diff options
| author | Daniel Baumann <daniel.baumann@progress-linux.org> | 2022-10-15 13:52:53 +0000 |
|---|---|---|
| committer | Daniel Baumann <daniel.baumann@progress-linux.org> | 2022-10-15 13:52:53 +0000 |
| commit | 97d3673ec2d668050912aa6aea1816885ca6c5ab (patch) | |
| tree | f391e30e039a3d22368e9696e171f759e104c765 | |
| parent | Adding upstream version 6.3.1. (diff) | |
| download | sqlglot-97d3673ec2d668050912aa6aea1816885ca6c5ab.tar.xz sqlglot-97d3673ec2d668050912aa6aea1816885ca6c5ab.zip | |
Adding upstream version 7.1.3.upstream/7.1.3
Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>
42 files changed, 1431 insertions, 254 deletions
diff --git a/CHANGELOG.md b/CHANGELOG.md index aee697a..cb76969 100644 --- a/CHANGELOG.md +++ b/CHANGELOG.md @@ -1,6 +1,32 @@ Changelog ========= +v7.1.0 +------ + +Changes: + +- Improvement: Pretty generator now takes max\_text\_width which breaks segments into new lines +- New: exp.to\_table helper to turn table names into table expression objects +- New: int[] type parsers +- New: annotations are now generated in sql + +v7.0.0 +------ + +Changes: + +- Breaking: DISTINCT within functions now take in multiple values eg. COUNT(DISTINCT a, b). + exp.Distinct no longer uses `this` and now uses the expressions property + +- New: Expression False kwargs are now excluded from equality checks + +- New: Parse DESCRIBE and CREATE SCHEMA + +- New: DELETE and VALUES builder + +- New: Unused CTE and JOINS are now removed in the optimizer + v6.3.0 ------ diff --git a/sqlglot/__init__.py b/sqlglot/__init__.py index 0228bdd..247085b 100644 --- a/sqlglot/__init__.py +++ b/sqlglot/__init__.py @@ -23,7 +23,7 @@ from sqlglot.generator import Generator from sqlglot.parser import Parser from sqlglot.tokens import Tokenizer, TokenType -__version__ = "6.3.1" +__version__ = "7.1.3" pretty = False diff --git a/sqlglot/dialects/clickhouse.py b/sqlglot/dialects/clickhouse.py index da5c856..f446e6d 100644 --- a/sqlglot/dialects/clickhouse.py +++ b/sqlglot/dialects/clickhouse.py @@ -1,7 +1,6 @@ from sqlglot import exp from sqlglot.dialects.dialect import Dialect, inline_array_sql, var_map_sql from sqlglot.generator import Generator -from sqlglot.helper import csv from sqlglot.parser import Parser, parse_var_map from sqlglot.tokens import Tokenizer, TokenType @@ -66,7 +65,7 @@ class ClickHouse(Dialect): TRANSFORMS = { **Generator.TRANSFORMS, exp.Array: inline_array_sql, - exp.StrPosition: lambda self, e: f"position({csv(self.sql(e, 'this'), self.sql(e, 'substr'), self.sql(e, 'position'))})", + exp.StrPosition: lambda self, e: f"position({self.format_args(e.this, e.args.get('substr'), e.args.get('position'))})", exp.Final: lambda self, e: f"{self.sql(e, 'this')} FINAL", exp.Map: lambda self, e: _lower_func(var_map_sql(self, e)), exp.VarMap: lambda self, e: _lower_func(var_map_sql(self, e)), diff --git a/sqlglot/dialects/dialect.py b/sqlglot/dialects/dialect.py index f7c6cb5..531c72a 100644 --- a/sqlglot/dialects/dialect.py +++ b/sqlglot/dialects/dialect.py @@ -2,7 +2,7 @@ from enum import Enum from sqlglot import exp from sqlglot.generator import Generator -from sqlglot.helper import csv, list_get +from sqlglot.helper import list_get from sqlglot.parser import Parser from sqlglot.time import format_time from sqlglot.tokens import Tokenizer @@ -177,11 +177,11 @@ class Dialect(metaclass=_Dialect): def rename_func(name): def _rename(self, expression): args = ( - self.expressions(expression, flat=True) + expression.expressions if isinstance(expression, exp.Func) and expression.is_var_len_args - else csv(*[self.sql(e) for e in expression.args.values()]) + else expression.args.values() ) - return f"{name}({args})" + return f"{name}({self.format_args(*args)})" return _rename @@ -189,15 +189,11 @@ def rename_func(name): def approx_count_distinct_sql(self, expression): if expression.args.get("accuracy"): self.unsupported("APPROX_COUNT_DISTINCT does not support accuracy") - return f"APPROX_COUNT_DISTINCT({self.sql(expression, 'this')})" + return f"APPROX_COUNT_DISTINCT({self.format_args(expression.this)})" def if_sql(self, expression): - expressions = csv( - self.sql(expression, "this"), - self.sql(expression, "true"), - self.sql(expression, "false"), - ) + expressions = self.format_args(expression.this, expression.args.get("true"), expression.args.get("false")) return f"IF({expressions})" @@ -254,6 +250,11 @@ def no_trycast_sql(self, expression): return self.cast_sql(expression) +def no_properties_sql(self, expression): + self.unsupported("Properties unsupported") + return "" + + def str_position_sql(self, expression): this = self.sql(expression, "this") substr = self.sql(expression, "substr") @@ -275,13 +276,13 @@ def var_map_sql(self, expression): if not isinstance(keys, exp.Array) or not isinstance(values, exp.Array): self.unsupported("Cannot convert array columns into map.") - return f"MAP({self.sql(keys)}, {self.sql(values)})" + return f"MAP({self.format_args(keys, values)})" args = [] for key, value in zip(keys.expressions, values.expressions): args.append(self.sql(key)) args.append(self.sql(value)) - return f"MAP({csv(*args)})" + return f"MAP({self.format_args(*args)})" def format_time_lambda(exp_class, dialect, default=None): diff --git a/sqlglot/dialects/duckdb.py b/sqlglot/dialects/duckdb.py index e09c3dd..f3ff6d3 100644 --- a/sqlglot/dialects/duckdb.py +++ b/sqlglot/dialects/duckdb.py @@ -6,6 +6,7 @@ from sqlglot.dialects.dialect import ( arrow_json_extract_sql, format_time_lambda, no_pivot_sql, + no_properties_sql, no_safe_divide_sql, no_tablesample_sql, rename_func, @@ -68,6 +69,12 @@ def _struct_pack_sql(self, expression): return f"STRUCT_PACK({', '.join(args)})" +def _datatype_sql(self, expression): + if expression.this == exp.DataType.Type.ARRAY: + return f"{self.expressions(expression, flat=True)}[]" + return self.datatype_sql(expression) + + class DuckDB(Dialect): class Tokenizer(Tokenizer): KEYWORDS = { @@ -106,6 +113,8 @@ class DuckDB(Dialect): } class Generator(Generator): + STRUCT_DELIMITER = ("(", ")") + TRANSFORMS = { **Generator.TRANSFORMS, exp.ApproxDistinct: approx_count_distinct_sql, @@ -113,8 +122,9 @@ class DuckDB(Dialect): exp.ArraySize: rename_func("ARRAY_LENGTH"), exp.ArraySort: _array_sort_sql, exp.ArraySum: rename_func("LIST_SUM"), + exp.DataType: _datatype_sql, exp.DateAdd: _date_add, - exp.DateDiff: lambda self, e: f"""DATE_DIFF({self.sql(e, 'unit') or "'day'"}, {self.sql(e, 'expression')}, {self.sql(e, 'this')})""", + exp.DateDiff: lambda self, e: f"""DATE_DIFF({self.format_args(e.args.get("unit") or "'day'", e.expression, e.this)})""", exp.DateStrToDate: lambda self, e: f"CAST({self.sql(e, 'this')} AS DATE)", exp.DateToDi: lambda self, e: f"CAST(STRFTIME({self.sql(e, 'this')}, {DuckDB.dateint_format}) AS INT)", exp.DiToDate: lambda self, e: f"CAST(STRPTIME(CAST({self.sql(e, 'this')} AS TEXT), {DuckDB.dateint_format}) AS DATE)", @@ -124,6 +134,7 @@ class DuckDB(Dialect): exp.JSONBExtract: arrow_json_extract_sql, exp.JSONBExtractScalar: arrow_json_extract_scalar_sql, exp.Pivot: no_pivot_sql, + exp.Properties: no_properties_sql, exp.RegexpLike: rename_func("REGEXP_MATCHES"), exp.RegexpSplit: rename_func("STR_SPLIT_REGEX"), exp.SafeDivide: no_safe_divide_sql, diff --git a/sqlglot/dialects/hive.py b/sqlglot/dialects/hive.py index 55d7bcc..8888df8 100644 --- a/sqlglot/dialects/hive.py +++ b/sqlglot/dialects/hive.py @@ -14,7 +14,7 @@ from sqlglot.dialects.dialect import ( var_map_sql, ) from sqlglot.generator import Generator -from sqlglot.helper import csv, list_get +from sqlglot.helper import list_get from sqlglot.parser import Parser, parse_var_map from sqlglot.tokens import Tokenizer @@ -32,7 +32,7 @@ def _property_sql(self, expression): def _str_to_unix(self, expression): - return f"UNIX_TIMESTAMP({csv(self.sql(expression, 'this'), _time_format(self, expression))})" + return f"UNIX_TIMESTAMP({self.format_args(expression.this, _time_format(self, expression))})" def _str_to_date(self, expression): @@ -226,7 +226,7 @@ class Hive(Dialect): exp.SchemaCommentProperty: lambda self, e: self.naked_property(e), exp.SetAgg: rename_func("COLLECT_SET"), exp.Split: lambda self, e: f"SPLIT({self.sql(e, 'this')}, CONCAT('\\\\Q', {self.sql(e, 'expression')}))", - exp.StrPosition: lambda self, e: f"LOCATE({csv(self.sql(e, 'substr'), self.sql(e, 'this'), self.sql(e, 'position'))})", + exp.StrPosition: lambda self, e: f"LOCATE({self.format_args(e.args.get('substr'), e.this, e.args.get('position'))})", exp.StrToDate: _str_to_date, exp.StrToTime: _str_to_time, exp.StrToUnix: _str_to_unix, @@ -241,7 +241,7 @@ class Hive(Dialect): exp.TsOrDsAdd: lambda self, e: f"DATE_ADD({self.sql(e, 'this')}, {self.sql(e, 'expression')})", exp.TsOrDsToDate: _to_date_sql, exp.TryCast: no_trycast_sql, - exp.UnixToStr: lambda self, e: f"FROM_UNIXTIME({csv(self.sql(e, 'this'), _time_format(self, e))})", + exp.UnixToStr: lambda self, e: f"FROM_UNIXTIME({self.format_args(e.this, _time_format(self, e))})", exp.UnixToTime: rename_func("FROM_UNIXTIME"), exp.UnixToTimeStr: rename_func("FROM_UNIXTIME"), exp.PartitionedByProperty: lambda self, e: f"PARTITIONED BY {self.sql(e, 'value')}", diff --git a/sqlglot/dialects/postgres.py b/sqlglot/dialects/postgres.py index 731e28e..c91ff4b 100644 --- a/sqlglot/dialects/postgres.py +++ b/sqlglot/dialects/postgres.py @@ -167,6 +167,7 @@ class Postgres(Dialect): **Tokenizer.KEYWORDS, "ALWAYS": TokenType.ALWAYS, "BY DEFAULT": TokenType.BY_DEFAULT, + "COMMENT ON": TokenType.COMMENT_ON, "IDENTITY": TokenType.IDENTITY, "GENERATED": TokenType.GENERATED, "DOUBLE PRECISION": TokenType.DOUBLE, diff --git a/sqlglot/dialects/presto.py b/sqlglot/dialects/presto.py index 85647c5..8dfb2fd 100644 --- a/sqlglot/dialects/presto.py +++ b/sqlglot/dialects/presto.py @@ -11,7 +11,7 @@ from sqlglot.dialects.dialect import ( ) from sqlglot.dialects.mysql import MySQL from sqlglot.generator import Generator -from sqlglot.helper import csv, list_get +from sqlglot.helper import list_get from sqlglot.parser import Parser from sqlglot.tokens import Tokenizer, TokenType @@ -26,7 +26,7 @@ def _concat_ws_sql(self, expression): sep, *args = expression.expressions sep = self.sql(sep) if len(args) > 1: - return f"ARRAY_JOIN(ARRAY[{csv(*(self.sql(e) for e in args))}], {sep})" + return f"ARRAY_JOIN(ARRAY[{self.format_args(*args)}], {sep})" return f"ARRAY_JOIN({self.sql(args[0])}, {sep})" @@ -66,7 +66,7 @@ def _no_sort_array(self, expression): comparator = "(a, b) -> CASE WHEN a < b THEN 1 WHEN a > b THEN -1 ELSE 0 END" else: comparator = None - args = csv(self.sql(expression, "this"), comparator) + args = self.format_args(expression.this, comparator) return f"ARRAY_SORT({args})" diff --git a/sqlglot/dialects/tableau.py b/sqlglot/dialects/tableau.py index e571749..45aa041 100644 --- a/sqlglot/dialects/tableau.py +++ b/sqlglot/dialects/tableau.py @@ -1,7 +1,6 @@ from sqlglot import exp from sqlglot.dialects.dialect import Dialect from sqlglot.generator import Generator -from sqlglot.helper import list_get from sqlglot.parser import Parser @@ -16,7 +15,7 @@ def _coalesce_sql(self, expression): def _count_sql(self, expression): this = expression.this if isinstance(this, exp.Distinct): - return f"COUNTD({self.sql(this, 'this')})" + return f"COUNTD({self.expressions(this, flat=True)})" return f"COUNT({self.sql(expression, 'this')})" @@ -33,5 +32,5 @@ class Tableau(Dialect): FUNCTIONS = { **Parser.FUNCTIONS, "IFNULL": exp.Coalesce.from_arg_list, - "COUNTD": lambda args: exp.Count(this=exp.Distinct(this=list_get(args, 0))), + "COUNTD": lambda args: exp.Count(this=exp.Distinct(expressions=args)), } diff --git a/sqlglot/expressions.py b/sqlglot/expressions.py index f2ffd12..39f4452 100644 --- a/sqlglot/expressions.py +++ b/sqlglot/expressions.py @@ -1,3 +1,4 @@ +import datetime import numbers import re from collections import deque @@ -508,7 +509,7 @@ class DerivedTable(Expression): return [select.alias_or_name for select in self.selects] -class Unionable: +class Unionable(Expression): def union(self, expression, distinct=True, dialect=None, **opts): """ Builds a UNION expression. @@ -614,6 +615,10 @@ class Create(Expression): } +class Describe(Expression): + pass + + class UserDefinedFunction(Expression): arg_types = {"this": True, "expressions": False} @@ -741,6 +746,11 @@ class Check(Expression): pass +class Directory(Expression): + # https://spark.apache.org/docs/3.0.0-preview/sql-ref-syntax-dml-insert-overwrite-directory-hive.html + arg_types = {"this": True, "local": False, "row_format": False} + + class ForeignKey(Expression): arg_types = { "expressions": True, @@ -804,6 +814,18 @@ class Introducer(Expression): arg_types = {"this": True, "expression": True} +class LoadData(Expression): + arg_types = { + "this": True, + "local": False, + "overwrite": False, + "inpath": True, + "partition": False, + "input_format": False, + "serde": False, + } + + class Partition(Expression): pass @@ -1037,6 +1059,18 @@ class Reference(Expression): arg_types = {"this": True, "expressions": True} +class RowFormat(Expression): + # https://cwiki.apache.org/confluence/display/hive/languagemanual+dml + arg_types = { + "fields": False, + "escaped": False, + "collection_items": False, + "map_keys": False, + "lines": False, + "null": False, + } + + class Tuple(Expression): arg_types = {"expressions": False} @@ -1071,6 +1105,14 @@ class Subqueryable(Unionable): return [] return with_.expressions + @property + def selects(self): + raise NotImplementedError("Subqueryable objects must implement `selects`") + + @property + def named_selects(self): + raise NotImplementedError("Subqueryable objects must implement `named_selects`") + def with_( self, alias, @@ -1158,7 +1200,7 @@ class Table(Expression): } -class Union(Subqueryable, Expression): +class Union(Subqueryable): arg_types = { "with": False, "this": True, @@ -1169,7 +1211,11 @@ class Union(Subqueryable, Expression): @property def named_selects(self): - return self.args["this"].unnest().named_selects + return self.this.unnest().named_selects + + @property + def selects(self): + return self.this.unnest().selects @property def left(self): @@ -1222,7 +1268,7 @@ class Schema(Expression): arg_types = {"this": False, "expressions": True} -class Select(Subqueryable, Expression): +class Select(Subqueryable): arg_types = { "with": False, "expressions": False, @@ -2075,7 +2121,7 @@ class Bracket(Condition): class Distinct(Expression): - arg_types = {"this": False, "on": False} + arg_types = {"expressions": False, "on": False} class In(Predicate): @@ -2233,6 +2279,14 @@ class Case(Func): class Cast(Func): arg_types = {"this": True, "to": True} + @property + def name(self): + return self.this.name + + @property + def to(self): + return self.args["to"] + class TryCast(Cast): pass @@ -2666,7 +2720,7 @@ def _norm_args(expression): else: arg = _norm_arg(arg) - if arg is not None: + if arg is not None and arg is not False: args[k] = arg return args @@ -3012,6 +3066,30 @@ def update(table, properties, where=None, from_=None, dialect=None, **opts): return update +def delete(table, where=None, dialect=None, **opts): + """ + Builds a delete statement. + + Example: + >>> delete("my_table", where="id > 1").sql() + 'DELETE FROM my_table WHERE id > 1' + + Args: + where (str|Condition): sql conditional parsed into a WHERE statement + dialect (str): the dialect used to parse the input expressions. + **opts: other options to use to parse the input expressions. + + Returns: + Delete: the syntax tree for the DELETE statement. + """ + return Delete( + this=maybe_parse(table, into=Table, dialect=dialect, **opts), + where=Where(this=where) + if isinstance(where, Condition) + else maybe_parse(where, into=Where, dialect=dialect, prefix="WHERE", **opts), + ) + + def condition(expression, dialect=None, **opts): """ Initialize a logical condition expression. @@ -3131,6 +3209,25 @@ def to_identifier(alias, quoted=None): return identifier +def to_table(sql_path, **kwargs): + """ + Create a table expression from a `[catalog].[schema].[table]` sql path. Catalog and schema are optional. + Example: + >>> to_table('catalog.db.table_name').sql() + 'catalog.db.table_name' + + Args: + sql_path(str): `[catalog].[schema].[table]` string + Returns: + Table: A table expression + """ + table_parts = sql_path.split(".") + catalog, db, table_name = [ + to_identifier(x) if x is not None else x for x in [None] * (3 - len(table_parts)) + table_parts + ] + return Table(this=table_name, db=db, catalog=catalog, **kwargs) + + def alias_(expression, alias, table=False, dialect=None, quoted=None, **opts): """ Create an Alias expression. @@ -3216,6 +3313,28 @@ def table_(table, db=None, catalog=None, quoted=None): ) +def values(values, alias=None): + """Build VALUES statement. + + Example: + >>> values([(1, '2')]).sql() + "VALUES (1, '2')" + + Args: + values (list[tuple[str | Expression]]): values statements that will be converted to SQL + alias (str): optional alias + dialect (str): the dialect used to parse the input expression. + **opts: other options to use to parse the input expressions. + + Returns: + Values: the Values expression object + """ + return Values( + expressions=[convert(tup) for tup in values], + alias=to_identifier(alias) if alias else None, + ) + + def convert(value): """Convert a python value into an expression object. @@ -3246,6 +3365,12 @@ def convert(value): keys=[convert(k) for k in value.keys()], values=[convert(v) for v in value.values()], ) + if isinstance(value, datetime.datetime): + datetime_literal = Literal.string(value.strftime("%Y-%m-%d %H:%M:%S")) + return TimeStrToTime(this=datetime_literal) + if isinstance(value, datetime.date): + date_literal = Literal.string(value.strftime("%Y-%m-%d")) + return DateStrToDate(this=date_literal) raise ValueError(f"Cannot convert {value}") diff --git a/sqlglot/generator.py b/sqlglot/generator.py index b7e295d..bb7fd71 100644 --- a/sqlglot/generator.py +++ b/sqlglot/generator.py @@ -2,7 +2,7 @@ import logging from sqlglot import exp from sqlglot.errors import ErrorLevel, UnsupportedError, concat_errors -from sqlglot.helper import apply_index_offset, csv, ensure_list +from sqlglot.helper import apply_index_offset, csv from sqlglot.time import format_time from sqlglot.tokens import TokenType @@ -43,14 +43,18 @@ class Generator: Default: 3 leading_comma (bool): if the the comma is leading or trailing in select statements Default: False + max_text_width: The max number of characters in a segment before creating new lines in pretty mode. + The default is on the smaller end because the length only represents a segment and not the true + line length. + Default: 80 """ TRANSFORMS = { exp.CharacterSetProperty: lambda self, e: f"{'DEFAULT ' if e.args['default'] else ''}CHARACTER SET={self.sql(e, 'value')}", - exp.DateAdd: lambda self, e: f"DATE_ADD({self.sql(e, 'this')}, {self.sql(e, 'expression')}, {self.sql(e, 'unit')})", - exp.DateDiff: lambda self, e: f"DATEDIFF({self.sql(e, 'this')}, {self.sql(e, 'expression')})", - exp.TsOrDsAdd: lambda self, e: f"TS_OR_DS_ADD({self.sql(e, 'this')}, {self.sql(e, 'expression')}, {self.sql(e, 'unit')})", - exp.VarMap: lambda self, e: f"MAP({self.sql(e.args['keys'])}, {self.sql(e.args['values'])})", + exp.DateAdd: lambda self, e: f"DATE_ADD({self.format_args(e.this, e.expression, e.args.get('unit'))})", + exp.DateDiff: lambda self, e: f"DATEDIFF({self.format_args(e.this, e.expression)})", + exp.TsOrDsAdd: lambda self, e: f"TS_OR_DS_ADD({self.format_args(e.this, e.expression, e.args.get('unit'))})", + exp.VarMap: lambda self, e: f"MAP({self.format_args(e.args['keys'], e.args['values'])})", exp.LanguageProperty: lambda self, e: self.naked_property(e), exp.LocationProperty: lambda self, e: self.naked_property(e), exp.ReturnsProperty: lambda self, e: self.naked_property(e), @@ -111,6 +115,7 @@ class Generator: "_replace_backslash", "_escaped_quote_end", "_leading_comma", + "_max_text_width", ) def __init__( @@ -135,6 +140,7 @@ class Generator: null_ordering=None, max_unsupported=3, leading_comma=False, + max_text_width=80, ): import sqlglot @@ -162,6 +168,7 @@ class Generator: self._replace_backslash = self.escape == "\\" self._escaped_quote_end = self.escape + self.quote_end self._leading_comma = leading_comma + self._max_text_width = max_text_width def generate(self, expression): """ @@ -268,7 +275,7 @@ class Generator: raise ValueError(f"Unsupported expression type {expression.