1(root->update_colnos); + if (parse->withCheckOptions) + withCheckOptionLists = list_make1(parse->withCheckOptions); + if (parse->returningList) + returningLists = list_make1(parse->returningList); + if (parse->mergeActionList) + mergeActionLists = list_make1(parse->mergeActionList); + } + } + else + { + /* Single-relation INSERT/UPDATE/DELETE/MERGE. */ + rootRelation = 0; /* there's no separate root rel */ + resultRelations = list_make1_int(parse->resultRelation); + if (parse->commandType == CMD_UPDATE) + updateColnosLists = list_make1(root->update_colnos); + if (parse->withCheckOptions) + withCheckOptionLists = list_make1(parse->withCheckOptions); + if (parse->returningList) + returningLists = list_make1(parse->returningList); + if (parse->mergeActionList) + mergeActionLists = list_make1(parse->mergeActionList); + } + + /* + * If there was a FOR [KEY] UPDATE/SHARE clause, the LockRows node + * will have dealt with fetching non-locked marked rows, else we + * need to have ModifyTable do that. + */ + if (parse->rowMarks) + rowMarks = NIL; + else + rowMarks = root->rowMarks; + + path = (Path *) + create_modifytable_path(root, final_rel, + path, + parse->commandType, + parse->canSetTag, + parse->resultRelation, + rootRelation, + root->partColsUpdated, + resultRelations, + updateColnosLists, + withCheckOptionLists, + returningLists, + rowMarks, + parse->onConflict, + mergeActionLists, + assign_special_exec_param(root)); + } + + /* And shove it into final_rel */ + add_path(final_rel, path); + } + + /* + * Generate partial paths for final_rel, too, if outer query levels might + * be able to make use of them. + */ + if (final_rel->consider_parallel && root->query_level > 1 && + !limit_needed(parse)) + { + Assert(!parse->rowMarks && parse->commandType == CMD_SELECT); + foreach(lc, current_rel->partial_pathlist) + { + Path *partial_path = (Path *) lfirst(lc); + + add_partial_path(final_rel, partial_path); + } + } + + extra.limit_needed = limit_needed(parse); + extra.limit_tuples = limit_tuples; + extra.count_est = count_est; + extra.offset_est = offset_est; + + /* + * If there is an FDW that's responsible for all baserels of the query, + * let it consider adding ForeignPaths. + */ + if (final_rel->fdwroutine && + final_rel->fdwroutine->GetForeignUpperPaths) + final_rel->fdwroutine->GetForeignUpperPaths(root, UPPERREL_FINAL, + current_rel, final_rel, + &extra); + + /* Let extensions possibly add some more paths */ + if (create_upper_paths_hook) + (*create_upper_paths_hook) (root, UPPERREL_FINAL, + current_rel, final_rel, &extra); + + /* Note: currently, we leave it to callers to do set_cheapest() */ +} + +/* + * Do preprocessing for groupingSets clause and related data. This handles the + * preliminary steps of expanding the grouping sets, organizing them into lists + * of rollups, and preparing annotations which will later be filled in with + * size estimates. + */ +static grouping_sets_data * +preprocess_grouping_sets(PlannerInfo *root) +{ + Query *parse = root->parse; + List *sets; + int maxref = 0; + ListCell *lc; + ListCell *lc_set; + grouping_sets_data *gd = palloc0(sizeof(grouping_sets_data)); + + parse->groupingSets = expand_grouping_sets(parse->groupingSets, parse->groupDistinct, -1); + + gd->any_hashable = false; + gd->unhashable_refs = NULL; + gd->unsortable_refs = NULL; + gd->unsortable_sets = NIL; + + if (parse->groupClause) + { + ListCell *lc; + + foreach(lc, parse->groupClause) + { + SortGroupClause *gc = lfirst_node(SortGroupClause, lc); + Index ref = gc->tleSortGroupRef; + + if (ref > maxref) + maxref = ref; + + if (!gc->hashable) + gd->unhashable_refs = bms_add_member(gd->unhashable_refs, ref); + + if (!OidIsValid(gc->sortop)) + gd->unsortable_refs = bms_add_member(gd->unsortable_refs, ref); + } + } + + /* Allocate workspace array for remapping */ + gd->tleref_to_colnum_map = (int *) palloc((maxref + 1) * sizeof(int)); + + /* + * If we have any unsortable sets, we must extract them before trying to + * prepare rollups. Unsortable sets don't go through + * reorder_grouping_sets, so we must apply the GroupingSetData annotation + * here. + */ + if (!bms_is_empty(gd->unsortable_refs)) + { + List *sortable_sets = NIL; + + foreach(lc, parse->groupingSets) + { + List *gset = (List *) lfirst(lc); + + if (bms_overlap_list(gd->unsortable_refs, gset)) + { + GroupingSetData *gs = makeNode(GroupingSetData); + + gs->set = gset; + gd->unsortable_sets = lappend(gd->unsortable_sets, gs); + + /* + * We must enforce here that an unsortable set is hashable; + * later code assumes this. Parse analysis only checks that + * every individual column is either hashable or sortable. + * + * Note that passing this test doesn't guarantee we can + * generate a plan; there might be other showstoppers. + */ + if (bms_overlap_list(gd->unhashable_refs, gset)) + ereport(ERROR, + (errcode(ERRCODE_FEATURE_NOT_SUPPORTED), + errmsg("could not implement GROUP BY"), + errdetail("Some of the datatypes only support hashing, while others only support sorting."))); + } + else + sortable_sets = lappend(sortable_sets, gset); + } + + if (sortable_sets) + sets = extract_rollup_sets(sortable_sets); + else + sets = NIL; + } + else + sets = extract_rollup_sets(parse->groupingSets); + + foreach(lc_set, sets) + { + List *current_sets = (List *) lfirst(lc_set); + RollupData *rollup = makeNode(RollupData); + GroupingSetData *gs; + + /* + * Reorder the current list of grouping sets into correct prefix + * order. If only one aggregation pass is needed, try to make the + * list match the ORDER BY clause; if more than one pass is needed, we + * don't bother with that. + * + * Note that this reorders the sets from smallest-member-first to + * largest-member-first, and applies the GroupingSetData annotations, + * though the data will be filled in later. + */ + current_sets = reorder_grouping_sets(current_sets, + (list_length(sets) == 1 + ? parse->sortClause + : NIL)); + + /* + * Get the initial (and therefore largest) grouping set. + */ + gs = linitial_node(GroupingSetData, current_sets); + + /* + * Order the groupClause appropriately. If the first grouping set is + * empty, then the groupClause must also be empty; otherwise we have + * to force the groupClause to match that grouping set's order. + * + * (The first grouping set can be empty even though parse->groupClause + * is not empty only if all non-empty grouping sets are unsortable. + * The groupClauses for hashed grouping sets are built later on.) + */ + if (gs->set) + rollup->groupClause = preprocess_groupclause(root, gs->set); + else + rollup->groupClause = NIL; + + /* + * Is it hashable? We pretend empty sets are hashable even though we + * actually force them not to be hashed later. But don't bother if + * there's nothing but empty sets (since in that case we can't hash + * anything). + */ + if (gs->set && + !bms_overlap_list(gd->unhashable_refs, gs->set)) + { + rollup->hashable = true; + gd->any_hashable = true; + } + + /* + * Now that we've pinned down an order for the groupClause for this + * list of grouping sets, we need to remap the entries in the grouping + * sets from sortgrouprefs to plain indices (0-based) into the + * groupClause for this collection of grouping sets. We keep the + * original form for later use, though. + */ + rollup->gsets = remap_to_groupclause_idx(rollup->groupClause, + current_sets, + gd->tleref_to_colnum_map); + rollup->gsets_data = current_sets; + + gd->rollups = lappend(gd->rollups, rollup); + } + + if (gd->unsortable_sets) + { + /* + * We have not yet pinned down a groupclause for this, but we will + * need index-based lists for estimation purposes. Construct + * hash_sets_idx based on the entire original groupclause for now. + */ + gd->hash_sets_idx = remap_to_groupclause_idx(parse->groupClause, + gd->unsortable_sets, + gd->tleref_to_colnum_map); + gd->any_hashable = true; + } + + return gd; +} + +/* + * Given a groupclause and a list of GroupingSetData, return equivalent sets + * (without annotation) mapped to indexes into the given groupclause. + */ +static List * +remap_to_groupclause_idx(List *groupClause, + List *gsets, + int *tleref_to_colnum_map) +{ + int ref = 0; + List *result = NIL; + ListCell *lc; + + foreach(lc, groupClause) + { + SortGroupClause *gc = lfirst_node(SortGroupClause, lc); + + tleref_to_colnum_map[gc->tleSortGroupRef] = ref++; + } + + foreach(lc, gsets) + { + List *set = NIL; + ListCell *lc2; + GroupingSetData *gs = lfirst_node(GroupingSetData, lc); + + foreach(lc2, gs->set) + { + set = lappend_int(set, tleref_to_colnum_map[lfirst_int(lc2)]); + } + + result = lappend(result, set); + } + + return result; +} + + +/* + * preprocess_rowmarks - set up PlanRowMarks if needed + */ +static void +preprocess_rowmarks(PlannerInfo *root) +{ + Query *parse = root->parse; + Bitmapset *rels; + List *prowmarks; + ListCell *l; + int i; + + if (parse->rowMarks) + { + /* + * We've got trouble if FOR [KEY] UPDATE/SHARE appears inside + * grouping, since grouping renders a reference to individual tuple + * CTIDs invalid. This is also checked at parse time, but that's + * insufficient because of rule substitution, query pullup, etc. + */ + CheckSelectLocking(parse, linitial_node(RowMarkClause, + parse->rowMarks)->strength); + } + else + { + /* + * We only need rowmarks for UPDATE, DELETE, MERGE, or FOR [KEY] + * UPDATE/SHARE. + */ + if (parse->commandType != CMD_UPDATE && + parse->commandType != CMD_DELETE && + parse->commandType != CMD_MERGE) + return; + } + + /* + * We need to have rowmarks for all base relations except the target. We + * make a bitmapset of all base rels and then remove the items we don't + * need or have FOR [KEY] UPDATE/SHARE marks for. + */ + rels = get_relids_in_jointree((Node *) parse->jointree, false); + if (parse->resultRelation) + rels = bms_del_member(rels, parse->resultRelation); + + /* + * Convert RowMarkClauses to PlanRowMark representation. + */ + prowmarks = NIL; + foreach(l, parse->rowMarks) + { + RowMarkClause *rc = lfirst_node(RowMarkClause, l); + RangeTblEntry *rte = rt_fetch(rc->rti, parse->rtable); + PlanRowMark *newrc; + + /* + * Currently, it is syntactically impossible to have FOR UPDATE et al + * applied to an update/delete target rel. If that ever becomes + * possible, we should drop the target from the PlanRowMark list. + */ + Assert(rc->rti != parse->resultRelation); + + /* + * Ignore RowMarkClauses for subqueries; they aren't real tables and + * can't support true locking. Subqueries that got flattened into the + * main query should be ignored completely. Any that didn't will get + * ROW_MARK_COPY items in the next loop. + */ + if (rte->rtekind != RTE_RELATION) + continue; + + rels = bms_del_member(rels, rc->rti); + + newrc = makeNode(PlanRowMark); + newrc->rti = newrc->prti = rc->rti; + newrc->rowmarkId = ++(root->glob->lastRowMarkId); + newrc->markType = select_rowmark_type(rte, rc->strength); + newrc->allMarkTypes = (1 << newrc->markType); + newrc->strength = rc->strength; + newrc->waitPolicy = rc->waitPolicy; + newrc->isParent = false; + + prowmarks = lappend(prowmarks, newrc); + } + + /* + * Now, add rowmarks for any non-target, non-locked base relations. + */ + i = 0; + foreach(l, parse->rtable) + { + RangeTblEntry *rte = lfirst_node(RangeTblEntry, l); + PlanRowMark *newrc; + + i++; + if (!bms_is_member(i, rels)) + continue; + + newrc = makeNode(PlanRowMark); + newrc->rti = newrc->prti = i; + newrc->rowmarkId = ++(root->glob->lastRowMarkId); + newrc->markType = select_rowmark_type(rte, LCS_NONE); + newrc->allMarkTypes = (1 << newrc->markType); + newrc->strength = LCS_NONE; + newrc->waitPolicy = LockWaitBlock; /* doesn't matter */ + newrc->isParent = false; + + prowmarks = lappend(prowmarks, newrc); + } + + root->rowMarks = prowmarks; +} + +/* + * Select RowMarkType to use for a given table + */ +RowMarkType +select_rowmark_type(RangeTblEntry *rte, LockClauseStrength strength) +{ + if (rte->rtekind != RTE_RELATION) + { + /* If it's not a table at all, use ROW_MARK_COPY */ + return ROW_MARK_COPY; + } + else if (rte->relkind == RELKIND_FOREIGN_TABLE) + { + /* Let the FDW select the rowmark type, if it wants to */ + FdwRoutine *fdwroutine = GetFdwRoutineByRelId(rte->relid); + + if (fdwroutine->GetForeignRowMarkType != NULL) + return fdwroutine->GetForeignRowMarkType(rte, strength); + /* Otherwise, use ROW_MARK_COPY by default */ + return ROW_MARK_COPY; + } + else + { + /* Regular table, apply the appropriate lock type */ + switch (strength) + { + case LCS_NONE: + + /* + * We don't need a tuple lock, only the ability to re-fetch + * the row. + */ + return ROW_MARK_REFERENCE; + break; + case LCS_FORKEYSHARE: + return ROW_MARK_KEYSHARE; + break; + case LCS_FORSHARE: + return ROW_MARK_SHARE; + break; + case LCS_FORNOKEYUPDATE: + return ROW_MARK_NOKEYEXCLUSIVE; + break; + case LCS_FORUPDATE: + return ROW_MARK_EXCLUSIVE; + break; + } + elog(ERROR, "unrecognized LockClauseStrength %d", (int) strength); + return ROW_MARK_EXCLUSIVE; /* keep compiler quiet */ + } +} + +/* + * preprocess_limit - do pre-estimation for LIMIT and/or OFFSET clauses + * + * We try to estimate the values of the LIMIT/OFFSET clauses, and pass the + * results back in *count_est and *offset_est. These variables are set to + * 0 if the corresponding clause is not present, and -1 if it's present + * but we couldn't estimate the value for it. (The "0" convention is OK + * for OFFSET but a little bit bogus for LIMIT: effectively we estimate + * LIMIT 0 as though it were LIMIT 1. But this is in line with the planner's + * usual practice of never estimating less than one row.) These values will + * be passed to create_limit_path, which see if you change this code. + * + * The return value is the suitably adjusted tuple_fraction to use for + * planning the query. This adjustment is not overridable, since it reflects + * plan actions that grouping_planner() will certainly take, not assumptions + * about context. + */ +static double +preprocess_limit(PlannerInfo *root, double tuple_fraction, + int64 *offset_est, int64 *count_est) +{ + Query *parse = root->parse; + Node *est; + double limit_fraction; + + /* Should not be called unless LIMIT or OFFSET */ + Assert(parse->limitCount || parse->limitOffset); + + /* + * Try to obtain the clause values. We use estimate_expression_value + * primarily because it can sometimes do something useful with Params. + */ + if (parse->limitCount) + { + est = estimate_expression_value(root, parse->limitCount); + if (est && IsA(est, Const)) + { + if (((Const *) est)->constisnull) + { + /* NULL indicates LIMIT ALL, ie, no limit */ + *count_est = 0; /* treat as not present */ + } + else + { + *count_est = DatumGetInt64(((Const *) est)->constvalue); + if (*count_est <= 0) + *count_est = 1; /* force to at least 1 */ + } + } + else + *count_est = -1; /* can't estimate */ + } + else + *count_est = 0; /* not present */ + + if (parse->limitOffset) + { + est = estimate_expression_value(root, parse->limitOffset); + if (est && IsA(est, Const)) + { + if (((Const *) est)->constisnull) + { + /* Treat NULL as no offset; the executor will too */ + *offset_est = 0; /* treat as not present */ + } + else + { + *offset_est = DatumGetInt64(((Const *) est)->constvalue); + if (*offset_est < 0) + *offset_est = 0; /* treat as not present */ + } + } + else + *offset_est = -1; /* can't estimate */ + } + else + *offset_est = 0; /* not present */ + + if (*count_est != 0) + { + /* + * A LIMIT clause limits the absolute number of tuples returned. + * However, if it's not a constant LIMIT then we have to guess; for + * lack of a better idea, assume 10% of the plan's result is wanted. + */ + if (*count_est < 0 || *offset_est < 0) + { + /* LIMIT or OFFSET is an expression ... punt ... */ + limit_fraction = 0.10; + } + else + { + /* LIMIT (plus OFFSET, if any) is max number of tuples needed */ + limit_fraction = (double) *count_est + (double) *offset_est; + } + + /* + * If we have absolute limits from both caller and LIMIT, use the + * smaller value; likewise if they are both fractional. If one is + * fractional and the other absolute, we can't easily determine which + * is smaller, but we use the heuristic that the absolute will usually + * be smaller. + */ + if (tuple_fraction >= 1.0) + { + if (limit_fraction >= 1.0) + { + /* both absolute */ + tuple_fraction = Min(tuple_fraction, limit_fraction); + } + else + { + /* caller absolute, limit fractional; use caller's value */ + } + } + else if (tuple_fraction > 0.0) + { + if (limit_fraction >= 1.0) + { + /* caller fractional, limit absolute; use limit */ + tuple_fraction = limit_fraction; + } + else + { + /* both fractional */ + tuple_fraction = Min(tuple_fraction, limit_fraction); + } + } + else + { + /* no info from caller, just use limit */ + tuple_fraction = limit_fraction; + } + } + else if (*offset_est != 0 && tuple_fraction > 0.0) + { + /* + * We have an OFFSET but no LIMIT. This acts entirely differently + * from the LIMIT case: here, we need to increase rather than decrease + * the caller's tuple_fraction, because the OFFSET acts to cause more + * tuples to be fetched instead of fewer. This only matters if we got + * a tuple_fraction > 0, however. + * + * As above, use 10% if OFFSET is present but unestimatable. + */ + if (*offset_est < 0) + limit_fraction = 0.10; + else + limit_fraction = (double) *offset_est; + + /* + * If we have absolute counts from both caller and OFFSET, add them + * together; likewise if they are both fractional. If one is + * fractional and the other absolute, we want to take the larger, and + * we heuristically assume that's the fractional one. + */ + if (tuple_fraction >= 1.0) + { + if (limit_fraction >= 1.0) + { + /* both absolute, so add them together */ + tuple_fraction += limit_fraction; + } + else + { + /* caller absolute, limit fractional; use limit */ + tuple_fraction = limit_fraction; + } + } + else + { + if (limit_fraction >= 1.0) + { + /* caller fractional, limit absolute; use caller's value */ + } + else + { + /* both fractional, so add them together */ + tuple_fraction += limit_fraction; + if (tuple_fraction >= 1.0) + tuple_fraction = 0.0; /* assume fetch all */ + } + } + } + + return tuple_fraction; +} + +/* + * limit_needed - do we actually need a Limit plan node? + * + * If we have constant-zero OFFSET and constant-null LIMIT, we can skip adding + * a Limit node. This is worth checking for because "OFFSET 0" is a common + * locution for an optimization fence. (Because other places in the planner + * merely check whether parse->limitOffset isn't NULL, it will still work as + * an optimization fence --- we're just suppressing unnecessary run-time + * overhead.) + * + * This might look like it could be merged into preprocess_limit, but there's + * a key distinction: here we need hard constants in OFFSET/LIMIT, whereas + * in preprocess_limit it's good enough to consider estimated values. + */ +bool +limit_needed(Query *parse) +{ + Node *node; + + node = parse->limitCount; + if (node) + { + if (IsA(node, Const)) + { + /* NULL indicates LIMIT ALL, ie, no limit */ + if (!((Const *) node)->constisnull) + return true; /* LIMIT with a constant value */ + } + else + return true; /* non-constant LIMIT */ + } + + node = parse->limitOffset; + if (node) + { + if (IsA(node, Const)) + { + /* Treat NULL as no offset; the executor would too */ + if (!((Const *) node)->constisnull) + { + int64 offset = DatumGetInt64(((Const *) node)->constvalue); + + if (offset != 0) + return true; /* OFFSET with a nonzero value */ + } + } + else + return true; /* non-constant OFFSET */ + } + + return false; /* don't need a Limit plan node */ +} + + +/* + * remove_useless_groupby_columns + * Remove any columns in the GROUP BY clause that are redundant due to + * being functionally dependent on other GROUP BY columns. + * + * Since some other DBMSes do not allow references to ungrouped columns, it's + * not unusual to find all columns listed in GROUP BY even though listing the + * primary-key columns would be sufficient. Deleting such excess columns + * avoids redundant sorting work, so it's worth doing. + * + * Relcache invalidations will ensure that cached plans become invalidated + * when the underlying index of the pkey constraint is dropped. + * + * Currently, we only make use of pkey constraints for this, however, we may + * wish to take this further in the future and also use unique constraints + * which have NOT NULL columns. In that case, plan invalidation will still + * work since relations will receive a relcache invalidation when a NOT NULL + * constraint is dropped. + */ +static void +remove_useless_groupby_columns(PlannerInfo *root) +{ + Query *parse = root->parse; + Bitmapset **groupbyattnos; + Bitmapset **surplusvars; + ListCell *lc; + int relid; + + /* No chance to do anything if there are less than two GROUP BY items */ + if (list_length(parse->groupClause) < 2) + return; + + /* Don't fiddle with the GROUP BY clause if the query has grouping sets */ + if (parse->groupingSets) + return; + + /* + * Scan the GROUP BY clause to find GROUP BY items that are simple Vars. + * Fill groupbyattnos[k] with a bitmapset of the column attnos of RTE k + * that are GROUP BY items. + */ + groupbyattnos = (Bitmapset **) palloc0(sizeof(Bitmapset *) * + (list_length(parse->rtable) + 1)); + foreach(lc, parse->groupClause) + { + SortGroupClause *sgc = lfirst_node(SortGroupClause, lc); + TargetEntry *tle = get_sortgroupclause_tle(sgc, parse->targetList); + Var *var = (Var *) tle->expr; + + /* + * Ignore non-Vars and Vars from other query levels. + * + * XXX in principle, stable expressions containing Vars could also be + * removed, if all the Vars are functionally dependent on other GROUP + * BY items. But it's not clear that such cases occur often enough to + * be worth troubling over. + */ + if (!IsA(var, Var) || + var->varlevelsup > 0) + continue; + + /* OK, remember we have this Var */ + relid = var->varno; + Assert(relid <= list_length(parse->rtable)); + groupbyattnos[relid] = bms_add_member(groupbyattnos[relid], + var->varattno - FirstLowInvalidHeapAttributeNumber); + } + + /* + * Consider each relation and see if it is possible to remove some of its + * Vars from GROUP BY. For simplicity and speed, we do the actual removal + * in a separate pass. Here, we just fill surplusvars[k] with a bitmapset + * of the column attnos of RTE k that are removable GROUP BY items. + */ + surplusvars = NULL; /* don't allocate array unless required */ + relid = 0; + foreach(lc, parse->rtable) + { + RangeTblEntry *rte = lfirst_node(RangeTblEntry, lc); + Bitmapset *relattnos; + Bitmapset *pkattnos; + Oid constraintOid; + + relid++; + + /* Only plain relations could have primary-key constraints */ + if (rte->rtekind != RTE_RELATION) + continue; + + /* + * We must skip inheritance parent tables as some of the child rels + * may cause duplicate rows. This cannot happen with partitioned + * tables, however. + */ + if (rte->inh && rte->relkind != RELKIND_PARTITIONED_TABLE) + continue; + + /* Nothing to do unless this rel has multiple Vars in GROUP BY */ + relattnos = groupbyattnos[relid]; + if (bms_membership(relattnos) != BMS_MULTIPLE) + continue; + + /* + * Can't remove any columns for this rel if there is no suitable + * (i.e., nondeferrable) primary key constraint. + */ + pkattnos = get_primary_key_attnos(rte->relid, false, &constraintOid); + if (pkattnos == NULL) + continue; + + /* + * If the primary key is a proper subset of relattnos then we have + * some items in the GROUP BY that can be removed. + */ + if (bms_subset_compare(pkattnos, relattnos) == BMS_SUBSET1) + { + /* + * To easily remember whether we've found anything to do, we don't + * allocate the surplusvars[] array until we find something. + */ + if (surplusvars == NULL) + surplusvars = (Bitmapset **) palloc0(sizeof(Bitmapset *) * + (list_length(parse->rtable) + 1)); + + /* Remember the attnos of the removable columns */ + surplusvars[relid] = bms_difference(relattnos, pkattnos); + } + } + + /* + * If we found any surplus Vars, build a new GROUP BY clause without them. + * (Note: this may leave some TLEs with unreferenced ressortgroupref + * markings, but that's harmless.) + */ + if (surplusvars != NULL) + { + List *new_groupby = NIL; + + foreach(lc, parse->groupClause) + { + SortGroupClause *sgc = lfirst_node(SortGroupClause, lc); + TargetEntry *tle = get_sortgroupclause_tle(sgc, parse->targetList); + Var *var = (Var *) tle->expr; + + /* + * New list must include non-Vars, outer Vars, and anything not + * marked as surplus. + */ + if (!IsA(var, Var) || + var->varlevelsup > 0 || + !bms_is_member(var->varattno - FirstLowInvalidHeapAttributeNumber, + surplusvars[var->varno])) + new_groupby = lappend(new_groupby, sgc); + } + + parse->groupClause = new_groupby; + } +} + +/* + * preprocess_groupclause - do preparatory work on GROUP BY clause + * + * The idea here is to adjust the ordering of the GROUP BY elements + * (which in itself is semantically insignificant) to match ORDER BY, + * thereby allowing a single sort operation to both implement the ORDER BY + * requirement and set up for a Unique step that implements GROUP BY. + * + * In principle it might be interesting to consider other orderings of the + * GROUP BY elements, which could match the sort ordering of other + * possible plans (eg an indexscan) and thereby reduce cost. We don't + * bother with that, though. Hashed grouping will frequently win anyway. + * + * Note: we need no comparable processing of the distinctClause because + * the parser already enforced that that matches ORDER BY. + * + * For grouping sets, the order of items is instead forced to agree with that + * of the grouping set (and items not in the grouping set are skipped). The + * work of sorting the order of grouping set elements to match the ORDER BY if + * possible is done elsewhere. + */ +static List * +preprocess_groupclause(PlannerInfo *root, List *force) +{ + Query *parse = root->parse; + List *new_groupclause = NIL; + bool partial_match; + ListCell *sl; + ListCell *gl; + + /* For grouping sets, we need to force the ordering */ + if (force) + { + foreach(sl, force) + { + Index ref = lfirst_int(sl); + SortGroupClause *cl = get_sortgroupref_clause(ref, parse->groupClause); + + new_groupclause = lappend(new_groupclause, cl); + } + + return new_groupclause; + } + + /* If no ORDER BY, nothing useful to do here */ + if (parse->sortClause == NIL) + return parse->groupClause; + + /* + * Scan the ORDER BY clause and construct a list of matching GROUP BY + * items, but only as far as we can make a matching prefix. + * + * This code assumes that the sortClause contains no duplicate items. + */ + foreach(sl, parse->sortClause) + { + SortGroupClause *sc = lfirst_node(SortGroupClause, sl); + + foreach(gl, parse->groupClause) + { + SortGroupClause *gc = lfirst_node(SortGroupClause, gl); + + if (equal(gc, sc)) + { + new_groupclause = lappend(new_groupclause, gc); + break; + } + } + if (gl == NULL) + break; /* no match, so stop scanning */ + } + + /* Did we match all of the ORDER BY list, or just some of it? */ + partial_match = (sl != NULL); + + /* If no match at all, no point in reordering GROUP BY */ + if (new_groupclause == NIL) + return parse->groupClause; + + /* + * Add any remaining GROUP BY items to the new list, but only if we were + * able to make a complete match. In other words, we only rearrange the + * GROUP BY list if the result is that one list is a prefix of the other + * --- otherwise there's no possibility of a common sort. Also, give up + * if there are any non-sortable GROUP BY items, since then there's no + * hope anyway. + */ + foreach(gl, parse->groupClause) + { + SortGroupClause *gc = lfirst_node(SortGroupClause, gl); + + if (list_member_ptr(new_groupclause, gc)) + continue; /* it matched an ORDER BY item */ + if (partial_match) + return parse->groupClause; /* give up, no common sort possible */ + if (!OidIsValid(gc->sortop)) + return parse->groupClause; /* give up, GROUP BY can't be sorted */ + new_groupclause = lappend(new_groupclause, gc); + } + + /* Success --- install the rearranged GROUP BY list */ + Assert(list_length(parse->groupClause) == list_length(new_groupclause)); + return new_groupclause; +} + +/* + * Extract lists of grouping sets that can be implemented using a single + * rollup-type aggregate pass each. Returns a list of lists of grouping sets. + * + * Input must be sorted with smallest sets first. Result has each sublist + * sorted with smallest sets first. + * + * We want to produce the absolute minimum possible number of lists here to + * avoid excess sorts. Fortunately, there is an algorithm for this; the problem + * of finding the minimal partition of a partially-ordered set into chains + * (which is what we need, taking the list of grouping sets as a poset ordered + * by set inclusion) can be mapped to the problem of finding the maximum + * cardinality matching on a bipartite graph, which is solvable in polynomial + * time with a worst case of no worse than O(n^2.5) and usually much + * better. Since our N is at most 4096, we don't need to consider fallbacks to + * heuristic or approximate methods. (Planning time for a 12-d cube is under + * half a second on my modest system even with optimization off and assertions + * on.) + */ +static List * +extract_rollup_sets(List *groupingSets) +{ + int num_sets_raw = list_length(groupingSets); + int num_empty = 0; + int num_sets = 0; /* distinct sets */ + int num_chains = 0; + List *result = NIL; + List **results; + List **orig_sets; + Bitmapset **set_masks; + int *chains; + short **adjacency; + short *adjacency_buf; + BipartiteMatchState *state; + int i; + int j; + int j_size; + ListCell *lc1 = list_head(groupingSets); + ListCell *lc; + + /* + * Start by stripping out empty sets. The algorithm doesn't require this, + * but the planner currently needs all empty sets to be returned in the + * first list, so we strip them here and add them back after. + */ + while (lc1 && lfirst(lc1) == NIL) + { + ++num_empty; + lc1 = lnext(groupingSets, lc1); + } + + /* bail out now if it turns out that all we had were empty sets. */ + if (!lc1) + return list_make1(groupingSets); + + /*---------- + * We don't strictly need to remove duplicate sets here, but if we don't, + * they tend to become scattered through the result, which is a bit + * confusing (and irritating if we ever decide to optimize them out). + * So we remove them here and add them back after. + * + * For each non-duplicate set, we fill in the following: + * + * orig_sets[i] = list of the original set lists + * set_masks[i] = bitmapset for testing inclusion + * adjacency[i] = array [n, v1, v2, ... vn] of adjacency indices + * + * chains[i] will be the result group this set is assigned to. + * + * We index all of these from 1 rather than 0 because it is convenient + * to leave 0 free for the NIL node in the graph algorithm. + *---------- + */ + orig_sets = palloc0((num_sets_raw + 1) * sizeof(List *)); + set_masks = palloc0((num_sets_raw + 1) * sizeof(Bitmapset *)); + adjacency = palloc0((num_sets_raw + 1) * sizeof(short *)); + adjacency_buf = palloc((num_sets_raw + 1) * sizeof(short)); + + j_size = 0; + j = 0; + i = 1; + + for_each_cell(lc, groupingSets, lc1) + { + List *candidate = (List *) lfirst(lc); + Bitmapset *candidate_set = NULL; + ListCell *lc2; + int dup_of = 0; + + foreach(lc2, candidate) + { + candidate_set = bms_add_member(candidate_set, lfirst_int(lc2)); + } + + /* we can only be a dup if we're the same length as a previous set */ + if (j_size == list_length(candidate)) + { + int k; + + for (k = j; k < i; ++k) + { + if (bms_equal(set_masks[k], candidate_set)) + { + dup_of = k; + break; + } + } + } + else if (j_size < list_length(candidate)) + { + j_size = list_length(candidate); + j = i; + } + + if (dup_of > 0) + { + orig_sets[dup_of] = lappend(orig_sets[dup_of], candidate); + bms_free(candidate_set); + } + else + { + int k; + int n_adj = 0; + + orig_sets[i] = list_make1(candidate); + set_masks[i] = candidate_set; + + /* fill in adjacency list; no need to compare equal-size sets */ + + for (k = j - 1; k > 0; --k) + { + if (bms_is_subset(set_masks[k], candidate_set)) + adjacency_buf[++n_adj] = k; + } + + if (n_adj > 0) + { + adjacency_buf[0] = n_adj; + adjacency[i] = palloc((n_adj + 1) * sizeof(short)); + memcpy(adjacency[i], adjacency_buf, (n_adj + 1) * sizeof(short)); + } + else + adjacency[i] = NULL; + + ++i; + } + } + + num_sets = i - 1; + + /* + * Apply the graph matching algorithm to do the work. + */ + state = BipartiteMatch(num_sets, num_sets, adjacency); + + /* + * Now, the state->pair* fields have the info we need to assign sets to + * chains. Two sets (u,v) belong to the same chain if pair_uv[u] = v or + * pair_vu[v] = u (both will be true, but we check both so that we can do + * it in one pass) + */ + chains = palloc0((num_sets + 1) * sizeof(int)); + + for (i = 1; i <= num_sets; ++i) + { + int u = state->pair_vu[i]; + int v = state->pair_uv[i]; + + if (u > 0 && u < i) + chains[i] = chains[u]; + else if (v > 0 && v < i) + chains[i] = chains[v]; + else + chains[i] = ++num_chains; + } + + /* build result lists. */ + results = palloc0((num_chains + 1) * sizeof(List *)); + + for (i = 1; i <= num_sets; ++i) + { + int c = chains[i]; + + Assert(c > 0); + + results[c] = list_concat(results[c], orig_sets[i]); + } + + /* push any empty sets back on the first list. */ + while (num_empty-- > 0) + results[1] = lcons(NIL, results[1]); + + /* make result list */ + for (i = 1; i <= num_chains; ++i) + result = lappend(result, results[i]); + + /* + * Free all the things. + * + * (This is over-fussy for small sets but for large sets we could have + * tied up a nontrivial amount of memory.) + */ + BipartiteMatchFree(state); + pfree(results); + pfree(chains); + for (i = 1; i <= num_sets; ++i) + if (adjacency[i]) + pfree(adjacency[i]); + pfree(adjacency); + pfree(adjacency_buf); + pfree(orig_sets); + for (i = 1; i <= num_sets; ++i) + bms_free(set_masks[i]); + pfree(set_masks); + + return result; +} + +/* + * Reorder the elements of a list of grouping sets such that they have correct + * prefix relationships. Also inserts the GroupingSetData annotations. + * + * The input must be ordered with smallest sets first; the result is returned + * with largest sets first. Note that the result shares no list substructure + * with the input, so it's safe for the caller to modify it later. + * + * If we're passed in a sortclause, we follow its order of columns to the + * extent possible, to minimize the chance that we add unnecessary sorts. + * (We're trying here to ensure that GROUPING SETS ((a,b,c),(c)) ORDER BY c,b,a + * gets implemented in one pass.) + */ +static List * +reorder_grouping_sets(List *groupingsets, List *sortclause) +{ + ListCell *lc; + List *previous = NIL; + List *result = NIL; + + foreach(lc, groupingsets) + { + List *candidate = (List *) lfirst(lc); + List *new_elems = list_difference_int(candidate, previous); + GroupingSetData *gs = makeNode(GroupingSetData); + + while (list_length(sortclause) > list_length(previous) && + list_length(new_elems) > 0) + { + SortGroupClause *sc = list_nth(sortclause, list_length(previous)); + int ref = sc->tleSortGroupRef; + + if (list_member_int(new_elems, ref)) + { + previous = lappend_int(previous, ref); + new_elems = list_delete_int(new_elems, ref); + } + else + { + /* diverged from the sortclause; give up on it */ + sortclause = NIL; + break; + } + } + + previous = list_concat(previous, new_elems); + + gs->set = list_copy(previous); + result = lcons(gs, result); + } + + list_free(previous); + + return result; +} + +/* + * Compute query_pathkeys and other pathkeys during plan generation + */ +static void +standard_qp_callback(PlannerInfo *root, void *extra) +{ + Query *parse = root->parse; + standard_qp_extra *qp_extra = (standard_qp_extra *) extra; + List *tlist = root->processed_tlist; + List *activeWindows = qp_extra->activeWindows; + + /* + * Calculate pathkeys that represent grouping/ordering requirements. The + * sortClause is certainly sort-able, but GROUP BY and DISTINCT might not + * be, in which case we just leave their pathkeys empty. + */ + if (qp_extra->groupClause && + grouping_is_sortable(qp_extra->groupClause)) + root->group_pathkeys = + make_pathkeys_for_sortclauses(root, + qp_extra->groupClause, + tlist); + else + root->group_pathkeys = NIL; + + /* We consider only the first (bottom) window in pathkeys logic */ + if (activeWindows != NIL) + { + WindowClause *wc = linitial_node(WindowClause, activeWindows); + + root->window_pathkeys = make_pathkeys_for_window(root, + wc, + tlist); + } + else + root->window_pathkeys = NIL; + + if (parse->distinctClause && + grouping_is_sortable(parse->distinctClause)) + root->distinct_pathkeys = + make_pathkeys_for_sortclauses(root, + parse->distinctClause, + tlist); + else + root->distinct_pathkeys = NIL; + + root->sort_pathkeys = + make_pathkeys_for_sortclauses(root, + parse->sortClause, + tlist); + + /* + * Figure out whether we want a sorted result from query_planner. + * + * If we have a sortable GROUP BY clause, then we want a result sorted + * properly for grouping. Otherwise, if we have window functions to + * evaluate, we try to sort for the first window. Otherwise, if there's a + * sortable DISTINCT clause that's more rigorous than the ORDER BY clause, + * we try to produce output that's sufficiently well sorted for the + * DISTINCT. Otherwise, if there is an ORDER BY clause, we want to sort + * by the ORDER BY clause. + * + * Note: if we have both ORDER BY and GROUP BY, and ORDER BY is a superset + * of GROUP BY, it would be tempting to request sort by ORDER BY --- but + * that might just leave us failing to exploit an available sort order at + * all. Needs more thought. The choice for DISTINCT versus ORDER BY is + * much easier, since we know that the parser ensured that one is a + * superset of the other. + */ + if (root->group_pathkeys) + root->query_pathkeys = root->group_pathkeys; + else if (root->window_pathkeys) + root->query_pathkeys = root->window_pathkeys; + else if (list_length(root->distinct_pathkeys) > + list_length(root->sort_pathkeys)) + root->query_pathkeys = root->distinct_pathkeys; + else if (root->sort_pathkeys) + root->query_pathkeys = root->sort_pathkeys; + else + root->query_pathkeys = NIL; +} + +/* + * Estimate number of groups produced by grouping clauses (1 if not grouping) + * + * path_rows: number of output rows from scan/join step + * gd: grouping sets data including list of grouping sets and their clauses + * target_list: target list containing group clause references + * + * If doing grouping sets, we also annotate the gsets data with the estimates + * for each set and each individual rollup list, with a view to later + * determining whether some combination of them could be hashed instead. + */ +static double +get_number_of_groups(PlannerInfo *root, + double path_rows, + grouping_sets_data *gd, + List *target_list) +{ + Query *parse = root->parse; + double dNumGroups; + + if (parse->groupClause) + { + List *groupExprs; + + if (parse->groupingSets) + { + /* Add up the estimates for each grouping set */ + ListCell *lc; + ListCell *lc2; + + Assert(gd); /* keep Coverity happy */ + + dNumGroups = 0; + + foreach(lc, gd->rollups) + { + RollupData *rollup = lfirst_node(RollupData, lc); + ListCell *lc; + + groupExprs = get_sortgrouplist_exprs(rollup->groupClause, + target_list); + + rollup->numGroups = 0.0; + + forboth(lc, rollup->gsets, lc2, rollup->gsets_data) + { + List *gset = (List *) lfirst(lc); + GroupingSetData *gs = lfirst_node(GroupingSetData, lc2); + double numGroups = estimate_num_groups(root, + groupExprs, + path_rows, + &gset, + NULL); + + gs->numGroups = numGroups; + rollup->numGroups += numGroups; + } + + dNumGroups += rollup->numGroups; + } + + if (gd->hash_sets_idx) + { + ListCell *lc; + + gd->dNumHashGroups = 0; + + groupExprs = get_sortgrouplist_exprs(parse->groupClause, + target_list); + + forboth(lc, gd->hash_sets_idx, lc2, gd->unsortable_sets) + { + List *gset = (List *) lfirst(lc); + GroupingSetData *gs = lfirst_node(GroupingSetData, lc2); + double numGroups = estimate_num_groups(root, + groupExprs, + path_rows, + &gset, + NULL); + + gs->numGroups = numGroups; + gd->dNumHashGroups += numGroups; + } + + dNumGroups += gd->dNumHashGroups; + } + } + else + { + /* Plain GROUP BY */ + groupExprs = get_sortgrouplist_exprs(parse->groupClause, + target_list); + + dNumGroups = estimate_num_groups(root, groupExprs, path_rows, + NULL, NULL); + } + } + else if (parse->groupingSets) + { + /* Empty grouping sets ... one result row for each one */ + dNumGroups = list_length(parse->groupingSets); + } + else if (parse->hasAggs || root->hasHavingQual) + { + /* Plain aggregation, one result row */ + dNumGroups = 1; + } + else + { + /* Not grouping */ + dNumGroups = 1; + } + + return dNumGroups; +} + +/* + * create_grouping_paths + * + * Build a new upperrel containing Paths for grouping and/or aggregation. + * Along the way, we also build an upperrel for Paths which are partially + * grouped and/or aggregated. A partially grouped and/or aggregated path + * needs a FinalizeAggregate node to complete the aggregation. Currently, + * the only partially grouped paths we build are also partial paths; that + * is, they need a Gather and then a FinalizeAggregate. + * + * input_rel: contains the source-data Paths + * target: the pathtarget for the result Paths to compute + * gd: grouping sets data including list of grouping sets and their clauses + * + * Note: all Paths in input_rel are expected to return the target computed + * by make_group_input_target. + */ +static RelOptInfo * +create_grouping_paths(PlannerInfo *root, + RelOptInfo *input_rel, + PathTarget *target, + bool target_parallel_safe, + grouping_sets_data *gd) +{ + Query *parse = root->parse; + RelOptInfo *grouped_rel; + RelOptInfo *partially_grouped_rel; + AggClauseCosts agg_costs; + + MemSet(&agg_costs, 0, sizeof(AggClauseCosts)); + get_agg_clause_costs(root, AGGSPLIT_SIMPLE, &agg_costs); + + /* + * Create grouping relation to hold fully aggregated grouping and/or + * aggregation paths. + */ + grouped_rel = make_grouping_rel(root, input_rel, target, + target_parallel_safe, parse->havingQual); + + /* + * Create either paths for a degenerate grouping or paths for ordinary + * grouping, as appropriate. + */ + if (is_degenerate_grouping(root)) + create_degenerate_grouping_paths(root, input_rel, grouped_rel); + else + { + int flags = 0; + GroupPathExtraData extra; + + /* + * Determine whether it's possible to perform sort-based + * implementations of grouping. (Note that if groupClause is empty, + * grouping_is_sortable() is trivially true, and all the + * pathkeys_contained_in() tests will succeed too, so that we'll + * consider every surviving input path.) + * + * If we have grouping sets, we might be able to sort some but not all + * of them; in this case, we need can_sort to be true as long as we + * must consider any sorted-input plan. + */ + if ((gd && gd->rollups != NIL) + || grouping_is_sortable(parse->groupClause)) + flags |= GROUPING_CAN_USE_SORT; + + /* + * Determine whether we should consider hash-based implementations of + * grouping. + * + * Hashed aggregation only applies if we're grouping. If we have + * grouping sets, some groups might be hashable but others not; in + * this case we set can_hash true as long as there is nothing globally + * preventing us from hashing (and we should therefore consider plans + * with hashes). + * + * Executor doesn't support hashed aggregation with DISTINCT or ORDER + * BY aggregates. (Doing so would imply storing *all* the input + * values in the hash table, and/or running many sorts in parallel, + * either of which seems like a certain loser.) We similarly don't + * support ordered-set aggregates in hashed aggregation, but that case + * is also included in the numOrderedAggs count. + * + * Note: grouping_is_hashable() is much more expensive to check than + * the other gating conditions, so we want to do it last. + */ + if ((parse->groupClause != NIL && + root->numOrderedAggs == 0 && + (gd ? gd->any_hashable : grouping_is_hashable(parse->groupClause)))) + flags |= GROUPING_CAN_USE_HASH; + + /* + * Determine whether partial aggregation is possible. + */ + if (can_partial_agg(root)) + flags |= GROUPING_CAN_PARTIAL_AGG; + + extra.flags = flags; + extra.target_parallel_safe = target_parallel_safe; + extra.havingQual = parse->havingQual; + extra.targetList = parse->targetList; + extra.partial_costs_set = false; + + /* + * Determine whether partitionwise aggregation is in theory possible. + * It can be disabled by the user, and for now, we don't try to + * support grouping sets. create_ordinary_grouping_paths() will check + * additional conditions, such as whether input_rel is partitioned. + */ + if (enable_partitionwise_aggregate && !parse->groupingSets) + extra.patype = PARTITIONWISE_AGGREGATE_FULL; + else + extra.patype = PARTITIONWISE_AGGREGATE_NONE; + + create_ordinary_grouping_paths(root, input_rel, grouped_rel, + &agg_costs, gd, &extra, + &partially_grouped_rel); + } + + set_cheapest(grouped_rel); + return grouped_rel; +} + +/* + * make_grouping_rel + * + * Create a new grouping rel and set basic properties. + * + * input_rel represents the underlying scan/join relation. + * target is the output expected from the grouping relation. + */ +static RelOptInfo * +make_grouping_rel(PlannerInfo *root, RelOptInfo *input_rel, + PathTarget *target, bool target_parallel_safe, + Node *havingQual) +{ + RelOptInfo *grouped_rel; + + if (IS_OTHER_REL(input_rel)) + { + grouped_rel = fetch_upper_rel(root, UPPERREL_GROUP_AGG, + input_rel->relids); + grouped_rel->reloptkind = RELOPT_OTHER_UPPER_REL; + } + else + { + /* + * By tradition, the relids set for the main grouping relation is + * NULL. (This could be changed, but might require adjustments + * elsewhere.) + */ + grouped_rel = fetch_upper_rel(root, UPPERREL_GROUP_AGG, NULL); + } + + /* Set target. */ + grouped_rel->reltarget = target; + + /* + * If the input relation is not parallel-safe, then the grouped relation + * can't be parallel-safe, either. Otherwise, it's parallel-safe if the + * target list and HAVING quals are parallel-safe. + */ + if (input_rel->consider_parallel && target_parallel_safe && + is_parallel_safe(root, (Node *) havingQual)) + grouped_rel->consider_parallel = true; + + /* + * If the input rel belongs to a single FDW, so does the grouped rel. + */ + grouped_rel->serverid = input_rel->serverid; + grouped_rel->userid = input_rel->userid; + grouped_rel->useridiscurrent = input_rel->useridiscurrent; + grouped_rel->fdwroutine = input_rel->fdwroutine; + + return grouped_rel; +} + +/* + * is_degenerate_grouping + * + * A degenerate grouping is one in which the query has a HAVING qual and/or + * grouping sets, but no aggregates and no GROUP BY (which implies that the + * grouping sets are all empty). + */ +static bool +is_degenerate_grouping(PlannerInfo *root) +{ + Query *parse = root->parse; + + return (root->hasHavingQual || parse->groupingSets) && + !parse->hasAggs && parse->groupClause == NIL; +} + +/* + * create_degenerate_grouping_paths + * + * When the grouping is degenerate (see is_degenerate_grouping), we are + * supposed to emit either zero or one row for each grouping set depending on + * whether HAVING succeeds. Furthermore, there cannot be any variables in + * either HAVING or the targetlist, so we actually do not need the FROM table + * at all! We can just throw away the plan-so-far and generate a Result node. + * This is a sufficiently unusual corner case that it's not worth contorting + * the structure of this module to avoid having to generate the earlier paths + * in the first place. + */ +static void +create_degenerate_grouping_paths(PlannerInfo *root, RelOptInfo *input_rel, + RelOptInfo *grouped_rel) +{ + Query *parse = root->parse; + int nrows; + Path *path; + + nrows = list_length(parse->groupingSets); + if (nrows > 1) + { + /* + * Doesn't seem worthwhile writing code to cons up a generate_series + * or a values scan to emit multiple rows. Instead just make N clones + * and append them. (With a volatile HAVING clause, this means you + * might get between 0 and N output rows. Offhand I think that's + * desired.) + */ + List *paths = NIL; + + while (--nrows >= 0) + { + path = (Path *) + create_group_result_path(root, grouped_rel, + grouped_rel->reltarget, + (List *) parse->havingQual); + paths = lappend(paths, path); + } + path = (Path *) + create_append_path(root, + grouped_rel, + paths, + NIL, + NIL, + NULL, + 0, + false, + -1); + } + else + { + /* No grouping sets, or just one, so one output row */ + path = (Path *) + create_group_result_path(root, grouped_rel, + grouped_rel->reltarget, + (List *) parse->havingQual); + } + + add_path(grouped_rel, path); +} + +/* + * create_ordinary_grouping_paths + * + * Create grouping paths for the ordinary (that is, non-degenerate) case. + * + * We need to consider sorted and hashed aggregation in the same function, + * because otherwise (1) it would be harder to throw an appropriate error + * message if neither way works, and (2) we should not allow hashtable size + * considerations to dissuade us from using hashing if sorting is not possible. + * + * *partially_grouped_rel_p will be set to the partially grouped rel which this + * function creates, or to NULL if it doesn't create one. + */ +static void +create_ordinary_grouping_paths(PlannerInfo *root, RelOptInfo *input_rel, + RelOptInfo *grouped_rel, + const AggClauseCosts *agg_costs, + grouping_sets_data *gd, + GroupPathExtraData *extra, + RelOptInfo **partially_grouped_rel_p) +{ + Path *cheapest_path = input_rel->cheapest_total_path; + RelOptInfo *partially_grouped_rel = NULL; + double dNumGroups; + PartitionwiseAggregateType patype = PARTITIONWISE_AGGREGATE_NONE; + + /* + * If this is the topmost grouping relation or if the parent relation is + * doing some form of partitionwise aggregation, then we may be able to do + * it at this level also. However, if the input relation is not + * partitioned, partitionwise aggregate is impossible. + */ + if (extra->patype != PARTITIONWISE_AGGREGATE_NONE && + IS_PARTITIONED_REL(input_rel)) + { + /* + * If this is the topmost relation or if the parent relation is doing + * full partitionwise aggregation, then we can do full partitionwise + * aggregation provided that the GROUP BY clause contains all of the + * partitioning columns at this level. Otherwise, we can do at most + * partial partitionwise aggregation. But if partial aggregation is + * not supported in general then we can't use it for partitionwise + * aggregation either. + */ + if (extra->patype == PARTITIONWISE_AGGREGATE_FULL && + group_by_has_partkey(input_rel, extra->targetList, + root->parse->groupClause)) + patype = PARTITIONWISE_AGGREGATE_FULL; + else if ((extra->flags & GROUPING_CAN_PARTIAL_AGG) != 0) + patype = PARTITIONWISE_AGGREGATE_PARTIAL; + else + patype = PARTITIONWISE_AGGREGATE_NONE; + } + + /* + * Before generating paths for grouped_rel, we first generate any possible + * partially grouped paths; that way, later code can easily consider both + * parallel and non-parallel approaches to grouping. + */ + if ((extra->flags & GROUPING_CAN_PARTIAL_AGG) != 0) + { + bool force_rel_creation; + + /* + * If we're doing partitionwise aggregation at this level, force + * creation of a partially_grouped_rel so we can add partitionwise + * paths to it. + */ + force_rel_creation = (patype == PARTITIONWISE_AGGREGATE_PARTIAL); + + partially_grouped_rel = + create_partial_grouping_paths(root, + grouped_rel, + input_rel, + gd, + extra, + force_rel_creation); + } + + /* Set out parameter. */ + *partially_grouped_rel_p = partially_grouped_rel; + + /* Apply partitionwise aggregation technique, if possible. */ + if (patype != PARTITIONWISE_AGGREGATE_NONE) + create_partitionwise_grouping_paths(root, input_rel, grouped_rel, + partially_grouped_rel, agg_costs, + gd, patype, extra); + + /* If we are doing partial aggregation only, return. */ + if (extra->patype == PARTITIONWISE_AGGREGATE_PARTIAL) + { + Assert(partially_grouped_rel); + + if (partially_grouped_rel->pathlist) + set_cheapest(partially_grouped_rel); + + return; + } + + /* Gather any partially grouped partial paths. */ + if (partially_grouped_rel && partially_grouped_rel->partial_pathlist) + { + gather_grouping_paths(root, partially_grouped_rel); + set_cheapest(partially_grouped_rel); + } + + /* + * Estimate number of groups. + */ + dNumGroups = get_number_of_groups(root, + cheapest_path->rows, + gd, + extra->targetList); + + /* Build final grouping paths */ + add_paths_to_grouping_rel(root, input_rel, grouped_rel, + partially_grouped_rel, agg_costs, gd, + dNumGroups, extra); + + /* Give a helpful error if we failed to find any implementation */ + if (grouped_rel->pathlist == NIL) + ereport(ERROR, + (errcode(ERRCODE_FEATURE_NOT_SUPPORTED), + errmsg("could not implement GROUP BY"), + errdetail("Some of the datatypes only support hashing, while others only support sorting."))); + + /* + * If there is an FDW that's responsible for all baserels of the query, + * let it consider adding ForeignPaths. + */ + if (grouped_rel->fdwroutine && + grouped_rel->fdwroutine->GetForeignUpperPaths) + grouped_rel->fdwroutine->GetForeignUpperPaths(root, UPPERREL_GROUP_AGG, + input_rel, grouped_rel, + extra); + + /* Let extensions possibly add some more paths */ + if (create_upper_paths_hook) + (*create_upper_paths_hook) (root, UPPERREL_GROUP_AGG, + input_rel, grouped_rel, + extra); +} + +/* + * For a given input path, consider the possible ways of doing grouping sets on + * it, by combinations of hashing and sorting. This can be called multiple + * times, so it's important that it not scribble on input. No result is + * returned, but any generated paths are added to grouped_rel. + */ +static void +consider_groupingsets_paths(PlannerInfo *root, + RelOptInfo *grouped_rel, + Path *path, + bool is_sorted, + bool can_hash, + grouping_sets_data *gd, + const AggClauseCosts *agg_costs, + double dNumGroups) +{ + Query *parse = root->parse; + Size hash_mem_limit = get_hash_memory_limit(); + + /* + * If we're not being offered sorted input, then only consider plans that + * can be done entirely by hashing. + * + * We can hash everything if it looks like it'll fit in hash_mem. But if + * the input is actually sorted despite not being advertised as such, we + * prefer to make use of that in order to use less memory. + * + * If none of the grouping sets are sortable, then ignore the hash_mem + * limit and generate a path anyway, since otherwise we'll just fail. + */ + if (!is_sorted) + { + List *new_rollups = NIL; + RollupData *unhashed_rollup = NULL; + List *sets_data; + List *empty_sets_data = NIL; + List *empty_sets = NIL; + ListCell *lc; + ListCell *l_start = list_head(gd->rollups); + AggStrategy strat = AGG_HASHED; + double hashsize; + double exclude_groups = 0.0; + + Assert(can_hash); + + /* + * If the input is coincidentally sorted usefully (which can happen + * even if is_sorted is false, since that only means that our caller + * has set up the sorting for us), then save some hashtable space by + * making use of that. But we need to watch out for degenerate cases: + * + * 1) If there are any empty grouping sets, then group_pathkeys might + * be NIL if all non-empty grouping sets are unsortable. In this case, + * there will be a rollup containing only empty groups, and the + * pathkeys_contained_in test is vacuously true; this is ok. + * + * XXX: the above relies on the fact that group_pathkeys is generated + * from the first rollup. If we add the ability to consider multiple + * sort orders for grouping input, this assumption might fail. + * + * 2) If there are no empty sets and only unsortable sets, then the + * rollups list will be empty (and thus l_start == NULL), and + * group_pathkeys will be NIL; we must ensure that the vacuously-true + * pathkeys_contained_in test doesn't cause us to crash. + */ + if (l_start != NULL && + pathkeys_contained_in(root->group_pathkeys, path->pathkeys)) + { + unhashed_rollup = lfirst_node(RollupData, l_start); + exclude_groups = unhashed_rollup->numGroups; + l_start = lnext(gd->rollups, l_start); + } + + hashsize = estimate_hashagg_tablesize(root, + path, + agg_costs, + dNumGroups - exclude_groups); + + /* + * gd->rollups is empty if we have only unsortable columns to work + * with. Override hash_mem in that case; otherwise, we'll rely on the + * sorted-input case to generate usable mixed paths. + */ + if (hashsize > hash_mem_limit && gd->rollups) + return; /* nope, won't fit */ + + /* + * We need to burst the existing rollups list into individual grouping + * sets and recompute a groupClause for each set. + */ + sets_data = list_copy(gd->unsortable_sets); + + for_each_cell(lc, gd->rollups, l_start) + { + RollupData *rollup = lfirst_node(RollupData, lc); + + /* + * If we find an unhashable rollup that's not been skipped by the + * "actually sorted" check above, we can't cope; we'd need sorted + * input (with a different sort order) but we can't get that here. + * So bail out; we'll get a valid path from the is_sorted case + * instead. + * + * The mere presence of empty grouping sets doesn't make a rollup + * unhashable (see preprocess_grouping_sets), we handle those + * specially below. + */ + if (!rollup->hashable) + return; + + sets_data = list_concat(sets_data, rollup->gsets_data); + } + foreach(lc, sets_data) + { + GroupingSetData *gs = lfirst_node(GroupingSetData, lc); + List *gset = gs->set; + RollupData *rollup; + + if (gset == NIL) + { + /* Empty grouping sets can't be hashed. */ + empty_sets_data = lappend(empty_sets_data, gs); + empty_sets = lappend(empty_sets, NIL); + } + else + { + rollup = makeNode(RollupData); + + rollup->groupClause = preprocess_groupclause(root, gset); + rollup->gsets_data = list_make1(gs); + rollup->gsets = remap_to_groupclause_idx(rollup->groupClause, + rollup->gsets_data, + gd->tleref_to_colnum_map); + rollup->numGroups = gs->numGroups; + rollup->hashable = true; + rollup->is_hashed = true; + new_rollups = lappend(new_rollups, rollup); + } + } + + /* + * If we didn't find anything nonempty to hash, then bail. We'll + * generate a path from the is_sorted case. + */ + if (new_rollups == NIL) + return; + + /* + * If there were empty grouping sets they should have been in the + * first rollup. + */ + Assert(!unhashed_rollup || !empty_sets); + + if (unhashed_rollup) + { + new_rollups = lappend(new_rollups, unhashed_rollup); + strat = AGG_MIXED; + } + else if (empty_sets) + { + RollupData *rollup = makeNode(RollupData); + + rollup->groupClause = NIL; + rollup->gsets_data = empty_sets_data; + rollup->gsets = empty_sets; + rollup->numGroups = list_length(empty_sets); + rollup->hashable = false; + rollup->is_hashed = false; + new_rollups = lappend(new_rollups, rollup); + strat = AGG_MIXED; + } + + add_path(grouped_rel, (Path *) + create_groupingsets_path(root, + grouped_rel, + path, + (List *) parse->havingQual, + strat, + new_rollups, + agg_costs, + dNumGroups)); + return; + } + + /* + * If we have sorted input but nothing we can do with it, bail. + */ + if (list_length(gd->rollups) == 0) + return; + + /* + * Given sorted input, we try and make two paths: one sorted and one mixed + * sort/hash. (We need to try both because hashagg might be disabled, or + * some columns might not be sortable.) + * + * can_hash is passed in as false if some obstacle elsewhere (such as + * ordered aggs) means that we shouldn't consider hashing at all. + */ + if (can_hash && gd->any_hashable) + { + List *rollups = NIL; + List *hash_sets = list_copy(gd->unsortable_sets); + double availspace = hash_mem_limit; + ListCell *lc; + + /* + * Account first for space needed for groups we can't sort at all. + */ + availspace -= estimate_hashagg_tablesize(root, + path, + agg_costs, + gd->dNumHashGroups); + + if (availspace > 0 && list_length(gd->rollups) > 1) + { + double scale; + int num_rollups = list_length(gd->rollups); + int k_capacity; + int *k_weights = palloc(num_rollups * sizeof(int)); + Bitmapset *hash_items = NULL; + int i; + + /* + * We treat this as a knapsack problem: the knapsack capacity + * represents hash_mem, the item weights are the estimated memory + * usage of the hashtables needed to implement a single rollup, + * and we really ought to use the cost saving as the item value; + * however, currently the costs assigned to sort nodes don't + * reflect the comparison costs well, and so we treat all items as + * of equal value (each rollup we hash instead saves us one sort). + * + * To use the discrete knapsack, we need to scale the values to a + * reasonably small bounded range. We choose to allow a 5% error + * margin; we have no more than 4096 rollups in the worst possible + * case, which with a 5% error margin will require a bit over 42MB + * of workspace. (Anyone wanting to plan queries that complex had + * better have the memory for it. In more reasonable cases, with + * no more than a couple of dozen rollups, the memory usage will + * be negligible.) + * + * k_capacity is naturally bounded, but we clamp the values for + * scale and weight (below) to avoid overflows or underflows (or + * uselessly trying to use a scale factor less than 1 byte). + */ + scale = Max(availspace / (20.0 * num_rollups), 1.0); + k_capacity = (int) floor(availspace / scale); + + /* + * We leave the first rollup out of consideration since it's the + * one that matches the input sort order. We assign indexes "i" + * to only those entries considered for hashing; the second loop, + * below, must use the same condition. + */ + i = 0; + for_each_from(lc, gd->rollups, 1) + { + RollupData *rollup = lfirst_node(RollupData, lc); + + if (rollup->hashable) + { + double sz = estimate_hashagg_tablesize(root, + path, + agg_costs, + rollup->numGroups); + + /* + * If sz is enormous, but hash_mem (and hence scale) is + * small, avoid integer overflow here. + */ + k_weights[i] = (int) Min(floor(sz / scale), + k_capacity + 1.0); + ++i; + } + } + + /* + * Apply knapsack algorithm; compute the set of items which + * maximizes the value stored (in this case the number of sorts + * saved) while keeping the total size (approximately) within + * capacity. + */ + if (i > 0) + hash_items = DiscreteKnapsack(k_capacity, i, k_weights, NULL); + + if (!bms_is_empty(hash_items)) + { + rollups = list_make1(linitial(gd->rollups)); + + i = 0; + for_each_from(lc, gd->rollups, 1) + { + RollupData *rollup = lfirst_node(RollupData, lc); + + if (rollup->hashable) + { + if (bms_is_member(i, hash_items)) + hash_sets = list_concat(hash_sets, + rollup->gsets_data); + else + rollups = lappend(rollups, rollup); + ++i; + } + else + rollups = lappend(rollups, rollup); + } + } + } + + if (!rollups && hash_sets) + rollups = list_copy(gd->rollups); + + foreach(lc, hash_sets) + { + GroupingSetData *gs = lfirst_node(GroupingSetData, lc); + RollupData *rollup = makeNode(RollupData); + + Assert(gs->set != NIL); + + rollup->groupClause = preprocess_groupclause(root, gs->set); + rollup->gsets_data = list_make1(gs); + rollup->gsets = remap_to_groupclause_idx(rollup->groupClause, + rollup->gsets_data, + gd->tleref_to_colnum_map); + rollup->numGroups = gs->numGroups; + rollup->hashable = true; + rollup->is_hashed = true; + rollups = lcons(rollup, rollups); + } + + if (rollups) + { + add_path(grouped_rel, (Path *) + create_groupingsets_path(root, + grouped_rel, + path, + (List *) parse->havingQual, + AGG_MIXED, + rollups, + agg_costs, + dNumGroups)); + } + } + + /* + * Now try the simple sorted case. + */ + if (!gd->unsortable_sets) + add_path(grouped_rel, (Path *) + create_groupingsets_path(root, + grouped_rel, + path, + (List *) parse->havingQual, + AGG_SORTED, + gd->rollups, + agg_costs, + dNumGroups)); +} + +/* + * create_window_paths + * + * Build a new upperrel containing Paths for window-function evaluation. + * + * input_rel: contains the source-data Paths + * input_target: result of make_window_input_target + * output_target: what the topmost WindowAggPath should return + * wflists: result of find_window_functions + * activeWindows: result of select_active_windows + * + * Note: all Paths in input_rel are expected to return input_target. + */ +static RelOptInfo * +create_window_paths(PlannerInfo *root, + RelOptInfo *input_rel, + PathTarget *input_target, + PathTarget *output_target, + bool output_target_parallel_safe, + WindowFuncLists *wflists, + List *activeWindows) +{ + RelOptInfo *window_rel; + ListCell *lc; + + /* For now, do all work in the (WINDOW, NULL) upperrel */ + window_rel = fetch_upper_rel(root, UPPERREL_WINDOW, NULL); + + /* + * If the input relation is not parallel-safe, then the window relation + * can't be parallel-safe, either. Otherwise, we need to examine the + * target list and active windows for non-parallel-safe constructs. + */ + if (input_rel->consider_parallel && output_target_parallel_safe && + is_parallel_safe(root, (Node *) activeWindows)) + window_rel->consider_parallel = true; + + /* + * If the input rel belongs to a single FDW, so does the window rel. + */ + window_rel->serverid = input_rel->serverid; + window_rel->userid = input_rel->userid; + window_rel->useridiscurrent = input_rel->useridiscurrent; + window_rel->fdwroutine = input_rel->fdwroutine; + + /* + * Consider computing window functions starting from the existing + * cheapest-total path (which will likely require a sort) as well as any + * existing paths that satisfy or partially satisfy root->window_pathkeys. + */ + foreach(lc, input_rel->pathlist) + { + Path *path = (Path *) lfirst(lc); + int presorted_keys; + + if (path == input_rel->cheapest_total_path || + pathkeys_count_contained_in(root->window_pathkeys, path->pathkeys, + &presorted_keys) || + presorted_keys > 0) + create_one_window_path(root, + window_rel, + path, + input_target, + output_target, + wflists, + activeWindows); + } + + /* + * If there is an FDW that's responsible for all baserels of the query, + * let it consider adding ForeignPaths. + */ + if (window_rel->fdwroutine && + window_rel->fdwroutine->GetForeignUpperPaths) + window_rel->fdwroutine->GetForeignUpperPaths(root, UPPERREL_WINDOW, + input_rel, window_rel, + NULL); + + /* Let extensions possibly add some more paths */ + if (create_upper_paths_hook) + (*create_upper_paths_hook) (root, UPPERREL_WINDOW, + input_rel, window_rel, NULL); + + /* Now choose the best path(s) */ + set_cheapest(window_rel); + + return window_rel; +} + +/* + * Stack window-function implementation steps atop the given Path, and + * add the result to window_rel. + * + * window_rel: upperrel to contain result + * path: input Path to use (must return input_target) + * input_target: result of make_window_input_target + * output_target: what the topmost WindowAggPath should return + * wflists: result of find_window_functions + * activeWindows: result of select_active_windows + */ +static void +create_one_window_path(PlannerInfo *root, + RelOptInfo *window_rel, + Path *path, + PathTarget *input_target, + PathTarget *output_target, + WindowFuncLists *wflists, + List *activeWindows) +{ + PathTarget *window_target; + ListCell *l; + List *topqual = NIL; + + /* + * Since each window clause could require a different sort order, we stack + * up a WindowAgg node for each clause, with sort steps between them as + * needed. (We assume that select_active_windows chose a good order for + * executing the clauses in.) + * + * input_target should contain all Vars and Aggs needed for the result. + * (In some cases we wouldn't need to propagate all of these all the way + * to the top, since they might only be needed as inputs to WindowFuncs. + * It's probably not worth trying to optimize that though.) It must also + * contain all window partitioning and sorting expressions, to ensure + * they're computed only once at the bottom of the stack (that's critical + * for volatile functions). As we climb up the stack, we'll add outputs + * for the WindowFuncs computed at each level. + */ + window_target = input_target; + + foreach(l, activeWindows) + { + WindowClause *wc = lfirst_node(WindowClause, l); + List *window_pathkeys; + int presorted_keys; + bool is_sorted; + bool topwindow; + + window_pathkeys = make_pathkeys_for_window(root, + wc, + root->processed_tlist); + + is_sorted = pathkeys_count_contained_in(window_pathkeys, + path->pathkeys, + &presorted_keys); + + /* Sort if necessary */ + if (!is_sorted) + { + /* + * No presorted keys or incremental sort disabled, just perform a + * complete sort. + */ + if (presorted_keys == 0 || !enable_incremental_sort) + path = (Path *) create_sort_path(root, window_rel, + path, + window_pathkeys, + -1.0); + else + { + /* + * Since we have presorted keys and incremental sort is + * enabled, just use incremental sort. + */ + path = (Path *) create_incremental_sort_path(root, + window_rel, + path, + window_pathkeys, + presorted_keys, + -1.0); + } + } + + if (lnext(activeWindows, l)) + { + /* + * Add the current WindowFuncs to the output target for this + * intermediate WindowAggPath. We must copy window_target to + * avoid changing the previous path's target. + * + * Note: a WindowFunc adds nothing to the target's eval costs; but + * we do need to account for the increase in tlist width. + */ + ListCell *lc2; + + window_target = copy_pathtarget(window_target); + foreach(lc2, wflists->windowFuncs[wc->winref]) + { + WindowFunc *wfunc = lfirst_node(WindowFunc, lc2); + + add_column_to_pathtarget(window_target, (Expr *) wfunc, 0); + window_target->width += get_typavgwidth(wfunc->wintype, -1); + } + } + else + { + /* Install the goal target in the topmost WindowAgg */ + window_target = output_target; + } + + /* mark the final item in the list as the top-level window */ + topwindow = foreach_current_index(l) == list_length(activeWindows) - 1; + + /* + * Accumulate all of the runConditions from each intermediate + * WindowClause. The top-level WindowAgg must pass these as a qual so + * that it filters out unwanted tuples correctly. + */ + if (!topwindow) + topqual = list_concat(topqual, wc->runCondition); + + path = (Path *) + create_windowagg_path(root, window_rel, path, window_target, + wflists->windowFuncs[wc->winref], + wc, topwindow ? topqual : NIL, topwindow); + } + + add_path(window_rel, path); +} + +/* + * create_distinct_paths + * + * Build a new upperrel containing Paths for SELECT DISTINCT evaluation. + * + * input_rel: contains the source-data Paths + * + * Note: input paths should already compute the desired pathtarget, since + * Sort/Unique won't project anything. + */ +static RelOptInfo * +create_distinct_paths(PlannerInfo *root, RelOptInfo *input_rel) +{ + RelOptInfo *distinct_rel; + + /* For now, do all work in the (DISTINCT, NULL) upperrel */ + distinct_rel = fetch_upper_rel(root, UPPERREL_DISTINCT, NULL); + + /* + * We don't compute anything at this level, so distinct_rel will be + * parallel-safe if the input rel is parallel-safe. In particular, if + * there is a DISTINCT ON (...) clause, any path for the input_rel will + * output those expressions, and will not be parallel-safe unless those + * expressions are parallel-safe. + */ + distinct_rel->consider_parallel = input_rel->consider_parallel; + + /* + * If the input rel belongs to a single FDW, so does the distinct_rel. + */ + distinct_rel->serverid = input_rel->serverid; + distinct_rel->userid = input_rel->userid; + distinct_rel->useridiscurrent = input_rel->useridiscurrent; + distinct_rel->fdwroutine = input_rel->fdwroutine; + + /* build distinct paths based on input_rel's pathlist */ + create_final_distinct_paths(root, input_rel, distinct_rel); + + /* now build distinct paths based on input_rel's partial_pathlist */ + create_partial_distinct_paths(root, input_rel, distinct_rel); + + /* Give a helpful error if we failed to create any paths */ + if (distinct_rel->pathlist == NIL) + ereport(ERROR, + (errcode(ERRCODE_FEATURE_NOT_SUPPORTED), + errmsg("could not implement DISTINCT"), + errdetail("Some of the datatypes only support hashing, while others only support sorting."))); + + /* + * If there is an FDW that's responsible for all baserels of the query, + * let it consider adding ForeignPaths. + */ + if (distinct_rel->fdwroutine && + distinct_rel->fdwroutine->GetForeignUpperPaths) + distinct_rel->fdwroutine->GetForeignUpperPaths(root, + UPPERREL_DISTINCT, + input_rel, + distinct_rel, + NULL); + + /* Let extensions possibly add some more paths */ + if (create_upper_paths_hook) + (*create_upper_paths_hook) (root, UPPERREL_DISTINCT, input_rel, + distinct_rel, NULL); + + /* Now choose the best path(s) */ + set_cheapest(distinct_rel); + + return distinct_rel; +} + +/* + * create_partial_distinct_paths + * + * Process 'input_rel' partial paths and add unique/aggregate paths to the + * UPPERREL_PARTIAL_DISTINCT rel. For paths created, add Gather/GatherMerge + * paths on top and add a final unique/aggregate path to remove any duplicate + * produced from combining rows from parallel workers. + */ +static void +create_partial_distinct_paths(PlannerInfo *root, RelOptInfo *input_rel, + RelOptInfo *final_distinct_rel) +{ + RelOptInfo *partial_distinct_rel; + Query *parse; + List *distinctExprs; + double numDistinctRows; + Path *cheapest_partial_path; + ListCell *lc; + + /* nothing to do when there are no partial paths in the input rel */ + if (!input_rel->consider_parallel || input_rel->partial_pathlist == NIL) + return; + + parse = root->parse; + + /* can't do parallel DISTINCT ON */ + if (parse->hasDistinctOn) + return; + + partial_distinct_rel = fetch_upper_rel(root, UPPERREL_PARTIAL_DISTINCT, + NULL); + partial_distinct_rel->reltarget = root->upper_targets[UPPERREL_PARTIAL_DISTINCT]; + partial_distinct_rel->consider_parallel = input_rel->consider_parallel; + + /* + * If input_rel belongs to a single FDW, so does the partial_distinct_rel. + */ + partial_distinct_rel->serverid = input_rel->serverid; + partial_distinct_rel->userid = input_rel->userid; + partial_distinct_rel->useridiscurrent = input_rel->useridiscurrent; + partial_distinct_rel->fdwroutine = input_rel->fdwroutine; + + cheapest_partial_path = linitial(input_rel->partial_pathlist); + + distinctExprs = get_sortgrouplist_exprs(parse->distinctClause, + parse->targetList); + + /* estimate how many distinct rows we'll get from each worker */ + numDistinctRows = estimate_num_groups(root, distinctExprs, + cheapest_partial_path->rows, + NULL, NULL); + + /* first try adding unique paths atop of sorted paths */ + if (grouping_is_sortable(parse->distinctClause)) + { + foreach(lc, input_rel->partial_pathlist) + { + Path *path = (Path *) lfirst(lc); + + if (pathkeys_contained_in(root->distinct_pathkeys, path->pathkeys)) + { + add_partial_path(partial_distinct_rel, (Path *) + create_upper_unique_path(root, + partial_distinct_rel, + path, + list_length(root->distinct_pathkeys), + numDistinctRows)); + } + } + } + + /* + * Now try hash aggregate paths, if enabled and hashing is possible. Since + * we're not on the hook to ensure we do our best to create at least one + * path here, we treat enable_hashagg as a hard off-switch rather than the + * slightly softer variant in create_final_distinct_paths. + */ + if (enable_hashagg && grouping_is_hashable(parse->distinctClause)) + { + add_partial_path(partial_distinct_rel, (Path *) + create_agg_path(root, + partial_distinct_rel, + cheapest_partial_path, + cheapest_partial_path->pathtarget, + AGG_HASHED, + AGGSPLIT_SIMPLE, + parse->distinctClause, + NIL, + NULL, + numDistinctRows)); + } + + /* + * If there is an FDW that's responsible for all baserels of the query, + * let it consider adding ForeignPaths. + */ + if (partial_distinct_rel->fdwroutine && + partial_distinct_rel->fdwroutine->GetForeignUpperPaths) + partial_distinct_rel->fdwroutine->GetForeignUpperPaths(root, + UPPERREL_PARTIAL_DISTINCT, + input_rel, + partial_distinct_rel, + NULL); + + /* Let extensions possibly add some more partial paths */ + if (create_upper_paths_hook) + (*create_upper_paths_hook) (root, UPPERREL_PARTIAL_DISTINCT, + input_rel, partial_distinct_rel, NULL); + + if (partial_distinct_rel->partial_pathlist != NIL) + { + generate_gather_paths(root, partial_distinct_rel, true); + set_cheapest(partial_distinct_rel); + + /* + * Finally, create paths to distinctify the final result. This step + * is needed to remove any duplicates due to combining rows from + * parallel workers. + */ + create_final_distinct_paths(root, partial_distinct_rel, + final_distinct_rel); + } +} + +/* + * create_final_distinct_paths + * Create distinct paths in 'distinct_rel' based on 'input_rel' pathlist + * + * input_rel: contains the source-data paths + * distinct_rel: destination relation for storing created paths + */ +static RelOptInfo * +create_final_distinct_paths(PlannerInfo *root, RelOptInfo *input_rel, + RelOptInfo *distinct_rel) +{ + Query *parse = root->parse; + Path *cheapest_input_path = input_rel->cheapest_total_path; + double numDistinctRows; + bool allow_hash; + Path *path; + ListCell *lc; + + /* Estimate number of distinct rows there will be */ + if (parse->groupClause || parse->groupingSets || parse->hasAggs || + root->hasHavingQual) + { + /* + * If there was grouping or aggregation, use the number of input rows + * as the estimated number of DISTINCT rows (ie, assume the input is + * already mostly unique). + */ + numDistinctRows = cheapest_input_path->rows; + } + else + { + /* + * Otherwise, the UNIQUE filter has effects comparable to GROUP BY. + */ + List *distinctExprs; + + distinctExprs = get_sortgrouplist_exprs(parse->distinctClause, + parse->targetList); + numDistinctRows = estimate_num_groups(root, distinctExprs, + cheapest_input_path->rows, + NULL, NULL); + } + + /* + * Consider sort-based implementations of DISTINCT, if possible. + */ + if (grouping_is_sortable(parse->distinctClause)) + { + /* + * First, if we have any adequately-presorted paths, just stick a + * Unique node on those. Then consider doing an explicit sort of the + * cheapest input path and Unique'ing that. + * + * When we have DISTINCT ON, we must sort by the more rigorous of + * DISTINCT and ORDER BY, else it won't have the desired behavior. + * Also, if we do have to do an explicit sort, we might as well use + * the more rigorous ordering to avoid a second sort later. (Note + * that the parser will have ensured that one clause is a prefix of + * the other.) + */ + List *needed_pathkeys; + + if (parse->hasDistinctOn && + list_length(root->distinct_pathkeys) < + list_length(root->sort_pathkeys)) + needed_pathkeys = root->sort_pathkeys; + else + needed_pathkeys = root->distinct_pathkeys; + + foreach(lc, input_rel->pathlist) + { + Path *path = (Path *) lfirst(lc); + + if (pathkeys_contained_in(needed_pathkeys, path->pathkeys)) + { + add_path(distinct_rel, (Path *) + create_upper_unique_path(root, distinct_rel, + path, + list_length(root->distinct_pathkeys), + numDistinctRows)); + } + } + + /* For explicit-sort case, always use the more rigorous clause */ + if (list_length(root->distinct_pathkeys) < + list_length(root->sort_pathkeys)) + { + needed_pathkeys = root->sort_pathkeys; + /* Assert checks that parser didn't mess up... */ + Assert(pathkeys_contained_in(root->distinct_pathkeys, + needed_pathkeys)); + } + else + needed_pathkeys = root->distinct_pathkeys; + + path = cheapest_input_path; + if (!pathkeys_contained_in(needed_pathkeys, path->pathkeys)) + path = (Path *) create_sort_path(root, distinct_rel, + path, + needed_pathkeys, + -1.0); + + add_path(distinct_rel, (Path *) + create_upper_unique_path(root, distinct_rel, + path, + list_length(root->distinct_pathkeys), + numDistinctRows)); + } + + /* + * Consider hash-based implementations of DISTINCT, if possible. + * + * If we were not able to make any other types of path, we *must* hash or + * die trying. If we do have other choices, there are two things that + * should prevent selection of hashing: if the query uses DISTINCT ON + * (because it won't really have the expected behavior if we hash), or if + * enable_hashagg is off. + * + * Note: grouping_is_hashable() is much more expensive to check than the + * other gating conditions, so we want to do it last. + */ + if (distinct_rel->pathlist == NIL) + allow_hash = true; /* we have no alternatives */ + else if (parse->hasDistinctOn || !enable_hashagg) + allow_hash = false; /* policy-based decision not to hash */ + else + allow_hash = true; /* default */ + + if (allow_hash && grouping_is_hashable(parse->distinctClause)) + { + /* Generate hashed aggregate path --- no sort needed */ + add_path(distinct_rel, (Path *) + create_agg_path(root, + distinct_rel, + cheapest_input_path, + cheapest_input_path->pathtarget, + AGG_HASHED, + AGGSPLIT_SIMPLE, + parse->distinctClause, + NIL, + NULL, + numDistinctRows)); + } + + return distinct_rel; +} + +/* + * create_ordered_paths + * + * Build a new upperrel containing Paths for ORDER BY evaluation. + * + * All paths in the result must satisfy the ORDER BY ordering. + * The only new paths we need consider are an explicit full sort + * and incremental sort on the cheapest-total existing path. + * + * input_rel: contains the source-data Paths + * target: the output tlist the result Paths must emit + * limit_tuples: estimated bound on the number of output tuples, + * or -1 if no LIMIT or couldn't estimate + * + * XXX This only looks at sort_pathkeys. I wonder if it needs to look at the + * other pathkeys (grouping, ...) like generate_useful_gather_paths. + */ +static RelOptInfo * +create_ordered_paths(PlannerInfo *root, + RelOptInfo *input_rel, + PathTarget *target, + bool target_parallel_safe, + double limit_tuples) +{ + Path *cheapest_input_path = input_rel->cheapest_total_path; + RelOptInfo *ordered_rel; + ListCell *lc; + + /* For now, do all work in the (ORDERED, NULL) upperrel */ + ordered_rel = fetch_upper_rel(root, UPPERREL_ORDERED, NULL); + + /* + * If the input relation is not parallel-safe, then the ordered relation + * can't be parallel-safe, either. Otherwise, it's parallel-safe if the + * target list is parallel-safe. + */ + if (input_rel->consider_parallel && target_parallel_safe) + ordered_rel->consider_parallel = true; + + /* + * If the input rel belongs to a single FDW, so does the ordered_rel. + */ + ordered_rel->serverid = input_rel->serverid; + ordered_rel->userid = input_rel->userid; + ordered_rel->useridiscurrent = input_rel->useridiscurrent; + ordered_rel->fdwroutine = input_rel->fdwroutine; + + foreach(lc, input_rel->pathlist) + { + Path *input_path = (Path *) lfirst(lc); + Path *sorted_path = input_path; + bool is_sorted; + int presorted_keys; + + is_sorted = pathkeys_count_contained_in(root->sort_pathkeys, + input_path->pathkeys, &presorted_keys); + + if (is_sorted) + { + /* Use the input path as is, but add a projection step if needed */ + if (sorted_path->pathtarget != target) + sorted_path = apply_projection_to_path(root, ordered_rel, + sorted_path, target); + + add_path(ordered_rel, sorted_path); + } + else + { + /* + * Try adding an explicit sort, but only to the cheapest total + * path since a full sort should generally add the same cost to + * all paths. + */ + if (input_path == cheapest_input_path) + { + /* + * Sort the cheapest input path. An explicit sort here can + * take advantage of LIMIT. + */ + sorted_path = (Path *) create_sort_path(root, + ordered_rel, + input_path, + root->sort_pathkeys, + limit_tuples); + /* Add projection step if needed */ + if (sorted_path->pathtarget != target) + sorted_path = apply_projection_to_path(root, ordered_rel, + sorted_path, target); + + add_path(ordered_rel, sorted_path); + } + + /* + * If incremental sort is enabled, then try it as well. Unlike + * with regular sorts, we can't just look at the cheapest path, + * because the cost of incremental sort depends on how well + * presorted the path is. Additionally incremental sort may enable + * a cheaper startup path to win out despite higher total cost. + */ + if (!enable_incremental_sort) + continue; + + /* Likewise, if the path can't be used for incremental sort. */ + if (!presorted_keys) + continue; + + /* Also consider incremental sort. */ + sorted_path = (Path *) create_incremental_sort_path(root, + ordered_rel, + input_path, + root->sort_pathkeys, + presorted_keys, + limit_tuples); + + /* Add projection step if needed */ + if (sorted_path->pathtarget != target) + sorted_path = apply_projection_to_path(root, ordered_rel, + sorted_path, target); + + add_path(ordered_rel, sorted_path); + } + } + + /* + * generate_gather_paths() will have already generated a simple Gather + * path for the best parallel path, if any, and the loop above will have + * considered sorting it. Similarly, generate_gather_paths() will also + * have generated order-preserving Gather Merge plans which can be used + * without sorting if they happen to match the sort_pathkeys, and the loop + * above will have handled those as well. However, there's one more + * possibility: it may make sense to sort the cheapest partial path + * according to the required output order and then use Gather Merge. + */ + if (ordered_rel->consider_parallel && root->sort_pathkeys != NIL && + input_rel->partial_pathlist != NIL) + { + Path *cheapest_partial_path; + + cheapest_partial_path = linitial(input_rel->partial_pathlist); + + /* + * If cheapest partial path doesn't need a sort, this is redundant + * with what's already been tried. + */ + if (!pathkeys_contained_in(root->sort_pathkeys, + cheapest_partial_path->pathkeys)) + { + Path *path; + double total_groups; + + path = (Path *) create_sort_path(root, + ordered_rel, + cheapest_partial_path, + root->sort_pathkeys, + limit_tuples); + + total_groups = cheapest_partial_path->rows * + cheapest_partial_path->parallel_workers; + path = (Path *) + create_gather_merge_path(root, ordered_rel, + path, + path->pathtarget, + root->sort_pathkeys, NULL, + &total_groups); + + /* Add projection step if needed */ + if (path->pathtarget != target) + path = apply_projection_to_path(root, ordered_rel, + path, target); + + add_path(ordered_rel, path); + } + + /* + * Consider incremental sort with a gather merge on partial paths. + * + * We can also skip the entire loop when we only have a single-item + * sort_pathkeys because then we can't possibly have a presorted + * prefix of the list without having the list be fully sorted. + */ + if (enable_incremental_sort && list_length(root->sort_pathkeys) > 1) + { + ListCell *lc; + + foreach(lc, input_rel->partial_pathlist) + { + Path *input_path = (Path *) lfirst(lc); + Path *sorted_path; + bool is_sorted; + int presorted_keys; + double total_groups; + + /* + * We don't care if this is the cheapest partial path - we + * can't simply skip it, because it may be partially sorted in + * which case we want to consider adding incremental sort + * (instead of full sort, which is what happens above). + */ + + is_sorted = pathkeys_count_contained_in(root->sort_pathkeys, + input_path->pathkeys, + &presorted_keys); + + /* No point in adding incremental sort on fully sorted paths. */ + if (is_sorted) + continue; + + if (presorted_keys == 0) + continue; + + /* Since we have presorted keys, consider incremental sort. */ + sorted_path = (Path *) create_incremental_sort_path(root, + ordered_rel, + input_path, + root->sort_pathkeys, + presorted_keys, + limit_tuples); + total_groups = input_path->rows * + input_path->parallel_workers; + sorted_path = (Path *) + create_gather_merge_path(root, ordered_rel, + sorted_path, + sorted_path->pathtarget, + root->sort_pathkeys, NULL, + &total_groups); + + /* Add projection step if needed */ + if (sorted_path->pathtarget != target) + sorted_path = apply_projection_to_path(root, ordered_rel, + sorted_path, target); + + add_path(ordered_rel, sorted_path); + } + } + } + + /* + * If there is an FDW that's responsible for all baserels of the query, + * let it consider adding ForeignPaths. + */ + if (ordered_rel->fdwroutine && + ordered_rel->fdwroutine->GetForeignUpperPaths) + ordered_rel->fdwroutine->GetForeignUpperPaths(root, UPPERREL_ORDERED, + input_rel, ordered_rel, + NULL); + + /* Let extensions possibly add some more paths */ + if (create_upper_paths_hook) + (*create_upper_paths_hook) (root, UPPERREL_ORDERED, + input_rel, ordered_rel, NULL); + + /* + * No need to bother with set_cheapest here; grouping_planner does not + * need us to do it. + */ + Assert(ordered_rel->pathlist != NIL); + + return ordered_rel; +} + + +/* + * make_group_input_target + * Generate appropriate PathTarget for initial input to grouping nodes. + * + * If there is grouping or aggregation, the scan/join subplan cannot emit + * the query's final targetlist; for example, it certainly can't emit any + * aggregate function calls. This routine generates the correct target + * for the scan/join subplan. + * + * The query target list passed from the parser already contains entries + * for all ORDER BY and GROUP BY expressions, but it will not have entries + * for variables used only in HAVING clauses; so we need to add those + * variables to the subplan target list. Also, we flatten all expressions + * except GROUP BY items into their component variables; other expressions + * will be computed by the upper plan nodes rather than by the subplan. + * For example, given a query like + * SELECT a+b,SUM(c+d) FROM table GROUP BY a+b; + * we want to pass this targetlist to the subplan: + * a+b,c,d + * where the a+b target will be used by the Sort/Group steps, and the + * other targets will be used for computing the final results. + * + * 'final_target' is the query's final target list (in PathTarget form) + * + * The result is the PathTarget to be computed by the Paths returned from + * query_planner(). + */ +static PathTarget * +make_group_input_target(PlannerInfo *root, PathTarget *final_target) +{ + Query *parse = root->parse; + PathTarget *input_target; + List *non_group_cols; + List *non_group_vars; + int i; + ListCell *lc; + + /* + * We must build a target containing all grouping columns, plus any other + * Vars mentioned in the query's targetlist and HAVING qual. + */ + input_target = create_empty_pathtarget(); + non_group_cols = NIL; + + i = 0; + foreach(lc, final_target->exprs) + { + Expr *expr = (Expr *) lfirst(lc); + Index sgref = get_pathtarget_sortgroupref(final_target, i); + + if (sgref && parse->groupClause && + get_sortgroupref_clause_noerr(sgref, parse->groupClause) != NULL) + { + /* + * It's a grouping column, so add it to the input target as-is. + */ + add_column_to_pathtarget(input_target, expr, sgref); + } + else + { + /* + * Non-grouping column, so just remember the expression for later + * call to pull_var_clause. + */ + non_group_cols = lappend(non_group_cols, expr); + } + + i++; + } + + /* + * If there's a HAVING clause, we'll need the Vars it uses, too. + */ + if (parse->havingQual) + non_group_cols = lappend(non_group_cols, parse->havingQual); + + /* + * Pull out all the Vars mentioned in non-group cols (plus HAVING), and + * add them to the input target if not already present. (A Var used + * directly as a GROUP BY item will be present already.) Note this + * includes Vars used in resjunk items, so we are covering the needs of + * ORDER BY and window specifications. Vars used within Aggrefs and + * WindowFuncs will be pulled out here, too. + */ + non_group_vars = pull_var_clause((Node *) non_group_cols, + PVC_RECURSE_AGGREGATES | + PVC_RECURSE_WINDOWFUNCS | + PVC_INCLUDE_PLACEHOLDERS); + add_new_columns_to_pathtarget(input_target, non_group_vars); + + /* clean up cruft */ + list_free(non_group_vars); + list_free(non_group_cols); + + /* XXX this causes some redundant cost calculation ... */ + return set_pathtarget_cost_width(root, input_target); +} + +/* + * make_partial_grouping_target + * Generate appropriate PathTarget for output of partial aggregate + * (or partial grouping, if there are no aggregates) nodes. + * + * A partial aggregation node needs to emit all the same aggregates that + * a regular aggregation node would, plus any aggregates used in HAVING; + * except that the Aggref nodes should be marked as partial aggregates. + * + * In addition, we'd better emit any Vars and PlaceHolderVars that are + * used outside of Aggrefs in the aggregation tlist and HAVING. (Presumably, + * these would be Vars that are grouped by or used in grouping expressions.) + * + * grouping_target is the tlist to be emitted by the topmost aggregation step. + * havingQual represents the HAVING clause. + */ +static PathTarget * +make_partial_grouping_target(PlannerInfo *root, + PathTarget *grouping_target, + Node *havingQual) +{ + Query *parse = root->parse; + PathTarget *partial_target; + List *non_group_cols; + List *non_group_exprs; + int i; + ListCell *lc; + + partial_target = create_empty_pathtarget(); + non_group_cols = NIL; + + i = 0; + foreach(lc, grouping_target->exprs) + { + Expr *expr = (Expr *) lfirst(lc); + Index sgref = get_pathtarget_sortgroupref(grouping_target, i); + + if (sgref && parse->groupClause && + get_sortgroupref_clause_noerr(sgref, parse->groupClause) != NULL) + { + /* + * It's a grouping column, so add it to the partial_target as-is. + * (This allows the upper agg step to repeat the grouping calcs.) + */ + add_column_to_pathtarget(partial_target, expr, sgref); + } + else + { + /* + * Non-grouping column, so just remember the expression for later + * call to pull_var_clause. + */ + non_group_cols = lappend(non_group_cols, expr); + } + + i++; + } + + /* + * If there's a HAVING clause, we'll need the Vars/Aggrefs it uses, too. + */ + if (havingQual) + non_group_cols = lappend(non_group_cols, havingQual); + + /* + * Pull out all the Vars, PlaceHolderVars, and Aggrefs mentioned in + * non-group cols (plus HAVING), and add them to the partial_target if not + * already present. (An expression used directly as a GROUP BY item will + * be present already.) Note this includes Vars used in resjunk items, so + * we are covering the needs of ORDER BY and window specifications. + */ + non_group_exprs = pull_var_clause((Node *) non_group_cols, + PVC_INCLUDE_AGGREGATES | + PVC_RECURSE_WINDOWFUNCS | + PVC_INCLUDE_PLACEHOLDERS); + + add_new_columns_to_pathtarget(partial_target, non_group_exprs); + + /* + * Adjust Aggrefs to put them in partial mode. At this point all Aggrefs + * are at the top level of the target list, so we can just scan the list + * rather than recursing through the expression trees. + */ + foreach(lc, partial_target->exprs) + { + Aggref *aggref = (Aggref *) lfirst(lc); + + if (IsA(aggref, Aggref)) + { + Aggref *newaggref; + + /* + * We shouldn't need to copy the substructure of the Aggref node, + * but flat-copy the node itself to avoid damaging other trees. + */ + newaggref = makeNode(Aggref); + memcpy(newaggref, aggref, sizeof(Aggref)); + + /* For now, assume serialization is required */ + mark_partial_aggref(newaggref, AGGSPLIT_INITIAL_SERIAL); + + lfirst(lc) = newaggref; + } + } + + /* clean up cruft */ + list_free(non_group_exprs); + list_free(non_group_cols); + + /* XXX this causes some redundant cost calculation ... */ + return set_pathtarget_cost_width(root, partial_target); +} + +/* + * mark_partial_aggref + * Adjust an Aggref to make it represent a partial-aggregation step. + * + * The Aggref node is modified in-place; caller must do any copying required. + */ +void +mark_partial_aggref(Aggref *agg, AggSplit aggsplit) +{ + /* aggtranstype should be computed by this point */ + Assert(OidIsValid(agg->aggtranstype)); + /* ... but aggsplit should still be as the parser left it */ + Assert(agg->aggsplit == AGGSPLIT_SIMPLE); + + /* Mark the Aggref with the intended partial-aggregation mode */ + agg->aggsplit = aggsplit; + + /* + * Adjust result type if needed. Normally, a partial aggregate returns + * the aggregate's transition type; but if that's INTERNAL and we're + * serializing, it returns BYTEA instead. + */ + if (DO_AGGSPLIT_SKIPFINAL(aggsplit)) + { + if (agg->aggtranstype == INTERNALOID && DO_AGGSPLIT_SERIALIZE(aggsplit)) + agg->aggtype = BYTEAOID; + else + agg->aggtype = agg->aggtranstype; + } +} + +/* + * postprocess_setop_tlist + * Fix up targetlist returned by plan_set_operations(). + * + * We need to transpose sort key info from the orig_tlist into new_tlist. + * NOTE: this would not be good enough if we supported resjunk sort keys + * for results of set operations --- then, we'd need to project a whole + * new tlist to evaluate the resjunk columns. For now, just ereport if we + * find any resjunk columns in orig_tlist. + */ +static List * +postprocess_setop_tlist(List *new_tlist, List *orig_tlist) +{ + ListCell *l; + ListCell *orig_tlist_item = list_head(orig_tlist); + + foreach(l, new_tlist) + { + TargetEntry *new_tle = lfirst_node(TargetEntry, l); + TargetEntry *orig_tle; + + /* ignore resjunk columns in setop result */ + if (new_tle->resjunk) + continue; + + Assert(orig_tlist_item != NULL); + orig_tle = lfirst_node(TargetEntry, orig_tlist_item); + orig_tlist_item = lnext(orig_tlist, orig_tlist_item); + if (orig_tle->resjunk) /* should not happen */ + elog(ERROR, "resjunk output columns are not implemented"); + Assert(new_tle->resno == orig_tle->resno); + new_tle->ressortgroupref = orig_tle->ressortgroupref; + } + if (orig_tlist_item != NULL) + elog(ERROR, "resjunk output columns are not implemented"); + return new_tlist; +} + +/* + * select_active_windows + * Create a list of the "active" window clauses (ie, those referenced + * by non-deleted WindowFuncs) in the order they are to be executed. + */ +static List * +select_active_windows(PlannerInfo *root, WindowFuncLists *wflists) +{ + List *windowClause = root->parse->windowClause; + List *result = NIL; + ListCell *lc; + int nActive = 0; + WindowClauseSortData *actives = palloc(sizeof(WindowClauseSortData) + * list_length(windowClause)); + + /* First, construct an array of the active windows */ + foreach(lc, windowClause) + { + WindowClause *wc = lfirst_node(WindowClause, lc); + + /* It's only active if wflists shows some related WindowFuncs */ + Assert(wc->winref <= wflists->maxWinRef); + if (wflists->windowFuncs[wc->winref] == NIL) + continue; + + actives[nActive].wc = wc; /* original clause */ + + /* + * For sorting, we want the list of partition keys followed by the + * list of sort keys. But pathkeys construction will remove duplicates + * between the two, so we can as well (even though we can't detect all + * of the duplicates, since some may come from ECs - that might mean + * we miss optimization chances here). We must, however, ensure that + * the order of entries is preserved with respect to the ones we do + * keep. + * + * partitionClause and orderClause had their own duplicates removed in + * parse analysis, so we're only concerned here with removing + * orderClause entries that also appear in partitionClause. + */ + actives[nActive].uniqueOrder = + list_concat_unique(list_copy(wc->partitionClause), + wc->orderClause); + nActive++; + } + + /* + * Sort active windows by their partitioning/ordering clauses, ignoring + * any framing clauses, so that the windows that need the same sorting are + * adjacent in the list. When we come to generate paths, this will avoid + * inserting additional Sort nodes. + * + * This is how we implement a specific requirement from the SQL standard, + * which says that when two or more windows are order-equivalent (i.e. + * have matching partition and order clauses, even if their names or + * framing clauses differ), then all peer rows must be presented in the + * same order in all of them. If we allowed multiple sort nodes for such + * cases, we'd risk having the peer rows end up in different orders in + * equivalent windows due to sort instability. (See General Rule 4 of + * <window clause> in SQL2008 - SQL2016.) + * + * Additionally, if the entire list of clauses of one window is a prefix + * of another, put first the window with stronger sorting requirements. + * This way we will first sort for stronger window, and won't have to sort + * again for the weaker one. + */ + qsort(actives, nActive, sizeof(WindowClauseSortData), common_prefix_cmp); + + /* build ordered list of the original WindowClause nodes */ + for (int i = 0; i < nActive; i++) + result = lappend(result, actives[i].wc); + + pfree(actives); + + return result; +} + +/* + * common_prefix_cmp + * QSort comparison function for WindowClauseSortData + * + * Sort the windows by the required sorting clauses. First, compare the sort + * clauses themselves. Second, if one window's clauses are a prefix of another + * one's clauses, put the window with more sort clauses first. + */ +static int +common_prefix_cmp(const void *a, const void *b) +{ + const WindowClauseSortData *wcsa = a; + const WindowClauseSortData *wcsb = b; + ListCell *item_a; + ListCell *item_b; + + forboth(item_a, wcsa->uniqueOrder, item_b, wcsb->uniqueOrder) + { + SortGroupClause *sca = lfirst_node(SortGroupClause, item_a); + SortGroupClause *scb = lfirst_node(SortGroupClause, item_b); + + if (sca->tleSortGroupRef > scb->tleSortGroupRef) + return -1; + else if (sca->tleSortGroupRef < scb->tleSortGroupRef) + return 1; + else if (sca->sortop > scb->sortop) + return -1; + else if (sca->sortop < scb->sortop) + return 1; + else if (sca->nulls_first && !scb->nulls_first) + return -1; + else if (!sca->nulls_first && scb->nulls_first) + return 1; + /* no need to compare eqop, since it is fully determined by sortop */ + } + + if (list_length(wcsa->uniqueOrder) > list_length(wcsb->uniqueOrder)) + return -1; + else if (list_length(wcsa->uniqueOrder) < list_length(wcsb->uniqueOrder)) + return 1; + + return 0; +} + +/* + * make_window_input_target + * Generate appropriate PathTarget for initial input to WindowAgg nodes. + * + * When the query has window functions, this function computes the desired + * target to be computed by the node just below the first WindowAgg. + * This tlist must contain all values needed to evaluate the window functions, + * compute the final target list, and perform any required final sort step. + * If multiple WindowAggs are needed, each intermediate one adds its window + * function results onto this base tlist; only the topmost WindowAgg computes + * the actual desired target list. + * + * This function is much like make_group_input_target, though not quite enough + * like it to share code. As in that function, we flatten most expressions + * into their component variables. But we do not want to flatten window + * PARTITION BY/ORDER BY clauses, since that might result in multiple + * evaluations of them, which would be bad (possibly even resulting in + * inconsistent answers, if they contain volatile functions). + * Also, we must not flatten GROUP BY clauses that were left unflattened by + * make_group_input_target, because we may no longer have access to the + * individual Vars in them. + * + * Another key difference from make_group_input_target is that we don't + * flatten Aggref expressions, since those are to be computed below the + * window functions and just referenced like Vars above that. + * + * 'final_target' is the query's final target list (in PathTarget form) + * 'activeWindows' is the list of active windows previously identified by + * select_active_windows. + * + * The result is the PathTarget to be computed by the plan node immediately + * below the first WindowAgg node. + */ +static PathTarget * +make_window_input_target(PlannerInfo *root, + PathTarget *final_target, + List *activeWindows) +{ + Query *parse = root->parse; + PathTarget *input_target; + Bitmapset *sgrefs; + List *flattenable_cols; + List *flattenable_vars; + int i; + ListCell *lc; + + Assert(parse->hasWindowFuncs); + + /* + * Collect the sortgroupref numbers of window PARTITION/ORDER BY clauses + * into a bitmapset for convenient reference below. + */ + sgrefs = NULL; + foreach(lc, activeWindows) + { + WindowClause *wc = lfirst_node(WindowClause, lc); + ListCell *lc2; + + foreach(lc2, wc->partitionClause) + { + SortGroupClause *sortcl = lfirst_node(SortGroupClause, lc2); + + sgrefs = bms_add_member(sgrefs, sortcl->tleSortGroupRef); + } + foreach(lc2, wc->orderClause) + { + SortGroupClause *sortcl = lfirst_node(SortGroupClause, lc2); + + sgrefs = bms_add_member(sgrefs, sortcl->tleSortGroupRef); + } + } + + /* Add in sortgroupref numbers of GROUP BY clauses, too */ + foreach(lc, parse->groupClause) + { + SortGroupClause *grpcl = lfirst_node(SortGroupClause, lc); + + sgrefs = bms_add_member(sgrefs, grpcl->tleSortGroupRef); + } + + /* + * Construct a target containing all the non-flattenable targetlist items, + * and save aside the others for a moment. + */ + input_target = create_empty_pathtarget(); + flattenable_cols = NIL; + + i = 0; + foreach(lc, final_target->exprs) + { + Expr *expr = (Expr *) lfirst(lc); + Index sgref = get_pathtarget_sortgroupref(final_target, i); + + /* + * Don't want to deconstruct window clauses or GROUP BY items. (Note + * that such items can't contain window functions, so it's okay to + * compute them below the WindowAgg nodes.) + */ + if (sgref != 0 && bms_is_member(sgref, sgrefs)) + { + /* + * Don't want to deconstruct this value, so add it to the input + * target as-is. + */ + add_column_to_pathtarget(input_target, expr, sgref); + } + else + { + /* + * Column is to be flattened, so just remember the expression for + * later call to pull_var_clause. + */ + flattenable_cols = lappend(flattenable_cols, expr); + } + + i++; + } + + /* + * Pull out all the Vars and Aggrefs mentioned in flattenable columns, and + * add them to the input target if not already present. (Some might be + * there already because they're used directly as window/group clauses.) + * + * Note: it's essential to use PVC_INCLUDE_AGGREGATES here, so that any + * Aggrefs are placed in the Agg node's tlist and not left to be computed + * at higher levels. On the other hand, we should recurse into + * WindowFuncs to make sure their input expressions are available. + */ + flattenable_vars = pull_var_clause((Node *) flattenable_cols, + PVC_INCLUDE_AGGREGATES | + PVC_RECURSE_WINDOWFUNCS | + PVC_INCLUDE_PLACEHOLDERS); + add_new_columns_to_pathtarget(input_target, flattenable_vars); + + /* clean up cruft */ + list_free(flattenable_vars); + list_free(flattenable_cols); + + /* XXX this causes some redundant cost calculation ... */ + return set_pathtarget_cost_width(root, input_target); +} + +/* + * make_pathkeys_for_window + * Create a pathkeys list describing the required input ordering + * for the given WindowClause. + * + * The required ordering is first the PARTITION keys, then the ORDER keys. + * In the future we might try to implement windowing using hashing, in which + * case the ordering could be relaxed, but for now we always sort. + */ +static List * +make_pathkeys_for_window(PlannerInfo *root, WindowClause *wc, + List *tlist) +{ + List *window_pathkeys; + List *window_sortclauses; + + /* Throw error if can't sort */ + if (!grouping_is_sortable(wc->partitionClause)) + ereport(ERROR, + (errcode(ERRCODE_FEATURE_NOT_SUPPORTED), + errmsg("could not implement window PARTITION BY"), + errdetail("Window partitioning columns must be of sortable datatypes."))); + if (!grouping_is_sortable(wc->orderClause)) + ereport(ERROR, + (errcode(ERRCODE_FEATURE_NOT_SUPPORTED), + errmsg("could not implement window ORDER BY"), + errdetail("Window ordering columns must be of sortable datatypes."))); + + /* Okay, make the combined pathkeys */ + window_sortclauses = list_concat_copy(wc->partitionClause, wc->orderClause); + window_pathkeys = make_pathkeys_for_sortclauses(root, + window_sortclauses, + tlist); + list_free(window_sortclauses); + return window_pathkeys; +} + +/* + * make_sort_input_target + * Generate appropriate PathTarget for initial input to Sort step. + * + * If the query has ORDER BY, this function chooses the target to be computed + * by the node just below the Sort (and DISTINCT, if any, since Unique can't + * project) steps. This might or might not be identical to the query's final + * output target. + * + * The main argument for keeping the sort-input tlist the same as the final + * is that we avoid a separate projection node (which will be needed if + * they're different, because Sort can't project). However, there are also + * advantages to postponing tlist evaluation till after the Sort: it ensures + * a consistent order of evaluation for any volatile functions in the tlist, + * and if there's also a LIMIT, we can stop the query without ever computing + * tlist functions for later rows, which is beneficial for both volatile and + * expensive functions. + * + * Our current policy is to postpone volatile expressions till after the sort + * unconditionally (assuming that that's possible, ie they are in plain tlist + * columns and not ORDER BY/GROUP BY/DISTINCT columns). We also prefer to + * postpone set-returning expressions, because running them beforehand would + * bloat the sort dataset, and because it might cause unexpected output order + * if the sort isn't stable. However there's a constraint on that: all SRFs + * in the tlist should be evaluated at the same plan step, so that they can + * run in sync in nodeProjectSet. So if any SRFs are in sort columns, we + * mustn't postpone any SRFs. (Note that in principle that policy should + * probably get applied to the group/window input targetlists too, but we + * have not done that historically.) Lastly, expensive expressions are + * postponed if there is a LIMIT, or if root->tuple_fraction shows that + * partial evaluation of the query is possible (if neither is true, we expect + * to have to evaluate the expressions for every row anyway), or if there are + * any volatile or set-returning expressions (since once we've put in a + * projection at all, it won't cost any more to postpone more stuff). + * + * Another issue that could potentially be considered here is that + * evaluating tlist expressions could result in data that's either wider + * or narrower than the input Vars, thus changing the volume of data that + * has to go through the Sort. However, we usually have only a very bad + * idea of the output width of any expression more complex than a Var, + * so for now it seems too risky to try to optimize on that basis. + * + * Note that if we do produce a modified sort-input target, and then the + * query ends up not using an explicit Sort, no particular harm is done: + * we'll initially use the modified target for the preceding path nodes, + * but then change them to the final target with apply_projection_to_path. + * Moreover, in such a case the guarantees about evaluation order of + * volatile functions still hold, since the rows are sorted already. + * + * This function has some things in common with make_group_input_target and + * make_window_input_target, though the detailed rules for what to do are + * different. We never flatten/postpone any grouping or ordering columns; + * those are needed before the sort. If we do flatten a particular + * expression, we leave Aggref and WindowFunc nodes alone, since those were + * computed earlier. + * + * 'final_target' is the query's final target list (in PathTarget form) + * 'have_postponed_srfs' is an output argument, see below + * + * The result is the PathTarget to be computed by the plan node immediately + * below the Sort step (and the Distinct step, if any). This will be + * exactly final_target if we decide a projection step wouldn't be helpful. + * + * In addition, *have_postponed_srfs is set to true if we choose to postpone + * any set-returning functions to after the Sort. + */ +static PathTarget * +make_sort_input_target(PlannerInfo *root, + PathTarget *final_target, + bool *have_postponed_srfs) +{ + Query *parse = root->parse; + PathTarget *input_target; + int ncols; + bool *col_is_srf; + bool *postpone_col; + bool have_srf; + bool have_volatile; + bool have_expensive; + bool have_srf_sortcols; + bool postpone_srfs; + List *postponable_cols; + List *postponable_vars; + int i; + ListCell *lc; + + /* Shouldn't get here unless query has ORDER BY */ + Assert(parse->sortClause); + + *have_postponed_srfs = false; /* default result */ + + /* Inspect tlist and collect per-column information */ + ncols = list_length(final_target->exprs); + col_is_srf = (bool *) palloc0(ncols * sizeof(bool)); + postpone_col = (bool *) palloc0(ncols * sizeof(bool)); + have_srf = have_volatile = have_expensive = have_srf_sortcols = false; + + i = 0; + foreach(lc, final_target->exprs) + { + Expr *expr = (Expr *) lfirst(lc); + + /* + * If the column has a sortgroupref, assume it has to be evaluated + * before sorting. Generally such columns would be ORDER BY, GROUP + * BY, etc targets. One exception is columns that were removed from + * GROUP BY by remove_useless_groupby_columns() ... but those would + * only be Vars anyway. There don't seem to be any cases where it + * would be worth the trouble to double-check. + */ + if (get_pathtarget_sortgroupref(final_target, i) == 0) + { + /* + * Check for SRF or volatile functions. Check the SRF case first + * because we must know whether we have any postponed SRFs. + */ + if (parse->hasTargetSRFs && + expression_returns_set((Node *) expr)) + { + /* We'll decide below whether these are postponable */ + col_is_srf[i] = true; + have_srf = true; + } + else if (contain_volatile_functions((Node *) expr)) + { + /* Unconditionally postpone */ + postpone_col[i] = true; + have_volatile = true; + } + else + { + /* + * Else check the cost. XXX it's annoying to have to do this + * when set_pathtarget_cost_width() just did it. Refactor to + * allow sharing the work? + */ + QualCost cost; + + cost_qual_eval_node(&cost, (Node *) expr, root); + + /* + * We arbitrarily define "expensive" as "more than 10X + * cpu_operator_cost". Note this will take in any PL function + * with default cost. + */ + if (cost.per_tuple > 10 * cpu_operator_cost) + { + postpone_col[i] = true; + have_expensive = true; + } + } + } + else + { + /* For sortgroupref cols, just check if any contain SRFs */ + if (!have_srf_sortcols && + parse->hasTargetSRFs && + expression_returns_set((Node *) expr)) + have_srf_sortcols = true; + } + + i++; + } + + /* + * We can postpone SRFs if we have some but none are in sortgroupref cols. + */ + postpone_srfs = (have_srf && !have_srf_sortcols); + + /* + * If we don't need a post-sort projection, just return final_target. + */ + if (!(postpone_srfs || have_volatile || + (have_expensive && + (parse->limitCount || root->tuple_fraction > 0)))) + return final_target; + + /* + * Report whether the post-sort projection will contain set-returning + * functions. This is important because it affects whether the Sort can + * rely on the query's LIMIT (if any) to bound the number of rows it needs + * to return. + */ + *have_postponed_srfs = postpone_srfs; + + /* + * Construct the sort-input target, taking all non-postponable columns and + * then adding Vars, PlaceHolderVars, Aggrefs, and WindowFuncs found in + * the postponable ones. + */ + input_target = create_empty_pathtarget(); + postponable_cols = NIL; + + i = 0; + foreach(lc, final_target->exprs) + { + Expr *expr = (Expr *) lfirst(lc); + + if (postpone_col[i] || (postpone_srfs && col_is_srf[i])) + postponable_cols = lappend(postponable_cols, expr); + else + add_column_to_pathtarget(input_target, expr, + get_pathtarget_sortgroupref(final_target, i)); + + i++; + } + + /* + * Pull out all the Vars, Aggrefs, and WindowFuncs mentioned in + * postponable columns, and add them to the sort-input target if not + * already present. (Some might be there already.) We mustn't + * deconstruct Aggrefs or WindowFuncs here, since the projection node + * would be unable to recompute them. + */ + postponable_vars = pull_var_clause((Node *) postponable_cols, + PVC_INCLUDE_AGGREGATES | + PVC_INCLUDE_WINDOWFUNCS | + PVC_INCLUDE_PLACEHOLDERS); + add_new_columns_to_pathtarget(input_target, postponable_vars); + + /* clean up cruft */ + list_free(postponable_vars); + list_free(postponable_cols); + + /* XXX this represents even more redundant cost calculation ... */ + return set_pathtarget_cost_width(root, input_target); +} + +/* + * get_cheapest_fractional_path + * Find the cheapest path for retrieving a specified fraction of all + * the tuples expected to be returned by the given relation. + * + * We interpret tuple_fraction the same way as grouping_planner. + * + * We assume set_cheapest() has been run on the given rel. + */ +Path * +get_cheapest_fractional_path(RelOptInfo *rel, double tuple_fraction) +{ + Path *best_path = rel->cheapest_total_path; + ListCell *l; + + /* If all tuples will be retrieved, just return the cheapest-total path */ + if (tuple_fraction <= 0.0) + return best_path; + + /* Convert absolute # of tuples to a fraction; no need to clamp to 0..1 */ + if (tuple_fraction >= 1.0 && best_path->rows > 0) + tuple_fraction /= best_path->rows; + + foreach(l, rel->pathlist) + { + Path *path = (Path *) lfirst(l); + + if (path == rel->cheapest_total_path || + compare_fractional_path_costs(best_path, path, tuple_fraction) <= 0) + continue; + + best_path = path; + } + + return best_path; +} + +/* + * adjust_paths_for_srfs + * Fix up the Paths of the given upperrel to handle tSRFs properly. + * + * The executor can only handle set-returning functions that appear at the + * top level of the targetlist of a ProjectSet plan node. If we have any SRFs + * that are not at top level, we need to split up the evaluation into multiple + * plan levels in which each level satisfies this constraint. This function + * modifies each Path of an upperrel that (might) compute any SRFs in its + * output tlist to insert appropriate projection steps. + * + * The given targets and targets_contain_srfs lists are from + * split_pathtarget_at_srfs(). We assume the existing Paths emit the first + * target in targets. + */ +static void +adjust_paths_for_srfs(PlannerInfo *root, RelOptInfo *rel, + List *targets, List *targets_contain_srfs) +{ + ListCell *lc; + + Assert(list_length(targets) == list_length(targets_contain_srfs)); + Assert(!linitial_int(targets_contain_srfs)); + + /* If no SRFs appear at this plan level, nothing to do */ + if (list_length(targets) == 1) + return; + + /* + * Stack SRF-evaluation nodes atop each path for the rel. + * + * In principle we should re-run set_cheapest() here to identify the + * cheapest path, but it seems unlikely that adding the same tlist eval + * costs to all the paths would change that, so we don't bother. Instead, + * just assume that the cheapest-startup and cheapest-total paths remain + * so. (There should be no parameterized paths anymore, so we needn't + * worry about updating cheapest_parameterized_paths.) + */ + foreach(lc, rel->pathlist) + { + Path *subpath = (Path *) lfirst(lc); + Path *newpath = subpath; + ListCell *lc1, + *lc2; + + Assert(subpath->param_info == NULL); + forboth(lc1, targets, lc2, targets_contain_srfs) + { + PathTarget *thistarget = lfirst_node(PathTarget, lc1); + bool contains_srfs = (bool) lfirst_int(lc2); + + /* If this level doesn't contain SRFs, do regular projection */ + if (contains_srfs) + newpath = (Path *) create_set_projection_path(root, + rel, + newpath, + thistarget); + else + newpath = (Path *) apply_projection_to_path(root, + rel, + newpath, + thistarget); + } + lfirst(lc) = newpath; + if (subpath == rel->cheapest_startup_path) + rel->cheapest_startup_path = newpath; + if (subpath == rel->cheapest_total_path) + rel->cheapest_total_path = newpath; + } + + /* Likewise for partial paths, if any */ + foreach(lc, rel->partial_pathlist) + { + Path *subpath = (Path *) lfirst(lc); + Path *newpath = subpath; + ListCell *lc1, + *lc2; + + Assert(subpath->param_info == NULL); + forboth(lc1, targets, lc2, targets_contain_srfs) + { + PathTarget *thistarget = lfirst_node(PathTarget, lc1); + bool contains_srfs = (bool) lfirst_int(lc2); + + /* If this level doesn't contain SRFs, do regular projection */ + if (contains_srfs) + newpath = (Path *) create_set_projection_path(root, + rel, + newpath, + thistarget); + else + { + /* avoid apply_projection_to_path, in case of multiple refs */ + newpath = (Path *) create_projection_path(root, + rel, + newpath, + thistarget); + } + } + lfirst(lc) = newpath; + } +} + +/* + * expression_planner + * Perform planner's transformations on a standalone expression. + * + * Various utility commands need to evaluate expressions that are not part + * of a plannable query. They can do so using the executor's regular + * expression-execution machinery, but first the expression has to be fed + * through here to transform it from parser output to something executable. + * + * Currently, we disallow sublinks in standalone expressions, so there's no + * real "planning" involved here. (That might not always be true though.) + * What we must do is run eval_const_expressions to ensure that any function + * calls are converted to positional notation and function default arguments + * get inserted. The fact that constant subexpressions get simplified is a + * side-effect that is useful when the expression will get evaluated more than + * once. Also, we must fix operator function IDs. + * + * This does not return any information about dependencies of the expression. + * Hence callers should use the results only for the duration of the current + * query. Callers that would like to cache the results for longer should use + * expression_planner_with_deps, probably via the plancache. + * + * Note: this must not make any damaging changes to the passed-in expression + * tree. (It would actually be okay to apply fix_opfuncids to it, but since + * we first do an expression_tree_mutator-based walk, what is returned will + * be a new node tree.) The result is constructed in the current memory + * context; beware that this can leak a lot of additional stuff there, too. + */ +Expr * +expression_planner(Expr *expr) +{ + Node *result; + + /* + * Convert named-argument function calls, insert default arguments and + * simplify constant subexprs + */ + result = eval_const_expressions(NULL, (Node *) expr); + + /* Fill in opfuncid values if missing */ + fix_opfuncids(result); + + return (Expr *) result; +} + +/* + * expression_planner_with_deps + * Perform planner's transformations on a standalone expression, + * returning expression dependency information along with the result. + * + * This is identical to expression_planner() except that it also returns + * information about possible dependencies of the expression, ie identities of + * objects whose definitions affect the result. As in a PlannedStmt, these + * are expressed as a list of relation Oids and a list of PlanInvalItems. + */ +Expr * +expression_planner_with_deps(Expr *expr, + List **relationOids, + List **invalItems) +{ + Node *result; + PlannerGlobal glob; + PlannerInfo root; + + /* Make up dummy planner state so we can use setrefs machinery */ + MemSet(&glob, 0, sizeof(glob)); + glob.type = T_PlannerGlobal; + glob.relationOids = NIL; + glob.invalItems = NIL; + + MemSet(&root, 0, sizeof(root)); + root.type = T_PlannerInfo; + root.glob = &glob; + + /* + * Convert named-argument function calls, insert default arguments and + * simplify constant subexprs. Collect identities of inlined functions + * and elided domains, too. + */ + result = eval_const_expressions(&root, (Node *) expr); + + /* Fill in opfuncid values if missing */ + fix_opfuncids(result); + + /* + * Now walk the finished expression to find anything else we ought to + * record as an expression dependency. + */ + (void) extract_query_dependencies_walker(result, &root); + + *relationOids = glob.relationOids; + *invalItems = glob.invalItems; + + return (Expr *) result; +} + + +/* + * plan_cluster_use_sort + * Use the planner to decide how CLUSTER should implement sorting + * + * tableOid is the OID of a table to be clustered on its index indexOid + * (which is already known to be a btree index). Decide whether it's + * cheaper to do an indexscan or a seqscan-plus-sort to execute the CLUSTER. + * Return true to use sorting, false to use an indexscan. + * + * Note: caller had better already hold some type of lock on the table. + */ +bool +plan_cluster_use_sort(Oid tableOid, Oid indexOid) +{ + PlannerInfo *root; + Query *query; + PlannerGlobal *glob; + RangeTblEntry *rte; + RelOptInfo *rel; + IndexOptInfo *indexInfo; + QualCost indexExprCost; + Cost comparisonCost; + Path *seqScanPath; + Path seqScanAndSortPath; + IndexPath *indexScanPath; + ListCell *lc; + + /* We can short-circuit the cost comparison if indexscans are disabled */ + if (!enable_indexscan) + return true; /* use sort */ + + /* Set up mostly-dummy planner state */ + query = makeNode(Query); + query->commandType = CMD_SELECT; + + glob = makeNode(PlannerGlobal); + + root = makeNode(PlannerInfo); + root->parse = query; + root->glob = glob; + root->query_level = 1; + root->planner_cxt = CurrentMemoryContext; + root->wt_param_id = -1; + + /* Build a minimal RTE for the rel */ + rte = makeNode(RangeTblEntry); + rte->rtekind = RTE_RELATION; + rte->relid = tableOid; + rte->relkind = RELKIND_RELATION; /* Don't be too picky. */ + rte->rellockmode = AccessShareLock; + rte->lateral = false; + rte->inh = false; + rte->inFromCl = true; + query->rtable = list_make1(rte); + + /* Set up RTE/RelOptInfo arrays */ + setup_simple_rel_arrays(root); + + /* Build RelOptInfo */ + rel = build_simple_rel(root, 1, NULL); + + /* Locate IndexOptInfo for the target index */ + indexInfo = NULL; + foreach(lc, rel->indexlist) + { + indexInfo = lfirst_node(IndexOptInfo, lc); + if (indexInfo->indexoid == indexOid) + break; + } + + /* + * It's possible that get_relation_info did not generate an IndexOptInfo + * for the desired index; this could happen if it's not yet reached its + * indcheckxmin usability horizon, or if it's a system index and we're + * ignoring system indexes. In such cases we should tell CLUSTER to not + * trust the index contents but use seqscan-and-sort. + */ + if (lc == NULL) /* not in the list? */ + return true; /* use sort */ + + /* + * Rather than doing all the pushups that would be needed to use + * set_baserel_size_estimates, just do a quick hack for rows and width. + */ + rel->rows = rel->tuples; + rel->reltarget->width = get_relation_data_width(tableOid, NULL); + + root->total_table_pages = rel->pages; + + /* + * Determine eval cost of the index expressions, if any. We need to + * charge twice that amount for each tuple comparison that happens during + * the sort, since tuplesort.c will have to re-evaluate the index + * expressions each time. (XXX that's pretty inefficient...) + */ + cost_qual_eval(&indexExprCost, indexInfo->indexprs, root); + comparisonCost = 2.0 * (indexExprCost.startup + indexExprCost.per_tuple); + + /* Estimate the cost of seq scan + sort */ + seqScanPath = create_seqscan_path(root, rel, NULL, 0); + cost_sort(&seqScanAndSortPath, root, NIL, + seqScanPath->total_cost, rel->tuples, rel->reltarget->width, + comparisonCost, maintenance_work_mem, -1.0); + + /* Estimate the cost of index scan */ + indexScanPath = create_index_path(root, indexInfo, + NIL, NIL, NIL, NIL, + ForwardScanDirection, false, + NULL, 1.0, false); + + return (seqScanAndSortPath.total_cost < indexScanPath->path.total_cost); +} + +/* + * plan_create_index_workers + * Use the planner to decide how many parallel worker processes + * CREATE INDEX should request for use + * + * tableOid is the table on which the index is to be built. indexOid is the + * OID of an index to be created or reindexed (which must be a btree index). + * + * Return value is the number of parallel worker processes to request. It + * may be unsafe to proceed if this is 0. Note that this does not include the + * leader participating as a worker (value is always a number of parallel + * worker processes). + * + * Note: caller had better already hold some type of lock on the table and + * index. + */ +int +plan_create_index_workers(Oid tableOid, Oid indexOid) +{ + PlannerInfo *root; + Query *query; + PlannerGlobal *glob; + RangeTblEntry *rte; + Relation heap; + Relation index; + RelOptInfo *rel; + int parallel_workers; + BlockNumber heap_blocks; + double reltuples; + double allvisfrac; + + /* + * We don't allow performing parallel operation in standalone backend or + * when parallelism is disabled. + */ + if (!IsUnderPostmaster || max_parallel_maintenance_workers == 0) + return 0; + + /* Set up largely-dummy planner state */ + query = makeNode(Query); + query->commandType = CMD_SELECT; + + glob = makeNode(PlannerGlobal); + + root = makeNode(PlannerInfo); + root->parse = query; + root->glob = glob; + root->query_level = 1; + root->planner_cxt = CurrentMemoryContext; + root->wt_param_id = -1; + + /* + * Build a minimal RTE. + * + * Mark the RTE with inh = true. This is a kludge to prevent + * get_relation_info() from fetching index info, which is necessary + * because it does not expect that any IndexOptInfo is currently + * undergoing REINDEX. + */ + rte = makeNode(RangeTblEntry); + rte->rtekind = RTE_RELATION; + rte->relid = tableOid; + rte->relkind = RELKIND_RELATION; /* Don't be too picky. */ + rte->rellockmode = AccessShareLock; + rte->lateral = false; + rte->inh = true; + rte->inFromCl = true; + query->rtable = list_make1(rte); + + /* Set up RTE/RelOptInfo arrays */ + setup_simple_rel_arrays(root); + + /* Build RelOptInfo */ + rel = build_simple_rel(root, 1, NULL); + + /* Rels are assumed already locked by the caller */ + heap = table_open(tableOid, NoLock); + index = index_open(indexOid, NoLock); + + /* + * Determine if it's safe to proceed. + * + * Currently, parallel workers can't access the leader's temporary tables. + * Furthermore, any index predicate or index expressions must be parallel + * safe. + */ + if (heap->rd_rel->relpersistence == RELPERSISTENCE_TEMP || + !is_parallel_safe(root, (Node *) RelationGetIndexExpressions(index)) || + !is_parallel_safe(root, (Node *) RelationGetIndexPredicate(index))) + { + parallel_workers = 0; + goto done; + } + + /* + * If parallel_workers storage parameter is set for the table, accept that + * as the number of parallel worker processes to launch (though still cap + * at max_parallel_maintenance_workers). Note that we deliberately do not + * consider any other factor when parallel_workers is set. (e.g., memory + * use by workers.) + */ + if (rel->rel_parallel_workers != -1) + { + parallel_workers = Min(rel->rel_parallel_workers, + max_parallel_maintenance_workers); + goto done; + } + + /* + * Estimate heap relation size ourselves, since rel->pages cannot be + * trusted (heap RTE was marked as inheritance parent) + */ + estimate_rel_size(heap, NULL, &heap_blocks, &reltuples, &allvisfrac); + + /* + * Determine number of workers to scan the heap relation using generic + * model + */ + parallel_workers = compute_parallel_worker(rel, heap_blocks, -1, + max_parallel_maintenance_workers); + + /* + * Cap workers based on available maintenance_work_mem as needed. + * + * Note that each tuplesort participant receives an even share of the + * total maintenance_work_mem budget. Aim to leave participants + * (including the leader as a participant) with no less than 32MB of + * memory. This leaves cases where maintenance_work_mem is set to 64MB + * immediately past the threshold of being capable of launching a single + * parallel worker to sort. + */ + while (parallel_workers > 0 && + maintenance_work_mem / (parallel_workers + 1) < 32768L) + parallel_workers--; + +done: + index_close(index, NoLock); + table_close(heap, NoLock); + + return parallel_workers; +} + +/* + * add_paths_to_grouping_rel + * + * Add non-partial paths to grouping relation. + */ +static void +add_paths_to_grouping_rel(PlannerInfo *root, RelOptInfo *input_rel, + RelOptInfo *grouped_rel, + RelOptInfo *partially_grouped_rel, + const AggClauseCosts *agg_costs, + grouping_sets_data *gd, double dNumGroups, + GroupPathExtraData *extra) +{ + Query *parse = root->parse; + Path *cheapest_path = input_rel->cheapest_total_path; + ListCell *lc; + bool can_hash = (extra->flags & GROUPING_CAN_USE_HASH) != 0; + bool can_sort = (extra->flags & GROUPING_CAN_USE_SORT) != 0; + List *havingQual = (List *) extra->havingQual; + AggClauseCosts *agg_final_costs = &extra->agg_final_costs; + + if (can_sort) + { + /* + * Use any available suitably-sorted path as input, and also consider + * sorting the cheapest-total path. + */ + foreach(lc, input_rel->pathlist) + { + Path *path = (Path *) lfirst(lc); + Path *path_original = path; + bool is_sorted; + int presorted_keys; + + is_sorted = pathkeys_count_contained_in(root->group_pathkeys, + path->pathkeys, + &presorted_keys); + + if (path == cheapest_path || is_sorted) + { + /* Sort the cheapest-total path if it isn't already sorted */ + if (!is_sorted) + path = (Path *) create_sort_path(root, + grouped_rel, + path, + root->group_pathkeys, + -1.0); + + /* Now decide what to stick atop it */ + if (parse->groupingSets) + { + consider_groupingsets_paths(root, grouped_rel, + path, true, can_hash, + gd, agg_costs, dNumGroups); + } + else if (parse->hasAggs) + { + /* + * We have aggregation, possibly with plain GROUP BY. Make + * an AggPath. + */ + add_path(grouped_rel, (Path *) + create_agg_path(root, + grouped_rel, + path, + grouped_rel->reltarget, + parse->groupClause ? AGG_SORTED : AGG_PLAIN, + AGGSPLIT_SIMPLE, + parse->groupClause, + havingQual, + agg_costs, + dNumGroups)); + } + else if (parse->groupClause) + { + /* + * We have GROUP BY without aggregation or grouping sets. + * Make a GroupPath. + */ + add_path(grouped_rel, (Path *) + create_group_path(root, + grouped_rel, + path, + parse->groupClause, + havingQual, + dNumGroups)); + } + else + { + /* Other cases should have been handled above */ + Assert(false); + } + } + + /* + * Now we may consider incremental sort on this path, but only + * when the path is not already sorted and when incremental sort + * is enabled. + */ + if (is_sorted || !enable_incremental_sort) + continue; + + /* Restore the input path (we might have added Sort on top). */ + path = path_original; + + /* no shared prefix, no point in building incremental sort */ + if (presorted_keys == 0) + continue; + + /* + * We should have already excluded pathkeys of length 1 because + * then presorted_keys > 0 would imply is_sorted was true. + */ + Assert(list_length(root->group_pathkeys) != 1); + + path = (Path *) create_incremental_sort_path(root, + grouped_rel, + path, + root->group_pathkeys, + presorted_keys, + -1.0); + + /* Now decide what to stick atop it */ + if (parse->groupingSets) + { + consider_groupingsets_paths(root, grouped_rel, + path, true, can_hash, + gd, agg_costs, dNumGroups); + } + else if (parse->hasAggs) + { + /* + * We have aggregation, possibly with plain GROUP BY. Make an + * AggPath. + */ + add_path(grouped_rel, (Path *) + create_agg_path(root, + grouped_rel, + path, + grouped_rel->reltarget, + parse->groupClause ? AGG_SORTED : AGG_PLAIN, + AGGSPLIT_SIMPLE, + parse->groupClause, + havingQual, + agg_costs, + dNumGroups)); + } + else if (parse->groupClause) + { + /* + * We have GROUP BY without aggregation or grouping sets. Make + * a GroupPath. + */ + add_path(grouped_rel, (Path *) + create_group_path(root, + grouped_rel, + path, + parse->groupClause, + havingQual, + dNumGroups)); + } + else + { + /* Other cases should have been handled above */ + Assert(false); + } + } + + /* + * Instead of operating directly on the input relation, we can + * consider finalizing a partially aggregated path. + */ + if (partially_grouped_rel != NULL) + { + foreach(lc, partially_grouped_rel->pathlist) + { + Path *path = (Path *) lfirst(lc); + Path *path_original = path; + bool is_sorted; + int presorted_keys; + + is_sorted = pathkeys_count_contained_in(root->group_pathkeys, + path->pathkeys, + &presorted_keys); + + /* + * Insert a Sort node, if required. But there's no point in + * sorting anything but the cheapest path. + */ + if (!is_sorted) + { + if (path != partially_grouped_rel->cheapest_total_path) + continue; + path = (Path *) create_sort_path(root, + grouped_rel, + path, + root->group_pathkeys, + -1.0); + } + + if (parse->hasAggs) + add_path(grouped_rel, (Path *) + create_agg_path(root, + grouped_rel, + path, + grouped_rel->reltarget, + parse->groupClause ? AGG_SORTED : AGG_PLAIN, + AGGSPLIT_FINAL_DESERIAL, + parse->groupClause, + havingQual, + agg_final_costs, + dNumGroups)); + else + add_path(grouped_rel, (Path *) + create_group_path(root, + grouped_rel, + path, + parse->groupClause, + havingQual, + dNumGroups)); + + /* + * Now we may consider incremental sort on this path, but only + * when the path is not already sorted and when incremental + * sort is enabled. + */ + if (is_sorted || !enable_incremental_sort) + continue; + + /* Restore the input path (we might have added Sort on top). */ + path = path_original; + + /* no shared prefix, not point in building incremental sort */ + if (presorted_keys == 0) + continue; + + /* + * We should have already excluded pathkeys of length 1 + * because then presorted_keys > 0 would imply is_sorted was + * true. + */ + Assert(list_length(root->group_pathkeys) != 1); + + path = (Path *) create_incremental_sort_path(root, + grouped_rel, + path, + root->group_pathkeys, + presorted_keys, + -1.0); + + if (parse->hasAggs) + add_path(grouped_rel, (Path *) + create_agg_path(root, + grouped_rel, + path, + grouped_rel->reltarget, + parse->groupClause ? AGG_SORTED : AGG_PLAIN, + AGGSPLIT_FINAL_DESERIAL, + parse->groupClause, + havingQual, + agg_final_costs, + dNumGroups)); + else + add_path(grouped_rel, (Path *) + create_group_path(root, + grouped_rel, + path, + parse->groupClause, + havingQual, + dNumGroups)); + } + } + } + + if (can_hash) + { + if (parse->groupingSets) + { + /* + * Try for a hash-only groupingsets path over unsorted input. + */ + consider_groupingsets_paths(root, grouped_rel, + cheapest_path, false, true, + gd, agg_costs, dNumGroups); + } + else + { + /* + * Generate a HashAgg Path. We just need an Agg over the + * cheapest-total input path, since input order won't matter. + */ + add_path(grouped_rel, (Path *) + create_agg_path(root, grouped_rel, + cheapest_path, + grouped_rel->reltarget, + AGG_HASHED, + AGGSPLIT_SIMPLE, + parse->groupClause, + havingQual, + agg_costs, + dNumGroups)); + } + + /* + * Generate a Finalize HashAgg Path atop of the cheapest partially + * grouped path, assuming there is one + */ + if (partially_grouped_rel && partially_grouped_rel->pathlist) + { + Path *path = partially_grouped_rel->cheapest_total_path; + + add_path(grouped_rel, (Path *) + create_agg_path(root, + grouped_rel, + path, + grouped_rel->reltarget, + AGG_HASHED, + AGGSPLIT_FINAL_DESERIAL, + parse->groupClause, + havingQual, + agg_final_costs, + dNumGroups)); + } + } + + /* + * When partitionwise aggregate is used, we might have fully aggregated + * paths in the partial pathlist, because add_paths_to_append_rel() will + * consider a path for grouped_rel consisting of a Parallel Append of + * non-partial paths from each child. + */ + if (grouped_rel->partial_pathlist != NIL) + gather_grouping_paths(root, grouped_rel); +} + +/* + * create_partial_grouping_paths + * + * Create a new upper relation representing the result of partial aggregation + * and populate it with appropriate paths. Note that we don't finalize the + * lists of paths here, so the caller can add additional partial or non-partial + * paths and must afterward call gather_grouping_paths and set_cheapest on + * the returned upper relation. + * + * All paths for this new upper relation -- both partial and non-partial -- + * have been partially aggregated but require a subsequent FinalizeAggregate + * step. + * + * NB: This function is allowed to return NULL if it determines that there is + * no real need to create a new RelOptInfo. + */ +static RelOptInfo * +create_partial_grouping_paths(PlannerInfo *root, + RelOptInfo *grouped_rel, + RelOptInfo *input_rel, + grouping_sets_data *gd, + GroupPathExtraData *extra, + bool force_rel_creation) +{ + Query *parse = root->parse; + RelOptInfo *partially_grouped_rel; + AggClauseCosts *agg_partial_costs = &extra->agg_partial_costs; + AggClauseCosts *agg_final_costs = &extra->agg_final_costs; + Path *cheapest_partial_path = NULL; + Path *cheapest_total_path = NULL; + double dNumPartialGroups = 0; + double dNumPartialPartialGroups = 0; + ListCell *lc; + bool can_hash = (extra->flags & GROUPING_CAN_USE_HASH) != 0; + bool can_sort = (extra->flags & GROUPING_CAN_USE_SORT) != 0; + + /* + * Consider whether we should generate partially aggregated non-partial + * paths. We can only do this if we have a non-partial path, and only if + * the parent of the input rel is performing partial partitionwise + * aggregation. (Note that extra->patype is the type of partitionwise + * aggregation being used at the parent level, not this level.) + */ + if (input_rel->pathlist != NIL && + extra->patype == PARTITIONWISE_AGGREGATE_PARTIAL) + cheapest_total_path = input_rel->cheapest_total_path; + + /* + * If parallelism is possible for grouped_rel, then we should consider + * generating partially-grouped partial paths. However, if the input rel + * has no partial paths, then we can't. + */ + if (grouped_rel->consider_parallel && input_rel->partial_pathlist != NIL) + cheapest_partial_path = linitial(input_rel->partial_pathlist); + + /* + * If we can't partially aggregate partial paths, and we can't partially + * aggregate non-partial paths, then don't bother creating the new + * RelOptInfo at all, unless the caller specified force_rel_creation. + */ + if (cheapest_total_path == NULL && + cheapest_partial_path == NULL && + !force_rel_creation) + return NULL; + + /* + * Build a new upper relation to represent the result of partially + * aggregating the rows from the input relation. + */ + partially_grouped_rel = fetch_upper_rel(root, + UPPERREL_PARTIAL_GROUP_AGG, + grouped_rel->relids); + partially_grouped_rel->consider_parallel = + grouped_rel->consider_parallel; + partially_grouped_rel->reloptkind = grouped_rel->reloptkind; + partially_grouped_rel->serverid = grouped_rel->serverid; + partially_grouped_rel->userid = grouped_rel->userid; + partially_grouped_rel->useridiscurrent = grouped_rel->useridiscurrent; + partially_grouped_rel->fdwroutine = grouped_rel->fdwroutine; + + /* + * Build target list for partial aggregate paths. These paths cannot just + * emit the same tlist as regular aggregate paths, because (1) we must + * include Vars and Aggrefs needed in HAVING, which might not appear in + * the result tlist, and (2) the Aggrefs must be set in partial mode. + */ + partially_grouped_rel->reltarget = + make_partial_grouping_target(root, grouped_rel->reltarget, + extra->havingQual); + + if (!extra->partial_costs_set) + { + /* + * Collect statistics about aggregates for estimating costs of + * performing aggregation in parallel. + */ + MemSet(agg_partial_costs, 0, sizeof(AggClauseCosts)); + MemSet(agg_final_costs, 0, sizeof(AggClauseCosts)); + if (parse->hasAggs) + { + /* partial phase */ + get_agg_clause_costs(root, AGGSPLIT_INITIAL_SERIAL, + agg_partial_costs); + + /* final phase */ + get_agg_clause_costs(root, AGGSPLIT_FINAL_DESERIAL, + agg_final_costs); + } + + extra->partial_costs_set = true; + } + + /* Estimate number of partial groups. */ + if (cheapest_total_path != NULL) + dNumPartialGroups = + get_number_of_groups(root, + cheapest_total_path->rows, + gd, + extra->targetList); + if (cheapest_partial_path != NULL) + dNumPartialPartialGroups = + get_number_of_groups(root, + cheapest_partial_path->rows, + gd, + extra->targetList); + + if (can_sort && cheapest_total_path != NULL) + { + /* This should have been checked previously */ + Assert(parse->hasAggs || parse->groupClause); + + /* + * Use any available suitably-sorted path as input, and also consider + * sorting the cheapest partial path. + */ + foreach(lc, input_rel->pathlist) + { + Path *path = (Path *) lfirst(lc); + bool is_sorted; + + is_sorted = pathkeys_contained_in(root->group_pathkeys, + path->pathkeys); + if (path == cheapest_total_path || is_sorted) + { + /* Sort the cheapest partial path, if it isn't already */ + if (!is_sorted) + path = (Path *) create_sort_path(root, + partially_grouped_rel, + path, + root->group_pathkeys, + -1.0); + + if (parse->hasAggs) + add_path(partially_grouped_rel, (Path *) + create_agg_path(root, + partially_grouped_rel, + path, + partially_grouped_rel->reltarget, + parse->groupClause ? AGG_SORTED : AGG_PLAIN, + AGGSPLIT_INITIAL_SERIAL, + parse->groupClause, + NIL, + agg_partial_costs, + dNumPartialGroups)); + else + add_path(partially_grouped_rel, (Path *) + create_group_path(root, + partially_grouped_rel, + path, + parse->groupClause, + NIL, + dNumPartialGroups)); + } + } + + /* + * Consider incremental sort on all partial paths, if enabled. + * + * We can also skip the entire loop when we only have a single-item + * group_pathkeys because then we can't possibly have a presorted + * prefix of the list without having the list be fully sorted. + */ + if (enable_incremental_sort && list_length(root->group_pathkeys) > 1) + { + foreach(lc, input_rel->pathlist) + { + Path *path = (Path *) lfirst(lc); + bool is_sorted; + int presorted_keys; + + is_sorted = pathkeys_count_contained_in(root->group_pathkeys, + path->pathkeys, + &presorted_keys); + + /* Ignore already sorted paths */ + if (is_sorted) + continue; + + if (presorted_keys == 0) + continue; + + /* Since we have presorted keys, consider incremental sort. */ + path = (Path *) create_incremental_sort_path(root, + partially_grouped_rel, + path, + root->group_pathkeys, + presorted_keys, + -1.0); + + if (parse->hasAggs) + add_path(partially_grouped_rel, (Path *) + create_agg_path(root, + partially_grouped_rel, + path, + partially_grouped_rel->reltarget, + parse->groupClause ? AGG_SORTED : AGG_PLAIN, + AGGSPLIT_INITIAL_SERIAL, + parse->groupClause, + NIL, + agg_partial_costs, + dNumPartialGroups)); + else + add_path(partially_grouped_rel, (Path *) + create_group_path(root, + partially_grouped_rel, + path, + parse->groupClause, + NIL, + dNumPartialGroups)); + } + } + } + + if (can_sort && cheapest_partial_path != NULL) + { + /* Similar to above logic, but for partial paths. */ + foreach(lc, input_rel->partial_pathlist) + { + Path *path = (Path *) lfirst(lc); + Path *path_original = path; + bool is_sorted; + int presorted_keys; + + is_sorted = pathkeys_count_contained_in(root->group_pathkeys, + path->pathkeys, + &presorted_keys); + + if (path == cheapest_partial_path || is_sorted) + { + /* Sort the cheapest partial path, if it isn't already */ + if (!is_sorted) + path = (Path *) create_sort_path(root, + partially_grouped_rel, + path, + root->group_pathkeys, + -1.0); + + if (parse->hasAggs) + add_partial_path(partially_grouped_rel, (Path *) + create_agg_path(root, + partially_grouped_rel, + path, + partially_grouped_rel->reltarget, + parse->groupClause ? AGG_SORTED : AGG_PLAIN, + AGGSPLIT_INITIAL_SERIAL, + parse->groupClause, + NIL, + agg_partial_costs, + dNumPartialPartialGroups)); + else + add_partial_path(partially_grouped_rel, (Path *) + create_group_path(root, + partially_grouped_rel, + path, + parse->groupClause, + NIL, + dNumPartialPartialGroups)); + } + + /* + * Now we may consider incremental sort on this path, but only + * when the path is not already sorted and when incremental sort + * is enabled. + */ + if (is_sorted || !enable_incremental_sort) + continue; + + /* Restore the input path (we might have added Sort on top). */ + path = path_original; + + /* no shared prefix, not point in building incremental sort */ + if (presorted_keys == 0) + continue; + + /* + * We should have already excluded pathkeys of length 1 because + * then presorted_keys > 0 would imply is_sorted was true. + */ + Assert(list_length(root->group_pathkeys) != 1); + + path = (Path *) create_incremental_sort_path(root, + partially_grouped_rel, + path, + root->group_pathkeys, + presorted_keys, + -1.0); + + if (parse->hasAggs) + add_partial_path(partially_grouped_rel, (Path *) + create_agg_path(root, + partially_grouped_rel, + path, + partially_grouped_rel->reltarget, + parse->groupClause ? AGG_SORTED : AGG_PLAIN, + AGGSPLIT_INITIAL_SERIAL, + parse->groupClause, + NIL, + agg_partial_costs, + dNumPartialPartialGroups)); + else + add_partial_path(partially_grouped_rel, (Path *) + create_group_path(root, + partially_grouped_rel, + path, + parse->groupClause, + NIL, + dNumPartialPartialGroups)); + } + } + + /* + * Add a partially-grouped HashAgg Path where possible + */ + if (can_hash && cheapest_total_path != NULL) + { + /* Checked above */ + Assert(parse->hasAggs || parse->groupClause); + + add_path(partially_grouped_rel, (Path *) + create_agg_path(root, + partially_grouped_rel, + cheapest_total_path, + partially_grouped_rel->reltarget, + AGG_HASHED, + AGGSPLIT_INITIAL_SERIAL, + parse->groupClause, + NIL, + agg_partial_costs, + dNumPartialGroups)); + } + + /* + * Now add a partially-grouped HashAgg partial Path where possible + */ + if (can_hash && cheapest_partial_path != NULL) + { + add_partial_path(partially_grouped_rel, (Path *) + create_agg_path(root, + partially_grouped_rel, + cheapest_partial_path, + partially_grouped_rel->reltarget, + AGG_HASHED, + AGGSPLIT_INITIAL_SERIAL, + parse->groupClause, + NIL, + agg_partial_costs, + dNumPartialPartialGroups)); + } + + /* + * If there is an FDW that's responsible for all baserels of the query, + * let it consider adding partially grouped ForeignPaths. + */ + if (partially_grouped_rel->fdwroutine && + partially_grouped_rel->fdwroutine->GetForeignUpperPaths) + { + FdwRoutine *fdwroutine = partially_grouped_rel->fdwroutine; + + fdwroutine->GetForeignUpperPaths(root, + UPPERREL_PARTIAL_GROUP_AGG, + input_rel, partially_grouped_rel, + extra); + } + + return partially_grouped_rel; +} + +/* + * Generate Gather and Gather Merge paths for a grouping relation or partial + * grouping relation. + * + * generate_useful_gather_paths does most of the work, but we also consider a + * special case: we could try sorting the data by the group_pathkeys and then + * applying Gather Merge. + * + * NB: This function shouldn't be used for anything other than a grouped or + * partially grouped relation not only because of the fact that it explicitly + * references group_pathkeys but we pass "true" as the third argument to + * generate_useful_gather_paths(). + */ +static void +gather_grouping_paths(PlannerInfo *root, RelOptInfo *rel) +{ + ListCell *lc; + Path *cheapest_partial_path; + + /* Try Gather for unordered paths and Gather Merge for ordered ones. */ + generate_useful_gather_paths(root, rel, true); + + /* Try cheapest partial path + explicit Sort + Gather Merge. */ + cheapest_partial_path = linitial(rel->partial_pathlist); + if (!pathkeys_contained_in(root->group_pathkeys, + cheapest_partial_path->pathkeys)) + { + Path *path; + double total_groups; + + total_groups = + cheapest_partial_path->rows * cheapest_partial_path->parallel_workers; + path = (Path *) create_sort_path(root, rel, cheapest_partial_path, + root->group_pathkeys, + -1.0); + path = (Path *) + create_gather_merge_path(root, + rel, + path, + rel->reltarget, + root->group_pathkeys, + NULL, + &total_groups); + + add_path(rel, path); + } + + /* + * Consider incremental sort on all partial paths, if enabled. + * + * We can also skip the entire loop when we only have a single-item + * group_pathkeys because then we can't possibly have a presorted prefix + * of the list without having the list be fully sorted. + */ + if (!enable_incremental_sort || list_length(root->group_pathkeys) == 1) + return; + + /* also consider incremental sort on partial paths, if enabled */ + foreach(lc, rel->partial_pathlist) + { + Path *path = (Path *) lfirst(lc); + bool is_sorted; + int presorted_keys; + double total_groups; + + is_sorted = pathkeys_count_contained_in(root->group_pathkeys, + path->pathkeys, + &presorted_keys); + + if (is_sorted) + continue; + + if (presorted_keys == 0) + continue; + + path = (Path *) create_incremental_sort_path(root, + rel, + path, + root->group_pathkeys, + presorted_keys, + -1.0); + + path = (Path *) + create_gather_merge_path(root, + rel, + path, + rel->reltarget, + root->group_pathkeys, + NULL, + &total_groups); + + add_path(rel, path); + } +} + +/* + * can_partial_agg + * + * Determines whether or not partial grouping and/or aggregation is possible. + * Returns true when possible, false otherwise. + */ +static bool +can_partial_agg(PlannerInfo *root) +{ + Query *parse = root->parse; + + if (!parse->hasAggs && parse->groupClause == NIL) + { + /* + * We don't know how to do parallel aggregation unless we have either + * some aggregates or a grouping clause. + */ + return false; + } + else if (parse->groupingSets) + { + /* We don't know how to do grouping sets in parallel. */ + return false; + } + else if (root->hasNonPartialAggs || root->hasNonSerialAggs) + { + /* Insufficient support for partial mode. */ + return false; + } + + /* Everything looks good. */ + return true; +} + +/* + * apply_scanjoin_target_to_paths + * + * Adjust the final scan/join relation, and recursively all of its children, + * to generate the final scan/join target. It would be more correct to model + * this as a separate planning step with a new RelOptInfo at the toplevel and + * for each child relation, but doing it this way is noticeably cheaper. + * Maybe that problem can be solved at some point, but for now we do this. + * + * If tlist_same_exprs is true, then the scan/join target to be applied has + * the same expressions as the existing reltarget, so we need only insert the + * appropriate sortgroupref information. By avoiding the creation of + * projection paths we save effort both immediately and at plan creation time. + */ +static void +apply_scanjoin_target_to_paths(PlannerInfo *root, + RelOptInfo *rel, + List *scanjoin_targets, + List *scanjoin_targets_contain_srfs, + bool scanjoin_target_parallel_safe, + bool tlist_same_exprs) +{ + bool rel_is_partitioned = IS_PARTITIONED_REL(rel); + PathTarget *scanjoin_target; + ListCell *lc; + + /* This recurses, so be paranoid. */ + check_stack_depth(); + + /* + * If the rel is partitioned, we want to drop its existing paths and + * generate new ones. This function would still be correct if we kept the + * existing paths: we'd modify them to generate the correct target above + * the partitioning Append, and then they'd compete on cost with paths + * generating the target below the Append. However, in our current cost + * model the latter way is always the same or cheaper cost, so modifying + * the existing paths would just be useless work. Moreover, when the cost + * is the same, varying roundoff errors might sometimes allow an existing + * path to be picked, resulting in undesirable cross-platform plan + * variations. So we drop old paths and thereby force the work to be done + * below the Append, except in the case of a non-parallel-safe target. + * + * Some care is needed, because we have to allow + * generate_useful_gather_paths to see the old partial paths in the next + * stanza. Hence, zap the main pathlist here, then allow + * generate_useful_gather_paths to add path(s) to the main list, and + * finally zap the partial pathlist. + */ + if (rel_is_partitioned) + rel->pathlist = NIL; + + /* + * If the scan/join target is not parallel-safe, partial paths cannot + * generate it. + */ + if (!scanjoin_target_parallel_safe) + { + /* + * Since we can't generate the final scan/join target in parallel + * workers, this is our last opportunity to use any partial paths that + * exist; so build Gather path(s) that use them and emit whatever the + * current reltarget is. We don't do this in the case where the + * target is parallel-safe, since we will be able to generate superior + * paths by doing it after the final scan/join target has been + * applied. + */ + generate_useful_gather_paths(root, rel, false); + + /* Can't use parallel query above this level. */ + rel->partial_pathlist = NIL; + rel->consider_parallel = false; + } + + /* Finish dropping old paths for a partitioned rel, per comment above */ + if (rel_is_partitioned) + rel->partial_pathlist = NIL; + + /* Extract SRF-free scan/join target. */ + scanjoin_target = linitial_node(PathTarget, scanjoin_targets); + + /* + * Apply the SRF-free scan/join target to each existing path. + * + * If the tlist exprs are the same, we can just inject the sortgroupref + * information into the existing pathtargets. Otherwise, replace each + * path with a projection path that generates the SRF-free scan/join + * target. This can't change the ordering of paths within rel->pathlist, + * so we just modify the list in place. + */ + foreach(lc, rel->pathlist) + { + Path *subpath = (Path *) lfirst(lc); + + /* Shouldn't have any parameterized paths anymore */ + Assert(subpath->param_info == NULL); + + if (tlist_same_exprs) + subpath->pathtarget->sortgrouprefs = + scanjoin_target->sortgrouprefs; + else + { + Path *newpath; + + newpath = (Path *) create_projection_path(root, rel, subpath, + scanjoin_target); + lfirst(lc) = newpath; + } + } + + /* Likewise adjust the targets for any partial paths. */ + foreach(lc, rel->partial_pathlist) + { + Path *subpath = (Path *) lfirst(lc); + + /* Shouldn't have any parameterized paths anymore */ + Assert(subpath->param_info == NULL); + + if (tlist_same_exprs) + subpath->pathtarget->sortgrouprefs = + scanjoin_target->sortgrouprefs; + else + { + Path *newpath; + + newpath = (Path *) create_projection_path(root, rel, subpath, + scanjoin_target); + lfirst(lc) = newpath; + } + } + + /* + * Now, if final scan/join target contains SRFs, insert ProjectSetPath(s) + * atop each existing path. (Note that this function doesn't look at the + * cheapest-path fields, which is a good thing because they're bogus right + * now.) + */ + if (root->parse->hasTargetSRFs) + adjust_paths_for_srfs(root, rel, + scanjoin_targets, + scanjoin_targets_contain_srfs); + + /* + * Update the rel's target to be the final (with SRFs) scan/join target. + * This now matches the actual output of all the paths, and we might get + * confused in createplan.c if they don't agree. We must do this now so + * that any append paths made in the next part will use the correct + * pathtarget (cf. create_append_path). + * + * Note that this is also necessary if GetForeignUpperPaths() gets called + * on the final scan/join relation or on any of its children, since the + * FDW might look at the rel's target to create ForeignPaths. + */ + rel->reltarget = llast_node(PathTarget, scanjoin_targets); + + /* + * If the relation is partitioned, recursively apply the scan/join target + * to all partitions, and generate brand-new Append paths in which the + * scan/join target is computed below the Append rather than above it. + * Since Append is not projection-capable, that might save a separate + * Result node, and it also is important for partitionwise aggregate. + */ + if (rel_is_partitioned) + { + List *live_children = NIL; + int i; + + /* Adjust each partition. */ + i = -1; + while ((i = bms_next_member(rel->live_parts, i)) >= 0) + { + RelOptInfo *child_rel = rel->part_rels[i]; + AppendRelInfo **appinfos; + int nappinfos; + List *child_scanjoin_targets = NIL; + ListCell *lc; + + Assert(child_rel != NULL); + + /* Dummy children can be ignored. */ + if (IS_DUMMY_REL(child_rel)) + continue; + + /* Translate scan/join targets for this child. */ + appinfos = find_appinfos_by_relids(root, child_rel->relids, + &nappinfos); + foreach(lc, scanjoin_targets) + { + PathTarget *target = lfirst_node(PathTarget, lc); + + target = copy_pathtarget(target); + target->exprs = (List *) + adjust_appendrel_attrs(root, + (Node *) target->exprs, + nappinfos, appinfos); + child_scanjoin_targets = lappend(child_scanjoin_targets, + target); + } + pfree(appinfos); + + /* Recursion does the real work. */ + apply_scanjoin_target_to_paths(root, child_rel, + child_scanjoin_targets, + scanjoin_targets_contain_srfs, + scanjoin_target_parallel_safe, + tlist_same_exprs); + + /* Save non-dummy children for Append paths. */ + if (!IS_DUMMY_REL(child_rel)) + live_children = lappend(live_children, child_rel); + } + + /* Build new paths for this relation by appending child paths. */ + add_paths_to_append_rel(root, rel, live_children); + } + + /* + * Consider generating Gather or Gather Merge paths. We must only do this + * if the relation is parallel safe, and we don't do it for child rels to + * avoid creating multiple Gather nodes within the same plan. We must do + * this after all paths have been generated and before set_cheapest, since + * one of the generated paths may turn out to be the cheapest one. + */ + if (rel->consider_parallel && !IS_OTHER_REL(rel)) + generate_useful_gather_paths(root, rel, false); + + /* + * Reassess which paths are the cheapest, now that we've potentially added + * new Gather (or Gather Merge) and/or Append (or MergeAppend) paths to + * this relation. + */ + set_cheapest(rel); +} + +/* + * create_partitionwise_grouping_paths + * + * If the partition keys of input relation are part of the GROUP BY clause, all + * the rows belonging to a given group come from a single partition. This + * allows aggregation/grouping over a partitioned relation to be broken down + * into aggregation/grouping on each partition. This should be no worse, and + * often better, than the normal approach. + * + * However, if the GROUP BY clause does not contain all the partition keys, + * rows from a given group may be spread across multiple partitions. In that + * case, we perform partial aggregation for each group, append the results, + * and then finalize aggregation. This is less certain to win than the + * previous case. It may win if the PartialAggregate stage greatly reduces + * the number of groups, because fewer rows will pass through the Append node. + * It may lose if we have lots of small groups. + */ +static void +create_partitionwise_grouping_paths(PlannerInfo *root, + RelOptInfo *input_rel, + RelOptInfo *grouped_rel, + RelOptInfo *partially_grouped_rel, + const AggClauseCosts *agg_costs, + grouping_sets_data *gd, + PartitionwiseAggregateType patype, + GroupPathExtraData *extra) +{ + List *grouped_live_children = NIL; + List *partially_grouped_live_children = NIL; + PathTarget *target = grouped_rel->reltarget; + bool partial_grouping_valid = true; + int i; + + Assert(patype != PARTITIONWISE_AGGREGATE_NONE); + Assert(patype != PARTITIONWISE_AGGREGATE_PARTIAL || + partially_grouped_rel != NULL); + + /* Add paths for partitionwise aggregation/grouping. */ + i = -1; + while ((i = bms_next_member(input_rel->live_parts, i)) >= 0) + { + RelOptInfo *child_input_rel = input_rel->part_rels[i]; + PathTarget *child_target; + AppendRelInfo **appinfos; + int nappinfos; + GroupPathExtraData child_extra; + RelOptInfo *child_grouped_rel; + RelOptInfo *child_partially_grouped_rel; + + Assert(child_input_rel != NULL); + + /* Dummy children can be ignored. */ + if (IS_DUMMY_REL(child_input_rel)) + continue; + + child_target = copy_pathtarget(target); + + /* + * Copy the given "extra" structure as is and then override the + * members specific to this child. + */ + memcpy(&child_extra, extra, sizeof(child_extra)); + + appinfos = find_appinfos_by_relids(root, child_input_rel->relids, + &nappinfos); + + child_target->exprs = (List *) + adjust_appendrel_attrs(root, + (Node *) target->exprs, + nappinfos, appinfos); + + /* Translate havingQual and targetList. */ + child_extra.havingQual = (Node *) + adjust_appendrel_attrs(root, + extra->havingQual, + nappinfos, appinfos); + child_extra.targetList = (List *) + adjust_appendrel_attrs(root, + (Node *) extra->targetList, + nappinfos, appinfos); + + /* + * extra->patype was the value computed for our parent rel; patype is + * the value for this relation. For the child, our value is its + * parent rel's value. + */ + child_extra.patype = patype; + + /* + * Create grouping relation to hold fully aggregated grouping and/or + * aggregation paths for the child. + */ + child_grouped_rel = make_grouping_rel(root, child_input_rel, + child_target, + extra->target_parallel_safe, + child_extra.havingQual); + + /* Create grouping paths for this child relation. */ + create_ordinary_grouping_paths(root, child_input_rel, + child_grouped_rel, + agg_costs, gd, &child_extra, + &child_partially_grouped_rel); + + if (child_partially_grouped_rel) + { + partially_grouped_live_children = + lappend(partially_grouped_live_children, + child_partially_grouped_rel); + } + else + partial_grouping_valid = false; + + if (patype == PARTITIONWISE_AGGREGATE_FULL) + { + set_cheapest(child_grouped_rel); + grouped_live_children = lappend(grouped_live_children, + child_grouped_rel); + } + + pfree(appinfos); + } + + /* + * Try to create append paths for partially grouped children. For full + * partitionwise aggregation, we might have paths in the partial_pathlist + * if parallel aggregation is possible. For partial partitionwise + * aggregation, we may have paths in both pathlist and partial_pathlist. + * + * NB: We must have a partially grouped path for every child in order to + * generate a partially grouped path for this relation. + */ + if (partially_grouped_rel && partial_grouping_valid) + { + Assert(partially_grouped_live_children != NIL); + + add_paths_to_append_rel(root, partially_grouped_rel, + partially_grouped_live_children); + + /* + * We need call set_cheapest, since the finalization step will use the + * cheapest path from the rel. + */ + if (partially_grouped_rel->pathlist) + set_cheapest(partially_grouped_rel); + } + + /* If possible, create append paths for fully grouped children. */ + if (patype == PARTITIONWISE_AGGREGATE_FULL) + { + Assert(grouped_live_children != NIL); + + add_paths_to_append_rel(root, grouped_rel, grouped_live_children); + } +} + +/* + * group_by_has_partkey + * + * Returns true, if all the partition keys of the given relation are part of + * the GROUP BY clauses, false otherwise. + */ +static bool +group_by_has_partkey(RelOptInfo *input_rel, + List *targetList, + List *groupClause) +{ + List *groupexprs = get_sortgrouplist_exprs(groupClause, targetList); + int cnt = 0; + int partnatts; + + /* Input relation should be partitioned. */ + Assert(input_rel->part_scheme); + + /* Rule out early, if there are no partition keys present. */ + if (!input_rel->partexprs) + return false; + + partnatts = input_rel->part_scheme->partnatts; + + for (cnt = 0; cnt < partnatts; cnt++) + { + List *partexprs = input_rel->partexprs[cnt]; + ListCell *lc; + bool found = false; + + foreach(lc, partexprs) + { + Expr *partexpr = lfirst(lc); + + if (list_member(groupexprs, partexpr)) + { + found = true; + break; + } + } + + /* + * If none of the partition key expressions match with any of the + * GROUP BY expression, return false. + */ + if (!found) + return false; + } + + return true; +} diff --git a/src/backend/optimizer/plan/setrefs.c b/src/backend/optimizer/plan/setrefs.c new file mode 100644 index 0000000..9d912a8 --- /dev/null +++ b/src/backend/optimizer/plan/setrefs.c @@ -0,0 +1,3398 @@ +/*------------------------------------------------------------------------- + * + * setrefs.c + * Post-processing of a completed plan tree: fix references to subplan + * vars, compute regproc values for operators, etc + * + * Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group + * Portions Copyright (c) 1994, Regents of the University of California + * + * + * IDENTIFICATION + * src/backend/optimizer/plan/setrefs.c + * + *------------------------------------------------------------------------- + */ +#include "postgres.h" + +#include "access/transam.h" +#include "catalog/pg_type.h" +#include "nodes/makefuncs.h" +#include "nodes/nodeFuncs.h" +#include "optimizer/optimizer.h" +#include "optimizer/pathnode.h" +#include "optimizer/planmain.h" +#include "optimizer/planner.h" +#include "optimizer/tlist.h" +#include "tcop/utility.h" +#include "utils/lsyscache.h" +#include "utils/syscache.h" + + +typedef struct +{ + int varno; /* RT index of Var */ + AttrNumber varattno; /* attr number of Var */ + AttrNumber resno; /* TLE position of Var */ +} tlist_vinfo; + +typedef struct +{ + List *tlist; /* underlying target list */ + int num_vars; /* number of plain Var tlist entries */ + bool has_ph_vars; /* are there PlaceHolderVar entries? */ + bool has_non_vars; /* are there other entries? */ + tlist_vinfo vars[FLEXIBLE_ARRAY_MEMBER]; /* has num_vars entries */ +} indexed_tlist; + +typedef struct +{ + PlannerInfo *root; + int rtoffset; + double num_exec; +} fix_scan_expr_context; + +typedef struct +{ + PlannerInfo *root; + indexed_tlist *outer_itlist; + indexed_tlist *inner_itlist; + Index acceptable_rel; + int rtoffset; + double num_exec; +} fix_join_expr_context; + +typedef struct +{ + PlannerInfo *root; + indexed_tlist *subplan_itlist; + int newvarno; + int rtoffset; + double num_exec; +} fix_upper_expr_context; + +typedef struct +{ + PlannerInfo *root; + indexed_tlist *subplan_itlist; + int newvarno; +} fix_windowagg_cond_context; + +/* + * Selecting the best alternative in an AlternativeSubPlan expression requires + * estimating how many times that expression will be evaluated. For an + * expression in a plan node's targetlist, the plan's estimated number of + * output rows is clearly what to use, but for an expression in a qual it's + * far less clear. Since AlternativeSubPlans aren't heavily used, we don't + * want to expend a lot of cycles making such estimates. What we use is twice + * the number of output rows. That's not entirely unfounded: we know that + * clause_selectivity() would fall back to a default selectivity estimate + * of 0.5 for any SubPlan, so if the qual containing the SubPlan is the last + * to be applied (which it likely would be, thanks to order_qual_clauses()), + * this matches what we could have estimated in a far more laborious fashion. + * Obviously there are many other scenarios, but it's probably not worth the + * trouble to try to improve on this estimate, especially not when we don't + * have a better estimate for the selectivity of the SubPlan qual itself. + */ +#define NUM_EXEC_TLIST(parentplan) ((parentplan)->plan_rows) +#define NUM_EXEC_QUAL(parentplan) ((parentplan)->plan_rows * 2.0) + +/* + * Check if a Const node is a regclass value. We accept plain OID too, + * since a regclass Const will get folded to that type if it's an argument + * to oideq or similar operators. (This might result in some extraneous + * values in a plan's list of relation dependencies, but the worst result + * would be occasional useless replans.) + */ +#define ISREGCLASSCONST(con) \ + (((con)->consttype == REGCLASSOID || (con)->consttype == OIDOID) && \ + !