__class__.__name__}") def annotation_sql(self, expression): - return self.sql(expression, "expression") + return f"{self.sql(expression, 'expression')} # {expression.name.strip()}" def uncache_sql(self, expression): table = self.sql(expression, "this") @@ -364,6 +371,9 @@ class Generator: ) return self.prepend_ctes(expression, expression_sql) + def describe_sql(self, expression): + return f"DESCRIBE {self.sql(expression, 'this')}" + def prepend_ctes(self, expression, sql): with_ = self.sql(expression, "with") if with_: @@ -405,6 +415,12 @@ class Generator: ) return f"{type_sql}{nested}" + def directory_sql(self, expression): + local = "LOCAL " if expression.args.get("local") else "" + row_format = self.sql(expression, "row_format") + row_format = f" {row_format}" if row_format else "" + return f"{local}DIRECTORY {self.sql(expression, 'this')}{row_format}" + def delete_sql(self, expression): this = self.sql(expression, "this") where_sql = self.sql(expression, "where") @@ -513,13 +529,19 @@ class Generator: return f"{key}={value}" def insert_sql(self, expression): - kind = "OVERWRITE TABLE" if expression.args.get("overwrite") else "INTO" - this = self.sql(expression, "this") + overwrite = expression.args.get("overwrite") + + if isinstance(expression.this, exp.Directory): + this = "OVERWRITE " if overwrite else "INTO " + else: + this = "OVERWRITE TABLE " if overwrite else "INTO " + + this = f"{this}{self.sql(expression, 'this')}" exists = " IF EXISTS " if expression.args.get("exists") else " " partition_sql = self.sql(expression, "partition") if expression.args.get("partition") else "" expression_sql = self.sql(expression, "expression") sep = self.sep() if partition_sql else "" - sql = f"INSERT {kind} {this}{exists}{partition_sql}{sep}{expression_sql}" + sql = f"INSERT {this}{exists}{partition_sql}{sep}{expression_sql}" return self.prepend_ctes(expression, sql) def intersect_sql(self, expression): @@ -534,6 +556,21 @@ class Generator: def introducer_sql(self, expression): return f"{self.sql(expression, 'this')} {self.sql(expression, 'expression')}" + def rowformat_sql(self, expression): + fields = expression.args.get("fields") + fields = f" FIELDS TERMINATED BY {fields}" if fields else "" + escaped = expression.args.get("escaped") + escaped = f" ESCAPED BY {escaped}" if escaped else "" + items = expression.args.get("collection_items") + items = f" COLLECTION ITEMS TERMINATED BY {items}" if items else "" + keys = expression.args.get("map_keys") + keys = f" MAP KEYS TERMINATED BY {keys}" if keys else "" + lines = expression.args.get("lines") + lines = f" LINES TERMINATED BY {lines}" if lines else "" + null = expression.args.get("null") + null = f" NULL DEFINED AS {null}" if null else "" + return f"ROW FORMAT DELIMITED{fields}{escaped}{items}{keys}{lines}{null}" + def table_sql(self, expression): table = ".".join( part @@ -688,6 +725,19 @@ class Generator: return f"{self.quote_start}{text}{self.quote_end}" return text + def loaddata_sql(self, expression): + local = " LOCAL" if expression.args.get("local") else "" + inpath = f" INPATH {self.sql(expression, 'inpath')}" + overwrite = " OVERWRITE" if expression.args.get("overwrite") else "" + this = f" INTO TABLE {self.sql(expression, 'this')}" + partition = self.sql(expression, "partition") + partition = f" {partition}" if partition else "" + input_format = self.sql(expression, "input_format") + input_format = f" INPUTFORMAT {input_format}" if input_format else "" + serde = self.sql(expression, "serde") + serde = f" SERDE {serde}" if serde else "" + return f"LOAD DATA{local}{inpath}{overwrite}{this}{partition}{input_format}{serde}" + def null_sql(self, *_): return "NULL" @@ -885,20 +935,24 @@ class Generator: return f"EXISTS{self.wrap(expression)}" def case_sql(self, expression): - this = self.indent(self.sql(expression, "this"), skip_first=True) - this = f" {this}" if this else "" - ifs = [] + this = self.sql(expression, "this") + statements = [f"CASE {this}" if this else "CASE"] for e in expression.args["ifs"]: - ifs.append(self.indent(f"WHEN {self.sql(e, 'this')}")) - ifs.append(self.indent(f"THEN {self.sql(e, 'true')}")) + statements.append(f"WHEN {self.sql(e, 'this')}") + statements.append(f"THEN {self.sql(e, 'true')}") + + default = self.sql(expression, "default") + + if default: + statements.append(f"ELSE {default}") - if expression.args.get("default") is not None: - ifs.append(self.indent(f"ELSE {self.sql(expression, 'default')}")) + statements.append("END") - ifs = "".join(self.seg(self.indent(e, skip_first=True)) for e in ifs) - statement = f"CASE{this}{ifs}{self.seg('END')}" - return statement + if self.pretty and self.text_width(statements) > self._max_text_width: + return self.indent("\n".join(statements), skip_first=True, skip_last=True) + + return " ".join(statements) def constraint_sql(self, expression): this = self.sql(expression, "this") @@ -970,7 +1024,7 @@ class Generator: return f"REFERENCES {this}({expressions})" def anonymous_sql(self, expression): - args = self.indent(self.expressions(expression, flat=True), skip_first=True, skip_last=True) + args = self.format_args(*expression.expressions) return f"{self.normalize_func(self.sql(expression, 'this'))}({args})" def paren_sql(self, expression): @@ -1008,7 +1062,9 @@ class Generator: if not self.pretty: return self.binary(expression, op) - return f"\n{op} ".join(self.sql(e) for e in expression.flatten(unnest=False)) + sqls = tuple(self.sql(e) for e in expression.flatten(unnest=False)) + sep = "\n" if self.text_width(sqls) > self._max_text_width else " " + return f"{sep}{op} ".join(sqls) def bitwiseand_sql(self, expression): return self.binary(expression, "&") @@ -1039,7 +1095,7 @@ class Generator: return f"{self.sql(expression, 'this').upper()} {expression.text('expression').strip()}" def distinct_sql(self, expression): - this = self.sql(expression, "this") + this = self.expressions(expression, flat=True) this = f" {this}" if this else "" on = self.sql(expression, "on") @@ -1128,13 +1184,23 @@ class Generator: def function_fallback_sql(self, expression): args = [] - for arg_key in expression.arg_types: - arg_value = ensure_list(expression.args.get(arg_key) or []) - for a in arg_value: - args.append(self.sql(a)) - - args_str = self.indent(", ".join(args), skip_first=True, skip_last=True) - return f"{self.normalize_func(expression.sql_name())}({args_str})" + for arg_value in expression.args.values(): + if isinstance(arg_value, list): + for value in arg_value: + args.append(value) + elif arg_value: + args.append(arg_value) + + return f"{self.normalize_func(expression.sql_name())}({self.format_args(*args)})" + + def format_args(self, *args): + args = tuple(self.sql(arg) for arg in args if arg is not None) + if self.pretty and self.text_width(args) > self._max_text_width: + return self.indent("\n" + f",\n".join(args) + "\n", skip_first=True, skip_last=True) + return ", ".join(args) + + def text_width(self, args): + return sum(len(arg) for arg in args) def format_time(self, expression): return format_time(self.sql(expression, "format"), self.time_mapping, self.time_trie) diff --git a/sqlglot/optimizer/eliminate_ctes.py b/sqlglot/optimizer/eliminate_ctes.py new file mode 100644 index 0000000..7b862c6 --- /dev/null +++ b/sqlglot/optimizer/eliminate_ctes.py @@ -0,0 +1,42 @@ +from sqlglot.optimizer.scope import Scope, build_scope + + +def eliminate_ctes(expression): + """ + Remove unused CTEs from an expression. + + Example: + >>> import sqlglot + >>> sql = "WITH y AS (SELECT a FROM x) SELECT a FROM z" + >>> expression = sqlglot.parse_one(sql) + >>> eliminate_ctes(expression).sql() + 'SELECT a FROM z' + + Args: + expression (sqlglot.Expression): expression to optimize + Returns: + sqlglot.Expression: optimized expression + """ + root = build_scope(expression) + + ref_count = root.ref_count() + + # Traverse the scope tree in reverse so we can remove chains of unused CTEs + for scope in reversed(list(root.traverse())): + if scope.is_cte: + count = ref_count[id(scope)] + if count <= 0: + cte_node = scope.expression.parent + with_node = cte_node.parent + cte_node.pop() + + # Pop the entire WITH clause if this is the last CTE + if len(with_node.expressions) <= 0: + with_node.pop() + + # Decrement the ref count for all sources this CTE selects from + for _, source in scope.selected_sources.values(): + if isinstance(source, Scope): + ref_count[id(source)] -= 1 + + return expression diff --git a/sqlglot/optimizer/eliminate_joins.py b/sqlglot/optimizer/eliminate_joins.py new file mode 100644 index 0000000..0854336 --- /dev/null +++ b/sqlglot/optimizer/eliminate_joins.py @@ -0,0 +1,160 @@ +from sqlglot import expressions as exp +from sqlglot.optimizer.normalize import normalized +from sqlglot.optimizer.scope import Scope, traverse_scope +from sqlglot.optimizer.simplify import simplify + + +def eliminate_joins(expression): + """ + Remove unused joins from an expression. + + This only removes joins when we know that the join condition doesn't produce duplicate rows. + + Example: + >>> import sqlglot + >>> sql = "SELECT x.a FROM x LEFT JOIN (SELECT DISTINCT y.b FROM y) AS y ON x.b = y.b" + >>> expression = sqlglot.parse_one(sql) + >>> eliminate_joins(expression).sql() + 'SELECT x.a FROM x' + + Args: + expression (sqlglot.Expression): expression to optimize + Returns: + sqlglot.Expression: optimized expression + """ + for scope in traverse_scope(expression): + # If any columns in this scope aren't qualified, it's hard to determine if a join isn't used. + # It's probably possible to infer this from the outputs of derived tables. + # But for now, let's just skip this rule. + if scope.unqualified_columns: + continue + + joins = scope.expression.args.get("joins", []) + + # Reverse the joins so we can remove chains of unused joins + for join in reversed(joins): + alias = join.this.alias_or_name + if _should_eliminate_join(scope, join, alias): + join.pop() + scope.remove_source(alias) + return expression + + +def _should_eliminate_join(scope, join, alias): + inner_source = scope.sources.get(alias) + return ( + isinstance(inner_source, Scope) + and not _join_is_used(scope, join, alias) + and ( + (join.side == "LEFT" and _is_joined_on_all_unique_outputs(inner_source, join)) + or (not join.args.get("on") and _has_single_output_row(inner_source)) + ) + ) + + +def _join_is_used(scope, join, alias): + # We need to find all columns that reference this join. + # But columns in the ON clause shouldn't count. + on = join.args.get("on") + if on: + on_clause_columns = set(id(column) for column in on.find_all(exp.Column)) + else: + on_clause_columns = set() + return any(column for column in scope.source_columns(alias) if id(column) not in on_clause_columns) + + +def _is_joined_on_all_unique_outputs(scope, join): + unique_outputs = _unique_outputs(scope) + if not unique_outputs: + return False + + _, join_keys, _ = join_condition(join) + remaining_unique_outputs = unique_outputs - set(c.name for c in join_keys) + return not remaining_unique_outputs + + +def _unique_outputs(scope): + """Determine output columns of `scope` that must have a unique combination per row""" + if scope.expression.args.get("distinct"): + return set(scope.expression.named_selects) + + group = scope.expression.args.get("group") + if group: + grouped_expressions = set(group.expressions) + grouped_outputs = set() + + unique_outputs = set() + for select in scope.selects: + output = select.unalias() + if output in grouped_expressions: + grouped_outputs.add(output) + unique_outputs.add(select.alias_or_name) + + # All the grouped expressions must be in the output + if not grouped_expressions.difference(grouped_outputs): + return unique_outputs + else: + return set() + + if _has_single_output_row(scope): + return set(scope.expression.named_selects) + + return set() + + +def _has_single_output_row(scope): + return isinstance(scope.expression, exp.Select) and ( + all(isinstance(e.unalias(), exp.AggFunc) for e in scope.selects) + or _is_limit_1(scope) + or not scope.expression.args.get("from") + ) + + +def _is_limit_1(scope): + limit = scope.expression.args.get("limit") + return limit and limit.expression.this == "1" + + +def join_condition(join): + """ + Extract the join condition from a join expression. + + Args: + join (exp.Join) + Returns: + tuple[list[str], list[str], exp.Expression]: + Tuple of (source key, join key, remaining predicate) + """ + name = join.this.alias_or_name + on = join.args.get("on") or exp.TRUE + on = on.copy() + source_key = [] + join_key = [] + + # find the join keys + # SELECT + # FROM x + # JOIN y + # ON x.a = y.b AND y.b > 1 + # + # should pull y.b as the join key and x.a as the source key + if normalized(on): + for condition in on.flatten() if isinstance(on, exp.And) else [on]: + if isinstance(condition, exp.EQ): + left, right = condition.unnest_operands() + left_tables = exp.column_table_names(left) + right_tables = exp.column_table_names(right) + + if name in left_tables and name not in right_tables: + join_key.append(left) + source_key.append(right) + condition.replace(exp.TRUE) + elif name in right_tables and name not in left_tables: + join_key.append(right) + source_key.append(left) + condition.replace(exp.TRUE) + + on = simplify(on) + remaining_condition = None if on == exp.TRUE else on + + return source_key, join_key, remaining_condition diff --git a/sqlglot/optimizer/eliminate_subqueries.py b/sqlglot/optimizer/eliminate_subqueries.py index 38e1299..44cdc94 100644 --- a/sqlglot/optimizer/eliminate_subqueries.py +++ b/sqlglot/optimizer/eliminate_subqueries.py @@ -8,7 +8,7 @@ from sqlglot.optimizer.simplify import simplify def eliminate_subqueries(expression): """ - Rewrite subqueries as CTES, deduplicating if possible. + Rewrite derived tables as CTES, deduplicating if possible. Example: >>> import sqlglot diff --git a/sqlglot/optimizer/merge_subqueries.py b/sqlglot/optimizer/merge_subqueries.py index 3e435f5..3c51c18 100644 --- a/sqlglot/optimizer/merge_subqueries.py +++ b/sqlglot/optimizer/merge_subqueries.py @@ -119,6 +119,23 @@ def _mergeable(outer_scope, inner_select, leave_tables_isolated, from_or_join): Returns: bool: True if can be merged """ + + def _is_a_window_expression_in_unmergable_operation(): + window_expressions = inner_select.find_all(exp.Window) + window_alias_names = {window.parent.alias_or_name for window in window_expressions} + inner_select_name = inner_select.parent.alias_or_name + unmergable_window_columns = [ + column + for column in outer_scope.columns + if column.find_ancestor(exp.Where, exp.Group, exp.Order, exp.Join, exp.Having, exp.AggFunc) + ] + window_expressions_in_unmergable = [ + column + for column in unmergable_window_columns + if column.table == inner_select_name and column.name in window_alias_names + ] + return any(window_expressions_in_unmergable) + return ( isinstance(outer_scope.expression, exp.Select) and isinstance(inner_select, exp.Select) @@ -137,6 +154,7 @@ def _mergeable(outer_scope, inner_select, leave_tables_isolated, from_or_join): and inner_select.args.get("where") and any(j.side in {"FULL", "RIGHT"} for j in outer_scope.expression.args.get("joins", [])) ) + and not _is_a_window_expression_in_unmergable_operation() ) diff --git a/sqlglot/optimizer/optimizer.py b/sqlglot/optimizer/optimizer.py index 9a09327..2c28ab8 100644 --- a/sqlglot/optimizer/optimizer.py +++ b/sqlglot/optimizer/optimizer.py @@ -1,3 +1,5 @@ +from sqlglot.optimizer.eliminate_ctes import eliminate_ctes +from sqlglot.optimizer.eliminate_joins import eliminate_joins from sqlglot.optimizer.eliminate_subqueries import eliminate_subqueries from sqlglot.optimizer.expand_multi_table_selects import expand_multi_table_selects from sqlglot.optimizer.isolate_table_selects import isolate_table_selects @@ -23,6 +25,8 @@ RULES = ( optimize_joins, eliminate_subqueries, merge_subqueries, + eliminate_joins, + eliminate_ctes, quote_identities, ) diff --git a/sqlglot/optimizer/pushdown_predicates.py b/sqlglot/optimizer/pushdown_predicates.py index 9c8d71d..583d059 100644 --- a/sqlglot/optimizer/pushdown_predicates.py +++ b/sqlglot/optimizer/pushdown_predicates.py @@ -1,8 +1,6 @@ -from collections import defaultdict - from sqlglot import exp from sqlglot.optimizer.normalize import normalized -from sqlglot.optimizer.scope import traverse_scope +from sqlglot.optimizer.scope import build_scope from sqlglot.optimizer.simplify import simplify @@ -22,15 +20,10 @@ def pushdown_predicates(expression): Returns: sqlglot.Expression: optimized expression """ - scope_ref_count = defaultdict(lambda: 0) - scopes = traverse_scope(expression) - scopes.reverse() - - for scope in scopes: - for _, source in scope.selected_sources.values(): - scope_ref_count[id(source)] += 1 + root = build_scope(expression) + scope_ref_count = root.ref_count() - for scope in scopes: + for scope in reversed(list(root.traverse())): select = scope.expression where = select.args.get("where") if where: @@ -152,9 +145,11 @@ def nodes_for_predicate(predicate, sources, scope_ref_count): return {} nodes[table] = node elif isinstance(node, exp.Select) and len(tables) == 1: + # We can't push down window expressions + has_window_expression = any(select for select in node.selects if select.find(exp.Window)) # we can't push down predicates to select statements if they are referenced in # multiple places. - if not node.args.get("group") and scope_ref_count[id(source)] < 2: + if not node.args.get("group") and scope_ref_count[id(source)] < 2 and not has_window_expression: nodes[table] = node return nodes diff --git a/sqlglot/optimizer/scope.py b/sqlglot/optimizer/scope.py index 89de517..68298a0 100644 --- a/sqlglot/optimizer/scope.py +++ b/sqlglot/optimizer/scope.py @@ -1,4 +1,5 @@ import itertools +from collections import defaultdict from enum import Enum, auto from sqlglot import exp @@ -315,6 +316,16 @@ class Scope: return self._external_columns @property + def unqualified_columns(self): + """ + Unqualified columns in the current scope. + + Returns: + list[exp.Column]: Unqualified columns + """ + return [c for c in self.columns if not c.table] + + @property def join_hints(self): """ Hints that exist in the scope that reference tables @@ -403,6 +414,21 @@ class Scope: yield from child_scope.traverse() yield self + def ref_count(self): + """ + Count the number of times each scope in this tree is referenced. + + Returns: + dict[int, int]: Mapping of Scope instance ID to reference count + """ + scope_ref_count = defaultdict(lambda: 0) + + for scope in self.traverse(): + for _, source in scope.selected_sources.values(): + scope_ref_count[id(source)] += 1 + + return scope_ref_count + def traverse_scope(expression): """ diff --git a/sqlglot/parser.py b/sqlglot/parser.py index c29e520..b378f12 100644 --- a/sqlglot/parser.py +++ b/sqlglot/parser.py @@ -135,11 +135,13 @@ class Parser: TokenType.BOTH, TokenType.BUCKET, TokenType.CACHE, + TokenType.CALL, TokenType.COLLATE, TokenType.COMMIT, TokenType.CONSTRAINT, TokenType.DEFAULT, TokenType.DELETE, + TokenType.DESCRIBE, TokenType.DETERMINISTIC, TokenType.EXECUTE, TokenType.ENGINE, @@ -160,6 +162,7 @@ class Parser: TokenType.LAZY, TokenType.LANGUAGE, TokenType.LEADING, + TokenType.LOCAL, TokenType.LOCATION, TokenType.MATERIALIZED, TokenType.NATURAL, @@ -176,6 +179,7 @@ class Parser: TokenType.REFERENCES, TokenType.RETURNS, TokenType.ROWS, + TokenType.SCHEMA, TokenType.SCHEMA_COMMENT, TokenType.SEED, TokenType.SEMI, @@ -294,6 +298,11 @@ class Parser: COLUMN_OPERATORS = { TokenType.DOT: None, + TokenType.DCOLON: lambda self, this, to: self.expression( + exp.Cast, + this=this, + to=to, + ), TokenType.ARROW: lambda self, this, path: self.expression( exp.JSONExtract, this=this, @@ -342,8 +351,10 @@ class Parser: STATEMENT_PARSERS = { TokenType.CREATE: lambda self: self._parse_create(), + TokenType.DESCRIBE: lambda self: self._parse_describe(), TokenType.DROP: lambda self: self._parse_drop(), TokenType.INSERT: lambda self: self._parse_insert(), + TokenType.LOAD_DATA: lambda self: self._parse_load_data(), TokenType.UPDATE: lambda self: self._parse_update(), TokenType.DELETE: lambda self: self._parse_delete(), TokenType.CACHE: lambda self: self._parse_cache(), @@ -449,7 +460,14 @@ class Parser: MODIFIABLES = (exp.Subquery, exp.Subqueryable, exp.Table) - CREATABLES = {TokenType.TABLE, TokenType.VIEW, TokenType.FUNCTION, TokenType.INDEX, TokenType.PROCEDURE} + CREATABLES = { + TokenType.TABLE, + TokenType.VIEW, + TokenType.FUNCTION, + TokenType.INDEX, + TokenType.PROCEDURE, + TokenType.SCHEMA, + } STRICT_CAST = True @@ -650,7 +668,7 @@ class Parser: materialized = self._match(TokenType.MATERIALIZED) kind = self._match_set(self.CREATABLES) and self._prev.text if not kind: - self.raise_error("Expected TABLE, VIEW, INDEX, FUNCTION, or PROCEDURE") + self.raise_error(f"Expected {self.CREATABLES}") return return self.expression( @@ -677,7 +695,7 @@ class Parser: create_token = self._match_set(self.CREATABLES) and self._prev if not create_token: - self.raise_error("Expected TABLE, VIEW, INDEX, FUNCTION, or PROCEDURE") + self.raise_error(f"Expected {self.CREATABLES}") return exists = self._parse_exists(not_=True) @@ -692,7 +710,7 @@ class Parser: expression = self._parse_select_or_expression() elif create_token.token_type == TokenType.INDEX: this = self._parse_index() - elif create_token.token_type in (TokenType.TABLE, TokenType.VIEW): + elif create_token.token_type in (TokenType.TABLE, TokenType.VIEW, TokenType.SCHEMA): this = self._parse_table(schema=True) properties = self._parse_properties() if self._match(TokenType.ALIAS): @@ -836,19 +854,74 @@ class Parser: return self.expression(exp.Properties, expressions=properties) return None + def _parse_describe(self): + self._match(TokenType.TABLE) + + return self.expression(exp.Describe, this=self._parse_id_var()) + def _parse_insert(self): overwrite = self._match(TokenType.OVERWRITE) - self._match(TokenType.INTO) - self._match(TokenType.TABLE) + local = self._match(TokenType.LOCAL) + if self._match_text("DIRECTORY"): + this = self.expression( + exp.Directory, + this=self._parse_var_or_string(), + local=local, + row_format=self._parse_row_format(), + ) + else: + self._match(TokenType.INTO) + self._match(TokenType.TABLE) + this = self._parse_table(schema=True) return self.expression( exp.Insert, - this=self._parse_table(schema=True), + this=this, exists=self._parse_exists(), partition=self._parse_partition(), expression=self._parse_select(nested=True), overwrite=overwrite, ) + def _parse_row_format(self): + if not self._match_pair(TokenType.ROW, TokenType.FORMAT): + return None + + self._match_text("DELIMITED") + + kwargs = {} + + if self._match_text("FIELDS", "TERMINATED", "BY"): + kwargs["fields"] = self._parse_string() + if self._match_text("ESCAPED", "BY"): + kwargs["escaped"] = self._parse_string() + if self._match_text("COLLECTION", "ITEMS", "TERMINATED", "BY"): + kwargs["collection_items"] = self._parse_string() + if self._match_text("MAP", "KEYS", "TERMINATED", "BY"): + kwargs["map_keys"] = self._parse_string() + if self._match_text("LINES", "TERMINATED", "BY"): + kwargs["lines"] = self._parse_string() + if self._match_text("NULL", "DEFINED", "AS"): + kwargs["null"] = self._parse_string() + return self.expression(exp.RowFormat, **kwargs) + + def _parse_load_data(self): + local = self._match(TokenType.LOCAL) + self._match_text("INPATH") + inpath = self._parse_string() + overwrite = self._match(TokenType.OVERWRITE) + self._match_pair(TokenType.INTO, TokenType.TABLE) + + return self.expression( + exp.LoadData, + this=self._parse_table(schema=True), + local=local, + overwrite=overwrite, + inpath=inpath, + partition=self._parse_partition(), + input_format=self._match_text("INPUTFORMAT") and self._parse_string(), + serde=self._match_text("SERDE") and self._parse_string(), + ) + def _parse_delete(self): self._match(TokenType.FROM) @@ -1484,6 +1557,14 @@ class Parser: if self._match_set(self.RANGE_PARSERS): this = self.RANGE_PARSERS[self._prev.token_type](self, this) + elif self._match(TokenType.ISNULL): + this = self.expression(exp.Is, this=this, expression=exp.Null()) + + # Postgres supports ISNULL and NOTNULL for conditions. + # https://blog.andreiavram.ro/postgresql-null-composite-type/ + if self._match(TokenType.NOTNULL): + this = self.expression(exp.Is, this=this, expression=exp.Null()) + this = self.expression(exp.Not, this=this) if negate: this = self.expression(exp.Not, this=this) @@ -1582,12 +1663,6 @@ class Parser: return self._parse_column() return type_token - while self._match(TokenType.DCOLON): - type_token = self._parse_types() - if not type_token: - self.raise_error("Expected type") - this = self.expression(exp.Cast, this=this, to=type_token) - return this def _parse_types(self): @@ -1601,6 +1676,11 @@ class Parser: is_struct = type_token == TokenType.STRUCT expressions = None + if not nested and self._match_pair(TokenType.L_BRACKET, TokenType.R_BRACKET): + return exp.DataType( + this=exp.DataType.Type.ARRAY, expressions=[exp.DataType.build(type_token.value)], nested=True + ) + if self._match(TokenType.L_BRACKET): self._retreat(index) return None @@ -1611,7 +1691,7 @@ class Parser: elif nested: expressions = self._parse_csv(self._parse_types) else: - expressions = self._parse_csv(self._parse_type) + expressions = self._parse_csv(self._parse_conjunction) if not expressions: self._retreat(index) @@ -1677,8 +1757,17 @@ class Parser: this = self._parse_bracket(this) while self._match_set(self.COLUMN_OPERATORS): - op = self.COLUMN_OPERATORS.get(self._prev.token_type) - field = self._parse_star() or self._parse_function() or self._parse_id_var() + op_token = self._prev.token_type + op = self.COLUMN_OPERATORS.get(op_token) + + if op_token == TokenType.DCOLON: + field = self._parse_types() + if not field: + self.raise_error("Expected type") + elif op: + field = exp.Literal.string(self._advance() or self._prev.text) + else: + field = self._parse_star() or self._parse_function() or self._parse_id_var() if isinstance(field, exp.Func): # bigquery allows function calls like x.y.count(...) @@ -1687,7 +1776,7 @@ class Parser: this = self._replace_columns_with_dots(this) if op: - this = op(self, this, exp.Literal.string(field.name)) + this = op(self, this, field) elif isinstance(this, exp.Column) and not this.table: this = self.expression(exp.Column, this=field, table=this.this) else: @@ -1808,11 +1897,10 @@ class Parser: if not self._match(TokenType.ARROW): self._retreat(index) - distinct = self._match(TokenType.DISTINCT) - this = self._parse_conjunction() - - if distinct: - this = self.expression(exp.Distinct, this=this) + if self._match(TokenType.DISTINCT): + this = self.expression(exp.Distinct, expressions=self._parse_csv(self._parse_conjunction)) + else: + this = self._parse_conjunction() if self._match(TokenType.IGNORE_NULLS): this = self.expression(exp.IgnoreNulls, this=this) @@ -2112,6 +2200,8 @@ class Parser: this = self.expression(exp.Filter, this=this, expression=self._parse_where()) self._match_r_paren() + # T-SQL allows the OVER (...) syntax after WITHIN GROUP. + # https://learn.microsoft.com/en-us/sql/t-sql/functions/percentile-disc-transact-sql?view=sql-server-ver16 if self._match(TokenType.WITHIN_GROUP): self._match_l_paren() this = self.expression( @@ -2120,7 +2210,6 @@ class Parser: expression=self._parse_order(), ) self._match_r_paren() - return this # SQL spec defines an optional [ { IGNORE | RESPECT } NULLS ] OVER # Some dialects choose to implement and some do not. @@ -2366,6 +2455,16 @@ class Parser: if not self._match(TokenType.R_PAREN): self.raise_error("Expecting )") + def _match_text(self, *texts): + index = self._index + for text in texts: + if self._curr and self._curr.text.upper() == text: + self._advance() + else: + self._retreat(index) + return False + return True + def _replace_columns_with_dots(self, this): if isinstance(this, exp.Dot): exp.replace_children(this, self._replace_columns_with_dots) diff --git a/sqlglot/planner.py b/sqlglot/planner.py index ed0b66c..efabc15 100644 --- a/sqlglot/planner.py +++ b/sqlglot/planner.py @@ -3,7 +3,7 @@ import math from sqlglot import alias, exp from sqlglot.errors import UnsupportedError -from sqlglot.optimizer.simplify import simplify +from sqlglot.optimizer.eliminate_joins import join_condition class Plan: @@ -236,40 +236,12 @@ class Join(Step): step = Join() for join in joins: - name = join.this.alias - on = join.args.get("on") or exp.TRUE - source_key = [] - join_key = [] - - # find the join keys - # SELECT - # FROM x - # JOIN y - # ON x.a = y.b AND y.b > 1 - # - # should pull y.b as the join key and x.a as the source key - for condition in on.flatten() if isinstance(on, exp.And) else [on]: - if isinstance(condition, exp.EQ): - left, right = condition.unnest_operands() - left_tables = exp.column_table_names(left) - right_tables = exp.column_table_names(right) - - if name in left_tables and name not in right_tables: - join_key.append(left) - source_key.append(right) - condition.replace(exp.TRUE) - elif name in right_tables and name not in left_tables: - join_key.append(right) - source_key.append(left) - condition.replace(exp.TRUE) - - on = simplify(on) - - step.joins[name] = { + source_key, join_key, condition = join_condition(join) + step.joins[join.this.alias_or_name] = { "side": join.side, "join_key": join_key, "source_key": source_key, - "condition": None if on == exp.TRUE else on, + "condition": condition, } step.add_dependency(Scan.from_expression(join.this, ctes)) diff --git a/sqlglot/tokens.py b/sqlglot/tokens.py index 17c038c..fc8e6e7 100644 --- a/sqlglot/tokens.py +++ b/sqlglot/tokens.py @@ -123,6 +123,7 @@ class TokenType(AutoName): CLUSTER_BY = auto() COLLATE = auto() COMMENT = auto() + COMMENT_ON = auto() COMMIT = auto() CONSTRAINT = auto() CREATE = auto() @@ -133,13 +134,14 @@ class TokenType(AutoName): CURRENT_ROW = auto() CURRENT_TIME = auto() CURRENT_TIMESTAMP = auto() - DIV = auto() DEFAULT = auto() DELETE = auto() DESC = auto() + DESCRIBE = auto() DETERMINISTIC = auto() DISTINCT = auto() DISTRIBUTE_BY = auto() + DIV = auto() DROP = auto() ELSE = auto() END = auto() @@ -189,6 +191,8 @@ class TokenType(AutoName): LEFT = auto() LIKE = auto() LIMIT = auto() + LOAD_DATA = auto() + LOCAL = auto() LOCATION = auto() MAP = auto() MATERIALIZED = auto() @@ -196,6 +200,7 @@ class TokenType(AutoName): NATURAL = auto() NEXT = auto() NO_ACTION = auto() + NOTNULL = auto() NULL = auto() NULLS_FIRST = auto() NULLS_LAST = auto() @@ -436,13 +441,14 @@ class Tokenizer(metaclass=_Tokenizer): "CURRENT_DATE": TokenType.CURRENT_DATE, "CURRENT ROW": TokenType.CURRENT_ROW, "CURRENT_TIMESTAMP": TokenType.CURRENT_TIMESTAMP, - "DIV": TokenType.DIV, "DEFAULT": TokenType.DEFAULT, "DELETE": TokenType.DELETE, "DESC": TokenType.DESC, + "DESCRIBE": TokenType.DESCRIBE, "DETERMINISTIC": TokenType.DETERMINISTIC, "DISTINCT": TokenType.DISTINCT, "DISTRIBUTE BY": TokenType.DISTRIBUTE_BY, + "DIV": TokenType.DIV, "DROP": TokenType.DROP, "ELSE": TokenType.ELSE, "END": TokenType.END, @@ -487,12 +493,15 @@ class Tokenizer(metaclass=_Tokenizer): "LEFT": TokenType.LEFT, "LIKE": TokenType.LIKE, "LIMIT": TokenType.LIMIT, + "LOAD DATA": TokenType.LOAD_DATA, + "LOCAL": TokenType.LOCAL, "LOCATION": TokenType.LOCATION, "MATERIALIZED": TokenType.MATERIALIZED, "NATURAL": TokenType.NATURAL, "NEXT": TokenType.NEXT, "NO ACTION": TokenType.NO_ACTION, "NOT": TokenType.NOT, + "NOTNULL": TokenType.NOTNULL, "NULL": TokenType.NULL, "NULLS FIRST": TokenType.NULLS_FIRST, "NULLS LAST": TokenType.NULLS_LAST, @@ -530,6 +539,7 @@ class Tokenizer(metaclass=_Tokenizer): "ROLLUP": TokenType.ROLLUP, "ROW": TokenType.ROW, "ROWS": TokenType.ROWS, + "SCHEMA": TokenType.SCHEMA, "SEED": TokenType.SEED, "SELECT": TokenType.SELECT, "SEMI": TokenType.SEMI, @@ -629,6 +639,7 @@ class Tokenizer(metaclass=_Tokenizer): TokenType.ANALYZE, TokenType.BEGIN, TokenType.CALL, + TokenType.COMMENT_ON, TokenType.COMMIT, TokenType.EXPLAIN, TokenType.OPTIMIZE, diff --git a/tests/dialects/test_bigquery.py b/tests/dialects/test_bigquery.py index 8921924..14fea9d 100644 --- a/tests/dialects/test_bigquery.py +++ b/tests/dialects/test_bigquery.py @@ -164,7 +164,7 @@ class TestBigQuery(Validator): "CREATE TABLE db.example_table (col_a struct<struct_col_a:int, struct_col_b:string>)", write={ "bigquery": "CREATE TABLE db.example_table (col_a STRUCT<struct_col_a INT64, struct_col_b STRING>)", - "duckdb": "CREATE TABLE db.example_table (col_a STRUCT<struct_col_a INT, struct_col_b TEXT>)", + "duckdb": "CREATE TABLE db.example_table (col_a STRUCT(struct_col_a INT, struct_col_b TEXT))", "presto": "CREATE TABLE db.example_table (col_a ROW(struct_col_a INTEGER, struct_col_b VARCHAR))", "hive": "CREATE TABLE db.example_table (col_a STRUCT<struct_col_a INT, struct_col_b STRING>)", "spark": "CREATE TABLE db.example_table (col_a STRUCT<struct_col_a: INT, struct_col_b: STRING>)", @@ -174,6 +174,7 @@ class TestBigQuery(Validator): "CREATE TABLE db.example_table (col_a STRUCT<struct_col_a INT64, struct_col_b STRUCT<nested_col_a STRING, nested_col_b STRING>>)", write={ "bigquery": "CREATE TABLE db.example_table (col_a STRUCT<struct_col_a INT64, struct_col_b STRUCT<nested_col_a STRING, nested_col_b STRING>>)", + "duckdb": "CREATE TABLE db.example_table (col_a STRUCT(struct_col_a BIGINT, struct_col_b STRUCT(nested_col_a TEXT, nested_col_b TEXT)))", "presto": "CREATE TABLE db.example_table (col_a ROW(struct_col_a BIGINT, struct_col_b ROW(nested_col_a VARCHAR, nested_col_b VARCHAR)))", "hive": "CREATE TABLE db.example_table (col_a STRUCT<struct_col_a BIGINT, struct_col_b STRUCT<nested_col_a STRING, nested_col_b STRING>>)", "spark": "CREATE TABLE db.example_table (col_a STRUCT<struct_col_a: LONG, struct_col_b: STRUCT<nested_col_a: STRING, nested_col_b: STRING>>)", diff --git a/tests/dialects/test_dialect.py b/tests/dialects/test_dialect.py index 53edb42..a1e1262 100644 --- a/tests/dialects/test_dialect.py +++ b/tests/dialects/test_dialect.py @@ -1056,6 +1056,7 @@ class TestDialect(Validator): self.validate_all( "CREATE TABLE t (c CHAR, nc NCHAR, v1 VARCHAR, v2 VARCHAR2, nv NVARCHAR, nv2 NVARCHAR2)", write={ + "duckdb": "CREATE TABLE t (c CHAR, nc CHAR, v1 TEXT, v2 TEXT, nv TEXT, nv2 TEXT)", "hive": "CREATE TABLE t (c CHAR, nc CHAR, v1 STRING, v2 STRING, nv STRING, nv2 STRING)", "oracle": "CREATE TABLE t (c CHAR, nc CHAR, v1 VARCHAR2, v2 VARCHAR2, nv NVARCHAR2, nv2 NVARCHAR2)", "postgres": "CREATE TABLE t (c CHAR, nc CHAR, v1 VARCHAR, v2 VARCHAR, nv VARCHAR, nv2 VARCHAR)", @@ -1096,6 +1097,7 @@ class TestDialect(Validator): self.validate_all( "CREATE TABLE t (b1 BINARY, b2 BINARY(1024), c1 TEXT, c2 TEXT(1024))", write={ + "duckdb": "CREATE TABLE t (b1 BINARY, b2 BINARY(1024), c1 TEXT, c2 TEXT(1024))", "hive": "CREATE TABLE t (b1 BINARY, b2 BINARY(1024), c1 STRING, c2 STRING(1024))", "oracle": "CREATE TABLE t (b1 BLOB, b2 BLOB(1024), c1 CLOB, c2 CLOB(1024))", "postgres": "CREATE TABLE t (b1 BYTEA, b2 BYTEA(1024), c1 TEXT, c2 TEXT(1024))", diff --git a/tests/dialects/test_duckdb.py b/tests/dialects/test_duckdb.py index 96e51df..298b3e9 100644 --- a/tests/dialects/test_duckdb.py +++ b/tests/dialects/test_duckdb.py @@ -66,6 +66,23 @@ class TestDuckDB(Validator): def test_duckdb(self): self.validate_all( + "CREATE TABLE IF NOT EXISTS table (cola INT, colb STRING) USING ICEBERG PARTITIONED BY (colb)", + write={ + "duckdb": "CREATE TABLE IF NOT EXISTS table (cola INT, colb TEXT)", + }, + ) + + self.validate_all( + "COL::BIGINT[]", + write={ + "duckdb": "CAST(COL AS BIGINT[])", + "presto": "CAST(COL AS ARRAY(BIGINT))", + "hive": "CAST(COL AS ARRAY<BIGINT>)", + "spark": "CAST(COL AS ARRAY<LONG>)", + }, + ) + + self.validate_all( "LIST_VALUE(0, 1, 2)", read={ "spark": "ARRAY(0, 1, 2)", diff --git a/tests/dialects/test_hive.py b/tests/dialects/test_hive.py index acb3be9..69c7630 100644 --- a/tests/dialects/test_hive.py +++ b/tests/dialects/test_hive.py @@ -127,21 +127,23 @@ class TestHive(Validator): def test_ddl(self): self.validate_all( - "CREATE TABLE test STORED AS parquet TBLPROPERTIES ('x'='1', 'Z'='2') AS SELECT 1", - write={ - "presto": "CREATE TABLE test WITH (FORMAT='parquet', x='1', Z='2') AS SELECT 1", - "hive": "CREATE TABLE test STORED AS PARQUET TBLPROPERTIES ('x'='1', 'Z'='2') AS SELECT 1", - "spark": "CREATE TABLE test USING PARQUET TBLPROPERTIES ('x'='1', 'Z'='2') AS SELECT 1", - }, - ) - self.validate_all( "CREATE TABLE x (w STRING) PARTITIONED BY (y INT, z INT)", write={ + "duckdb": "CREATE TABLE x (w TEXT)", # Partition columns should exist in table "presto": "CREATE TABLE x (w VARCHAR, y INTEGER, z INTEGER) WITH (PARTITIONED_BY=ARRAY['y', 'z'])", "hive": "CREATE TABLE x (w STRING) PARTITIONED BY (y INT, z INT)", "spark": "CREATE TABLE x (w STRING) PARTITIONED BY (y INT, z INT)", }, ) + self.validate_all( + "CREATE TABLE test STORED AS parquet TBLPROPERTIES ('x'='1', 'Z'='2') AS SELECT 1", + write={ + "duckdb": "CREATE TABLE test AS SELECT 1", + "presto": "CREATE TABLE test WITH (FORMAT='parquet', x='1', Z='2') AS SELECT 1", + "hive": "CREATE TABLE test STORED AS PARQUET TBLPROPERTIES ('x'='1', 'Z'='2') AS SELECT 1", + "spark": "CREATE TABLE test USING PARQUET TBLPROPERTIES ('x'='1', 'Z'='2') AS SELECT 1", + }, + ) def test_lateral_view(self): self.validate_all( diff --git a/tests/dialects/test_mysql.py b/tests/dialects/test_mysql.py index 02dc1ad..723e27c 100644 --- a/tests/dialects/test_mysql.py +++ b/tests/dialects/test_mysql.py @@ -8,6 +8,7 @@ class TestMySQL(Validator): self.validate_all( "CREATE TABLE z (a INT) ENGINE=InnoDB AUTO_INCREMENT=1 DEFAULT CHARACTER SET=utf8 COLLATE=utf8_bin COMMENT='x'", write={ + "duckdb": "CREATE TABLE z (a INT)", "mysql": "CREATE TABLE z (a INT) ENGINE=InnoDB AUTO_INCREMENT=1 DEFAULT CHARACTER SET=utf8 COLLATE=utf8_bin COMMENT='x'", "spark": "CREATE TABLE z (a INT) COMMENT 'x'", }, diff --git a/tests/dialects/test_postgres.py b/tests/dialects/test_postgres.py index dc93c3a..4b8f3c3 100644 --- a/tests/dialects/test_postgres.py +++ b/tests/dialects/test_postgres.py @@ -68,10 +68,12 @@ class TestPostgres(Validator): self.validate_identity("SELECT TRIM(' X' FROM ' XXX ')") self.validate_identity("SELECT TRIM(LEADING 'bla' FROM ' XXX ' COLLATE utf8_bin)") self.validate_identity("SELECT TO_TIMESTAMP(1284352323.5), TO_TIMESTAMP('05 Dec 2000', 'DD Mon YYYY')") + self.validate_identity("COMMENT ON TABLE mytable IS 'this'") self.validate_all( "CREATE TABLE x (a UUID, b BYTEA)", write={ + "duckdb": "CREATE TABLE x (a UUID, b BINARY)", "presto": "CREATE TABLE x (a UUID, b VARBINARY)", "hive": "CREATE TABLE x (a UUID, b BINARY)", "spark": "CREATE TABLE x (a UUID, b BINARY)", @@ -163,3 +165,42 @@ class TestPostgres(Validator): "postgres": "SELECT p1.id, p2.id, v1, v2 FROM polygons p1, polygons p2, LATERAL VERTICES(p1.poly) v1, LATERAL VERTICES(p2.poly) v2 WHERE (v1 <-> v2) < 10 AND p1.id != p2.id", }, ) + self.validate_all( + "SELECT id, email, CAST(deleted AS TEXT) FROM users WHERE NOT deleted IS NULL", + read={"postgres": "SELECT id, email, CAST(deleted AS TEXT) FROM users WHERE deleted NOTNULL"}, + ) + self.validate_all( + "SELECT id, email, CAST(deleted AS TEXT) FROM users WHERE NOT deleted IS NULL", + read={"postgres": "SELECT id, email, CAST(deleted AS TEXT) FROM users WHERE NOT deleted ISNULL"}, + ) + self.validate_all( + "'[1,2,3]'::json->2", + write={"postgres": "CAST('[1,2,3]' AS JSON)->'2'"}, + ) + self.validate_all( + """'{"a":1,"b":2}'::json->'b'""", + write={"postgres": """CAST('{"a":1,"b":2}' AS JSON)->'b'"""}, + ) + self.validate_all( + """'{"x": {"y": 1}}'::json->'x'->'y'""", write={"postgres": """CAST('{"x": {"y": 1}}' AS JSON)->'x'->'y'"""} + ) + self.validate_all( + """'{"x": {"y": 1}}'::json->'x'::json->'y'""", + write={"postgres": """CAST(CAST('{"x": {"y": 1}}' AS JSON)->'x' AS JSON)->'y'"""}, + ) + self.validate_all( + """'[1,2,3]'::json->>2""", + write={"postgres": "CAST('[1,2,3]' AS JSON)->>'2'"}, + ) + self.validate_all( + """'{"a":1,"b":2}'::json->>'b'""", + write={"postgres": """CAST('{"a":1,"b":2}' AS JSON)->>'b'"""}, + ) + self.validate_all( + """'{"a":[1,2,3],"b":[4,5,6]}'::json#>'{a,2}'""", + write={"postgres": """CAST('{"a":[1,2,3],"b":[4,5,6]}' AS JSON)#>'{a,2}'"""}, + ) + self.validate_all( + """'{"a":[1,2,3],"b":[4,5,6]}'::json#>>'{a,2}'""", + write={"postgres": """CAST('{"a":[1,2,3],"b":[4,5,6]}' AS JSON)#>>'{a,2}'"""}, + ) diff --git a/tests/dialects/test_presto.py b/tests/dialects/test_presto.py index b0d9ad9..10c9d35 100644 --- a/tests/dialects/test_presto.py +++ b/tests/dialects/test_presto.py @@ -10,7 +10,7 @@ class TestPresto(Validator): "CAST(a AS ARRAY(INT))", write={ "bigquery": "CAST(a AS ARRAY<INT64>)", - "duckdb": "CAST(a AS ARRAY<INT>)", + "duckdb": "CAST(a AS INT[])", "presto": "CAST(a AS ARRAY(INTEGER))", "spark": "CAST(a AS ARRAY<INT>)", }, @@ -28,7 +28,7 @@ class TestPresto(Validator): "CAST(ARRAY[1, 2] AS ARRAY(BIGINT))", write={ "bigquery": "CAST([1, 2] AS ARRAY<INT64>)", - "duckdb": "CAST(LIST_VALUE(1, 2) AS ARRAY<BIGINT>)", + "duckdb": "CAST(LIST_VALUE(1, 2) AS BIGINT[])", "presto": "CAST(ARRAY[1, 2] AS ARRAY(BIGINT))", "spark": "CAST(ARRAY(1, 2) AS ARRAY<LONG>)", }, @@ -37,7 +37,7 @@ class TestPresto(Validator): "CAST(MAP(ARRAY[1], ARRAY[1]) AS MAP(INT,INT))", write={ "bigquery": "CAST(MAP([1], [1]) AS MAP<INT64, INT64>)", - "duckdb": "CAST(MAP(LIST_VALUE(1), LIST_VALUE(1)) AS MAP<INT, INT>)", + "duckdb": "CAST(MAP(LIST_VALUE(1), LIST_VALUE(1)) AS MAP(INT, INT))", "presto": "CAST(MAP(ARRAY[1], ARRAY[1]) AS MAP(INTEGER, INTEGER))", "hive": "CAST(MAP(1, 1) AS MAP<INT, INT>)", "spark": "CAST(MAP_FROM_ARRAYS(ARRAY(1), ARRAY(1)) AS MAP<INT, INT>)", @@ -47,7 +47,7 @@ class TestPresto(Validator): "CAST(MAP(ARRAY['a','b','c'], ARRAY[ARRAY[1], ARRAY[2], ARRAY[3]]) AS MAP(VARCHAR, ARRAY(INT)))", write={ "bigquery": "CAST(MAP(['a', 'b', 'c'], [[1], [2], [3]]) AS MAP<STRING, ARRAY<INT64>>)", - "duckdb": "CAST(MAP(LIST_VALUE('a', 'b', 'c'), LIST_VALUE(LIST_VALUE(1), LIST_VALUE(2), LIST_VALUE(3))) AS MAP<TEXT, ARRAY<INT>>)", + "duckdb": "CAST(MAP(LIST_VALUE('a', 'b', 'c'), LIST_VALUE(LIST_VALUE(1), LIST_VALUE(2), LIST_VALUE(3))) AS MAP(TEXT, INT[]))", "presto": "CAST(MAP(ARRAY['a', 'b', 'c'], ARRAY[ARRAY[1], ARRAY[2], ARRAY[3]]) AS MAP(VARCHAR, ARRAY(INTEGER)))", "hive": "CAST(MAP('a', ARRAY(1), 'b', ARRAY(2), 'c', ARRAY(3)) AS MAP<STRING, ARRAY<INT>>)", "spark": "CAST(MAP_FROM_ARRAYS(ARRAY('a', 'b', 'c'), ARRAY(ARRAY(1), ARRAY(2), ARRAY(3))) AS MAP<STRING, ARRAY<INT>>)", @@ -171,6 +171,7 @@ class TestPresto(Validator): self.validate_all( "CREATE TABLE test WITH (FORMAT = 'PARQUET') AS SELECT 1", write={ + "duckdb": "CREATE TABLE test AS SELECT 1", "presto": "CREATE TABLE test WITH (FORMAT='PARQUET') AS SELECT 1", "hive": "CREATE TABLE test STORED AS PARQUET AS SELECT 1", "spark": "CREATE TABLE test USING PARQUET AS SELECT 1", @@ -179,6 +180,7 @@ class TestPresto(Validator): self.validate_all( "CREATE TABLE test WITH (FORMAT = 'PARQUET', X = '1', Z = '2') AS SELECT 1", write={ + "duckdb": "CREATE TABLE test AS SELECT 1", "presto": "CREATE TABLE test WITH (FORMAT='PARQUET', X='1', Z='2') AS SELECT 1", "hive": "CREATE TABLE test STORED AS PARQUET TBLPROPERTIES ('X'='1', 'Z'='2') AS SELECT 1", "spark": "CREATE TABLE test USING PARQUET TBLPROPERTIES ('X'='1', 'Z'='2') AS SELECT 1", @@ -187,6 +189,7 @@ class TestPresto(Validator): self.validate_all( "CREATE TABLE x (w VARCHAR, y INTEGER, z INTEGER) WITH (PARTITIONED_BY=ARRAY['y', 'z'])", write={ + "duckdb": "CREATE TABLE x (w TEXT, y INT, z INT)", "presto": "CREATE TABLE x (w VARCHAR, y INTEGER, z INTEGER) WITH (PARTITIONED_BY=ARRAY['y', 'z'])", "hive": "CREATE TABLE x (w STRING) PARTITIONED BY (y INT, z INT)", "spark": "CREATE TABLE x (w STRING) PARTITIONED BY (y INT, z INT)", @@ -195,6 +198,7 @@ class TestPresto(Validator): self.validate_all( "CREATE TABLE x WITH (bucket_by = ARRAY['y'], bucket_count = 64) AS SELECT 1 AS y", write={ + "duckdb": "CREATE TABLE x AS SELECT 1 AS y", "presto": "CREATE TABLE x WITH (bucket_by=ARRAY['y'], bucket_count=64) AS SELECT 1 AS y", "hive": "CREATE TABLE x TBLPROPERTIES ('bucket_by'=ARRAY('y'), 'bucket_count'=64) AS SELECT 1 AS y", "spark": "CREATE TABLE x TBLPROPERTIES ('bucket_by'=ARRAY('y'), 'bucket_count'=64) AS SELECT 1 AS y", @@ -203,6 +207,7 @@ class TestPresto(Validator): self.validate_all( "CREATE TABLE db.example_table (col_a ROW(struct_col_a INTEGER, struct_col_b VARCHAR))", write={ + "duckdb": "CREATE TABLE db.example_table (col_a STRUCT(struct_col_a INT, struct_col_b TEXT))", "presto": "CREATE TABLE db.example_table (col_a ROW(struct_col_a INTEGER, struct_col_b VARCHAR))", "hive": "CREATE TABLE db.example_table (col_a STRUCT<struct_col_a INT, struct_col_b STRING>)", "spark": "CREATE TABLE db.example_table (col_a STRUCT<struct_col_a: INT, struct_col_b: STRING>)", @@ -211,6 +216,7 @@ class TestPresto(Validator): self.validate_all( "CREATE TABLE db.example_table (col_a ROW(struct_col_a INTEGER, struct_col_b ROW(nested_col_a VARCHAR, nested_col_b VARCHAR)))", write={ + "duckdb": "CREATE TABLE db.example_table (col_a STRUCT(struct_col_a INT, struct_col_b STRUCT(nested_col_a TEXT, nested_col_b TEXT)))", "presto": "CREATE TABLE db.example_table (col_a ROW(struct_col_a INTEGER, struct_col_b ROW(nested_col_a VARCHAR, nested_col_b VARCHAR)))", "hive": "CREATE TABLE db.example_table (col_a STRUCT<struct_col_a INT, struct_col_b STRUCT<nested_col_a STRING, nested_col_b STRING>>)", "spark": "CREATE TABLE db.example_table (col_a STRUCT<struct_col_a: INT, struct_col_b: STRUCT<nested_col_a: STRING, nested_col_b: STRING>>)", diff --git a/tests/dialects/test_spark.py b/tests/dialects/test_spark.py index 9a7e64c..339d1a6 100644 --- a/tests/dialects/test_spark.py +++ b/tests/dialects/test_spark.py @@ -8,6 +8,7 @@ class TestSpark(Validator): self.validate_all( "CREATE TABLE db.example_table (col_a struct<struct_col_a:int, struct_col_b:string>)", write={ + "duckdb": "CREATE TABLE db.example_table (col_a STRUCT(struct_col_a INT, struct_col_b TEXT))", "presto": "CREATE TABLE db.example_table (col_a ROW(struct_col_a INTEGER, struct_col_b VARCHAR))", "hive": "CREATE TABLE db.example_table (col_a STRUCT<struct_col_a INT, struct_col_b STRING>)", "spark": "CREATE TABLE db.example_table (col_a STRUCT<struct_col_a: INT, struct_col_b: STRING>)", @@ -17,6 +18,7 @@ class TestSpark(Validator): "CREATE TABLE db.example_table (col_a struct<struct_col_a:int, struct_col_b:struct<nested_col_a:string, nested_col_b:string>>)", write={ "bigquery": "CREATE TABLE db.example_table (col_a STRUCT<struct_col_a INT64, struct_col_b STRUCT<nested_col_a STRING, nested_col_b STRING>>)", + "duckdb": "CREATE TABLE db.example_table (col_a STRUCT(struct_col_a INT, struct_col_b STRUCT(nested_col_a TEXT, nested_col_b TEXT)))", "presto": "CREATE TABLE db.example_table (col_a ROW(struct_col_a INTEGER, struct_col_b ROW(nested_col_a VARCHAR, nested_col_b VARCHAR)))", "hive": "CREATE TABLE db.example_table (col_a STRUCT<struct_col_a INT, struct_col_b STRUCT<nested_col_a STRING, nested_col_b STRING>>)", "spark": "CREATE TABLE db.example_table (col_a STRUCT<struct_col_a: INT, struct_col_b: STRUCT<nested_col_a: STRING, nested_col_b: STRING>>)", @@ -26,6 +28,7 @@ class TestSpark(Validator): "CREATE TABLE db.example_table (col_a array<int>, col_b array<array<int>>)", write={ "bigquery": "CREATE TABLE db.example_table (col_a ARRAY<INT64>, col_b ARRAY<ARRAY<INT64>>)", + "duckdb": "CREATE TABLE db.example_table (col_a INT[], col_b INT[][])", "presto": "CREATE TABLE db.example_table (col_a ARRAY(INTEGER), col_b ARRAY(ARRAY(INTEGER)))", "hive": "CREATE TABLE db.example_table (col_a ARRAY<INT>, col_b ARRAY<ARRAY<INT>>)", "spark": "CREATE TABLE db.example_table (col_a ARRAY<INT>, col_b ARRAY<ARRAY<INT>>)", @@ -34,6 +37,7 @@ class TestSpark(Validator): self.validate_all( "CREATE TABLE x USING ICEBERG PARTITIONED BY (MONTHS(y)) LOCATION 's3://z'", write={ + "duckdb": "CREATE TABLE x", "presto": "CREATE TABLE x WITH (TABLE_FORMAT = 'ICEBERG', PARTITIONED_BY=ARRAY['MONTHS'])", "hive": "CREATE TABLE x USING ICEBERG PARTITIONED BY (MONTHS(y)) LOCATION 's3://z'", "spark": "CREATE TABLE x USING ICEBERG PARTITIONED BY (MONTHS(y)) LOCATION 's3://z'", @@ -42,6 +46,7 @@ class TestSpark(Validator): self.validate_all( "CREATE TABLE test STORED AS PARQUET AS SELECT 1", write={ + "duckdb": "CREATE TABLE test AS SELECT 1", "presto": "CREATE TABLE test WITH (FORMAT='PARQUET') AS SELECT 1", "hive": "CREATE TABLE test STORED AS PARQUET AS SELECT 1", "spark": "CREATE TABLE test USING PARQUET AS SELECT 1", @@ -50,6 +55,9 @@ class TestSpark(Validator): self.validate_all( """CREATE TABLE blah (col_a INT) COMMENT "Test comment: blah" PARTITIONED BY (date STRING) STORED AS ICEBERG TBLPROPERTIES('x' = '1')""", write={ + "duckdb": """CREATE TABLE blah ( + col_a INT +)""", # Partition columns should exist in table "presto": """CREATE TABLE blah ( col_a INTEGER, date VARCHAR diff --git a/tests/dialects/test_tsql.py b/tests/dialects/test_tsql.py index 6b0b39b..b061784 100644 --- a/tests/dialects/test_tsql.py +++ b/tests/dialects/test_tsql.py @@ -6,6 +6,9 @@ class TestTSQL(Validator): def test_tsql(self): self.validate_identity('SELECT "x"."y" FROM foo') + self.validate_identity( + "SELECT DISTINCT DepartmentName, PERCENTILE_CONT(0.5) WITHIN GROUP (ORDER BY BaseRate) OVER (PARTITION BY DepartmentName) AS MedianCont FROM dbo.DimEmployee" + ) self.validate_all( "SELECT CAST([a].