(con)->constisnull) + +#define fix_scan_list(root, lst, rtoffset, num_exec) \ + ((List *) fix_scan_expr(root, (Node *) (lst), rtoffset, num_exec)) + +static void add_rtes_to_flat_rtable(PlannerInfo *root, bool recursing); +static void flatten_unplanned_rtes(PlannerGlobal *glob, RangeTblEntry *rte); +static bool flatten_rtes_walker(Node *node, PlannerGlobal *glob); +static void add_rte_to_flat_rtable(PlannerGlobal *glob, RangeTblEntry *rte); +static Plan *set_plan_refs(PlannerInfo *root, Plan *plan, int rtoffset); +static Plan *set_indexonlyscan_references(PlannerInfo *root, + IndexOnlyScan *plan, + int rtoffset); +static Plan *set_subqueryscan_references(PlannerInfo *root, + SubqueryScan *plan, + int rtoffset); +static Plan *clean_up_removed_plan_level(Plan *parent, Plan *child); +static void set_foreignscan_references(PlannerInfo *root, + ForeignScan *fscan, + int rtoffset); +static void set_customscan_references(PlannerInfo *root, + CustomScan *cscan, + int rtoffset); +static Plan *set_append_references(PlannerInfo *root, + Append *aplan, + int rtoffset); +static Plan *set_mergeappend_references(PlannerInfo *root, + MergeAppend *mplan, + int rtoffset); +static void set_hash_references(PlannerInfo *root, Plan *plan, int rtoffset); +static Relids offset_relid_set(Relids relids, int rtoffset); +static Node *fix_scan_expr(PlannerInfo *root, Node *node, + int rtoffset, double num_exec); +static Node *fix_scan_expr_mutator(Node *node, fix_scan_expr_context *context); +static bool fix_scan_expr_walker(Node *node, fix_scan_expr_context *context); +static void set_join_references(PlannerInfo *root, Join *join, int rtoffset); +static void set_upper_references(PlannerInfo *root, Plan *plan, int rtoffset); +static void set_param_references(PlannerInfo *root, Plan *plan); +static Node *convert_combining_aggrefs(Node *node, void *context); +static void set_dummy_tlist_references(Plan *plan, int rtoffset); +static indexed_tlist *build_tlist_index(List *tlist); +static Var *search_indexed_tlist_for_var(Var *var, + indexed_tlist *itlist, + int newvarno, + int rtoffset); +static Var *search_indexed_tlist_for_non_var(Expr *node, + indexed_tlist *itlist, + int newvarno); +static Var *search_indexed_tlist_for_sortgroupref(Expr *node, + Index sortgroupref, + indexed_tlist *itlist, + int newvarno); +static List *fix_join_expr(PlannerInfo *root, + List *clauses, + indexed_tlist *outer_itlist, + indexed_tlist *inner_itlist, + Index acceptable_rel, + int rtoffset, double num_exec); +static Node *fix_join_expr_mutator(Node *node, + fix_join_expr_context *context); +static Node *fix_upper_expr(PlannerInfo *root, + Node *node, + indexed_tlist *subplan_itlist, + int newvarno, + int rtoffset, double num_exec); +static Node *fix_upper_expr_mutator(Node *node, + fix_upper_expr_context *context); +static List *set_returning_clause_references(PlannerInfo *root, + List *rlist, + Plan *topplan, + Index resultRelation, + int rtoffset); +static List *set_windowagg_runcondition_references(PlannerInfo *root, + List *runcondition, + Plan *plan); + + +/***************************************************************************** + * + * SUBPLAN REFERENCES + * + *****************************************************************************/ + +/* + * set_plan_references + * + * This is the final processing pass of the planner/optimizer. The plan + * tree is complete; we just have to adjust some representational details + * for the convenience of the executor: + * + * 1. We flatten the various subquery rangetables into a single list, and + * zero out RangeTblEntry fields that are not useful to the executor. + * + * 2. We adjust Vars in scan nodes to be consistent with the flat rangetable. + * + * 3. We adjust Vars in upper plan nodes to refer to the outputs of their + * subplans. + * + * 4. Aggrefs in Agg plan nodes need to be adjusted in some cases involving + * partial aggregation or minmax aggregate optimization. + * + * 5. PARAM_MULTIEXPR Params are replaced by regular PARAM_EXEC Params, + * now that we have finished planning all MULTIEXPR subplans. + * + * 6. AlternativeSubPlan expressions are replaced by just one of their + * alternatives, using an estimate of how many times they'll be executed. + * + * 7. We compute regproc OIDs for operators (ie, we look up the function + * that implements each op). + * + * 8. We create lists of specific objects that the plan depends on. + * This will be used by plancache.c to drive invalidation of cached plans. + * Relation dependencies are represented by OIDs, and everything else by + * PlanInvalItems (this distinction is motivated by the shared-inval APIs). + * Currently, relations, user-defined functions, and domains are the only + * types of objects that are explicitly tracked this way. + * + * 9. We assign every plan node in the tree a unique ID. + * + * We also perform one final optimization step, which is to delete + * SubqueryScan, Append, and MergeAppend plan nodes that aren't doing + * anything useful. The reason for doing this last is that + * it can't readily be done before set_plan_references, because it would + * break set_upper_references: the Vars in the child plan's top tlist + * wouldn't match up with the Vars in the outer plan tree. A SubqueryScan + * serves a necessary function as a buffer between outer query and subquery + * variable numbering ... but after we've flattened the rangetable this is + * no longer a problem, since then there's only one rtindex namespace. + * Likewise, Append and MergeAppend buffer between the parent and child vars + * of an appendrel, but we don't need to worry about that once we've done + * set_plan_references. + * + * set_plan_references recursively traverses the whole plan tree. + * + * The return value is normally the same Plan node passed in, but can be + * different when the passed-in Plan is a node we decide isn't needed. + * + * The flattened rangetable entries are appended to root->glob->finalrtable. + * Also, rowmarks entries are appended to root->glob->finalrowmarks, and the + * RT indexes of ModifyTable result relations to root->glob->resultRelations, + * and flattened AppendRelInfos are appended to root->glob->appendRelations. + * Plan dependencies are appended to root->glob->relationOids (for relations) + * and root->glob->invalItems (for everything else). + * + * Notice that we modify Plan nodes in-place, but use expression_tree_mutator + * to process targetlist and qual expressions. We can assume that the Plan + * nodes were just built by the planner and are not multiply referenced, but + * it's not so safe to assume that for expression tree nodes. + */ +Plan * +set_plan_references(PlannerInfo *root, Plan *plan) +{ + Plan *result; + PlannerGlobal *glob = root->glob; + int rtoffset = list_length(glob->finalrtable); + ListCell *lc; + + /* + * Add all the query's RTEs to the flattened rangetable. The live ones + * will have their rangetable indexes increased by rtoffset. (Additional + * RTEs, not referenced by the Plan tree, might get added after those.) + */ + add_rtes_to_flat_rtable(root, false); + + /* + * Adjust RT indexes of PlanRowMarks and add to final rowmarks list + */ + foreach(lc, root->rowMarks) + { + PlanRowMark *rc = lfirst_node(PlanRowMark, lc); + PlanRowMark *newrc; + + /* flat copy is enough since all fields are scalars */ + newrc = (PlanRowMark *) palloc(sizeof(PlanRowMark)); + memcpy(newrc, rc, sizeof(PlanRowMark)); + + /* adjust indexes ... but *not* the rowmarkId */ + newrc->rti += rtoffset; + newrc->prti += rtoffset; + + glob->finalrowmarks = lappend(glob->finalrowmarks, newrc); + } + + /* + * Adjust RT indexes of AppendRelInfos and add to final appendrels list. + * We assume the AppendRelInfos were built during planning and don't need + * to be copied. + */ + foreach(lc, root->append_rel_list) + { + AppendRelInfo *appinfo = lfirst_node(AppendRelInfo, lc); + + /* adjust RT indexes */ + appinfo->parent_relid += rtoffset; + appinfo->child_relid += rtoffset; + + /* + * Rather than adjust the translated_vars entries, just drop 'em. + * Neither the executor nor EXPLAIN currently need that data. + */ + appinfo->translated_vars = NIL; + + glob->appendRelations = lappend(glob->appendRelations, appinfo); + } + + /* If needed, create workspace for processing AlternativeSubPlans */ + if (root->hasAlternativeSubPlans) + { + root->isAltSubplan = (bool *) + palloc0(list_length(glob->subplans) * sizeof(bool)); + root->isUsedSubplan = (bool *) + palloc0(list_length(glob->subplans) * sizeof(bool)); + } + + /* Now fix the Plan tree */ + result = set_plan_refs(root, plan, rtoffset); + + /* + * If we have AlternativeSubPlans, it is likely that we now have some + * unreferenced subplans in glob->subplans. To avoid expending cycles on + * those subplans later, get rid of them by setting those list entries to + * NULL. (Note: we can't do this immediately upon processing an + * AlternativeSubPlan, because there may be multiple copies of the + * AlternativeSubPlan, and they can get resolved differently.) + */ + if (root->hasAlternativeSubPlans) + { + foreach(lc, glob->subplans) + { + int ndx = foreach_current_index(lc); + + /* + * If it was used by some AlternativeSubPlan in this query level, + * but wasn't selected as best by any AlternativeSubPlan, then we + * don't need it. Do not touch subplans that aren't parts of + * AlternativeSubPlans. + */ + if (root->isAltSubplan[ndx] && !root->isUsedSubplan[ndx]) + lfirst(lc) = NULL; + } + } + + return result; +} + +/* + * Extract RangeTblEntries from the plan's rangetable, and add to flat rtable + * + * This can recurse into subquery plans; "recursing" is true if so. + */ +static void +add_rtes_to_flat_rtable(PlannerInfo *root, bool recursing) +{ + PlannerGlobal *glob = root->glob; + Index rti; + ListCell *lc; + + /* + * Add the query's own RTEs to the flattened rangetable. + * + * At top level, we must add all RTEs so that their indexes in the + * flattened rangetable match up with their original indexes. When + * recursing, we only care about extracting relation RTEs. + */ + foreach(lc, root->parse->rtable) + { + RangeTblEntry *rte = (RangeTblEntry *) lfirst(lc); + + if (!recursing || rte->rtekind == RTE_RELATION) + add_rte_to_flat_rtable(glob, rte); + } + + /* + * If there are any dead subqueries, they are not referenced in the Plan + * tree, so we must add RTEs contained in them to the flattened rtable + * separately. (If we failed to do this, the executor would not perform + * expected permission checks for tables mentioned in such subqueries.) + * + * Note: this pass over the rangetable can't be combined with the previous + * one, because that would mess up the numbering of the live RTEs in the + * flattened rangetable. + */ + rti = 1; + foreach(lc, root->parse->rtable) + { + RangeTblEntry *rte = (RangeTblEntry *) lfirst(lc); + + /* + * We should ignore inheritance-parent RTEs: their contents have been + * pulled up into our rangetable already. Also ignore any subquery + * RTEs without matching RelOptInfos, as they likewise have been + * pulled up. + */ + if (rte->rtekind == RTE_SUBQUERY && !rte->inh && + rti < root->simple_rel_array_size) + { + RelOptInfo *rel = root->simple_rel_array[rti]; + + if (rel != NULL) + { + Assert(rel->relid == rti); /* sanity check on array */ + + /* + * The subquery might never have been planned at all, if it + * was excluded on the basis of self-contradictory constraints + * in our query level. In this case apply + * flatten_unplanned_rtes. + * + * If it was planned but the result rel is dummy, we assume + * that it has been omitted from our plan tree (see + * set_subquery_pathlist), and recurse to pull up its RTEs. + * + * Otherwise, it should be represented by a SubqueryScan node + * somewhere in our plan tree, and we'll pull up its RTEs when + * we process that plan node. + * + * However, if we're recursing, then we should pull up RTEs + * whether the subquery is dummy or not, because we've found + * that some upper query level is treating this one as dummy, + * and so we won't scan this level's plan tree at all. + */ + if (rel->subroot == NULL) + flatten_unplanned_rtes(glob, rte); + else if (recursing || + IS_DUMMY_REL(fetch_upper_rel(rel->subroot, + UPPERREL_FINAL, NULL))) + add_rtes_to_flat_rtable(rel->subroot, true); + } + } + rti++; + } +} + +/* + * Extract RangeTblEntries from a subquery that was never planned at all + */ +static void +flatten_unplanned_rtes(PlannerGlobal *glob, RangeTblEntry *rte) +{ + /* Use query_tree_walker to find all RTEs in the parse tree */ + (void) query_tree_walker(rte->subquery, + flatten_rtes_walker, + (void *) glob, + QTW_EXAMINE_RTES_BEFORE); +} + +static bool +flatten_rtes_walker(Node *node, PlannerGlobal *glob) +{ + if (node == NULL) + return false; + if (IsA(node, RangeTblEntry)) + { + RangeTblEntry *rte = (RangeTblEntry *) node; + + /* As above, we need only save relation RTEs */ + if (rte->rtekind == RTE_RELATION) + add_rte_to_flat_rtable(glob, rte); + return false; + } + if (IsA(node, Query)) + { + /* Recurse into subselects */ + return query_tree_walker((Query *) node, + flatten_rtes_walker, + (void *) glob, + QTW_EXAMINE_RTES_BEFORE); + } + return expression_tree_walker(node, flatten_rtes_walker, + (void *) glob); +} + +/* + * Add (a copy of) the given RTE to the final rangetable + * + * In the flat rangetable, we zero out substructure pointers that are not + * needed by the executor; this reduces the storage space and copying cost + * for cached plans. We keep only the ctename, alias and eref Alias fields, + * which are needed by EXPLAIN, and the selectedCols, insertedCols, + * updatedCols, and extraUpdatedCols bitmaps, which are needed for + * executor-startup permissions checking and for trigger event checking. + */ +static void +add_rte_to_flat_rtable(PlannerGlobal *glob, RangeTblEntry *rte) +{ + RangeTblEntry *newrte; + + /* flat copy to duplicate all the scalar fields */ + newrte = (RangeTblEntry *) palloc(sizeof(RangeTblEntry)); + memcpy(newrte, rte, sizeof(RangeTblEntry)); + + /* zap unneeded sub-structure */ + newrte->tablesample = NULL; + newrte->subquery = NULL; + newrte->joinaliasvars = NIL; + newrte->joinleftcols = NIL; + newrte->joinrightcols = NIL; + newrte->join_using_alias = NULL; + newrte->functions = NIL; + newrte->tablefunc = NULL; + newrte->values_lists = NIL; + newrte->coltypes = NIL; + newrte->coltypmods = NIL; + newrte->colcollations = NIL; + newrte->securityQuals = NIL; + + glob->finalrtable = lappend(glob->finalrtable, newrte); + + /* + * If it's a plain relation RTE, add the table to relationOids. + * + * We do this even though the RTE might be unreferenced in the plan tree; + * this would correspond to cases such as views that were expanded, child + * tables that were eliminated by constraint exclusion, etc. Schema + * invalidation on such a rel must still force rebuilding of the plan. + * + * Note we don't bother to avoid making duplicate list entries. We could, + * but it would probably cost more cycles than it would save. + */ + if (newrte->rtekind == RTE_RELATION) + glob->relationOids = lappend_oid(glob->relationOids, newrte->relid); +} + +/* + * set_plan_refs: recurse through the Plan nodes of a single subquery level + */ +static Plan * +set_plan_refs(PlannerInfo *root, Plan *plan, int rtoffset) +{ + ListCell *l; + + if (plan == NULL) + return NULL; + + /* Assign this node a unique ID. */ + plan->plan_node_id = root->glob->lastPlanNodeId++; + + /* + * Plan-type-specific fixes + */ + switch (nodeTag(plan)) + { + case T_SeqScan: + { + SeqScan *splan = (SeqScan *) plan; + + splan->scan.scanrelid += rtoffset; + splan->scan.plan.targetlist = + fix_scan_list(root, splan->scan.plan.targetlist, + rtoffset, NUM_EXEC_TLIST(plan)); + splan->scan.plan.qual = + fix_scan_list(root, splan->scan.plan.qual, + rtoffset, NUM_EXEC_QUAL(plan)); + } + break; + case T_SampleScan: + { + SampleScan *splan = (SampleScan *) plan; + + splan->scan.scanrelid += rtoffset; + splan->scan.plan.targetlist = + fix_scan_list(root, splan->scan.plan.targetlist, + rtoffset, NUM_EXEC_TLIST(plan)); + splan->scan.plan.qual = + fix_scan_list(root, splan->scan.plan.qual, + rtoffset, NUM_EXEC_QUAL(plan)); + splan->tablesample = (TableSampleClause *) + fix_scan_expr(root, (Node *) splan->tablesample, + rtoffset, 1); + } + break; + case T_IndexScan: + { + IndexScan *splan = (IndexScan *) plan; + + splan->scan.scanrelid += rtoffset; + splan->scan.plan.targetlist = + fix_scan_list(root, splan->scan.plan.targetlist, + rtoffset, NUM_EXEC_TLIST(plan)); + splan->scan.plan.qual = + fix_scan_list(root, splan->scan.plan.qual, + rtoffset, NUM_EXEC_QUAL(plan)); + splan->indexqual = + fix_scan_list(root, splan->indexqual, + rtoffset, 1); + splan->indexqualorig = + fix_scan_list(root, splan->indexqualorig, + rtoffset, NUM_EXEC_QUAL(plan)); + splan->indexorderby = + fix_scan_list(root, splan->indexorderby, + rtoffset, 1); + splan->indexorderbyorig = + fix_scan_list(root, splan->indexorderbyorig, + rtoffset, NUM_EXEC_QUAL(plan)); + } + break; + case T_IndexOnlyScan: + { + IndexOnlyScan *splan = (IndexOnlyScan *) plan; + + return set_indexonlyscan_references(root, splan, rtoffset); + } + break; + case T_BitmapIndexScan: + { + BitmapIndexScan *splan = (BitmapIndexScan *) plan; + + splan->scan.scanrelid += rtoffset; + /* no need to fix targetlist and qual */ + Assert(splan->scan.plan.targetlist == NIL); + Assert(splan->scan.plan.qual == NIL); + splan->indexqual = + fix_scan_list(root, splan->indexqual, rtoffset, 1); + splan->indexqualorig = + fix_scan_list(root, splan->indexqualorig, + rtoffset, NUM_EXEC_QUAL(plan)); + } + break; + case T_BitmapHeapScan: + { + BitmapHeapScan *splan = (BitmapHeapScan *) plan; + + splan->scan.scanrelid += rtoffset; + splan->scan.plan.targetlist = + fix_scan_list(root, splan->scan.plan.targetlist, + rtoffset, NUM_EXEC_TLIST(plan)); + splan->scan.plan.qual = + fix_scan_list(root, splan->scan.plan.qual, + rtoffset, NUM_EXEC_QUAL(plan)); + splan->bitmapqualorig = + fix_scan_list(root, splan->bitmapqualorig, + rtoffset, NUM_EXEC_QUAL(plan)); + } + break; + case T_TidScan: + { + TidScan *splan = (TidScan *) plan; + + splan->scan.scanrelid += rtoffset; + splan->scan.plan.targetlist = + fix_scan_list(root, splan->scan.plan.targetlist, + rtoffset, NUM_EXEC_TLIST(plan)); + splan->scan.plan.qual = + fix_scan_list(root, splan->scan.plan.qual, + rtoffset, NUM_EXEC_QUAL(plan)); + splan->tidquals = + fix_scan_list(root, splan->tidquals, + rtoffset, 1); + } + break; + case T_TidRangeScan: + { + TidRangeScan *splan = (TidRangeScan *) plan; + + splan->scan.scanrelid += rtoffset; + splan->scan.plan.targetlist = + fix_scan_list(root, splan->scan.plan.targetlist, + rtoffset, NUM_EXEC_TLIST(plan)); + splan->scan.plan.qual = + fix_scan_list(root, splan->scan.plan.qual, + rtoffset, NUM_EXEC_QUAL(plan)); + splan->tidrangequals = + fix_scan_list(root, splan->tidrangequals, + rtoffset, 1); + } + break; + case T_SubqueryScan: + /* Needs special treatment, see comments below */ + return set_subqueryscan_references(root, + (SubqueryScan *) plan, + rtoffset); + case T_FunctionScan: + { + FunctionScan *splan = (FunctionScan *) plan; + + splan->scan.scanrelid += rtoffset; + splan->scan.plan.targetlist = + fix_scan_list(root, splan->scan.plan.targetlist, + rtoffset, NUM_EXEC_TLIST(plan)); + splan->scan.plan.qual = + fix_scan_list(root, splan->scan.plan.qual, + rtoffset, NUM_EXEC_QUAL(plan)); + splan->functions = + fix_scan_list(root, splan->functions, rtoffset, 1); + } + break; + case T_TableFuncScan: + { + TableFuncScan *splan = (TableFuncScan *) plan; + + splan->scan.scanrelid += rtoffset; + splan->scan.plan.targetlist = + fix_scan_list(root, splan->scan.plan.targetlist, + rtoffset, NUM_EXEC_TLIST(plan)); + splan->scan.plan.qual = + fix_scan_list(root, splan->scan.plan.qual, + rtoffset, NUM_EXEC_QUAL(plan)); + splan->tablefunc = (TableFunc *) + fix_scan_expr(root, (Node *) splan->tablefunc, + rtoffset, 1); + } + break; + case T_ValuesScan: + { + ValuesScan *splan = (ValuesScan *) plan; + + splan->scan.scanrelid += rtoffset; + splan->scan.plan.targetlist = + fix_scan_list(root, splan->scan.plan.targetlist, + rtoffset, NUM_EXEC_TLIST(plan)); + splan->scan.plan.qual = + fix_scan_list(root, splan->scan.plan.qual, + rtoffset, NUM_EXEC_QUAL(plan)); + splan->values_lists = + fix_scan_list(root, splan->values_lists, + rtoffset, 1); + } + break; + case T_CteScan: + { + CteScan *splan = (CteScan *) plan; + + splan->scan.scanrelid += rtoffset; + splan->scan.plan.targetlist = + fix_scan_list(root, splan->scan.plan.targetlist, + rtoffset, NUM_EXEC_TLIST(plan)); + splan->scan.plan.qual = + fix_scan_list(root, splan->scan.plan.qual, + rtoffset, NUM_EXEC_QUAL(plan)); + } + break; + case T_NamedTuplestoreScan: + { + NamedTuplestoreScan *splan = (NamedTuplestoreScan *) plan; + + splan->scan.scanrelid += rtoffset; + splan->scan.plan.targetlist = + fix_scan_list(root, splan->scan.plan.targetlist, + rtoffset, NUM_EXEC_TLIST(plan)); + splan->scan.plan.qual = + fix_scan_list(root, splan->scan.plan.qual, + rtoffset, NUM_EXEC_QUAL(plan)); + } + break; + case T_WorkTableScan: + { + WorkTableScan *splan = (WorkTableScan *) plan; + + splan->scan.scanrelid += rtoffset; + splan->scan.plan.targetlist = + fix_scan_list(root, splan->scan.plan.targetlist, + rtoffset, NUM_EXEC_TLIST(plan)); + splan->scan.plan.qual = + fix_scan_list(root, splan->scan.plan.qual, + rtoffset, NUM_EXEC_QUAL(plan)); + } + break; + case T_ForeignScan: + set_foreignscan_references(root, (ForeignScan *) plan, rtoffset); + break; + case T_CustomScan: + set_customscan_references(root, (CustomScan *) plan, rtoffset); + break; + + case T_NestLoop: + case T_MergeJoin: + case T_HashJoin: + set_join_references(root, (Join *) plan, rtoffset); + break; + + case T_Gather: + case T_GatherMerge: + { + set_upper_references(root, plan, rtoffset); + set_param_references(root, plan); + } + break; + + case T_Hash: + set_hash_references(root, plan, rtoffset); + break; + + case T_Memoize: + { + Memoize *mplan = (Memoize *) plan; + + /* + * Memoize does not evaluate its targetlist. It just uses the + * same targetlist from its outer subnode. + */ + set_dummy_tlist_references(plan, rtoffset); + + mplan->param_exprs = fix_scan_list(root, mplan->param_exprs, + rtoffset, + NUM_EXEC_TLIST(plan)); + break; + } + + case T_Material: + case T_Sort: + case T_IncrementalSort: + case T_Unique: + case T_SetOp: + + /* + * These plan types don't actually bother to evaluate their + * targetlists, because they just return their unmodified input + * tuples. Even though the targetlist won't be used by the + * executor, we fix it up for possible use by EXPLAIN (not to + * mention ease of debugging --- wrong varnos are very confusing). + */ + set_dummy_tlist_references(plan, rtoffset); + + /* + * Since these plan types don't check quals either, we should not + * find any qual expression attached to them. + */ + Assert(plan->qual == NIL); + break; + case T_LockRows: + { + LockRows *splan = (LockRows *) plan; + + /* + * Like the plan types above, LockRows doesn't evaluate its + * tlist or quals. But we have to fix up the RT indexes in + * its rowmarks. + */ + set_dummy_tlist_references(plan, rtoffset); + Assert(splan->plan.qual == NIL); + + foreach(l, splan->rowMarks) + { + PlanRowMark *rc = (PlanRowMark *) lfirst(l); + + rc->rti += rtoffset; + rc->prti += rtoffset; + } + } + break; + case T_Limit: + { + Limit *splan = (Limit *) plan; + + /* + * Like the plan types above, Limit doesn't evaluate its tlist + * or quals. It does have live expressions for limit/offset, + * however; and those cannot contain subplan variable refs, so + * fix_scan_expr works for them. + */ + set_dummy_tlist_references(plan, rtoffset); + Assert(splan->plan.qual == NIL); + + splan->limitOffset = + fix_scan_expr(root, splan->limitOffset, rtoffset, 1); + splan->limitCount = + fix_scan_expr(root, splan->limitCount, rtoffset, 1); + } + break; + case T_Agg: + { + Agg *agg = (Agg *) plan; + + /* + * If this node is combining partial-aggregation results, we + * must convert its Aggrefs to contain references to the + * partial-aggregate subexpressions that will be available + * from the child plan node. + */ + if (DO_AGGSPLIT_COMBINE(agg->aggsplit)) + { + plan->targetlist = (List *) + convert_combining_aggrefs((Node *) plan->targetlist, + NULL); + plan->qual = (List *) + convert_combining_aggrefs((Node *) plan->qual, + NULL); + } + + set_upper_references(root, plan, rtoffset); + } + break; + case T_Group: + set_upper_references(root, plan, rtoffset); + break; + case T_WindowAgg: + { + WindowAgg *wplan = (WindowAgg *) plan; + + /* + * Adjust the WindowAgg's run conditions by swapping the + * WindowFuncs references out to instead reference the Var in + * the scan slot so that when the executor evaluates the + * runCondition, it receives the WindowFunc's value from the + * slot that the result has just been stored into rather than + * evaluating the WindowFunc all over again. + */ + wplan->runCondition = set_windowagg_runcondition_references(root, + wplan->runCondition, + (Plan *) wplan); + + set_upper_references(root, plan, rtoffset); + + /* + * Like Limit node limit/offset expressions, WindowAgg has + * frame offset expressions, which cannot contain subplan + * variable refs, so fix_scan_expr works for them. + */ + wplan->startOffset = + fix_scan_expr(root, wplan->startOffset, rtoffset, 1); + wplan->endOffset = + fix_scan_expr(root, wplan->endOffset, rtoffset, 1); + wplan->runCondition = fix_scan_list(root, + wplan->runCondition, + rtoffset, + NUM_EXEC_TLIST(plan)); + wplan->runConditionOrig = fix_scan_list(root, + wplan->runConditionOrig, + rtoffset, + NUM_EXEC_TLIST(plan)); + } + break; + case T_Result: + { + Result *splan = (Result *) plan; + + /* + * Result may or may not have a subplan; if not, it's more + * like a scan node than an upper node. + */ + if (splan->plan.lefttree != NULL) + set_upper_references(root, plan, rtoffset); + else + { + /* + * The tlist of a childless Result could contain + * unresolved ROWID_VAR Vars, in case it's representing a + * target relation which is completely empty because of + * constraint exclusion. Replace any such Vars by null + * constants, as though they'd been resolved for a leaf + * scan node that doesn't support them. We could have + * fix_scan_expr do this, but since the case is only + * expected to occur here, it seems safer to special-case + * it here and keep the assertions that ROWID_VARs + * shouldn't be seen by fix_scan_expr. + */ + foreach(l, splan->plan.targetlist) + { + TargetEntry *tle = (TargetEntry *) lfirst(l); + Var *var = (Var *) tle->expr; + + if (var && IsA(var, Var) && var->varno == ROWID_VAR) + tle->expr = (Expr *) makeNullConst(var->vartype, + var->vartypmod, + var->varcollid); + } + + splan->plan.targetlist = + fix_scan_list(root, splan->plan.targetlist, + rtoffset, NUM_EXEC_TLIST(plan)); + splan->plan.qual = + fix_scan_list(root, splan->plan.qual, + rtoffset, NUM_EXEC_QUAL(plan)); + } + /* resconstantqual can't contain any subplan variable refs */ + splan->resconstantqual = + fix_scan_expr(root, splan->resconstantqual, rtoffset, 1); + } + break; + case T_ProjectSet: + set_upper_references(root, plan, rtoffset); + break; + case T_ModifyTable: + { + ModifyTable *splan = (ModifyTable *) plan; + Plan *subplan = outerPlan(splan); + + Assert(splan->plan.targetlist == NIL); + Assert(splan->plan.qual == NIL); + + splan->withCheckOptionLists = + fix_scan_list(root, splan->withCheckOptionLists, + rtoffset, 1); + + if (splan->returningLists) + { + List *newRL = NIL; + ListCell *lcrl, + *lcrr; + + /* + * Pass each per-resultrel returningList through + * set_returning_clause_references(). + */ + Assert(list_length(splan->returningLists) == list_length(splan->resultRelations)); + forboth(lcrl, splan->returningLists, + lcrr, splan->resultRelations) + { + List *rlist = (List *) lfirst(lcrl); + Index resultrel = lfirst_int(lcrr); + + rlist = set_returning_clause_references(root, + rlist, + subplan, + resultrel, + rtoffset); + newRL = lappend(newRL, rlist); + } + splan->returningLists = newRL; + + /* + * Set up the visible plan targetlist as being the same as + * the first RETURNING list. This is for the use of + * EXPLAIN; the executor won't pay any attention to the + * targetlist. We postpone this step until here so that + * we don't have to do set_returning_clause_references() + * twice on identical targetlists. + */ + splan->plan.targetlist = copyObject(linitial(newRL)); + } + + /* + * We treat ModifyTable with ON CONFLICT as a form of 'pseudo + * join', where the inner side is the EXCLUDED tuple. + * Therefore use fix_join_expr to setup the relevant variables + * to INNER_VAR. We explicitly don't create any OUTER_VARs as + * those are already used by RETURNING and it seems better to + * be non-conflicting. + */ + if (splan->onConflictSet) + { + indexed_tlist *itlist; + + itlist = build_tlist_index(splan->exclRelTlist); + + splan->onConflictSet = + fix_join_expr(root, splan->onConflictSet, + NULL, itlist, + linitial_int(splan->resultRelations), + rtoffset, NUM_EXEC_QUAL(plan)); + + splan->onConflictWhere = (Node *) + fix_join_expr(root, (List *) splan->onConflictWhere, + NULL, itlist, + linitial_int(splan->resultRelations), + rtoffset, NUM_EXEC_QUAL(plan)); + + pfree(itlist); + + splan->exclRelTlist = + fix_scan_list(root, splan->exclRelTlist, rtoffset, 1); + } + + /* + * The MERGE statement produces the target rows by performing + * a right join between the target relation and the source + * relation (which could be a plain relation or a subquery). + * The INSERT and UPDATE actions of the MERGE statement + * require access to the columns from the source relation. We + * arrange things so that the source relation attributes are + * available as INNER_VAR and the target relation attributes + * are available from the scan tuple. + */ + if (splan->mergeActionLists != NIL) + { + ListCell *lca, + *lcr; + + /* + * Fix the targetList of individual action nodes so that + * the so-called "source relation" Vars are referenced as + * INNER_VAR. Note that for this to work correctly during + * execution, the ecxt_innertuple must be set to the tuple + * obtained by executing the subplan, which is what + * constitutes the "source relation". + * + * We leave the Vars from the result relation (i.e. the + * target relation) unchanged i.e. those Vars would be + * picked from the scan slot. So during execution, we must + * ensure that ecxt_scantuple is setup correctly to refer + * to the tuple from the target relation. + */ + indexed_tlist *itlist; + + itlist = build_tlist_index(subplan->targetlist); + + forboth(lca, splan->mergeActionLists, + lcr, splan->resultRelations) + { + List *mergeActionList = lfirst(lca); + Index resultrel = lfirst_int(lcr); + + foreach(l, mergeActionList) + { + MergeAction *action = (MergeAction *) lfirst(l); + + /* Fix targetList of each action. */ + action->targetList = fix_join_expr(root, + action->targetList, + NULL, itlist, + resultrel, + rtoffset, + NUM_EXEC_TLIST(plan)); + + /* Fix quals too. */ + action->qual = (Node *) fix_join_expr(root, + (List *) action->qual, + NULL, itlist, + resultrel, + rtoffset, + NUM_EXEC_QUAL(plan)); + } + } + } + + splan->nominalRelation += rtoffset; + if (splan->rootRelation) + splan->rootRelation += rtoffset; + splan->exclRelRTI += rtoffset; + + foreach(l, splan->resultRelations) + { + lfirst_int(l) += rtoffset; + } + foreach(l, splan->rowMarks) + { + PlanRowMark *rc = (PlanRowMark *) lfirst(l); + + rc->rti += rtoffset; + rc->prti += rtoffset; + } + + /* + * Append this ModifyTable node's final result relation RT + * index(es) to the global list for the plan. + */ + root->glob->resultRelations = + list_concat(root->glob->resultRelations, + splan->resultRelations); + if (splan->rootRelation) + { + root->glob->resultRelations = + lappend_int(root->glob->resultRelations, + splan->rootRelation); + } + } + break; + case T_Append: + /* Needs special treatment, see comments below */ + return set_append_references(root, + (Append *) plan, + rtoffset); + case T_MergeAppend: + /* Needs special treatment, see comments below */ + return set_mergeappend_references(root, + (MergeAppend *) plan, + rtoffset); + case T_RecursiveUnion: + /* This doesn't evaluate targetlist or check quals either */ + set_dummy_tlist_references(plan, rtoffset); + Assert(plan->qual == NIL); + break; + case T_BitmapAnd: + { + BitmapAnd *splan = (BitmapAnd *) plan; + + /* BitmapAnd works like Append, but has no tlist */ + Assert(splan->plan.targetlist == NIL); + Assert(splan->plan.qual == NIL); + foreach(l, splan->bitmapplans) + { + lfirst(l) = set_plan_refs(root, + (Plan *) lfirst(l), + rtoffset); + } + } + break; + case T_BitmapOr: + { + BitmapOr *splan = (BitmapOr *) plan; + + /* BitmapOr works like Append, but has no tlist */ + Assert(splan->plan.targetlist == NIL); + Assert(splan->plan.qual == NIL); + foreach(l, splan->bitmapplans) + { + lfirst(l) = set_plan_refs(root, + (Plan *) lfirst(l), + rtoffset); + } + } + break; + default: + elog(ERROR, "unrecognized node type: %d", + (int) nodeTag(plan)); + break; + } + + /* + * Now recurse into child plans, if any + * + * NOTE: it is essential that we recurse into child plans AFTER we set + * subplan references in this plan's tlist and quals. If we did the + * reference-adjustments bottom-up, then we would fail to match this + * plan's var nodes against the already-modified nodes of the children. + */ + plan->lefttree = set_plan_refs(root, plan->lefttree, rtoffset); + plan->righttree = set_plan_refs(root, plan->righttree, rtoffset); + + return plan; +} + +/* + * set_indexonlyscan_references + * Do set_plan_references processing on an IndexOnlyScan + * + * This is unlike the handling of a plain IndexScan because we have to + * convert Vars referencing the heap into Vars referencing the index. + * We can use the fix_upper_expr machinery for that, by working from a + * targetlist describing the index columns. + */ +static Plan * +set_indexonlyscan_references(PlannerInfo *root, + IndexOnlyScan *plan, + int rtoffset) +{ + indexed_tlist *index_itlist; + List *stripped_indextlist; + ListCell *lc; + + /* + * Vars in the plan node's targetlist, qual, and recheckqual must only + * reference columns that the index AM can actually return. To ensure + * this, remove non-returnable columns (which are marked as resjunk) from + * the indexed tlist. We can just drop them because the indexed_tlist + * machinery pays attention to TLE resnos, not physical list position. + */ + stripped_indextlist = NIL; + foreach(lc, plan->indextlist) + { + TargetEntry *indextle = (TargetEntry *) lfirst(lc); + + if (!indextle->resjunk) + stripped_indextlist = lappend(stripped_indextlist, indextle); + } + + index_itlist = build_tlist_index(stripped_indextlist); + + plan->scan.scanrelid += rtoffset; + plan->scan.plan.targetlist = (List *) + fix_upper_expr(root, + (Node *) plan->scan.plan.targetlist, + index_itlist, + INDEX_VAR, + rtoffset, + NUM_EXEC_TLIST((Plan *) plan)); + plan->scan.plan.qual = (List *) + fix_upper_expr(root, + (Node *) plan->scan.plan.qual, + index_itlist, + INDEX_VAR, + rtoffset, + NUM_EXEC_QUAL((Plan *) plan)); + plan->recheckqual = (List *) + fix_upper_expr(root, + (Node *) plan->recheckqual, + index_itlist, + INDEX_VAR, + rtoffset, + NUM_EXEC_QUAL((Plan *) plan)); + /* indexqual is already transformed to reference index columns */ + plan->indexqual = fix_scan_list(root, plan->indexqual, + rtoffset, 1); + /* indexorderby is already transformed to reference index columns */ + plan->indexorderby = fix_scan_list(root, plan->indexorderby, + rtoffset, 1); + /* indextlist must NOT be transformed to reference index columns */ + plan->indextlist = fix_scan_list(root, plan->indextlist, + rtoffset, NUM_EXEC_TLIST((Plan *) plan)); + + pfree(index_itlist); + + return (Plan *) plan; +} + +/* + * set_subqueryscan_references + * Do set_plan_references processing on a SubqueryScan + * + * We try to strip out the SubqueryScan entirely; if we can't, we have + * to do the normal processing on it. + */ +static Plan * +set_subqueryscan_references(PlannerInfo *root, + SubqueryScan *plan, + int rtoffset) +{ + RelOptInfo *rel; + Plan *result; + + /* Need to look up the subquery's RelOptInfo, since we need its subroot */ + rel = find_base_rel(root, plan->scan.scanrelid); + + /* Recursively process the subplan */ + plan->subplan = set_plan_references(rel->subroot, plan->subplan); + + if (trivial_subqueryscan(plan)) + { + /* + * We can omit the SubqueryScan node and just pull up the subplan. + */ + result = clean_up_removed_plan_level((Plan *) plan, plan->subplan); + } + else + { + /* + * Keep the SubqueryScan node. We have to do the processing that + * set_plan_references would otherwise have done on it. Notice we do + * not do set_upper_references() here, because a SubqueryScan will + * always have been created with correct references to its subplan's + * outputs to begin with. + */ + plan->scan.scanrelid += rtoffset; + plan->scan.plan.targetlist = + fix_scan_list(root, plan->scan.plan.targetlist, + rtoffset, NUM_EXEC_TLIST((Plan *) plan)); + plan->scan.plan.qual = + fix_scan_list(root, plan->scan.plan.qual, + rtoffset, NUM_EXEC_QUAL((Plan *) plan)); + + result = (Plan *) plan; + } + + return result; +} + +/* + * trivial_subqueryscan + * Detect whether a SubqueryScan can be deleted from the plan tree. + * + * We can delete it if it has no qual to check and the targetlist just + * regurgitates the output of the child plan. + * + * This can be called from mark_async_capable_plan(), a helper function for + * create_append_plan(), before set_subqueryscan_references(), to determine + * triviality of a SubqueryScan that is a child of an Append node. So we + * cache the result in the SubqueryScan node to avoid repeated computation. + * + * Note: when called from mark_async_capable_plan(), we determine the result + * before running finalize_plan() on the SubqueryScan node (if needed) and + * set_plan_references() on the subplan tree, but this would be safe, because + * 1) finalize_plan() doesn't modify the tlist or quals for the SubqueryScan + * node (or that for any plan node in the subplan tree), and + * 2) set_plan_references() modifies the tlist for every plan node in the + * subplan tree, but keeps const/resjunk columns as const/resjunk ones and + * preserves the length and order of the tlist, and + * 3) set_plan_references() might delete the topmost plan node like an Append + * or MergeAppend from the subplan tree and pull up the child plan node, + * but in that case, the tlist for the child plan node exactly matches the + * parent. + */ +bool +trivial_subqueryscan(SubqueryScan *plan) +{ + int attrno; + ListCell *lp, + *lc; + + /* We might have detected this already; in which case reuse the result */ + if (plan->scanstatus == SUBQUERY_SCAN_TRIVIAL) + return true; + if (plan->scanstatus == SUBQUERY_SCAN_NONTRIVIAL) + return false; + Assert(plan->scanstatus == SUBQUERY_SCAN_UNKNOWN); + /* Initially, mark the SubqueryScan as non-deletable from the plan tree */ + plan->scanstatus = SUBQUERY_SCAN_NONTRIVIAL; + + if (plan->scan.plan.qual != NIL) + return false; + + if (list_length(plan->scan.plan.targetlist) != + list_length(plan->subplan->targetlist)) + return false; /* tlists not same length */ + + attrno = 1; + forboth(lp, plan->scan.plan.targetlist, lc, plan->subplan->targetlist) + { + TargetEntry *ptle = (TargetEntry *) lfirst(lp); + TargetEntry *ctle = (TargetEntry *) lfirst(lc); + + if (ptle->resjunk != ctle->resjunk) + return false; /* tlist doesn't match junk status */ + + /* + * We accept either a Var referencing the corresponding element of the + * subplan tlist, or a Const equaling the subplan element. See + * generate_setop_tlist() for motivation. + */ + if (ptle->expr && IsA(ptle->expr, Var)) + { + Var *var = (Var *) ptle->expr; + + Assert(var->varno == plan->scan.scanrelid); + Assert(var->varlevelsup == 0); + if (var->varattno != attrno) + return false; /* out of order */ + } + else if (ptle->expr && IsA(ptle->expr, Const)) + { + if (!equal(ptle->expr, ctle->expr)) + return false; + } + else + return false; + + attrno++; + } + + /* Re-mark the SubqueryScan as deletable from the plan tree */ + plan->scanstatus = SUBQUERY_SCAN_TRIVIAL; + + return true; +} + +/* + * clean_up_removed_plan_level + * Do necessary cleanup when we strip out a SubqueryScan, Append, etc + * + * We are dropping the "parent" plan in favor of returning just its "child". + * A few small tweaks are needed. + */ +static Plan * +clean_up_removed_plan_level(Plan *parent, Plan *child) +{ + /* + * We have to be sure we don't lose any initplans, so move any that were + * attached to the parent plan to the child. If we do move any, the child + * is no longer parallel-safe. + */ + if (parent->initPlan) + child->parallel_safe = false; + + /* + * Attach plans this way so that parent's initplans are processed before + * any pre-existing initplans of the child. Probably doesn't matter, but + * let's preserve the ordering just in case. + */ + child->initPlan = list_concat(parent->initPlan, + child->initPlan); + + /* + * We also have to transfer the parent's column labeling info into the + * child, else columns sent to client will be improperly labeled if this + * is the topmost plan level. resjunk and so on may be important too. + */ + apply_tlist_labeling(child->targetlist, parent->targetlist); + + return child; +} + +/* + * set_foreignscan_references + * Do set_plan_references processing on a ForeignScan + */ +static void +set_foreignscan_references(PlannerInfo *root, + ForeignScan *fscan, + int rtoffset) +{ + /* Adjust scanrelid if it's valid */ + if (fscan->scan.scanrelid > 0) + fscan->scan.scanrelid += rtoffset; + + if (fscan->fdw_scan_tlist != NIL || fscan->scan.scanrelid == 0) + { + /* + * Adjust tlist, qual, fdw_exprs, fdw_recheck_quals to reference + * foreign scan tuple + */ + indexed_tlist *itlist = build_tlist_index(fscan->fdw_scan_tlist); + + fscan->scan.plan.targetlist = (List *) + fix_upper_expr(root, + (Node *) fscan->scan.plan.targetlist, + itlist, + INDEX_VAR, + rtoffset, + NUM_EXEC_TLIST((Plan *) fscan)); + fscan->scan.plan.qual = (List *) + fix_upper_expr(root, + (Node *) fscan->scan.plan.qual, + itlist, + INDEX_VAR, + rtoffset, + NUM_EXEC_QUAL((Plan *) fscan)); + fscan->fdw_exprs = (List *) + fix_upper_expr(root, + (Node *) fscan->fdw_exprs, + itlist, + INDEX_VAR, + rtoffset, + NUM_EXEC_QUAL((Plan *) fscan)); + fscan->fdw_recheck_quals = (List *) + fix_upper_expr(root, + (Node *) fscan->fdw_recheck_quals, + itlist, + INDEX_VAR, + rtoffset, + NUM_EXEC_QUAL((Plan *) fscan)); + pfree(itlist); + /* fdw_scan_tlist itself just needs fix_scan_list() adjustments */ + fscan->fdw_scan_tlist = + fix_scan_list(root, fscan->fdw_scan_tlist, + rtoffset, NUM_EXEC_TLIST((Plan *) fscan)); + } + else + { + /* + * Adjust tlist, qual, fdw_exprs, fdw_recheck_quals in the standard + * way + */ + fscan->scan.plan.targetlist = + fix_scan_list(root, fscan->scan.plan.targetlist, + rtoffset, NUM_EXEC_TLIST((Plan *) fscan)); + fscan->scan.plan.qual = + fix_scan_list(root, fscan->scan.plan.qual, + rtoffset, NUM_EXEC_QUAL((Plan *) fscan)); + fscan->fdw_exprs = + fix_scan_list(root, fscan->fdw_exprs, + rtoffset, NUM_EXEC_QUAL((Plan *) fscan)); + fscan->fdw_recheck_quals = + fix_scan_list(root, fscan->fdw_recheck_quals, + rtoffset, NUM_EXEC_QUAL((Plan *) fscan)); + } + + fscan->fs_relids = offset_relid_set(fscan->fs_relids, rtoffset); + + /* Adjust resultRelation if it's valid */ + if (fscan->resultRelation > 0) + fscan->resultRelation += rtoffset; +} + +/* + * set_customscan_references + * Do set_plan_references processing on a CustomScan + */ +static void +set_customscan_references(PlannerInfo *root, + CustomScan *cscan, + int rtoffset) +{ + ListCell *lc; + + /* Adjust scanrelid if it's valid */ + if (cscan->scan.scanrelid > 0) + cscan->scan.scanrelid += rtoffset; + + if (cscan->custom_scan_tlist != NIL || cscan->scan.scanrelid == 0) + { + /* Adjust tlist, qual, custom_exprs to reference custom scan tuple */ + indexed_tlist *itlist = build_tlist_index(cscan->custom_scan_tlist); + + cscan->scan.plan.targetlist = (List *) + fix_upper_expr(root, + (Node *) cscan->scan.plan.targetlist, + itlist, + INDEX_VAR, + rtoffset, + NUM_EXEC_TLIST((Plan *) cscan)); + cscan->scan.plan.qual = (List *) + fix_upper_expr(root, + (Node *) cscan->scan.plan.qual, + itlist, + INDEX_VAR, + rtoffset, + NUM_EXEC_QUAL((Plan *) cscan)); + cscan->custom_exprs = (List *) + fix_upper_expr(root, + (Node *) cscan->custom_exprs, + itlist, + INDEX_VAR, + rtoffset, + NUM_EXEC_QUAL((Plan *) cscan)); + pfree(itlist); + /* custom_scan_tlist itself just needs fix_scan_list() adjustments */ + cscan->custom_scan_tlist = + fix_scan_list(root, cscan->custom_scan_tlist, + rtoffset, NUM_EXEC_TLIST((Plan *) cscan)); + } + else + { + /* Adjust tlist, qual, custom_exprs in the standard way */ + cscan->scan.plan.targetlist = + fix_scan_list(root, cscan->scan.plan.targetlist, + rtoffset, NUM_EXEC_TLIST((Plan *) cscan)); + cscan->scan.plan.qual = + fix_scan_list(root, cscan->scan.plan.qual, + rtoffset, NUM_EXEC_QUAL((Plan *) cscan)); + cscan->custom_exprs = + fix_scan_list(root, cscan->custom_exprs, + rtoffset, NUM_EXEC_QUAL((Plan *) cscan)); + } + + /* Adjust child plan-nodes recursively, if needed */ + foreach(lc, cscan->custom_plans) + { + lfirst(lc) = set_plan_refs(root, (Plan *) lfirst(lc), rtoffset); + } + + cscan->custom_relids = offset_relid_set(cscan->custom_relids, rtoffset); +} + +/* + * set_append_references + * Do set_plan_references processing on an Append + * + * We try to strip out the Append entirely; if we can't, we have + * to do the normal processing on it. + */ +static Plan * +set_append_references(PlannerInfo *root, + Append *aplan, + int rtoffset) +{ + ListCell *l; + + /* + * Append, like Sort et al, doesn't actually evaluate its targetlist or + * check quals. If it's got exactly one child plan, then it's not doing + * anything useful at all, and we can strip it out. + */ + Assert(aplan->plan.qual == NIL); + + /* First, we gotta recurse on the children */ + foreach(l, aplan->appendplans) + { + lfirst(l) = set_plan_refs(root, (Plan *) lfirst(l), rtoffset); + } + + /* + * See if it's safe to get rid of the Append entirely. For this to be + * safe, there must be only one child plan and that child plan's parallel + * awareness must match that of the Append's. The reason for the latter + * is that the if the Append is parallel aware and the child is not then + * the calling plan may execute the non-parallel aware child multiple + * times. + */ + if (list_length(aplan->appendplans) == 1 && + ((Plan *) linitial(aplan->appendplans))->parallel_aware == aplan->plan.parallel_aware) + return clean_up_removed_plan_level((Plan *) aplan, + (Plan *) linitial(aplan->appendplans)); + + /* + * Otherwise, clean up the Append as needed. It's okay to do this after + * recursing to the children, because set_dummy_tlist_references doesn't + * look at those. + */ + set_dummy_tlist_references((Plan *) aplan, rtoffset); + + aplan->apprelids = offset_relid_set(aplan->apprelids, rtoffset); + + if (aplan->part_prune_info) + { + foreach(l, aplan->part_prune_info->prune_infos) + { + List *prune_infos = lfirst(l); + ListCell *l2; + + foreach(l2, prune_infos) + { + PartitionedRelPruneInfo *pinfo = lfirst(l2); + + pinfo->rtindex += rtoffset; + } + } + } + + /* We don't need to recurse to lefttree or righttree ... */ + Assert(aplan->plan.lefttree == NULL); + Assert(aplan->plan.righttree == NULL); + + return (Plan *) aplan; +} + +/* + * set_mergeappend_references + * Do set_plan_references processing on a MergeAppend + * + * We try to strip out the MergeAppend entirely; if we can't, we have + * to do the normal processing on it. + */ +static Plan * +set_mergeappend_references(PlannerInfo *root, + MergeAppend *mplan, + int rtoffset) +{ + ListCell *l; + + /* + * MergeAppend, like Sort et al, doesn't actually evaluate its targetlist + * or check quals. If it's got exactly one child plan, then it's not + * doing anything useful at all, and we can strip it out. + */ + Assert(mplan->plan.qual == NIL); + + /* First, we gotta recurse on the children */ + foreach(l, mplan->mergeplans) + { + lfirst(l) = set_plan_refs(root, (Plan *) lfirst(l), rtoffset); + } + + /* + * See if it's safe to get rid of the MergeAppend entirely. For this to + * be safe, there must be only one child plan and that child plan's + * parallel awareness must match that of the MergeAppend's. The reason + * for the latter is that the if the MergeAppend is parallel aware and the + * child is not then the calling plan may execute the non-parallel aware + * child multiple times. + */ + if (list_length(mplan->mergeplans) == 1 && + ((Plan *) linitial(mplan->mergeplans))->parallel_aware == mplan->plan.parallel_aware) + return clean_up_removed_plan_level((Plan *) mplan, + (Plan *) linitial(mplan->mergeplans)); + + /* + * Otherwise, clean up the MergeAppend as needed. It's okay to do this + * after recursing to the children, because set_dummy_tlist_references + * doesn't look at those. + */ + set_dummy_tlist_references((Plan *) mplan, rtoffset); + + mplan->apprelids = offset_relid_set(mplan->apprelids, rtoffset); + + if (mplan->part_prune_info) + { + foreach(l, mplan->part_prune_info->prune_infos) + { + List *prune_infos = lfirst(l); + ListCell *l2; + + foreach(l2, prune_infos) + { + PartitionedRelPruneInfo *pinfo = lfirst(l2); + + pinfo->rtindex += rtoffset; + } + } + } + + /* We don't need to recurse to lefttree or righttree ... */ + Assert(mplan->plan.lefttree == NULL); + Assert(mplan->plan.righttree == NULL); + + return (Plan *) mplan; +} + +/* + * set_hash_references + * Do set_plan_references processing on a Hash node + */ +static void +set_hash_references(PlannerInfo *root, Plan *plan, int rtoffset) +{ + Hash *hplan = (Hash *) plan; + Plan *outer_plan = plan->lefttree; + indexed_tlist *outer_itlist; + + /* + * Hash's hashkeys are used when feeding tuples into the hashtable, + * therefore have them reference Hash's outer plan (which itself is the + * inner plan of the HashJoin). + */ + outer_itlist = build_tlist_index(outer_plan->targetlist); + hplan->hashkeys = (List *) + fix_upper_expr(root, + (Node *) hplan->hashkeys, + outer_itlist, + OUTER_VAR, + rtoffset, + NUM_EXEC_QUAL(plan)); + + /* Hash doesn't project */ + set_dummy_tlist_references(plan, rtoffset); + + /* Hash nodes don't have their own quals */ + Assert(plan->qual == NIL); +} + +/* + * offset_relid_set + * Apply rtoffset to the members of a Relids set. + */ +static Relids +offset_relid_set(Relids relids, int rtoffset) +{ + Relids result = NULL; + int rtindex; + + /* If there's no offset to apply, we needn't recompute the value */ + if (rtoffset == 0) + return relids; + rtindex = -1; + while ((rtindex = bms_next_member(relids, rtindex)) >= 0) + result = bms_add_member(result, rtindex + rtoffset); + return result; +} + +/* + * copyVar + * Copy a Var node. + * + * fix_scan_expr and friends do this enough times that it's worth having + * a bespoke routine instead of using the generic copyObject() function. + */ +static inline Var * +copyVar(Var *var) +{ + Var *newvar = (Var *) palloc(sizeof(Var)); + + *newvar = *var; + return newvar; +} + +/* + * fix_expr_common + * Do generic set_plan_references processing on an expression node + * + * This is code that is common to all variants of expression-fixing. + * We must look up operator opcode info for OpExpr and related nodes, + * add OIDs from regclass Const nodes into root->glob->relationOids, and + * add PlanInvalItems for user-defined functions into root->glob->invalItems. + * We also fill in column index lists for GROUPING() expressions. + * + * We assume it's okay to update opcode info in-place. So this could possibly + * scribble on the planner's input data structures, but it's OK. + */ +static void +fix_expr_common(PlannerInfo *root, Node *node) +{ + /* We assume callers won't call us on a NULL pointer */ + if (IsA(node, Aggref)) + { + record_plan_function_dependency(root, + ((Aggref *) node)->aggfnoid); + } + else if (IsA(node, WindowFunc)) + { + record_plan_function_dependency(root, + ((WindowFunc *) node)->winfnoid); + } + else if (IsA(node, FuncExpr)) + { + record_plan_function_dependency(root, + ((FuncExpr *) node)->funcid); + } + else if (IsA(node, OpExpr)) + { + set_opfuncid((OpExpr *) node); + record_plan_function_dependency(root, + ((OpExpr *) node)->opfuncid); + } + else if (IsA(node, DistinctExpr)) + { + set_opfuncid((OpExpr *) node); /* rely on struct equivalence */ + record_plan_function_dependency(root, + ((DistinctExpr *) node)->opfuncid); + } + else if (IsA(node, NullIfExpr)) + { + set_opfuncid((OpExpr *) node); /* rely on struct equivalence */ + record_plan_function_dependency(root, + ((NullIfExpr *) node)->opfuncid); + } + else if (IsA(node, ScalarArrayOpExpr)) + { + ScalarArrayOpExpr *saop = (ScalarArrayOpExpr *) node; + + set_sa_opfuncid(saop); + record_plan_function_dependency(root, saop->opfuncid); + + if (OidIsValid(saop->hashfuncid)) + record_plan_function_dependency(root, saop->hashfuncid); + + if (OidIsValid(saop->negfuncid)) + record_plan_function_dependency(root, saop->negfuncid); + } + else if (IsA(node, Const)) + { + Const *con = (Const *) node; + + /* Check for regclass reference */ + if (ISREGCLASSCONST(con)) + root->glob->relationOids = + lappend_oid(root->glob->relationOids, + DatumGetObjectId(con->constvalue)); + } + else if (IsA(node, GroupingFunc)) + { + GroupingFunc *g = (GroupingFunc *) node; + AttrNumber *grouping_map = root->grouping_map; + + /* If there are no grouping sets, we don't need this. */ + + Assert(grouping_map || g->cols == NIL); + + if (grouping_map) + { + ListCell *lc; + List *cols = NIL; + + foreach(lc, g->refs) + { + cols = lappend_int(cols, grouping_map[lfirst_int(lc)]); + } + + Assert(!g->cols || equal(cols, g->cols)); + + if (!g->cols) + g->cols = cols; + } + } +} + +/* + * fix_param_node + * Do set_plan_references processing on a Param + * + * If it's a PARAM_MULTIEXPR, replace it with the appropriate Param from + * root->multiexpr_params; otherwise no change is needed. + * Just for paranoia's sake, we make a copy of the node in either case. + */ +static Node * +fix_param_node(PlannerInfo *root, Param *p) +{ + if (p->paramkind == PARAM_MULTIEXPR) + { + int subqueryid = p->paramid >> 16; + int colno = p->paramid & 0xFFFF; + List *params; + + if (subqueryid <= 0 || + subqueryid > list_length(root->multiexpr_params)) + elog(ERROR, "unexpected PARAM_MULTIEXPR ID: %d", p->paramid); + params = (List *) list_nth(root->multiexpr_params, subqueryid - 1); + if (colno <= 0 || colno > list_length(params)) + elog(ERROR, "unexpected PARAM_MULTIEXPR ID: %d", p->paramid); + return copyObject(list_nth(params, colno - 1)); + } + return (Node *) copyObject(p); +} + +/* + * fix_alternative_subplan + * Do set_plan_references processing on an AlternativeSubPlan + * + * Choose one of the alternative implementations and return just that one, + * discarding the rest of the AlternativeSubPlan structure. + * Note: caller must still recurse into the result! + * + * We don't make any attempt to fix up cost estimates in the parent plan + * node or higher-level nodes. + */ +static Node * +fix_alternative_subplan(PlannerInfo *root, AlternativeSubPlan *asplan, + double num_exec) +{ + SubPlan *bestplan = NULL; + Cost bestcost = 0; + ListCell *lc; + + /* + * Compute the estimated cost of each subplan assuming num_exec + * executions, and keep the cheapest one. In event of exact equality of + * estimates, we prefer the later plan; this is a bit arbitrary, but in + * current usage it biases us to break ties against fast-start subplans. + */ + Assert(asplan->subplans != NIL); + + foreach(lc, asplan->subplans) + { + SubPlan *curplan = (SubPlan *) lfirst(lc); + Cost curcost; + + curcost = curplan->startup_cost + num_exec * curplan->per_call_cost; + if (bestplan == NULL || curcost <= bestcost) + { + bestplan = curplan; + bestcost = curcost; + } + + /* Also mark all subplans that are in AlternativeSubPlans */ + root->isAltSubplan[curplan->plan_id - 1] = true; + } + + /* Mark the subplan we selected */ + root->isUsedSubplan[bestplan->plan_id - 1] = true; + + return (Node *) bestplan; +} + +/* + * fix_scan_expr + * Do set_plan_references processing on a scan-level expression + * + * This consists of incrementing all Vars' varnos by rtoffset, + * replacing PARAM_MULTIEXPR Params, expanding PlaceHolderVars, + * replacing Aggref nodes that should be replaced by initplan output Params, + * choosing the best implementation for AlternativeSubPlans, + * looking up operator opcode info for OpExpr and related nodes, + * and adding OIDs from regclass Const nodes into root->glob->relationOids. + * + * 'node': the expression to be modified + * 'rtoffset': how much to increment varnos by + * 'num_exec': estimated number of executions of expression + * + * The expression tree is either copied-and-modified, or modified in-place + * if that seems safe. + */ +static Node * +fix_scan_expr(PlannerInfo *root, Node *node, int rtoffset, double num_exec) +{ + fix_scan_expr_context context; + + context.root = root; + context.rtoffset = rtoffset; + context.num_exec = num_exec; + + if (rtoffset != 0 || + root->multiexpr_params != NIL || + root->glob->lastPHId != 0 || + root->minmax_aggs != NIL || + root->hasAlternativeSubPlans) + { + return fix_scan_expr_mutator(node, &context); + } + else + { + /* + * If rtoffset == 0, we don't need to change any Vars, and if there + * are no MULTIEXPR subqueries then we don't need to replace + * PARAM_MULTIEXPR Params, and if there are no placeholders anywhere + * we won't need to remove them, and if there are no minmax Aggrefs we + * won't need to replace them, and if there are no AlternativeSubPlans + * we won't need to remove them. Then it's OK to just scribble on the + * input node tree instead of copying (since the only change, filling + * in any unset opfuncid fields, is harmless). This saves just enough + * cycles to be noticeable on trivial queries. + */ + (void) fix_scan_expr_walker(node, &context); + return node; + } +} + +static Node * +fix_scan_expr_mutator(Node *node, fix_scan_expr_context *context) +{ + if (node == NULL) + return NULL; + if (IsA(node, Var)) + { + Var *var = copyVar((Var *) node); + + Assert(var->varlevelsup == 0); + + /* + * We should not see Vars marked INNER_VAR, OUTER_VAR, or ROWID_VAR. + * But an indexqual expression could contain INDEX_VAR Vars. + */ + Assert(var->varno != INNER_VAR); + Assert(var->varno != OUTER_VAR); + Assert(var->varno != ROWID_VAR); + if (!IS_SPECIAL_VARNO(var->varno)) + var->varno += context->rtoffset; + if (var->varnosyn > 0) + var->varnosyn += context->rtoffset; + return (Node *) var; + } + if (IsA(node, Param)) + return fix_param_node(context->root, (Param *) node); + if (IsA(node, Aggref)) + { + Aggref *aggref = (Aggref *) node; + + /* See if the Aggref should be replaced by a Param */ + if (context->root->minmax_aggs != NIL && + list_length(aggref->args) == 1) + { + TargetEntry *curTarget = (TargetEntry *) linitial(aggref->args); + ListCell *lc; + + foreach(lc, context->root->minmax_aggs) + { + MinMaxAggInfo *mminfo = (MinMaxAggInfo *) lfirst(lc); + + if (mminfo->aggfnoid == aggref->aggfnoid && + equal(mminfo->target, curTarget->expr)) + return (Node *) copyObject(mminfo->param); + } + } + /* If no match, just fall through to process it normally */ + } + if (IsA(node, CurrentOfExpr)) + { + CurrentOfExpr *cexpr = (CurrentOfExpr *) copyObject(node); + + Assert(!IS_SPECIAL_VARNO(cexpr->cvarno)); + cexpr->cvarno += context->rtoffset; + return (Node *) cexpr; + } + if (IsA(node, PlaceHolderVar)) + { + /* At scan level, we should always just evaluate the contained expr */ + PlaceHolderVar *phv = (PlaceHolderVar *) node; + + return fix_scan_expr_mutator((Node *) phv->phexpr, context); + } + if (IsA(node, AlternativeSubPlan)) + return fix_scan_expr_mutator(fix_alternative_subplan(context->root, + (AlternativeSubPlan *) node, + context->num_exec), + context); + fix_expr_common(context->root, node); + return expression_tree_mutator(node, fix_scan_expr_mutator, + (void *) context); +} + +static bool +fix_scan_expr_walker(Node *node, fix_scan_expr_context *context) +{ + if (node == NULL) + return false; + Assert(!(IsA(node, Var) && ((Var *) node)->varno == ROWID_VAR)); + Assert(!IsA(node, PlaceHolderVar)); + Assert(!IsA(node, AlternativeSubPlan)); + fix_expr_common(context->root, node); + return expression_tree_walker(node, fix_scan_expr_walker, + (void *) context); +} + +/* + * set_join_references + * Modify the target list and quals of a join node to reference its + * subplans, by setting the varnos to OUTER_VAR or INNER_VAR and setting + * attno values to the result domain number of either the corresponding + * outer or inner join tuple item. Also perform opcode lookup for these + * expressions, and add regclass OIDs to root->glob->relationOids. + */ +static void +set_join_references(PlannerInfo *root, Join *join, int rtoffset) +{ + Plan *outer_plan = join->plan.lefttree; + Plan *inner_plan = join->plan.righttree; + indexed_tlist *outer_itlist; + indexed_tlist *inner_itlist; + + outer_itlist = build_tlist_index(outer_plan->targetlist); + inner_itlist = build_tlist_index(inner_plan->targetlist); + + /* + * First process the joinquals (including merge or hash clauses). These + * are logically below the join so they can always use all values + * available from the input tlists. It's okay to also handle + * NestLoopParams now, because those couldn't refer to nullable + * subexpressions. + */ + join->joinqual = fix_join_expr(root, + join->joinqual, + outer_itlist, + inner_itlist, + (Index) 0, + rtoffset, + NUM_EXEC_QUAL((Plan *) join)); + + /* Now do join-type-specific stuff */ + if (IsA(join, NestLoop)) + { + NestLoop *nl = (NestLoop *) join; + ListCell *lc; + + foreach(lc, nl->nestParams) + { + NestLoopParam *nlp = (NestLoopParam *) lfirst(lc); + + nlp->paramval = (Var *) fix_upper_expr(root, + (Node *) nlp->paramval, + outer_itlist, + OUTER_VAR, + rtoffset, + NUM_EXEC_TLIST(outer_plan)); + /* Check we replaced any PlaceHolderVar with simple Var */ + if (!(IsA(nlp->paramval, Var) && + nlp->paramval->varno == OUTER_VAR)) + elog(ERROR, "NestLoopParam was not reduced to a simple Var"); + } + } + else if (IsA(join, MergeJoin)) + { + MergeJoin *mj = (MergeJoin *) join; + + mj->mergeclauses = fix_join_expr(root, + mj->mergeclauses, + outer_itlist, + inner_itlist, + (Index) 0, + rtoffset, + NUM_EXEC_QUAL((Plan *) join)); + } + else if (IsA(join, HashJoin)) + { + HashJoin *hj = (HashJoin *) join; + + hj->hashclauses = fix_join_expr(root, + hj->hashclauses, + outer_itlist, + inner_itlist, + (Index) 0, + rtoffset, + NUM_EXEC_QUAL((Plan *) join)); + + /* + * HashJoin's hashkeys are used to look for matching tuples from its + * outer plan (not the Hash node!) in the hashtable. + */ + hj->hashkeys = (List *) fix_upper_expr(root, + (Node *) hj->hashkeys, + outer_itlist, + OUTER_VAR, + rtoffset, + NUM_EXEC_QUAL((Plan *) join)); + } + + /* + * Now we need to fix up the targetlist and qpqual, which are logically + * above the join. This means they should not re-use any input expression + * that was computed in the nullable side of an outer join. Vars and + * PlaceHolderVars are fine, so we can implement this restriction just by + * clearing has_non_vars in the indexed_tlist structs. + * + * XXX This is a grotty workaround for the fact that we don't clearly + * distinguish between a Var appearing below an outer join and the "same" + * Var appearing above it. If we did, we'd not need to hack the matching + * rules this way. + */ + switch (join->jointype) + { + case JOIN_LEFT: + case JOIN_SEMI: + case JOIN_ANTI: + inner_itlist->has_non_vars = false; + break; + case JOIN_RIGHT: + outer_itlist->has_non_vars = false; + break; + case JOIN_FULL: + outer_itlist->has_non_vars = false; + inner_itlist->has_non_vars = false; + break; + default: + break; + } + + join->plan.targetlist = fix_join_expr(root, + join->plan.targetlist, + outer_itlist, + inner_itlist, + (Index) 0, + rtoffset, + NUM_EXEC_TLIST((Plan *) join)); + join->plan.qual = fix_join_expr(root, + join->plan.qual, + outer_itlist, + inner_itlist, + (Index) 0, + rtoffset, + NUM_EXEC_QUAL((Plan *) join)); + + pfree(outer_itlist); + pfree(inner_itlist); +} + +/* + * set_upper_references + * Update the targetlist and quals of an upper-level plan node + * to refer to the tuples returned by its lefttree subplan. + * Also perform opcode lookup for these expressions, and + * add regclass OIDs to root->glob->relationOids. + * + * This is used for single-input plan types like Agg, Group, Result. + * + * In most cases, we have to match up individual Vars in the tlist and + * qual expressions with elements of the subplan's tlist (which was + * generated by flattening these selfsame expressions, so it should have all + * the required variables). There is an important exception, however: + * depending on where we are in the plan tree, sort/group columns may have + * been pushed into the subplan tlist unflattened. If these values are also + * needed in the output then we want to reference the subplan tlist element + * rather than recomputing the expression. + */ +static void +set_upper_references(PlannerInfo *root, Plan *plan, int rtoffset) +{ + Plan *subplan = plan->lefttree; + indexed_tlist *subplan_itlist; + List *output_targetlist; + ListCell *l; + + subplan_itlist = build_tlist_index(subplan->targetlist); + + output_targetlist = NIL; + foreach(l, plan->targetlist) + { + TargetEntry *tle = (TargetEntry *) lfirst(l); + Node *newexpr; + + /* If it's a sort/group item, first try to match by sortref */ + if (tle->ressortgroupref != 0) + { + newexpr = (Node *) + search_indexed_tlist_for_sortgroupref(tle->expr, + tle->ressortgroupref, + subplan_itlist, + OUTER_VAR); + if (!newexpr) + newexpr = fix_upper_expr(root, + (Node *) tle->expr, + subplan_itlist, + OUTER_VAR, + rtoffset, + NUM_EXEC_TLIST(plan)); + } + else + newexpr = fix_upper_expr(root, + (Node *) tle->expr, + subplan_itlist, + OUTER_VAR, + rtoffset, + NUM_EXEC_TLIST(plan)); + tle = flatCopyTargetEntry(tle); + tle->expr = (Expr *) newexpr; + output_targetlist = lappend(output_targetlist, tle); + } + plan->targetlist = output_targetlist; + + plan->qual = (List *) + fix_upper_expr(root, + (Node *) plan->qual, + subplan_itlist, + OUTER_VAR, + rtoffset, + NUM_EXEC_QUAL(plan)); + + pfree(subplan_itlist); +} + +/* + * set_param_references + * Initialize the initParam list in Gather or Gather merge node such that + * it contains reference of all the params that needs to be evaluated + * before execution of the node. It contains the initplan params that are + * being passed to the plan nodes below it. + */ +static void +set_param_references(PlannerInfo *root, Plan *plan) +{ + Assert(IsA(plan, Gather) || IsA(plan, GatherMerge)); + + if (plan->lefttree->extParam) + { + PlannerInfo *proot; + Bitmapset *initSetParam = NULL; + ListCell *l; + + for (proot = root; proot != NULL; proot = proot->parent_root) + { + foreach(l, proot->init_plans) + { + SubPlan *initsubplan = (SubPlan *) lfirst(l); + ListCell *l2; + + foreach(l2, initsubplan->setParam) + { + initSetParam = bms_add_member(initSetParam, lfirst_int(l2)); + } + } + } + + /* + * Remember the list of all external initplan params that are used by + * the children of Gather or Gather merge node. + */ + if (IsA(plan, Gather)) + ((Gather *) plan)->initParam = + bms_intersect(plan->lefttree->extParam, initSetParam); + else + ((GatherMerge *) plan)->initParam = + bms_intersect(plan->lefttree->extParam, initSetParam); + } +} + +/* + * Recursively scan an expression tree and convert Aggrefs to the proper + * intermediate form for combining aggregates. This means (1) replacing each + * one's argument list with a single argument that is the original Aggref + * modified to show partial aggregation and (2) changing the upper Aggref to + * show combining aggregation. + * + * After this step, set_upper_references will replace the partial Aggrefs + * with Vars referencing the lower Agg plan node's outputs, so that the final + * form seen by the executor is a combining Aggref with a Var as input. + * + * It's rather messy to postpone this step until setrefs.c; ideally it'd be + * done in createplan.c. The difficulty is that once we modify the Aggref + * expressions, they will no longer be equal() to their original form and + * so cross-plan-node-level matches will fail. So this has to happen after + * the plan node above the Agg has resolved its subplan references. + */ +static Node * +convert_combining_aggrefs(Node *node, void *context) +{ + if (node == NULL) + return NULL; + if (IsA(node, Aggref)) + { + Aggref *orig_agg = (Aggref *) node; + Aggref *child_agg; + Aggref *parent_agg; + + /* Assert we've not chosen to partial-ize any unsupported cases */ + Assert(orig_agg->aggorder == NIL); + Assert(orig_agg->aggdistinct == NIL); + + /* + * Since aggregate calls can't be nested, we needn't recurse into the + * arguments. But for safety, flat-copy the Aggref node itself rather + * than modifying it in-place. + */ + child_agg = makeNode(Aggref); + memcpy(child_agg, orig_agg, sizeof(Aggref)); + + /* + * For the parent Aggref, we want to copy all the fields of the + * original aggregate *except* the args list, which we'll replace + * below, and the aggfilter expression, which should be applied only + * by the child not the parent. Rather than explicitly knowing about + * all the other fields here, we can momentarily modify child_agg to + * provide a suitable source for copyObject. + */ + child_agg->args = NIL; + child_agg->aggfilter = NULL; + parent_agg = copyObject(child_agg); + child_agg->args = orig_agg->args; + child_agg->aggfilter = orig_agg->aggfilter; + + /* + * Now, set up child_agg to represent the first phase of partial + * aggregation. For now, assume serialization is required. + */ + mark_partial_aggref(child_agg, AGGSPLIT_INITIAL_SERIAL); + + /* + * And set up parent_agg to represent the second phase. + */ + parent_agg->args = list_make1(makeTargetEntry((Expr *) child_agg, + 1, NULL, false)); + mark_partial_aggref(parent_agg, AGGSPLIT_FINAL_DESERIAL); + + return (Node *) parent_agg; + } + return expression_tree_mutator(node, convert_combining_aggrefs, + (void *) context); +} + +/* + * set_dummy_tlist_references + * Replace the targetlist of an upper-level plan node with a simple + * list of OUTER_VAR references to its child. + * + * This is used for plan types like Sort and Append that don't evaluate + * their targetlists. Although the executor doesn't care at all what's in + * the tlist, EXPLAIN needs it to be realistic. + * + * Note: we could almost use set_upper_references() here, but it fails for + * Append for lack of a lefttree subplan. Single-purpose code is faster + * anyway. + */ +static void +set_dummy_tlist_references(Plan *plan, int rtoffset) +{ + List *output_targetlist; + ListCell *l; + + output_targetlist = NIL; + foreach(l, plan->targetlist) + { + TargetEntry *tle = (TargetEntry *) lfirst(l); + Var *oldvar = (Var *) tle->expr; + Var *newvar; + + /* + * As in search_indexed_tlist_for_non_var(), we prefer to keep Consts + * as Consts, not Vars referencing Consts. Here, there's no speed + * advantage to be had, but it makes EXPLAIN output look cleaner, and + * again it avoids confusing the executor. + */ + if (IsA(oldvar, Const)) + { + /* just reuse the existing TLE node */ + output_targetlist = lappend(output_targetlist, tle); + continue; + } + + newvar = makeVar(OUTER_VAR, + tle->resno, + exprType((Node *) oldvar), + exprTypmod((Node *) oldvar), + exprCollation((Node *) oldvar), + 0); + if (IsA(oldvar, Var) && + oldvar->varnosyn > 0) + { + newvar->varnosyn = oldvar->varnosyn + rtoffset; + newvar->varattnosyn = oldvar->varattnosyn; + } + else + { + newvar->varnosyn = 0; /* wasn't ever a plain Var */ + newvar->varattnosyn = 0; + } + + tle = flatCopyTargetEntry(tle); + tle->expr = (Expr *) newvar; + output_targetlist = lappend(output_targetlist, tle); + } + plan->targetlist = output_targetlist; + + /* We don't touch plan->qual here */ +} + + +/* + * build_tlist_index --- build an index data structure for a child tlist + * + * In most cases, subplan tlists will be "flat" tlists with only Vars, + * so we try to optimize that case by extracting information about Vars + * in advance. Matching a parent tlist to a child is still an O(N^2) + * operation, but at least with a much smaller constant factor than plain + * tlist_member() searches. + * + * The result of this function is an indexed_tlist struct to pass to + * search_indexed_tlist_for_var() or search_indexed_tlist_for_non_var(). + * When done, the indexed_tlist may be freed with a single pfree(). + */ +static indexed_tlist * +build_tlist_index(List *tlist) +{ + indexed_tlist *itlist; + tlist_vinfo *vinfo; + ListCell *l; + + /* Create data structure with enough slots for all tlist entries */ + itlist = (indexed_tlist *) + palloc(offsetof(indexed_tlist, vars) + + list_length(tlist) * sizeof(tlist_vinfo)); + + itlist->tlist = tlist; + itlist->has_ph_vars = false; + itlist->has_non_vars = false; + + /* Find the Vars and fill in the index array */ + vinfo = itlist->vars; + foreach(l, tlist) + { + TargetEntry *tle = (TargetEntry *) lfirst(l); + + if (tle->expr && IsA(tle->expr, Var)) + { + Var *var = (Var *) tle->expr; + + vinfo->varno = var->varno; + vinfo->varattno = var->varattno; + vinfo->resno = tle->resno; + vinfo++; + } + else if (tle->expr && IsA(tle->expr, PlaceHolderVar)) + itlist->has_ph_vars = true; + else + itlist->has_non_vars = true; + } + + itlist->num_vars = (vinfo - itlist->vars); + + return itlist; +} + +/* + * build_tlist_index_other_vars --- build a restricted tlist index + * + * This is like build_tlist_index, but we only index tlist entries that + * are Vars belonging to some rel other than the one specified. We will set + * has_ph_vars (allowing PlaceHolderVars to be matched), but not has_non_vars + * (so nothing other than Vars and PlaceHolderVars can be matched). + */ +static indexed_tlist * +build_tlist_index_other_vars(List *tlist, int ignore_rel) +{ + indexed_tlist *itlist; + tlist_vinfo *vinfo; + ListCell *l; + + /* Create data structure with enough slots for all tlist entries */ + itlist = (indexed_tlist *) + palloc(offsetof(indexed_tlist, vars) + + list_length(tlist) * sizeof(tlist_vinfo)); + + itlist->tlist = tlist; + itlist->has_ph_vars = false; + itlist->has_non_vars = false; + + /* Find the desired Vars and fill in the index array */ + vinfo = itlist->vars; + foreach(l, tlist) + { + TargetEntry *tle = (TargetEntry *) lfirst(l); + + if (tle->expr && IsA(tle->expr, Var)) + { + Var *var = (Var *) tle->expr; + + if (var->varno != ignore_rel) + { + vinfo->varno = var->varno; + vinfo->varattno = var->varattno; + vinfo->resno = tle->resno; + vinfo++; + } + } + else if (tle->expr && IsA(tle->expr, PlaceHolderVar)) + itlist->has_ph_vars = true; + } + + itlist->num_vars = (vinfo - itlist->vars); + + return itlist; +} + +/* + * search_indexed_tlist_for_var --- find a Var in an indexed tlist + * + * If a match is found, return a copy of the given Var with suitably + * modified varno/varattno (to wit, newvarno and the resno of the TLE entry). + * Also ensure that varnosyn is incremented by rtoffset. + * If no match, return NULL. + */ +static Var * +search_indexed_tlist_for_var(Var *var, indexed_tlist *itlist, + int newvarno, int rtoffset) +{ + int varno = var->varno; + AttrNumber varattno = var->varattno; + tlist_vinfo *vinfo; + int i; + + vinfo = itlist->vars; + i = itlist->num_vars; + while (i-- > 0) + { + if (vinfo->varno == varno && vinfo->varattno == varattno) + { + /* Found a match */ + Var *newvar = copyVar(var); + + newvar->varno = newvarno; + newvar->varattno = vinfo->resno; + if (newvar->varnosyn > 0) + newvar->varnosyn += rtoffset; + return newvar; + } + vinfo++; + } + return NULL; /* no match */ +} + +/* + * search_indexed_tlist_for_non_var --- find a non-Var in an indexed tlist + * + * If a match is found, return a Var constructed to reference the tlist item. + * If no match, return NULL. + * + * NOTE: it is a waste of time to call this unless itlist->has_ph_vars or + * itlist->has_non_vars. Furthermore, set_join_references() relies on being + * able to prevent matching of non-Vars by clearing itlist->has_non_vars, + * so there's a correctness reason not to call it unless that's set. + */ +static Var * +search_indexed_tlist_for_non_var(Expr *node, + indexed_tlist *itlist, int newvarno) +{ + TargetEntry *tle; + + /* + * If it's a simple Const, replacing it with a Var is silly, even if there + * happens to be an identical Const below; a Var is more expensive to + * execute than a Const. What's more, replacing it could confuse some + * places in the executor that expect to see simple Consts for, eg, + * dropped columns. + */ + if (IsA(node, Const)) + return NULL; + + tle = tlist_member(node, itlist->tlist); + if (tle) + { + /* Found a matching subplan output expression */ + Var *newvar; + + newvar = makeVarFromTargetEntry(newvarno, tle); + newvar->varnosyn = 0; /* wasn't ever a plain Var */ + newvar->varattnosyn = 0; + return newvar; + } + return NULL; /* no match */ +} + +/* + * search_indexed_tlist_for_sortgroupref --- find a sort/group expression + * + * If a match is found, return a Var constructed to reference the tlist item. + * If no match, return NULL. + * + * This is needed to ensure that we select the right subplan TLE in cases + * where there are multiple textually-equal()-but-volatile sort expressions. + * And it's also faster than search_indexed_tlist_for_non_var. + */ +static Var * +search_indexed_tlist_for_sortgroupref(Expr *node, + Index sortgroupref, + indexed_tlist *itlist, + int newvarno) +{ + ListCell *lc; + + foreach(lc, itlist->tlist) + { + TargetEntry *tle = (TargetEntry *) lfirst(lc); + + /* The equal() check should be redundant, but let's be paranoid */ + if (tle->ressortgroupref == sortgroupref && + equal(node, tle->expr)) + { + /* Found a matching subplan output expression */ + Var *newvar; + + newvar = makeVarFromTargetEntry(newvarno, tle); + newvar->varnosyn = 0; /* wasn't ever a plain Var */ + newvar->varattnosyn = 0; + return newvar; + } + } + return NULL; /* no match */ +} + +/* + * fix_join_expr + * Create a new set of targetlist entries or join qual clauses by + * changing the varno/varattno values of variables in the clauses + * to reference target list values from the outer and inner join + * relation target lists. Also perform opcode lookup and add + * regclass OIDs to root->glob->relationOids. + * + * This is used in four different scenarios: + * 1) a normal join clause, where all the Vars in the clause *must* be + * replaced by OUTER_VAR or INNER_VAR references. In this case + * acceptable_rel should be zero so that any failure to match a Var will be + * reported as an error. + * 2) RETURNING clauses, which may contain both Vars of the target relation + * and Vars of other relations. In this case we want to replace the + * other-relation Vars by OUTER_VAR references, while leaving target Vars + * alone. Thus inner_itlist = NULL and acceptable_rel = the ID of the + * target relation should be passed. + * 3) ON CONFLICT UPDATE SET/WHERE clauses. Here references to EXCLUDED are + * to be replaced with INNER_VAR references, while leaving target Vars (the + * to-be-updated relation) alone. Correspondingly inner_itlist is to be + * EXCLUDED elements, outer_itlist = NULL and acceptable_rel the target + * relation. + * 4) MERGE. In this case, references to the source relation are to be + * replaced with INNER_VAR references, leaving Vars of the target + * relation (the to-be-modified relation) alone. So inner_itlist is to be + * the source relation elements, outer_itlist = NULL and acceptable_rel + * the target relation. + * + * 'clauses' is the targetlist or list of join clauses + * 'outer_itlist' is the indexed target list of the outer join relation, + * or NULL + * 'inner_itlist' is the indexed target list of the inner join relation, + * or NULL + * 'acceptable_rel' is either zero or the rangetable index of a relation + * whose Vars may appear in the clause without provoking an error + * 'rtoffset': how much to increment varnos by + * 'num_exec': estimated number of executions of expression + * + * Returns the new expression tree. The original clause structure is + * not modified. + */ +static List * +fix_join_expr(PlannerInfo *root, + List *clauses, + indexed_tlist *outer_itlist, + indexed_tlist *inner_itlist, + Index acceptable_rel, + int rtoffset, + double num_exec) +{ + fix_join_expr_context context; + + context.root = root; + context.outer_itlist = outer_itlist; + context.inner_itlist = inner_itlist; + context.acceptable_rel = acceptable_rel; + context.rtoffset = rtoffset; + context.num_exec = num_exec; + return (List *) fix_join_expr_mutator((Node *) clauses, &context); +} + +static Node * +fix_join_expr_mutator(Node *node, fix_join_expr_context *context) +{ + Var *newvar; + + if (node == NULL) + return NULL; + if (IsA(node, Var)) + { + Var *var = (Var *) node; + + /* Look for the var in the input tlists, first in the outer */ + if (context->outer_itlist) + { + newvar = search_indexed_tlist_for_var(var, + context->outer_itlist, + OUTER_VAR, + context->rtoffset); + if (newvar) + return (Node *) newvar; + } + + /* then in the inner. */ + if (context->inner_itlist) + { + newvar = search_indexed_tlist_for_var(var, + context->inner_itlist, + INNER_VAR, + context->rtoffset); + if (newvar) + return (Node *) newvar; + } + + /* If it's for acceptable_rel, adjust and return it */ + if (var->varno == context->acceptable_rel) + { + var = copyVar(var); + var->varno += context->rtoffset; + if (var->varnosyn > 0) + var->varnosyn += context->rtoffset; + return (Node *) var; + } + + /* No referent found for Var */ + elog(ERROR, "variable not found in subplan target lists"); + } + if (IsA(node, PlaceHolderVar)) + { + PlaceHolderVar *phv = (PlaceHolderVar *) node; + + /* See if the PlaceHolderVar has bubbled up from a lower plan node */ + if (context->outer_itlist && context->outer_itlist->has_ph_vars) + { + newvar = search_indexed_tlist_for_non_var((Expr *) phv, + context->outer_itlist, + OUTER_VAR); + if (newvar) + return (Node *) newvar; + } + if (context->inner_itlist && context->inner_itlist->has_ph_vars) + { + newvar = search_indexed_tlist_for_non_var((Expr *) phv, + context->inner_itlist, + INNER_VAR); + if (newvar) + return (Node *) newvar; + } + + /* If not supplied by input plans, evaluate the contained expr */ + return fix_join_expr_mutator((Node *) phv->phexpr, context); + } + /* Try matching more complex expressions too, if tlists have any */ + if (context->outer_itlist && context->outer_itlist->has_non_vars) + { + newvar = search_indexed_tlist_for_non_var((Expr *) node, + context->outer_itlist, + OUTER_VAR); + if (newvar) + return (Node *) newvar; + } + if (context->inner_itlist && context->inner_itlist->has_non_vars) + { + newvar = search_indexed_tlist_for_non_var((Expr *) node, + context->inner_itlist, + INNER_VAR); + if (newvar) + return (Node *) newvar; + } + /* Special cases (apply only AFTER failing to match to lower tlist) */ + if (IsA(node, Param)) + return fix_param_node(context->root, (Param *) node); + if (IsA(node, AlternativeSubPlan)) + return fix_join_expr_mutator(fix_alternative_subplan(context->root, + (AlternativeSubPlan *) node, + context->num_exec), + context); + fix_expr_common(context->root, node); + return expression_tree_mutator(node, + fix_join_expr_mutator, + (void *) context); +} + +/* + * fix_upper_expr + * Modifies an expression tree so that all Var nodes reference outputs + * of a subplan. Also looks for Aggref nodes that should be replaced + * by initplan output Params. Also performs opcode lookup, and adds + * regclass OIDs to root->glob->relationOids. + * + * This is used to fix up target and qual expressions of non-join upper-level + * plan nodes, as well as index-only scan nodes. + * + * An error is raised if no matching var can be found in the subplan tlist + * --- so this routine should only be applied to nodes whose subplans' + * targetlists were generated by flattening the expressions used in the + * parent node. + * + * If itlist->has_non_vars is true, then we try to match whole subexpressions + * against elements of the subplan tlist, so that we can avoid recomputing + * expressions that were already computed by the subplan. (This is relatively + * expensive, so we don't want to try it in the common case where the + * subplan tlist is just a flattened list of Vars.) + * + * 'node': the tree to be fixed (a target item or qual) + * 'subplan_itlist': indexed target list for subplan (or index) + * 'newvarno': varno to use for Vars referencing tlist elements + * 'rtoffset': how much to increment varnos by + * 'num_exec': estimated number of executions of expression + * + * The resulting tree is a copy of the original in which all Var nodes have + * varno = newvarno, varattno = resno of corresponding targetlist element. + * The original tree is not modified. + */ +static Node * +fix_upper_expr(PlannerInfo *root, + Node *node, + indexed_tlist *subplan_itlist, + int newvarno, + int rtoffset, + double num_exec) +{ + fix_upper_expr_context context; + + context.root = root; + context.subplan_itlist = subplan_itlist; + context.newvarno = newvarno; + context.rtoffset = rtoffset; + context.num_exec = num_exec; + return fix_upper_expr_mutator(node, &context); +} + +static Node * +fix_upper_expr_mutator(Node *node, fix_upper_expr_context *context) +{ + Var *newvar; + + if (node == NULL) + return NULL; + if (IsA(node, Var)) + { + Var *var = (Var *) node; + + newvar = search_indexed_tlist_for_var(var, + context->subplan_itlist, + context->newvarno, + context->rtoffset); + if (!newvar) + elog(ERROR, "variable not found in subplan target list"); + return (Node *) newvar; + } + if (IsA(node, PlaceHolderVar)) + { + PlaceHolderVar *phv = (PlaceHolderVar *) node; + + /* See if the PlaceHolderVar has bubbled up from a lower plan node */ + if (context->subplan_itlist->has_ph_vars) + { + newvar = search_indexed_tlist_for_non_var((Expr *) phv, + context->subplan_itlist, + context->newvarno); + if (newvar) + return (Node *) newvar; + } + /* If not supplied by input plan, evaluate the contained expr */ + return fix_upper_expr_mutator((Node *) phv->phexpr, context); + } + /* Try matching more complex expressions too, if tlist has any */ + if (context->subplan_itlist->has_non_vars) + { + newvar = search_indexed_tlist_for_non_var((Expr *) node, + context->subplan_itlist, + context->newvarno); + if (newvar) + return (Node *) newvar; + } + /* Special cases (apply only AFTER failing to match to lower tlist) */ + if (IsA(node, Param)) + return fix_param_node(context->root, (Param *) node); + if (IsA(node, Aggref)) + { + Aggref *aggref = (Aggref *) node; + + /* See if the Aggref should be replaced by a Param */ + if (context->root->minmax_aggs != NIL && + list_length(aggref->args) == 1) + { + TargetEntry *curTarget = (TargetEntry *) linitial(aggref->args); + ListCell *lc; + + foreach(lc, context->root->minmax_aggs) + { + MinMaxAggInfo *mminfo = (MinMaxAggInfo *) lfirst(lc); + + if (mminfo->aggfnoid == aggref->aggfnoid && + equal(mminfo->target, curTarget->expr)) + return (Node *) copyObject(mminfo->param); + } + } + /* If no match, just fall through to process it normally */ + } + if (IsA(node, AlternativeSubPlan)) + return fix_upper_expr_mutator(fix_alternative_subplan(context->root, + (AlternativeSubPlan *) node, + context->num_exec), + context); + fix_expr_common(context->root, node); + return expression_tree_mutator(node, + fix_upper_expr_mutator, + (void *) context); +} + +/* + * set_returning_clause_references + * Perform setrefs.c's work on a RETURNING targetlist + * + * If the query involves more than just the result table, we have to + * adjust any Vars that refer to other tables to reference junk tlist + * entries in the top subplan's targetlist. Vars referencing the result + * table should be left alone, however (the executor will evaluate them + * using the actual heap tuple, after firing triggers if any). In the + * adjusted RETURNING list, result-table Vars will have their original + * varno (plus rtoffset), but Vars for other rels will have varno OUTER_VAR. + * + * We also must perform opcode lookup and add regclass OIDs to + * root->glob->relationOids. + * + * 'rlist': the RETURNING targetlist to be fixed + * 'topplan': the top subplan node that will be just below the ModifyTable + * node (note it's not yet passed through set_plan_refs) + * 'resultRelation': RT index of the associated result relation + * 'rtoffset': how much to increment varnos by + * + * Note: the given 'root' is for the parent query level, not the 'topplan'. + * This does not matter currently since we only access the dependency-item + * lists in root->glob, but it would need some hacking if we wanted a root + * that actually matches the subplan. + * + * Note: resultRelation is not yet adjusted by rtoffset. + */ +static List * +set_returning_clause_references(PlannerInfo *root, + List *rlist, + Plan *topplan, + Index resultRelation, + int rtoffset) +{ + indexed_tlist *itlist; + + /* + * We can perform the desired Var fixup by abusing the fix_join_expr + * machinery that formerly handled inner indexscan fixup. We search the + * top plan's targetlist for Vars of non-result relations, and use + * fix_join_expr to convert RETURNING Vars into references to those tlist + * entries, while leaving result-rel Vars as-is. + * + * PlaceHolderVars will also be sought in the targetlist, but no + * more-complex expressions will be. Note that it is not possible for a + * PlaceHolderVar to refer to the result relation, since the result is + * never below an outer join. If that case could happen, we'd have to be + * prepared to pick apart the PlaceHolderVar and evaluate its contained + * expression instead. + */ + itlist = build_tlist_index_other_vars(topplan->targetlist, resultRelation); + + rlist = fix_join_expr(root, + rlist, + itlist, + NULL, + resultRelation, + rtoffset, + NUM_EXEC_TLIST(topplan)); + + pfree(itlist); + + return rlist; +} + +/* + * fix_windowagg_condition_expr_mutator + * Mutator function for replacing WindowFuncs with the corresponding Var + * in the targetlist which references that WindowFunc. + */ +static Node * +fix_windowagg_condition_expr_mutator(Node *node, + fix_windowagg_cond_context *context) +{ + if (node == NULL) + return NULL; + + if (IsA(node, WindowFunc)) + { + Var *newvar; + + newvar = search_indexed_tlist_for_non_var((Expr *) node, + context->subplan_itlist, + context->newvarno); + if (newvar) + return (Node *) newvar; + elog(ERROR, "WindowFunc not found in subplan target lists"); + } + + return expression_tree_mutator(node, + fix_windowagg_condition_expr_mutator, + (void *) context); +} + +/* + * fix_windowagg_condition_expr + * Converts references in 'runcondition' so that any WindowFunc + * references are swapped out for a Var which references the matching + * WindowFunc in 'subplan_itlist'. + */ +static List * +fix_windowagg_condition_expr(PlannerInfo *root, + List *runcondition, + indexed_tlist *subplan_itlist) +{ + fix_windowagg_cond_context context; + + context.root = root; + context.subplan_itlist = subplan_itlist; + context.newvarno = 0; + + return (List *) fix_windowagg_condition_expr_mutator((Node *) runcondition, + &context); +} + +/* + * set_windowagg_runcondition_references + * Converts references in 'runcondition' so that any WindowFunc + * references are swapped out for a Var which references the matching + * WindowFunc in 'plan' targetlist. + */ +static List * +set_windowagg_runcondition_references(PlannerInfo *root, + List *runcondition, + Plan *plan) +{ + List *newlist; + indexed_tlist *itlist; + + itlist = build_tlist_index(plan->targetlist); + + newlist = fix_windowagg_condition_expr(root, runcondition, itlist); + + pfree(itlist); + + return newlist; +} + +/***************************************************************************** + * QUERY DEPENDENCY MANAGEMENT + *****************************************************************************/ + +/* + * record_plan_function_dependency + * Mark the current plan as depending on a particular function. + * + * This is exported so that the function-inlining code can record a + * dependency on a function that it's removed from the plan tree. + */ +void +record_plan_function_dependency(PlannerInfo *root, Oid funcid) +{ + /* + * For performance reasons, we don't bother to track built-in functions; + * we just assume they'll never change (or at least not in ways that'd + * invalidate plans using them). For this purpose we can consider a + * built-in function to be one with OID less than FirstUnpinnedObjectId. + * Note that the OID generator guarantees never to generate such an OID + * after startup, even at OID wraparound. + */ + if (funcid >= (Oid) FirstUnpinnedObjectId) + { + PlanInvalItem *inval_item = makeNode(PlanInvalItem); + + /* + * It would work to use any syscache on pg_proc, but the easiest is + * PROCOID since we already have the function's OID at hand. Note + * that plancache.c knows we use PROCOID. + */ + inval_item->cacheId = PROCOID; + inval_item->hashValue = GetSysCacheHashValue1(PROCOID, + ObjectIdGetDatum(funcid)); + + root->glob->invalItems = lappend(root->glob->invalItems, inval_item); + } +} + +/* + * record_plan_type_dependency + * Mark the current plan as depending on a particular type. + * + * This is exported so that eval_const_expressions can record a + * dependency on a domain that it's removed a CoerceToDomain node for. + * + * We don't currently need to record dependencies on domains that the + * plan contains CoerceToDomain nodes for, though that might change in + * future. Hence, this isn't actually called in this module, though + * someday fix_expr_common might call it. + */ +void +record_plan_type_dependency(PlannerInfo *root, Oid typid) +{ + /* + * As in record_plan_function_dependency, ignore the possibility that + * someone would change a built-in domain. + */ + if (typid >= (Oid) FirstUnpinnedObjectId) + { + PlanInvalItem *inval_item = makeNode(PlanInvalItem); + + /* + * It would work to use any syscache on pg_type, but the easiest is + * TYPEOID since we already have the type's OID at hand. Note that + * plancache.c knows we use TYPEOID. + */ + inval_item->cacheId = TYPEOID; + inval_item->hashValue = GetSysCacheHashValue1(TYPEOID, + ObjectIdGetDatum(typid)); + + root->glob->invalItems = lappend(root->glob->invalItems, inval_item); + } +} + +/* + * extract_query_dependencies + * Given a rewritten, but not yet planned, query or queries + * (i.e. a Query node or list of Query nodes), extract dependencies + * just as set_plan_references would do. Also detect whether any + * rewrite steps were affected by RLS. + * + * This is needed by plancache.c to handle invalidation of cached unplanned + * queries. + * + * Note: this does not go through eval_const_expressions, and hence doesn't + * reflect its additions of inlined functions and elided CoerceToDomain nodes + * to the invalItems list. This is obviously OK for functions, since we'll + * see them in the original query tree anyway. For domains, it's OK because + * we don't care about domains unless they get elided. That is, a plan might + * have domain dependencies that the query tree doesn't. + */ +void +extract_query_dependencies(Node *query, + List **relationOids, + List **invalItems, + bool *hasRowSecurity) +{ + PlannerGlobal glob; + PlannerInfo root; + + /* Make up dummy planner state so we can use this module's machinery */ + MemSet(&glob, 0, sizeof(glob)); + glob.type = T_PlannerGlobal; + glob.relationOids = NIL; + glob.invalItems = NIL; + /* Hack: we use glob.dependsOnRole to collect hasRowSecurity flags */ + glob.dependsOnRole = false; + + MemSet(&root, 0, sizeof(root)); + root.type = T_PlannerInfo; + root.glob = &glob; + + (void) extract_query_dependencies_walker(query, &root); + + *relationOids = glob.relationOids; + *invalItems = glob.invalItems; + *hasRowSecurity = glob.dependsOnRole; +} + +/* + * Tree walker for extract_query_dependencies. + * + * This is exported so that expression_planner_with_deps can call it on + * simple expressions (post-planning, not before planning, in that case). + * In that usage, glob.dependsOnRole isn't meaningful, but the relationOids + * and invalItems lists are added to as needed. + */ +bool +extract_query_dependencies_walker(Node *node, PlannerInfo *context) +{ + if (node == NULL) + return false; + Assert(!IsA(node, PlaceHolderVar)); + if (IsA(node, Query)) + { + Query *query = (Query *) node; + ListCell *lc; + + if (query->commandType == CMD_UTILITY) + { + /* + * This logic must handle any utility command for which parse + * analysis was nontrivial (cf. stmt_requires_parse_analysis). + * + * Notably, CALL requires its own processing. + */ + if (IsA(query->utilityStmt, CallStmt)) + { + CallStmt *callstmt = (CallStmt *) query->utilityStmt; + + /* We need not examine funccall, just the transformed exprs */ + (void) extract_query_dependencies_walker((Node *) callstmt->funcexpr, + context); + (void) extract_query_dependencies_walker((Node *) callstmt->outargs, + context); + return false; + } + + /* + * Ignore other utility statements, except those (such as EXPLAIN) + * that contain a parsed-but-not-planned query. For those, we + * just need to transfer our attention to the contained query. + */ + query = UtilityContainsQuery(query->utilityStmt); + if (query == NULL) + return false; + } + + /* Remember if any Query has RLS quals applied by rewriter */ + if (query->hasRowSecurity) + context->glob->dependsOnRole = true; + + /* Collect relation OIDs in this Query's rtable */ + foreach(lc, query->rtable) + { + RangeTblEntry *rte = (RangeTblEntry *) lfirst(lc); + + if (rte->rtekind == RTE_RELATION) + context->glob->relationOids = + lappend_oid(context->glob->relationOids, rte->relid); + else if (rte->rtekind == RTE_NAMEDTUPLESTORE && + OidIsValid(rte->relid)) + context->glob->relationOids = + lappend_oid(context->glob->relationOids, + rte->relid); + } + + /* And recurse into the query's subexpressions */ + return query_tree_walker(query, extract_query_dependencies_walker, + (void *) context, 0); + } + /* Extract function dependencies and check for regclass Consts */ + fix_expr_common(context, node); + return expression_tree_walker(node, extract_query_dependencies_walker, + (void *) context); +} diff --git a/src/backend/optimizer/plan/subselect.c b/src/backend/optimizer/plan/subselect.c new file mode 100644 index 0000000..a195788 --- /dev/null +++ b/src/backend/optimizer/plan/subselect.c @@ -0,0 +1,2999 @@ +/*------------------------------------------------------------------------- + * + * subselect.c + * Planning routines for subselects. + * + * This module deals with SubLinks and CTEs, but not subquery RTEs (i.e., + * not sub-SELECT-in-FROM cases). + * + * Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group + * Portions Copyright (c) 1994, Regents of the University of California + * + * IDENTIFICATION + * src/backend/optimizer/plan/subselect.c + * + *------------------------------------------------------------------------- + */ +#include "postgres.h" + +#include "access/htup_details.h" +#include "catalog/pg_operator.h" +#include "catalog/pg_type.h" +#include "executor/executor.h" +#include "miscadmin.h" +#include "nodes/makefuncs.h" +#include "nodes/nodeFuncs.h" +#include "optimizer/clauses.h" +#include "optimizer/cost.h" +#include "optimizer/optimizer.h" +#include "optimizer/paramassign.h" +#include "optimizer/pathnode.h" +#include "optimizer/planmain.h" +#include "optimizer/planner.h" +#include "optimizer/prep.h" +#include "optimizer/subselect.h" +#include "parser/parse_relation.h" +#include "rewrite/rewriteManip.h" +#include "utils/builtins.h" +#include "utils/lsyscache.h" +#include "utils/syscache.h" + + +typedef struct convert_testexpr_context +{ + PlannerInfo *root; + List *subst_nodes; /* Nodes to substitute for Params */ +} convert_testexpr_context; + +typedef struct process_sublinks_context +{ + PlannerInfo *root; + bool isTopQual; +} process_sublinks_context; + +typedef struct finalize_primnode_context +{ + PlannerInfo *root; + Bitmapset *paramids; /* Non-local PARAM_EXEC paramids found */ +} finalize_primnode_context; + +typedef struct inline_cte_walker_context +{ + const char *ctename; /* name and relative level of target CTE */ + int levelsup; + Query *ctequery; /* query to substitute */ +} inline_cte_walker_context; + + +static Node *build_subplan(PlannerInfo *root, Plan *plan, PlannerInfo *subroot, + List *plan_params, + SubLinkType subLinkType, int subLinkId, + Node *testexpr, List *testexpr_paramids, + bool unknownEqFalse); +static List *generate_subquery_params(PlannerInfo *root, List *tlist, + List **paramIds); +static List *generate_subquery_vars(PlannerInfo *root, List *tlist, + Index varno); +static Node *convert_testexpr(PlannerInfo *root, + Node *testexpr, + List *subst_nodes); +static Node *convert_testexpr_mutator(Node *node, + convert_testexpr_context *context); +static bool subplan_is_hashable(Plan *plan); +static bool subpath_is_hashable(Path *path); +static bool testexpr_is_hashable(Node *testexpr, List *param_ids); +static bool test_opexpr_is_hashable(OpExpr *testexpr, List *param_ids); +static bool hash_ok_operator(OpExpr *expr); +static bool contain_dml(Node *node); +static bool contain_dml_walker(Node *node, void *context); +static bool contain_outer_selfref(Node *node); +static bool contain_outer_selfref_walker(Node *node, Index *depth); +static void inline_cte(PlannerInfo *root, CommonTableExpr *cte); +static bool inline_cte_walker(Node *node, inline_cte_walker_context *context); +static bool simplify_EXISTS_query(PlannerInfo *root, Query *query); +static Query *convert_EXISTS_to_ANY(PlannerInfo *root, Query *subselect, + Node **testexpr, List **paramIds); +static Node *replace_correlation_vars_mutator(Node *node, PlannerInfo *root); +static Node *process_sublinks_mutator(Node *node, + process_sublinks_context *context); +static Bitmapset *finalize_plan(PlannerInfo *root, + Plan *plan, + int gather_param, + Bitmapset *valid_params, + Bitmapset *scan_params); +static bool finalize_primnode(Node *node, finalize_primnode_context *context); +static bool finalize_agg_primnode(Node *node, finalize_primnode_context *context); + + +/* + * Get the datatype/typmod/collation of the first column of the plan's output. + * + * This information is stored for ARRAY_SUBLINK execution and for + * exprType()/exprTypmod()/exprCollation(), which have no way to get at the + * plan associated with a SubPlan node. We really only need the info for + * EXPR_SUBLINK and ARRAY_SUBLINK subplans, but for consistency we save it + * always. + */ +static void +get_first_col_type(Plan *plan, Oid *coltype, int32 *coltypmod, + Oid *colcollation) +{ + /* In cases such as EXISTS, tlist might be empty; arbitrarily use VOID */ + if (plan->targetlist) + { + TargetEntry *tent = linitial_node(TargetEntry, plan->targetlist); + + if (!tent->resjunk) + { + *coltype = exprType((Node *) tent->expr); + *coltypmod = exprTypmod((Node *) tent->expr); + *colcollation = exprCollation((Node *) tent->expr); + return; + } + } + *coltype = VOIDOID; + *coltypmod = -1; + *colcollation = InvalidOid; +} + +/* + * Convert a SubLink (as created by the parser) into a SubPlan. + * + * We are given the SubLink's contained query, type, ID, and testexpr. We are + * also told if this expression appears at top level of a WHERE/HAVING qual. + * + * Note: we assume that the testexpr has been AND/OR flattened (actually, + * it's been through eval_const_expressions), but not converted to + * implicit-AND form; and any SubLinks in it should already have been + * converted to SubPlans. The subquery is as yet untouched, however. + * + * The result is whatever we need to substitute in place of the SubLink node + * in the executable expression. If we're going to do the subplan as a + * regular subplan, this will be the constructed SubPlan node. If we're going + * to do the subplan as an InitPlan, the SubPlan node instead goes into + * root->init_plans, and what we return here is an expression tree + * representing the InitPlan's result: usually just a Param node representing + * a single scalar result, but possibly a row comparison tree containing + * multiple Param nodes, or for a MULTIEXPR subquery a simple NULL constant + * (since the real output Params are elsewhere in the tree, and the MULTIEXPR + * subquery itself is in a resjunk tlist entry whose value is uninteresting). + */ +static Node * +make_subplan(PlannerInfo *root, Query *orig_subquery, + SubLinkType subLinkType, int subLinkId, + Node *testexpr, bool isTopQual) +{ + Query *subquery; + bool simple_exists = false; + double tuple_fraction; + PlannerInfo *subroot; + RelOptInfo *final_rel; + Path *best_path; + Plan *plan; + List *plan_params; + Node *result; + + /* + * Copy the source Query node. This is a quick and dirty kluge to resolve + * the fact that the parser can generate trees with multiple links to the + * same sub-Query node, but the planner wants to scribble on the Query. + * Try to clean this up when we do querytree redesign... + */ + subquery = copyObject(orig_subquery); + + /* + * If it's an EXISTS subplan, we might be able to simplify it. + */ + if (subLinkType == EXISTS_SUBLINK) + simple_exists = simplify_EXISTS_query(root, subquery); + + /* + * For an EXISTS subplan, tell lower-level planner to expect that only the + * first tuple will be retrieved. For ALL and ANY subplans, we will be + * able to stop evaluating if the test condition fails or matches, so very + * often not all the tuples will be retrieved; for lack of a better idea, + * specify 50% retrieval. For EXPR, MULTIEXPR, and ROWCOMPARE subplans, + * use default behavior (we're only expecting one row out, anyway). + * + * NOTE: if you change these numbers, also change cost_subplan() in + * path/costsize.c. + * + * XXX If an ANY subplan is uncorrelated, build_subplan may decide to hash + * its output. In that case it would've been better to specify full + * retrieval. At present, however, we can only check hashability after + * we've made the subplan :-(. (Determining whether it'll fit in hash_mem + * is the really hard part.) Therefore, we don't want to be too + * optimistic about the percentage of tuples retrieved, for fear of + * selecting a plan that's bad for the materialization case. + */ + if (subLinkType == EXISTS_SUBLINK) + tuple_fraction = 1.0; /* just like a LIMIT 1 */ + else if (subLinkType == ALL_SUBLINK || + subLinkType == ANY_SUBLINK) + tuple_fraction = 0.5; /* 50% */ + else + tuple_fraction = 0.0; /* default behavior */ + + /* plan_params should not be in use in current query level */ + Assert(root->plan_params == NIL); + + /* Generate Paths for the subquery */ + subroot = subquery_planner(root->glob, subquery, + root, + false, tuple_fraction); + + /* Isolate the params needed by this specific subplan */ + plan_params = root->plan_params; + root->plan_params = NIL; + + /* + * Select best Path and turn it into a Plan. At least for now, there + * seems no reason to postpone doing that. + */ + final_rel = fetch_upper_rel(subroot, UPPERREL_FINAL, NULL); + best_path = get_cheapest_fractional_path(final_rel, tuple_fraction); + + plan = create_plan(subroot, best_path); + + /* And convert to SubPlan or InitPlan format. */ + result = build_subplan(root, plan, subroot, plan_params, + subLinkType, subLinkId, + testexpr, NIL, isTopQual); + + /* + * If it's a correlated EXISTS with an unimportant targetlist, we might be + * able to transform it to the equivalent of an IN and then implement it + * by hashing. We don't have enough information yet to tell which way is + * likely to be better (it depends on the expected number of executions of + * the EXISTS qual, and we are much too early in planning the outer query + * to be able to guess that). So we generate both plans, if possible, and + * leave it to setrefs.c to decide which to use. + */ + if (simple_exists && IsA(result, SubPlan)) + { + Node *newtestexpr; + List *paramIds; + + /* Make a second copy of the original subquery */ + subquery = copyObject(orig_subquery); + /* and re-simplify */ + simple_exists = simplify_EXISTS_query(root, subquery); + Assert(simple_exists); + /* See if it can be converted to an ANY query */ + subquery = convert_EXISTS_to_ANY(root, subquery, + &newtestexpr, ¶mIds); + if (subquery) + { + /* Generate Paths for the ANY subquery; we'll need all rows */ + subroot = subquery_planner(root->glob, subquery, + root, + false, 0.0); + + /* Isolate the params needed by this specific subplan */ + plan_params = root->plan_params; + root->plan_params = NIL; + + /* Select best Path */ + final_rel = fetch_upper_rel(subroot, UPPERREL_FINAL, NULL); + best_path = final_rel->cheapest_total_path; + + /* Now we can check if it'll fit in hash_mem */ + if (subpath_is_hashable(best_path)) + { + SubPlan *hashplan; + AlternativeSubPlan *asplan; + + /* OK, finish planning the ANY subquery */ + plan = create_plan(subroot, best_path); + + /* ... and convert to SubPlan format */ + hashplan = castNode(SubPlan, + build_subplan(root, plan, subroot, + plan_params, + ANY_SUBLINK, 0, + newtestexpr, + paramIds, + true)); + /* Check we got what we expected */ + Assert(hashplan->parParam == NIL); + Assert(hashplan->useHashTable); + + /* Leave it to setrefs.c to decide which plan to use */ + asplan = makeNode(AlternativeSubPlan); + asplan->subplans = list_make2(result, hashplan); + result = (Node *) asplan; + root->hasAlternativeSubPlans = true; + } + } + } + + return result; +} + +/* + * Build a SubPlan node given the raw inputs --- subroutine for make_subplan + * + * Returns either the SubPlan, or a replacement expression if we decide to + * make it an InitPlan, as explained in the comments for make_subplan. + */ +static Node * +build_subplan(PlannerInfo *root, Plan *plan, PlannerInfo *subroot, + List *plan_params, + SubLinkType subLinkType, int subLinkId, + Node *testexpr, List *testexpr_paramids, + bool unknownEqFalse) +{ + Node *result; + SubPlan *splan; + bool isInitPlan; + ListCell *lc; + + /* + * Initialize the SubPlan node. Note plan_id, plan_name, and cost fields + * are set further down. + */ + splan = makeNode(SubPlan); + splan->subLinkType = subLinkType; + splan->testexpr = NULL; + splan->paramIds = NIL; + get_first_col_type(plan, &splan->firstColType, &splan->firstColTypmod, + &splan->firstColCollation); + splan->useHashTable = false; + splan->unknownEqFalse = unknownEqFalse; + splan->parallel_safe = plan->parallel_safe; + splan->setParam = NIL; + splan->parParam = NIL; + splan->args = NIL; + + /* + * Make parParam and args lists of param IDs and expressions that current + * query level will pass to this child plan. + */ + foreach(lc, plan_params) + { + PlannerParamItem *pitem = (PlannerParamItem *) lfirst(lc); + Node *arg = pitem->item; + + /* + * The Var, PlaceHolderVar, Aggref or GroupingFunc has already been + * adjusted to have the correct varlevelsup, phlevelsup, or + * agglevelsup. + * + * If it's a PlaceHolderVar, Aggref or GroupingFunc, its arguments + * might contain SubLinks, which have not yet been processed (see the + * comments for SS_replace_correlation_vars). Do that now. + */ + if (IsA(arg, PlaceHolderVar) || + IsA(arg, Aggref) || + IsA(arg, GroupingFunc)) + arg = SS_process_sublinks(root, arg, false); + + splan->parParam = lappend_int(splan->parParam, pitem->paramId); + splan->args = lappend(splan->args, arg); + } + + /* + * Un-correlated or undirect correlated plans of EXISTS, EXPR, ARRAY, + * ROWCOMPARE, or MULTIEXPR types can be used as initPlans. For EXISTS, + * EXPR, or ARRAY, we return a Param referring to the result of evaluating + * the initPlan. For ROWCOMPARE, we must modify the testexpr tree to + * contain PARAM_EXEC Params instead of the PARAM_SUBLINK Params emitted + * by the parser, and then return that tree. For MULTIEXPR, we return a + * null constant: the resjunk targetlist item containing the SubLink does + * not need to return anything useful, since the referencing Params are + * elsewhere. + */ + if (splan->parParam == NIL && subLinkType == EXISTS_SUBLINK) + { + Param *prm; + + Assert(testexpr == NULL); + prm = generate_new_exec_param(root, BOOLOID, -1, InvalidOid); + splan->setParam = list_make1_int(prm->paramid); + isInitPlan = true; + result = (Node *) prm; + } + else if (splan->parParam == NIL && subLinkType == EXPR_SUBLINK) + { + TargetEntry *te = linitial(plan->targetlist); + Param *prm; + + Assert(!te->resjunk); + Assert(testexpr == NULL); + prm = generate_new_exec_param(root, + exprType((Node *) te->expr), + exprTypmod((Node *) te->expr), + exprCollation((Node *) te->expr)); + splan->setParam = list_make1_int(prm->paramid); + isInitPlan = true; + result = (Node *) prm; + } + else if (splan->parParam == NIL && subLinkType == ARRAY_SUBLINK) + { + TargetEntry *te = linitial(plan->targetlist); + Oid arraytype; + Param *prm; + + Assert(!te->resjunk); + Assert(testexpr == NULL); + arraytype = get_promoted_array_type(exprType((Node *) te->expr)); + if (!OidIsValid(arraytype)) + elog(ERROR, "could not find array type for datatype %s", + format_type_be(exprType((Node *) te->expr))); + prm = generate_new_exec_param(root, + arraytype, + exprTypmod((Node *) te->expr), + exprCollation((Node *) te->expr)); + splan->setParam = list_make1_int(prm->paramid); + isInitPlan = true; + result = (Node *) prm; + } + else if (splan->parParam == NIL && subLinkType == ROWCOMPARE_SUBLINK) + { + /* Adjust the Params */ + List *params; + + Assert(testexpr != NULL); + params = generate_subquery_params(root, + plan->targetlist, + &splan->paramIds); + result = convert_testexpr(root, + testexpr, + params); + splan->setParam = list_copy(splan->paramIds); + isInitPlan = true; + + /* + * The executable expression is returned to become part of the outer + * plan's expression tree; it is not kept in the initplan node. + */ + } + else if (subLinkType == MULTIEXPR_SUBLINK) + { + /* + * Whether it's an initplan or not, it needs to set a PARAM_EXEC Param + * for each output column. + */ + List *params; + + Assert(testexpr == NULL); + params = generate_subquery_params(root, + plan->targetlist, + &splan->setParam); + + /* + * Save the list of replacement Params in the n'th cell of + * root->multiexpr_params; setrefs.c will use it to replace + * PARAM_MULTIEXPR Params. + */ + while (list_length(root->multiexpr_params) < subLinkId) + root->multiexpr_params = lappend(root->multiexpr_params, NIL); + lc = list_nth_cell(root->multiexpr_params, subLinkId - 1); + Assert(lfirst(lc) == NIL); + lfirst(lc) = params; + + /* It can be an initplan if there are no parParams. */ + if (splan->parParam == NIL) + { + isInitPlan = true; + result = (Node *) makeNullConst(RECORDOID, -1, InvalidOid); + } + else + { + isInitPlan = false; + result = (Node *) splan; + } + } + else + { + /* + * Adjust the Params in the testexpr, unless caller already took care + * of it (as indicated by passing a list of Param IDs). + */ + if (testexpr && testexpr_paramids == NIL) + { + List *params; + + params = generate_subquery_params(root, + plan->targetlist, + &splan->paramIds); + splan->testexpr = convert_testexpr(root, + testexpr, + params); + } + else + { + splan->testexpr = testexpr; + splan->paramIds = testexpr_paramids; + } + + /* + * We can't convert subplans of ALL_SUBLINK or ANY_SUBLINK types to + * initPlans, even when they are uncorrelated or undirect correlated, + * because we need to scan the output of the subplan for each outer + * tuple. But if it's a not-direct-correlated IN (= ANY) test, we + * might be able to use a hashtable to avoid comparing all the tuples. + */ + if (subLinkType == ANY_SUBLINK && + splan->parParam == NIL && + subplan_is_hashable(plan) && + testexpr_is_hashable(splan->testexpr, splan->paramIds)) + splan->useHashTable = true; + + /* + * Otherwise, we have the option to tack a Material node onto the top + * of the subplan, to reduce the cost of reading it repeatedly. This + * is pointless for a direct-correlated subplan, since we'd have to + * recompute its results each time anyway. For uncorrelated/undirect + * correlated subplans, we add Material unless the subplan's top plan + * node would materialize its output anyway. Also, if enable_material + * is false, then the user does not want us to materialize anything + * unnecessarily, so we don't. + */ + else if (splan->parParam == NIL && enable_material && + !ExecMaterializesOutput(nodeTag(plan))) + plan = materialize_finished_plan(plan); + + result = (Node *) splan; + isInitPlan = false; + } + + /* + * Add the subplan and its PlannerInfo to the global lists. + */ + root->glob->subplans = lappend(root->glob->subplans, plan); + root->glob->subroots = lappend(root->glob->subroots, subroot); + splan->plan_id = list_length(root->glob->subplans); + + if (isInitPlan) + root->init_plans = lappend(root->init_plans, splan); + + /* + * A parameterless subplan (not initplan) should be prepared to handle + * REWIND efficiently. If it has direct parameters then there's no point + * since it'll be reset on each scan anyway; and if it's an initplan then + * there's no point since it won't get re-run without parameter changes + * anyway. The input of a hashed subplan doesn't need REWIND either. + */ + if (splan->parParam == NIL && !isInitPlan && !splan->useHashTable) + root->glob->rewindPlanIDs = bms_add_member(root->glob->rewindPlanIDs, + splan->plan_id); + + /* Label the subplan for EXPLAIN purposes */ + splan->plan_name = palloc(32 + 12 * list_length(splan->setParam)); + sprintf(splan->plan_name, "%s %d", + isInitPlan ? "InitPlan" : "SubPlan", + splan->plan_id); + if (splan->setParam) + { + char *ptr = splan->plan_name + strlen(splan->plan_name); + + ptr += sprintf(ptr, " (returns "); + foreach(lc, splan->setParam) + { + ptr += sprintf(ptr, "$%d%s", + lfirst_int(lc), + lnext(splan->setParam, lc) ? "," : ")"); + } + } + + /* Lastly, fill in the cost estimates for use later */ + cost_subplan(root, splan, plan); + + return result; +} + +/* + * generate_subquery_params: build a list of Params representing the output + * columns of a sublink's sub-select, given the sub-select's targetlist. + * + * We also return an integer list of the paramids of the Params. + */ +static List * +generate_subquery_params(PlannerInfo *root, List *tlist, List **paramIds) +{ + List *result; + List *ids; + ListCell *lc; + + result = ids = NIL; + foreach(lc, tlist) + { + TargetEntry *tent = (TargetEntry *) lfirst(lc); + Param *param; + + if (tent->resjunk) + continue; + + param = generate_new_exec_param(root, + exprType((Node *) tent->expr), + exprTypmod((Node *) tent->expr), + exprCollation((Node *) tent->expr)); + result = lappend(result, param); + ids = lappend_int(ids, param->paramid); + } + + *paramIds = ids; + return result; +} + +/* + * generate_subquery_vars: build a list of Vars representing the output + * columns of a sublink's sub-select, given the sub-select's targetlist. + * The Vars have the specified varno (RTE index). + */ +static List * +generate_subquery_vars(PlannerInfo *root, List *tlist, Index varno) +{ + List *result; + ListCell *lc; + + result = NIL; + foreach(lc, tlist) + { + TargetEntry *tent = (TargetEntry *) lfirst(lc); + Var *var; + + if (tent->resjunk) + continue; + + var = makeVarFromTargetEntry(varno, tent); + result = lappend(result, var); + } + + return result; +} + +/* + * convert_testexpr: convert the testexpr given by the parser into + * actually executable form. This entails replacing PARAM_SUBLINK Params + * with Params or Vars representing the results of the sub-select. The + * nodes to be substituted are passed in as the List result from + * generate_subquery_params or generate_subquery_vars. + */ +static Node * +convert_testexpr(PlannerInfo *root, + Node *testexpr, + List *subst_nodes) +{ + convert_testexpr_context context; + + context.root = root; + context.subst_nodes = subst_nodes; + return convert_testexpr_mutator(testexpr, &context); +} + +static Node * +convert_testexpr_mutator(Node *node, + convert_testexpr_context *context) +{ + if (node == NULL) + return NULL; + if (IsA(node, Param)) + { + Param *param = (Param *) node; + + if (param->paramkind == PARAM_SUBLINK) + { + if (param->paramid <= 0 || + param->paramid > list_length(context->subst_nodes)) + elog(ERROR, "unexpected PARAM_SUBLINK ID: %d", param->paramid); + + /* + * We copy the list item to avoid having doubly-linked + * substructure in the modified parse tree. This is probably + * unnecessary when it's a Param, but be safe. + */ + return (Node *) copyObject(list_nth(context->subst_nodes, + param->paramid - 1)); + } + } + if (IsA(node, SubLink)) + { + /* + * If we come across a nested SubLink, it is neither necessary nor + * correct to recurse into it: any PARAM_SUBLINKs we might find inside + * belong to the inner SubLink not the outer. So just return it as-is. + * + * This reasoning depends on the assumption that nothing will pull + * subexpressions into or out of the testexpr field of a SubLink, at + * least not without replacing PARAM_SUBLINKs first. If we did want + * to do that we'd need to rethink the parser-output representation + * altogether, since currently PARAM_SUBLINKs are only unique per + * SubLink not globally across the query. The whole point of + * replacing them with Vars or PARAM_EXEC nodes is to make them + * globally unique before they escape from the SubLink's testexpr. + * + * Note: this can't happen when called during SS_process_sublinks, + * because that recursively processes inner SubLinks first. It can + * happen when called from convert_ANY_sublink_to_join, though. + */ + return node; + } + return expression_tree_mutator(node, + convert_testexpr_mutator, + (void *) context); +} + +/* + * subplan_is_hashable: can we implement an ANY subplan by hashing? + * + * This is not responsible for checking whether the combining testexpr + * is suitable for hashing. We only look at the subquery itself. + */ +static bool +subplan_is_hashable(Plan *plan) +{ + double subquery_size; + + /* + * The estimated size of the subquery result must fit in hash_mem. (Note: + * we use heap tuple overhead here even though the tuples will actually be + * stored as MinimalTuples; this provides some fudge factor for hashtable + * overhead.) + */ + subquery_size = plan->plan_rows * + (MAXALIGN(plan->plan_width) + MAXALIGN(SizeofHeapTupleHeader)); + if (subquery_size > get_hash_memory_limit()) + return false; + + return true; +} + +/* + * subpath_is_hashable: can we implement an ANY subplan by hashing? + * + * Identical to subplan_is_hashable, but work from a Path for the subplan. + */ +static bool +subpath_is_hashable(Path *path) +{ + double subquery_size; + + /* + * The estimated size of the subquery result must fit in hash_mem. (Note: + * we use heap tuple overhead here even though the tuples will actually be + * stored as MinimalTuples; this provides some fudge factor for hashtable + * overhead.) + */ + subquery_size = path->rows * + (MAXALIGN(path->pathtarget->width) + MAXALIGN(SizeofHeapTupleHeader)); + if (subquery_size > get_hash_memory_limit()) + return false; + + return true; +} + +/* + * testexpr_is_hashable: is an ANY SubLink's test expression hashable? + * + * To identify LHS vs RHS of the hash expression, we must be given the + * list of output Param IDs of the SubLink's subquery. + */ +static bool +testexpr_is_hashable(Node *testexpr, List *param_ids) +{ + /* + * The testexpr must be a single OpExpr, or an AND-clause containing only + * OpExprs, each of which satisfy test_opexpr_is_hashable(). + */ + if (testexpr && IsA(testexpr, OpExpr)) + { + if (test_opexpr_is_hashable((OpExpr *) testexpr, param_ids)) + return true; + } + else if (is_andclause(testexpr)) + { + ListCell *l; + + foreach(l, ((BoolExpr *) testexpr)->args) + { + Node *andarg = (Node *) lfirst(l); + + if (!IsA(andarg, OpExpr)) + return false; + if (!test_opexpr_is_hashable((OpExpr *) andarg, param_ids)) + return false; + } + return true; + } + + return false; +} + +static bool +test_opexpr_is_hashable(OpExpr *testexpr, List *param_ids) +{ + /* + * The combining operator must be hashable and strict. The need for + * hashability is obvious, since we want to use hashing. Without + * strictness, behavior in the presence of nulls is too unpredictable. We + * actually must assume even more than plain strictness: it can't yield + * NULL for non-null inputs, either (see nodeSubplan.c). However, hash + * indexes and hash joins assume that too. + */ + if (!hash_ok_operator(testexpr)) + return false; + + /* + * The left and right inputs must belong to the outer and inner queries + * respectively; hence Params that will be supplied by the subquery must + * not appear in the LHS, and Vars of the outer query must not appear in + * the RHS. (Ordinarily, this must be true because of the way that the + * parser builds an ANY SubLink's testexpr ... but inlining of functions + * could have changed the expression's structure, so we have to check. + * Such cases do not occur often enough to be worth trying to optimize, so + * we don't worry about trying to commute the clause or anything like + * that; we just need to be sure not to build an invalid plan.) + */ + if (list_length(testexpr->args) != 2) + return false; + if (contain_exec_param((Node *) linitial(testexpr->args), param_ids)) + return false; + if (contain_var_clause((Node *) lsecond(testexpr->args))) + return false; + return true; +} + +/* + * Check expression is hashable + strict + * + * We could use op_hashjoinable() and op_strict(), but do it like this to + * avoid a redundant cache lookup. + */ +static bool +hash_ok_operator(OpExpr *expr) +{ + Oid opid = expr->opno; + + /* quick out if not a binary operator */ + if (list_length(expr->args) != 2) + return false; + if (opid == ARRAY_EQ_OP || + opid == RECORD_EQ_OP) + { + /* these are strict, but must check input type to ensure hashable */ + Node *leftarg = linitial(expr->args); + + return op_hashjoinable(opid, exprType(leftarg)); + } + else + { + /* else must look up the operator properties */ + HeapTuple tup; + Form_pg_operator optup; + + tup = SearchSysCache1(OPEROID, ObjectIdGetDatum(opid)); + if (!HeapTupleIsValid(tup)) + elog(ERROR, "cache lookup failed for operator %u", opid); + optup = (Form_pg_operator) GETSTRUCT(tup); + if (!optup->oprcanhash || !func_strict(optup->oprcode)) + { + ReleaseSysCache(tup); + return false; + } + ReleaseSysCache(tup); + return true; + } +} + + +/* + * SS_process_ctes: process a query's WITH list + * + * Consider each CTE in the WITH list and either ignore it (if it's an + * unreferenced SELECT), "inline" it to create a regular sub-SELECT-in-FROM, + * or convert it to an initplan. + * + * A side effect is to fill in root->cte_plan_ids with a list that + * parallels root->parse->cteList and provides the subplan ID for + * each CTE's initplan, or a dummy ID (-1) if we didn't make an initplan. + */ +void +SS_process_ctes(PlannerInfo *root) +{ + ListCell *lc; + + Assert(root->cte_plan_ids == NIL); + + foreach(lc, root->parse->cteList) + { + CommonTableExpr *cte = (CommonTableExpr *) lfirst(lc); + CmdType cmdType = ((Query *) cte->ctequery)->commandType; + Query *subquery; + PlannerInfo *subroot; + RelOptInfo *final_rel; + Path *best_path; + Plan *plan; + SubPlan *splan; + int paramid; + + /* + * Ignore SELECT CTEs that are not actually referenced anywhere. + */ + if (cte->cterefcount == 0 && cmdType == CMD_SELECT) + { + /* Make a dummy entry in cte_plan_ids */ + root->cte_plan_ids = lappend_int(root->cte_plan_ids, -1); + continue; + } + + /* + * Consider inlining the CTE (creating RTE_SUBQUERY RTE(s)) instead of + * implementing it as a separately-planned CTE. + * + * We cannot inline if any of these conditions hold: + * + * 1. The user said not to (the CTEMaterializeAlways option). + * + * 2. The CTE is recursive. + * + * 3. The CTE has side-effects; this includes either not being a plain + * SELECT, or containing volatile functions. Inlining might change + * the side-effects, which would be bad. + * + * 4. The CTE is multiply-referenced and contains a self-reference to + * a recursive CTE outside itself. Inlining would result in multiple + * recursive self-references, which we don't support. + * + * Otherwise, we have an option whether to inline or not. That should + * always be a win if there's just a single reference, but if the CTE + * is multiply-referenced then it's unclear: inlining adds duplicate + * computations, but the ability to absorb restrictions from the outer + * query level could outweigh that. We do not have nearly enough + * information at this point to tell whether that's true, so we let + * the user express a preference. Our default behavior is to inline + * only singly-referenced CTEs, but a CTE marked CTEMaterializeNever + * will be inlined even if multiply referenced. + * + * Note: we check for volatile functions last, because that's more + * expensive than the other tests needed. + */ + if ((cte->ctematerialized == CTEMaterializeNever || + (cte->ctematerialized == CTEMaterializeDefault && + cte->cterefcount == 1)) && + !cte->cterecursive && + cmdType == CMD_SELECT && + !contain_dml(cte->ctequery) && + (cte->cterefcount <= 1 || + !contain_outer_selfref(cte->ctequery)) && + !contain_volatile_functions(cte->ctequery)) + { + inline_cte(root, cte); + /* Make a dummy entry in cte_plan_ids */ + root->cte_plan_ids = lappend_int(root->cte_plan_ids, -1); + continue; + } + + /* + * Copy the source Query node. Probably not necessary, but let's keep + * this similar to make_subplan. + */ + subquery = (Query *) copyObject(cte->ctequery); + + /* plan_params should not be in use in current query level */ + Assert(root->plan_params == NIL); + + /* + * Generate Paths for the CTE query. Always plan for full retrieval + * --- we don't have enough info to predict otherwise. + */ + subroot = subquery_planner(root->glob, subquery, + root, + cte->cterecursive, 0.0); + + /* + * Since the current query level doesn't yet contain any RTEs, it + * should not be possible for the CTE to have requested parameters of + * this level. + */ + if (root->plan_params) + elog(ERROR, "unexpected outer reference in CTE query"); + + /* + * Select best Path and turn it into a Plan. At least for now, there + * seems no reason to postpone doing that. + */ + final_rel = fetch_upper_rel(subroot, UPPERREL_FINAL, NULL); + best_path = final_rel->cheapest_total_path; + + plan = create_plan(subroot, best_path); + + /* + * Make a SubPlan node for it. This is just enough unlike + * build_subplan that we can't share code. + * + * Note plan_id, plan_name, and cost fields are set further down. + */ + splan = makeNode(SubPlan); + splan->subLinkType = CTE_SUBLINK; + splan->testexpr = NULL; + splan->paramIds = NIL; + get_first_col_type(plan, &splan->firstColType, &splan->firstColTypmod, + &splan->firstColCollation); + splan->useHashTable = false; + splan->unknownEqFalse = false; + + /* + * CTE scans are not considered for parallelism (cf + * set_rel_consider_parallel), and even if they were, initPlans aren't + * parallel-safe. + */ + splan->parallel_safe = false; + splan->setParam = NIL; + splan->parParam = NIL; + splan->args = NIL; + + /* + * The node can't have any inputs (since it's an initplan), so the + * parParam and args lists remain empty. (It could contain references + * to earlier CTEs' output param IDs, but CTE outputs are not + * propagated via the args list.) + */ + + /* + * Assign a param ID to represent the CTE's output. No ordinary + * "evaluation" of this param slot ever happens, but we use the param + * ID for setParam/chgParam signaling just as if the CTE plan were + * returning a simple scalar output. (Also, the executor abuses the + * ParamExecData slot for this param ID for communication among + * multiple CteScan nodes that might be scanning this CTE.) + */ + paramid = assign_special_exec_param(root); + splan->setParam = list_make1_int(paramid); + + /* + * Add the subplan and its PlannerInfo to the global lists. + */ + root->glob->subplans = lappend(root->glob->subplans, plan); + root->glob->subroots = lappend(root->glob->subroots, subroot); + splan->plan_id = list_length(root->glob->subplans); + + root->init_plans = lappend(root->init_plans, splan); + + root->cte_plan_ids = lappend_int(root->cte_plan_ids, splan->plan_id); + + /* Label the subplan for EXPLAIN purposes */ + splan->plan_name = psprintf("CTE %s", cte->ctename); + + /* Lastly, fill in the cost estimates for use later */ + cost_subplan(root, splan, plan); + } +} + +/* + * contain_dml: is any subquery not a plain SELECT? + * + * We reject SELECT FOR UPDATE/SHARE as well as INSERT etc. + */ +static bool +contain_dml(Node *node) +{ + return contain_dml_walker(node, NULL); +} + +static bool +contain_dml_walker(Node *node, void *context) +{ + if (node == NULL) + return false; + if (IsA(node, Query)) + { + Query *query = (Query *) node; + + if (query->commandType != CMD_SELECT || + query->rowMarks != NIL) + return true; + + return query_tree_walker(query, contain_dml_walker, context, 0); + } + return expression_tree_walker(node, contain_dml_walker, context); +} + +/* + * contain_outer_selfref: is there an external recursive self-reference? + */ +static bool +contain_outer_selfref(Node *node) +{ + Index depth = 0; + + /* + * We should be starting with a Query, so that depth will be 1 while + * examining its immediate contents. + */ + Assert(IsA(node, Query)); + + return contain_outer_selfref_walker(node, &depth); +} + +static bool +contain_outer_selfref_walker(Node *node, Index *depth) +{ + if (node == NULL) + return false; + if (IsA(node, RangeTblEntry)) + { + RangeTblEntry *rte = (RangeTblEntry *) node; + + /* + * Check for a self-reference to a CTE that's above the Query that our + * search started at. + */ + if (rte->rtekind == RTE_CTE && + rte->self_reference && + rte->ctelevelsup >= *depth) + return true; + return false; /* allow range_table_walker to continue */ + } + if (IsA(node, Query)) + { + /* Recurse into subquery, tracking nesting depth properly */ + Query *query = (Query *) node; + bool result; + + (*depth)++; + + result = query_tree_walker(query, contain_outer_selfref_walker, + (void *) depth, QTW_EXAMINE_RTES_BEFORE); + + (*depth)--; + + return result; + } + return expression_tree_walker(node, contain_outer_selfref_walker, + (void *) depth); +} + +/* + * inline_cte: convert RTE_CTE references to given CTE into RTE_SUBQUERYs + */ +static void +inline_cte(PlannerInfo *root, CommonTableExpr *cte) +{ + struct inline_cte_walker_context context; + + context.ctename = cte->ctename; + /* Start at levelsup = -1 because we'll immediately increment it */ + context.levelsup = -1; + context.ctequery = castNode(Query, cte->ctequery); + + (void) inline_cte_walker((Node *) root->parse, &context); +} + +static bool +inline_cte_walker(Node *node, inline_cte_walker_context *context) +{ + if (node == NULL) + return false; + if (IsA(node, Query)) + { + Query *query = (Query *) node; + + context->levelsup++; + + /* + * Visit the query's RTE nodes after their contents; otherwise + * query_tree_walker would descend into the newly inlined CTE query, + * which we don't want. + */ + (void) query_tree_walker(query, inline_cte_walker, context, + QTW_EXAMINE_RTES_AFTER); + + context->levelsup--; + + return false; + } + else if (IsA(node, RangeTblEntry)) + { + RangeTblEntry *rte = (RangeTblEntry *) node; + + if (rte->rtekind == RTE_CTE && + strcmp(rte->ctename, context->ctename) == 0 && + rte->ctelevelsup == context->levelsup) + { + /* + * Found a reference to replace. Generate a copy of the CTE query + * with appropriate level adjustment for outer references (e.g., + * to other CTEs). + */ + Query *newquery = copyObject(context->ctequery); + + if (context->levelsup > 0) + IncrementVarSublevelsUp((Node *) newquery, context->levelsup, 1); + + /* + * Convert the RTE_CTE RTE into a RTE_SUBQUERY. + * + * Historically, a FOR UPDATE clause has been treated as extending + * into views and subqueries, but not into CTEs. We preserve this + * distinction by not trying to push rowmarks into the new + * subquery. + */ + rte->rtekind = RTE_SUBQUERY; + rte->subquery = newquery; + rte->security_barrier = false; + + /* Zero out CTE-specific fields */ + rte->ctename = NULL; + rte->ctelevelsup = 0; + rte->self_reference = false; + rte->coltypes = NIL; + rte->coltypmods = NIL; + rte->colcollations = NIL; + } + + return false; + } + + return expression_tree_walker(node, inline_cte_walker, context); +} + + +/* + * convert_ANY_sublink_to_join: try to convert an ANY SubLink to a join + * + * The caller has found an ANY SubLink at the top level of one of the query's + * qual clauses, but has not checked the properties of the SubLink further. + * Decide whether it is appropriate to process this SubLink in join style. + * If so, form a JoinExpr and return it. Return NULL if the SubLink cannot + * be converted to a join. + * + * The only non-obvious input parameter is available_rels: this is the set + * of query rels that can safely be referenced in the sublink expression. + * (We must restrict this to avoid changing the semantics when a sublink + * is present in an outer join's ON qual.) The conversion must fail if + * the converted qual would reference any but these parent-query relids. + * + * On success, the returned JoinExpr has larg = NULL and rarg = the jointree + * item representing the pulled-up subquery. The caller must set larg to + * represent the relation(s) on the lefthand side of the new join, and insert + * the JoinExpr into the upper query's jointree at an appropriate place + * (typically, where the lefthand relation(s) had been). Note that the + * passed-in SubLink must also be removed from its original position in the + * query quals, since the quals of the returned JoinExpr replace it. + * (Notionally, we replace the SubLink with a constant TRUE, then elide the + * redundant constant from the qual.) + * + * On success, the caller is also responsible for recursively applying + * pull_up_sublinks processing to the rarg and quals of the returned JoinExpr. + * (On failure, there is no need to do anything, since pull_up_sublinks will + * be applied when we recursively plan the sub-select.) + * + * Side effects of a successful conversion include adding the SubLink's + * subselect to the query's rangetable, so that it can be referenced in + * the JoinExpr's rarg. + */ +JoinExpr * +convert_ANY_sublink_to_join(PlannerInfo *root, SubLink *sublink, + Relids available_rels) +{ + JoinExpr *result; + Query *parse = root->parse; + Query *subselect = (Query *) sublink->subselect; + Relids upper_varnos; + int rtindex; + ParseNamespaceItem *nsitem; + RangeTblEntry *rte; + RangeTblRef *rtr; + List *subquery_vars; + Node *quals; + ParseState *pstate; + + Assert(sublink->subLinkType == ANY_SUBLINK); + + /* + * The sub-select must not refer to any Vars of the parent query. (Vars of + * higher levels should be okay, though.) + */ + if (contain_vars_of_level((Node *) subselect, 1)) + return NULL; + + /* + * The test expression must contain some Vars of the parent query, else + * it's not gonna be a join. (Note that it won't have Vars referring to + * the subquery, rather Params.) + */ + upper_varnos = pull_varnos(root, sublink->testexpr); + if (bms_is_empty(upper_varnos)) + return NULL; + + /* + * However, it can't refer to anything outside available_rels. + */ + if (!bms_is_subset(upper_varnos, available_rels)) + return NULL; + + /* + * The combining operators and left-hand expressions mustn't be volatile. + */ + if (contain_volatile_functions(sublink->testexpr)) + return NULL; + + /* Create a dummy ParseState for addRangeTableEntryForSubquery */ + pstate = make_parsestate(NULL); + + /* + * Okay, pull up the sub-select into upper range table. + * + * We rely here on the assumption that the outer query has no references + * to the inner (necessarily true, other than the Vars that we build + * below). Therefore this is a lot easier than what pull_up_subqueries has + * to go through. + */ + nsitem = addRangeTableEntryForSubquery(pstate, + subselect, + makeAlias("ANY_subquery", NIL), + false, + false); + rte = nsitem->p_rte; + parse->rtable = lappend(parse->rtable, rte); + rtindex = list_length(parse->rtable); + + /* + * Form a RangeTblRef for the pulled-up sub-select. + */ + rtr = makeNode(RangeTblRef); + rtr->rtindex = rtindex; + + /* + * Build a list of Vars representing the subselect outputs. + */ + subquery_vars = generate_subquery_vars(root, + subselect->targetList, + rtindex); + + /* + * Build the new join's qual expression, replacing Params with these Vars. + */ + quals = convert_testexpr(root, sublink->testexpr, subquery_vars); + + /* + * And finally, build the JoinExpr node. + */ + result = makeNode(JoinExpr); + result->jointype = JOIN_SEMI; + result->isNatural = false; + result->larg = NULL; /* caller must fill this in */ + result->rarg = (Node *) rtr; + result->usingClause = NIL; + result->join_using_alias = NULL; + result->quals = quals; + result->alias = NULL; + result->rtindex = 0; /* we don't need an RTE for it */ + + return result; +} + +/* + * convert_EXISTS_sublink_to_join: try to convert an EXISTS SubLink to a join + * + * The API of this function is identical to convert_ANY_sublink_to_join's, + * except that we also support the case where the caller has found NOT EXISTS, + * so we need an additional input parameter "under_not". + */ +JoinExpr * +convert_EXISTS_sublink_to_join(PlannerInfo *root, SubLink *sublink, + bool under_not, Relids available_rels) +{ + JoinExpr *result; + Query *parse = root->parse; + Query *subselect = (Query *) sublink->subselect; + Node *whereClause; + int rtoffset; + int varno; + Relids clause_varnos; + Relids upper_varnos; + + Assert(sublink->subLinkType == EXISTS_SUBLINK); + + /* + * Can't flatten if it contains WITH. (We could arrange to pull up the + * WITH into the parent query's cteList, but that risks changing the + * semantics, since a WITH ought to be executed once per associated query + * call.) Note that convert_ANY_sublink_to_join doesn't have to reject + * this case, since it just produces a subquery RTE that doesn't have to + * get flattened into the parent query. + */ + if (subselect->cteList) + return NULL; + + /* + * Copy the subquery so we can modify it safely (see comments in + * make_subplan). + */ + subselect = copyObject(subselect); + + /* + * See if the subquery can be simplified based on the knowledge that it's + * being used in EXISTS(). If we aren't able to get rid of its + * targetlist, we have to fail, because the pullup operation leaves us + * with noplace to evaluate the targetlist. + */ + if (!simplify_EXISTS_query(root, subselect)) + return NULL; + + /* + * Separate out the WHERE clause. (We could theoretically also remove + * top-level plain JOIN/ON clauses, but it's probably not worth the + * trouble.) + */ + whereClause = subselect->jointree->quals; + subselect->jointree->quals = NULL; + + /* + * The rest of the sub-select must not refer to any Vars of the parent + * query. (Vars of higher levels should be okay, though.) + */ + if (contain_vars_of_level((Node *) subselect, 1)) + return NULL; + + /* + * On the other hand, the WHERE clause must contain some Vars of the + * parent query, else it's not gonna be a join. + */ + if (!contain_vars_of_level(whereClause, 1)) + return NULL; + + /* + * We don't risk optimizing if the WHERE clause is volatile, either. + */ + if (contain_volatile_functions(whereClause)) + return NULL; + + /* + * The subquery must have a nonempty jointree, but we can make it so. + */ + replace_empty_jointree(subselect); + + /* + * Prepare to pull up the sub-select into top range table. + * + * We rely here on the assumption that the outer query has no references + * to the inner (necessarily true). Therefore this is a lot easier than + * what pull_up_subqueries has to go through. + * + * In fact, it's even easier than what convert_ANY_sublink_to_join has to + * do. The machinations of simplify_EXISTS_query ensured that there is + * nothing interesting in the subquery except an rtable and jointree, and + * even the jointree FromExpr no longer has quals. So we can just append + * the rtable to our own and use the FromExpr in our jointree. But first, + * adjust all level-zero varnos in the subquery to account for the rtable + * merger. + */ + rtoffset = list_length(parse->rtable); + OffsetVarNodes((Node *) subselect, rtoffset, 0); + OffsetVarNodes(whereClause, rtoffset, 0); + + /* + * Upper-level vars in subquery will now be one level closer to their + * parent than before; in particular, anything that had been level 1 + * becomes level zero. + */ + IncrementVarSublevelsUp((Node *) subselect, -1, 1); + IncrementVarSublevelsUp(whereClause, -1, 1); + + /* + * Now that the WHERE clause is adjusted to match the parent query + * environment, we can easily identify all the level-zero rels it uses. + * The ones <= rtoffset belong to the upper query; the ones > rtoffset do + * not. + */ + clause_varnos = pull_varnos(root, whereClause); + upper_varnos = NULL; + while ((varno = bms_first_member(clause_varnos)) >= 0) + { + if (varno <= rtoffset) + upper_varnos = bms_add_member(upper_varnos, varno); + } + bms_free(clause_varnos); + Assert(!bms_is_empty(upper_varnos)); + + /* + * Now that we've got the set of upper-level varnos, we can make the last + * check: only available_rels can be referenced. + */ + if (!bms_is_subset(upper_varnos, available_rels)) + return NULL; + + /* Now we can attach the modified subquery rtable to the parent */ + parse->rtable = list_concat(parse->rtable, subselect->rtable); + + /* + * And finally, build the JoinExpr node. + */ + result = makeNode(JoinExpr); + result->jointype = under_not ? JOIN_ANTI : JOIN_SEMI; + result->isNatural = false; + result->larg = NULL; /* caller must fill this in */ + /* flatten out the FromExpr node if it's useless */ + if (list_length(subselect->jointree->fromlist) == 1) + result->rarg = (Node *) linitial(subselect->jointree->fromlist); + else + result->rarg = (Node *) subselect->jointree; + result->usingClause = NIL; + result->join_using_alias = NULL; + result->quals = whereClause; + result->alias = NULL; + result->rtindex = 0; /* we don't need an RTE for it */ + + return result; +} + +/* + * simplify_EXISTS_query: remove any useless stuff in an EXISTS's subquery + * + * The only thing that matters about an EXISTS query is whether it returns + * zero or more than zero rows. Therefore, we can remove certain SQL features + * that won't affect that. The only part that is really likely to matter in + * typical usage is simplifying the targetlist: it's a common habit to write + * "SELECT * FROM" even though there is no need to evaluate any columns. + * + * Note: by suppressing the targetlist we could cause an observable behavioral + * change, namely that any errors that might occur in evaluating the tlist + * won't occur, nor will other side-effects of volatile functions. This seems + * unlikely to bother anyone in practice. + * + * Returns true if was able to discard the targetlist, else false. + */ +static bool +simplify_EXISTS_query(PlannerInfo *root, Query *query) +{ + /* + * We don't try to simplify at all if the query uses set operations, + * aggregates, grouping sets, SRFs, modifying CTEs, HAVING, OFFSET, or FOR + * UPDATE/SHARE; none of these seem likely in normal usage and their + * possible effects are complex. (Note: we could ignore an "OFFSET 0" + * clause, but that traditionally is used as an optimization fence, so we + * don't.) + */ + if (query->commandType != CMD_SELECT || + query->setOperations || + query->hasAggs || + query->groupingSets || + query->hasWindowFuncs || + query->hasTargetSRFs || + query->hasModifyingCTE || + query->havingQual || + query->limitOffset || + query->rowMarks) + return false; + + /* + * LIMIT with a constant positive (or NULL) value doesn't affect the + * semantics of EXISTS, so let's ignore such clauses. This is worth doing + * because people accustomed to certain other DBMSes may be in the habit + * of writing EXISTS(SELECT ... LIMIT 1) as an optimization. If there's a + * LIMIT with anything else as argument, though, we can't simplify. + */ + if (query->limitCount) + { + /* + * The LIMIT clause has not yet been through eval_const_expressions, + * so we have to apply that here. It might seem like this is a waste + * of cycles, since the only case plausibly worth worrying about is + * "LIMIT 1" ... but what we'll actually see is "LIMIT int8(1::int4)", + * so we have to fold constants or we're not going to recognize it. + */ + Node *node = eval_const_expressions(root, query->limitCount); + Const *limit; + + /* Might as well update the query if we simplified the clause. */ + query->limitCount = node; + + if (!IsA(node, Const)) + return false; + + limit = (Const *) node; + Assert(limit->consttype == INT8OID); + if (!limit->constisnull && DatumGetInt64(limit->constvalue) <= 0) + return false; + + /* Whether or not the targetlist is safe, we can drop the LIMIT. */ + query->limitCount = NULL; + } + + /* + * Otherwise, we can throw away the targetlist, as well as any GROUP, + * WINDOW, DISTINCT, and ORDER BY clauses; none of those clauses will + * change a nonzero-rows result to zero rows or vice versa. (Furthermore, + * since our parsetree representation of these clauses depends on the + * targetlist, we'd better throw them away if we drop the targetlist.) + */ + query->targetList = NIL; + query->groupClause = NIL; + query->windowClause = NIL; + query->distinctClause = NIL; + query->sortClause = NIL; + query->hasDistinctOn = false; + + return true; +} + +/* + * convert_EXISTS_to_ANY: try to convert EXISTS to a hashable ANY sublink + * + * The subselect is expected to be a fresh copy that we can munge up, + * and to have been successfully passed through simplify_EXISTS_query. + * + * On success, the modified subselect is returned, and we store a suitable + * upper-level test expression at *testexpr, plus a list of the subselect's + * output Params at *paramIds. (The test expression is already Param-ified + * and hence need not go through convert_testexpr, which is why we have to + * deal with the Param IDs specially.) + * + * On failure, returns NULL. + */ +static Query * +convert_EXISTS_to_ANY(PlannerInfo *root, Query *subselect, + Node **testexpr, List **paramIds) +{ + Node *whereClause; + List *leftargs, + *rightargs, + *opids, + *opcollations, + *newWhere, + *tlist, + *testlist, + *paramids; + ListCell *lc, + *rc, + *oc, + *cc; + AttrNumber resno; + + /* + * Query must not require a targetlist, since we have to insert a new one. + * Caller should have dealt with the case already. + */ + Assert(subselect->targetList == NIL); + + /* + * Separate out the WHERE clause. (We could theoretically also remove + * top-level plain JOIN/ON clauses, but it's probably not worth the + * trouble.) + */ + whereClause = subselect->jointree->quals; + subselect->jointree->quals = NULL; + + /* + * The rest of the sub-select must not refer to any Vars of the parent + * query. (Vars of higher levels should be okay, though.) + * + * Note: we need not check for Aggrefs separately because we know the + * sub-select is as yet unoptimized; any uplevel Aggref must therefore + * contain an uplevel Var reference. This is not the case below ... + */ + if (contain_vars_of_level((Node *) subselect, 1)) + return NULL; + + /* + * We don't risk optimizing if the WHERE clause is volatile, either. + */ + if (contain_volatile_functions(whereClause)) + return NULL; + + /* + * Clean up the WHERE clause by doing const-simplification etc on it. + * Aside from simplifying the processing we're about to do, this is + * important for being able to pull chunks of the WHERE clause up into the + * parent query. Since we are invoked partway through the parent's + * preprocess_expression() work, earlier steps of preprocess_expression() + * wouldn't get applied to the pulled-up stuff unless we do them here. For + * the parts of the WHERE clause that get put back into the child query, + * this work is partially duplicative, but it shouldn't hurt. + * + * Note: we do not run flatten_join_alias_vars. This is OK because any + * parent aliases were flattened already, and we're not going to pull any + * child Vars (of any description) into the parent. + * + * Note: passing the parent's root to eval_const_expressions is + * technically wrong, but we can get away with it since only the + * boundParams (if any) are used, and those would be the same in a + * subroot. + */ + whereClause = eval_const_expressions(root, whereClause); + whereClause = (Node *) canonicalize_qual((Expr *) whereClause, false); + whereClause = (Node *) make_ands_implicit((Expr *) whereClause); + + /* + * We now have a flattened implicit-AND list of clauses, which we try to + * break apart into "outervar = innervar" hash clauses. Anything that + * can't be broken apart just goes back into the newWhere list. Note that + * we aren't trying hard yet to ensure that we have only outer or only + * inner on each side; we'll check that if we get to the end. + */ + leftargs = rightargs = opids = opcollations = newWhere = NIL; + foreach(lc, (List *) whereClause) + { + OpExpr *expr = (OpExpr *) lfirst(lc); + + if (IsA(expr, OpExpr) && + hash_ok_operator(expr)) + { + Node *leftarg = (Node *) linitial(expr->args); + Node *rightarg = (Node *) lsecond(expr->args); + + if (contain_vars_of_level(leftarg, 1)) + { + leftargs = lappend(leftargs, leftarg); + rightargs = lappend(rightargs, rightarg); + opids = lappend_oid(opids, expr->opno); + opcollations = lappend_oid(opcollations, expr->inputcollid); + continue; + } + if (contain_vars_of_level(rightarg, 1)) + { + /* + * We must commute the clause to put the outer var on the + * left, because the hashing code in nodeSubplan.c expects + * that. This probably shouldn't ever fail, since hashable + * operators ought to have commutators, but be paranoid. + */ + expr->opno = get_commutator(expr->opno); + if (OidIsValid(expr->opno) && hash_ok_operator(expr)) + { + leftargs = lappend(leftargs, rightarg); + rightargs = lappend(rightargs, leftarg); + opids = lappend_oid(opids, expr->opno); + opcollations = lappend_oid(opcollations, expr->inputcollid); + continue; + } + /* If no commutator, no chance to optimize the WHERE clause */ + return NULL; + } + } + /* Couldn't handle it as a hash clause */ + newWhere = lappend(newWhere, expr); + } + + /* + * If we didn't find anything we could convert, fail. + */ + if (leftargs == NIL) + return NULL; + + /* + * There mustn't be any parent Vars or Aggs in the stuff that we intend to + * put back into the child query. Note: you might think we don't need to + * check for Aggs separately, because an uplevel Agg must contain an + * uplevel Var in its argument. But it is possible that the uplevel Var + * got optimized away by eval_const_expressions. Consider + * + * SUM(CASE WHEN false THEN uplevelvar ELSE 0 END) + */ + if (contain_vars_of_level((Node *) newWhere, 1) || + contain_vars_of_level((Node *) rightargs, 1)) + return NULL; + if (root->parse->hasAggs && + (contain_aggs_of_level((Node *) newWhere, 1) || + contain_aggs_of_level((Node *) rightargs, 1))) + return NULL; + + /* + * And there can't be any child Vars in the stuff we intend to pull up. + * (Note: we'd need to check for child Aggs too, except we know the child + * has no aggs at all because of simplify_EXISTS_query's check. The same + * goes for window functions.) + */ + if (contain_vars_of_level((Node *) leftargs, 0)) + return NULL; + + /* + * Also reject sublinks in the stuff we intend to pull up. (It might be + * possible to support this, but doesn't seem worth the complication.) + */ + if (contain_subplans((Node *) leftargs)) + return NULL; + + /* + * Okay, adjust the sublevelsup in the stuff we're pulling up. + */ + IncrementVarSublevelsUp((Node *) leftargs, -1, 1); + + /* + * Put back any child-level-only WHERE clauses. + */ + if (newWhere) + subselect->jointree->quals = (Node *) make_ands_explicit(newWhere); + + /* + * Build a new targetlist for the child that emits the expressions we + * need. Concurrently, build a testexpr for the parent using Params to + * reference the child outputs. (Since we generate Params directly here, + * there will be no need to convert the testexpr in build_subplan.) + */ + tlist = testlist = paramids = NIL; + resno = 1; + forfour(lc, leftargs, rc, rightargs, oc, opids, cc, opcollations) + { + Node *leftarg = (Node *) lfirst(lc); + Node *rightarg = (Node *) lfirst(rc); + Oid opid = lfirst_oid(oc); + Oid opcollation = lfirst_oid(cc); + Param *param; + + param = generate_new_exec_param(root, + exprType(rightarg), + exprTypmod(rightarg), + exprCollation(rightarg)); + tlist = lappend(tlist, + makeTargetEntry((Expr *) rightarg, + resno++, + NULL, + false)); + testlist = lappend(testlist, + make_opclause(opid, BOOLOID, false, + (Expr *) leftarg, (Expr *) param, + InvalidOid, opcollation)); + paramids = lappend_int(paramids, param->paramid); + } + + /* Put everything where it should go, and we're done */ + subselect->targetList = tlist; + *testexpr = (Node *) make_ands_explicit(testlist); + *paramIds = paramids; + + return subselect; +} + + +/* + * Replace correlation vars (uplevel vars) with Params. + * + * Uplevel PlaceHolderVars and aggregates are replaced, too. + * + * Note: it is critical that this runs immediately after SS_process_sublinks. + * Since we do not recurse into the arguments of uplevel PHVs and aggregates, + * they will get copied to the appropriate subplan args list in the parent + * query with uplevel vars not replaced by Params, but only adjusted in level + * (see replace_outer_placeholdervar and replace_outer_agg). That's exactly + * what we want for the vars of the parent level --- but if a PHV's or + * aggregate's argument contains any further-up variables, they have to be + * replaced with Params in their turn. That will happen when the parent level + * runs SS_replace_correlation_vars. Therefore it must do so after expanding + * its sublinks to subplans. And we don't want any steps in between, else + * those steps would never get applied to the argument expressions, either in + * the parent or the child level. + * + * Another fairly tricky thing going on here is the handling of SubLinks in + * the arguments of uplevel PHVs/aggregates. Those are not touched inside the + * intermediate query level, either. Instead, SS_process_sublinks recurses on + * them after copying the PHV or Aggref expression into the parent plan level + * (this is actually taken care of in build_subplan). + */ +Node * +SS_replace_correlation_vars(PlannerInfo *root, Node *expr) +{ + /* No setup needed for tree walk, so away we go */ + return replace_correlation_vars_mutator(expr, root); +} + +static Node * +replace_correlation_vars_mutator(Node *node, PlannerInfo *root) +{ + if (node == NULL) + return NULL; + if (IsA(node, Var)) + { + if (((Var *) node)->varlevelsup > 0) + return (Node *) replace_outer_var(root, (Var *) node); + } + if (IsA(node, PlaceHolderVar)) + { + if (((PlaceHolderVar *) node)->phlevelsup > 0) + return (Node *) replace_outer_placeholdervar(root, + (PlaceHolderVar *) node); + } + if (IsA(node, Aggref)) + { + if (((Aggref *) node)->agglevelsup > 0) + return (Node *) replace_outer_agg(root, (Aggref *) node); + } + if (IsA(node, GroupingFunc)) + { + if (((GroupingFunc *) node)->agglevelsup > 0) + return (Node *) replace_outer_grouping(root, (GroupingFunc *) node); + } + return expression_tree_mutator(node, + replace_correlation_vars_mutator, + (void *) root); +} + +/* + * Expand SubLinks to SubPlans in the given expression. + * + * The isQual argument tells whether or not this expression is a WHERE/HAVING + * qualifier expression. If it is, any sublinks appearing at top level need + * not distinguish FALSE from UNKNOWN return values. + */ +Node * +SS_process_sublinks(PlannerInfo *root, Node *expr, bool isQual) +{ + process_sublinks_context context; + + context.root = root; + context.isTopQual = isQual; + return process_sublinks_mutator(expr, &context); +} + +static Node * +process_sublinks_mutator(Node *node, process_sublinks_context *context) +{ + process_sublinks_context locContext; + + locContext.root = context->root; + + if (node == NULL) + return NULL; + if (IsA(node, SubLink)) + { + SubLink *sublink = (SubLink *) node; + Node *testexpr; + + /* + * First, recursively process the lefthand-side expressions, if any. + * They're not top-level anymore. + */ + locContext.isTopQual = false; + testexpr = process_sublinks_mutator(sublink->testexpr, &locContext); + + /* + * Now build the SubPlan node and make the expr to return. + */ + return make_subplan(context->root, + (Query *) sublink->subselect, + sublink->subLinkType, + sublink->subLinkId, + testexpr, + context->isTopQual); + } + + /* + * Don't recurse into the arguments of an outer PHV, Aggref or + * GroupingFunc here. Any SubLinks in the arguments have to be dealt with + * at the outer query level; they'll be handled when build_subplan + * collects the PHV, Aggref or GroupingFunc into the arguments to be + * passed down to the current subplan. + */ + if (IsA(node, PlaceHolderVar)) + { + if (((PlaceHolderVar *) node)->phlevelsup > 0) + return node; + } + else if (IsA(node, Aggref)) + { + if (((Aggref *) node)->agglevelsup > 0) + return node; + } + else if (IsA(node, GroupingFunc)) + { + if (((GroupingFunc *) node)->agglevelsup > 0) + return node; + } + + /* + * We should never see a SubPlan expression in the input (since this is + * the very routine that creates 'em to begin with). We shouldn't find + * ourselves invoked directly on a Query, either. + */ + Assert(!IsA(node, SubPlan)); + Assert(!IsA(node, AlternativeSubPlan)); + Assert(!IsA(node, Query)); + + /* + * Because make_subplan() could return an AND or OR clause, we have to + * take steps to preserve AND/OR flatness of a qual. We assume the input + * has been AND/OR flattened and so we need no recursion here. + * + * (Due to the coding here, we will not get called on the List subnodes of + * an AND; and the input is *not* yet in implicit-AND format. So no check + * is needed for a bare List.) + * + * Anywhere within the top-level AND/OR clause structure, we can tell + * make_subplan() that NULL and FALSE are interchangeable. So isTopQual + * propagates down in both cases. (Note that this is unlike the meaning + * of "top level qual" used in most other places in Postgres.) + */ + if (is_andclause(node)) + { + List *newargs = NIL; + ListCell *l; + + /* Still at qual top-level */ + locContext.isTopQual = context->isTopQual; + + foreach(l, ((BoolExpr *) node)->args) + { + Node *newarg; + + newarg = process_sublinks_mutator(lfirst(l), &locContext); + if (is_andclause(newarg)) + newargs = list_concat(newargs, ((BoolExpr *) newarg)->args); + else + newargs = lappend(newargs, newarg); + } + return (Node *) make_andclause(newargs); + } + + if (is_orclause(node)) + { + List *newargs = NIL; + ListCell *l; + + /* Still at qual top-level */ + locContext.isTopQual = context->isTopQual; + + foreach(l, ((BoolExpr *) node)->args) + { + Node *newarg; + + newarg = process_sublinks_mutator(lfirst(l), &locContext); + if (is_orclause(newarg)) + newargs = list_concat(newargs, ((BoolExpr *) newarg)->args); + else + newargs = lappend(newargs, newarg); + } + return (Node *) make_orclause(newargs); + } + + /* + * If we recurse down through anything other than an AND or OR node, we + * are definitely not at top qual level anymore. + */ + locContext.isTopQual = false; + + return expression_tree_mutator(node, + process_sublinks_mutator, + (void *) &locContext); +} + +/* + * SS_identify_outer_params - identify the Params available from outer levels + * + * This must be run after SS_replace_correlation_vars and SS_process_sublinks + * processing is complete in a given query level as well as all of its + * descendant levels (which means it's most practical to do it at the end of + * processing the query level). We compute the set of paramIds that outer + * levels will make available to this level+descendants, and record it in + * root->outer_params for use while computing extParam/allParam sets in final + * plan cleanup. (We can't just compute it then, because the upper levels' + * plan_params lists are transient and will be gone by then.) + */ +void +SS_identify_outer_params(PlannerInfo *root) +{ + Bitmapset *outer_params; + PlannerInfo *proot; + ListCell *l; + + /* + * If no parameters have been assigned anywhere in the tree, we certainly + * don't need to do anything here. + */ + if (root->glob->paramExecTypes == NIL) + return; + + /* + * Scan all query levels above this one to see which parameters are due to + * be available from them, either because lower query levels have + * requested them (via plan_params) or because they will be available from + * initPlans of those levels. + */ + outer_params = NULL; + for (proot = root->parent_root; proot != NULL; proot = proot->parent_root) + { + /* Include ordinary Var/PHV/Aggref/GroupingFunc params */ + foreach(l, proot->plan_params) + { + PlannerParamItem *pitem = (PlannerParamItem *) lfirst(l); + + outer_params = bms_add_member(outer_params, pitem->paramId); + } + /* Include any outputs of outer-level initPlans */ + foreach(l, proot->init_plans) + { + SubPlan *initsubplan = (SubPlan *) lfirst(l); + ListCell *l2; + + foreach(l2, initsubplan->setParam) + { + outer_params = bms_add_member(outer_params, lfirst_int(l2)); + } + } + /* Include worktable ID, if a recursive query is being planned */ + if (proot->wt_param_id >= 0) + outer_params = bms_add_member(outer_params, proot->wt_param_id); + } + root->outer_params = outer_params; +} + +/* + * SS_charge_for_initplans - account for initplans in Path costs & parallelism + * + * If any initPlans have been created in the current query level, they will + * get attached to the Plan tree created from whichever Path we select from + * the given rel. Increment all that rel's Paths' costs to account for them, + * and make sure the paths get marked as parallel-unsafe, since we can't + * currently transmit initPlans to parallel workers. + * + * This is separate from SS_attach_initplans because we might conditionally + * create more initPlans during create_plan(), depending on which Path we + * select. However, Paths that would generate such initPlans are expected + * to have included their cost and parallel-safety effects already. + */ +void +SS_charge_for_initplans(PlannerInfo *root, RelOptInfo *final_rel) +{ + Cost initplan_cost; + ListCell *lc; + + /* Nothing to do if no initPlans */ + if (root->init_plans == NIL) + return; + + /* + * Compute the cost increment just once, since it will be the same for all + * Paths. We assume each initPlan gets run once during top plan startup. + * This is a conservative overestimate, since in fact an initPlan might be + * executed later than plan startup, or even not at all. + */ + initplan_cost = 0; + foreach(lc, root->init_plans) + { + SubPlan *initsubplan = (SubPlan *) lfirst(lc); + + initplan_cost += initsubplan->startup_cost + initsubplan->per_call_cost; + } + + /* + * Now adjust the costs and parallel_safe flags. + */ + foreach(lc, final_rel->pathlist) + { + Path *path = (Path *) lfirst(lc); + + path->startup_cost += initplan_cost; + path->total_cost += initplan_cost; + path->parallel_safe = false; + } + + /* + * Forget about any partial paths and clear consider_parallel, too; + * they're not usable if we attached an initPlan. + */ + final_rel->partial_pathlist = NIL; + final_rel->consider_parallel = false; + + /* We needn't do set_cheapest() here, caller will do it */ +} + +/* + * SS_attach_initplans - attach initplans to topmost plan node + * + * Attach any initplans created in the current query level to the specified + * plan node, which should normally be the topmost node for the query level. + * (In principle the initPlans could go in any node at or above where they're + * referenced; but there seems no reason to put them any lower than the + * topmost node, so we don't bother to track exactly where they came from.) + * + * We do not touch the plan node's cost or parallel_safe flag. The initplans + * must have been accounted for in SS_charge_for_initplans, or by any later + * code that adds initplans via SS_make_initplan_from_plan. + */ +void +SS_attach_initplans(PlannerInfo *root, Plan *plan) +{ + plan->initPlan = root->init_plans; +} + +/* + * SS_finalize_plan - do final parameter processing for a completed Plan. + * + * This recursively computes the extParam and allParam sets for every Plan + * node in the given plan tree. (Oh, and RangeTblFunction.funcparams too.) + * + * We assume that SS_finalize_plan has already been run on any initplans or + * subplans the plan tree could reference. + */ +void +SS_finalize_plan(PlannerInfo *root, Plan *plan) +{ + /* No setup needed, just recurse through plan tree. */ + (void) finalize_plan(root, plan, -1, root->outer_params, NULL); +} + +/* + * Recursive processing of all nodes in the plan tree + * + * gather_param is the rescan_param of an ancestral Gather/GatherMerge, + * or -1 if there is none. + * + * valid_params is the set of param IDs supplied by outer plan levels + * that are valid to reference in this plan node or its children. + * + * scan_params is a set of param IDs to force scan plan nodes to reference. + * This is for EvalPlanQual support, and is always NULL at the top of the + * recursion. + * + * The return value is the computed allParam set for the given Plan node. + * This is just an internal notational convenience: we can add a child + * plan's allParams to the set of param IDs of interest to this level + * in the same statement that recurses to that child. + * + * Do not scribble on caller's values of valid_params or scan_params! + * + * Note: although we attempt to deal with initPlans anywhere in the tree, the + * logic is not really right. The problem is that a plan node might return an + * output Param of its initPlan as a targetlist item, in which case it's valid + * for the parent plan level to reference that same Param; the parent's usage + * will be converted into a Var referencing the child plan node by setrefs.c. + * But this function would see the parent's reference as out of scope and + * complain about it. For now, this does not matter because the planner only + * attaches initPlans to the topmost plan node in a query level, so the case + * doesn't arise. If we ever merge this processing into setrefs.c, maybe it + * can be handled more cleanly. + */ +static Bitmapset * +finalize_plan(PlannerInfo *root, Plan *plan, + int gather_param, + Bitmapset *valid_params, + Bitmapset *scan_params) +{ + finalize_primnode_context context; + int locally_added_param; + Bitmapset *nestloop_params; + Bitmapset *initExtParam; + Bitmapset *initSetParam; + Bitmapset *child_params; + ListCell *l; + + if (plan == NULL) + return NULL; + + context.root = root; + context.paramids = NULL; /* initialize set to empty */ + locally_added_param = -1; /* there isn't one */ + nestloop_params = NULL; /* there aren't any */ + + /* + * Examine any initPlans to determine the set of external params they + * reference and the set of output params they supply. (We assume + * SS_finalize_plan was run on them already.) + */ + initExtParam = initSetParam = NULL; + foreach(l, plan->initPlan) + { + SubPlan *initsubplan = (SubPlan *) lfirst(l); + Plan *initplan = planner_subplan_get_plan(root, initsubplan); + ListCell *l2; + + initExtParam = bms_add_members(initExtParam, initplan->extParam); + foreach(l2, initsubplan->setParam) + { + initSetParam = bms_add_member(initSetParam, lfirst_int(l2)); + } + } + + /* Any setParams are validly referenceable in this node and children */ + if (initSetParam) + valid_params = bms_union(valid_params, initSetParam); + + /* + * When we call finalize_primnode, context.paramids sets are automatically + * merged together. But when recursing to self, we have to do it the hard + * way. We want the paramids set to include params in subplans as well as + * at this level. + */ + + /* Find params in targetlist and qual */ + finalize_primnode((Node *) plan->targetlist, &context); + finalize_primnode((Node *) plan->qual, &context); + + /* + * If it's a parallel-aware scan node, mark it as dependent on the parent + * Gather/GatherMerge's rescan Param. + */ + if (plan->parallel_aware) + { + if (gather_param < 0) + elog(ERROR, "parallel-aware plan node is not below a Gather"); + context.paramids = + bms_add_member(context.paramids, gather_param); + } + + /* Check additional node-type-specific fields */ + switch (nodeTag(plan)) + { + case T_Result: + finalize_primnode(((Result *) plan)->resconstantqual, + &context); + break; + + case T_SeqScan: + context.paramids = bms_add_members(context.paramids, scan_params); + break; + + case T_SampleScan: + finalize_primnode((Node *) ((SampleScan *) plan)->tablesample, + &context); + context.paramids = bms_add_members(context.paramids, scan_params); + break; + + case T_IndexScan: + finalize_primnode((Node *) ((IndexScan *) plan)->indexqual, + &context); + finalize_primnode((Node *) ((IndexScan *) plan)->indexorderby, + &context); + + /* + * we need not look at indexqualorig, since it will have the same + * param references as indexqual. Likewise, we can ignore + * indexorderbyorig. + */ + context.paramids = bms_add_members(context.paramids, scan_params); + break; + + case T_IndexOnlyScan: + finalize_primnode((Node *) ((IndexOnlyScan *) plan)->indexqual, + &context); + finalize_primnode((Node *) ((IndexOnlyScan *) plan)->recheckqual, + &context); + finalize_primnode((Node *) ((IndexOnlyScan *) plan)->indexorderby, + &context); + + /* + * we need not look at indextlist, since it cannot contain Params. + */ + context.paramids = bms_add_members(context.paramids, scan_params); + break; + + case T_BitmapIndexScan: + finalize_primnode((Node *) ((BitmapIndexScan *) plan)->indexqual, + &context); + + /* + * we need not look at indexqualorig, since it will have the same + * param references as indexqual. + */ + break; + + case T_BitmapHeapScan: + finalize_primnode((Node *) ((BitmapHeapScan *) plan)->bitmapqualorig, + &context); + context.paramids = bms_add_members(context.paramids, scan_params); + break; + + case T_TidScan: + finalize_primnode((Node *) ((TidScan *) plan)->tidquals, + &context); + context.paramids = bms_add_members(context.paramids, scan_params); + break; + + case T_TidRangeScan: + finalize_primnode((Node *) ((TidRangeScan *) plan)->tidrangequals, + &context); + context.paramids = bms_add_members(context.paramids, scan_params); + break; + + case T_SubqueryScan: + { + SubqueryScan *sscan = (SubqueryScan *) plan; + RelOptInfo *rel; + Bitmapset *subquery_params; + + /* We must run finalize_plan on the subquery */ + rel = find_base_rel(root, sscan->scan.scanrelid); + subquery_params = rel->subroot->outer_params; + if (gather_param >= 0) + subquery_params = bms_add_member(bms_copy(subquery_params), + gather_param); + finalize_plan(rel->subroot, sscan->subplan, gather_param, + subquery_params, NULL); + + /* Now we can add its extParams to the parent's params */ + context.paramids = bms_add_members(context.paramids, + sscan->subplan->extParam); + /* We need scan_params too, though */ + context.paramids = bms_add_members(context.paramids, + scan_params); + } + break; + + case T_FunctionScan: + { + FunctionScan *fscan = (FunctionScan *) plan; + ListCell *lc; + + /* + * Call finalize_primnode independently on each function + * expression, so that we can record which params are + * referenced in each, in order to decide which need + * re-evaluating during rescan. + */ + foreach(lc, fscan->functions) + { + RangeTblFunction *rtfunc = (RangeTblFunction *) lfirst(lc); + finalize_primnode_context funccontext; + + funccontext = context; + funccontext.paramids = NULL; + + finalize_primnode(rtfunc->funcexpr, &funccontext); + + /* remember results for execution */ + rtfunc->funcparams = funccontext.paramids; + + /* add the function's params to the overall set */ + context.paramids = bms_add_members(context.paramids, + funccontext.paramids); + } + + context.paramids = bms_add_members(context.paramids, + scan_params); + } + break; + + case T_TableFuncScan: + finalize_primnode((Node *) ((TableFuncScan *) plan)->tablefunc, + &context); + context.paramids = bms_add_members(context.paramids, scan_params); + break; + + case T_ValuesScan: + finalize_primnode((Node *) ((ValuesScan *) plan)->values_lists, + &context); + context.paramids = bms_add_members(context.paramids, scan_params); + break; + + case T_CteScan: + { + /* + * You might think we should add the node's cteParam to + * paramids, but we shouldn't because that param is just a + * linkage mechanism for multiple CteScan nodes for the same + * CTE; it is never used for changed-param signaling. What we + * have to do instead is to find the referenced CTE plan and + * incorporate its external paramids, so that the correct + * things will happen if the CTE references outer-level + * variables. See test cases for bug #4902. (We assume + * SS_finalize_plan was run on the CTE plan already.) + */ + int plan_id = ((CteScan *) plan)->ctePlanId; + Plan *cteplan; + + /* so, do this ... */ + if (plan_id < 1 || plan_id > list_length(root->glob->subplans)) + elog(ERROR, "could not find plan for CteScan referencing plan ID %d", + plan_id); + cteplan = (Plan *) list_nth(root->glob->subplans, plan_id - 1); + context.paramids = + bms_add_members(context.paramids, cteplan->extParam); + +#ifdef NOT_USED + /* ... but not this */ + context.paramids = + bms_add_member(context.paramids, + ((CteScan *) plan)->cteParam); +#endif + + context.paramids = bms_add_members(context.paramids, + scan_params); + } + break; + + case T_WorkTableScan: + context.paramids = + bms_add_member(context.paramids, + ((WorkTableScan *) plan)->wtParam); + context.paramids = bms_add_members(context.paramids, scan_params); + break; + + case T_NamedTuplestoreScan: + context.paramids = bms_add_members(context.paramids, scan_params); + break; + + case T_ForeignScan: + { + ForeignScan *fscan = (ForeignScan *) plan; + + finalize_primnode((Node *) fscan->fdw_exprs, + &context); + finalize_primnode((Node *) fscan->fdw_recheck_quals, + &context); + + /* We assume fdw_scan_tlist cannot contain Params */ + context.paramids = bms_add_members(context.paramids, + scan_params); + } + break; + + case T_CustomScan: + { + CustomScan *cscan = (CustomScan *) plan; + ListCell *lc; + + finalize_primnode((Node *) cscan->custom_exprs, + &context); + /* We assume custom_scan_tlist cannot contain Params */ + context.paramids = + bms_add_members(context.paramids, scan_params); + + /* child nodes if any */ + foreach(lc, cscan->custom_plans) + { + context.paramids = + bms_add_members(context.paramids, + finalize_plan(root, + (Plan *) lfirst(lc), + gather_param, + valid_params, + scan_params)); + } + } + break; + + case T_ModifyTable: + { + ModifyTable *mtplan = (ModifyTable *) plan; + + /* Force descendant scan nodes to reference epqParam */ + locally_added_param = mtplan->epqParam; + valid_params = bms_add_member(bms_copy(valid_params), + locally_added_param); + scan_params = bms_add_member(bms_copy(scan_params), + locally_added_param); + finalize_primnode((Node *) mtplan->returningLists, + &context); + finalize_primnode((Node *) mtplan->onConflictSet, + &context); + finalize_primnode((Node *) mtplan->onConflictWhere, + &context); + /* exclRelTlist contains only Vars, doesn't need examination */ + } + break; + + case T_Append: + { + ListCell *l; + + foreach(l, ((Append *) plan)->appendplans) + { + context.paramids = + bms_add_members(context.paramids, + finalize_plan(root, + (Plan *) lfirst(l), + gather_param, + valid_params, + scan_params)); + } + } + break; + + case T_MergeAppend: + { + ListCell *l; + + foreach(l, ((MergeAppend *) plan)->mergeplans) + { + context.paramids = + bms_add_members(context.paramids, + finalize_plan(root, + (Plan *) lfirst(l), + gather_param, + valid_params, + scan_params)); + } + } + break; + + case T_BitmapAnd: + { + ListCell *l; + + foreach(l, ((BitmapAnd *) plan)->bitmapplans) + { + context.paramids = + bms_add_members(context.paramids, + finalize_plan(root, + (Plan *) lfirst(l), + gather_param, + valid_params, + scan_params)); + } + } + break; + + case T_BitmapOr: + { + ListCell *l; + + foreach(l, ((BitmapOr *) plan)->bitmapplans) + { + context.paramids = + bms_add_members(context.paramids, + finalize_plan(root, + (Plan *) lfirst(l), + gather_param, + valid_params, + scan_params)); + } + } + break; + + case T_NestLoop: + { + ListCell *l; + + finalize_primnode((Node *) ((Join *) plan)->joinqual, + &context); + /* collect set of params that will be passed to right child */ + foreach(l, ((NestLoop *) plan)->nestParams) + { + NestLoopParam *nlp = (NestLoopParam *) lfirst(l); + + nestloop_params = bms_add_member(nestloop_params, + nlp->paramno); + } + } + break; + + case T_MergeJoin: + finalize_primnode((Node *) ((Join *) plan)->joinqual, + &context); + finalize_primnode((Node *) ((MergeJoin *) plan)->mergeclauses, + &context); + break; + + case T_HashJoin: + finalize_primnode((Node *) ((Join *) plan)->joinqual, + &context); + finalize_primnode((Node *) ((HashJoin *) plan)->hashclauses, + &context); + break; + + case T_Hash: + finalize_primnode((Node *) ((Hash *) plan)->hashkeys, + &context); + break; + + case T_Limit: + finalize_primnode(((Limit *) plan)->limitOffset, + &context); + finalize_primnode(((Limit *) plan)->limitCount, + &context); + break; + + case T_RecursiveUnion: + /* child nodes are allowed to reference wtParam */ + locally_added_param = ((RecursiveUnion *) plan)->wtParam; + valid_params = bms_add_member(bms_copy(valid_params), + locally_added_param); + /* wtParam does *not* get added to scan_params */ + break; + + case T_LockRows: + /* Force descendant scan nodes to reference epqParam */ + locally_added_param = ((LockRows *) plan)->epqParam; + valid_params = bms_add_member(bms_copy(valid_params), + locally_added_param); + scan_params = bms_add_member(bms_copy(scan_params), + locally_added_param); + break; + + case T_Agg: + { + Agg *agg = (Agg *) plan; + + /* + * AGG_HASHED plans need to know which Params are referenced + * in aggregate calls. Do a separate scan to identify them. + */ + if (agg->aggstrategy == AGG_HASHED) + { + finalize_primnode_context aggcontext; + + aggcontext.root = root; + aggcontext.paramids = NULL; + finalize_agg_primnode((Node *) agg->plan.targetlist, + &aggcontext); + finalize_agg_primnode((Node *) agg->plan.qual, + &aggcontext); + agg->aggParams = aggcontext.paramids; + } + } + break; + + case T_WindowAgg: + finalize_primnode(((WindowAgg *) plan)->startOffset, + &context); + finalize_primnode(((WindowAgg *) plan)->endOffset, + &context); + break; + + case T_Gather: + /* child nodes are allowed to reference rescan_param, if any */ + locally_added_param = ((Gather *) plan)->rescan_param; + if (locally_added_param >= 0) + { + valid_params = bms_add_member(bms_copy(valid_params), + locally_added_param); + + /* + * We currently don't support nested Gathers. The issue so + * far as this function is concerned would be how to identify + * which child nodes depend on which Gather. + */ + Assert(gather_param < 0); + /* Pass down rescan_param to child parallel-aware nodes */ + gather_param = locally_added_param; + } + /* rescan_param does *not* get added to scan_params */ + break; + + case T_GatherMerge: + /* child nodes are allowed to reference rescan_param, if any */ + locally_added_param = ((GatherMerge *) plan)->rescan_param; + if (locally_added_param >= 0) + { + valid_params = bms_add_member(bms_copy(valid_params), + locally_added_param); + + /* + * We currently don't support nested Gathers. The issue so + * far as this function is concerned would be how to identify + * which child nodes depend on which Gather. + */ + Assert(gather_param < 0); + /* Pass down rescan_param to child parallel-aware nodes */ + gather_param = locally_added_param; + } + /* rescan_param does *not* get added to scan_params */ + break; + + case T_Memoize: + finalize_primnode((Node *) ((Memoize *) plan)->param_exprs, + &context); + break; + + case T_ProjectSet: + case T_Material: + case T_Sort: + case T_IncrementalSort: + case T_Unique: + case T_SetOp: + case T_Group: + /* no node-type-specific fields need fixing */ + break; + + default: + elog(ERROR, "unrecognized node type: %d", + (int) nodeTag(plan)); + } + + /* Process left and right child plans, if any */ + child_params = finalize_plan(root, + plan->lefttree, + gather_param, + valid_params, + scan_params); + context.paramids = bms_add_members(context.paramids, child_params); + + if (nestloop_params) + { + /* right child can reference nestloop_params as well as valid_params */ + child_params = finalize_plan(root, + plan->righttree, + gather_param, + bms_union(nestloop_params, valid_params), + scan_params); + /* ... and they don't count as parameters used at my level */ + child_params = bms_difference(child_params, nestloop_params); + bms_free(nestloop_params); + } + else + { + /* easy case */ + child_params = finalize_plan(root, + plan->righttree, + gather_param, + valid_params, + scan_params); + } + context.paramids = bms_add_members(context.paramids, child_params); + + /* + * Any locally generated parameter doesn't count towards its generating + * plan node's external dependencies. (Note: if we changed valid_params + * and/or scan_params, we leak those bitmapsets; not worth the notational + * trouble to clean them up.) + */ + if (locally_added_param >= 0) + { + context.paramids = bms_del_member(context.paramids, + locally_added_param); + } + + /* Now we have all the paramids referenced in this node and children */ + + if (!bms_is_subset(context.paramids, valid_params)) + elog(ERROR, "plan should not reference subplan's variable"); + + /* + * The plan node's allParam and extParam fields should include all its + * referenced paramids, plus contributions from any child initPlans. + * However, any setParams of the initPlans should not be present in the + * parent node's extParams, only in its allParams. (It's possible that + * some initPlans have extParams that are setParams of other initPlans.) + */ + + /* allParam must include initplans' extParams and setParams */ + plan->allParam = bms_union(context.paramids, initExtParam); + plan->allParam = bms_add_members(plan->allParam, initSetParam); + /* extParam must include any initplan extParams */ + plan->extParam = bms_union(context.paramids, initExtParam); + /* but not any initplan setParams */ + plan->extParam = bms_del_members(plan->extParam, initSetParam); + + /* + * For speed at execution time, make sure extParam/allParam are actually + * NULL if they are empty sets. + */ + if (bms_is_empty(plan->extParam)) + plan->extParam = NULL; + if (bms_is_empty(plan->allParam)) + plan->allParam = NULL; + + return plan->allParam; +} + +/* + * finalize_primnode: add IDs of all PARAM_EXEC params appearing in the given + * expression tree to the result set. + */ +static bool +finalize_primnode(Node *node, finalize_primnode_context *context) +{ + if (node == NULL) + return false; + if (IsA(node, Param)) + { + if (((Param *) node)->paramkind == PARAM_EXEC) + { + int paramid = ((Param *) node)->paramid; + + context->paramids = bms_add_member(context->paramids, paramid); + } + return false; /* no more to do here */ + } + if (IsA(node, SubPlan)) + { + SubPlan *subplan = (SubPlan *) node; + Plan *plan = planner_subplan_get_plan(context->root, subplan); + ListCell *lc; + Bitmapset *subparamids; + + /* Recurse into the testexpr, but not into the Plan */ + finalize_primnode(subplan->testexpr, context); + + /* + * Remove any param IDs of output parameters of the subplan that were + * referenced in the testexpr. These are not interesting for + * parameter change signaling since we always re-evaluate the subplan. + * Note that this wouldn't work too well if there might be uses of the + * same param IDs elsewhere in the plan, but that can't happen because + * generate_new_exec_param never tries to merge params. + */ + foreach(lc, subplan->paramIds) + { + context->paramids = bms_del_member(context->paramids, + lfirst_int(lc)); + } + + /* Also examine args list */ + finalize_primnode((Node *) subplan->args, context); + + /* + * Add params needed by the subplan to paramids, but excluding those + * we will pass down to it. (We assume SS_finalize_plan was run on + * the subplan already.) + */ + subparamids = bms_copy(plan->extParam); + foreach(lc, subplan->parParam) + { + subparamids = bms_del_member(subparamids, lfirst_int(lc)); + } + context->paramids = bms_join(context->paramids, subparamids); + + return false; /* no more to do here */ + } + return expression_tree_walker(node, finalize_primnode, + (void *) context); +} + +/* + * finalize_agg_primnode: find all Aggref nodes in the given expression tree, + * and add IDs of all PARAM_EXEC params appearing within their aggregated + * arguments to the result set. + */ +static bool +finalize_agg_primnode(Node *node, finalize_primnode_context *context) +{ + if (node == NULL) + return false; + if (IsA(node, Aggref)) + { + Aggref *agg = (Aggref *) node; + + /* we should not consider the direct arguments, if any */ + finalize_primnode((Node *) agg->args, context); + finalize_primnode((Node *) agg->aggfilter, context); + return false; /* there can't be any Aggrefs below here */ + } + return expression_tree_walker(node, finalize_agg_primnode, + (void *) context); +} + +/* + * SS_make_initplan_output_param - make a Param for an initPlan's output + * + * The plan is expected to return a scalar value of the given type/collation. + * + * Note that in some cases the initplan may not ever appear in the finished + * plan tree. If that happens, we'll have wasted a PARAM_EXEC slot, which + * is no big deal. + */ +Param * +SS_make_initplan_output_param(PlannerInfo *root, + Oid resulttype, int32 resulttypmod, + Oid resultcollation) +{ + return generate_new_exec_param(root, resulttype, + resulttypmod, resultcollation); +} + +/* + * SS_make_initplan_from_plan - given a plan tree, make it an InitPlan + * + * We build an EXPR_SUBLINK SubPlan node and put it into the initplan + * list for the outer query level. A Param that represents the initplan's + * output has already been assigned using SS_make_initplan_output_param. + */ +void +SS_make_initplan_from_plan(PlannerInfo *root, + PlannerInfo *subroot, Plan *plan, + Param *prm) +{ + SubPlan *node; + + /* + * Add the subplan and its PlannerInfo to the global lists. + */ + root->glob->subplans = lappend(root->glob->subplans, plan); + root->glob->subroots = lappend(root->glob->subroots, subroot); + + /* + * Create a SubPlan node and add it to the outer list of InitPlans. Note + * it has to appear after any other InitPlans it might depend on (see + * comments in ExecReScan). + */ + node = makeNode(SubPlan); + node->subLinkType = EXPR_SUBLINK; + node->plan_id = list_length(root->glob->subplans); + node->plan_name = psprintf("InitPlan %d (returns $%d)", + node->plan_id, prm->paramid); + get_first_col_type(plan, &node->firstColType, &node->firstColTypmod, + &node->firstColCollation); + node->setParam = list_make1_int(prm->paramid); + + root->init_plans = lappend(root->init_plans, node); + + /* + * The node can't have any inputs (since it's an initplan), so the + * parParam and args lists remain empty. + */ + + /* Set costs of SubPlan using info from the plan tree */ + cost_subplan(subroot, node, plan); +} diff --git a/src/backend/optimizer/prep/Makefile b/src/backend/optimizer/prep/Makefile new file mode 100644 index 0000000..6f8c6c8 --- /dev/null +++ b/src/backend/optimizer/prep/Makefile @@ -0,0 +1,22 @@ +#------------------------------------------------------------------------- +# +# Makefile-- +# Makefile for optimizer/prep +# +# IDENTIFICATION +# src/backend/optimizer/prep/Makefile +# +#------------------------------------------------------------------------- + +subdir = src/backend/optimizer/prep +top_builddir = ../../../.. +include $(top_builddir)/src/Makefile.global + +OBJS = \ + prepagg.o \ + prepjointree.o \ + prepqual.o \ + preptlist.o \ + prepunion.o + +include $(top_srcdir)/src/backend/common.mk diff --git a/src/backend/optimizer/prep/prepagg.c b/src/backend/optimizer/prep/prepagg.c new file mode 100644 index 0000000..404a5f1 --- /dev/null +++ b/src/backend/optimizer/prep/prepagg.c @@ -0,0 +1,674 @@ +/*------------------------------------------------------------------------- + * + * prepagg.c + * Routines to preprocess aggregate function calls + * + * If there are identical aggregate calls in the query, they only need to + * be computed once. Also, some aggregate functions can share the same + * transition state, so that we only need to call the final function for + * them separately. These optimizations are independent of how the + * aggregates are executed. + * + * preprocess_aggrefs() detects those cases, creates AggInfo and + * AggTransInfo structs for each aggregate and transition state that needs + * to be computed, and sets the 'aggno' and 'transno' fields in the Aggrefs + * accordingly. It also resolves polymorphic transition types, and sets + * the 'aggtranstype' fields accordingly. + * + * XXX: The AggInfo and AggTransInfo structs are thrown away after + * planning, so executor startup has to perform some of the same lookups + * of transition functions and initial values that we do here. One day, we + * might want to carry that information to the Agg nodes to save the effort + * at executor startup. The Agg nodes are constructed much later in the + * planning, however, so it's not trivial. + * + * Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group + * Portions Copyright (c) 1994, Regents of the University of California + * + * + * IDENTIFICATION + * src/backend/optimizer/prep/prepagg.c + * + *------------------------------------------------------------------------- + */ + +#include "postgres.h" + +#include "access/htup_details.h" +#include "catalog/pg_aggregate.h" +#include "catalog/pg_type.h" +#include "nodes/nodeFuncs.h" +#include "nodes/pathnodes.h" +#include "optimizer/clauses.h" +#include "optimizer/cost.h" +#include "optimizer/optimizer.h" +#include "optimizer/plancat.h" +#include "optimizer/prep.h" +#include "parser/parse_agg.h" +#include "utils/builtins.h" +#include "utils/datum.h" +#include "utils/fmgroids.h" +#include "utils/lsyscache.h" +#include "utils/memutils.h" +#include "utils/syscache.h" + +static bool preprocess_aggrefs_walker(Node *node, PlannerInfo *root); +static int find_compatible_agg(PlannerInfo *root, Aggref *newagg, + List **same_input_transnos); +static int find_compatible_trans(PlannerInfo *root, Aggref *newagg, + bool shareable, + Oid aggtransfn, Oid aggtranstype, + int transtypeLen, bool transtypeByVal, + Oid aggcombinefn, + Oid aggserialfn, Oid aggdeserialfn, + Datum initValue, bool initValueIsNull, + List *transnos); +static Datum GetAggInitVal(Datum textInitVal, Oid transtype); + +/* ----------------- + * Resolve the transition type of all Aggrefs, and determine which Aggrefs + * can share aggregate or transition state. + * + * Information about the aggregates and transition functions are collected + * in the root->agginfos and root->aggtransinfos lists. The 'aggtranstype', + * 'aggno', and 'aggtransno' fields of each Aggref are filled in. + * + * NOTE: This modifies the Aggrefs in the input expression in-place! + * + * We try to optimize by detecting duplicate aggregate functions so that + * their state and final values are re-used, rather than needlessly being + * re-calculated independently. We also detect aggregates that are not + * the same, but which can share the same transition state. + * + * Scenarios: + * + * 1. Identical aggregate function calls appear in the query: + * + * SELECT SUM(x) FROM ... HAVING SUM(x) > 0 + * + * Since these aggregates are identical, we only need to calculate + * the value once. Both aggregates will share the same 'aggno' value. + * + * 2. Two different aggregate functions appear in the query, but the + * aggregates have the same arguments, transition functions and + * initial values (and, presumably, different final functions): + * + * SELECT AVG(x), STDDEV(x) FROM ... + * + * In this case we must create a new AggInfo for the varying aggregate, + * and we need to call the final functions separately, but we need + * only run the transition function once. (This requires that the + * final functions be nondestructive of the transition state, but + * that's required anyway for other reasons.) + * + * For either of these optimizations to be valid, all aggregate properties + * used in the transition phase must be the same, including any modifiers + * such as ORDER BY, DISTINCT and FILTER, and the arguments mustn't + * contain any volatile functions. + * ----------------- + */ +void +preprocess_aggrefs(PlannerInfo *root, Node *clause) +{ + (void) preprocess_aggrefs_walker(clause, root); +} + +static void +preprocess_aggref(Aggref *aggref, PlannerInfo *root) +{ + HeapTuple aggTuple; + Form_pg_aggregate aggform; + Oid aggtransfn; + Oid aggfinalfn; + Oid aggcombinefn; + Oid aggserialfn; + Oid aggdeserialfn; + Oid aggtranstype; + int32 aggtranstypmod; + int32 aggtransspace; + bool shareable; + int aggno; + int transno; + List *same_input_transnos; + int16 resulttypeLen; + bool resulttypeByVal; + Datum textInitVal; + Datum initValue; + bool initValueIsNull; + bool transtypeByVal; + int16 transtypeLen; + Oid inputTypes[FUNC_MAX_ARGS]; + int numArguments; + + Assert(aggref->agglevelsup == 0); + + /* + * Fetch info about the aggregate from pg_aggregate. Note it's correct to + * ignore the moving-aggregate variant, since what we're concerned with + * here is aggregates not window functions. + */ + aggTuple = SearchSysCache1(AGGFNOID, + ObjectIdGetDatum(aggref->aggfnoid)); + if (!HeapTupleIsValid(aggTuple)) + elog(ERROR, "cache lookup failed for aggregate %u", + aggref->aggfnoid); + aggform = (Form_pg_aggregate) GETSTRUCT(aggTuple); + aggtransfn = aggform->aggtransfn; + aggfinalfn = aggform->aggfinalfn; + aggcombinefn = aggform->aggcombinefn; + aggserialfn = aggform->aggserialfn; + aggdeserialfn = aggform->aggdeserialfn; + aggtranstype = aggform->aggtranstype; + aggtransspace = aggform->aggtransspace; + + /* + * Resolve the possibly-polymorphic aggregate transition type. + */ + + /* extract argument types (ignoring any ORDER BY expressions) */ + numArguments = get_aggregate_argtypes(aggref, inputTypes); + + /* resolve actual type of transition state, if polymorphic */ + aggtranstype = resolve_aggregate_transtype(aggref->aggfnoid, + aggtranstype, + inputTypes, + numArguments); + aggref->aggtranstype = aggtranstype; + + /* + * If transition state is of same type as first aggregated input, assume + * it's the same typmod (same width) as well. This works for cases like + * MAX/MIN and is probably somewhat reasonable otherwise. + */ + aggtranstypmod = -1; + if (aggref->args) + { + TargetEntry *tle = (TargetEntry *) linitial(aggref->args); + + if (aggtranstype == exprType((Node *) tle->expr)) + aggtranstypmod = exprTypmod((Node *) tle->expr); + } + + /* + * If finalfn is marked read-write, we can't share transition states; but + * it is okay to share states for AGGMODIFY_SHAREABLE aggs. + * + * In principle, in a partial aggregate, we could share the transition + * state even if the final function is marked as read-write, because the + * partial aggregate doesn't execute the final function. But it's too + * early to know whether we're going perform a partial aggregate. + */ + shareable = (aggform->aggfinalmodify != AGGMODIFY_READ_WRITE); + + /* get info about the output value's datatype */ + get_typlenbyval(aggref->aggtype, + &resulttypeLen, + &resulttypeByVal); + + /* get initial value */ + textInitVal = SysCacheGetAttr(AGGFNOID, aggTuple, + Anum_pg_aggregate_agginitval, + &initValueIsNull); + if (initValueIsNull) + initValue = (Datum) 0; + else + initValue = GetAggInitVal(textInitVal, aggtranstype); + + ReleaseSysCache(aggTuple); + + /* + * 1. See if this is identical to another aggregate function call that + * we've seen already. + */ + aggno = find_compatible_agg(root, aggref, &same_input_transnos); + if (aggno != -1) + { + AggInfo *agginfo = list_nth(root->agginfos, aggno); + + transno = agginfo->transno; + } + else + { + AggInfo *agginfo = palloc(sizeof(AggInfo)); + + agginfo->finalfn_oid = aggfinalfn; + agginfo->representative_aggref = aggref; + agginfo->shareable = shareable; + + aggno = list_length(root->agginfos); + root->agginfos = lappend(root->agginfos, agginfo); + + /* + * Count it, and check for cases requiring ordered input. Note that + * ordered-set aggs always have nonempty aggorder. Any ordered-input + * case also defeats partial aggregation. + */ + if (aggref->aggorder != NIL || aggref->aggdistinct != NIL) + { + root->numOrderedAggs++; + root->hasNonPartialAggs = true; + } + + get_typlenbyval(aggtranstype, + &transtypeLen, + &transtypeByVal); + + /* + * 2. See if this aggregate can share transition state with another + * aggregate that we've initialized already. + */ + transno = find_compatible_trans(root, aggref, shareable, + aggtransfn, aggtranstype, + transtypeLen, transtypeByVal, + aggcombinefn, + aggserialfn, aggdeserialfn, + initValue, initValueIsNull, + same_input_transnos); + if (transno == -1) + { + AggTransInfo *transinfo = palloc(sizeof(AggTransInfo)); + + transinfo->args = aggref->args; + transinfo->aggfilter = aggref->aggfilter; + transinfo->transfn_oid = aggtransfn; + transinfo->combinefn_oid = aggcombinefn; + transinfo->serialfn_oid = aggserialfn; + transinfo->deserialfn_oid = aggdeserialfn; + transinfo->aggtranstype = aggtranstype; + transinfo->aggtranstypmod = aggtranstypmod; + transinfo->transtypeLen = transtypeLen; + transinfo->transtypeByVal = transtypeByVal; + transinfo->aggtransspace = aggtransspace; + transinfo->initValue = initValue; + transinfo->initValueIsNull = initValueIsNull; + + transno = list_length(root->aggtransinfos); + root->aggtransinfos = lappend(root->aggtransinfos, transinfo); + + /* + * Check whether partial aggregation is feasible, unless we + * already found out that we can't do it. + */ + if (!root->hasNonPartialAggs) + { + /* + * If there is no combine function, then partial aggregation + * is not possible. + */ + if (!OidIsValid(transinfo->combinefn_oid)) + root->hasNonPartialAggs = true; + + /* + * If we have any aggs with transtype INTERNAL then we must + * check whether they have serialization/deserialization + * functions; if not, we can't serialize partial-aggregation + * results. + */ + else if (transinfo->aggtranstype == INTERNALOID && + (!OidIsValid(transinfo->serialfn_oid) || + !OidIsValid(transinfo->deserialfn_oid))) + root->hasNonSerialAggs = true; + } + } + agginfo->transno = transno; + } + + /* + * Fill in the fields in the Aggref (aggtranstype was set above already) + */ + aggref->aggno = aggno; + aggref->aggtransno = transno; +} + +static bool +preprocess_aggrefs_walker(Node *node, PlannerInfo *root) +{ + if (node == NULL) + return false; + if (IsA(node, Aggref)) + { + Aggref *aggref = (Aggref *) node; + + preprocess_aggref(aggref, root); + + /* + * We assume that the parser checked that there are no aggregates (of + * this level anyway) in the aggregated arguments, direct arguments, + * or filter clause. Hence, we need not recurse into any of them. + */ + return false; + } + Assert(!IsA(node, SubLink)); + return expression_tree_walker(node, preprocess_aggrefs_walker, + (void *) root); +} + + +/* + * find_compatible_agg - search for a previously initialized per-Agg struct + * + * Searches the previously looked at aggregates to find one which is compatible + * with this one, with the same input parameters. If no compatible aggregate + * can be found, returns -1. + * + * As a side-effect, this also collects a list of existing, shareable per-Trans + * structs with matching inputs. If no identical Aggref is found, the list is + * passed later to find_compatible_trans, to see if we can at least reuse + * the state value of another aggregate. + */ +static int +find_compatible_agg(PlannerInfo *root, Aggref *newagg, + List **same_input_transnos) +{ + ListCell *lc; + int aggno; + + *same_input_transnos = NIL; + + /* we mustn't reuse the aggref if it contains volatile function calls */ + if (contain_volatile_functions((Node *) newagg)) + return -1; + + /* + * Search through the list of already seen aggregates. If we find an + * existing identical aggregate call, then we can re-use that one. While + * searching, we'll also collect a list of Aggrefs with the same input + * parameters. If no matching Aggref is found, the caller can potentially + * still re-use the transition state of one of them. (At this stage we + * just compare the parsetrees; whether different aggregates share the + * same transition function will be checked later.) + */ + aggno = -1; + foreach(lc, root->agginfos) + { + AggInfo *agginfo = (AggInfo *) lfirst(lc); + Aggref *existingRef; + + aggno++; + + existingRef = agginfo->representative_aggref; + + /* all of the following must be the same or it's no match */ + if (newagg->inputcollid != existingRef->inputcollid || + newagg->aggtranstype != existingRef->aggtranstype || + newagg->aggstar != existingRef->aggstar || + newagg->aggvariadic != existingRef->aggvariadic || + newagg->aggkind != existingRef->aggkind || + !equal(newagg->args, existingRef->args) || + !equal(newagg->aggorder, existingRef->aggorder) || + !equal(newagg->aggdistinct, existingRef->aggdistinct) || + !equal(newagg->aggfilter, existingRef->aggfilter)) + continue; + + /* if it's the same aggregate function then report exact match */ + if (newagg->aggfnoid == existingRef->aggfnoid && + newagg->aggtype == existingRef->aggtype && + newagg->aggcollid == existingRef->aggcollid && + equal(newagg->aggdirectargs, existingRef->aggdirectargs)) + { + list_free(*same_input_transnos); + *same_input_transnos = NIL; + return aggno; + } + + /* + * Not identical, but it had the same inputs. If the final function + * permits sharing, return its transno to the caller, in case we can + * re-use its per-trans state. (If there's already sharing going on, + * we might report a transno more than once. find_compatible_trans is + * cheap enough that it's not worth spending cycles to avoid that.) + */ + if (agginfo->shareable) + *same_input_transnos = lappend_int(*same_input_transnos, + agginfo->transno); + } + + return -1; +} + +/* + * find_compatible_trans - search for a previously initialized per-Trans + * struct + * + * Searches the list of transnos for a per-Trans struct with the same + * transition function and initial condition. (The inputs have already been + * verified to match.) + */ +static int +find_compatible_trans(PlannerInfo *root, Aggref *newagg, bool shareable, + Oid aggtransfn, Oid aggtranstype, + int transtypeLen, bool transtypeByVal, + Oid aggcombinefn, + Oid aggserialfn, Oid aggdeserialfn, + Datum initValue, bool initValueIsNull, + List *transnos) +{ + ListCell *lc; + + /* If this aggregate can't share transition states, give up */ + if (!shareable) + return -1; + + foreach(lc, transnos) + { + int transno = lfirst_int(lc); + AggTransInfo *pertrans = (AggTransInfo *) list_nth(root->aggtransinfos, transno); + + /* + * if the transfns or transition state types are not the same then the + * state can't be shared. + */ + if (aggtransfn != pertrans->transfn_oid || + aggtranstype != pertrans->aggtranstype) + continue; + + /* + * The serialization and deserialization functions must match, if + * present, as we're unable to share the trans state for aggregates + * which will serialize or deserialize into different formats. + * Remember that these will be InvalidOid if they're not required for + * this agg node. + */ + if (aggserialfn != pertrans->serialfn_oid || + aggdeserialfn != pertrans->deserialfn_oid) + continue; + + /* + * Combine function must also match. We only care about the combine + * function with partial aggregates, but it's too early in the + * planning to know if we will do partial aggregation, so be + * conservative. + */ + if (aggcombinefn != pertrans->combinefn_oid) + continue; + + /* + * Check that the initial condition matches, too. + */ + if (initValueIsNull && pertrans->initValueIsNull) + return transno; + + if (!initValueIsNull && !pertrans->initValueIsNull && + datumIsEqual(initValue, pertrans->initValue, + transtypeByVal, transtypeLen)) + return transno; + } + return -1; +} + +static Datum +GetAggInitVal(Datum textInitVal, Oid transtype) +{ + Oid typinput, + typioparam; + char *strInitVal; + Datum initVal; + + getTypeInputInfo(transtype, &typinput, &typioparam); + strInitVal = TextDatumGetCString(textInitVal); + initVal = OidInputFunctionCall(typinput, strInitVal, + typioparam, -1); + pfree(strInitVal); + return initVal; +} + + +/* + * get_agg_clause_costs + * Process the PlannerInfo's 'aggtransinfos' and 'agginfos' lists + * accumulating the cost information about them. + * + * 'aggsplit' tells us the expected partial-aggregation mode, which affects + * the cost estimates. + * + * NOTE that the costs are ADDED to those already in *costs ... so the caller + * is responsible for zeroing the struct initially. + * + * For each AggTransInfo, we add the cost of an aggregate transition using + * either the transfn or combinefn depending on the 'aggsplit' value. We also + * account for the costs of any aggfilters and any serializations and + * deserializations of the transition state and also estimate the total space + * needed for the transition states as if each aggregate's state was stored in + * memory concurrently (as would be done in a HashAgg plan). + * + * For each AggInfo in the 'agginfos' list we add the cost of running the + * final function and the direct args, if any. + */ +void +get_agg_clause_costs(PlannerInfo *root, AggSplit aggsplit, AggClauseCosts *costs) +{ + ListCell *lc; + + foreach(lc, root->aggtransinfos) + { + AggTransInfo *transinfo = (AggTransInfo *) lfirst(lc); + + /* + * Add the appropriate component function execution costs to + * appropriate totals. + */ + if (DO_AGGSPLIT_COMBINE(aggsplit)) + { + /* charge for combining previously aggregated states */ + add_function_cost(root, transinfo->combinefn_oid, NULL, + &costs->transCost); + } + else + add_function_cost(root, transinfo->transfn_oid, NULL, + &costs->transCost); + if (DO_AGGSPLIT_DESERIALIZE(aggsplit) && + OidIsValid(transinfo->deserialfn_oid)) + add_function_cost(root, transinfo->deserialfn_oid, NULL, + &costs->transCost); + if (DO_AGGSPLIT_SERIALIZE(aggsplit) && + OidIsValid(transinfo->serialfn_oid)) + add_function_cost(root, transinfo->serialfn_oid, NULL, + &costs->finalCost); + + /* + * These costs are incurred only by the initial aggregate node, so we + * mustn't include them again at upper levels. + */ + if (!DO_AGGSPLIT_COMBINE(aggsplit)) + { + /* add the input expressions' cost to per-input-row costs */ + QualCost argcosts; + + cost_qual_eval_node(&argcosts, (Node *) transinfo->args, root); + costs->transCost.startup += argcosts.startup; + costs->transCost.per_tuple += argcosts.per_tuple; + + /* + * Add any filter's cost to per-input-row costs. + * + * XXX Ideally we should reduce input expression costs according + * to filter selectivity, but it's not clear it's worth the + * trouble. + */ + if (transinfo->aggfilter) + { + cost_qual_eval_node(&argcosts, (Node *) transinfo->aggfilter, + root); + costs->transCost.startup += argcosts.startup; + costs->transCost.per_tuple += argcosts.per_tuple; + } + } + + /* + * If the transition type is pass-by-value then it doesn't add + * anything to the required size of the hashtable. If it is + * pass-by-reference then we have to add the estimated size of the + * value itself, plus palloc overhead. + */ + if (!transinfo->transtypeByVal) + { + int32 avgwidth; + + /* Use average width if aggregate definition gave one */ + if (transinfo->aggtransspace > 0) + avgwidth = transinfo->aggtransspace; + else if (transinfo->transfn_oid == F_ARRAY_APPEND) + { + /* + * If the transition function is array_append(), it'll use an + * expanded array as transvalue, which will occupy at least + * ALLOCSET_SMALL_INITSIZE and possibly more. Use that as the + * estimate for lack of a better idea. + */ + avgwidth = ALLOCSET_SMALL_INITSIZE; + } + else + { + avgwidth = get_typavgwidth(transinfo->aggtranstype, transinfo->aggtranstypmod); + } + + avgwidth = MAXALIGN(avgwidth); + costs->transitionSpace += avgwidth + 2 * sizeof(void *); + } + else if (transinfo->aggtranstype == INTERNALOID) + { + /* + * INTERNAL transition type is a special case: although INTERNAL + * is pass-by-value, it's almost certainly being used as a pointer + * to some large data structure. The aggregate definition can + * provide an estimate of the size. If it doesn't, then we assume + * ALLOCSET_DEFAULT_INITSIZE, which is a good guess if the data is + * being kept in a private memory context, as is done by + * array_agg() for instance. + */ + if (transinfo->aggtransspace > 0) + costs->transitionSpace += transinfo->aggtransspace; + else + costs->transitionSpace += ALLOCSET_DEFAULT_INITSIZE; + } + } + + foreach(lc, root->agginfos) + { + AggInfo *agginfo = (AggInfo *) lfirst(lc); + Aggref *aggref = agginfo->representative_aggref; + + /* + * Add the appropriate component function execution costs to + * appropriate totals. + */ + if (!DO_AGGSPLIT_SKIPFINAL(aggsplit) && + OidIsValid(agginfo->finalfn_oid)) + add_function_cost(root, agginfo->finalfn_oid, NULL, + &costs->finalCost); + + /* + * If there are direct arguments, treat their evaluation cost like the + * cost of the finalfn. + */ + if (aggref->aggdirectargs) + { + QualCost argcosts; + + cost_qual_eval_node(&argcosts, (Node *) aggref->aggdirectargs, + root); + costs->finalCost.startup += argcosts.startup; + costs->finalCost.per_tuple += argcosts.per_tuple; + } + } +} diff --git a/src/backend/optimizer/prep/prepjointree.c b/src/backend/optimizer/prep/prepjointree.c new file mode 100644 index 0000000..7f4bb7b --- /dev/null +++ b/src/backend/optimizer/prep/prepjointree.c @@ -0,0 +1,3746 @@ +/*------------------------------------------------------------------------- + * + * prepjointree.c + * Planner preprocessing for subqueries and join tree manipulation. + * + * NOTE: the intended sequence for invoking these operations is + * replace_empty_jointree + * pull_up_sublinks + * preprocess_function_rtes + * pull_up_subqueries + * flatten_simple_union_all + * do expression preprocessing (including flattening JOIN alias vars) + * reduce_outer_joins + * remove_useless_result_rtes + * + * + * Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group + * Portions Copyright (c) 1994, Regents of the University of California + * + * + * IDENTIFICATION + * src/backend/optimizer/prep/prepjointree.c + * + *------------------------------------------------------------------------- + */ +#include "postgres.h" + +#include "catalog/pg_type.h" +#include "funcapi.h" +#include "miscadmin.h" +#include "nodes/makefuncs.h" +#include "nodes/nodeFuncs.h" +#include "optimizer/clauses.h" +#include "optimizer/optimizer.h" +#include "optimizer/placeholder.h" +#include "optimizer/prep.h" +#include "optimizer/subselect.h" +#include "optimizer/tlist.h" +#include "parser/parse_relation.h" +#include "parser/parsetree.h" +#include "rewrite/rewriteManip.h" + + +typedef struct pullup_replace_vars_context +{ + PlannerInfo *root; + List *targetlist; /* tlist of subquery being pulled up */ + RangeTblEntry *target_rte; /* RTE of subquery */ + Relids relids; /* relids within subquery, as numbered after + * pullup (set only if target_rte->lateral) */ + bool *outer_hasSubLinks; /* -> outer query's hasSubLinks */ + int varno; /* varno of subquery */ + bool need_phvs; /* do we need PlaceHolderVars? */ + bool wrap_non_vars; /* do we need 'em on *all* non-Vars? */ + Node **rv_cache; /* cache for results with PHVs */ +} pullup_replace_vars_context; + +typedef struct reduce_outer_joins_state +{ + Relids relids; /* base relids within this subtree */ + bool contains_outer; /* does subtree contain outer join(s)? */ + List *sub_states; /* List of states for subtree components */ +} reduce_outer_joins_state; + +static Node *pull_up_sublinks_jointree_recurse(PlannerInfo *root, Node *jtnode, + Relids *relids); +static Node *pull_up_sublinks_qual_recurse(PlannerInfo *root, Node *node, + Node **jtlink1, Relids available_rels1, + Node **jtlink2, Relids available_rels2); +static Node *pull_up_subqueries_recurse(PlannerInfo *root, Node *jtnode, + JoinExpr *lowest_outer_join, + JoinExpr *lowest_nulling_outer_join, + AppendRelInfo *containing_appendrel); +static Node *pull_up_simple_subquery(PlannerInfo *root, Node *jtnode, + RangeTblEntry *rte, + JoinExpr *lowest_outer_join, + JoinExpr *lowest_nulling_outer_join, + AppendRelInfo *containing_appendrel); +static Node *pull_up_simple_union_all(PlannerInfo *root, Node *jtnode, + RangeTblEntry *rte); +static void pull_up_union_leaf_queries(Node *setOp, PlannerInfo *root, + int parentRTindex, Query *setOpQuery, + int childRToffset); +static void make_setop_translation_list(Query *query, int newvarno, + AppendRelInfo *appinfo); +static bool is_simple_subquery(PlannerInfo *root, Query *subquery, + RangeTblEntry *rte, + JoinExpr *lowest_outer_join); +static Node *pull_up_simple_values(PlannerInfo *root, Node *jtnode, + RangeTblEntry *rte); +static bool is_simple_values(PlannerInfo *root, RangeTblEntry *rte); +static Node *pull_up_constant_function(PlannerInfo *root, Node *jtnode, + RangeTblEntry *rte, + JoinExpr *lowest_nulling_outer_join, + AppendRelInfo *containing_appendrel); +static bool is_simple_union_all(Query *subquery); +static bool is_simple_union_all_recurse(Node *setOp, Query *setOpQuery, + List *colTypes); +static bool is_safe_append_member(Query *subquery); +static bool jointree_contains_lateral_outer_refs(PlannerInfo *root, + Node *jtnode, bool restricted, + Relids safe_upper_varnos); +static void perform_pullup_replace_vars(PlannerInfo *root, + pullup_replace_vars_context *rvcontext, + JoinExpr *lowest_nulling_outer_join, + AppendRelInfo *containing_appendrel); +static void replace_vars_in_jointree(Node *jtnode, + pullup_replace_vars_context *context, + JoinExpr *lowest_nulling_outer_join); +static Node *pullup_replace_vars(Node *expr, + pullup_replace_vars_context *context); +static Node *pullup_replace_vars_callback(Var *var, + replace_rte_variables_context *context); +static Query *pullup_replace_vars_subquery(Query *query, + pullup_replace_vars_context *context); +static reduce_outer_joins_state *reduce_outer_joins_pass1(Node *jtnode); +static void reduce_outer_joins_pass2(Node *jtnode, + reduce_outer_joins_state *state, + PlannerInfo *root, + Relids nonnullable_rels, + List *nonnullable_vars, + List *forced_null_vars); +static Node *remove_useless_results_recurse(PlannerInfo *root, Node *jtnode); +static int get_result_relid(PlannerInfo *root, Node *jtnode); +static void remove_result_refs(PlannerInfo *root, int varno, Node *newjtloc); +static bool find_dependent_phvs(PlannerInfo *root, int varno); +static bool find_dependent_phvs_in_jointree(PlannerInfo *root, + Node *node, int varno); +static void substitute_phv_relids(Node *node, + int varno, Relids subrelids); +static void fix_append_rel_relids(List *append_rel_list, int varno, + Relids subrelids); +static Node *find_jointree_node_for_rel(Node *jtnode, int relid); + + +/* + * transform_MERGE_to_join + * Replace a MERGE's jointree to also include the target relation. + */ +void +transform_MERGE_to_join(Query *parse) +{ + RangeTblEntry *joinrte; + JoinExpr *joinexpr; + JoinType jointype; + int joinrti; + List *vars; + + if (parse->commandType != CMD_MERGE) + return; + + /* XXX probably bogus */ + vars = NIL; + + /* + * When any WHEN NOT MATCHED THEN INSERT clauses exist, we need to use an + * outer join so that we process all unmatched tuples from the source + * relation. If none exist, we can use an inner join. + */ + if (parse->mergeUseOuterJoin) + jointype = JOIN_RIGHT; + else + jointype = JOIN_INNER; + + /* Manufacture a join RTE to use. */ + joinrte = makeNode(RangeTblEntry); + joinrte->rtekind = RTE_JOIN; + joinrte->jointype = jointype; + joinrte->joinmergedcols = 0; + joinrte->joinaliasvars = vars; + joinrte->joinleftcols = NIL; /* MERGE does not allow JOIN USING */ + joinrte->joinrightcols = NIL; /* ditto */ + joinrte->join_using_alias = NULL; + + joinrte->alias = NULL; + joinrte->eref = makeAlias("*MERGE*", NIL); + joinrte->lateral = false; + joinrte->inh = false; + joinrte->inFromCl = true; + joinrte->requiredPerms = 0; + joinrte->checkAsUser = InvalidOid; + joinrte->selectedCols = NULL; + joinrte->insertedCols = NULL; + joinrte->updatedCols = NULL; + joinrte->extraUpdatedCols = NULL; + joinrte->securityQuals = NIL; + + /* + * Add completed RTE to pstate's range table list, so that we know its + * index. + */ + parse->rtable = lappend(parse->rtable, joinrte); + joinrti = list_length(parse->rtable); + + /* + * Create a JOIN between the target and the source relation. + */ + joinexpr = makeNode(JoinExpr); + joinexpr->jointype = jointype; + joinexpr->isNatural = false; + joinexpr->larg = (Node *) makeNode(RangeTblRef); + ((RangeTblRef *) joinexpr->larg)->rtindex = parse->resultRelation; + joinexpr->rarg = linitial(parse->jointree->fromlist); /* original join */ + joinexpr->usingClause = NIL; + joinexpr->join_using_alias = NULL; + /* The quals are removed from the jointree and into this specific join */ + joinexpr->quals = parse->jointree->quals; + joinexpr->alias = NULL; + joinexpr->rtindex = joinrti; + + /* Make the new join be the sole entry in the query's jointree */ + parse->jointree->fromlist = list_make1(joinexpr); + parse->jointree->quals = NULL; +} + +/* + * replace_empty_jointree + * If the Query's jointree is empty, replace it with a dummy RTE_RESULT + * relation. + * + * By doing this, we can avoid a bunch of corner cases that formerly existed + * for SELECTs with omitted FROM clauses. An example is that a subquery + * with empty jointree previously could not be pulled up, because that would + * have resulted in an empty relid set, making the subquery not uniquely + * identifiable for join or PlaceHolderVar processing. + * + * Unlike most other functions in this file, this function doesn't recurse; + * we rely on other processing to invoke it on sub-queries at suitable times. + */ +void +replace_empty_jointree(Query *parse) +{ + RangeTblEntry *rte; + Index rti; + RangeTblRef *rtr; + + /* Nothing to do if jointree is already nonempty */ + if (parse->jointree->fromlist != NIL) + return; + + /* We mustn't change it in the top level of a setop tree, either */ + if (parse->setOperations) + return; + + /* Create suitable RTE */ + rte = makeNode(RangeTblEntry); + rte->rtekind = RTE_RESULT; + rte->eref = makeAlias("*RESULT*", NIL); + + /* Add it to rangetable */ + parse->rtable = lappend(parse->rtable, rte); + rti = list_length(parse->rtable); + + /* And jam a reference into the jointree */ + rtr = makeNode(RangeTblRef); + rtr->rtindex = rti; + parse->jointree->fromlist = list_make1(rtr); +} + +/* + * pull_up_sublinks + * Attempt to pull up ANY and EXISTS SubLinks to be treated as + * semijoins or anti-semijoins. + * + * A clause "foo op ANY (sub-SELECT)" can be processed by pulling the + * sub-SELECT up to become a rangetable entry and treating the implied + * comparisons as quals of a semijoin. However, this optimization *only* + * works at the top level of WHERE or a JOIN/ON clause, because we cannot + * distinguish whether the ANY ought to return FALSE or NULL in cases + * involving NULL inputs. Also, in an outer join's ON clause we can only + * do this if the sublink is degenerate (ie, references only the nullable + * side of the join). In that case it is legal to push the semijoin + * down into the nullable side of the join. If the sublink references any + * nonnullable-side variables then it would have to be evaluated as part + * of the outer join, which makes things way too complicated. + * + * Under similar conditions, EXISTS and NOT EXISTS clauses can be handled + * by pulling up the sub-SELECT and creating a semijoin or anti-semijoin. + * + * This routine searches for such clauses and does the necessary parsetree + * transformations if any are found. + * + * This routine has to run before preprocess_expression(), so the quals + * clauses are not yet reduced to implicit-AND format, and are not guaranteed + * to be AND/OR-flat either. That means we need to recursively search through + * explicit AND clauses. We stop as soon as we hit a non-AND item. + */ +void +pull_up_sublinks(PlannerInfo *root) +{ + Node *jtnode; + Relids relids; + + /* Begin recursion through the jointree */ + jtnode = pull_up_sublinks_jointree_recurse(root, + (Node *) root->parse->jointree, + &relids); + + /* + * root->parse->jointree must always be a FromExpr, so insert a dummy one + * if we got a bare RangeTblRef or JoinExpr out of the recursion. + */ + if (IsA(jtnode, FromExpr)) + root->parse->jointree = (FromExpr *) jtnode; + else + root->parse->jointree = makeFromExpr(list_make1(jtnode), NULL); +} + +/* + * Recurse through jointree nodes for pull_up_sublinks() + * + * In addition to returning the possibly-modified jointree node, we return + * a relids set of the contained rels into *relids. + */ +static Node * +pull_up_sublinks_jointree_recurse(PlannerInfo *root, Node *jtnode, + Relids *relids) +{ + /* Since this function recurses, it could be driven to stack overflow. */ + check_stack_depth(); + + if (jtnode == NULL) + { + *relids = NULL; + } + else if (IsA(jtnode, RangeTblRef)) + { + int varno = ((RangeTblRef *) jtnode)->rtindex; + + *relids = bms_make_singleton(varno); + /* jtnode is returned unmodified */ + } + else if (IsA(jtnode, FromExpr)) + { + FromExpr *f = (FromExpr *) jtnode; + List *newfromlist = NIL; + Relids frelids = NULL; + FromExpr *newf; + Node *jtlink; + ListCell *l; + + /* First, recurse to process children and collect their relids */ + foreach(l, f->fromlist) + { + Node *newchild; + Relids childrelids; + + newchild = pull_up_sublinks_jointree_recurse(root, + lfirst(l), + &childrelids); + newfromlist = lappend(newfromlist, newchild); + frelids = bms_join(frelids, childrelids); + } + /* Build the replacement FromExpr; no quals yet */ + newf = makeFromExpr(newfromlist, NULL); + /* Set up a link representing the rebuilt jointree */ + jtlink = (Node *) newf; + /* Now process qual --- all children are available for use */ + newf->quals = pull_up_sublinks_qual_recurse(root, f->quals, + &jtlink, frelids, + NULL, NULL); + + /* + * Note that the result will be either newf, or a stack of JoinExprs + * with newf at the base. We rely on subsequent optimization steps to + * flatten this and rearrange the joins as needed. + * + * Although we could include the pulled-up subqueries in the returned + * relids, there's no need since upper quals couldn't refer to their + * outputs anyway. + */ + *relids = frelids; + jtnode = jtlink; + } + else if (IsA(jtnode, JoinExpr)) + { + JoinExpr *j; + Relids leftrelids; + Relids rightrelids; + Node *jtlink; + + /* + * Make a modifiable copy of join node, but don't bother copying its + * subnodes (yet). + */ + j = (JoinExpr *) palloc(sizeof(JoinExpr)); + memcpy(j, jtnode, sizeof(JoinExpr)); + jtlink = (Node *) j; + + /* Recurse to process children and collect their relids */ + j->larg = pull_up_sublinks_jointree_recurse(root, j->larg, + &leftrelids); + j->rarg = pull_up_sublinks_jointree_recurse(root, j->rarg, + &rightrelids); + + /* + * Now process qual, showing appropriate child relids as available, + * and attach any pulled-up jointree items at the right place. In the + * inner-join case we put new JoinExprs above the existing one (much + * as for a FromExpr-style join). In outer-join cases the new + * JoinExprs must go into the nullable side of the outer join. The + * point of the available_rels machinations is to ensure that we only + * pull up quals for which that's okay. + * + * We don't expect to see any pre-existing JOIN_SEMI or JOIN_ANTI + * nodes here. + */ + switch (j->jointype) + { + case JOIN_INNER: + j->quals = pull_up_sublinks_qual_recurse(root, j->quals, + &jtlink, + bms_union(leftrelids, + rightrelids), + NULL, NULL); + break; + case JOIN_LEFT: + j->quals = pull_up_sublinks_qual_recurse(root, j->quals, + &j->rarg, + rightrelids, + NULL, NULL); + break; + case JOIN_FULL: + /* can't do anything with full-join quals */ + break; + case JOIN_RIGHT: + j->quals = pull_up_sublinks_qual_recurse(root, j->quals, + &j->larg, + leftrelids, + NULL, NULL); + break; + default: + elog(ERROR, "unrecognized join type: %d", + (int) j->jointype); + break; + } + + /* + * Although we could include the pulled-up subqueries in the returned + * relids, there's no need since upper quals couldn't refer to their + * outputs anyway. But we *do* need to include the join's own rtindex + * because we haven't yet collapsed join alias variables, so upper + * levels would mistakenly think they couldn't use references to this + * join. + */ + *relids = bms_join(leftrelids, rightrelids); + if (j->rtindex) + *relids = bms_add_member(*relids, j->rtindex); + jtnode = jtlink; + } + else + elog(ERROR, "unrecognized node type: %d", + (int) nodeTag(jtnode)); + return jtnode; +} + +/* + * Recurse through top-level qual nodes for pull_up_sublinks() + * + * jtlink1 points to the link in the jointree where any new JoinExprs should + * be inserted if they reference available_rels1 (i.e., available_rels1 + * denotes the relations present underneath jtlink1). Optionally, jtlink2 can + * point to a second link where new JoinExprs should be inserted if they + * reference available_rels2 (pass NULL for both those arguments if not used). + * Note that SubLinks referencing both sets of variables cannot be optimized. + * If we find multiple pull-up-able SubLinks, they'll get stacked onto jtlink1 + * and/or jtlink2 in the order we encounter them. We rely on subsequent + * optimization to rearrange the stack if appropriate. + * + * Returns the replacement qual node, or NULL if the qual should be removed. + */ +static Node * +pull_up_sublinks_qual_recurse(PlannerInfo *root, Node *node, + Node **jtlink1, Relids available_rels1, + Node **jtlink2, Relids available_rels2) +{ + if (node == NULL) + return NULL; + if (IsA(node, SubLink)) + { + SubLink *sublink = (SubLink *) node; + JoinExpr *j; + Relids child_rels; + + /* Is it a convertible ANY or EXISTS clause? */ + if (sublink->subLinkType == ANY_SUBLINK) + { + if ((j = convert_ANY_sublink_to_join(root, sublink, + available_rels1)) != NULL) + { + /* Yes; insert the new join node into the join tree */ + j->larg = *jtlink1; + *jtlink1 = (Node *) j; + /* Recursively process pulled-up jointree nodes */ + j->rarg = pull_up_sublinks_jointree_recurse(root, + j->rarg, + &child_rels); + + /* + * Now recursively process the pulled-up quals. Any inserted + * joins can get stacked onto either j->larg or j->rarg, + * depending on which rels they reference. + */ + j->quals = pull_up_sublinks_qual_recurse(root, + j->quals, + &j->larg, + available_rels1, + &j->rarg, + child_rels); + /* Return NULL representing constant TRUE */ + return NULL; + } + if (available_rels2 != NULL && + (j = convert_ANY_sublink_to_join(root, sublink, + available_rels2)) != NULL) + { + /* Yes; insert the new join node into the join tree */ + j->larg = *jtlink2; + *jtlink2 = (Node *) j; + /* Recursively process pulled-up jointree nodes */ + j->rarg = pull_up_sublinks_jointree_recurse(root, + j->rarg, + &child_rels); + + /* + * Now recursively process the pulled-up quals. Any inserted + * joins can get stacked onto either j->larg or j->rarg, + * depending on which rels they reference. + */ + j->quals = pull_up_sublinks_qual_recurse(root, + j->quals, + &j->larg, + available_rels2, + &j->rarg, + child_rels); + /* Return NULL representing constant TRUE */ + return NULL; + } + } + else if (sublink->subLinkType == EXISTS_SUBLINK) + { + if ((j = convert_EXISTS_sublink_to_join(root, sublink, false, + available_rels1)) != NULL) + { + /* Yes; insert the new join node into the join tree */ + j->larg = *jtlink1; + *jtlink1 = (Node *) j; + /* Recursively process pulled-up jointree nodes */ + j->rarg = pull_up_sublinks_jointree_recurse(root, + j->rarg, + &child_rels); + + /* + * Now recursively process the pulled-up quals. Any inserted + * joins can get stacked onto either j->larg or j->rarg, + * depending on which rels they reference. + */ + j->quals = pull_up_sublinks_qual_recurse(root, + j->quals, + &j->larg, + available_rels1, + &j->rarg, + child_rels); + /* Return NULL representing constant TRUE */ + return NULL; + } + if (available_rels2 != NULL && + (j = convert_EXISTS_sublink_to_join(root, sublink, false, + available_rels2)) != NULL) + { + /* Yes; insert the new join node into the join tree */ + j->larg = *jtlink2; + *jtlink2 = (Node *) j; + /* Recursively process pulled-up jointree nodes */ + j->rarg = pull_up_sublinks_jointree_recurse(root, + j->rarg, + &child_rels); + + /* + * Now recursively process the pulled-up quals. Any inserted + * joins can get stacked onto either j->larg or j->rarg, + * depending on which rels they reference. + */ + j->quals = pull_up_sublinks_qual_recurse(root, + j->quals, + &j->larg, + available_rels2, + &j->rarg, + child_rels); + /* Return NULL representing constant TRUE */ + return NULL; + } + } + /* Else return it unmodified */ + return node; + } + if (is_notclause(node)) + { + /* If the immediate argument of NOT is EXISTS, try to convert */ + SubLink *sublink = (SubLink *) get_notclausearg((Expr *) node); + JoinExpr *j; + Relids child_rels; + + if (sublink && IsA(sublink, SubLink)) + { + if (sublink->subLinkType == EXISTS_SUBLINK) + { + if ((j = convert_EXISTS_sublink_to_join(root, sublink, true, + available_rels1)) != NULL) + { + /* Yes; insert the new join node into the join tree */ + j->larg = *jtlink1; + *jtlink1 = (Node *) j; + /* Recursively process pulled-up jointree nodes */ + j->rarg = pull_up_sublinks_jointree_recurse(root, + j->rarg, + &child_rels); + + /* + * Now recursively process the pulled-up quals. Because + * we are underneath a NOT, we can't pull up sublinks that + * reference the left-hand stuff, but it's still okay to + * pull up sublinks referencing j->rarg. + */ + j->quals = pull_up_sublinks_qual_recurse(root, + j->quals, + &j->rarg, + child_rels, + NULL, NULL); + /* Return NULL representing constant TRUE */ + return NULL; + } + if (available_rels2 != NULL && + (j = convert_EXISTS_sublink_to_join(root, sublink, true, + available_rels2)) != NULL) + { + /* Yes; insert the new join node into the join tree */ + j->larg = *jtlink2; + *jtlink2 = (Node *) j; + /* Recursively process pulled-up jointree nodes */ + j->rarg = pull_up_sublinks_jointree_recurse(root, + j->rarg, + &child_rels); + + /* + * Now recursively process the pulled-up quals. Because + * we are underneath a NOT, we can't pull up sublinks that + * reference the left-hand stuff, but it's still okay to + * pull up sublinks referencing j->rarg. + */ + j->quals = pull_up_sublinks_qual_recurse(root, + j->quals, + &j->rarg, + child_rels, + NULL, NULL); + /* Return NULL representing constant TRUE */ + return NULL; + } + } + } + /* Else return it unmodified */ + return node; + } + if (is_andclause(node)) + { + /* Recurse into AND clause */ + List *newclauses = NIL; + ListCell *l; + + foreach(l, ((BoolExpr *) node)->args) + { + Node *oldclause = (Node *) lfirst(l); + Node *newclause; + + newclause = pull_up_sublinks_qual_recurse(root, + oldclause, + jtlink1, + available_rels1, + jtlink2, + available_rels2); + if (newclause) + newclauses = lappend(newclauses, newclause); + } + /* We might have got back fewer clauses than we started with */ + if (newclauses == NIL) + return NULL; + else if (list_length(newclauses) == 1) + return (Node *) linitial(newclauses); + else + return (Node *) make_andclause(newclauses); + } + /* Stop if not an AND */ + return node; +} + +/* + * preprocess_function_rtes + * Constant-simplify any FUNCTION RTEs in the FROM clause, and then + * attempt to "inline" any that are set-returning functions. + * + * If an RTE_FUNCTION rtable entry invokes a set-returning function that + * contains just a simple SELECT, we can convert the rtable entry to an + * RTE_SUBQUERY entry exposing the SELECT directly. This is especially + * useful if the subquery can then be "pulled up" for further optimization, + * but we do it even if not, to reduce executor overhead. + * + * This has to be done before we have started to do any optimization of + * subqueries, else any such steps wouldn't get applied to subqueries + * obtained via inlining. However, we do it after pull_up_sublinks + * so that we can inline any functions used in SubLink subselects. + * + * The reason for applying const-simplification at this stage is that + * (a) we'd need to do it anyway to inline a SRF, and (b) by doing it now, + * we can be sure that pull_up_constant_function() will see constants + * if there are constants to be seen. This approach also guarantees + * that every FUNCTION RTE has been const-simplified, allowing planner.c's + * preprocess_expression() to skip doing it again. + * + * Like most of the planner, this feels free to scribble on its input data + * structure. + */ +void +preprocess_function_rtes(PlannerInfo *root) +{ + ListCell *rt; + + foreach(rt, root->parse->rtable) + { + RangeTblEntry *rte = (RangeTblEntry *) lfirst(rt); + + if (rte->rtekind == RTE_FUNCTION) + { + Query *funcquery; + + /* Apply const-simplification */ + rte->functions = (List *) + eval_const_expressions(root, (Node *) rte->functions); + + /* Check safety of expansion, and expand if possible */ + funcquery = inline_set_returning_function(root, rte); + if (funcquery) + { + /* Successful expansion, convert the RTE to a subquery */ + rte->rtekind = RTE_SUBQUERY; + rte->subquery = funcquery; + rte->security_barrier = false; + /* Clear fields that should not be set in a subquery RTE */ + rte->functions = NIL; + rte->funcordinality = false; + } + } + } +} + +/* + * pull_up_subqueries + * Look for subqueries in the rangetable that can be pulled up into + * the parent query. If the subquery has no special features like + * grouping/aggregation then we can merge it into the parent's jointree. + * Also, subqueries that are simple UNION ALL structures can be + * converted into "append relations". + */ +void +pull_up_subqueries(PlannerInfo *root) +{ + /* Top level of jointree must always be a FromExpr */ + Assert(IsA(root->parse->jointree, FromExpr)); + /* Recursion starts with no containing join nor appendrel */ + root->parse->jointree = (FromExpr *) + pull_up_subqueries_recurse(root, (Node *) root->parse->jointree, + NULL, NULL, NULL); + /* We should still have a FromExpr */ + Assert(IsA(root->parse->jointree, FromExpr)); +} + +/* + * pull_up_subqueries_recurse + * Recursive guts of pull_up_subqueries. + * + * This recursively processes the jointree and returns a modified jointree. + * + * If this jointree node is within either side of an outer join, then + * lowest_outer_join references the lowest such JoinExpr node; otherwise + * it is NULL. We use this to constrain the effects of LATERAL subqueries. + * + * If this jointree node is within the nullable side of an outer join, then + * lowest_nulling_outer_join references the lowest such JoinExpr node; + * otherwise it is NULL. This forces use of the PlaceHolderVar mechanism for + * references to non-nullable targetlist items, but only for references above + * that join. + * + * If we are looking at a member subquery of an append relation, + * containing_appendrel describes that relation; else it is NULL. + * This forces use of the PlaceHolderVar mechanism for all non-Var targetlist + * items, and puts some additional restrictions on what can be pulled up. + * + * A tricky aspect of this code is that if we pull up a subquery we have + * to replace Vars that reference the subquery's outputs throughout the + * parent query, including quals attached to jointree nodes above the one + * we are currently processing! We handle this by being careful to maintain + * validity of the jointree structure while recursing, in the following sense: + * whenever we recurse, all qual expressions in the tree must be reachable + * from the top level, in case the recursive call needs to modify them. + * + * Notice also that we can't turn pullup_replace_vars loose on the whole + * jointree, because it'd return a mutated copy of the tree; we have to + * invoke it just on the quals, instead. This behavior is what makes it + * reasonable to pass lowest_outer_join and lowest_nulling_outer_join as + * pointers rather than some more-indirect way of identifying the lowest + * OJs. Likewise, we don't replace append_rel_list members but only their + * substructure, so the containing_appendrel reference is safe to use. + */ +static Node * +pull_up_subqueries_recurse(PlannerInfo *root, Node *jtnode, + JoinExpr *lowest_outer_join, + JoinExpr *lowest_nulling_outer_join, + AppendRelInfo *containing_appendrel) +{ + /* Since this function recurses, it could be driven to stack overflow. */ + check_stack_depth(); + /* Also, since it's a bit expensive, let's check for query cancel. */ + CHECK_FOR_INTERRUPTS(); + + Assert(jtnode != NULL); + if (IsA(jtnode, RangeTblRef)) + { + int varno = ((RangeTblRef *) jtnode)->rtindex; + RangeTblEntry *rte = rt_fetch(varno, root->parse->rtable); + + /* + * Is this a subquery RTE, and if so, is the subquery simple enough to + * pull up? + * + * If we are looking at an append-relation member, we can't pull it up + * unless is_safe_append_member says so. + */ + if (rte->rtekind == RTE_SUBQUERY && + is_simple_subquery(root, rte->subquery, rte, lowest_outer_join) && + (containing_appendrel == NULL || + is_safe_append_member(rte->subquery))) + return pull_up_simple_subquery(root, jtnode, rte, + lowest_outer_join, + lowest_nulling_outer_join, + containing_appendrel); + + /* + * Alternatively, is it a simple UNION ALL subquery? If so, flatten + * into an "append relation". + * + * It's safe to do this regardless of whether this query is itself an + * appendrel member. (If you're thinking we should try to flatten the + * two levels of appendrel together, you're right; but we handle that + * in set_append_rel_pathlist, not here.) + */ + if (rte->rtekind == RTE_SUBQUERY && + is_simple_union_all(rte->subquery)) + return pull_up_simple_union_all(root, jtnode, rte); + + /* + * Or perhaps it's a simple VALUES RTE? + * + * We don't allow VALUES pullup below an outer join nor into an + * appendrel (such cases are impossible anyway at the moment). + */ + if (rte->rtekind == RTE_VALUES && + lowest_outer_join == NULL && + containing_appendrel == NULL && + is_simple_values(root, rte)) + return pull_up_simple_values(root, jtnode, rte); + + /* + * Or perhaps it's a FUNCTION RTE that we could inline? + */ + if (rte->rtekind == RTE_FUNCTION) + return pull_up_constant_function(root, jtnode, rte, + lowest_nulling_outer_join, + containing_appendrel); + + /* Otherwise, do nothing at this node. */ + } + else if (IsA(jtnode, FromExpr)) + { + FromExpr *f = (FromExpr *) jtnode; + ListCell *l; + + Assert(containing_appendrel == NULL); + /* Recursively transform all the child nodes */ + foreach(l, f->fromlist) + { + lfirst(l) = pull_up_subqueries_recurse(root, lfirst(l), + lowest_outer_join, + lowest_nulling_outer_join, + NULL); + } + } + else if (IsA(jtnode, JoinExpr)) + { + JoinExpr *j = (JoinExpr *) jtnode; + + Assert(containing_appendrel == NULL); + /* Recurse, being careful to tell myself when inside outer join */ + switch (j->jointype) + { + case JOIN_INNER: + j->larg = pull_up_subqueries_recurse(root, j->larg, + lowest_outer_join, + lowest_nulling_outer_join, + NULL); + j->rarg = pull_up_subqueries_recurse(root, j->rarg, + lowest_outer_join, + lowest_nulling_outer_join, + NULL); + break; + case JOIN_LEFT: + case JOIN_SEMI: + case JOIN_ANTI: + j->larg = pull_up_subqueries_recurse(root, j->larg, + j, + lowest_nulling_outer_join, + NULL); + j->rarg = pull_up_subqueries_recurse(root, j->rarg, + j, + j, + NULL); + break; + case JOIN_FULL: + j->larg = pull_up_subqueries_recurse(root, j->larg, + j, + j, + NULL); + j->rarg = pull_up_subqueries_recurse(root, j->rarg, + j, + j, + NULL); + break; + case JOIN_RIGHT: + j->larg = pull_up_subqueries_recurse(root, j->larg, + j, + j, + NULL); + j->rarg = pull_up_subqueries_recurse(root, j->rarg, + j, + lowest_nulling_outer_join, + NULL); + break; + default: + elog(ERROR, "unrecognized join type: %d", + (int) j->jointype); + break; + } + } + else + elog(ERROR, "unrecognized node type: %d", + (int) nodeTag(jtnode)); + return jtnode; +} + +/* + * pull_up_simple_subquery + * Attempt to pull up a single simple subquery. + * + * jtnode is a RangeTblRef that has been tentatively identified as a simple + * subquery by pull_up_subqueries. We return the replacement jointree node, + * or jtnode itself if we determine that the subquery can't be pulled up + * after all. + * + * rte is the RangeTblEntry referenced by jtnode. Remaining parameters are + * as for pull_up_subqueries_recurse. + */ +static Node * +pull_up_simple_subquery(PlannerInfo *root, Node *jtnode, RangeTblEntry *rte, + JoinExpr *lowest_outer_join, + JoinExpr *lowest_nulling_outer_join, + AppendRelInfo *containing_appendrel) +{ + Query *parse = root->parse; + int varno = ((RangeTblRef *) jtnode)->rtindex; + Query *subquery; + PlannerInfo *subroot; + int rtoffset; + pullup_replace_vars_context rvcontext; + ListCell *lc; + + /* + * Make a modifiable copy of the subquery to hack on, so that the RTE will + * be left unchanged in case we decide below that we can't pull it up + * after all. + */ + subquery = copyObject(rte->subquery); + + /* + * Create a PlannerInfo data structure for this subquery. + * + * NOTE: the next few steps should match the first processing in + * subquery_planner(). Can we refactor to avoid code duplication, or + * would that just make things uglier? + */ + subroot = makeNode(PlannerInfo); + subroot->parse = subquery; + subroot->glob = root->glob; + subroot->query_level = root->query_level; + subroot->parent_root = root->parent_root; + subroot->plan_params = NIL; + subroot->outer_params = NULL; + subroot->planner_cxt = CurrentMemoryContext; + subroot->init_plans = NIL; + subroot->cte_plan_ids = NIL; + subroot->multiexpr_params = NIL; + subroot->eq_classes = NIL; + subroot->ec_merging_done = false; + subroot->all_result_relids = NULL; + subroot->leaf_result_relids = NULL; + subroot->append_rel_list = NIL; + subroot->row_identity_vars = NIL; + subroot->rowMarks = NIL; + memset(subroot->upper_rels, 0, sizeof(subroot->upper_rels)); + memset(subroot->upper_targets, 0, sizeof(subroot->upper_targets)); + subroot->processed_tlist = NIL; + subroot->update_colnos = NIL; + subroot->grouping_map = NULL; + subroot->minmax_aggs = NIL; + subroot->qual_security_level = 0; + subroot->hasRecursion = false; + subroot->wt_param_id = -1; + subroot->non_recursive_path = NULL; + + /* No CTEs to worry about */ + Assert(subquery->cteList == NIL); + + /* + * If the FROM clause is empty, replace it with a dummy RTE_RESULT RTE, so + * that we don't need so many special cases to deal with that situation. + */ + replace_empty_jointree(subquery); + + /* + * Pull up any SubLinks within the subquery's quals, so that we don't + * leave unoptimized SubLinks behind. + */ + if (subquery->hasSubLinks) + pull_up_sublinks(subroot); + + /* + * Similarly, preprocess its function RTEs to inline any set-returning + * functions in its rangetable. + */ + preprocess_function_rtes(subroot); + + /* + * Recursively pull up the subquery's subqueries, so that + * pull_up_subqueries' processing is complete for its jointree and + * rangetable. + * + * Note: it's okay that the subquery's recursion starts with NULL for + * containing-join info, even if we are within an outer join in the upper + * query; the lower query starts with a clean slate for outer-join + * semantics. Likewise, we needn't pass down appendrel state. + */ + pull_up_subqueries(subroot); + + /* + * Now we must recheck whether the subquery is still simple enough to pull + * up. If not, abandon processing it. + * + * We don't really need to recheck all the conditions involved, but it's + * easier just to keep this "if" looking the same as the one in + * pull_up_subqueries_recurse. + */ + if (is_simple_subquery(root, subquery, rte, lowest_outer_join) && + (containing_appendrel == NULL || is_safe_append_member(subquery))) + { + /* good to go */ + } + else + { + /* + * Give up, return unmodified RangeTblRef. + * + * Note: The work we just did will be redone when the subquery gets + * planned on its own. Perhaps we could avoid that by storing the + * modified subquery back into the rangetable, but I'm not gonna risk + * it now. + */ + return jtnode; + } + + /* + * We must flatten any join alias Vars in the subquery's targetlist, + * because pulling up the subquery's subqueries might have changed their + * expansions into arbitrary expressions, which could affect + * pullup_replace_vars' decisions about whether PlaceHolderVar wrappers + * are needed for tlist entries. (Likely it'd be better to do + * flatten_join_alias_vars on the whole query tree at some earlier stage, + * maybe even in the rewriter; but for now let's just fix this case here.) + */ + subquery->targetList = (List *) + flatten_join_alias_vars(subroot->parse, (Node *) subquery->targetList); + + /* + * Adjust level-0 varnos in subquery so that we can append its rangetable + * to upper query's. We have to fix the subquery's append_rel_list as + * well. + */ + rtoffset = list_length(parse->rtable); + OffsetVarNodes((Node *) subquery, rtoffset, 0); + OffsetVarNodes((Node *) subroot->append_rel_list, rtoffset, 0); + + /* + * Upper-level vars in subquery are now one level closer to their parent + * than before. + */ + IncrementVarSublevelsUp((Node *) subquery, -1, 1); + IncrementVarSublevelsUp((Node *) subroot->append_rel_list, -1, 1); + + /* + * The subquery's targetlist items are now in the appropriate form to + * insert into the top query, except that we may need to wrap them in + * PlaceHolderVars. Set up required context data for pullup_replace_vars. + */ + rvcontext.root = root; + rvcontext.targetlist = subquery->targetList; + rvcontext.target_rte = rte; + if (rte->lateral) + rvcontext.relids = get_relids_in_jointree((Node *) subquery->jointree, + true); + else /* won't need relids */ + rvcontext.relids = NULL; + rvcontext.outer_hasSubLinks = &parse->hasSubLinks; + rvcontext.varno = varno; + /* these flags will be set below, if needed */ + rvcontext.need_phvs = false; + rvcontext.wrap_non_vars = false; + /* initialize cache array with indexes 0 .. length(tlist) */ + rvcontext.rv_cache = palloc0((list_length(subquery->targetList) + 1) * + sizeof(Node *)); + + /* + * If we are under an outer join then non-nullable items and lateral + * references may have to be turned into PlaceHolderVars. + */ + if (lowest_nulling_outer_join != NULL) + rvcontext.need_phvs = true; + + /* + * If we are dealing with an appendrel member then anything that's not a + * simple Var has to be turned into a PlaceHolderVar. We force this to + * ensure that what we pull up doesn't get merged into a surrounding + * expression during later processing and then fail to match the + * expression actually available from the appendrel. + */ + if (containing_appendrel != NULL) + { + rvcontext.need_phvs = true; + rvcontext.wrap_non_vars = true; + } + + /* + * If the parent query uses grouping sets, we need a PlaceHolderVar for + * anything that's not a simple Var. Again, this ensures that expressions + * retain their separate identity so that they will match grouping set + * columns when appropriate. (It'd be sufficient to wrap values used in + * grouping set columns, and do so only in non-aggregated portions of the + * tlist and havingQual, but that would require a lot of infrastructure + * that pullup_replace_vars hasn't currently got.) + */ + if (parse->groupingSets) + { + rvcontext.need_phvs = true; + rvcontext.wrap_non_vars = true; + } + + /* + * Replace all of the top query's references to the subquery's outputs + * with copies of the adjusted subtlist items, being careful not to + * replace any of the jointree structure. + */ + perform_pullup_replace_vars(root, &rvcontext, + lowest_nulling_outer_join, + containing_appendrel); + + /* + * If the subquery had a LATERAL marker, propagate that to any of its + * child RTEs that could possibly now contain lateral cross-references. + * The children might or might not contain any actual lateral + * cross-references, but we have to mark the pulled-up child RTEs so that + * later planner stages will check for such. + */ + if (rte->lateral) + { + foreach(lc, subquery->rtable) + { + RangeTblEntry *child_rte = (RangeTblEntry *) lfirst(lc); + + switch (child_rte->rtekind) + { + case RTE_RELATION: + if (child_rte->tablesample) + child_rte->lateral = true; + break; + case RTE_SUBQUERY: + case RTE_FUNCTION: + case RTE_VALUES: + case RTE_TABLEFUNC: + child_rte->lateral = true; + break; + case RTE_JOIN: + case RTE_CTE: + case RTE_NAMEDTUPLESTORE: + case RTE_RESULT: + /* these can't contain any lateral references */ + break; + } + } + } + + /* + * Now append the adjusted rtable entries to upper query. (We hold off + * until after fixing the upper rtable entries; no point in running that + * code on the subquery ones too.) + */ + parse->rtable = list_concat(parse->rtable, subquery->rtable); + + /* + * Pull up any FOR UPDATE/SHARE markers, too. (OffsetVarNodes already + * adjusted the marker rtindexes, so just concat the lists.) + */ + parse->rowMarks = list_concat(parse->rowMarks, subquery->rowMarks); + + /* + * We also have to fix the relid sets of any PlaceHolderVar nodes in the + * parent query. (This could perhaps be done by pullup_replace_vars(), + * but it seems cleaner to use two passes.) Note in particular that any + * PlaceHolderVar nodes just created by pullup_replace_vars() will be + * adjusted, so having created them with the subquery's varno is correct. + * + * Likewise, relids appearing in AppendRelInfo nodes have to be fixed. We + * already checked that this won't require introducing multiple subrelids + * into the single-slot AppendRelInfo structs. + */ + if (parse->hasSubLinks || root->glob->lastPHId != 0 || + root->append_rel_list) + { + Relids subrelids; + + subrelids = get_relids_in_jointree((Node *) subquery->jointree, false); + substitute_phv_relids((Node *) parse, varno, subrelids); + fix_append_rel_relids(root->append_rel_list, varno, subrelids); + } + + /* + * And now add subquery's AppendRelInfos to our list. + */ + root->append_rel_list = list_concat(root->append_rel_list, + subroot->append_rel_list); + + /* + * We don't have to do the equivalent bookkeeping for outer-join info, + * because that hasn't been set up yet. placeholder_list likewise. + */ + Assert(root->join_info_list == NIL); + Assert(subroot->join_info_list == NIL); + Assert(root->placeholder_list == NIL); + Assert(subroot->placeholder_list == NIL); + + /* + * We no longer need the RTE's copy of the subquery's query tree. Getting + * rid of it saves nothing in particular so far as this level of query is + * concerned; but if this query level is in turn pulled up into a parent, + * we'd waste cycles copying the now-unused query tree. + */ + rte->subquery = NULL; + + /* + * Miscellaneous housekeeping. + * + * Although replace_rte_variables() faithfully updated parse->hasSubLinks + * if it copied any SubLinks out of the subquery's targetlist, we still + * could have SubLinks added to the query in the expressions of FUNCTION + * and VALUES RTEs copied up from the subquery. So it's necessary to copy + * subquery->hasSubLinks anyway. Perhaps this can be improved someday. + */ + parse->hasSubLinks |= subquery->hasSubLinks; + + /* If subquery had any RLS conditions, now main query does too */ + parse->hasRowSecurity |= subquery->hasRowSecurity; + + /* + * subquery won't be pulled up if it hasAggs, hasWindowFuncs, or + * hasTargetSRFs, so no work needed on those flags + */ + + /* + * Return the adjusted subquery jointree to replace the RangeTblRef entry + * in parent's jointree; or, if the FromExpr is degenerate, just return + * its single member. + */ + Assert(IsA(subquery->jointree, FromExpr)); + Assert(subquery->jointree->fromlist != NIL); + if (subquery->jointree->quals == NULL && + list_length(subquery->jointree->fromlist) == 1) + return (Node *) linitial(subquery->jointree->fromlist); + + return (Node *) subquery->jointree; +} + +/* + * pull_up_simple_union_all + * Pull up a single simple UNION ALL subquery. + * + * jtnode is a RangeTblRef that has been identified as a simple UNION ALL + * subquery by pull_up_subqueries. We pull up the leaf subqueries and + * build an "append relation" for the union set. The result value is just + * jtnode, since we don't actually need to change the query jointree. + */ +static Node * +pull_up_simple_union_all(PlannerInfo *root, Node *jtnode, RangeTblEntry *rte) +{ + int varno = ((RangeTblRef *) jtnode)->rtindex; + Query *subquery = rte->subquery; + int rtoffset = list_length(root->parse->rtable); + List *rtable; + + /* + * Make a modifiable copy of the subquery's rtable, so we can adjust + * upper-level Vars in it. There are no such Vars in the setOperations + * tree proper, so fixing the rtable should be sufficient. + */ + rtable = copyObject(subquery->rtable); + + /* + * Upper-level vars in subquery are now one level closer to their parent + * than before. We don't have to worry about offsetting varnos, though, + * because the UNION leaf queries can't cross-reference each other. + */ + IncrementVarSublevelsUp_rtable(rtable, -1, 1); + + /* + * If the UNION ALL subquery had a LATERAL marker, propagate that to all + * its children. The individual children might or might not contain any + * actual lateral cross-references, but we have to mark the pulled-up + * child RTEs so that later planner stages will check for such. + */ + if (rte->lateral) + { + ListCell *rt; + + foreach(rt, rtable) + { + RangeTblEntry *child_rte = (RangeTblEntry *) lfirst(rt); + + Assert(child_rte->rtekind == RTE_SUBQUERY); + child_rte->lateral = true; + } + } + + /* + * Append child RTEs to parent rtable. + */ + root->parse->rtable = list_concat(root->parse->rtable, rtable); + + /* + * Recursively scan the subquery's setOperations tree and add + * AppendRelInfo nodes for leaf subqueries to the parent's + * append_rel_list. Also apply pull_up_subqueries to the leaf subqueries. + */ + Assert(subquery->setOperations); + pull_up_union_leaf_queries(subquery->setOperations, root, varno, subquery, + rtoffset); + + /* + * Mark the parent as an append relation. + */ + rte->inh = true; + + return jtnode; +} + +/* + * pull_up_union_leaf_queries -- recursive guts of pull_up_simple_union_all + * + * Build an AppendRelInfo for each leaf query in the setop tree, and then + * apply pull_up_subqueries to the leaf query. + * + * Note that setOpQuery is the Query containing the setOp node, whose tlist + * contains references to all the setop output columns. When called from + * pull_up_simple_union_all, this is *not* the same as root->parse, which is + * the parent Query we are pulling up into. + * + * parentRTindex is the appendrel parent's index in root->parse->rtable. + * + * The child RTEs have already been copied to the parent. childRToffset + * tells us where in the parent's range table they were copied. When called + * from flatten_simple_union_all, childRToffset is 0 since the child RTEs + * were already in root->parse->rtable and no RT index adjustment is needed. + */ +static void +pull_up_union_leaf_queries(Node *setOp, PlannerInfo *root, int parentRTindex, + Query *setOpQuery, int childRToffset) +{ + if (IsA(setOp, RangeTblRef)) + { + RangeTblRef *rtr = (RangeTblRef *) setOp; + int childRTindex; + AppendRelInfo *appinfo; + + /* + * Calculate the index in the parent's range table + */ + childRTindex = childRToffset + rtr->rtindex; + + /* + * Build a suitable AppendRelInfo, and attach to parent's list. + */ + appinfo = makeNode(AppendRelInfo); + appinfo->parent_relid = parentRTindex; + appinfo->child_relid = childRTindex; + appinfo->parent_reltype = InvalidOid; + appinfo->child_reltype = InvalidOid; + make_setop_translation_list(setOpQuery, childRTindex, appinfo); + appinfo->parent_reloid = InvalidOid; + root->append_rel_list = lappend(root->append_rel_list, appinfo); + + /* + * Recursively apply pull_up_subqueries to the new child RTE. (We + * must build the AppendRelInfo first, because this will modify it.) + * Note that we can pass NULL for containing-join info even if we're + * actually under an outer join, because the child's expressions + * aren't going to propagate up to the join. Also, we ignore the + * possibility that pull_up_subqueries_recurse() returns a different + * jointree node than what we pass it; if it does, the important thing + * is that it replaced the child relid in the AppendRelInfo node. + */ + rtr = makeNode(RangeTblRef); + rtr->rtindex = childRTindex; + (void) pull_up_subqueries_recurse(root, (Node *) rtr, + NULL, NULL, appinfo); + } + else if (IsA(setOp, SetOperationStmt)) + { + SetOperationStmt *op = (SetOperationStmt *) setOp; + + /* Recurse to reach leaf queries */ + pull_up_union_leaf_queries(op->larg, root, parentRTindex, setOpQuery, + childRToffset); + pull_up_union_leaf_queries(op->rarg, root, parentRTindex, setOpQuery, + childRToffset); + } + else + { + elog(ERROR, "unrecognized node type: %d", + (int) nodeTag(setOp)); + } +} + +/* + * make_setop_translation_list + * Build the list of translations from parent Vars to child Vars for + * a UNION ALL member. (At this point it's just a simple list of + * referencing Vars, but if we succeed in pulling up the member + * subquery, the Vars will get replaced by pulled-up expressions.) + * Also create the rather trivial reverse-translation array. + */ +static void +make_setop_translation_list(Query *query, int newvarno, + AppendRelInfo *appinfo) +{ + List *vars = NIL; + AttrNumber *pcolnos; + ListCell *l; + + /* Initialize reverse-translation array with all entries zero */ + /* (entries for resjunk columns will stay that way) */ + appinfo->num_child_cols = list_length(query->targetList); + appinfo->parent_colnos = pcolnos = + (AttrNumber *) palloc0(appinfo->num_child_cols * sizeof(AttrNumber)); + + foreach(l, query->targetList) + { + TargetEntry *tle = (TargetEntry *) lfirst(l); + + if (tle->resjunk) + continue; + + vars = lappend(vars, makeVarFromTargetEntry(newvarno, tle)); + pcolnos[tle->resno - 1] = tle->resno; + } + + appinfo->translated_vars = vars; +} + +/* + * is_simple_subquery + * Check a subquery in the range table to see if it's simple enough + * to pull up into the parent query. + * + * rte is the RTE_SUBQUERY RangeTblEntry that contained the subquery. + * (Note subquery is not necessarily equal to rte->subquery; it could be a + * processed copy of that.) + * lowest_outer_join is the lowest outer join above the subquery, or NULL. + */ +static bool +is_simple_subquery(PlannerInfo *root, Query *subquery, RangeTblEntry *rte, + JoinExpr *lowest_outer_join) +{ + /* + * Let's just make sure it's a valid subselect ... + */ + if (!IsA(subquery, Query) || + subquery->commandType != CMD_SELECT) + elog(ERROR, "subquery is bogus"); + + /* + * Can't currently pull up a query with setops (unless it's simple UNION + * ALL, which is handled by a different code path). Maybe after querytree + * redesign... + */ + if (subquery->setOperations) + return false; + + /* + * Can't pull up a subquery involving grouping, aggregation, SRFs, + * sorting, limiting, or WITH. (XXX WITH could possibly be allowed later) + * + * We also don't pull up a subquery that has explicit FOR UPDATE/SHARE + * clauses, because pullup would cause the locking to occur semantically + * higher than it should. Implicit FOR UPDATE/SHARE is okay because in + * that case the locking was originally declared in the upper query + * anyway. + */ + if (subquery->hasAggs || + subquery->hasWindowFuncs || + subquery->hasTargetSRFs || + subquery->groupClause || + subquery->groupingSets || + subquery->havingQual || + subquery->sortClause || + subquery->distinctClause || + subquery->limitOffset || + subquery->limitCount || + subquery->hasForUpdate || + subquery->cteList) + return false; + + /* + * Don't pull up if the RTE represents a security-barrier view; we + * couldn't prevent information leakage once the RTE's Vars are scattered + * about in the upper query. + */ + if (rte->security_barrier) + return false; + + /* + * If the subquery is LATERAL, check for pullup restrictions from that. + */ + if (rte->lateral) + { + bool restricted; + Relids safe_upper_varnos; + + /* + * The subquery's WHERE and JOIN/ON quals mustn't contain any lateral + * references to rels outside a higher outer join (including the case + * where the outer join is within the subquery itself). In such a + * case, pulling up would result in a situation where we need to + * postpone quals from below an outer join to above it, which is + * probably completely wrong and in any case is a complication that + * doesn't seem worth addressing at the moment. + */ + if (lowest_outer_join != NULL) + { + restricted = true; + safe_upper_varnos = get_relids_in_jointree((Node *) lowest_outer_join, + true); + } + else + { + restricted = false; + safe_upper_varnos = NULL; /* doesn't matter */ + } + + if (jointree_contains_lateral_outer_refs(root, + (Node *) subquery->jointree, + restricted, safe_upper_varnos)) + return false; + + /* + * If there's an outer join above the LATERAL subquery, also disallow + * pullup if the subquery's targetlist has any references to rels + * outside the outer join, since these might get pulled into quals + * above the subquery (but in or below the outer join) and then lead + * to qual-postponement issues similar to the case checked for above. + * (We wouldn't need to prevent pullup if no such references appear in + * outer-query quals, but we don't have enough info here to check + * that. Also, maybe this restriction could be removed if we forced + * such refs to be wrapped in PlaceHolderVars, even when they're below + * the nearest outer join? But it's a pretty hokey usage, so not + * clear this is worth sweating over.) + */ + if (lowest_outer_join != NULL) + { + Relids lvarnos = pull_varnos_of_level(root, + (Node *) subquery->targetList, + 1); + + if (!bms_is_subset(lvarnos, safe_upper_varnos)) + return false; + } + } + + /* + * Don't pull up a subquery that has any volatile functions in its + * targetlist. Otherwise we might introduce multiple evaluations of these + * functions, if they get copied to multiple places in the upper query, + * leading to surprising results. (Note: the PlaceHolderVar mechanism + * doesn't quite guarantee single evaluation; else we could pull up anyway + * and just wrap such items in PlaceHolderVars ...) + */ + if (contain_volatile_functions((Node *) subquery->targetList)) + return false; + + return true; +} + +/* + * pull_up_simple_values + * Pull up a single simple VALUES RTE. + * + * jtnode is a RangeTblRef that has been identified as a simple VALUES RTE + * by pull_up_subqueries. We always return a RangeTblRef representing a + * RESULT RTE to replace it (all failure cases should have been detected by + * is_simple_values()). Actually, what we return is just jtnode, because + * we replace the VALUES RTE in the rangetable with the RESULT RTE. + * + * rte is the RangeTblEntry referenced by jtnode. Because of the limited + * possible usage of VALUES RTEs, we do not need the remaining parameters + * of pull_up_subqueries_recurse. + */ +static Node * +pull_up_simple_values(PlannerInfo *root, Node *jtnode, RangeTblEntry *rte) +{ + Query *parse = root->parse; + int varno = ((RangeTblRef *) jtnode)->rtindex; + List *values_list; + List *tlist; + AttrNumber attrno; + pullup_replace_vars_context rvcontext; + ListCell *lc; + + Assert(rte->rtekind == RTE_VALUES); + Assert(list_length(rte->values_lists) == 1); + + /* + * Need a modifiable copy of the VALUES list to hack on, just in case it's + * multiply referenced. + */ + values_list = copyObject(linitial(rte->values_lists)); + + /* + * The VALUES RTE can't contain any Vars of level zero, let alone any that + * are join aliases, so no need to flatten join alias Vars. + */ + Assert(!contain_vars_of_level((Node *) values_list, 0)); + + /* + * Set up required context data for pullup_replace_vars. In particular, + * we have to make the VALUES list look like a subquery targetlist. + */ + tlist = NIL; + attrno = 1; + foreach(lc, values_list) + { + tlist = lappend(tlist, + makeTargetEntry((Expr *) lfirst(lc), + attrno, + NULL, + false)); + attrno++; + } + rvcontext.root = root; + rvcontext.targetlist = tlist; + rvcontext.target_rte = rte; + rvcontext.relids = NULL; + rvcontext.outer_hasSubLinks = &parse->hasSubLinks; + rvcontext.varno = varno; + rvcontext.need_phvs = false; + rvcontext.wrap_non_vars = false; + /* initialize cache array with indexes 0 .. length(tlist) */ + rvcontext.rv_cache = palloc0((list_length(tlist) + 1) * + sizeof(Node *)); + + /* + * Replace all of the top query's references to the RTE's outputs with + * copies of the adjusted VALUES expressions, being careful not to replace + * any of the jointree structure. We can assume there's no outer joins or + * appendrels in the dummy Query that surrounds a VALUES RTE. + */ + perform_pullup_replace_vars(root, &rvcontext, NULL, NULL); + + /* + * There should be no appendrels to fix, nor any outer joins and hence no + * PlaceHolderVars. + */ + Assert(root->append_rel_list == NIL); + Assert(root->join_info_list == NIL); + Assert(root->placeholder_list == NIL); + + /* + * Replace the VALUES RTE with a RESULT RTE. The VALUES RTE is the only + * rtable entry in the current query level, so this is easy. + */ + Assert(list_length(parse->rtable) == 1); + + /* Create suitable RTE */ + rte = makeNode(RangeTblEntry); + rte->rtekind = RTE_RESULT; + rte->eref = makeAlias("*RESULT*", NIL); + + /* Replace rangetable */ + parse->rtable = list_make1(rte); + + /* We could manufacture a new RangeTblRef, but the one we have is fine */ + Assert(varno == 1); + + return jtnode; +} + +/* + * is_simple_values + * Check a VALUES RTE in the range table to see if it's simple enough + * to pull up into the parent query. + * + * rte is the RTE_VALUES RangeTblEntry to check. + */ +static bool +is_simple_values(PlannerInfo *root, RangeTblEntry *rte) +{ + Assert(rte->rtekind == RTE_VALUES); + + /* + * There must be exactly one VALUES list, else it's not semantically + * correct to replace the VALUES RTE with a RESULT RTE, nor would we have + * a unique set of expressions to substitute into the parent query. + */ + if (list_length(rte->values_lists) != 1) + return false; + + /* + * Because VALUES can't appear under an outer join (or at least, we won't + * try to pull it up if it does), we need not worry about LATERAL, nor + * about validity of PHVs for the VALUES' outputs. + */ + + /* + * Don't pull up a VALUES that contains any set-returning or volatile + * functions. The considerations here are basically identical to the + * restrictions on a pull-able subquery's targetlist. + */ + if (expression_returns_set((Node *) rte->values_lists) || + contain_volatile_functions((Node *) rte->values_lists)) + return false; + + /* + * Do not pull up a VALUES that's not the only RTE in its parent query. + * This is actually the only case that the parser will generate at the + * moment, and assuming this is true greatly simplifies + * pull_up_simple_values(). + */ + if (list_length(root->parse->rtable) != 1 || + rte != (RangeTblEntry *) linitial(root->parse->rtable)) + return false; + + return true; +} + +/* + * pull_up_constant_function + * Pull up an RTE_FUNCTION expression that was simplified to a constant. + * + * jtnode is a RangeTblRef that has been identified as a FUNCTION RTE by + * pull_up_subqueries. If its expression is just a Const, hoist that value + * up into the parent query, and replace the RTE_FUNCTION with RTE_RESULT. + * + * In principle we could pull up any immutable expression, but we don't. + * That might result in multiple evaluations of the expression, which could + * be costly if it's not just a Const. Also, the main value of this is + * to let the constant participate in further const-folding, and of course + * that won't happen for a non-Const. + * + * The pulled-up value might need to be wrapped in a PlaceHolderVar if the + * RTE is below an outer join or is part of an appendrel; the extra + * parameters show whether that's needed. + */ +static Node * +pull_up_constant_function(PlannerInfo *root, Node *jtnode, + RangeTblEntry *rte, + JoinExpr *lowest_nulling_outer_join, + AppendRelInfo *containing_appendrel) +{ + Query *parse = root->parse; + RangeTblFunction *rtf; + TypeFuncClass functypclass; + Oid funcrettype; + TupleDesc tupdesc; + pullup_replace_vars_context rvcontext; + + /* Fail if the RTE has ORDINALITY - we don't implement that here. */ + if (rte->funcordinality) + return jtnode; + + /* Fail if RTE isn't a single, simple Const expr */ + if (list_length(rte->functions) != 1) + return jtnode; + rtf = linitial_node(RangeTblFunction, rte->functions); + if (!IsA(rtf->funcexpr, Const)) + return jtnode; + + /* + * If the function's result is not a scalar, we punt. In principle we + * could break the composite constant value apart into per-column + * constants, but for now it seems not worth the work. + */ + if (rtf->funccolcount != 1) + return jtnode; /* definitely composite */ + + functypclass = get_expr_result_type(rtf->funcexpr, + &funcrettype, + &tupdesc); + if (functypclass != TYPEFUNC_SCALAR) + return jtnode; /* must be a one-column composite type */ + + /* Create context for applying pullup_replace_vars */ + rvcontext.root = root; + rvcontext.targetlist = list_make1(makeTargetEntry((Expr *) rtf->funcexpr, + 1, /* resno */ + NULL, /* resname */ + false)); /* resjunk */ + rvcontext.target_rte = rte; + + /* + * Since this function was reduced to a Const, it doesn't contain any + * lateral references, even if it's marked as LATERAL. This means we + * don't need to fill relids. + */ + rvcontext.relids = NULL; + + rvcontext.outer_hasSubLinks = &parse->hasSubLinks; + rvcontext.varno = ((RangeTblRef *) jtnode)->rtindex; + /* these flags will be set below, if needed */ + rvcontext.need_phvs = false; + rvcontext.wrap_non_vars = false; + /* initialize cache array with indexes 0 .. length(tlist) */ + rvcontext.rv_cache = palloc0((list_length(rvcontext.targetlist) + 1) * + sizeof(Node *)); + + /* + * If we are under an outer join then non-nullable items and lateral + * references may have to be turned into PlaceHolderVars. + */ + if (lowest_nulling_outer_join != NULL) + rvcontext.need_phvs = true; + + /* + * If we are dealing with an appendrel member then anything that's not a + * simple Var has to be turned into a PlaceHolderVar. (See comments in + * pull_up_simple_subquery().) + */ + if (containing_appendrel != NULL) + { + rvcontext.need_phvs = true; + rvcontext.wrap_non_vars = true; + } + + /* + * If the parent query uses grouping sets, we need a PlaceHolderVar for + * anything that's not a simple Var. + */ + if (parse->groupingSets) + { + rvcontext.need_phvs = true; + rvcontext.wrap_non_vars = true; + } + + /* + * Replace all of the top query's references to the RTE's output with + * copies of the funcexpr, being careful not to replace any of the + * jointree structure. + */ + perform_pullup_replace_vars(root, &rvcontext, + lowest_nulling_outer_join, + containing_appendrel); + + /* + * We don't need to bother with changing PlaceHolderVars in the parent + * query. Their references to the RT index are still good for now, and + * will get removed later if we're able to drop the RTE_RESULT. + */ + + /* + * Convert the RTE to be RTE_RESULT type, signifying that we don't need to + * scan it anymore, and zero out RTE_FUNCTION-specific fields. Also make + * sure the RTE is not marked LATERAL, since elsewhere we don't expect + * RTE_RESULTs to be LATERAL. + */ + rte->rtekind = RTE_RESULT; + rte->functions = NIL; + rte->lateral = false; + + /* + * We can reuse the RangeTblRef node. + */ + return jtnode; +} + +/* + * is_simple_union_all + * Check a subquery to see if it's a simple UNION ALL. + * + * We require all the setops to be UNION ALL (no mixing) and there can't be + * any datatype coercions involved, ie, all the leaf queries must emit the + * same datatypes. + */ +static bool +is_simple_union_all(Query *subquery) +{ + SetOperationStmt *topop; + + /* Let's just make sure it's a valid subselect ... */ + if (!IsA(subquery, Query) || + subquery->commandType != CMD_SELECT) + elog(ERROR, "subquery is bogus"); + + /* Is it a set-operation query at all? */ + topop = castNode(SetOperationStmt, subquery->setOperations); + if (!topop) + return false; + + /* Can't handle ORDER BY, LIMIT/OFFSET, locking, or WITH */ + if (subquery->sortClause || + subquery->limitOffset || + subquery->limitCount || + subquery->rowMarks || + subquery->cteList) + return false; + + /* Recursively check the tree of set operations */ + return is_simple_union_all_recurse((Node *) topop, subquery, + topop->colTypes); +} + +static bool +is_simple_union_all_recurse(Node *setOp, Query *setOpQuery, List *colTypes) +{ + /* Since this function recurses, it could be driven to stack overflow. */ + check_stack_depth(); + + if (IsA(setOp, RangeTblRef)) + { + RangeTblRef *rtr = (RangeTblRef *) setOp; + RangeTblEntry *rte = rt_fetch(rtr->rtindex, setOpQuery->rtable); + Query *subquery = rte->subquery; + + Assert(subquery != NULL); + + /* Leaf nodes are OK if they match the toplevel column types */ + /* We don't have to compare typmods or collations here */ + return tlist_same_datatypes(subquery->targetList, colTypes, true); + } + else if (IsA(setOp, SetOperationStmt)) + { + SetOperationStmt *op = (SetOperationStmt *) setOp; + + /* Must be UNION ALL */ + if (op->op != SETOP_UNION || !op->all) + return false; + + /* Recurse to check inputs */ + return is_simple_union_all_recurse(op->larg, setOpQuery, colTypes) && + is_simple_union_all_recurse(op->rarg, setOpQuery, colTypes); + } + else + { + elog(ERROR, "unrecognized node type: %d", + (int) nodeTag(setOp)); + return false; /* keep compiler quiet */ + } +} + +/* + * is_safe_append_member + * Check a subquery that is a leaf of a UNION ALL appendrel to see if it's + * safe to pull up. + */ +static bool +is_safe_append_member(Query *subquery) +{ + FromExpr *jtnode; + + /* + * It's only safe to pull up the child if its jointree contains exactly + * one RTE, else the AppendRelInfo data structure breaks. The one base RTE + * could be buried in several levels of FromExpr, however. Also, if the + * child's jointree is completely empty, we can pull up because + * pull_up_simple_subquery will insert a single RTE_RESULT RTE instead. + * + * Also, the child can't have any WHERE quals because there's no place to + * put them in an appendrel. (This is a bit annoying...) If we didn't + * need to check this, we'd just test whether get_relids_in_jointree() + * yields a singleton set, to be more consistent with the coding of + * fix_append_rel_relids(). + */ + jtnode = subquery->jointree; + Assert(IsA(jtnode, FromExpr)); + /* Check the completely-empty case */ + if (jtnode->fromlist == NIL && jtnode->quals == NULL) + return true; + /* Check the more general case */ + while (IsA(jtnode, FromExpr)) + { + if (jtnode->quals != NULL) + return false; + if (list_length(jtnode->fromlist) != 1) + return false; + jtnode = linitial(jtnode->fromlist); + } + if (!IsA(jtnode, RangeTblRef)) + return false; + + return true; +} + +/* + * jointree_contains_lateral_outer_refs + * Check for disallowed lateral references in a jointree's quals + * + * If restricted is false, all level-1 Vars are allowed (but we still must + * search the jointree, since it might contain outer joins below which there + * will be restrictions). If restricted is true, return true when any qual + * in the jointree contains level-1 Vars coming from outside the rels listed + * in safe_upper_varnos. + */ +static bool +jointree_contains_lateral_outer_refs(PlannerInfo *root, Node *jtnode, + bool restricted, + Relids safe_upper_varnos) +{ + if (jtnode == NULL) + return false; + if (IsA(jtnode, RangeTblRef)) + return false; + else if (IsA(jtnode, FromExpr)) + { + FromExpr *f = (FromExpr *) jtnode; + ListCell *l; + + /* First, recurse to check child joins */ + foreach(l, f->fromlist) + { + if (jointree_contains_lateral_outer_refs(root, + lfirst(l), + restricted, + safe_upper_varnos)) + return true; + } + + /* Then check the top-level quals */ + if (restricted && + !bms_is_subset(pull_varnos_of_level(root, f->quals, 1), + safe_upper_varnos)) + return true; + } + else if (IsA(jtnode, JoinExpr)) + { + JoinExpr *j = (JoinExpr *) jtnode; + + /* + * If this is an outer join, we mustn't allow any upper lateral + * references in or below it. + */ + if (j->jointype != JOIN_INNER) + { + restricted = true; + safe_upper_varnos = NULL; + } + + /* Check the child joins */ + if (jointree_contains_lateral_outer_refs(root, + j->larg, + restricted, + safe_upper_varnos)) + return true; + if (jointree_contains_lateral_outer_refs(root, + j->rarg, + restricted, + safe_upper_varnos)) + return true; + + /* Check the JOIN's qual clauses */ + if (restricted && + !bms_is_subset(pull_varnos_of_level(root, j->quals, 1), + safe_upper_varnos)) + return true; + } + else + elog(ERROR, "unrecognized node type: %d", + (int) nodeTag(jtnode)); + return false; +} + +/* + * Perform pullup_replace_vars everyplace it's needed in the query tree. + * + * Caller has already filled *rvcontext with data describing what to + * substitute for Vars referencing the target subquery. In addition + * we need the identity of the lowest outer join that can null the + * target subquery, and its containing appendrel if any. + */ +static void +perform_pullup_replace_vars(PlannerInfo *root, + pullup_replace_vars_context *rvcontext, + JoinExpr *lowest_nulling_outer_join, + AppendRelInfo *containing_appendrel) +{ + Query *parse = root->parse; + ListCell *lc; + + /* + * Replace all of the top query's references to the subquery's outputs + * with copies of the adjusted subtlist items, being careful not to + * replace any of the jointree structure. (This'd be a lot cleaner if we + * could use query_tree_mutator.) We have to use PHVs in the targetList, + * returningList, and havingQual, since those are certainly above any + * outer join. replace_vars_in_jointree tracks its location in the + * jointree and uses PHVs or not appropriately. + */ + parse->targetList = (List *) + pullup_replace_vars((Node *) parse->targetList, rvcontext); + parse->returningList = (List *) + pullup_replace_vars((Node *) parse->returningList, rvcontext); + + foreach(lc, parse->windowClause) + { + WindowClause *wc = lfirst_node(WindowClause, lc); + + if (wc->runCondition != NIL) + wc->runCondition = (List *) + pullup_replace_vars((Node *) wc->runCondition, rvcontext); + } + if (parse->onConflict) + { + parse->onConflict->onConflictSet = (List *) + pullup_replace_vars((Node *) parse->onConflict->onConflictSet, + rvcontext); + parse->onConflict->onConflictWhere = + pullup_replace_vars(parse->onConflict->onConflictWhere, + rvcontext); + + /* + * We assume ON CONFLICT's arbiterElems, arbiterWhere, exclRelTlist + * can't contain any references to a subquery. + */ + } + if (parse->mergeActionList) + { + foreach(lc, parse->mergeActionList) + { + MergeAction *action = lfirst(lc); + + action->qual = pullup_replace_vars(action->qual, rvcontext); + action->targetList = (List *) + pullup_replace_vars((Node *) action->targetList, rvcontext); + } + } + replace_vars_in_jointree((Node *) parse->jointree, rvcontext, + lowest_nulling_outer_join); + Assert(parse->setOperations == NULL); + parse->havingQual = pullup_replace_vars(parse->havingQual, rvcontext); + + /* + * Replace references in the translated_vars lists of appendrels. When + * pulling up an appendrel member, we do not need PHVs in the list of the + * parent appendrel --- there isn't any outer join between. Elsewhere, + * use PHVs for safety. (This analysis could be made tighter but it seems + * unlikely to be worth much trouble.) + */ + foreach(lc, root->append_rel_list) + { + AppendRelInfo *appinfo = (AppendRelInfo *) lfirst(lc); + bool save_need_phvs = rvcontext->need_phvs; + + if (appinfo == containing_appendrel) + rvcontext->need_phvs = false; + appinfo->translated_vars = (List *) + pullup_replace_vars((Node *) appinfo->translated_vars, rvcontext); + rvcontext->need_phvs = save_need_phvs; + } + + /* + * Replace references in the joinaliasvars lists of join RTEs. + * + * You might think that we could avoid using PHVs for alias vars of joins + * below lowest_nulling_outer_join, but that doesn't work because the + * alias vars could be referenced above that join; we need the PHVs to be + * present in such references after the alias vars get flattened. (It + * might be worth trying to be smarter here, someday.) + */ + foreach(lc, parse->rtable) + { + RangeTblEntry *otherrte = (RangeTblEntry *) lfirst(lc); + + if (otherrte->rtekind == RTE_JOIN) + otherrte->joinaliasvars = (List *) + pullup_replace_vars((Node *) otherrte->joinaliasvars, + rvcontext); + } +} + +/* + * Helper routine for perform_pullup_replace_vars: do pullup_replace_vars on + * every expression in the jointree, without changing the jointree structure + * itself. Ugly, but there's no other way... + * + * If we are at or below lowest_nulling_outer_join, we can suppress use of + * PlaceHolderVars wrapped around the replacement expressions. + */ +static void +replace_vars_in_jointree(Node *jtnode, + pullup_replace_vars_context *context, + JoinExpr *lowest_nulling_outer_join) +{ + if (jtnode == NULL) + return; + if (IsA(jtnode, RangeTblRef)) + { + /* + * If the RangeTblRef refers to a LATERAL subquery (that isn't the + * same subquery we're pulling up), it might contain references to the + * target subquery, which we must replace. We drive this from the + * jointree scan, rather than a scan of the rtable, for a couple of + * reasons: we can avoid processing no-longer-referenced RTEs, and we + * can use the appropriate setting of need_phvs depending on whether + * the RTE is above possibly-nulling outer joins or not. + */ + int varno = ((RangeTblRef *) jtnode)->rtindex; + + if (varno != context->varno) /* ignore target subquery itself */ + { + RangeTblEntry *rte = rt_fetch(varno, context->root->parse->rtable); + + Assert(rte != context->target_rte); + if (rte->lateral) + { + switch (rte->rtekind) + { + case RTE_RELATION: + /* shouldn't be marked LATERAL unless tablesample */ + Assert(rte->tablesample); + rte->tablesample = (TableSampleClause *) + pullup_replace_vars((Node *) rte->tablesample, + context); + break; + case RTE_SUBQUERY: + rte->subquery = + pullup_replace_vars_subquery(rte->subquery, + context); + break; + case RTE_FUNCTION: + rte->functions = (List *) + pullup_replace_vars((Node *) rte->functions, + context); + break; + case RTE_TABLEFUNC: + rte->tablefunc = (TableFunc *) + pullup_replace_vars((Node *) rte->tablefunc, + context); + break; + case RTE_VALUES: + rte->values_lists = (List *) + pullup_replace_vars((Node *) rte->values_lists, + context); + break; + case RTE_JOIN: + case RTE_CTE: + case RTE_NAMEDTUPLESTORE: + case RTE_RESULT: + /* these shouldn't be marked LATERAL */ + Assert(false); + break; + } + } + } + } + else if (IsA(jtnode, FromExpr)) + { + FromExpr *f = (FromExpr *) jtnode; + ListCell *l; + + foreach(l, f->fromlist) + replace_vars_in_jointree(lfirst(l), context, + lowest_nulling_outer_join); + f->quals = pullup_replace_vars(f->quals, context); + } + else if (IsA(jtnode, JoinExpr)) + { + JoinExpr *j = (JoinExpr *) jtnode; + bool save_need_phvs = context->need_phvs; + + if (j == lowest_nulling_outer_join) + { + /* no more PHVs in or below this join */ + context->need_phvs = false; + lowest_nulling_outer_join = NULL; + } + replace_vars_in_jointree(j->larg, context, lowest_nulling_outer_join); + replace_vars_in_jointree(j->rarg, context, lowest_nulling_outer_join); + + /* + * Use PHVs within the join quals of a full join, even when it's the + * lowest nulling outer join. Otherwise, we cannot identify which + * side of the join a pulled-up var-free expression came from, which + * can lead to failure to make a plan at all because none of the quals + * appear to be mergeable or hashable conditions. For this purpose we + * don't care about the state of wrap_non_vars, so leave it alone. + */ + if (j->jointype == JOIN_FULL) + context->need_phvs = true; + + j->quals = pullup_replace_vars(j->quals, context); + + /* + * We don't bother to update the colvars list, since it won't be used + * again ... + */ + context->need_phvs = save_need_phvs; + } + else + elog(ERROR, "unrecognized node type: %d", + (int) nodeTag(jtnode)); +} + +/* + * Apply pullup variable replacement throughout an expression tree + * + * Returns a modified copy of the tree, so this can't be used where we + * need to do in-place replacement. + */ +static Node * +pullup_replace_vars(Node *expr, pullup_replace_vars_context *context) +{ + return replace_rte_variables(expr, + context->varno, 0, + pullup_replace_vars_callback, + (void *) context, + context->outer_hasSubLinks); +} + +static Node * +pullup_replace_vars_callback(Var *var, + replace_rte_variables_context *context) +{ + pullup_replace_vars_context *rcon = (pullup_replace_vars_context *) context->callback_arg; + int varattno = var->varattno; + Node *newnode; + + /* + * If PlaceHolderVars are needed, we cache the modified expressions in + * rcon->rv_cache[]. This is not in hopes of any material speed gain + * within this function, but to avoid generating identical PHVs with + * different IDs. That would result in duplicate evaluations at runtime, + * and possibly prevent optimizations that rely on recognizing different + * references to the same subquery output as being equal(). So it's worth + * a bit of extra effort to avoid it. + */ + if (rcon->need_phvs && + varattno >= InvalidAttrNumber && + varattno <= list_length(rcon->targetlist) && + rcon->rv_cache[varattno] != NULL) + { + /* Just copy the entry and fall through to adjust its varlevelsup */ + newnode = copyObject(rcon->rv_cache[varattno]); + } + else if (varattno == InvalidAttrNumber) + { + /* Must expand whole-tuple reference into RowExpr */ + RowExpr *rowexpr; + List *colnames; + List *fields; + bool save_need_phvs = rcon->need_phvs; + int save_sublevelsup = context->sublevels_up; + + /* + * If generating an expansion for a var of a named rowtype (ie, this + * is a plain relation RTE), then we must include dummy items for + * dropped columns. If the var is RECORD (ie, this is a JOIN), then + * omit dropped columns. In the latter case, attach column names to + * the RowExpr for use of the executor and ruleutils.c. + * + * In order to be able to cache the results, we always generate the + * expansion with varlevelsup = 0, and then adjust if needed. + */ + expandRTE(rcon->target_rte, + var->varno, 0 /* not varlevelsup */ , var->location, + (var->vartype != RECORDOID), + &colnames, &fields); + /* Adjust the generated per-field Vars, but don't insert PHVs */ + rcon->need_phvs = false; + context->sublevels_up = 0; /* to match the expandRTE output */ + fields = (List *) replace_rte_variables_mutator((Node *) fields, + context); + rcon->need_phvs = save_need_phvs; + context->sublevels_up = save_sublevelsup; + + rowexpr = makeNode(RowExpr); + rowexpr->args = fields; + rowexpr->row_typeid = var->vartype; + rowexpr->row_format = COERCE_IMPLICIT_CAST; + rowexpr->colnames = (var->vartype == RECORDOID) ? colnames : NIL; + rowexpr->location = var->location; + newnode = (Node *) rowexpr; + + /* + * Insert PlaceHolderVar if needed. Notice that we are wrapping one + * PlaceHolderVar around the whole RowExpr, rather than putting one + * around each element of the row. This is because we need the + * expression to yield NULL, not ROW(NULL,NULL,...) when it is forced + * to null by an outer join. + */ + if (rcon->need_phvs) + { + /* RowExpr is certainly not strict, so always need PHV */ + newnode = (Node *) + make_placeholder_expr(rcon->root, + (Expr *) newnode, + bms_make_singleton(rcon->varno)); + /* cache it with the PHV, and with varlevelsup still zero */ + rcon->rv_cache[InvalidAttrNumber] = copyObject(newnode); + } + } + else + { + /* Normal case referencing one targetlist element */ + TargetEntry *tle = get_tle_by_resno(rcon->targetlist, varattno); + + if (tle == NULL) /* shouldn't happen */ + elog(ERROR, "could not find attribute %d in subquery targetlist", + varattno); + + /* Make a copy of the tlist item to return */ + newnode = (Node *) copyObject(tle->expr); + + /* Insert PlaceHolderVar if needed */ + if (rcon->need_phvs) + { + bool wrap; + + if (newnode && IsA(newnode, Var) && + ((Var *) newnode)->varlevelsup == 0) + { + /* + * Simple Vars always escape being wrapped, unless they are + * lateral references to something outside the subquery being + * pulled up. (Even then, we could omit the PlaceHolderVar if + * the referenced rel is under the same lowest outer join, but + * it doesn't seem worth the trouble to check that.) + */ + if (rcon->target_rte->lateral && + !bms_is_member(((Var *) newnode)->varno, rcon->relids)) + wrap = true; + else + wrap = false; + } + else if (newnode && IsA(newnode, PlaceHolderVar) && + ((PlaceHolderVar *) newnode)->phlevelsup == 0) + { + /* No need to wrap a PlaceHolderVar with another one, either */ + wrap = false; + } + else if (rcon->wrap_non_vars) + { + /* Wrap all non-Vars in a PlaceHolderVar */ + wrap = true; + } + else + { + /* + * If it contains a Var of the subquery being pulled up, and + * does not contain any non-strict constructs, then it's + * certainly nullable so we don't need to insert a + * PlaceHolderVar. + * + * This analysis could be tighter: in particular, a non-strict + * construct hidden within a lower-level PlaceHolderVar is not + * reason to add another PHV. But for now it doesn't seem + * worth the code to be more exact. + * + * Note: in future maybe we should insert a PlaceHolderVar + * anyway, if the tlist item is expensive to evaluate? + * + * For a LATERAL subquery, we have to check the actual var + * membership of the node, but if it's non-lateral then any + * level-zero var must belong to the subquery. + */ + if ((rcon->target_rte->lateral ? + bms_overlap(pull_varnos(rcon->root, (Node *) newnode), + rcon->relids) : + contain_vars_of_level((Node *) newnode, 0)) && + !contain_nonstrict_functions((Node *) newnode)) + { + /* No wrap needed */ + wrap = false; + } + else + { + /* Else wrap it in a PlaceHolderVar */ + wrap = true; + } + } + + if (wrap) + newnode = (Node *) + make_placeholder_expr(rcon->root, + (Expr *) newnode, + bms_make_singleton(rcon->varno)); + + /* + * Cache it if possible (ie, if the attno is in range, which it + * probably always should be). We can cache the value even if we + * decided we didn't need a PHV, since this result will be + * suitable for any request that has need_phvs. + */ + if (varattno > InvalidAttrNumber && + varattno <= list_length(rcon->targetlist)) + rcon->rv_cache[varattno] = copyObject(newnode); + } + } + + /* Must adjust varlevelsup if tlist item is from higher query */ + if (var->varlevelsup > 0) + IncrementVarSublevelsUp(newnode, var->varlevelsup, 0); + + return newnode; +} + +/* + * Apply pullup variable replacement to a subquery + * + * This needs to be different from pullup_replace_vars() because + * replace_rte_variables will think that it shouldn't increment sublevels_up + * before entering the Query; so we need to call it with sublevels_up == 1. + */ +static Query * +pullup_replace_vars_subquery(Query *query, + pullup_replace_vars_context *context) +{ + Assert(IsA(query, Query)); + return (Query *) replace_rte_variables((Node *) query, + context->varno, 1, + pullup_replace_vars_callback, + (void *) context, + NULL); +} + + +/* + * flatten_simple_union_all + * Try to optimize top-level UNION ALL structure into an appendrel + * + * If a query's setOperations tree consists entirely of simple UNION ALL + * operations, flatten it into an append relation, which we can process more + * intelligently than the general setops case. Otherwise, do nothing. + * + * In most cases, this can succeed only for a top-level query, because for a + * subquery in FROM, the parent query's invocation of pull_up_subqueries would + * already have flattened the UNION via pull_up_simple_union_all. But there + * are a few cases we can support here but not in that code path, for example + * when the subquery also contains ORDER BY. + */ +void +flatten_simple_union_all(PlannerInfo *root) +{ + Query *parse = root->parse; + SetOperationStmt *topop; + Node *leftmostjtnode; + int leftmostRTI; + RangeTblEntry *leftmostRTE; + int childRTI; + RangeTblEntry *childRTE; + RangeTblRef *rtr; + + /* Shouldn't be called unless query has setops */ + topop = castNode(SetOperationStmt, parse->setOperations); + Assert(topop); + + /* Can't optimize away a recursive UNION */ + if (root->hasRecursion) + return; + + /* + * Recursively check the tree of set operations. If not all UNION ALL + * with identical column types, punt. + */ + if (!is_simple_union_all_recurse((Node *) topop, parse, topop->colTypes)) + return; + + /* + * Locate the leftmost leaf query in the setops tree. The upper query's + * Vars all refer to this RTE (see transformSetOperationStmt). + */ + leftmostjtnode = topop->larg; + while (leftmostjtnode && IsA(leftmostjtnode, SetOperationStmt)) + leftmostjtnode = ((SetOperationStmt *) leftmostjtnode)->larg; + Assert(leftmostjtnode && IsA(leftmostjtnode, RangeTblRef)); + leftmostRTI = ((RangeTblRef *) leftmostjtnode)->rtindex; + leftmostRTE = rt_fetch(leftmostRTI, parse->rtable); + Assert(leftmostRTE->rtekind == RTE_SUBQUERY); + + /* + * Make a copy of the leftmost RTE and add it to the rtable. This copy + * will represent the leftmost leaf query in its capacity as a member of + * the appendrel. The original will represent the appendrel as a whole. + * (We must do things this way because the upper query's Vars have to be + * seen as referring to the whole appendrel.) + */ + childRTE = copyObject(leftmostRTE); + parse->rtable = lappend(parse->rtable, childRTE); + childRTI = list_length(parse->rtable); + + /* Modify the setops tree to reference the child copy */ + ((RangeTblRef *) leftmostjtnode)->rtindex = childRTI; + + /* Modify the formerly-leftmost RTE to mark it as an appendrel parent */ + leftmostRTE->inh = true; + + /* + * Form a RangeTblRef for the appendrel, and insert it into FROM. The top + * Query of a setops tree should have had an empty FromClause initially. + */ + rtr = makeNode(RangeTblRef); + rtr->rtindex = leftmostRTI; + Assert(parse->jointree->fromlist == NIL); + parse->jointree->fromlist = list_make1(rtr); + + /* + * Now pretend the query has no setops. We must do this before trying to + * do subquery pullup, because of Assert in pull_up_simple_subquery. + */ + parse->setOperations = NULL; + + /* + * Build AppendRelInfo information, and apply pull_up_subqueries to the + * leaf queries of the UNION ALL. (We must do that now because they + * weren't previously referenced by the jointree, and so were missed by + * the main invocation of pull_up_subqueries.) + */ + pull_up_union_leaf_queries((Node *) topop, root, leftmostRTI, parse, 0); +} + + +/* + * reduce_outer_joins + * Attempt to reduce outer joins to plain inner joins. + * + * The idea here is that given a query like + * SELECT ... FROM a LEFT JOIN b ON (...) WHERE b.y = 42; + * we can reduce the LEFT JOIN to a plain JOIN if the "=" operator in WHERE + * is strict. The strict operator will always return NULL, causing the outer + * WHERE to fail, on any row where the LEFT JOIN filled in NULLs for b's + * columns. Therefore, there's no need for the join to produce null-extended + * rows in the first place --- which makes it a plain join not an outer join. + * (This scenario may not be very likely in a query written out by hand, but + * it's reasonably likely when pushing quals down into complex views.) + * + * More generally, an outer join can be reduced in strength if there is a + * strict qual above it in the qual tree that constrains a Var from the + * nullable side of the join to be non-null. (For FULL joins this applies + * to each side separately.) + * + * Another transformation we apply here is to recognize cases like + * SELECT ... FROM a LEFT JOIN b ON (a.x = b.y) WHERE b.y IS NULL; + * If the join clause is strict for b.y, then only null-extended rows could + * pass the upper WHERE, and we can conclude that what the query is really + * specifying is an anti-semijoin. We change the join type from JOIN_LEFT + * to JOIN_ANTI. The IS NULL clause then becomes redundant, and must be + * removed to prevent bogus selectivity calculations, but we leave it to + * distribute_qual_to_rels to get rid of such clauses. + * + * Also, we get rid of JOIN_RIGHT cases by flipping them around to become + * JOIN_LEFT. This saves some code here and in some later planner routines, + * but the main reason to do it is to not need to invent a JOIN_REVERSE_ANTI + * join type. + * + * To ease recognition of strict qual clauses, we require this routine to be + * run after expression preprocessing (i.e., qual canonicalization and JOIN + * alias-var expansion). + */ +void +reduce_outer_joins(PlannerInfo *root) +{ + reduce_outer_joins_state *state; + + /* + * To avoid doing strictness checks on more quals than necessary, we want + * to stop descending the jointree as soon as there are no outer joins + * below our current point. This consideration forces a two-pass process. + * The first pass gathers information about which base rels appear below + * each side of each join clause, and about whether there are outer + * join(s) below each side of each join clause. The second pass examines + * qual clauses and changes join types as it descends the tree. + */ + state = reduce_outer_joins_pass1((Node *) root->parse->jointree); + + /* planner.c shouldn't have called me if no outer joins */ + if (state == NULL || !state->contains_outer) + elog(ERROR, "so where are the outer joins?"); + + reduce_outer_joins_pass2((Node *) root->parse->jointree, + state, root, NULL, NIL, NIL); +} + +/* + * reduce_outer_joins_pass1 - phase 1 data collection + * + * Returns a state node describing the given jointree node. + */ +static reduce_outer_joins_state * +reduce_outer_joins_pass1(Node *jtnode) +{ + reduce_outer_joins_state *result; + + result = (reduce_outer_joins_state *) + palloc(sizeof(reduce_outer_joins_state)); + result->relids = NULL; + result->contains_outer = false; + result->sub_states = NIL; + + if (jtnode == NULL) + return result; + if (IsA(jtnode, RangeTblRef)) + { + int varno = ((RangeTblRef *) jtnode)->rtindex; + + result->relids = bms_make_singleton(varno); + } + else if (IsA(jtnode, FromExpr)) + { + FromExpr *f = (FromExpr *) jtnode; + ListCell *l; + + foreach(l, f->fromlist) + { + reduce_outer_joins_state *sub_state; + + sub_state = reduce_outer_joins_pass1(lfirst(l)); + result->relids = bms_add_members(result->relids, + sub_state->relids); + result->contains_outer |= sub_state->contains_outer; + result->sub_states = lappend(result->sub_states, sub_state); + } + } + else if (IsA(jtnode, JoinExpr)) + { + JoinExpr *j = (JoinExpr *) jtnode; + reduce_outer_joins_state *sub_state; + + /* join's own RT index is not wanted in result->relids */ + if (IS_OUTER_JOIN(j->jointype)) + result->contains_outer = true; + + sub_state = reduce_outer_joins_pass1(j->larg); + result->relids = bms_add_members(result->relids, + sub_state->relids); + result->contains_outer |= sub_state->contains_outer; + result->sub_states = lappend(result->sub_states, sub_state); + + sub_state = reduce_outer_joins_pass1(j->rarg); + result->relids = bms_add_members(result->relids, + sub_state->relids); + result->contains_outer |= sub_state->contains_outer; + result->sub_states = lappend(result->sub_states, sub_state); + } + else + elog(ERROR, "unrecognized node type: %d", + (int) nodeTag(jtnode)); + return result; +} + +/* + * reduce_outer_joins_pass2 - phase 2 processing + * + * jtnode: current jointree node + * state: state data collected by phase 1 for this node + * root: toplevel planner state + * nonnullable_rels: set of base relids forced non-null by upper quals + * nonnullable_vars: list of Vars forced non-null by upper quals + * forced_null_vars: list of Vars forced null by upper quals + */ +static void +reduce_outer_joins_pass2(Node *jtnode, + reduce_outer_joins_state *state, + PlannerInfo *root, + Relids nonnullable_rels, + List *nonnullable_vars, + List *forced_null_vars) +{ + /* + * pass 2 should never descend as far as an empty subnode or base rel, + * because it's only called on subtrees marked as contains_outer. + */ + if (jtnode == NULL) + elog(ERROR, "reached empty jointree"); + if (IsA(jtnode, RangeTblRef)) + elog(ERROR, "reached base rel"); + else if (IsA(jtnode, FromExpr)) + { + FromExpr *f = (FromExpr *) jtnode; + ListCell *l; + ListCell *s; + Relids pass_nonnullable_rels; + List *pass_nonnullable_vars; + List *pass_forced_null_vars; + + /* Scan quals to see if we can add any constraints */ + pass_nonnullable_rels = find_nonnullable_rels(f->quals); + pass_nonnullable_rels = bms_add_members(pass_nonnullable_rels, + nonnullable_rels); + pass_nonnullable_vars = find_nonnullable_vars(f->quals); + pass_nonnullable_vars = list_concat(pass_nonnullable_vars, + nonnullable_vars); + pass_forced_null_vars = find_forced_null_vars(f->quals); + pass_forced_null_vars = list_concat(pass_forced_null_vars, + forced_null_vars); + /* And recurse --- but only into interesting subtrees */ + Assert(list_length(f->fromlist) == list_length(state->sub_states)); + forboth(l, f->fromlist, s, state->sub_states) + { + reduce_outer_joins_state *sub_state = lfirst(s); + + if (sub_state->contains_outer) + reduce_outer_joins_pass2(lfirst(l), sub_state, root, + pass_nonnullable_rels, + pass_nonnullable_vars, + pass_forced_null_vars); + } + bms_free(pass_nonnullable_rels); + /* can't so easily clean up var lists, unfortunately */ + } + else if (IsA(jtnode, JoinExpr)) + { + JoinExpr *j = (JoinExpr *) jtnode; + int rtindex = j->rtindex; + JoinType jointype = j->jointype; + reduce_outer_joins_state *left_state = linitial(state->sub_states); + reduce_outer_joins_state *right_state = lsecond(state->sub_states); + List *local_nonnullable_vars = NIL; + bool computed_local_nonnullable_vars = false; + + /* Can we simplify this join? */ + switch (jointype) + { + case JOIN_INNER: + break; + case JOIN_LEFT: + if (bms_overlap(nonnullable_rels, right_state->relids)) + jointype = JOIN_INNER; + break; + case JOIN_RIGHT: + if (bms_overlap(nonnullable_rels, left_state->relids)) + jointype = JOIN_INNER; + break; + case JOIN_FULL: + if (bms_overlap(nonnullable_rels, left_state->relids)) + { + if (bms_overlap(nonnullable_rels, right_state->relids)) + jointype = JOIN_INNER; + else + jointype = JOIN_LEFT; + } + else + { + if (bms_overlap(nonnullable_rels, right_state->relids)) + jointype = JOIN_RIGHT; + } + break; + case JOIN_SEMI: + case JOIN_ANTI: + + /* + * These could only have been introduced by pull_up_sublinks, + * so there's no way that upper quals could refer to their + * righthand sides, and no point in checking. + */ + break; + default: + elog(ERROR, "unrecognized join type: %d", + (int) jointype); + break; + } + + /* + * Convert JOIN_RIGHT to JOIN_LEFT. Note that in the case where we + * reduced JOIN_FULL to JOIN_RIGHT, this will mean the JoinExpr no + * longer matches the internal ordering of any CoalesceExpr's built to + * represent merged join variables. We don't care about that at + * present, but be wary of it ... + */ + if (jointype == JOIN_RIGHT) + { + Node *tmparg; + + tmparg = j->larg; + j->larg = j->rarg; + j->rarg = tmparg; + jointype = JOIN_LEFT; + right_state = linitial(state->sub_states); + left_state = lsecond(state->sub_states); + } + + /* + * See if we can reduce JOIN_LEFT to JOIN_ANTI. This is the case if + * the join's own quals are strict for any var that was forced null by + * higher qual levels. NOTE: there are other ways that we could + * detect an anti-join, in particular if we were to check whether Vars + * coming from the RHS must be non-null because of table constraints. + * That seems complicated and expensive though (in particular, one + * would have to be wary of lower outer joins). For the moment this + * seems sufficient. + */ + if (jointype == JOIN_LEFT) + { + List *overlap; + + local_nonnullable_vars = find_nonnullable_vars(j->quals); + computed_local_nonnullable_vars = true; + + /* + * It's not sufficient to check whether local_nonnullable_vars and + * forced_null_vars overlap: we need to know if the overlap + * includes any RHS variables. + */ + overlap = list_intersection(local_nonnullable_vars, + forced_null_vars); + if (overlap != NIL && + bms_overlap(pull_varnos(root, (Node *) overlap), + right_state->relids)) + jointype = JOIN_ANTI; + } + + /* Apply the jointype change, if any, to both jointree node and RTE */ + if (rtindex && jointype != j->jointype) + { + RangeTblEntry *rte = rt_fetch(rtindex, root->parse->rtable); + + Assert(rte->rtekind == RTE_JOIN); + Assert(rte->jointype == j->jointype); + rte->jointype = jointype; + } + j->jointype = jointype; + + /* Only recurse if there's more to do below here */ + if (left_state->contains_outer || right_state->contains_outer) + { + Relids local_nonnullable_rels; + List *local_forced_null_vars; + Relids pass_nonnullable_rels; + List *pass_nonnullable_vars; + List *pass_forced_null_vars; + + /* + * If this join is (now) inner, we can add any constraints its + * quals provide to those we got from above. But if it is outer, + * we can pass down the local constraints only into the nullable + * side, because an outer join never eliminates any rows from its + * non-nullable side. Also, there is no point in passing upper + * constraints into the nullable side, since if there were any + * we'd have been able to reduce the join. (In the case of upper + * forced-null constraints, we *must not* pass them into the + * nullable side --- they either applied here, or not.) The upshot + * is that we pass either the local or the upper constraints, + * never both, to the children of an outer join. + * + * Note that a SEMI join works like an inner join here: it's okay + * to pass down both local and upper constraints. (There can't be + * any upper constraints affecting its inner side, but it's not + * worth having a separate code path to avoid passing them.) + * + * At a FULL join we just punt and pass nothing down --- is it + * possible to be smarter? + */ + if (jointype != JOIN_FULL) + { + local_nonnullable_rels = find_nonnullable_rels(j->quals); + if (!computed_local_nonnullable_vars) + local_nonnullable_vars = find_nonnullable_vars(j->quals); + local_forced_null_vars = find_forced_null_vars(j->quals); + if (jointype == JOIN_INNER || jointype == JOIN_SEMI) + { + /* OK to merge upper and local constraints */ + local_nonnullable_rels = bms_add_members(local_nonnullable_rels, + nonnullable_rels); + local_nonnullable_vars = list_concat(local_nonnullable_vars, + nonnullable_vars); + local_forced_null_vars = list_concat(local_forced_null_vars, + forced_null_vars); + } + } + else + { + /* no use in calculating these */ + local_nonnullable_rels = NULL; + local_forced_null_vars = NIL; + } + + if (left_state->contains_outer) + { + if (jointype == JOIN_INNER || jointype == JOIN_SEMI) + { + /* pass union of local and upper constraints */ + pass_nonnullable_rels = local_nonnullable_rels; + pass_nonnullable_vars = local_nonnullable_vars; + pass_forced_null_vars = local_forced_null_vars; + } + else if (jointype != JOIN_FULL) /* ie, LEFT or ANTI */ + { + /* can't pass local constraints to non-nullable side */ + pass_nonnullable_rels = nonnullable_rels; + pass_nonnullable_vars = nonnullable_vars; + pass_forced_null_vars = forced_null_vars; + } + else + { + /* no constraints pass through JOIN_FULL */ + pass_nonnullable_rels = NULL; + pass_nonnullable_vars = NIL; + pass_forced_null_vars = NIL; + } + reduce_outer_joins_pass2(j->larg, left_state, root, + pass_nonnullable_rels, + pass_nonnullable_vars, + pass_forced_null_vars); + } + + if (right_state->contains_outer) + { + if (jointype != JOIN_FULL) /* ie, INNER/LEFT/SEMI/ANTI */ + { + /* pass appropriate constraints, per comment above */ + pass_nonnullable_rels = local_nonnullable_rels; + pass_nonnullable_vars = local_nonnullable_vars; + pass_forced_null_vars = local_forced_null_vars; + } + else + { + /* no constraints pass through JOIN_FULL */ + pass_nonnullable_rels = NULL; + pass_nonnullable_vars = NIL; + pass_forced_null_vars = NIL; + } + reduce_outer_joins_pass2(j->rarg, right_state, root, + pass_nonnullable_rels, + pass_nonnullable_vars, + pass_forced_null_vars); + } + bms_free(local_nonnullable_rels); + } + } + else + elog(ERROR, "unrecognized node type: %d", + (int) nodeTag(jtnode)); +} + + +/* + * remove_useless_result_rtes + * Attempt to remove RTE_RESULT RTEs from the join tree. + * + * We can remove RTE_RESULT entries from the join tree using the knowledge + * that RTE_RESULT returns exactly one row and has no output columns. Hence, + * if one is inner-joined to anything else, we can delete it. Optimizations + * are also possible for some outer-join cases, as detailed below. + * + * Some of these optimizations depend on recognizing empty (constant-true) + * quals for FromExprs and JoinExprs. That makes it useful to apply this + * optimization pass after expression preprocessing, since that will have + * eliminated constant-true quals, allowing more cases to be recognized as + * optimizable. What's more, the usual reason for an RTE_RESULT to be present + * is that we pulled up a subquery or VALUES clause, thus very possibly + * replacing Vars with constants, making it more likely that a qual can be + * reduced to constant true. Also, because some optimizations depend on + * the outer-join type, it's best to have done reduce_outer_joins() first. + * + * A PlaceHolderVar referencing an RTE_RESULT RTE poses an obstacle to this + * process: we must remove the RTE_RESULT's relid from the PHV's phrels, but + * we must not reduce the phrels set to empty. If that would happen, and + * the RTE_RESULT is an immediate child of an outer join, we have to give up + * and not remove the RTE_RESULT: there is noplace else to evaluate the + * PlaceHolderVar. (That is, in such cases the RTE_RESULT *does* have output + * columns.) But if the RTE_RESULT is an immediate child of an inner join, + * we can usually change the PlaceHolderVar's phrels so as to evaluate it at + * the inner join instead. This is OK because we really only care that PHVs + * are evaluated above or below the correct outer joins. We can't, however, + * postpone the evaluation of a PHV to above where it is used; so there are + * some checks below on whether output PHVs are laterally referenced in the + * other join input rel(s). + * + * We used to try to do this work as part of pull_up_subqueries() where the + * potentially-optimizable cases get introduced; but it's way simpler, and + * more effective, to do it separately. + */ +void +remove_useless_result_rtes(PlannerInfo *root) +{ + ListCell *cell; + + /* Top level of jointree must always be a FromExpr */ + Assert(IsA(root->parse->jointree, FromExpr)); + /* Recurse ... */ + root->parse->jointree = (FromExpr *) + remove_useless_results_recurse(root, (Node *) root->parse->jointree); + /* We should still have a FromExpr */ + Assert(IsA(root->parse->jointree, FromExpr)); + + /* + * Remove any PlanRowMark referencing an RTE_RESULT RTE. We obviously + * must do that for any RTE_RESULT that we just removed. But one for a + * RTE that we did not remove can be dropped anyway: since the RTE has + * only one possible output row, there is no need for EPQ to mark and + * restore that row. + * + * It's necessary, not optional, to remove the PlanRowMark for a surviving + * RTE_RESULT RTE; otherwise we'll generate a whole-row Var for the + * RTE_RESULT, which the executor has no support for. + */ + foreach(cell, root->rowMarks) + { + PlanRowMark *rc = (PlanRowMark *) lfirst(cell); + + if (rt_fetch(rc->rti, root->parse->rtable)->rtekind == RTE_RESULT) + root->rowMarks = foreach_delete_current(root->rowMarks, cell); + } +} + +/* + * remove_useless_results_recurse + * Recursive guts of remove_useless_result_rtes. + * + * This recursively processes the jointree and returns a modified jointree. + */ +static Node * +remove_useless_results_recurse(PlannerInfo *root, Node *jtnode) +{ + Assert(jtnode != NULL); + if (IsA(jtnode, RangeTblRef)) + { + /* Can't immediately do anything with a RangeTblRef */ + } + else if (IsA(jtnode, FromExpr)) + { + FromExpr *f = (FromExpr *) jtnode; + Relids result_relids = NULL; + ListCell *cell; + + /* + * We can drop RTE_RESULT rels from the fromlist so long as at least + * one child remains, since joining to a one-row table changes + * nothing. (But we can't drop a RTE_RESULT that computes PHV(s) that + * are needed by some sibling. The cleanup transformation below would + * reassign the PHVs to be computed at the join, which is too late for + * the sibling's use.) The easiest way to mechanize this rule is to + * modify the list in-place. + */ + foreach(cell, f->fromlist) + { + Node *child = (Node *) lfirst(cell); + int varno; + + /* Recursively transform child ... */ + child = remove_useless_results_recurse(root, child); + /* ... and stick it back into the tree */ + lfirst(cell) = child; + + /* + * If it's an RTE_RESULT with at least one sibling, and no sibling + * references dependent PHVs, we can drop it. We don't yet know + * what the inner join's final relid set will be, so postpone + * cleanup of PHVs etc till after this loop. + */ + if (list_length(f->fromlist) > 1 && + (varno = get_result_relid(root, child)) != 0 && + !find_dependent_phvs_in_jointree(root, (Node *) f, varno)) + { + f->fromlist = foreach_delete_current(f->fromlist, cell); + result_relids = bms_add_member(result_relids, varno); + } + } + + /* + * Clean up if we dropped any RTE_RESULT RTEs. This is a bit + * inefficient if there's more than one, but it seems better to + * optimize the support code for the single-relid case. + */ + if (result_relids) + { + int varno = -1; + + while ((varno = bms_next_member(result_relids, varno)) >= 0) + remove_result_refs(root, varno, (Node *) f); + } + + /* + * If we're not at the top of the jointree, it's valid to simplify a + * degenerate FromExpr into its single child. (At the top, we must + * keep the FromExpr since Query.jointree is required to point to a + * FromExpr.) + */ + if (f != root->parse->jointree && + f->quals == NULL && + list_length(f->fromlist) == 1) + return (Node *) linitial(f->fromlist); + } + else if (IsA(jtnode, JoinExpr)) + { + JoinExpr *j = (JoinExpr *) jtnode; + int varno; + + /* First, recurse */ + j->larg = remove_useless_results_recurse(root, j->larg); + j->rarg = remove_useless_results_recurse(root, j->rarg); + + /* Apply join-type-specific optimization rules */ + switch (j->jointype) + { + case JOIN_INNER: + + /* + * An inner join is equivalent to a FromExpr, so if either + * side was simplified to an RTE_RESULT rel, we can replace + * the join with a FromExpr with just the other side; and if + * the qual is empty (JOIN ON TRUE) then we can omit the + * FromExpr as well. + * + * Just as in the FromExpr case, we can't simplify if the + * other input rel references any PHVs that are marked as to + * be evaluated at the RTE_RESULT rel, because we can't + * postpone their evaluation in that case. But we only have + * to check this in cases where it's syntactically legal for + * the other input to have a LATERAL reference to the + * RTE_RESULT rel. Only RHSes of inner and left joins are + * allowed to have such refs. + */ + if ((varno = get_result_relid(root, j->larg)) != 0 && + !find_dependent_phvs_in_jointree(root, j->rarg, varno)) + { + remove_result_refs(root, varno, j->rarg); + if (j->quals) + jtnode = (Node *) + makeFromExpr(list_make1(j->rarg), j->quals); + else + jtnode = j->rarg; + } + else if ((varno = get_result_relid(root, j->rarg)) != 0) + { + remove_result_refs(root, varno, j->larg); + if (j->quals) + jtnode = (Node *) + makeFromExpr(list_make1(j->larg), j->quals); + else + jtnode = j->larg; + } + break; + case JOIN_LEFT: + + /* + * We can simplify this case if the RHS is an RTE_RESULT, with + * two different possibilities: + * + * If the qual is empty (JOIN ON TRUE), then the join can be + * strength-reduced to a plain inner join, since each LHS row + * necessarily has exactly one join partner. So we can always + * discard the RHS, much as in the JOIN_INNER case above. + * (Again, the LHS could not contain a lateral reference to + * the RHS.) + * + * Otherwise, it's still true that each LHS row should be + * returned exactly once, and since the RHS returns no columns + * (unless there are PHVs that have to be evaluated there), we + * don't much care if it's null-extended or not. So in this + * case also, we can just ignore the qual and discard the left + * join. + */ + if ((varno = get_result_relid(root, j->rarg)) != 0 && + (j->quals == NULL || + !find_dependent_phvs(root, varno))) + { + remove_result_refs(root, varno, j->larg); + jtnode = j->larg; + } + break; + case JOIN_SEMI: + + /* + * We may simplify this case if the RHS is an RTE_RESULT; the + * join qual becomes effectively just a filter qual for the + * LHS, since we should either return the LHS row or not. For + * simplicity we inject the filter qual into a new FromExpr. + * + * There is a fine point about PHVs that are supposed to be + * evaluated at the RHS. Such PHVs could only appear in the + * semijoin's qual, since the rest of the query cannot + * reference any outputs of the semijoin's RHS. Therefore, + * they can't actually go to null before being examined, and + * it'd be OK to just remove the PHV wrapping. We don't have + * infrastructure for that, but remove_result_refs() will + * relabel them as to be evaluated at the LHS, which is fine. + */ + if ((varno = get_result_relid(root, j->rarg)) != 0) + { + remove_result_refs(root, varno, j->larg); + if (j->quals) + jtnode = (Node *) + makeFromExpr(list_make1(j->larg), j->quals); + else + jtnode = j->larg; + } + break; + case JOIN_FULL: + case JOIN_ANTI: + /* We have no special smarts for these cases */ + break; + default: + /* Note: JOIN_RIGHT should be gone at this point */ + elog(ERROR, "unrecognized join type: %d", + (int) j->jointype); + break; + } + } + else + elog(ERROR, "unrecognized node type: %d", + (int) nodeTag(jtnode)); + return jtnode; +} + +/* + * get_result_relid + * If jtnode is a RangeTblRef for an RTE_RESULT RTE, return its relid; + * otherwise return 0. + */ +static int +get_result_relid(PlannerInfo *root, Node *jtnode) +{ + int varno; + + if (!IsA(jtnode, RangeTblRef)) + return 0; + varno = ((RangeTblRef *) jtnode)->rtindex; + if (rt_fetch(varno, root->parse->rtable)->rtekind != RTE_RESULT) + return 0; + return varno; +} + +/* + * remove_result_refs + * Helper routine for dropping an unneeded RTE_RESULT RTE. + * + * This doesn't physically remove the RTE from the jointree, because that's + * more easily handled in remove_useless_results_recurse. What it does do + * is the necessary cleanup in the rest of the tree: we must adjust any PHVs + * that may reference the RTE. Be sure to call this at a point where the + * jointree is valid (no disconnected nodes). + * + * Note that we don't need to process the append_rel_list, since RTEs + * referenced directly in the jointree won't be appendrel members. + * + * varno is the RTE_RESULT's relid. + * newjtloc is the jointree location at which any PHVs referencing the + * RTE_RESULT should be evaluated instead. + */ +static void +remove_result_refs(PlannerInfo *root, int varno, Node *newjtloc) +{ + /* Fix up PlaceHolderVars as needed */ + /* If there are no PHVs anywhere, we can skip this bit */ + if (root->glob->lastPHId != 0) + { + Relids subrelids; + + subrelids = get_relids_in_jointree(newjtloc, false); + Assert(!bms_is_empty(subrelids)); + substitute_phv_relids((Node *) root->parse, varno, subrelids); + fix_append_rel_relids(root->append_rel_list, varno, subrelids); + } + + /* + * We also need to remove any PlanRowMark referencing the RTE, but we + * postpone that work until we return to remove_useless_result_rtes. + */ +} + + +/* + * find_dependent_phvs - are there any PlaceHolderVars whose relids are + * exactly the given varno? + * + * find_dependent_phvs should be used when we want to see if there are + * any such PHVs anywhere in the Query. Another use-case is to see if + * a subtree of the join tree contains such PHVs; but for that, we have + * to look not only at the join tree nodes themselves but at the + * referenced RTEs. For that, use find_dependent_phvs_in_jointree. + */ + +typedef struct +{ + Relids relids; + int sublevels_up; +} find_dependent_phvs_context; + +static bool +find_dependent_phvs_walker(Node *node, + find_dependent_phvs_context *context) +{ + if (node == NULL) + return false; + if (IsA(node, PlaceHolderVar)) + { + PlaceHolderVar *phv = (PlaceHolderVar *) node; + + if (phv->phlevelsup == context->sublevels_up && + bms_equal(context->relids, phv->phrels)) + return true; + /* fall through to examine children */ + } + if (IsA(node, Query)) + { + /* Recurse into subselects */ + bool result; + + context->sublevels_up++; + result = query_tree_walker((Query *) node, + find_dependent_phvs_walker, + (void *) context, 0); + context->sublevels_up--; + return result; + } + /* Shouldn't need to handle planner auxiliary nodes here */ + Assert(!IsA(node, SpecialJoinInfo)); + Assert(!IsA(node, AppendRelInfo)); + Assert(!IsA(node, PlaceHolderInfo)); + Assert(!IsA(node, MinMaxAggInfo)); + + return expression_tree_walker(node, find_dependent_phvs_walker, + (void *) context); +} + +static bool +find_dependent_phvs(PlannerInfo *root, int varno) +{ + find_dependent_phvs_context context; + + /* If there are no PHVs anywhere, we needn't work hard */ + if (root->glob->lastPHId == 0) + return false; + + context.relids = bms_make_singleton(varno); + context.sublevels_up = 0; + + return query_tree_walker(root->parse, + find_dependent_phvs_walker, + (void *) &context, + 0); +} + +static bool +find_dependent_phvs_in_jointree(PlannerInfo *root, Node *node, int varno) +{ + find_dependent_phvs_context context; + Relids subrelids; + int relid; + + /* If there are no PHVs anywhere, we needn't work hard */ + if (root->glob->lastPHId == 0) + return false; + + context.relids = bms_make_singleton(varno); + context.sublevels_up = 0; + + /* + * See if the jointree fragment itself contains references (in join quals) + */ + if (find_dependent_phvs_walker(node, &context)) + return true; + + /* + * Otherwise, identify the set of referenced RTEs (we can ignore joins, + * since they should be flattened already, so their join alias lists no + * longer matter), and tediously check each RTE. We can ignore RTEs that + * are not marked LATERAL, though, since they couldn't possibly contain + * any cross-references to other RTEs. + */ + subrelids = get_relids_in_jointree(node, false); + relid = -1; + while ((relid = bms_next_member(subrelids, relid)) >= 0) + { + RangeTblEntry *rte = rt_fetch(relid, root->parse->rtable); + + if (rte->lateral && + range_table_entry_walker(rte, + find_dependent_phvs_walker, + (void *) &context, + 0)) + return true; + } + + return false; +} + +/* + * substitute_phv_relids - adjust PlaceHolderVar relid sets after pulling up + * a subquery or removing an RTE_RESULT jointree item + * + * Find any PlaceHolderVar nodes in the given tree that reference the + * pulled-up relid, and change them to reference the replacement relid(s). + * + * NOTE: although this has the form of a walker, we cheat and modify the + * nodes in-place. This should be OK since the tree was copied by + * pullup_replace_vars earlier. Avoid scribbling on the original values of + * the bitmapsets, though, because expression_tree_mutator doesn't copy those. + */ + +typedef struct +{ + int varno; + int sublevels_up; + Relids subrelids; +} substitute_phv_relids_context; + +static bool +substitute_phv_relids_walker(Node *node, + substitute_phv_relids_context *context) +{ + if (node == NULL) + return false; + if (IsA(node, PlaceHolderVar)) + { + PlaceHolderVar *phv = (PlaceHolderVar *) node; + + if (phv->phlevelsup == context->sublevels_up && + bms_is_member(context->varno, phv->phrels)) + { + phv->phrels = bms_union(phv->phrels, + context->subrelids); + phv->phrels = bms_del_member(phv->phrels, + context->varno); + /* Assert we haven't broken the PHV */ + Assert(!bms_is_empty(phv->phrels)); + } + /* fall through to examine children */ + } + if (IsA(node, Query)) + { + /* Recurse into subselects */ + bool result; + + context->sublevels_up++; + result = query_tree_walker((Query *) node, + substitute_phv_relids_walker, + (void *) context, 0); + context->sublevels_up--; + return result; + } + /* Shouldn't need to handle planner auxiliary nodes here */ + Assert(!IsA(node, SpecialJoinInfo)); + Assert(!IsA(node, AppendRelInfo)); + Assert(!IsA(node, PlaceHolderInfo)); + Assert(!IsA(node, MinMaxAggInfo)); + + return expression_tree_walker(node, substitute_phv_relids_walker, + (void *) context); +} + +static void +substitute_phv_relids(Node *node, int varno, Relids subrelids) +{ + substitute_phv_relids_context context; + + context.varno = varno; + context.sublevels_up = 0; + context.subrelids = subrelids; + + /* + * Must be prepared to start with a Query or a bare expression tree. + */ + query_or_expression_tree_walker(node, + substitute_phv_relids_walker, + (void *) &context, + 0); +} + +/* + * fix_append_rel_relids: update RT-index fields of AppendRelInfo nodes + * + * When we pull up a subquery, any AppendRelInfo references to the subquery's + * RT index have to be replaced by the substituted relid (and there had better + * be only one). We also need to apply substitute_phv_relids to their + * translated_vars lists, since those might contain PlaceHolderVars. + * + * We assume we may modify the AppendRelInfo nodes in-place. + */ +static void +fix_append_rel_relids(List *append_rel_list, int varno, Relids subrelids) +{ + ListCell *l; + int subvarno = -1; + + /* + * We only want to extract the member relid once, but we mustn't fail + * immediately if there are multiple members; it could be that none of the + * AppendRelInfo nodes refer to it. So compute it on first use. Note that + * bms_singleton_member will complain if set is not singleton. + */ + foreach(l, append_rel_list) + { + AppendRelInfo *appinfo = (AppendRelInfo *) lfirst(l); + + /* The parent_relid shouldn't ever be a pullup target */ + Assert(appinfo->parent_relid != varno); + + if (appinfo->child_relid == varno) + { + if (subvarno < 0) + subvarno = bms_singleton_member(subrelids); + appinfo->child_relid = subvarno; + } + + /* Also fix up any PHVs in its translated vars */ + substitute_phv_relids((Node *) appinfo->translated_vars, + varno, subrelids); + } +} + +/* + * get_relids_in_jointree: get set of RT indexes present in a jointree + * + * If include_joins is true, join RT indexes are included; if false, + * only base rels are included. + */ +Relids +get_relids_in_jointree(Node *jtnode, bool include_joins) +{ + Relids result = NULL; + + if (jtnode == NULL) + return result; + if (IsA(jtnode, RangeTblRef)) + { + int varno = ((RangeTblRef *) jtnode)->rtindex; + + result = bms_make_singleton(varno); + } + else if (IsA(jtnode, FromExpr)) + { + FromExpr *f = (FromExpr *) jtnode; + ListCell *l; + + foreach(l, f->fromlist) + { + result = bms_join(result, + get_relids_in_jointree(lfirst(l), + include_joins)); + } + } + else if (IsA(jtnode, JoinExpr)) + { + JoinExpr *j = (JoinExpr *) jtnode; + + result = get_relids_in_jointree(j->larg, include_joins); + result = bms_join(result, + get_relids_in_jointree(j->rarg, include_joins)); + if (include_joins && j->rtindex) + result = bms_add_member(result, j->rtindex); + } + else + elog(ERROR, "unrecognized node type: %d", + (int) nodeTag(jtnode)); + return result; +} + +/* + * get_relids_for_join: get set of base RT indexes making up a join + */ +Relids +get_relids_for_join(Query *query, int joinrelid) +{ + Node *jtnode; + + jtnode = find_jointree_node_for_rel((Node *) query->jointree, + joinrelid); + if (!jtnode) + elog(ERROR, "could not find join node %d", joinrelid); + return get_relids_in_jointree(jtnode, false); +} + +/* + * find_jointree_node_for_rel: locate jointree node for a base or join RT index + * + * Returns NULL if not found + */ +static Node * +find_jointree_node_for_rel(Node *jtnode, int relid) +{ + if (jtnode == NULL) + return NULL; + if (IsA(jtnode, RangeTblRef)) + { + int varno = ((RangeTblRef *) jtnode)->rtindex; + + if (relid == varno) + return jtnode; + } + else if (IsA(jtnode, FromExpr)) + { + FromExpr *f = (FromExpr *) jtnode; + ListCell *l; + + foreach(l, f->fromlist) + { + jtnode = find_jointree_node_for_rel(lfirst(l), relid); + if (jtnode) + return jtnode; + } + } + else if (IsA(jtnode, JoinExpr)) + { + JoinExpr *j = (JoinExpr *) jtnode; + + if (relid == j->rtindex) + return jtnode; + jtnode = find_jointree_node_for_rel(j->larg, relid); + if (jtnode) + return jtnode; + jtnode = find_jointree_node_for_rel(j->rarg, relid); + if (jtnode) + return jtnode; + } + else + elog(ERROR, "unrecognized node type: %d", + (int) nodeTag(jtnode)); + return NULL; +} diff --git a/src/backend/optimizer/prep/prepqual.c b/src/backend/optimizer/prep/prepqual.c new file mode 100644 index 0000000..da01234 --- /dev/null +++ b/src/backend/optimizer/prep/prepqual.c @@ -0,0 +1,677 @@ +/*------------------------------------------------------------------------- + * + * prepqual.c + * Routines for preprocessing qualification expressions + * + * + * While the parser will produce flattened (N-argument) AND/OR trees from + * simple sequences of AND'ed or OR'ed clauses, there might be an AND clause + * directly underneath another AND, or OR underneath OR, if the input was + * oddly parenthesized. Also, rule expansion and subquery flattening could + * produce such parsetrees. The planner wants to flatten all such cases + * to ensure consistent optimization behavior. + * + * Formerly, this module was responsible for doing the initial flattening, + * but now we leave it to eval_const_expressions to do that since it has to + * make a complete pass over the expression tree anyway. Instead, we just + * have to ensure that our manipulations preserve AND/OR flatness. + * pull_ands() and pull_ors() are used to maintain flatness of the AND/OR + * tree after local transformations that might introduce nested AND/ORs. + * + * + * Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group + * Portions Copyright (c) 1994, Regents of the University of California + * + * + * IDENTIFICATION + * src/backend/optimizer/prep/prepqual.c + * + *------------------------------------------------------------------------- + */ + +#include "postgres.h" + +#include "nodes/makefuncs.h" +#include "nodes/nodeFuncs.h" +#include "optimizer/optimizer.h" +#include "optimizer/prep.h" +#include "utils/lsyscache.h" + + +static List *pull_ands(List *andlist); +static List *pull_ors(List *orlist); +static Expr *find_duplicate_ors(Expr *qual, bool is_check); +static Expr *process_duplicate_ors(List *orlist); + + +/* + * negate_clause + * Negate a Boolean expression. + * + * Input is a clause to be negated (e.g., the argument of a NOT clause). + * Returns a new clause equivalent to the negation of the given clause. + * + * Although this can be invoked on its own, it's mainly intended as a helper + * for eval_const_expressions(), and that context drives several design + * decisions. In particular, if the input is already AND/OR flat, we must + * preserve that property. We also don't bother to recurse in situations + * where we can assume that lower-level executions of eval_const_expressions + * would already have simplified sub-clauses of the input. + * + * The difference between this and a simple make_notclause() is that this + * tries to get rid of the NOT node by logical simplification. It's clearly + * always a win if the NOT node can be eliminated altogether. However, our + * use of DeMorgan's laws could result in having more NOT nodes rather than + * fewer. We do that unconditionally anyway, because in WHERE clauses it's + * important to expose as much top-level AND/OR structure as possible. + * Also, eliminating an intermediate NOT may allow us to flatten two levels + * of AND or OR together that we couldn't have otherwise. Finally, one of + * the motivations for doing this is to ensure that logically equivalent + * expressions will be seen as physically equal(), so we should always apply + * the same transformations. + */ +Node * +negate_clause(Node *node) +{ + if (node == NULL) /* should not happen */ + elog(ERROR, "can't negate an empty subexpression"); + switch (nodeTag(node)) + { + case T_Const: + { + Const *c = (Const *) node; + + /* NOT NULL is still NULL */ + if (c->constisnull) + return makeBoolConst(false, true); + /* otherwise pretty easy */ + return makeBoolConst(!DatumGetBool(c->constvalue), false); + } + break; + case T_OpExpr: + { + /* + * Negate operator if possible: (NOT (< A B)) => (>= A B) + */ + OpExpr *opexpr = (OpExpr *) node; + Oid negator = get_negator(opexpr->opno); + + if (negator) + { + OpExpr *newopexpr = makeNode(OpExpr); + + newopexpr->opno = negator; + newopexpr->opfuncid = InvalidOid; + newopexpr->opresulttype = opexpr->opresulttype; + newopexpr->opretset = opexpr->opretset; + newopexpr->opcollid = opexpr->opcollid; + newopexpr->inputcollid = opexpr->inputcollid; + newopexpr->args = opexpr->args; + newopexpr->location = opexpr->location; + return (Node *) newopexpr; + } + } + break; + case T_ScalarArrayOpExpr: + { + /* + * Negate a ScalarArrayOpExpr if its operator has a negator; + * for example x = ANY (list) becomes x <> ALL (list) + */ + ScalarArrayOpExpr *saopexpr = (ScalarArrayOpExpr *) node; + Oid negator = get_negator(saopexpr->opno); + + if (negator) + { + ScalarArrayOpExpr *newopexpr = makeNode(ScalarArrayOpExpr); + + newopexpr->opno = negator; + newopexpr->opfuncid = InvalidOid; + newopexpr->hashfuncid = InvalidOid; + newopexpr->negfuncid = InvalidOid; + newopexpr->useOr = !saopexpr->useOr; + newopexpr->inputcollid = saopexpr->inputcollid; + newopexpr->args = saopexpr->args; + newopexpr->location = saopexpr->location; + return (Node *) newopexpr; + } + } + break; + case T_BoolExpr: + { + BoolExpr *expr = (BoolExpr *) node; + + switch (expr->boolop) + { + /*-------------------- + * Apply DeMorgan's Laws: + * (NOT (AND A B)) => (OR (NOT A) (NOT B)) + * (NOT (OR A B)) => (AND (NOT A) (NOT B)) + * i.e., swap AND for OR and negate each subclause. + * + * If the input is already AND/OR flat and has no NOT + * directly above AND or OR, this transformation preserves + * those properties. For example, if no direct child of + * the given AND clause is an AND or a NOT-above-OR, then + * the recursive calls of negate_clause() can't return any + * OR clauses. So we needn't call pull_ors() before + * building a new OR clause. Similarly for the OR case. + *-------------------- + */ + case AND_EXPR: + { + List *nargs = NIL; + ListCell *lc; + + foreach(lc, expr->args) + { + nargs = lappend(nargs, + negate_clause(lfirst(lc))); + } + return (Node *) make_orclause(nargs); + } + break; + case OR_EXPR: + { + List *nargs = NIL; + ListCell *lc; + + foreach(lc, expr->args) + { + nargs = lappend(nargs, + negate_clause(lfirst(lc))); + } + return (Node *) make_andclause(nargs); + } + break; + case NOT_EXPR: + + /* + * NOT underneath NOT: they cancel. We assume the + * input is already simplified, so no need to recurse. + */ + return (Node *) linitial(expr->args); + default: + elog(ERROR, "unrecognized boolop: %d", + (int) expr->boolop); + break; + } + } + break; + case T_NullTest: + { + NullTest *expr = (NullTest *) node; + + /* + * In the rowtype case, the two flavors of NullTest are *not* + * logical inverses, so we can't simplify. But it does work + * for scalar datatypes. + */ + if (!expr->argisrow) + { + NullTest *newexpr = makeNode(NullTest); + + newexpr->arg = expr->arg; + newexpr->nulltesttype = (expr->nulltesttype == IS_NULL ? + IS_NOT_NULL : IS_NULL); + newexpr->argisrow = expr->argisrow; + newexpr->location = expr->location; + return (Node *) newexpr; + } + } + break; + case T_BooleanTest: + { + BooleanTest *expr = (BooleanTest *) node; + BooleanTest *newexpr = makeNode(BooleanTest); + + newexpr->arg = expr->arg; + switch (expr->booltesttype) + { + case IS_TRUE: + newexpr->booltesttype = IS_NOT_TRUE; + break; + case IS_NOT_TRUE: + newexpr->booltesttype = IS_TRUE; + break; + case IS_FALSE: + newexpr->booltesttype = IS_NOT_FALSE; + break; + case IS_NOT_FALSE: + newexpr->booltesttype = IS_FALSE; + break; + case IS_UNKNOWN: + newexpr->booltesttype = IS_NOT_UNKNOWN; + break; + case IS_NOT_UNKNOWN: + newexpr->booltesttype = IS_UNKNOWN; + break; + default: + elog(ERROR, "unrecognized booltesttype: %d", + (int) expr->booltesttype); + break; + } + newexpr->location = expr->location; + return (Node *) newexpr; + } + break; + default: + /* else fall through */ + break; + } + + /* + * Otherwise we don't know how to simplify this, so just tack on an + * explicit NOT node. + */ + return (Node *) make_notclause((Expr *) node); +} + + +/* + * canonicalize_qual + * Convert a qualification expression to the most useful form. + * + * This is primarily intended to be used on top-level WHERE (or JOIN/ON) + * clauses. It can also be used on top-level CHECK constraints, for which + * pass is_check = true. DO NOT call it on any expression that is not known + * to be one or the other, as it might apply inappropriate simplifications. + * + * The name of this routine is a holdover from a time when it would try to + * force the expression into canonical AND-of-ORs or OR-of-ANDs form. + * Eventually, we recognized that that had more theoretical purity than + * actual usefulness, and so now the transformation doesn't involve any + * notion of reaching a canonical form. + * + * NOTE: we assume the input has already been through eval_const_expressions + * and therefore possesses AND/OR flatness. Formerly this function included + * its own flattening logic, but that requires a useless extra pass over the + * tree. + * + * Returns the modified qualification. + */ +Expr * +canonicalize_qual(Expr *qual, bool is_check) +{ + Expr *newqual; + + /* Quick exit for empty qual */ + if (qual == NULL) + return NULL; + + /* This should not be invoked on quals in implicit-AND format */ + Assert(!IsA(qual, List)); + + /* + * Pull up redundant subclauses in OR-of-AND trees. We do this only + * within the top-level AND/OR structure; there's no point in looking + * deeper. Also remove any NULL constants in the top-level structure. + */ + newqual = find_duplicate_ors(qual, is_check); + + return newqual; +} + + +/* + * pull_ands + * Recursively flatten nested AND clauses into a single and-clause list. + * + * Input is the arglist of an AND clause. + * Returns the rebuilt arglist (note original list structure is not touched). + */ +static List * +pull_ands(List *andlist) +{ + List *out_list = NIL; + ListCell *arg; + + foreach(arg, andlist) + { + Node *subexpr = (Node *) lfirst(arg); + + if (is_andclause(subexpr)) + out_list = list_concat(out_list, + pull_ands(((BoolExpr *) subexpr)->args)); + else + out_list = lappend(out_list, subexpr); + } + return out_list; +} + +/* + * pull_ors + * Recursively flatten nested OR clauses into a single or-clause list. + * + * Input is the arglist of an OR clause. + * Returns the rebuilt arglist (note original list structure is not touched). + */ +static List * +pull_ors(List *orlist) +{ + List *out_list = NIL; + ListCell *arg; + + foreach(arg, orlist) + { + Node *subexpr = (Node *) lfirst(arg); + + if (is_orclause(subexpr)) + out_list = list_concat(out_list, + pull_ors(((BoolExpr *) subexpr)->args)); + else + out_list = lappend(out_list, subexpr); + } + return out_list; +} + + +/*-------------------- + * The following code attempts to apply the inverse OR distributive law: + * ((A AND B) OR (A AND C)) => (A AND (B OR C)) + * That is, locate OR clauses in which every subclause contains an + * identical term, and pull out the duplicated terms. + * + * This may seem like a fairly useless activity, but it turns out to be + * applicable to many machine-generated queries, and there are also queries + * in some of the TPC benchmarks that need it. This was in fact almost the + * sole useful side-effect of the old prepqual code that tried to force + * the query into canonical AND-of-ORs form: the canonical equivalent of + * ((A AND B) OR (A AND C)) + * is + * ((A OR A) AND (A OR C) AND (B OR A) AND (B OR C)) + * which the code was able to simplify to + * (A AND (A OR C) AND (B OR A) AND (B OR C)) + * thus successfully extracting the common condition A --- but at the cost + * of cluttering the qual with many redundant clauses. + *-------------------- + */ + +/* + * find_duplicate_ors + * Given a qualification tree with the NOTs pushed down, search for + * OR clauses to which the inverse OR distributive law might apply. + * Only the top-level AND/OR structure is searched. + * + * While at it, we remove any NULL constants within the top-level AND/OR + * structure, eg in a WHERE clause, "x OR NULL::boolean" is reduced to "x". + * In general that would change the result, so eval_const_expressions can't + * do it; but at top level of WHERE, we don't need to distinguish between + * FALSE and NULL results, so it's valid to treat NULL::boolean the same + * as FALSE and then simplify AND/OR accordingly. Conversely, in a top-level + * CHECK constraint, we may treat a NULL the same as TRUE. + * + * Returns the modified qualification. AND/OR flatness is preserved. + */ +static Expr * +find_duplicate_ors(Expr *qual, bool is_check) +{ + if (is_orclause(qual)) + { + List *orlist = NIL; + ListCell *temp; + + /* Recurse */ + foreach(temp, ((BoolExpr *) qual)->args) + { + Expr *arg = (Expr *) lfirst(temp); + + arg = find_duplicate_ors(arg, is_check); + + /* Get rid of any constant inputs */ + if (arg && IsA(arg, Const)) + { + Const *carg = (Const *) arg; + + if (is_check) + { + /* Within OR in CHECK, drop constant FALSE */ + if (!carg->constisnull && !DatumGetBool(carg->constvalue)) + continue; + /* Constant TRUE or NULL, so OR reduces to TRUE */ + return (Expr *) makeBoolConst(true, false); + } + else + { + /* Within OR in WHERE, drop constant FALSE or NULL */ + if (carg->constisnull || !DatumGetBool(carg->constvalue)) + continue; + /* Constant TRUE, so OR reduces to TRUE */ + return arg; + } + } + + orlist = lappend(orlist, arg); + } + + /* Flatten any ORs pulled up to just below here */ + orlist = pull_ors(orlist); + + /* Now we can look for duplicate ORs */ + return process_duplicate_ors(orlist); + } + else if (is_andclause(qual)) + { + List *andlist = NIL; + ListCell *temp; + + /* Recurse */ + foreach(temp, ((BoolExpr *) qual)->args) + { + Expr *arg = (Expr *) lfirst(temp); + + arg = find_duplicate_ors(arg, is_check); + + /* Get rid of any constant inputs */ + if (arg && IsA(arg, Const)) + { + Const *carg = (Const *) arg; + + if (is_check) + { + /* Within AND in CHECK, drop constant TRUE or NULL */ + if (carg->constisnull || DatumGetBool(carg->constvalue)) + continue; + /* Constant FALSE, so AND reduces to FALSE */ + return arg; + } + else + { + /* Within AND in WHERE, drop constant TRUE */ + if (!carg->constisnull && DatumGetBool(carg->constvalue)) + continue; + /* Constant FALSE or NULL, so AND reduces to FALSE */ + return (Expr *) makeBoolConst(false, false); + } + } + + andlist = lappend(andlist, arg); + } + + /* Flatten any ANDs introduced just below here */ + andlist = pull_ands(andlist); + + /* AND of no inputs reduces to TRUE */ + if (andlist == NIL) + return (Expr *) makeBoolConst(true, false); + + /* Single-expression AND just reduces to that expression */ + if (list_length(andlist) == 1) + return (Expr *) linitial(andlist); + + /* Else we still need an AND node */ + return make_andclause(andlist); + } + else + return qual; +} + +/* + * process_duplicate_ors + * Given a list of exprs which are ORed together, try to apply + * the inverse OR distributive law. + * + * Returns the resulting expression (could be an AND clause, an OR + * clause, or maybe even a single subexpression). + */ +static Expr * +process_duplicate_ors(List *orlist) +{ + List *reference = NIL; + int num_subclauses = 0; + List *winners; + List *neworlist; + ListCell *temp; + + /* OR of no inputs reduces to FALSE */ + if (orlist == NIL) + return (Expr *) makeBoolConst(false, false); + + /* Single-expression OR just reduces to that expression */ + if (list_length(orlist) == 1) + return (Expr *) linitial(orlist); + + /* + * Choose the shortest AND clause as the reference list --- obviously, any + * subclause not in this clause isn't in all the clauses. If we find a + * clause that's not an AND, we can treat it as a one-element AND clause, + * which necessarily wins as shortest. + */ + foreach(temp, orlist) + { + Expr *clause = (Expr *) lfirst(temp); + + if (is_andclause(clause)) + { + List *subclauses = ((BoolExpr *) clause)->args; + int nclauses = list_length(subclauses); + + if (reference == NIL || nclauses < num_subclauses) + { + reference = subclauses; + num_subclauses = nclauses; + } + } + else + { + reference = list_make1(clause); + break; + } + } + + /* + * Just in case, eliminate any duplicates in the reference list. + */ + reference = list_union(NIL, reference); + + /* + * Check each element of the reference list to see if it's in all the OR + * clauses. Build a new list of winning clauses. + */ + winners = NIL; + foreach(temp, reference) + { + Expr *refclause = (Expr *) lfirst(temp); + bool win = true; + ListCell *temp2; + + foreach(temp2, orlist) + { + Expr *clause = (Expr *) lfirst(temp2); + + if (is_andclause(clause)) + { + if (!list_member(((BoolExpr *) clause)->args, refclause)) + { + win = false; + break; + } + } + else + { + if (!equal(refclause, clause)) + { + win = false; + break; + } + } + } + + if (win) + winners = lappend(winners, refclause); + } + + /* + * If no winners, we can't transform the OR + */ + if (winners == NIL) + return make_orclause(orlist); + + /* + * Generate new OR list consisting of the remaining sub-clauses. + * + * If any clause degenerates to empty, then we have a situation like (A + * AND B) OR (A), which can be reduced to just A --- that is, the + * additional conditions in other arms of the OR are irrelevant. + * + * Note that because we use list_difference, any multiple occurrences of a + * winning clause in an AND sub-clause will be removed automatically. + */ + neworlist = NIL; + foreach(temp, orlist) + { + Expr *clause = (Expr *) lfirst(temp); + + if (is_andclause(clause)) + { + List *subclauses = ((BoolExpr *) clause)->args; + + subclauses = list_difference(subclauses, winners); + if (subclauses != NIL) + { + if (list_length(subclauses) == 1) + neworlist = lappend(neworlist, linitial(subclauses)); + else + neworlist = lappend(neworlist, make_andclause(subclauses)); + } + else + { + neworlist = NIL; /* degenerate case, see above */ + break; + } + } + else + { + if (!list_member(winners, clause)) + neworlist = lappend(neworlist, clause); + else + { + neworlist = NIL; /* degenerate case, see above */ + break; + } + } + } + + /* + * Append reduced OR to the winners list, if it's not degenerate, handling + * the special case of one element correctly (can that really happen?). + * Also be careful to maintain AND/OR flatness in case we pulled up a + * sub-sub-OR-clause. + */ + if (neworlist != NIL) + { + if (list_length(neworlist) == 1) + winners = lappend(winners, linitial(neworlist)); + else + winners = lappend(winners, make_orclause(pull_ors(neworlist))); + } + + /* + * And return the constructed AND clause, again being wary of a single + * element and AND/OR flatness. + */ + if (list_length(winners) == 1) + return (Expr *) linitial(winners); + else + return make_andclause(pull_ands(winners)); +} diff --git a/src/backend/optimizer/prep/preptlist.c b/src/backend/optimizer/prep/preptlist.c new file mode 100644 index 0000000..133966b --- /dev/null +++ b/src/backend/optimizer/prep/preptlist.c @@ -0,0 +1,497 @@ +/*------------------------------------------------------------------------- + * + * preptlist.c + * Routines to preprocess the parse tree target list + * + * For an INSERT, the targetlist must contain an entry for each attribute of + * the target relation in the correct order. + * + * For an UPDATE, the targetlist just contains the expressions for the new + * column values. + * + * For UPDATE and DELETE queries, the targetlist must also contain "junk" + * tlist entries needed to allow the executor to identify the rows to be + * updated or deleted; for example, the ctid of a heap row. (The planner + * adds these; they're not in what we receive from the planner/rewriter.) + * + * For all query types, there can be additional junk tlist entries, such as + * sort keys, Vars needed for a RETURNING list, and row ID information needed + * for SELECT FOR UPDATE locking and/or EvalPlanQual checking. + * + * The query rewrite phase also does preprocessing of the targetlist (see + * rewriteTargetListIU). The division of labor between here and there is + * partially historical, but it's not entirely arbitrary. The stuff done + * here is closely connected to physical access to tables, whereas the + * rewriter's work is more concerned with SQL semantics. + * + * + * Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group + * Portions Copyright (c) 1994, Regents of the University of California + * + * IDENTIFICATION + * src/backend/optimizer/prep/preptlist.c + * + *------------------------------------------------------------------------- + */ + +#include "postgres.h" + +#include "access/table.h" +#include "nodes/makefuncs.h" +#include "optimizer/appendinfo.h" +#include "optimizer/optimizer.h" +#include "optimizer/prep.h" +#include "optimizer/tlist.h" +#include "parser/parse_coerce.h" +#include "parser/parsetree.h" +#include "utils/rel.h" + +static List *expand_insert_targetlist(List *tlist, Relation rel); + + +/* + * preprocess_targetlist + * Driver for preprocessing the parse tree targetlist. + * + * The preprocessed targetlist is returned in root->processed_tlist. + * Also, if this is an UPDATE, we return a list of target column numbers + * in root->update_colnos. (Resnos in processed_tlist will be consecutive, + * so do not look at that to find out which columns are targets!) + */ +void +preprocess_targetlist(PlannerInfo *root) +{ + Query *parse = root->parse; + int result_relation = parse->resultRelation; + List *range_table = parse->rtable; + CmdType command_type = parse->commandType; + RangeTblEntry *target_rte = NULL; + Relation target_relation = NULL; + List *tlist; + ListCell *lc; + + /* + * If there is a result relation, open it so we can look for missing + * columns and so on. We assume that previous code already acquired at + * least AccessShareLock on the relation, so we need no lock here. + */ + if (result_relation) + { + target_rte = rt_fetch(result_relation, range_table); + + /* + * Sanity check: it'd better be a real relation not, say, a subquery. + * Else parser or rewriter messed up. + */ + if (target_rte->rtekind != RTE_RELATION) + elog(ERROR, "result relation must be a regular relation"); + + target_relation = table_open(target_rte->relid, NoLock); + } + else + Assert(command_type == CMD_SELECT); + + /* + * In an INSERT, the executor expects the targetlist to match the exact + * order of the target table's attributes, including entries for + * attributes not mentioned in the source query. + * + * In an UPDATE, we don't rearrange the tlist order, but we need to make a + * separate list of the target attribute numbers, in tlist order, and then + * renumber the processed_tlist entries to be consecutive. + */ + tlist = parse->targetList; + if (command_type == CMD_INSERT) + tlist = expand_insert_targetlist(tlist, target_relation); + else if (command_type == CMD_UPDATE) + root->update_colnos = extract_update_targetlist_colnos(tlist); + + /* + * For non-inherited UPDATE/DELETE/MERGE, register any junk column(s) + * needed to allow the executor to identify the rows to be updated or + * deleted. In the inheritance case, we do nothing now, leaving this to + * be dealt with when expand_inherited_rtentry() makes the leaf target + * relations. (But there might not be any leaf target relations, in which + * case we must do this in distribute_row_identity_vars().) + */ + if ((command_type == CMD_UPDATE || command_type == CMD_DELETE || + command_type == CMD_MERGE) && + !target_rte->inh) + { + /* row-identity logic expects to add stuff to processed_tlist */ + root->processed_tlist = tlist; + add_row_identity_columns(root, result_relation, + target_rte, target_relation); + tlist = root->processed_tlist; + } + + /* + * For MERGE we also need to handle the target list for each INSERT and + * UPDATE action separately. In addition, we examine the qual of each + * action and add any Vars there (other than those of the target rel) to + * the subplan targetlist. + */ + if (command_type == CMD_MERGE) + { + ListCell *l; + + /* + * For MERGE, handle targetlist of each MergeAction separately. Give + * the same treatment to MergeAction->targetList as we would have + * given to a regular INSERT. For UPDATE, collect the column numbers + * being modified. + */ + foreach(l, parse->mergeActionList) + { + MergeAction *action = (MergeAction *) lfirst(l); + List *vars; + ListCell *l2; + + if (action->commandType == CMD_INSERT) + action->targetList = expand_insert_targetlist(action->targetList, + target_relation); + else if (action->commandType == CMD_UPDATE) + action->updateColnos = + extract_update_targetlist_colnos(action->targetList); + + /* + * Add resjunk entries for any Vars and PlaceHolderVars used in + * each action's targetlist and WHEN condition that belong to + * relations other than the target. We don't expect to see any + * aggregates or window functions here. + */ + vars = pull_var_clause((Node *) + list_concat_copy((List *) action->qual, + action->targetList), + PVC_INCLUDE_PLACEHOLDERS); + foreach(l2, vars) + { + Var *var = (Var *) lfirst(l2); + TargetEntry *tle; + + if (IsA(var, Var) && var->varno == result_relation) + continue; /* don't need it */ + + if (tlist_member((Expr *) var, tlist)) + continue; /* already got it */ + + tle = makeTargetEntry((Expr *) var, + list_length(tlist) + 1, + NULL, true); + tlist = lappend(tlist, tle); + } + list_free(vars); + } + } + + /* + * Add necessary junk columns for rowmarked rels. These values are needed + * for locking of rels selected FOR UPDATE/SHARE, and to do EvalPlanQual + * rechecking. See comments for PlanRowMark in plannodes.h. If you + * change this stanza, see also expand_inherited_rtentry(), which has to + * be able to add on junk columns equivalent to these. + * + * (Someday it might be useful to fold these resjunk columns into the + * row-identity-column management used for UPDATE/DELETE. Today is not + * that day, however. One notable issue is that it seems important that + * the whole-row Vars made here use the real table rowtype, not RECORD, so + * that conversion to/from child relations' rowtypes will happen. Also, + * since these entries don't potentially bloat with more and more child + * relations, there's not really much need for column sharing.) + */ + foreach(lc, root->rowMarks) + { + PlanRowMark *rc = (PlanRowMark *) lfirst(lc); + Var *var; + char resname[32]; + TargetEntry *tle; + + /* child rels use the same junk attrs as their parents */ + if (rc->rti != rc->prti) + continue; + + if (rc->allMarkTypes & ~(1 << ROW_MARK_COPY)) + { + /* Need to fetch TID */ + var = makeVar(rc->rti, + SelfItemPointerAttributeNumber, + TIDOID, + -1, + InvalidOid, + 0); + snprintf(resname, sizeof(resname), "ctid%u", rc->rowmarkId); + tle = makeTargetEntry((Expr *) var, + list_length(tlist) + 1, + pstrdup(resname), + true); + tlist = lappend(tlist, tle); + } + if (rc->allMarkTypes & (1 << ROW_MARK_COPY)) + { + /* Need the whole row as a junk var */ + var = makeWholeRowVar(rt_fetch(rc->rti, range_table), + rc->rti, + 0, + false); + snprintf(resname, sizeof(resname), "wholerow%u", rc->rowmarkId); + tle = makeTargetEntry((Expr *) var, + list_length(tlist) + 1, + pstrdup(resname), + true); + tlist = lappend(tlist, tle); + } + + /* If parent of inheritance tree, always fetch the tableoid too. */ + if (rc->isParent) + { + var = makeVar(rc->rti, + TableOidAttributeNumber, + OIDOID, + -1, + InvalidOid, + 0); + snprintf(resname, sizeof(resname), "tableoid%u", rc->rowmarkId); + tle = makeTargetEntry((Expr *) var, + list_length(tlist) + 1, + pstrdup(resname), + true); + tlist = lappend(tlist, tle); + } + } + + /* + * If the query has a RETURNING list, add resjunk entries for any Vars + * used in RETURNING that belong to other relations. We need to do this + * to make these Vars available for the RETURNING calculation. Vars that + * belong to the result rel don't need to be added, because they will be + * made to refer to the actual heap tuple. + */ + if (parse->returningList && list_length(parse->rtable) > 1) + { + List *vars; + ListCell *l; + + vars = pull_var_clause((Node *) parse->returningList, + PVC_RECURSE_AGGREGATES | + PVC_RECURSE_WINDOWFUNCS | + PVC_INCLUDE_PLACEHOLDERS); + foreach(l, vars) + { + Var *var = (Var *) lfirst(l); + TargetEntry *tle; + + if (IsA(var, Var) && + var->varno == result_relation) + continue; /* don't need it */ + + if (tlist_member((Expr *) var, tlist)) + continue; /* already got it */ + + tle = makeTargetEntry((Expr *) var, + list_length(tlist) + 1, + NULL, + true); + + tlist = lappend(tlist, tle); + } + list_free(vars); + } + + root->processed_tlist = tlist; + + if (target_relation) + table_close(target_relation, NoLock); +} + +/* + * extract_update_targetlist_colnos + * Extract a list of the target-table column numbers that + * an UPDATE's targetlist wants to assign to, then renumber. + * + * The convention in the parser and rewriter is that the resnos in an + * UPDATE's non-resjunk TLE entries are the target column numbers + * to assign to. Here, we extract that info into a separate list, and + * then convert the tlist to the sequential-numbering convention that's + * used by all other query types. + * + * This is also applied to the tlist associated with INSERT ... ON CONFLICT + * ... UPDATE, although not till much later in planning. + */ +List * +extract_update_targetlist_colnos(List *tlist) +{ + List *update_colnos = NIL; + AttrNumber nextresno = 1; + ListCell *lc; + + foreach(lc, tlist) + { + TargetEntry *tle = (TargetEntry *) lfirst(lc); + + if (!tle->resjunk) + update_colnos = lappend_int(update_colnos, tle->resno); + tle->resno = nextresno++; + } + return update_colnos; +} + + +/***************************************************************************** + * + * TARGETLIST EXPANSION + * + *****************************************************************************/ + +/* + * expand_insert_targetlist + * Given a target list as generated by the parser and a result relation, + * add targetlist entries for any missing attributes, and ensure the + * non-junk attributes appear in proper field order. + * + * Once upon a time we also did more or less this with UPDATE targetlists, + * but now this code is only applied to INSERT targetlists. + */ +static List * +expand_insert_targetlist(List *tlist, Relation rel) +{ + List *new_tlist = NIL; + ListCell *tlist_item; + int attrno, + numattrs; + + tlist_item = list_head(tlist); + + /* + * The rewriter should have already ensured that the TLEs are in correct + * order; but we have to insert TLEs for any missing attributes. + * + * Scan the tuple description in the relation's relcache entry to make + * sure we have all the user attributes in the right order. + */ + numattrs = RelationGetNumberOfAttributes(rel); + + for (attrno = 1; attrno <= numattrs; attrno++) + { + Form_pg_attribute att_tup = TupleDescAttr(rel->rd_att, attrno - 1); + TargetEntry *new_tle = NULL; + + if (tlist_item != NULL) + { + TargetEntry *old_tle = (TargetEntry *) lfirst(tlist_item); + + if (!old_tle->resjunk && old_tle->resno == attrno) + { + new_tle = old_tle; + tlist_item = lnext(tlist, tlist_item); + } + } + + if (new_tle == NULL) + { + /* + * Didn't find a matching tlist entry, so make one. + * + * INSERTs should insert NULL in this case. (We assume the + * rewriter would have inserted any available non-NULL default + * value.) Also, if the column isn't dropped, apply any domain + * constraints that might exist --- this is to catch domain NOT + * NULL. + * + * When generating a NULL constant for a dropped column, we label + * it INT4 (any other guaranteed-to-exist datatype would do as + * well). We can't label it with the dropped column's datatype + * since that might not exist anymore. It does not really matter + * what we claim the type is, since NULL is NULL --- its + * representation is datatype-independent. This could perhaps + * confuse code comparing the finished plan to the target + * relation, however. + */ + Oid atttype = att_tup->atttypid; + Oid attcollation = att_tup->attcollation; + Node *new_expr; + + if (!att_tup->attisdropped) + { + new_expr = (Node *) makeConst(atttype, + -1, + attcollation, + att_tup->attlen, + (Datum) 0, + true, /* isnull */ + att_tup->attbyval); + new_expr = coerce_to_domain(new_expr, + InvalidOid, -1, + atttype, + COERCION_IMPLICIT, + COERCE_IMPLICIT_CAST, + -1, + false); + } + else + { + /* Insert NULL for dropped column */ + new_expr = (Node *) makeConst(INT4OID, + -1, + InvalidOid, + sizeof(int32), + (Datum) 0, + true, /* isnull */ + true /* byval */ ); + } + + new_tle = makeTargetEntry((Expr *) new_expr, + attrno, + pstrdup(NameStr(att_tup->attname)), + false); + } + + new_tlist = lappend(new_tlist, new_tle); + } + + /* + * The remaining tlist entries should be resjunk; append them all to the + * end of the new tlist, making sure they have resnos higher than the last + * real attribute. (Note: although the rewriter already did such + * renumbering, we have to do it again here in case we added NULL entries + * above.) + */ + while (tlist_item) + { + TargetEntry *old_tle = (TargetEntry *) lfirst(tlist_item); + + if (!old_tle->resjunk) + elog(ERROR, "targetlist is not sorted correctly"); + /* Get the resno right, but don't copy unnecessarily */ + if (old_tle->resno != attrno) + { + old_tle = flatCopyTargetEntry(old_tle); + old_tle->resno = attrno; + } + new_tlist = lappend(new_tlist, old_tle); + attrno++; + tlist_item = lnext(tlist, tlist_item); + } + + return new_tlist; +} + + +/* + * Locate PlanRowMark for given RT index, or return NULL if none + * + * This probably ought to be elsewhere, but there's no very good place + */ +PlanRowMark * +get_plan_rowmark(List *rowmarks, Index rtindex) +{ + ListCell *l; + + foreach(l, rowmarks) + { + PlanRowMark *rc = (PlanRowMark *) lfirst(l); + + if (rc->rti == rtindex) + return rc; + } + return NULL; +} diff --git a/src/backend/optimizer/prep/prepunion.c b/src/backend/optimizer/prep/prepunion.c new file mode 100644 index 0000000..f004fad --- /dev/null +++ b/src/backend/optimizer/prep/prepunion.c @@ -0,0 +1,1420 @@ +/*------------------------------------------------------------------------- + * + * prepunion.c + * Routines to plan set-operation queries. The filename is a leftover + * from a time when only UNIONs were implemented. + * + * There are two code paths in the planner for set-operation queries. + * If a subquery consists entirely of simple UNION ALL operations, it + * is converted into an "append relation". Otherwise, it is handled + * by the general code in this module (plan_set_operations and its + * subroutines). There is some support code here for the append-relation + * case, but most of the heavy lifting for that is done elsewhere, + * notably in prepjointree.c and allpaths.c. + * + * Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group + * Portions Copyright (c) 1994, Regents of the University of California + * + * + * IDENTIFICATION + * src/backend/optimizer/prep/prepunion.c + * + *------------------------------------------------------------------------- + */ +#include "postgres.h" + +#include "access/htup_details.h" +#include "access/sysattr.h" +#include "catalog/partition.h" +#include "catalog/pg_inherits.h" +#include "catalog/pg_type.h" +#include "miscadmin.h" +#include "nodes/makefuncs.h" +#include "nodes/nodeFuncs.h" +#include "optimizer/cost.h" +#include "optimizer/pathnode.h" +#include "optimizer/paths.h" +#include "optimizer/planmain.h" +#include "optimizer/planner.h" +#include "optimizer/prep.h" +#include "optimizer/tlist.h" +#include "parser/parse_coerce.h" +#include "parser/parsetree.h" +#include "utils/lsyscache.h" +#include "utils/rel.h" +#include "utils/selfuncs.h" +#include "utils/syscache.h" + + +static RelOptInfo *recurse_set_operations(Node *setOp, PlannerInfo *root, + List *colTypes, List *colCollations, + bool junkOK, + int flag, List *refnames_tlist, + List **pTargetList, + double *pNumGroups); +static RelOptInfo *generate_recursion_path(SetOperationStmt *setOp, + PlannerInfo *root, + List *refnames_tlist, + List **pTargetList); +static RelOptInfo *generate_union_paths(SetOperationStmt *op, PlannerInfo *root, + List *refnames_tlist, + List **pTargetList); +static RelOptInfo *generate_nonunion_paths(SetOperationStmt *op, PlannerInfo *root, + List *refnames_tlist, + List **pTargetList); +static List *plan_union_children(PlannerInfo *root, + SetOperationStmt *top_union, + List *refnames_tlist, + List **tlist_list); +static Path *make_union_unique(SetOperationStmt *op, Path *path, List *tlist, + PlannerInfo *root); +static void postprocess_setop_rel(PlannerInfo *root, RelOptInfo *rel); +static bool choose_hashed_setop(PlannerInfo *root, List *groupClauses, + Path *input_path, + double dNumGroups, double dNumOutputRows, + const char *construct); +static List *generate_setop_tlist(List *colTypes, List *colCollations, + int flag, + Index varno, + bool hack_constants, + List *input_tlist, + List *refnames_tlist); +static List *generate_append_tlist(List *colTypes, List *colCollations, + bool flag, + List *input_tlists, + List *refnames_tlist); +static List *generate_setop_grouplist(SetOperationStmt *op, List *targetlist); + + +/* + * plan_set_operations + * + * Plans the queries for a tree of set operations (UNION/INTERSECT/EXCEPT) + * + * This routine only deals with the setOperations tree of the given query. + * Any top-level ORDER BY requested in root->parse->sortClause will be handled + * when we return to grouping_planner; likewise for LIMIT. + * + * What we return is an "upperrel" RelOptInfo containing at least one Path + * that implements the set-operation tree. In addition, root->processed_tlist + * receives a targetlist representing the output of the topmost setop node. + */ +RelOptInfo * +plan_set_operations(PlannerInfo *root) +{ + Query *parse = root->parse; + SetOperationStmt *topop = castNode(SetOperationStmt, parse->setOperations); + Node *node; + RangeTblEntry *leftmostRTE; + Query *leftmostQuery; + RelOptInfo *setop_rel; + List *top_tlist; + + Assert(topop); + + /* check for unsupported stuff */ + Assert(parse->jointree->fromlist == NIL); + Assert(parse->jointree->quals == NULL); + Assert(parse->groupClause == NIL); + Assert(parse->havingQual == NULL); + Assert(parse->windowClause == NIL); + Assert(parse->distinctClause == NIL); + + /* + * In the outer query level, we won't have any true equivalences to deal + * with; but we do want to be able to make pathkeys, which will require + * single-member EquivalenceClasses. Indicate that EC merging is complete + * so that pathkeys.c won't complain. + */ + Assert(root->eq_classes == NIL); + root->ec_merging_done = true; + + /* + * We'll need to build RelOptInfos for each of the leaf subqueries, which + * are RTE_SUBQUERY rangetable entries in this Query. Prepare the index + * arrays for those, and for AppendRelInfos in case they're needed. + */ + setup_simple_rel_arrays(root); + + /* + * Find the leftmost component Query. We need to use its column names for + * all generated tlists (else SELECT INTO won't work right). + */ + node = topop->larg; + while (node && IsA(node, SetOperationStmt)) + node = ((SetOperationStmt *) node)->larg; + Assert(node && IsA(node, RangeTblRef)); + leftmostRTE = root->simple_rte_array[((RangeTblRef *) node)->rtindex]; + leftmostQuery = leftmostRTE->subquery; + Assert(leftmostQuery != NULL); + + /* + * If the topmost node is a recursive union, it needs special processing. + */ + if (root->hasRecursion) + { + setop_rel = generate_recursion_path(topop, root, + leftmostQuery->targetList, + &top_tlist); + } + else + { + /* + * Recurse on setOperations tree to generate paths for set ops. The + * final output paths should have just the column types shown as the + * output from the top-level node, plus possibly resjunk working + * columns (we can rely on upper-level nodes to deal with that). + */ + setop_rel = recurse_set_operations((Node *) topop, root, + topop->colTypes, topop->colCollations, + true, -1, + leftmostQuery->targetList, + &top_tlist, + NULL); + } + + /* Must return the built tlist into root->processed_tlist. */ + root->processed_tlist = top_tlist; + + return setop_rel; +} + +/* + * recurse_set_operations + * Recursively handle one step in a tree of set operations + * + * colTypes: OID list of set-op's result column datatypes + * colCollations: OID list of set-op's result column collations + * junkOK: if true, child resjunk columns may be left in the result + * flag: if >= 0, add a resjunk output column indicating value of flag + * refnames_tlist: targetlist to take column names from + * + * Returns a RelOptInfo for the subtree, as well as these output parameters: + * *pTargetList: receives the fully-fledged tlist for the subtree's top plan + * *pNumGroups: if not NULL, we estimate the number of distinct groups + * in the result, and store it there + * + * The pTargetList output parameter is mostly redundant with the pathtarget + * of the returned RelOptInfo, but for the moment we need it because much of + * the logic in this file depends on flag columns being marked resjunk. + * Pending a redesign of how that works, this is the easy way out. + * + * We don't have to care about typmods here: the only allowed difference + * between set-op input and output typmods is input is a specific typmod + * and output is -1, and that does not require a coercion. + */ +static RelOptInfo * +recurse_set_operations(Node *setOp, PlannerInfo *root, + List *colTypes, List *colCollations, + bool junkOK, + int flag, List *refnames_tlist, + List **pTargetList, + double *pNumGroups) +{ + RelOptInfo *rel = NULL; /* keep compiler quiet */ + + /* Guard against stack overflow due to overly complex setop nests */ + check_stack_depth(); + + if (IsA(setOp, RangeTblRef)) + { + RangeTblRef *rtr = (RangeTblRef *) setOp; + RangeTblEntry *rte = root->simple_rte_array[rtr->rtindex]; + Query *subquery = rte->subquery; + PlannerInfo *subroot; + RelOptInfo *final_rel; + Path *subpath; + Path *path; + List *tlist; + + Assert(subquery != NULL); + + /* Build a RelOptInfo for this leaf subquery. */ + rel = build_simple_rel(root, rtr->rtindex, NULL); + + /* plan_params should not be in use in current query level */ + Assert(root->plan_params == NIL); + + /* Generate a subroot and Paths for the subquery */ + subroot = rel->subroot = subquery_planner(root->glob, subquery, + root, + false, + root->tuple_fraction); + + /* + * It should not be possible for the primitive query to contain any + * cross-references to other primitive queries in the setop tree. + */ + if (root->plan_params) + elog(ERROR, "unexpected outer reference in set operation subquery"); + + /* Figure out the appropriate target list for this subquery. */ + tlist = generate_setop_tlist(colTypes, colCollations, + flag, + rtr->rtindex, + true, + subroot->processed_tlist, + refnames_tlist); + rel->reltarget = create_pathtarget(root, tlist); + + /* Return the fully-fledged tlist to caller, too */ + *pTargetList = tlist; + + /* + * Mark rel with estimated output rows, width, etc. Note that we have + * to do this before generating outer-query paths, else + * cost_subqueryscan is not happy. + */ + set_subquery_size_estimates(root, rel); + + /* + * Since we may want to add a partial path to this relation, we must + * set its consider_parallel flag correctly. + */ + final_rel = fetch_upper_rel(subroot, UPPERREL_FINAL, NULL); + rel->consider_parallel = final_rel->consider_parallel; + + /* + * For the moment, we consider only a single Path for the subquery. + * This should change soon (make it look more like + * set_subquery_pathlist). + */ + subpath = get_cheapest_fractional_path(final_rel, + root->tuple_fraction); + + /* + * Stick a SubqueryScanPath atop that. + * + * We don't bother to determine the subquery's output ordering since + * it won't be reflected in the set-op result anyhow; so just label + * the SubqueryScanPath with nil pathkeys. (XXX that should change + * soon too, likely.) + */ + path = (Path *) create_subqueryscan_path(root, rel, subpath, + NIL, NULL); + + add_path(rel, path); + + /* + * If we have a partial path for the child relation, we can use that + * to build a partial path for this relation. But there's no point in + * considering any path but the cheapest. + */ + if (rel->consider_parallel && bms_is_empty(rel->lateral_relids) && + final_rel->partial_pathlist != NIL) + { + Path *partial_subpath; + Path *partial_path; + + partial_subpath = linitial(final_rel->partial_pathlist); + partial_path = (Path *) + create_subqueryscan_path(root, rel, partial_subpath, + NIL, NULL); + add_partial_path(rel, partial_path); + } + + /* + * Estimate number of groups if caller wants it. If the subquery used + * grouping or aggregation, its output is probably mostly unique + * anyway; otherwise do statistical estimation. + * + * XXX you don't really want to know about this: we do the estimation + * using the subquery's original targetlist expressions, not the + * subroot->processed_tlist which might seem more appropriate. The + * reason is that if the subquery is itself a setop, it may return a + * processed_tlist containing "varno 0" Vars generated by + * generate_append_tlist, and those would confuse estimate_num_groups + * mightily. We ought to get rid of the "varno 0" hack, but that + * requires a redesign of the parsetree representation of setops, so + * that there can be an RTE corresponding to each setop's output. + */ + if (pNumGroups) + { + if (subquery->groupClause || subquery->groupingSets || + subquery->distinctClause || + subroot->hasHavingQual || subquery->hasAggs) + *pNumGroups = subpath->rows; + else + *pNumGroups = estimate_num_groups(subroot, + get_tlist_exprs(subquery->targetList, false), + subpath->rows, + NULL, + NULL); + } + } + else if (IsA(setOp, SetOperationStmt)) + { + SetOperationStmt *op = (SetOperationStmt *) setOp; + + /* UNIONs are much different from INTERSECT/EXCEPT */ + if (op->op == SETOP_UNION) + rel = generate_union_paths(op, root, + refnames_tlist, + pTargetList); + else + rel = generate_nonunion_paths(op, root, + refnames_tlist, + pTargetList); + if (pNumGroups) + *pNumGroups = rel->rows; + + /* + * If necessary, add a Result node to project the caller-requested + * output columns. + * + * XXX you don't really want to know about this: setrefs.c will apply + * fix_upper_expr() to the Result node's tlist. This would fail if the + * Vars generated by generate_setop_tlist() were not exactly equal() + * to the corresponding tlist entries of the subplan. However, since + * the subplan was generated by generate_union_paths() or + * generate_nonunion_paths(), and hence its tlist was generated by + * generate_append_tlist(), this will work. We just tell + * generate_setop_tlist() to use varno 0. + */ + if (flag >= 0 || + !tlist_same_datatypes(*pTargetList, colTypes, junkOK) || + !tlist_same_collations(*pTargetList, colCollations, junkOK)) + { + PathTarget *target; + ListCell *lc; + + *pTargetList = generate_setop_tlist(colTypes, colCollations, + flag, + 0, + false, + *pTargetList, + refnames_tlist); + target = create_pathtarget(root, *pTargetList); + + /* Apply projection to each path */ + foreach(lc, rel->pathlist) + { + Path *subpath = (Path *) lfirst(lc); + Path *path; + + Assert(subpath->param_info == NULL); + path = apply_projection_to_path(root, subpath->parent, + subpath, target); + /* If we had to add a Result, path is different from subpath */ + if (path != subpath) + lfirst(lc) = path; + } + + /* Apply projection to each partial path */ + foreach(lc, rel->partial_pathlist) + { + Path *subpath = (Path *) lfirst(lc); + Path *path; + + Assert(subpath->param_info == NULL); + + /* avoid apply_projection_to_path, in case of multiple refs */ + path = (Path *) create_projection_path(root, subpath->parent, + subpath, target); + lfirst(lc) = path; + } + } + } + else + { + elog(ERROR, "unrecognized node type: %d", + (int) nodeTag(setOp)); + *pTargetList = NIL; + } + + postprocess_setop_rel(root, rel); + + return rel; +} + +/* + * Generate paths for a recursive UNION node + */ +static RelOptInfo * +generate_recursion_path(SetOperationStmt *setOp, PlannerInfo *root, + List *refnames_tlist, + List **pTargetList) +{ + RelOptInfo *result_rel; + Path *path; + RelOptInfo *lrel, + *rrel; + Path *lpath; + Path *rpath; + List *lpath_tlist; + List *rpath_tlist; + List *tlist; + List *groupList; + double dNumGroups; + + /* Parser should have rejected other cases */ + if (setOp->op != SETOP_UNION) + elog(ERROR, "only UNION queries can be recursive"); + /* Worktable ID should be assigned */ + Assert(root->wt_param_id >= 0); + + /* + * Unlike a regular UNION node, process the left and right inputs + * separately without any intention of combining them into one Append. + */ + lrel = recurse_set_operations(setOp->larg, root, + setOp->colTypes, setOp->colCollations, + false, -1, + refnames_tlist, + &lpath_tlist, + NULL); + lpath = lrel->cheapest_total_path; + /* The right path will want to look at the left one ... */ + root->non_recursive_path = lpath; + rrel = recurse_set_operations(setOp->rarg, root, + setOp->colTypes, setOp->colCollations, + false, -1, + refnames_tlist, + &rpath_tlist, + NULL); + rpath = rrel->cheapest_total_path; + root->non_recursive_path = NULL; + + /* + * Generate tlist for RecursiveUnion path node --- same as in Append cases + */ + tlist = generate_append_tlist(setOp->colTypes, setOp->colCollations, false, + list_make2(lpath_tlist, rpath_tlist), + refnames_tlist); + + *pTargetList = tlist; + + /* Build result relation. */ + result_rel = fetch_upper_rel(root, UPPERREL_SETOP, + bms_union(lrel->relids, rrel->relids)); + result_rel->reltarget = create_pathtarget(root, tlist); + + /* + * If UNION, identify the grouping operators + */ + if (setOp->all) + { + groupList = NIL; + dNumGroups = 0; + } + else + { + /* Identify the grouping semantics */ + groupList = generate_setop_grouplist(setOp, tlist); + + /* We only support hashing here */ + if (!grouping_is_hashable(groupList)) + ereport(ERROR, + (errcode(ERRCODE_FEATURE_NOT_SUPPORTED), + errmsg("could not implement recursive UNION"), + errdetail("All column datatypes must be hashable."))); + + /* + * For the moment, take the number of distinct groups as equal to the + * total input size, ie, the worst case. + */ + dNumGroups = lpath->rows + rpath->rows * 10; + } + + /* + * And make the path node. + */ + path = (Path *) create_recursiveunion_path(root, + result_rel, + lpath, + rpath, + result_rel->reltarget, + groupList, + root->wt_param_id, + dNumGroups); + + add_path(result_rel, path); + postprocess_setop_rel(root, result_rel); + return result_rel; +} + +/* + * Generate paths for a UNION or UNION ALL node + */ +static RelOptInfo * +generate_union_paths(SetOperationStmt *op, PlannerInfo *root, + List *refnames_tlist, + List **pTargetList) +{ + Relids relids = NULL; + RelOptInfo *result_rel; + double save_fraction = root->tuple_fraction; + ListCell *lc; + List *pathlist = NIL; + List *partial_pathlist = NIL; + bool partial_paths_valid = true; + bool consider_parallel = true; + List *rellist; + List *tlist_list; + List *tlist; + Path *path; + + /* + * If plain UNION, tell children to fetch all tuples. + * + * Note: in UNION ALL, we pass the top-level tuple_fraction unmodified to + * each arm of the UNION ALL. One could make a case for reducing the + * tuple fraction for later arms (discounting by the expected size of the + * earlier arms' results) but it seems not worth the trouble. The normal + * case where tuple_fraction isn't already zero is a LIMIT at top level, + * and passing it down as-is is usually enough to get the desired result + * of preferring fast-start plans. + */ + if (!op->all) + root->tuple_fraction = 0.0; + + /* + * If any of my children are identical UNION nodes (same op, all-flag, and + * colTypes) then they can be merged into this node so that we generate + * only one Append and unique-ification for the lot. Recurse to find such + * nodes and compute their children's paths. + */ + rellist = plan_union_children(root, op, refnames_tlist, &tlist_list); + + /* + * Generate tlist for Append plan node. + * + * The tlist for an Append plan isn't important as far as the Append is + * concerned, but we must make it look real anyway for the benefit of the + * next plan level up. + */ + tlist = generate_append_tlist(op->colTypes, op->colCollations, false, + tlist_list, refnames_tlist); + + *pTargetList = tlist; + + /* Build path lists and relid set. */ + foreach(lc, rellist) + { + RelOptInfo *rel = lfirst(lc); + + pathlist = lappend(pathlist, rel->cheapest_total_path); + + if (consider_parallel) + { + if (!rel->consider_parallel) + { + consider_parallel = false; + partial_paths_valid = false; + } + else if (rel->partial_pathlist == NIL) + partial_paths_valid = false; + else + partial_pathlist = lappend(partial_pathlist, + linitial(rel->partial_pathlist)); + } + + relids = bms_union(relids, rel->relids); + } + + /* Build result relation. */ + result_rel = fetch_upper_rel(root, UPPERREL_SETOP, relids); + result_rel->reltarget = create_pathtarget(root, tlist); + result_rel->consider_parallel = consider_parallel; + + /* + * Append the child results together. + */ + path = (Path *) create_append_path(root, result_rel, pathlist, NIL, + NIL, NULL, 0, false, -1); + + /* + * For UNION ALL, we just need the Append path. For UNION, need to add + * node(s) to remove duplicates. + */ + if (!op->all) + path = make_union_unique(op, path, tlist, root); + + add_path(result_rel, path); + + /* + * Estimate number of groups. For now we just assume the output is unique + * --- this is certainly true for the UNION case, and we want worst-case + * estimates anyway. + */ + result_rel->rows = path->rows; + + /* + * Now consider doing the same thing using the partial paths plus Append + * plus Gather. + */ + if (partial_paths_valid) + { + Path *ppath; + ListCell *lc; + int parallel_workers = 0; + + /* Find the highest number of workers requested for any subpath. */ + foreach(lc, partial_pathlist) + { + Path *path = lfirst(lc); + + parallel_workers = Max(parallel_workers, path->parallel_workers); + } + Assert(parallel_workers > 0); + + /* + * If the use of parallel append is permitted, always request at least + * log2(# of children) paths. We assume it can be useful to have + * extra workers in this case because they will be spread out across + * the children. The precise formula is just a guess; see + * add_paths_to_append_rel. + */ + if (enable_parallel_append) + { + parallel_workers = Max(parallel_workers, + fls(list_length(partial_pathlist))); + parallel_workers = Min(parallel_workers, + max_parallel_workers_per_gather); + } + Assert(parallel_workers > 0); + + ppath = (Path *) + create_append_path(root, result_rel, NIL, partial_pathlist, + NIL, NULL, + parallel_workers, enable_parallel_append, + -1); + ppath = (Path *) + create_gather_path(root, result_rel, ppath, + result_rel->reltarget, NULL, NULL); + if (!op->all) + ppath = make_union_unique(op, ppath, tlist, root); + add_path(result_rel, ppath); + } + + /* Undo effects of possibly forcing tuple_fraction to 0 */ + root->tuple_fraction = save_fraction; + + return result_rel; +} + +/* + * Generate paths for an INTERSECT, INTERSECT ALL, EXCEPT, or EXCEPT ALL node + */ +static RelOptInfo * +generate_nonunion_paths(SetOperationStmt *op, PlannerInfo *root, + List *refnames_tlist, + List **pTargetList) +{ + RelOptInfo *result_rel; + RelOptInfo *lrel, + *rrel; + double save_fraction = root->tuple_fraction; + Path *lpath, + *rpath, + *path; + List *lpath_tlist, + *rpath_tlist, + *tlist_list, + *tlist, + *groupList, + *pathlist; + double dLeftGroups, + dRightGroups, + dNumGroups, + dNumOutputRows; + bool use_hash; + SetOpCmd cmd; + int firstFlag; + + /* + * Tell children to fetch all tuples. + */ + root->tuple_fraction = 0.0; + + /* Recurse on children, ensuring their outputs are marked */ + lrel = recurse_set_operations(op->larg, root, + op->colTypes, op->colCollations, + false, 0, + refnames_tlist, + &lpath_tlist, + &dLeftGroups); + lpath = lrel->cheapest_total_path; + rrel = recurse_set_operations(op->rarg, root, + op->colTypes, op->colCollations, + false, 1, + refnames_tlist, + &rpath_tlist, + &dRightGroups); + rpath = rrel->cheapest_total_path; + + /* Undo effects of forcing tuple_fraction to 0 */ + root->tuple_fraction = save_fraction; + + /* + * For EXCEPT, we must put the left input first. For INTERSECT, either + * order should give the same results, and we prefer to put the smaller + * input first in order to minimize the size of the hash table in the + * hashing case. "Smaller" means the one with the fewer groups. + */ + if (op->op == SETOP_EXCEPT || dLeftGroups <= dRightGroups) + { + pathlist = list_make2(lpath, rpath); + tlist_list = list_make2(lpath_tlist, rpath_tlist); + firstFlag = 0; + } + else + { + pathlist = list_make2(rpath, lpath); + tlist_list = list_make2(rpath_tlist, lpath_tlist); + firstFlag = 1; + } + + /* + * Generate tlist for Append plan node. + * + * The tlist for an Append plan isn't important as far as the Append is + * concerned, but we must make it look real anyway for the benefit of the + * next plan level up. In fact, it has to be real enough that the flag + * column is shown as a variable not a constant, else setrefs.c will get + * confused. + */ + tlist = generate_append_tlist(op->colTypes, op->colCollations, true, + tlist_list, refnames_tlist); + + *pTargetList = tlist; + + /* Build result relation. */ + result_rel = fetch_upper_rel(root, UPPERREL_SETOP, + bms_union(lrel->relids, rrel->relids)); + result_rel->reltarget = create_pathtarget(root, tlist); + + /* + * Append the child results together. + */ + path = (Path *) create_append_path(root, result_rel, pathlist, NIL, + NIL, NULL, 0, false, -1); + + /* Identify the grouping semantics */ + groupList = generate_setop_grouplist(op, tlist); + + /* + * Estimate number of distinct groups that we'll need hashtable entries + * for; this is the size of the left-hand input for EXCEPT, or the smaller + * input for INTERSECT. Also estimate the number of eventual output rows. + * In non-ALL cases, we estimate each group produces one output row; in + * ALL cases use the relevant relation size. These are worst-case + * estimates, of course, but we need to be conservative. + */ + if (op->op == SETOP_EXCEPT) + { + dNumGroups = dLeftGroups; + dNumOutputRows = op->all ? lpath->rows : dNumGroups; + } + else + { + dNumGroups = Min(dLeftGroups, dRightGroups); + dNumOutputRows = op->all ? Min(lpath->rows, rpath->rows) : dNumGroups; + } + + /* + * Decide whether to hash or sort, and add a sort node if needed. + */ + use_hash = choose_hashed_setop(root, groupList, path, + dNumGroups, dNumOutputRows, + (op->op == SETOP_INTERSECT) ? "INTERSECT" : "EXCEPT"); + + if (groupList && !use_hash) + path = (Path *) create_sort_path(root, + result_rel, + path, + make_pathkeys_for_sortclauses(root, + groupList, + tlist), + -1.0); + + /* + * Finally, add a SetOp path node to generate the correct output. + */ + switch (op->op) + { + case SETOP_INTERSECT: + cmd = op->all ? SETOPCMD_INTERSECT_ALL : SETOPCMD_INTERSECT; + break; + case SETOP_EXCEPT: + cmd = op->all ? SETOPCMD_EXCEPT_ALL : SETOPCMD_EXCEPT; + break; + default: + elog(ERROR, "unrecognized set op: %d", (int) op->op); + cmd = SETOPCMD_INTERSECT; /* keep compiler quiet */ + break; + } + path = (Path *) create_setop_path(root, + result_rel, + path, + cmd, + use_hash ? SETOP_HASHED : SETOP_SORTED, + groupList, + list_length(op->colTypes) + 1, + use_hash ? firstFlag : -1, + dNumGroups, + dNumOutputRows); + + result_rel->rows = path->rows; + add_path(result_rel, path); + return result_rel; +} + +/* + * Pull up children of a UNION node that are identically-propertied UNIONs. + * + * NOTE: we can also pull a UNION ALL up into a UNION, since the distinct + * output rows will be lost anyway. + * + * NOTE: currently, we ignore collations while determining if a child has + * the same properties. This is semantically sound only so long as all + * collations have the same notion of equality. It is valid from an + * implementation standpoint because we don't care about the ordering of + * a UNION child's result: UNION ALL results are always unordered, and + * generate_union_paths will force a fresh sort if the top level is a UNION. + */ +static List * +plan_union_children(PlannerInfo *root, + SetOperationStmt *top_union, + List *refnames_tlist, + List **tlist_list) +{ + List *pending_rels = list_make1(top_union); + List *result = NIL; + List *child_tlist; + + *tlist_list = NIL; + + while (pending_rels != NIL) + { + Node *setOp = linitial(pending_rels); + + pending_rels = list_delete_first(pending_rels); + + if (IsA(setOp, SetOperationStmt)) + { + SetOperationStmt *op = (SetOperationStmt *) setOp; + + if (op->op == top_union->op && + (op->all == top_union->all || op->all) && + equal(op->colTypes, top_union->colTypes)) + { + /* Same UNION, so fold children into parent */ + pending_rels = lcons(op->rarg, pending_rels); + pending_rels = lcons(op->larg, pending_rels); + continue; + } + } + + /* + * Not same, so plan this child separately. + * + * Note we disallow any resjunk columns in child results. This is + * necessary since the Append node that implements the union won't do + * any projection, and upper levels will get confused if some of our + * output tuples have junk and some don't. This case only arises when + * we have an EXCEPT or INTERSECT as child, else there won't be + * resjunk anyway. + */ + result = lappend(result, recurse_set_operations(setOp, root, + top_union->colTypes, + top_union->colCollations, + false, -1, + refnames_tlist, + &child_tlist, + NULL)); + *tlist_list = lappend(*tlist_list, child_tlist); + } + + return result; +} + +/* + * Add nodes to the given path tree to unique-ify the result of a UNION. + */ +static Path * +make_union_unique(SetOperationStmt *op, Path *path, List *tlist, + PlannerInfo *root) +{ + RelOptInfo *result_rel = fetch_upper_rel(root, UPPERREL_SETOP, NULL); + List *groupList; + double dNumGroups; + + /* Identify the grouping semantics */ + groupList = generate_setop_grouplist(op, tlist); + + /* + * XXX for the moment, take the number of distinct groups as equal to the + * total input size, ie, the worst case. This is too conservative, but + * it's not clear how to get a decent estimate of the true size. One + * should note as well the propensity of novices to write UNION rather + * than UNION ALL even when they don't expect any duplicates... + */ + dNumGroups = path->rows; + + /* Decide whether to hash or sort */ + if (choose_hashed_setop(root, groupList, path, + dNumGroups, dNumGroups, + "UNION")) + { + /* Hashed aggregate plan --- no sort needed */ + path = (Path *) create_agg_path(root, + result_rel, + path, + create_pathtarget(root, tlist), + AGG_HASHED, + AGGSPLIT_SIMPLE, + groupList, + NIL, + NULL, + dNumGroups); + } + else + { + /* Sort and Unique */ + if (groupList) + path = (Path *) + create_sort_path(root, + result_rel, + path, + make_pathkeys_for_sortclauses(root, + groupList, + tlist), + -1.0); + path = (Path *) create_upper_unique_path(root, + result_rel, + path, + list_length(path->pathkeys), + dNumGroups); + } + + return path; +} + +/* + * postprocess_setop_rel - perform steps required after adding paths + */ +static void +postprocess_setop_rel(PlannerInfo *root, RelOptInfo *rel) +{ + /* + * We don't currently worry about allowing FDWs to contribute paths to + * this relation, but give extensions a chance. + */ + if (create_upper_paths_hook) + (*create_upper_paths_hook) (root, UPPERREL_SETOP, + NULL, rel, NULL); + + /* Select cheapest path */ + set_cheapest(rel); +} + +/* + * choose_hashed_setop - should we use hashing for a set operation? + */ +static bool +choose_hashed_setop(PlannerInfo *root, List *groupClauses, + Path *input_path, + double dNumGroups, double dNumOutputRows, + const char *construct) +{ + int numGroupCols = list_length(groupClauses); + Size hash_mem_limit = get_hash_memory_limit(); + bool can_sort; + bool can_hash; + Size hashentrysize; + Path hashed_p; + Path sorted_p; + double tuple_fraction; + + /* Check whether the operators support sorting or hashing */ + can_sort = grouping_is_sortable(groupClauses); + can_hash = grouping_is_hashable(groupClauses); + if (can_hash && can_sort) + { + /* we have a meaningful choice to make, continue ... */ + } + else if (can_hash) + return true; + else if (can_sort) + return false; + else + ereport(ERROR, + (errcode(ERRCODE_FEATURE_NOT_SUPPORTED), + /* translator: %s is UNION, INTERSECT, or EXCEPT */ + errmsg("could not implement %s", construct), + errdetail("Some of the datatypes only support hashing, while others only support sorting."))); + + /* Prefer sorting when enable_hashagg is off */ + if (!enable_hashagg) + return false; + + /* + * Don't do it if it doesn't look like the hashtable will fit into + * hash_mem. + */ + hashentrysize = MAXALIGN(input_path->pathtarget->width) + MAXALIGN(SizeofMinimalTupleHeader); + + if (hashentrysize * dNumGroups > hash_mem_limit) + return false; + + /* + * See if the estimated cost is no more than doing it the other way. + * + * We need to consider input_plan + hashagg versus input_plan + sort + + * group. Note that the actual result plan might involve a SetOp or + * Unique node, not Agg or Group, but the cost estimates for Agg and Group + * should be close enough for our purposes here. + * + * These path variables are dummies that just hold cost fields; we don't + * make actual Paths for these steps. + */ + cost_agg(&hashed_p, root, AGG_HASHED, NULL, + numGroupCols, dNumGroups, + NIL, + input_path->startup_cost, input_path->total_cost, + input_path->rows, input_path->pathtarget->width); + + /* + * Now for the sorted case. Note that the input is *always* unsorted, + * since it was made by appending unrelated sub-relations together. + */ + sorted_p.startup_cost = input_path->startup_cost; + sorted_p.total_cost = input_path->total_cost; + /* XXX cost_sort doesn't actually look at pathkeys, so just pass NIL */ + cost_sort(&sorted_p, root, NIL, sorted_p.total_cost, + input_path->rows, input_path->pathtarget->width, + 0.0, work_mem, -1.0); + cost_group(&sorted_p, root, numGroupCols, dNumGroups, + NIL, + sorted_p.startup_cost, sorted_p.total_cost, + input_path->rows); + + /* + * Now make the decision using the top-level tuple fraction. First we + * have to convert an absolute count (LIMIT) into fractional form. + */ + tuple_fraction = root->tuple_fraction; + if (tuple_fraction >= 1.0) + tuple_fraction /= dNumOutputRows; + + if (compare_fractional_path_costs(&hashed_p, &sorted_p, + tuple_fraction) < 0) + { + /* Hashed is cheaper, so use it */ + return true; + } + return false; +} + +/* + * Generate targetlist for a set-operation plan node + * + * colTypes: OID list of set-op's result column datatypes + * colCollations: OID list of set-op's result column collations + * flag: -1 if no flag column needed, 0 or 1 to create a const flag column + * varno: varno to use in generated Vars + * hack_constants: true to copy up constants (see comments in code) + * input_tlist: targetlist of this node's input node + * refnames_tlist: targetlist to take column names from + */ +static List * +generate_setop_tlist(List *colTypes, List *colCollations, + int flag, + Index varno, + bool hack_constants, + List *input_tlist, + List *refnames_tlist) +{ + List *tlist = NIL; + int resno = 1; + ListCell *ctlc, + *cclc, + *itlc, + *rtlc; + TargetEntry *tle; + Node *expr; + + forfour(ctlc, colTypes, cclc, colCollations, + itlc, input_tlist, rtlc, refnames_tlist) + { + Oid colType = lfirst_oid(ctlc); + Oid colColl = lfirst_oid(cclc); + TargetEntry *inputtle = (TargetEntry *) lfirst(itlc); + TargetEntry *reftle = (TargetEntry *) lfirst(rtlc); + + Assert(inputtle->resno == resno); + Assert(reftle->resno == resno); + Assert(!inputtle->resjunk); + Assert(!reftle->resjunk); + + /* + * Generate columns referencing input columns and having appropriate + * data types and column names. Insert datatype coercions where + * necessary. + * + * HACK: constants in the input's targetlist are copied up as-is + * rather than being referenced as subquery outputs. This is mainly + * to ensure that when we try to coerce them to the output column's + * datatype, the right things happen for UNKNOWN constants. But do + * this only at the first level of subquery-scan plans; we don't want + * phony constants appearing in the output tlists of upper-level + * nodes! + */ + if (hack_constants && inputtle->expr && IsA(inputtle->expr, Const)) + expr = (Node *) inputtle->expr; + else + expr = (Node *) makeVar(varno, + inputtle->resno, + exprType((Node *) inputtle->expr), + exprTypmod((Node *) inputtle->expr), + exprCollation((Node *) inputtle->expr), + 0); + + if (exprType(expr) != colType) + { + /* + * Note: it's not really cool to be applying coerce_to_common_type + * here; one notable point is that assign_expr_collations never + * gets run on any generated nodes. For the moment that's not a + * problem because we force the correct exposed collation below. + * It would likely be best to make the parser generate the correct + * output tlist for every set-op to begin with, though. + */ + expr = coerce_to_common_type(NULL, /* no UNKNOWNs here */ + expr, + colType, + "UNION/INTERSECT/EXCEPT"); + } + + /* + * Ensure the tlist entry's exposed collation matches the set-op. This + * is necessary because plan_set_operations() reports the result + * ordering as a list of SortGroupClauses, which don't carry collation + * themselves but just refer to tlist entries. If we don't show the + * right collation then planner.c might do the wrong thing in + * higher-level queries. + * + * Note we use RelabelType, not CollateExpr, since this expression + * will reach the executor without any further processing. + */ + if (exprCollation(expr) != colColl) + expr = applyRelabelType(expr, + exprType(expr), exprTypmod(expr), colColl, + COERCE_IMPLICIT_CAST, -1, false); + + tle = makeTargetEntry((Expr *) expr, + (AttrNumber) resno++, + pstrdup(reftle->resname), + false); + + /* + * By convention, all non-resjunk columns in a setop tree have + * ressortgroupref equal to their resno. In some cases the ref isn't + * needed, but this is a cleaner way than modifying the tlist later. + */ + tle->ressortgroupref = tle->resno; + + tlist = lappend(tlist, tle); + } + + if (flag >= 0) + { + /* Add a resjunk flag column */ + /* flag value is the given constant */ + expr = (Node *) makeConst(INT4OID, + -1, + InvalidOid, + sizeof(int32), + Int32GetDatum(flag), + false, + true); + tle = makeTargetEntry((Expr *) expr, + (AttrNumber) resno++, + pstrdup("flag"), + true); + tlist = lappend(tlist, tle); + } + + return tlist; +} + +/* + * Generate targetlist for a set-operation Append node + * + * colTypes: OID list of set-op's result column datatypes + * colCollations: OID list of set-op's result column collations + * flag: true to create a flag column copied up from subplans + * input_tlists: list of tlists for sub-plans of the Append + * refnames_tlist: targetlist to take column names from + * + * The entries in the Append's targetlist should always be simple Vars; + * we just have to make sure they have the right datatypes/typmods/collations. + * The Vars are always generated with varno 0. + * + * XXX a problem with the varno-zero approach is that set_pathtarget_cost_width + * cannot figure out a realistic width for the tlist we make here. But we + * ought to refactor this code to produce a PathTarget directly, anyway. + */ +static List * +generate_append_tlist(List *colTypes, List *colCollations, + bool flag, + List *input_tlists, + List *refnames_tlist) +{ + List *tlist = NIL; + int resno = 1; + ListCell *curColType; + ListCell *curColCollation; + ListCell *ref_tl_item; + int colindex; + TargetEntry *tle; + Node *expr; + ListCell *tlistl; + int32 *colTypmods; + + /* + * First extract typmods to use. + * + * If the inputs all agree on type and typmod of a particular column, use + * that typmod; else use -1. + */ + colTypmods = (int32 *) palloc(list_length(colTypes) * sizeof(int32)); + + foreach(tlistl, input_tlists) + { + List *subtlist = (List *) lfirst(tlistl); + ListCell *subtlistl; + + curColType = list_head(colTypes); + colindex = 0; + foreach(subtlistl, subtlist) + { + TargetEntry *subtle = (TargetEntry *) lfirst(subtlistl); + + if (subtle->resjunk) + continue; + Assert(curColType != NULL); + if (exprType((Node *) subtle->expr) == lfirst_oid(curColType)) + { + /* If first subplan, copy the typmod; else compare */ + int32 subtypmod = exprTypmod((Node *) subtle->expr); + + if (tlistl == list_head(input_tlists)) + colTypmods[colindex] = subtypmod; + else if (subtypmod != colTypmods[colindex]) + colTypmods[colindex] = -1; + } + else + { + /* types disagree, so force typmod to -1 */ + colTypmods[colindex] = -1; + } + curColType = lnext(colTypes, curColType); + colindex++; + } + Assert(curColType == NULL); + } + + /* + * Now we can build the tlist for the Append. + */ + colindex = 0; + forthree(curColType, colTypes, curColCollation, colCollations, + ref_tl_item, refnames_tlist) + { + Oid colType = lfirst_oid(curColType); + int32 colTypmod = colTypmods[colindex++]; + Oid colColl = lfirst_oid(curColCollation); + TargetEntry *reftle = (TargetEntry *) lfirst(ref_tl_item); + + Assert(reftle->resno == resno); + Assert(!reftle->resjunk); + expr = (Node *) makeVar(0, + resno, + colType, + colTypmod, + colColl, + 0); + tle = makeTargetEntry((Expr *) expr, + (AttrNumber) resno++, + pstrdup(reftle->resname), + false); + + /* + * By convention, all non-resjunk columns in a setop tree have + * ressortgroupref equal to their resno. In some cases the ref isn't + * needed, but this is a cleaner way than modifying the tlist later. + */ + tle->ressortgroupref = tle->resno; + + tlist = lappend(tlist, tle); + } + + if (flag) + { + /* Add a resjunk flag column */ + /* flag value is shown as copied up from subplan */ + expr = (Node *) makeVar(0, + resno, + INT4OID, + -1, + InvalidOid, + 0); + tle = makeTargetEntry((Expr *) expr, + (AttrNumber) resno++, + pstrdup("flag"), + true); + tlist = lappend(tlist, tle); + } + + pfree(colTypmods); + + return tlist; +} + +/* + * generate_setop_grouplist + * Build a SortGroupClause list defining the sort/grouping properties + * of the setop's output columns. + * + * Parse analysis already determined the properties and built a suitable + * list, except that the entries do not have sortgrouprefs set because + * the parser output representation doesn't include a tlist for each + * setop. So what we need to do here is copy that list and install + * proper sortgrouprefs into it (copying those from the targetlist). + */ +static List * +generate_setop_grouplist(SetOperationStmt *op, List *targetlist) +{ + List *grouplist = copyObject(op->groupClauses); + ListCell *lg; + ListCell *lt; + + lg = list_head(grouplist); + foreach(lt, targetlist) + { + TargetEntry *tle = (TargetEntry *) lfirst(lt); + SortGroupClause *sgc; + + if (tle->resjunk) + { + /* resjunk columns should not have sortgrouprefs */ + Assert(tle->ressortgroupref == 0); + continue; /* ignore resjunk columns */ + } + + /* non-resjunk columns should have sortgroupref = resno */ + Assert(tle->ressortgroupref == tle->resno); + + /* non-resjunk columns should have grouping clauses */ + Assert(lg != NULL); + sgc = (SortGroupClause *) lfirst(lg); + lg = lnext(grouplist, lg); + Assert(sgc->tleSortGroupRef == 0); + + sgc->tleSortGroupRef = tle->ressortgroupref; + } + Assert(lg == NULL); + return grouplist; +} diff --git a/src/backend/optimizer/util/Makefile b/src/backend/optimizer/util/Makefile new file mode 100644 index 0000000..4fb115c --- /dev/null +++ b/src/backend/optimizer/util/Makefile @@ -0,0 +1,31 @@ +#------------------------------------------------------------------------- +# +# Makefile-- +# Makefile for optimizer/util +# +# IDENTIFICATION +# src/backend/optimizer/util/Makefile +# +#------------------------------------------------------------------------- + +subdir = src/backend/optimizer/util +top_builddir = ../../../.. +include $(top_builddir)/src/Makefile.global + +OBJS = \ + appendinfo.o \ + clauses.o \ + inherit.o \ + joininfo.o \ + orclauses.o \ + paramassign.o \ + pathnode.o \ + placeholder.o \ + plancat.o \ + predtest.o \ + relnode.o \ + restrictinfo.o \ + tlist.o \ + var.o + +include $(top_srcdir)/src/backend/common.mk diff --git a/src/backend/optimizer/util/appendinfo.c b/src/backend/optimizer/util/appendinfo.c new file mode 100644 index 0000000..5c3d5a7 --- /dev/null +++ b/src/backend/optimizer/util/appendinfo.c @@ -0,0 +1,1011 @@ +/*------------------------------------------------------------------------- + * + * appendinfo.c + * Routines for mapping between append parent(s) and children + * + * Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group + * Portions Copyright (c) 1994, Regents of the University of California + * + * + * IDENTIFICATION + * src/backend/optimizer/util/appendinfo.c + * + *------------------------------------------------------------------------- + */ +#include "postgres.h" + +#include "access/htup_details.h" +#include "access/table.h" +#include "foreign/fdwapi.h" +#include "nodes/makefuncs.h" +#include "nodes/nodeFuncs.h" +#include "optimizer/appendinfo.h" +#include "optimizer/pathnode.h" +#include "optimizer/planmain.h" +#include "parser/parsetree.h" +#include "utils/lsyscache.h" +#include "utils/rel.h" +#include "utils/syscache.h" + + +typedef struct +{ + PlannerInfo *root; + int nappinfos; + AppendRelInfo **appinfos; +} adjust_appendrel_attrs_context; + +static void make_inh_translation_list(Relation oldrelation, + Relation newrelation, + Index newvarno, + AppendRelInfo *appinfo); +static Node *adjust_appendrel_attrs_mutator(Node *node, + adjust_appendrel_attrs_context *context); + + +/* + * make_append_rel_info + * Build an AppendRelInfo for the parent-child pair + */ +AppendRelInfo * +make_append_rel_info(Relation parentrel, Relation childrel, + Index parentRTindex, Index childRTindex) +{ + AppendRelInfo *appinfo = makeNode(AppendRelInfo); + + appinfo->parent_relid = parentRTindex; + appinfo->child_relid = childRTindex; + appinfo->parent_reltype = parentrel->rd_rel->reltype; + appinfo->child_reltype = childrel->rd_rel->reltype; + make_inh_translation_list(parentrel, childrel, childRTindex, appinfo); + appinfo->parent_reloid = RelationGetRelid(parentrel); + + return appinfo; +} + +/* + * make_inh_translation_list + * Build the list of translations from parent Vars to child Vars for + * an inheritance child, as well as a reverse-translation array. + * + * The reverse-translation array has an entry for each child relation + * column, which is either the 1-based index of the corresponding parent + * column, or 0 if there's no match (that happens for dropped child columns, + * as well as child columns beyond those of the parent, which are allowed in + * traditional inheritance though not partitioning). + * + * For paranoia's sake, we match type/collation as well as attribute name. + */ +static void +make_inh_translation_list(Relation oldrelation, Relation newrelation, + Index newvarno, + AppendRelInfo *appinfo) +{ + List *vars = NIL; + AttrNumber *pcolnos; + TupleDesc old_tupdesc = RelationGetDescr(oldrelation); + TupleDesc new_tupdesc = RelationGetDescr(newrelation); + Oid new_relid = RelationGetRelid(newrelation); + int oldnatts = old_tupdesc->natts; + int newnatts = new_tupdesc->natts; + int old_attno; + int new_attno = 0; + + /* Initialize reverse-translation array with all entries zero */ + appinfo->num_child_cols = newnatts; + appinfo->parent_colnos = pcolnos = + (AttrNumber *) palloc0(newnatts * sizeof(AttrNumber)); + + for (old_attno = 0; old_attno < oldnatts; old_attno++) + { + Form_pg_attribute att; + char *attname; + Oid atttypid; + int32 atttypmod; + Oid attcollation; + + att = TupleDescAttr(old_tupdesc, old_attno); + if (att->attisdropped) + { + /* Just put NULL into this list entry */ + vars = lappend(vars, NULL); + continue; + } + attname = NameStr(att->attname); + atttypid = att->atttypid; + atttypmod = att->atttypmod; + attcollation = att->attcollation; + + /* + * When we are generating the "translation list" for the parent table + * of an inheritance set, no need to search for matches. + */ + if (oldrelation == newrelation) + { + vars = lappend(vars, makeVar(newvarno, + (AttrNumber) (old_attno + 1), + atttypid, + atttypmod, + attcollation, + 0)); + pcolnos[old_attno] = old_attno + 1; + continue; + } + + /* + * Otherwise we have to search for the matching column by name. + * There's no guarantee it'll have the same column position, because + * of cases like ALTER TABLE ADD COLUMN and multiple inheritance. + * However, in simple cases, the relative order of columns is mostly + * the same in both relations, so try the column of newrelation that + * follows immediately after the one that we just found, and if that + * fails, let syscache handle it. + */ + if (new_attno >= newnatts || + (att = TupleDescAttr(new_tupdesc, new_attno))->attisdropped || + strcmp(attname, NameStr(att->attname)) != 0) + { + HeapTuple newtup; + + newtup = SearchSysCacheAttName(new_relid, attname); + if (!HeapTupleIsValid(newtup)) + elog(ERROR, "could not find inherited attribute \"%s\" of relation \"%s\"", + attname, RelationGetRelationName(newrelation)); + new_attno = ((Form_pg_attribute) GETSTRUCT(newtup))->attnum - 1; + Assert(new_attno >= 0 && new_attno < newnatts); + ReleaseSysCache(newtup); + + att = TupleDescAttr(new_tupdesc, new_attno); + } + + /* Found it, check type and collation match */ + if (atttypid != att->atttypid || atttypmod != att->atttypmod) + elog(ERROR, "attribute \"%s\" of relation \"%s\" does not match parent's type", + attname, RelationGetRelationName(newrelation)); + if (attcollation != att->attcollation) + elog(ERROR, "attribute \"%s\" of relation \"%s\" does not match parent's collation", + attname, RelationGetRelationName(newrelation)); + + vars = lappend(vars, makeVar(newvarno, + (AttrNumber) (new_attno + 1), + atttypid, + atttypmod, + attcollation, + 0)); + pcolnos[new_attno] = old_attno + 1; + new_attno++; + } + + appinfo->translated_vars = vars; +} + +/* + * adjust_appendrel_attrs + * Copy the specified query or expression and translate Vars referring to a + * parent rel to refer to the corresponding child rel instead. We also + * update rtindexes appearing outside Vars, such as resultRelation and + * jointree relids. + * + * Note: this is only applied after conversion of sublinks to subplans, + * so we don't need to cope with recursion into sub-queries. + * + * Note: this is not hugely different from what pullup_replace_vars() does; + * maybe we should try to fold the two routines together. + */ +Node * +adjust_appendrel_attrs(PlannerInfo *root, Node *node, int nappinfos, + AppendRelInfo **appinfos) +{ + adjust_appendrel_attrs_context context; + + context.root = root; + context.nappinfos = nappinfos; + context.appinfos = appinfos; + + /* If there's nothing to adjust, don't call this function. */ + Assert(nappinfos >= 1 && appinfos != NULL); + + /* Should never be translating a Query tree. */ + Assert(node == NULL || !IsA(node, Query)); + + return adjust_appendrel_attrs_mutator(node, &context); +} + +static Node * +adjust_appendrel_attrs_mutator(Node *node, + adjust_appendrel_attrs_context *context) +{ + AppendRelInfo **appinfos = context->appinfos; + int nappinfos = context->nappinfos; + int cnt; + + if (node == NULL) + return NULL; + if (IsA(node, Var)) + { + Var *var = (Var *) copyObject(node); + AppendRelInfo *appinfo = NULL; + + if (var->varlevelsup != 0) + return (Node *) var; /* no changes needed */ + + for (cnt = 0; cnt < nappinfos; cnt++) + { + if (var->varno == appinfos[cnt]->parent_relid) + { + appinfo = appinfos[cnt]; + break; + } + } + + if (appinfo) + { + var->varno = appinfo->child_relid; + /* it's now a generated Var, so drop any syntactic labeling */ + var->varnosyn = 0; + var->varattnosyn = 0; + if (var->varattno > 0) + { + Node *newnode; + + if (var->varattno > list_length(appinfo->translated_vars)) + elog(ERROR, "attribute %d of relation \"%s\" does not exist", + var->varattno, get_rel_name(appinfo->parent_reloid)); + newnode = copyObject(list_nth(appinfo->translated_vars, + var->varattno - 1)); + if (newnode == NULL) + elog(ERROR, "attribute %d of relation \"%s\" does not exist", + var->varattno, get_rel_name(appinfo->parent_reloid)); + return newnode; + } + else if (var->varattno == 0) + { + /* + * Whole-row Var: if we are dealing with named rowtypes, we + * can use a whole-row Var for the child table plus a coercion + * step to convert the tuple layout to the parent's rowtype. + * Otherwise we have to generate a RowExpr. + */ + if (OidIsValid(appinfo->child_reltype)) + { + Assert(var->vartype == appinfo->parent_reltype); + if (appinfo->parent_reltype != appinfo->child_reltype) + { + ConvertRowtypeExpr *r = makeNode(ConvertRowtypeExpr); + + r->arg = (Expr *) var; + r->resulttype = appinfo->parent_reltype; + r->convertformat = COERCE_IMPLICIT_CAST; + r->location = -1; + /* Make sure the Var node has the right type ID, too */ + var->vartype = appinfo->child_reltype; + return (Node *) r; + } + } + else + { + /* + * Build a RowExpr containing the translated variables. + * + * In practice var->vartype will always be RECORDOID here, + * so we need to come up with some suitable column names. + * We use the parent RTE's column names. + * + * Note: we can't get here for inheritance cases, so there + * is no need to worry that translated_vars might contain + * some dummy NULLs. + */ + RowExpr *rowexpr; + List *fields; + RangeTblEntry *rte; + + rte = rt_fetch(appinfo->parent_relid, + context->root->parse->rtable); + fields = copyObject(appinfo->translated_vars); + rowexpr = makeNode(RowExpr); + rowexpr->args = fields; + rowexpr->row_typeid = var->vartype; + rowexpr->row_format = COERCE_IMPLICIT_CAST; + rowexpr->colnames = copyObject(rte->eref->colnames); + rowexpr->location = -1; + + return (Node *) rowexpr; + } + } + /* system attributes don't need any other translation */ + } + else if (var->varno == ROWID_VAR) + { + /* + * If it's a ROWID_VAR placeholder, see if we've reached a leaf + * target rel, for which we can translate the Var to a specific + * instantiation. We should never be asked to translate to a set + * of relids containing more than one leaf target rel, so the + * answer will be unique. If we're still considering non-leaf + * inheritance levels, return the ROWID_VAR Var as-is. + */ + Relids leaf_result_relids = context->root->leaf_result_relids; + Index leaf_relid = 0; + + for (cnt = 0; cnt < nappinfos; cnt++) + { + if (bms_is_member(appinfos[cnt]->child_relid, + leaf_result_relids)) + { + if (leaf_relid) + elog(ERROR, "cannot translate to multiple leaf relids"); + leaf_relid = appinfos[cnt]->child_relid; + } + } + + if (leaf_relid) + { + RowIdentityVarInfo *ridinfo = (RowIdentityVarInfo *) + list_nth(context->root->row_identity_vars, var->varattno - 1); + + if (bms_is_member(leaf_relid, ridinfo->rowidrels)) + { + /* Substitute the Var given in the RowIdentityVarInfo */ + var = copyObject(ridinfo->rowidvar); + /* ... but use the correct relid */ + var->varno = leaf_relid; + /* varnosyn in the RowIdentityVarInfo is probably wrong */ + var->varnosyn = 0; + var->varattnosyn = 0; + } + else + { + /* + * This leaf rel can't return the desired value, so + * substitute a NULL of the correct type. + */ + return (Node *) makeNullConst(var->vartype, + var->vartypmod, + var->varcollid); + } + } + } + return (Node *) var; + } + if (IsA(node, CurrentOfExpr)) + { + CurrentOfExpr *cexpr = (CurrentOfExpr *) copyObject(node); + + for (cnt = 0; cnt < nappinfos; cnt++) + { + AppendRelInfo *appinfo = appinfos[cnt]; + + if (cexpr->cvarno == appinfo->parent_relid) + { + cexpr->cvarno = appinfo->child_relid; + break; + } + } + return (Node *) cexpr; + } + if (IsA(node, PlaceHolderVar)) + { + /* Copy the PlaceHolderVar node with correct mutation of subnodes */ + PlaceHolderVar *phv; + + phv = (PlaceHolderVar *) expression_tree_mutator(node, + adjust_appendrel_attrs_mutator, + (void *) context); + /* now fix PlaceHolderVar's relid sets */ + if (phv->phlevelsup == 0) + phv->phrels = adjust_child_relids(phv->phrels, context->nappinfos, + context->appinfos); + return (Node *) phv; + } + /* Shouldn't need to handle planner auxiliary nodes here */ + Assert(!IsA(node, SpecialJoinInfo)); + Assert(!IsA(node, AppendRelInfo)); + Assert(!IsA(node, PlaceHolderInfo)); + Assert(!IsA(node, MinMaxAggInfo)); + + /* + * We have to process RestrictInfo nodes specially. (Note: although + * set_append_rel_pathlist will hide RestrictInfos in the parent's + * baserestrictinfo list from us, it doesn't hide those in joininfo.) + */ + if (IsA(node, RestrictInfo)) + { + RestrictInfo *oldinfo = (RestrictInfo *) node; + RestrictInfo *newinfo = makeNode(RestrictInfo); + + /* Copy all flat-copiable fields */ + memcpy(newinfo, oldinfo, sizeof(RestrictInfo)); + + /* Recursively fix the clause itself */ + newinfo->clause = (Expr *) + adjust_appendrel_attrs_mutator((Node *) oldinfo->clause, context); + + /* and the modified version, if an OR clause */ + newinfo->orclause = (Expr *) + adjust_appendrel_attrs_mutator((Node *) oldinfo->orclause, context); + + /* adjust relid sets too */ + newinfo->clause_relids = adjust_child_relids(oldinfo->clause_relids, + context->nappinfos, + context->appinfos); + newinfo->required_relids = adjust_child_relids(oldinfo->required_relids, + context->nappinfos, + context->appinfos); + newinfo->outer_relids = adjust_child_relids(oldinfo->outer_relids, + context->nappinfos, + context->appinfos); + newinfo->nullable_relids = adjust_child_relids(oldinfo->nullable_relids, + context->nappinfos, + context->appinfos); + newinfo->left_relids = adjust_child_relids(oldinfo->left_relids, + context->nappinfos, + context->appinfos); + newinfo->right_relids = adjust_child_relids(oldinfo->right_relids, + context->nappinfos, + context->appinfos); + + /* + * Reset cached derivative fields, since these might need to have + * different values when considering the child relation. Note we + * don't reset left_ec/right_ec: each child variable is implicitly + * equivalent to its parent, so still a member of the same EC if any. + */ + newinfo->eval_cost.startup = -1; + newinfo->norm_selec = -1; + newinfo->outer_selec = -1; + newinfo->left_em = NULL; + newinfo->right_em = NULL; + newinfo->scansel_cache = NIL; + newinfo->left_bucketsize = -1; + newinfo->right_bucketsize = -1; + newinfo->left_mcvfreq = -1; + newinfo->right_mcvfreq = -1; + + return (Node *) newinfo; + } + + /* + * NOTE: we do not need to recurse into sublinks, because they should + * already have been converted to subplans before we see them. + */ + Assert(!IsA(node, SubLink)); + Assert(!IsA(node, Query)); + /* We should never see these Query substructures, either. */ + Assert(!IsA(node, RangeTblRef)); + Assert(!IsA(node, JoinExpr)); + + return expression_tree_mutator(node, adjust_appendrel_attrs_mutator, + (void *) context); +} + +/* + * adjust_appendrel_attrs_multilevel + * Apply Var translations from a toplevel appendrel parent down to a child. + * + * In some cases we need to translate expressions referencing a parent relation + * to reference an appendrel child that's multiple levels removed from it. + */ +Node * +adjust_appendrel_attrs_multilevel(PlannerInfo *root, Node *node, + Relids child_relids, + Relids top_parent_relids) +{ + AppendRelInfo **appinfos; + Bitmapset *parent_relids = NULL; + int nappinfos; + int cnt; + + Assert(bms_num_members(child_relids) == bms_num_members(top_parent_relids)); + + appinfos = find_appinfos_by_relids(root, child_relids, &nappinfos); + + /* Construct relids set for the immediate parent of given child. */ + for (cnt = 0; cnt < nappinfos; cnt++) + { + AppendRelInfo *appinfo = appinfos[cnt]; + + parent_relids = bms_add_member(parent_relids, appinfo->parent_relid); + } + + /* Recurse if immediate parent is not the top parent. */ + if (!bms_equal(parent_relids, top_parent_relids)) + node = adjust_appendrel_attrs_multilevel(root, node, parent_relids, + top_parent_relids); + + /* Now translate for this child */ + node = adjust_appendrel_attrs(root, node, nappinfos, appinfos); + + pfree(appinfos); + + return node; +} + +/* + * Substitute child relids for parent relids in a Relid set. The array of + * appinfos specifies the substitutions to be performed. + */ +Relids +adjust_child_relids(Relids relids, int nappinfos, AppendRelInfo **appinfos) +{ + Bitmapset *result = NULL; + int cnt; + + for (cnt = 0; cnt < nappinfos; cnt++) + { + AppendRelInfo *appinfo = appinfos[cnt]; + + /* Remove parent, add child */ + if (bms_is_member(appinfo->parent_relid, relids)) + { + /* Make a copy if we are changing the set. */ + if (!result) + result = bms_copy(relids); + + result = bms_del_member(result, appinfo->parent_relid); + result = bms_add_member(result, appinfo->child_relid); + } + } + + /* If we made any changes, return the modified copy. */ + if (result) + return result; + + /* Otherwise, return the original set without modification. */ + return relids; +} + +/* + * Replace any relid present in top_parent_relids with its child in + * child_relids. Members of child_relids can be multiple levels below top + * parent in the partition hierarchy. + */ +Relids +adjust_child_relids_multilevel(PlannerInfo *root, Relids relids, + Relids child_relids, Relids top_parent_relids) +{ + AppendRelInfo **appinfos; + int nappinfos; + Relids parent_relids = NULL; + Relids result; + Relids tmp_result = NULL; + int cnt; + + /* + * If the given relids set doesn't contain any of the top parent relids, + * it will remain unchanged. + */ + if (!bms_overlap(relids, top_parent_relids)) + return relids; + + appinfos = find_appinfos_by_relids(root, child_relids, &nappinfos); + + /* Construct relids set for the immediate parent of the given child. */ + for (cnt = 0; cnt < nappinfos; cnt++) + { + AppendRelInfo *appinfo = appinfos[cnt]; + + parent_relids = bms_add_member(parent_relids, appinfo->parent_relid); + } + + /* Recurse if immediate parent is not the top parent. */ + if (!bms_equal(parent_relids, top_parent_relids)) + { + tmp_result = adjust_child_relids_multilevel(root, relids, + parent_relids, + top_parent_relids); + relids = tmp_result; + } + + result = adjust_child_relids(relids, nappinfos, appinfos); + + /* Free memory consumed by any intermediate result. */ + if (tmp_result) + bms_free(tmp_result); + bms_free(parent_relids); + pfree(appinfos); + + return result; +} + +/* + * adjust_inherited_attnums + * Translate an integer list of attribute numbers from parent to child. + */ +List * +adjust_inherited_attnums(List *attnums, AppendRelInfo *context) +{ + List *result = NIL; + ListCell *lc; + + /* This should only happen for an inheritance case, not UNION ALL */ + Assert(OidIsValid(context->parent_reloid)); + + /* Look up each attribute in the AppendRelInfo's translated_vars list */ + foreach(lc, attnums) + { + AttrNumber parentattno = lfirst_int(lc); + Var *childvar; + + /* Look up the translation of this column: it must be a Var */ + if (parentattno <= 0 || + parentattno > list_length(context->translated_vars)) + elog(ERROR, "attribute %d of relation \"%s\" does not exist", + parentattno, get_rel_name(context->parent_reloid)); + childvar = (Var *) list_nth(context->translated_vars, parentattno - 1); + if (childvar == NULL || !IsA(childvar, Var)) + elog(ERROR, "attribute %d of relation \"%s\" does not exist", + parentattno, get_rel_name(context->parent_reloid)); + + result = lappend_int(result, childvar->varattno); + } + return result; +} + +/* + * adjust_inherited_attnums_multilevel + * As above, but traverse multiple inheritance levels as needed. + */ +List * +adjust_inherited_attnums_multilevel(PlannerInfo *root, List *attnums, + Index child_relid, Index top_parent_relid) +{ + AppendRelInfo *appinfo = root->append_rel_array[child_relid]; + + if (!appinfo) + elog(ERROR, "child rel %d not found in append_rel_array", child_relid); + + /* Recurse if immediate parent is not the top parent. */ + if (appinfo->parent_relid != top_parent_relid) + attnums = adjust_inherited_attnums_multilevel(root, attnums, + appinfo->parent_relid, + top_parent_relid); + + /* Now translate for this child */ + return adjust_inherited_attnums(attnums, appinfo); +} + +/* + * get_translated_update_targetlist + * Get the processed_tlist of an UPDATE query, translated as needed to + * match a child target relation. + * + * Optionally also return the list of target column numbers translated + * to this target relation. (The resnos in processed_tlist MUST NOT be + * relied on for this purpose.) + */ +void +get_translated_update_targetlist(PlannerInfo *root, Index relid, + List **processed_tlist, List **update_colnos) +{ + /* This is pretty meaningless for commands other than UPDATE. */ + Assert(root->parse->commandType == CMD_UPDATE); + if (relid == root->parse->resultRelation) + { + /* + * Non-inheritance case, so it's easy. The caller might be expecting + * a tree it can scribble on, though, so copy. + */ + *processed_tlist = copyObject(root->processed_tlist); + if (update_colnos) + *update_colnos = copyObject(root->update_colnos); + } + else + { + Assert(bms_is_member(relid, root->all_result_relids)); + *processed_tlist = (List *) + adjust_appendrel_attrs_multilevel(root, + (Node *) root->processed_tlist, + bms_make_singleton(relid), + bms_make_singleton(root->parse->resultRelation)); + if (update_colnos) + *update_colnos = + adjust_inherited_attnums_multilevel(root, root->update_colnos, + relid, + root->parse->resultRelation); + } +} + +/* + * find_appinfos_by_relids + * Find AppendRelInfo structures for all relations specified by relids. + * + * The AppendRelInfos are returned in an array, which can be pfree'd by the + * caller. *nappinfos is set to the number of entries in the array. + */ +AppendRelInfo ** +find_appinfos_by_relids(PlannerInfo *root, Relids relids, int *nappinfos) +{ + AppendRelInfo **appinfos; + int cnt = 0; + int i; + + *nappinfos = bms_num_members(relids); + appinfos = (AppendRelInfo **) palloc(sizeof(AppendRelInfo *) * *nappinfos); + + i = -1; + while ((i = bms_next_member(relids, i)) >= 0) + { + AppendRelInfo *appinfo = root->append_rel_array[i]; + + if (!appinfo) + elog(ERROR, "child rel %d not found in append_rel_array", i); + + appinfos[cnt++] = appinfo; + } + return appinfos; +} + + +/***************************************************************************** + * + * ROW-IDENTITY VARIABLE MANAGEMENT + * + * This code lacks a good home, perhaps. We choose to keep it here because + * adjust_appendrel_attrs_mutator() is its principal co-conspirator. That + * function does most of what is needed to expand ROWID_VAR Vars into the + * right things. + * + *****************************************************************************/ + +/* + * add_row_identity_var + * Register a row-identity column to be used in UPDATE/DELETE/MERGE. + * + * The Var must be equal(), aside from varno, to any other row-identity + * column with the same rowid_name. Thus, for example, "wholerow" + * row identities had better use vartype == RECORDOID. + * + * rtindex is currently redundant with rowid_var->varno, but we specify + * it as a separate parameter in case this is ever generalized to support + * non-Var expressions. (We could reasonably handle expressions over + * Vars of the specified rtindex, but for now that seems unnecessary.) + */ +void +add_row_identity_var(PlannerInfo *root, Var *orig_var, + Index rtindex, const char *rowid_name) +{ + TargetEntry *tle; + Var *rowid_var; + RowIdentityVarInfo *ridinfo; + ListCell *lc; + + /* For now, the argument must be just a Var of the given rtindex */ + Assert(IsA(orig_var, Var)); + Assert(orig_var->varno == rtindex); + Assert(orig_var->varlevelsup == 0); + + /* + * If we're doing non-inherited UPDATE/DELETE/MERGE, there's little need + * for ROWID_VAR shenanigans. Just shove the presented Var into the + * processed_tlist, and we're done. + */ + if (rtindex == root->parse->resultRelation) + { + tle = makeTargetEntry((Expr *) orig_var, + list_length(root->processed_tlist) + 1, + pstrdup(rowid_name), + true); + root->processed_tlist = lappend(root->processed_tlist, tle); + return; + } + + /* + * Otherwise, rtindex should reference a leaf target relation that's being + * added to the query during expand_inherited_rtentry(). + */ + Assert(bms_is_member(rtindex, root->leaf_result_relids)); + Assert(root->append_rel_array[rtindex] != NULL); + + /* + * We have to find a matching RowIdentityVarInfo, or make one if there is + * none. To allow using equal() to match the vars, change the varno to + * ROWID_VAR, leaving all else alone. + */ + rowid_var = copyObject(orig_var); + /* This could eventually become ChangeVarNodes() */ + rowid_var->varno = ROWID_VAR; + + /* Look for an existing row-id column of the same name */ + foreach(lc, root->row_identity_vars) + { + ridinfo = (RowIdentityVarInfo *) lfirst(lc); + if (strcmp(rowid_name, ridinfo->rowidname) != 0) + continue; + if (equal(rowid_var, ridinfo->rowidvar)) + { + /* Found a match; we need only record that rtindex needs it too */ + ridinfo->rowidrels = bms_add_member(ridinfo->rowidrels, rtindex); + return; + } + else + { + /* Ooops, can't handle this */ + elog(ERROR, "conflicting uses of row-identity name \"%s\"", + rowid_name); + } + } + + /* No request yet, so add a new RowIdentityVarInfo */ + ridinfo = makeNode(RowIdentityVarInfo); + ridinfo->rowidvar = copyObject(rowid_var); + /* for the moment, estimate width using just the datatype info */ + ridinfo->rowidwidth = get_typavgwidth(exprType((Node *) rowid_var), + exprTypmod((Node *) rowid_var)); + ridinfo->rowidname = pstrdup(rowid_name); + ridinfo->rowidrels = bms_make_singleton(rtindex); + + root->row_identity_vars = lappend(root->row_identity_vars, ridinfo); + + /* Change rowid_var into a reference to this row_identity_vars entry */ + rowid_var->varattno = list_length(root->row_identity_vars); + + /* Push the ROWID_VAR reference variable into processed_tlist */ + tle = makeTargetEntry((Expr *) rowid_var, + list_length(root->processed_tlist) + 1, + pstrdup(rowid_name), + true); + root->processed_tlist = lappend(root->processed_tlist, tle); +} + +/* + * add_row_identity_columns + * + * This function adds the row identity columns needed by the core code. + * FDWs might call add_row_identity_var() for themselves to add nonstandard + * columns. (Duplicate requests are fine.) + */ +void +add_row_identity_columns(PlannerInfo *root, Index rtindex, + RangeTblEntry *target_rte, + Relation target_relation) +{ + CmdType commandType = root->parse->commandType; + char relkind = target_relation->rd_rel->relkind; + Var *var; + + Assert(commandType == CMD_UPDATE || commandType == CMD_DELETE || commandType == CMD_MERGE); + + if (commandType == CMD_MERGE || + relkind == RELKIND_RELATION || + relkind == RELKIND_MATVIEW || + relkind == RELKIND_PARTITIONED_TABLE) + { + /* + * Emit CTID so that executor can find the row to merge, update or + * delete. + */ + var = makeVar(rtindex, + SelfItemPointerAttributeNumber, + TIDOID, + -1, + InvalidOid, + 0); + add_row_identity_var(root, var, rtindex, "ctid"); + } + else if (relkind == RELKIND_FOREIGN_TABLE) + { + /* + * Let the foreign table's FDW add whatever junk TLEs it wants. + */ + FdwRoutine *fdwroutine; + + fdwroutine = GetFdwRoutineForRelation(target_relation, false); + + if (fdwroutine->AddForeignUpdateTargets != NULL) + fdwroutine->AddForeignUpdateTargets(root, rtindex, + target_rte, target_relation); + + /* + * For UPDATE, we need to make the FDW fetch unchanged columns by + * asking it to fetch a whole-row Var. That's because the top-level + * targetlist only contains entries for changed columns, but + * ExecUpdate will need to build the complete new tuple. (Actually, + * we only really need this in UPDATEs that are not pushed to the + * remote side, but it's hard to tell if that will be the case at the + * point when this function is called.) + * + * We will also need the whole row if there are any row triggers, so + * that the executor will have the "old" row to pass to the trigger. + * Alas, this misses system columns. + */ + if (commandType == CMD_UPDATE || + (target_relation->trigdesc && + (target_relation->trigdesc->trig_delete_after_row || + target_relation->trigdesc->trig_delete_before_row))) + { + var = makeVar(rtindex, + InvalidAttrNumber, + RECORDOID, + -1, + InvalidOid, + 0); + add_row_identity_var(root, var, rtindex, "wholerow"); + } + } +} + +/* + * distribute_row_identity_vars + * + * After we have finished identifying all the row identity columns + * needed by an inherited UPDATE/DELETE/MERGE query, make sure that + * these columns will be generated by all the target relations. + * + * This is more or less like what build_base_rel_tlists() does, + * except that it would not understand what to do with ROWID_VAR Vars. + * Since that function runs before inheritance relations are expanded, + * it will never see any such Vars anyway. + */ +void +distribute_row_identity_vars(PlannerInfo *root) +{ + Query *parse = root->parse; + int result_relation = parse->resultRelation; + RangeTblEntry *target_rte; + RelOptInfo *target_rel; + ListCell *lc; + + /* + * There's nothing to do if this isn't an inherited UPDATE/DELETE/MERGE. + */ + if (parse->commandType != CMD_UPDATE && parse->commandType != CMD_DELETE && + parse->commandType != CMD_MERGE) + { + Assert(root->row_identity_vars == NIL); + return; + } + target_rte = rt_fetch(result_relation, parse->rtable); + if (!target_rte->inh) + { + Assert(root->row_identity_vars == NIL); + return; + } + + /* + * Ordinarily, we expect that leaf result relation(s) will have added some + * ROWID_VAR Vars to the query. However, it's possible that constraint + * exclusion suppressed every leaf relation. The executor will get upset + * if the plan has no row identity columns at all, even though it will + * certainly process no rows. Handle this edge case by re-opening the top + * result relation and adding the row identity columns it would have used, + * as preprocess_targetlist() would have done if it weren't marked "inh". + * Then re-run build_base_rel_tlists() to ensure that the added columns + * get propagated to the relation's reltarget. (This is a bit ugly, but + * it seems better to confine the ugliness and extra cycles to this + * unusual corner case.) + */ + if (root->row_identity_vars == NIL) + { + Relation target_relation; + + target_relation = table_open(target_rte->relid, NoLock); + add_row_identity_columns(root, result_relation, + target_rte, target_relation); + table_close(target_relation, NoLock); + build_base_rel_tlists(root, root->processed_tlist); + /* There are no ROWID_VAR Vars in this case, so we're done. */ + return; + } + + /* + * Dig through the processed_tlist to find the ROWID_VAR reference Vars, + * and forcibly copy them into the reltarget list of the topmost target + * relation. That's sufficient because they'll be copied to the + * individual leaf target rels (with appropriate translation) later, + * during appendrel expansion --- see set_append_rel_size(). + */ + target_rel = find_base_rel(root, result_relation); + + foreach(lc, root->processed_tlist) + { + TargetEntry *tle = lfirst(lc); + Var *var = (Var *) tle->expr; + + if (var && IsA(var, Var) && var->varno == ROWID_VAR) + { + target_rel->reltarget->exprs = + lappend(target_rel->reltarget->exprs, copyObject(var)); + /* reltarget cost and width will be computed later */ + } + } +} diff --git a/src/backend/optimizer/util/clauses.c b/src/backend/optimizer/util/clauses.c new file mode 100644 index 0000000..f2216f5 --- /dev/null +++ b/src/backend/optimizer/util/clauses.c @@ -0,0 +1,5255 @@ +/*------------------------------------------------------------------------- + * + * clauses.c + * routines to manipulate qualification clauses + * + * Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group + * Portions Copyright (c) 1994, Regents of the University of California + * + * + * IDENTIFICATION + * src/backend/optimizer/util/clauses.c + * + * HISTORY + * AUTHOR DATE MAJOR EVENT + * Andrew Yu Nov 3, 1994 clause.c and clauses.c combined + * + *------------------------------------------------------------------------- + */ + +#include "postgres.h" + +#include "access/htup_details.h" +#include "catalog/pg_aggregate.h" +#include "catalog/pg_class.h" +#include "catalog/pg_language.h" +#include "catalog/pg_operator.h" +#include "catalog/pg_proc.h" +#include "catalog/pg_type.h" +#include "executor/executor.h" +#include "executor/functions.h" +#include "funcapi.h" +#include "miscadmin.h" +#include "nodes/makefuncs.h" +#include "nodes/nodeFuncs.h" +#include "nodes/subscripting.h" +#include "nodes/supportnodes.h" +#include "optimizer/clauses.h" +#include "optimizer/cost.h" +#include "optimizer/optimizer.h" +#include "optimizer/plancat.h" +#include "optimizer/planmain.h" +#include "parser/analyze.h" +#include "parser/parse_agg.h" +#include "parser/parse_coerce.h" +#include "parser/parse_func.h" +#include "rewrite/rewriteHandler.h" +#include "rewrite/rewriteManip.h" +#include "tcop/tcopprot.h" +#include "utils/acl.h" +#include "utils/builtins.h" +#include "utils/datum.h" +#include "utils/fmgroids.h" +#include "utils/lsyscache.h" +#include "utils/memutils.h" +#include "utils/syscache.h" +#include "utils/typcache.h" + +typedef struct +{ + ParamListInfo boundParams; + PlannerInfo *root; + List *active_fns; + Node *case_val; + bool estimate; +} eval_const_expressions_context; + +typedef struct +{ + int nargs; + List *args; + int *usecounts; +} substitute_actual_parameters_context; + +typedef struct +{ + int nargs; + List *args; + int sublevels_up; +} substitute_actual_srf_parameters_context; + +typedef struct +{ + char *proname; + char *prosrc; +} inline_error_callback_arg; + +typedef struct +{ + char max_hazard; /* worst proparallel hazard found so far */ + char max_interesting; /* worst proparallel hazard of interest */ + List *safe_param_ids; /* PARAM_EXEC Param IDs to treat as safe */ +} max_parallel_hazard_context; + +static bool contain_agg_clause_walker(Node *node, void *context); +static bool find_window_functions_walker(Node *node, WindowFuncLists *lists); +static bool contain_subplans_walker(Node *node, void *context); +static bool contain_mutable_functions_walker(Node *node, void *context); +static bool contain_volatile_functions_walker(Node *node, void *context); +static bool contain_volatile_functions_not_nextval_walker(Node *node, void *context); +static bool max_parallel_hazard_walker(Node *node, + max_parallel_hazard_context *context); +static bool contain_nonstrict_functions_walker(Node *node, void *context); +static bool contain_exec_param_walker(Node *node, List *param_ids); +static bool contain_context_dependent_node(Node *clause); +static bool contain_context_dependent_node_walker(Node *node, int *flags); +static bool contain_leaked_vars_walker(Node *node, void *context); +static Relids find_nonnullable_rels_walker(Node *node, bool top_level); +static List *find_nonnullable_vars_walker(Node *node, bool top_level); +static bool is_strict_saop(ScalarArrayOpExpr *expr, bool falseOK); +static bool convert_saop_to_hashed_saop_walker(Node *node, void *context); +static Node *eval_const_expressions_mutator(Node *node, + eval_const_expressions_context *context); +static bool contain_non_const_walker(Node *node, void *context); +static bool ece_function_is_safe(Oid funcid, + eval_const_expressions_context *context); +static List *simplify_or_arguments(List *args, + eval_const_expressions_context *context, + bool *haveNull, bool *forceTrue); +static List *simplify_and_arguments(List *args, + eval_const_expressions_context *context, + bool *haveNull, bool *forceFalse); +static Node *simplify_boolean_equality(Oid opno, List *args); +static Expr *simplify_function(Oid funcid, + Oid result_type, int32 result_typmod, + Oid result_collid, Oid input_collid, List **args_p, + bool funcvariadic, bool process_args, bool allow_non_const, + eval_const_expressions_context *context); +static List *reorder_function_arguments(List *args, int pronargs, + HeapTuple func_tuple); +static List *add_function_defaults(List *args, int pronargs, + HeapTuple func_tuple); +static List *fetch_function_defaults(HeapTuple func_tuple); +static void recheck_cast_function_args(List *args, Oid result_type, + Oid *proargtypes, int pronargs, + HeapTuple func_tuple); +static Expr *evaluate_function(Oid funcid, Oid result_type, int32 result_typmod, + Oid result_collid, Oid input_collid, List *args, + bool funcvariadic, + HeapTuple func_tuple, + eval_const_expressions_context *context); +static Expr *inline_function(Oid funcid, Oid result_type, Oid result_collid, + Oid input_collid, List *args, + bool funcvariadic, + HeapTuple func_tuple, + eval_const_expressions_context *context); +static Node *substitute_actual_parameters(Node *expr, int nargs, List *args, + int *usecounts); +static Node *substitute_actual_parameters_mutator(Node *node, + substitute_actual_parameters_context *context); +static void sql_inline_error_callback(void *arg); +static Query *substitute_actual_srf_parameters(Query *expr, + int nargs, List *args); +static Node *substitute_actual_srf_parameters_mutator(Node *node, + substitute_actual_srf_parameters_context *context); +static bool pull_paramids_walker(Node *node, Bitmapset **context); + + +/***************************************************************************** + * Aggregate-function clause manipulation + *****************************************************************************/ + +/* + * contain_agg_clause + * Recursively search for Aggref/GroupingFunc nodes within a clause. + * + * Returns true if any aggregate found. + * + * This does not descend into subqueries, and so should be used only after + * reduction of sublinks to subplans, or in contexts where it's known there + * are no subqueries. There mustn't be outer-aggregate references either. + * + * (If you want something like this but able to deal with subqueries, + * see rewriteManip.c's contain_aggs_of_level().) + */ +bool +contain_agg_clause(Node *clause) +{ + return contain_agg_clause_walker(clause, NULL); +} + +static bool +contain_agg_clause_walker(Node *node, void *context) +{ + if (node == NULL) + return false; + if (IsA(node, Aggref)) + { + Assert(((Aggref *) node)->agglevelsup == 0); + return true; /* abort the tree traversal and return true */ + } + if (IsA(node, GroupingFunc)) + { + Assert(((GroupingFunc *) node)->agglevelsup == 0); + return true; /* abort the tree traversal and return true */ + } + Assert(!IsA(node, SubLink)); + return expression_tree_walker(node, contain_agg_clause_walker, context); +} + +/***************************************************************************** + * Window-function clause manipulation + *****************************************************************************/ + +/* + * contain_window_function + * Recursively search for WindowFunc nodes within a clause. + * + * Since window functions don't have level fields, but are hard-wired to + * be associated with the current query level, this is just the same as + * rewriteManip.c's function. + */ +bool +contain_window_function(Node *clause) +{ + return contain_windowfuncs(clause); +} + +/* + * find_window_functions + * Locate all the WindowFunc nodes in an expression tree, and organize + * them by winref ID number. + * + * Caller must provide an upper bound on the winref IDs expected in the tree. + */ +WindowFuncLists * +find_window_functions(Node *clause, Index maxWinRef) +{ + WindowFuncLists *lists = palloc(sizeof(WindowFuncLists)); + + lists->numWindowFuncs = 0; + lists->maxWinRef = maxWinRef; + lists->windowFuncs = (List **) palloc0((maxWinRef + 1) * sizeof(List *)); + (void) find_window_functions_walker(clause, lists); + return lists; +} + +static bool +find_window_functions_walker(Node *node, WindowFuncLists *lists) +{ + if (node == NULL) + return false; + if (IsA(node, WindowFunc)) + { + WindowFunc *wfunc = (WindowFunc *) node; + + /* winref is unsigned, so one-sided test is OK */ + if (wfunc->winref > lists->maxWinRef) + elog(ERROR, "WindowFunc contains out-of-range winref %u", + wfunc->winref); + /* eliminate duplicates, so that we avoid repeated computation */ + if (!list_member(lists->windowFuncs[wfunc->winref], wfunc)) + { + lists->windowFuncs[wfunc->winref] = + lappend(lists->windowFuncs[wfunc->winref], wfunc); + lists->numWindowFuncs++; + } + + /* + * We assume that the parser checked that there are no window + * functions in the arguments or filter clause. Hence, we need not + * recurse into them. (If either the parser or the planner screws up + * on this point, the executor will still catch it; see ExecInitExpr.) + */ + return false; + } + Assert(!IsA(node, SubLink)); + return expression_tree_walker(node, find_window_functions_walker, + (void *) lists); +} + + +/***************************************************************************** + * Support for expressions returning sets + *****************************************************************************/ + +/* + * expression_returns_set_rows + * Estimate the number of rows returned by a set-returning expression. + * The result is 1 if it's not a set-returning expression. + * + * We should only examine the top-level function or operator; it used to be + * appropriate to recurse, but not anymore. (Even if there are more SRFs in + * the function's inputs, their multipliers are accounted for separately.) + * + * Note: keep this in sync with expression_returns_set() in nodes/nodeFuncs.c. + */ +double +expression_returns_set_rows(PlannerInfo *root, Node *clause) +{ + if (clause == NULL) + return 1.0; + if (IsA(clause, FuncExpr)) + { + FuncExpr *expr = (FuncExpr *) clause; + + if (expr->funcretset) + return clamp_row_est(get_function_rows(root, expr->funcid, clause)); + } + if (IsA(clause, OpExpr)) + { + OpExpr *expr = (OpExpr *) clause; + + if (expr->opretset) + { + set_opfuncid(expr); + return clamp_row_est(get_function_rows(root, expr->opfuncid, clause)); + } + } + return 1.0; +} + + +/***************************************************************************** + * Subplan clause manipulation + *****************************************************************************/ + +/* + * contain_subplans + * Recursively search for subplan nodes within a clause. + * + * If we see a SubLink node, we will return true. This is only possible if + * the expression tree hasn't yet been transformed by subselect.c. We do not + * know whether the node will produce a true subplan or just an initplan, + * but we make the conservative assumption that it will be a subplan. + * + * Returns true if any subplan found. + */ +bool +contain_subplans(Node *clause) +{ + return contain_subplans_walker(clause, NULL); +} + +static bool +contain_subplans_walker(Node *node, void *context) +{ + if (node == NULL) + return false; + if (IsA(node, SubPlan) || + IsA(node, AlternativeSubPlan) || + IsA(node, SubLink)) + return true; /* abort the tree traversal and return true */ + return expression_tree_walker(node, contain_subplans_walker, context); +} + + +/***************************************************************************** + * Check clauses for mutable functions + *****************************************************************************/ + +/* + * contain_mutable_functions + * Recursively search for mutable functions within a clause. + * + * Returns true if any mutable function (or operator implemented by a + * mutable function) is found. This test is needed so that we don't + * mistakenly think that something like "WHERE random() < 0.5" can be treated + * as a constant qualification. + * + * We will recursively look into Query nodes (i.e., SubLink sub-selects) + * but not into SubPlans. See comments for contain_volatile_functions(). + */ +bool +contain_mutable_functions(Node *clause) +{ + return contain_mutable_functions_walker(clause, NULL); +} + +static bool +contain_mutable_functions_checker(Oid func_id, void *context) +{ + return (func_volatile(func_id) != PROVOLATILE_IMMUTABLE); +} + +static bool +contain_mutable_functions_walker(Node *node, void *context) +{ + if (node == NULL) + return false; + /* Check for mutable functions in node itself */ + if (check_functions_in_node(node, contain_mutable_functions_checker, + context)) + return true; + + if (IsA(node, SQLValueFunction)) + { + /* all variants of SQLValueFunction are stable */ + return true; + } + + if (IsA(node, NextValueExpr)) + { + /* NextValueExpr is volatile */ + return true; + } + + /* + * It should be safe to treat MinMaxExpr as immutable, because it will + * depend on a non-cross-type btree comparison function, and those should + * always be immutable. Treating XmlExpr as immutable is more dubious, + * and treating CoerceToDomain as immutable is outright dangerous. But we + * have done so historically, and changing this would probably cause more + * problems than it would fix. In practice, if you have a non-immutable + * domain constraint you are in for pain anyhow. + */ + + /* Recurse to check arguments */ + if (IsA(node, Query)) + { + /* Recurse into subselects */ + return query_tree_walker((Query *) node, + contain_mutable_functions_walker, + context, 0); + } + return expression_tree_walker(node, contain_mutable_functions_walker, + context); +} + + +/***************************************************************************** + * Check clauses for volatile functions + *****************************************************************************/ + +/* + * contain_volatile_functions + * Recursively search for volatile functions within a clause. + * + * Returns true if any volatile function (or operator implemented by a + * volatile function) is found. This test prevents, for example, + * invalid conversions of volatile expressions into indexscan quals. + * + * We will recursively look into Query nodes (i.e., SubLink sub-selects) + * but not into SubPlans. This is a bit odd, but intentional. If we are + * looking at a SubLink, we are probably deciding whether a query tree + * transformation is safe, and a contained sub-select should affect that; + * for example, duplicating a sub-select containing a volatile function + * would be bad. However, once we've got to the stage of having SubPlans, + * subsequent planning need not consider volatility within those, since + * the executor won't change its evaluation rules for a SubPlan based on + * volatility. + * + * For some node types, for example, RestrictInfo and PathTarget, we cache + * whether we found any volatile functions or not and reuse that value in any + * future checks for that node. All of the logic for determining if the + * cached value should be set to VOLATILITY_NOVOLATILE or VOLATILITY_VOLATILE + * belongs in this function. Any code which makes changes to these nodes + * which could change the outcome this function must set the cached value back + * to VOLATILITY_UNKNOWN. That allows this function to redetermine the + * correct value during the next call, should we need to redetermine if the + * node contains any volatile functions again in the future. + */ +bool +contain_volatile_functions(Node *clause) +{ + return contain_volatile_functions_walker(clause, NULL); +} + +static bool +contain_volatile_functions_checker(Oid func_id, void *context) +{ + return (func_volatile(func_id) == PROVOLATILE_VOLATILE); +} + +static bool +contain_volatile_functions_walker(Node *node, void *context) +{ + if (node == NULL) + return false; + /* Check for volatile functions in node itself */ + if (check_functions_in_node(node, contain_volatile_functions_checker, + context)) + return true; + + if (IsA(node, NextValueExpr)) + { + /* NextValueExpr is volatile */ + return true; + } + + if (IsA(node, RestrictInfo)) + { + RestrictInfo *rinfo = (RestrictInfo *) node; + + /* + * For RestrictInfo, check if we've checked the volatility of it + * before. If so, we can just use the cached value and not bother + * checking it again. Otherwise, check it and cache if whether we + * found any volatile functions. + */ + if (rinfo->has_volatile == VOLATILITY_NOVOLATILE) + return false; + else if (rinfo->has_volatile == VOLATILITY_VOLATILE) + return true; + else + { + bool hasvolatile; + + hasvolatile = contain_volatile_functions_walker((Node *) rinfo->clause, + context); + if (hasvolatile) + rinfo->has_volatile = VOLATILITY_VOLATILE; + else + rinfo->has_volatile = VOLATILITY_NOVOLATILE; + + return hasvolatile; + } + } + + if (IsA(node, PathTarget)) + { + PathTarget *target = (PathTarget *) node; + + /* + * We also do caching for PathTarget the same as we do above for + * RestrictInfos. + */ + if (target->has_volatile_expr == VOLATILITY_NOVOLATILE) + return false; + else if (target->has_volatile_expr == VOLATILITY_VOLATILE) + return true; + else + { + bool hasvolatile; + + hasvolatile = contain_volatile_functions_walker((Node *) target->exprs, + context); + + if (hasvolatile) + target->has_volatile_expr = VOLATILITY_VOLATILE; + else + target->has_volatile_expr = VOLATILITY_NOVOLATILE; + + return hasvolatile; + } + } + + /* + * See notes in contain_mutable_functions_walker about why we treat + * MinMaxExpr, XmlExpr, and CoerceToDomain as immutable, while + * SQLValueFunction is stable. Hence, none of them are of interest here. + */ + + /* Recurse to check arguments */ + if (IsA(node, Query)) + { + /* Recurse into subselects */ + return query_tree_walker((Query *) node, + contain_volatile_functions_walker, + context, 0); + } + return expression_tree_walker(node, contain_volatile_functions_walker, + context); +} + +/* + * Special purpose version of contain_volatile_functions() for use in COPY: + * ignore nextval(), but treat all other functions normally. + */ +bool +contain_volatile_functions_not_nextval(Node *clause) +{ + return contain_volatile_functions_not_nextval_walker(clause, NULL); +} + +static bool +contain_volatile_functions_not_nextval_checker(Oid func_id, void *context) +{ + return (func_id != F_NEXTVAL && + func_volatile(func_id) == PROVOLATILE_VOLATILE); +} + +static bool +contain_volatile_functions_not_nextval_walker(Node *node, void *context) +{ + if (node == NULL) + return false; + /* Check for volatile functions in node itself */ + if (check_functions_in_node(node, + contain_volatile_functions_not_nextval_checker, + context)) + return true; + + /* + * See notes in contain_mutable_functions_walker about why we treat + * MinMaxExpr, XmlExpr, and CoerceToDomain as immutable, while + * SQLValueFunction is stable. Hence, none of them are of interest here. + * Also, since we're intentionally ignoring nextval(), presumably we + * should ignore NextValueExpr. + */ + + /* Recurse to check arguments */ + if (IsA(node, Query)) + { + /* Recurse into subselects */ + return query_tree_walker((Query *) node, + contain_volatile_functions_not_nextval_walker, + context, 0); + } + return expression_tree_walker(node, + contain_volatile_functions_not_nextval_walker, + context); +} + + +/***************************************************************************** + * Check queries for parallel unsafe and/or restricted constructs + *****************************************************************************/ + +/* + * max_parallel_hazard + * Find the worst parallel-hazard level in the given query + * + * Returns the worst function hazard property (the earliest in this list: + * PROPARALLEL_UNSAFE, PROPARALLEL_RESTRICTED, PROPARALLEL_SAFE) that can + * be found in the given parsetree. We use this to find out whether the query + * can be parallelized at all. The caller will also save the result in + * PlannerGlobal so as to short-circuit checks of portions of the querytree + * later, in the common case where everything is SAFE. + */ +char +max_parallel_hazard(Query *parse) +{ + max_parallel_hazard_context context; + + context.max_hazard = PROPARALLEL_SAFE; + context.max_interesting = PROPARALLEL_UNSAFE; + context.safe_param_ids = NIL; + (void) max_parallel_hazard_walker((Node *) parse, &context); + return context.max_hazard; +} + +/* + * is_parallel_safe + * Detect whether the given expr contains only parallel-safe functions + * + * root->glob->maxParallelHazard must previously have been set to the + * result of max_parallel_hazard() on the whole query. + */ +bool +is_parallel_safe(PlannerInfo *root, Node *node) +{ + max_parallel_hazard_context context; + PlannerInfo *proot; + ListCell *l; + + /* + * Even if the original querytree contained nothing unsafe, we need to + * search the expression if we have generated any PARAM_EXEC Params while + * planning, because those are parallel-restricted and there might be one + * in this expression. But otherwise we don't need to look. + */ + if (root->glob->maxParallelHazard == PROPARALLEL_SAFE && + root->glob->paramExecTypes == NIL) + return true; + /* Else use max_parallel_hazard's search logic, but stop on RESTRICTED */ + context.max_hazard = PROPARALLEL_SAFE; + context.max_interesting = PROPARALLEL_RESTRICTED; + context.safe_param_ids = NIL; + + /* + * The params that refer to the same or parent query level are considered + * parallel-safe. The idea is that we compute such params at Gather or + * Gather Merge node and pass their value to workers. + */ + for (proot = root; proot != NULL; proot = proot->parent_root) + { + foreach(l, proot->init_plans) + { + SubPlan *initsubplan = (SubPlan *) lfirst(l); + + context.safe_param_ids = list_concat(context.safe_param_ids, + initsubplan->setParam); + } + } + + return !max_parallel_hazard_walker(node, &context); +} + +/* core logic for all parallel-hazard checks */ +static bool +max_parallel_hazard_test(char proparallel, max_parallel_hazard_context *context) +{ + switch (proparallel) + { + case PROPARALLEL_SAFE: + /* nothing to see here, move along */ + break; + case PROPARALLEL_RESTRICTED: + /* increase max_hazard to RESTRICTED */ + Assert(context->max_hazard != PROPARALLEL_UNSAFE); + context->max_hazard = proparallel; + /* done if we are not expecting any unsafe functions */ + if (context->max_interesting == proparallel) + return true; + break; + case PROPARALLEL_UNSAFE: + context->max_hazard = proparallel; + /* we're always done at the first unsafe construct */ + return true; + default: + elog(ERROR, "unrecognized proparallel value \"%c\"", proparallel); + break; + } + return false; +} + +/* check_functions_in_node callback */ +static bool +max_parallel_hazard_checker(Oid func_id, void *context) +{ + return max_parallel_hazard_test(func_parallel(func_id), + (max_parallel_hazard_context *) context); +} + +static bool +max_parallel_hazard_walker(Node *node, max_parallel_hazard_context *context) +{ + if (node == NULL) + return false; + + /* Check for hazardous functions in node itself */ + if (check_functions_in_node(node, max_parallel_hazard_checker, + context)) + return true; + + /* + * It should be OK to treat MinMaxExpr as parallel-safe, since btree + * opclass support functions are generally parallel-safe. XmlExpr is a + * bit more dubious but we can probably get away with it. We err on the + * side of caution by treating CoerceToDomain as parallel-restricted. + * (Note: in principle that's wrong because a domain constraint could + * contain a parallel-unsafe function; but useful constraints probably + * never would have such, and assuming they do would cripple use of + * parallel query in the presence of domain types.) SQLValueFunction + * should be safe in all cases. NextValueExpr is parallel-unsafe. + */ + if (IsA(node, CoerceToDomain)) + { + if (max_parallel_hazard_test(PROPARALLEL_RESTRICTED, context)) + return true; + } + + else if (IsA(node, NextValueExpr)) + { + if (max_parallel_hazard_test(PROPARALLEL_UNSAFE, context)) + return true; + } + + /* + * Treat window functions as parallel-restricted because we aren't sure + * whether the input row ordering is fully deterministic, and the output + * of window functions might vary across workers if not. (In some cases, + * like where the window frame orders by a primary key, we could relax + * this restriction. But it doesn't currently seem worth expending extra + * effort to do so.) + */ + else if (IsA(node, WindowFunc)) + { + if (max_parallel_hazard_test(PROPARALLEL_RESTRICTED, context)) + return true; + } + + /* + * As a notational convenience for callers, look through RestrictInfo. + */ + else if (IsA(node, RestrictInfo)) + { + RestrictInfo *rinfo = (RestrictInfo *) node; + + return max_parallel_hazard_walker((Node *) rinfo->clause, context); + } + + /* + * Really we should not see SubLink during a max_interesting == restricted + * scan, but if we do, return true. + */ + else if (IsA(node, SubLink)) + { + if (max_parallel_hazard_test(PROPARALLEL_RESTRICTED, context)) + return true; + } + + /* + * Only parallel-safe SubPlans can be sent to workers. Within the + * testexpr of the SubPlan, Params representing the output columns of the + * subplan can be treated as parallel-safe, so temporarily add their IDs + * to the safe_param_ids list while examining the testexpr. + */ + else if (IsA(node, SubPlan)) + { + SubPlan *subplan = (SubPlan *) node; + List *save_safe_param_ids; + + if (!subplan->parallel_safe && + max_parallel_hazard_test(PROPARALLEL_RESTRICTED, context)) + return true; + save_safe_param_ids = context->safe_param_ids; + context->safe_param_ids = list_concat_copy(context->safe_param_ids, + subplan->paramIds); + if (max_parallel_hazard_walker(subplan->testexpr, context)) + return true; /* no need to restore safe_param_ids */ + list_free(context->safe_param_ids); + context->safe_param_ids = save_safe_param_ids; + /* we must also check args, but no special Param treatment there */ + if (max_parallel_hazard_walker((Node *) subplan->args, context)) + return true; + /* don't want to recurse normally, so we're done */ + return false; + } + + /* + * We can't pass Params to workers at the moment either, so they are also + * parallel-restricted, unless they are PARAM_EXTERN Params or are + * PARAM_EXEC Params listed in safe_param_ids, meaning they could be + * either generated within workers or can be computed by the leader and + * then their value can be passed to workers. + */ + else if (IsA(node, Param)) + { + Param *param = (Param *) node; + + if (param->paramkind == PARAM_EXTERN) + return false; + + if (param->paramkind != PARAM_EXEC || + !list_member_int(context->safe_param_ids, param->paramid)) + { + if (max_parallel_hazard_test(PROPARALLEL_RESTRICTED, context)) + return true; + } + return false; /* nothing to recurse to */ + } + + /* + * When we're first invoked on a completely unplanned tree, we must + * recurse into subqueries so to as to locate parallel-unsafe constructs + * anywhere in the tree. + */ + else if (IsA(node, Query)) + { + Query *query = (Query *) node; + + /* SELECT FOR UPDATE/SHARE must be treated as unsafe */ + if (query->rowMarks != NULL) + { + context->max_hazard = PROPARALLEL_UNSAFE; + return true; + } + + /* Recurse into subselects */ + return query_tree_walker(query, + max_parallel_hazard_walker, + context, 0); + } + + /* Recurse to check arguments */ + return expression_tree_walker(node, + max_parallel_hazard_walker, + context); +} + + +/***************************************************************************** + * Check clauses for nonstrict functions + *****************************************************************************/ + +/* + * contain_nonstrict_functions + * Recursively search for nonstrict functions within a clause. + * + * Returns true if any nonstrict construct is found --- ie, anything that + * could produce non-NULL output with a NULL input. + * + * The idea here is that the caller has verified that the expression contains + * one or more Var or Param nodes (as appropriate for the caller's need), and + * now wishes to prove that the expression result will be NULL if any of these + * inputs is NULL. If we return false, then the proof succeeded. + */ +bool +contain_nonstrict_functions(Node *clause) +{ + return contain_nonstrict_functions_walker(clause, NULL); +} + +static bool +contain_nonstrict_functions_checker(Oid func_id, void *context) +{ + return !func_strict(func_id); +} + +static bool +contain_nonstrict_functions_walker(Node *node, void *context) +{ + if (node == NULL) + return false; + if (IsA(node, Aggref)) + { + /* an aggregate could return non-null with null input */ + return true; + } + if (IsA(node, GroupingFunc)) + { + /* + * A GroupingFunc doesn't evaluate its arguments, and therefore must + * be treated as nonstrict. + */ + return true; + } + if (IsA(node, WindowFunc)) + { + /* a window function could return non-null with null input */ + return true; + } + if (IsA(node, SubscriptingRef)) + { + SubscriptingRef *sbsref = (SubscriptingRef *) node; + const SubscriptRoutines *sbsroutines; + + /* Subscripting assignment is always presumed nonstrict */ + if (sbsref->refassgnexpr != NULL) + return true; + /* Otherwise we must look up the subscripting support methods */ + sbsroutines = getSubscriptingRoutines(sbsref->refcontainertype, NULL); + if (!(sbsroutines && sbsroutines->fetch_strict)) + return true; + /* else fall through to check args */ + } + if (IsA(node, DistinctExpr)) + { + /* IS DISTINCT FROM is inherently non-strict */ + return true; + } + if (IsA(node, NullIfExpr)) + { + /* NULLIF is inherently non-strict */ + return true; + } + if (IsA(node, BoolExpr)) + { + BoolExpr *expr = (BoolExpr *) node; + + switch (expr->boolop) + { + case AND_EXPR: + case OR_EXPR: + /* AND, OR are inherently non-strict */ + return true; + default: + break; + } + } + if (IsA(node, SubLink)) + { + /* In some cases a sublink might be strict, but in general not */ + return true; + } + if (IsA(node, SubPlan)) + return true; + if (IsA(node, AlternativeSubPlan)) + return true; + if (IsA(node, FieldStore)) + return true; + if (IsA(node, CoerceViaIO)) + { + /* + * CoerceViaIO is strict regardless of whether the I/O functions are, + * so just go look at its argument; asking check_functions_in_node is + * useless expense and could deliver the wrong answer. + */ + return contain_nonstrict_functions_walker((Node *) ((CoerceViaIO *) node)->arg, + context); + } + if (IsA(node, ArrayCoerceExpr)) + { + /* + * ArrayCoerceExpr is strict at the array level, regardless of what + * the per-element expression is; so we should ignore elemexpr and + * recurse only into the arg. + */ + return contain_nonstrict_functions_walker((Node *) ((ArrayCoerceExpr *) node)->arg, + context); + } + if (IsA(node, CaseExpr)) + return true; + if (IsA(node, ArrayExpr)) + return true; + if (IsA(node, RowExpr)) + return true; + if (IsA(node, RowCompareExpr)) + return true; + if (IsA(node, CoalesceExpr)) + return true; + if (IsA(node, MinMaxExpr)) + return true; + if (IsA(node, XmlExpr)) + return true; + if (IsA(node, NullTest)) + return true; + if (IsA(node, BooleanTest)) + return true; + + /* Check other function-containing nodes */ + if (check_functions_in_node(node, contain_nonstrict_functions_checker, + context)) + return true; + + return expression_tree_walker(node, contain_nonstrict_functions_walker, + context); +} + +/***************************************************************************** + * Check clauses for Params + *****************************************************************************/ + +/* + * contain_exec_param + * Recursively search for PARAM_EXEC Params within a clause. + * + * Returns true if the clause contains any PARAM_EXEC Param with a paramid + * appearing in the given list of Param IDs. Does not descend into + * subqueries! + */ +bool +contain_exec_param(Node *clause, List *param_ids) +{ + return contain_exec_param_walker(clause, param_ids); +} + +static bool +contain_exec_param_walker(Node *node, List *param_ids) +{ + if (node == NULL) + return false; + if (IsA(node, Param)) + { + Param *p = (Param *) node; + + if (p->paramkind == PARAM_EXEC && + list_member_int(param_ids, p->paramid)) + return true; + } + return expression_tree_walker(node, contain_exec_param_walker, param_ids); +} + +/***************************************************************************** + * Check clauses for context-dependent nodes + *****************************************************************************/ + +/* + * contain_context_dependent_node + * Recursively search for context-dependent nodes within a clause. + * + * CaseTestExpr nodes must appear directly within the corresponding CaseExpr, + * not nested within another one, or they'll see the wrong test value. If one + * appears "bare" in the arguments of a SQL function, then we can't inline the + * SQL function for fear of creating such a situation. The same applies for + * CaseTestExpr used within the elemexpr of an ArrayCoerceExpr. + * + * CoerceToDomainValue would have the same issue if domain CHECK expressions + * could get inlined into larger expressions, but presently that's impossible. + * Still, it might be allowed in future, or other node types with similar + * issues might get invented. So give this function a generic name, and set + * up the recursion state to allow multiple flag bits. + */ +static bool +contain_context_dependent_node(Node *clause) +{ + int flags = 0; + + return contain_context_dependent_node_walker(clause, &flags); +} + +#define CCDN_CASETESTEXPR_OK 0x0001 /* CaseTestExpr okay here? */ + +static bool +contain_context_dependent_node_walker(Node *node, int *flags) +{ + if (node == NULL) + return false; + if (IsA(node, CaseTestExpr)) + return !(*flags & CCDN_CASETESTEXPR_OK); + else if (IsA(node, CaseExpr)) + { + CaseExpr *caseexpr = (CaseExpr *) node; + + /* + * If this CASE doesn't have a test expression, then it doesn't create + * a context in which CaseTestExprs should appear, so just fall + * through and treat it as a generic expression node. + */ + if (caseexpr->arg) + { + int save_flags = *flags; + bool res; + + /* + * Note: in principle, we could distinguish the various sub-parts + * of a CASE construct and set the flag bit only for some of them, + * since we are only expecting CaseTestExprs to appear in the + * "expr" subtree of the CaseWhen nodes. But it doesn't really + * seem worth any extra code. If there are any bare CaseTestExprs + * elsewhere in the CASE, something's wrong already. + */ + *flags |= CCDN_CASETESTEXPR_OK; + res = expression_tree_walker(node, + contain_context_dependent_node_walker, + (void *) flags); + *flags = save_flags; + return res; + } + } + else if (IsA(node, ArrayCoerceExpr)) + { + ArrayCoerceExpr *ac = (ArrayCoerceExpr *) node; + int save_flags; + bool res; + + /* Check the array expression */ + if (contain_context_dependent_node_walker((Node *) ac->arg, flags)) + return true; + + /* Check the elemexpr, which is allowed to contain CaseTestExpr */ + save_flags = *flags; + *flags |= CCDN_CASETESTEXPR_OK; + res = contain_context_dependent_node_walker((Node *) ac->elemexpr, + flags); + *flags = save_flags; + return res; + } + return expression_tree_walker(node, contain_context_dependent_node_walker, + (void *) flags); +} + +/***************************************************************************** + * Check clauses for Vars passed to non-leakproof functions + *****************************************************************************/ + +/* + * contain_leaked_vars + * Recursively scan a clause to discover whether it contains any Var + * nodes (of the current query level) that are passed as arguments to + * leaky functions. + * + * Returns true if the clause contains any non-leakproof functions that are + * passed Var nodes of the current query level, and which might therefore leak + * data. Such clauses must be applied after any lower-level security barrier + * clauses. + */ +bool +contain_leaked_vars(Node *clause) +{ + return contain_leaked_vars_walker(clause, NULL); +} + +static bool +contain_leaked_vars_checker(Oid func_id, void *context) +{ + return !get_func_leakproof(func_id); +} + +static bool +contain_leaked_vars_walker(Node *node, void *context) +{ + if (node == NULL) + return false; + + switch (nodeTag(node)) + { + case T_Var: + case T_Const: + case T_Param: + case T_ArrayExpr: + case T_FieldSelect: + case T_FieldStore: + case T_NamedArgExpr: + case T_BoolExpr: + case T_RelabelType: + case T_CollateExpr: + case T_CaseExpr: + case T_CaseTestExpr: + case T_RowExpr: + case T_SQLValueFunction: + case T_NullTest: + case T_BooleanTest: + case T_NextValueExpr: + case T_List: + + /* + * We know these node types don't contain function calls; but + * something further down in the node tree might. + */ + break; + + case T_FuncExpr: + case T_OpExpr: + case T_DistinctExpr: + case T_NullIfExpr: + case T_ScalarArrayOpExpr: + case T_CoerceViaIO: + case T_ArrayCoerceExpr: + + /* + * If node contains a leaky function call, and there's any Var + * underneath it, reject. + */ + if (check_functions_in_node(node, contain_leaked_vars_checker, + context) && + contain_var_clause(node)) + return true; + break; + + case T_SubscriptingRef: + { + SubscriptingRef *sbsref = (SubscriptingRef *) node; + const SubscriptRoutines *sbsroutines; + + /* Consult the subscripting support method info */ + sbsroutines = getSubscriptingRoutines(sbsref->refcontainertype, + NULL); + if (!sbsroutines || + !(sbsref->refassgnexpr != NULL ? + sbsroutines->store_leakproof : + sbsroutines->fetch_leakproof)) + { + /* Node is leaky, so reject if it contains Vars */ + if (contain_var_clause(node)) + return true; + } + } + break; + + case T_RowCompareExpr: + { + /* + * It's worth special-casing this because a leaky comparison + * function only compromises one pair of row elements, which + * might not contain Vars while others do. + */ + RowCompareExpr *rcexpr = (RowCompareExpr *) node; + ListCell *opid; + ListCell *larg; + ListCell *rarg; + + forthree(opid, rcexpr->opnos, + larg, rcexpr->largs, + rarg, rcexpr->rargs) + { + Oid funcid = get_opcode(lfirst_oid(opid)); + + if (!get_func_leakproof(funcid) && + (contain_var_clause((Node *) lfirst(larg)) || + contain_var_clause((Node *) lfirst(rarg)))) + return true; + } + } + break; + + case T_MinMaxExpr: + { + /* + * MinMaxExpr is leakproof if the comparison function it calls + * is leakproof. + */ + MinMaxExpr *minmaxexpr = (MinMaxExpr *) node; + TypeCacheEntry *typentry; + bool leakproof; + + /* Look up the btree comparison function for the datatype */ + typentry = lookup_type_cache(minmaxexpr->minmaxtype, + TYPECACHE_CMP_PROC); + if (OidIsValid(typentry->cmp_proc)) + leakproof = get_func_leakproof(typentry->cmp_proc); + else + { + /* + * The executor will throw an error, but here we just + * treat the missing function as leaky. + */ + leakproof = false; + } + + if (!leakproof && + contain_var_clause((Node *) minmaxexpr->args)) + return true; + } + break; + + case T_CurrentOfExpr: + + /* + * WHERE CURRENT OF doesn't contain leaky function calls. + * Moreover, it is essential that this is considered non-leaky, + * since the planner must always generate a TID scan when CURRENT + * OF is present -- cf. cost_tidscan. + */ + return false; + + default: + + /* + * If we don't recognize the node tag, assume it might be leaky. + * This prevents an unexpected security hole if someone adds a new + * node type that can call a function. + */ + return true; + } + return expression_tree_walker(node, contain_leaked_vars_walker, + context); +} + +/* + * find_nonnullable_rels + * Determine which base rels are forced nonnullable by given clause. + * + * Returns the set of all Relids that are referenced in the clause in such + * a way that the clause cannot possibly return TRUE if any of these Relids + * is an all-NULL row. (It is OK to err on the side of conservatism; hence + * the analysis here is simplistic.) + * + * The semantics here are subtly different from contain_nonstrict_functions: + * that function is concerned with NULL results from arbitrary expressions, + * but here we assume that the input is a Boolean expression, and wish to + * see if NULL inputs will provably cause a FALSE-or-NULL result. We expect + * the expression to have been AND/OR flattened and converted to implicit-AND + * format. + * + * Note: this function is largely duplicative of find_nonnullable_vars(). + * The reason not to simplify this function into a thin wrapper around + * find_nonnullable_vars() is that the tested conditions really are different: + * a clause like "t1.v1 IS NOT NULL OR t1.v2 IS NOT NULL" does not prove + * that either v1 or v2 can't be NULL, but it does prove that the t1 row + * as a whole can't be all-NULL. Also, the behavior for PHVs is different. + * + * top_level is true while scanning top-level AND/OR structure; here, showing + * the result is either FALSE or NULL is good enough. top_level is false when + * we have descended below a NOT or a strict function: now we must be able to + * prove that the subexpression goes to NULL. + * + * We don't use expression_tree_walker here because we don't want to descend + * through very many kinds of nodes; only the ones we can be sure are strict. + */ +Relids +find_nonnullable_rels(Node *clause) +{ + return find_nonnullable_rels_walker(clause, true); +} + +static Relids +find_nonnullable_rels_walker(Node *node, bool top_level) +{ + Relids result = NULL; + ListCell *l; + + if (node == NULL) + return NULL; + if (IsA(node, Var)) + { + Var *var = (Var *) node; + + if (var->varlevelsup == 0) + result = bms_make_singleton(var->varno); + } + else if (IsA(node, List)) + { + /* + * At top level, we are examining an implicit-AND list: if any of the + * arms produces FALSE-or-NULL then the result is FALSE-or-NULL. If + * not at top level, we are examining the arguments of a strict + * function: if any of them produce NULL then the result of the + * function must be NULL. So in both cases, the set of nonnullable + * rels is the union of those found in the arms, and we pass down the + * top_level flag unmodified. + */ + foreach(l, (List *) node) + { + result = bms_join(result, + find_nonnullable_rels_walker(lfirst(l), + top_level)); + } + } + else if (IsA(node, FuncExpr)) + { + FuncExpr *expr = (FuncExpr *) node; + + if (func_strict(expr->funcid)) + result = find_nonnullable_rels_walker((Node *) expr->args, false); + } + else if (IsA(node, OpExpr)) + { + OpExpr *expr = (OpExpr *) node; + + set_opfuncid(expr); + if (func_strict(expr->opfuncid)) + result = find_nonnullable_rels_walker((Node *) expr->args, false); + } + else if (IsA(node, ScalarArrayOpExpr)) + { + ScalarArrayOpExpr *expr = (ScalarArrayOpExpr *) node; + + if (is_strict_saop(expr, true)) + result = find_nonnullable_rels_walker((Node *) expr->args, false); + } + else if (IsA(node, BoolExpr)) + { + BoolExpr *expr = (BoolExpr *) node; + + switch (expr->boolop) + { + case AND_EXPR: + /* At top level we can just recurse (to the List case) */ + if (top_level) + { + result = find_nonnullable_rels_walker((Node *) expr->args, + top_level); + break; + } + + /* + * Below top level, even if one arm produces NULL, the result + * could be FALSE (hence not NULL). However, if *all* the + * arms produce NULL then the result is NULL, so we can take + * the intersection of the sets of nonnullable rels, just as + * for OR. Fall through to share code. + */ + /* FALL THRU */ + case OR_EXPR: + + /* + * OR is strict if all of its arms are, so we can take the + * intersection of the sets of nonnullable rels for each arm. + * This works for both values of top_level. + */ + foreach(l, expr->args) + { + Relids subresult; + + subresult = find_nonnullable_rels_walker(lfirst(l), + top_level); + if (result == NULL) /* first subresult? */ + result = subresult; + else + result = bms_int_members(result, subresult); + + /* + * If the intersection is empty, we can stop looking. This + * also justifies the test for first-subresult above. + */ + if (bms_is_empty(result)) + break; + } + break; + case NOT_EXPR: + /* NOT will return null if its arg is null */ + result = find_nonnullable_rels_walker((Node *) expr->args, + false); + break; + default: + elog(ERROR, "unrecognized boolop: %d", (int) expr->boolop); + break; + } + } + else if (IsA(node, RelabelType)) + { + RelabelType *expr = (RelabelType *) node; + + result = find_nonnullable_rels_walker((Node *) expr->arg, top_level); + } + else if (IsA(node, CoerceViaIO)) + { + /* not clear this is useful, but it can't hurt */ + CoerceViaIO *expr = (CoerceViaIO *) node; + + result = find_nonnullable_rels_walker((Node *) expr->arg, top_level); + } + else if (IsA(node, ArrayCoerceExpr)) + { + /* ArrayCoerceExpr is strict at the array level; ignore elemexpr */ + ArrayCoerceExpr *expr = (ArrayCoerceExpr *) node; + + result = find_nonnullable_rels_walker((Node *) expr->arg, top_level); + } + else if (IsA(node, ConvertRowtypeExpr)) + { + /* not clear this is useful, but it can't hurt */ + ConvertRowtypeExpr *expr = (ConvertRowtypeExpr *) node; + + result = find_nonnullable_rels_walker((Node *) expr->arg, top_level); + } + else if (IsA(node, CollateExpr)) + { + CollateExpr *expr = (CollateExpr *) node; + + result = find_nonnullable_rels_walker((Node *) expr->arg, top_level); + } + else if (IsA(node, NullTest)) + { + /* IS NOT NULL can be considered strict, but only at top level */ + NullTest *expr = (NullTest *) node; + + if (top_level && expr->nulltesttype == IS_NOT_NULL && !expr->argisrow) + result = find_nonnullable_rels_walker((Node *) expr->arg, false); + } + else if (IsA(node, BooleanTest)) + { + /* Boolean tests that reject NULL are strict at top level */ + BooleanTest *expr = (BooleanTest *) node; + + if (top_level && + (expr->booltesttype == IS_TRUE || + expr->booltesttype == IS_FALSE || + expr->booltesttype == IS_NOT_UNKNOWN)) + result = find_nonnullable_rels_walker((Node *) expr->arg, false); + } + else if (IsA(node, PlaceHolderVar)) + { + PlaceHolderVar *phv = (PlaceHolderVar *) node; + + /* + * If the contained expression forces any rels non-nullable, so does + * the PHV. + */ + result = find_nonnullable_rels_walker((Node *) phv->phexpr, top_level); + + /* + * If the PHV's syntactic scope is exactly one rel, it will be forced + * to be evaluated at that rel, and so it will behave like a Var of + * that rel: if the rel's entire output goes to null, so will the PHV. + * (If the syntactic scope is a join, we know that the PHV will go to + * null if the whole join does; but that is AND semantics while we + * need OR semantics for find_nonnullable_rels' result, so we can't do + * anything with the knowledge.) + */ + if (phv->phlevelsup == 0 && + bms_membership(phv->phrels) == BMS_SINGLETON) + result = bms_add_members(result, phv->phrels); + } + return result; +} + +/* + * find_nonnullable_vars + * Determine which Vars are forced nonnullable by given clause. + * + * Returns a list of all level-zero Vars that are referenced in the clause in + * such a way that the clause cannot possibly return TRUE if any of these Vars + * is NULL. (It is OK to err on the side of conservatism; hence the analysis + * here is simplistic.) + * + * The semantics here are subtly different from contain_nonstrict_functions: + * that function is concerned with NULL results from arbitrary expressions, + * but here we assume that the input is a Boolean expression, and wish to + * see if NULL inputs will provably cause a FALSE-or-NULL result. We expect + * the expression to have been AND/OR flattened and converted to implicit-AND + * format. + * + * The result is a palloc'd List, but we have not copied the member Var nodes. + * Also, we don't bother trying to eliminate duplicate entries. + * + * top_level is true while scanning top-level AND/OR structure; here, showing + * the result is either FALSE or NULL is good enough. top_level is false when + * we have descended below a NOT or a strict function: now we must be able to + * prove that the subexpression goes to NULL. + * + * We don't use expression_tree_walker here because we don't want to descend + * through very many kinds of nodes; only the ones we can be sure are strict. + */ +List * +find_nonnullable_vars(Node *clause) +{ + return find_nonnullable_vars_walker(clause, true); +} + +static List * +find_nonnullable_vars_walker(Node *node, bool top_level) +{ + List *result = NIL; + ListCell *l; + + if (node == NULL) + return NIL; + if (IsA(node, Var)) + { + Var *var = (Var *) node; + + if (var->varlevelsup == 0) + result = list_make1(var); + } + else if (IsA(node, List)) + { + /* + * At top level, we are examining an implicit-AND list: if any of the + * arms produces FALSE-or-NULL then the result is FALSE-or-NULL. If + * not at top level, we are examining the arguments of a strict + * function: if any of them produce NULL then the result of the + * function must be NULL. So in both cases, the set of nonnullable + * vars is the union of those found in the arms, and we pass down the + * top_level flag unmodified. + */ + foreach(l, (List *) node) + { + result = list_concat(result, + find_nonnullable_vars_walker(lfirst(l), + top_level)); + } + } + else if (IsA(node, FuncExpr)) + { + FuncExpr *expr = (FuncExpr *) node; + + if (func_strict(expr->funcid)) + result = find_nonnullable_vars_walker((Node *) expr->args, false); + } + else if (IsA(node, OpExpr)) + { + OpExpr *expr = (OpExpr *) node; + + set_opfuncid(expr); + if (func_strict(expr->opfuncid)) + result = find_nonnullable_vars_walker((Node *) expr->args, false); + } + else if (IsA(node, ScalarArrayOpExpr)) + { + ScalarArrayOpExpr *expr = (ScalarArrayOpExpr *) node; + + if (is_strict_saop(expr, true)) + result = find_nonnullable_vars_walker((Node *) expr->args, false); + } + else if (IsA(node, BoolExpr)) + { + BoolExpr *expr = (BoolExpr *) node; + + switch (expr->boolop) + { + case AND_EXPR: + /* At top level we can just recurse (to the List case) */ + if (top_level) + { + result = find_nonnullable_vars_walker((Node *) expr->args, + top_level); + break; + } + + /* + * Below top level, even if one arm produces NULL, the result + * could be FALSE (hence not NULL). However, if *all* the + * arms produce NULL then the result is NULL, so we can take + * the intersection of the sets of nonnullable vars, just as + * for OR. Fall through to share code. + */ + /* FALL THRU */ + case OR_EXPR: + + /* + * OR is strict if all of its arms are, so we can take the + * intersection of the sets of nonnullable vars for each arm. + * This works for both values of top_level. + */ + foreach(l, expr->args) + { + List *subresult; + + subresult = find_nonnullable_vars_walker(lfirst(l), + top_level); + if (result == NIL) /* first subresult? */ + result = subresult; + else + result = list_intersection(result, subresult); + + /* + * If the intersection is empty, we can stop looking. This + * also justifies the test for first-subresult above. + */ + if (result == NIL) + break; + } + break; + case NOT_EXPR: + /* NOT will return null if its arg is null */ + result = find_nonnullable_vars_walker((Node *) expr->args, + false); + break; + default: + elog(ERROR, "unrecognized boolop: %d", (int) expr->boolop); + break; + } + } + else if (IsA(node, RelabelType)) + { + RelabelType *expr = (RelabelType *) node; + + result = find_nonnullable_vars_walker((Node *) expr->arg, top_level); + } + else if (IsA(node, CoerceViaIO)) + { + /* not clear this is useful, but it can't hurt */ + CoerceViaIO *expr = (CoerceViaIO *) node; + + result = find_nonnullable_vars_walker((Node *) expr->arg, false); + } + else if (IsA(node, ArrayCoerceExpr)) + { + /* ArrayCoerceExpr is strict at the array level; ignore elemexpr */ + ArrayCoerceExpr *expr = (ArrayCoerceExpr *) node; + + result = find_nonnullable_vars_walker((Node *) expr->arg, top_level); + } + else if (IsA(node, ConvertRowtypeExpr)) + { + /* not clear this is useful, but it can't hurt */ + ConvertRowtypeExpr *expr = (ConvertRowtypeExpr *) node; + + result = find_nonnullable_vars_walker((Node *) expr->arg, top_level); + } + else if (IsA(node, CollateExpr)) + { + CollateExpr *expr = (CollateExpr *) node; + + result = find_nonnullable_vars_walker((Node *) expr->arg, top_level); + } + else if (IsA(node, NullTest)) + { + /* IS NOT NULL can be considered strict, but only at top level */ + NullTest *expr = (NullTest *) node; + + if (top_level && expr->nulltesttype == IS_NOT_NULL && !expr->argisrow) + result = find_nonnullable_vars_walker((Node *) expr->arg, false); + } + else if (IsA(node, BooleanTest)) + { + /* Boolean tests that reject NULL are strict at top level */ + BooleanTest *expr = (BooleanTest *) node; + + if (top_level && + (expr->booltesttype == IS_TRUE || + expr->booltesttype == IS_FALSE || + expr->booltesttype == IS_NOT_UNKNOWN)) + result = find_nonnullable_vars_walker((Node *) expr->arg, false); + } + else if (IsA(node, PlaceHolderVar)) + { + PlaceHolderVar *phv = (PlaceHolderVar *) node; + + result = find_nonnullable_vars_walker((Node *) phv->phexpr, top_level); + } + return result; +} + +/* + * find_forced_null_vars + * Determine which Vars must be NULL for the given clause to return TRUE. + * + * This is the complement of find_nonnullable_vars: find the level-zero Vars + * that must be NULL for the clause to return TRUE. (It is OK to err on the + * side of conservatism; hence the analysis here is simplistic. In fact, + * we only detect simple "var IS NULL" tests at the top level.) + * + * The result is a palloc'd List, but we have not copied the member Var nodes. + * Also, we don't bother trying to eliminate duplicate entries. + */ +List * +find_forced_null_vars(Node *node) +{ + List *result = NIL; + Var *var; + ListCell *l; + + if (node == NULL) + return NIL; + /* Check single-clause cases using subroutine */ + var = find_forced_null_var(node); + if (var) + { + result = list_make1(var); + } + /* Otherwise, handle AND-conditions */ + else if (IsA(node, List)) + { + /* + * At top level, we are examining an implicit-AND list: if any of the + * arms produces FALSE-or-NULL then the result is FALSE-or-NULL. + */ + foreach(l, (List *) node) + { + result = list_concat(result, + find_forced_null_vars(lfirst(l))); + } + } + else if (IsA(node, BoolExpr)) + { + BoolExpr *expr = (BoolExpr *) node; + + /* + * We don't bother considering the OR case, because it's fairly + * unlikely anyone would write "v1 IS NULL OR v1 IS NULL". Likewise, + * the NOT case isn't worth expending code on. + */ + if (expr->boolop == AND_EXPR) + { + /* At top level we can just recurse (to the List case) */ + result = find_forced_null_vars((Node *) expr->args); + } + } + return result; +} + +/* + * find_forced_null_var + * Return the Var forced null by the given clause, or NULL if it's + * not an IS NULL-type clause. For success, the clause must enforce + * *only* nullness of the particular Var, not any other conditions. + * + * This is just the single-clause case of find_forced_null_vars(), without + * any allowance for AND conditions. It's used by initsplan.c on individual + * qual clauses. The reason for not just applying find_forced_null_vars() + * is that if an AND of an IS NULL clause with something else were to somehow + * survive AND/OR flattening, initsplan.c might get fooled into discarding + * the whole clause when only the IS NULL part of it had been proved redundant. + */ +Var * +find_forced_null_var(Node *node) +{ + if (node == NULL) + return NULL; + if (IsA(node, NullTest)) + { + /* check for var IS NULL */ + NullTest *expr = (NullTest *) node; + + if (expr->nulltesttype == IS_NULL && !expr->argisrow) + { + Var *var = (Var *) expr->arg; + + if (var && IsA(var, Var) && + var->varlevelsup == 0) + return var; + } + } + else if (IsA(node, BooleanTest)) + { + /* var IS UNKNOWN is equivalent to var IS NULL */ + BooleanTest *expr = (BooleanTest *) node; + + if (expr->booltesttype == IS_UNKNOWN) + { + Var *var = (Var *) expr->arg; + + if (var && IsA(var, Var) && + var->varlevelsup == 0) + return var; + } + } + return NULL; +} + +/* + * Can we treat a ScalarArrayOpExpr as strict? + * + * If "falseOK" is true, then a "false" result can be considered strict, + * else we need to guarantee an actual NULL result for NULL input. + * + * "foo op ALL array" is strict if the op is strict *and* we can prove + * that the array input isn't an empty array. We can check that + * for the cases of an array constant and an ARRAY[] construct. + * + * "foo op ANY array" is strict in the falseOK sense if the op is strict. + * If not falseOK, the test is the same as for "foo op ALL array". + */ +static bool +is_strict_saop(ScalarArrayOpExpr *expr, bool falseOK) +{ + Node *rightop; + + /* The contained operator must be strict. */ + set_sa_opfuncid(expr); + if (!func_strict(expr->opfuncid)) + return false; + /* If ANY and falseOK, that's all we need to check. */ + if (expr->useOr && falseOK) + return true; + /* Else, we have to see if the array is provably non-empty. */ + Assert(list_length(expr->args) == 2); + rightop = (Node *) lsecond(expr->args); + if (rightop && IsA(rightop, Const)) + { + Datum arraydatum = ((Const *) rightop)->constvalue; + bool arrayisnull = ((Const *) rightop)->constisnull; + ArrayType *arrayval; + int nitems; + + if (arrayisnull) + return false; + arrayval = DatumGetArrayTypeP(arraydatum); + nitems = ArrayGetNItems(ARR_NDIM(arrayval), ARR_DIMS(arrayval)); + if (nitems > 0) + return true; + } + else if (rightop && IsA(rightop, ArrayExpr)) + { + ArrayExpr *arrayexpr = (ArrayExpr *) rightop; + + if (arrayexpr->elements != NIL && !arrayexpr->multidims) + return true; + } + return false; +} + + +/***************************************************************************** + * Check for "pseudo-constant" clauses + *****************************************************************************/ + +/* + * is_pseudo_constant_clause + * Detect whether an expression is "pseudo constant", ie, it contains no + * variables of the current query level and no uses of volatile functions. + * Such an expr is not necessarily a true constant: it can still contain + * Params and outer-level Vars, not to mention functions whose results + * may vary from one statement to the next. However, the expr's value + * will be constant over any one scan of the current query, so it can be + * used as, eg, an indexscan key. (Actually, the condition for indexscan + * keys is weaker than this; see is_pseudo_constant_for_index().) + * + * CAUTION: this function omits to test for one very important class of + * not-constant expressions, namely aggregates (Aggrefs). In current usage + * this is only applied to WHERE clauses and so a check for Aggrefs would be + * a waste of cycles; but be sure to also check contain_agg_clause() if you + * want to know about pseudo-constness in other contexts. The same goes + * for window functions (WindowFuncs). + */ +bool +is_pseudo_constant_clause(Node *clause) +{ + /* + * We could implement this check in one recursive scan. But since the + * check for volatile functions is both moderately expensive and unlikely + * to fail, it seems better to look for Vars first and only check for + * volatile functions if we find no Vars. + */ + if (!contain_var_clause(clause) && + !contain_volatile_functions(clause)) + return true; + return false; +} + +/* + * is_pseudo_constant_clause_relids + * Same as above, except caller already has available the var membership + * of the expression; this lets us avoid the contain_var_clause() scan. + */ +bool +is_pseudo_constant_clause_relids(Node *clause, Relids relids) +{ + if (bms_is_empty(relids) && + !contain_volatile_functions(clause)) + return true; + return false; +} + + +/***************************************************************************** + * * + * General clause-manipulating routines * + * * + *****************************************************************************/ + +/* + * NumRelids + * (formerly clause_relids) + * + * Returns the number of different relations referenced in 'clause'. + */ +int +NumRelids(PlannerInfo *root, Node *clause) +{ + Relids varnos = pull_varnos(root, clause); + int result = bms_num_members(varnos); + + bms_free(varnos); + return result; +} + +/* + * CommuteOpExpr: commute a binary operator clause + * + * XXX the clause is destructively modified! + */ +void +CommuteOpExpr(OpExpr *clause) +{ + Oid opoid; + Node *temp; + + /* Sanity checks: caller is at fault if these fail */ + if (!is_opclause(clause) || + list_length(clause->args) != 2) + elog(ERROR, "cannot commute non-binary-operator clause"); + + opoid = get_commutator(clause->opno); + + if (!OidIsValid(opoid)) + elog(ERROR, "could not find commutator for operator %u", + clause->opno); + + /* + * modify the clause in-place! + */ + clause->opno = opoid; + clause->opfuncid = InvalidOid; + /* opresulttype, opretset, opcollid, inputcollid need not change */ + + temp = linitial(clause->args); + linitial(clause->args) = lsecond(clause->args); + lsecond(clause->args) = temp; +} + +/* + * Helper for eval_const_expressions: check that datatype of an attribute + * is still what it was when the expression was parsed. This is needed to + * guard against improper simplification after ALTER COLUMN TYPE. (XXX we + * may well need to make similar checks elsewhere?) + * + * rowtypeid may come from a whole-row Var, and therefore it can be a domain + * over composite, but for this purpose we only care about checking the type + * of a contained field. + */ +static bool +rowtype_field_matches(Oid rowtypeid, int fieldnum, + Oid expectedtype, int32 expectedtypmod, + Oid expectedcollation) +{ + TupleDesc tupdesc; + Form_pg_attribute attr; + + /* No issue for RECORD, since there is no way to ALTER such a type */ + if (rowtypeid == RECORDOID) + return true; + tupdesc = lookup_rowtype_tupdesc_domain(rowtypeid, -1, false); + if (fieldnum <= 0 || fieldnum > tupdesc->natts) + { + ReleaseTupleDesc(tupdesc); + return false; + } + attr = TupleDescAttr(tupdesc, fieldnum - 1); + if (attr->attisdropped || + attr->atttypid != expectedtype || + attr->atttypmod != expectedtypmod || + attr->attcollation != expectedcollation) + { + ReleaseTupleDesc(tupdesc); + return false; + } + ReleaseTupleDesc(tupdesc); + return true; +} + + +/*-------------------- + * eval_const_expressions + * + * Reduce any recognizably constant subexpressions of the given + * expression tree, for example "2 + 2" => "4". More interestingly, + * we can reduce certain boolean expressions even when they contain + * non-constant subexpressions: "x OR true" => "true" no matter what + * the subexpression x is. (XXX We assume that no such subexpression + * will have important side-effects, which is not necessarily a good + * assumption in the presence of user-defined functions; do we need a + * pg_proc flag that prevents discarding the execution of a function?) + * + * We do understand that certain functions may deliver non-constant + * results even with constant inputs, "nextval()" being the classic + * example. Functions that are not marked "immutable" in pg_proc + * will not be pre-evaluated here, although we will reduce their + * arguments as far as possible. + * + * Whenever a function is eliminated from the expression by means of + * constant-expression evaluation or inlining, we add the function to + * root->glob->invalItems. This ensures the plan is known to depend on + * such functions, even though they aren't referenced anymore. + * + * We assume that the tree has already been type-checked and contains + * only operators and functions that are reasonable to try to execute. + * + * NOTE: "root" can be passed as NULL if the caller never wants to do any + * Param substitutions nor receive info about inlined functions. + * + * NOTE: the planner assumes that this will always flatten nested AND and + * OR clauses into N-argument form. See comments in prepqual.c. + * + * NOTE: another critical effect is that any function calls that require + * default arguments will be expanded, and named-argument calls will be + * converted to positional notation. The executor won't handle either. + *-------------------- + */ +Node * +eval_const_expressions(PlannerInfo *root, Node *node) +{ + eval_const_expressions_context context; + + if (root) + context.boundParams = root->glob->boundParams; /* bound Params */ + else + context.boundParams = NULL; + context.root = root; /* for inlined-function dependencies */ + context.active_fns = NIL; /* nothing being recursively simplified */ + context.case_val = NULL; /* no CASE being examined */ + context.estimate = false; /* safe transformations only */ + return eval_const_expressions_mutator(node, &context); +} + +#define MIN_ARRAY_SIZE_FOR_HASHED_SAOP 9 +/*-------------------- + * convert_saop_to_hashed_saop + * + * Recursively search 'node' for ScalarArrayOpExprs and fill in the hash + * function for any ScalarArrayOpExpr that looks like it would be useful to + * evaluate using a hash table rather than a linear search. + * + * We'll use a hash table if all of the following conditions are met: + * 1. The 2nd argument of the array contain only Consts. + * 2. useOr is true or there is a valid negator operator for the + * ScalarArrayOpExpr's opno. + * 3. There's valid hash function for both left and righthand operands and + * these hash functions are the same. + * 4. If the array contains enough elements for us to consider it to be + * worthwhile using a hash table rather than a linear search. + */ +void +convert_saop_to_hashed_saop(Node *node) +{ + (void) convert_saop_to_hashed_saop_walker(node, NULL); +} + +static bool +convert_saop_to_hashed_saop_walker(Node *node, void *context) +{ + if (node == NULL) + return false; + + if (IsA(node, ScalarArrayOpExpr)) + { + ScalarArrayOpExpr *saop = (ScalarArrayOpExpr *) node; + Expr *arrayarg = (Expr *) lsecond(saop->args); + Oid lefthashfunc; + Oid righthashfunc; + + if (arrayarg && IsA(arrayarg, Const) && + !((Const *) arrayarg)->constisnull) + { + if (saop->useOr) + { + if (get_op_hash_functions(saop->opno, &lefthashfunc, &righthashfunc) && + lefthashfunc == righthashfunc) + { + Datum arrdatum = ((Const *) arrayarg)->constvalue; + ArrayType *arr = (ArrayType *) DatumGetPointer(arrdatum); + int nitems; + + /* + * Only fill in the hash functions if the array looks + * large enough for it to be worth hashing instead of + * doing a linear search. + */ + nitems = ArrayGetNItems(ARR_NDIM(arr), ARR_DIMS(arr)); + + if (nitems >= MIN_ARRAY_SIZE_FOR_HASHED_SAOP) + { + /* Looks good. Fill in the hash functions */ + saop->hashfuncid = lefthashfunc; + } + return true; + } + } + else /* !saop->useOr */ + { + Oid negator = get_negator(saop->opno); + + /* + * Check if this is a NOT IN using an operator whose negator + * is hashable. If so we can still build a hash table and + * just ensure the lookup items are not in the hash table. + */ + if (OidIsValid(negator) && + get_op_hash_functions(negator, &lefthashfunc, &righthashfunc) && + lefthashfunc == righthashfunc) + { + Datum arrdatum = ((Const *) arrayarg)->constvalue; + ArrayType *arr = (ArrayType *) DatumGetPointer(arrdatum); + int nitems; + + /* + * Only fill in the hash functions if the array looks + * large enough for it to be worth hashing instead of + * doing a linear search. + */ + nitems = ArrayGetNItems(ARR_NDIM(arr), ARR_DIMS(arr)); + + if (nitems >= MIN_ARRAY_SIZE_FOR_HASHED_SAOP) + { + /* Looks good. Fill in the hash functions */ + saop->hashfuncid = lefthashfunc; + + /* + * Also set the negfuncid. The executor will need + * that to perform hashtable lookups. + */ + saop->negfuncid = get_opcode(negator); + } + return true; + } + } + } + } + + return expression_tree_walker(node, convert_saop_to_hashed_saop_walker, NULL); +} + + +/*-------------------- + * estimate_expression_value + * + * This function attempts to estimate the value of an expression for + * planning purposes. It is in essence a more aggressive version of + * eval_const_expressions(): we will perform constant reductions that are + * not necessarily 100% safe, but are reasonable for estimation purposes. + * + * Currently the extra steps that are taken in this mode are: + * 1. Substitute values for Params, where a bound Param value has been made + * available by the caller of planner(), even if the Param isn't marked + * constant. This effectively means that we plan using the first supplied + * value of the Param. + * 2. Fold stable, as well as immutable, functions to constants. + * 3. Reduce PlaceHolderVar nodes to their contained expressions. + *-------------------- + */ +Node * +estimate_expression_value(PlannerInfo *root, Node *node) +{ + eval_const_expressions_context context; + + context.boundParams = root->glob->boundParams; /* bound Params */ + /* we do not need to mark the plan as depending on inlined functions */ + context.root = NULL; + context.active_fns = NIL; /* nothing being recursively simplified */ + context.case_val = NULL; /* no CASE being examined */ + context.estimate = true; /* unsafe transformations OK */ + return eval_const_expressions_mutator(node, &context); +} + +/* + * The generic case in eval_const_expressions_mutator is to recurse using + * expression_tree_mutator, which will copy the given node unchanged but + * const-simplify its arguments (if any) as far as possible. If the node + * itself does immutable processing, and each of its arguments were reduced + * to a Const, we can then reduce it to a Const using evaluate_expr. (Some + * node types need more complicated logic; for example, a CASE expression + * might be reducible to a constant even if not all its subtrees are.) + */ +#define ece_generic_processing(node) \ + expression_tree_mutator((Node *) (node), eval_const_expressions_mutator, \ + (void *) context) + +/* + * Check whether all arguments of the given node were reduced to Consts. + * By going directly to expression_tree_walker, contain_non_const_walker + * is not applied to the node itself, only to its children. + */ +#define ece_all_arguments_const(node) \ + (!expression_tree_walker((Node *) (node), contain_non_const_walker, NULL)) + +/* Generic macro for applying evaluate_expr */ +#define ece_evaluate_expr(node) \ + ((Node *) evaluate_expr((Expr *) (node), \ + exprType((Node *) (node)), \ + exprTypmod((Node *) (node)), \ + exprCollation((Node *) (node)))) + +/* + * Recursive guts of eval_const_expressions/estimate_expression_value + */ +static Node * +eval_const_expressions_mutator(Node *node, + eval_const_expressions_context *context) +{ + if (node == NULL) + return NULL; + switch (nodeTag(node)) + { + case T_Param: + { + Param *param = (Param *) node; + ParamListInfo paramLI = context->boundParams; + + /* Look to see if we've been given a value for this Param */ + if (param->paramkind == PARAM_EXTERN && + paramLI != NULL && + param->paramid > 0 && + param->paramid <= paramLI->numParams) + { + ParamExternData *prm; + ParamExternData prmdata; + + /* + * Give hook a chance in case parameter is dynamic. Tell + * it that this fetch is speculative, so it should avoid + * erroring out if parameter is unavailable. + */ + if (paramLI->paramFetch != NULL) + prm = paramLI->paramFetch(paramLI, param->paramid, + true, &prmdata); + else + prm = ¶mLI->params[param->paramid - 1]; + + /* + * We don't just check OidIsValid, but insist that the + * fetched type match the Param, just in case the hook did + * something unexpected. No need to throw an error here + * though; leave that for runtime. + */ + if (OidIsValid(prm->ptype) && + prm->ptype == param->paramtype) + { + /* OK to substitute parameter value? */ + if (context->estimate || + (prm->pflags & PARAM_FLAG_CONST)) + { + /* + * Return a Const representing the param value. + * Must copy pass-by-ref datatypes, since the + * Param might be in a memory context + * shorter-lived than our output plan should be. + */ + int16 typLen; + bool typByVal; + Datum pval; + Const *con; + + get_typlenbyval(param->paramtype, + &typLen, &typByVal); + if (prm->isnull || typByVal) + pval = prm->value; + else + pval = datumCopy(prm->value, typByVal, typLen); + con = makeConst(param->paramtype, + param->paramtypmod, + param->paramcollid, + (int) typLen, + pval, + prm->isnull, + typByVal); + con->location = param->location; + return (Node *) con; + } + } + } + + /* + * Not replaceable, so just copy the Param (no need to + * recurse) + */ + return (Node *) copyObject(param); + } + case T_WindowFunc: + { + WindowFunc *expr = (WindowFunc *) node; + Oid funcid = expr->winfnoid; + List *args; + Expr *aggfilter; + HeapTuple func_tuple; + WindowFunc *newexpr; + + /* + * We can't really simplify a WindowFunc node, but we mustn't + * just fall through to the default processing, because we + * have to apply expand_function_arguments to its argument + * list. That takes care of inserting default arguments and + * expanding named-argument notation. + */ + func_tuple = SearchSysCache1(PROCOID, ObjectIdGetDatum(funcid)); + if (!HeapTupleIsValid(func_tuple)) + elog(ERROR, "cache lookup failed for function %u", funcid); + + args = expand_function_arguments(expr->args, + false, expr->wintype, + func_tuple); + + ReleaseSysCache(func_tuple); + + /* Now, recursively simplify the args (which are a List) */ + args = (List *) + expression_tree_mutator((Node *) args, + eval_const_expressions_mutator, + (void *) context); + /* ... and the filter expression, which isn't */ + aggfilter = (Expr *) + eval_const_expressions_mutator((Node *) expr->aggfilter, + context); + + /* And build the replacement WindowFunc node */ + newexpr = makeNode(WindowFunc); + newexpr->winfnoid = expr->winfnoid; + newexpr->wintype = expr->wintype; + newexpr->wincollid = expr->wincollid; + newexpr->inputcollid = expr->inputcollid; + newexpr->args = args; + newexpr->aggfilter = aggfilter; + newexpr->winref = expr->winref; + newexpr->winstar = expr->winstar; + newexpr->winagg = expr->winagg; + newexpr->location = expr->location; + + return (Node *) newexpr; + } + case T_FuncExpr: + { + FuncExpr *expr = (FuncExpr *) node; + List *args = expr->args; + Expr *simple; + FuncExpr *newexpr; + + /* + * Code for op/func reduction is pretty bulky, so split it out + * as a separate function. Note: exprTypmod normally returns + * -1 for a FuncExpr, but not when the node is recognizably a + * length coercion; we want to preserve the typmod in the + * eventual Const if so. + */ + simple = simplify_function(expr->funcid, + expr->funcresulttype, + exprTypmod(node), + expr->funccollid, + expr->inputcollid, + &args, + expr->funcvariadic, + true, + true, + context); + if (simple) /* successfully simplified it */ + return (Node *) simple; + + /* + * The expression cannot be simplified any further, so build + * and return a replacement FuncExpr node using the + * possibly-simplified arguments. Note that we have also + * converted the argument list to positional notation. + */ + newexpr = makeNode(FuncExpr); + newexpr->funcid = expr->funcid; + newexpr->funcresulttype = expr->funcresulttype; + newexpr->funcretset = expr->funcretset; + newexpr->funcvariadic = expr->funcvariadic; + newexpr->funcformat = expr->funcformat; + newexpr->funccollid = expr->funccollid; + newexpr->inputcollid = expr->inputcollid; + newexpr->args = args; + newexpr->location = expr->location; + return (Node *) newexpr; + } + case T_OpExpr: + { + OpExpr *expr = (OpExpr *) node; + List *args = expr->args; + Expr *simple; + OpExpr *newexpr; + + /* + * Need to get OID of underlying function. Okay to scribble + * on input to this extent. + */ + set_opfuncid(expr); + + /* + * Code for op/func reduction is pretty bulky, so split it out + * as a separate function. + */ + simple = simplify_function(expr->opfuncid, + expr->opresulttype, -1, + expr->opcollid, + expr->inputcollid, + &args, + false, + true, + true, + context); + if (simple) /* successfully simplified it */ + return (Node *) simple; + + /* + * If the operator is boolean equality or inequality, we know + * how to simplify cases involving one constant and one + * non-constant argument. + */ + if (expr->opno == BooleanEqualOperator || + expr->opno == BooleanNotEqualOperator) + { + simple = (Expr *) simplify_boolean_equality(expr->opno, + args); + if (simple) /* successfully simplified it */ + return (Node *) simple; + } + + /* + * The expression cannot be simplified any further, so build + * and return a replacement OpExpr node using the + * possibly-simplified arguments. + */ + newexpr = makeNode(OpExpr); + newexpr->opno = expr->opno; + newexpr->opfuncid = expr->opfuncid; + newexpr->opresulttype = expr->opresulttype; + newexpr->opretset = expr->opretset; + newexpr->opcollid = expr->opcollid; + newexpr->inputcollid = expr->inputcollid; + newexpr->args = args; + newexpr->location = expr->location; + return (Node *) newexpr; + } + case T_DistinctExpr: + { + DistinctExpr *expr = (DistinctExpr *) node; + List *args; + ListCell *arg; + bool has_null_input = false; + bool all_null_input = true; + bool has_nonconst_input = false; + Expr *simple; + DistinctExpr *newexpr; + + /* + * Reduce constants in the DistinctExpr's arguments. We know + * args is either NIL or a List node, so we can call + * expression_tree_mutator directly rather than recursing to + * self. + */ + args = (List *) expression_tree_mutator((Node *) expr->args, + eval_const_expressions_mutator, + (void *) context); + + /* + * We must do our own check for NULLs because DistinctExpr has + * different results for NULL input than the underlying + * operator does. + */ + foreach(arg, args) + { + if (IsA(lfirst(arg), Const)) + { + has_null_input |= ((Const *) lfirst(arg))->constisnull; + all_null_input &= ((Const *) lfirst(arg))->constisnull; + } + else + has_nonconst_input = true; + } + + /* all constants? then can optimize this out */ + if (!has_nonconst_input) + { + /* all nulls? then not distinct */ + if (all_null_input) + return makeBoolConst(false, false); + + /* one null? then distinct */ + if (has_null_input) + return makeBoolConst(true, false); + + /* otherwise try to evaluate the '=' operator */ + /* (NOT okay to try to inline it, though!) */ + + /* + * Need to get OID of underlying function. Okay to + * scribble on input to this extent. + */ + set_opfuncid((OpExpr *) expr); /* rely on struct + * equivalence */ + + /* + * Code for op/func reduction is pretty bulky, so split it + * out as a separate function. + */ + simple = simplify_function(expr->opfuncid, + expr->opresulttype, -1, + expr->opcollid, + expr->inputcollid, + &args, + false, + false, + false, + context); + if (simple) /* successfully simplified it */ + { + /* + * Since the underlying operator is "=", must negate + * its result + */ + Const *csimple = castNode(Const, simple); + + csimple->constvalue = + BoolGetDatum(!DatumGetBool(csimple->constvalue)); + return (Node *) csimple; + } + } + + /* + * The expression cannot be simplified any further, so build + * and return a replacement DistinctExpr node using the + * possibly-simplified arguments. + */ + newexpr = makeNode(DistinctExpr); + newexpr->opno = expr->opno; + newexpr->opfuncid = expr->opfuncid; + newexpr->opresulttype = expr->opresulttype; + newexpr->opretset = expr->opretset; + newexpr->opcollid = expr->opcollid; + newexpr->inputcollid = expr->inputcollid; + newexpr->args = args; + newexpr->location = expr->location; + return (Node *) newexpr; + } + case T_NullIfExpr: + { + NullIfExpr *expr; + ListCell *arg; + bool has_nonconst_input = false; + + /* Copy the node and const-simplify its arguments */ + expr = (NullIfExpr *) ece_generic_processing(node); + + /* If either argument is NULL they can't be equal */ + foreach(arg, expr->args) + { + if (!IsA(lfirst(arg), Const)) + has_nonconst_input = true; + else if (((Const *) lfirst(arg))->constisnull) + return (Node *) linitial(expr->args); + } + + /* + * Need to get OID of underlying function before checking if + * the function is OK to evaluate. + */ + set_opfuncid((OpExpr *) expr); + + if (!has_nonconst_input && + ece_function_is_safe(expr->opfuncid, context)) + return ece_evaluate_expr(expr); + + return (Node *) expr; + } + case T_ScalarArrayOpExpr: + { + ScalarArrayOpExpr *saop; + + /* Copy the node and const-simplify its arguments */ + saop = (ScalarArrayOpExpr *) ece_generic_processing(node); + + /* Make sure we know underlying function */ + set_sa_opfuncid(saop); + + /* + * If all arguments are Consts, and it's a safe function, we + * can fold to a constant + */ + if (ece_all_arguments_const(saop) && + ece_function_is_safe(saop->opfuncid, context)) + return ece_evaluate_expr(saop); + return (Node *) saop; + } + case T_BoolExpr: + { + BoolExpr *expr = (BoolExpr *) node; + + switch (expr->boolop) + { + case OR_EXPR: + { + List *newargs; + bool haveNull = false; + bool forceTrue = false; + + newargs = simplify_or_arguments(expr->args, + context, + &haveNull, + &forceTrue); + if (forceTrue) + return makeBoolConst(true, false); + if (haveNull) + newargs = lappend(newargs, + makeBoolConst(false, true)); + /* If all the inputs are FALSE, result is FALSE */ + if (newargs == NIL) + return makeBoolConst(false, false); + + /* + * If only one nonconst-or-NULL input, it's the + * result + */ + if (list_length(newargs) == 1) + return (Node *) linitial(newargs); + /* Else we still need an OR node */ + return (Node *) make_orclause(newargs); + } + case AND_EXPR: + { + List *newargs; + bool haveNull = false; + bool forceFalse = false; + + newargs = simplify_and_arguments(expr->args, + context, + &haveNull, + &forceFalse); + if (forceFalse) + return makeBoolConst(false, false); + if (haveNull) + newargs = lappend(newargs, + makeBoolConst(false, true)); + /* If all the inputs are TRUE, result is TRUE */ + if (newargs == NIL) + return makeBoolConst(true, false); + + /* + * If only one nonconst-or-NULL input, it's the + * result + */ + if (list_length(newargs) == 1) + return (Node *) linitial(newargs); + /* Else we still need an AND node */ + return (Node *) make_andclause(newargs); + } + case NOT_EXPR: + { + Node *arg; + + Assert(list_length(expr->args) == 1); + arg = eval_const_expressions_mutator(linitial(expr->args), + context); + + /* + * Use negate_clause() to see if we can simplify + * away the NOT. + */ + return negate_clause(arg); + } + default: + elog(ERROR, "unrecognized boolop: %d", + (int) expr->boolop); + break; + } + break; + } + case T_SubPlan: + case T_AlternativeSubPlan: + + /* + * Return a SubPlan unchanged --- too late to do anything with it. + * + * XXX should we ereport() here instead? Probably this routine + * should never be invoked after SubPlan creation. + */ + return node; + case T_RelabelType: + { + RelabelType *relabel = (RelabelType *) node; + Node *arg; + + /* Simplify the input ... */ + arg = eval_const_expressions_mutator((Node *) relabel->arg, + context); + /* ... and attach a new RelabelType node, if needed */ + return applyRelabelType(arg, + relabel->resulttype, + relabel->resulttypmod, + relabel->resultcollid, + relabel->relabelformat, + relabel->location, + true); + } + case T_CoerceViaIO: + { + CoerceViaIO *expr = (CoerceViaIO *) node; + List *args; + Oid outfunc; + bool outtypisvarlena; + Oid infunc; + Oid intypioparam; + Expr *simple; + CoerceViaIO *newexpr; + + /* Make a List so we can use simplify_function */ + args = list_make1(expr->arg); + + /* + * CoerceViaIO represents calling the source type's output + * function then the result type's input function. So, try to + * simplify it as though it were a stack of two such function + * calls. First we need to know what the functions are. + * + * Note that the coercion functions are assumed not to care + * about input collation, so we just pass InvalidOid for that. + */ + getTypeOutputInfo(exprType((Node *) expr->arg), + &outfunc, &outtypisvarlena); + getTypeInputInfo(expr->resulttype, + &infunc, &intypioparam); + + simple = simplify_function(outfunc, + CSTRINGOID, -1, + InvalidOid, + InvalidOid, + &args, + false, + true, + true, + context); + if (simple) /* successfully simplified output fn */ + { + /* + * Input functions may want 1 to 3 arguments. We always + * supply all three, trusting that nothing downstream will + * complain. + */ + args = list_make3(simple, + makeConst(OIDOID, + -1, + InvalidOid, + sizeof(Oid), + ObjectIdGetDatum(intypioparam), + false, + true), + makeConst(INT4OID, + -1, + InvalidOid, + sizeof(int32), + Int32GetDatum(-1), + false, + true)); + + simple = simplify_function(infunc, + expr->resulttype, -1, + expr->resultcollid, + InvalidOid, + &args, + false, + false, + true, + context); + if (simple) /* successfully simplified input fn */ + return (Node *) simple; + } + + /* + * The expression cannot be simplified any further, so build + * and return a replacement CoerceViaIO node using the + * possibly-simplified argument. + */ + newexpr = makeNode(CoerceViaIO); + newexpr->arg = (Expr *) linitial(args); + newexpr->resulttype = expr->resulttype; + newexpr->resultcollid = expr->resultcollid; + newexpr->coerceformat = expr->coerceformat; + newexpr->location = expr->location; + return (Node *) newexpr; + } + case T_ArrayCoerceExpr: + { + ArrayCoerceExpr *ac = makeNode(ArrayCoerceExpr); + Node *save_case_val; + + /* + * Copy the node and const-simplify its arguments. We can't + * use ece_generic_processing() here because we need to mess + * with case_val only while processing the elemexpr. + */ + memcpy(ac, node, sizeof(ArrayCoerceExpr)); + ac->arg = (Expr *) + eval_const_expressions_mutator((Node *) ac->arg, + context); + + /* + * Set up for the CaseTestExpr node contained in the elemexpr. + * We must prevent it from absorbing any outer CASE value. + */ + save_case_val = context->case_val; + context->case_val = NULL; + + ac->elemexpr = (Expr *) + eval_const_expressions_mutator((Node *) ac->elemexpr, + context); + + context->case_val = save_case_val; + + /* + * If constant argument and the per-element expression is + * immutable, we can simplify the whole thing to a constant. + * Exception: although contain_mutable_functions considers + * CoerceToDomain immutable for historical reasons, let's not + * do so here; this ensures coercion to an array-over-domain + * does not apply the domain's constraints until runtime. + */ + if (ac->arg && IsA(ac->arg, Const) && + ac->elemexpr && !IsA(ac->elemexpr, CoerceToDomain) && + !contain_mutable_functions((Node *) ac->elemexpr)) + return ece_evaluate_expr(ac); + + return (Node *) ac; + } + case T_CollateExpr: + { + /* + * We replace CollateExpr with RelabelType, so as to improve + * uniformity of expression representation and thus simplify + * comparison of expressions. Hence this looks very nearly + * the same as the RelabelType case, and we can apply the same + * optimizations to avoid unnecessary RelabelTypes. + */ + CollateExpr *collate = (CollateExpr *) node; + Node *arg; + + /* Simplify the input ... */ + arg = eval_const_expressions_mutator((Node *) collate->arg, + context); + /* ... and attach a new RelabelType node, if needed */ + return applyRelabelType(arg, + exprType(arg), + exprTypmod(arg), + collate->collOid, + COERCE_IMPLICIT_CAST, + collate->location, + true); + } + case T_CaseExpr: + { + /*---------- + * CASE expressions can be simplified if there are constant + * condition clauses: + * FALSE (or NULL): drop the alternative + * TRUE: drop all remaining alternatives + * If the first non-FALSE alternative is a constant TRUE, + * we can simplify the entire CASE to that alternative's + * expression. If there are no non-FALSE alternatives, + * we simplify the entire CASE to the default result (ELSE). + * + * If we have a simple-form CASE with constant test + * expression, we substitute the constant value for contained + * CaseTestExpr placeholder nodes, so that we have the + * opportunity to reduce constant test conditions. For + * example this allows + * CASE 0 WHEN 0 THEN 1 ELSE 1/0 END + * to reduce to 1 rather than drawing a divide-by-0 error. + * Note that when the test expression is constant, we don't + * have to include it in the resulting CASE; for example + * CASE 0 WHEN x THEN y ELSE z END + * is transformed by the parser to + * CASE 0 WHEN CaseTestExpr = x THEN y ELSE z END + * which we can simplify to + * CASE WHEN 0 = x THEN y ELSE z END + * It is not necessary for the executor to evaluate the "arg" + * expression when executing the CASE, since any contained + * CaseTestExprs that might have referred to it will have been + * replaced by the constant. + *---------- + */ + CaseExpr *caseexpr = (CaseExpr *) node; + CaseExpr *newcase; + Node *save_case_val; + Node *newarg; + List *newargs; + bool const_true_cond; + Node *defresult = NULL; + ListCell *arg; + + /* Simplify the test expression, if any */ + newarg = eval_const_expressions_mutator((Node *) caseexpr->arg, + context); + + /* Set up for contained CaseTestExpr nodes */ + save_case_val = context->case_val; + if (newarg && IsA(newarg, Const)) + { + context->case_val = newarg; + newarg = NULL; /* not needed anymore, see above */ + } + else + context->case_val = NULL; + + /* Simplify the WHEN clauses */ + newargs = NIL; + const_true_cond = false; + foreach(arg, caseexpr->args) + { + CaseWhen *oldcasewhen = lfirst_node(CaseWhen, arg); + Node *casecond; + Node *caseresult; + + /* Simplify this alternative's test condition */ + casecond = eval_const_expressions_mutator((Node *) oldcasewhen->expr, + context); + + /* + * If the test condition is constant FALSE (or NULL), then + * drop this WHEN clause completely, without processing + * the result. + */ + if (casecond && IsA(casecond, Const)) + { + Const *const_input = (Const *) casecond; + + if (const_input->constisnull || + !DatumGetBool(const_input->constvalue)) + continue; /* drop alternative with FALSE cond */ + /* Else it's constant TRUE */ + const_true_cond = true; + } + + /* Simplify this alternative's result value */ + caseresult = eval_const_expressions_mutator((Node *) oldcasewhen->result, + context); + + /* If non-constant test condition, emit a new WHEN node */ + if (!const_true_cond) + { + CaseWhen *newcasewhen = makeNode(CaseWhen); + + newcasewhen->expr = (Expr *) casecond; + newcasewhen->result = (Expr *) caseresult; + newcasewhen->location = oldcasewhen->location; + newargs = lappend(newargs, newcasewhen); + continue; + } + + /* + * Found a TRUE condition, so none of the remaining + * alternatives can be reached. We treat the result as + * the default result. + */ + defresult = caseresult; + break; + } + + /* Simplify the default result, unless we replaced it above */ + if (!const_true_cond) + defresult = eval_const_expressions_mutator((Node *) caseexpr->defresult, + context); + + context->case_val = save_case_val; + + /* + * If no non-FALSE alternatives, CASE reduces to the default + * result + */ + if (newargs == NIL) + return defresult; + /* Otherwise we need a new CASE node */ + newcase = makeNode(CaseExpr); + newcase->casetype = caseexpr->casetype; + newcase->casecollid = caseexpr->casecollid; + newcase->arg = (Expr *) newarg; + newcase->args = newargs; + newcase->defresult = (Expr *) defresult; + newcase->location = caseexpr->location; + return (Node *) newcase; + } + case T_CaseTestExpr: + { + /* + * If we know a constant test value for the current CASE + * construct, substitute it for the placeholder. Else just + * return the placeholder as-is. + */ + if (context->case_val) + return copyObject(context->case_val); + else + return copyObject(node); + } + case T_SubscriptingRef: + case T_ArrayExpr: + case T_RowExpr: + case T_MinMaxExpr: + { + /* + * Generic handling for node types whose own processing is + * known to be immutable, and for which we need no smarts + * beyond "simplify if all inputs are constants". + * + * Treating SubscriptingRef this way assumes that subscripting + * fetch and assignment are both immutable. This constrains + * type-specific subscripting implementations; maybe we should + * relax it someday. + * + * Treating MinMaxExpr this way amounts to assuming that the + * btree comparison function it calls is immutable; see the + * reasoning in contain_mutable_functions_walker. + */ + + /* Copy the node and const-simplify its arguments */ + node = ece_generic_processing(node); + /* If all arguments are Consts, we can fold to a constant */ + if (ece_all_arguments_const(node)) + return ece_evaluate_expr(node); + return node; + } + case T_CoalesceExpr: + { + CoalesceExpr *coalesceexpr = (CoalesceExpr *) node; + CoalesceExpr *newcoalesce; + List *newargs; + ListCell *arg; + + newargs = NIL; + foreach(arg, coalesceexpr->args) + { + Node *e; + + e = eval_const_expressions_mutator((Node *) lfirst(arg), + context); + + /* + * We can remove null constants from the list. For a + * non-null constant, if it has not been preceded by any + * other non-null-constant expressions then it is the + * result. Otherwise, it's the next argument, but we can + * drop following arguments since they will never be + * reached. + */ + if (IsA(e, Const)) + { + if (((Const *) e)->constisnull) + continue; /* drop null constant */ + if (newargs == NIL) + return e; /* first expr */ + newargs = lappend(newargs, e); + break; + } + newargs = lappend(newargs, e); + } + + /* + * If all the arguments were constant null, the result is just + * null + */ + if (newargs == NIL) + return (Node *) makeNullConst(coalesceexpr->coalescetype, + -1, + coalesceexpr->coalescecollid); + + newcoalesce = makeNode(CoalesceExpr); + newcoalesce->coalescetype = coalesceexpr->coalescetype; + newcoalesce->coalescecollid = coalesceexpr->coalescecollid; + newcoalesce->args = newargs; + newcoalesce->location = coalesceexpr->location; + return (Node *) newcoalesce; + } + case T_SQLValueFunction: + { + /* + * All variants of SQLValueFunction are stable, so if we are + * estimating the expression's value, we should evaluate the + * current function value. Otherwise just copy. + */ + SQLValueFunction *svf = (SQLValueFunction *) node; + + if (context->estimate) + return (Node *) evaluate_expr((Expr *) svf, + svf->type, + svf->typmod, + InvalidOid); + else + return copyObject((Node *) svf); + } + case T_FieldSelect: + { + /* + * We can optimize field selection from a whole-row Var into a + * simple Var. (This case won't be generated directly by the + * parser, because ParseComplexProjection short-circuits it. + * But it can arise while simplifying functions.) Also, we + * can optimize field selection from a RowExpr construct, or + * of course from a constant. + * + * However, replacing a whole-row Var in this way has a + * pitfall: if we've already built the rel targetlist for the + * source relation, then the whole-row Var is scheduled to be + * produced by the relation scan, but the simple Var probably + * isn't, which will lead to a failure in setrefs.c. This is + * not a problem when handling simple single-level queries, in + * which expression simplification always happens first. It + * is a risk for lateral references from subqueries, though. + * To avoid such failures, don't optimize uplevel references. + * + * We must also check that the declared type of the field is + * still the same as when the FieldSelect was created --- this + * can change if someone did ALTER COLUMN TYPE on the rowtype. + * If it isn't, we skip the optimization; the case will + * probably fail at runtime, but that's not our problem here. + */ + FieldSelect *fselect = (FieldSelect *) node; + FieldSelect *newfselect; + Node *arg; + + arg = eval_const_expressions_mutator((Node *) fselect->arg, + context); + if (arg && IsA(arg, Var) && + ((Var *) arg)->varattno == InvalidAttrNumber && + ((Var *) arg)->varlevelsup == 0) + { + if (rowtype_field_matches(((Var *) arg)->vartype, + fselect->fieldnum, + fselect->resulttype, + fselect->resulttypmod, + fselect->resultcollid)) + return (Node *) makeVar(((Var *) arg)->varno, + fselect->fieldnum, + fselect->resulttype, + fselect->resulttypmod, + fselect->resultcollid, + ((Var *) arg)->varlevelsup); + } + if (arg && IsA(arg, RowExpr)) + { + RowExpr *rowexpr = (RowExpr *) arg; + + if (fselect->fieldnum > 0 && + fselect->fieldnum <= list_length(rowexpr->args)) + { + Node *fld = (Node *) list_nth(rowexpr->args, + fselect->fieldnum - 1); + + if (rowtype_field_matches(rowexpr->row_typeid, + fselect->fieldnum, + fselect->resulttype, + fselect->resulttypmod, + fselect->resultcollid) && + fselect->resulttype == exprType(fld) && + fselect->resulttypmod == exprTypmod(fld) && + fselect->resultcollid == exprCollation(fld)) + return fld; + } + } + newfselect = makeNode(FieldSelect); + newfselect->arg = (Expr *) arg; + newfselect->fieldnum = fselect->fieldnum; + newfselect->resulttype = fselect->resulttype; + newfselect->resulttypmod = fselect->resulttypmod; + newfselect->resultcollid = fselect->resultcollid; + if (arg && IsA(arg, Const)) + { + Const *con = (Const *) arg; + + if (rowtype_field_matches(con->consttype, + newfselect->fieldnum, + newfselect->resulttype, + newfselect->resulttypmod, + newfselect->resultcollid)) + return ece_evaluate_expr(newfselect); + } + return (Node *) newfselect; + } + case T_NullTest: + { + NullTest *ntest = (NullTest *) node; + NullTest *newntest; + Node *arg; + + arg = eval_const_expressions_mutator((Node *) ntest->arg, + context); + if (ntest->argisrow && arg && IsA(arg, RowExpr)) + { + /* + * We break ROW(...) IS [NOT] NULL into separate tests on + * its component fields. This form is usually more + * efficient to evaluate, as well as being more amenable + * to optimization. + */ + RowExpr *rarg = (RowExpr *) arg; + List *newargs = NIL; + ListCell *l; + + foreach(l, rarg->args) + { + Node *relem = (Node *) lfirst(l); + + /* + * A constant field refutes the whole NullTest if it's + * of the wrong nullness; else we can discard it. + */ + if (relem && IsA(relem, Const)) + { + Const *carg = (Const *) relem; + + if (carg->constisnull ? + (ntest->nulltesttype == IS_NOT_NULL) : + (ntest->nulltesttype == IS_NULL)) + return makeBoolConst(false, false); + continue; + } + + /* + * Else, make a scalar (argisrow == false) NullTest + * for this field. Scalar semantics are required + * because IS [NOT] NULL doesn't recurse; see comments + * in ExecEvalRowNullInt(). + */ + newntest = makeNode(NullTest); + newntest->arg = (Expr *) relem; + newntest->nulltesttype = ntest->nulltesttype; + newntest->argisrow = false; + newntest->location = ntest->location; + newargs = lappend(newargs, newntest); + } + /* If all the inputs were constants, result is TRUE */ + if (newargs == NIL) + return makeBoolConst(true, false); + /* If only one nonconst input, it's the result */ + if (list_length(newargs) == 1) + return (Node *) linitial(newargs); + /* Else we need an AND node */ + return (Node *) make_andclause(newargs); + } + if (!ntest->argisrow && arg && IsA(arg, Const)) + { + Const *carg = (Const *) arg; + bool result; + + switch (ntest->nulltesttype) + { + case IS_NULL: + result = carg->constisnull; + break; + case IS_NOT_NULL: + result = !carg->constisnull; + break; + default: + elog(ERROR, "unrecognized nulltesttype: %d", + (int) ntest->nulltesttype); + result = false; /* keep compiler quiet */ + break; + } + + return makeBoolConst(result, false); + } + + newntest = makeNode(NullTest); + newntest->arg = (Expr *) arg; + newntest->nulltesttype = ntest->nulltesttype; + newntest->argisrow = ntest->argisrow; + newntest->location = ntest->location; + return (Node *) newntest; + } + case T_BooleanTest: + { + /* + * This case could be folded into the generic handling used + * for ArrayExpr etc. But because the simplification logic is + * so trivial, applying evaluate_expr() to perform it would be + * a heavy overhead. BooleanTest is probably common enough to + * justify keeping this bespoke implementation. + */ + BooleanTest *btest = (BooleanTest *) node; + BooleanTest *newbtest; + Node *arg; + + arg = eval_const_expressions_mutator((Node *) btest->arg, + context); + if (arg && IsA(arg, Const)) + { + Const *carg = (Const *) arg; + bool result; + + switch (btest->booltesttype) + { + case IS_TRUE: + result = (!carg->constisnull && + DatumGetBool(carg->constvalue)); + break; + case IS_NOT_TRUE: + result = (carg->constisnull || + !DatumGetBool(carg->constvalue)); + break; + case IS_FALSE: + result = (!carg->constisnull && + !DatumGetBool(carg->constvalue)); + break; + case IS_NOT_FALSE: + result = (carg->constisnull || + DatumGetBool(carg->constvalue)); + break; + case IS_UNKNOWN: + result = carg->constisnull; + break; + case IS_NOT_UNKNOWN: + result = !carg->constisnull; + break; + default: + elog(ERROR, "unrecognized booltesttype: %d", + (int) btest->booltesttype); + result = false; /* keep compiler quiet */ + break; + } + + return makeBoolConst(result, false); + } + + newbtest = makeNode(BooleanTest); + newbtest->arg = (Expr *) arg; + newbtest->booltesttype = btest->booltesttype; + newbtest->location = btest->location; + return (Node *) newbtest; + } + case T_CoerceToDomain: + { + /* + * If the domain currently has no constraints, we replace the + * CoerceToDomain node with a simple RelabelType, which is + * both far faster to execute and more amenable to later + * optimization. We must then mark the plan as needing to be + * rebuilt if the domain's constraints change. + * + * Also, in estimation mode, always replace CoerceToDomain + * nodes, effectively assuming that the coercion will succeed. + */ + CoerceToDomain *cdomain = (CoerceToDomain *) node; + CoerceToDomain *newcdomain; + Node *arg; + + arg = eval_const_expressions_mutator((Node *) cdomain->arg, + context); + if (context->estimate || + !DomainHasConstraints(cdomain->resulttype)) + { + /* Record dependency, if this isn't estimation mode */ + if (context->root && !context->estimate) + record_plan_type_dependency(context->root, + cdomain->resulttype); + + /* Generate RelabelType to substitute for CoerceToDomain */ + return applyRelabelType(arg, + cdomain->resulttype, + cdomain->resulttypmod, + cdomain->resultcollid, + cdomain->coercionformat, + cdomain->location, + true); + } + + newcdomain = makeNode(CoerceToDomain); + newcdomain->arg = (Expr *) arg; + newcdomain->resulttype = cdomain->resulttype; + newcdomain->resulttypmod = cdomain->resulttypmod; + newcdomain->resultcollid = cdomain->resultcollid; + newcdomain->coercionformat = cdomain->coercionformat; + newcdomain->location = cdomain->location; + return (Node *) newcdomain; + } + case T_PlaceHolderVar: + + /* + * In estimation mode, just strip the PlaceHolderVar node + * altogether; this amounts to estimating that the contained value + * won't be forced to null by an outer join. In regular mode we + * just use the default behavior (ie, simplify the expression but + * leave the PlaceHolderVar node intact). + */ + if (context->estimate) + { + PlaceHolderVar *phv = (PlaceHolderVar *) node; + + return eval_const_expressions_mutator((Node *) phv->phexpr, + context); + } + break; + case T_ConvertRowtypeExpr: + { + ConvertRowtypeExpr *cre = castNode(ConvertRowtypeExpr, node); + Node *arg; + ConvertRowtypeExpr *newcre; + + arg = eval_const_expressions_mutator((Node *) cre->arg, + context); + + newcre = makeNode(ConvertRowtypeExpr); + newcre->resulttype = cre->resulttype; + newcre->convertformat = cre->convertformat; + newcre->location = cre->location; + + /* + * In case of a nested ConvertRowtypeExpr, we can convert the + * leaf row directly to the topmost row format without any + * intermediate conversions. (This works because + * ConvertRowtypeExpr is used only for child->parent + * conversion in inheritance trees, which works by exact match + * of column name, and a column absent in an intermediate + * result can't be present in the final result.) + * + * No need to check more than one level deep, because the + * above recursion will have flattened anything else. + */ + if (arg != NULL && IsA(arg, ConvertRowtypeExpr)) + { + ConvertRowtypeExpr *argcre = (ConvertRowtypeExpr *) arg; + + arg = (Node *) argcre->arg; + + /* + * Make sure an outer implicit conversion can't hide an + * inner explicit one. + */ + if (newcre->convertformat == COERCE_IMPLICIT_CAST) + newcre->convertformat = argcre->convertformat; + } + + newcre->arg = (Expr *) arg; + + if (arg != NULL && IsA(arg, Const)) + return ece_evaluate_expr((Node *) newcre); + return (Node *) newcre; + } + default: + break; + } + + /* + * For any node type not handled above, copy the node unchanged but + * const-simplify its subexpressions. This is the correct thing for node + * types whose behavior might change between planning and execution, such + * as CurrentOfExpr. It's also a safe default for new node types not + * known to this routine. + */ + return ece_generic_processing(node); +} + +/* + * Subroutine for eval_const_expressions: check for non-Const nodes. + * + * We can abort recursion immediately on finding a non-Const node. This is + * critical for performance, else eval_const_expressions_mutator would take + * O(N^2) time on non-simplifiable trees. However, we do need to descend + * into List nodes since expression_tree_walker sometimes invokes the walker + * function directly on List subtrees. + */ +static bool +contain_non_const_walker(Node *node, void *context) +{ + if (node == NULL) + return false; + if (IsA(node, Const)) + return false; + if (IsA(node, List)) + return expression_tree_walker(node, contain_non_const_walker, context); + /* Otherwise, abort the tree traversal and return true */ + return true; +} + +/* + * Subroutine for eval_const_expressions: check if a function is OK to evaluate + */ +static bool +ece_function_is_safe(Oid funcid, eval_const_expressions_context *context) +{ + char provolatile = func_volatile(funcid); + + /* + * Ordinarily we are only allowed to simplify immutable functions. But for + * purposes of estimation, we consider it okay to simplify functions that + * are merely stable; the risk that the result might change from planning + * time to execution time is worth taking in preference to not being able + * to estimate the value at all. + */ + if (provolatile == PROVOLATILE_IMMUTABLE) + return true; + if (context->estimate && provolatile == PROVOLATILE_STABLE) + return true; + return false; +} + +/* + * Subroutine for eval_const_expressions: process arguments of an OR clause + * + * This includes flattening of nested ORs as well as recursion to + * eval_const_expressions to simplify the OR arguments. + * + * After simplification, OR arguments are handled as follows: + * non constant: keep + * FALSE: drop (does not affect result) + * TRUE: force result to TRUE + * NULL: keep only one + * We must keep one NULL input because OR expressions evaluate to NULL when no + * input is TRUE and at least one is NULL. We don't actually include the NULL + * here, that's supposed to be done by the caller. + * + * The output arguments *haveNull and *forceTrue must be initialized false + * by the caller. They will be set true if a NULL constant or TRUE constant, + * respectively, is detected anywhere in the argument list. + */ +static List * +simplify_or_arguments(List *args, + eval_const_expressions_context *context, + bool *haveNull, bool *forceTrue) +{ + List *newargs = NIL; + List *unprocessed_args; + + /* + * We want to ensure that any OR immediately beneath another OR gets + * flattened into a single OR-list, so as to simplify later reasoning. + * + * To avoid stack overflow from recursion of eval_const_expressions, we + * resort to some tenseness here: we keep a list of not-yet-processed + * inputs, and handle flattening of nested ORs by prepending to the to-do + * list instead of recursing. Now that the parser generates N-argument + * ORs from simple lists, this complexity is probably less necessary than + * it once was, but we might as well keep the logic. + */ + unprocessed_args = list_copy(args); + while (unprocessed_args) + { + Node *arg = (Node *) linitial(unprocessed_args); + + unprocessed_args = list_delete_first(unprocessed_args); + + /* flatten nested ORs as per above comment */ + if (is_orclause(arg)) + { + List *subargs = ((BoolExpr *) arg)->args; + List *oldlist = unprocessed_args; + + unprocessed_args = list_concat_copy(subargs, unprocessed_args); + /* perhaps-overly-tense code to avoid leaking old lists */ + list_free(oldlist); + continue; + } + + /* If it's not an OR, simplify it */ + arg = eval_const_expressions_mutator(arg, context); + + /* + * It is unlikely but not impossible for simplification of a non-OR + * clause to produce an OR. Recheck, but don't be too tense about it + * since it's not a mainstream case. In particular we don't worry + * about const-simplifying the input twice, nor about list leakage. + */ + if (is_orclause(arg)) + { + List *subargs = ((BoolExpr *) arg)->args; + + unprocessed_args = list_concat_copy(subargs, unprocessed_args); + continue; + } + + /* + * OK, we have a const-simplified non-OR argument. Process it per + * comments above. + */ + if (IsA(arg, Const)) + { + Const *const_input = (Const *) arg; + + if (const_input->constisnull) + *haveNull = true; + else if (DatumGetBool(const_input->constvalue)) + { + *forceTrue = true; + + /* + * Once we detect a TRUE result we can just exit the loop + * immediately. However, if we ever add a notion of + * non-removable functions, we'd need to keep scanning. + */ + return NIL; + } + /* otherwise, we can drop the constant-false input */ + continue; + } + + /* else emit the simplified arg into the result list */ + newargs = lappend(newargs, arg); + } + + return newargs; +} + +/* + * Subroutine for eval_const_expressions: process arguments of an AND clause + * + * This includes flattening of nested ANDs as well as recursion to + * eval_const_expressions to simplify the AND arguments. + * + * After simplification, AND arguments are handled as follows: + * non constant: keep + * TRUE: drop (does not affect result) + * FALSE: force result to FALSE + * NULL: keep only one + * We must keep one NULL input because AND expressions evaluate to NULL when + * no input is FALSE and at least one is NULL. We don't actually include the + * NULL here, that's supposed to be done by the caller. + * + * The output arguments *haveNull and *forceFalse must be initialized false + * by the caller. They will be set true if a null constant or false constant, + * respectively, is detected anywhere in the argument list. + */ +static List * +simplify_and_arguments(List *args, + eval_const_expressions_context *context, + bool *haveNull, bool *forceFalse) +{ + List *newargs = NIL; + List *unprocessed_args; + + /* See comments in simplify_or_arguments */ + unprocessed_args = list_copy(args); + while (unprocessed_args) + { + Node *arg = (Node *) linitial(unprocessed_args); + + unprocessed_args = list_delete_first(unprocessed_args); + + /* flatten nested ANDs as per above comment */ + if (is_andclause(arg)) + { + List *subargs = ((BoolExpr *) arg)->args; + List *oldlist = unprocessed_args; + + unprocessed_args = list_concat_copy(subargs, unprocessed_args); + /* perhaps-overly-tense code to avoid leaking old lists */ + list_free(oldlist); + continue; + } + + /* If it's not an AND, simplify it */ + arg = eval_const_expressions_mutator(arg, context); + + /* + * It is unlikely but not impossible for simplification of a non-AND + * clause to produce an AND. Recheck, but don't be too tense about it + * since it's not a mainstream case. In particular we don't worry + * about const-simplifying the input twice, nor about list leakage. + */ + if (is_andclause(arg)) + { + List *subargs = ((BoolExpr *) arg)->args; + + unprocessed_args = list_concat_copy(subargs, unprocessed_args); + continue; + } + + /* + * OK, we have a const-simplified non-AND argument. Process it per + * comments above. + */ + if (IsA(arg, Const)) + { + Const *const_input = (Const *) arg; + + if (const_input->constisnull) + *haveNull = true; + else if (!DatumGetBool(const_input->constvalue)) + { + *forceFalse = true; + + /* + * Once we detect a FALSE result we can just exit the loop + * immediately. However, if we ever add a notion of + * non-removable functions, we'd need to keep scanning. + */ + return NIL; + } + /* otherwise, we can drop the constant-true input */ + continue; + } + + /* else emit the simplified arg into the result list */ + newargs = lappend(newargs, arg); + } + + return newargs; +} + +/* + * Subroutine for eval_const_expressions: try to simplify boolean equality + * or inequality condition + * + * Inputs are the operator OID and the simplified arguments to the operator. + * Returns a simplified expression if successful, or NULL if cannot + * simplify the expression. + * + * The idea here is to reduce "x = true" to "x" and "x = false" to "NOT x", + * or similarly "x <> true" to "NOT x" and "x <> false" to "x". + * This is only marginally useful in itself, but doing it in constant folding + * ensures that we will recognize these forms as being equivalent in, for + * example, partial index matching. + * + * We come here only if simplify_function has failed; therefore we cannot + * see two constant inputs, nor a constant-NULL input. + */ +static Node * +simplify_boolean_equality(Oid opno, List *args) +{ + Node *leftop; + Node *rightop; + + Assert(list_length(args) == 2); + leftop = linitial(args); + rightop = lsecond(args); + if (leftop && IsA(leftop, Const)) + { + Assert(!((Const *) leftop)->constisnull); + if (opno == BooleanEqualOperator) + { + if (DatumGetBool(((Const *) leftop)->constvalue)) + return rightop; /* true = foo */ + else + return negate_clause(rightop); /* false = foo */ + } + else + { + if (DatumGetBool(((Const *) leftop)->constvalue)) + return negate_clause(rightop); /* true <> foo */ + else + return rightop; /* false <> foo */ + } + } + if (rightop && IsA(rightop, Const)) + { + Assert(!((Const *) rightop)->constisnull); + if (opno == BooleanEqualOperator) + { + if (DatumGetBool(((Const *) rightop)->constvalue)) + return leftop; /* foo = true */ + else + return negate_clause(leftop); /* foo = false */ + } + else + { + if (DatumGetBool(((Const *) rightop)->constvalue)) + return negate_clause(leftop); /* foo <> true */ + else + return leftop; /* foo <> false */ + } + } + return NULL; +} + +/* + * Subroutine for eval_const_expressions: try to simplify a function call + * (which might originally have been an operator; we don't care) + * + * Inputs are the function OID, actual result type OID (which is needed for + * polymorphic functions), result typmod, result collation, the input + * collation to use for the function, the original argument list (not + * const-simplified yet, unless process_args is false), and some flags; + * also the context data for eval_const_expressions. + * + * Returns a simplified expression if successful, or NULL if cannot + * simplify the function call. + * + * This function is also responsible for converting named-notation argument + * lists into positional notation and/or adding any needed default argument + * expressions; which is a bit grotty, but it avoids extra fetches of the + * function's pg_proc tuple. For this reason, the args list is + * pass-by-reference. Conversion and const-simplification of the args list + * will be done even if simplification of the function call itself is not + * possible. + */ +static Expr * +simplify_function(Oid funcid, Oid result_type, int32 result_typmod, + Oid result_collid, Oid input_collid, List **args_p, + bool funcvariadic, bool process_args, bool allow_non_const, + eval_const_expressions_context *context) +{ + List *args = *args_p; + HeapTuple func_tuple; + Form_pg_proc func_form; + Expr *newexpr; + + /* + * We have three strategies for simplification: execute the function to + * deliver a constant result, use a transform function to generate a + * substitute node tree, or expand in-line the body of the function + * definition (which only works for simple SQL-language functions, but + * that is a common case). Each case needs access to the function's + * pg_proc tuple, so fetch it just once. + * + * Note: the allow_non_const flag suppresses both the second and third + * strategies; so if !allow_non_const, simplify_function can only return a + * Const or NULL. Argument-list rewriting happens anyway, though. + */ + func_tuple = SearchSysCache1(PROCOID, ObjectIdGetDatum(funcid)); + if (!HeapTupleIsValid(func_tuple)) + elog(ERROR, "cache lookup failed for function %u", funcid); + func_form = (Form_pg_proc) GETSTRUCT(func_tuple); + + /* + * Process the function arguments, unless the caller did it already. + * + * Here we must deal with named or defaulted arguments, and then + * recursively apply eval_const_expressions to the whole argument list. + */ + if (process_args) + { + args = expand_function_arguments(args, false, result_type, func_tuple); + args = (List *) expression_tree_mutator((Node *) args, + eval_const_expressions_mutator, + (void *) context); + /* Argument processing done, give it back to the caller */ + *args_p = args; + } + + /* Now attempt simplification of the function call proper. */ + + newexpr = evaluate_function(funcid, result_type, result_typmod, + result_collid, input_collid, + args, funcvariadic, + func_tuple, context); + + if (!newexpr && allow_non_const && OidIsValid(func_form->prosupport)) + { + /* + * Build a SupportRequestSimplify node to pass to the support + * function, pointing to a dummy FuncExpr node containing the + * simplified arg list. We use this approach to present a uniform + * interface to the support function regardless of how the target + * function is actually being invoked. + */ + SupportRequestSimplify req; + FuncExpr fexpr; + + fexpr.xpr.type = T_FuncExpr; + fexpr.funcid = funcid; + fexpr.funcresulttype = result_type; + fexpr.funcretset = func_form->proretset; + fexpr.funcvariadic = funcvariadic; + fexpr.funcformat = COERCE_EXPLICIT_CALL; + fexpr.funccollid = result_collid; + fexpr.inputcollid = input_collid; + fexpr.args = args; + fexpr.location = -1; + + req.type = T_SupportRequestSimplify; + req.root = context->root; + req.fcall = &fexpr; + + newexpr = (Expr *) + DatumGetPointer(OidFunctionCall1(func_form->prosupport, + PointerGetDatum(&req))); + + /* catch a possible API misunderstanding */ + Assert(newexpr != (Expr *) &fexpr); + } + + if (!newexpr && allow_non_const) + newexpr = inline_function(funcid, result_type, result_collid, + input_collid, args, funcvariadic, + func_tuple, context); + + ReleaseSysCache(func_tuple); + + return newexpr; +} + +/* + * expand_function_arguments: convert named-notation args to positional args + * and/or insert default args, as needed + * + * Returns a possibly-transformed version of the args list. + * + * If include_out_arguments is true, then the args list and the result + * include OUT arguments. + * + * The expected result type of the call must be given, for sanity-checking + * purposes. Also, we ask the caller to provide the function's actual + * pg_proc tuple, not just its OID. + * + * If we need to change anything, the input argument list is copied, not + * modified. + * + * Note: this gets applied to operator argument lists too, even though the + * cases it handles should never occur there. This should be OK since it + * will fall through very quickly if there's nothing to do. + */ +List * +expand_function_arguments(List *args, bool include_out_arguments, + Oid result_type, HeapTuple func_tuple) +{ + Form_pg_proc funcform = (Form_pg_proc) GETSTRUCT(func_tuple); + Oid *proargtypes = funcform->proargtypes.values; + int pronargs = funcform->pronargs; + bool has_named_args = false; + ListCell *lc; + + /* + * If we are asked to match to OUT arguments, then use the proallargtypes + * array (which includes those); otherwise use proargtypes (which + * doesn't). Of course, if proallargtypes is null, we always use + * proargtypes. (Fetching proallargtypes is annoyingly expensive + * considering that we may have nothing to do here, but fortunately the + * common case is include_out_arguments == false.) + */ + if (include_out_arguments) + { + Datum proallargtypes; + bool isNull; + + proallargtypes = SysCacheGetAttr(PROCOID, func_tuple, + Anum_pg_proc_proallargtypes, + &isNull); + if (!isNull) + { + ArrayType *arr = DatumGetArrayTypeP(proallargtypes); + + pronargs = ARR_DIMS(arr)[0]; + if (ARR_NDIM(arr) != 1 || + pronargs < 0 || + ARR_HASNULL(arr) || + ARR_ELEMTYPE(arr) != OIDOID) + elog(ERROR, "proallargtypes is not a 1-D Oid array or it contains nulls"); + Assert(pronargs >= funcform->pronargs); + proargtypes = (Oid *) ARR_DATA_PTR(arr); + } + } + + /* Do we have any named arguments? */ + foreach(lc, args) + { + Node *arg = (Node *) lfirst(lc); + + if (IsA(arg, NamedArgExpr)) + { + has_named_args = true; + break; + } + } + + /* If so, we must apply reorder_function_arguments */ + if (has_named_args) + { + args = reorder_function_arguments(args, pronargs, func_tuple); + /* Recheck argument types and add casts if needed */ + recheck_cast_function_args(args, result_type, + proargtypes, pronargs, + func_tuple); + } + else if (list_length(args) < pronargs) + { + /* No named args, but we seem to be short some defaults */ + args = add_function_defaults(args, pronargs, func_tuple); + /* Recheck argument types and add casts if needed */ + recheck_cast_function_args(args, result_type, + proargtypes, pronargs, + func_tuple); + } + + return args; +} + +/* + * reorder_function_arguments: convert named-notation args to positional args + * + * This function also inserts default argument values as needed, since it's + * impossible to form a truly valid positional call without that. + */ +static List * +reorder_function_arguments(List *args, int pronargs, HeapTuple func_tuple) +{ + Form_pg_proc funcform = (Form_pg_proc) GETSTRUCT(func_tuple); + int nargsprovided = list_length(args); + Node *argarray[FUNC_MAX_ARGS]; + ListCell *lc; + int i; + + Assert(nargsprovided <= pronargs); + if (pronargs < 0 || pronargs > FUNC_MAX_ARGS) + elog(ERROR, "too many function arguments"); + memset(argarray, 0, pronargs * sizeof(Node *)); + + /* Deconstruct the argument list into an array indexed by argnumber */ + i = 0; + foreach(lc, args) + { + Node *arg = (Node *) lfirst(lc); + + if (!IsA(arg, NamedArgExpr)) + { + /* positional argument, assumed to precede all named args */ + Assert(argarray[i] == NULL); + argarray[i++] = arg; + } + else + { + NamedArgExpr *na = (NamedArgExpr *) arg; + + Assert(na->argnumber >= 0 && na->argnumber < pronargs); + Assert(argarray[na->argnumber] == NULL); + argarray[na->argnumber] = (Node *) na->arg; + } + } + + /* + * Fetch default expressions, if needed, and insert into array at proper + * locations (they aren't necessarily consecutive or all used) + */ + if (nargsprovided < pronargs) + { + List *defaults = fetch_function_defaults(func_tuple); + + i = pronargs - funcform->pronargdefaults; + foreach(lc, defaults) + { + if (argarray[i] == NULL) + argarray[i] = (Node *) lfirst(lc); + i++; + } + } + + /* Now reconstruct the args list in proper order */ + args = NIL; + for (i = 0; i < pronargs; i++) + { + Assert(argarray[i] != NULL); + args = lappend(args, argarray[i]); + } + + return args; +} + +/* + * add_function_defaults: add missing function arguments from its defaults + * + * This is used only when the argument list was positional to begin with, + * and so we know we just need to add defaults at the end. + */ +static List * +add_function_defaults(List *args, int pronargs, HeapTuple func_tuple) +{ + int nargsprovided = list_length(args); + List *defaults; + int ndelete; + + /* Get all the default expressions from the pg_proc tuple */ + defaults = fetch_function_defaults(func_tuple); + + /* Delete any unused defaults from the list */ + ndelete = nargsprovided + list_length(defaults) - pronargs; + if (ndelete < 0) + elog(ERROR, "not enough default arguments"); + if (ndelete > 0) + defaults = list_delete_first_n(defaults, ndelete); + + /* And form the combined argument list, not modifying the input list */ + return list_concat_copy(args, defaults); +} + +/* + * fetch_function_defaults: get function's default arguments as expression list + */ +static List * +fetch_function_defaults(HeapTuple func_tuple) +{ + List *defaults; + Datum proargdefaults; + bool isnull; + char *str; + + /* The error cases here shouldn't happen, but check anyway */ + proargdefaults = SysCacheGetAttr(PROCOID, func_tuple, + Anum_pg_proc_proargdefaults, + &isnull); + if (isnull) + elog(ERROR, "not enough default arguments"); + str = TextDatumGetCString(proargdefaults); + defaults = castNode(List, stringToNode(str)); + pfree(str); + return defaults; +} + +/* + * recheck_cast_function_args: recheck function args and typecast as needed + * after adding defaults. + * + * It is possible for some of the defaulted arguments to be polymorphic; + * therefore we can't assume that the default expressions have the correct + * data types already. We have to re-resolve polymorphics and do coercion + * just like the parser did. + * + * This should be a no-op if there are no polymorphic arguments, + * but we do it anyway to be sure. + * + * Note: if any casts are needed, the args list is modified in-place; + * caller should have already copied the list structure. + */ +static void +recheck_cast_function_args(List *args, Oid result_type, + Oid *proargtypes, int pronargs, + HeapTuple func_tuple) +{ + Form_pg_proc funcform = (Form_pg_proc) GETSTRUCT(func_tuple); + int nargs; + Oid actual_arg_types[FUNC_MAX_ARGS]; + Oid declared_arg_types[FUNC_MAX_ARGS]; + Oid rettype; + ListCell *lc; + + if (list_length(args) > FUNC_MAX_ARGS) + elog(ERROR, "too many function arguments"); + nargs = 0; + foreach(lc, args) + { + actual_arg_types[nargs++] = exprType((Node *) lfirst(lc)); + } + Assert(nargs == pronargs); + memcpy(declared_arg_types, proargtypes, pronargs * sizeof(Oid)); + rettype = enforce_generic_type_consistency(actual_arg_types, + declared_arg_types, + nargs, + funcform->prorettype, + false); + /* let's just check we got the same answer as the parser did ... */ + if (rettype != result_type) + elog(ERROR, "function's resolved result type changed during planning"); + + /* perform any necessary typecasting of arguments */ + make_fn_arguments(NULL, args, actual_arg_types, declared_arg_types); +} + +/* + * evaluate_function: try to pre-evaluate a function call + * + * We can do this if the function is strict and has any constant-null inputs + * (just return a null constant), or if the function is immutable and has all + * constant inputs (call it and return the result as a Const node). In + * estimation mode we are willing to pre-evaluate stable functions too. + * + * Returns a simplified expression if successful, or NULL if cannot + * simplify the function. + */ +static Expr * +evaluate_function(Oid funcid, Oid result_type, int32 result_typmod, + Oid result_collid, Oid input_collid, List *args, + bool funcvariadic, + HeapTuple func_tuple, + eval_const_expressions_context *context) +{ + Form_pg_proc funcform = (Form_pg_proc) GETSTRUCT(func_tuple); + bool has_nonconst_input = false; + bool has_null_input = false; + ListCell *arg; + FuncExpr *newexpr; + + /* + * Can't simplify if it returns a set. + */ + if (funcform->proretset) + return NULL; + + /* + * Can't simplify if it returns RECORD. The immediate problem is that it + * will be needing an expected tupdesc which we can't supply here. + * + * In the case where it has OUT parameters, it could get by without an + * expected tupdesc, but we still have issues: get_expr_result_type() + * doesn't know how to extract type info from a RECORD constant, and in + * the case of a NULL function result there doesn't seem to be any clean + * way to fix that. In view of the likelihood of there being still other + * gotchas, seems best to leave the function call unreduced. + */ + if (funcform->prorettype == RECORDOID) + return NULL; + + /* + * Check for constant inputs and especially constant-NULL inputs. + */ + foreach(arg, args) + { + if (IsA(lfirst(arg), Const)) + has_null_input |= ((Const *) lfirst(arg))->constisnull; + else + has_nonconst_input = true; + } + + /* + * If the function is strict and has a constant-NULL input, it will never + * be called at all, so we can replace the call by a NULL constant, even + * if there are other inputs that aren't constant, and even if the + * function is not otherwise immutable. + */ + if (funcform->proisstrict && has_null_input) + return (Expr *) makeNullConst(result_type, result_typmod, + result_collid); + + /* + * Otherwise, can simplify only if all inputs are constants. (For a + * non-strict function, constant NULL inputs are treated the same as + * constant non-NULL inputs.) + */ + if (has_nonconst_input) + return NULL; + + /* + * Ordinarily we are only allowed to simplify immutable functions. But for + * purposes of estimation, we consider it okay to simplify functions that + * are merely stable; the risk that the result might change from planning + * time to execution time is worth taking in preference to not being able + * to estimate the value at all. + */ + if (funcform->provolatile == PROVOLATILE_IMMUTABLE) + /* okay */ ; + else if (context->estimate && funcform->provolatile == PROVOLATILE_STABLE) + /* okay */ ; + else + return NULL; + + /* + * OK, looks like we can simplify this operator/function. + * + * Build a new FuncExpr node containing the already-simplified arguments. + */ + newexpr = makeNode(FuncExpr); + newexpr->funcid = funcid; + newexpr->funcresulttype = result_type; + newexpr->funcretset = false; + newexpr->funcvariadic = funcvariadic; + newexpr->funcformat = COERCE_EXPLICIT_CALL; /* doesn't matter */ + newexpr->funccollid = result_collid; /* doesn't matter */ + newexpr->inputcollid = input_collid; + newexpr->args = args; + newexpr->location = -1; + + return evaluate_expr((Expr *) newexpr, result_type, result_typmod, + result_collid); +} + +/* + * inline_function: try to expand a function call inline + * + * If the function is a sufficiently simple SQL-language function + * (just "SELECT expression"), then we can inline it and avoid the rather + * high per-call overhead of SQL functions. Furthermore, this can expose + * opportunities for constant-folding within the function expression. + * + * We have to beware of some special cases however. A directly or + * indirectly recursive function would cause us to recurse forever, + * so we keep track of which functions we are already expanding and + * do not re-expand them. Also, if a parameter is used more than once + * in the SQL-function body, we require it not to contain any volatile + * functions (volatiles might deliver inconsistent answers) nor to be + * unreasonably expensive to evaluate. The expensiveness check not only + * prevents us from doing multiple evaluations of an expensive parameter + * at runtime, but is a safety value to limit growth of an expression due + * to repeated inlining. + * + * We must also beware of changing the volatility or strictness status of + * functions by inlining them. + * + * Also, at the moment we can't inline functions returning RECORD. This + * doesn't work in the general case because it discards information such + * as OUT-parameter declarations. + * + * Also, context-dependent expression nodes in the argument list are trouble. + * + * Returns a simplified expression if successful, or NULL if cannot + * simplify the function. + */ +static Expr * +inline_function(Oid funcid, Oid result_type, Oid result_collid, + Oid input_collid, List *args, + bool funcvariadic, + HeapTuple func_tuple, + eval_const_expressions_context *context) +{ + Form_pg_proc funcform = (Form_pg_proc) GETSTRUCT(func_tuple); + char *src; + Datum tmp; + bool isNull; + MemoryContext oldcxt; + MemoryContext mycxt; + inline_error_callback_arg callback_arg; + ErrorContextCallback sqlerrcontext; + FuncExpr *fexpr; + SQLFunctionParseInfoPtr pinfo; + TupleDesc rettupdesc; + ParseState *pstate; + List *raw_parsetree_list; + List *querytree_list; + Query *querytree; + Node *newexpr; + int *usecounts; + ListCell *arg; + int i; + + /* + * Forget it if the function is not SQL-language or has other showstopper + * properties. (The prokind and nargs checks are just paranoia.) + */ + if (funcform->prolang != SQLlanguageId || + funcform->prokind != PROKIND_FUNCTION || + funcform->prosecdef || + funcform->proretset || + funcform->prorettype == RECORDOID || + !heap_attisnull(func_tuple, Anum_pg_proc_proconfig, NULL) || + funcform->pronargs != list_length(args)) + return NULL; + + /* Check for recursive function, and give up trying to expand if so */ + if (list_member_oid(context->active_fns, funcid)) + return NULL; + + /* Check permission to call function (fail later, if not) */ + if (pg_proc_aclcheck(funcid, GetUserId(), ACL_EXECUTE) != ACLCHECK_OK) + return NULL; + + /* Check whether a plugin wants to hook function entry/exit */ + if (FmgrHookIsNeeded(funcid)) + return NULL; + + /* + * Make a temporary memory context, so that we don't leak all the stuff + * that parsing might create. + */ + mycxt = AllocSetContextCreate(CurrentMemoryContext, + "inline_function", + ALLOCSET_DEFAULT_SIZES); + oldcxt = MemoryContextSwitchTo(mycxt); + + /* + * We need a dummy FuncExpr node containing the already-simplified + * arguments. (In some cases we don't really need it, but building it is + * cheap enough that it's not worth contortions to avoid.) + */ + fexpr = makeNode(FuncExpr); + fexpr->funcid = funcid; + fexpr->funcresulttype = result_type; + fexpr->funcretset = false; + fexpr->funcvariadic = funcvariadic; + fexpr->funcformat = COERCE_EXPLICIT_CALL; /* doesn't matter */ + fexpr->funccollid = result_collid; /* doesn't matter */ + fexpr->inputcollid = input_collid; + fexpr->args = args; + fexpr->location = -1; + + /* Fetch the function body */ + tmp = SysCacheGetAttr(PROCOID, + func_tuple, + Anum_pg_proc_prosrc, + &isNull); + if (isNull) + elog(ERROR, "null prosrc for function %u", funcid); + src = TextDatumGetCString(tmp); + + /* + * Setup error traceback support for ereport(). This is so that we can + * finger the function that bad information came from. + */ + callback_arg.proname = NameStr(funcform->proname); + callback_arg.prosrc = src; + + sqlerrcontext.callback = sql_inline_error_callback; + sqlerrcontext.arg = (void *) &callback_arg; + sqlerrcontext.previous = error_context_stack; + error_context_stack = &sqlerrcontext; + + /* If we have prosqlbody, pay attention to that not prosrc */ + tmp = SysCacheGetAttr(PROCOID, + func_tuple, + Anum_pg_proc_prosqlbody, + &isNull); + if (!isNull) + { + Node *n; + List *querytree_list; + + n = stringToNode(TextDatumGetCString(tmp)); + if (IsA(n, List)) + querytree_list = linitial_node(List, castNode(List, n)); + else + querytree_list = list_make1(n); + if (list_length(querytree_list) != 1) + goto fail; + querytree = linitial(querytree_list); + + /* + * Because we'll insist below that the querytree have an empty rtable + * and no sublinks, it cannot have any relation references that need + * to be locked or rewritten. So we can omit those steps. + */ + } + else + { + /* Set up to handle parameters while parsing the function body. */ + pinfo = prepare_sql_fn_parse_info(func_tuple, + (Node *) fexpr, + input_collid); + + /* + * We just do parsing and parse analysis, not rewriting, because + * rewriting will not affect table-free-SELECT-only queries, which is + * all that we care about. Also, we can punt as soon as we detect + * more than one command in the function body. + */ + raw_parsetree_list = pg_parse_query(src); + if (list_length(raw_parsetree_list) != 1) + goto fail; + + pstate = make_parsestate(NULL); + pstate->p_sourcetext = src; + sql_fn_parser_setup(pstate, pinfo); + + querytree = transformTopLevelStmt(pstate, linitial(raw_parsetree_list)); + + free_parsestate(pstate); + } + + /* + * The single command must be a simple "SELECT expression". + * + * Note: if you change the tests involved in this, see also plpgsql's + * exec_simple_check_plan(). That generally needs to have the same idea + * of what's a "simple expression", so that inlining a function that + * previously wasn't inlined won't change plpgsql's conclusion. + */ + if (!IsA(querytree, Query) || + querytree->commandType != CMD_SELECT || + querytree->hasAggs || + querytree->hasWindowFuncs || + querytree->hasTargetSRFs || + querytree->hasSubLinks || + querytree->cteList || + querytree->rtable || + querytree->jointree->fromlist || + querytree->jointree->quals || + querytree->groupClause || + querytree->groupingSets || + querytree->havingQual || + querytree->windowClause || + querytree->distinctClause || + querytree->sortClause || + querytree->limitOffset || + querytree->limitCount || + querytree->setOperations || + list_length(querytree->targetList) != 1) + goto fail; + + /* If the function result is composite, resolve it */ + (void) get_expr_result_type((Node *) fexpr, + NULL, + &rettupdesc); + + /* + * Make sure the function (still) returns what it's declared to. This + * will raise an error if wrong, but that's okay since the function would + * fail at runtime anyway. Note that check_sql_fn_retval will also insert + * a coercion if needed to make the tlist expression match the declared + * type of the function. + * + * Note: we do not try this until we have verified that no rewriting was + * needed; that's probably not important, but let's be careful. + */ + querytree_list = list_make1(querytree); + if (check_sql_fn_retval(list_make1(querytree_list), + result_type, rettupdesc, + false, NULL)) + goto fail; /* reject whole-tuple-result cases */ + + /* + * Given the tests above, check_sql_fn_retval shouldn't have decided to + * inject a projection step, but let's just make sure. + */ + if (querytree != linitial(querytree_list)) + goto fail; + + /* Now we can grab the tlist expression */ + newexpr = (Node *) ((TargetEntry *) linitial(querytree->targetList))->expr; + + /* + * If the SQL function returns VOID, we can only inline it if it is a + * SELECT of an expression returning VOID (ie, it's just a redirection to + * another VOID-returning function). In all non-VOID-returning cases, + * check_sql_fn_retval should ensure that newexpr returns the function's + * declared result type, so this test shouldn't fail otherwise; but we may + * as well cope gracefully if it does. + */ + if (exprType(newexpr) != result_type) + goto fail; + + /* + * Additional validity checks on the expression. It mustn't be more + * volatile than the surrounding function (this is to avoid breaking hacks + * that involve pretending a function is immutable when it really ain't). + * If the surrounding function is declared strict, then the expression + * must contain only strict constructs and must use all of the function + * parameters (this is overkill, but an exact analysis is hard). + */ + if (funcform->provolatile == PROVOLATILE_IMMUTABLE && + contain_mutable_functions(newexpr)) + goto fail; + else if (funcform->provolatile == PROVOLATILE_STABLE && + contain_volatile_functions(newexpr)) + goto fail; + + if (funcform->proisstrict && + contain_nonstrict_functions(newexpr)) + goto fail; + + /* + * If any parameter expression contains a context-dependent node, we can't + * inline, for fear of putting such a node into the wrong context. + */ + if (contain_context_dependent_node((Node *) args)) + goto fail; + + /* + * We may be able to do it; there are still checks on parameter usage to + * make, but those are most easily done in combination with the actual + * substitution of the inputs. So start building expression with inputs + * substituted. + */ + usecounts = (int *) palloc0(funcform->pronargs * sizeof(int)); + newexpr = substitute_actual_parameters(newexpr, funcform->pronargs, + args, usecounts); + + /* Now check for parameter usage */ + i = 0; + foreach(arg, args) + { + Node *param = lfirst(arg); + + if (usecounts[i] == 0) + { + /* Param not used at all: uncool if func is strict */ + if (funcform->proisstrict) + goto fail; + } + else if (usecounts[i] != 1) + { + /* Param used multiple times: uncool if expensive or volatile */ + QualCost eval_cost; + + /* + * We define "expensive" as "contains any subplan or more than 10 + * operators". Note that the subplan search has to be done + * explicitly, since cost_qual_eval() will barf on unplanned + * subselects. + */ + if (contain_subplans(param)) + goto fail; + cost_qual_eval(&eval_cost, list_make1(param), NULL); + if (eval_cost.startup + eval_cost.per_tuple > + 10 * cpu_operator_cost) + goto fail; + + /* + * Check volatility last since this is more expensive than the + * above tests + */ + if (contain_volatile_functions(param)) + goto fail; + } + i++; + } + + /* + * Whew --- we can make the substitution. Copy the modified expression + * out of the temporary memory context, and clean up. + */ + MemoryContextSwitchTo(oldcxt); + + newexpr = copyObject(newexpr); + + MemoryContextDelete(mycxt); + + /* + * If the result is of a collatable type, force the result to expose the + * correct collation. In most cases this does not matter, but it's + * possible that the function result is used directly as a sort key or in + * other places where we expect exprCollation() to tell the truth. + */ + if (OidIsValid(result_collid)) + { + Oid exprcoll = exprCollation(newexpr); + + if (OidIsValid(exprcoll) && exprcoll != result_collid) + { + CollateExpr *newnode = makeNode(CollateExpr); + + newnode->arg = (Expr *) newexpr; + newnode->collOid = result_collid; + newnode->location = -1; + + newexpr = (Node *) newnode; + } + } + + /* + * Since there is now no trace of the function in the plan tree, we must + * explicitly record the plan's dependency on the function. + */ + if (context->root) + record_plan_function_dependency(context->root, funcid); + + /* + * Recursively try to simplify the modified expression. Here we must add + * the current function to the context list of active functions. + */ + context->active_fns = lappend_oid(context->active_fns, funcid); + newexpr = eval_const_expressions_mutator(newexpr, context); + context->active_fns = list_delete_last(context->active_fns); + + error_context_stack = sqlerrcontext.previous; + + return (Expr *) newexpr; + + /* Here if func is not inlinable: release temp memory and return NULL */ +fail: + MemoryContextSwitchTo(oldcxt); + MemoryContextDelete(mycxt); + error_context_stack = sqlerrcontext.previous; + + return NULL; +} + +/* + * Replace Param nodes by appropriate actual parameters + */ +static Node * +substitute_actual_parameters(Node *expr, int nargs, List *args, + int *usecounts) +{ + substitute_actual_parameters_context context; + + context.nargs = nargs; + context.args = args; + context.usecounts = usecounts; + + return substitute_actual_parameters_mutator(expr, &context); +} + +static Node * +substitute_actual_parameters_mutator(Node *node, + substitute_actual_parameters_context *context) +{ + if (node == NULL) + return NULL; + if (IsA(node, Param)) + { + Param *param = (Param *) node; + + if (param->paramkind != PARAM_EXTERN) + elog(ERROR, "unexpected paramkind: %d", (int) param->paramkind); + if (param->paramid <= 0 || param->paramid > context->nargs) + elog(ERROR, "invalid paramid: %d", param->paramid); + + /* Count usage of parameter */ + context->usecounts[param->paramid - 1]++; + + /* Select the appropriate actual arg and replace the Param with it */ + /* We don't need to copy at this time (it'll get done later) */ + return list_nth(context->args, param->paramid - 1); + } + return expression_tree_mutator(node, substitute_actual_parameters_mutator, + (void *) context); +} + +/* + * error context callback to let us supply a call-stack traceback + */ +static void +sql_inline_error_callback(void *arg) +{ + inline_error_callback_arg *callback_arg = (inline_error_callback_arg *) arg; + int syntaxerrposition; + + /* If it's a syntax error, convert to internal syntax error report */ + syntaxerrposition = geterrposition(); + if (syntaxerrposition > 0) + { + errposition(0); + internalerrposition(syntaxerrposition); + internalerrquery(callback_arg->prosrc); + } + + errcontext("SQL function \"%s\" during inlining", callback_arg->proname); +} + +/* + * evaluate_expr: pre-evaluate a constant expression + * + * We use the executor's routine ExecEvalExpr() to avoid duplication of + * code and ensure we get the same result as the executor would get. + */ +Expr * +evaluate_expr(Expr *expr, Oid result_type, int32 result_typmod, + Oid result_collation) +{ + EState *estate; + ExprState *exprstate; + MemoryContext oldcontext; + Datum const_val; + bool const_is_null; + int16 resultTypLen; + bool resultTypByVal; + + /* + * To use the executor, we need an EState. + */ + estate = CreateExecutorState(); + + /* We can use the estate's working context to avoid memory leaks. */ + oldcontext = MemoryContextSwitchTo(estate->es_query_cxt); + + /* Make sure any opfuncids are filled in. */ + fix_opfuncids((Node *) expr); + + /* + * Prepare expr for execution. (Note: we can't use ExecPrepareExpr + * because it'd result in recursively invoking eval_const_expressions.) + */ + exprstate = ExecInitExpr(expr, NULL); + + /* + * And evaluate it. + * + * It is OK to use a default econtext because none of the ExecEvalExpr() + * code used in this situation will use econtext. That might seem + * fortuitous, but it's not so unreasonable --- a constant expression does + * not depend on context, by definition, n'est ce pas? + */ + const_val = ExecEvalExprSwitchContext(exprstate, + GetPerTupleExprContext(estate), + &const_is_null); + + /* Get info needed about result datatype */ + get_typlenbyval(result_type, &resultTypLen, &resultTypByVal); + + /* Get back to outer memory context */ + MemoryContextSwitchTo(oldcontext); + + /* + * Must copy result out of sub-context used by expression eval. + * + * Also, if it's varlena, forcibly detoast it. This protects us against + * storing TOAST pointers into plans that might outlive the referenced + * data. (makeConst would handle detoasting anyway, but it's worth a few + * extra lines here so that we can do the copy and detoast in one step.) + */ + if (!const_is_null) + { + if (resultTypLen == -1) + const_val = PointerGetDatum(PG_DETOAST_DATUM_COPY(const_val)); + else + const_val = datumCopy(const_val, resultTypByVal, resultTypLen); + } + + /* Release all the junk we just created */ + FreeExecutorState(estate); + + /* + * Make the constant result node. + */ + return (Expr *) makeConst(result_type, result_typmod, result_collation, + resultTypLen, + const_val, const_is_null, + resultTypByVal); +} + + +/* + * inline_set_returning_function + * Attempt to "inline" a set-returning function in the FROM clause. + * + * "rte" is an RTE_FUNCTION rangetable entry. If it represents a call of a + * set-returning SQL function that can safely be inlined, expand the function + * and return the substitute Query structure. Otherwise, return NULL. + * + * We assume that the RTE's expression has already been put through + * eval_const_expressions(), which among other things will take care of + * default arguments and named-argument notation. + * + * This has a good deal of similarity to inline_function(), but that's + * for the non-set-returning case, and there are enough differences to + * justify separate functions. + */ +Query * +inline_set_returning_function(PlannerInfo *root, RangeTblEntry *rte) +{ + RangeTblFunction *rtfunc; + FuncExpr *fexpr; + Oid func_oid; + HeapTuple func_tuple; + Form_pg_proc funcform; + char *src; + Datum tmp; + bool isNull; + MemoryContext oldcxt; + MemoryContext mycxt; + inline_error_callback_arg callback_arg; + ErrorContextCallback sqlerrcontext; + SQLFunctionParseInfoPtr pinfo; + TypeFuncClass functypclass; + TupleDesc rettupdesc; + List *raw_parsetree_list; + List *querytree_list; + Query *querytree; + + Assert(rte->rtekind == RTE_FUNCTION); + + /* + * It doesn't make a lot of sense for a SQL SRF to refer to itself in its + * own FROM clause, since that must cause infinite recursion at runtime. + * It will cause this code to recurse too, so check for stack overflow. + * (There's no need to do more.) + */ + check_stack_depth(); + + /* Fail if the RTE has ORDINALITY - we don't implement that here. */ + if (rte->funcordinality) + return NULL; + + /* Fail if RTE isn't a single, simple FuncExpr */ + if (list_length(rte->functions) != 1) + return NULL; + rtfunc = (RangeTblFunction *) linitial(rte->functions); + + if (!IsA(rtfunc->funcexpr, FuncExpr)) + return NULL; + fexpr = (FuncExpr *) rtfunc->funcexpr; + + func_oid = fexpr->funcid; + + /* + * The function must be declared to return a set, else inlining would + * change the results if the contained SELECT didn't return exactly one + * row. + */ + if (!fexpr->funcretset) + return NULL; + + /* + * Refuse to inline if the arguments contain any volatile functions or + * sub-selects. Volatile functions are rejected because inlining may + * result in the arguments being evaluated multiple times, risking a + * change in behavior. Sub-selects are rejected partly for implementation + * reasons (pushing them down another level might change their behavior) + * and partly because they're likely to be expensive and so multiple + * evaluation would be bad. + */ + if (contain_volatile_functions((Node *) fexpr->args) || + contain_subplans((Node *) fexpr->args)) + return NULL; + + /* Check permission to call function (fail later, if not) */ + if (pg_proc_aclcheck(func_oid, GetUserId(), ACL_EXECUTE) != ACLCHECK_OK) + return NULL; + + /* Check whether a plugin wants to hook function entry/exit */ + if (FmgrHookIsNeeded(func_oid)) + return NULL; + + /* + * OK, let's take a look at the function's pg_proc entry. + */ + func_tuple = SearchSysCache1(PROCOID, ObjectIdGetDatum(func_oid)); + if (!HeapTupleIsValid(func_tuple)) + elog(ERROR, "cache lookup failed for function %u", func_oid); + funcform = (Form_pg_proc) GETSTRUCT(func_tuple); + + /* + * Forget it if the function is not SQL-language or has other showstopper + * properties. In particular it mustn't be declared STRICT, since we + * couldn't enforce that. It also mustn't be VOLATILE, because that is + * supposed to cause it to be executed with its own snapshot, rather than + * sharing the snapshot of the calling query. We also disallow returning + * SETOF VOID, because inlining would result in exposing the actual result + * of the function's last SELECT, which should not happen in that case. + * (Rechecking prokind, proretset, and pronargs is just paranoia.) + */ + if (funcform->prolang != SQLlanguageId || + funcform->prokind != PROKIND_FUNCTION || + funcform->proisstrict || + funcform->provolatile == PROVOLATILE_VOLATILE || + funcform->prorettype == VOIDOID || + funcform->prosecdef || + !funcform->proretset || + list_length(fexpr->args) != funcform->pronargs || + !heap_attisnull(func_tuple, Anum_pg_proc_proconfig, NULL)) + { + ReleaseSysCache(func_tuple); + return NULL; + } + + /* + * Make a temporary memory context, so that we don't leak all the stuff + * that parsing might create. + */ + mycxt = AllocSetContextCreate(CurrentMemoryContext, + "inline_set_returning_function", + ALLOCSET_DEFAULT_SIZES); + oldcxt = MemoryContextSwitchTo(mycxt); + + /* Fetch the function body */ + tmp = SysCacheGetAttr(PROCOID, + func_tuple, + Anum_pg_proc_prosrc, + &isNull); + if (isNull) + elog(ERROR, "null prosrc for function %u", func_oid); + src = TextDatumGetCString(tmp); + + /* + * Setup error traceback support for ereport(). This is so that we can + * finger the function that bad information came from. + */ + callback_arg.proname = NameStr(funcform->proname); + callback_arg.prosrc = src; + + sqlerrcontext.callback = sql_inline_error_callback; + sqlerrcontext.arg = (void *) &callback_arg; + sqlerrcontext.previous = error_context_stack; + error_context_stack = &sqlerrcontext; + + /* If we have prosqlbody, pay attention to that not prosrc */ + tmp = SysCacheGetAttr(PROCOID, + func_tuple, + Anum_pg_proc_prosqlbody, + &isNull); + if (!isNull) + { + Node *n; + + n = stringToNode(TextDatumGetCString(tmp)); + if (IsA(n, List)) + querytree_list = linitial_node(List, castNode(List, n)); + else + querytree_list = list_make1(n); + if (list_length(querytree_list) != 1) + goto fail; + querytree = linitial(querytree_list); + + /* Acquire necessary locks, then apply rewriter. */ + AcquireRewriteLocks(querytree, true, false); + querytree_list = pg_rewrite_query(querytree); + if (list_length(querytree_list) != 1) + goto fail; + querytree = linitial(querytree_list); + } + else + { + /* + * Set up to handle parameters while parsing the function body. We + * can use the FuncExpr just created as the input for + * prepare_sql_fn_parse_info. + */ + pinfo = prepare_sql_fn_parse_info(func_tuple, + (Node *) fexpr, + fexpr->inputcollid); + + /* + * Parse, analyze, and rewrite (unlike inline_function(), we can't + * skip rewriting here). We can fail as soon as we find more than one + * query, though. + */ + raw_parsetree_list = pg_parse_query(src); + if (list_length(raw_parsetree_list) != 1) + goto fail; + + querytree_list = pg_analyze_and_rewrite_withcb(linitial(raw_parsetree_list), + src, + (ParserSetupHook) sql_fn_parser_setup, + pinfo, NULL); + if (list_length(querytree_list) != 1) + goto fail; + querytree = linitial(querytree_list); + } + + /* + * Also resolve the actual function result tupdesc, if composite. If the + * function is just declared to return RECORD, dig the info out of the AS + * clause. + */ + functypclass = get_expr_result_type((Node *) fexpr, NULL, &rettupdesc); + if (functypclass == TYPEFUNC_RECORD) + rettupdesc = BuildDescFromLists(rtfunc->funccolnames, + rtfunc->funccoltypes, + rtfunc->funccoltypmods, + rtfunc->funccolcollations); + + /* + * The single command must be a plain SELECT. + */ + if (!IsA(querytree, Query) || + querytree->commandType != CMD_SELECT) + goto fail; + + /* + * Make sure the function (still) returns what it's declared to. This + * will raise an error if wrong, but that's okay since the function would + * fail at runtime anyway. Note that check_sql_fn_retval will also insert + * coercions if needed to make the tlist expression(s) match the declared + * type of the function. We also ask it to insert dummy NULL columns for + * any dropped columns in rettupdesc, so that the elements of the modified + * tlist match up to the attribute numbers. + * + * If the function returns a composite type, don't inline unless the check + * shows it's returning a whole tuple result; otherwise what it's + * returning is a single composite column which is not what we need. + */ + if (!check_sql_fn_retval(list_make1(querytree_list), + fexpr->funcresulttype, rettupdesc, + true, NULL) && + (functypclass == TYPEFUNC_COMPOSITE || + functypclass == TYPEFUNC_COMPOSITE_DOMAIN || + functypclass == TYPEFUNC_RECORD)) + goto fail; /* reject not-whole-tuple-result cases */ + + /* + * check_sql_fn_retval might've inserted a projection step, but that's + * fine; just make sure we use the upper Query. + */ + querytree = linitial_node(Query, querytree_list); + + /* + * Looks good --- substitute parameters into the query. + */ + querytree = substitute_actual_srf_parameters(querytree, + funcform->pronargs, + fexpr->args); + + /* + * Copy the modified query out of the temporary memory context, and clean + * up. + */ + MemoryContextSwitchTo(oldcxt); + + querytree = copyObject(querytree); + + MemoryContextDelete(mycxt); + error_context_stack = sqlerrcontext.previous; + ReleaseSysCache(func_tuple); + + /* + * We don't have to fix collations here because the upper query is already + * parsed, ie, the collations in the RTE are what count. + */ + + /* + * Since there is now no trace of the function in the plan tree, we must + * explicitly record the plan's dependency on the function. + */ + record_plan_function_dependency(root, func_oid); + + /* + * We must also notice if the inserted query adds a dependency on the + * calling role due to RLS quals. + */ + if (querytree->hasRowSecurity) + root->glob->dependsOnRole = true; + + return querytree; + + /* Here if func is not inlinable: release temp memory and return NULL */ +fail: + MemoryContextSwitchTo(oldcxt); + MemoryContextDelete(mycxt); + error_context_stack = sqlerrcontext.previous; + ReleaseSysCache(func_tuple); + + return NULL; +} + +/* + * Replace Param nodes by appropriate actual parameters + * + * This is just enough different from substitute_actual_parameters() + * that it needs its own code. + */ +static Query * +substitute_actual_srf_parameters(Query *expr, int nargs, List *args) +{ + substitute_actual_srf_parameters_context context; + + context.nargs = nargs; + context.args = args; + context.sublevels_up = 1; + + return query_tree_mutator(expr, + substitute_actual_srf_parameters_mutator, + &context, + 0); +} + +static Node * +substitute_actual_srf_parameters_mutator(Node *node, + substitute_actual_srf_parameters_context *context) +{ + Node *result; + + if (node == NULL) + return NULL; + if (IsA(node, Query)) + { + context->sublevels_up++; + result = (Node *) query_tree_mutator((Query *) node, + substitute_actual_srf_parameters_mutator, + (void *) context, + 0); + context->sublevels_up--; + return result; + } + if (IsA(node, Param)) + { + Param *param = (Param *) node; + + if (param->paramkind == PARAM_EXTERN) + { + if (param->paramid <= 0 || param->paramid > context->nargs) + elog(ERROR, "invalid paramid: %d", param->paramid); + + /* + * Since the parameter is being inserted into a subquery, we must + * adjust levels. + */ + result = copyObject(list_nth(context->args, param->paramid - 1)); + IncrementVarSublevelsUp(result, context->sublevels_up, 0); + return result; + } + } + return expression_tree_mutator(node, + substitute_actual_srf_parameters_mutator, + (void *) context); +} + +/* + * pull_paramids + * Returns a Bitmapset containing the paramids of all Params in 'expr'. + */ +Bitmapset * +pull_paramids(Expr *expr) +{ + Bitmapset *result = NULL; + + (void) pull_paramids_walker((Node *) expr, &result); + + return result; +} + +static bool +pull_paramids_walker(Node *node, Bitmapset **context) +{ + if (node == NULL) + return false; + if (IsA(node, Param)) + { + Param *param = (Param *) node; + + *context = bms_add_member(*context, param->paramid); + return false; + } + return expression_tree_walker(node, pull_paramids_walker, + (void *) context); +} diff --git a/src/backend/optimizer/util/inherit.c b/src/backend/optimizer/util/inherit.c new file mode 100644 index 0000000..3c11f5d --- /dev/null +++ b/src/backend/optimizer/util/inherit.c @@ -0,0 +1,949 @@ +/*------------------------------------------------------------------------- + * + * inherit.c + * Routines to process child relations in inheritance trees + * + * Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group + * Portions Copyright (c) 1994, Regents of the University of California + * + * + * IDENTIFICATION + * src/backend/optimizer/util/inherit.c + * + *------------------------------------------------------------------------- + */ +#include "postgres.h" + +#include "access/sysattr.h" +#include "access/table.h" +#include "catalog/partition.h" +#include "catalog/pg_inherits.h" +#include "catalog/pg_type.h" +#include "miscadmin.h" +#include "nodes/makefuncs.h" +#include "optimizer/appendinfo.h" +#include "optimizer/inherit.h" +#include "optimizer/optimizer.h" +#include "optimizer/pathnode.h" +#include "optimizer/plancat.h" +#include "optimizer/planmain.h" +#include "optimizer/planner.h" +#include "optimizer/prep.h" +#include "optimizer/restrictinfo.h" +#include "parser/parsetree.h" +#include "partitioning/partdesc.h" +#include "partitioning/partprune.h" +#include "utils/rel.h" + + +static void expand_partitioned_rtentry(PlannerInfo *root, RelOptInfo *relinfo, + RangeTblEntry *parentrte, + Index parentRTindex, Relation parentrel, + PlanRowMark *top_parentrc, LOCKMODE lockmode); +static void expand_single_inheritance_child(PlannerInfo *root, + RangeTblEntry *parentrte, + Index parentRTindex, Relation parentrel, + PlanRowMark *top_parentrc, Relation childrel, + RangeTblEntry **childrte_p, + Index *childRTindex_p); +static Bitmapset *translate_col_privs(const Bitmapset *parent_privs, + List *translated_vars); +static Bitmapset *translate_col_privs_multilevel(PlannerInfo *root, + RelOptInfo *rel, + RelOptInfo *parent_rel, + Bitmapset *parent_cols); +static void expand_appendrel_subquery(PlannerInfo *root, RelOptInfo *rel, + RangeTblEntry *rte, Index rti); + + +/* + * expand_inherited_rtentry + * Expand a rangetable entry that has the "inh" bit set. + * + * "inh" is only allowed in two cases: RELATION and SUBQUERY RTEs. + * + * "inh" on a plain RELATION RTE means that it is a partitioned table or the + * parent of a traditional-inheritance set. In this case we must add entries + * for all the interesting child tables to the query's rangetable, and build + * additional planner data structures for them, including RelOptInfos, + * AppendRelInfos, and possibly PlanRowMarks. + * + * Note that the original RTE is considered to represent the whole inheritance + * set. In the case of traditional inheritance, the first of the generated + * RTEs is an RTE for the same table, but with inh = false, to represent the + * parent table in its role as a simple member of the inheritance set. For + * partitioning, we don't need a second RTE because the partitioned table + * itself has no data and need not be scanned. + * + * "inh" on a SUBQUERY RTE means that it's the parent of a UNION ALL group, + * which is treated as an appendrel similarly to inheritance cases; however, + * we already made RTEs and AppendRelInfos for the subqueries. We only need + * to build RelOptInfos for them, which is done by expand_appendrel_subquery. + */ +void +expand_inherited_rtentry(PlannerInfo *root, RelOptInfo *rel, + RangeTblEntry *rte, Index rti) +{ + Oid parentOID; + Relation oldrelation; + LOCKMODE lockmode; + PlanRowMark *oldrc; + bool old_isParent = false; + int old_allMarkTypes = 0; + + Assert(rte->inh); /* else caller error */ + + if (rte->rtekind == RTE_SUBQUERY) + { + expand_appendrel_subquery(root, rel, rte, rti); + return; + } + + Assert(rte->rtekind == RTE_RELATION); + + parentOID = rte->relid; + + /* + * We used to check has_subclass() here, but there's no longer any need + * to, because subquery_planner already did. + */ + + /* + * The rewriter should already have obtained an appropriate lock on each + * relation named in the query, so we can open the parent relation without + * locking it. However, for each child relation we add to the query, we + * must obtain an appropriate lock, because this will be the first use of + * those relations in the parse/rewrite/plan pipeline. Child rels should + * use the same lockmode as their parent. + */ + oldrelation = table_open(parentOID, NoLock); + lockmode = rte->rellockmode; + + /* + * If parent relation is selected FOR UPDATE/SHARE, we need to mark its + * PlanRowMark as isParent = true, and generate a new PlanRowMark for each + * child. + */ + oldrc = get_plan_rowmark(root->rowMarks, rti); + if (oldrc) + { + old_isParent = oldrc->isParent; + oldrc->isParent = true; + /* Save initial value of allMarkTypes before children add to it */ + old_allMarkTypes = oldrc->allMarkTypes; + } + + /* Scan the inheritance set and expand it */ + if (oldrelation->rd_rel->relkind == RELKIND_PARTITIONED_TABLE) + { + /* + * Partitioned table, so set up for partitioning. + */ + Assert(rte->relkind == RELKIND_PARTITIONED_TABLE); + + /* + * Recursively expand and lock the partitions. While at it, also + * extract the partition key columns of all the partitioned tables. + */ + expand_partitioned_rtentry(root, rel, rte, rti, + oldrelation, oldrc, lockmode); + } + else + { + /* + * Ordinary table, so process traditional-inheritance children. (Note + * that partitioned tables are not allowed to have inheritance + * children, so it's not possible for both cases to apply.) + */ + List *inhOIDs; + ListCell *l; + + /* Scan for all members of inheritance set, acquire needed locks */ + inhOIDs = find_all_inheritors(parentOID, lockmode, NULL); + + /* + * We used to special-case the situation where the table no longer has + * any children, by clearing rte->inh and exiting. That no longer + * works, because this function doesn't get run until after decisions + * have been made that depend on rte->inh. We have to treat such + * situations as normal inheritance. The table itself should always + * have been found, though. + */ + Assert(inhOIDs != NIL); + Assert(linitial_oid(inhOIDs) == parentOID); + + /* Expand simple_rel_array and friends to hold child objects. */ + expand_planner_arrays(root, list_length(inhOIDs)); + + /* + * Expand inheritance children in the order the OIDs were returned by + * find_all_inheritors. + */ + foreach(l, inhOIDs) + { + Oid childOID = lfirst_oid(l); + Relation newrelation; + RangeTblEntry *childrte; + Index childRTindex; + + /* Open rel if needed; we already have required locks */ + if (childOID != parentOID) + newrelation = table_open(childOID, NoLock); + else + newrelation = oldrelation; + + /* + * It is possible that the parent table has children that are temp + * tables of other backends. We cannot safely access such tables + * (because of buffering issues), and the best thing to do seems + * to be to silently ignore them. + */ + if (childOID != parentOID && RELATION_IS_OTHER_TEMP(newrelation)) + { + table_close(newrelation, lockmode); + continue; + } + + /* Create RTE and AppendRelInfo, plus PlanRowMark if needed. */ + expand_single_inheritance_child(root, rte, rti, oldrelation, + oldrc, newrelation, + &childrte, &childRTindex); + + /* Create the otherrel RelOptInfo too. */ + (void) build_simple_rel(root, childRTindex, rel); + + /* Close child relations, but keep locks */ + if (childOID != parentOID) + table_close(newrelation, NoLock); + } + } + + /* + * Some children might require different mark types, which would've been + * reported into oldrc. If so, add relevant entries to the top-level + * targetlist and update parent rel's reltarget. This should match what + * preprocess_targetlist() would have added if the mark types had been + * requested originally. + * + * (Someday it might be useful to fold these resjunk columns into the + * row-identity-column management used for UPDATE/DELETE. Today is not + * that day, however.) + */ + if (oldrc) + { + int new_allMarkTypes = oldrc->allMarkTypes; + Var *var; + TargetEntry *tle; + char resname[32]; + List *newvars = NIL; + + /* Add TID junk Var if needed, unless we had it already */ + if (new_allMarkTypes & ~(1 << ROW_MARK_COPY) && + !(old_allMarkTypes & ~(1 << ROW_MARK_COPY))) + { + /* Need to fetch TID */ + var = makeVar(oldrc->rti, + SelfItemPointerAttributeNumber, + TIDOID, + -1, + InvalidOid, + 0); + snprintf(resname, sizeof(resname), "ctid%u", oldrc->rowmarkId); + tle = makeTargetEntry((Expr *) var, + list_length(root->processed_tlist) + 1, + pstrdup(resname), + true); + root->processed_tlist = lappend(root->processed_tlist, tle); + newvars = lappend(newvars, var); + } + + /* Add whole-row junk Var if needed, unless we had it already */ + if ((new_allMarkTypes & (1 << ROW_MARK_COPY)) && + !(old_allMarkTypes & (1 << ROW_MARK_COPY))) + { + var = makeWholeRowVar(planner_rt_fetch(oldrc->rti, root), + oldrc->rti, + 0, + false); + snprintf(resname, sizeof(resname), "wholerow%u", oldrc->rowmarkId); + tle = makeTargetEntry((Expr *) var, + list_length(root->processed_tlist) + 1, + pstrdup(resname), + true); + root->processed_tlist = lappend(root->processed_tlist, tle); + newvars = lappend(newvars, var); + } + + /* Add tableoid junk Var, unless we had it already */ + if (!old_isParent) + { + var = makeVar(oldrc->rti, + TableOidAttributeNumber, + OIDOID, + -1, + InvalidOid, + 0); + snprintf(resname, sizeof(resname), "tableoid%u", oldrc->rowmarkId); + tle = makeTargetEntry((Expr *) var, + list_length(root->processed_tlist) + 1, + pstrdup(resname), + true); + root->processed_tlist = lappend(root->processed_tlist, tle); + newvars = lappend(newvars, var); + } + + /* + * Add the newly added Vars to parent's reltarget. We needn't worry + * about the children's reltargets, they'll be made later. + */ + add_vars_to_targetlist(root, newvars, bms_make_singleton(0), false); + } + + table_close(oldrelation, NoLock); +} + +/* + * expand_partitioned_rtentry + * Recursively expand an RTE for a partitioned table. + */ +static void +expand_partitioned_rtentry(PlannerInfo *root, RelOptInfo *relinfo, + RangeTblEntry *parentrte, + Index parentRTindex, Relation parentrel, + PlanRowMark *top_parentrc, LOCKMODE lockmode) +{ + PartitionDesc partdesc; + Bitmapset *live_parts; + int num_live_parts; + int i; + + check_stack_depth(); + + Assert(parentrte->inh); + + partdesc = PartitionDirectoryLookup(root->glob->partition_directory, + parentrel); + + /* A partitioned table should always have a partition descriptor. */ + Assert(partdesc); + + /* + * Note down whether any partition key cols are being updated. Though it's + * the root partitioned table's updatedCols we are interested in, we + * instead use parentrte to get the updatedCols. This is convenient + * because parentrte already has the root partrel's updatedCols translated + * to match the attribute ordering of parentrel. + */ + if (!root->partColsUpdated) + root->partColsUpdated = + has_partition_attrs(parentrel, parentrte->updatedCols, NULL); + + /* Nothing further to do here if there are no partitions. */ + if (partdesc->nparts == 0) + return; + + /* + * Perform partition pruning using restriction clauses assigned to parent + * relation. live_parts will contain PartitionDesc indexes of partitions + * that survive pruning. Below, we will initialize child objects for the + * surviving partitions. + */ + relinfo->live_parts = live_parts = prune_append_rel_partitions(relinfo); + + /* Expand simple_rel_array and friends to hold child objects. */ + num_live_parts = bms_num_members(live_parts); + if (num_live_parts > 0) + expand_planner_arrays(root, num_live_parts); + + /* + * We also store partition RelOptInfo pointers in the parent relation. + * Since we're palloc0'ing, slots corresponding to pruned partitions will + * contain NULL. + */ + Assert(relinfo->part_rels == NULL); + relinfo->part_rels = (RelOptInfo **) + palloc0(relinfo->nparts * sizeof(RelOptInfo *)); + + /* + * Create a child RTE for each live partition. Note that unlike + * traditional inheritance, we don't need a child RTE for the partitioned + * table itself, because it's not going to be scanned. + */ + i = -1; + while ((i = bms_next_member(live_parts, i)) >= 0) + { + Oid childOID = partdesc->oids[i]; + Relation childrel; + RangeTblEntry *childrte; + Index childRTindex; + RelOptInfo *childrelinfo; + + /* Open rel, acquiring required locks */ + childrel = table_open(childOID, lockmode); + + /* + * Temporary partitions belonging to other sessions should have been + * disallowed at definition, but for paranoia's sake, let's double + * check. + */ + if (RELATION_IS_OTHER_TEMP(childrel)) + elog(ERROR, "temporary relation from another session found as partition"); + + /* Create RTE and AppendRelInfo, plus PlanRowMark if needed. */ + expand_single_inheritance_child(root, parentrte, parentRTindex, + parentrel, top_parentrc, childrel, + &childrte, &childRTindex); + + /* Create the otherrel RelOptInfo too. */ + childrelinfo = build_simple_rel(root, childRTindex, relinfo); + relinfo->part_rels[i] = childrelinfo; + relinfo->all_partrels = bms_add_members(relinfo->all_partrels, + childrelinfo->relids); + + /* If this child is itself partitioned, recurse */ + if (childrel->rd_rel->relkind == RELKIND_PARTITIONED_TABLE) + expand_partitioned_rtentry(root, childrelinfo, + childrte, childRTindex, + childrel, top_parentrc, lockmode); + + /* Close child relation, but keep locks */ + table_close(childrel, NoLock); + } +} + +/* + * expand_single_inheritance_child + * Build a RangeTblEntry and an AppendRelInfo, plus maybe a PlanRowMark. + * + * We now expand the partition hierarchy level by level, creating a + * corresponding hierarchy of AppendRelInfos and RelOptInfos, where each + * partitioned descendant acts as a parent of its immediate partitions. + * (This is a difference from what older versions of PostgreSQL did and what + * is still done in the case of table inheritance for unpartitioned tables, + * where the hierarchy is flattened during RTE expansion.) + * + * PlanRowMarks still carry the top-parent's RTI, and the top-parent's + * allMarkTypes field still accumulates values from all descendents. + * + * "parentrte" and "parentRTindex" are immediate parent's RTE and + * RTI. "top_parentrc" is top parent's PlanRowMark. + * + * The child RangeTblEntry and its RTI are returned in "childrte_p" and + * "childRTindex_p" resp. + */ +static void +expand_single_inheritance_child(PlannerInfo *root, RangeTblEntry *parentrte, + Index parentRTindex, Relation parentrel, + PlanRowMark *top_parentrc, Relation childrel, + RangeTblEntry **childrte_p, + Index *childRTindex_p) +{ + Query *parse = root->parse; + Oid parentOID = RelationGetRelid(parentrel); + Oid childOID = RelationGetRelid(childrel); + RangeTblEntry *childrte; + Index childRTindex; + AppendRelInfo *appinfo; + TupleDesc child_tupdesc; + List *parent_colnames; + List *child_colnames; + + /* + * Build an RTE for the child, and attach to query's rangetable list. We + * copy most scalar fields of the parent's RTE, but replace relation OID, + * relkind, and inh for the child. Also, set requiredPerms to zero since + * all required permissions checks are done on the original RTE. Likewise, + * set the child's securityQuals to empty, because we only want to apply + * the parent's RLS conditions regardless of what RLS properties + * individual children may have. (This is an intentional choice to make + * inherited RLS work like regular permissions checks.) The parent + * securityQuals will be propagated to children along with other base + * restriction clauses, so we don't need to do it here. Other + * infrastructure of the parent RTE has to be translated to match the + * child table's column ordering, which we do below, so a "flat" copy is + * sufficient to start with. + */ + childrte = makeNode(RangeTblEntry); + memcpy(childrte, parentrte, sizeof(RangeTblEntry)); + Assert(parentrte->rtekind == RTE_RELATION); /* else this is dubious */ + childrte->relid = childOID; + childrte->relkind = childrel->rd_rel->relkind; + /* A partitioned child will need to be expanded further. */ + if (childrte->relkind == RELKIND_PARTITIONED_TABLE) + { + Assert(childOID != parentOID); + childrte->inh = true; + } + else + childrte->inh = false; + childrte->requiredPerms = 0; + childrte->securityQuals = NIL; + + /* Link not-yet-fully-filled child RTE into data structures */ + parse->rtable = lappend(parse->rtable, childrte); + childRTindex = list_length(parse->rtable); + *childrte_p = childrte; + *childRTindex_p = childRTindex; + + /* + * Build an AppendRelInfo struct for each parent/child pair. + */ + appinfo = make_append_rel_info(parentrel, childrel, + parentRTindex, childRTindex); + root->append_rel_list = lappend(root->append_rel_list, appinfo); + + /* tablesample is probably null, but copy it */ + childrte->tablesample = copyObject(parentrte->tablesample); + + /* + * Construct an alias clause for the child, which we can also use as eref. + * This is important so that EXPLAIN will print the right column aliases + * for child-table columns. (Since ruleutils.c doesn't have any easy way + * to reassociate parent and child columns, we must get the child column + * aliases right to start with. Note that setting childrte->alias forces + * ruleutils.c to use these column names, which it otherwise would not.) + */ + child_tupdesc = RelationGetDescr(childrel); + parent_colnames = parentrte->eref->colnames; + child_colnames = NIL; + for (int cattno = 0; cattno < child_tupdesc->natts; cattno++) + { + Form_pg_attribute att = TupleDescAttr(child_tupdesc, cattno); + const char *attname; + + if (att->attisdropped) + { + /* Always insert an empty string for a dropped column */ + attname = ""; + } + else if (appinfo->parent_colnos[cattno] > 0 && + appinfo->parent_colnos[cattno] <= list_length(parent_colnames)) + { + /* Duplicate the query-assigned name for the parent column */ + attname = strVal(list_nth(parent_colnames, + appinfo->parent_colnos[cattno] - 1)); + } + else + { + /* New column, just use its real name */ + attname = NameStr(att->attname); + } + child_colnames = lappend(child_colnames, makeString(pstrdup(attname))); + } + + /* + * We just duplicate the parent's table alias name for each child. If the + * plan gets printed, ruleutils.c has to sort out unique table aliases to + * use, which it can handle. + */ + childrte->alias = childrte->eref = makeAlias(parentrte->eref->aliasname, + child_colnames); + + /* + * Translate the column permissions bitmaps to the child's attnums (we + * have to build the translated_vars list before we can do this). But if + * this is the parent table, we can just duplicate the parent's bitmaps. + * + * Note: we need to do this even though the executor won't run any + * permissions checks on the child RTE. The insertedCols/updatedCols + * bitmaps may be examined for trigger-firing purposes. + */ + if (childOID != parentOID) + { + childrte->selectedCols = translate_col_privs(parentrte->selectedCols, + appinfo->translated_vars); + childrte->insertedCols = translate_col_privs(parentrte->insertedCols, + appinfo->translated_vars); + childrte->updatedCols = translate_col_privs(parentrte->updatedCols, + appinfo->translated_vars); + } + else + { + childrte->selectedCols = bms_copy(parentrte->selectedCols); + childrte->insertedCols = bms_copy(parentrte->insertedCols); + childrte->updatedCols = bms_copy(parentrte->updatedCols); + } + + /* + * Store the RTE and appinfo in the respective PlannerInfo arrays, which + * the caller must already have allocated space for. + */ + Assert(childRTindex < root->simple_rel_array_size); + Assert(root->simple_rte_array[childRTindex] == NULL); + root->simple_rte_array[childRTindex] = childrte; + Assert(root->append_rel_array[childRTindex] == NULL); + root->append_rel_array[childRTindex] = appinfo; + + /* + * Build a PlanRowMark if parent is marked FOR UPDATE/SHARE. + */ + if (top_parentrc) + { + PlanRowMark *childrc = makeNode(PlanRowMark); + + childrc->rti = childRTindex; + childrc->prti = top_parentrc->rti; + childrc->rowmarkId = top_parentrc->rowmarkId; + /* Reselect rowmark type, because relkind might not match parent */ + childrc->markType = select_rowmark_type(childrte, + top_parentrc->strength); + childrc->allMarkTypes = (1 << childrc->markType); + childrc->strength = top_parentrc->strength; + childrc->waitPolicy = top_parentrc->waitPolicy; + + /* + * We mark RowMarks for partitioned child tables as parent RowMarks so + * that the executor ignores them (except their existence means that + * the child tables will be locked using the appropriate mode). + */ + childrc->isParent = (childrte->relkind == RELKIND_PARTITIONED_TABLE); + + /* Include child's rowmark type in top parent's allMarkTypes */ + top_parentrc->allMarkTypes |= childrc->allMarkTypes; + + root->rowMarks = lappend(root->rowMarks, childrc); + } + + /* + * If we are creating a child of the query target relation (only possible + * in UPDATE/DELETE/MERGE), add it to all_result_relids, as well as + * leaf_result_relids if appropriate, and make sure that we generate + * required row-identity data. + */ + if (bms_is_member(parentRTindex, root->all_result_relids)) + { + /* OK, record the child as a result rel too. */ + root->all_result_relids = bms_add_member(root->all_result_relids, + childRTindex); + + /* Non-leaf partitions don't need any row identity info. */ + if (childrte->relkind != RELKIND_PARTITIONED_TABLE) + { + Var *rrvar; + + root->leaf_result_relids = bms_add_member(root->leaf_result_relids, + childRTindex); + + /* + * If we have any child target relations, assume they all need to + * generate a junk "tableoid" column. (If only one child survives + * pruning, we wouldn't really need this, but it's not worth + * thrashing about to avoid it.) + */ + rrvar = makeVar(childRTindex, + TableOidAttributeNumber, + OIDOID, + -1, + InvalidOid, + 0); + add_row_identity_var(root, rrvar, childRTindex, "tableoid"); + + /* Register any row-identity columns needed by this child. */ + add_row_identity_columns(root, childRTindex, + childrte, childrel); + } + } +} + +/* + * get_rel_all_updated_cols + * Returns the set of columns of a given "simple" relation that are + * updated by this query. + */ +Bitmapset * +get_rel_all_updated_cols(PlannerInfo *root, RelOptInfo *rel) +{ + Index relid; + RangeTblEntry *rte; + Bitmapset *updatedCols, + *extraUpdatedCols; + + Assert(root->parse->commandType == CMD_UPDATE); + Assert(IS_SIMPLE_REL(rel)); + + /* + * We obtain updatedCols for the query's result relation. Then, if + * necessary, we map it to the column numbers of the relation for which + * they were requested. + */ + relid = root->parse->resultRelation; + rte = planner_rt_fetch(relid, root); + + updatedCols = rte->updatedCols; + + if (rel->relid != relid) + { + RelOptInfo *top_parent_rel = find_base_rel(root, relid); + + Assert(IS_OTHER_REL(rel)); + + updatedCols = translate_col_privs_multilevel(root, rel, top_parent_rel, + updatedCols); + } + + /* + * Now we must check to see if there are any generated columns that depend + * on the updatedCols, and add them to the result. + */ + extraUpdatedCols = get_dependent_generated_columns(root, rel->relid, + updatedCols); + + return bms_union(updatedCols, extraUpdatedCols); +} + +/* + * translate_col_privs + * Translate a bitmapset representing per-column privileges from the + * parent rel's attribute numbering to the child's. + * + * The only surprise here is that we don't translate a parent whole-row + * reference into a child whole-row reference. That would mean requiring + * permissions on all child columns, which is overly strict, since the + * query is really only going to reference the inherited columns. Instead + * we set the per-column bits for all inherited columns. + */ +static Bitmapset * +translate_col_privs(const Bitmapset *parent_privs, + List *translated_vars) +{ + Bitmapset *child_privs = NULL; + bool whole_row; + int attno; + ListCell *lc; + + /* System attributes have the same numbers in all tables */ + for (attno = FirstLowInvalidHeapAttributeNumber + 1; attno < 0; attno++) + { + if (bms_is_member(attno - FirstLowInvalidHeapAttributeNumber, + parent_privs)) + child_privs = bms_add_member(child_privs, + attno - FirstLowInvalidHeapAttributeNumber); + } + + /* Check if parent has whole-row reference */ + whole_row = bms_is_member(InvalidAttrNumber - FirstLowInvalidHeapAttributeNumber, + parent_privs); + + /* And now translate the regular user attributes, using the vars list */ + attno = InvalidAttrNumber; + foreach(lc, translated_vars) + { + Var *var = lfirst_node(Var, lc); + + attno++; + if (var == NULL) /* ignore dropped columns */ + continue; + if (whole_row || + bms_is_member(attno - FirstLowInvalidHeapAttributeNumber, + parent_privs)) + child_privs = bms_add_member(child_privs, + var->varattno - FirstLowInvalidHeapAttributeNumber); + } + + return child_privs; +} + +/* + * translate_col_privs_multilevel + * Recursively translates the column numbers contained in 'parent_cols' + * to the column numbers of a descendant relation given by 'rel' + * + * Note that because this is based on translate_col_privs, it will expand + * a whole-row reference into all inherited columns. This is not an issue + * for current usages, but beware. + */ +static Bitmapset * +translate_col_privs_multilevel(PlannerInfo *root, RelOptInfo *rel, + RelOptInfo *parent_rel, + Bitmapset *parent_cols) +{ + AppendRelInfo *appinfo; + + /* Fast path for easy case. */ + if (parent_cols == NULL) + return NULL; + + Assert(root->append_rel_array != NULL); + appinfo = root->append_rel_array[rel->relid]; + Assert(appinfo != NULL); + + /* Recurse if immediate parent is not the top parent. */ + if (appinfo->parent_relid != parent_rel->relid) + { + RelOptInfo *next_parent = find_base_rel(root, appinfo->parent_relid); + + parent_cols = translate_col_privs_multilevel(root, next_parent, + parent_rel, + parent_cols); + } + + /* Now translate for this child. */ + return translate_col_privs(parent_cols, appinfo->translated_vars); +} + +/* + * expand_appendrel_subquery + * Add "other rel" RelOptInfos for the children of an appendrel baserel + * + * "rel" is a subquery relation that has the rte->inh flag set, meaning it + * is a UNION ALL subquery that's been flattened into an appendrel, with + * child subqueries listed in root->append_rel_list. We need to build + * a RelOptInfo for each child relation so that we can plan scans on them. + */ +static void +expand_appendrel_subquery(PlannerInfo *root, RelOptInfo *rel, + RangeTblEntry *rte, Index rti) +{ + ListCell *l; + + foreach(l, root->append_rel_list) + { + AppendRelInfo *appinfo = (AppendRelInfo *) lfirst(l); + Index childRTindex = appinfo->child_relid; + RangeTblEntry *childrte; + RelOptInfo *childrel; + + /* append_rel_list contains all append rels; ignore others */ + if (appinfo->parent_relid != rti) + continue; + + /* find the child RTE, which should already exist */ + Assert(childRTindex < root->simple_rel_array_size); + childrte = root->simple_rte_array[childRTindex]; + Assert(childrte != NULL); + + /* Build the child RelOptInfo. */ + childrel = build_simple_rel(root, childRTindex, rel); + + /* Child may itself be an inherited rel, either table or subquery. */ + if (childrte->inh) + expand_inherited_rtentry(root, childrel, childrte, childRTindex); + } +} + + +/* + * apply_child_basequals + * Populate childrel's base restriction quals from parent rel's quals, + * translating them using appinfo. + * + * If any of the resulting clauses evaluate to constant false or NULL, we + * return false and don't apply any quals. Caller should mark the relation as + * a dummy rel in this case, since it doesn't need to be scanned. + */ +bool +apply_child_basequals(PlannerInfo *root, RelOptInfo *parentrel, + RelOptInfo *childrel, RangeTblEntry *childRTE, + AppendRelInfo *appinfo) +{ + List *childquals; + Index cq_min_security; + ListCell *lc; + + /* + * The child rel's targetlist might contain non-Var expressions, which + * means that substitution into the quals could produce opportunities for + * const-simplification, and perhaps even pseudoconstant quals. Therefore, + * transform each RestrictInfo separately to see if it reduces to a + * constant or pseudoconstant. (We must process them separately to keep + * track of the security level of each qual.) + */ + childquals = NIL; + cq_min_security = UINT_MAX; + foreach(lc, parentrel->baserestrictinfo) + { + RestrictInfo *rinfo = (RestrictInfo *) lfirst(lc); + Node *childqual; + ListCell *lc2; + + Assert(IsA(rinfo, RestrictInfo)); + childqual = adjust_appendrel_attrs(root, + (Node *) rinfo->clause, + 1, &appinfo); + childqual = eval_const_expressions(root, childqual); + /* check for flat-out constant */ + if (childqual && IsA(childqual, Const)) + { + if (((Const *) childqual)->constisnull || + !DatumGetBool(((Const *) childqual)->constvalue)) + { + /* Restriction reduces to constant FALSE or NULL */ + return false; + } + /* Restriction reduces to constant TRUE, so drop it */ + continue; + } + /* might have gotten an AND clause, if so flatten it */ + foreach(lc2, make_ands_implicit((Expr *) childqual)) + { + Node *onecq = (Node *) lfirst(lc2); + bool pseudoconstant; + + /* check for pseudoconstant (no Vars or volatile functions) */ + pseudoconstant = + !contain_vars_of_level(onecq, 0) && + !contain_volatile_functions(onecq); + if (pseudoconstant) + { + /* tell createplan.c to check for gating quals */ + root->hasPseudoConstantQuals = true; + } + /* reconstitute RestrictInfo with appropriate properties */ + childquals = lappend(childquals, + make_restrictinfo(root, + (Expr *) onecq, + rinfo->is_pushed_down, + rinfo->outerjoin_delayed, + pseudoconstant, + rinfo->security_level, + NULL, NULL, NULL)); + /* track minimum security level among child quals */ + cq_min_security = Min(cq_min_security, rinfo->security_level); + } + } + + /* + * In addition to the quals inherited from the parent, we might have + * securityQuals associated with this particular child node. (Currently + * this can only happen in appendrels originating from UNION ALL; + * inheritance child tables don't have their own securityQuals, see + * expand_single_inheritance_child().) Pull any such securityQuals up + * into the baserestrictinfo for the child. This is similar to + * process_security_barrier_quals() for the parent rel, except that we + * can't make any general deductions from such quals, since they don't + * hold for the whole appendrel. + */ + if (childRTE->securityQuals) + { + Index security_level = 0; + + foreach(lc, childRTE->securityQuals) + { + List *qualset = (List *) lfirst(lc); + ListCell *lc2; + + foreach(lc2, qualset) + { + Expr *qual = (Expr *) lfirst(lc2); + + /* not likely that we'd see constants here, so no check */ + childquals = lappend(childquals, + make_restrictinfo(root, qual, + true, false, false, + security_level, + NULL, NULL, NULL)); + cq_min_security = Min(cq_min_security, security_level); + } + security_level++; + } + Assert(security_level <= root->qual_security_level); + } + + /* + * OK, we've got all the baserestrictinfo quals for this child. + */ + childrel->baserestrictinfo = childquals; + childrel->baserestrict_min_security = cq_min_security; + + return true; +} diff --git a/src/backend/optimizer/util/joininfo.c b/src/backend/optimizer/util/joininfo.c new file mode 100644 index 0000000..d4cffdb --- /dev/null +++ b/src/backend/optimizer/util/joininfo.c @@ -0,0 +1,140 @@ +/*------------------------------------------------------------------------- + * + * joininfo.c + * joininfo list manipulation routines + * + * Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group + * Portions Copyright (c) 1994, Regents of the University of California + * + * + * IDENTIFICATION + * src/backend/optimizer/util/joininfo.c + * + *------------------------------------------------------------------------- + */ +#include "postgres.h" + +#include "optimizer/joininfo.h" +#include "optimizer/pathnode.h" +#include "optimizer/paths.h" + + +/* + * have_relevant_joinclause + * Detect whether there is a joinclause that involves + * the two given relations. + * + * Note: the joinclause does not have to be evaluable with only these two + * relations. This is intentional. For example consider + * SELECT * FROM a, b, c WHERE a.x = (b.y + c.z) + * If a is much larger than the other tables, it may be worthwhile to + * cross-join b and c and then use an inner indexscan on a.x. Therefore + * we should consider this joinclause as reason to join b to c, even though + * it can't be applied at that join step. + */ +bool +have_relevant_joinclause(PlannerInfo *root, + RelOptInfo *rel1, RelOptInfo *rel2) +{ + bool result = false; + List *joininfo; + Relids other_relids; + ListCell *l; + + /* + * We could scan either relation's joininfo list; may as well use the + * shorter one. + */ + if (list_length(rel1->joininfo) <= list_length(rel2->joininfo)) + { + joininfo = rel1->joininfo; + other_relids = rel2->relids; + } + else + { + joininfo = rel2->joininfo; + other_relids = rel1->relids; + } + + foreach(l, joininfo) + { + RestrictInfo *rinfo = (RestrictInfo *) lfirst(l); + + if (bms_overlap(other_relids, rinfo->required_relids)) + { + result = true; + break; + } + } + + /* + * We also need to check the EquivalenceClass data structure, which might + * contain relationships not emitted into the joininfo lists. + */ + if (!result && rel1->has_eclass_joins && rel2->has_eclass_joins) + result = have_relevant_eclass_joinclause(root, rel1, rel2); + + return result; +} + + +/* + * add_join_clause_to_rels + * Add 'restrictinfo' to the joininfo list of each relation it requires. + * + * Note that the same copy of the restrictinfo node is linked to by all the + * lists it is in. This allows us to exploit caching of information about + * the restriction clause (but we must be careful that the information does + * not depend on context). + * + * 'restrictinfo' describes the join clause + * 'join_relids' is the list of relations participating in the join clause + * (there must be more than one) + */ +void +add_join_clause_to_rels(PlannerInfo *root, + RestrictInfo *restrictinfo, + Relids join_relids) +{ + int cur_relid; + + cur_relid = -1; + while ((cur_relid = bms_next_member(join_relids, cur_relid)) >= 0) + { + RelOptInfo *rel = find_base_rel(root, cur_relid); + + rel->joininfo = lappend(rel->joininfo, restrictinfo); + } +} + +/* + * remove_join_clause_from_rels + * Delete 'restrictinfo' from all the joininfo lists it is in + * + * This reverses the effect of add_join_clause_to_rels. It's used when we + * discover that a relation need not be joined at all. + * + * 'restrictinfo' describes the join clause + * 'join_relids' is the list of relations participating in the join clause + * (there must be more than one) + */ +void +remove_join_clause_from_rels(PlannerInfo *root, + RestrictInfo *restrictinfo, + Relids join_relids) +{ + int cur_relid; + + cur_relid = -1; + while ((cur_relid = bms_next_member(join_relids, cur_relid)) >= 0) + { + RelOptInfo *rel = find_base_rel(root, cur_relid); + + /* + * Remove the restrictinfo from the list. Pointer comparison is + * sufficient. + */ + Assert(list_member_ptr(rel->joininfo, restrictinfo)); + rel->joininfo = list_delete_ptr(rel->joininfo, restrictinfo); + } +} diff --git a/src/backend/optimizer/util/orclauses.c b/src/backend/optimizer/util/orclauses.c new file mode 100644 index 0000000..b1363df --- /dev/null +++ b/src/backend/optimizer/util/orclauses.c @@ -0,0 +1,360 @@ +/*------------------------------------------------------------------------- + * + * orclauses.c + * Routines to extract restriction OR clauses from join OR clauses + * + * Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group + * Portions Copyright (c) 1994, Regents of the University of California + * + * + * IDENTIFICATION + * src/backend/optimizer/util/orclauses.c + * + *------------------------------------------------------------------------- + */ + +#include "postgres.h" + +#include "nodes/makefuncs.h" +#include "nodes/nodeFuncs.h" +#include "optimizer/clauses.h" +#include "optimizer/cost.h" +#include "optimizer/optimizer.h" +#include "optimizer/orclauses.h" +#include "optimizer/restrictinfo.h" + + +static bool is_safe_restriction_clause_for(RestrictInfo *rinfo, RelOptInfo *rel); +static Expr *extract_or_clause(RestrictInfo *or_rinfo, RelOptInfo *rel); +static void consider_new_or_clause(PlannerInfo *root, RelOptInfo *rel, + Expr *orclause, RestrictInfo *join_or_rinfo); + + +/* + * extract_restriction_or_clauses + * Examine join OR-of-AND clauses to see if any useful restriction OR + * clauses can be extracted. If so, add them to the query. + * + * Although a join clause must reference multiple relations overall, + * an OR of ANDs clause might contain sub-clauses that reference just one + * relation and can be used to build a restriction clause for that rel. + * For example consider + * WHERE ((a.x = 42 AND b.y = 43) OR (a.x = 44 AND b.z = 45)); + * We can transform this into + * WHERE ((a.x = 42 AND b.y = 43) OR (a.x = 44 AND b.z = 45)) + * AND (a.x = 42 OR a.x = 44) + * AND (b.y = 43 OR b.z = 45); + * which allows the latter clauses to be applied during the scans of a and b, + * perhaps as index qualifications, and in any case reducing the number of + * rows arriving at the join. In essence this is a partial transformation to + * CNF (AND of ORs format). It is not complete, however, because we do not + * unravel the original OR --- doing so would usually bloat the qualification + * expression to little gain. + * + * The added quals are partially redundant with the original OR, and therefore + * would cause the size of the joinrel to be underestimated when it is finally + * formed. (This would be true of a full transformation to CNF as well; the + * fault is not really in the transformation, but in clauselist_selectivity's + * inability to recognize redundant conditions.) We can compensate for this + * redundancy by changing the cached selectivity of the original OR clause, + * canceling out the (valid) reduction in the estimated sizes of the base + * relations so that the estimated joinrel size remains the same. This is + * a MAJOR HACK: it depends on the fact that clause selectivities are cached + * and on the fact that the same RestrictInfo node will appear in every + * joininfo list that might be used when the joinrel is formed. + * And it doesn't work in cases where the size estimation is nonlinear + * (i.e., outer and IN joins). But it beats not doing anything. + * + * We examine each base relation to see if join clauses associated with it + * contain extractable restriction conditions. If so, add those conditions + * to the rel's baserestrictinfo and update the cached selectivities of the + * join clauses. Note that the same join clause will be examined afresh + * from the point of view of each baserel that participates in it, so its + * cached selectivity may get updated multiple times. + */ +void +extract_restriction_or_clauses(PlannerInfo *root) +{ + Index rti; + + /* Examine each baserel for potential join OR clauses */ + for (rti = 1; rti < root->simple_rel_array_size; rti++) + { + RelOptInfo *rel = root->simple_rel_array[rti]; + ListCell *lc; + + /* there may be empty slots corresponding to non-baserel RTEs */ + if (rel == NULL) + continue; + + Assert(rel->relid == rti); /* sanity check on array */ + + /* ignore RTEs that are "other rels" */ + if (rel->reloptkind != RELOPT_BASEREL) + continue; + + /* + * Find potentially interesting OR joinclauses. We can use any + * joinclause that is considered safe to move to this rel by the + * parameterized-path machinery, even though what we are going to do + * with it is not exactly a parameterized path. + * + * However, it seems best to ignore clauses that have been marked + * redundant (by setting norm_selec > 1). That likely can't happen + * for OR clauses, but let's be safe. + */ + foreach(lc, rel->joininfo) + { + RestrictInfo *rinfo = (RestrictInfo *) lfirst(lc); + + if (restriction_is_or_clause(rinfo) && + join_clause_is_movable_to(rinfo, rel) && + rinfo->norm_selec <= 1) + { + /* Try to extract a qual for this rel only */ + Expr *orclause = extract_or_clause(rinfo, rel); + + /* + * If successful, decide whether we want to use the clause, + * and insert it into the rel's restrictinfo list if so. + */ + if (orclause) + consider_new_or_clause(root, rel, orclause, rinfo); + } + } + } +} + +/* + * Is the given primitive (non-OR) RestrictInfo safe to move to the rel? + */ +static bool +is_safe_restriction_clause_for(RestrictInfo *rinfo, RelOptInfo *rel) +{ + /* + * We want clauses that mention the rel, and only the rel. So in + * particular pseudoconstant clauses can be rejected quickly. Then check + * the clause's Var membership. + */ + if (rinfo->pseudoconstant) + return false; + if (!bms_equal(rinfo->clause_relids, rel->relids)) + return false; + + /* We don't want extra evaluations of any volatile functions */ + if (contain_volatile_functions((Node *) rinfo->clause)) + return false; + + return true; +} + +/* + * Try to extract a restriction clause mentioning only "rel" from the given + * join OR-clause. + * + * We must be able to extract at least one qual for this rel from each of + * the arms of the OR, else we can't use it. + * + * Returns an OR clause (not a RestrictInfo!) pertaining to rel, or NULL + * if no OR clause could be extracted. + */ +static Expr * +extract_or_clause(RestrictInfo *or_rinfo, RelOptInfo *rel) +{ + List *clauselist = NIL; + ListCell *lc; + + /* + * Scan each arm of the input OR clause. Notice we descend into + * or_rinfo->orclause, which has RestrictInfo nodes embedded below the + * toplevel OR/AND structure. This is useful because we can use the info + * in those nodes to make is_safe_restriction_clause_for()'s checks + * cheaper. We'll strip those nodes from the returned tree, though, + * meaning that fresh ones will be built if the clause is accepted as a + * restriction clause. This might seem wasteful --- couldn't we re-use + * the existing RestrictInfos? But that'd require assuming that + * selectivity and other cached data is computed exactly the same way for + * a restriction clause as for a join clause, which seems undesirable. + */ + Assert(is_orclause(or_rinfo->orclause)); + foreach(lc, ((BoolExpr *) or_rinfo->orclause)->args) + { + Node *orarg = (Node *) lfirst(lc); + List *subclauses = NIL; + Node *subclause; + + /* OR arguments should be ANDs or sub-RestrictInfos */ + if (is_andclause(orarg)) + { + List *andargs = ((BoolExpr *) orarg)->args; + ListCell *lc2; + + foreach(lc2, andargs) + { + RestrictInfo *rinfo = lfirst_node(RestrictInfo, lc2); + + if (restriction_is_or_clause(rinfo)) + { + /* + * Recurse to deal with nested OR. Note we *must* recurse + * here, this isn't just overly-tense optimization: we + * have to descend far enough to find and strip all + * RestrictInfos in the expression. + */ + Expr *suborclause; + + suborclause = extract_or_clause(rinfo, rel); + if (suborclause) + subclauses = lappend(subclauses, suborclause); + } + else if (is_safe_restriction_clause_for(rinfo, rel)) + subclauses = lappend(subclauses, rinfo->clause); + } + } + else + { + RestrictInfo *rinfo = castNode(RestrictInfo, orarg); + + Assert(!restriction_is_or_clause(rinfo)); + if (is_safe_restriction_clause_for(rinfo, rel)) + subclauses = lappend(subclauses, rinfo->clause); + } + + /* + * If nothing could be extracted from this arm, we can't do anything + * with this OR clause. + */ + if (subclauses == NIL) + return NULL; + + /* + * OK, add subclause(s) to the result OR. If we found more than one, + * we need an AND node. But if we found only one, and it is itself an + * OR node, add its subclauses to the result instead; this is needed + * to preserve AND/OR flatness (ie, no OR directly underneath OR). + */ + subclause = (Node *) make_ands_explicit(subclauses); + if (is_orclause(subclause)) + clauselist = list_concat(clauselist, + ((BoolExpr *) subclause)->args); + else + clauselist = lappend(clauselist, subclause); + } + + /* + * If we got a restriction clause from every arm, wrap them up in an OR + * node. (In theory the OR node might be unnecessary, if there was only + * one arm --- but then the input OR node was also redundant.) + */ + if (clauselist != NIL) + return make_orclause(clauselist); + return NULL; +} + +/* + * Consider whether a successfully-extracted restriction OR clause is + * actually worth using. If so, add it to the planner's data structures, + * and adjust the original join clause (join_or_rinfo) to compensate. + */ +static void +consider_new_or_clause(PlannerInfo *root, RelOptInfo *rel, + Expr *orclause, RestrictInfo *join_or_rinfo) +{ + RestrictInfo *or_rinfo; + Selectivity or_selec, + orig_selec; + + /* + * Build a RestrictInfo from the new OR clause. We can assume it's valid + * as a base restriction clause. + */ + or_rinfo = make_restrictinfo(root, + orclause, + true, + false, + false, + join_or_rinfo->security_level, + NULL, + NULL, + NULL); + + /* + * Estimate its selectivity. (We could have done this earlier, but doing + * it on the RestrictInfo representation allows the result to get cached, + * saving work later.) + */ + or_selec = clause_selectivity(root, (Node *) or_rinfo, + 0, JOIN_INNER, NULL); + + /* + * The clause is only worth adding to the query if it rejects a useful + * fraction of the base relation's rows; otherwise, it's just going to + * cause duplicate computation (since we will still have to check the + * original OR clause when the join is formed). Somewhat arbitrarily, we + * set the selectivity threshold at 0.9. + */ + if (or_selec > 0.9) + return; /* forget it */ + + /* + * OK, add it to the rel's restriction-clause list. + */ + rel->baserestrictinfo = lappend(rel->baserestrictinfo, or_rinfo); + rel->baserestrict_min_security = Min(rel->baserestrict_min_security, + or_rinfo->security_level); + + /* + * Adjust the original join OR clause's cached selectivity to compensate + * for the selectivity of the added (but redundant) lower-level qual. This + * should result in the join rel getting approximately the same rows + * estimate as it would have gotten without all these shenanigans. + * + * XXX major hack alert: this depends on the assumption that the + * selectivity will stay cached. + * + * XXX another major hack: we adjust only norm_selec, the cached + * selectivity for JOIN_INNER semantics, even though the join clause + * might've been an outer-join clause. This is partly because we can't + * easily identify the relevant SpecialJoinInfo here, and partly because + * the linearity assumption we're making would fail anyway. (If it is an + * outer-join clause, "rel" must be on the nullable side, else we'd not + * have gotten here. So the computation of the join size is going to be + * quite nonlinear with respect to the size of "rel", so it's not clear + * how we ought to adjust outer_selec even if we could compute its + * original value correctly.) + */ + if (or_selec > 0) + { + SpecialJoinInfo sjinfo; + + /* + * Make up a SpecialJoinInfo for JOIN_INNER semantics. (Compare + * approx_tuple_count() in costsize.c.) + */ + sjinfo.type = T_SpecialJoinInfo; + sjinfo.min_lefthand = bms_difference(join_or_rinfo->clause_relids, + rel->relids); + sjinfo.min_righthand = rel->relids; + sjinfo.syn_lefthand = sjinfo.min_lefthand; + sjinfo.syn_righthand = sjinfo.min_righthand; + sjinfo.jointype = JOIN_INNER; + /* we don't bother trying to make the remaining fields valid */ + sjinfo.lhs_strict = false; + sjinfo.delay_upper_joins = false; + sjinfo.semi_can_btree = false; + sjinfo.semi_can_hash = false; + sjinfo.semi_operators = NIL; + sjinfo.semi_rhs_exprs = NIL; + + /* Compute inner-join size */ + orig_selec = clause_selectivity(root, (Node *) join_or_rinfo, + 0, JOIN_INNER, &sjinfo); + + /* And hack cached selectivity so join size remains the same */ + join_or_rinfo->norm_selec = orig_selec / or_selec; + /* ensure result stays in sane range, in particular not "redundant" */ + if (join_or_rinfo->norm_selec > 1) + join_or_rinfo->norm_selec = 1; + /* as explained above, we don't touch outer_selec */ + } +} diff --git a/src/backend/optimizer/util/paramassign.c b/src/backend/optimizer/util/paramassign.c new file mode 100644 index 0000000..12486cb --- /dev/null +++ b/src/backend/optimizer/util/paramassign.c @@ -0,0 +1,591 @@ +/*------------------------------------------------------------------------- + * + * paramassign.c + * Functions for assigning PARAM_EXEC slots during planning. + * + * This module is responsible for managing three planner data structures: + * + * root->glob->paramExecTypes: records actual assignments of PARAM_EXEC slots. + * The i'th list element holds the data type OID of the i'th parameter slot. + * (Elements can be InvalidOid if they represent slots that are needed for + * chgParam signaling, but will never hold a value at runtime.) This list is + * global to the whole plan since the executor has only one PARAM_EXEC array. + * Assignments are permanent for the plan: we never remove entries once added. + * + * root->plan_params: a list of PlannerParamItem nodes, recording Vars and + * PlaceHolderVars that the root's query level needs to supply to lower-level + * subqueries, along with the PARAM_EXEC number to use for each such value. + * Elements are added to this list while planning a subquery, and the list + * is reset to empty after completion of each subquery. + * + * root->curOuterParams: a list of NestLoopParam nodes, recording Vars and + * PlaceHolderVars that some outer level of nestloop needs to pass down to + * a lower-level plan node in its righthand side. Elements are added to this + * list as createplan.c creates lower Plan nodes that need such Params, and + * are removed when it creates a NestLoop Plan node that will supply those + * values. + * + * The latter two data structures are used to prevent creating multiple + * PARAM_EXEC slots (each requiring work to fill) when the same upper + * SubPlan or NestLoop supplies a value that is referenced in more than + * one place in its child plan nodes. However, when the same Var has to + * be supplied to different subplan trees by different SubPlan or NestLoop + * parent nodes, we don't recognize any commonality; a fresh plan_params or + * curOuterParams entry will be made (since the old one has been removed + * when we finished processing the earlier SubPlan or NestLoop) and a fresh + * PARAM_EXEC number will be assigned. At one time we tried to avoid + * allocating duplicate PARAM_EXEC numbers in such cases, but it's harder + * than it seems to avoid bugs due to overlapping Param lifetimes, so we + * don't risk that anymore. Minimizing the number of PARAM_EXEC slots + * doesn't really save much executor work anyway. + * + * + * Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group + * Portions Copyright (c) 1994, Regents of the University of California + * + * IDENTIFICATION + * src/backend/optimizer/util/paramassign.c + * + *------------------------------------------------------------------------- + */ +#include "postgres.h" + +#include "nodes/nodeFuncs.h" +#include "nodes/plannodes.h" +#include "optimizer/paramassign.h" +#include "optimizer/placeholder.h" +#include "rewrite/rewriteManip.h" + + +/* + * Select a PARAM_EXEC number to identify the given Var as a parameter for + * the current subquery. (It might already have one.) + * Record the need for the Var in the proper upper-level root->plan_params. + */ +static int +assign_param_for_var(PlannerInfo *root, Var *var) +{ + ListCell *ppl; + PlannerParamItem *pitem; + Index levelsup; + + /* Find the query level the Var belongs to */ + for (levelsup = var->varlevelsup; levelsup > 0; levelsup--) + root = root->parent_root; + + /* If there's already a matching PlannerParamItem there, just use it */ + foreach(ppl, root->plan_params) + { + pitem = (PlannerParamItem *) lfirst(ppl); + if (IsA(pitem->item, Var)) + { + Var *pvar = (Var *) pitem->item; + + /* + * This comparison must match _equalVar(), except for ignoring + * varlevelsup. Note that _equalVar() ignores varnosyn, + * varattnosyn, and location, so this does too. + */ + if (pvar->varno == var->varno && + pvar->varattno == var->varattno && + pvar->vartype == var->vartype && + pvar->vartypmod == var->vartypmod && + pvar->varcollid == var->varcollid) + return pitem->paramId; + } + } + + /* Nope, so make a new one */ + var = copyObject(var); + var->varlevelsup = 0; + + pitem = makeNode(PlannerParamItem); + pitem->item = (Node *) var; + pitem->paramId = list_length(root->glob->paramExecTypes); + root->glob->paramExecTypes = lappend_oid(root->glob->paramExecTypes, + var->vartype); + + root->plan_params = lappend(root->plan_params, pitem); + + return pitem->paramId; +} + +/* + * Generate a Param node to replace the given Var, + * which is expected to have varlevelsup > 0 (ie, it is not local). + * Record the need for the Var in the proper upper-level root->plan_params. + */ +Param * +replace_outer_var(PlannerInfo *root, Var *var) +{ + Param *retval; + int i; + + Assert(var->varlevelsup > 0 && var->varlevelsup < root->query_level); + + /* Find the Var in the appropriate plan_params, or add it if not present */ + i = assign_param_for_var(root, var); + + retval = makeNode(Param); + retval->paramkind = PARAM_EXEC; + retval->paramid = i; + retval->paramtype = var->vartype; + retval->paramtypmod = var->vartypmod; + retval->paramcollid = var->varcollid; + retval->location = var->location; + + return retval; +} + +/* + * Select a PARAM_EXEC number to identify the given PlaceHolderVar as a + * parameter for the current subquery. (It might already have one.) + * Record the need for the PHV in the proper upper-level root->plan_params. + * + * This is just like assign_param_for_var, except for PlaceHolderVars. + */ +static int +assign_param_for_placeholdervar(PlannerInfo *root, PlaceHolderVar *phv) +{ + ListCell *ppl; + PlannerParamItem *pitem; + Index levelsup; + + /* Find the query level the PHV belongs to */ + for (levelsup = phv->phlevelsup; levelsup > 0; levelsup--) + root = root->parent_root; + + /* If there's already a matching PlannerParamItem there, just use it */ + foreach(ppl, root->plan_params) + { + pitem = (PlannerParamItem *) lfirst(ppl); + if (IsA(pitem->item, PlaceHolderVar)) + { + PlaceHolderVar *pphv = (PlaceHolderVar *) pitem->item; + + /* We assume comparing the PHIDs is sufficient */ + if (pphv->phid == phv->phid) + return pitem->paramId; + } + } + + /* Nope, so make a new one */ + phv = copyObject(phv); + IncrementVarSublevelsUp((Node *) phv, -((int) phv->phlevelsup), 0); + Assert(phv->phlevelsup == 0); + + pitem = makeNode(PlannerParamItem); + pitem->item = (Node *) phv; + pitem->paramId = list_length(root->glob->paramExecTypes); + root->glob->paramExecTypes = lappend_oid(root->glob->paramExecTypes, + exprType((Node *) phv->phexpr)); + + root->plan_params = lappend(root->plan_params, pitem); + + return pitem->paramId; +} + +/* + * Generate a Param node to replace the given PlaceHolderVar, + * which is expected to have phlevelsup > 0 (ie, it is not local). + * Record the need for the PHV in the proper upper-level root->plan_params. + * + * This is just like replace_outer_var, except for PlaceHolderVars. + */ +Param * +replace_outer_placeholdervar(PlannerInfo *root, PlaceHolderVar *phv) +{ + Param *retval; + int i; + + Assert(phv->phlevelsup > 0 && phv->phlevelsup < root->query_level); + + /* Find the PHV in the appropriate plan_params, or add it if not present */ + i = assign_param_for_placeholdervar(root, phv); + + retval = makeNode(Param); + retval->paramkind = PARAM_EXEC; + retval->paramid = i; + retval->paramtype = exprType((Node *) phv->phexpr); + retval->paramtypmod = exprTypmod((Node *) phv->phexpr); + retval->paramcollid = exprCollation((Node *) phv->phexpr); + retval->location = -1; + + return retval; +} + +/* + * Generate a Param node to replace the given Aggref + * which is expected to have agglevelsup > 0 (ie, it is not local). + * Record the need for the Aggref in the proper upper-level root->plan_params. + */ +Param * +replace_outer_agg(PlannerInfo *root, Aggref *agg) +{ + Param *retval; + PlannerParamItem *pitem; + Index levelsup; + + Assert(agg->agglevelsup > 0 && agg->agglevelsup < root->query_level); + + /* Find the query level the Aggref belongs to */ + for (levelsup = agg->agglevelsup; levelsup > 0; levelsup--) + root = root->parent_root; + + /* + * It does not seem worthwhile to try to de-duplicate references to outer + * aggs. Just make a new slot every time. + */ + agg = copyObject(agg); + IncrementVarSublevelsUp((Node *) agg, -((int) agg->agglevelsup), 0); + Assert(agg->agglevelsup == 0); + + pitem = makeNode(PlannerParamItem); + pitem->item = (Node *) agg; + pitem->paramId = list_length(root->glob->paramExecTypes); + root->glob->paramExecTypes = lappend_oid(root->glob->paramExecTypes, + agg->aggtype); + + root->plan_params = lappend(root->plan_params, pitem); + + retval = makeNode(Param); + retval->paramkind = PARAM_EXEC; + retval->paramid = pitem->paramId; + retval->paramtype = agg->aggtype; + retval->paramtypmod = -1; + retval->paramcollid = agg->aggcollid; + retval->location = agg->location; + + return retval; +} + +/* + * Generate a Param node to replace the given GroupingFunc expression which is + * expected to have agglevelsup > 0 (ie, it is not local). + * Record the need for the GroupingFunc in the proper upper-level + * root->plan_params. + */ +Param * +replace_outer_grouping(PlannerInfo *root, GroupingFunc *grp) +{ + Param *retval; + PlannerParamItem *pitem; + Index levelsup; + Oid ptype = exprType((Node *) grp); + + Assert(grp->agglevelsup > 0 && grp->agglevelsup < root->query_level); + + /* Find the query level the GroupingFunc belongs to */ + for (levelsup = grp->agglevelsup; levelsup > 0; levelsup--) + root = root->parent_root; + + /* + * It does not seem worthwhile to try to de-duplicate references to outer + * aggs. Just make a new slot every time. + */ + grp = copyObject(grp); + IncrementVarSublevelsUp((Node *) grp, -((int) grp->agglevelsup), 0); + Assert(grp->agglevelsup == 0); + + pitem = makeNode(PlannerParamItem); + pitem->item = (Node *) grp; + pitem->paramId = list_length(root->glob->paramExecTypes); + root->glob->paramExecTypes = lappend_oid(root->glob->paramExecTypes, + ptype); + + root->plan_params = lappend(root->plan_params, pitem); + + retval = makeNode(Param); + retval->paramkind = PARAM_EXEC; + retval->paramid = pitem->paramId; + retval->paramtype = ptype; + retval->paramtypmod = -1; + retval->paramcollid = InvalidOid; + retval->location = grp->location; + + return retval; +} + +/* + * Generate a Param node to replace the given Var, + * which is expected to come from some upper NestLoop plan node. + * Record the need for the Var in root->curOuterParams. + */ +Param * +replace_nestloop_param_var(PlannerInfo *root, Var *var) +{ + Param *param; + NestLoopParam *nlp; + ListCell *lc; + + /* Is this Var already listed in root->curOuterParams? */ + foreach(lc, root->curOuterParams) + { + nlp = (NestLoopParam *) lfirst(lc); + if (equal(var, nlp->paramval)) + { + /* Yes, so just make a Param referencing this NLP's slot */ + param = makeNode(Param); + param->paramkind = PARAM_EXEC; + param->paramid = nlp->paramno; + param->paramtype = var->vartype; + param->paramtypmod = var->vartypmod; + param->paramcollid = var->varcollid; + param->location = var->location; + return param; + } + } + + /* No, so assign a PARAM_EXEC slot for a new NLP */ + param = generate_new_exec_param(root, + var->vartype, + var->vartypmod, + var->varcollid); + param->location = var->location; + + /* Add it to the list of required NLPs */ + nlp = makeNode(NestLoopParam); + nlp->paramno = param->paramid; + nlp->paramval = copyObject(var); + root->curOuterParams = lappend(root->curOuterParams, nlp); + + /* And return the replacement Param */ + return param; +} + +/* + * Generate a Param node to replace the given PlaceHolderVar, + * which is expected to come from some upper NestLoop plan node. + * Record the need for the PHV in root->curOuterParams. + * + * This is just like replace_nestloop_param_var, except for PlaceHolderVars. + */ +Param * +replace_nestloop_param_placeholdervar(PlannerInfo *root, PlaceHolderVar *phv) +{ + Param *param; + NestLoopParam *nlp; + ListCell *lc; + + /* Is this PHV already listed in root->curOuterParams? */ + foreach(lc, root->curOuterParams) + { + nlp = (NestLoopParam *) lfirst(lc); + if (equal(phv, nlp->paramval)) + { + /* Yes, so just make a Param referencing this NLP's slot */ + param = makeNode(Param); + param->paramkind = PARAM_EXEC; + param->paramid = nlp->paramno; + param->paramtype = exprType((Node *) phv->phexpr); + param->paramtypmod = exprTypmod((Node *) phv->phexpr); + param->paramcollid = exprCollation((Node *) phv->phexpr); + param->location = -1; + return param; + } + } + + /* No, so assign a PARAM_EXEC slot for a new NLP */ + param = generate_new_exec_param(root, + exprType((Node *) phv->phexpr), + exprTypmod((Node *) phv->phexpr), + exprCollation((Node *) phv->phexpr)); + + /* Add it to the list of required NLPs */ + nlp = makeNode(NestLoopParam); + nlp->paramno = param->paramid; + nlp->paramval = (Var *) copyObject(phv); + root->curOuterParams = lappend(root->curOuterParams, nlp); + + /* And return the replacement Param */ + return param; +} + +/* + * process_subquery_nestloop_params + * Handle params of a parameterized subquery that need to be fed + * from an outer nestloop. + * + * Currently, that would be *all* params that a subquery in FROM has demanded + * from the current query level, since they must be LATERAL references. + * + * subplan_params is a list of PlannerParamItems that we intend to pass to + * a subquery-in-FROM. (This was constructed in root->plan_params while + * planning the subquery, but isn't there anymore when this is called.) + * + * The subplan's references to the outer variables are already represented + * as PARAM_EXEC Params, since that conversion was done by the routines above + * while planning the subquery. So we need not modify the subplan or the + * PlannerParamItems here. What we do need to do is add entries to + * root->curOuterParams to signal the parent nestloop plan node that it must + * provide these values. This differs from replace_nestloop_param_var in + * that the PARAM_EXEC slots to use have already been determined. + * + * Note that we also use root->curOuterRels as an implicit parameter for + * sanity checks. + */ +void +process_subquery_nestloop_params(PlannerInfo *root, List *subplan_params) +{ + ListCell *lc; + + foreach(lc, subplan_params) + { + PlannerParamItem *pitem = lfirst_node(PlannerParamItem, lc); + + if (IsA(pitem->item, Var)) + { + Var *var = (Var *) pitem->item; + NestLoopParam *nlp; + ListCell *lc; + + /* If not from a nestloop outer rel, complain */ + if (!bms_is_member(var->varno, root->curOuterRels)) + elog(ERROR, "non-LATERAL parameter required by subquery"); + + /* Is this param already listed in root->curOuterParams? */ + foreach(lc, root->curOuterParams) + { + nlp = (NestLoopParam *) lfirst(lc); + if (nlp->paramno == pitem->paramId) + { + Assert(equal(var, nlp->paramval)); + /* Present, so nothing to do */ + break; + } + } + if (lc == NULL) + { + /* No, so add it */ + nlp = makeNode(NestLoopParam); + nlp->paramno = pitem->paramId; + nlp->paramval = copyObject(var); + root->curOuterParams = lappend(root->curOuterParams, nlp); + } + } + else if (IsA(pitem->item, PlaceHolderVar)) + { + PlaceHolderVar *phv = (PlaceHolderVar *) pitem->item; + NestLoopParam *nlp; + ListCell *lc; + + /* If not from a nestloop outer rel, complain */ + if (!bms_is_subset(find_placeholder_info(root, phv, false)->ph_eval_at, + root->curOuterRels)) + elog(ERROR, "non-LATERAL parameter required by subquery"); + + /* Is this param already listed in root->curOuterParams? */ + foreach(lc, root->curOuterParams) + { + nlp = (NestLoopParam *) lfirst(lc); + if (nlp->paramno == pitem->paramId) + { + Assert(equal(phv, nlp->paramval)); + /* Present, so nothing to do */ + break; + } + } + if (lc == NULL) + { + /* No, so add it */ + nlp = makeNode(NestLoopParam); + nlp->paramno = pitem->paramId; + nlp->paramval = (Var *) copyObject(phv); + root->curOuterParams = lappend(root->curOuterParams, nlp); + } + } + else + elog(ERROR, "unexpected type of subquery parameter"); + } +} + +/* + * Identify any NestLoopParams that should be supplied by a NestLoop plan + * node with the specified lefthand rels. Remove them from the active + * root->curOuterParams list and return them as the result list. + */ +List * +identify_current_nestloop_params(PlannerInfo *root, Relids leftrelids) +{ + List *result; + ListCell *cell; + + result = NIL; + foreach(cell, root->curOuterParams) + { + NestLoopParam *nlp = (NestLoopParam *) lfirst(cell); + + /* + * We are looking for Vars and PHVs that can be supplied by the + * lefthand rels. The "bms_overlap" test is just an optimization to + * allow skipping find_placeholder_info() if the PHV couldn't match. + */ + if (IsA(nlp->paramval, Var) && + bms_is_member(nlp->paramval->varno, leftrelids)) + { + root->curOuterParams = foreach_delete_current(root->curOuterParams, + cell); + result = lappend(result, nlp); + } + else if (IsA(nlp->paramval, PlaceHolderVar) && + bms_overlap(((PlaceHolderVar *) nlp->paramval)->phrels, + leftrelids) && + bms_is_subset(find_placeholder_info(root, + (PlaceHolderVar *) nlp->paramval, + false)->ph_eval_at, + leftrelids)) + { + root->curOuterParams = foreach_delete_current(root->curOuterParams, + cell); + result = lappend(result, nlp); + } + } + return result; +} + +/* + * Generate a new Param node that will not conflict with any other. + * + * This is used to create Params representing subplan outputs or + * NestLoop parameters. + * + * We don't need to build a PlannerParamItem for such a Param, but we do + * need to make sure we record the type in paramExecTypes (otherwise, + * there won't be a slot allocated for it). + */ +Param * +generate_new_exec_param(PlannerInfo *root, Oid paramtype, int32 paramtypmod, + Oid paramcollation) +{ + Param *retval; + + retval = makeNode(Param); + retval->paramkind = PARAM_EXEC; + retval->paramid = list_length(root->glob->paramExecTypes); + root->glob->paramExecTypes = lappend_oid(root->glob->paramExecTypes, + paramtype); + retval->paramtype = paramtype; + retval->paramtypmod = paramtypmod; + retval->paramcollid = paramcollation; + retval->location = -1; + + return retval; +} + +/* + * Assign a (nonnegative) PARAM_EXEC ID for a special parameter (one that + * is not actually used to carry a value at runtime). Such parameters are + * used for special runtime signaling purposes, such as connecting a + * recursive union node to its worktable scan node or forcing plan + * re-evaluation within the EvalPlanQual mechanism. No actual Param node + * exists with this ID, however. + */ +int +assign_special_exec_param(PlannerInfo *root) +{ + int paramId = list_length(root->glob->paramExecTypes); + + root->glob->paramExecTypes = lappend_oid(root->glob->paramExecTypes, + InvalidOid); + return paramId; +} diff --git a/src/backend/optimizer/util/pathnode.c b/src/backend/optimizer/util/pathnode.c new file mode 100644 index 0000000..33affaf --- /dev/null +++ b/src/backend/optimizer/util/pathnode.c @@ -0,0 +1,4298 @@ +/*------------------------------------------------------------------------- + * + * pathnode.c + * Routines to manipulate pathlists and create path nodes + * + * Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group + * Portions Copyright (c) 1994, Regents of the University of California + * + * + * IDENTIFICATION + * src/backend/optimizer/util/pathnode.c + * + *------------------------------------------------------------------------- + */ +#include "postgres.h" + +#include <math.h> + +#include "foreign/fdwapi.h" +#include "miscadmin.h" +#include "nodes/extensible.h" +#include "nodes/nodeFuncs.h" +#include "optimizer/appendinfo.h" +#include "optimizer/clauses.h" +#include "optimizer/cost.h" +#include "optimizer/optimizer.h" +#include "optimizer/pathnode.h" +#include "optimizer/paths.h" +#include "optimizer/planmain.h" +#include "optimizer/prep.h" +#include "optimizer/restrictinfo.h" +#include "optimizer/tlist.h" +#include "parser/parsetree.h" +#include "utils/lsyscache.h" +#include "utils/memutils.h" +#include "utils/selfuncs.h" + +typedef enum +{ + COSTS_EQUAL, /* path costs are fuzzily equal */ + COSTS_BETTER1, /* first path is cheaper than second */ + COSTS_BETTER2, /* second path is cheaper than first */ + COSTS_DIFFERENT /* neither path dominates the other on cost */ +} PathCostComparison; + +/* + * STD_FUZZ_FACTOR is the normal fuzz factor for compare_path_costs_fuzzily. + * XXX is it worth making this user-controllable? It provides a tradeoff + * between planner runtime and the accuracy of path cost comparisons. + */ +#define STD_FUZZ_FACTOR 1.01 + +static List *translate_sub_tlist(List *tlist, int relid); +static int append_total_cost_compare(const ListCell *a, const ListCell *b); +static int append_startup_cost_compare(const ListCell *a, const ListCell *b); +static List *reparameterize_pathlist_by_child(PlannerInfo *root, + List *pathlist, + RelOptInfo *child_rel); + + +/***************************************************************************** + * MISC. PATH UTILITIES + *****************************************************************************/ + +/* + * compare_path_costs + * Return -1, 0, or +1 according as path1 is cheaper, the same cost, + * or more expensive than path2 for the specified criterion. + */ +int +compare_path_costs(Path *path1, Path *path2, CostSelector criterion) +{ + if (criterion == STARTUP_COST) + { + if (path1->startup_cost < path2->startup_cost) + return -1; + if (path1->startup_cost > path2->startup_cost) + return +1; + + /* + * If paths have the same startup cost (not at all unlikely), order + * them by total cost. + */ + if (path1->total_cost < path2->total_cost) + return -1; + if (path1->total_cost > path2->total_cost) + return +1; + } + else + { + if (path1->total_cost < path2->total_cost) + return -1; + if (path1->total_cost > path2->total_cost) + return +1; + + /* + * If paths have the same total cost, order them by startup cost. + */ + if (path1->startup_cost < path2->startup_cost) + return -1; + if (path1->startup_cost > path2->startup_cost) + return +1; + } + return 0; +} + +/* + * compare_fractional_path_costs + * Return -1, 0, or +1 according as path1 is cheaper, the same cost, + * or more expensive than path2 for fetching the specified fraction + * of the total tuples. + * + * If fraction is <= 0 or > 1, we interpret it as 1, ie, we select the + * path with the cheaper total_cost. + */ +int +compare_fractional_path_costs(Path *path1, Path *path2, + double fraction) +{ + Cost cost1, + cost2; + + if (fraction <= 0.0 || fraction >= 1.0) + return compare_path_costs(path1, path2, TOTAL_COST); + cost1 = path1->startup_cost + + fraction * (path1->total_cost - path1->startup_cost); + cost2 = path2->startup_cost + + fraction * (path2->total_cost - path2->startup_cost); + if (cost1 < cost2) + return -1; + if (cost1 > cost2) + return +1; + return 0; +} + +/* + * compare_path_costs_fuzzily + * Compare the costs of two paths to see if either can be said to + * dominate the other. + * + * We use fuzzy comparisons so that add_path() can avoid keeping both of + * a pair of paths that really have insignificantly different cost. + * + * The fuzz_factor argument must be 1.0 plus delta, where delta is the + * fraction of the smaller cost that is considered to be a significant + * difference. For example, fuzz_factor = 1.01 makes the fuzziness limit + * be 1% of the smaller cost. + * + * The two paths are said to have "equal" costs if both startup and total + * costs are fuzzily the same. Path1 is said to be better than path2 if + * it has fuzzily better startup cost and fuzzily no worse total cost, + * or if it has fuzzily better total cost and fuzzily no worse startup cost. + * Path2 is better than path1 if the reverse holds. Finally, if one path + * is fuzzily better than the other on startup cost and fuzzily worse on + * total cost, we just say that their costs are "different", since neither + * dominates the other across the whole performance spectrum. + * + * This function also enforces a policy rule that paths for which the relevant + * one of parent->consider_startup and parent->consider_param_startup is false + * cannot survive comparisons solely on the grounds of good startup cost, so + * we never return COSTS_DIFFERENT when that is true for the total-cost loser. + * (But if total costs are fuzzily equal, we compare startup costs anyway, + * in hopes of eliminating one path or the other.) + */ +static PathCostComparison +compare_path_costs_fuzzily(Path *path1, Path *path2, double fuzz_factor) +{ +#define CONSIDER_PATH_STARTUP_COST(p) \ + ((p)->param_info == NULL ? (p)->parent->consider_startup : (p)->parent->consider_param_startup) + + /* + * Check total cost first since it's more likely to be different; many + * paths have zero startup cost. + */ + if (path1->total_cost > path2->total_cost * fuzz_factor) + { + /* path1 fuzzily worse on total cost */ + if (CONSIDER_PATH_STARTUP_COST(path1) && + path2->startup_cost > path1->startup_cost * fuzz_factor) + { + /* ... but path2 fuzzily worse on startup, so DIFFERENT */ + return COSTS_DIFFERENT; + } + /* else path2 dominates */ + return COSTS_BETTER2; + } + if (path2->total_cost > path1->total_cost * fuzz_factor) + { + /* path2 fuzzily worse on total cost */ + if (CONSIDER_PATH_STARTUP_COST(path2) && + path1->startup_cost > path2->startup_cost * fuzz_factor) + { + /* ... but path1 fuzzily worse on startup, so DIFFERENT */ + return COSTS_DIFFERENT; + } + /* else path1 dominates */ + return COSTS_BETTER1; + } + /* fuzzily the same on total cost ... */ + if (path1->startup_cost > path2->startup_cost * fuzz_factor) + { + /* ... but path1 fuzzily worse on startup, so path2 wins */ + return COSTS_BETTER2; + } + if (path2->startup_cost > path1->startup_cost * fuzz_factor) + { + /* ... but path2 fuzzily worse on startup, so path1 wins */ + return COSTS_BETTER1; + } + /* fuzzily the same on both costs */ + return COSTS_EQUAL; + +#undef CONSIDER_PATH_STARTUP_COST +} + +/* + * set_cheapest + * Find the minimum-cost paths from among a relation's paths, + * and save them in the rel's cheapest-path fields. + * + * cheapest_total_path is normally the cheapest-total-cost unparameterized + * path; but if there are no unparameterized paths, we assign it to be the + * best (cheapest least-parameterized) parameterized path. However, only + * unparameterized paths are considered candidates for cheapest_startup_path, + * so that will be NULL if there are no unparameterized paths. + * + * The cheapest_parameterized_paths list collects all parameterized paths + * that have survived the add_path() tournament for this relation. (Since + * add_path ignores pathkeys for a parameterized path, these will be paths + * that have best cost or best row count for their parameterization. We + * may also have both a parallel-safe and a non-parallel-safe path in some + * cases for the same parameterization in some cases, but this should be + * relatively rare since, most typically, all paths for the same relation + * will be parallel-safe or none of them will.) + * + * cheapest_parameterized_paths always includes the cheapest-total + * unparameterized path, too, if there is one; the users of that list find + * it more convenient if that's included. + * + * This is normally called only after we've finished constructing the path + * list for the rel node. + */ +void +set_cheapest(RelOptInfo *parent_rel) +{ + Path *cheapest_startup_path; + Path *cheapest_total_path; + Path *best_param_path; + List *parameterized_paths; + ListCell *p; + + Assert(IsA(parent_rel, RelOptInfo)); + + if (parent_rel->pathlist == NIL) + elog(ERROR, "could not devise a query plan for the given query"); + + cheapest_startup_path = cheapest_total_path = best_param_path = NULL; + parameterized_paths = NIL; + + foreach(p, parent_rel->pathlist) + { + Path *path = (Path *) lfirst(p); + int cmp; + + if (path->param_info) + { + /* Parameterized path, so add it to parameterized_paths */ + parameterized_paths = lappend(parameterized_paths, path); + + /* + * If we have an unparameterized cheapest-total, we no longer care + * about finding the best parameterized path, so move on. + */ + if (cheapest_total_path) + continue; + + /* + * Otherwise, track the best parameterized path, which is the one + * with least total cost among those of the minimum + * parameterization. + */ + if (best_param_path == NULL) + best_param_path = path; + else + { + switch (bms_subset_compare(PATH_REQ_OUTER(path), + PATH_REQ_OUTER(best_param_path))) + { + case BMS_EQUAL: + /* keep the cheaper one */ + if (compare_path_costs(path, best_param_path, + TOTAL_COST) < 0) + best_param_path = path; + break; + case BMS_SUBSET1: + /* new path is less-parameterized */ + best_param_path = path; + break; + case BMS_SUBSET2: + /* old path is less-parameterized, keep it */ + break; + case BMS_DIFFERENT: + + /* + * This means that neither path has the least possible + * parameterization for the rel. We'll sit on the old + * path until something better comes along. + */ + break; + } + } + } + else + { + /* Unparameterized path, so consider it for cheapest slots */ + if (cheapest_total_path == NULL) + { + cheapest_startup_path = cheapest_total_path = path; + continue; + } + + /* + * If we find two paths of identical costs, try to keep the + * better-sorted one. The paths might have unrelated sort + * orderings, in which case we can only guess which might be + * better to keep, but if one is superior then we definitely + * should keep that one. + */ + cmp = compare_path_costs(cheapest_startup_path, path, STARTUP_COST); + if (cmp > 0 || + (cmp == 0 && + compare_pathkeys(cheapest_startup_path->pathkeys, + path->pathkeys) == PATHKEYS_BETTER2)) + cheapest_startup_path = path; + + cmp = compare_path_costs(cheapest_total_path, path, TOTAL_COST); + if (cmp > 0 || + (cmp == 0 && + compare_pathkeys(cheapest_total_path->pathkeys, + path->pathkeys) == PATHKEYS_BETTER2)) + cheapest_total_path = path; + } + } + + /* Add cheapest unparameterized path, if any, to parameterized_paths */ + if (cheapest_total_path) + parameterized_paths = lcons(cheapest_total_path, parameterized_paths); + + /* + * If there is no unparameterized path, use the best parameterized path as + * cheapest_total_path (but not as cheapest_startup_path). + */ + if (cheapest_total_path == NULL) + cheapest_total_path = best_param_path; + Assert(cheapest_total_path != NULL); + + parent_rel->cheapest_startup_path = cheapest_startup_path; + parent_rel->cheapest_total_path = cheapest_total_path; + parent_rel->cheapest_unique_path = NULL; /* computed only if needed */ + parent_rel->cheapest_parameterized_paths = parameterized_paths; +} + +/* + * add_path + * Consider a potential implementation path for the specified parent rel, + * and add it to the rel's pathlist if it is worthy of consideration. + * A path is worthy if it has a better sort order (better pathkeys) or + * cheaper cost (on either dimension), or generates fewer rows, than any + * existing path that has the same or superset parameterization rels. + * We also consider parallel-safe paths more worthy than others. + * + * We also remove from the rel's pathlist any old paths that are dominated + * by new_path --- that is, new_path is cheaper, at least as well ordered, + * generates no more rows, requires no outer rels not required by the old + * path, and is no less parallel-safe. + * + * In most cases, a path with a superset parameterization will generate + * fewer rows (since it has more join clauses to apply), so that those two + * figures of merit move in opposite directions; this means that a path of + * one parameterization can seldom dominate a path of another. But such + * cases do arise, so we make the full set of checks anyway. + * + * There are two policy decisions embedded in this function, along with + * its sibling add_path_precheck. First, we treat all parameterized paths + * as having NIL pathkeys, so that they cannot win comparisons on the + * basis of sort order. This is to reduce the number of parameterized + * paths that are kept; see discussion in src/backend/optimizer/README. + * + * Second, we only consider cheap startup cost to be interesting if + * parent_rel->consider_startup is true for an unparameterized path, or + * parent_rel->consider_param_startup is true for a parameterized one. + * Again, this allows discarding useless paths sooner. + * + * The pathlist is kept sorted by total_cost, with cheaper paths + * at the front. Within this routine, that's simply a speed hack: + * doing it that way makes it more likely that we will reject an inferior + * path after a few comparisons, rather than many comparisons. + * However, add_path_precheck relies on this ordering to exit early + * when possible. + * + * NOTE: discarded Path objects are immediately pfree'd to reduce planner + * memory consumption. We dare not try to free the substructure of a Path, + * since much of it may be shared with other Paths or the query tree itself; + * but just recycling discarded Path nodes is a very useful savings in + * a large join tree. We can recycle the List nodes of pathlist, too. + * + * As noted in optimizer/README, deleting a previously-accepted Path is + * safe because we know that Paths of this rel cannot yet be referenced + * from any other rel, such as a higher-level join. However, in some cases + * it is possible that a Path is referenced by another Path for its own + * rel; we must not delete such a Path, even if it is dominated by the new + * Path. Currently this occurs only for IndexPath objects, which may be + * referenced as children of BitmapHeapPaths as well as being paths in + * their own right. Hence, we don't pfree IndexPaths when rejecting them. + * + * 'parent_rel' is the relation entry to which the path corresponds. + * 'new_path' is a potential path for parent_rel. + * + * Returns nothing, but modifies parent_rel->pathlist. + */ +void +add_path(RelOptInfo *parent_rel, Path *new_path) +{ + bool accept_new = true; /* unless we find a superior old path */ + int insert_at = 0; /* where to insert new item */ + List *new_path_pathkeys; + ListCell *p1; + + /* + * This is a convenient place to check for query cancel --- no part of the + * planner goes very long without calling add_path(). + */ + CHECK_FOR_INTERRUPTS(); + + /* Pretend parameterized paths have no pathkeys, per comment above */ + new_path_pathkeys = new_path->param_info ? NIL : new_path->pathkeys; + + /* + * Loop to check proposed new path against old paths. Note it is possible + * for more than one old path to be tossed out because new_path dominates + * it. + */ + foreach(p1, parent_rel->pathlist) + { + Path *old_path = (Path *) lfirst(p1); + bool remove_old = false; /* unless new proves superior */ + PathCostComparison costcmp; + PathKeysComparison keyscmp; + BMS_Comparison outercmp; + + /* + * Do a fuzzy cost comparison with standard fuzziness limit. + */ + costcmp = compare_path_costs_fuzzily(new_path, old_path, + STD_FUZZ_FACTOR); + + /* + * If the two paths compare differently for startup and total cost, + * then we want to keep both, and we can skip comparing pathkeys and + * required_outer rels. If they compare the same, proceed with the + * other comparisons. Row count is checked last. (We make the tests + * in this order because the cost comparison is most likely to turn + * out "different", and the pathkeys comparison next most likely. As + * explained above, row count very seldom makes a difference, so even + * though it's cheap to compare there's not much point in checking it + * earlier.) + */ + if (costcmp != COSTS_DIFFERENT) + { + /* Similarly check to see if either dominates on pathkeys */ + List *old_path_pathkeys; + + old_path_pathkeys = old_path->param_info ? NIL : old_path->pathkeys; + keyscmp = compare_pathkeys(new_path_pathkeys, + old_path_pathkeys); + if (keyscmp != PATHKEYS_DIFFERENT) + { + switch (costcmp) + { + case COSTS_EQUAL: + outercmp = bms_subset_compare(PATH_REQ_OUTER(new_path), + PATH_REQ_OUTER(old_path)); + if (keyscmp == PATHKEYS_BETTER1) + { + if ((outercmp == BMS_EQUAL || + outercmp == BMS_SUBSET1) && + new_path->rows <= old_path->rows && + new_path->parallel_safe >= old_path->parallel_safe) + remove_old = true; /* new dominates old */ + } + else if (keyscmp == PATHKEYS_BETTER2) + { + if ((outercmp == BMS_EQUAL || + outercmp == BMS_SUBSET2) && + new_path->rows >= old_path->rows && + new_path->parallel_safe <= old_path->parallel_safe) + accept_new = false; /* old dominates new */ + } + else /* keyscmp == PATHKEYS_EQUAL */ + { + if (outercmp == BMS_EQUAL) + { + /* + * Same pathkeys and outer rels, and fuzzily + * the same cost, so keep just one; to decide + * which, first check parallel-safety, then + * rows, then do a fuzzy cost comparison with + * very small fuzz limit. (We used to do an + * exact cost comparison, but that results in + * annoying platform-specific plan variations + * due to roundoff in the cost estimates.) If + * things are still tied, arbitrarily keep + * only the old path. Notice that we will + * keep only the old path even if the + * less-fuzzy comparison decides the startup + * and total costs compare differently. + */ + if (new_path->parallel_safe > + old_path->parallel_safe) + remove_old = true; /* new dominates old */ + else if (new_path->parallel_safe < + old_path->parallel_safe) + accept_new = false; /* old dominates new */ + else if (new_path->rows < old_path->rows) + remove_old = true; /* new dominates old */ + else if (new_path->rows > old_path->rows) + accept_new = false; /* old dominates new */ + else if (compare_path_costs_fuzzily(new_path, + old_path, + 1.0000000001) == COSTS_BETTER1) + remove_old = true; /* new dominates old */ + else + accept_new = false; /* old equals or + * dominates new */ + } + else if (outercmp == BMS_SUBSET1 && + new_path->rows <= old_path->rows && + new_path->parallel_safe >= old_path->parallel_safe) + remove_old = true; /* new dominates old */ + else if (outercmp == BMS_SUBSET2 && + new_path->rows >= old_path->rows && + new_path->parallel_safe <= old_path->parallel_safe) + accept_new = false; /* old dominates new */ + /* else different parameterizations, keep both */ + } + break; + case COSTS_BETTER1: + if (keyscmp != PATHKEYS_BETTER2) + { + outercmp = bms_subset_compare(PATH_REQ_OUTER(new_path), + PATH_REQ_OUTER(old_path)); + if ((outercmp == BMS_EQUAL || + outercmp == BMS_SUBSET1) && + new_path->rows <= old_path->rows && + new_path->parallel_safe >= old_path->parallel_safe) + remove_old = true; /* new dominates old */ + } + break; + case COSTS_BETTER2: + if (keyscmp != PATHKEYS_BETTER1) + { + outercmp = bms_subset_compare(PATH_REQ_OUTER(new_path), + PATH_REQ_OUTER(old_path)); + if ((outercmp == BMS_EQUAL || + outercmp == BMS_SUBSET2) && + new_path->rows >= old_path->rows && + new_path->parallel_safe <= old_path->parallel_safe) + accept_new = false; /* old dominates new */ + } + break; + case COSTS_DIFFERENT: + + /* + * can't get here, but keep this case to keep compiler + * quiet + */ + break; + } + } + } + + /* + * Remove current element from pathlist if dominated by new. + */ + if (remove_old) + { + parent_rel->pathlist = foreach_delete_current(parent_rel->pathlist, + p1); + + /* + * Delete the data pointed-to by the deleted cell, if possible + */ + if (!IsA(old_path, IndexPath)) + pfree(old_path); + } + else + { + /* new belongs after this old path if it has cost >= old's */ + if (new_path->total_cost >= old_path->total_cost) + insert_at = foreach_current_index(p1) + 1; + } + + /* + * If we found an old path that dominates new_path, we can quit + * scanning the pathlist; we will not add new_path, and we assume + * new_path cannot dominate any other elements of the pathlist. + */ + if (!accept_new) + break; + } + + if (accept_new) + { + /* Accept the new path: insert it at proper place in pathlist */ + parent_rel->pathlist = + list_insert_nth(parent_rel->pathlist, insert_at, new_path); + } + else + { + /* Reject and recycle the new path */ + if (!IsA(new_path, IndexPath)) + pfree(new_path); + } +} + +/* + * add_path_precheck + * Check whether a proposed new path could possibly get accepted. + * We assume we know the path's pathkeys and parameterization accurately, + * and have lower bounds for its costs. + * + * Note that we do not know the path's rowcount, since getting an estimate for + * that is too expensive to do before prechecking. We assume here that paths + * of a superset parameterization will generate fewer rows; if that holds, + * then paths with different parameterizations cannot dominate each other + * and so we can simply ignore existing paths of another parameterization. + * (In the infrequent cases where that rule of thumb fails, add_path will + * get rid of the inferior path.) + * + * At the time this is called, we haven't actually built a Path structure, + * so the required information has to be passed piecemeal. + */ +bool +add_path_precheck(RelOptInfo *parent_rel, + Cost startup_cost, Cost total_cost, + List *pathkeys, Relids required_outer) +{ + List *new_path_pathkeys; + bool consider_startup; + ListCell *p1; + + /* Pretend parameterized paths have no pathkeys, per add_path policy */ + new_path_pathkeys = required_outer ? NIL : pathkeys; + + /* Decide whether new path's startup cost is interesting */ + consider_startup = required_outer ? parent_rel->consider_param_startup : parent_rel->consider_startup; + + foreach(p1, parent_rel->pathlist) + { + Path *old_path = (Path *) lfirst(p1); + PathKeysComparison keyscmp; + + /* + * We are looking for an old_path with the same parameterization (and + * by assumption the same rowcount) that dominates the new path on + * pathkeys as well as both cost metrics. If we find one, we can + * reject the new path. + * + * Cost comparisons here should match compare_path_costs_fuzzily. + */ + if (total_cost > old_path->total_cost * STD_FUZZ_FACTOR) + { + /* new path can win on startup cost only if consider_startup */ + if (startup_cost > old_path->startup_cost * STD_FUZZ_FACTOR || + !consider_startup) + { + /* new path loses on cost, so check pathkeys... */ + List *old_path_pathkeys; + + old_path_pathkeys = old_path->param_info ? NIL : old_path->pathkeys; + keyscmp = compare_pathkeys(new_path_pathkeys, + old_path_pathkeys); + if (keyscmp == PATHKEYS_EQUAL || + keyscmp == PATHKEYS_BETTER2) + { + /* new path does not win on pathkeys... */ + if (bms_equal(required_outer, PATH_REQ_OUTER(old_path))) + { + /* Found an old path that dominates the new one */ + return false; + } + } + } + } + else + { + /* + * Since the pathlist is sorted by total_cost, we can stop looking + * once we reach a path with a total_cost larger than the new + * path's. + */ + break; + } + } + + return true; +} + +/* + * add_partial_path + * Like add_path, our goal here is to consider whether a path is worthy + * of being kept around, but the considerations here are a bit different. + * A partial path is one which can be executed in any number of workers in + * parallel such that each worker will generate a subset of the path's + * overall result. + * + * As in add_path, the partial_pathlist is kept sorted with the cheapest + * total path in front. This is depended on by multiple places, which + * just take the front entry as the cheapest path without searching. + * + * We don't generate parameterized partial paths for several reasons. Most + * importantly, they're not safe to execute, because there's nothing to + * make sure that a parallel scan within the parameterized portion of the + * plan is running with the same value in every worker at the same time. + * Fortunately, it seems unlikely to be worthwhile anyway, because having + * each worker scan the entire outer relation and a subset of the inner + * relation will generally be a terrible plan. The inner (parameterized) + * side of the plan will be small anyway. There could be rare cases where + * this wins big - e.g. if join order constraints put a 1-row relation on + * the outer side of the topmost join with a parameterized plan on the inner + * side - but we'll have to be content not to handle such cases until + * somebody builds an executor infrastructure that can cope with them. + * + * Because we don't consider parameterized paths here, we also don't + * need to consider the row counts as a measure of quality: every path will + * produce the same number of rows. Neither do we need to consider startup + * costs: parallelism is only used for plans that will be run to completion. + * Therefore, this routine is much simpler than add_path: it needs to + * consider only pathkeys and total cost. + * + * As with add_path, we pfree paths that are found to be dominated by + * another partial path; this requires that there be no other references to + * such paths yet. Hence, GatherPaths must not be created for a rel until + * we're done creating all partial paths for it. Unlike add_path, we don't + * take an exception for IndexPaths as partial index paths won't be + * referenced by partial BitmapHeapPaths. + */ +void +add_partial_path(RelOptInfo *parent_rel, Path *new_path) +{ + bool accept_new = true; /* unless we find a superior old path */ + int insert_at = 0; /* where to insert new item */ + ListCell *p1; + + /* Check for query cancel. */ + CHECK_FOR_INTERRUPTS(); + + /* Path to be added must be parallel safe. */ + Assert(new_path->parallel_safe); + + /* Relation should be OK for parallelism, too. */ + Assert(parent_rel->consider_parallel); + + /* + * As in add_path, throw out any paths which are dominated by the new + * path, but throw out the new path if some existing path dominates it. + */ + foreach(p1, parent_rel->partial_pathlist) + { + Path *old_path = (Path *) lfirst(p1); + bool remove_old = false; /* unless new proves superior */ + PathKeysComparison keyscmp; + + /* Compare pathkeys. */ + keyscmp = compare_pathkeys(new_path->pathkeys, old_path->pathkeys); + + /* Unless pathkeys are incompatible, keep just one of the two paths. */ + if (keyscmp != PATHKEYS_DIFFERENT) + { + if (new_path->total_cost > old_path->total_cost * STD_FUZZ_FACTOR) + { + /* New path costs more; keep it only if pathkeys are better. */ + if (keyscmp != PATHKEYS_BETTER1) + accept_new = false; + } + else if (old_path->total_cost > new_path->total_cost + * STD_FUZZ_FACTOR) + { + /* Old path costs more; keep it only if pathkeys are better. */ + if (keyscmp != PATHKEYS_BETTER2) + remove_old = true; + } + else if (keyscmp == PATHKEYS_BETTER1) + { + /* Costs are about the same, new path has better pathkeys. */ + remove_old = true; + } + else if (keyscmp == PATHKEYS_BETTER2) + { + /* Costs are about the same, old path has better pathkeys. */ + accept_new = false; + } + else if (old_path->total_cost > new_path->total_cost * 1.0000000001) + { + /* Pathkeys are the same, and the old path costs more. */ + remove_old = true; + } + else + { + /* + * Pathkeys are the same, and new path isn't materially + * cheaper. + */ + accept_new = false; + } + } + + /* + * Remove current element from partial_pathlist if dominated by new. + */ + if (remove_old) + { + parent_rel->partial_pathlist = + foreach_delete_current(parent_rel->partial_pathlist, p1); + pfree(old_path); + } + else + { + /* new belongs after this old path if it has cost >= old's */ + if (new_path->total_cost >= old_path->total_cost) + insert_at = foreach_current_index(p1) + 1; + } + + /* + * If we found an old path that dominates new_path, we can quit + * scanning the partial_pathlist; we will not add new_path, and we + * assume new_path cannot dominate any later path. + */ + if (!accept_new) + break; + } + + if (accept_new) + { + /* Accept the new path: insert it at proper place */ + parent_rel->partial_pathlist = + list_insert_nth(parent_rel->partial_pathlist, insert_at, new_path); + } + else + { + /* Reject and recycle the new path */ + pfree(new_path); + } +} + +/* + * add_partial_path_precheck + * Check whether a proposed new partial path could possibly get accepted. + * + * Unlike add_path_precheck, we can ignore startup cost and parameterization, + * since they don't matter for partial paths (see add_partial_path). But + * we do want to make sure we don't add a partial path if there's already + * a complete path that dominates it, since in that case the proposed path + * is surely a loser. + */ +bool +add_partial_path_precheck(RelOptInfo *parent_rel, Cost total_cost, + List *pathkeys) +{ + ListCell *p1; + + /* + * Our goal here is twofold. First, we want to find out whether this path + * is clearly inferior to some existing partial path. If so, we want to + * reject it immediately. Second, we want to find out whether this path + * is clearly superior to some existing partial path -- at least, modulo + * final cost computations. If so, we definitely want to consider it. + * + * Unlike add_path(), we always compare pathkeys here. This is because we + * expect partial_pathlist to be very short, and getting a definitive + * answer at this stage avoids the need to call add_path_precheck. + */ + foreach(p1, parent_rel->partial_pathlist) + { + Path *old_path = (Path *) lfirst(p1); + PathKeysComparison keyscmp; + + keyscmp = compare_pathkeys(pathkeys, old_path->pathkeys); + if (keyscmp != PATHKEYS_DIFFERENT) + { + if (total_cost > old_path->total_cost * STD_FUZZ_FACTOR && + keyscmp != PATHKEYS_BETTER1) + return false; + if (old_path->total_cost > total_cost * STD_FUZZ_FACTOR && + keyscmp != PATHKEYS_BETTER2) + return true; + } + } + + /* + * This path is neither clearly inferior to an existing partial path nor + * clearly good enough that it might replace one. Compare it to + * non-parallel plans. If it loses even before accounting for the cost of + * the Gather node, we should definitely reject it. + * + * Note that we pass the total_cost to add_path_precheck twice. This is + * because it's never advantageous to consider the startup cost of a + * partial path; the resulting plans, if run in parallel, will be run to + * completion. + */ + if (!add_path_precheck(parent_rel, total_cost, total_cost, pathkeys, + NULL)) + return false; + + return true; +} + + +/***************************************************************************** + * PATH NODE CREATION ROUTINES + *****************************************************************************/ + +/* + * create_seqscan_path + * Creates a path corresponding to a sequential scan, returning the + * pathnode. + */ +Path * +create_seqscan_path(PlannerInfo *root, RelOptInfo *rel, + Relids required_outer, int parallel_workers) +{ + Path *pathnode = makeNode(Path); + + pathnode->pathtype = T_SeqScan; + pathnode->parent = rel; + pathnode->pathtarget = rel->reltarget; + pathnode->param_info = get_baserel_parampathinfo(root, rel, + required_outer); + pathnode->parallel_aware = (parallel_workers > 0); + pathnode->parallel_safe = rel->consider_parallel; + pathnode->parallel_workers = parallel_workers; + pathnode->pathkeys = NIL; /* seqscan has unordered result */ + + cost_seqscan(pathnode, root, rel, pathnode->param_info); + + return pathnode; +} + +/* + * create_samplescan_path + * Creates a path node for a sampled table scan. + */ +Path * +create_samplescan_path(PlannerInfo *root, RelOptInfo *rel, Relids required_outer) +{ + Path *pathnode = makeNode(Path); + + pathnode->pathtype = T_SampleScan; + pathnode->parent = rel; + pathnode->pathtarget = rel->reltarget; + pathnode->param_info = get_baserel_parampathinfo(root, rel, + required_outer); + pathnode->parallel_aware = false; + pathnode->parallel_safe = rel->consider_parallel; + pathnode->parallel_workers = 0; + pathnode->pathkeys = NIL; /* samplescan has unordered result */ + + cost_samplescan(pathnode, root, rel, pathnode->param_info); + + return pathnode; +} + +/* + * create_index_path + * Creates a path node for an index scan. + * + * 'index' is a usable index. + * 'indexclauses' is a list of IndexClause nodes representing clauses + * to be enforced as qual conditions in the scan. + * 'indexorderbys' is a list of bare expressions (no RestrictInfos) + * to be used as index ordering operators in the scan. + * 'indexorderbycols' is an integer list of index column numbers (zero based) + * the ordering operators can be used with. + * 'pathkeys' describes the ordering of the path. + * 'indexscandir' is ForwardScanDirection or BackwardScanDirection + * for an ordered index, or NoMovementScanDirection for + * an unordered index. + * 'indexonly' is true if an index-only scan is wanted. + * 'required_outer' is the set of outer relids for a parameterized path. + * 'loop_count' is the number of repetitions of the indexscan to factor into + * estimates of caching behavior. + * 'partial_path' is true if constructing a parallel index scan path. + * + * Returns the new path node. + */ +IndexPath * +create_index_path(PlannerInfo *root, + IndexOptInfo *index, + List *indexclauses, + List *indexorderbys, + List *indexorderbycols, + List *pathkeys, + ScanDirection indexscandir, + bool indexonly, + Relids required_outer, + double loop_count, + bool partial_path) +{ + IndexPath *pathnode = makeNode(IndexPath); + RelOptInfo *rel = index->rel; + + pathnode->path.pathtype = indexonly ? T_IndexOnlyScan : T_IndexScan; + pathnode->path.parent = rel; + pathnode->path.pathtarget = rel->reltarget; + pathnode->path.param_info = get_baserel_parampathinfo(root, rel, + required_outer); + pathnode->path.parallel_aware = false; + pathnode->path.parallel_safe = rel->consider_parallel; + pathnode->path.parallel_workers = 0; + pathnode->path.pathkeys = pathkeys; + + pathnode->indexinfo = index; + pathnode->indexclauses = indexclauses; + pathnode->indexorderbys = indexorderbys; + pathnode->indexorderbycols = indexorderbycols; + pathnode->indexscandir = indexscandir; + + cost_index(pathnode, root, loop_count, partial_path); + + return pathnode; +} + +/* + * create_bitmap_heap_path + * Creates a path node for a bitmap scan. + * + * 'bitmapqual' is a tree of IndexPath, BitmapAndPath, and BitmapOrPath nodes. + * 'required_outer' is the set of outer relids for a parameterized path. + * 'loop_count' is the number of repetitions of the indexscan to factor into + * estimates of caching behavior. + * + * loop_count should match the value used when creating the component + * IndexPaths. + */ +BitmapHeapPath * +create_bitmap_heap_path(PlannerInfo *root, + RelOptInfo *rel, + Path *bitmapqual, + Relids required_outer, + double loop_count, + int parallel_degree) +{ + BitmapHeapPath *pathnode = makeNode(BitmapHeapPath); + + pathnode->path.pathtype = T_BitmapHeapScan; + pathnode->path.parent = rel; + pathnode->path.pathtarget = rel->reltarget; + pathnode->path.param_info = get_baserel_parampathinfo(root, rel, + required_outer); + pathnode->path.parallel_aware = (parallel_degree > 0); + pathnode->path.parallel_safe = rel->consider_parallel; + pathnode->path.parallel_workers = parallel_degree; + pathnode->path.pathkeys = NIL; /* always unordered */ + + pathnode->bitmapqual = bitmapqual; + + cost_bitmap_heap_scan(&pathnode->path, root, rel, + pathnode->path.param_info, + bitmapqual, loop_count); + + return pathnode; +} + +/* + * create_bitmap_and_path + * Creates a path node representing a BitmapAnd. + */ +BitmapAndPath * +create_bitmap_and_path(PlannerInfo *root, + RelOptInfo *rel, + List *bitmapquals) +{ + BitmapAndPath *pathnode = makeNode(BitmapAndPath); + Relids required_outer = NULL; + ListCell *lc; + + pathnode->path.pathtype = T_BitmapAnd; + pathnode->path.parent = rel; + pathnode->path.pathtarget = rel->reltarget; + + /* + * Identify the required outer rels as the union of what the child paths + * depend on. (Alternatively, we could insist that the caller pass this + * in, but it's more convenient and reliable to compute it here.) + */ + foreach(lc, bitmapquals) + { + Path *bitmapqual = (Path *) lfirst(lc); + + required_outer = bms_add_members(required_outer, + PATH_REQ_OUTER(bitmapqual)); + } + pathnode->path.param_info = get_baserel_parampathinfo(root, rel, + required_outer); + + /* + * Currently, a BitmapHeapPath, BitmapAndPath, or BitmapOrPath will be + * parallel-safe if and only if rel->consider_parallel is set. So, we can + * set the flag for this path based only on the relation-level flag, + * without actually iterating over the list of children. + */ + pathnode->path.parallel_aware = false; + pathnode->path.parallel_safe = rel->consider_parallel; + pathnode->path.parallel_workers = 0; + + pathnode->path.pathkeys = NIL; /* always unordered */ + + pathnode->bitmapquals = bitmapquals; + + /* this sets bitmapselectivity as well as the regular cost fields: */ + cost_bitmap_and_node(pathnode, root); + + return pathnode; +} + +/* + * create_bitmap_or_path + * Creates a path node representing a BitmapOr. + */ +BitmapOrPath * +create_bitmap_or_path(PlannerInfo *root, + RelOptInfo *rel, + List *bitmapquals) +{ + BitmapOrPath *pathnode = makeNode(BitmapOrPath); + Relids required_outer = NULL; + ListCell *lc; + + pathnode->path.pathtype = T_BitmapOr; + pathnode->path.parent = rel; + pathnode->path.pathtarget = rel->reltarget; + + /* + * Identify the required outer rels as the union of what the child paths + * depend on. (Alternatively, we could insist that the caller pass this + * in, but it's more convenient and reliable to compute it here.) + */ + foreach(lc, bitmapquals) + { + Path *bitmapqual = (Path *) lfirst(lc); + + required_outer = bms_add_members(required_outer, + PATH_REQ_OUTER(bitmapqual)); + } + pathnode->path.param_info = get_baserel_parampathinfo(root, rel, + required_outer); + + /* + * Currently, a BitmapHeapPath, BitmapAndPath, or BitmapOrPath will be + * parallel-safe if and only if rel->consider_parallel is set. So, we can + * set the flag for this path based only on the relation-level flag, + * without actually iterating over the list of children. + */ + pathnode->path.parallel_aware = false; + pathnode->path.parallel_safe = rel->consider_parallel; + pathnode->path.parallel_workers = 0; + + pathnode->path.pathkeys = NIL; /* always unordered */ + + pathnode->bitmapquals = bitmapquals; + + /* this sets bitmapselectivity as well as the regular cost fields: */ + cost_bitmap_or_node(pathnode, root); + + return pathnode; +} + +/* + * create_tidscan_path + * Creates a path corresponding to a scan by TID, returning the pathnode. + */ +TidPath * +create_tidscan_path(PlannerInfo *root, RelOptInfo *rel, List *tidquals, + Relids required_outer) +{ + TidPath *pathnode = makeNode(TidPath); + + pathnode->path.pathtype = T_TidScan; + pathnode->path.parent = rel; + pathnode->path.pathtarget = rel->reltarget; + pathnode->path.param_info = get_baserel_parampathinfo(root, rel, + required_outer); + pathnode->path.parallel_aware = false; + pathnode->path.parallel_safe = rel->consider_parallel; + pathnode->path.parallel_workers = 0; + pathnode->path.pathkeys = NIL; /* always unordered */ + + pathnode->tidquals = tidquals; + + cost_tidscan(&pathnode->path, root, rel, tidquals, + pathnode->path.param_info); + + return pathnode; +} + +/* + * create_tidrangescan_path + * Creates a path corresponding to a scan by a range of TIDs, returning + * the pathnode. + */ +TidRangePath * +create_tidrangescan_path(PlannerInfo *root, RelOptInfo *rel, + List *tidrangequals, Relids required_outer) +{ + TidRangePath *pathnode = makeNode(TidRangePath); + + pathnode->path.pathtype = T_TidRangeScan; + pathnode->path.parent = rel; + pathnode->path.pathtarget = rel->reltarget; + pathnode->path.param_info = get_baserel_parampathinfo(root, rel, + required_outer); + pathnode->path.parallel_aware = false; + pathnode->path.parallel_safe = rel->consider_parallel; + pathnode->path.parallel_workers = 0; + pathnode->path.pathkeys = NIL; /* always unordered */ + + pathnode->tidrangequals = tidrangequals; + + cost_tidrangescan(&pathnode->path, root, rel, tidrangequals, + pathnode->path.param_info); + + return pathnode; +} + +/* + * create_append_path + * Creates a path corresponding to an Append plan, returning the + * pathnode. + * + * Note that we must handle subpaths = NIL, representing a dummy access path. + * Also, there are callers that pass root = NULL. + */ +AppendPath * +create_append_path(PlannerInfo *root, + RelOptInfo *rel, + List *subpaths, List *partial_subpaths, + List *pathkeys, Relids required_outer, + int parallel_workers, bool parallel_aware, + double rows) +{ + AppendPath *pathnode = makeNode(AppendPath); + ListCell *l; + + Assert(!parallel_aware || parallel_workers > 0); + + pathnode->path.pathtype = T_Append; + pathnode->path.parent = rel; + pathnode->path.pathtarget = rel->reltarget; + + /* + * When generating an Append path for a partitioned table, there may be + * parameterized quals that are useful for run-time pruning. Hence, + * compute path.param_info the same way as for any other baserel, so that + * such quals will be available for make_partition_pruneinfo(). (This + * would not work right for a non-baserel, ie a scan on a non-leaf child + * partition, and it's not necessary anyway in that case. Must skip it if + * we don't have "root", too.) + */ + if (root && rel->reloptkind == RELOPT_BASEREL && IS_PARTITIONED_REL(rel)) + pathnode->path.param_info = get_baserel_parampathinfo(root, + rel, + required_outer); + else + pathnode->path.param_info = get_appendrel_parampathinfo(rel, + required_outer); + + pathnode->path.parallel_aware = parallel_aware; + pathnode->path.parallel_safe = rel->consider_parallel; + pathnode->path.parallel_workers = parallel_workers; + pathnode->path.pathkeys = pathkeys; + + /* + * For parallel append, non-partial paths are sorted by descending total + * costs. That way, the total time to finish all non-partial paths is + * minimized. Also, the partial paths are sorted by descending startup + * costs. There may be some paths that require to do startup work by a + * single worker. In such case, it's better for workers to choose the + * expensive ones first, whereas the leader should choose the cheapest + * startup plan. + */ + if (pathnode->path.parallel_aware) + { + /* + * We mustn't fiddle with the order of subpaths when the Append has + * pathkeys. The order they're listed in is critical to keeping the + * pathkeys valid. + */ + Assert(pathkeys == NIL); + + list_sort(subpaths, append_total_cost_compare); + list_sort(partial_subpaths, append_startup_cost_compare); + } + pathnode->first_partial_path = list_length(subpaths); + pathnode->subpaths = list_concat(subpaths, partial_subpaths); + + /* + * Apply query-wide LIMIT if known and path is for sole base relation. + * (Handling this at this low level is a bit klugy.) + */ + if (root != NULL && bms_equal(rel->relids, root->all_baserels)) + pathnode->limit_tuples = root->limit_tuples; + else + pathnode->limit_tuples = -1.0; + + foreach(l, pathnode->subpaths) + { + Path *subpath = (Path *) lfirst(l); + + pathnode->path.parallel_safe = pathnode->path.parallel_safe && + subpath->parallel_safe; + + /* All child paths must have same parameterization */ + Assert(bms_equal(PATH_REQ_OUTER(subpath), required_outer)); + } + + Assert(!parallel_aware || pathnode->path.parallel_safe); + + /* + * If there's exactly one child path, the Append is a no-op and will be + * discarded later (in setrefs.c); therefore, we can inherit the child's + * size and cost, as well as its pathkeys if any (overriding whatever the + * caller might've said). Otherwise, we must do the normal costsize + * calculation. + */ + if (list_length(pathnode->subpaths) == 1) + { + Path *child = (Path *) linitial(pathnode->subpaths); + + pathnode->path.rows = child->rows; + pathnode->path.startup_cost = child->startup_cost; + pathnode->path.total_cost = child->total_cost; + pathnode->path.pathkeys = child->pathkeys; + } + else + cost_append(pathnode); + + /* If the caller provided a row estimate, override the computed value. */ + if (rows >= 0) + pathnode->path.rows = rows; + + return pathnode; +} + +/* + * append_total_cost_compare + * list_sort comparator for sorting append child paths + * by total_cost descending + * + * For equal total costs, we fall back to comparing startup costs; if those + * are equal too, break ties using bms_compare on the paths' relids. + * (This is to avoid getting unpredictable results from list_sort.) + */ +static int +append_total_cost_compare(const ListCell *a, const ListCell *b) +{ + Path *path1 = (Path *) lfirst(a); + Path *path2 = (Path *) lfirst(b); + int cmp; + + cmp = compare_path_costs(path1, path2, TOTAL_COST); + if (cmp != 0) + return -cmp; + return bms_compare(path1->parent->relids, path2->parent->relids); +} + +/* + * append_startup_cost_compare + * list_sort comparator for sorting append child paths + * by startup_cost descending + * + * For equal startup costs, we fall back to comparing total costs; if those + * are equal too, break ties using bms_compare on the paths' relids. + * (This is to avoid getting unpredictable results from list_sort.) + */ +static int +append_startup_cost_compare(const ListCell *a, const ListCell *b) +{ + Path *path1 = (Path *) lfirst(a); + Path *path2 = (Path *) lfirst(b); + int cmp; + + cmp = compare_path_costs(path1, path2, STARTUP_COST); + if (cmp != 0) + return -cmp; + return bms_compare(path1->parent->relids, path2->parent->relids); +} + +/* + * create_merge_append_path + * Creates a path corresponding to a MergeAppend plan, returning the + * pathnode. + */ +MergeAppendPath * +create_merge_append_path(PlannerInfo *root, + RelOptInfo *rel, + List *subpaths, + List *pathkeys, + Relids required_outer) +{ + MergeAppendPath *pathnode = makeNode(MergeAppendPath); + Cost input_startup_cost; + Cost input_total_cost; + ListCell *l; + + pathnode->path.pathtype = T_MergeAppend; + pathnode->path.parent = rel; + pathnode->path.pathtarget = rel->reltarget; + pathnode->path.param_info = get_appendrel_parampathinfo(rel, + required_outer); + pathnode->path.parallel_aware = false; + pathnode->path.parallel_safe = rel->consider_parallel; + pathnode->path.parallel_workers = 0; + pathnode->path.pathkeys = pathkeys; + pathnode->subpaths = subpaths; + + /* + * Apply query-wide LIMIT if known and path is for sole base relation. + * (Handling this at this low level is a bit klugy.) + */ + if (bms_equal(rel->relids, root->all_baserels)) + pathnode->limit_tuples = root->limit_tuples; + else + pathnode->limit_tuples = -1.0; + + /* + * Add up the sizes and costs of the input paths. + */ + pathnode->path.rows = 0; + input_startup_cost = 0; + input_total_cost = 0; + foreach(l, subpaths) + { + Path *subpath = (Path *) lfirst(l); + + pathnode->path.rows += subpath->rows; + pathnode->path.parallel_safe = pathnode->path.parallel_safe && + subpath->parallel_safe; + + if (pathkeys_contained_in(pathkeys, subpath->pathkeys)) + { + /* Subpath is adequately ordered, we won't need to sort it */ + input_startup_cost += subpath->startup_cost; + input_total_cost += subpath->total_cost; + } + else + { + /* We'll need to insert a Sort node, so include cost for that */ + Path sort_path; /* dummy for result of cost_sort */ + + cost_sort(&sort_path, + root, + pathkeys, + subpath->total_cost, + subpath->parent->tuples, + subpath->pathtarget->width, + 0.0, + work_mem, + pathnode->limit_tuples); + input_startup_cost += sort_path.startup_cost; + input_total_cost += sort_path.total_cost; + } + + /* All child paths must have same parameterization */ + Assert(bms_equal(PATH_REQ_OUTER(subpath), required_outer)); + } + + /* + * Now we can compute total costs of the MergeAppend. If there's exactly + * one child path, the MergeAppend is a no-op and will be discarded later + * (in setrefs.c); otherwise we do the normal cost calculation. + */ + if (list_length(subpaths) == 1) + { + pathnode->path.startup_cost = input_startup_cost; + pathnode->path.total_cost = input_total_cost; + } + else + cost_merge_append(&pathnode->path, root, + pathkeys, list_length(subpaths), + input_startup_cost, input_total_cost, + pathnode->path.rows); + + return pathnode; +} + +/* + * create_group_result_path + * Creates a path representing a Result-and-nothing-else plan. + * + * This is only used for degenerate grouping cases, in which we know we + * need to produce one result row, possibly filtered by a HAVING qual. + */ +GroupResultPath * +create_group_result_path(PlannerInfo *root, RelOptInfo *rel, + PathTarget *target, List *havingqual) +{ + GroupResultPath *pathnode = makeNode(GroupResultPath); + + pathnode->path.pathtype = T_Result; + pathnode->path.parent = rel; + pathnode->path.pathtarget = target; + pathnode->path.param_info = NULL; /* there are no other rels... */ + pathnode->path.parallel_aware = false; + pathnode->path.parallel_safe = rel->consider_parallel; + pathnode->path.parallel_workers = 0; + pathnode->path.pathkeys = NIL; + pathnode->quals = havingqual; + + /* + * We can't quite use cost_resultscan() because the quals we want to + * account for are not baserestrict quals of the rel. Might as well just + * hack it here. + */ + pathnode->path.rows = 1; + pathnode->path.startup_cost = target->cost.startup; + pathnode->path.total_cost = target->cost.startup + + cpu_tuple_cost + target->cost.per_tuple; + + /* + * Add cost of qual, if any --- but we ignore its selectivity, since our + * rowcount estimate should be 1 no matter what the qual is. + */ + if (havingqual) + { + QualCost qual_cost; + + cost_qual_eval(&qual_cost, havingqual, root); + /* havingqual is evaluated once at startup */ + pathnode->path.startup_cost += qual_cost.startup + qual_cost.per_tuple; + pathnode->path.total_cost += qual_cost.startup + qual_cost.per_tuple; + } + + return pathnode; +} + +/* + * create_material_path + * Creates a path corresponding to a Material plan, returning the + * pathnode. + */ +MaterialPath * +create_material_path(RelOptInfo *rel, Path *subpath) +{ + MaterialPath *pathnode = makeNode(MaterialPath); + + Assert(subpath->parent == rel); + + pathnode->path.pathtype = T_Material; + pathnode->path.parent = rel; + pathnode->path.pathtarget = rel->reltarget; + pathnode->path.param_info = subpath->param_info; + pathnode->path.parallel_aware = false; + pathnode->path.parallel_safe = rel->consider_parallel && + subpath->parallel_safe; + pathnode->path.parallel_workers = subpath->parallel_workers; + pathnode->path.pathkeys = subpath->pathkeys; + + pathnode->subpath = subpath; + + cost_material(&pathnode->path, + subpath->startup_cost, + subpath->total_cost, + subpath->rows, + subpath->pathtarget->width); + + return pathnode; +} + +/* + * create_memoize_path + * Creates a path corresponding to a Memoize plan, returning the pathnode. + */ +MemoizePath * +create_memoize_path(PlannerInfo *root, RelOptInfo *rel, Path *subpath, + List *param_exprs, List *hash_operators, + bool singlerow, bool binary_mode, double calls) +{ + MemoizePath *pathnode = makeNode(MemoizePath); + + Assert(subpath->parent == rel); + + pathnode->path.pathtype = T_Memoize; + pathnode->path.parent = rel; + pathnode->path.pathtarget = rel->reltarget; + pathnode->path.param_info = subpath->param_info; + pathnode->path.parallel_aware = false; + pathnode->path.parallel_safe = rel->consider_parallel && + subpath->parallel_safe; + pathnode->path.parallel_workers = subpath->parallel_workers; + pathnode->path.pathkeys = subpath->pathkeys; + + pathnode->subpath = subpath; + pathnode->hash_operators = hash_operators; + pathnode->param_exprs = param_exprs; + pathnode->singlerow = singlerow; + pathnode->binary_mode = binary_mode; + pathnode->calls = calls; + + /* + * For now we set est_entries to 0. cost_memoize_rescan() does all the + * hard work to determine how many cache entries there are likely to be, + * so it seems best to leave it up to that function to fill this field in. + * If left at 0, the executor will make a guess at a good value. + */ + pathnode->est_entries = 0; + + /* + * Add a small additional charge for caching the first entry. All the + * harder calculations for rescans are performed in cost_memoize_rescan(). + */ + pathnode->path.startup_cost = subpath->startup_cost + cpu_tuple_cost; + pathnode->path.total_cost = subpath->total_cost + cpu_tuple_cost; + pathnode->path.rows = subpath->rows; + + return pathnode; +} + +/* + * create_unique_path + * Creates a path representing elimination of distinct rows from the + * input data. Distinct-ness is defined according to the needs of the + * semijoin represented by sjinfo. If it is not possible to identify + * how to make the data unique, NULL is returned. + * + * If used at all, this is likely to be called repeatedly on the same rel; + * and the input subpath should always be the same (the cheapest_total path + * for the rel). So we cache the result. + */ +UniquePath * +create_unique_path(PlannerInfo *root, RelOptInfo *rel, Path *subpath, + SpecialJoinInfo *sjinfo) +{ + UniquePath *pathnode; + Path sort_path; /* dummy for result of cost_sort */ + Path agg_path; /* dummy for result of cost_agg */ + MemoryContext oldcontext; + int numCols; + + /* Caller made a mistake if subpath isn't cheapest_total ... */ + Assert(subpath == rel->cheapest_total_path); + Assert(subpath->parent == rel); + /* ... or if SpecialJoinInfo is the wrong one */ + Assert(sjinfo->jointype == JOIN_SEMI); + Assert(bms_equal(rel->relids, sjinfo->syn_righthand)); + + /* If result already cached, return it */ + if (rel->cheapest_unique_path) + return (UniquePath *) rel->cheapest_unique_path; + + /* If it's not possible to unique-ify, return NULL */ + if (!(sjinfo->semi_can_btree || sjinfo->semi_can_hash)) + return NULL; + + /* + * When called during GEQO join planning, we are in a short-lived memory + * context. We must make sure that the path and any subsidiary data + * structures created for a baserel survive the GEQO cycle, else the + * baserel is trashed for future GEQO cycles. On the other hand, when we + * are creating those for a joinrel during GEQO, we don't want them to + * clutter the main planning context. Upshot is that the best solution is + * to explicitly allocate memory in the same context the given RelOptInfo + * is in. + */ + oldcontext = MemoryContextSwitchTo(GetMemoryChunkContext(rel)); + + pathnode = makeNode(UniquePath); + + pathnode->path.pathtype = T_Unique; + pathnode->path.parent = rel; + pathnode->path.pathtarget = rel->reltarget; + pathnode->path.param_info = subpath->param_info; + pathnode->path.parallel_aware = false; + pathnode->path.parallel_safe = rel->consider_parallel && + subpath->parallel_safe; + pathnode->path.parallel_workers = subpath->parallel_workers; + + /* + * Assume the output is unsorted, since we don't necessarily have pathkeys + * to represent it. (This might get overridden below.) + */ + pathnode->path.pathkeys = NIL; + + pathnode->subpath = subpath; + pathnode->in_operators = sjinfo->semi_operators; + pathnode->uniq_exprs = sjinfo->semi_rhs_exprs; + + /* + * If the input is a relation and it has a unique index that proves the + * semi_rhs_exprs are unique, then we don't need to do anything. Note + * that relation_has_unique_index_for automatically considers restriction + * clauses for the rel, as well. + */ + if (rel->rtekind == RTE_RELATION && sjinfo->semi_can_btree && + relation_has_unique_index_for(root, rel, NIL, + sjinfo->semi_rhs_exprs, + sjinfo->semi_operators)) + { + pathnode->umethod = UNIQUE_PATH_NOOP; + pathnode->path.rows = rel->rows; + pathnode->path.startup_cost = subpath->startup_cost; + pathnode->path.total_cost = subpath->total_cost; + pathnode->path.pathkeys = subpath->pathkeys; + + rel->cheapest_unique_path = (Path *) pathnode; + + MemoryContextSwitchTo(oldcontext); + + return pathnode; + } + + /* + * If the input is a subquery whose output must be unique already, then we + * don't need to do anything. The test for uniqueness has to consider + * exactly which columns we are extracting; for example "SELECT DISTINCT + * x,y" doesn't guarantee that x alone is distinct. So we cannot check for + * this optimization unless semi_rhs_exprs consists only of simple Vars + * referencing subquery outputs. (Possibly we could do something with + * expressions in the subquery outputs, too, but for now keep it simple.) + */ + if (rel->rtekind == RTE_SUBQUERY) + { + RangeTblEntry *rte = planner_rt_fetch(rel->relid, root); + + if (query_supports_distinctness(rte->subquery)) + { + List *sub_tlist_colnos; + + sub_tlist_colnos = translate_sub_tlist(sjinfo->semi_rhs_exprs, + rel->relid); + + if (sub_tlist_colnos && + query_is_distinct_for(rte->subquery, + sub_tlist_colnos, + sjinfo->semi_operators)) + { + pathnode->umethod = UNIQUE_PATH_NOOP; + pathnode->path.rows = rel->rows; + pathnode->path.startup_cost = subpath->startup_cost; + pathnode->path.total_cost = subpath->total_cost; + pathnode->path.pathkeys = subpath->pathkeys; + + rel->cheapest_unique_path = (Path *) pathnode; + + MemoryContextSwitchTo(oldcontext); + + return pathnode; + } + } + } + + /* Estimate number of output rows */ + pathnode->path.rows = estimate_num_groups(root, + sjinfo->semi_rhs_exprs, + rel->rows, + NULL, + NULL); + numCols = list_length(sjinfo->semi_rhs_exprs); + + if (sjinfo->semi_can_btree) + { + /* + * Estimate cost for sort+unique implementation + */ + cost_sort(&sort_path, root, NIL, + subpath->total_cost, + rel->rows, + subpath->pathtarget->width, + 0.0, + work_mem, + -1.0); + + /* + * Charge one cpu_operator_cost per comparison per input tuple. We + * assume all columns get compared at most of the tuples. (XXX + * probably this is an overestimate.) This should agree with + * create_upper_unique_path. + */ + sort_path.total_cost += cpu_operator_cost * rel->rows * numCols; + } + + if (sjinfo->semi_can_hash) + { + /* + * Estimate the overhead per hashtable entry at 64 bytes (same as in + * planner.c). + */ + int hashentrysize = subpath->pathtarget->width + 64; + + if (hashentrysize * pathnode->path.rows > get_hash_memory_limit()) + { + /* + * We should not try to hash. Hack the SpecialJoinInfo to + * remember this, in case we come through here again. + */ + sjinfo->semi_can_hash = false; + } + else + cost_agg(&agg_path, root, + AGG_HASHED, NULL, + numCols, pathnode->path.rows, + NIL, + subpath->startup_cost, + subpath->total_cost, + rel->rows, + subpath->pathtarget->width); + } + + if (sjinfo->semi_can_btree && sjinfo->semi_can_hash) + { + if (agg_path.total_cost < sort_path.total_cost) + pathnode->umethod = UNIQUE_PATH_HASH; + else + pathnode->umethod = UNIQUE_PATH_SORT; + } + else if (sjinfo->semi_can_btree) + pathnode->umethod = UNIQUE_PATH_SORT; + else if (sjinfo->semi_can_hash) + pathnode->umethod = UNIQUE_PATH_HASH; + else + { + /* we can get here only if we abandoned hashing above */ + MemoryContextSwitchTo(oldcontext); + return NULL; + } + + if (pathnode->umethod == UNIQUE_PATH_HASH) + { + pathnode->path.startup_cost = agg_path.startup_cost; + pathnode->path.total_cost = agg_path.total_cost; + } + else + { + pathnode->path.startup_cost = sort_path.startup_cost; + pathnode->path.total_cost = sort_path.total_cost; + } + + rel->cheapest_unique_path = (Path *) pathnode; + + MemoryContextSwitchTo(oldcontext); + + return pathnode; +} + +/* + * create_gather_merge_path + * + * Creates a path corresponding to a gather merge scan, returning + * the pathnode. + */ +GatherMergePath * +create_gather_merge_path(PlannerInfo *root, RelOptInfo *rel, Path *subpath, + PathTarget *target, List *pathkeys, + Relids required_outer, double *rows) +{ + GatherMergePath *pathnode = makeNode(GatherMergePath); + Cost input_startup_cost = 0; + Cost input_total_cost = 0; + + Assert(subpath->parallel_safe); + Assert(pathkeys); + + pathnode->path.pathtype = T_GatherMerge; + pathnode->path.parent = rel; + pathnode->path.param_info = get_baserel_parampathinfo(root, rel, + required_outer); + pathnode->path.parallel_aware = false; + + pathnode->subpath = subpath; + pathnode->num_workers = subpath->parallel_workers; + pathnode->path.pathkeys = pathkeys; + pathnode->path.pathtarget = target ? target : rel->reltarget; + pathnode->path.rows += subpath->rows; + + if (pathkeys_contained_in(pathkeys, subpath->pathkeys)) + { + /* Subpath is adequately ordered, we won't need to sort it */ + input_startup_cost += subpath->startup_cost; + input_total_cost += subpath->total_cost; + } + else + { + /* We'll need to insert a Sort node, so include cost for that */ + Path sort_path; /* dummy for result of cost_sort */ + + cost_sort(&sort_path, + root, + pathkeys, + subpath->total_cost, + subpath->rows, + subpath->pathtarget->width, + 0.0, + work_mem, + -1); + input_startup_cost += sort_path.startup_cost; + input_total_cost += sort_path.total_cost; + } + + cost_gather_merge(pathnode, root, rel, pathnode->path.param_info, + input_startup_cost, input_total_cost, rows); + + return pathnode; +} + +/* + * translate_sub_tlist - get subquery column numbers represented by tlist + * + * The given targetlist usually contains only Vars referencing the given relid. + * Extract their varattnos (ie, the column numbers of the subquery) and return + * as an integer List. + * + * If any of the tlist items is not a simple Var, we cannot determine whether + * the subquery's uniqueness condition (if any) matches ours, so punt and + * return NIL. + */ +static List * +translate_sub_tlist(List *tlist, int relid) +{ + List *result = NIL; + ListCell *l; + + foreach(l, tlist) + { + Var *var = (Var *) lfirst(l); + + if (!var || !IsA(var, Var) || + var->varno != relid) + return NIL; /* punt */ + + result = lappend_int(result, var->varattno); + } + return result; +} + +/* + * create_gather_path + * Creates a path corresponding to a gather scan, returning the + * pathnode. + * + * 'rows' may optionally be set to override row estimates from other sources. + */ +GatherPath * +create_gather_path(PlannerInfo *root, RelOptInfo *rel, Path *subpath, + PathTarget *target, Relids required_outer, double *rows) +{ + GatherPath *pathnode = makeNode(GatherPath); + + Assert(subpath->parallel_safe); + + pathnode->path.pathtype = T_Gather; + pathnode->path.parent = rel; + pathnode->path.pathtarget = target; + pathnode->path.param_info = get_baserel_parampathinfo(root, rel, + required_outer); + pathnode->path.parallel_aware = false; + pathnode->path.parallel_safe = false; + pathnode->path.parallel_workers = 0; + pathnode->path.pathkeys = NIL; /* Gather has unordered result */ + + pathnode->subpath = subpath; + pathnode->num_workers = subpath->parallel_workers; + pathnode->single_copy = false; + + if (pathnode->num_workers == 0) + { + pathnode->path.pathkeys = subpath->pathkeys; + pathnode->num_workers = 1; + pathnode->single_copy = true; + } + + cost_gather(pathnode, root, rel, pathnode->path.param_info, rows); + + return pathnode; +} + +/* + * create_subqueryscan_path + * Creates a path corresponding to a scan of a subquery, + * returning the pathnode. + */ +SubqueryScanPath * +create_subqueryscan_path(PlannerInfo *root, RelOptInfo *rel, Path *subpath, + List *pathkeys, Relids required_outer) +{ + SubqueryScanPath *pathnode = makeNode(SubqueryScanPath); + + pathnode->path.pathtype = T_SubqueryScan; + pathnode->path.parent = rel; + pathnode->path.pathtarget = rel->reltarget; + pathnode->path.param_info = get_baserel_parampathinfo(root, rel, + required_outer); + pathnode->path.parallel_aware = false; + pathnode->path.parallel_safe = rel->consider_parallel && + subpath->parallel_safe; + pathnode->path.parallel_workers = subpath->parallel_workers; + pathnode->path.pathkeys = pathkeys; + pathnode->subpath = subpath; + + cost_subqueryscan(pathnode, root, rel, pathnode->path.param_info); + + return pathnode; +} + +/* + * create_functionscan_path + * Creates a path corresponding to a sequential scan of a function, + * returning the pathnode. + */ +Path * +create_functionscan_path(PlannerInfo *root, RelOptInfo *rel, + List *pathkeys, Relids required_outer) +{ + Path *pathnode = makeNode(Path); + + pathnode->pathtype = T_FunctionScan; + pathnode->parent = rel; + pathnode->pathtarget = rel->reltarget; + pathnode->param_info = get_baserel_parampathinfo(root, rel, + required_outer); + pathnode->parallel_aware = false; + pathnode->parallel_safe = rel->consider_parallel; + pathnode->parallel_workers = 0; + pathnode->pathkeys = pathkeys; + + cost_functionscan(pathnode, root, rel, pathnode->param_info); + + return pathnode; +} + +/* + * create_tablefuncscan_path + * Creates a path corresponding to a sequential scan of a table function, + * returning the pathnode. + */ +Path * +create_tablefuncscan_path(PlannerInfo *root, RelOptInfo *rel, + Relids required_outer) +{ + Path *pathnode = makeNode(Path); + + pathnode->pathtype = T_TableFuncScan; + pathnode->parent = rel; + pathnode->pathtarget = rel->reltarget; + pathnode->param_info = get_baserel_parampathinfo(root, rel, + required_outer); + pathnode->parallel_aware = false; + pathnode->parallel_safe = rel->consider_parallel; + pathnode->parallel_workers = 0; + pathnode->pathkeys = NIL; /* result is always unordered */ + + cost_tablefuncscan(pathnode, root, rel, pathnode->param_info); + + return pathnode; +} + +/* + * create_valuesscan_path + * Creates a path corresponding to a scan of a VALUES list, + * returning the pathnode. + */ +Path * +create_valuesscan_path(PlannerInfo *root, RelOptInfo *rel, + Relids required_outer) +{ + Path *pathnode = makeNode(Path); + + pathnode->pathtype = T_ValuesScan; + pathnode->parent = rel; + pathnode->pathtarget = rel->reltarget; + pathnode->param_info = get_baserel_parampathinfo(root, rel, + required_outer); + pathnode->parallel_aware = false; + pathnode->parallel_safe = rel->consider_parallel; + pathnode->parallel_workers = 0; + pathnode->pathkeys = NIL; /* result is always unordered */ + + cost_valuesscan(pathnode, root, rel, pathnode->param_info); + + return pathnode; +} + +/* + * create_ctescan_path + * Creates a path corresponding to a scan of a non-self-reference CTE, + * returning the pathnode. + */ +Path * +create_ctescan_path(PlannerInfo *root, RelOptInfo *rel, Relids required_outer) +{ + Path *pathnode = makeNode(Path); + + pathnode->pathtype = T_CteScan; + pathnode->parent = rel; + pathnode->pathtarget = rel->reltarget; + pathnode->param_info = get_baserel_parampathinfo(root, rel, + required_outer); + pathnode->parallel_aware = false; + pathnode->parallel_safe = rel->consider_parallel; + pathnode->parallel_workers = 0; + pathnode->pathkeys = NIL; /* XXX for now, result is always unordered */ + + cost_ctescan(pathnode, root, rel, pathnode->param_info); + + return pathnode; +} + +/* + * create_namedtuplestorescan_path + * Creates a path corresponding to a scan of a named tuplestore, returning + * the pathnode. + */ +Path * +create_namedtuplestorescan_path(PlannerInfo *root, RelOptInfo *rel, + Relids required_outer) +{ + Path *pathnode = makeNode(Path); + + pathnode->pathtype = T_NamedTuplestoreScan; + pathnode->parent = rel; + pathnode->pathtarget = rel->reltarget; + pathnode->param_info = get_baserel_parampathinfo(root, rel, + required_outer); + pathnode->parallel_aware = false; + pathnode->parallel_safe = rel->consider_parallel; + pathnode->parallel_workers = 0; + pathnode->pathkeys = NIL; /* result is always unordered */ + + cost_namedtuplestorescan(pathnode, root, rel, pathnode->param_info); + + return pathnode; +} + +/* + * create_resultscan_path + * Creates a path corresponding to a scan of an RTE_RESULT relation, + * returning the pathnode. + */ +Path * +create_resultscan_path(PlannerInfo *root, RelOptInfo *rel, + Relids required_outer) +{ + Path *pathnode = makeNode(Path); + + pathnode->pathtype = T_Result; + pathnode->parent = rel; + pathnode->pathtarget = rel->reltarget; + pathnode->param_info = get_baserel_parampathinfo(root, rel, + required_outer); + pathnode->parallel_aware = false; + pathnode->parallel_safe = rel->consider_parallel; + pathnode->parallel_workers = 0; + pathnode->pathkeys = NIL; /* result is always unordered */ + + cost_resultscan(pathnode, root, rel, pathnode->param_info); + + return pathnode; +} + +/* + * create_worktablescan_path + * Creates a path corresponding to a scan of a self-reference CTE, + * returning the pathnode. + */ +Path * +create_worktablescan_path(PlannerInfo *root, RelOptInfo *rel, + Relids required_outer) +{ + Path *pathnode = makeNode(Path); + + pathnode->pathtype = T_WorkTableScan; + pathnode->parent = rel; + pathnode->pathtarget = rel->reltarget; + pathnode->param_info = get_baserel_parampathinfo(root, rel, + required_outer); + pathnode->parallel_aware = false; + pathnode->parallel_safe = rel->consider_parallel; + pathnode->parallel_workers = 0; + pathnode->pathkeys = NIL; /* result is always unordered */ + + /* Cost is the same as for a regular CTE scan */ + cost_ctescan(pathnode, root, rel, pathnode->param_info); + + return pathnode; +} + +/* + * create_foreignscan_path + * Creates a path corresponding to a scan of a foreign base table, + * returning the pathnode. + * + * This function is never called from core Postgres; rather, it's expected + * to be called by the GetForeignPaths function of a foreign data wrapper. + * We make the FDW supply all fields of the path, since we do not have any way + * to calculate them in core. However, there is a usually-sane default for + * the pathtarget (rel->reltarget), so we let a NULL for "target" select that. + */ +ForeignPath * +create_foreignscan_path(PlannerInfo *root, RelOptInfo *rel, + PathTarget *target, + double rows, Cost startup_cost, Cost total_cost, + List *pathkeys, + Relids required_outer, + Path *fdw_outerpath, + List *fdw_private) +{ + ForeignPath *pathnode = makeNode(ForeignPath); + + /* Historically some FDWs were confused about when to use this */ + Assert(IS_SIMPLE_REL(rel)); + + pathnode->path.pathtype = T_ForeignScan; + pathnode->path.parent = rel; + pathnode->path.pathtarget = target ? target : rel->reltarget; + pathnode->path.param_info = get_baserel_parampathinfo(root, rel, + required_outer); + pathnode->path.parallel_aware = false; + pathnode->path.parallel_safe = rel->consider_parallel; + pathnode->path.parallel_workers = 0; + pathnode->path.rows = rows; + pathnode->path.startup_cost = startup_cost; + pathnode->path.total_cost = total_cost; + pathnode->path.pathkeys = pathkeys; + + pathnode->fdw_outerpath = fdw_outerpath; + pathnode->fdw_private = fdw_private; + + return pathnode; +} + +/* + * create_foreign_join_path + * Creates a path corresponding to a scan of a foreign join, + * returning the pathnode. + * + * This function is never called from core Postgres; rather, it's expected + * to be called by the GetForeignJoinPaths function of a foreign data wrapper. + * We make the FDW supply all fields of the path, since we do not have any way + * to calculate them in core. However, there is a usually-sane default for + * the pathtarget (rel->reltarget), so we let a NULL for "target" select that. + */ +ForeignPath * +create_foreign_join_path(PlannerInfo *root, RelOptInfo *rel, + PathTarget *target, + double rows, Cost startup_cost, Cost total_cost, + List *pathkeys, + Relids required_outer, + Path *fdw_outerpath, + List *fdw_private) +{ + ForeignPath *pathnode = makeNode(ForeignPath); + + /* + * We should use get_joinrel_parampathinfo to handle parameterized paths, + * but the API of this function doesn't support it, and existing + * extensions aren't yet trying to build such paths anyway. For the + * moment just throw an error if someone tries it; eventually we should + * revisit this. + */ + if (!bms_is_empty(required_outer) || !bms_is_empty(rel->lateral_relids)) + elog(ERROR, "parameterized foreign joins are not supported yet"); + + pathnode->path.pathtype = T_ForeignScan; + pathnode->path.parent = rel; + pathnode->path.pathtarget = target ? target : rel->reltarget; + pathnode->path.param_info = NULL; /* XXX see above */ + pathnode->path.parallel_aware = false; + pathnode->path.parallel_safe = rel->consider_parallel; + pathnode->path.parallel_workers = 0; + pathnode->path.rows = rows; + pathnode->path.startup_cost = startup_cost; + pathnode->path.total_cost = total_cost; + pathnode->path.pathkeys = pathkeys; + + pathnode->fdw_outerpath = fdw_outerpath; + pathnode->fdw_private = fdw_private; + + return pathnode; +} + +/* + * create_foreign_upper_path + * Creates a path corresponding to an upper relation that's computed + * directly by an FDW, returning the pathnode. + * + * This function is never called from core Postgres; rather, it's expected to + * be called by the GetForeignUpperPaths function of a foreign data wrapper. + * We make the FDW supply all fields of the path, since we do not have any way + * to calculate them in core. However, there is a usually-sane default for + * the pathtarget (rel->reltarget), so we let a NULL for "target" select that. + */ +ForeignPath * +create_foreign_upper_path(PlannerInfo *root, RelOptInfo *rel, + PathTarget *target, + double rows, Cost startup_cost, Cost total_cost, + List *pathkeys, + Path *fdw_outerpath, + List *fdw_private) +{ + ForeignPath *pathnode = makeNode(ForeignPath); + + /* + * Upper relations should never have any lateral references, since joining + * is complete. + */ + Assert(bms_is_empty(rel->lateral_relids)); + + pathnode->path.pathtype = T_ForeignScan; + pathnode->path.parent = rel; + pathnode->path.pathtarget = target ? target : rel->reltarget; + pathnode->path.param_info = NULL; + pathnode->path.parallel_aware = false; + pathnode->path.parallel_safe = rel->consider_parallel; + pathnode->path.parallel_workers = 0; + pathnode->path.rows = rows; + pathnode->path.startup_cost = startup_cost; + pathnode->path.total_cost = total_cost; + pathnode->path.pathkeys = pathkeys; + + pathnode->fdw_outerpath = fdw_outerpath; + pathnode->fdw_private = fdw_private; + + return pathnode; +} + +/* + * calc_nestloop_required_outer + * Compute the required_outer set for a nestloop join path + * + * Note: result must not share storage with either input + */ +Relids +calc_nestloop_required_outer(Relids outerrelids, + Relids outer_paramrels, + Relids innerrelids, + Relids inner_paramrels) +{ + Relids required_outer; + + /* inner_path can require rels from outer path, but not vice versa */ + Assert(!bms_overlap(outer_paramrels, innerrelids)); + /* easy case if inner path is not parameterized */ + if (!inner_paramrels) + return bms_copy(outer_paramrels); + /* else, form the union ... */ + required_outer = bms_union(outer_paramrels, inner_paramrels); + /* ... and remove any mention of now-satisfied outer rels */ + required_outer = bms_del_members(required_outer, + outerrelids); + /* maintain invariant that required_outer is exactly NULL if empty */ + if (bms_is_empty(required_outer)) + { + bms_free(required_outer); + required_outer = NULL; + } + return required_outer; +} + +/* + * calc_non_nestloop_required_outer + * Compute the required_outer set for a merge or hash join path + * + * Note: result must not share storage with either input + */ +Relids +calc_non_nestloop_required_outer(Path *outer_path, Path *inner_path) +{ + Relids outer_paramrels = PATH_REQ_OUTER(outer_path); + Relids inner_paramrels = PATH_REQ_OUTER(inner_path); + Relids required_outer; + + /* neither path can require rels from the other */ + Assert(!bms_overlap(outer_paramrels, inner_path->parent->relids)); + Assert(!bms_overlap(inner_paramrels, outer_path->parent->relids)); + /* form the union ... */ + required_outer = bms_union(outer_paramrels, inner_paramrels); + /* we do not need an explicit test for empty; bms_union gets it right */ + return required_outer; +} + +/* + * create_nestloop_path + * Creates a pathnode corresponding to a nestloop join between two + * relations. + * + * 'joinrel' is the join relation. + * 'jointype' is the type of join required + * 'workspace' is the result from initial_cost_nestloop + * 'extra' contains various information about the join + * 'outer_path' is the outer path + * 'inner_path' is the inner path + * 'restrict_clauses' are the RestrictInfo nodes to apply at the join + * 'pathkeys' are the path keys of the new join path + * 'required_outer' is the set of required outer rels + * + * Returns the resulting path node. + */ +NestPath * +create_nestloop_path(PlannerInfo *root, + RelOptInfo *joinrel, + JoinType jointype, + JoinCostWorkspace *workspace, + JoinPathExtraData *extra, + Path *outer_path, + Path *inner_path, + List *restrict_clauses, + List *pathkeys, + Relids required_outer) +{ + NestPath *pathnode = makeNode(NestPath); + Relids inner_req_outer = PATH_REQ_OUTER(inner_path); + + /* + * If the inner path is parameterized by the outer, we must drop any + * restrict_clauses that are due to be moved into the inner path. We have + * to do this now, rather than postpone the work till createplan time, + * because the restrict_clauses list can affect the size and cost + * estimates for this path. + */ + if (bms_overlap(inner_req_outer, outer_path->parent->relids)) + { + Relids inner_and_outer = bms_union(inner_path->parent->relids, + inner_req_outer); + List *jclauses = NIL; + ListCell *lc; + + foreach(lc, restrict_clauses) + { + RestrictInfo *rinfo = (RestrictInfo *) lfirst(lc); + + if (!join_clause_is_movable_into(rinfo, + inner_path->parent->relids, + inner_and_outer)) + jclauses = lappend(jclauses, rinfo); + } + restrict_clauses = jclauses; + } + + pathnode->jpath.path.pathtype = T_NestLoop; + pathnode->jpath.path.parent = joinrel; + pathnode->jpath.path.pathtarget = joinrel->reltarget; + pathnode->jpath.path.param_info = + get_joinrel_parampathinfo(root, + joinrel, + outer_path, + inner_path, + extra->sjinfo, + required_outer, + &restrict_clauses); + pathnode->jpath.path.parallel_aware = false; + pathnode->jpath.path.parallel_safe = joinrel->consider_parallel && + outer_path->parallel_safe && inner_path->parallel_safe; + /* This is a foolish way to estimate parallel_workers, but for now... */ + pathnode->jpath.path.parallel_workers = outer_path->parallel_workers; + pathnode->jpath.path.pathkeys = pathkeys; + pathnode->jpath.jointype = jointype; + pathnode->jpath.inner_unique = extra->inner_unique; + pathnode->jpath.outerjoinpath = outer_path; + pathnode->jpath.innerjoinpath = inner_path; + pathnode->jpath.joinrestrictinfo = restrict_clauses; + + final_cost_nestloop(root, pathnode, workspace, extra); + + return pathnode; +} + +/* + * create_mergejoin_path + * Creates a pathnode corresponding to a mergejoin join between + * two relations + * + * 'joinrel' is the join relation + * 'jointype' is the type of join required + * 'workspace' is the result from initial_cost_mergejoin + * 'extra' contains various information about the join + * 'outer_path' is the outer path + * 'inner_path' is the inner path + * 'restrict_clauses' are the RestrictInfo nodes to apply at the join + * 'pathkeys' are the path keys of the new join path + * 'required_outer' is the set of required outer rels + * 'mergeclauses' are the RestrictInfo nodes to use as merge clauses + * (this should be a subset of the restrict_clauses list) + * 'outersortkeys' are the sort varkeys for the outer relation + * 'innersortkeys' are the sort varkeys for the inner relation + */ +MergePath * +create_mergejoin_path(PlannerInfo *root, + RelOptInfo *joinrel, + JoinType jointype, + JoinCostWorkspace *workspace, + JoinPathExtraData *extra, + Path *outer_path, + Path *inner_path, + List *restrict_clauses, + List *pathkeys, + Relids required_outer, + List *mergeclauses, + List *outersortkeys, + List *innersortkeys) +{ + MergePath *pathnode = makeNode(MergePath); + + pathnode->jpath.path.pathtype = T_MergeJoin; + pathnode->jpath.path.parent = joinrel; + pathnode->jpath.path.pathtarget = joinrel->reltarget; + pathnode->jpath.path.param_info = + get_joinrel_parampathinfo(root, + joinrel, + outer_path, + inner_path, + extra->sjinfo, + required_outer, + &restrict_clauses); + pathnode->jpath.path.parallel_aware = false; + pathnode->jpath.path.parallel_safe = joinrel->consider_parallel && + outer_path->parallel_safe && inner_path->parallel_safe; + /* This is a foolish way to estimate parallel_workers, but for now... */ + pathnode->jpath.path.parallel_workers = outer_path->parallel_workers; + pathnode->jpath.path.pathkeys = pathkeys; + pathnode->jpath.jointype = jointype; + pathnode->jpath.inner_unique = extra->inner_unique; + pathnode->jpath.outerjoinpath = outer_path; + pathnode->jpath.innerjoinpath = inner_path; + pathnode->jpath.joinrestrictinfo = restrict_clauses; + pathnode->path_mergeclauses = mergeclauses; + pathnode->outersortkeys = outersortkeys; + pathnode->innersortkeys = innersortkeys; + /* pathnode->skip_mark_restore will be set by final_cost_mergejoin */ + /* pathnode->materialize_inner will be set by final_cost_mergejoin */ + + final_cost_mergejoin(root, pathnode, workspace, extra); + + return pathnode; +} + +/* + * create_hashjoin_path + * Creates a pathnode corresponding to a hash join between two relations. + * + * 'joinrel' is the join relation + * 'jointype' is the type of join required + * 'workspace' is the result from initial_cost_hashjoin + * 'extra' contains various information about the join + * 'outer_path' is the cheapest outer path + * 'inner_path' is the cheapest inner path + * 'parallel_hash' to select Parallel Hash of inner path (shared hash table) + * 'restrict_clauses' are the RestrictInfo nodes to apply at the join + * 'required_outer' is the set of required outer rels + * 'hashclauses' are the RestrictInfo nodes to use as hash clauses + * (this should be a subset of the restrict_clauses list) + */ +HashPath * +create_hashjoin_path(PlannerInfo *root, + RelOptInfo *joinrel, + JoinType jointype, + JoinCostWorkspace *workspace, + JoinPathExtraData *extra, + Path *outer_path, + Path *inner_path, + bool parallel_hash, + List *restrict_clauses, + Relids required_outer, + List *hashclauses) +{ + HashPath *pathnode = makeNode(HashPath); + + pathnode->jpath.path.pathtype = T_HashJoin; + pathnode->jpath.path.parent = joinrel; + pathnode->jpath.path.pathtarget = joinrel->reltarget; + pathnode->jpath.path.param_info = + get_joinrel_parampathinfo(root, + joinrel, + outer_path, + inner_path, + extra->sjinfo, + required_outer, + &restrict_clauses); + pathnode->jpath.path.parallel_aware = + joinrel->consider_parallel && parallel_hash; + pathnode->jpath.path.parallel_safe = joinrel->consider_parallel && + outer_path->parallel_safe && inner_path->parallel_safe; + /* This is a foolish way to estimate parallel_workers, but for now... */ + pathnode->jpath.path.parallel_workers = outer_path->parallel_workers; + + /* + * A hashjoin never has pathkeys, since its output ordering is + * unpredictable due to possible batching. XXX If the inner relation is + * small enough, we could instruct the executor that it must not batch, + * and then we could assume that the output inherits the outer relation's + * ordering, which might save a sort step. However there is considerable + * downside if our estimate of the inner relation size is badly off. For + * the moment we don't risk it. (Note also that if we wanted to take this + * seriously, joinpath.c would have to consider many more paths for the + * outer rel than it does now.) + */ + pathnode->jpath.path.pathkeys = NIL; + pathnode->jpath.jointype = jointype; + pathnode->jpath.inner_unique = extra->inner_unique; + pathnode->jpath.outerjoinpath = outer_path; + pathnode->jpath.innerjoinpath = inner_path; + pathnode->jpath.joinrestrictinfo = restrict_clauses; + pathnode->path_hashclauses = hashclauses; + /* final_cost_hashjoin will fill in pathnode->num_batches */ + + final_cost_hashjoin(root, pathnode, workspace, extra); + + return pathnode; +} + +/* + * create_projection_path + * Creates a pathnode that represents performing a projection. + * + * 'rel' is the parent relation associated with the result + * 'subpath' is the path representing the source of data + * 'target' is the PathTarget to be computed + */ +ProjectionPath * +create_projection_path(PlannerInfo *root, + RelOptInfo *rel, + Path *subpath, + PathTarget *target) +{ + ProjectionPath *pathnode = makeNode(ProjectionPath); + PathTarget *oldtarget; + + /* + * We mustn't put a ProjectionPath directly above another; it's useless + * and will confuse create_projection_plan. Rather than making sure all + * callers handle that, let's implement it here, by stripping off any + * ProjectionPath in what we're given. Given this rule, there won't be + * more than one. + */ + if (IsA(subpath, ProjectionPath)) + { + ProjectionPath *subpp = (ProjectionPath *) subpath; + + Assert(subpp->path.parent == rel); + subpath = subpp->subpath; + Assert(!IsA(subpath, ProjectionPath)); + } + + pathnode->path.pathtype = T_Result; + pathnode->path.parent = rel; + pathnode->path.pathtarget = target; + /* For now, assume we are above any joins, so no parameterization */ + pathnode->path.param_info = NULL; + pathnode->path.parallel_aware = false; + pathnode->path.parallel_safe = rel->consider_parallel && + subpath->parallel_safe && + is_parallel_safe(root, (Node *) target->exprs); + pathnode->path.parallel_workers = subpath->parallel_workers; + /* Projection does not change the sort order */ + pathnode->path.pathkeys = subpath->pathkeys; + + pathnode->subpath = subpath; + + /* + * We might not need a separate Result node. If the input plan node type + * can project, we can just tell it to project something else. Or, if it + * can't project but the desired target has the same expression list as + * what the input will produce anyway, we can still give it the desired + * tlist (possibly changing its ressortgroupref labels, but nothing else). + * Note: in the latter case, create_projection_plan has to recheck our + * conclusion; see comments therein. + */ + oldtarget = subpath->pathtarget; + if (is_projection_capable_path(subpath) || + equal(oldtarget->exprs, target->exprs)) + { + /* No separate Result node needed */ + pathnode->dummypp = true; + + /* + * Set cost of plan as subpath's cost, adjusted for tlist replacement. + */ + pathnode->path.rows = subpath->rows; + pathnode->path.startup_cost = subpath->startup_cost + + (target->cost.startup - oldtarget->cost.startup); + pathnode->path.total_cost = subpath->total_cost + + (target->cost.startup - oldtarget->cost.startup) + + (target->cost.per_tuple - oldtarget->cost.per_tuple) * subpath->rows; + } + else + { + /* We really do need the Result node */ + pathnode->dummypp = false; + + /* + * The Result node's cost is cpu_tuple_cost per row, plus the cost of + * evaluating the tlist. There is no qual to worry about. + */ + pathnode->path.rows = subpath->rows; + pathnode->path.startup_cost = subpath->startup_cost + + target->cost.startup; + pathnode->path.total_cost = subpath->total_cost + + target->cost.startup + + (cpu_tuple_cost + target->cost.per_tuple) * subpath->rows; + } + + return pathnode; +} + +/* + * apply_projection_to_path + * Add a projection step, or just apply the target directly to given path. + * + * This has the same net effect as create_projection_path(), except that if + * a separate Result plan node isn't needed, we just replace the given path's + * pathtarget with the desired one. This must be used only when the caller + * knows that the given path isn't referenced elsewhere and so can be modified + * in-place. + * + * If the input path is a GatherPath or GatherMergePath, we try to push the + * new target down to its input as well; this is a yet more invasive + * modification of the input path, which create_projection_path() can't do. + * + * Note that we mustn't change the source path's parent link; so when it is + * add_path'd to "rel" things will be a bit inconsistent. So far that has + * not caused any trouble. + * + * 'rel' is the parent relation associated with the result + * 'path' is the path representing the source of data + * 'target' is the PathTarget to be computed + */ +Path * +apply_projection_to_path(PlannerInfo *root, + RelOptInfo *rel, + Path *path, + PathTarget *target) +{ + QualCost oldcost; + + /* + * If given path can't project, we might need a Result node, so make a + * separate ProjectionPath. + */ + if (!is_projection_capable_path(path)) + return (Path *) create_projection_path(root, rel, path, target); + + /* + * We can just jam the desired tlist into the existing path, being sure to + * update its cost estimates appropriately. + */ + oldcost = path->pathtarget->cost; + path->pathtarget = target; + + path->startup_cost += target->cost.startup - oldcost.startup; + path->total_cost += target->cost.startup - oldcost.startup + + (target->cost.per_tuple - oldcost.per_tuple) * path->rows; + + /* + * If the path happens to be a Gather or GatherMerge path, we'd like to + * arrange for the subpath to return the required target list so that + * workers can help project. But if there is something that is not + * parallel-safe in the target expressions, then we can't. + */ + if ((IsA(path, GatherPath) || IsA(path, GatherMergePath)) && + is_parallel_safe(root, (Node *) target->exprs)) + { + /* + * We always use create_projection_path here, even if the subpath is + * projection-capable, so as to avoid modifying the subpath in place. + * It seems unlikely at present that there could be any other + * references to the subpath, but better safe than sorry. + * + * Note that we don't change the parallel path's cost estimates; it + * might be appropriate to do so, to reflect the fact that the bulk of + * the target evaluation will happen in workers. + */ + if (IsA(path, GatherPath)) + { + GatherPath *gpath = (GatherPath *) path; + + gpath->subpath = (Path *) + create_projection_path(root, + gpath->subpath->parent, + gpath->subpath, + target); + } + else + { + GatherMergePath *gmpath = (GatherMergePath *) path; + + gmpath->subpath = (Path *) + create_projection_path(root, + gmpath->subpath->parent, + gmpath->subpath, + target); + } + } + else if (path->parallel_safe && + !is_parallel_safe(root, (Node *) target->exprs)) + { + /* + * We're inserting a parallel-restricted target list into a path + * currently marked parallel-safe, so we have to mark it as no longer + * safe. + */ + path->parallel_safe = false; + } + + return path; +} + +/* + * create_set_projection_path + * Creates a pathnode that represents performing a projection that + * includes set-returning functions. + * + * 'rel' is the parent relation associated with the result + * 'subpath' is the path representing the source of data + * 'target' is the PathTarget to be computed + */ +ProjectSetPath * +create_set_projection_path(PlannerInfo *root, + RelOptInfo *rel, + Path *subpath, + PathTarget *target) +{ + ProjectSetPath *pathnode = makeNode(ProjectSetPath); + double tlist_rows; + ListCell *lc; + + pathnode->path.pathtype = T_ProjectSet; + pathnode->path.parent = rel; + pathnode->path.pathtarget = target; + /* For now, assume we are above any joins, so no parameterization */ + pathnode->path.param_info = NULL; + pathnode->path.parallel_aware = false; + pathnode->path.parallel_safe = rel->consider_parallel && + subpath->parallel_safe && + is_parallel_safe(root, (Node *) target->exprs); + pathnode->path.parallel_workers = subpath->parallel_workers; + /* Projection does not change the sort order XXX? */ + pathnode->path.pathkeys = subpath->pathkeys; + + pathnode->subpath = subpath; + + /* + * Estimate number of rows produced by SRFs for each row of input; if + * there's more than one in this node, use the maximum. + */ + tlist_rows = 1; + foreach(lc, target->exprs) + { + Node *node = (Node *) lfirst(lc); + double itemrows; + + itemrows = expression_returns_set_rows(root, node); + if (tlist_rows < itemrows) + tlist_rows = itemrows; + } + + /* + * In addition to the cost of evaluating the tlist, charge cpu_tuple_cost + * per input row, and half of cpu_tuple_cost for each added output row. + * This is slightly bizarre maybe, but it's what 9.6 did; we may revisit + * this estimate later. + */ + pathnode->path.rows = subpath->rows * tlist_rows; + pathnode->path.startup_cost = subpath->startup_cost + + target->cost.startup; + pathnode->path.total_cost = subpath->total_cost + + target->cost.startup + + (cpu_tuple_cost + target->cost.per_tuple) * subpath->rows + + (pathnode->path.rows - subpath->rows) * cpu_tuple_cost / 2; + + return pathnode; +} + +/* + * create_incremental_sort_path + * Creates a pathnode that represents performing an incremental sort. + * + * 'rel' is the parent relation associated with the result + * 'subpath' is the path representing the source of data + * 'pathkeys' represents the desired sort order + * 'presorted_keys' is the number of keys by which the input path is + * already sorted + * 'limit_tuples' is the estimated bound on the number of output tuples, + * or -1 if no LIMIT or couldn't estimate + */ +IncrementalSortPath * +create_incremental_sort_path(PlannerInfo *root, + RelOptInfo *rel, + Path *subpath, + List *pathkeys, + int presorted_keys, + double limit_tuples) +{ + IncrementalSortPath *sort = makeNode(IncrementalSortPath); + SortPath *pathnode = &sort->spath; + + pathnode->path.pathtype = T_IncrementalSort; + pathnode->path.parent = rel; + /* Sort doesn't project, so use source path's pathtarget */ + pathnode->path.pathtarget = subpath->pathtarget; + /* For now, assume we are above any joins, so no parameterization */ + pathnode->path.param_info = NULL; + pathnode->path.parallel_aware = false; + pathnode->path.parallel_safe = rel->consider_parallel && + subpath->parallel_safe; + pathnode->path.parallel_workers = subpath->parallel_workers; + pathnode->path.pathkeys = pathkeys; + + pathnode->subpath = subpath; + + cost_incremental_sort(&pathnode->path, + root, pathkeys, presorted_keys, + subpath->startup_cost, + subpath->total_cost, + subpath->rows, + subpath->pathtarget->width, + 0.0, /* XXX comparison_cost shouldn't be 0? */ + work_mem, limit_tuples); + + sort->nPresortedCols = presorted_keys; + + return sort; +} + +/* + * create_sort_path + * Creates a pathnode that represents performing an explicit sort. + * + * 'rel' is the parent relation associated with the result + * 'subpath' is the path representing the source of data + * 'pathkeys' represents the desired sort order + * 'limit_tuples' is the estimated bound on the number of output tuples, + * or -1 if no LIMIT or couldn't estimate + */ +SortPath * +create_sort_path(PlannerInfo *root, + RelOptInfo *rel, + Path *subpath, + List *pathkeys, + double limit_tuples) +{ + SortPath *pathnode = makeNode(SortPath); + + pathnode->path.pathtype = T_Sort; + pathnode->path.parent = rel; + /* Sort doesn't project, so use source path's pathtarget */ + pathnode->path.pathtarget = subpath->pathtarget; + /* For now, assume we are above any joins, so no parameterization */ + pathnode->path.param_info = NULL; + pathnode->path.parallel_aware = false; + pathnode->path.parallel_safe = rel->consider_parallel && + subpath->parallel_safe; + pathnode->path.parallel_workers = subpath->parallel_workers; + pathnode->path.pathkeys = pathkeys; + + pathnode->subpath = subpath; + + cost_sort(&pathnode->path, root, pathkeys, + subpath->total_cost, + subpath->rows, + subpath->pathtarget->width, + 0.0, /* XXX comparison_cost shouldn't be 0? */ + work_mem, limit_tuples); + + return pathnode; +} + +/* + * create_group_path + * Creates a pathnode that represents performing grouping of presorted input + * + * 'rel' is the parent relation associated with the result + * 'subpath' is the path representing the source of data + * 'target' is the PathTarget to be computed + * 'groupClause' is a list of SortGroupClause's representing the grouping + * 'qual' is the HAVING quals if any + * 'numGroups' is the estimated number of groups + */ +GroupPath * +create_group_path(PlannerInfo *root, + RelOptInfo *rel, + Path *subpath, + List *groupClause, + List *qual, + double numGroups) +{ + GroupPath *pathnode = makeNode(GroupPath); + PathTarget *target = rel->reltarget; + + pathnode->path.pathtype = T_Group; + pathnode->path.parent = rel; + pathnode->path.pathtarget = target; + /* For now, assume we are above any joins, so no parameterization */ + pathnode->path.param_info = NULL; + pathnode->path.parallel_aware = false; + pathnode->path.parallel_safe = rel->consider_parallel && + subpath->parallel_safe; + pathnode->path.parallel_workers = subpath->parallel_workers; + /* Group doesn't change sort ordering */ + pathnode->path.pathkeys = subpath->pathkeys; + + pathnode->subpath = subpath; + + pathnode->groupClause = groupClause; + pathnode->qual = qual; + + cost_group(&pathnode->path, root, + list_length(groupClause), + numGroups, + qual, + subpath->startup_cost, subpath->total_cost, + subpath->rows); + + /* add tlist eval cost for each output row */ + pathnode->path.startup_cost += target->cost.startup; + pathnode->path.total_cost += target->cost.startup + + target->cost.per_tuple * pathnode->path.rows; + + return pathnode; +} + +/* + * create_upper_unique_path + * Creates a pathnode that represents performing an explicit Unique step + * on presorted input. + * + * This produces a Unique plan node, but the use-case is so different from + * create_unique_path that it doesn't seem worth trying to merge the two. + * + * 'rel' is the parent relation associated with the result + * 'subpath' is the path representing the source of data + * 'numCols' is the number of grouping columns + * 'numGroups' is the estimated number of groups + * + * The input path must be sorted on the grouping columns, plus possibly + * additional columns; so the first numCols pathkeys are the grouping columns + */ +UpperUniquePath * +create_upper_unique_path(PlannerInfo *root, + RelOptInfo *rel, + Path *subpath, + int numCols, + double numGroups) +{ + UpperUniquePath *pathnode = makeNode(UpperUniquePath); + + pathnode->path.pathtype = T_Unique; + pathnode->path.parent = rel; + /* Unique doesn't project, so use source path's pathtarget */ + pathnode->path.pathtarget = subpath->pathtarget; + /* For now, assume we are above any joins, so no parameterization */ + pathnode->path.param_info = NULL; + pathnode->path.parallel_aware = false; + pathnode->path.parallel_safe = rel->consider_parallel && + subpath->parallel_safe; + pathnode->path.parallel_workers = subpath->parallel_workers; + /* Unique doesn't change the input ordering */ + pathnode->path.pathkeys = subpath->pathkeys; + + pathnode->subpath = subpath; + pathnode->numkeys = numCols; + + /* + * Charge one cpu_operator_cost per comparison per input tuple. We assume + * all columns get compared at most of the tuples. (XXX probably this is + * an overestimate.) + */ + pathnode->path.startup_cost = subpath->startup_cost; + pathnode->path.total_cost = subpath->total_cost + + cpu_operator_cost * subpath->rows * numCols; + pathnode->path.rows = numGroups; + + return pathnode; +} + +/* + * create_agg_path + * Creates a pathnode that represents performing aggregation/grouping + * + * 'rel' is the parent relation associated with the result + * 'subpath' is the path representing the source of data + * 'target' is the PathTarget to be computed + * 'aggstrategy' is the Agg node's basic implementation strategy + * 'aggsplit' is the Agg node's aggregate-splitting mode + * 'groupClause' is a list of SortGroupClause's representing the grouping + * 'qual' is the HAVING quals if any + * 'aggcosts' contains cost info about the aggregate functions to be computed + * 'numGroups' is the estimated number of groups (1 if not grouping) + */ +AggPath * +create_agg_path(PlannerInfo *root, + RelOptInfo *rel, + Path *subpath, + PathTarget *target, + AggStrategy aggstrategy, + AggSplit aggsplit, + List *groupClause, + List *qual, + const AggClauseCosts *aggcosts, + double numGroups) +{ + AggPath *pathnode = makeNode(AggPath); + + pathnode->path.pathtype = T_Agg; + pathnode->path.parent = rel; + pathnode->path.pathtarget = target; + /* For now, assume we are above any joins, so no parameterization */ + pathnode->path.param_info = NULL; + pathnode->path.parallel_aware = false; + pathnode->path.parallel_safe = rel->consider_parallel && + subpath->parallel_safe; + pathnode->path.parallel_workers = subpath->parallel_workers; + if (aggstrategy == AGG_SORTED) + pathnode->path.pathkeys = subpath->pathkeys; /* preserves order */ + else + pathnode->path.pathkeys = NIL; /* output is unordered */ + pathnode->subpath = subpath; + + pathnode->aggstrategy = aggstrategy; + pathnode->aggsplit = aggsplit; + pathnode->numGroups = numGroups; + pathnode->transitionSpace = aggcosts ? aggcosts->transitionSpace : 0; + pathnode->groupClause = groupClause; + pathnode->qual = qual; + + cost_agg(&pathnode->path, root, + aggstrategy, aggcosts, + list_length(groupClause), numGroups, + qual, + subpath->startup_cost, subpath->total_cost, + subpath->rows, subpath->pathtarget->width); + + /* add tlist eval cost for each output row */ + pathnode->path.startup_cost += target->cost.startup; + pathnode->path.total_cost += target->cost.startup + + target->cost.per_tuple * pathnode->path.rows; + + return pathnode; +} + +/* + * create_groupingsets_path + * Creates a pathnode that represents performing GROUPING SETS aggregation + * + * GroupingSetsPath represents sorted grouping with one or more grouping sets. + * The input path's result must be sorted to match the last entry in + * rollup_groupclauses. + * + * 'rel' is the parent relation associated with the result + * 'subpath' is the path representing the source of data + * 'target' is the PathTarget to be computed + * 'having_qual' is the HAVING quals if any + * 'rollups' is a list of RollupData nodes + * 'agg_costs' contains cost info about the aggregate functions to be computed + * 'numGroups' is the estimated total number of groups + */ +GroupingSetsPath * +create_groupingsets_path(PlannerInfo *root, + RelOptInfo *rel, + Path *subpath, + List *having_qual, + AggStrategy aggstrategy, + List *rollups, + const AggClauseCosts *agg_costs, + double numGroups) +{ + GroupingSetsPath *pathnode = makeNode(GroupingSetsPath); + PathTarget *target = rel->reltarget; + ListCell *lc; + bool is_first = true; + bool is_first_sort = true; + + /* The topmost generated Plan node will be an Agg */ + pathnode->path.pathtype = T_Agg; + pathnode->path.parent = rel; + pathnode->path.pathtarget = target; + pathnode->path.param_info = subpath->param_info; + pathnode->path.parallel_aware = false; + pathnode->path.parallel_safe = rel->consider_parallel && + subpath->parallel_safe; + pathnode->path.parallel_workers = subpath->parallel_workers; + pathnode->subpath = subpath; + + /* + * Simplify callers by downgrading AGG_SORTED to AGG_PLAIN, and AGG_MIXED + * to AGG_HASHED, here if possible. + */ + if (aggstrategy == AGG_SORTED && + list_length(rollups) == 1 && + ((RollupData *) linitial(rollups))->groupClause == NIL) + aggstrategy = AGG_PLAIN; + + if (aggstrategy == AGG_MIXED && + list_length(rollups) == 1) + aggstrategy = AGG_HASHED; + + /* + * Output will be in sorted order by group_pathkeys if, and only if, there + * is a single rollup operation on a non-empty list of grouping + * expressions. + */ + if (aggstrategy == AGG_SORTED && list_length(rollups) == 1) + pathnode->path.pathkeys = root->group_pathkeys; + else + pathnode->path.pathkeys = NIL; + + pathnode->aggstrategy = aggstrategy; + pathnode->rollups = rollups; + pathnode->qual = having_qual; + pathnode->transitionSpace = agg_costs ? agg_costs->transitionSpace : 0; + + Assert(rollups != NIL); + Assert(aggstrategy != AGG_PLAIN || list_length(rollups) == 1); + Assert(aggstrategy != AGG_MIXED || list_length(rollups) > 1); + + foreach(lc, rollups) + { + RollupData *rollup = lfirst(lc); + List *gsets = rollup->gsets; + int numGroupCols = list_length(linitial(gsets)); + + /* + * In AGG_SORTED or AGG_PLAIN mode, the first rollup takes the + * (already-sorted) input, and following ones do their own sort. + * + * In AGG_HASHED mode, there is one rollup for each grouping set. + * + * In AGG_MIXED mode, the first rollups are hashed, the first + * non-hashed one takes the (already-sorted) input, and following ones + * do their own sort. + */ + if (is_first) + { + cost_agg(&pathnode->path, root, + aggstrategy, + agg_costs, + numGroupCols, + rollup->numGroups, + having_qual, + subpath->startup_cost, + subpath->total_cost, + subpath->rows, + subpath->pathtarget->width); + is_first = false; + if (!rollup->is_hashed) + is_first_sort = false; + } + else + { + Path sort_path; /* dummy for result of cost_sort */ + Path agg_path; /* dummy for result of cost_agg */ + + if (rollup->is_hashed || is_first_sort) + { + /* + * Account for cost of aggregation, but don't charge input + * cost again + */ + cost_agg(&agg_path, root, + rollup->is_hashed ? AGG_HASHED : AGG_SORTED, + agg_costs, + numGroupCols, + rollup->numGroups, + having_qual, + 0.0, 0.0, + subpath->rows, + subpath->pathtarget->width); + if (!rollup->is_hashed) + is_first_sort = false; + } + else + { + /* Account for cost of sort, but don't charge input cost again */ + cost_sort(&sort_path, root, NIL, + 0.0, + subpath->rows, + subpath->pathtarget->width, + 0.0, + work_mem, + -1.0); + + /* Account for cost of aggregation */ + + cost_agg(&agg_path, root, + AGG_SORTED, + agg_costs, + numGroupCols, + rollup->numGroups, + having_qual, + sort_path.startup_cost, + sort_path.total_cost, + sort_path.rows, + subpath->pathtarget->width); + } + + pathnode->path.total_cost += agg_path.total_cost; + pathnode->path.rows += agg_path.rows; + } + } + + /* add tlist eval cost for each output row */ + pathnode->path.startup_cost += target->cost.startup; + pathnode->path.total_cost += target->cost.startup + + target->cost.per_tuple * pathnode->path.rows; + + return pathnode; +} + +/* + * create_minmaxagg_path + * Creates a pathnode that represents computation of MIN/MAX aggregates + * + * 'rel' is the parent relation associated with the result + * 'target' is the PathTarget to be computed + * 'mmaggregates' is a list of MinMaxAggInfo structs + * 'quals' is the HAVING quals if any + */ +MinMaxAggPath * +create_minmaxagg_path(PlannerInfo *root, + RelOptInfo *rel, + PathTarget *target, + List *mmaggregates, + List *quals) +{ + MinMaxAggPath *pathnode = makeNode(MinMaxAggPath); + Cost initplan_cost; + ListCell *lc; + + /* The topmost generated Plan node will be a Result */ + pathnode->path.pathtype = T_Result; + pathnode->path.parent = rel; + pathnode->path.pathtarget = target; + /* For now, assume we are above any joins, so no parameterization */ + pathnode->path.param_info = NULL; + pathnode->path.parallel_aware = false; + /* A MinMaxAggPath implies use of initplans, so cannot be parallel-safe */ + pathnode->path.parallel_safe = false; + pathnode->path.parallel_workers = 0; + /* Result is one unordered row */ + pathnode->path.rows = 1; + pathnode->path.pathkeys = NIL; + + pathnode->mmaggregates = mmaggregates; + pathnode->quals = quals; + + /* Calculate cost of all the initplans ... */ + initplan_cost = 0; + foreach(lc, mmaggregates) + { + MinMaxAggInfo *mminfo = (MinMaxAggInfo *) lfirst(lc); + + initplan_cost += mminfo->pathcost; + } + + /* add tlist eval cost for each output row, plus cpu_tuple_cost */ + pathnode->path.startup_cost = initplan_cost + target->cost.startup; + pathnode->path.total_cost = initplan_cost + target->cost.startup + + target->cost.per_tuple + cpu_tuple_cost; + + /* + * Add cost of qual, if any --- but we ignore its selectivity, since our + * rowcount estimate should be 1 no matter what the qual is. + */ + if (quals) + { + QualCost qual_cost; + + cost_qual_eval(&qual_cost, quals, root); + pathnode->path.startup_cost += qual_cost.startup; + pathnode->path.total_cost += qual_cost.startup + qual_cost.per_tuple; + } + + return pathnode; +} + +/* + * create_windowagg_path + * Creates a pathnode that represents computation of window functions + * + * 'rel' is the parent relation associated with the result + * 'subpath' is the path representing the source of data + * 'target' is the PathTarget to be computed + * 'windowFuncs' is a list of WindowFunc structs + * 'winclause' is a WindowClause that is common to all the WindowFuncs + * 'qual' WindowClause.runconditions from lower-level WindowAggPaths. + * Must always be NIL when topwindow == false + * 'topwindow' pass as true only for the top-level WindowAgg. False for all + * intermediate WindowAggs. + * + * The input must be sorted according to the WindowClause's PARTITION keys + * plus ORDER BY keys. + */ +WindowAggPath * +create_windowagg_path(PlannerInfo *root, + RelOptInfo *rel, + Path *subpath, + PathTarget *target, + List *windowFuncs, + WindowClause *winclause, + List *qual, + bool topwindow) +{ + WindowAggPath *pathnode = makeNode(WindowAggPath); + + /* qual can only be set for the topwindow */ + Assert(qual == NIL || topwindow); + + pathnode->path.pathtype = T_WindowAgg; + pathnode->path.parent = rel; + pathnode->path.pathtarget = target; + /* For now, assume we are above any joins, so no parameterization */ + pathnode->path.param_info = NULL; + pathnode->path.parallel_aware = false; + pathnode->path.parallel_safe = rel->consider_parallel && + subpath->parallel_safe; + pathnode->path.parallel_workers = subpath->parallel_workers; + /* WindowAgg preserves the input sort order */ + pathnode->path.pathkeys = subpath->pathkeys; + + pathnode->subpath = subpath; + pathnode->winclause = winclause; + pathnode->qual = qual; + pathnode->topwindow = topwindow; + + /* + * For costing purposes, assume that there are no redundant partitioning + * or ordering columns; it's not worth the trouble to deal with that + * corner case here. So we just pass the unmodified list lengths to + * cost_windowagg. + */ + cost_windowagg(&pathnode->path, root, + windowFuncs, + list_length(winclause->partitionClause), + list_length(winclause->orderClause), + subpath->startup_cost, + subpath->total_cost, + subpath->rows); + + /* add tlist eval cost for each output row */ + pathnode->path.startup_cost += target->cost.startup; + pathnode->path.total_cost += target->cost.startup + + target->cost.per_tuple * pathnode->path.rows; + + return pathnode; +} + +/* + * create_setop_path + * Creates a pathnode that represents computation of INTERSECT or EXCEPT + * + * 'rel' is the parent relation associated with the result + * 'subpath' is the path representing the source of data + * 'cmd' is the specific semantics (INTERSECT or EXCEPT, with/without ALL) + * 'strategy' is the implementation strategy (sorted or hashed) + * 'distinctList' is a list of SortGroupClause's representing the grouping + * 'flagColIdx' is the column number where the flag column will be, if any + * 'firstFlag' is the flag value for the first input relation when hashing; + * or -1 when sorting + * 'numGroups' is the estimated number of distinct groups + * 'outputRows' is the estimated number of output rows + */ +SetOpPath * +create_setop_path(PlannerInfo *root, + RelOptInfo *rel, + Path *subpath, + SetOpCmd cmd, + SetOpStrategy strategy, + List *distinctList, + AttrNumber flagColIdx, + int firstFlag, + double numGroups, + double outputRows) +{ + SetOpPath *pathnode = makeNode(SetOpPath); + + pathnode->path.pathtype = T_SetOp; + pathnode->path.parent = rel; + /* SetOp doesn't project, so use source path's pathtarget */ + pathnode->path.pathtarget = subpath->pathtarget; + /* For now, assume we are above any joins, so no parameterization */ + pathnode->path.param_info = NULL; + pathnode->path.parallel_aware = false; + pathnode->path.parallel_safe = rel->consider_parallel && + subpath->parallel_safe; + pathnode->path.parallel_workers = subpath->parallel_workers; + /* SetOp preserves the input sort order if in sort mode */ + pathnode->path.pathkeys = + (strategy == SETOP_SORTED) ? subpath->pathkeys : NIL; + + pathnode->subpath = subpath; + pathnode->cmd = cmd; + pathnode->strategy = strategy; + pathnode->distinctList = distinctList; + pathnode->flagColIdx = flagColIdx; + pathnode->firstFlag = firstFlag; + pathnode->numGroups = numGroups; + + /* + * Charge one cpu_operator_cost per comparison per input tuple. We assume + * all columns get compared at most of the tuples. + */ + pathnode->path.startup_cost = subpath->startup_cost; + pathnode->path.total_cost = subpath->total_cost + + cpu_operator_cost * subpath->rows * list_length(distinctList); + pathnode->path.rows = outputRows; + + return pathnode; +} + +/* + * create_recursiveunion_path + * Creates a pathnode that represents a recursive UNION node + * + * 'rel' is the parent relation associated with the result + * 'leftpath' is the source of data for the non-recursive term + * 'rightpath' is the source of data for the recursive term + * 'target' is the PathTarget to be computed + * 'distinctList' is a list of SortGroupClause's representing the grouping + * 'wtParam' is the ID of Param representing work table + * 'numGroups' is the estimated number of groups + * + * For recursive UNION ALL, distinctList is empty and numGroups is zero + */ +RecursiveUnionPath * +create_recursiveunion_path(PlannerInfo *root, + RelOptInfo *rel, + Path *leftpath, + Path *rightpath, + PathTarget *target, + List *distinctList, + int wtParam, + double numGroups) +{ + RecursiveUnionPath *pathnode = makeNode(RecursiveUnionPath); + + pathnode->path.pathtype = T_RecursiveUnion; + pathnode->path.parent = rel; + pathnode->path.pathtarget = target; + /* For now, assume we are above any joins, so no parameterization */ + pathnode->path.param_info = NULL; + pathnode->path.parallel_aware = false; + pathnode->path.parallel_safe = rel->consider_parallel && + leftpath->parallel_safe && rightpath->parallel_safe; + /* Foolish, but we'll do it like joins for now: */ + pathnode->path.parallel_workers = leftpath->parallel_workers; + /* RecursiveUnion result is always unsorted */ + pathnode->path.pathkeys = NIL; + + pathnode->leftpath = leftpath; + pathnode->rightpath = rightpath; + pathnode->distinctList = distinctList; + pathnode->wtParam = wtParam; + pathnode->numGroups = numGroups; + + cost_recursive_union(&pathnode->path, leftpath, rightpath); + + return pathnode; +} + +/* + * create_lockrows_path + * Creates a pathnode that represents acquiring row locks + * + * 'rel' is the parent relation associated with the result + * 'subpath' is the path representing the source of data + * 'rowMarks' is a list of PlanRowMark's + * 'epqParam' is the ID of Param for EvalPlanQual re-eval + */ +LockRowsPath * +create_lockrows_path(PlannerInfo *root, RelOptInfo *rel, + Path *subpath, List *rowMarks, int epqParam) +{ + LockRowsPath *pathnode = makeNode(LockRowsPath); + + pathnode->path.pathtype = T_LockRows; + pathnode->path.parent = rel; + /* LockRows doesn't project, so use source path's pathtarget */ + pathnode->path.pathtarget = subpath->pathtarget; + /* For now, assume we are above any joins, so no parameterization */ + pathnode->path.param_info = NULL; + pathnode->path.parallel_aware = false; + pathnode->path.parallel_safe = false; + pathnode->path.parallel_workers = 0; + pathnode->path.rows = subpath->rows; + + /* + * The result cannot be assumed sorted, since locking might cause the sort + * key columns to be replaced with new values. + */ + pathnode->path.pathkeys = NIL; + + pathnode->subpath = subpath; + pathnode->rowMarks = rowMarks; + pathnode->epqParam = epqParam; + + /* + * We should charge something extra for the costs of row locking and + * possible refetches, but it's hard to say how much. For now, use + * cpu_tuple_cost per row. + */ + pathnode->path.startup_cost = subpath->startup_cost; + pathnode->path.total_cost = subpath->total_cost + + cpu_tuple_cost * subpath->rows; + + return pathnode; +} + +/* + * create_modifytable_path + * Creates a pathnode that represents performing INSERT/UPDATE/DELETE/MERGE + * mods + * + * 'rel' is the parent relation associated with the result + * 'subpath' is a Path producing source data + * 'operation' is the operation type + * 'canSetTag' is true if we set the command tag/es_processed + * 'nominalRelation' is the parent RT index for use of EXPLAIN + * 'rootRelation' is the partitioned/inherited table root RTI, or 0 if none + * 'partColsUpdated' is true if any partitioning columns are being updated, + * either from the target relation or a descendent partitioned table. + * 'resultRelations' is an integer list of actual RT indexes of target rel(s) + * 'updateColnosLists' is a list of UPDATE target column number lists + * (one sublist per rel); or NIL if not an UPDATE + * 'withCheckOptionLists' is a list of WCO lists (one per rel) + * 'returningLists' is a list of RETURNING tlists (one per rel) + * 'rowMarks' is a list of PlanRowMarks (non-locking only) + * 'onconflict' is the ON CONFLICT clause, or NULL + * 'epqParam' is the ID of Param for EvalPlanQual re-eval + * 'mergeActionLists' is a list of lists of MERGE actions (one per rel) + */ +ModifyTablePath * +create_modifytable_path(PlannerInfo *root, RelOptInfo *rel, + Path *subpath, + CmdType operation, bool canSetTag, + Index nominalRelation, Index rootRelation, + bool partColsUpdated, + List *resultRelations, + List *updateColnosLists, + List *withCheckOptionLists, List *returningLists, + List *rowMarks, OnConflictExpr *onconflict, + List *mergeActionLists, int epqParam) +{ + ModifyTablePath *pathnode = makeNode(ModifyTablePath); + + Assert(operation == CMD_MERGE || + (operation == CMD_UPDATE ? + list_length(resultRelations) == list_length(updateColnosLists) : + updateColnosLists == NIL)); + Assert(withCheckOptionLists == NIL || + list_length(resultRelations) == list_length(withCheckOptionLists)); + Assert(returningLists == NIL || + list_length(resultRelations) == list_length(returningLists)); + + pathnode->path.pathtype = T_ModifyTable; + pathnode->path.parent = rel; + /* pathtarget is not interesting, just make it minimally valid */ + pathnode->path.pathtarget = rel->reltarget; + /* For now, assume we are above any joins, so no parameterization */ + pathnode->path.param_info = NULL; + pathnode->path.parallel_aware = false; + pathnode->path.parallel_safe = false; + pathnode->path.parallel_workers = 0; + pathnode->path.pathkeys = NIL; + + /* + * Compute cost & rowcount as subpath cost & rowcount (if RETURNING) + * + * Currently, we don't charge anything extra for the actual table + * modification work, nor for the WITH CHECK OPTIONS or RETURNING + * expressions if any. It would only be window dressing, since + * ModifyTable is always a top-level node and there is no way for the + * costs to change any higher-level planning choices. But we might want + * to make it look better sometime. + */ + pathnode->path.startup_cost = subpath->startup_cost; + pathnode->path.total_cost = subpath->total_cost; + if (returningLists != NIL) + { + pathnode->path.rows = subpath->rows; + + /* + * Set width to match the subpath output. XXX this is totally wrong: + * we should return an average of the RETURNING tlist widths. But + * it's what happened historically, and improving it is a task for + * another day. (Again, it's mostly window dressing.) + */ + pathnode->path.pathtarget->width = subpath->pathtarget->width; + } + else + { + pathnode->path.rows = 0; + pathnode->path.pathtarget->width = 0; + } + + pathnode->subpath = subpath; + pathnode->operation = operation; + pathnode->canSetTag = canSetTag; + pathnode->nominalRelation = nominalRelation; + pathnode->rootRelation = rootRelation; + pathnode->partColsUpdated = partColsUpdated; + pathnode->resultRelations = resultRelations; + pathnode->updateColnosLists = updateColnosLists; + pathnode->withCheckOptionLists = withCheckOptionLists; + pathnode->returningLists = returningLists; + pathnode->rowMarks = rowMarks; + pathnode->onconflict = onconflict; + pathnode->epqParam = epqParam; + pathnode->mergeActionLists = mergeActionLists; + + return pathnode; +} + +/* + * create_limit_path + * Creates a pathnode that represents performing LIMIT/OFFSET + * + * In addition to providing the actual OFFSET and LIMIT expressions, + * the caller must provide estimates of their values for costing purposes. + * The estimates are as computed by preprocess_limit(), ie, 0 represents + * the clause not being present, and -1 means it's present but we could + * not estimate its value. + * + * 'rel' is the parent relation associated with the result + * 'subpath' is the path representing the source of data + * 'limitOffset' is the actual OFFSET expression, or NULL + * 'limitCount' is the actual LIMIT expression, or NULL + * 'offset_est' is the estimated value of the OFFSET expression + * 'count_est' is the estimated value of the LIMIT expression + */ +LimitPath * +create_limit_path(PlannerInfo *root, RelOptInfo *rel, + Path *subpath, + Node *limitOffset, Node *limitCount, + LimitOption limitOption, + int64 offset_est, int64 count_est) +{ + LimitPath *pathnode = makeNode(LimitPath); + + pathnode->path.pathtype = T_Limit; + pathnode->path.parent = rel; + /* Limit doesn't project, so use source path's pathtarget */ + pathnode->path.pathtarget = subpath->pathtarget; + /* For now, assume we are above any joins, so no parameterization */ + pathnode->path.param_info = NULL; + pathnode->path.parallel_aware = false; + pathnode->path.parallel_safe = rel->consider_parallel && + subpath->parallel_safe; + pathnode->path.parallel_workers = subpath->parallel_workers; + pathnode->path.rows = subpath->rows; + pathnode->path.startup_cost = subpath->startup_cost; + pathnode->path.total_cost = subpath->total_cost; + pathnode->path.pathkeys = subpath->pathkeys; + pathnode->subpath = subpath; + pathnode->limitOffset = limitOffset; + pathnode->limitCount = limitCount; + pathnode->limitOption = limitOption; + + /* + * Adjust the output rows count and costs according to the offset/limit. + */ + adjust_limit_rows_costs(&pathnode->path.rows, + &pathnode->path.startup_cost, + &pathnode->path.total_cost, + offset_est, count_est); + + return pathnode; +} + +/* + * adjust_limit_rows_costs + * Adjust the size and cost estimates for a LimitPath node according to the + * offset/limit. + * + * This is only a cosmetic issue if we are at top level, but if we are + * building a subquery then it's important to report correct info to the outer + * planner. + * + * When the offset or count couldn't be estimated, use 10% of the estimated + * number of rows emitted from the subpath. + * + * XXX we don't bother to add eval costs of the offset/limit expressions + * themselves to the path costs. In theory we should, but in most cases those + * expressions are trivial and it's just not worth the trouble. + */ +void +adjust_limit_rows_costs(double *rows, /* in/out parameter */ + Cost *startup_cost, /* in/out parameter */ + Cost *total_cost, /* in/out parameter */ + int64 offset_est, + int64 count_est) +{ + double input_rows = *rows; + Cost input_startup_cost = *startup_cost; + Cost input_total_cost = *total_cost; + + if (offset_est != 0) + { + double offset_rows; + + if (offset_est > 0) + offset_rows = (double) offset_est; + else + offset_rows = clamp_row_est(input_rows * 0.10); + if (offset_rows > *rows) + offset_rows = *rows; + if (input_rows > 0) + *startup_cost += + (input_total_cost - input_startup_cost) + * offset_rows / input_rows; + *rows -= offset_rows; + if (*rows < 1) + *rows = 1; + } + + if (count_est != 0) + { + double count_rows; + + if (count_est > 0) + count_rows = (double) count_est; + else + count_rows = clamp_row_est(input_rows * 0.10); + if (count_rows > *rows) + count_rows = *rows; + if (input_rows > 0) + *total_cost = *startup_cost + + (input_total_cost - input_startup_cost) + * count_rows / input_rows; + *rows = count_rows; + if (*rows < 1) + *rows = 1; + } +} + + +/* + * reparameterize_path + * Attempt to modify a Path to have greater parameterization + * + * We use this to attempt to bring all child paths of an appendrel to the + * same parameterization level, ensuring that they all enforce the same set + * of join quals (and thus that that parameterization can be attributed to + * an append path built from such paths). Currently, only a few path types + * are supported here, though more could be added at need. We return NULL + * if we can't reparameterize the given path. + * + * Note: we intentionally do not pass created paths to add_path(); it would + * possibly try to delete them on the grounds of being cost-inferior to the + * paths they were made from, and we don't want that. Paths made here are + * not necessarily of general-purpose usefulness, but they can be useful + * as members of an append path. + */ +Path * +reparameterize_path(PlannerInfo *root, Path *path, + Relids required_outer, + double loop_count) +{ + RelOptInfo *rel = path->parent; + + /* Can only increase, not decrease, path's parameterization */ + if (!bms_is_subset(PATH_REQ_OUTER(path), required_outer)) + return NULL; + switch (path->pathtype) + { + case T_SeqScan: + return create_seqscan_path(root, rel, required_outer, 0); + case T_SampleScan: + return (Path *) create_samplescan_path(root, rel, required_outer); + case T_IndexScan: + case T_IndexOnlyScan: + { + IndexPath *ipath = (IndexPath *) path; + IndexPath *newpath = makeNode(IndexPath); + + /* + * We can't use create_index_path directly, and would not want + * to because it would re-compute the indexqual conditions + * which is wasted effort. Instead we hack things a bit: + * flat-copy the path node, revise its param_info, and redo + * the cost estimate. + */ + memcpy(newpath, ipath, sizeof(IndexPath)); + newpath->path.param_info = + get_baserel_parampathinfo(root, rel, required_outer); + cost_index(newpath, root, loop_count, false); + return (Path *) newpath; + } + case T_BitmapHeapScan: + { + BitmapHeapPath *bpath = (BitmapHeapPath *) path; + + return (Path *) create_bitmap_heap_path(root, + rel, + bpath->bitmapqual, + required_outer, + loop_count, 0); + } + case T_SubqueryScan: + { + SubqueryScanPath *spath = (SubqueryScanPath *) path; + + return (Path *) create_subqueryscan_path(root, + rel, + spath->subpath, + spath->path.pathkeys, + required_outer); + } + case T_Result: + /* Supported only for RTE_RESULT scan paths */ + if (IsA(path, Path)) + return create_resultscan_path(root, rel, required_outer); + break; + case T_Append: + { + AppendPath *apath = (AppendPath *) path; + List *childpaths = NIL; + List *partialpaths = NIL; + int i; + ListCell *lc; + + /* Reparameterize the children */ + i = 0; + foreach(lc, apath->subpaths) + { + Path *spath = (Path *) lfirst(lc); + + spath = reparameterize_path(root, spath, + required_outer, + loop_count); + if (spath == NULL) + return NULL; + /* We have to re-split the regular and partial paths */ + if (i < apath->first_partial_path) + childpaths = lappend(childpaths, spath); + else + partialpaths = lappend(partialpaths, spath); + i++; + } + return (Path *) + create_append_path(root, rel, childpaths, partialpaths, + apath->path.pathkeys, required_outer, + apath->path.parallel_workers, + apath->path.parallel_aware, + -1); + } + case T_Memoize: + { + MemoizePath *mpath = (MemoizePath *) path; + Path *spath = mpath->subpath; + + spath = reparameterize_path(root, spath, + required_outer, + loop_count); + if (spath == NULL) + return NULL; + return (Path *) create_memoize_path(root, rel, + spath, + mpath->param_exprs, + mpath->hash_operators, + mpath->singlerow, + mpath->binary_mode, + mpath->calls); + } + default: + break; + } + return NULL; +} + +/* + * reparameterize_path_by_child + * Given a path parameterized by the parent of the given child relation, + * translate the path to be parameterized by the given child relation. + * + * The function creates a new path of the same type as the given path, but + * parameterized by the given child relation. Most fields from the original + * path can simply be flat-copied, but any expressions must be adjusted to + * refer to the correct varnos, and any paths must be recursively + * reparameterized. Other fields that refer to specific relids also need + * adjustment. + * + * The cost, number of rows, width and parallel path properties depend upon + * path->parent, which does not change during the translation. Hence those + * members are copied as they are. + * + * If the given path can not be reparameterized, the function returns NULL. + */ +Path * +reparameterize_path_by_child(PlannerInfo *root, Path *path, + RelOptInfo *child_rel) +{ + +#define FLAT_COPY_PATH(newnode, node, nodetype) \ + ( (newnode) = makeNode(nodetype), \ + memcpy((newnode), (node), sizeof(nodetype)) ) + +#define ADJUST_CHILD_ATTRS(node) \ + ((node) = \ + (List *) adjust_appendrel_attrs_multilevel(root, (Node *) (node), \ + child_rel->relids, \ + child_rel->top_parent_relids)) + +#define REPARAMETERIZE_CHILD_PATH(path) \ +do { \ + (path) = reparameterize_path_by_child(root, (path), child_rel); \ + if ((path) == NULL) \ + return NULL; \ +} while(0) + +#define REPARAMETERIZE_CHILD_PATH_LIST(pathlist) \ +do { \ + if ((pathlist) != NIL) \ + { \ + (pathlist) = reparameterize_pathlist_by_child(root, (pathlist), \ + child_rel); \ + if ((pathlist) == NIL) \ + return NULL; \ + } \ +} while(0) + + Path *new_path; + ParamPathInfo *new_ppi; + ParamPathInfo *old_ppi; + Relids required_outer; + + /* + * If the path is not parameterized by parent of the given relation, it + * doesn't need reparameterization. + */ + if (!path->param_info || + !bms_overlap(PATH_REQ_OUTER(path), child_rel->top_parent_relids)) + return path; + + /* + * If possible, reparameterize the given path, making a copy. + * + * This function is currently only applied to the inner side of a nestloop + * join that is being partitioned by the partitionwise-join code. Hence, + * we need only support path types that plausibly arise in that context. + * (In particular, supporting sorted path types would be a waste of code + * and cycles: even if we translated them here, they'd just lose in + * subsequent cost comparisons.) If we do see an unsupported path type, + * that just means we won't be able to generate a partitionwise-join plan + * using that path type. + */ + switch (nodeTag(path)) + { + case T_Path: + FLAT_COPY_PATH(new_path, path, Path); + break; + + case T_IndexPath: + { + IndexPath *ipath; + + FLAT_COPY_PATH(ipath, path, IndexPath); + ADJUST_CHILD_ATTRS(ipath->indexclauses); + new_path = (Path *) ipath; + } + break; + + case T_BitmapHeapPath: + { + BitmapHeapPath *bhpath; + + FLAT_COPY_PATH(bhpath, path, BitmapHeapPath); + REPARAMETERIZE_CHILD_PATH(bhpath->bitmapqual); + new_path = (Path *) bhpath; + } + break; + + case T_BitmapAndPath: + { + BitmapAndPath *bapath; + + FLAT_COPY_PATH(bapath, path, BitmapAndPath); + REPARAMETERIZE_CHILD_PATH_LIST(bapath->bitmapquals); + new_path = (Path *) bapath; + } + break; + + case T_BitmapOrPath: + { + BitmapOrPath *bopath; + + FLAT_COPY_PATH(bopath, path, BitmapOrPath); + REPARAMETERIZE_CHILD_PATH_LIST(bopath->bitmapquals); + new_path = (Path *) bopath; + } + break; + + case T_ForeignPath: + { + ForeignPath *fpath; + ReparameterizeForeignPathByChild_function rfpc_func; + + FLAT_COPY_PATH(fpath, path, ForeignPath); + if (fpath->fdw_outerpath) + REPARAMETERIZE_CHILD_PATH(fpath->fdw_outerpath); + + /* Hand over to FDW if needed. */ + rfpc_func = + path->parent->fdwroutine->ReparameterizeForeignPathByChild; + if (rfpc_func) + fpath->fdw_private = rfpc_func(root, fpath->fdw_private, + child_rel); + new_path = (Path *) fpath; + } + break; + + case T_CustomPath: + { + CustomPath *cpath; + + FLAT_COPY_PATH(cpath, path, CustomPath); + REPARAMETERIZE_CHILD_PATH_LIST(cpath->custom_paths); + if (cpath->methods && + cpath->methods->ReparameterizeCustomPathByChild) + cpath->custom_private = + cpath->methods->ReparameterizeCustomPathByChild(root, + cpath->custom_private, + child_rel); + new_path = (Path *) cpath; + } + break; + + case T_NestPath: + { + JoinPath *jpath; + NestPath *npath; + + FLAT_COPY_PATH(npath, path, NestPath); + + jpath = (JoinPath *) npath; + REPARAMETERIZE_CHILD_PATH(jpath->outerjoinpath); + REPARAMETERIZE_CHILD_PATH(jpath->innerjoinpath); + ADJUST_CHILD_ATTRS(jpath->joinrestrictinfo); + new_path = (Path *) npath; + } + break; + + case T_MergePath: + { + JoinPath *jpath; + MergePath *mpath; + + FLAT_COPY_PATH(mpath, path, MergePath); + + jpath = (JoinPath *) mpath; + REPARAMETERIZE_CHILD_PATH(jpath->outerjoinpath); + REPARAMETERIZE_CHILD_PATH(jpath->innerjoinpath); + ADJUST_CHILD_ATTRS(jpath->joinrestrictinfo); + ADJUST_CHILD_ATTRS(mpath->path_mergeclauses); + new_path = (Path *) mpath; + } + break; + + case T_HashPath: + { + JoinPath *jpath; + HashPath *hpath; + + FLAT_COPY_PATH(hpath, path, HashPath); + + jpath = (JoinPath *) hpath; + REPARAMETERIZE_CHILD_PATH(jpath->outerjoinpath); + REPARAMETERIZE_CHILD_PATH(jpath->innerjoinpath); + ADJUST_CHILD_ATTRS(jpath->joinrestrictinfo); + ADJUST_CHILD_ATTRS(hpath->path_hashclauses); + new_path = (Path *) hpath; + } + break; + + case T_AppendPath: + { + AppendPath *apath; + + FLAT_COPY_PATH(apath, path, AppendPath); + REPARAMETERIZE_CHILD_PATH_LIST(apath->subpaths); + new_path = (Path *) apath; + } + break; + + case T_MemoizePath: + { + MemoizePath *mpath; + + FLAT_COPY_PATH(mpath, path, MemoizePath); + REPARAMETERIZE_CHILD_PATH(mpath->subpath); + ADJUST_CHILD_ATTRS(mpath->param_exprs); + new_path = (Path *) mpath; + } + break; + + case T_GatherPath: + { + GatherPath *gpath; + + FLAT_COPY_PATH(gpath, path, GatherPath); + REPARAMETERIZE_CHILD_PATH(gpath->subpath); + new_path = (Path *) gpath; + } + break; + + default: + + /* We don't know how to reparameterize this path. */ + return NULL; + } + + /* + * Adjust the parameterization information, which refers to the topmost + * parent. The topmost parent can be multiple levels away from the given + * child, hence use multi-level expression adjustment routines. + */ + old_ppi = new_path->param_info; + required_outer = + adjust_child_relids_multilevel(root, old_ppi->ppi_req_outer, + child_rel->relids, + child_rel->top_parent_relids); + + /* If we already have a PPI for this parameterization, just return it */ + new_ppi = find_param_path_info(new_path->parent, required_outer); + + /* + * If not, build a new one and link it to the list of PPIs. For the same + * reason as explained in mark_dummy_rel(), allocate new PPI in the same + * context the given RelOptInfo is in. + */ + if (new_ppi == NULL) + { + MemoryContext oldcontext; + RelOptInfo *rel = path->parent; + + oldcontext = MemoryContextSwitchTo(GetMemoryChunkContext(rel)); + + new_ppi = makeNode(ParamPathInfo); + new_ppi->ppi_req_outer = bms_copy(required_outer); + new_ppi->ppi_rows = old_ppi->ppi_rows; + new_ppi->ppi_clauses = old_ppi->ppi_clauses; + ADJUST_CHILD_ATTRS(new_ppi->ppi_clauses); + rel->ppilist = lappend(rel->ppilist, new_ppi); + + MemoryContextSwitchTo(oldcontext); + } + bms_free(required_outer); + + new_path->param_info = new_ppi; + + /* + * Adjust the path target if the parent of the outer relation is + * referenced in the targetlist. This can happen when only the parent of + * outer relation is laterally referenced in this relation. + */ + if (bms_overlap(path->parent->lateral_relids, + child_rel->top_parent_relids)) + { + new_path->pathtarget = copy_pathtarget(new_path->pathtarget); + ADJUST_CHILD_ATTRS(new_path->pathtarget->exprs); + } + + return new_path; +} + +/* + * reparameterize_pathlist_by_child + * Helper function to reparameterize a list of paths by given child rel. + */ +static List * +reparameterize_pathlist_by_child(PlannerInfo *root, + List *pathlist, + RelOptInfo *child_rel) +{ + ListCell *lc; + List *result = NIL; + + foreach(lc, pathlist) + { + Path *path = reparameterize_path_by_child(root, lfirst(lc), + child_rel); + + if (path == NULL) + { + list_free(result); + return NIL; + } + + result = lappend(result, path); + } + + return result; +} diff --git a/src/backend/optimizer/util/placeholder.c b/src/backend/optimizer/util/placeholder.c new file mode 100644 index 0000000..3b0f058 --- /dev/null +++ b/src/backend/optimizer/util/placeholder.c @@ -0,0 +1,477 @@ +/*------------------------------------------------------------------------- + * + * placeholder.c + * PlaceHolderVar and PlaceHolderInfo manipulation routines + * + * + * Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group + * Portions Copyright (c) 1994, Regents of the University of California + * + * + * IDENTIFICATION + * src/backend/optimizer/util/placeholder.c + * + *------------------------------------------------------------------------- + */ +#include "postgres.h" + +#include "nodes/nodeFuncs.h" +#include "optimizer/cost.h" +#include "optimizer/optimizer.h" +#include "optimizer/pathnode.h" +#include "optimizer/placeholder.h" +#include "optimizer/planmain.h" +#include "utils/lsyscache.h" + +/* Local functions */ +static void find_placeholders_recurse(PlannerInfo *root, Node *jtnode); +static void find_placeholders_in_expr(PlannerInfo *root, Node *expr); + + +/* + * make_placeholder_expr + * Make a PlaceHolderVar for the given expression. + * + * phrels is the syntactic location (as a set of baserels) to attribute + * to the expression. + */ +PlaceHolderVar * +make_placeholder_expr(PlannerInfo *root, Expr *expr, Relids phrels) +{ + PlaceHolderVar *phv = makeNode(PlaceHolderVar); + + phv->phexpr = expr; + phv->phrels = phrels; + phv->phid = ++(root->glob->lastPHId); + phv->phlevelsup = 0; + + return phv; +} + +/* + * find_placeholder_info + * Fetch the PlaceHolderInfo for the given PHV + * + * If the PlaceHolderInfo doesn't exist yet, create it if create_new_ph is + * true, else throw an error. + * + * This is separate from make_placeholder_expr because subquery pullup has + * to make PlaceHolderVars for expressions that might not be used at all in + * the upper query, or might not remain after const-expression simplification. + * We build PlaceHolderInfos only for PHVs that are still present in the + * simplified query passed to query_planner(). + * + * Note: this should only be called after query_planner() has started. Also, + * create_new_ph must not be true after deconstruct_jointree begins, because + * make_outerjoininfo assumes that we already know about all placeholders. + */ +PlaceHolderInfo * +find_placeholder_info(PlannerInfo *root, PlaceHolderVar *phv, + bool create_new_ph) +{ + PlaceHolderInfo *phinfo; + Relids rels_used; + ListCell *lc; + + /* if this ever isn't true, we'd need to be able to look in parent lists */ + Assert(phv->phlevelsup == 0); + + foreach(lc, root->placeholder_list) + { + phinfo = (PlaceHolderInfo *) lfirst(lc); + if (phinfo->phid == phv->phid) + return phinfo; + } + + /* Not found, so create it */ + if (!create_new_ph) + elog(ERROR, "too late to create a new PlaceHolderInfo"); + + phinfo = makeNode(PlaceHolderInfo); + + phinfo->phid = phv->phid; + phinfo->ph_var = copyObject(phv); + + /* + * Any referenced rels that are outside the PHV's syntactic scope are + * LATERAL references, which should be included in ph_lateral but not in + * ph_eval_at. If no referenced rels are within the syntactic scope, + * force evaluation at the syntactic location. + */ + rels_used = pull_varnos(root, (Node *) phv->phexpr); + phinfo->ph_lateral = bms_difference(rels_used, phv->phrels); + if (bms_is_empty(phinfo->ph_lateral)) + phinfo->ph_lateral = NULL; /* make it exactly NULL if empty */ + phinfo->ph_eval_at = bms_int_members(rels_used, phv->phrels); + /* If no contained vars, force evaluation at syntactic location */ + if (bms_is_empty(phinfo->ph_eval_at)) + { + phinfo->ph_eval_at = bms_copy(phv->phrels); + Assert(!bms_is_empty(phinfo->ph_eval_at)); + } + /* ph_eval_at may change later, see update_placeholder_eval_levels */ + phinfo->ph_needed = NULL; /* initially it's unused */ + /* for the moment, estimate width using just the datatype info */ + phinfo->ph_width = get_typavgwidth(exprType((Node *) phv->phexpr), + exprTypmod((Node *) phv->phexpr)); + + root->placeholder_list = lappend(root->placeholder_list, phinfo); + + /* + * The PHV's contained expression may contain other, lower-level PHVs. We + * now know we need to get those into the PlaceHolderInfo list, too, so we + * may as well do that immediately. + */ + find_placeholders_in_expr(root, (Node *) phinfo->ph_var->phexpr); + + return phinfo; +} + +/* + * find_placeholders_in_jointree + * Search the jointree for PlaceHolderVars, and build PlaceHolderInfos + * + * We don't need to look at the targetlist because build_base_rel_tlists() + * will already have made entries for any PHVs in the tlist. + * + * This is called before we begin deconstruct_jointree. Once we begin + * deconstruct_jointree, all active placeholders must be present in + * root->placeholder_list, because make_outerjoininfo and + * update_placeholder_eval_levels require this info to be available + * while we crawl up the join tree. + */ +void +find_placeholders_in_jointree(PlannerInfo *root) +{ + /* We need do nothing if the query contains no PlaceHolderVars */ + if (root->glob->lastPHId != 0) + { + /* Start recursion at top of jointree */ + Assert(root->parse->jointree != NULL && + IsA(root->parse->jointree, FromExpr)); + find_placeholders_recurse(root, (Node *) root->parse->jointree); + } +} + +/* + * find_placeholders_recurse + * One recursion level of find_placeholders_in_jointree. + * + * jtnode is the current jointree node to examine. + */ +static void +find_placeholders_recurse(PlannerInfo *root, Node *jtnode) +{ + if (jtnode == NULL) + return; + if (IsA(jtnode, RangeTblRef)) + { + /* No quals to deal with here */ + } + else if (IsA(jtnode, FromExpr)) + { + FromExpr *f = (FromExpr *) jtnode; + ListCell *l; + + /* + * First, recurse to handle child joins. + */ + foreach(l, f->fromlist) + { + find_placeholders_recurse(root, lfirst(l)); + } + + /* + * Now process the top-level quals. + */ + find_placeholders_in_expr(root, f->quals); + } + else if (IsA(jtnode, JoinExpr)) + { + JoinExpr *j = (JoinExpr *) jtnode; + + /* + * First, recurse to handle child joins. + */ + find_placeholders_recurse(root, j->larg); + find_placeholders_recurse(root, j->rarg); + + /* Process the qual clauses */ + find_placeholders_in_expr(root, j->quals); + } + else + elog(ERROR, "unrecognized node type: %d", + (int) nodeTag(jtnode)); +} + +/* + * find_placeholders_in_expr + * Find all PlaceHolderVars in the given expression, and create + * PlaceHolderInfo entries for them. + */ +static void +find_placeholders_in_expr(PlannerInfo *root, Node *expr) +{ + List *vars; + ListCell *vl; + + /* + * pull_var_clause does more than we need here, but it'll do and it's + * convenient to use. + */ + vars = pull_var_clause(expr, + PVC_RECURSE_AGGREGATES | + PVC_RECURSE_WINDOWFUNCS | + PVC_INCLUDE_PLACEHOLDERS); + foreach(vl, vars) + { + PlaceHolderVar *phv = (PlaceHolderVar *) lfirst(vl); + + /* Ignore any plain Vars */ + if (!IsA(phv, PlaceHolderVar)) + continue; + + /* Create a PlaceHolderInfo entry if there's not one already */ + (void) find_placeholder_info(root, phv, true); + } + list_free(vars); +} + +/* + * update_placeholder_eval_levels + * Adjust the target evaluation levels for placeholders + * + * The initial eval_at level set by find_placeholder_info was the set of + * rels used in the placeholder's expression (or the whole subselect below + * the placeholder's syntactic location, if the expr is variable-free). + * If the query contains any outer joins that can null any of those rels, + * we must delay evaluation to above those joins. + * + * We repeat this operation each time we add another outer join to + * root->join_info_list. It's somewhat annoying to have to do that, but + * since we don't have very much information on the placeholders' locations, + * it's hard to avoid. Each placeholder's eval_at level must be correct + * by the time it starts to figure in outer-join delay decisions for higher + * outer joins. + * + * In future we might want to put additional policy/heuristics here to + * try to determine an optimal evaluation level. The current rules will + * result in evaluation at the lowest possible level. However, pushing a + * placeholder eval up the tree is likely to further constrain evaluation + * order for outer joins, so it could easily be counterproductive; and we + * don't have enough information at this point to make an intelligent choice. + */ +void +update_placeholder_eval_levels(PlannerInfo *root, SpecialJoinInfo *new_sjinfo) +{ + ListCell *lc1; + + foreach(lc1, root->placeholder_list) + { + PlaceHolderInfo *phinfo = (PlaceHolderInfo *) lfirst(lc1); + Relids syn_level = phinfo->ph_var->phrels; + Relids eval_at; + bool found_some; + ListCell *lc2; + + /* + * We don't need to do any work on this placeholder unless the + * newly-added outer join is syntactically beneath its location. + */ + if (!bms_is_subset(new_sjinfo->syn_lefthand, syn_level) || + !bms_is_subset(new_sjinfo->syn_righthand, syn_level)) + continue; + + /* + * Check for delays due to lower outer joins. This is the same logic + * as in check_outerjoin_delay in initsplan.c, except that we don't + * have anything to do with the delay_upper_joins flags; delay of + * upper outer joins will be handled later, based on the eval_at + * values we compute now. + */ + eval_at = phinfo->ph_eval_at; + + do + { + found_some = false; + foreach(lc2, root->join_info_list) + { + SpecialJoinInfo *sjinfo = (SpecialJoinInfo *) lfirst(lc2); + + /* disregard joins not within the PHV's sub-select */ + if (!bms_is_subset(sjinfo->syn_lefthand, syn_level) || + !bms_is_subset(sjinfo->syn_righthand, syn_level)) + continue; + + /* do we reference any nullable rels of this OJ? */ + if (bms_overlap(eval_at, sjinfo->min_righthand) || + (sjinfo->jointype == JOIN_FULL && + bms_overlap(eval_at, sjinfo->min_lefthand))) + { + /* yes; have we included all its rels in eval_at? */ + if (!bms_is_subset(sjinfo->min_lefthand, eval_at) || + !bms_is_subset(sjinfo->min_righthand, eval_at)) + { + /* no, so add them in */ + eval_at = bms_add_members(eval_at, + sjinfo->min_lefthand); + eval_at = bms_add_members(eval_at, + sjinfo->min_righthand); + /* we'll need another iteration */ + found_some = true; + } + } + } + } while (found_some); + + /* Can't move the PHV's eval_at level to above its syntactic level */ + Assert(bms_is_subset(eval_at, syn_level)); + + phinfo->ph_eval_at = eval_at; + } +} + +/* + * fix_placeholder_input_needed_levels + * Adjust the "needed at" levels for placeholder inputs + * + * This is called after we've finished determining the eval_at levels for + * all placeholders. We need to make sure that all vars and placeholders + * needed to evaluate each placeholder will be available at the scan or join + * level where the evaluation will be done. (It might seem that scan-level + * evaluations aren't interesting, but that's not so: a LATERAL reference + * within a placeholder's expression needs to cause the referenced var or + * placeholder to be marked as needed in the scan where it's evaluated.) + * Note that this loop can have side-effects on the ph_needed sets of other + * PlaceHolderInfos; that's okay because we don't examine ph_needed here, so + * there are no ordering issues to worry about. + */ +void +fix_placeholder_input_needed_levels(PlannerInfo *root) +{ + ListCell *lc; + + foreach(lc, root->placeholder_list) + { + PlaceHolderInfo *phinfo = (PlaceHolderInfo *) lfirst(lc); + List *vars = pull_var_clause((Node *) phinfo->ph_var->phexpr, + PVC_RECURSE_AGGREGATES | + PVC_RECURSE_WINDOWFUNCS | + PVC_INCLUDE_PLACEHOLDERS); + + add_vars_to_targetlist(root, vars, phinfo->ph_eval_at, false); + list_free(vars); + } +} + +/* + * add_placeholders_to_base_rels + * Add any required PlaceHolderVars to base rels' targetlists. + * + * If any placeholder can be computed at a base rel and is needed above it, + * add it to that rel's targetlist. This might look like it could be merged + * with fix_placeholder_input_needed_levels, but it must be separate because + * join removal happens in between, and can change the ph_eval_at sets. There + * is essentially the same logic in add_placeholders_to_joinrel, but we can't + * do that part until joinrels are formed. + */ +void +add_placeholders_to_base_rels(PlannerInfo *root) +{ + ListCell *lc; + + foreach(lc, root->placeholder_list) + { + PlaceHolderInfo *phinfo = (PlaceHolderInfo *) lfirst(lc); + Relids eval_at = phinfo->ph_eval_at; + int varno; + + if (bms_get_singleton_member(eval_at, &varno) && + bms_nonempty_difference(phinfo->ph_needed, eval_at)) + { + RelOptInfo *rel = find_base_rel(root, varno); + + rel->reltarget->exprs = lappend(rel->reltarget->exprs, + copyObject(phinfo->ph_var)); + /* reltarget's cost and width fields will be updated later */ + } + } +} + +/* + * add_placeholders_to_joinrel + * Add any required PlaceHolderVars to a join rel's targetlist; + * and if they contain lateral references, add those references to the + * joinrel's direct_lateral_relids. + * + * A join rel should emit a PlaceHolderVar if (a) the PHV can be computed + * at or below this join level and (b) the PHV is needed above this level. + * However, condition (a) is sufficient to add to direct_lateral_relids, + * as explained below. + */ +void +add_placeholders_to_joinrel(PlannerInfo *root, RelOptInfo *joinrel, + RelOptInfo *outer_rel, RelOptInfo *inner_rel) +{ + Relids relids = joinrel->relids; + ListCell *lc; + + foreach(lc, root->placeholder_list) + { + PlaceHolderInfo *phinfo = (PlaceHolderInfo *) lfirst(lc); + + /* Is it computable here? */ + if (bms_is_subset(phinfo->ph_eval_at, relids)) + { + /* Is it still needed above this joinrel? */ + if (bms_nonempty_difference(phinfo->ph_needed, relids)) + { + /* Yup, add it to the output */ + joinrel->reltarget->exprs = lappend(joinrel->reltarget->exprs, + phinfo->ph_var); + joinrel->reltarget->width += phinfo->ph_width; + + /* + * Charge the cost of evaluating the contained expression if + * the PHV can be computed here but not in either input. This + * is a bit bogus because we make the decision based on the + * first pair of possible input relations considered for the + * joinrel. With other pairs, it might be possible to compute + * the PHV in one input or the other, and then we'd be double + * charging the PHV's cost for some join paths. For now, live + * with that; but we might want to improve it later by + * refiguring the reltarget costs for each pair of inputs. + */ + if (!bms_is_subset(phinfo->ph_eval_at, outer_rel->relids) && + !bms_is_subset(phinfo->ph_eval_at, inner_rel->relids)) + { + QualCost cost; + + cost_qual_eval_node(&cost, (Node *) phinfo->ph_var->phexpr, + root); + joinrel->reltarget->cost.startup += cost.startup; + joinrel->reltarget->cost.per_tuple += cost.per_tuple; + } + } + + /* + * Also adjust joinrel's direct_lateral_relids to include the + * PHV's source rel(s). We must do this even if we're not + * actually going to emit the PHV, otherwise join_is_legal() will + * reject valid join orderings. (In principle maybe we could + * instead remove the joinrel's lateral_relids dependency; but + * that's complicated to get right, and cases where we're not + * going to emit the PHV are too rare to justify the work.) + * + * In principle we should only do this if the join doesn't yet + * include the PHV's source rel(s). But our caller + * build_join_rel() will clean things up by removing the join's + * own relids from its direct_lateral_relids, so we needn't + * account for that here. + */ + joinrel->direct_lateral_relids = + bms_add_members(joinrel->direct_lateral_relids, + phinfo->ph_lateral); + } + } +} diff --git a/src/backend/optimizer/util/plancat.c b/src/backend/optimizer/util/plancat.c new file mode 100644 index 0000000..419f2ac --- /dev/null +++ b/src/backend/optimizer/util/plancat.c @@ -0,0 +1,2509 @@ +/*------------------------------------------------------------------------- + * + * plancat.c + * routines for accessing the system catalogs + * + * + * Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group + * Portions Copyright (c) 1994, Regents of the University of California + * + * + * IDENTIFICATION + * src/backend/optimizer/util/plancat.c + * + *------------------------------------------------------------------------- + */ +#include "postgres.h" + +#include <math.h> + +#include "access/genam.h" +#include "access/htup_details.h" +#include "access/nbtree.h" +#include "access/sysattr.h" +#include "access/table.h" +#include "access/tableam.h" +#include "access/transam.h" +#include "access/xlog.h" +#include "catalog/catalog.h" +#include "catalog/heap.h" +#include "catalog/pg_am.h" +#include "catalog/pg_proc.h" +#include "catalog/pg_statistic_ext.h" +#include "catalog/pg_statistic_ext_data.h" +#include "foreign/fdwapi.h" +#include "miscadmin.h" +#include "nodes/makefuncs.h" +#include "nodes/nodeFuncs.h" +#include "nodes/supportnodes.h" +#include "optimizer/clauses.h" +#include "optimizer/cost.h" +#include "optimizer/optimizer.h" +#include "optimizer/plancat.h" +#include "optimizer/prep.h" +#include "parser/parse_relation.h" +#include "parser/parsetree.h" +#include "partitioning/partdesc.h" +#include "rewrite/rewriteManip.h" +#include "statistics/statistics.h" +#include "storage/bufmgr.h" +#include "utils/builtins.h" +#include "utils/lsyscache.h" +#include "utils/partcache.h" +#include "utils/rel.h" +#include "utils/snapmgr.h" +#include "utils/syscache.h" + +/* GUC parameter */ +int constraint_exclusion = CONSTRAINT_EXCLUSION_PARTITION; + +/* Hook for plugins to get control in get_relation_info() */ +get_relation_info_hook_type get_relation_info_hook = NULL; + + +static void get_relation_foreign_keys(PlannerInfo *root, RelOptInfo *rel, + Relation relation, bool inhparent); +static bool infer_collation_opclass_match(InferenceElem *elem, Relation idxRel, + List *idxExprs); +static List *get_relation_constraints(PlannerInfo *root, + Oid relationObjectId, RelOptInfo *rel, + bool include_noinherit, + bool include_notnull, + bool include_partition); +static List *build_index_tlist(PlannerInfo *root, IndexOptInfo *index, + Relation heapRelation); +static List *get_relation_statistics(RelOptInfo *rel, Relation relation); +static void set_relation_partition_info(PlannerInfo *root, RelOptInfo *rel, + Relation relation); +static PartitionScheme find_partition_scheme(PlannerInfo *root, Relation rel); +static void set_baserel_partition_key_exprs(Relation relation, + RelOptInfo *rel); +static void set_baserel_partition_constraint(Relation relation, + RelOptInfo *rel); + + +/* + * get_relation_info - + * Retrieves catalog information for a given relation. + * + * Given the Oid of the relation, return the following info into fields + * of the RelOptInfo struct: + * + * min_attr lowest valid AttrNumber + * max_attr highest valid AttrNumber + * indexlist list of IndexOptInfos for relation's indexes + * statlist list of StatisticExtInfo for relation's statistic objects + * serverid if it's a foreign table, the server OID + * fdwroutine if it's a foreign table, the FDW function pointers + * pages number of pages + * tuples number of tuples + * rel_parallel_workers user-defined number of parallel workers + * + * Also, add information about the relation's foreign keys to root->fkey_list. + * + * Also, initialize the attr_needed[] and attr_widths[] arrays. In most + * cases these are left as zeroes, but sometimes we need to compute attr + * widths here, and we may as well cache the results for costsize.c. + * + * If inhparent is true, all we need to do is set up the attr arrays: + * the RelOptInfo actually represents the appendrel formed by an inheritance + * tree, and so the parent rel's physical size and index information isn't + * important for it. + */ +void +get_relation_info(PlannerInfo *root, Oid relationObjectId, bool inhparent, + RelOptInfo *rel) +{ + Index varno = rel->relid; + Relation relation; + bool hasindex; + List *indexinfos = NIL; + + /* + * We need not lock the relation since it was already locked, either by + * the rewriter or when expand_inherited_rtentry() added it to the query's + * rangetable. + */ + relation = table_open(relationObjectId, NoLock); + + /* + * Relations without a table AM can be used in a query only if they are of + * special-cased relkinds. This check prevents us from crashing later if, + * for example, a view's ON SELECT rule has gone missing. Note that + * table_open() already rejected indexes and composite types; spell the + * error the same way it does. + */ + if (!relation->rd_tableam) + { + if (!(relation->rd_rel->relkind == RELKIND_FOREIGN_TABLE || + relation->rd_rel->relkind == RELKIND_PARTITIONED_TABLE)) + ereport(ERROR, + (errcode(ERRCODE_WRONG_OBJECT_TYPE), + errmsg("cannot open relation \"%s\"", + RelationGetRelationName(relation)), + errdetail_relkind_not_supported(relation->rd_rel->relkind))); + } + + /* Temporary and unlogged relations are inaccessible during recovery. */ + if (!RelationIsPermanent(relation) && RecoveryInProgress()) + ereport(ERROR, + (errcode(ERRCODE_FEATURE_NOT_SUPPORTED), + errmsg("cannot access temporary or unlogged relations during recovery"))); + + rel->min_attr = FirstLowInvalidHeapAttributeNumber + 1; + rel->max_attr = RelationGetNumberOfAttributes(relation); + rel->reltablespace = RelationGetForm(relation)->reltablespace; + + Assert(rel->max_attr >= rel->min_attr); + rel->attr_needed = (Relids *) + palloc0((rel->max_attr - rel->min_attr + 1) * sizeof(Relids)); + rel->attr_widths = (int32 *) + palloc0((rel->max_attr - rel->min_attr + 1) * sizeof(int32)); + + /* + * Estimate relation size --- unless it's an inheritance parent, in which + * case the size we want is not the rel's own size but the size of its + * inheritance tree. That will be computed in set_append_rel_size(). + */ + if (!inhparent) + estimate_rel_size(relation, rel->attr_widths - rel->min_attr, + &rel->pages, &rel->tuples, &rel->allvisfrac); + + /* Retrieve the parallel_workers reloption, or -1 if not set. */ + rel->rel_parallel_workers = RelationGetParallelWorkers(relation, -1); + + /* + * Make list of indexes. Ignore indexes on system catalogs if told to. + * Don't bother with indexes for an inheritance parent, either. + */ + if (inhparent || + (IgnoreSystemIndexes && IsSystemRelation(relation))) + hasindex = false; + else + hasindex = relation->rd_rel->relhasindex; + + if (hasindex) + { + List *indexoidlist; + LOCKMODE lmode; + ListCell *l; + + indexoidlist = RelationGetIndexList(relation); + + /* + * For each index, we get the same type of lock that the executor will + * need, and do not release it. This saves a couple of trips to the + * shared lock manager while not creating any real loss of + * concurrency, because no schema changes could be happening on the + * index while we hold lock on the parent rel, and no lock type used + * for queries blocks any other kind of index operation. + */ + lmode = root->simple_rte_array[varno]->rellockmode; + + foreach(l, indexoidlist) + { + Oid indexoid = lfirst_oid(l); + Relation indexRelation; + Form_pg_index index; + IndexAmRoutine *amroutine; + IndexOptInfo *info; + int ncolumns, + nkeycolumns; + int i; + + /* + * Extract info from the relation descriptor for the index. + */ + indexRelation = index_open(indexoid, lmode); + index = indexRelation->rd_index; + + /* + * Ignore invalid indexes, since they can't safely be used for + * queries. Note that this is OK because the data structure we + * are constructing is only used by the planner --- the executor + * still needs to insert into "invalid" indexes, if they're marked + * indisready. + */ + if (!index->indisvalid) + { + index_close(indexRelation, NoLock); + continue; + } + + /* + * Ignore partitioned indexes, since they are not usable for + * queries. + */ + if (indexRelation->rd_rel->relkind == RELKIND_PARTITIONED_INDEX) + { + index_close(indexRelation, NoLock); + continue; + } + + /* + * If the index is valid, but cannot yet be used, ignore it; but + * mark the plan we are generating as transient. See + * src/backend/access/heap/README.HOT for discussion. + */ + if (index->indcheckxmin && + !TransactionIdPrecedes(HeapTupleHeaderGetXmin(indexRelation->rd_indextuple->t_data), + TransactionXmin)) + { + root->glob->transientPlan = true; + index_close(indexRelation, NoLock); + continue; + } + + info = makeNode(IndexOptInfo); + + info->indexoid = index->indexrelid; + info->reltablespace = + RelationGetForm(indexRelation)->reltablespace; + info->rel = rel; + info->ncolumns = ncolumns = index->indnatts; + info->nkeycolumns = nkeycolumns = index->indnkeyatts; + + info->indexkeys = (int *) palloc(sizeof(int) * ncolumns); + info->indexcollations = (Oid *) palloc(sizeof(Oid) * nkeycolumns); + info->opfamily = (Oid *) palloc(sizeof(Oid) * nkeycolumns); + info->opcintype = (Oid *) palloc(sizeof(Oid) * nkeycolumns); + info->canreturn = (bool *) palloc(sizeof(bool) * ncolumns); + + for (i = 0; i < ncolumns; i++) + { + info->indexkeys[i] = index->indkey.values[i]; + info->canreturn[i] = index_can_return(indexRelation, i + 1); + } + + for (i = 0; i < nkeycolumns; i++) + { + info->opfamily[i] = indexRelation->rd_opfamily[i]; + info->opcintype[i] = indexRelation->rd_opcintype[i]; + info->indexcollations[i] = indexRelation->rd_indcollation[i]; + } + + info->relam = indexRelation->rd_rel->relam; + + /* We copy just the fields we need, not all of rd_indam */ + amroutine = indexRelation->rd_indam; + info->amcanorderbyop = amroutine->amcanorderbyop; + info->amoptionalkey = amroutine->amoptionalkey; + info->amsearcharray = amroutine->amsearcharray; + info->amsearchnulls = amroutine->amsearchnulls; + info->amcanparallel = amroutine->amcanparallel; + info->amhasgettuple = (amroutine->amgettuple != NULL); + info->amhasgetbitmap = amroutine->amgetbitmap != NULL && + relation->rd_tableam->scan_bitmap_next_block != NULL; + info->amcanmarkpos = (amroutine->ammarkpos != NULL && + amroutine->amrestrpos != NULL); + info->amcostestimate = amroutine->amcostestimate; + Assert(info->amcostestimate != NULL); + + /* Fetch index opclass options */ + info->opclassoptions = RelationGetIndexAttOptions(indexRelation, true); + + /* + * Fetch the ordering information for the index, if any. + */ + if (info->relam == BTREE_AM_OID) + { + /* + * If it's a btree index, we can use its opfamily OIDs + * directly as the sort ordering opfamily OIDs. + */ + Assert(amroutine->amcanorder); + + info->sortopfamily = info->opfamily; + info->reverse_sort = (bool *) palloc(sizeof(bool) * nkeycolumns); + info->nulls_first = (bool *) palloc(sizeof(bool) * nkeycolumns); + + for (i = 0; i < nkeycolumns; i++) + { + int16 opt = indexRelation->rd_indoption[i]; + + info->reverse_sort[i] = (opt & INDOPTION_DESC) != 0; + info->nulls_first[i] = (opt & INDOPTION_NULLS_FIRST) != 0; + } + } + else if (amroutine->amcanorder) + { + /* + * Otherwise, identify the corresponding btree opfamilies by + * trying to map this index's "<" operators into btree. Since + * "<" uniquely defines the behavior of a sort order, this is + * a sufficient test. + * + * XXX This method is rather slow and also requires the + * undesirable assumption that the other index AM numbers its + * strategies the same as btree. It'd be better to have a way + * to explicitly declare the corresponding btree opfamily for + * each opfamily of the other index type. But given the lack + * of current or foreseeable amcanorder index types, it's not + * worth expending more effort on now. + */ + info->sortopfamily = (Oid *) palloc(sizeof(Oid) * nkeycolumns); + info->reverse_sort = (bool *) palloc(sizeof(bool) * nkeycolumns); + info->nulls_first = (bool *) palloc(sizeof(bool) * nkeycolumns); + + for (i = 0; i < nkeycolumns; i++) + { + int16 opt = indexRelation->rd_indoption[i]; + Oid ltopr; + Oid btopfamily; + Oid btopcintype; + int16 btstrategy; + + info->reverse_sort[i] = (opt & INDOPTION_DESC) != 0; + info->nulls_first[i] = (opt & INDOPTION_NULLS_FIRST) != 0; + + ltopr = get_opfamily_member(info->opfamily[i], + info->opcintype[i], + info->opcintype[i], + BTLessStrategyNumber); + if (OidIsValid(ltopr) && + get_ordering_op_properties(ltopr, + &btopfamily, + &btopcintype, + &btstrategy) && + btopcintype == info->opcintype[i] && + btstrategy == BTLessStrategyNumber) + { + /* Successful mapping */ + info->sortopfamily[i] = btopfamily; + } + else + { + /* Fail ... quietly treat index as unordered */ + info->sortopfamily = NULL; + info->reverse_sort = NULL; + info->nulls_first = NULL; + break; + } + } + } + else + { + info->sortopfamily = NULL; + info->reverse_sort = NULL; + info->nulls_first = NULL; + } + + /* + * Fetch the index expressions and predicate, if any. We must + * modify the copies we obtain from the relcache to have the + * correct varno for the parent relation, so that they match up + * correctly against qual clauses. + */ + info->indexprs = RelationGetIndexExpressions(indexRelation); + info->indpred = RelationGetIndexPredicate(indexRelation); + if (info->indexprs && varno != 1) + ChangeVarNodes((Node *) info->indexprs, 1, varno, 0); + if (info->indpred && varno != 1) + ChangeVarNodes((Node *) info->indpred, 1, varno, 0); + + /* Build targetlist using the completed indexprs data */ + info->indextlist = build_index_tlist(root, info, relation); + + info->indrestrictinfo = NIL; /* set later, in indxpath.c */ + info->predOK = false; /* set later, in indxpath.c */ + info->unique = index->indisunique; + info->immediate = index->indimmediate; + info->hypothetical = false; + + /* + * Estimate the index size. If it's not a partial index, we lock + * the number-of-tuples estimate to equal the parent table; if it + * is partial then we have to use the same methods as we would for + * a table, except we can be sure that the index is not larger + * than the table. + */ + if (info->indpred == NIL) + { + info->pages = RelationGetNumberOfBlocks(indexRelation); + info->tuples = rel->tuples; + } + else + { + double allvisfrac; /* dummy */ + + estimate_rel_size(indexRelation, NULL, + &info->pages, &info->tuples, &allvisfrac); + if (info->tuples > rel->tuples) + info->tuples = rel->tuples; + } + + if (info->relam == BTREE_AM_OID) + { + /* For btrees, get tree height while we have the index open */ + info->tree_height = _bt_getrootheight(indexRelation); + } + else + { + /* For other index types, just set it to "unknown" for now */ + info->tree_height = -1; + } + + index_close(indexRelation, NoLock); + + /* + * We've historically used lcons() here. It'd make more sense to + * use lappend(), but that causes the planner to change behavior + * in cases where two indexes seem equally attractive. For now, + * stick with lcons() --- few tables should have so many indexes + * that the O(N^2) behavior of lcons() is really a problem. + */ + indexinfos = lcons(info, indexinfos); + } + + list_free(indexoidlist); + } + + rel->indexlist = indexinfos; + + rel->statlist = get_relation_statistics(rel, relation); + + /* Grab foreign-table info using the relcache, while we have it */ + if (relation->rd_rel->relkind == RELKIND_FOREIGN_TABLE) + { + rel->serverid = GetForeignServerIdByRelId(RelationGetRelid(relation)); + rel->fdwroutine = GetFdwRoutineForRelation(relation, true); + } + else + { + rel->serverid = InvalidOid; + rel->fdwroutine = NULL; + } + + /* Collect info about relation's foreign keys, if relevant */ + get_relation_foreign_keys(root, rel, relation, inhparent); + + /* Collect info about functions implemented by the rel's table AM. */ + if (relation->rd_tableam && + relation->rd_tableam->scan_set_tidrange != NULL && + relation->rd_tableam->scan_getnextslot_tidrange != NULL) + rel->amflags |= AMFLAG_HAS_TID_RANGE; + + /* + * Collect info about relation's partitioning scheme, if any. Only + * inheritance parents may be partitioned. + */ + if (inhparent && relation->rd_rel->relkind == RELKIND_PARTITIONED_TABLE) + set_relation_partition_info(root, rel, relation); + + table_close(relation, NoLock); + + /* + * Allow a plugin to editorialize on the info we obtained from the + * catalogs. Actions might include altering the assumed relation size, + * removing an index, or adding a hypothetical index to the indexlist. + */ + if (get_relation_info_hook) + (*get_relation_info_hook) (root, relationObjectId, inhparent, rel); +} + +/* + * get_relation_foreign_keys - + * Retrieves foreign key information for a given relation. + * + * ForeignKeyOptInfos for relevant foreign keys are created and added to + * root->fkey_list. We do this now while we have the relcache entry open. + * We could sometimes avoid making useless ForeignKeyOptInfos if we waited + * until all RelOptInfos have been built, but the cost of re-opening the + * relcache entries would probably exceed any savings. + */ +static void +get_relation_foreign_keys(PlannerInfo *root, RelOptInfo *rel, + Relation relation, bool inhparent) +{ + List *rtable = root->parse->rtable; + List *cachedfkeys; + ListCell *lc; + + /* + * If it's not a baserel, we don't care about its FKs. Also, if the query + * references only a single relation, we can skip the lookup since no FKs + * could satisfy the requirements below. + */ + if (rel->reloptkind != RELOPT_BASEREL || + list_length(rtable) < 2) + return; + + /* + * If it's the parent of an inheritance tree, ignore its FKs. We could + * make useful FK-based deductions if we found that all members of the + * inheritance tree have equivalent FK constraints, but detecting that + * would require code that hasn't been written. + */ + if (inhparent) + return; + + /* + * Extract data about relation's FKs from the relcache. Note that this + * list belongs to the relcache and might disappear in a cache flush, so + * we must not do any further catalog access within this function. + */ + cachedfkeys = RelationGetFKeyList(relation); + + /* + * Figure out which FKs are of interest for this query, and create + * ForeignKeyOptInfos for them. We want only FKs that reference some + * other RTE of the current query. In queries containing self-joins, + * there might be more than one other RTE for a referenced table, and we + * should make a ForeignKeyOptInfo for each occurrence. + * + * Ideally, we would ignore RTEs that correspond to non-baserels, but it's + * too hard to identify those here, so we might end up making some useless + * ForeignKeyOptInfos. If so, match_foreign_keys_to_quals() will remove + * them again. + */ + foreach(lc, cachedfkeys) + { + ForeignKeyCacheInfo *cachedfk = (ForeignKeyCacheInfo *) lfirst(lc); + Index rti; + ListCell *lc2; + + /* conrelid should always be that of the table we're considering */ + Assert(cachedfk->conrelid == RelationGetRelid(relation)); + + /* Scan to find other RTEs matching confrelid */ + rti = 0; + foreach(lc2, rtable) + { + RangeTblEntry *rte = (RangeTblEntry *) lfirst(lc2); + ForeignKeyOptInfo *info; + + rti++; + /* Ignore if not the correct table */ + if (rte->rtekind != RTE_RELATION || + rte->relid != cachedfk->confrelid) + continue; + /* Ignore if it's an inheritance parent; doesn't really match */ + if (rte->inh) + continue; + /* Ignore self-referential FKs; we only care about joins */ + if (rti == rel->relid) + continue; + + /* OK, let's make an entry */ + info = makeNode(ForeignKeyOptInfo); + info->con_relid = rel->relid; + info->ref_relid = rti; + info->nkeys = cachedfk->nkeys; + memcpy(info->conkey, cachedfk->conkey, sizeof(info->conkey)); + memcpy(info->confkey, cachedfk->confkey, sizeof(info->confkey)); + memcpy(info->conpfeqop, cachedfk->conpfeqop, sizeof(info->conpfeqop)); + /* zero out fields to be filled by match_foreign_keys_to_quals */ + info->nmatched_ec = 0; + info->nconst_ec = 0; + info->nmatched_rcols = 0; + info->nmatched_ri = 0; + memset(info->eclass, 0, sizeof(info->eclass)); + memset(info->fk_eclass_member, 0, sizeof(info->fk_eclass_member)); + memset(info->rinfos, 0, sizeof(info->rinfos)); + + root->fkey_list = lappend(root->fkey_list, info); + } + } +} + +/* + * infer_arbiter_indexes - + * Determine the unique indexes used to arbitrate speculative insertion. + * + * Uses user-supplied inference clause expressions and predicate to match a + * unique index from those defined and ready on the heap relation (target). + * An exact match is required on columns/expressions (although they can appear + * in any order). However, the predicate given by the user need only restrict + * insertion to a subset of some part of the table covered by some particular + * unique index (in particular, a partial unique index) in order to be + * inferred. + * + * The implementation does not consider which B-Tree operator class any + * particular available unique index attribute uses, unless one was specified + * in the inference specification. The same is true of collations. In + * particular, there is no system dependency on the default operator class for + * the purposes of inference. If no opclass (or collation) is specified, then + * all matching indexes (that may or may not match the default in terms of + * each attribute opclass/collation) are used for inference. + */ +List * +infer_arbiter_indexes(PlannerInfo *root) +{ + OnConflictExpr *onconflict = root->parse->onConflict; + + /* Iteration state */ + RangeTblEntry *rte; + Relation relation; + Oid indexOidFromConstraint = InvalidOid; + List *indexList; + ListCell *l; + + /* Normalized inference attributes and inference expressions: */ + Bitmapset *inferAttrs = NULL; + List *inferElems = NIL; + + /* Results */ + List *results = NIL; + + /* + * Quickly return NIL for ON CONFLICT DO NOTHING without an inference + * specification or named constraint. ON CONFLICT DO UPDATE statements + * must always provide one or the other (but parser ought to have caught + * that already). + */ + if (onconflict->arbiterElems == NIL && + onconflict->constraint == InvalidOid) + return NIL; + + /* + * We need not lock the relation since it was already locked, either by + * the rewriter or when expand_inherited_rtentry() added it to the query's + * rangetable. + */ + rte = rt_fetch(root->parse->resultRelation, root->parse->rtable); + + relation = table_open(rte->relid, NoLock); + + /* + * Build normalized/BMS representation of plain indexed attributes, as + * well as a separate list of expression items. This simplifies matching + * the cataloged definition of indexes. + */ + foreach(l, onconflict->arbiterElems) + { + InferenceElem *elem = (InferenceElem *) lfirst(l); + Var *var; + int attno; + + if (!IsA(elem->expr, Var)) + { + /* If not a plain Var, just shove it in inferElems for now */ + inferElems = lappend(inferElems, elem->expr); + continue; + } + + var = (Var *) elem->expr; + attno = var->varattno; + + if (attno == 0) + ereport(ERROR, + (errcode(ERRCODE_FEATURE_NOT_SUPPORTED), + errmsg("whole row unique index inference specifications are not supported"))); + + inferAttrs = bms_add_member(inferAttrs, + attno - FirstLowInvalidHeapAttributeNumber); + } + + /* + * Lookup named constraint's index. This is not immediately returned + * because some additional sanity checks are required. + */ + if (onconflict->constraint != InvalidOid) + { + indexOidFromConstraint = get_constraint_index(onconflict->constraint); + + if (indexOidFromConstraint == InvalidOid) + ereport(ERROR, + (errcode(ERRCODE_WRONG_OBJECT_TYPE), + errmsg("constraint in ON CONFLICT clause has no associated index"))); + } + + /* + * Using that representation, iterate through the list of indexes on the + * target relation to try and find a match + */ + indexList = RelationGetIndexList(relation); + + foreach(l, indexList) + { + Oid indexoid = lfirst_oid(l); + Relation idxRel; + Form_pg_index idxForm; + Bitmapset *indexedAttrs; + List *idxExprs; + List *predExprs; + AttrNumber natt; + ListCell *el; + + /* + * Extract info from the relation descriptor for the index. Obtain + * the same lock type that the executor will ultimately use. + * + * Let executor complain about !indimmediate case directly, because + * enforcement needs to occur there anyway when an inference clause is + * omitted. + */ + idxRel = index_open(indexoid, rte->rellockmode); + idxForm = idxRel->rd_index; + + if (!idxForm->indisvalid) + goto next; + + /* + * Note that we do not perform a check against indcheckxmin (like e.g. + * get_relation_info()) here to eliminate candidates, because + * uniqueness checking only cares about the most recently committed + * tuple versions. + */ + + /* + * Look for match on "ON constraint_name" variant, which may not be + * unique constraint. This can only be a constraint name. + */ + if (indexOidFromConstraint == idxForm->indexrelid) + { + if (!idxForm->indisunique && onconflict->action == ONCONFLICT_UPDATE) + ereport(ERROR, + (errcode(ERRCODE_WRONG_OBJECT_TYPE), + errmsg("ON CONFLICT DO UPDATE not supported with exclusion constraints"))); + + results = lappend_oid(results, idxForm->indexrelid); + list_free(indexList); + index_close(idxRel, NoLock); + table_close(relation, NoLock); + return results; + } + else if (indexOidFromConstraint != InvalidOid) + { + /* No point in further work for index in named constraint case */ + goto next; + } + + /* + * Only considering conventional inference at this point (not named + * constraints), so index under consideration can be immediately + * skipped if it's not unique + */ + if (!idxForm->indisunique) + goto next; + + /* Build BMS representation of plain (non expression) index attrs */ + indexedAttrs = NULL; + for (natt = 0; natt < idxForm->indnkeyatts; natt++) + { + int attno = idxRel->rd_index->indkey.values[natt]; + + if (attno != 0) + indexedAttrs = bms_add_member(indexedAttrs, + attno - FirstLowInvalidHeapAttributeNumber); + } + + /* Non-expression attributes (if any) must match */ + if (!bms_equal(indexedAttrs, inferAttrs)) + goto next; + + /* Expression attributes (if any) must match */ + idxExprs = RelationGetIndexExpressions(idxRel); + foreach(el, onconflict->arbiterElems) + { + InferenceElem *elem = (InferenceElem *) lfirst(el); + + /* + * Ensure that collation/opclass aspects of inference expression + * element match. Even though this loop is primarily concerned + * with matching expressions, it is a convenient point to check + * this for both expressions and ordinary (non-expression) + * attributes appearing as inference elements. + */ + if (!infer_collation_opclass_match(elem, idxRel, idxExprs)) + goto next; + + /* + * Plain Vars don't factor into count of expression elements, and + * the question of whether or not they satisfy the index + * definition has already been considered (they must). + */ + if (IsA(elem->expr, Var)) + continue; + + /* + * Might as well avoid redundant check in the rare cases where + * infer_collation_opclass_match() is required to do real work. + * Otherwise, check that element expression appears in cataloged + * index definition. + */ + if (elem->infercollid != InvalidOid || + elem->inferopclass != InvalidOid || + list_member(idxExprs, elem->expr)) + continue; + + goto next; + } + + /* + * Now that all inference elements were matched, ensure that the + * expression elements from inference clause are not missing any + * cataloged expressions. This does the right thing when unique + * indexes redundantly repeat the same attribute, or if attributes + * redundantly appear multiple times within an inference clause. + */ + if (list_difference(idxExprs, inferElems) != NIL) + goto next; + + /* + * If it's a partial index, its predicate must be implied by the ON + * CONFLICT's WHERE clause. + */ + predExprs = RelationGetIndexPredicate(idxRel); + + if (!predicate_implied_by(predExprs, (List *) onconflict->arbiterWhere, false)) + goto next; + + results = lappend_oid(results, idxForm->indexrelid); +next: + index_close(idxRel, NoLock); + } + + list_free(indexList); + table_close(relation, NoLock); + + if (results == NIL) + ereport(ERROR, + (errcode(ERRCODE_INVALID_COLUMN_REFERENCE), + errmsg("there is no unique or exclusion constraint matching the ON CONFLICT specification"))); + + return results; +} + +/* + * infer_collation_opclass_match - ensure infer element opclass/collation match + * + * Given unique index inference element from inference specification, if + * collation was specified, or if opclass was specified, verify that there is + * at least one matching indexed attribute (occasionally, there may be more). + * Skip this in the common case where inference specification does not include + * collation or opclass (instead matching everything, regardless of cataloged + * collation/opclass of indexed attribute). + * + * At least historically, Postgres has not offered collations or opclasses + * with alternative-to-default notions of equality, so these additional + * criteria should only be required infrequently. + * + * Don't give up immediately when an inference element matches some attribute + * cataloged as indexed but not matching additional opclass/collation + * criteria. This is done so that the implementation is as forgiving as + * possible of redundancy within cataloged index attributes (or, less + * usefully, within inference specification elements). If collations actually + * differ between apparently redundantly indexed attributes (redundant within + * or across indexes), then there really is no redundancy as such. + * + * Note that if an inference element specifies an opclass and a collation at + * once, both must match in at least one particular attribute within index + * catalog definition in order for that inference element to be considered + * inferred/satisfied. + */ +static bool +infer_collation_opclass_match(InferenceElem *elem, Relation idxRel, + List *idxExprs) +{ + AttrNumber natt; + Oid inferopfamily = InvalidOid; /* OID of opclass opfamily */ + Oid inferopcinputtype = InvalidOid; /* OID of opclass input type */ + int nplain = 0; /* # plain attrs observed */ + + /* + * If inference specification element lacks collation/opclass, then no + * need to check for exact match. + */ + if (elem->infercollid == InvalidOid && elem->inferopclass == InvalidOid) + return true; + + /* + * Lookup opfamily and input type, for matching indexes + */ + if (elem->inferopclass) + { + inferopfamily = get_opclass_family(elem->inferopclass); + inferopcinputtype = get_opclass_input_type(elem->inferopclass); + } + + for (natt = 1; natt <= idxRel->rd_att->natts; natt++) + { + Oid opfamily = idxRel->rd_opfamily[natt - 1]; + Oid opcinputtype = idxRel->rd_opcintype[natt - 1]; + Oid collation = idxRel->rd_indcollation[natt - 1]; + int attno = idxRel->rd_index->indkey.values[natt - 1]; + + if (attno != 0) + nplain++; + + if (elem->inferopclass != InvalidOid && + (inferopfamily != opfamily || inferopcinputtype != opcinputtype)) + { + /* Attribute needed to match opclass, but didn't */ + continue; + } + + if (elem->infercollid != InvalidOid && + elem->infercollid != collation) + { + /* Attribute needed to match collation, but didn't */ + continue; + } + + /* If one matching index att found, good enough -- return true */ + if (IsA(elem->expr, Var)) + { + if (((Var *) elem->expr)->varattno == attno) + return true; + } + else if (attno == 0) + { + Node *nattExpr = list_nth(idxExprs, (natt - 1) - nplain); + + /* + * Note that unlike routines like match_index_to_operand() we + * don't need to care about RelabelType. Neither the index + * definition nor the inference clause should contain them. + */ + if (equal(elem->expr, nattExpr)) + return true; + } + } + + return false; +} + +/* + * estimate_rel_size - estimate # pages and # tuples in a table or index + * + * We also estimate the fraction of the pages that are marked all-visible in + * the visibility map, for use in estimation of index-only scans. + * + * If attr_widths isn't NULL, it points to the zero-index entry of the + * relation's attr_widths[] cache; we fill this in if we have need to compute + * the attribute widths for estimation purposes. + */ +void +estimate_rel_size(Relation rel, int32 *attr_widths, + BlockNumber *pages, double *tuples, double *allvisfrac) +{ + BlockNumber curpages; + BlockNumber relpages; + double reltuples; + BlockNumber relallvisible; + double density; + + if (RELKIND_HAS_TABLE_AM(rel->rd_rel->relkind)) + { + table_relation_estimate_size(rel, attr_widths, pages, tuples, + allvisfrac); + } + else if (rel->rd_rel->relkind == RELKIND_INDEX) + { + /* + * XXX: It'd probably be good to move this into a callback, individual + * index types e.g. know if they have a metapage. + */ + + /* it has storage, ok to call the smgr */ + curpages = RelationGetNumberOfBlocks(rel); + + /* report estimated # pages */ + *pages = curpages; + /* quick exit if rel is clearly empty */ + if (curpages == 0) + { + *tuples = 0; + *allvisfrac = 0; + return; + } + + /* coerce values in pg_class to more desirable types */ + relpages = (BlockNumber) rel->rd_rel->relpages; + reltuples = (double) rel->rd_rel->reltuples; + relallvisible = (BlockNumber) rel->rd_rel->relallvisible; + + /* + * Discount the metapage while estimating the number of tuples. This + * is a kluge because it assumes more than it ought to about index + * structure. Currently it's OK for btree, hash, and GIN indexes but + * suspect for GiST indexes. + */ + if (relpages > 0) + { + curpages--; + relpages--; + } + + /* estimate number of tuples from previous tuple density */ + if (reltuples >= 0 && relpages > 0) + density = reltuples / (double) relpages; + else + { + /* + * If we have no data because the relation was never vacuumed, + * estimate tuple width from attribute datatypes. We assume here + * that the pages are completely full, which is OK for tables + * (since they've presumably not been VACUUMed yet) but is + * probably an overestimate for indexes. Fortunately + * get_relation_info() can clamp the overestimate to the parent + * table's size. + * + * Note: this code intentionally disregards alignment + * considerations, because (a) that would be gilding the lily + * considering how crude the estimate is, and (b) it creates + * platform dependencies in the default plans which are kind of a + * headache for regression testing. + * + * XXX: Should this logic be more index specific? + */ + int32 tuple_width; + + tuple_width = get_rel_data_width(rel, attr_widths); + tuple_width += MAXALIGN(SizeofHeapTupleHeader); + tuple_width += sizeof(ItemIdData); + /* note: integer division is intentional here */ + density = (BLCKSZ - SizeOfPageHeaderData) / tuple_width; + } + *tuples = rint(density * (double) curpages); + + /* + * We use relallvisible as-is, rather than scaling it up like we do + * for the pages and tuples counts, on the theory that any pages added + * since the last VACUUM are most likely not marked all-visible. But + * costsize.c wants it converted to a fraction. + */ + if (relallvisible == 0 || curpages <= 0) + *allvisfrac = 0; + else if ((double) relallvisible >= curpages) + *allvisfrac = 1; + else + *allvisfrac = (double) relallvisible / curpages; + } + else + { + /* + * Just use whatever's in pg_class. This covers foreign tables, + * sequences, and also relkinds without storage (shouldn't get here?); + * see initializations in AddNewRelationTuple(). Note that FDW must + * cope if reltuples is -1! + */ + *pages = rel->rd_rel->relpages; + *tuples = rel->rd_rel->reltuples; + *allvisfrac = 0; + } +} + + +/* + * get_rel_data_width + * + * Estimate the average width of (the data part of) the relation's tuples. + * + * If attr_widths isn't NULL, it points to the zero-index entry of the + * relation's attr_widths[] cache; use and update that cache as appropriate. + * + * Currently we ignore dropped columns. Ideally those should be included + * in the result, but we haven't got any way to get info about them; and + * since they might be mostly NULLs, treating them as zero-width is not + * necessarily the wrong thing anyway. + */ +int32 +get_rel_data_width(Relation rel, int32 *attr_widths) +{ + int32 tuple_width = 0; + int i; + + for (i = 1; i <= RelationGetNumberOfAttributes(rel); i++) + { + Form_pg_attribute att = TupleDescAttr(rel->rd_att, i - 1); + int32 item_width; + + if (att->attisdropped) + continue; + + /* use previously cached data, if any */ + if (attr_widths != NULL && attr_widths[i] > 0) + { + tuple_width += attr_widths[i]; + continue; + } + + /* This should match set_rel_width() in costsize.c */ + item_width = get_attavgwidth(RelationGetRelid(rel), i); + if (item_width <= 0) + { + item_width = get_typavgwidth(att->atttypid, att->atttypmod); + Assert(item_width > 0); + } + if (attr_widths != NULL) + attr_widths[i] = item_width; + tuple_width += item_width; + } + + return tuple_width; +} + +/* + * get_relation_data_width + * + * External API for get_rel_data_width: same behavior except we have to + * open the relcache entry. + */ +int32 +get_relation_data_width(Oid relid, int32 *attr_widths) +{ + int32 result; + Relation relation; + + /* As above, assume relation is already locked */ + relation = table_open(relid, NoLock); + + result = get_rel_data_width(relation, attr_widths); + + table_close(relation, NoLock); + + return result; +} + + +/* + * get_relation_constraints + * + * Retrieve the applicable constraint expressions of the given relation. + * + * Returns a List (possibly empty) of constraint expressions. Each one + * has been canonicalized, and its Vars are changed to have the varno + * indicated by rel->relid. This allows the expressions to be easily + * compared to expressions taken from WHERE. + * + * If include_noinherit is true, it's okay to include constraints that + * are marked NO INHERIT. + * + * If include_notnull is true, "col IS NOT NULL" expressions are generated + * and added to the result for each column that's marked attnotnull. + * + * If include_partition is true, and the relation is a partition, + * also include the partitioning constraints. + * + * Note: at present this is invoked at most once per relation per planner + * run, and in many cases it won't be invoked at all, so there seems no + * point in caching the data in RelOptInfo. + */ +static List * +get_relation_constraints(PlannerInfo *root, + Oid relationObjectId, RelOptInfo *rel, + bool include_noinherit, + bool include_notnull, + bool include_partition) +{ + List *result = NIL; + Index varno = rel->relid; + Relation relation; + TupleConstr *constr; + + /* + * We assume the relation has already been safely locked. + */ + relation = table_open(relationObjectId, NoLock); + + constr = relation->rd_att->constr; + if (constr != NULL) + { + int num_check = constr->num_check; + int i; + + for (i = 0; i < num_check; i++) + { + Node *cexpr; + + /* + * If this constraint hasn't been fully validated yet, we must + * ignore it here. Also ignore if NO INHERIT and we weren't told + * that that's safe. + */ + if (!constr->check[i].ccvalid) + continue; + if (constr->check[i].ccnoinherit && !include_noinherit) + continue; + + cexpr = stringToNode(constr->check[i].ccbin); + + /* + * Run each expression through const-simplification and + * canonicalization. This is not just an optimization, but is + * necessary, because we will be comparing it to + * similarly-processed qual clauses, and may fail to detect valid + * matches without this. This must match the processing done to + * qual clauses in preprocess_expression()! (We can skip the + * stuff involving subqueries, however, since we don't allow any + * in check constraints.) + */ + cexpr = eval_const_expressions(root, cexpr); + + cexpr = (Node *) canonicalize_qual((Expr *) cexpr, true); + + /* Fix Vars to have the desired varno */ + if (varno != 1) + ChangeVarNodes(cexpr, 1, varno, 0); + + /* + * Finally, convert to implicit-AND format (that is, a List) and + * append the resulting item(s) to our output list. + */ + result = list_concat(result, + make_ands_implicit((Expr *) cexpr)); + } + + /* Add NOT NULL constraints in expression form, if requested */ + if (include_notnull && constr->has_not_null) + { + int natts = relation->rd_att->natts; + + for (i = 1; i <= natts; i++) + { + Form_pg_attribute att = TupleDescAttr(relation->rd_att, i - 1); + + if (att->attnotnull && !att->attisdropped) + { + NullTest *ntest = makeNode(NullTest); + + ntest->arg = (Expr *) makeVar(varno, + i, + att->atttypid, + att->atttypmod, + att->attcollation, + 0); + ntest->nulltesttype = IS_NOT_NULL; + + /* + * argisrow=false is correct even for a composite column, + * because attnotnull does not represent a SQL-spec IS NOT + * NULL test in such a case, just IS DISTINCT FROM NULL. + */ + ntest->argisrow = false; + ntest->location = -1; + result = lappend(result, ntest); + } + } + } + } + + /* + * Add partitioning constraints, if requested. + */ + if (include_partition && relation->rd_rel->relispartition) + { + /* make sure rel->partition_qual is set */ + set_baserel_partition_constraint(relation, rel); + result = list_concat(result, rel->partition_qual); + } + + table_close(relation, NoLock); + + return result; +} + +/* + * Try loading data for the statistics object. + * + * We don't know if the data (specified by statOid and inh value) exist. + * The result is stored in stainfos list. + */ +static void +get_relation_statistics_worker(List **stainfos, RelOptInfo *rel, + Oid statOid, bool inh, + Bitmapset *keys, List *exprs) +{ + Form_pg_statistic_ext_data dataForm; + HeapTuple dtup; + + dtup = SearchSysCache2(STATEXTDATASTXOID, + ObjectIdGetDatum(statOid), BoolGetDatum(inh)); + if (!HeapTupleIsValid(dtup)) + return; + + dataForm = (Form_pg_statistic_ext_data) GETSTRUCT(dtup); + + /* add one StatisticExtInfo for each kind built */ + if (statext_is_kind_built(dtup, STATS_EXT_NDISTINCT)) + { + StatisticExtInfo *info = makeNode(StatisticExtInfo); + + info->statOid = statOid; + info->inherit = dataForm->stxdinherit; + info->rel = rel; + info->kind = STATS_EXT_NDISTINCT; + info->keys = bms_copy(keys); + info->exprs = exprs; + + *stainfos = lappend(*stainfos, info); + } + + if (statext_is_kind_built(dtup, STATS_EXT_DEPENDENCIES)) + { + StatisticExtInfo *info = makeNode(StatisticExtInfo); + + info->statOid = statOid; + info->inherit = dataForm->stxdinherit; + info->rel = rel; + info->kind = STATS_EXT_DEPENDENCIES; + info->keys = bms_copy(keys); + info->exprs = exprs; + + *stainfos = lappend(*stainfos, info); + } + + if (statext_is_kind_built(dtup, STATS_EXT_MCV)) + { + StatisticExtInfo *info = makeNode(StatisticExtInfo); + + info->statOid = statOid; + info->inherit = dataForm->stxdinherit; + info->rel = rel; + info->kind = STATS_EXT_MCV; + info->keys = bms_copy(keys); + info->exprs = exprs; + + *stainfos = lappend(*stainfos, info); + } + + if (statext_is_kind_built(dtup, STATS_EXT_EXPRESSIONS)) + { + StatisticExtInfo *info = makeNode(StatisticExtInfo); + + info->statOid = statOid; + info->inherit = dataForm->stxdinherit; + info->rel = rel; + info->kind = STATS_EXT_EXPRESSIONS; + info->keys = bms_copy(keys); + info->exprs = exprs; + + *stainfos = lappend(*stainfos, info); + } + + ReleaseSysCache(dtup); +} + +/* + * get_relation_statistics + * Retrieve extended statistics defined on the table. + * + * Returns a List (possibly empty) of StatisticExtInfo objects describing + * the statistics. Note that this doesn't load the actual statistics data, + * just the identifying metadata. Only stats actually built are considered. + */ +static List * +get_relation_statistics(RelOptInfo *rel, Relation relation) +{ + Index varno = rel->relid; + List *statoidlist; + List *stainfos = NIL; + ListCell *l; + + statoidlist = RelationGetStatExtList(relation); + + foreach(l, statoidlist) + { + Oid statOid = lfirst_oid(l); + Form_pg_statistic_ext staForm; + HeapTuple htup; + Bitmapset *keys = NULL; + List *exprs = NIL; + int i; + + htup = SearchSysCache1(STATEXTOID, ObjectIdGetDatum(statOid)); + if (!HeapTupleIsValid(htup)) + elog(ERROR, "cache lookup failed for statistics object %u", statOid); + staForm = (Form_pg_statistic_ext) GETSTRUCT(htup); + + /* + * First, build the array of columns covered. This is ultimately + * wasted if no stats within the object have actually been built, but + * it doesn't seem worth troubling over that case. + */ + for (i = 0; i < staForm->stxkeys.dim1; i++) + keys = bms_add_member(keys, staForm->stxkeys.values[i]); + + /* + * Preprocess expressions (if any). We read the expressions, run them + * through eval_const_expressions, and fix the varnos. + * + * XXX We don't know yet if there are any data for this stats object, + * with either stxdinherit value. But it's reasonable to assume there + * is at least one of those, possibly both. So it's better to process + * keys and expressions here. + */ + { + bool isnull; + Datum datum; + + /* decode expression (if any) */ + datum = SysCacheGetAttr(STATEXTOID, htup, + Anum_pg_statistic_ext_stxexprs, &isnull); + + if (!isnull) + { + char *exprsString; + + exprsString = TextDatumGetCString(datum); + exprs = (List *) stringToNode(exprsString); + pfree(exprsString); + + /* + * Run the expressions through eval_const_expressions. This is + * not just an optimization, but is necessary, because the + * planner will be comparing them to similarly-processed qual + * clauses, and may fail to detect valid matches without this. + * We must not use canonicalize_qual, however, since these + * aren't qual expressions. + */ + exprs = (List *) eval_const_expressions(NULL, (Node *) exprs); + + /* May as well fix opfuncids too */ + fix_opfuncids((Node *) exprs); + + /* + * Modify the copies we obtain from the relcache to have the + * correct varno for the parent relation, so that they match + * up correctly against qual clauses. + */ + if (varno != 1) + ChangeVarNodes((Node *) exprs, 1, varno, 0); + } + } + + /* extract statistics for possible values of stxdinherit flag */ + + get_relation_statistics_worker(&stainfos, rel, statOid, true, keys, exprs); + + get_relation_statistics_worker(&stainfos, rel, statOid, false, keys, exprs); + + ReleaseSysCache(htup); + bms_free(keys); + } + + list_free(statoidlist); + + return stainfos; +} + +/* + * relation_excluded_by_constraints + * + * Detect whether the relation need not be scanned because it has either + * self-inconsistent restrictions, or restrictions inconsistent with the + * relation's applicable constraints. + * + * Note: this examines only rel->relid, rel->reloptkind, and + * rel->baserestrictinfo; therefore it can be called before filling in + * other fields of the RelOptInfo. + */ +bool +relation_excluded_by_constraints(PlannerInfo *root, + RelOptInfo *rel, RangeTblEntry *rte) +{ + bool include_noinherit; + bool include_notnull; + bool include_partition = false; + List *safe_restrictions; + List *constraint_pred; + List *safe_constraints; + ListCell *lc; + + /* As of now, constraint exclusion works only with simple relations. */ + Assert(IS_SIMPLE_REL(rel)); + + /* + * If there are no base restriction clauses, we have no hope of proving + * anything below, so fall out quickly. + */ + if (rel->baserestrictinfo == NIL) + return false; + + /* + * Regardless of the setting of constraint_exclusion, detect + * constant-FALSE-or-NULL restriction clauses. Because const-folding will + * reduce "anything AND FALSE" to just "FALSE", any such case should + * result in exactly one baserestrictinfo entry. This doesn't fire very + * often, but it seems cheap enough to be worth doing anyway. (Without + * this, we'd miss some optimizations that 9.5 and earlier found via much + * more roundabout methods.) + */ + if (list_length(rel->baserestrictinfo) == 1) + { + RestrictInfo *rinfo = (RestrictInfo *) linitial(rel->baserestrictinfo); + Expr *clause = rinfo->clause; + + if (clause && IsA(clause, Const) && + (((Const *) clause)->constisnull || + !DatumGetBool(((Const *) clause)->constvalue))) + return true; + } + + /* + * Skip further tests, depending on constraint_exclusion. + */ + switch (constraint_exclusion) + { + case CONSTRAINT_EXCLUSION_OFF: + /* In 'off' mode, never make any further tests */ + return false; + + case CONSTRAINT_EXCLUSION_PARTITION: + + /* + * When constraint_exclusion is set to 'partition' we only handle + * appendrel members. Partition pruning has already been applied, + * so there is no need to consider the rel's partition constraints + * here. + */ + if (rel->reloptkind == RELOPT_OTHER_MEMBER_REL) + break; /* appendrel member, so process it */ + return false; + + case CONSTRAINT_EXCLUSION_ON: + + /* + * In 'on' mode, always apply constraint exclusion. If we are + * considering a baserel that is a partition (i.e., it was + * directly named rather than expanded from a parent table), then + * its partition constraints haven't been considered yet, so + * include them in the processing here. + */ + if (rel->reloptkind == RELOPT_BASEREL) + include_partition = true; + break; /* always try to exclude */ + } + + /* + * Check for self-contradictory restriction clauses. We dare not make + * deductions with non-immutable functions, but any immutable clauses that + * are self-contradictory allow us to conclude the scan is unnecessary. + * + * Note: strip off RestrictInfo because predicate_refuted_by() isn't + * expecting to see any in its predicate argument. + */ + safe_restrictions = NIL; + foreach(lc, rel->baserestrictinfo) + { + RestrictInfo *rinfo = (RestrictInfo *) lfirst(lc); + + if (!contain_mutable_functions((Node *) rinfo->clause)) + safe_restrictions = lappend(safe_restrictions, rinfo->clause); + } + + /* + * We can use weak refutation here, since we're comparing restriction + * clauses with restriction clauses. + */ + if (predicate_refuted_by(safe_restrictions, safe_restrictions, true)) + return true; + + /* + * Only plain relations have constraints, so stop here for other rtekinds. + */ + if (rte->rtekind != RTE_RELATION) + return false; + + /* + * If we are scanning just this table, we can use NO INHERIT constraints, + * but not if we're scanning its children too. (Note that partitioned + * tables should never have NO INHERIT constraints; but it's not necessary + * for us to assume that here.) + */ + include_noinherit = !rte->inh; + + /* + * Currently, attnotnull constraints must be treated as NO INHERIT unless + * this is a partitioned table. In future we might track their + * inheritance status more accurately, allowing this to be refined. + */ + include_notnull = (!rte->inh || rte->relkind == RELKIND_PARTITIONED_TABLE); + + /* + * Fetch the appropriate set of constraint expressions. + */ + constraint_pred = get_relation_constraints(root, rte->relid, rel, + include_noinherit, + include_notnull, + include_partition); + + /* + * We do not currently enforce that CHECK constraints contain only + * immutable functions, so it's necessary to check here. We daren't draw + * conclusions from plan-time evaluation of non-immutable functions. Since + * they're ANDed, we can just ignore any mutable constraints in the list, + * and reason about the rest. + */ + safe_constraints = NIL; + foreach(lc, constraint_pred) + { + Node *pred = (Node *) lfirst(lc); + + if (!contain_mutable_functions(pred)) + safe_constraints = lappend(safe_constraints, pred); + } + + /* + * The constraints are effectively ANDed together, so we can just try to + * refute the entire collection at once. This may allow us to make proofs + * that would fail if we took them individually. + * + * Note: we use rel->baserestrictinfo, not safe_restrictions as might seem + * an obvious optimization. Some of the clauses might be OR clauses that + * have volatile and nonvolatile subclauses, and it's OK to make + * deductions with the nonvolatile parts. + * + * We need strong refutation because we have to prove that the constraints + * would yield false, not just NULL. + */ + if (predicate_refuted_by(safe_constraints, rel->baserestrictinfo, false)) + return true; + + return false; +} + + +/* + * build_physical_tlist + * + * Build a targetlist consisting of exactly the relation's user attributes, + * in order. The executor can special-case such tlists to avoid a projection + * step at runtime, so we use such tlists preferentially for scan nodes. + * + * Exception: if there are any dropped or missing columns, we punt and return + * NIL. Ideally we would like to handle these cases too. However this + * creates problems for ExecTypeFromTL, which may be asked to build a tupdesc + * for a tlist that includes vars of no-longer-existent types. In theory we + * could dig out the required info from the pg_attribute entries of the + * relation, but that data is not readily available to ExecTypeFromTL. + * For now, we don't apply the physical-tlist optimization when there are + * dropped cols. + * + * We also support building a "physical" tlist for subqueries, functions, + * values lists, table expressions, and CTEs, since the same optimization can + * occur in SubqueryScan, FunctionScan, ValuesScan, CteScan, TableFunc, + * NamedTuplestoreScan, and WorkTableScan nodes. + */ +List * +build_physical_tlist(PlannerInfo *root, RelOptInfo *rel) +{ + List *tlist = NIL; + Index varno = rel->relid; + RangeTblEntry *rte = planner_rt_fetch(varno, root); + Relation relation; + Query *subquery; + Var *var; + ListCell *l; + int attrno, + numattrs; + List *colvars; + + switch (rte->rtekind) + { + case RTE_RELATION: + /* Assume we already have adequate lock */ + relation = table_open(rte->relid, NoLock); + + numattrs = RelationGetNumberOfAttributes(relation); + for (attrno = 1; attrno <= numattrs; attrno++) + { + Form_pg_attribute att_tup = TupleDescAttr(relation->rd_att, + attrno - 1); + + if (att_tup->attisdropped || att_tup->atthasmissing) + { + /* found a dropped or missing col, so punt */ + tlist = NIL; + break; + } + + var = makeVar(varno, + attrno, + att_tup->atttypid, + att_tup->atttypmod, + att_tup->attcollation, + 0); + + tlist = lappend(tlist, + makeTargetEntry((Expr *) var, + attrno, + NULL, + false)); + } + + table_close(relation, NoLock); + break; + + case RTE_SUBQUERY: + subquery = rte->subquery; + foreach(l, subquery->targetList) + { + TargetEntry *tle = (TargetEntry *) lfirst(l); + + /* + * A resjunk column of the subquery can be reflected as + * resjunk in the physical tlist; we need not punt. + */ + var = makeVarFromTargetEntry(varno, tle); + + tlist = lappend(tlist, + makeTargetEntry((Expr *) var, + tle->resno, + NULL, + tle->resjunk)); + } + break; + + case RTE_FUNCTION: + case RTE_TABLEFUNC: + case RTE_VALUES: + case RTE_CTE: + case RTE_NAMEDTUPLESTORE: + case RTE_RESULT: + /* Not all of these can have dropped cols, but share code anyway */ + expandRTE(rte, varno, 0, -1, true /* include dropped */ , + NULL, &colvars); + foreach(l, colvars) + { + var = (Var *) lfirst(l); + + /* + * A non-Var in expandRTE's output means a dropped column; + * must punt. + */ + if (!IsA(var, Var)) + { + tlist = NIL; + break; + } + + tlist = lappend(tlist, + makeTargetEntry((Expr *) var, + var->varattno, + NULL, + false)); + } + break; + + default: + /* caller error */ + elog(ERROR, "unsupported RTE kind %d in build_physical_tlist", + (int) rte->rtekind); + break; + } + + return tlist; +} + +/* + * build_index_tlist + * + * Build a targetlist representing the columns of the specified index. + * Each column is represented by a Var for the corresponding base-relation + * column, or an expression in base-relation Vars, as appropriate. + * + * There are never any dropped columns in indexes, so unlike + * build_physical_tlist, we need no failure case. + */ +static List * +build_index_tlist(PlannerInfo *root, IndexOptInfo *index, + Relation heapRelation) +{ + List *tlist = NIL; + Index varno = index->rel->relid; + ListCell *indexpr_item; + int i; + + indexpr_item = list_head(index->indexprs); + for (i = 0; i < index->ncolumns; i++) + { + int indexkey = index->indexkeys[i]; + Expr *indexvar; + + if (indexkey != 0) + { + /* simple column */ + const FormData_pg_attribute *att_tup; + + if (indexkey < 0) + att_tup = SystemAttributeDefinition(indexkey); + else + att_tup = TupleDescAttr(heapRelation->rd_att, indexkey - 1); + + indexvar = (Expr *) makeVar(varno, + indexkey, + att_tup->atttypid, + att_tup->atttypmod, + att_tup->attcollation, + 0); + } + else + { + /* expression column */ + if (indexpr_item == NULL) + elog(ERROR, "wrong number of index expressions"); + indexvar = (Expr *) lfirst(indexpr_item); + indexpr_item = lnext(index->indexprs, indexpr_item); + } + + tlist = lappend(tlist, + makeTargetEntry(indexvar, + i + 1, + NULL, + false)); + } + if (indexpr_item != NULL) + elog(ERROR, "wrong number of index expressions"); + + return tlist; +} + +/* + * restriction_selectivity + * + * Returns the selectivity of a specified restriction operator clause. + * This code executes registered procedures stored in the + * operator relation, by calling the function manager. + * + * See clause_selectivity() for the meaning of the additional parameters. + */ +Selectivity +restriction_selectivity(PlannerInfo *root, + Oid operatorid, + List *args, + Oid inputcollid, + int varRelid) +{ + RegProcedure oprrest = get_oprrest(operatorid); + float8 result; + + /* + * if the oprrest procedure is missing for whatever reason, use a + * selectivity of 0.5 + */ + if (!oprrest) + return (Selectivity) 0.5; + + result = DatumGetFloat8(OidFunctionCall4Coll(oprrest, + inputcollid, + PointerGetDatum(root), + ObjectIdGetDatum(operatorid), + PointerGetDatum(args), + Int32GetDatum(varRelid))); + + if (result < 0.0 || result > 1.0) + elog(ERROR, "invalid restriction selectivity: %f", result); + + return (Selectivity) result; +} + +/* + * join_selectivity + * + * Returns the selectivity of a specified join operator clause. + * This code executes registered procedures stored in the + * operator relation, by calling the function manager. + * + * See clause_selectivity() for the meaning of the additional parameters. + */ +Selectivity +join_selectivity(PlannerInfo *root, + Oid operatorid, + List *args, + Oid inputcollid, + JoinType jointype, + SpecialJoinInfo *sjinfo) +{ + RegProcedure oprjoin = get_oprjoin(operatorid); + float8 result; + + /* + * if the oprjoin procedure is missing for whatever reason, use a + * selectivity of 0.5 + */ + if (!oprjoin) + return (Selectivity) 0.5; + + result = DatumGetFloat8(OidFunctionCall5Coll(oprjoin, + inputcollid, + PointerGetDatum(root), + ObjectIdGetDatum(operatorid), + PointerGetDatum(args), + Int16GetDatum(jointype), + PointerGetDatum(sjinfo))); + + if (result < 0.0 || result > 1.0) + elog(ERROR, "invalid join selectivity: %f", result); + + return (Selectivity) result; +} + +/* + * function_selectivity + * + * Returns the selectivity of a specified boolean function clause. + * This code executes registered procedures stored in the + * pg_proc relation, by calling the function manager. + * + * See clause_selectivity() for the meaning of the additional parameters. + */ +Selectivity +function_selectivity(PlannerInfo *root, + Oid funcid, + List *args, + Oid inputcollid, + bool is_join, + int varRelid, + JoinType jointype, + SpecialJoinInfo *sjinfo) +{ + RegProcedure prosupport = get_func_support(funcid); + SupportRequestSelectivity req; + SupportRequestSelectivity *sresult; + + /* + * If no support function is provided, use our historical default + * estimate, 0.3333333. This seems a pretty unprincipled choice, but + * Postgres has been using that estimate for function calls since 1992. + * The hoariness of this behavior suggests that we should not be in too + * much hurry to use another value. + */ + if (!prosupport) + return (Selectivity) 0.3333333; + + req.type = T_SupportRequestSelectivity; + req.root = root; + req.funcid = funcid; + req.args = args; + req.inputcollid = inputcollid; + req.is_join = is_join; + req.varRelid = varRelid; + req.jointype = jointype; + req.sjinfo = sjinfo; + req.selectivity = -1; /* to catch failure to set the value */ + + sresult = (SupportRequestSelectivity *) + DatumGetPointer(OidFunctionCall1(prosupport, + PointerGetDatum(&req))); + + /* If support function fails, use default */ + if (sresult != &req) + return (Selectivity) 0.3333333; + + if (req.selectivity < 0.0 || req.selectivity > 1.0) + elog(ERROR, "invalid function selectivity: %f", req.selectivity); + + return (Selectivity) req.selectivity; +} + +/* + * add_function_cost + * + * Get an estimate of the execution cost of a function, and *add* it to + * the contents of *cost. The estimate may include both one-time and + * per-tuple components, since QualCost does. + * + * The funcid must always be supplied. If it is being called as the + * implementation of a specific parsetree node (FuncExpr, OpExpr, + * WindowFunc, etc), pass that as "node", else pass NULL. + * + * In some usages root might be NULL, too. + */ +void +add_function_cost(PlannerInfo *root, Oid funcid, Node *node, + QualCost *cost) +{ + HeapTuple proctup; + Form_pg_proc procform; + + proctup = SearchSysCache1(PROCOID, ObjectIdGetDatum(funcid)); + if (!HeapTupleIsValid(proctup)) + elog(ERROR, "cache lookup failed for function %u", funcid); + procform = (Form_pg_proc) GETSTRUCT(proctup); + + if (OidIsValid(procform->prosupport)) + { + SupportRequestCost req; + SupportRequestCost *sresult; + + req.type = T_SupportRequestCost; + req.root = root; + req.funcid = funcid; + req.node = node; + + /* Initialize cost fields so that support function doesn't have to */ + req.startup = 0; + req.per_tuple = 0; + + sresult = (SupportRequestCost *) + DatumGetPointer(OidFunctionCall1(procform->prosupport, + PointerGetDatum(&req))); + + if (sresult == &req) + { + /* Success, so accumulate support function's estimate into *cost */ + cost->startup += req.startup; + cost->per_tuple += req.per_tuple; + ReleaseSysCache(proctup); + return; + } + } + + /* No support function, or it failed, so rely on procost */ + cost->per_tuple += procform->procost * cpu_operator_cost; + + ReleaseSysCache(proctup); +} + +/* + * get_function_rows + * + * Get an estimate of the number of rows returned by a set-returning function. + * + * The funcid must always be supplied. In current usage, the calling node + * will always be supplied, and will be either a FuncExpr or OpExpr. + * But it's a good idea to not fail if it's NULL. + * + * In some usages root might be NULL, too. + * + * Note: this returns the unfiltered result of the support function, if any. + * It's usually a good idea to apply clamp_row_est() to the result, but we + * leave it to the caller to do so. + */ +double +get_function_rows(PlannerInfo *root, Oid funcid, Node *node) +{ + HeapTuple proctup; + Form_pg_proc procform; + double result; + + proctup = SearchSysCache1(PROCOID, ObjectIdGetDatum(funcid)); + if (!HeapTupleIsValid(proctup)) + elog(ERROR, "cache lookup failed for function %u", funcid); + procform = (Form_pg_proc) GETSTRUCT(proctup); + + Assert(procform->proretset); /* else caller error */ + + if (OidIsValid(procform->prosupport)) + { + SupportRequestRows req; + SupportRequestRows *sresult; + + req.type = T_SupportRequestRows; + req.root = root; + req.funcid = funcid; + req.node = node; + + req.rows = 0; /* just for sanity */ + + sresult = (SupportRequestRows *) + DatumGetPointer(OidFunctionCall1(procform->prosupport, + PointerGetDatum(&req))); + + if (sresult == &req) + { + /* Success */ + ReleaseSysCache(proctup); + return req.rows; + } + } + + /* No support function, or it failed, so rely on prorows */ + result = procform->prorows; + + ReleaseSysCache(proctup); + + return result; +} + +/* + * has_unique_index + * + * Detect whether there is a unique index on the specified attribute + * of the specified relation, thus allowing us to conclude that all + * the (non-null) values of the attribute are distinct. + * + * This function does not check the index's indimmediate property, which + * means that uniqueness may transiently fail to hold intra-transaction. + * That's appropriate when we are making statistical estimates, but beware + * of using this for any correctness proofs. + */ +bool +has_unique_index(RelOptInfo *rel, AttrNumber attno) +{ + ListCell *ilist; + + foreach(ilist, rel->indexlist) + { + IndexOptInfo *index = (IndexOptInfo *) lfirst(ilist); + + /* + * Note: ignore partial indexes, since they don't allow us to conclude + * that all attr values are distinct, *unless* they are marked predOK + * which means we know the index's predicate is satisfied by the + * query. We don't take any interest in expressional indexes either. + * Also, a multicolumn unique index doesn't allow us to conclude that + * just the specified attr is unique. + */ + if (index->unique && + index->nkeycolumns == 1 && + index->indexkeys[0] == attno && + (index->indpred == NIL || index->predOK)) + return true; + } + return false; +} + + +/* + * has_row_triggers + * + * Detect whether the specified relation has any row-level triggers for event. + */ +bool +has_row_triggers(PlannerInfo *root, Index rti, CmdType event) +{ + RangeTblEntry *rte = planner_rt_fetch(rti, root); + Relation relation; + TriggerDesc *trigDesc; + bool result = false; + + /* Assume we already have adequate lock */ + relation = table_open(rte->relid, NoLock); + + trigDesc = relation->trigdesc; + switch (event) + { + case CMD_INSERT: + if (trigDesc && + (trigDesc->trig_insert_after_row || + trigDesc->trig_insert_before_row)) + result = true; + break; + case CMD_UPDATE: + if (trigDesc && + (trigDesc->trig_update_after_row || + trigDesc->trig_update_before_row)) + result = true; + break; + case CMD_DELETE: + if (trigDesc && + (trigDesc->trig_delete_after_row || + trigDesc->trig_delete_before_row)) + result = true; + break; + /* There is no separate event for MERGE, only INSERT/UPDATE/DELETE */ + case CMD_MERGE: + result = false; + break; + default: + elog(ERROR, "unrecognized CmdType: %d", (int) event); + break; + } + + table_close(relation, NoLock); + return result; +} + +/* + * has_stored_generated_columns + * + * Does table identified by RTI have any STORED GENERATED columns? + */ +bool +has_stored_generated_columns(PlannerInfo *root, Index rti) +{ + RangeTblEntry *rte = planner_rt_fetch(rti, root); + Relation relation; + TupleDesc tupdesc; + bool result = false; + + /* Assume we already have adequate lock */ + relation = table_open(rte->relid, NoLock); + + tupdesc = RelationGetDescr(relation); + result = tupdesc->constr && tupdesc->constr->has_generated_stored; + + table_close(relation, NoLock); + + return result; +} + +/* + * get_dependent_generated_columns + * + * Get the column numbers of any STORED GENERATED columns of the relation + * that depend on any column listed in target_cols. Both the input and + * result bitmapsets contain column numbers offset by + * FirstLowInvalidHeapAttributeNumber. + */ +Bitmapset * +get_dependent_generated_columns(PlannerInfo *root, Index rti, + Bitmapset *target_cols) +{ + Bitmapset *dependentCols = NULL; + RangeTblEntry *rte = planner_rt_fetch(rti, root); + Relation relation; + TupleDesc tupdesc; + TupleConstr *constr; + + /* Assume we already have adequate lock */ + relation = table_open(rte->relid, NoLock); + + tupdesc = RelationGetDescr(relation); + constr = tupdesc->constr; + + if (constr && constr->has_generated_stored) + { + for (int i = 0; i < constr->num_defval; i++) + { + AttrDefault *defval = &constr->defval[i]; + Node *expr; + Bitmapset *attrs_used = NULL; + + /* skip if not generated column */ + if (!TupleDescAttr(tupdesc, defval->adnum - 1)->attgenerated) + continue; + + /* identify columns this generated column depends on */ + expr = stringToNode(defval->adbin); + pull_varattnos(expr, 1, &attrs_used); + + if (bms_overlap(target_cols, attrs_used)) + dependentCols = bms_add_member(dependentCols, + defval->adnum - FirstLowInvalidHeapAttributeNumber); + } + } + + table_close(relation, NoLock); + + return dependentCols; +} + +/* + * set_relation_partition_info + * + * Set partitioning scheme and related information for a partitioned table. + */ +static void +set_relation_partition_info(PlannerInfo *root, RelOptInfo *rel, + Relation relation) +{ + PartitionDesc partdesc; + + /* + * Create the PartitionDirectory infrastructure if we didn't already. + */ + if (root->glob->partition_directory == NULL) + { + root->glob->partition_directory = + CreatePartitionDirectory(CurrentMemoryContext, true); + } + + partdesc = PartitionDirectoryLookup(root->glob->partition_directory, + relation); + rel->part_scheme = find_partition_scheme(root, relation); + Assert(partdesc != NULL && rel->part_scheme != NULL); + rel->boundinfo = partdesc->boundinfo; + rel->nparts = partdesc->nparts; + set_baserel_partition_key_exprs(relation, rel); + set_baserel_partition_constraint(relation, rel); +} + +/* + * find_partition_scheme + * + * Find or create a PartitionScheme for this Relation. + */ +static PartitionScheme +find_partition_scheme(PlannerInfo *root, Relation relation) +{ + PartitionKey partkey = RelationGetPartitionKey(relation); + ListCell *lc; + int partnatts, + i; + PartitionScheme part_scheme; + + /* A partitioned table should have a partition key. */ + Assert(partkey != NULL); + + partnatts = partkey->partnatts; + + /* Search for a matching partition scheme and return if found one. */ + foreach(lc, root->part_schemes) + { + part_scheme = lfirst(lc); + + /* Match partitioning strategy and number of keys. */ + if (partkey->strategy != part_scheme->strategy || + partnatts != part_scheme->partnatts) + continue; + + /* Match partition key type properties. */ + if (memcmp(partkey->partopfamily, part_scheme->partopfamily, + sizeof(Oid) * partnatts) != 0 || + memcmp(partkey->partopcintype, part_scheme->partopcintype, + sizeof(Oid) * partnatts) != 0 || + memcmp(partkey->partcollation, part_scheme->partcollation, + sizeof(Oid) * partnatts) != 0) + continue; + + /* + * Length and byval information should match when partopcintype + * matches. + */ + Assert(memcmp(partkey->parttyplen, part_scheme->parttyplen, + sizeof(int16) * partnatts) == 0); + Assert(memcmp(partkey->parttypbyval, part_scheme->parttypbyval, + sizeof(bool) * partnatts) == 0); + + /* + * If partopfamily and partopcintype matched, must have the same + * partition comparison functions. Note that we cannot reliably + * Assert the equality of function structs themselves for they might + * be different across PartitionKey's, so just Assert for the function + * OIDs. + */ +#ifdef USE_ASSERT_CHECKING + for (i = 0; i < partkey->partnatts; i++) + Assert(partkey->partsupfunc[i].fn_oid == + part_scheme->partsupfunc[i].fn_oid); +#endif + + /* Found matching partition scheme. */ + return part_scheme; + } + + /* + * Did not find matching partition scheme. Create one copying relevant + * information from the relcache. We need to copy the contents of the + * array since the relcache entry may not survive after we have closed the + * relation. + */ + part_scheme = (PartitionScheme) palloc0(sizeof(PartitionSchemeData)); + part_scheme->strategy = partkey->strategy; + part_scheme->partnatts = partkey->partnatts; + + part_scheme->partopfamily = (Oid *) palloc(sizeof(Oid) * partnatts); + memcpy(part_scheme->partopfamily, partkey->partopfamily, + sizeof(Oid) * partnatts); + + part_scheme->partopcintype = (Oid *) palloc(sizeof(Oid) * partnatts); + memcpy(part_scheme->partopcintype, partkey->partopcintype, + sizeof(Oid) * partnatts); + + part_scheme->partcollation = (Oid *) palloc(sizeof(Oid) * partnatts); + memcpy(part_scheme->partcollation, partkey->partcollation, + sizeof(Oid) * partnatts); + + part_scheme->parttyplen = (int16 *) palloc(sizeof(int16) * partnatts); + memcpy(part_scheme->parttyplen, partkey->parttyplen, + sizeof(int16) * partnatts); + + part_scheme->parttypbyval = (bool *) palloc(sizeof(bool) * partnatts); + memcpy(part_scheme->parttypbyval, partkey->parttypbyval, + sizeof(bool) * partnatts); + + part_scheme->partsupfunc = (FmgrInfo *) + palloc(sizeof(FmgrInfo) * partnatts); + for (i = 0; i < partnatts; i++) + fmgr_info_copy(&part_scheme->partsupfunc[i], &partkey->partsupfunc[i], + CurrentMemoryContext); + + /* Add the partitioning scheme to PlannerInfo. */ + root->part_schemes = lappend(root->part_schemes, part_scheme); + + return part_scheme; +} + +/* + * set_baserel_partition_key_exprs + * + * Builds partition key expressions for the given base relation and fills + * rel->partexprs. + */ +static void +set_baserel_partition_key_exprs(Relation relation, + RelOptInfo *rel) +{ + PartitionKey partkey = RelationGetPartitionKey(relation); + int partnatts; + int cnt; + List **partexprs; + ListCell *lc; + Index varno = rel->relid; + + Assert(IS_SIMPLE_REL(rel) && rel->relid > 0); + + /* A partitioned table should have a partition key. */ + Assert(partkey != NULL); + + partnatts = partkey->partnatts; + partexprs = (List **) palloc(sizeof(List *) * partnatts); + lc = list_head(partkey->partexprs); + + for (cnt = 0; cnt < partnatts; cnt++) + { + Expr *partexpr; + AttrNumber attno = partkey->partattrs[cnt]; + + if (attno != InvalidAttrNumber) + { + /* Single column partition key is stored as a Var node. */ + Assert(attno > 0); + + partexpr = (Expr *) makeVar(varno, attno, + partkey->parttypid[cnt], + partkey->parttypmod[cnt], + partkey->parttypcoll[cnt], 0); + } + else + { + if (lc == NULL) + elog(ERROR, "wrong number of partition key expressions"); + + /* Re-stamp the expression with given varno. */ + partexpr = (Expr *) copyObject(lfirst(lc)); + ChangeVarNodes((Node *) partexpr, 1, varno, 0); + lc = lnext(partkey->partexprs, lc); + } + + /* Base relations have a single expression per key. */ + partexprs[cnt] = list_make1(partexpr); + } + + rel->partexprs = partexprs; + + /* + * A base relation does not have nullable partition key expressions, since + * no outer join is involved. We still allocate an array of empty + * expression lists to keep partition key expression handling code simple. + * See build_joinrel_partition_info() and match_expr_to_partition_keys(). + */ + rel->nullable_partexprs = (List **) palloc0(sizeof(List *) * partnatts); +} + +/* + * set_baserel_partition_constraint + * + * Builds the partition constraint for the given base relation and sets it + * in the given RelOptInfo. All Var nodes are restamped with the relid of the + * given relation. + */ +static void +set_baserel_partition_constraint(Relation relation, RelOptInfo *rel) +{ + List *partconstr; + + if (rel->partition_qual) /* already done */ + return; + + /* + * Run the partition quals through const-simplification similar to check + * constraints. We skip canonicalize_qual, though, because partition + * quals should be in canonical form already; also, since the qual is in + * implicit-AND format, we'd have to explicitly convert it to explicit-AND + * format and back again. + */ + partconstr = RelationGetPartitionQual(relation); + if (partconstr) + { + partconstr = (List *) expression_planner((Expr *) partconstr); + if (rel->relid != 1) + ChangeVarNodes((Node *) partconstr, 1, rel->relid, 0); + rel->partition_qual = partconstr; + } +} diff --git a/src/backend/optimizer/util/predtest.c b/src/backend/optimizer/util/predtest.c new file mode 100644 index 0000000..182d5b1 --- /dev/null +++ b/src/backend/optimizer/util/predtest.c @@ -0,0 +1,2224 @@ +/*------------------------------------------------------------------------- + * + * predtest.c + * Routines to attempt to prove logical implications between predicate + * expressions. + * + * Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group + * Portions Copyright (c) 1994, Regents of the University of California + * + * + * IDENTIFICATION + * src/backend/optimizer/util/predtest.c + * + *------------------------------------------------------------------------- + */ +#include "postgres.h" + +#include "catalog/pg_proc.h" +#include "catalog/pg_type.h" +#include "executor/executor.h" +#include "miscadmin.h" +#include "nodes/makefuncs.h" +#include "nodes/nodeFuncs.h" +#include "nodes/pathnodes.h" +#include "optimizer/optimizer.h" +#include "utils/array.h" +#include "utils/inval.h" +#include "utils/lsyscache.h" +#include "utils/syscache.h" + + +/* + * Proof attempts involving large arrays in ScalarArrayOpExpr nodes are + * likely to require O(N^2) time, and more often than not fail anyway. + * So we set an arbitrary limit on the number of array elements that + * we will allow to be treated as an AND or OR clause. + * XXX is it worth exposing this as a GUC knob? + */ +#define MAX_SAOP_ARRAY_SIZE 100 + +/* + * To avoid redundant coding in predicate_implied_by_recurse and + * predicate_refuted_by_recurse, we need to abstract out the notion of + * iterating over the components of an expression that is logically an AND + * or OR structure. There are multiple sorts of expression nodes that can + * be treated as ANDs or ORs, and we don't want to code each one separately. + * Hence, these types and support routines. + */ +typedef enum +{ + CLASS_ATOM, /* expression that's not AND or OR */ + CLASS_AND, /* expression with AND semantics */ + CLASS_OR /* expression with OR semantics */ +} PredClass; + +typedef struct PredIterInfoData *PredIterInfo; + +typedef struct PredIterInfoData +{ + /* node-type-specific iteration state */ + void *state; + List *state_list; + /* initialize to do the iteration */ + void (*startup_fn) (Node *clause, PredIterInfo info); + /* next-component iteration function */ + Node *(*next_fn) (PredIterInfo info); + /* release resources when done with iteration */ + void (*cleanup_fn) (PredIterInfo info); +} PredIterInfoData; + +#define iterate_begin(item, clause, info) \ + do { \ + Node *item; \ + (info).startup_fn((clause), &(info)); \ + while ((item = (info).next_fn(&(info))) != NULL) + +#define iterate_end(info) \ + (info).cleanup_fn(&(info)); \ + } while (0) + + +static bool predicate_implied_by_recurse(Node *clause, Node *predicate, + bool weak); +static bool predicate_refuted_by_recurse(Node *clause, Node *predicate, + bool weak); +static PredClass predicate_classify(Node *clause, PredIterInfo info); +static void list_startup_fn(Node *clause, PredIterInfo info); +static Node *list_next_fn(PredIterInfo info); +static void list_cleanup_fn(PredIterInfo info); +static void boolexpr_startup_fn(Node *clause, PredIterInfo info); +static void arrayconst_startup_fn(Node *clause, PredIterInfo info); +static Node *arrayconst_next_fn(PredIterInfo info); +static void arrayconst_cleanup_fn(PredIterInfo info); +static void arrayexpr_startup_fn(Node *clause, PredIterInfo info); +static Node *arrayexpr_next_fn(PredIterInfo info); +static void arrayexpr_cleanup_fn(PredIterInfo info); +static bool predicate_implied_by_simple_clause(Expr *predicate, Node *clause, + bool weak); +static bool predicate_refuted_by_simple_clause(Expr *predicate, Node *clause, + bool weak); +static Node *extract_not_arg(Node *clause); +static Node *extract_strong_not_arg(Node *clause); +static bool clause_is_strict_for(Node *clause, Node *subexpr, bool allow_false); +static bool operator_predicate_proof(Expr *predicate, Node *clause, + bool refute_it, bool weak); +static bool operator_same_subexprs_proof(Oid pred_op, Oid clause_op, + bool refute_it); +static bool operator_same_subexprs_lookup(Oid pred_op, Oid clause_op, + bool refute_it); +static Oid get_btree_test_op(Oid pred_op, Oid clause_op, bool refute_it); +static void InvalidateOprProofCacheCallBack(Datum arg, int cacheid, uint32 hashvalue); + + +/* + * predicate_implied_by + * Recursively checks whether the clauses in clause_list imply that the + * given predicate is true. + * + * We support two definitions of implication: + * + * "Strong" implication: A implies B means that truth of A implies truth of B. + * We use this to prove that a row satisfying one WHERE clause or index + * predicate must satisfy another one. + * + * "Weak" implication: A implies B means that non-falsity of A implies + * non-falsity of B ("non-false" means "either true or NULL"). We use this to + * prove that a row satisfying one CHECK constraint must satisfy another one. + * + * Strong implication can also be used to prove that a WHERE clause implies a + * CHECK constraint, although it will fail to prove a few cases where we could + * safely conclude that the implication holds. There's no support for proving + * the converse case, since only a few kinds of CHECK constraint would allow + * deducing anything. + * + * The top-level List structure of each list corresponds to an AND list. + * We assume that eval_const_expressions() has been applied and so there + * are no un-flattened ANDs or ORs (e.g., no AND immediately within an AND, + * including AND just below the top-level List structure). + * If this is not true we might fail to prove an implication that is + * valid, but no worse consequences will ensue. + * + * We assume the predicate has already been checked to contain only + * immutable functions and operators. (In many current uses this is known + * true because the predicate is part of an index predicate that has passed + * CheckPredicate(); otherwise, the caller must check it.) We dare not make + * deductions based on non-immutable functions, because they might change + * answers between the time we make the plan and the time we execute the plan. + * Immutability of functions in the clause_list is checked here, if necessary. + */ +bool +predicate_implied_by(List *predicate_list, List *clause_list, + bool weak) +{ + Node *p, + *c; + + if (predicate_list == NIL) + return true; /* no predicate: implication is vacuous */ + if (clause_list == NIL) + return false; /* no restriction: implication must fail */ + + /* + * If either input is a single-element list, replace it with its lone + * member; this avoids one useless level of AND-recursion. We only need + * to worry about this at top level, since eval_const_expressions should + * have gotten rid of any trivial ANDs or ORs below that. + */ + if (list_length(predicate_list) == 1) + p = (Node *) linitial(predicate_list); + else + p = (Node *) predicate_list; + if (list_length(clause_list) == 1) + c = (Node *) linitial(clause_list); + else + c = (Node *) clause_list; + + /* And away we go ... */ + return predicate_implied_by_recurse(c, p, weak); +} + +/* + * predicate_refuted_by + * Recursively checks whether the clauses in clause_list refute the given + * predicate (that is, prove it false). + * + * This is NOT the same as !(predicate_implied_by), though it is similar + * in the technique and structure of the code. + * + * We support two definitions of refutation: + * + * "Strong" refutation: A refutes B means truth of A implies falsity of B. + * We use this to disprove a CHECK constraint given a WHERE clause, i.e., + * prove that any row satisfying the WHERE clause would violate the CHECK + * constraint. (Observe we must prove B yields false, not just not-true.) + * + * "Weak" refutation: A refutes B means truth of A implies non-truth of B + * (i.e., B must yield false or NULL). We use this to detect mutually + * contradictory WHERE clauses. + * + * Weak refutation can be proven in some cases where strong refutation doesn't + * hold, so it's useful to use it when possible. We don't currently have + * support for disproving one CHECK constraint based on another one, nor for + * disproving WHERE based on CHECK. (As with implication, the last case + * doesn't seem very practical. CHECK-vs-CHECK might be useful, but isn't + * currently needed anywhere.) + * + * The top-level List structure of each list corresponds to an AND list. + * We assume that eval_const_expressions() has been applied and so there + * are no un-flattened ANDs or ORs (e.g., no AND immediately within an AND, + * including AND just below the top-level List structure). + * If this is not true we might fail to prove an implication that is + * valid, but no worse consequences will ensue. + * + * We assume the predicate has already been checked to contain only + * immutable functions and operators. We dare not make deductions based on + * non-immutable functions, because they might change answers between the + * time we make the plan and the time we execute the plan. + * Immutability of functions in the clause_list is checked here, if necessary. + */ +bool +predicate_refuted_by(List *predicate_list, List *clause_list, + bool weak) +{ + Node *p, + *c; + + if (predicate_list == NIL) + return false; /* no predicate: no refutation is possible */ + if (clause_list == NIL) + return false; /* no restriction: refutation must fail */ + + /* + * If either input is a single-element list, replace it with its lone + * member; this avoids one useless level of AND-recursion. We only need + * to worry about this at top level, since eval_const_expressions should + * have gotten rid of any trivial ANDs or ORs below that. + */ + if (list_length(predicate_list) == 1) + p = (Node *) linitial(predicate_list); + else + p = (Node *) predicate_list; + if (list_length(clause_list) == 1) + c = (Node *) linitial(clause_list); + else + c = (Node *) clause_list; + + /* And away we go ... */ + return predicate_refuted_by_recurse(c, p, weak); +} + +/*---------- + * predicate_implied_by_recurse + * Does the predicate implication test for non-NULL restriction and + * predicate clauses. + * + * The logic followed here is ("=>" means "implies"): + * atom A => atom B iff: predicate_implied_by_simple_clause says so + * atom A => AND-expr B iff: A => each of B's components + * atom A => OR-expr B iff: A => any of B's components + * AND-expr A => atom B iff: any of A's components => B + * AND-expr A => AND-expr B iff: A => each of B's components + * AND-expr A => OR-expr B iff: A => any of B's components, + * *or* any of A's components => B + * OR-expr A => atom B iff: each of A's components => B + * OR-expr A => AND-expr B iff: A => each of B's components + * OR-expr A => OR-expr B iff: each of A's components => any of B's + * + * An "atom" is anything other than an AND or OR node. Notice that we don't + * have any special logic to handle NOT nodes; these should have been pushed + * down or eliminated where feasible during eval_const_expressions(). + * + * All of these rules apply equally to strong or weak implication. + * + * We can't recursively expand either side first, but have to interleave + * the expansions per the above rules, to be sure we handle all of these + * examples: + * (x OR y) => (x OR y OR z) + * (x AND y AND z) => (x AND y) + * (x AND y) => ((x AND y) OR z) + * ((x OR y) AND z) => (x OR y) + * This is still not an exhaustive test, but it handles most normal cases + * under the assumption that both inputs have been AND/OR flattened. + * + * We have to be prepared to handle RestrictInfo nodes in the restrictinfo + * tree, though not in the predicate tree. + *---------- + */ +static bool +predicate_implied_by_recurse(Node *clause, Node *predicate, + bool weak) +{ + PredIterInfoData clause_info; + PredIterInfoData pred_info; + PredClass pclass; + bool result; + + /* skip through RestrictInfo */ + Assert(clause != NULL); + if (IsA(clause, RestrictInfo)) + clause = (Node *) ((RestrictInfo *) clause)->clause; + + pclass = predicate_classify(predicate, &pred_info); + + switch (predicate_classify(clause, &clause_info)) + { + case CLASS_AND: + switch (pclass) + { + case CLASS_AND: + + /* + * AND-clause => AND-clause if A implies each of B's items + */ + result = true; + iterate_begin(pitem, predicate, pred_info) + { + if (!predicate_implied_by_recurse(clause, pitem, + weak)) + { + result = false; + break; + } + } + iterate_end(pred_info); + return result; + + case CLASS_OR: + + /* + * AND-clause => OR-clause if A implies any of B's items + * + * Needed to handle (x AND y) => ((x AND y) OR z) + */ + result = false; + iterate_begin(pitem, predicate, pred_info) + { + if (predicate_implied_by_recurse(clause, pitem, + weak)) + { + result = true; + break; + } + } + iterate_end(pred_info); + if (result) + return result; + + /* + * Also check if any of A's items implies B + * + * Needed to handle ((x OR y) AND z) => (x OR y) + */ + iterate_begin(citem, clause, clause_info) + { + if (predicate_implied_by_recurse(citem, predicate, + weak)) + { + result = true; + break; + } + } + iterate_end(clause_info); + return result; + + case CLASS_ATOM: + + /* + * AND-clause => atom if any of A's items implies B + */ + result = false; + iterate_begin(citem, clause, clause_info) + { + if (predicate_implied_by_recurse(citem, predicate, + weak)) + { + result = true; + break; + } + } + iterate_end(clause_info); + return result; + } + break; + + case CLASS_OR: + switch (pclass) + { + case CLASS_OR: + + /* + * OR-clause => OR-clause if each of A's items implies any + * of B's items. Messy but can't do it any more simply. + */ + result = true; + iterate_begin(citem, clause, clause_info) + { + bool presult = false; + + iterate_begin(pitem, predicate, pred_info) + { + if (predicate_implied_by_recurse(citem, pitem, + weak)) + { + presult = true; + break; + } + } + iterate_end(pred_info); + if (!presult) + { + result = false; /* doesn't imply any of B's */ + break; + } + } + iterate_end(clause_info); + return result; + + case CLASS_AND: + case CLASS_ATOM: + + /* + * OR-clause => AND-clause if each of A's items implies B + * + * OR-clause => atom if each of A's items implies B + */ + result = true; + iterate_begin(citem, clause, clause_info) + { + if (!predicate_implied_by_recurse(citem, predicate, + weak)) + { + result = false; + break; + } + } + iterate_end(clause_info); + return result; + } + break; + + case CLASS_ATOM: + switch (pclass) + { + case CLASS_AND: + + /* + * atom => AND-clause if A implies each of B's items + */ + result = true; + iterate_begin(pitem, predicate, pred_info) + { + if (!predicate_implied_by_recurse(clause, pitem, + weak)) + { + result = false; + break; + } + } + iterate_end(pred_info); + return result; + + case CLASS_OR: + + /* + * atom => OR-clause if A implies any of B's items + */ + result = false; + iterate_begin(pitem, predicate, pred_info) + { + if (predicate_implied_by_recurse(clause, pitem, + weak)) + { + result = true; + break; + } + } + iterate_end(pred_info); + return result; + + case CLASS_ATOM: + + /* + * atom => atom is the base case + */ + return + predicate_implied_by_simple_clause((Expr *) predicate, + clause, + weak); + } + break; + } + + /* can't get here */ + elog(ERROR, "predicate_classify returned a bogus value"); + return false; +} + +/*---------- + * predicate_refuted_by_recurse + * Does the predicate refutation test for non-NULL restriction and + * predicate clauses. + * + * The logic followed here is ("R=>" means "refutes"): + * atom A R=> atom B iff: predicate_refuted_by_simple_clause says so + * atom A R=> AND-expr B iff: A R=> any of B's components + * atom A R=> OR-expr B iff: A R=> each of B's components + * AND-expr A R=> atom B iff: any of A's components R=> B + * AND-expr A R=> AND-expr B iff: A R=> any of B's components, + * *or* any of A's components R=> B + * AND-expr A R=> OR-expr B iff: A R=> each of B's components + * OR-expr A R=> atom B iff: each of A's components R=> B + * OR-expr A R=> AND-expr B iff: each of A's components R=> any of B's + * OR-expr A R=> OR-expr B iff: A R=> each of B's components + * + * All of the above rules apply equally to strong or weak refutation. + * + * In addition, if the predicate is a NOT-clause then we can use + * A R=> NOT B if: A => B + * This works for several different SQL constructs that assert the non-truth + * of their argument, ie NOT, IS FALSE, IS NOT TRUE, IS UNKNOWN, although some + * of them require that we prove strong implication. Likewise, we can use + * NOT A R=> B if: B => A + * but here we must be careful about strong vs. weak refutation and make + * the appropriate type of implication proof (weak or strong respectively). + * + * Other comments are as for predicate_implied_by_recurse(). + *---------- + */ +static bool +predicate_refuted_by_recurse(Node *clause, Node *predicate, + bool weak) +{ + PredIterInfoData clause_info; + PredIterInfoData pred_info; + PredClass pclass; + Node *not_arg; + bool result; + + /* skip through RestrictInfo */ + Assert(clause != NULL); + if (IsA(clause, RestrictInfo)) + clause = (Node *) ((RestrictInfo *) clause)->clause; + + pclass = predicate_classify(predicate, &pred_info); + + switch (predicate_classify(clause, &clause_info)) + { + case CLASS_AND: + switch (pclass) + { + case CLASS_AND: + + /* + * AND-clause R=> AND-clause if A refutes any of B's items + * + * Needed to handle (x AND y) R=> ((!x OR !y) AND z) + */ + result = false; + iterate_begin(pitem, predicate, pred_info) + { + if (predicate_refuted_by_recurse(clause, pitem, + weak)) + { + result = true; + break; + } + } + iterate_end(pred_info); + if (result) + return result; + + /* + * Also check if any of A's items refutes B + * + * Needed to handle ((x OR y) AND z) R=> (!x AND !y) + */ + iterate_begin(citem, clause, clause_info) + { + if (predicate_refuted_by_recurse(citem, predicate, + weak)) + { + result = true; + break; + } + } + iterate_end(clause_info); + return result; + + case CLASS_OR: + + /* + * AND-clause R=> OR-clause if A refutes each of B's items + */ + result = true; + iterate_begin(pitem, predicate, pred_info) + { + if (!predicate_refuted_by_recurse(clause, pitem, + weak)) + { + result = false; + break; + } + } + iterate_end(pred_info); + return result; + + case CLASS_ATOM: + + /* + * If B is a NOT-type clause, A R=> B if A => B's arg + * + * Since, for either type of refutation, we are starting + * with the premise that A is true, we can use a strong + * implication test in all cases. That proves B's arg is + * true, which is more than we need for weak refutation if + * B is a simple NOT, but it allows not worrying about + * exactly which kind of negation clause we have. + */ + not_arg = extract_not_arg(predicate); + if (not_arg && + predicate_implied_by_recurse(clause, not_arg, + false)) + return true; + + /* + * AND-clause R=> atom if any of A's items refutes B + */ + result = false; + iterate_begin(citem, clause, clause_info) + { + if (predicate_refuted_by_recurse(citem, predicate, + weak)) + { + result = true; + break; + } + } + iterate_end(clause_info); + return result; + } + break; + + case CLASS_OR: + switch (pclass) + { + case CLASS_OR: + + /* + * OR-clause R=> OR-clause if A refutes each of B's items + */ + result = true; + iterate_begin(pitem, predicate, pred_info) + { + if (!predicate_refuted_by_recurse(clause, pitem, + weak)) + { + result = false; + break; + } + } + iterate_end(pred_info); + return result; + + case CLASS_AND: + + /* + * OR-clause R=> AND-clause if each of A's items refutes + * any of B's items. + */ + result = true; + iterate_begin(citem, clause, clause_info) + { + bool presult = false; + + iterate_begin(pitem, predicate, pred_info) + { + if (predicate_refuted_by_recurse(citem, pitem, + weak)) + { + presult = true; + break; + } + } + iterate_end(pred_info); + if (!presult) + { + result = false; /* citem refutes nothing */ + break; + } + } + iterate_end(clause_info); + return result; + + case CLASS_ATOM: + + /* + * If B is a NOT-type clause, A R=> B if A => B's arg + * + * Same logic as for the AND-clause case above. + */ + not_arg = extract_not_arg(predicate); + if (not_arg && + predicate_implied_by_recurse(clause, not_arg, + false)) + return true; + + /* + * OR-clause R=> atom if each of A's items refutes B + */ + result = true; + iterate_begin(citem, clause, clause_info) + { + if (!predicate_refuted_by_recurse(citem, predicate, + weak)) + { + result = false; + break; + } + } + iterate_end(clause_info); + return result; + } + break; + + case CLASS_ATOM: + + /* + * If A is a strong NOT-clause, A R=> B if B => A's arg + * + * Since A is strong, we may assume A's arg is false (not just + * not-true). If B weakly implies A's arg, then B can be neither + * true nor null, so that strong refutation is proven. If B + * strongly implies A's arg, then B cannot be true, so that weak + * refutation is proven. + */ + not_arg = extract_strong_not_arg(clause); + if (not_arg && + predicate_implied_by_recurse(predicate, not_arg, + !weak)) + return true; + + switch (pclass) + { + case CLASS_AND: + + /* + * atom R=> AND-clause if A refutes any of B's items + */ + result = false; + iterate_begin(pitem, predicate, pred_info) + { + if (predicate_refuted_by_recurse(clause, pitem, + weak)) + { + result = true; + break; + } + } + iterate_end(pred_info); + return result; + + case CLASS_OR: + + /* + * atom R=> OR-clause if A refutes each of B's items + */ + result = true; + iterate_begin(pitem, predicate, pred_info) + { + if (!predicate_refuted_by_recurse(clause, pitem, + weak)) + { + result = false; + break; + } + } + iterate_end(pred_info); + return result; + + case CLASS_ATOM: + + /* + * If B is a NOT-type clause, A R=> B if A => B's arg + * + * Same logic as for the AND-clause case above. + */ + not_arg = extract_not_arg(predicate); + if (not_arg && + predicate_implied_by_recurse(clause, not_arg, + false)) + return true; + + /* + * atom R=> atom is the base case + */ + return + predicate_refuted_by_simple_clause((Expr *) predicate, + clause, + weak); + } + break; + } + + /* can't get here */ + elog(ERROR, "predicate_classify returned a bogus value"); + return false; +} + + +/* + * predicate_classify + * Classify an expression node as AND-type, OR-type, or neither (an atom). + * + * If the expression is classified as AND- or OR-type, then *info is filled + * in with the functions needed to iterate over its components. + * + * This function also implements enforcement of MAX_SAOP_ARRAY_SIZE: if a + * ScalarArrayOpExpr's array has too many elements, we just classify it as an + * atom. (This will result in its being passed as-is to the simple_clause + * functions, many of which will fail to prove anything about it.) Note that we + * cannot just stop after considering MAX_SAOP_ARRAY_SIZE elements; in general + * that would result in wrong proofs, rather than failing to prove anything. + */ +static PredClass +predicate_classify(Node *clause, PredIterInfo info) +{ + /* Caller should not pass us NULL, nor a RestrictInfo clause */ + Assert(clause != NULL); + Assert(!IsA(clause, RestrictInfo)); + + /* + * If we see a List, assume it's an implicit-AND list; this is the correct + * semantics for lists of RestrictInfo nodes. + */ + if (IsA(clause, List)) + { + info->startup_fn = list_startup_fn; + info->next_fn = list_next_fn; + info->cleanup_fn = list_cleanup_fn; + return CLASS_AND; + } + + /* Handle normal AND and OR boolean clauses */ + if (is_andclause(clause)) + { + info->startup_fn = boolexpr_startup_fn; + info->next_fn = list_next_fn; + info->cleanup_fn = list_cleanup_fn; + return CLASS_AND; + } + if (is_orclause(clause)) + { + info->startup_fn = boolexpr_startup_fn; + info->next_fn = list_next_fn; + info->cleanup_fn = list_cleanup_fn; + return CLASS_OR; + } + + /* Handle ScalarArrayOpExpr */ + if (IsA(clause, ScalarArrayOpExpr)) + { + ScalarArrayOpExpr *saop = (ScalarArrayOpExpr *) clause; + Node *arraynode = (Node *) lsecond(saop->args); + + /* + * We can break this down into an AND or OR structure, but only if we + * know how to iterate through expressions for the array's elements. + * We can do that if the array operand is a non-null constant or a + * simple ArrayExpr. + */ + if (arraynode && IsA(arraynode, Const) && + !((Const *) arraynode)->constisnull) + { + ArrayType *arrayval; + int nelems; + + arrayval = DatumGetArrayTypeP(((Const *) arraynode)->constvalue); + nelems = ArrayGetNItems(ARR_NDIM(arrayval), ARR_DIMS(arrayval)); + if (nelems <= MAX_SAOP_ARRAY_SIZE) + { + info->startup_fn = arrayconst_startup_fn; + info->next_fn = arrayconst_next_fn; + info->cleanup_fn = arrayconst_cleanup_fn; + return saop->useOr ? CLASS_OR : CLASS_AND; + } + } + else if (arraynode && IsA(arraynode, ArrayExpr) && + !((ArrayExpr *) arraynode)->multidims && + list_length(((ArrayExpr *) arraynode)->elements) <= MAX_SAOP_ARRAY_SIZE) + { + info->startup_fn = arrayexpr_startup_fn; + info->next_fn = arrayexpr_next_fn; + info->cleanup_fn = arrayexpr_cleanup_fn; + return saop->useOr ? CLASS_OR : CLASS_AND; + } + } + + /* None of the above, so it's an atom */ + return CLASS_ATOM; +} + +/* + * PredIterInfo routines for iterating over regular Lists. The iteration + * state variable is the next ListCell to visit. + */ +static void +list_startup_fn(Node *clause, PredIterInfo info) +{ + info->state_list = (List *) clause; + info->state = (void *) list_head(info->state_list); +} + +static Node * +list_next_fn(PredIterInfo info) +{ + ListCell *l = (ListCell *) info->state; + Node *n; + + if (l == NULL) + return NULL; + n = lfirst(l); + info->state = (void *) lnext(info->state_list, l); + return n; +} + +static void +list_cleanup_fn(PredIterInfo info) +{ + /* Nothing to clean up */ +} + +/* + * BoolExpr needs its own startup function, but can use list_next_fn and + * list_cleanup_fn. + */ +static void +boolexpr_startup_fn(Node *clause, PredIterInfo info) +{ + info->state_list = ((BoolExpr *) clause)->args; + info->state = (void *) list_head(info->state_list); +} + +/* + * PredIterInfo routines for iterating over a ScalarArrayOpExpr with a + * constant array operand. + */ +typedef struct +{ + OpExpr opexpr; + Const constexpr; + int next_elem; + int num_elems; + Datum *elem_values; + bool *elem_nulls; +} ArrayConstIterState; + +static void +arrayconst_startup_fn(Node *clause, PredIterInfo info) +{ + ScalarArrayOpExpr *saop = (ScalarArrayOpExpr *) clause; + ArrayConstIterState *state; + Const *arrayconst; + ArrayType *arrayval; + int16 elmlen; + bool elmbyval; + char elmalign; + + /* Create working state struct */ + state = (ArrayConstIterState *) palloc(sizeof(ArrayConstIterState)); + info->state = (void *) state; + + /* Deconstruct the array literal */ + arrayconst = (Const *) lsecond(saop->args); + arrayval = DatumGetArrayTypeP(arrayconst->constvalue); + get_typlenbyvalalign(ARR_ELEMTYPE(arrayval), + &elmlen, &elmbyval, &elmalign); + deconstruct_array(arrayval, + ARR_ELEMTYPE(arrayval), + elmlen, elmbyval, elmalign, + &state->elem_values, &state->elem_nulls, + &state->num_elems); + + /* Set up a dummy OpExpr to return as the per-item node */ + state->opexpr.xpr.type = T_OpExpr; + state->opexpr.opno = saop->opno; + state->opexpr.opfuncid = saop->opfuncid; + state->opexpr.opresulttype = BOOLOID; + state->opexpr.opretset = false; + state->opexpr.opcollid = InvalidOid; + state->opexpr.inputcollid = saop->inputcollid; + state->opexpr.args = list_copy(saop->args); + + /* Set up a dummy Const node to hold the per-element values */ + state->constexpr.xpr.type = T_Const; + state->constexpr.consttype = ARR_ELEMTYPE(arrayval); + state->constexpr.consttypmod = -1; + state->constexpr.constcollid = arrayconst->constcollid; + state->constexpr.constlen = elmlen; + state->constexpr.constbyval = elmbyval; + lsecond(state->opexpr.args) = &state->constexpr; + + /* Initialize iteration state */ + state->next_elem = 0; +} + +static Node * +arrayconst_next_fn(PredIterInfo info) +{ + ArrayConstIterState *state = (ArrayConstIterState *) info->state; + + if (state->next_elem >= state->num_elems) + return NULL; + state->constexpr.constvalue = state->elem_values[state->next_elem]; + state->constexpr.constisnull = state->elem_nulls[state->next_elem]; + state->next_elem++; + return (Node *) &(state->opexpr); +} + +static void +arrayconst_cleanup_fn(PredIterInfo info) +{ + ArrayConstIterState *state = (ArrayConstIterState *) info->state; + + pfree(state->elem_values); + pfree(state->elem_nulls); + list_free(state->opexpr.args); + pfree(state); +} + +/* + * PredIterInfo routines for iterating over a ScalarArrayOpExpr with a + * one-dimensional ArrayExpr array operand. + */ +typedef struct +{ + OpExpr opexpr; + ListCell *next; +} ArrayExprIterState; + +static void +arrayexpr_startup_fn(Node *clause, PredIterInfo info) +{ + ScalarArrayOpExpr *saop = (ScalarArrayOpExpr *) clause; + ArrayExprIterState *state; + ArrayExpr *arrayexpr; + + /* Create working state struct */ + state = (ArrayExprIterState *) palloc(sizeof(ArrayExprIterState)); + info->state = (void *) state; + + /* Set up a dummy OpExpr to return as the per-item node */ + state->opexpr.xpr.type = T_OpExpr; + state->opexpr.opno = saop->opno; + state->opexpr.opfuncid = saop->opfuncid; + state->opexpr.opresulttype = BOOLOID; + state->opexpr.opretset = false; + state->opexpr.opcollid = InvalidOid; + state->opexpr.inputcollid = saop->inputcollid; + state->opexpr.args = list_copy(saop->args); + + /* Initialize iteration variable to first member of ArrayExpr */ + arrayexpr = (ArrayExpr *) lsecond(saop->args); + info->state_list = arrayexpr->elements; + state->next = list_head(arrayexpr->elements); +} + +static Node * +arrayexpr_next_fn(PredIterInfo info) +{ + ArrayExprIterState *state = (ArrayExprIterState *) info->state; + + if (state->next == NULL) + return NULL; + lsecond(state->opexpr.args) = lfirst(state->next); + state->next = lnext(info->state_list, state->next); + return (Node *) &(state->opexpr); +} + +static void +arrayexpr_cleanup_fn(PredIterInfo info) +{ + ArrayExprIterState *state = (ArrayExprIterState *) info->state; + + list_free(state->opexpr.args); + pfree(state); +} + + +/*---------- + * predicate_implied_by_simple_clause + * Does the predicate implication test for a "simple clause" predicate + * and a "simple clause" restriction. + * + * We return true if able to prove the implication, false if not. + * + * We have three strategies for determining whether one simple clause + * implies another: + * + * A simple and general way is to see if they are equal(); this works for any + * kind of expression, and for either implication definition. (Actually, + * there is an implied assumption that the functions in the expression are + * immutable --- but this was checked for the predicate by the caller.) + * + * If the predicate is of the form "foo IS NOT NULL", and we are considering + * strong implication, we can conclude that the predicate is implied if the + * clause is strict for "foo", i.e., it must yield false or NULL when "foo" + * is NULL. In that case truth of the clause ensures that "foo" isn't NULL. + * (Again, this is a safe conclusion because "foo" must be immutable.) + * This doesn't work for weak implication, though. + * + * Finally, if both clauses are binary operator expressions, we may be able + * to prove something using the system's knowledge about operators; those + * proof rules are encapsulated in operator_predicate_proof(). + *---------- + */ +static bool +predicate_implied_by_simple_clause(Expr *predicate, Node *clause, + bool weak) +{ + /* Allow interrupting long proof attempts */ + CHECK_FOR_INTERRUPTS(); + + /* First try the equal() test */ + if (equal((Node *) predicate, clause)) + return true; + + /* Next try the IS NOT NULL case */ + if (!weak && + predicate && IsA(predicate, NullTest)) + { + NullTest *ntest = (NullTest *) predicate; + + /* row IS NOT NULL does not act in the simple way we have in mind */ + if (ntest->nulltesttype == IS_NOT_NULL && + !ntest->argisrow) + { + /* strictness of clause for foo implies foo IS NOT NULL */ + if (clause_is_strict_for(clause, (Node *) ntest->arg, true)) + return true; + } + return false; /* we can't succeed below... */ + } + + /* Else try operator-related knowledge */ + return operator_predicate_proof(predicate, clause, false, weak); +} + +/*---------- + * predicate_refuted_by_simple_clause + * Does the predicate refutation test for a "simple clause" predicate + * and a "simple clause" restriction. + * + * We return true if able to prove the refutation, false if not. + * + * Unlike the implication case, checking for equal() clauses isn't helpful. + * But relation_excluded_by_constraints() checks for self-contradictions in a + * list of clauses, so that we may get here with predicate and clause being + * actually pointer-equal, and that is worth eliminating quickly. + * + * When the predicate is of the form "foo IS NULL", we can conclude that + * the predicate is refuted if the clause is strict for "foo" (see notes for + * implication case), or is "foo IS NOT NULL". That works for either strong + * or weak refutation. + * + * A clause "foo IS NULL" refutes a predicate "foo IS NOT NULL" in all cases. + * If we are considering weak refutation, it also refutes a predicate that + * is strict for "foo", since then the predicate must yield false or NULL + * (and since "foo" appears in the predicate, it's known immutable). + * + * (The main motivation for covering these IS [NOT] NULL cases is to support + * using IS NULL/IS NOT NULL as partition-defining constraints.) + * + * Finally, if both clauses are binary operator expressions, we may be able + * to prove something using the system's knowledge about operators; those + * proof rules are encapsulated in operator_predicate_proof(). + *---------- + */ +static bool +predicate_refuted_by_simple_clause(Expr *predicate, Node *clause, + bool weak) +{ + /* Allow interrupting long proof attempts */ + CHECK_FOR_INTERRUPTS(); + + /* A simple clause can't refute itself */ + /* Worth checking because of relation_excluded_by_constraints() */ + if ((Node *) predicate == clause) + return false; + + /* Try the predicate-IS-NULL case */ + if (predicate && IsA(predicate, NullTest) && + ((NullTest *) predicate)->nulltesttype == IS_NULL) + { + Expr *isnullarg = ((NullTest *) predicate)->arg; + + /* row IS NULL does not act in the simple way we have in mind */ + if (((NullTest *) predicate)->argisrow) + return false; + + /* strictness of clause for foo refutes foo IS NULL */ + if (clause_is_strict_for(clause, (Node *) isnullarg, true)) + return true; + + /* foo IS NOT NULL refutes foo IS NULL */ + if (clause && IsA(clause, NullTest) && + ((NullTest *) clause)->nulltesttype == IS_NOT_NULL && + !((NullTest *) clause)->argisrow && + equal(((NullTest *) clause)->arg, isnullarg)) + return true; + + return false; /* we can't succeed below... */ + } + + /* Try the clause-IS-NULL case */ + if (clause && IsA(clause, NullTest) && + ((NullTest *) clause)->nulltesttype == IS_NULL) + { + Expr *isnullarg = ((NullTest *) clause)->arg; + + /* row IS NULL does not act in the simple way we have in mind */ + if (((NullTest *) clause)->argisrow) + return false; + + /* foo IS NULL refutes foo IS NOT NULL */ + if (predicate && IsA(predicate, NullTest) && + ((NullTest *) predicate)->nulltesttype == IS_NOT_NULL && + !((NullTest *) predicate)->argisrow && + equal(((NullTest *) predicate)->arg, isnullarg)) + return true; + + /* foo IS NULL weakly refutes any predicate that is strict for foo */ + if (weak && + clause_is_strict_for((Node *) predicate, (Node *) isnullarg, true)) + return true; + + return false; /* we can't succeed below... */ + } + + /* Else try operator-related knowledge */ + return operator_predicate_proof(predicate, clause, true, weak); +} + + +/* + * If clause asserts the non-truth of a subclause, return that subclause; + * otherwise return NULL. + */ +static Node * +extract_not_arg(Node *clause) +{ + if (clause == NULL) + return NULL; + if (IsA(clause, BoolExpr)) + { + BoolExpr *bexpr = (BoolExpr *) clause; + + if (bexpr->boolop == NOT_EXPR) + return (Node *) linitial(bexpr->args); + } + else if (IsA(clause, BooleanTest)) + { + BooleanTest *btest = (BooleanTest *) clause; + + if (btest->booltesttype == IS_NOT_TRUE || + btest->booltesttype == IS_FALSE || + btest->booltesttype == IS_UNKNOWN) + return (Node *) btest->arg; + } + return NULL; +} + +/* + * If clause asserts the falsity of a subclause, return that subclause; + * otherwise return NULL. + */ +static Node * +extract_strong_not_arg(Node *clause) +{ + if (clause == NULL) + return NULL; + if (IsA(clause, BoolExpr)) + { + BoolExpr *bexpr = (BoolExpr *) clause; + + if (bexpr->boolop == NOT_EXPR) + return (Node *) linitial(bexpr->args); + } + else if (IsA(clause, BooleanTest)) + { + BooleanTest *btest = (BooleanTest *) clause; + + if (btest->booltesttype == IS_FALSE) + return (Node *) btest->arg; + } + return NULL; +} + + +/* + * Can we prove that "clause" returns NULL (or FALSE) if "subexpr" is + * assumed to yield NULL? + * + * In most places in the planner, "strictness" refers to a guarantee that + * an expression yields NULL output for a NULL input, and that's mostly what + * we're looking for here. However, at top level where the clause is known + * to yield boolean, it may be sufficient to prove that it cannot return TRUE + * when "subexpr" is NULL. The caller should pass allow_false = true when + * this weaker property is acceptable. (When this function recurses + * internally, we pass down allow_false = false since we need to prove actual + * nullness of the subexpression.) + * + * We assume that the caller checked that least one of the input expressions + * is immutable. All of the proof rules here involve matching "subexpr" to + * some portion of "clause", so that this allows assuming that "subexpr" is + * immutable without a separate check. + * + * The base case is that clause and subexpr are equal(). + * + * We can also report success if the subexpr appears as a subexpression + * of "clause" in a place where it'd force nullness of the overall result. + */ +static bool +clause_is_strict_for(Node *clause, Node *subexpr, bool allow_false) +{ + ListCell *lc; + + /* safety checks */ + if (clause == NULL || subexpr == NULL) + return false; + + /* + * Look through any RelabelType nodes, so that we can match, say, + * varcharcol with lower(varcharcol::text). (In general we could recurse + * through any nullness-preserving, immutable operation.) We should not + * see stacked RelabelTypes here. + */ + if (IsA(clause, RelabelType)) + clause = (Node *) ((RelabelType *) clause)->arg; + if (IsA(subexpr, RelabelType)) + subexpr = (Node *) ((RelabelType *) subexpr)->arg; + + /* Base case */ + if (equal(clause, subexpr)) + return true; + + /* + * If we have a strict operator or function, a NULL result is guaranteed + * if any input is forced NULL by subexpr. This is OK even if the op or + * func isn't immutable, since it won't even be called on NULL input. + */ + if (is_opclause(clause) && + op_strict(((OpExpr *) clause)->opno)) + { + foreach(lc, ((OpExpr *) clause)->args) + { + if (clause_is_strict_for((Node *) lfirst(lc), subexpr, false)) + return true; + } + return false; + } + if (is_funcclause(clause) && + func_strict(((FuncExpr *) clause)->funcid)) + { + foreach(lc, ((FuncExpr *) clause)->args) + { + if (clause_is_strict_for((Node *) lfirst(lc), subexpr, false)) + return true; + } + return false; + } + + /* + * CoerceViaIO is strict (whether or not the I/O functions it calls are). + * Likewise, ArrayCoerceExpr is strict for its array argument (regardless + * of what the per-element expression is), ConvertRowtypeExpr is strict at + * the row level, and CoerceToDomain is strict too. These are worth + * checking mainly because it saves us having to explain to users why some + * type coercions are known strict and others aren't. + */ + if (IsA(clause, CoerceViaIO)) + return clause_is_strict_for((Node *) ((CoerceViaIO *) clause)->arg, + subexpr, false); + if (IsA(clause, ArrayCoerceExpr)) + return clause_is_strict_for((Node *) ((ArrayCoerceExpr *) clause)->arg, + subexpr, false); + if (IsA(clause, ConvertRowtypeExpr)) + return clause_is_strict_for((Node *) ((ConvertRowtypeExpr *) clause)->arg, + subexpr, false); + if (IsA(clause, CoerceToDomain)) + return clause_is_strict_for((Node *) ((CoerceToDomain *) clause)->arg, + subexpr, false); + + /* + * ScalarArrayOpExpr is a special case. Note that we'd only reach here + * with a ScalarArrayOpExpr clause if we failed to deconstruct it into an + * AND or OR tree, as for example if it has too many array elements. + */ + if (IsA(clause, ScalarArrayOpExpr)) + { + ScalarArrayOpExpr *saop = (ScalarArrayOpExpr *) clause; + Node *scalarnode = (Node *) linitial(saop->args); + Node *arraynode = (Node *) lsecond(saop->args); + + /* + * If we can prove the scalar input to be null, and the operator is + * strict, then the SAOP result has to be null --- unless the array is + * empty. For an empty array, we'd get either false (for ANY) or true + * (for ALL). So if allow_false = true then the proof succeeds anyway + * for the ANY case; otherwise we can only make the proof if we can + * prove the array non-empty. + */ + if (clause_is_strict_for(scalarnode, subexpr, false) && + op_strict(saop->opno)) + { + int nelems = 0; + + if (allow_false && saop->useOr) + return true; /* can succeed even if array is empty */ + + if (arraynode && IsA(arraynode, Const)) + { + Const *arrayconst = (Const *) arraynode; + ArrayType *arrval; + + /* + * If array is constant NULL then we can succeed, as in the + * case below. + */ + if (arrayconst->constisnull) + return true; + + /* Otherwise, we can compute the number of elements. */ + arrval = DatumGetArrayTypeP(arrayconst->constvalue); + nelems = ArrayGetNItems(ARR_NDIM(arrval), ARR_DIMS(arrval)); + } + else if (arraynode && IsA(arraynode, ArrayExpr) && + !((ArrayExpr *) arraynode)->multidims) + { + /* + * We can also reliably count the number of array elements if + * the input is a non-multidim ARRAY[] expression. + */ + nelems = list_length(((ArrayExpr *) arraynode)->elements); + } + + /* Proof succeeds if array is definitely non-empty */ + if (nelems > 0) + return true; + } + + /* + * If we can prove the array input to be null, the proof succeeds in + * all cases, since ScalarArrayOpExpr will always return NULL for a + * NULL array. Otherwise, we're done here. + */ + return clause_is_strict_for(arraynode, subexpr, false); + } + + /* + * When recursing into an expression, we might find a NULL constant. + * That's certainly NULL, whether it matches subexpr or not. + */ + if (IsA(clause, Const)) + return ((Const *) clause)->constisnull; + + return false; +} + + +/* + * Define "operator implication tables" for btree operators ("strategies"), + * and similar tables for refutation. + * + * The strategy numbers defined by btree indexes (see access/stratnum.h) are: + * 1 < 2 <= 3 = 4 >= 5 > + * and in addition we use 6 to represent <>. <> is not a btree-indexable + * operator, but we assume here that if an equality operator of a btree + * opfamily has a negator operator, the negator behaves as <> for the opfamily. + * (This convention is also known to get_op_btree_interpretation().) + * + * BT_implies_table[] and BT_refutes_table[] are used for cases where we have + * two identical subexpressions and we want to know whether one operator + * expression implies or refutes the other. That is, if the "clause" is + * EXPR1 clause_op EXPR2 and the "predicate" is EXPR1 pred_op EXPR2 for the + * same two (immutable) subexpressions: + * BT_implies_table[clause_op-1][pred_op-1] + * is true if the clause implies the predicate + * BT_refutes_table[clause_op-1][pred_op-1] + * is true if the clause refutes the predicate + * where clause_op and pred_op are strategy numbers (from 1 to 6) in the + * same btree opfamily. For example, "x < y" implies "x <= y" and refutes + * "x > y". + * + * BT_implic_table[] and BT_refute_table[] are used where we have two + * constants that we need to compare. The interpretation of: + * + * test_op = BT_implic_table[clause_op-1][pred_op-1] + * + * where test_op, clause_op and pred_op are strategy numbers (from 1 to 6) + * of btree operators, is as follows: + * + * If you know, for some EXPR, that "EXPR clause_op CONST1" is true, and you + * want to determine whether "EXPR pred_op CONST2" must also be true, then + * you can use "CONST2 test_op CONST1" as a test. If this test returns true, + * then the predicate expression must be true; if the test returns false, + * then the predicate expression may be false. + * + * For example, if clause is "Quantity > 10" and pred is "Quantity > 5" + * then we test "5 <= 10" which evals to true, so clause implies pred. + * + * Similarly, the interpretation of a BT_refute_table entry is: + * + * If you know, for some EXPR, that "EXPR clause_op CONST1" is true, and you + * want to determine whether "EXPR pred_op CONST2" must be false, then + * you can use "CONST2 test_op CONST1" as a test. If this test returns true, + * then the predicate expression must be false; if the test returns false, + * then the predicate expression may be true. + * + * For example, if clause is "Quantity > 10" and pred is "Quantity < 5" + * then we test "5 <= 10" which evals to true, so clause refutes pred. + * + * An entry where test_op == 0 means the implication cannot be determined. + */ + +#define BTLT BTLessStrategyNumber +#define BTLE BTLessEqualStrategyNumber +#define BTEQ BTEqualStrategyNumber +#define BTGE BTGreaterEqualStrategyNumber +#define BTGT BTGreaterStrategyNumber +#define BTNE ROWCOMPARE_NE + +/* We use "none" for 0/false to make the tables align nicely */ +#define none 0 + +static const bool BT_implies_table[6][6] = { +/* + * The predicate operator: + * LT LE EQ GE GT NE + */ + {true, true, none, none, none, true}, /* LT */ + {none, true, none, none, none, none}, /* LE */ + {none, true, true, true, none, none}, /* EQ */ + {none, none, none, true, none, none}, /* GE */ + {none, none, none, true, true, true}, /* GT */ + {none, none, none, none, none, true} /* NE */ +}; + +static const bool BT_refutes_table[6][6] = { +/* + * The predicate operator: + * LT LE EQ GE GT NE + */ + {none, none, true, true, true, none}, /* LT */ + {none, none, none, none, true, none}, /* LE */ + {true, none, none, none, true, true}, /* EQ */ + {true, none, none, none, none, none}, /* GE */ + {true, true, true, none, none, none}, /* GT */ + {none, none, true, none, none, none} /* NE */ +}; + +static const StrategyNumber BT_implic_table[6][6] = { +/* + * The predicate operator: + * LT LE EQ GE GT NE + */ + {BTGE, BTGE, none, none, none, BTGE}, /* LT */ + {BTGT, BTGE, none, none, none, BTGT}, /* LE */ + {BTGT, BTGE, BTEQ, BTLE, BTLT, BTNE}, /* EQ */ + {none, none, none, BTLE, BTLT, BTLT}, /* GE */ + {none, none, none, BTLE, BTLE, BTLE}, /* GT */ + {none, none, none, none, none, BTEQ} /* NE */ +}; + +static const StrategyNumber BT_refute_table[6][6] = { +/* + * The predicate operator: + * LT LE EQ GE GT NE + */ + {none, none, BTGE, BTGE, BTGE, none}, /* LT */ + {none, none, BTGT, BTGT, BTGE, none}, /* LE */ + {BTLE, BTLT, BTNE, BTGT, BTGE, BTEQ}, /* EQ */ + {BTLE, BTLT, BTLT, none, none, none}, /* GE */ + {BTLE, BTLE, BTLE, none, none, none}, /* GT */ + {none, none, BTEQ, none, none, none} /* NE */ +}; + + +/* + * operator_predicate_proof + * Does the predicate implication or refutation test for a "simple clause" + * predicate and a "simple clause" restriction, when both are operator + * clauses using related operators and identical input expressions. + * + * When refute_it == false, we want to prove the predicate true; + * when refute_it == true, we want to prove the predicate false. + * (There is enough common code to justify handling these two cases + * in one routine.) We return true if able to make the proof, false + * if not able to prove it. + * + * We mostly need not distinguish strong vs. weak implication/refutation here. + * This depends on the assumption that a pair of related operators (i.e., + * commutators, negators, or btree opfamily siblings) will not return one NULL + * and one non-NULL result for the same inputs. Then, for the proof types + * where we start with an assumption of truth of the clause, the predicate + * operator could not return NULL either, so it doesn't matter whether we are + * trying to make a strong or weak proof. For weak implication, it could be + * that the clause operator returned NULL, but then the predicate operator + * would as well, so that the weak implication still holds. This argument + * doesn't apply in the case where we are considering two different constant + * values, since then the operators aren't being given identical inputs. But + * we only support that for btree operators, for which we can assume that all + * non-null inputs result in non-null outputs, so that it doesn't matter which + * two non-null constants we consider. If either constant is NULL, we have + * to think harder, but sometimes the proof still works, as explained below. + * + * We can make proofs involving several expression forms (here "foo" and "bar" + * represent subexpressions that are identical according to equal()): + * "foo op1 bar" refutes "foo op2 bar" if op1 is op2's negator + * "foo op1 bar" implies "bar op2 foo" if op1 is op2's commutator + * "foo op1 bar" refutes "bar op2 foo" if op1 is negator of op2's commutator + * "foo op1 bar" can imply/refute "foo op2 bar" based on btree semantics + * "foo op1 bar" can imply/refute "bar op2 foo" based on btree semantics + * "foo op1 const1" can imply/refute "foo op2 const2" based on btree semantics + * + * For the last three cases, op1 and op2 have to be members of the same btree + * operator family. When both subexpressions are identical, the idea is that, + * for instance, x < y implies x <= y, independently of exactly what x and y + * are. If we have two different constants compared to the same expression + * foo, we have to execute a comparison between the two constant values + * in order to determine the result; for instance, foo < c1 implies foo < c2 + * if c1 <= c2. We assume it's safe to compare the constants at plan time + * if the comparison operator is immutable. + * + * Note: all the operators and subexpressions have to be immutable for the + * proof to be safe. We assume the predicate expression is entirely immutable, + * so no explicit check on the subexpressions is needed here, but in some + * cases we need an extra check of operator immutability. In particular, + * btree opfamilies can contain cross-type operators that are merely stable, + * and we dare not make deductions with those. + */ +static bool +operator_predicate_proof(Expr *predicate, Node *clause, + bool refute_it, bool weak) +{ + OpExpr *pred_opexpr, + *clause_opexpr; + Oid pred_collation, + clause_collation; + Oid pred_op, + clause_op, + test_op; + Node *pred_leftop, + *pred_rightop, + *clause_leftop, + *clause_rightop; + Const *pred_const, + *clause_const; + Expr *test_expr; + ExprState *test_exprstate; + Datum test_result; + bool isNull; + EState *estate; + MemoryContext oldcontext; + + /* + * Both expressions must be binary opclauses, else we can't do anything. + * + * Note: in future we might extend this logic to other operator-based + * constructs such as DistinctExpr. But the planner isn't very smart + * about DistinctExpr in general, and this probably isn't the first place + * to fix if you want to improve that. + */ + if (!is_opclause(predicate)) + return false; + pred_opexpr = (OpExpr *) predicate; + if (list_length(pred_opexpr->args) != 2) + return false; + if (!is_opclause(clause)) + return false; + clause_opexpr = (OpExpr *) clause; + if (list_length(clause_opexpr->args) != 2) + return false; + + /* + * If they're marked with different collations then we can't do anything. + * This is a cheap test so let's get it out of the way early. + */ + pred_collation = pred_opexpr->inputcollid; + clause_collation = clause_opexpr->inputcollid; + if (pred_collation != clause_collation) + return false; + + /* Grab the operator OIDs now too. We may commute these below. */ + pred_op = pred_opexpr->opno; + clause_op = clause_opexpr->opno; + + /* + * We have to match up at least one pair of input expressions. + */ + pred_leftop = (Node *) linitial(pred_opexpr->args); + pred_rightop = (Node *) lsecond(pred_opexpr->args); + clause_leftop = (Node *) linitial(clause_opexpr->args); + clause_rightop = (Node *) lsecond(clause_opexpr->args); + + if (equal(pred_leftop, clause_leftop)) + { + if (equal(pred_rightop, clause_rightop)) + { + /* We have x op1 y and x op2 y */ + return operator_same_subexprs_proof(pred_op, clause_op, refute_it); + } + else + { + /* Fail unless rightops are both Consts */ + if (pred_rightop == NULL || !IsA(pred_rightop, Const)) + return false; + pred_const = (Const *) pred_rightop; + if (clause_rightop == NULL || !IsA(clause_rightop, Const)) + return false; + clause_const = (Const *) clause_rightop; + } + } + else if (equal(pred_rightop, clause_rightop)) + { + /* Fail unless leftops are both Consts */ + if (pred_leftop == NULL || !IsA(pred_leftop, Const)) + return false; + pred_const = (Const *) pred_leftop; + if (clause_leftop == NULL || !IsA(clause_leftop, Const)) + return false; + clause_const = (Const *) clause_leftop; + /* Commute both operators so we can assume Consts are on the right */ + pred_op = get_commutator(pred_op); + if (!OidIsValid(pred_op)) + return false; + clause_op = get_commutator(clause_op); + if (!OidIsValid(clause_op)) + return false; + } + else if (equal(pred_leftop, clause_rightop)) + { + if (equal(pred_rightop, clause_leftop)) + { + /* We have x op1 y and y op2 x */ + /* Commute pred_op that we can treat this like a straight match */ + pred_op = get_commutator(pred_op); + if (!OidIsValid(pred_op)) + return false; + return operator_same_subexprs_proof(pred_op, clause_op, refute_it); + } + else + { + /* Fail unless pred_rightop/clause_leftop are both Consts */ + if (pred_rightop == NULL || !IsA(pred_rightop, Const)) + return false; + pred_const = (Const *) pred_rightop; + if (clause_leftop == NULL || !IsA(clause_leftop, Const)) + return false; + clause_const = (Const *) clause_leftop; + /* Commute clause_op so we can assume Consts are on the right */ + clause_op = get_commutator(clause_op); + if (!OidIsValid(clause_op)) + return false; + } + } + else if (equal(pred_rightop, clause_leftop)) + { + /* Fail unless pred_leftop/clause_rightop are both Consts */ + if (pred_leftop == NULL || !IsA(pred_leftop, Const)) + return false; + pred_const = (Const *) pred_leftop; + if (clause_rightop == NULL || !IsA(clause_rightop, Const)) + return false; + clause_const = (Const *) clause_rightop; + /* Commute pred_op so we can assume Consts are on the right */ + pred_op = get_commutator(pred_op); + if (!OidIsValid(pred_op)) + return false; + } + else + { + /* Failed to match up any of the subexpressions, so we lose */ + return false; + } + + /* + * We have two identical subexpressions, and two other subexpressions that + * are not identical but are both Consts; and we have commuted the + * operators if necessary so that the Consts are on the right. We'll need + * to compare the Consts' values. If either is NULL, we can't do that, so + * usually the proof fails ... but in some cases we can claim success. + */ + if (clause_const->constisnull) + { + /* If clause_op isn't strict, we can't prove anything */ + if (!op_strict(clause_op)) + return false; + + /* + * At this point we know that the clause returns NULL. For proof + * types that assume truth of the clause, this means the proof is + * vacuously true (a/k/a "false implies anything"). That's all proof + * types except weak implication. + */ + if (!(weak && !refute_it)) + return true; + + /* + * For weak implication, it's still possible for the proof to succeed, + * if the predicate can also be proven NULL. In that case we've got + * NULL => NULL which is valid for this proof type. + */ + if (pred_const->constisnull && op_strict(pred_op)) + return true; + /* Else the proof fails */ + return false; + } + if (pred_const->constisnull) + { + /* + * If the pred_op is strict, we know the predicate yields NULL, which + * means the proof succeeds for either weak implication or weak + * refutation. + */ + if (weak && op_strict(pred_op)) + return true; + /* Else the proof fails */ + return false; + } + + /* + * Lookup the constant-comparison operator using the system catalogs and + * the operator implication tables. + */ + test_op = get_btree_test_op(pred_op, clause_op, refute_it); + + if (!OidIsValid(test_op)) + { + /* couldn't find a suitable comparison operator */ + return false; + } + + /* + * Evaluate the test. For this we need an EState. + */ + estate = CreateExecutorState(); + + /* We can use the estate's working context to avoid memory leaks. */ + oldcontext = MemoryContextSwitchTo(estate->es_query_cxt); + + /* Build expression tree */ + test_expr = make_opclause(test_op, + BOOLOID, + false, + (Expr *) pred_const, + (Expr *) clause_const, + InvalidOid, + pred_collation); + + /* Fill in opfuncids */ + fix_opfuncids((Node *) test_expr); + + /* Prepare it for execution */ + test_exprstate = ExecInitExpr(test_expr, NULL); + + /* And execute it. */ + test_result = ExecEvalExprSwitchContext(test_exprstate, + GetPerTupleExprContext(estate), + &isNull); + + /* Get back to outer memory context */ + MemoryContextSwitchTo(oldcontext); + + /* Release all the junk we just created */ + FreeExecutorState(estate); + + if (isNull) + { + /* Treat a null result as non-proof ... but it's a tad fishy ... */ + elog(DEBUG2, "null predicate test result"); + return false; + } + return DatumGetBool(test_result); +} + + +/* + * operator_same_subexprs_proof + * Assuming that EXPR1 clause_op EXPR2 is true, try to prove or refute + * EXPR1 pred_op EXPR2. + * + * Return true if able to make the proof, false if not able to prove it. + */ +static bool +operator_same_subexprs_proof(Oid pred_op, Oid clause_op, bool refute_it) +{ + /* + * A simple and general rule is that the predicate is proven if clause_op + * and pred_op are the same, or refuted if they are each other's negators. + * We need not check immutability since the pred_op is already known + * immutable. (Actually, by this point we may have the commutator of a + * known-immutable pred_op, but that should certainly be immutable too. + * Likewise we don't worry whether the pred_op's negator is immutable.) + * + * Note: the "same" case won't get here if we actually had EXPR1 clause_op + * EXPR2 and EXPR1 pred_op EXPR2, because the overall-expression-equality + * test in predicate_implied_by_simple_clause would have caught it. But + * we can see the same operator after having commuted the pred_op. + */ + if (refute_it) + { + if (get_negator(pred_op) == clause_op) + return true; + } + else + { + if (pred_op == clause_op) + return true; + } + + /* + * Otherwise, see if we can determine the implication by finding the + * operators' relationship via some btree opfamily. + */ + return operator_same_subexprs_lookup(pred_op, clause_op, refute_it); +} + + +/* + * We use a lookaside table to cache the result of btree proof operator + * lookups, since the actual lookup is pretty expensive and doesn't change + * for any given pair of operators (at least as long as pg_amop doesn't + * change). A single hash entry stores both implication and refutation + * results for a given pair of operators; but note we may have determined + * only one of those sets of results as yet. + */ +typedef struct OprProofCacheKey +{ + Oid pred_op; /* predicate operator */ + Oid clause_op; /* clause operator */ +} OprProofCacheKey; + +typedef struct OprProofCacheEntry +{ + /* the hash lookup key MUST BE FIRST */ + OprProofCacheKey key; + + bool have_implic; /* do we know the implication result? */ + bool have_refute; /* do we know the refutation result? */ + bool same_subexprs_implies; /* X clause_op Y implies X pred_op Y? */ + bool same_subexprs_refutes; /* X clause_op Y refutes X pred_op Y? */ + Oid implic_test_op; /* OID of the test operator, or 0 if none */ + Oid refute_test_op; /* OID of the test operator, or 0 if none */ +} OprProofCacheEntry; + +static HTAB *OprProofCacheHash = NULL; + + +/* + * lookup_proof_cache + * Get, and fill in if necessary, the appropriate cache entry. + */ +static OprProofCacheEntry * +lookup_proof_cache(Oid pred_op, Oid clause_op, bool refute_it) +{ + OprProofCacheKey key; + OprProofCacheEntry *cache_entry; + bool cfound; + bool same_subexprs = false; + Oid test_op = InvalidOid; + bool found = false; + List *pred_op_infos, + *clause_op_infos; + ListCell *lcp, + *lcc; + + /* + * Find or make a cache entry for this pair of operators. + */ + if (OprProofCacheHash == NULL) + { + /* First time through: initialize the hash table */ + HASHCTL ctl; + + ctl.keysize = sizeof(OprProofCacheKey); + ctl.entrysize = sizeof(OprProofCacheEntry); + OprProofCacheHash = hash_create("Btree proof lookup cache", 256, + &ctl, HASH_ELEM | HASH_BLOBS); + + /* Arrange to flush cache on pg_amop changes */ + CacheRegisterSyscacheCallback(AMOPOPID, + InvalidateOprProofCacheCallBack, + (Datum) 0); + } + + key.pred_op = pred_op; + key.clause_op = clause_op; + cache_entry = (OprProofCacheEntry *) hash_search(OprProofCacheHash, + (void *) &key, + HASH_ENTER, &cfound); + if (!cfound) + { + /* new cache entry, set it invalid */ + cache_entry->have_implic = false; + cache_entry->have_refute = false; + } + else + { + /* pre-existing cache entry, see if we know the answer yet */ + if (refute_it ? cache_entry->have_refute : cache_entry->have_implic) + return cache_entry; + } + + /* + * Try to find a btree opfamily containing the given operators. + * + * We must find a btree opfamily that contains both operators, else the + * implication can't be determined. Also, the opfamily must contain a + * suitable test operator taking the operators' righthand datatypes. + * + * If there are multiple matching opfamilies, assume we can use any one to + * determine the logical relationship of the two operators and the correct + * corresponding test operator. This should work for any logically + * consistent opfamilies. + * + * Note that we can determine the operators' relationship for + * same-subexprs cases even from an opfamily that lacks a usable test + * operator. This can happen in cases with incomplete sets of cross-type + * comparison operators. + */ + clause_op_infos = get_op_btree_interpretation(clause_op); + if (clause_op_infos) + pred_op_infos = get_op_btree_interpretation(pred_op); + else /* no point in looking */ + pred_op_infos = NIL; + + foreach(lcp, pred_op_infos) + { + OpBtreeInterpretation *pred_op_info = lfirst(lcp); + Oid opfamily_id = pred_op_info->opfamily_id; + + foreach(lcc, clause_op_infos) + { + OpBtreeInterpretation *clause_op_info = lfirst(lcc); + StrategyNumber pred_strategy, + clause_strategy, + test_strategy; + + /* Must find them in same opfamily */ + if (opfamily_id != clause_op_info->opfamily_id) + continue; + /* Lefttypes should match */ + Assert(clause_op_info->oplefttype == pred_op_info->oplefttype); + + pred_strategy = pred_op_info->strategy; + clause_strategy = clause_op_info->strategy; + + /* + * Check to see if we can make a proof for same-subexpressions + * cases based on the operators' relationship in this opfamily. + */ + if (refute_it) + same_subexprs |= BT_refutes_table[clause_strategy - 1][pred_strategy - 1]; + else + same_subexprs |= BT_implies_table[clause_strategy - 1][pred_strategy - 1]; + + /* + * Look up the "test" strategy number in the implication table + */ + if (refute_it) + test_strategy = BT_refute_table[clause_strategy - 1][pred_strategy - 1]; + else + test_strategy = BT_implic_table[clause_strategy - 1][pred_strategy - 1]; + + if (test_strategy == 0) + { + /* Can't determine implication using this interpretation */ + continue; + } + + /* + * See if opfamily has an operator for the test strategy and the + * datatypes. + */ + if (test_strategy == BTNE) + { + test_op = get_opfamily_member(opfamily_id, + pred_op_info->oprighttype, + clause_op_info->oprighttype, + BTEqualStrategyNumber); + if (OidIsValid(test_op)) + test_op = get_negator(test_op); + } + else + { + test_op = get_opfamily_member(opfamily_id, + pred_op_info->oprighttype, + clause_op_info->oprighttype, + test_strategy); + } + + if (!OidIsValid(test_op)) + continue; + + /* + * Last check: test_op must be immutable. + * + * Note that we require only the test_op to be immutable, not the + * original clause_op. (pred_op is assumed to have been checked + * immutable by the caller.) Essentially we are assuming that the + * opfamily is consistent even if it contains operators that are + * merely stable. + */ + if (op_volatile(test_op) == PROVOLATILE_IMMUTABLE) + { + found = true; + break; + } + } + + if (found) + break; + } + + list_free_deep(pred_op_infos); + list_free_deep(clause_op_infos); + + if (!found) + { + /* couldn't find a suitable comparison operator */ + test_op = InvalidOid; + } + + /* + * If we think we were able to prove something about same-subexpressions + * cases, check to make sure the clause_op is immutable before believing + * it completely. (Usually, the clause_op would be immutable if the + * pred_op is, but it's not entirely clear that this must be true in all + * cases, so let's check.) + */ + if (same_subexprs && + op_volatile(clause_op) != PROVOLATILE_IMMUTABLE) + same_subexprs = false; + + /* Cache the results, whether positive or negative */ + if (refute_it) + { + cache_entry->refute_test_op = test_op; + cache_entry->same_subexprs_refutes = same_subexprs; + cache_entry->have_refute = true; + } + else + { + cache_entry->implic_test_op = test_op; + cache_entry->same_subexprs_implies = same_subexprs; + cache_entry->have_implic = true; + } + + return cache_entry; +} + +/* + * operator_same_subexprs_lookup + * Convenience subroutine to look up the cached answer for + * same-subexpressions cases. + */ +static bool +operator_same_subexprs_lookup(Oid pred_op, Oid clause_op, bool refute_it) +{ + OprProofCacheEntry *cache_entry; + + cache_entry = lookup_proof_cache(pred_op, clause_op, refute_it); + if (refute_it) + return cache_entry->same_subexprs_refutes; + else + return cache_entry->same_subexprs_implies; +} + +/* + * get_btree_test_op + * Identify the comparison operator needed for a btree-operator + * proof or refutation involving comparison of constants. + * + * Given the truth of a clause "var clause_op const1", we are attempting to + * prove or refute a predicate "var pred_op const2". The identities of the + * two operators are sufficient to determine the operator (if any) to compare + * const2 to const1 with. + * + * Returns the OID of the operator to use, or InvalidOid if no proof is + * possible. + */ +static Oid +get_btree_test_op(Oid pred_op, Oid clause_op, bool refute_it) +{ + OprProofCacheEntry *cache_entry; + + cache_entry = lookup_proof_cache(pred_op, clause_op, refute_it); + if (refute_it) + return cache_entry->refute_test_op; + else + return cache_entry->implic_test_op; +} + + +/* + * Callback for pg_amop inval events + */ +static void +InvalidateOprProofCacheCallBack(Datum arg, int cacheid, uint32 hashvalue) +{ + HASH_SEQ_STATUS status; + OprProofCacheEntry *hentry; + + Assert(OprProofCacheHash != NULL); + + /* Currently we just reset all entries; hard to be smarter ... */ + hash_seq_init(&status, OprProofCacheHash); + + while ((hentry = (OprProofCacheEntry *) hash_seq_search(&status)) != NULL) + { + hentry->have_implic = false; + hentry->have_refute = false; + } +} diff --git a/src/backend/optimizer/util/relnode.c b/src/backend/optimizer/util/relnode.c new file mode 100644 index 0000000..3c75fd5 --- /dev/null +++ b/src/backend/optimizer/util/relnode.c @@ -0,0 +1,2047 @@ +/*------------------------------------------------------------------------- + * + * relnode.c + * Relation-node lookup/construction routines + * + * Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group + * Portions Copyright (c) 1994, Regents of the University of California + * + * + * IDENTIFICATION + * src/backend/optimizer/util/relnode.c + * + *------------------------------------------------------------------------- + */ +#include "postgres.h" + +#include <limits.h> + +#include "miscadmin.h" +#include "nodes/nodeFuncs.h" +#include "optimizer/appendinfo.h" +#include "optimizer/clauses.h" +#include "optimizer/cost.h" +#include "optimizer/inherit.h" +#include "optimizer/pathnode.h" +#include "optimizer/paths.h" +#include "optimizer/placeholder.h" +#include "optimizer/plancat.h" +#include "optimizer/restrictinfo.h" +#include "optimizer/tlist.h" +#include "utils/hsearch.h" +#include "utils/lsyscache.h" + + +typedef struct JoinHashEntry +{ + Relids join_relids; /* hash key --- MUST BE FIRST */ + RelOptInfo *join_rel; +} JoinHashEntry; + +static void build_joinrel_tlist(PlannerInfo *root, RelOptInfo *joinrel, + RelOptInfo *input_rel); +static List *build_joinrel_restrictlist(PlannerInfo *root, + RelOptInfo *joinrel, + RelOptInfo *outer_rel, + RelOptInfo *inner_rel); +static void build_joinrel_joinlist(RelOptInfo *joinrel, + RelOptInfo *outer_rel, + RelOptInfo *inner_rel); +static List *subbuild_joinrel_restrictlist(RelOptInfo *joinrel, + List *joininfo_list, + List *new_restrictlist); +static List *subbuild_joinrel_joinlist(RelOptInfo *joinrel, + List *joininfo_list, + List *new_joininfo); +static void set_foreign_rel_properties(RelOptInfo *joinrel, + RelOptInfo *outer_rel, RelOptInfo *inner_rel); +static void add_join_rel(PlannerInfo *root, RelOptInfo *joinrel); +static void build_joinrel_partition_info(RelOptInfo *joinrel, + RelOptInfo *outer_rel, RelOptInfo *inner_rel, + List *restrictlist, JoinType jointype); +static bool have_partkey_equi_join(RelOptInfo *joinrel, + RelOptInfo *rel1, RelOptInfo *rel2, + JoinType jointype, List *restrictlist); +static int match_expr_to_partition_keys(Expr *expr, RelOptInfo *rel, + bool strict_op); +static void set_joinrel_partition_key_exprs(RelOptInfo *joinrel, + RelOptInfo *outer_rel, RelOptInfo *inner_rel, + JoinType jointype); +static void build_child_join_reltarget(PlannerInfo *root, + RelOptInfo *parentrel, + RelOptInfo *childrel, + int nappinfos, + AppendRelInfo **appinfos); + + +/* + * setup_simple_rel_arrays + * Prepare the arrays we use for quickly accessing base relations + * and AppendRelInfos. + */ +void +setup_simple_rel_arrays(PlannerInfo *root) +{ + int size; + Index rti; + ListCell *lc; + + /* Arrays are accessed using RT indexes (1..N) */ + size = list_length(root->parse->rtable) + 1; + root->simple_rel_array_size = size; + + /* + * simple_rel_array is initialized to all NULLs, since no RelOptInfos + * exist yet. It'll be filled by later calls to build_simple_rel(). + */ + root->simple_rel_array = (RelOptInfo **) + palloc0(size * sizeof(RelOptInfo *)); + + /* simple_rte_array is an array equivalent of the rtable list */ + root->simple_rte_array = (RangeTblEntry **) + palloc0(size * sizeof(RangeTblEntry *)); + rti = 1; + foreach(lc, root->parse->rtable) + { + RangeTblEntry *rte = (RangeTblEntry *) lfirst(lc); + + root->simple_rte_array[rti++] = rte; + } + + /* append_rel_array is not needed if there are no AppendRelInfos */ + if (root->append_rel_list == NIL) + { + root->append_rel_array = NULL; + return; + } + + root->append_rel_array = (AppendRelInfo **) + palloc0(size * sizeof(AppendRelInfo *)); + + /* + * append_rel_array is filled with any already-existing AppendRelInfos, + * which currently could only come from UNION ALL flattening. We might + * add more later during inheritance expansion, but it's the + * responsibility of the expansion code to update the array properly. + */ + foreach(lc, root->append_rel_list) + { + AppendRelInfo *appinfo = lfirst_node(AppendRelInfo, lc); + int child_relid = appinfo->child_relid; + + /* Sanity check */ + Assert(child_relid < size); + + if (root->append_rel_array[child_relid]) + elog(ERROR, "child relation already exists"); + + root->append_rel_array[child_relid] = appinfo; + } +} + +/* + * expand_planner_arrays + * Expand the PlannerInfo's per-RTE arrays by add_size members + * and initialize the newly added entries to NULLs + * + * Note: this causes the append_rel_array to become allocated even if + * it was not before. This is okay for current uses, because we only call + * this when adding child relations, which always have AppendRelInfos. + */ +void +expand_planner_arrays(PlannerInfo *root, int add_size) +{ + int new_size; + + Assert(add_size > 0); + + new_size = root->simple_rel_array_size + add_size; + + root->simple_rel_array = (RelOptInfo **) + repalloc(root->simple_rel_array, + sizeof(RelOptInfo *) * new_size); + MemSet(root->simple_rel_array + root->simple_rel_array_size, + 0, sizeof(RelOptInfo *) * add_size); + + root->simple_rte_array = (RangeTblEntry **) + repalloc(root->simple_rte_array, + sizeof(RangeTblEntry *) * new_size); + MemSet(root->simple_rte_array + root->simple_rel_array_size, + 0, sizeof(RangeTblEntry *) * add_size); + + if (root->append_rel_array) + { + root->append_rel_array = (AppendRelInfo **) + repalloc(root->append_rel_array, + sizeof(AppendRelInfo *) * new_size); + MemSet(root->append_rel_array + root->simple_rel_array_size, + 0, sizeof(AppendRelInfo *) * add_size); + } + else + { + root->append_rel_array = (AppendRelInfo **) + palloc0(sizeof(AppendRelInfo *) * new_size); + } + + root->simple_rel_array_size = new_size; +} + +/* + * build_simple_rel + * Construct a new RelOptInfo for a base relation or 'other' relation. + */ +RelOptInfo * +build_simple_rel(PlannerInfo *root, int relid, RelOptInfo *parent) +{ + RelOptInfo *rel; + RangeTblEntry *rte; + + /* Rel should not exist already */ + Assert(relid > 0 && relid < root->simple_rel_array_size); + if (root->simple_rel_array[relid] != NULL) + elog(ERROR, "rel %d already exists", relid); + + /* Fetch RTE for relation */ + rte = root->simple_rte_array[relid]; + Assert(rte != NULL); + + rel = makeNode(RelOptInfo); + rel->reloptkind = parent ? RELOPT_OTHER_MEMBER_REL : RELOPT_BASEREL; + rel->relids = bms_make_singleton(relid); + rel->rows = 0; + /* cheap startup cost is interesting iff not all tuples to be retrieved */ + rel->consider_startup = (root->tuple_fraction > 0); + rel->consider_param_startup = false; /* might get changed later */ + rel->consider_parallel = false; /* might get changed later */ + rel->reltarget = create_empty_pathtarget(); + rel->pathlist = NIL; + rel->ppilist = NIL; + rel->partial_pathlist = NIL; + rel->cheapest_startup_path = NULL; + rel->cheapest_total_path = NULL; + rel->cheapest_unique_path = NULL; + rel->cheapest_parameterized_paths = NIL; + rel->relid = relid; + rel->rtekind = rte->rtekind; + /* min_attr, max_attr, attr_needed, attr_widths are set below */ + rel->lateral_vars = NIL; + rel->indexlist = NIL; + rel->statlist = NIL; + rel->pages = 0; + rel->tuples = 0; + rel->allvisfrac = 0; + rel->eclass_indexes = NULL; + rel->subroot = NULL; + rel->subplan_params = NIL; + rel->rel_parallel_workers = -1; /* set up in get_relation_info */ + rel->amflags = 0; + rel->serverid = InvalidOid; + rel->userid = rte->checkAsUser; + rel->useridiscurrent = false; + rel->fdwroutine = NULL; + rel->fdw_private = NULL; + rel->unique_for_rels = NIL; + rel->non_unique_for_rels = NIL; + rel->baserestrictinfo = NIL; + rel->baserestrictcost.startup = 0; + rel->baserestrictcost.per_tuple = 0; + rel->baserestrict_min_security = UINT_MAX; + rel->joininfo = NIL; + rel->has_eclass_joins = false; + rel->consider_partitionwise_join = false; /* might get changed later */ + rel->part_scheme = NULL; + rel->nparts = -1; + rel->boundinfo = NULL; + rel->partbounds_merged = false; + rel->partition_qual = NIL; + rel->part_rels = NULL; + rel->live_parts = NULL; + rel->all_partrels = NULL; + rel->partexprs = NULL; + rel->nullable_partexprs = NULL; + + /* + * Pass assorted information down the inheritance hierarchy. + */ + if (parent) + { + /* + * Each direct or indirect child wants to know the relids of its + * topmost parent. + */ + if (parent->top_parent_relids) + rel->top_parent_relids = parent->top_parent_relids; + else + rel->top_parent_relids = bms_copy(parent->relids); + + /* + * Also propagate lateral-reference information from appendrel parent + * rels to their child rels. We intentionally give each child rel the + * same minimum parameterization, even though it's quite possible that + * some don't reference all the lateral rels. This is because any + * append path for the parent will have to have the same + * parameterization for every child anyway, and there's no value in + * forcing extra reparameterize_path() calls. Similarly, a lateral + * reference to the parent prevents use of otherwise-movable join rels + * for each child. + * + * It's possible for child rels to have their own children, in which + * case the topmost parent's lateral info propagates all the way down. + */ + rel->direct_lateral_relids = parent->direct_lateral_relids; + rel->lateral_relids = parent->lateral_relids; + rel->lateral_referencers = parent->lateral_referencers; + } + else + { + rel->top_parent_relids = NULL; + rel->direct_lateral_relids = NULL; + rel->lateral_relids = NULL; + rel->lateral_referencers = NULL; + } + + /* Check type of rtable entry */ + switch (rte->rtekind) + { + case RTE_RELATION: + /* Table --- retrieve statistics from the system catalogs */ + get_relation_info(root, rte->relid, rte->inh, rel); + break; + case RTE_SUBQUERY: + case RTE_FUNCTION: + case RTE_TABLEFUNC: + case RTE_VALUES: + case RTE_CTE: + case RTE_NAMEDTUPLESTORE: + + /* + * Subquery, function, tablefunc, values list, CTE, or ENR --- set + * up attr range and arrays + * + * Note: 0 is included in range to support whole-row Vars + */ + rel->min_attr = 0; + rel->max_attr = list_length(rte->eref->colnames); + rel->attr_needed = (Relids *) + palloc0((rel->max_attr - rel->min_attr + 1) * sizeof(Relids)); + rel->attr_widths = (int32 *) + palloc0((rel->max_attr - rel->min_attr + 1) * sizeof(int32)); + break; + case RTE_RESULT: + /* RTE_RESULT has no columns, nor could it have whole-row Var */ + rel->min_attr = 0; + rel->max_attr = -1; + rel->attr_needed = NULL; + rel->attr_widths = NULL; + break; + default: + elog(ERROR, "unrecognized RTE kind: %d", + (int) rte->rtekind); + break; + } + + /* + * Copy the parent's quals to the child, with appropriate substitution of + * variables. If any constant false or NULL clauses turn up, we can mark + * the child as dummy right away. (We must do this immediately so that + * pruning works correctly when recursing in expand_partitioned_rtentry.) + */ + if (parent) + { + AppendRelInfo *appinfo = root->append_rel_array[relid]; + + Assert(appinfo != NULL); + if (!apply_child_basequals(root, parent, rel, rte, appinfo)) + { + /* + * Some restriction clause reduced to constant FALSE or NULL after + * substitution, so this child need not be scanned. + */ + mark_dummy_rel(rel); + } + } + + /* Save the finished struct in the query's simple_rel_array */ + root->simple_rel_array[relid] = rel; + + return rel; +} + +/* + * find_base_rel + * Find a base or other relation entry, which must already exist. + */ +RelOptInfo * +find_base_rel(PlannerInfo *root, int relid) +{ + RelOptInfo *rel; + + Assert(relid > 0); + + if (relid < root->simple_rel_array_size) + { + rel = root->simple_rel_array[relid]; + if (rel) + return rel; + } + + elog(ERROR, "no relation entry for relid %d", relid); + + return NULL; /* keep compiler quiet */ +} + +/* + * build_join_rel_hash + * Construct the auxiliary hash table for join relations. + */ +static void +build_join_rel_hash(PlannerInfo *root) +{ + HTAB *hashtab; + HASHCTL hash_ctl; + ListCell *l; + + /* Create the hash table */ + hash_ctl.keysize = sizeof(Relids); + hash_ctl.entrysize = sizeof(JoinHashEntry); + hash_ctl.hash = bitmap_hash; + hash_ctl.match = bitmap_match; + hash_ctl.hcxt = CurrentMemoryContext; + hashtab = hash_create("JoinRelHashTable", + 256L, + &hash_ctl, + HASH_ELEM | HASH_FUNCTION | HASH_COMPARE | HASH_CONTEXT); + + /* Insert all the already-existing joinrels */ + foreach(l, root->join_rel_list) + { + RelOptInfo *rel = (RelOptInfo *) lfirst(l); + JoinHashEntry *hentry; + bool found; + + hentry = (JoinHashEntry *) hash_search(hashtab, + &(rel->relids), + HASH_ENTER, + &found); + Assert(!found); + hentry->join_rel = rel; + } + + root->join_rel_hash = hashtab; +} + +/* + * find_join_rel + * Returns relation entry corresponding to 'relids' (a set of RT indexes), + * or NULL if none exists. This is for join relations. + */ +RelOptInfo * +find_join_rel(PlannerInfo *root, Relids relids) +{ + /* + * Switch to using hash lookup when list grows "too long". The threshold + * is arbitrary and is known only here. + */ + if (!root->join_rel_hash && list_length(root->join_rel_list) > 32) + build_join_rel_hash(root); + + /* + * Use either hashtable lookup or linear search, as appropriate. + * + * Note: the seemingly redundant hashkey variable is used to avoid taking + * the address of relids; unless the compiler is exceedingly smart, doing + * so would force relids out of a register and thus probably slow down the + * list-search case. + */ + if (root->join_rel_hash) + { + Relids hashkey = relids; + JoinHashEntry *hentry; + + hentry = (JoinHashEntry *) hash_search(root->join_rel_hash, + &hashkey, + HASH_FIND, + NULL); + if (hentry) + return hentry->join_rel; + } + else + { + ListCell *l; + + foreach(l, root->join_rel_list) + { + RelOptInfo *rel = (RelOptInfo *) lfirst(l); + + if (bms_equal(rel->relids, relids)) + return rel; + } + } + + return NULL; +} + +/* + * set_foreign_rel_properties + * Set up foreign-join fields if outer and inner relation are foreign + * tables (or joins) belonging to the same server and assigned to the same + * user to check access permissions as. + * + * In addition to an exact match of userid, we allow the case where one side + * has zero userid (implying current user) and the other side has explicit + * userid that happens to equal the current user; but in that case, pushdown of + * the join is only valid for the current user. The useridiscurrent field + * records whether we had to make such an assumption for this join or any + * sub-join. + * + * Otherwise these fields are left invalid, so GetForeignJoinPaths will not be + * called for the join relation. + * + */ +static void +set_foreign_rel_properties(RelOptInfo *joinrel, RelOptInfo *outer_rel, + RelOptInfo *inner_rel) +{ + if (OidIsValid(outer_rel->serverid) && + inner_rel->serverid == outer_rel->serverid) + { + if (inner_rel->userid == outer_rel->userid) + { + joinrel->serverid = outer_rel->serverid; + joinrel->userid = outer_rel->userid; + joinrel->useridiscurrent = outer_rel->useridiscurrent || inner_rel->useridiscurrent; + joinrel->fdwroutine = outer_rel->fdwroutine; + } + else if (!OidIsValid(inner_rel->userid) && + outer_rel->userid == GetUserId()) + { + joinrel->serverid = outer_rel->serverid; + joinrel->userid = outer_rel->userid; + joinrel->useridiscurrent = true; + joinrel->fdwroutine = outer_rel->fdwroutine; + } + else if (!OidIsValid(outer_rel->userid) && + inner_rel->userid == GetUserId()) + { + joinrel->serverid = outer_rel->serverid; + joinrel->userid = inner_rel->userid; + joinrel->useridiscurrent = true; + joinrel->fdwroutine = outer_rel->fdwroutine; + } + } +} + +/* + * add_join_rel + * Add given join relation to the list of join relations in the given + * PlannerInfo. Also add it to the auxiliary hashtable if there is one. + */ +static void +add_join_rel(PlannerInfo *root, RelOptInfo *joinrel) +{ + /* GEQO requires us to append the new joinrel to the end of the list! */ + root->join_rel_list = lappend(root->join_rel_list, joinrel); + + /* store it into the auxiliary hashtable if there is one. */ + if (root->join_rel_hash) + { + JoinHashEntry *hentry; + bool found; + + hentry = (JoinHashEntry *) hash_search(root->join_rel_hash, + &(joinrel->relids), + HASH_ENTER, + &found); + Assert(!found); + hentry->join_rel = joinrel; + } +} + +/* + * build_join_rel + * Returns relation entry corresponding to the union of two given rels, + * creating a new relation entry if none already exists. + * + * 'joinrelids' is the Relids set that uniquely identifies the join + * 'outer_rel' and 'inner_rel' are relation nodes for the relations to be + * joined + * 'sjinfo': join context info + * 'restrictlist_ptr': result variable. If not NULL, *restrictlist_ptr + * receives the list of RestrictInfo nodes that apply to this + * particular pair of joinable relations. + * + * restrictlist_ptr makes the routine's API a little grotty, but it saves + * duplicated calculation of the restrictlist... + */ +RelOptInfo * +build_join_rel(PlannerInfo *root, + Relids joinrelids, + RelOptInfo *outer_rel, + RelOptInfo *inner_rel, + SpecialJoinInfo *sjinfo, + List **restrictlist_ptr) +{ + RelOptInfo *joinrel; + List *restrictlist; + + /* This function should be used only for join between parents. */ + Assert(!IS_OTHER_REL(outer_rel) && !IS_OTHER_REL(inner_rel)); + + /* + * See if we already have a joinrel for this set of base rels. + */ + joinrel = find_join_rel(root, joinrelids); + + if (joinrel) + { + /* + * Yes, so we only need to figure the restrictlist for this particular + * pair of component relations. + */ + if (restrictlist_ptr) + *restrictlist_ptr = build_joinrel_restrictlist(root, + joinrel, + outer_rel, + inner_rel); + return joinrel; + } + + /* + * Nope, so make one. + */ + joinrel = makeNode(RelOptInfo); + joinrel->reloptkind = RELOPT_JOINREL; + joinrel->relids = bms_copy(joinrelids); + joinrel->rows = 0; + /* cheap startup cost is interesting iff not all tuples to be retrieved */ + joinrel->consider_startup = (root->tuple_fraction > 0); + joinrel->consider_param_startup = false; + joinrel->consider_parallel = false; + joinrel->reltarget = create_empty_pathtarget(); + joinrel->pathlist = NIL; + joinrel->ppilist = NIL; + joinrel->partial_pathlist = NIL; + joinrel->cheapest_startup_path = NULL; + joinrel->cheapest_total_path = NULL; + joinrel->cheapest_unique_path = NULL; + joinrel->cheapest_parameterized_paths = NIL; + /* init direct_lateral_relids from children; we'll finish it up below */ + joinrel->direct_lateral_relids = + bms_union(outer_rel->direct_lateral_relids, + inner_rel->direct_lateral_relids); + joinrel->lateral_relids = min_join_parameterization(root, joinrel->relids, + outer_rel, inner_rel); + joinrel->relid = 0; /* indicates not a baserel */ + joinrel->rtekind = RTE_JOIN; + joinrel->min_attr = 0; + joinrel->max_attr = 0; + joinrel->attr_needed = NULL; + joinrel->attr_widths = NULL; + joinrel->lateral_vars = NIL; + joinrel->lateral_referencers = NULL; + joinrel->indexlist = NIL; + joinrel->statlist = NIL; + joinrel->pages = 0; + joinrel->tuples = 0; + joinrel->allvisfrac = 0; + joinrel->eclass_indexes = NULL; + joinrel->subroot = NULL; + joinrel->subplan_params = NIL; + joinrel->rel_parallel_workers = -1; + joinrel->amflags = 0; + joinrel->serverid = InvalidOid; + joinrel->userid = InvalidOid; + joinrel->useridiscurrent = false; + joinrel->fdwroutine = NULL; + joinrel->fdw_private = NULL; + joinrel->unique_for_rels = NIL; + joinrel->non_unique_for_rels = NIL; + joinrel->baserestrictinfo = NIL; + joinrel->baserestrictcost.startup = 0; + joinrel->baserestrictcost.per_tuple = 0; + joinrel->baserestrict_min_security = UINT_MAX; + joinrel->joininfo = NIL; + joinrel->has_eclass_joins = false; + joinrel->consider_partitionwise_join = false; /* might get changed later */ + joinrel->top_parent_relids = NULL; + joinrel->part_scheme = NULL; + joinrel->nparts = -1; + joinrel->boundinfo = NULL; + joinrel->partbounds_merged = false; + joinrel->partition_qual = NIL; + joinrel->part_rels = NULL; + joinrel->live_parts = NULL; + joinrel->all_partrels = NULL; + joinrel->partexprs = NULL; + joinrel->nullable_partexprs = NULL; + + /* Compute information relevant to the foreign relations. */ + set_foreign_rel_properties(joinrel, outer_rel, inner_rel); + + /* + * Create a new tlist containing just the vars that need to be output from + * this join (ie, are needed for higher joinclauses or final output). + * + * NOTE: the tlist order for a join rel will depend on which pair of outer + * and inner rels we first try to build it from. But the contents should + * be the same regardless. + */ + build_joinrel_tlist(root, joinrel, outer_rel); + build_joinrel_tlist(root, joinrel, inner_rel); + add_placeholders_to_joinrel(root, joinrel, outer_rel, inner_rel); + + /* + * add_placeholders_to_joinrel also took care of adding the ph_lateral + * sets of any PlaceHolderVars computed here to direct_lateral_relids, so + * now we can finish computing that. This is much like the computation of + * the transitively-closed lateral_relids in min_join_parameterization, + * except that here we *do* have to consider the added PHVs. + */ + joinrel->direct_lateral_relids = + bms_del_members(joinrel->direct_lateral_relids, joinrel->relids); + if (bms_is_empty(joinrel->direct_lateral_relids)) + joinrel->direct_lateral_relids = NULL; + + /* + * Construct restrict and join clause lists for the new joinrel. (The + * caller might or might not need the restrictlist, but I need it anyway + * for set_joinrel_size_estimates().) + */ + restrictlist = build_joinrel_restrictlist(root, joinrel, + outer_rel, inner_rel); + if (restrictlist_ptr) + *restrictlist_ptr = restrictlist; + build_joinrel_joinlist(joinrel, outer_rel, inner_rel); + + /* + * This is also the right place to check whether the joinrel has any + * pending EquivalenceClass joins. + */ + joinrel->has_eclass_joins = has_relevant_eclass_joinclause(root, joinrel); + + /* Store the partition information. */ + build_joinrel_partition_info(joinrel, outer_rel, inner_rel, restrictlist, + sjinfo->jointype); + + /* + * Set estimates of the joinrel's size. + */ + set_joinrel_size_estimates(root, joinrel, outer_rel, inner_rel, + sjinfo, restrictlist); + + /* + * Set the consider_parallel flag if this joinrel could potentially be + * scanned within a parallel worker. If this flag is false for either + * inner_rel or outer_rel, then it must be false for the joinrel also. + * Even if both are true, there might be parallel-restricted expressions + * in the targetlist or quals. + * + * Note that if there are more than two rels in this relation, they could + * be divided between inner_rel and outer_rel in any arbitrary way. We + * assume this doesn't matter, because we should hit all the same baserels + * and joinclauses while building up to this joinrel no matter which we + * take; therefore, we should make the same decision here however we get + * here. + */ + if (inner_rel->consider_parallel && outer_rel->consider_parallel && + is_parallel_safe(root, (Node *) restrictlist) && + is_parallel_safe(root, (Node *) joinrel->reltarget->exprs)) + joinrel->consider_parallel = true; + + /* Add the joinrel to the PlannerInfo. */ + add_join_rel(root, joinrel); + + /* + * Also, if dynamic-programming join search is active, add the new joinrel + * to the appropriate sublist. Note: you might think the Assert on number + * of members should be for equality, but some of the level 1 rels might + * have been joinrels already, so we can only assert <=. + */ + if (root->join_rel_level) + { + Assert(root->join_cur_level > 0); + Assert(root->join_cur_level <= bms_num_members(joinrel->relids)); + root->join_rel_level[root->join_cur_level] = + lappend(root->join_rel_level[root->join_cur_level], joinrel); + } + + return joinrel; +} + +/* + * build_child_join_rel + * Builds RelOptInfo representing join between given two child relations. + * + * 'outer_rel' and 'inner_rel' are the RelOptInfos of child relations being + * joined + * 'parent_joinrel' is the RelOptInfo representing the join between parent + * relations. Some of the members of new RelOptInfo are produced by + * translating corresponding members of this RelOptInfo + * 'sjinfo': child-join context info + * 'restrictlist': list of RestrictInfo nodes that apply to this particular + * pair of joinable relations + * 'jointype' is the join type (inner, left, full, etc) + */ +RelOptInfo * +build_child_join_rel(PlannerInfo *root, RelOptInfo *outer_rel, + RelOptInfo *inner_rel, RelOptInfo *parent_joinrel, + List *restrictlist, SpecialJoinInfo *sjinfo, + JoinType jointype) +{ + RelOptInfo *joinrel = makeNode(RelOptInfo); + AppendRelInfo **appinfos; + int nappinfos; + + /* Only joins between "other" relations land here. */ + Assert(IS_OTHER_REL(outer_rel) && IS_OTHER_REL(inner_rel)); + + /* The parent joinrel should have consider_partitionwise_join set. */ + Assert(parent_joinrel->consider_partitionwise_join); + + joinrel->reloptkind = RELOPT_OTHER_JOINREL; + joinrel->relids = bms_union(outer_rel->relids, inner_rel->relids); + joinrel->rows = 0; + /* cheap startup cost is interesting iff not all tuples to be retrieved */ + joinrel->consider_startup = (root->tuple_fraction > 0); + joinrel->consider_param_startup = false; + joinrel->consider_parallel = false; + joinrel->reltarget = create_empty_pathtarget(); + joinrel->pathlist = NIL; + joinrel->ppilist = NIL; + joinrel->partial_pathlist = NIL; + joinrel->cheapest_startup_path = NULL; + joinrel->cheapest_total_path = NULL; + joinrel->cheapest_unique_path = NULL; + joinrel->cheapest_parameterized_paths = NIL; + joinrel->direct_lateral_relids = NULL; + joinrel->lateral_relids = NULL; + joinrel->relid = 0; /* indicates not a baserel */ + joinrel->rtekind = RTE_JOIN; + joinrel->min_attr = 0; + joinrel->max_attr = 0; + joinrel->attr_needed = NULL; + joinrel->attr_widths = NULL; + joinrel->lateral_vars = NIL; + joinrel->lateral_referencers = NULL; + joinrel->indexlist = NIL; + joinrel->pages = 0; + joinrel->tuples = 0; + joinrel->allvisfrac = 0; + joinrel->eclass_indexes = NULL; + joinrel->subroot = NULL; + joinrel->subplan_params = NIL; + joinrel->amflags = 0; + joinrel->serverid = InvalidOid; + joinrel->userid = InvalidOid; + joinrel->useridiscurrent = false; + joinrel->fdwroutine = NULL; + joinrel->fdw_private = NULL; + joinrel->baserestrictinfo = NIL; + joinrel->baserestrictcost.startup = 0; + joinrel->baserestrictcost.per_tuple = 0; + joinrel->joininfo = NIL; + joinrel->has_eclass_joins = false; + joinrel->consider_partitionwise_join = false; /* might get changed later */ + joinrel->top_parent_relids = NULL; + joinrel->part_scheme = NULL; + joinrel->nparts = -1; + joinrel->boundinfo = NULL; + joinrel->partbounds_merged = false; + joinrel->partition_qual = NIL; + joinrel->part_rels = NULL; + joinrel->live_parts = NULL; + joinrel->all_partrels = NULL; + joinrel->partexprs = NULL; + joinrel->nullable_partexprs = NULL; + + joinrel->top_parent_relids = bms_union(outer_rel->top_parent_relids, + inner_rel->top_parent_relids); + + /* Compute information relevant to foreign relations. */ + set_foreign_rel_properties(joinrel, outer_rel, inner_rel); + + /* Compute information needed for mapping Vars to the child rel */ + appinfos = find_appinfos_by_relids(root, joinrel->relids, &nappinfos); + + /* Set up reltarget struct */ + build_child_join_reltarget(root, parent_joinrel, joinrel, + nappinfos, appinfos); + + /* Construct joininfo list. */ + joinrel->joininfo = (List *) adjust_appendrel_attrs(root, + (Node *) parent_joinrel->joininfo, + nappinfos, + appinfos); + + /* + * Lateral relids referred in child join will be same as that referred in + * the parent relation. + */ + joinrel->direct_lateral_relids = (Relids) bms_copy(parent_joinrel->direct_lateral_relids); + joinrel->lateral_relids = (Relids) bms_copy(parent_joinrel->lateral_relids); + + /* + * If the parent joinrel has pending equivalence classes, so does the + * child. + */ + joinrel->has_eclass_joins = parent_joinrel->has_eclass_joins; + + /* Is the join between partitions itself partitioned? */ + build_joinrel_partition_info(joinrel, outer_rel, inner_rel, restrictlist, + jointype); + + /* Child joinrel is parallel safe if parent is parallel safe. */ + joinrel->consider_parallel = parent_joinrel->consider_parallel; + + /* Set estimates of the child-joinrel's size. */ + set_joinrel_size_estimates(root, joinrel, outer_rel, inner_rel, + sjinfo, restrictlist); + + /* We build the join only once. */ + Assert(!find_join_rel(root, joinrel->relids)); + + /* Add the relation to the PlannerInfo. */ + add_join_rel(root, joinrel); + + /* + * We might need EquivalenceClass members corresponding to the child join, + * so that we can represent sort pathkeys for it. As with children of + * baserels, we shouldn't need this unless there are relevant eclass joins + * (implying that a merge join might be possible) or pathkeys to sort by. + */ + if (joinrel->has_eclass_joins || has_useful_pathkeys(root, parent_joinrel)) + add_child_join_rel_equivalences(root, + nappinfos, appinfos, + parent_joinrel, joinrel); + + pfree(appinfos); + + return joinrel; +} + +/* + * min_join_parameterization + * + * Determine the minimum possible parameterization of a joinrel, that is, the + * set of other rels it contains LATERAL references to. We save this value in + * the join's RelOptInfo. This function is split out of build_join_rel() + * because join_is_legal() needs the value to check a prospective join. + */ +Relids +min_join_parameterization(PlannerInfo *root, + Relids joinrelids, + RelOptInfo *outer_rel, + RelOptInfo *inner_rel) +{ + Relids result; + + /* + * Basically we just need the union of the inputs' lateral_relids, less + * whatever is already in the join. + * + * It's not immediately obvious that this is a valid way to compute the + * result, because it might seem that we're ignoring possible lateral refs + * of PlaceHolderVars that are due to be computed at the join but not in + * either input. However, because create_lateral_join_info() already + * charged all such PHV refs to each member baserel of the join, they'll + * be accounted for already in the inputs' lateral_relids. Likewise, we + * do not need to worry about doing transitive closure here, because that + * was already accounted for in the original baserel lateral_relids. + */ + result = bms_union(outer_rel->lateral_relids, inner_rel->lateral_relids); + result = bms_del_members(result, joinrelids); + + /* Maintain invariant that result is exactly NULL if empty */ + if (bms_is_empty(result)) + result = NULL; + + return result; +} + +/* + * build_joinrel_tlist + * Builds a join relation's target list from an input relation. + * (This is invoked twice to handle the two input relations.) + * + * The join's targetlist includes all Vars of its member relations that + * will still be needed above the join. This subroutine adds all such + * Vars from the specified input rel's tlist to the join rel's tlist. + * + * We also compute the expected width of the join's output, making use + * of data that was cached at the baserel level by set_rel_width(). + */ +static void +build_joinrel_tlist(PlannerInfo *root, RelOptInfo *joinrel, + RelOptInfo *input_rel) +{ + Relids relids = joinrel->relids; + ListCell *vars; + + foreach(vars, input_rel->reltarget->exprs) + { + Var *var = (Var *) lfirst(vars); + + /* + * Ignore PlaceHolderVars in the input tlists; we'll make our own + * decisions about whether to copy them. + */ + if (IsA(var, PlaceHolderVar)) + continue; + + /* + * Otherwise, anything in a baserel or joinrel targetlist ought to be + * a Var. (More general cases can only appear in appendrel child + * rels, which will never be seen here.) + */ + if (!IsA(var, Var)) + elog(ERROR, "unexpected node type in rel targetlist: %d", + (int) nodeTag(var)); + + if (var->varno == ROWID_VAR) + { + /* UPDATE/DELETE/MERGE row identity vars are always needed */ + RowIdentityVarInfo *ridinfo = (RowIdentityVarInfo *) + list_nth(root->row_identity_vars, var->varattno - 1); + + joinrel->reltarget->exprs = lappend(joinrel->reltarget->exprs, + var); + /* Vars have cost zero, so no need to adjust reltarget->cost */ + joinrel->reltarget->width += ridinfo->rowidwidth; + } + else + { + RelOptInfo *baserel; + int ndx; + + /* Get the Var's original base rel */ + baserel = find_base_rel(root, var->varno); + + /* Is it still needed above this joinrel? */ + ndx = var->varattno - baserel->min_attr; + if (bms_nonempty_difference(baserel->attr_needed[ndx], relids)) + { + /* Yup, add it to the output */ + joinrel->reltarget->exprs = lappend(joinrel->reltarget->exprs, + var); + /* Vars have cost zero, so no need to adjust reltarget->cost */ + joinrel->reltarget->width += baserel->attr_widths[ndx]; + } + } + } +} + +/* + * build_joinrel_restrictlist + * build_joinrel_joinlist + * These routines build lists of restriction and join clauses for a + * join relation from the joininfo lists of the relations it joins. + * + * These routines are separate because the restriction list must be + * built afresh for each pair of input sub-relations we consider, whereas + * the join list need only be computed once for any join RelOptInfo. + * The join list is fully determined by the set of rels making up the + * joinrel, so we should get the same results (up to ordering) from any + * candidate pair of sub-relations. But the restriction list is whatever + * is not handled in the sub-relations, so it depends on which + * sub-relations are considered. + * + * If a join clause from an input relation refers to base rels still not + * present in the joinrel, then it is still a join clause for the joinrel; + * we put it into the joininfo list for the joinrel. Otherwise, + * the clause is now a restrict clause for the joined relation, and we + * return it to the caller of build_joinrel_restrictlist() to be stored in + * join paths made from this pair of sub-relations. (It will not need to + * be considered further up the join tree.) + * + * In many cases we will find the same RestrictInfos in both input + * relations' joinlists, so be careful to eliminate duplicates. + * Pointer equality should be a sufficient test for dups, since all + * the various joinlist entries ultimately refer to RestrictInfos + * pushed into them by distribute_restrictinfo_to_rels(). + * + * 'joinrel' is a join relation node + * 'outer_rel' and 'inner_rel' are a pair of relations that can be joined + * to form joinrel. + * + * build_joinrel_restrictlist() returns a list of relevant restrictinfos, + * whereas build_joinrel_joinlist() stores its results in the joinrel's + * joininfo list. One or the other must accept each given clause! + * + * NB: Formerly, we made deep(!) copies of each input RestrictInfo to pass + * up to the join relation. I believe this is no longer necessary, because + * RestrictInfo nodes are no longer context-dependent. Instead, just include + * the original nodes in the lists made for the join relation. + */ +static List * +build_joinrel_restrictlist(PlannerInfo *root, + RelOptInfo *joinrel, + RelOptInfo *outer_rel, + RelOptInfo *inner_rel) +{ + List *result; + + /* + * Collect all the clauses that syntactically belong at this level, + * eliminating any duplicates (important since we will see many of the + * same clauses arriving from both input relations). + */ + result = subbuild_joinrel_restrictlist(joinrel, outer_rel->joininfo, NIL); + result = subbuild_joinrel_restrictlist(joinrel, inner_rel->joininfo, result); + + /* + * Add on any clauses derived from EquivalenceClasses. These cannot be + * redundant with the clauses in the joininfo lists, so don't bother + * checking. + */ + result = list_concat(result, + generate_join_implied_equalities(root, + joinrel->relids, + outer_rel->relids, + inner_rel)); + + return result; +} + +static void +build_joinrel_joinlist(RelOptInfo *joinrel, + RelOptInfo *outer_rel, + RelOptInfo *inner_rel) +{ + List *result; + + /* + * Collect all the clauses that syntactically belong above this level, + * eliminating any duplicates (important since we will see many of the + * same clauses arriving from both input relations). + */ + result = subbuild_joinrel_joinlist(joinrel, outer_rel->joininfo, NIL); + result = subbuild_joinrel_joinlist(joinrel, inner_rel->joininfo, result); + + joinrel->joininfo = result; +} + +static List * +subbuild_joinrel_restrictlist(RelOptInfo *joinrel, + List *joininfo_list, + List *new_restrictlist) +{ + ListCell *l; + + foreach(l, joininfo_list) + { + RestrictInfo *rinfo = (RestrictInfo *) lfirst(l); + + if (bms_is_subset(rinfo->required_relids, joinrel->relids)) + { + /* + * This clause becomes a restriction clause for the joinrel, since + * it refers to no outside rels. Add it to the list, being + * careful to eliminate duplicates. (Since RestrictInfo nodes in + * different joinlists will have been multiply-linked rather than + * copied, pointer equality should be a sufficient test.) + */ + new_restrictlist = list_append_unique_ptr(new_restrictlist, rinfo); + } + else + { + /* + * This clause is still a join clause at this level, so we ignore + * it in this routine. + */ + } + } + + return new_restrictlist; +} + +static List * +subbuild_joinrel_joinlist(RelOptInfo *joinrel, + List *joininfo_list, + List *new_joininfo) +{ + ListCell *l; + + /* Expected to be called only for join between parent relations. */ + Assert(joinrel->reloptkind == RELOPT_JOINREL); + + foreach(l, joininfo_list) + { + RestrictInfo *rinfo = (RestrictInfo *) lfirst(l); + + if (bms_is_subset(rinfo->required_relids, joinrel->relids)) + { + /* + * This clause becomes a restriction clause for the joinrel, since + * it refers to no outside rels. So we can ignore it in this + * routine. + */ + } + else + { + /* + * This clause is still a join clause at this level, so add it to + * the new joininfo list, being careful to eliminate duplicates. + * (Since RestrictInfo nodes in different joinlists will have been + * multiply-linked rather than copied, pointer equality should be + * a sufficient test.) + */ + new_joininfo = list_append_unique_ptr(new_joininfo, rinfo); + } + } + + return new_joininfo; +} + + +/* + * fetch_upper_rel + * Build a RelOptInfo describing some post-scan/join query processing, + * or return a pre-existing one if somebody already built it. + * + * An "upper" relation is identified by an UpperRelationKind and a Relids set. + * The meaning of the Relids set is not specified here, and very likely will + * vary for different relation kinds. + * + * Most of the fields in an upper-level RelOptInfo are not used and are not + * set here (though makeNode should ensure they're zeroes). We basically only + * care about fields that are of interest to add_path() and set_cheapest(). + */ +RelOptInfo * +fetch_upper_rel(PlannerInfo *root, UpperRelationKind kind, Relids relids) +{ + RelOptInfo *upperrel; + ListCell *lc; + + /* + * For the moment, our indexing data structure is just a List for each + * relation kind. If we ever get so many of one kind that this stops + * working well, we can improve it. No code outside this function should + * assume anything about how to find a particular upperrel. + */ + + /* If we already made this upperrel for the query, return it */ + foreach(lc, root->upper_rels[kind]) + { + upperrel = (RelOptInfo *) lfirst(lc); + + if (bms_equal(upperrel->relids, relids)) + return upperrel; + } + + upperrel = makeNode(RelOptInfo); + upperrel->reloptkind = RELOPT_UPPER_REL; + upperrel->relids = bms_copy(relids); + + /* cheap startup cost is interesting iff not all tuples to be retrieved */ + upperrel->consider_startup = (root->tuple_fraction > 0); + upperrel->consider_param_startup = false; + upperrel->consider_parallel = false; /* might get changed later */ + upperrel->reltarget = create_empty_pathtarget(); + upperrel->pathlist = NIL; + upperrel->cheapest_startup_path = NULL; + upperrel->cheapest_total_path = NULL; + upperrel->cheapest_unique_path = NULL; + upperrel->cheapest_parameterized_paths = NIL; + + root->upper_rels[kind] = lappend(root->upper_rels[kind], upperrel); + + return upperrel; +} + + +/* + * find_childrel_parents + * Compute the set of parent relids of an appendrel child rel. + * + * Since appendrels can be nested, a child could have multiple levels of + * appendrel ancestors. This function computes a Relids set of all the + * parent relation IDs. + */ +Relids +find_childrel_parents(PlannerInfo *root, RelOptInfo *rel) +{ + Relids result = NULL; + + Assert(rel->reloptkind == RELOPT_OTHER_MEMBER_REL); + Assert(rel->relid > 0 && rel->relid < root->simple_rel_array_size); + + do + { + AppendRelInfo *appinfo = root->append_rel_array[rel->relid]; + Index prelid = appinfo->parent_relid; + + result = bms_add_member(result, prelid); + + /* traverse up to the parent rel, loop if it's also a child rel */ + rel = find_base_rel(root, prelid); + } while (rel->reloptkind == RELOPT_OTHER_MEMBER_REL); + + Assert(rel->reloptkind == RELOPT_BASEREL); + + return result; +} + + +/* + * get_baserel_parampathinfo + * Get the ParamPathInfo for a parameterized path for a base relation, + * constructing one if we don't have one already. + * + * This centralizes estimating the rowcounts for parameterized paths. + * We need to cache those to be sure we use the same rowcount for all paths + * of the same parameterization for a given rel. This is also a convenient + * place to determine which movable join clauses the parameterized path will + * be responsible for evaluating. + */ +ParamPathInfo * +get_baserel_parampathinfo(PlannerInfo *root, RelOptInfo *baserel, + Relids required_outer) +{ + ParamPathInfo *ppi; + Relids joinrelids; + List *pclauses; + double rows; + ListCell *lc; + + /* If rel has LATERAL refs, every path for it should account for them */ + Assert(bms_is_subset(baserel->lateral_relids, required_outer)); + + /* Unparameterized paths have no ParamPathInfo */ + if (bms_is_empty(required_outer)) + return NULL; + + Assert(!bms_overlap(baserel->relids, required_outer)); + + /* If we already have a PPI for this parameterization, just return it */ + if ((ppi = find_param_path_info(baserel, required_outer))) + return ppi; + + /* + * Identify all joinclauses that are movable to this base rel given this + * parameterization. + */ + joinrelids = bms_union(baserel->relids, required_outer); + pclauses = NIL; + foreach(lc, baserel->joininfo) + { + RestrictInfo *rinfo = (RestrictInfo *) lfirst(lc); + + if (join_clause_is_movable_into(rinfo, + baserel->relids, + joinrelids)) + pclauses = lappend(pclauses, rinfo); + } + + /* + * Add in joinclauses generated by EquivalenceClasses, too. (These + * necessarily satisfy join_clause_is_movable_into.) + */ + pclauses = list_concat(pclauses, + generate_join_implied_equalities(root, + joinrelids, + required_outer, + baserel)); + + /* Estimate the number of rows returned by the parameterized scan */ + rows = get_parameterized_baserel_size(root, baserel, pclauses); + + /* And now we can build the ParamPathInfo */ + ppi = makeNode(ParamPathInfo); + ppi->ppi_req_outer = required_outer; + ppi->ppi_rows = rows; + ppi->ppi_clauses = pclauses; + baserel->ppilist = lappend(baserel->ppilist, ppi); + + return ppi; +} + +/* + * get_joinrel_parampathinfo + * Get the ParamPathInfo for a parameterized path for a join relation, + * constructing one if we don't have one already. + * + * This centralizes estimating the rowcounts for parameterized paths. + * We need to cache those to be sure we use the same rowcount for all paths + * of the same parameterization for a given rel. This is also a convenient + * place to determine which movable join clauses the parameterized path will + * be responsible for evaluating. + * + * outer_path and inner_path are a pair of input paths that can be used to + * construct the join, and restrict_clauses is the list of regular join + * clauses (including clauses derived from EquivalenceClasses) that must be + * applied at the join node when using these inputs. + * + * Unlike the situation for base rels, the set of movable join clauses to be + * enforced at a join varies with the selected pair of input paths, so we + * must calculate that and pass it back, even if we already have a matching + * ParamPathInfo. We handle this by adding any clauses moved down to this + * join to *restrict_clauses, which is an in/out parameter. (The addition + * is done in such a way as to not modify the passed-in List structure.) + * + * Note: when considering a nestloop join, the caller must have removed from + * restrict_clauses any movable clauses that are themselves scheduled to be + * pushed into the right-hand path. We do not do that here since it's + * unnecessary for other join types. + */ +ParamPathInfo * +get_joinrel_parampathinfo(PlannerInfo *root, RelOptInfo *joinrel, + Path *outer_path, + Path *inner_path, + SpecialJoinInfo *sjinfo, + Relids required_outer, + List **restrict_clauses) +{ + ParamPathInfo *ppi; + Relids join_and_req; + Relids outer_and_req; + Relids inner_and_req; + List *pclauses; + List *eclauses; + List *dropped_ecs; + double rows; + ListCell *lc; + + /* If rel has LATERAL refs, every path for it should account for them */ + Assert(bms_is_subset(joinrel->lateral_relids, required_outer)); + + /* Unparameterized paths have no ParamPathInfo or extra join clauses */ + if (bms_is_empty(required_outer)) + return NULL; + + Assert(!bms_overlap(joinrel->relids, required_outer)); + + /* + * Identify all joinclauses that are movable to this join rel given this + * parameterization. These are the clauses that are movable into this + * join, but not movable into either input path. Treat an unparameterized + * input path as not accepting parameterized clauses (because it won't, + * per the shortcut exit above), even though the joinclause movement rules + * might allow the same clauses to be moved into a parameterized path for + * that rel. + */ + join_and_req = bms_union(joinrel->relids, required_outer); + if (outer_path->param_info) + outer_and_req = bms_union(outer_path->parent->relids, + PATH_REQ_OUTER(outer_path)); + else + outer_and_req = NULL; /* outer path does not accept parameters */ + if (inner_path->param_info) + inner_and_req = bms_union(inner_path->parent->relids, + PATH_REQ_OUTER(inner_path)); + else + inner_and_req = NULL; /* inner path does not accept parameters */ + + pclauses = NIL; + foreach(lc, joinrel->joininfo) + { + RestrictInfo *rinfo = (RestrictInfo *) lfirst(lc); + + if (join_clause_is_movable_into(rinfo, + joinrel->relids, + join_and_req) && + !join_clause_is_movable_into(rinfo, + outer_path->parent->relids, + outer_and_req) && + !join_clause_is_movable_into(rinfo, + inner_path->parent->relids, + inner_and_req)) + pclauses = lappend(pclauses, rinfo); + } + + /* Consider joinclauses generated by EquivalenceClasses, too */ + eclauses = generate_join_implied_equalities(root, + join_and_req, + required_outer, + joinrel); + /* We only want ones that aren't movable to lower levels */ + dropped_ecs = NIL; + foreach(lc, eclauses) + { + RestrictInfo *rinfo = (RestrictInfo *) lfirst(lc); + + /* + * In principle, join_clause_is_movable_into() should accept anything + * returned by generate_join_implied_equalities(); but because its + * analysis is only approximate, sometimes it doesn't. So we + * currently cannot use this Assert; instead just assume it's okay to + * apply the joinclause at this level. + */ +#ifdef NOT_USED + Assert(join_clause_is_movable_into(rinfo, + joinrel->relids, + join_and_req)); +#endif + if (join_clause_is_movable_into(rinfo, + outer_path->parent->relids, + outer_and_req)) + continue; /* drop if movable into LHS */ + if (join_clause_is_movable_into(rinfo, + inner_path->parent->relids, + inner_and_req)) + { + /* drop if movable into RHS, but remember EC for use below */ + Assert(rinfo->left_ec == rinfo->right_ec); + dropped_ecs = lappend(dropped_ecs, rinfo->left_ec); + continue; + } + pclauses = lappend(pclauses, rinfo); + } + + /* + * EquivalenceClasses are harder to deal with than we could wish, because + * of the fact that a given EC can generate different clauses depending on + * context. Suppose we have an EC {X.X, Y.Y, Z.Z} where X and Y are the + * LHS and RHS of the current join and Z is in required_outer, and further + * suppose that the inner_path is parameterized by both X and Z. The code + * above will have produced either Z.Z = X.X or Z.Z = Y.Y from that EC, + * and in the latter case will have discarded it as being movable into the + * RHS. However, the EC machinery might have produced either Y.Y = X.X or + * Y.Y = Z.Z as the EC enforcement clause within the inner_path; it will + * not have produced both, and we can't readily tell from here which one + * it did pick. If we add no clause to this join, we'll end up with + * insufficient enforcement of the EC; either Z.Z or X.X will fail to be + * constrained to be equal to the other members of the EC. (When we come + * to join Z to this X/Y path, we will certainly drop whichever EC clause + * is generated at that join, so this omission won't get fixed later.) + * + * To handle this, for each EC we discarded such a clause from, try to + * generate a clause connecting the required_outer rels to the join's LHS + * ("Z.Z = X.X" in the terms of the above example). If successful, and if + * the clause can't be moved to the LHS, add it to the current join's + * restriction clauses. (If an EC cannot generate such a clause then it + * has nothing that needs to be enforced here, while if the clause can be + * moved into the LHS then it should have been enforced within that path.) + * + * Note that we don't need similar processing for ECs whose clause was + * considered to be movable into the LHS, because the LHS can't refer to + * the RHS so there is no comparable ambiguity about what it might + * actually be enforcing internally. + */ + if (dropped_ecs) + { + Relids real_outer_and_req; + + real_outer_and_req = bms_union(outer_path->parent->relids, + required_outer); + eclauses = + generate_join_implied_equalities_for_ecs(root, + dropped_ecs, + real_outer_and_req, + required_outer, + outer_path->parent); + foreach(lc, eclauses) + { + RestrictInfo *rinfo = (RestrictInfo *) lfirst(lc); + + /* As above, can't quite assert this here */ +#ifdef NOT_USED + Assert(join_clause_is_movable_into(rinfo, + outer_path->parent->relids, + real_outer_and_req)); +#endif + if (!join_clause_is_movable_into(rinfo, + outer_path->parent->relids, + outer_and_req)) + pclauses = lappend(pclauses, rinfo); + } + } + + /* + * Now, attach the identified moved-down clauses to the caller's + * restrict_clauses list. By using list_concat in this order, we leave + * the original list structure of restrict_clauses undamaged. + */ + *restrict_clauses = list_concat(pclauses, *restrict_clauses); + + /* If we already have a PPI for this parameterization, just return it */ + if ((ppi = find_param_path_info(joinrel, required_outer))) + return ppi; + + /* Estimate the number of rows returned by the parameterized join */ + rows = get_parameterized_joinrel_size(root, joinrel, + outer_path, + inner_path, + sjinfo, + *restrict_clauses); + + /* + * And now we can build the ParamPathInfo. No point in saving the + * input-pair-dependent clause list, though. + * + * Note: in GEQO mode, we'll be called in a temporary memory context, but + * the joinrel structure is there too, so no problem. + */ + ppi = makeNode(ParamPathInfo); + ppi->ppi_req_outer = required_outer; + ppi->ppi_rows = rows; + ppi->ppi_clauses = NIL; + joinrel->ppilist = lappend(joinrel->ppilist, ppi); + + return ppi; +} + +/* + * get_appendrel_parampathinfo + * Get the ParamPathInfo for a parameterized path for an append relation. + * + * For an append relation, the rowcount estimate will just be the sum of + * the estimates for its children. However, we still need a ParamPathInfo + * to flag the fact that the path requires parameters. So this just creates + * a suitable struct with zero ppi_rows (and no ppi_clauses either, since + * the Append node isn't responsible for checking quals). + */ +ParamPathInfo * +get_appendrel_parampathinfo(RelOptInfo *appendrel, Relids required_outer) +{ + ParamPathInfo *ppi; + + /* If rel has LATERAL refs, every path for it should account for them */ + Assert(bms_is_subset(appendrel->lateral_relids, required_outer)); + + /* Unparameterized paths have no ParamPathInfo */ + if (bms_is_empty(required_outer)) + return NULL; + + Assert(!bms_overlap(appendrel->relids, required_outer)); + + /* If we already have a PPI for this parameterization, just return it */ + if ((ppi = find_param_path_info(appendrel, required_outer))) + return ppi; + + /* Else build the ParamPathInfo */ + ppi = makeNode(ParamPathInfo); + ppi->ppi_req_outer = required_outer; + ppi->ppi_rows = 0; + ppi->ppi_clauses = NIL; + appendrel->ppilist = lappend(appendrel->ppilist, ppi); + + return ppi; +} + +/* + * Returns a ParamPathInfo for the parameterization given by required_outer, if + * already available in the given rel. Returns NULL otherwise. + */ +ParamPathInfo * +find_param_path_info(RelOptInfo *rel, Relids required_outer) +{ + ListCell *lc; + + foreach(lc, rel->ppilist) + { + ParamPathInfo *ppi = (ParamPathInfo *) lfirst(lc); + + if (bms_equal(ppi->ppi_req_outer, required_outer)) + return ppi; + } + + return NULL; +} + +/* + * build_joinrel_partition_info + * Checks if the two relations being joined can use partitionwise join + * and if yes, initialize partitioning information of the resulting + * partitioned join relation. + */ +static void +build_joinrel_partition_info(RelOptInfo *joinrel, RelOptInfo *outer_rel, + RelOptInfo *inner_rel, List *restrictlist, + JoinType jointype) +{ + PartitionScheme part_scheme; + + /* Nothing to do if partitionwise join technique is disabled. */ + if (!enable_partitionwise_join) + { + Assert(!IS_PARTITIONED_REL(joinrel)); + return; + } + + /* + * We can only consider this join as an input to further partitionwise + * joins if (a) the input relations are partitioned and have + * consider_partitionwise_join=true, (b) the partition schemes match, and + * (c) we can identify an equi-join between the partition keys. Note that + * if it were possible for have_partkey_equi_join to return different + * answers for the same joinrel depending on which join ordering we try + * first, this logic would break. That shouldn't happen, though, because + * of the way the query planner deduces implied equalities and reorders + * the joins. Please see optimizer/README for details. + */ + if (outer_rel->part_scheme == NULL || inner_rel->part_scheme == NULL || + !outer_rel->consider_partitionwise_join || + !inner_rel->consider_partitionwise_join || + outer_rel->part_scheme != inner_rel->part_scheme || + !have_partkey_equi_join(joinrel, outer_rel, inner_rel, + jointype, restrictlist)) + { + Assert(!IS_PARTITIONED_REL(joinrel)); + return; + } + + part_scheme = outer_rel->part_scheme; + + /* + * This function will be called only once for each joinrel, hence it + * should not have partitioning fields filled yet. + */ + Assert(!joinrel->part_scheme && !joinrel->partexprs && + !joinrel->nullable_partexprs && !joinrel->part_rels && + !joinrel->boundinfo); + + /* + * If the join relation is partitioned, it uses the same partitioning + * scheme as the joining relations. + * + * Note: we calculate the partition bounds, number of partitions, and + * child-join relations of the join relation in try_partitionwise_join(). + */ + joinrel->part_scheme = part_scheme; + set_joinrel_partition_key_exprs(joinrel, outer_rel, inner_rel, jointype); + + /* + * Set the consider_partitionwise_join flag. + */ + Assert(outer_rel->consider_partitionwise_join); + Assert(inner_rel->consider_partitionwise_join); + joinrel->consider_partitionwise_join = true; +} + +/* + * have_partkey_equi_join + * + * Returns true if there exist equi-join conditions involving pairs + * of matching partition keys of the relations being joined for all + * partition keys. + */ +static bool +have_partkey_equi_join(RelOptInfo *joinrel, + RelOptInfo *rel1, RelOptInfo *rel2, + JoinType jointype, List *restrictlist) +{ + PartitionScheme part_scheme = rel1->part_scheme; + ListCell *lc; + int cnt_pks; + bool pk_has_clause[PARTITION_MAX_KEYS]; + bool strict_op; + + /* + * This function must only be called when the joined relations have same + * partitioning scheme. + */ + Assert(rel1->part_scheme == rel2->part_scheme); + Assert(part_scheme); + + memset(pk_has_clause, 0, sizeof(pk_has_clause)); + foreach(lc, restrictlist) + { + RestrictInfo *rinfo = lfirst_node(RestrictInfo, lc); + OpExpr *opexpr; + Expr *expr1; + Expr *expr2; + int ipk1; + int ipk2; + + /* If processing an outer join, only use its own join clauses. */ + if (IS_OUTER_JOIN(jointype) && + RINFO_IS_PUSHED_DOWN(rinfo, joinrel->relids)) + continue; + + /* Skip clauses which can not be used for a join. */ + if (!rinfo->can_join) + continue; + + /* Skip clauses which are not equality conditions. */ + if (!rinfo->mergeopfamilies && !OidIsValid(rinfo->hashjoinoperator)) + continue; + + /* Should be OK to assume it's an OpExpr. */ + opexpr = castNode(OpExpr, rinfo->clause); + + /* Match the operands to the relation. */ + if (bms_is_subset(rinfo->left_relids, rel1->relids) && + bms_is_subset(rinfo->right_relids, rel2->relids)) + { + expr1 = linitial(opexpr->args); + expr2 = lsecond(opexpr->args); + } + else if (bms_is_subset(rinfo->left_relids, rel2->relids) && + bms_is_subset(rinfo->right_relids, rel1->relids)) + { + expr1 = lsecond(opexpr->args); + expr2 = linitial(opexpr->args); + } + else + continue; + + /* + * Now we need to know whether the join operator is strict; see + * comments in pathnodes.h. + */ + strict_op = op_strict(opexpr->opno); + + /* + * Only clauses referencing the partition keys are useful for + * partitionwise join. + */ + ipk1 = match_expr_to_partition_keys(expr1, rel1, strict_op); + if (ipk1 < 0) + continue; + ipk2 = match_expr_to_partition_keys(expr2, rel2, strict_op); + if (ipk2 < 0) + continue; + + /* + * If the clause refers to keys at different ordinal positions, it can + * not be used for partitionwise join. + */ + if (ipk1 != ipk2) + continue; + + /* + * The clause allows partitionwise join only if it uses the same + * operator family as that specified by the partition key. + */ + if (rel1->part_scheme->strategy == PARTITION_STRATEGY_HASH) + { + if (!OidIsValid(rinfo->hashjoinoperator) || + !op_in_opfamily(rinfo->hashjoinoperator, + part_scheme->partopfamily[ipk1])) + continue; + } + else if (!list_member_oid(rinfo->mergeopfamilies, + part_scheme->partopfamily[ipk1])) + continue; + + /* Mark the partition key as having an equi-join clause. */ + pk_has_clause[ipk1] = true; + } + + /* Check whether every partition key has an equi-join condition. */ + for (cnt_pks = 0; cnt_pks < part_scheme->partnatts; cnt_pks++) + { + if (!pk_has_clause[cnt_pks]) + return false; + } + + return true; +} + +/* + * match_expr_to_partition_keys + * + * Tries to match an expression to one of the nullable or non-nullable + * partition keys of "rel". Returns the matched key's ordinal position, + * or -1 if the expression could not be matched to any of the keys. + * + * strict_op must be true if the expression will be compared with the + * partition key using a strict operator. This allows us to consider + * nullable as well as nonnullable partition keys. + */ +static int +match_expr_to_partition_keys(Expr *expr, RelOptInfo *rel, bool strict_op) +{ + int cnt; + + /* This function should be called only for partitioned relations. */ + Assert(rel->part_scheme); + Assert(rel->partexprs); + Assert(rel->nullable_partexprs); + + /* Remove any relabel decorations. */ + while (IsA(expr, RelabelType)) + expr = (Expr *) (castNode(RelabelType, expr))->arg; + + for (cnt = 0; cnt < rel->part_scheme->partnatts; cnt++) + { + ListCell *lc; + + /* We can always match to the non-nullable partition keys. */ + foreach(lc, rel->partexprs[cnt]) + { + if (equal(lfirst(lc), expr)) + return cnt; + } + + if (!strict_op) + continue; + + /* + * If it's a strict join operator then a NULL partition key on one + * side will not join to any partition key on the other side, and in + * particular such a row can't join to a row from a different + * partition on the other side. So, it's okay to search the nullable + * partition keys as well. + */ + foreach(lc, rel->nullable_partexprs[cnt]) + { + if (equal(lfirst(lc), expr)) + return cnt; + } + } + + return -1; +} + +/* + * set_joinrel_partition_key_exprs + * Initialize partition key expressions for a partitioned joinrel. + */ +static void +set_joinrel_partition_key_exprs(RelOptInfo *joinrel, + RelOptInfo *outer_rel, RelOptInfo *inner_rel, + JoinType jointype) +{ + PartitionScheme part_scheme = joinrel->part_scheme; + int partnatts = part_scheme->partnatts; + + joinrel->partexprs = (List **) palloc0(sizeof(List *) * partnatts); + joinrel->nullable_partexprs = + (List **) palloc0(sizeof(List *) * partnatts); + + /* + * The joinrel's partition expressions are the same as those of the input + * rels, but we must properly classify them as nullable or not in the + * joinrel's output. (Also, we add some more partition expressions if + * it's a FULL JOIN.) + */ + for (int cnt = 0; cnt < partnatts; cnt++) + { + /* mark these const to enforce that we copy them properly */ + const List *outer_expr = outer_rel->partexprs[cnt]; + const List *outer_null_expr = outer_rel->nullable_partexprs[cnt]; + const List *inner_expr = inner_rel->partexprs[cnt]; + const List *inner_null_expr = inner_rel->nullable_partexprs[cnt]; + List *partexpr = NIL; + List *nullable_partexpr = NIL; + ListCell *lc; + + switch (jointype) + { + /* + * A join relation resulting from an INNER join may be + * regarded as partitioned by either of the inner and outer + * relation keys. For example, A INNER JOIN B ON A.a = B.b + * can be regarded as partitioned on either A.a or B.b. So we + * add both keys to the joinrel's partexpr lists. However, + * anything that was already nullable still has to be treated + * as nullable. + */ + case JOIN_INNER: + partexpr = list_concat_copy(outer_expr, inner_expr); + nullable_partexpr = list_concat_copy(outer_null_expr, + inner_null_expr); + break; + + /* + * A join relation resulting from a SEMI or ANTI join may be + * regarded as partitioned by the outer relation keys. The + * inner relation's keys are no longer interesting; since they + * aren't visible in the join output, nothing could join to + * them. + */ + case JOIN_SEMI: + case JOIN_ANTI: + partexpr = list_copy(outer_expr); + nullable_partexpr = list_copy(outer_null_expr); + break; + + /* + * A join relation resulting from a LEFT OUTER JOIN likewise + * may be regarded as partitioned on the (non-nullable) outer + * relation keys. The inner (nullable) relation keys are okay + * as partition keys for further joins as long as they involve + * strict join operators. + */ + case JOIN_LEFT: + partexpr = list_copy(outer_expr); + nullable_partexpr = list_concat_copy(inner_expr, + outer_null_expr); + nullable_partexpr = list_concat(nullable_partexpr, + inner_null_expr); + break; + + /* + * For FULL OUTER JOINs, both relations are nullable, so the + * resulting join relation may be regarded as partitioned on + * either of inner and outer relation keys, but only for joins + * that involve strict join operators. + */ + case JOIN_FULL: + nullable_partexpr = list_concat_copy(outer_expr, + inner_expr); + nullable_partexpr = list_concat(nullable_partexpr, + outer_null_expr); + nullable_partexpr = list_concat(nullable_partexpr, + inner_null_expr); + + /* + * Also add CoalesceExprs corresponding to each possible + * full-join output variable (that is, left side coalesced to + * right side), so that we can match equijoin expressions + * using those variables. We really only need these for + * columns merged by JOIN USING, and only with the pairs of + * input items that correspond to the data structures that + * parse analysis would build for such variables. But it's + * hard to tell which those are, so just make all the pairs. + * Extra items in the nullable_partexprs list won't cause big + * problems. (It's possible that such items will get matched + * to user-written COALESCEs, but it should still be valid to + * partition on those, since they're going to be either the + * partition column or NULL; it's the same argument as for + * partitionwise nesting of any outer join.) We assume no + * type coercions are needed to make the coalesce expressions, + * since columns of different types won't have gotten + * classified as the same PartitionScheme. + */ + foreach(lc, list_concat_copy(outer_expr, outer_null_expr)) + { + Node *larg = (Node *) lfirst(lc); + ListCell *lc2; + + foreach(lc2, list_concat_copy(inner_expr, inner_null_expr)) + { + Node *rarg = (Node *) lfirst(lc2); + CoalesceExpr *c = makeNode(CoalesceExpr); + + c->coalescetype = exprType(larg); + c->coalescecollid = exprCollation(larg); + c->args = list_make2(larg, rarg); + c->location = -1; + nullable_partexpr = lappend(nullable_partexpr, c); + } + } + break; + + default: + elog(ERROR, "unrecognized join type: %d", (int) jointype); + } + + joinrel->partexprs[cnt] = partexpr; + joinrel->nullable_partexprs[cnt] = nullable_partexpr; + } +} + +/* + * build_child_join_reltarget + * Set up a child-join relation's reltarget from a parent-join relation. + */ +static void +build_child_join_reltarget(PlannerInfo *root, + RelOptInfo *parentrel, + RelOptInfo *childrel, + int nappinfos, + AppendRelInfo **appinfos) +{ + /* Build the targetlist */ + childrel->reltarget->exprs = (List *) + adjust_appendrel_attrs(root, + (Node *) parentrel->reltarget->exprs, + nappinfos, appinfos); + + /* Set the cost and width fields */ + childrel->reltarget->cost.startup = parentrel->reltarget->cost.startup; + childrel->reltarget->cost.per_tuple = parentrel->reltarget->cost.per_tuple; + childrel->reltarget->width = parentrel->reltarget->width; +} diff --git a/src/backend/optimizer/util/restrictinfo.c b/src/backend/optimizer/util/restrictinfo.c new file mode 100644 index 0000000..e7764f4 --- /dev/null +++ b/src/backend/optimizer/util/restrictinfo.c @@ -0,0 +1,655 @@ +/*------------------------------------------------------------------------- + * + * restrictinfo.c + * RestrictInfo node manipulation routines. + * + * Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group + * Portions Copyright (c) 1994, Regents of the University of California + * + * + * IDENTIFICATION + * src/backend/optimizer/util/restrictinfo.c + * + *------------------------------------------------------------------------- + */ +#include "postgres.h" + +#include "nodes/makefuncs.h" +#include "nodes/nodeFuncs.h" +#include "optimizer/clauses.h" +#include "optimizer/optimizer.h" +#include "optimizer/restrictinfo.h" + + +static RestrictInfo *make_restrictinfo_internal(PlannerInfo *root, + Expr *clause, + Expr *orclause, + bool is_pushed_down, + bool outerjoin_delayed, + bool pseudoconstant, + Index security_level, + Relids required_relids, + Relids outer_relids, + Relids nullable_relids); +static Expr *make_sub_restrictinfos(PlannerInfo *root, + Expr *clause, + bool is_pushed_down, + bool outerjoin_delayed, + bool pseudoconstant, + Index security_level, + Relids required_relids, + Relids outer_relids, + Relids nullable_relids); + + +/* + * make_restrictinfo + * + * Build a RestrictInfo node containing the given subexpression. + * + * The is_pushed_down, outerjoin_delayed, and pseudoconstant flags for the + * RestrictInfo must be supplied by the caller, as well as the correct values + * for security_level, outer_relids, and nullable_relids. + * required_relids can be NULL, in which case it defaults to the actual clause + * contents (i.e., clause_relids). + * + * We initialize fields that depend only on the given subexpression, leaving + * others that depend on context (or may never be needed at all) to be filled + * later. + */ +RestrictInfo * +make_restrictinfo(PlannerInfo *root, + Expr *clause, + bool is_pushed_down, + bool outerjoin_delayed, + bool pseudoconstant, + Index security_level, + Relids required_relids, + Relids outer_relids, + Relids nullable_relids) +{ + /* + * If it's an OR clause, build a modified copy with RestrictInfos inserted + * above each subclause of the top-level AND/OR structure. + */ + if (is_orclause(clause)) + return (RestrictInfo *) make_sub_restrictinfos(root, + clause, + is_pushed_down, + outerjoin_delayed, + pseudoconstant, + security_level, + required_relids, + outer_relids, + nullable_relids); + + /* Shouldn't be an AND clause, else AND/OR flattening messed up */ + Assert(!is_andclause(clause)); + + return make_restrictinfo_internal(root, + clause, + NULL, + is_pushed_down, + outerjoin_delayed, + pseudoconstant, + security_level, + required_relids, + outer_relids, + nullable_relids); +} + +/* + * make_restrictinfo_internal + * + * Common code for the main entry points and the recursive cases. + */ +static RestrictInfo * +make_restrictinfo_internal(PlannerInfo *root, + Expr *clause, + Expr *orclause, + bool is_pushed_down, + bool outerjoin_delayed, + bool pseudoconstant, + Index security_level, + Relids required_relids, + Relids outer_relids, + Relids nullable_relids) +{ + RestrictInfo *restrictinfo = makeNode(RestrictInfo); + + restrictinfo->clause = clause; + restrictinfo->orclause = orclause; + restrictinfo->is_pushed_down = is_pushed_down; + restrictinfo->outerjoin_delayed = outerjoin_delayed; + restrictinfo->pseudoconstant = pseudoconstant; + restrictinfo->can_join = false; /* may get set below */ + restrictinfo->security_level = security_level; + restrictinfo->outer_relids = outer_relids; + restrictinfo->nullable_relids = nullable_relids; + + /* + * If it's potentially delayable by lower-level security quals, figure out + * whether it's leakproof. We can skip testing this for level-zero quals, + * since they would never get delayed on security grounds anyway. + */ + if (security_level > 0) + restrictinfo->leakproof = !contain_leaked_vars((Node *) clause); + else + restrictinfo->leakproof = false; /* really, "don't know" */ + + /* + * Mark volatility as unknown. The contain_volatile_functions function + * will determine if there are any volatile functions when called for the + * first time with this RestrictInfo. + */ + restrictinfo->has_volatile = VOLATILITY_UNKNOWN; + + /* + * If it's a binary opclause, set up left/right relids info. In any case + * set up the total clause relids info. + */ + if (is_opclause(clause) && list_length(((OpExpr *) clause)->args) == 2) + { + restrictinfo->left_relids = pull_varnos(root, get_leftop(clause)); + restrictinfo->right_relids = pull_varnos(root, get_rightop(clause)); + + restrictinfo->clause_relids = bms_union(restrictinfo->left_relids, + restrictinfo->right_relids); + + /* + * Does it look like a normal join clause, i.e., a binary operator + * relating expressions that come from distinct relations? If so we + * might be able to use it in a join algorithm. Note that this is a + * purely syntactic test that is made regardless of context. + */ + if (!bms_is_empty(restrictinfo->left_relids) && + !bms_is_empty(restrictinfo->right_relids) && + !bms_overlap(restrictinfo->left_relids, + restrictinfo->right_relids)) + { + restrictinfo->can_join = true; + /* pseudoconstant should certainly not be true */ + Assert(!restrictinfo->pseudoconstant); + } + } + else + { + /* Not a binary opclause, so mark left/right relid sets as empty */ + restrictinfo->left_relids = NULL; + restrictinfo->right_relids = NULL; + /* and get the total relid set the hard way */ + restrictinfo->clause_relids = pull_varnos(root, (Node *) clause); + } + + /* required_relids defaults to clause_relids */ + if (required_relids != NULL) + restrictinfo->required_relids = required_relids; + else + restrictinfo->required_relids = restrictinfo->clause_relids; + + /* + * Fill in all the cacheable fields with "not yet set" markers. None of + * these will be computed until/unless needed. Note in particular that we + * don't mark a binary opclause as mergejoinable or hashjoinable here; + * that happens only if it appears in the right context (top level of a + * joinclause list). + */ + restrictinfo->parent_ec = NULL; + + restrictinfo->eval_cost.startup = -1; + restrictinfo->norm_selec = -1; + restrictinfo->outer_selec = -1; + + restrictinfo->mergeopfamilies = NIL; + + restrictinfo->left_ec = NULL; + restrictinfo->right_ec = NULL; + restrictinfo->left_em = NULL; + restrictinfo->right_em = NULL; + restrictinfo->scansel_cache = NIL; + + restrictinfo->outer_is_left = false; + + restrictinfo->hashjoinoperator = InvalidOid; + + restrictinfo->left_bucketsize = -1; + restrictinfo->right_bucketsize = -1; + restrictinfo->left_mcvfreq = -1; + restrictinfo->right_mcvfreq = -1; + + restrictinfo->left_hasheqoperator = InvalidOid; + restrictinfo->right_hasheqoperator = InvalidOid; + + return restrictinfo; +} + +/* + * Recursively insert sub-RestrictInfo nodes into a boolean expression. + * + * We put RestrictInfos above simple (non-AND/OR) clauses and above + * sub-OR clauses, but not above sub-AND clauses, because there's no need. + * This may seem odd but it is closely related to the fact that we use + * implicit-AND lists at top level of RestrictInfo lists. Only ORs and + * simple clauses are valid RestrictInfos. + * + * The same is_pushed_down, outerjoin_delayed, and pseudoconstant flag + * values can be applied to all RestrictInfo nodes in the result. Likewise + * for security_level, outer_relids, and nullable_relids. + * + * The given required_relids are attached to our top-level output, + * but any OR-clause constituents are allowed to default to just the + * contained rels. + */ +static Expr * +make_sub_restrictinfos(PlannerInfo *root, + Expr *clause, + bool is_pushed_down, + bool outerjoin_delayed, + bool pseudoconstant, + Index security_level, + Relids required_relids, + Relids outer_relids, + Relids nullable_relids) +{ + if (is_orclause(clause)) + { + List *orlist = NIL; + ListCell *temp; + + foreach(temp, ((BoolExpr *) clause)->args) + orlist = lappend(orlist, + make_sub_restrictinfos(root, + lfirst(temp), + is_pushed_down, + outerjoin_delayed, + pseudoconstant, + security_level, + NULL, + outer_relids, + nullable_relids)); + return (Expr *) make_restrictinfo_internal(root, + clause, + make_orclause(orlist), + is_pushed_down, + outerjoin_delayed, + pseudoconstant, + security_level, + required_relids, + outer_relids, + nullable_relids); + } + else if (is_andclause(clause)) + { + List *andlist = NIL; + ListCell *temp; + + foreach(temp, ((BoolExpr *) clause)->args) + andlist = lappend(andlist, + make_sub_restrictinfos(root, + lfirst(temp), + is_pushed_down, + outerjoin_delayed, + pseudoconstant, + security_level, + required_relids, + outer_relids, + nullable_relids)); + return make_andclause(andlist); + } + else + return (Expr *) make_restrictinfo_internal(root, + clause, + NULL, + is_pushed_down, + outerjoin_delayed, + pseudoconstant, + security_level, + required_relids, + outer_relids, + nullable_relids); +} + +/* + * commute_restrictinfo + * + * Given a RestrictInfo containing a binary opclause, produce a RestrictInfo + * representing the commutation of that clause. The caller must pass the + * OID of the commutator operator (which it's presumably looked up, else + * it would not know this is valid). + * + * Beware that the result shares sub-structure with the given RestrictInfo. + * That's okay for the intended usage with derived index quals, but might + * be hazardous if the source is subject to change. Also notice that we + * assume without checking that the commutator op is a member of the same + * btree and hash opclasses as the original op. + */ +RestrictInfo * +commute_restrictinfo(RestrictInfo *rinfo, Oid comm_op) +{ + RestrictInfo *result; + OpExpr *newclause; + OpExpr *clause = castNode(OpExpr, rinfo->clause); + + Assert(list_length(clause->args) == 2); + + /* flat-copy all the fields of clause ... */ + newclause = makeNode(OpExpr); + memcpy(newclause, clause, sizeof(OpExpr)); + + /* ... and adjust those we need to change to commute it */ + newclause->opno = comm_op; + newclause->opfuncid = InvalidOid; + newclause->args = list_make2(lsecond(clause->args), + linitial(clause->args)); + + /* likewise, flat-copy all the fields of rinfo ... */ + result = makeNode(RestrictInfo); + memcpy(result, rinfo, sizeof(RestrictInfo)); + + /* + * ... and adjust those we need to change. Note in particular that we can + * preserve any cached selectivity or cost estimates, since those ought to + * be the same for the new clause. Likewise we can keep the source's + * parent_ec. + */ + result->clause = (Expr *) newclause; + result->left_relids = rinfo->right_relids; + result->right_relids = rinfo->left_relids; + Assert(result->orclause == NULL); + result->left_ec = rinfo->right_ec; + result->right_ec = rinfo->left_ec; + result->left_em = rinfo->right_em; + result->right_em = rinfo->left_em; + result->scansel_cache = NIL; /* not worth updating this */ + if (rinfo->hashjoinoperator == clause->opno) + result->hashjoinoperator = comm_op; + else + result->hashjoinoperator = InvalidOid; + result->left_bucketsize = rinfo->right_bucketsize; + result->right_bucketsize = rinfo->left_bucketsize; + result->left_mcvfreq = rinfo->right_mcvfreq; + result->right_mcvfreq = rinfo->left_mcvfreq; + result->left_hasheqoperator = InvalidOid; + result->right_hasheqoperator = InvalidOid; + + return result; +} + +/* + * restriction_is_or_clause + * + * Returns t iff the restrictinfo node contains an 'or' clause. + */ +bool +restriction_is_or_clause(RestrictInfo *restrictinfo) +{ + if (restrictinfo->orclause != NULL) + return true; + else + return false; +} + +/* + * restriction_is_securely_promotable + * + * Returns true if it's okay to evaluate this clause "early", that is before + * other restriction clauses attached to the specified relation. + */ +bool +restriction_is_securely_promotable(RestrictInfo *restrictinfo, + RelOptInfo *rel) +{ + /* + * It's okay if there are no baserestrictinfo clauses for the rel that + * would need to go before this one, *or* if this one is leakproof. + */ + if (restrictinfo->security_level <= rel->baserestrict_min_security || + restrictinfo->leakproof) + return true; + else + return false; +} + +/* + * get_actual_clauses + * + * Returns a list containing the bare clauses from 'restrictinfo_list'. + * + * This is only to be used in cases where none of the RestrictInfos can + * be pseudoconstant clauses (for instance, it's OK on indexqual lists). + */ +List * +get_actual_clauses(List *restrictinfo_list) +{ + List *result = NIL; + ListCell *l; + + foreach(l, restrictinfo_list) + { + RestrictInfo *rinfo = lfirst_node(RestrictInfo, l); + + Assert(!rinfo->pseudoconstant); + + result = lappend(result, rinfo->clause); + } + return result; +} + +/* + * extract_actual_clauses + * + * Extract bare clauses from 'restrictinfo_list', returning either the + * regular ones or the pseudoconstant ones per 'pseudoconstant'. + */ +List * +extract_actual_clauses(List *restrictinfo_list, + bool pseudoconstant) +{ + List *result = NIL; + ListCell *l; + + foreach(l, restrictinfo_list) + { + RestrictInfo *rinfo = lfirst_node(RestrictInfo, l); + + if (rinfo->pseudoconstant == pseudoconstant) + result = lappend(result, rinfo->clause); + } + return result; +} + +/* + * extract_actual_join_clauses + * + * Extract bare clauses from 'restrictinfo_list', separating those that + * semantically match the join level from those that were pushed down. + * Pseudoconstant clauses are excluded from the results. + * + * This is only used at outer joins, since for plain joins we don't care + * about pushed-down-ness. + */ +void +extract_actual_join_clauses(List *restrictinfo_list, + Relids joinrelids, + List **joinquals, + List **otherquals) +{ + ListCell *l; + + *joinquals = NIL; + *otherquals = NIL; + + foreach(l, restrictinfo_list) + { + RestrictInfo *rinfo = lfirst_node(RestrictInfo, l); + + if (RINFO_IS_PUSHED_DOWN(rinfo, joinrelids)) + { + if (!rinfo->pseudoconstant) + *otherquals = lappend(*otherquals, rinfo->clause); + } + else + { + /* joinquals shouldn't have been marked pseudoconstant */ + Assert(!rinfo->pseudoconstant); + *joinquals = lappend(*joinquals, rinfo->clause); + } + } +} + +/* + * has_pseudoconstant_clauses + * + * Returns true if 'restrictinfo_list' includes pseudoconstant clauses. + * + * This is used when we determine whether to allow extensions to consider + * pushing down joins in add_paths_to_joinrel(). + */ +bool +has_pseudoconstant_clauses(PlannerInfo *root, + List *restrictinfo_list) +{ + ListCell *l; + + /* No need to look if we know there are no pseudoconstants */ + if (!root->hasPseudoConstantQuals) + return false; + + /* See if there are pseudoconstants in the RestrictInfo list */ + foreach(l, restrictinfo_list) + { + RestrictInfo *rinfo = lfirst_node(RestrictInfo, l); + + if (rinfo->pseudoconstant) + return true; + } + return false; +} + + +/* + * join_clause_is_movable_to + * Test whether a join clause is a safe candidate for parameterization + * of a scan on the specified base relation. + * + * A movable join clause is one that can safely be evaluated at a rel below + * its normal semantic level (ie, its required_relids), if the values of + * variables that it would need from other rels are provided. + * + * We insist that the clause actually reference the target relation; this + * prevents undesirable movement of degenerate join clauses, and ensures + * that there is a unique place that a clause can be moved down to. + * + * We cannot move an outer-join clause into the non-nullable side of its + * outer join, as that would change the results (rows would be suppressed + * rather than being null-extended). + * + * Also there must not be an outer join below the clause that would null the + * Vars coming from the target relation. Otherwise the clause might give + * results different from what it would give at its normal semantic level. + * + * Also, the join clause must not use any relations that have LATERAL + * references to the target relation, since we could not put such rels on + * the outer side of a nestloop with the target relation. + */ +bool +join_clause_is_movable_to(RestrictInfo *rinfo, RelOptInfo *baserel) +{ + /* Clause must physically reference target rel */ + if (!bms_is_member(baserel->relid, rinfo->clause_relids)) + return false; + + /* Cannot move an outer-join clause into the join's outer side */ + if (bms_is_member(baserel->relid, rinfo->outer_relids)) + return false; + + /* Target rel must not be nullable below the clause */ + if (bms_is_member(baserel->relid, rinfo->nullable_relids)) + return false; + + /* Clause must not use any rels with LATERAL references to this rel */ + if (bms_overlap(baserel->lateral_referencers, rinfo->clause_relids)) + return false; + + return true; +} + +/* + * join_clause_is_movable_into + * Test whether a join clause is movable and can be evaluated within + * the current join context. + * + * currentrelids: the relids of the proposed evaluation location + * current_and_outer: the union of currentrelids and the required_outer + * relids (parameterization's outer relations) + * + * The API would be a bit clearer if we passed the current relids and the + * outer relids separately and did bms_union internally; but since most + * callers need to apply this function to multiple clauses, we make the + * caller perform the union. + * + * Obviously, the clause must only refer to Vars available from the current + * relation plus the outer rels. We also check that it does reference at + * least one current Var, ensuring that the clause will be pushed down to + * a unique place in a parameterized join tree. And we check that we're + * not pushing the clause into its outer-join outer side, nor down into + * a lower outer join's inner side. + * + * The check about pushing a clause down into a lower outer join's inner side + * is only approximate; it sometimes returns "false" when actually it would + * be safe to use the clause here because we're still above the outer join + * in question. This is okay as long as the answers at different join levels + * are consistent: it just means we might sometimes fail to push a clause as + * far down as it could safely be pushed. It's unclear whether it would be + * worthwhile to do this more precisely. (But if it's ever fixed to be + * exactly accurate, there's an Assert in get_joinrel_parampathinfo() that + * should be re-enabled.) + * + * There's no check here equivalent to join_clause_is_movable_to's test on + * lateral_referencers. We assume the caller wouldn't be inquiring unless + * it'd verified that the proposed outer rels don't have lateral references + * to the current rel(s). (If we are considering join paths with the outer + * rels on the outside and the current rels on the inside, then this should + * have been checked at the outset of such consideration; see join_is_legal + * and the path parameterization checks in joinpath.c.) On the other hand, + * in join_clause_is_movable_to we are asking whether the clause could be + * moved for some valid set of outer rels, so we don't have the benefit of + * relying on prior checks for lateral-reference validity. + * + * Note: if this returns true, it means that the clause could be moved to + * this join relation, but that doesn't mean that this is the lowest join + * it could be moved to. Caller may need to make additional calls to verify + * that this doesn't succeed on either of the inputs of a proposed join. + * + * Note: get_joinrel_parampathinfo depends on the fact that if + * current_and_outer is NULL, this function will always return false + * (since one or the other of the first two tests must fail). + */ +bool +join_clause_is_movable_into(RestrictInfo *rinfo, + Relids currentrelids, + Relids current_and_outer) +{ + /* Clause must be evaluable given available context */ + if (!bms_is_subset(rinfo->clause_relids, current_and_outer)) + return false; + + /* Clause must physically reference at least one target rel */ + if (!bms_overlap(currentrelids, rinfo->clause_relids)) + return false; + + /* Cannot move an outer-join clause into the join's outer side */ + if (bms_overlap(currentrelids, rinfo->outer_relids)) + return false; + + /* + * Target rel(s) must not be nullable below the clause. This is + * approximate, in the safe direction, because the current join might be + * above the join where the nulling would happen, in which case the clause + * would work correctly here. But we don't have enough info to be sure. + */ + if (bms_overlap(currentrelids, rinfo->nullable_relids)) + return false; + + return true; +} diff --git a/src/backend/optimizer/util/tlist.c b/src/backend/optimizer/util/tlist.c new file mode 100644 index 0000000..eed3e3f --- /dev/null +++ b/src/backend/optimizer/util/tlist.c @@ -0,0 +1,1258 @@ +/*------------------------------------------------------------------------- + * + * tlist.c + * Target list manipulation routines + * + * Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group + * Portions Copyright (c) 1994, Regents of the University of California + * + * + * IDENTIFICATION + * src/backend/optimizer/util/tlist.c + * + *------------------------------------------------------------------------- + */ +#include "postgres.h" + +#include "nodes/makefuncs.h" +#include "nodes/nodeFuncs.h" +#include "optimizer/cost.h" +#include "optimizer/optimizer.h" +#include "optimizer/tlist.h" + + +/* + * Test if an expression node represents a SRF call. Beware multiple eval! + * + * Please note that this is only meant for use in split_pathtarget_at_srfs(); + * if you use it anywhere else, your code is almost certainly wrong for SRFs + * nested within expressions. Use expression_returns_set() instead. + */ +#define IS_SRF_CALL(node) \ + ((IsA(node, FuncExpr) && ((FuncExpr *) (node))->funcretset) || \ + (IsA(node, OpExpr) && ((OpExpr *) (node))->opretset)) + +/* + * Data structures for split_pathtarget_at_srfs(). To preserve the identity + * of sortgroupref items even if they are textually equal(), what we track is + * not just bare expressions but expressions plus their sortgroupref indexes. + */ +typedef struct +{ + Node *expr; /* some subexpression of a PathTarget */ + Index sortgroupref; /* its sortgroupref, or 0 if none */ +} split_pathtarget_item; + +typedef struct +{ + /* This is a List of bare expressions: */ + List *input_target_exprs; /* exprs available from input */ + /* These are Lists of Lists of split_pathtarget_items: */ + List *level_srfs; /* SRF exprs to evaluate at each level */ + List *level_input_vars; /* input vars needed at each level */ + List *level_input_srfs; /* input SRFs needed at each level */ + /* These are Lists of split_pathtarget_items: */ + List *current_input_vars; /* vars needed in current subexpr */ + List *current_input_srfs; /* SRFs needed in current subexpr */ + /* Auxiliary data for current split_pathtarget_walker traversal: */ + int current_depth; /* max SRF depth in current subexpr */ + Index current_sgref; /* current subexpr's sortgroupref, or 0 */ +} split_pathtarget_context; + +static bool split_pathtarget_walker(Node *node, + split_pathtarget_context *context); +static void add_sp_item_to_pathtarget(PathTarget *target, + split_pathtarget_item *item); +static void add_sp_items_to_pathtarget(PathTarget *target, List *items); + + +/***************************************************************************** + * Target list creation and searching utilities + *****************************************************************************/ + +/* + * tlist_member + * Finds the (first) member of the given tlist whose expression is + * equal() to the given expression. Result is NULL if no such member. + */ +TargetEntry * +tlist_member(Expr *node, List *targetlist) +{ + ListCell *temp; + + foreach(temp, targetlist) + { + TargetEntry *tlentry = (TargetEntry *) lfirst(temp); + + if (equal(node, tlentry->expr)) + return tlentry; + } + return NULL; +} + +/* + * tlist_member_match_var + * Same as above, except that we match the provided Var on the basis + * of varno/varattno/varlevelsup/vartype only, rather than full equal(). + * + * This is needed in some cases where we can't be sure of an exact typmod + * match. For safety, though, we insist on vartype match. + */ +static TargetEntry * +tlist_member_match_var(Var *var, List *targetlist) +{ + ListCell *temp; + + foreach(temp, targetlist) + { + TargetEntry *tlentry = (TargetEntry *) lfirst(temp); + Var *tlvar = (Var *) tlentry->expr; + + if (!tlvar || !IsA(tlvar, Var)) + continue; + if (var->varno == tlvar->varno && + var->varattno == tlvar->varattno && + var->varlevelsup == tlvar->varlevelsup && + var->vartype == tlvar->vartype) + return tlentry; + } + return NULL; +} + +/* + * add_to_flat_tlist + * Add more items to a flattened tlist (if they're not already in it) + * + * 'tlist' is the flattened tlist + * 'exprs' is a list of expressions (usually, but not necessarily, Vars) + * + * Returns the extended tlist. + */ +List * +add_to_flat_tlist(List *tlist, List *exprs) +{ + int next_resno = list_length(tlist) + 1; + ListCell *lc; + + foreach(lc, exprs) + { + Expr *expr = (Expr *) lfirst(lc); + + if (!tlist_member(expr, tlist)) + { + TargetEntry *tle; + + tle = makeTargetEntry(copyObject(expr), /* copy needed?? */ + next_resno++, + NULL, + false); + tlist = lappend(tlist, tle); + } + } + return tlist; +} + + +/* + * get_tlist_exprs + * Get just the expression subtrees of a tlist + * + * Resjunk columns are ignored unless includeJunk is true + */ +List * +get_tlist_exprs(List *tlist, bool includeJunk) +{ + List *result = NIL; + ListCell *l; + + foreach(l, tlist) + { + TargetEntry *tle = (TargetEntry *) lfirst(l); + + if (tle->resjunk && !includeJunk) + continue; + + result = lappend(result, tle->expr); + } + return result; +} + + +/* + * count_nonjunk_tlist_entries + * What it says ... + */ +int +count_nonjunk_tlist_entries(List *tlist) +{ + int len = 0; + ListCell *l; + + foreach(l, tlist) + { + TargetEntry *tle = (TargetEntry *) lfirst(l); + + if (!tle->resjunk) + len++; + } + return len; +} + + +/* + * tlist_same_exprs + * Check whether two target lists contain the same expressions + * + * Note: this function is used to decide whether it's safe to jam a new tlist + * into a non-projection-capable plan node. Obviously we can't do that unless + * the node's tlist shows it already returns the column values we want. + * However, we can ignore the TargetEntry attributes resname, ressortgroupref, + * resorigtbl, resorigcol, and resjunk, because those are only labelings that + * don't affect the row values computed by the node. (Moreover, if we didn't + * ignore them, we'd frequently fail to make the desired optimization, since + * the planner tends to not bother to make resname etc. valid in intermediate + * plan nodes.) Note that on success, the caller must still jam the desired + * tlist into the plan node, else it won't have the desired labeling fields. + */ +bool +tlist_same_exprs(List *tlist1, List *tlist2) +{ + ListCell *lc1, + *lc2; + + if (list_length(tlist1) != list_length(tlist2)) + return false; /* not same length, so can't match */ + + forboth(lc1, tlist1, lc2, tlist2) + { + TargetEntry *tle1 = (TargetEntry *) lfirst(lc1); + TargetEntry *tle2 = (TargetEntry *) lfirst(lc2); + + if (!equal(tle1->expr, tle2->expr)) + return false; + } + + return true; +} + + +/* + * Does tlist have same output datatypes as listed in colTypes? + * + * Resjunk columns are ignored if junkOK is true; otherwise presence of + * a resjunk column will always cause a 'false' result. + * + * Note: currently no callers care about comparing typmods. + */ +bool +tlist_same_datatypes(List *tlist, List *colTypes, bool junkOK) +{ + ListCell *l; + ListCell *curColType = list_head(colTypes); + + foreach(l, tlist) + { + TargetEntry *tle = (TargetEntry *) lfirst(l); + + if (tle->resjunk) + { + if (!junkOK) + return false; + } + else + { + if (curColType == NULL) + return false; /* tlist longer than colTypes */ + if (exprType((Node *) tle->expr) != lfirst_oid(curColType)) + return false; + curColType = lnext(colTypes, curColType); + } + } + if (curColType != NULL) + return false; /* tlist shorter than colTypes */ + return true; +} + +/* + * Does tlist have same exposed collations as listed in colCollations? + * + * Identical logic to the above, but for collations. + */ +bool +tlist_same_collations(List *tlist, List *colCollations, bool junkOK) +{ + ListCell *l; + ListCell *curColColl = list_head(colCollations); + + foreach(l, tlist) + { + TargetEntry *tle = (TargetEntry *) lfirst(l); + + if (tle->resjunk) + { + if (!junkOK) + return false; + } + else + { + if (curColColl == NULL) + return false; /* tlist longer than colCollations */ + if (exprCollation((Node *) tle->expr) != lfirst_oid(curColColl)) + return false; + curColColl = lnext(colCollations, curColColl); + } + } + if (curColColl != NULL) + return false; /* tlist shorter than colCollations */ + return true; +} + +/* + * apply_tlist_labeling + * Apply the TargetEntry labeling attributes of src_tlist to dest_tlist + * + * This is useful for reattaching column names etc to a plan's final output + * targetlist. + */ +void +apply_tlist_labeling(List *dest_tlist, List *src_tlist) +{ + ListCell *ld, + *ls; + + Assert(list_length(dest_tlist) == list_length(src_tlist)); + forboth(ld, dest_tlist, ls, src_tlist) + { + TargetEntry *dest_tle = (TargetEntry *) lfirst(ld); + TargetEntry *src_tle = (TargetEntry *) lfirst(ls); + + Assert(dest_tle->resno == src_tle->resno); + dest_tle->resname = src_tle->resname; + dest_tle->ressortgroupref = src_tle->ressortgroupref; + dest_tle->resorigtbl = src_tle->resorigtbl; + dest_tle->resorigcol = src_tle->resorigcol; + dest_tle->resjunk = src_tle->resjunk; + } +} + + +/* + * get_sortgroupref_tle + * Find the targetlist entry matching the given SortGroupRef index, + * and return it. + */ +TargetEntry * +get_sortgroupref_tle(Index sortref, List *targetList) +{ + ListCell *l; + + foreach(l, targetList) + { + TargetEntry *tle = (TargetEntry *) lfirst(l); + + if (tle->ressortgroupref == sortref) + return tle; + } + + elog(ERROR, "ORDER/GROUP BY expression not found in targetlist"); + return NULL; /* keep compiler quiet */ +} + +/* + * get_sortgroupclause_tle + * Find the targetlist entry matching the given SortGroupClause + * by ressortgroupref, and return it. + */ +TargetEntry * +get_sortgroupclause_tle(SortGroupClause *sgClause, + List *targetList) +{ + return get_sortgroupref_tle(sgClause->tleSortGroupRef, targetList); +} + +/* + * get_sortgroupclause_expr + * Find the targetlist entry matching the given SortGroupClause + * by ressortgroupref, and return its expression. + */ +Node * +get_sortgroupclause_expr(SortGroupClause *sgClause, List *targetList) +{ + TargetEntry *tle = get_sortgroupclause_tle(sgClause, targetList); + + return (Node *) tle->expr; +} + +/* + * get_sortgrouplist_exprs + * Given a list of SortGroupClauses, build a list + * of the referenced targetlist expressions. + */ +List * +get_sortgrouplist_exprs(List *sgClauses, List *targetList) +{ + List *result = NIL; + ListCell *l; + + foreach(l, sgClauses) + { + SortGroupClause *sortcl = (SortGroupClause *) lfirst(l); + Node *sortexpr; + + sortexpr = get_sortgroupclause_expr(sortcl, targetList); + result = lappend(result, sortexpr); + } + return result; +} + + +/***************************************************************************** + * Functions to extract data from a list of SortGroupClauses + * + * These don't really belong in tlist.c, but they are sort of related to the + * functions just above, and they don't seem to deserve their own file. + *****************************************************************************/ + +/* + * get_sortgroupref_clause + * Find the SortGroupClause matching the given SortGroupRef index, + * and return it. + */ +SortGroupClause * +get_sortgroupref_clause(Index sortref, List *clauses) +{ + ListCell *l; + + foreach(l, clauses) + { + SortGroupClause *cl = (SortGroupClause *) lfirst(l); + + if (cl->tleSortGroupRef == sortref) + return cl; + } + + elog(ERROR, "ORDER/GROUP BY expression not found in list"); + return NULL; /* keep compiler quiet */ +} + +/* + * get_sortgroupref_clause_noerr + * As above, but return NULL rather than throwing an error if not found. + */ +SortGroupClause * +get_sortgroupref_clause_noerr(Index sortref, List *clauses) +{ + ListCell *l; + + foreach(l, clauses) + { + SortGroupClause *cl = (SortGroupClause *) lfirst(l); + + if (cl->tleSortGroupRef == sortref) + return cl; + } + + return NULL; +} + +/* + * extract_grouping_ops - make an array of the equality operator OIDs + * for a SortGroupClause list + */ +Oid * +extract_grouping_ops(List *groupClause) +{ + int numCols = list_length(groupClause); + int colno = 0; + Oid *groupOperators; + ListCell *glitem; + + groupOperators = (Oid *) palloc(sizeof(Oid) * numCols); + + foreach(glitem, groupClause) + { + SortGroupClause *groupcl = (SortGroupClause *) lfirst(glitem); + + groupOperators[colno] = groupcl->eqop; + Assert(OidIsValid(groupOperators[colno])); + colno++; + } + + return groupOperators; +} + +/* + * extract_grouping_collations - make an array of the grouping column collations + * for a SortGroupClause list + */ +Oid * +extract_grouping_collations(List *groupClause, List *tlist) +{ + int numCols = list_length(groupClause); + int colno = 0; + Oid *grpCollations; + ListCell *glitem; + + grpCollations = (Oid *) palloc(sizeof(Oid) * numCols); + + foreach(glitem, groupClause) + { + SortGroupClause *groupcl = (SortGroupClause *) lfirst(glitem); + TargetEntry *tle = get_sortgroupclause_tle(groupcl, tlist); + + grpCollations[colno++] = exprCollation((Node *) tle->expr); + } + + return grpCollations; +} + +/* + * extract_grouping_cols - make an array of the grouping column resnos + * for a SortGroupClause list + */ +AttrNumber * +extract_grouping_cols(List *groupClause, List *tlist) +{ + AttrNumber *grpColIdx; + int numCols = list_length(groupClause); + int colno = 0; + ListCell *glitem; + + grpColIdx = (AttrNumber *) palloc(sizeof(AttrNumber) * numCols); + + foreach(glitem, groupClause) + { + SortGroupClause *groupcl = (SortGroupClause *) lfirst(glitem); + TargetEntry *tle = get_sortgroupclause_tle(groupcl, tlist); + + grpColIdx[colno++] = tle->resno; + } + + return grpColIdx; +} + +/* + * grouping_is_sortable - is it possible to implement grouping list by sorting? + * + * This is easy since the parser will have included a sortop if one exists. + */ +bool +grouping_is_sortable(List *groupClause) +{ + ListCell *glitem; + + foreach(glitem, groupClause) + { + SortGroupClause *groupcl = (SortGroupClause *) lfirst(glitem); + + if (!OidIsValid(groupcl->sortop)) + return false; + } + return true; +} + +/* + * grouping_is_hashable - is it possible to implement grouping list by hashing? + * + * We rely on the parser to have set the hashable flag correctly. + */ +bool +grouping_is_hashable(List *groupClause) +{ + ListCell *glitem; + + foreach(glitem, groupClause) + { + SortGroupClause *groupcl = (SortGroupClause *) lfirst(glitem); + + if (!groupcl->hashable) + return false; + } + return true; +} + + +/***************************************************************************** + * PathTarget manipulation functions + * + * PathTarget is a somewhat stripped-down version of a full targetlist; it + * omits all the TargetEntry decoration except (optionally) sortgroupref data, + * and it adds evaluation cost and output data width info. + *****************************************************************************/ + +/* + * make_pathtarget_from_tlist + * Construct a PathTarget equivalent to the given targetlist. + * + * This leaves the cost and width fields as zeroes. Most callers will want + * to use create_pathtarget(), so as to get those set. + */ +PathTarget * +make_pathtarget_from_tlist(List *tlist) +{ + PathTarget *target = makeNode(PathTarget); + int i; + ListCell *lc; + + target->sortgrouprefs = (Index *) palloc(list_length(tlist) * sizeof(Index)); + + i = 0; + foreach(lc, tlist) + { + TargetEntry *tle = (TargetEntry *) lfirst(lc); + + target->exprs = lappend(target->exprs, tle->expr); + target->sortgrouprefs[i] = tle->ressortgroupref; + i++; + } + + /* + * Mark volatility as unknown. The contain_volatile_functions function + * will determine if there are any volatile functions when called for the + * first time with this PathTarget. + */ + target->has_volatile_expr = VOLATILITY_UNKNOWN; + + return target; +} + +/* + * make_tlist_from_pathtarget + * Construct a targetlist from a PathTarget. + */ +List * +make_tlist_from_pathtarget(PathTarget *target) +{ + List *tlist = NIL; + int i; + ListCell *lc; + + i = 0; + foreach(lc, target->exprs) + { + Expr *expr = (Expr *) lfirst(lc); + TargetEntry *tle; + + tle = makeTargetEntry(expr, + i + 1, + NULL, + false); + if (target->sortgrouprefs) + tle->ressortgroupref = target->sortgrouprefs[i]; + tlist = lappend(tlist, tle); + i++; + } + + return tlist; +} + +/* + * copy_pathtarget + * Copy a PathTarget. + * + * The new PathTarget has its own exprs List, but shares the underlying + * target expression trees with the old one. + */ +PathTarget * +copy_pathtarget(PathTarget *src) +{ + PathTarget *dst = makeNode(PathTarget); + + /* Copy scalar fields */ + memcpy(dst, src, sizeof(PathTarget)); + /* Shallow-copy the expression list */ + dst->exprs = list_copy(src->exprs); + /* Duplicate sortgrouprefs if any (if not, the memcpy handled this) */ + if (src->sortgrouprefs) + { + Size nbytes = list_length(src->exprs) * sizeof(Index); + + dst->sortgrouprefs = (Index *) palloc(nbytes); + memcpy(dst->sortgrouprefs, src->sortgrouprefs, nbytes); + } + return dst; +} + +/* + * create_empty_pathtarget + * Create an empty (zero columns, zero cost) PathTarget. + */ +PathTarget * +create_empty_pathtarget(void) +{ + /* This is easy, but we don't want callers to hard-wire this ... */ + return makeNode(PathTarget); +} + +/* + * add_column_to_pathtarget + * Append a target column to the PathTarget. + * + * As with make_pathtarget_from_tlist, we leave it to the caller to update + * the cost and width fields. + */ +void +add_column_to_pathtarget(PathTarget *target, Expr *expr, Index sortgroupref) +{ + /* Updating the exprs list is easy ... */ + target->exprs = lappend(target->exprs, expr); + /* ... the sortgroupref data, a bit less so */ + if (target->sortgrouprefs) + { + int nexprs = list_length(target->exprs); + + /* This might look inefficient, but actually it's usually cheap */ + target->sortgrouprefs = (Index *) + repalloc(target->sortgrouprefs, nexprs * sizeof(Index)); + target->sortgrouprefs[nexprs - 1] = sortgroupref; + } + else if (sortgroupref) + { + /* Adding sortgroupref labeling to a previously unlabeled target */ + int nexprs = list_length(target->exprs); + + target->sortgrouprefs = (Index *) palloc0(nexprs * sizeof(Index)); + target->sortgrouprefs[nexprs - 1] = sortgroupref; + } + + /* + * Reset has_volatile_expr to UNKNOWN. We just leave it up to + * contain_volatile_functions to set this properly again. Technically we + * could save some effort here and just check the new Expr, but it seems + * better to keep the logic for setting this flag in one location rather + * than duplicating the logic here. + */ + if (target->has_volatile_expr == VOLATILITY_NOVOLATILE) + target->has_volatile_expr = VOLATILITY_UNKNOWN; +} + +/* + * add_new_column_to_pathtarget + * Append a target column to the PathTarget, but only if it's not + * equal() to any pre-existing target expression. + * + * The caller cannot specify a sortgroupref, since it would be unclear how + * to merge that with a pre-existing column. + * + * As with make_pathtarget_from_tlist, we leave it to the caller to update + * the cost and width fields. + */ +void +add_new_column_to_pathtarget(PathTarget *target, Expr *expr) +{ + if (!list_member(target->exprs, expr)) + add_column_to_pathtarget(target, expr, 0); +} + +/* + * add_new_columns_to_pathtarget + * Apply add_new_column_to_pathtarget() for each element of the list. + */ +void +add_new_columns_to_pathtarget(PathTarget *target, List *exprs) +{ + ListCell *lc; + + foreach(lc, exprs) + { + Expr *expr = (Expr *) lfirst(lc); + + add_new_column_to_pathtarget(target, expr); + } +} + +/* + * apply_pathtarget_labeling_to_tlist + * Apply any sortgrouprefs in the PathTarget to matching tlist entries + * + * Here, we do not assume that the tlist entries are one-for-one with the + * PathTarget. The intended use of this function is to deal with cases + * where createplan.c has decided to use some other tlist and we have + * to identify what matches exist. + */ +void +apply_pathtarget_labeling_to_tlist(List *tlist, PathTarget *target) +{ + int i; + ListCell *lc; + + /* Nothing to do if PathTarget has no sortgrouprefs data */ + if (target->sortgrouprefs == NULL) + return; + + i = 0; + foreach(lc, target->exprs) + { + Expr *expr = (Expr *) lfirst(lc); + TargetEntry *tle; + + if (target->sortgrouprefs[i]) + { + /* + * For Vars, use tlist_member_match_var's weakened matching rule; + * this allows us to deal with some cases where a set-returning + * function has been inlined, so that we now have more knowledge + * about what it returns than we did when the original Var was + * created. Otherwise, use regular equal() to find the matching + * TLE. (In current usage, only the Var case is actually needed; + * but it seems best to have sane behavior here for non-Vars too.) + */ + if (expr && IsA(expr, Var)) + tle = tlist_member_match_var((Var *) expr, tlist); + else + tle = tlist_member(expr, tlist); + + /* + * Complain if noplace for the sortgrouprefs label, or if we'd + * have to label a column twice. (The case where it already has + * the desired label probably can't happen, but we may as well + * allow for it.) + */ + if (!tle) + elog(ERROR, "ORDER/GROUP BY expression not found in targetlist"); + if (tle->ressortgroupref != 0 && + tle->ressortgroupref != target->sortgrouprefs[i]) + elog(ERROR, "targetlist item has multiple sortgroupref labels"); + + tle->ressortgroupref = target->sortgrouprefs[i]; + } + i++; + } +} + +/* + * split_pathtarget_at_srfs + * Split given PathTarget into multiple levels to position SRFs safely + * + * The executor can only handle set-returning functions that appear at the + * top level of the targetlist of a ProjectSet plan node. If we have any SRFs + * that are not at top level, we need to split up the evaluation into multiple + * plan levels in which each level satisfies this constraint. This function + * creates appropriate PathTarget(s) for each level. + * + * As an example, consider the tlist expression + * x + srf1(srf2(y + z)) + * This expression should appear as-is in the top PathTarget, but below that + * we must have a PathTarget containing + * x, srf1(srf2(y + z)) + * and below that, another PathTarget containing + * x, srf2(y + z) + * and below that, another PathTarget containing + * x, y, z + * When these tlists are processed by setrefs.c, subexpressions that match + * output expressions of the next lower tlist will be replaced by Vars, + * so that what the executor gets are tlists looking like + * Var1 + Var2 + * Var1, srf1(Var2) + * Var1, srf2(Var2 + Var3) + * x, y, z + * which satisfy the desired property. + * + * Another example is + * srf1(x), srf2(srf3(y)) + * That must appear as-is in the top PathTarget, but below that we need + * srf1(x), srf3(y) + * That is, each SRF must be computed at a level corresponding to the nesting + * depth of SRFs within its arguments. + * + * In some cases, a SRF has already been evaluated in some previous plan level + * and we shouldn't expand it again (that is, what we see in the target is + * already meant as a reference to a lower subexpression). So, don't expand + * any tlist expressions that appear in input_target, if that's not NULL. + * + * It's also important that we preserve any sortgroupref annotation appearing + * in the given target, especially on expressions matching input_target items. + * + * The outputs of this function are two parallel lists, one a list of + * PathTargets and the other an integer list of bool flags indicating + * whether the corresponding PathTarget contains any evaluable SRFs. + * The lists are given in the order they'd need to be evaluated in, with + * the "lowest" PathTarget first. So the last list entry is always the + * originally given PathTarget, and any entries before it indicate evaluation + * levels that must be inserted below it. The first list entry must not + * contain any SRFs (other than ones duplicating input_target entries), since + * it will typically be attached to a plan node that cannot evaluate SRFs. + * + * Note: using a list for the flags may seem like overkill, since there + * are only a few possible patterns for which levels contain SRFs. + * But this representation decouples callers from that knowledge. + */ +void +split_pathtarget_at_srfs(PlannerInfo *root, + PathTarget *target, PathTarget *input_target, + List **targets, List **targets_contain_srfs) +{ + split_pathtarget_context context; + int max_depth; + bool need_extra_projection; + List *prev_level_tlist; + int lci; + ListCell *lc, + *lc1, + *lc2, + *lc3; + + /* + * It's not unusual for planner.c to pass us two physically identical + * targets, in which case we can conclude without further ado that all + * expressions are available from the input. (The logic below would + * arrive at the same conclusion, but much more tediously.) + */ + if (target == input_target) + { + *targets = list_make1(target); + *targets_contain_srfs = list_make1_int(false); + return; + } + + /* Pass any input_target exprs down to split_pathtarget_walker() */ + context.input_target_exprs = input_target ? input_target->exprs : NIL; + + /* + * Initialize with empty level-zero lists, and no levels after that. + * (Note: we could dispense with representing level zero explicitly, since + * it will never receive any SRFs, but then we'd have to special-case that + * level when we get to building result PathTargets. Level zero describes + * the SRF-free PathTarget that will be given to the input plan node.) + */ + context.level_srfs = list_make1(NIL); + context.level_input_vars = list_make1(NIL); + context.level_input_srfs = list_make1(NIL); + + /* Initialize data we'll accumulate across all the target expressions */ + context.current_input_vars = NIL; + context.current_input_srfs = NIL; + max_depth = 0; + need_extra_projection = false; + + /* Scan each expression in the PathTarget looking for SRFs */ + lci = 0; + foreach(lc, target->exprs) + { + Node *node = (Node *) lfirst(lc); + + /* Tell split_pathtarget_walker about this expr's sortgroupref */ + context.current_sgref = get_pathtarget_sortgroupref(target, lci); + lci++; + + /* + * Find all SRFs and Vars (and Var-like nodes) in this expression, and + * enter them into appropriate lists within the context struct. + */ + context.current_depth = 0; + split_pathtarget_walker(node, &context); + + /* An expression containing no SRFs is of no further interest */ + if (context.current_depth == 0) + continue; + + /* + * Track max SRF nesting depth over the whole PathTarget. Also, if + * this expression establishes a new max depth, we no longer care + * whether previous expressions contained nested SRFs; we can handle + * any required projection for them in the final ProjectSet node. + */ + if (max_depth < context.current_depth) + { + max_depth = context.current_depth; + need_extra_projection = false; + } + + /* + * If any maximum-depth SRF is not at the top level of its expression, + * we'll need an extra Result node to compute the top-level scalar + * expression. + */ + if (max_depth == context.current_depth && !IS_SRF_CALL(node)) + need_extra_projection = true; + } + + /* + * If we found no SRFs needing evaluation (maybe they were all present in + * input_target, or maybe they were all removed by const-simplification), + * then no ProjectSet is needed; fall out. + */ + if (max_depth == 0) + { + *targets = list_make1(target); + *targets_contain_srfs = list_make1_int(false); + return; + } + + /* + * The Vars and SRF outputs needed at top level can be added to the last + * level_input lists if we don't need an extra projection step. If we do + * need one, add a SRF-free level to the lists. + */ + if (need_extra_projection) + { + context.level_srfs = lappend(context.level_srfs, NIL); + context.level_input_vars = lappend(context.level_input_vars, + context.current_input_vars); + context.level_input_srfs = lappend(context.level_input_srfs, + context.current_input_srfs); + } + else + { + lc = list_nth_cell(context.level_input_vars, max_depth); + lfirst(lc) = list_concat(lfirst(lc), context.current_input_vars); + lc = list_nth_cell(context.level_input_srfs, max_depth); + lfirst(lc) = list_concat(lfirst(lc), context.current_input_srfs); + } + + /* + * Now construct the output PathTargets. The original target can be used + * as-is for the last one, but we need to construct a new SRF-free target + * representing what the preceding plan node has to emit, as well as a + * target for each intermediate ProjectSet node. + */ + *targets = *targets_contain_srfs = NIL; + prev_level_tlist = NIL; + + forthree(lc1, context.level_srfs, + lc2, context.level_input_vars, + lc3, context.level_input_srfs) + { + List *level_srfs = (List *) lfirst(lc1); + PathTarget *ntarget; + + if (lnext(context.level_srfs, lc1) == NULL) + { + ntarget = target; + } + else + { + ntarget = create_empty_pathtarget(); + + /* + * This target should actually evaluate any SRFs of the current + * level, and it needs to propagate forward any Vars needed by + * later levels, as well as SRFs computed earlier and needed by + * later levels. + */ + add_sp_items_to_pathtarget(ntarget, level_srfs); + for_each_cell(lc, context.level_input_vars, + lnext(context.level_input_vars, lc2)) + { + List *input_vars = (List *) lfirst(lc); + + add_sp_items_to_pathtarget(ntarget, input_vars); + } + for_each_cell(lc, context.level_input_srfs, + lnext(context.level_input_srfs, lc3)) + { + List *input_srfs = (List *) lfirst(lc); + ListCell *lcx; + + foreach(lcx, input_srfs) + { + split_pathtarget_item *item = lfirst(lcx); + + if (list_member(prev_level_tlist, item->expr)) + add_sp_item_to_pathtarget(ntarget, item); + } + } + set_pathtarget_cost_width(root, ntarget); + } + + /* + * Add current target and does-it-compute-SRFs flag to output lists. + */ + *targets = lappend(*targets, ntarget); + *targets_contain_srfs = lappend_int(*targets_contain_srfs, + (level_srfs != NIL)); + + /* Remember this level's output for next pass */ + prev_level_tlist = ntarget->exprs; + } +} + +/* + * Recursively examine expressions for split_pathtarget_at_srfs. + * + * Note we make no effort here to prevent duplicate entries in the output + * lists. Duplicates will be gotten rid of later. + */ +static bool +split_pathtarget_walker(Node *node, split_pathtarget_context *context) +{ + if (node == NULL) + return false; + + /* + * A subexpression that matches an expression already computed in + * input_target can be treated like a Var (which indeed it will be after + * setrefs.c gets done with it), even if it's actually a SRF. Record it + * as being needed for the current expression, and ignore any + * substructure. (Note in particular that this preserves the identity of + * any expressions that appear as sortgrouprefs in input_target.) + */ + if (list_member(context->input_target_exprs, node)) + { + split_pathtarget_item *item = palloc(sizeof(split_pathtarget_item)); + + item->expr = node; + item->sortgroupref = context->current_sgref; + context->current_input_vars = lappend(context->current_input_vars, + item); + return false; + } + + /* + * Vars and Var-like constructs are expected to be gotten from the input, + * too. We assume that these constructs cannot contain any SRFs (if one + * does, there will be an executor failure from a misplaced SRF). + */ + if (IsA(node, Var) || + IsA(node, PlaceHolderVar) || + IsA(node, Aggref) || + IsA(node, GroupingFunc) || + IsA(node, WindowFunc)) + { + split_pathtarget_item *item = palloc(sizeof(split_pathtarget_item)); + + item->expr = node; + item->sortgroupref = context->current_sgref; + context->current_input_vars = lappend(context->current_input_vars, + item); + return false; + } + + /* + * If it's a SRF, recursively examine its inputs, determine its level, and + * make appropriate entries in the output lists. + */ + if (IS_SRF_CALL(node)) + { + split_pathtarget_item *item = palloc(sizeof(split_pathtarget_item)); + List *save_input_vars = context->current_input_vars; + List *save_input_srfs = context->current_input_srfs; + int save_current_depth = context->current_depth; + int srf_depth; + ListCell *lc; + + item->expr = node; + item->sortgroupref = context->current_sgref; + + context->current_input_vars = NIL; + context->current_input_srfs = NIL; + context->current_depth = 0; + context->current_sgref = 0; /* subexpressions are not sortgroup items */ + + (void) expression_tree_walker(node, split_pathtarget_walker, + (void *) context); + + /* Depth is one more than any SRF below it */ + srf_depth = context->current_depth + 1; + + /* If new record depth, initialize another level of output lists */ + if (srf_depth >= list_length(context->level_srfs)) + { + context->level_srfs = lappend(context->level_srfs, NIL); + context->level_input_vars = lappend(context->level_input_vars, NIL); + context->level_input_srfs = lappend(context->level_input_srfs, NIL); + } + + /* Record this SRF as needing to be evaluated at appropriate level */ + lc = list_nth_cell(context->level_srfs, srf_depth); + lfirst(lc) = lappend(lfirst(lc), item); + + /* Record its inputs as being needed at the same level */ + lc = list_nth_cell(context->level_input_vars, srf_depth); + lfirst(lc) = list_concat(lfirst(lc), context->current_input_vars); + lc = list_nth_cell(context->level_input_srfs, srf_depth); + lfirst(lc) = list_concat(lfirst(lc), context->current_input_srfs); + + /* + * Restore caller-level state and update it for presence of this SRF. + * Notice we report the SRF itself as being needed for evaluation of + * surrounding expression. + */ + context->current_input_vars = save_input_vars; + context->current_input_srfs = lappend(save_input_srfs, item); + context->current_depth = Max(save_current_depth, srf_depth); + + /* We're done here */ + return false; + } + + /* + * Otherwise, the node is a scalar (non-set) expression, so recurse to + * examine its inputs. + */ + context->current_sgref = 0; /* subexpressions are not sortgroup items */ + return expression_tree_walker(node, split_pathtarget_walker, + (void *) context); +} + +/* + * Add a split_pathtarget_item to the PathTarget, unless a matching item is + * already present. This is like add_new_column_to_pathtarget, but allows + * for sortgrouprefs to be handled. An item having zero sortgroupref can + * be merged with one that has a sortgroupref, acquiring the latter's + * sortgroupref. + * + * Note that we don't worry about possibly adding duplicate sortgrouprefs + * to the PathTarget. That would be bad, but it should be impossible unless + * the target passed to split_pathtarget_at_srfs already had duplicates. + * As long as it didn't, we can have at most one split_pathtarget_item with + * any particular nonzero sortgroupref. + */ +static void +add_sp_item_to_pathtarget(PathTarget *target, split_pathtarget_item *item) +{ + int lci; + ListCell *lc; + + /* + * Look for a pre-existing entry that is equal() and does not have a + * conflicting sortgroupref already. + */ + lci = 0; + foreach(lc, target->exprs) + { + Node *node = (Node *) lfirst(lc); + Index sgref = get_pathtarget_sortgroupref(target, lci); + + if ((item->sortgroupref == sgref || + item->sortgroupref == 0 || + sgref == 0) && + equal(item->expr, node)) + { + /* Found a match. Assign item's sortgroupref if it has one. */ + if (item->sortgroupref) + { + if (target->sortgrouprefs == NULL) + { + target->sortgrouprefs = (Index *) + palloc0(list_length(target->exprs) * sizeof(Index)); + } + target->sortgrouprefs[lci] = item->sortgroupref; + } + return; + } + lci++; + } + + /* + * No match, so add item to PathTarget. Copy the expr for safety. + */ + add_column_to_pathtarget(target, (Expr *) copyObject(item->expr), + item->sortgroupref); +} + +/* + * Apply add_sp_item_to_pathtarget to each element of list. + */ +static void +add_sp_items_to_pathtarget(PathTarget *target, List *items) +{ + ListCell *lc; + + foreach(lc, items) + { + split_pathtarget_item *item = lfirst(lc); + + add_sp_item_to_pathtarget(target, item); + } +} diff --git a/src/backend/optimizer/util/var.c b/src/backend/optimizer/util/var.c new file mode 100644 index 0000000..ebc6ce8 --- /dev/null +++ b/src/backend/optimizer/util/var.c @@ -0,0 +1,903 @@ +/*------------------------------------------------------------------------- + * + * var.c + * Var node manipulation routines + * + * Note: for most purposes, PlaceHolderVar is considered a Var too, + * even if its contained expression is variable-free. Also, CurrentOfExpr + * is treated as a Var for purposes of determining whether an expression + * contains variables. + * + * + * Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group + * Portions Copyright (c) 1994, Regents of the University of California + * + * + * IDENTIFICATION + * src/backend/optimizer/util/var.c + * + *------------------------------------------------------------------------- + */ +#include "postgres.h" + +#include "access/sysattr.h" +#include "nodes/nodeFuncs.h" +#include "optimizer/optimizer.h" +#include "optimizer/placeholder.h" +#include "optimizer/prep.h" +#include "parser/parsetree.h" +#include "rewrite/rewriteManip.h" + + +typedef struct +{ + Relids varnos; + PlannerInfo *root; + int sublevels_up; +} pull_varnos_context; + +typedef struct +{ + Bitmapset *varattnos; + Index varno; +} pull_varattnos_context; + +typedef struct +{ + List *vars; + int sublevels_up; +} pull_vars_context; + +typedef struct +{ + int var_location; + int sublevels_up; +} locate_var_of_level_context; + +typedef struct +{ + List *varlist; + int flags; +} pull_var_clause_context; + +typedef struct +{ + Query *query; /* outer Query */ + int sublevels_up; + bool possible_sublink; /* could aliases include a SubLink? */ + bool inserted_sublink; /* have we inserted a SubLink? */ +} flatten_join_alias_vars_context; + +static bool pull_varnos_walker(Node *node, + pull_varnos_context *context); +static bool pull_varattnos_walker(Node *node, pull_varattnos_context *context); +static bool pull_vars_walker(Node *node, pull_vars_context *context); +static bool contain_var_clause_walker(Node *node, void *context); +static bool contain_vars_of_level_walker(Node *node, int *sublevels_up); +static bool locate_var_of_level_walker(Node *node, + locate_var_of_level_context *context); +static bool pull_var_clause_walker(Node *node, + pull_var_clause_context *context); +static Node *flatten_join_alias_vars_mutator(Node *node, + flatten_join_alias_vars_context *context); +static Relids alias_relid_set(Query *query, Relids relids); + + +/* + * pull_varnos + * Create a set of all the distinct varnos present in a parsetree. + * Only varnos that reference level-zero rtable entries are considered. + * + * "root" can be passed as NULL if it is not necessary to process + * PlaceHolderVars. + * + * NOTE: this is used on not-yet-planned expressions. It may therefore find + * bare SubLinks, and if so it needs to recurse into them to look for uplevel + * references to the desired rtable level! But when we find a completed + * SubPlan, we only need to look at the parameters passed to the subplan. + */ +Relids +pull_varnos(PlannerInfo *root, Node *node) +{ + pull_varnos_context context; + + context.varnos = NULL; + context.root = root; + context.sublevels_up = 0; + + /* + * Must be prepared to start with a Query or a bare expression tree; if + * it's a Query, we don't want to increment sublevels_up. + */ + query_or_expression_tree_walker(node, + pull_varnos_walker, + (void *) &context, + 0); + + return context.varnos; +} + +/* + * pull_varnos_of_level + * Create a set of all the distinct varnos present in a parsetree. + * Only Vars of the specified level are considered. + */ +Relids +pull_varnos_of_level(PlannerInfo *root, Node *node, int levelsup) +{ + pull_varnos_context context; + + context.varnos = NULL; + context.root = root; + context.sublevels_up = levelsup; + + /* + * Must be prepared to start with a Query or a bare expression tree; if + * it's a Query, we don't want to increment sublevels_up. + */ + query_or_expression_tree_walker(node, + pull_varnos_walker, + (void *) &context, + 0); + + return context.varnos; +} + +static bool +pull_varnos_walker(Node *node, pull_varnos_context *context) +{ + if (node == NULL) + return false; + if (IsA(node, Var)) + { + Var *var = (Var *) node; + + if (var->varlevelsup == context->sublevels_up) + context->varnos = bms_add_member(context->varnos, var->varno); + return false; + } + if (IsA(node, CurrentOfExpr)) + { + CurrentOfExpr *cexpr = (CurrentOfExpr *) node; + + if (context->sublevels_up == 0) + context->varnos = bms_add_member(context->varnos, cexpr->cvarno); + return false; + } + if (IsA(node, PlaceHolderVar)) + { + PlaceHolderVar *phv = (PlaceHolderVar *) node; + + /* + * If a PlaceHolderVar is not of the target query level, ignore it, + * instead recursing into its expression to see if it contains any + * vars that are of the target level. We'll also do that when the + * caller doesn't pass a "root" pointer. (We probably shouldn't see + * PlaceHolderVars at all in such cases, but if we do, this is a + * reasonable behavior.) + */ + if (phv->phlevelsup == context->sublevels_up && + context->root != NULL) + { + /* + * Ideally, the PHV's contribution to context->varnos is its + * ph_eval_at set. However, this code can be invoked before + * that's been computed. If we cannot find a PlaceHolderInfo, + * fall back to the conservative assumption that the PHV will be + * evaluated at its syntactic level (phv->phrels). + * + * There is a second hazard: this code is also used to examine + * qual clauses during deconstruct_jointree, when we may have a + * PlaceHolderInfo but its ph_eval_at value is not yet final, so + * that theoretically we could obtain a relid set that's smaller + * than we'd see later on. That should never happen though, + * because we deconstruct the jointree working upwards. Any outer + * join that forces delay of evaluation of a given qual clause + * will be processed before we examine that clause here, so the + * ph_eval_at value should have been updated to include it. + * + * Another problem is that a PlaceHolderVar can appear in quals or + * tlists that have been translated for use in a child appendrel. + * Typically such a PHV is a parameter expression sourced by some + * other relation, so that the translation from parent appendrel + * to child doesn't change its phrels, and we should still take + * ph_eval_at at face value. But in corner cases, the PHV's + * original phrels can include the parent appendrel itself, in + * which case the translated PHV will have the child appendrel in + * phrels, and we must translate ph_eval_at to match. + */ + PlaceHolderInfo *phinfo = NULL; + + if (phv->phlevelsup == 0) + { + ListCell *lc; + + foreach(lc, context->root->placeholder_list) + { + phinfo = (PlaceHolderInfo *) lfirst(lc); + if (phinfo->phid == phv->phid) + break; + phinfo = NULL; + } + } + if (phinfo == NULL) + { + /* No PlaceHolderInfo yet, use phrels */ + context->varnos = bms_add_members(context->varnos, + phv->phrels); + } + else if (bms_equal(phv->phrels, phinfo->ph_var->phrels)) + { + /* Normal case: use ph_eval_at */ + context->varnos = bms_add_members(context->varnos, + phinfo->ph_eval_at); + } + else + { + /* Translated PlaceHolderVar: translate ph_eval_at to match */ + Relids newevalat, + delta; + + /* remove what was removed from phv->phrels ... */ + delta = bms_difference(phinfo->ph_var->phrels, phv->phrels); + newevalat = bms_difference(phinfo->ph_eval_at, delta); + /* ... then if that was in fact part of ph_eval_at ... */ + if (!bms_equal(newevalat, phinfo->ph_eval_at)) + { + /* ... add what was added */ + delta = bms_difference(phv->phrels, phinfo->ph_var->phrels); + newevalat = bms_join(newevalat, delta); + } + context->varnos = bms_join(context->varnos, + newevalat); + } + return false; /* don't recurse into expression */ + } + } + else if (IsA(node, Query)) + { + /* Recurse into RTE subquery or not-yet-planned sublink subquery */ + bool result; + + context->sublevels_up++; + result = query_tree_walker((Query *) node, pull_varnos_walker, + (void *) context, 0); + context->sublevels_up--; + return result; + } + return expression_tree_walker(node, pull_varnos_walker, + (void *) context); +} + + +/* + * pull_varattnos + * Find all the distinct attribute numbers present in an expression tree, + * and add them to the initial contents of *varattnos. + * Only Vars of the given varno and rtable level zero are considered. + * + * Attribute numbers are offset by FirstLowInvalidHeapAttributeNumber so that + * we can include system attributes (e.g., OID) in the bitmap representation. + * + * Currently, this does not support unplanned subqueries; that is not needed + * for current uses. It will handle already-planned SubPlan nodes, though, + * looking into only the "testexpr" and the "args" list. (The subplan cannot + * contain any other references to Vars of the current level.) + */ +void +pull_varattnos(Node *node, Index varno, Bitmapset **varattnos) +{ + pull_varattnos_context context; + + context.varattnos = *varattnos; + context.varno = varno; + + (void) pull_varattnos_walker(node, &context); + + *varattnos = context.varattnos; +} + +static bool +pull_varattnos_walker(Node *node, pull_varattnos_context *context) +{ + if (node == NULL) + return false; + if (IsA(node, Var)) + { + Var *var = (Var *) node; + + if (var->varno == context->varno && var->varlevelsup == 0) + context->varattnos = + bms_add_member(context->varattnos, + var->varattno - FirstLowInvalidHeapAttributeNumber); + return false; + } + + /* Should not find an unplanned subquery */ + Assert(!IsA(node, Query)); + + return expression_tree_walker(node, pull_varattnos_walker, + (void *) context); +} + + +/* + * pull_vars_of_level + * Create a list of all Vars (and PlaceHolderVars) referencing the + * specified query level in the given parsetree. + * + * Caution: the Vars are not copied, only linked into the list. + */ +List * +pull_vars_of_level(Node *node, int levelsup) +{ + pull_vars_context context; + + context.vars = NIL; + context.sublevels_up = levelsup; + + /* + * Must be prepared to start with a Query or a bare expression tree; if + * it's a Query, we don't want to increment sublevels_up. + */ + query_or_expression_tree_walker(node, + pull_vars_walker, + (void *) &context, + 0); + + return context.vars; +} + +static bool +pull_vars_walker(Node *node, pull_vars_context *context) +{ + if (node == NULL) + return false; + if (IsA(node, Var)) + { + Var *var = (Var *) node; + + if (var->varlevelsup == context->sublevels_up) + context->vars = lappend(context->vars, var); + return false; + } + if (IsA(node, PlaceHolderVar)) + { + PlaceHolderVar *phv = (PlaceHolderVar *) node; + + if (phv->phlevelsup == context->sublevels_up) + context->vars = lappend(context->vars, phv); + /* we don't want to look into the contained expression */ + return false; + } + if (IsA(node, Query)) + { + /* Recurse into RTE subquery or not-yet-planned sublink subquery */ + bool result; + + context->sublevels_up++; + result = query_tree_walker((Query *) node, pull_vars_walker, + (void *) context, 0); + context->sublevels_up--; + return result; + } + return expression_tree_walker(node, pull_vars_walker, + (void *) context); +} + + +/* + * contain_var_clause + * Recursively scan a clause to discover whether it contains any Var nodes + * (of the current query level). + * + * Returns true if any varnode found. + * + * Does not examine subqueries, therefore must only be used after reduction + * of sublinks to subplans! + */ +bool +contain_var_clause(Node *node) +{ + return contain_var_clause_walker(node, NULL); +} + +static bool +contain_var_clause_walker(Node *node, void *context) +{ + if (node == NULL) + return false; + if (IsA(node, Var)) + { + if (((Var *) node)->varlevelsup == 0) + return true; /* abort the tree traversal and return true */ + return false; + } + if (IsA(node, CurrentOfExpr)) + return true; + if (IsA(node, PlaceHolderVar)) + { + if (((PlaceHolderVar *) node)->phlevelsup == 0) + return true; /* abort the tree traversal and return true */ + /* else fall through to check the contained expr */ + } + return expression_tree_walker(node, contain_var_clause_walker, context); +} + + +/* + * contain_vars_of_level + * Recursively scan a clause to discover whether it contains any Var nodes + * of the specified query level. + * + * Returns true if any such Var found. + * + * Will recurse into sublinks. Also, may be invoked directly on a Query. + */ +bool +contain_vars_of_level(Node *node, int levelsup) +{ + int sublevels_up = levelsup; + + return query_or_expression_tree_walker(node, + contain_vars_of_level_walker, + (void *) &sublevels_up, + 0); +} + +static bool +contain_vars_of_level_walker(Node *node, int *sublevels_up) +{ + if (node == NULL) + return false; + if (IsA(node, Var)) + { + if (((Var *) node)->varlevelsup == *sublevels_up) + return true; /* abort tree traversal and return true */ + return false; + } + if (IsA(node, CurrentOfExpr)) + { + if (*sublevels_up == 0) + return true; + return false; + } + if (IsA(node, PlaceHolderVar)) + { + if (((PlaceHolderVar *) node)->phlevelsup == *sublevels_up) + return true; /* abort the tree traversal and return true */ + /* else fall through to check the contained expr */ + } + if (IsA(node, Query)) + { + /* Recurse into subselects */ + bool result; + + (*sublevels_up)++; + result = query_tree_walker((Query *) node, + contain_vars_of_level_walker, + (void *) sublevels_up, + 0); + (*sublevels_up)--; + return result; + } + return expression_tree_walker(node, + contain_vars_of_level_walker, + (void *) sublevels_up); +} + + +/* + * locate_var_of_level + * Find the parse location of any Var of the specified query level. + * + * Returns -1 if no such Var is in the querytree, or if they all have + * unknown parse location. (The former case is probably caller error, + * but we don't bother to distinguish it from the latter case.) + * + * Will recurse into sublinks. Also, may be invoked directly on a Query. + * + * Note: it might seem appropriate to merge this functionality into + * contain_vars_of_level, but that would complicate that function's API. + * Currently, the only uses of this function are for error reporting, + * and so shaving cycles probably isn't very important. + */ +int +locate_var_of_level(Node *node, int levelsup) +{ + locate_var_of_level_context context; + + context.var_location = -1; /* in case we find nothing */ + context.sublevels_up = levelsup; + + (void) query_or_expression_tree_walker(node, + locate_var_of_level_walker, + (void *) &context, + 0); + + return context.var_location; +} + +static bool +locate_var_of_level_walker(Node *node, + locate_var_of_level_context *context) +{ + if (node == NULL) + return false; + if (IsA(node, Var)) + { + Var *var = (Var *) node; + + if (var->varlevelsup == context->sublevels_up && + var->location >= 0) + { + context->var_location = var->location; + return true; /* abort tree traversal and return true */ + } + return false; + } + if (IsA(node, CurrentOfExpr)) + { + /* since CurrentOfExpr doesn't carry location, nothing we can do */ + return false; + } + /* No extra code needed for PlaceHolderVar; just look in contained expr */ + if (IsA(node, Query)) + { + /* Recurse into subselects */ + bool result; + + context->sublevels_up++; + result = query_tree_walker((Query *) node, + locate_var_of_level_walker, + (void *) context, + 0); + context->sublevels_up--; + return result; + } + return expression_tree_walker(node, + locate_var_of_level_walker, + (void *) context); +} + + +/* + * pull_var_clause + * Recursively pulls all Var nodes from an expression clause. + * + * Aggrefs are handled according to these bits in 'flags': + * PVC_INCLUDE_AGGREGATES include Aggrefs in output list + * PVC_RECURSE_AGGREGATES recurse into Aggref arguments + * neither flag throw error if Aggref found + * Vars within an Aggref's expression are included in the result only + * when PVC_RECURSE_AGGREGATES is specified. + * + * WindowFuncs are handled according to these bits in 'flags': + * PVC_INCLUDE_WINDOWFUNCS include WindowFuncs in output list + * PVC_RECURSE_WINDOWFUNCS recurse into WindowFunc arguments + * neither flag throw error if WindowFunc found + * Vars within a WindowFunc's expression are included in the result only + * when PVC_RECURSE_WINDOWFUNCS is specified. + * + * PlaceHolderVars are handled according to these bits in 'flags': + * PVC_INCLUDE_PLACEHOLDERS include PlaceHolderVars in output list + * PVC_RECURSE_PLACEHOLDERS recurse into PlaceHolderVar arguments + * neither flag throw error if PlaceHolderVar found + * Vars within a PHV's expression are included in the result only + * when PVC_RECURSE_PLACEHOLDERS is specified. + * + * GroupingFuncs are treated exactly like Aggrefs, and so do not need + * their own flag bits. + * + * CurrentOfExpr nodes are ignored in all cases. + * + * Upper-level vars (with varlevelsup > 0) should not be seen here, + * likewise for upper-level Aggrefs and PlaceHolderVars. + * + * Returns list of nodes found. Note the nodes themselves are not + * copied, only referenced. + * + * Does not examine subqueries, therefore must only be used after reduction + * of sublinks to subplans! + */ +List * +pull_var_clause(Node *node, int flags) +{ + pull_var_clause_context context; + + /* Assert that caller has not specified inconsistent flags */ + Assert((flags & (PVC_INCLUDE_AGGREGATES | PVC_RECURSE_AGGREGATES)) + != (PVC_INCLUDE_AGGREGATES | PVC_RECURSE_AGGREGATES)); + Assert((flags & (PVC_INCLUDE_WINDOWFUNCS | PVC_RECURSE_WINDOWFUNCS)) + != (PVC_INCLUDE_WINDOWFUNCS | PVC_RECURSE_WINDOWFUNCS)); + Assert((flags & (PVC_INCLUDE_PLACEHOLDERS | PVC_RECURSE_PLACEHOLDERS)) + != (PVC_INCLUDE_PLACEHOLDERS | PVC_RECURSE_PLACEHOLDERS)); + + context.varlist = NIL; + context.flags = flags; + + pull_var_clause_walker(node, &context); + return context.varlist; +} + +static bool +pull_var_clause_walker(Node *node, pull_var_clause_context *context) +{ + if (node == NULL) + return false; + if (IsA(node, Var)) + { + if (((Var *) node)->varlevelsup != 0) + elog(ERROR, "Upper-level Var found where not expected"); + context->varlist = lappend(context->varlist, node); + return false; + } + else if (IsA(node, Aggref)) + { + if (((Aggref *) node)->agglevelsup != 0) + elog(ERROR, "Upper-level Aggref found where not expected"); + if (context->flags & PVC_INCLUDE_AGGREGATES) + { + context->varlist = lappend(context->varlist, node); + /* we do NOT descend into the contained expression */ + return false; + } + else if (context->flags & PVC_RECURSE_AGGREGATES) + { + /* fall through to recurse into the aggregate's arguments */ + } + else + elog(ERROR, "Aggref found where not expected"); + } + else if (IsA(node, GroupingFunc)) + { + if (((GroupingFunc *) node)->agglevelsup != 0) + elog(ERROR, "Upper-level GROUPING found where not expected"); + if (context->flags & PVC_INCLUDE_AGGREGATES) + { + context->varlist = lappend(context->varlist, node); + /* we do NOT descend into the contained expression */ + return false; + } + else if (context->flags & PVC_RECURSE_AGGREGATES) + { + /* fall through to recurse into the GroupingFunc's arguments */ + } + else + elog(ERROR, "GROUPING found where not expected"); + } + else if (IsA(node, WindowFunc)) + { + /* WindowFuncs have no levelsup field to check ... */ + if (context->flags & PVC_INCLUDE_WINDOWFUNCS) + { + context->varlist = lappend(context->varlist, node); + /* we do NOT descend into the contained expressions */ + return false; + } + else if (context->flags & PVC_RECURSE_WINDOWFUNCS) + { + /* fall through to recurse into the windowfunc's arguments */ + } + else + elog(ERROR, "WindowFunc found where not expected"); + } + else if (IsA(node, PlaceHolderVar)) + { + if (((PlaceHolderVar *) node)->phlevelsup != 0) + elog(ERROR, "Upper-level PlaceHolderVar found where not expected"); + if (context->flags & PVC_INCLUDE_PLACEHOLDERS) + { + context->varlist = lappend(context->varlist, node); + /* we do NOT descend into the contained expression */ + return false; + } + else if (context->flags & PVC_RECURSE_PLACEHOLDERS) + { + /* fall through to recurse into the placeholder's expression */ + } + else + elog(ERROR, "PlaceHolderVar found where not expected"); + } + return expression_tree_walker(node, pull_var_clause_walker, + (void *) context); +} + + +/* + * flatten_join_alias_vars + * Replace Vars that reference JOIN outputs with references to the original + * relation variables instead. This allows quals involving such vars to be + * pushed down. Whole-row Vars that reference JOIN relations are expanded + * into RowExpr constructs that name the individual output Vars. This + * is necessary since we will not scan the JOIN as a base relation, which + * is the only way that the executor can directly handle whole-row Vars. + * + * This also adjusts relid sets found in some expression node types to + * substitute the contained base rels for any join relid. + * + * If a JOIN contains sub-selects that have been flattened, its join alias + * entries might now be arbitrary expressions, not just Vars. This affects + * this function in one important way: we might find ourselves inserting + * SubLink expressions into subqueries, and we must make sure that their + * Query.hasSubLinks fields get set to true if so. If there are any + * SubLinks in the join alias lists, the outer Query should already have + * hasSubLinks = true, so this is only relevant to un-flattened subqueries. + * + * NOTE: this is used on not-yet-planned expressions. We do not expect it + * to be applied directly to the whole Query, so if we see a Query to start + * with, we do want to increment sublevels_up (this occurs for LATERAL + * subqueries). + */ +Node * +flatten_join_alias_vars(Query *query, Node *node) +{ + flatten_join_alias_vars_context context; + + context.query = query; + context.sublevels_up = 0; + /* flag whether join aliases could possibly contain SubLinks */ + context.possible_sublink = query->hasSubLinks; + /* if hasSubLinks is already true, no need to work hard */ + context.inserted_sublink = query->hasSubLinks; + + return flatten_join_alias_vars_mutator(node, &context); +} + +static Node * +flatten_join_alias_vars_mutator(Node *node, + flatten_join_alias_vars_context *context) +{ + if (node == NULL) + return NULL; + if (IsA(node, Var)) + { + Var *var = (Var *) node; + RangeTblEntry *rte; + Node *newvar; + + /* No change unless Var belongs to a JOIN of the target level */ + if (var->varlevelsup != context->sublevels_up) + return node; /* no need to copy, really */ + rte = rt_fetch(var->varno, context->query->rtable); + if (rte->rtekind != RTE_JOIN) + return node; + if (var->varattno == InvalidAttrNumber) + { + /* Must expand whole-row reference */ + RowExpr *rowexpr; + List *fields = NIL; + List *colnames = NIL; + ListCell *lv; + ListCell *ln; + + Assert(list_length(rte->joinaliasvars) == list_length(rte->eref->colnames)); + forboth(lv, rte->joinaliasvars, ln, rte->eref->colnames) + { + newvar = (Node *) lfirst(lv); + /* Ignore dropped columns */ + if (newvar == NULL) + continue; + newvar = copyObject(newvar); + + /* + * If we are expanding an alias carried down from an upper + * query, must adjust its varlevelsup fields. + */ + if (context->sublevels_up != 0) + IncrementVarSublevelsUp(newvar, context->sublevels_up, 0); + /* Preserve original Var's location, if possible */ + if (IsA(newvar, Var)) + ((Var *) newvar)->location = var->location; + /* Recurse in case join input is itself a join */ + /* (also takes care of setting inserted_sublink if needed) */ + newvar = flatten_join_alias_vars_mutator(newvar, context); + fields = lappend(fields, newvar); + /* We need the names of non-dropped columns, too */ + colnames = lappend(colnames, copyObject((Node *) lfirst(ln))); + } + rowexpr = makeNode(RowExpr); + rowexpr->args = fields; + rowexpr->row_typeid = var->vartype; + rowexpr->row_format = COERCE_IMPLICIT_CAST; + /* vartype will always be RECORDOID, so we always need colnames */ + rowexpr->colnames = colnames; + rowexpr->location = var->location; + + return (Node *) rowexpr; + } + + /* Expand join alias reference */ + Assert(var->varattno > 0); + newvar = (Node *) list_nth(rte->joinaliasvars, var->varattno - 1); + Assert(newvar != NULL); + newvar = copyObject(newvar); + + /* + * If we are expanding an alias carried down from an upper query, must + * adjust its varlevelsup fields. + */ + if (context->sublevels_up != 0) + IncrementVarSublevelsUp(newvar, context->sublevels_up, 0); + + /* Preserve original Var's location, if possible */ + if (IsA(newvar, Var)) + ((Var *) newvar)->location = var->location; + + /* Recurse in case join input is itself a join */ + newvar = flatten_join_alias_vars_mutator(newvar, context); + + /* Detect if we are adding a sublink to query */ + if (context->possible_sublink && !context->inserted_sublink) + context->inserted_sublink = checkExprHasSubLink(newvar); + + return newvar; + } + if (IsA(node, PlaceHolderVar)) + { + /* Copy the PlaceHolderVar node with correct mutation of subnodes */ + PlaceHolderVar *phv; + + phv = (PlaceHolderVar *) expression_tree_mutator(node, + flatten_join_alias_vars_mutator, + (void *) context); + /* now fix PlaceHolderVar's relid sets */ + if (phv->phlevelsup == context->sublevels_up) + { + phv->phrels = alias_relid_set(context->query, + phv->phrels); + } + return (Node *) phv; + } + + if (IsA(node, Query)) + { + /* Recurse into RTE subquery or not-yet-planned sublink subquery */ + Query *newnode; + bool save_inserted_sublink; + + context->sublevels_up++; + save_inserted_sublink = context->inserted_sublink; + context->inserted_sublink = ((Query *) node)->hasSubLinks; + newnode = query_tree_mutator((Query *) node, + flatten_join_alias_vars_mutator, + (void *) context, + QTW_IGNORE_JOINALIASES); + newnode->hasSubLinks |= context->inserted_sublink; + context->inserted_sublink = save_inserted_sublink; + context->sublevels_up--; + return (Node *) newnode; + } + /* Already-planned tree not supported */ + Assert(!IsA(node, SubPlan)); + /* Shouldn't need to handle these planner auxiliary nodes here */ + Assert(!IsA(node, SpecialJoinInfo)); + Assert(!IsA(node, PlaceHolderInfo)); + Assert(!IsA(node, MinMaxAggInfo)); + + return expression_tree_mutator(node, flatten_join_alias_vars_mutator, + (void *) context); +} + +/* + * alias_relid_set: in a set of RT indexes, replace joins by their + * underlying base relids + */ +static Relids +alias_relid_set(Query *query, Relids relids) +{ + Relids result = NULL; + int rtindex; + + rtindex = -1; + while ((rtindex = bms_next_member(relids, rtindex)) >= 0) + { + RangeTblEntry *rte = rt_fetch(rtindex, query->rtable); + + if (rte->rtekind == RTE_JOIN) + result = bms_join(result, get_relids_for_join(query, rtindex)); + else + result = bms_add_member(result, rtindex); + } + return result; +} |