[b] AS SMALLINT) FROM foo", diff --git a/tests/fixtures/identity.sql b/tests/fixtures/identity.sql index 40e7cc1..57e51e0 100644 --- a/tests/fixtures/identity.sql +++ b/tests/fixtures/identity.sql @@ -50,7 +50,9 @@ a.B() a['x'].C() int.x map.x +SELECT call.x a.b.INT(1.234) +INT(x / 100) x IN (-1, 1) x IN ('a', 'a''a') x IN ((1)) @@ -147,6 +149,7 @@ SELECT 1 AS count FROM test SELECT 1 AS comment FROM test SELECT 1 AS numeric FROM test SELECT 1 AS number FROM test +SELECT 1 AS number # annotation SELECT t.count SELECT DISTINCT x FROM test SELECT DISTINCT x, y FROM test @@ -159,6 +162,8 @@ SELECT TIMESTAMP(DATE_TRUNC(DATE(time_field), MONTH)) AS time_value FROM "table" SELECT GREATEST((3 + 1), LEAST(3, 4)) SELECT TRANSFORM(a, b -> b) AS x SELECT AGGREGATE(a, (a, b) -> a + b) AS x +SELECT COUNT(DISTINCT a, b) +SELECT COUNT(DISTINCT a, b + 1) SELECT SUM(DISTINCT x) SELECT SUM(x IGNORE NULLS) AS x SELECT ARRAY_AGG(DISTINCT x IGNORE NULLS ORDER BY a, b DESC LIMIT 10) AS x @@ -485,6 +490,9 @@ CREATE INDEX abc ON t (a) CREATE INDEX abc ON t (a, b, b) CREATE UNIQUE INDEX abc ON t (a, b, b) CREATE UNIQUE INDEX IF NOT EXISTS my_idx ON tbl (a, b) +CREATE SCHEMA x +CREATE SCHEMA IF NOT EXISTS y +DESCRIBE x DROP INDEX a.b.c CACHE TABLE x CACHE LAZY TABLE x @@ -524,6 +532,15 @@ INSERT INTO x VALUES (1, 'a', 2.0), (1, 'a', 3.0), (X(), y[1], z.x) INSERT INTO y (a, b, c) SELECT a, b, c FROM x INSERT OVERWRITE TABLE x IF EXISTS SELECT * FROM y INSERT OVERWRITE TABLE a.b IF EXISTS SELECT * FROM y +INSERT OVERWRITE DIRECTORY 'x' SELECT 1 +INSERT OVERWRITE LOCAL DIRECTORY 'x' SELECT 1 +INSERT OVERWRITE LOCAL DIRECTORY 'x' ROW FORMAT DELIMITED FIELDS TERMINATED BY '1' COLLECTION ITEMS TERMINATED BY '2' MAP KEYS TERMINATED BY '3' LINES TERMINATED BY '4' NULL DEFINED AS '5' SELECT 1 +LOAD DATA INPATH 'x' INTO TABLE y PARTITION(ds='yyyy') +LOAD DATA LOCAL INPATH 'x' INTO TABLE y PARTITION(ds='yyyy') +LOAD DATA LOCAL INPATH 'x' INTO TABLE y PARTITION(ds='yyyy') INPUTFORMAT 'y' +LOAD DATA LOCAL INPATH 'x' INTO TABLE y PARTITION(ds='yyyy') INPUTFORMAT 'y' SERDE 'z' +LOAD DATA INPATH 'x' INTO TABLE y INPUTFORMAT 'y' SERDE 'z' +LOAD DATA INPATH 'x' INTO TABLE y.b INPUTFORMAT 'y' SERDE 'z' SELECT 1 FROM PARQUET_SCAN('/x/y/*') AS y UNCACHE TABLE x UNCACHE TABLE IF EXISTS x diff --git a/tests/fixtures/optimizer/eliminate_ctes.sql b/tests/fixtures/optimizer/eliminate_ctes.sql new file mode 100644 index 0000000..11e5e4f --- /dev/null +++ b/tests/fixtures/optimizer/eliminate_ctes.sql @@ -0,0 +1,48 @@ +# title: CTE +WITH q AS ( + SELECT + a + FROM x +) +SELECT + a +FROM x; +SELECT + a +FROM x; + +# title: Nested CTE +SELECT + a +FROM ( + WITH q AS ( + SELECT + a + FROM x + ) + SELECT a FROM x +); +SELECT + a +FROM ( + SELECT + a + FROM x +); + +# title: Chained CTE +WITH q AS ( + SELECT + a + FROM x +), r AS ( + SELECT + a + FROM q +) +SELECT + a +FROM x; +SELECT + a +FROM x; diff --git a/tests/fixtures/optimizer/eliminate_joins.sql b/tests/fixtures/optimizer/eliminate_joins.sql new file mode 100644 index 0000000..ac3d92d --- /dev/null +++ b/tests/fixtures/optimizer/eliminate_joins.sql @@ -0,0 +1,317 @@ +# title: Remove left join on distinct derived table +SELECT + x.a +FROM x +LEFT JOIN ( + SELECT DISTINCT + y.b + FROM y +) AS y + ON x.b = y.b; +SELECT + x.a +FROM x; + +# title: Remove left join on grouped derived table +SELECT + x.a +FROM x +LEFT JOIN ( + SELECT + y.b, + SUM(y.c) + FROM y + GROUP BY y.b +) AS y + ON x.b = y.b; +SELECT + x.a +FROM x; + +# title: Remove left join on aggregate derived table +SELECT + x.a +FROM x +LEFT JOIN ( + SELECT + SUM(y.b) AS b + FROM y +) AS y + ON x.b = y.b; +SELECT + x.a +FROM x; + +# title: Noop - not all distinct columns in condition +SELECT + x.a +FROM x +LEFT JOIN ( + SELECT DISTINCT + y.b, + y.c + FROM y +) AS y + ON x.b = y.b; +SELECT + x.a +FROM x +LEFT JOIN ( + SELECT DISTINCT + y.b, + y.c + FROM y +) AS y + ON x.b = y.b; + +# title: Noop - not all grouped columns in condition +SELECT + x.a +FROM x +LEFT JOIN ( + SELECT + y.b, + y.c + FROM y + GROUP BY + y.b, + y.c +) AS y + ON x.b = y.b; +SELECT + x.a +FROM x +LEFT JOIN ( + SELECT + y.b, + y.c + FROM y + GROUP BY + y.b, + y.c +) AS y + ON x.b = y.b; + +# title: Noop - not left join +SELECT + x.a +FROM x +JOIN ( + SELECT DISTINCT + y.b + FROM y +) AS y + ON x.b = y.b; +SELECT + x.a +FROM x +JOIN ( + SELECT DISTINCT + y.b + FROM y +) AS y + ON x.b = y.b; + +# title: Noop - unqualified columns +SELECT + a +FROM x +LEFT JOIN ( + SELECT DISTINCT + y.b + FROM y +) AS y + ON x.b = y.b; +SELECT + a +FROM x +LEFT JOIN ( + SELECT DISTINCT + y.b + FROM y +) AS y + ON x.b = y.b; + +# title: Noop - cross join +SELECT + a +FROM x +CROSS JOIN ( + SELECT DISTINCT + y.b + FROM y +) AS y; +SELECT + a +FROM x +CROSS JOIN ( + SELECT DISTINCT + y.b + FROM y +) AS y; + +# title: Noop - column is used +SELECT + x.a, + y.b +FROM x +LEFT JOIN ( + SELECT DISTINCT + y.b + FROM y +) AS y + ON x.b = y.b; +SELECT + x.a, + y.b +FROM x +LEFT JOIN ( + SELECT DISTINCT + y.b + FROM y +) AS y + ON x.b = y.b; + +# title: Multiple group by columns +SELECT + x.a +FROM x +LEFT JOIN ( + SELECT + y.b AS b, + y.c + 1 AS d, + COUNT(1) + FROM y + GROUP BY y.b, y.c + 1 +) AS y + ON x.b = y.b + AND 1 = y.d; +SELECT + x.a +FROM x; + +# title: Chained left joins +SELECT + x.a +FROM x +LEFT JOIN ( + SELECT + y.b AS b + FROM y + GROUP BY y.b +) AS y + ON x.b = y.b +LEFT JOIN ( + SELECT + y.b AS c + FROM y + GROUP BY y.b +) AS z + ON y.b = z.c; +SELECT + x.a +FROM x; + +# title: CTE +WITH z AS ( + SELECT DISTINCT + y.b + FROM y +) +SELECT + x.a +FROM x +LEFT JOIN z + ON x.b = z.b; +WITH z AS ( + SELECT DISTINCT + y.b + FROM y +) +SELECT + x.a +FROM x; + +# title: Noop - Not all grouped expressions are in outputs +SELECT + x.a +FROM x +LEFT JOIN ( + SELECT + y.b + FROM y + GROUP BY + y.b, + y.c +) AS y + ON x.b = y.b; +SELECT + x.a +FROM x +LEFT JOIN ( + SELECT + y.b + FROM y + GROUP BY + y.b, + y.c +) AS y + ON x.b = y.b; + +# title: Cross join on aggregate derived table +SELECT + x.a +FROM x +CROSS JOIN ( + SELECT + SUM(y.b) AS b + FROM y +) AS y; +SELECT + x.a +FROM x; + +# title: Cross join on derived table with LIMIT 1 +SELECT + x.a +FROM x +CROSS JOIN ( + SELECT + y.b AS b + FROM y + LIMIT 1 +) AS y; +SELECT + x.a +FROM x; + +# title: Cross join on derived table with no FROM clause +SELECT + x.a +FROM x +CROSS JOIN ( + SELECT + 1 AS b, + 2 AS c +) AS y; +SELECT + x.a +FROM x; + +# title: Noop - cross join on non-aggregate subquery +SELECT + x.a +FROM x +CROSS JOIN ( + SELECT + y.b + FROM y +) AS y; +SELECT + x.a +FROM x +CROSS JOIN ( + SELECT + y.b + FROM y +) AS y; diff --git a/tests/fixtures/optimizer/merge_subqueries.sql b/tests/fixtures/optimizer/merge_subqueries.sql index c8186cc..a82e1ed 100644 --- a/tests/fixtures/optimizer/merge_subqueries.sql +++ b/tests/fixtures/optimizer/merge_subqueries.sql @@ -187,3 +187,103 @@ FROM ( ) AS x ) AS x; SELECT /*+ BROADCAST(x) */ x.a AS a, x.c AS c FROM (SELECT x.a AS a, COUNT(1) AS c FROM x AS x GROUP BY x.a) AS x; + +# title: Test preventing merge of window expressions where clause +with t1 as ( + SELECT + x.a, + x.b, + ROW_NUMBER() OVER (PARTITION BY x.a ORDER BY x.a) as row_num + FROM + x +) +SELECT + t1.a, + t1.b +FROM + t1 +WHERE + row_num = 1; +WITH t1 AS (SELECT x.a AS a, x.b AS b, ROW_NUMBER() OVER (PARTITION BY x.a ORDER BY x.a) AS row_num FROM x AS x) SELECT t1.a AS a, t1.b AS b FROM t1 WHERE t1.row_num = 1; + +# title: Test preventing merge of window expressions join clause +with t1 as ( + SELECT + x.a, + x.b, + ROW_NUMBER() OVER (PARTITION BY x.a ORDER BY x.a) as row_num + FROM + x +) +SELECT + t1.a, + t1.b +FROM t1 JOIN y ON t1.a = y.c AND t1.row_num = 1; +WITH t1 AS (SELECT x.a AS a, x.b AS b, ROW_NUMBER() OVER (PARTITION BY x.a ORDER BY x.a) AS row_num FROM x AS x) SELECT t1.a AS a, t1.b AS b FROM t1 JOIN y AS y ON t1.a = y.c AND t1.row_num = 1; + +# title: Test preventing merge of window expressions agg function +with t1 as ( + SELECT + x.a, + x.b, + ROW_NUMBER() OVER (PARTITION BY x.a ORDER BY x.a) as row_num + FROM + x +) +SELECT + SUM(t1.row_num) as total_rows +FROM + t1; +WITH t1 AS (SELECT x.a AS a, x.b AS b, ROW_NUMBER() OVER (PARTITION BY x.a ORDER BY x.a) AS row_num FROM x AS x) SELECT SUM(t1.row_num) AS total_rows FROM t1; + +# title: Test prevent merging of window if in group by func +with t1 as ( + SELECT + x.a, + x.b, + ROW_NUMBER() OVER (PARTITION BY x.a ORDER BY x.a) as row_num + FROM + x +) +SELECT + t1.row_num AS row_num, + SUM(t1.a) AS total +FROM + t1 +GROUP BY t1.row_num +ORDER BY t1.row_num; +WITH t1 AS (SELECT x.a AS a, x.b AS b, ROW_NUMBER() OVER (PARTITION BY x.a ORDER BY x.a) AS row_num FROM x AS x) SELECT t1.row_num AS row_num, SUM(t1.a) AS total FROM t1 GROUP BY t1.row_num ORDER BY t1.row_num; + +# title: Test prevent merging of window if in order by func +with t1 as ( + SELECT + x.a, + x.b, + ROW_NUMBER() OVER (PARTITION BY x.a ORDER BY x.a) as row_num + FROM + x +) +SELECT + t1.row_num AS row_num, + t1.a AS a +FROM + t1 +ORDER BY t1.row_num, t1.a; +WITH t1 AS (SELECT x.a AS a, x.b AS b, ROW_NUMBER() OVER (PARTITION BY x.a ORDER BY x.a) AS row_num FROM x AS x) SELECT t1.row_num AS row_num, t1.a AS a FROM t1 ORDER BY t1.row_num, t1.a; + +# title: Test allow merging of window function +with t1 as ( + SELECT + x.a, + x.b, + ROW_NUMBER() OVER (PARTITION BY x.a ORDER BY x.a) as row_num + FROM + x +) +SELECT + t1.a, + t1.b, + t1.row_num +FROM + t1; +SELECT x.a AS a, x.b AS b, ROW_NUMBER() OVER (PARTITION BY x.a ORDER BY x.a) AS row_num FROM x AS x; diff --git a/tests/fixtures/optimizer/optimizer.sql b/tests/fixtures/optimizer/optimizer.sql index ab4f769..eb7e9cb 100644 --- a/tests/fixtures/optimizer/optimizer.sql +++ b/tests/fixtures/optimizer/optimizer.sql @@ -105,9 +105,7 @@ LEFT JOIN "_u_0" AS "_u_0" JOIN "y" AS "y" ON "x"."b" = "y"."b" WHERE - "_u_0"."_col_0" >= 0 - AND "x"."a" > 1 - AND NOT "_u_0"."_u_1" IS NULL + "_u_0"."_col_0" >= 0 AND "x"."a" > 1 AND NOT "_u_0"."_u_1" IS NULL GROUP BY "x"."a"; diff --git a/tests/fixtures/optimizer/pushdown_predicates.sql b/tests/fixtures/optimizer/pushdown_predicates.sql index 676cb96..ef591ec 100644 --- a/tests/fixtures/optimizer/pushdown_predicates.sql +++ b/tests/fixtures/optimizer/pushdown_predicates.sql @@ -30,3 +30,6 @@ SELECT x.a AS a FROM x AS x JOIN (SELECT * FROM y AS y WHERE y.a = 1) AS y ON y. SELECT x.a AS a FROM x AS x CROSS JOIN (SELECT * FROM y AS y) AS y WHERE x.a = 1 AND x.b = 1 AND y.a = x.a AND y.a = 1; SELECT x.a AS a FROM x AS x JOIN (SELECT * FROM y AS y WHERE y.a = 1) AS y ON y.a = x.a AND TRUE WHERE x.a = 1 AND x.b = 1 AND TRUE AND TRUE; + +with t1 as (SELECT x.a, x.b, ROW_NUMBER() OVER (PARTITION BY x.a ORDER BY x.a) as row_num FROM x) SELECT t1.a, t1.b FROM t1 WHERE row_num = 1; +WITH t1 AS (SELECT x.a, x.b, ROW_NUMBER() OVER (PARTITION BY x.a ORDER BY x.a) AS row_num FROM x) SELECT t1.a, t1.b FROM t1 WHERE row_num = 1; diff --git a/tests/fixtures/optimizer/tpc-h/tpc-h.sql b/tests/fixtures/optimizer/tpc-h/tpc-h.sql index 936a0af..b91205c 100644 --- a/tests/fixtures/optimizer/tpc-h/tpc-h.sql +++ b/tests/fixtures/optimizer/tpc-h/tpc-h.sql @@ -28,13 +28,15 @@ SELECT SUM("lineitem"."l_quantity") AS "sum_qty", SUM("lineitem"."l_extendedprice") AS "sum_base_price", SUM("lineitem"."l_extendedprice" * ( - 1 - "lineitem"."l_discount" + 1 - "lineitem"."l_discount" )) AS "sum_disc_price", - SUM("lineitem"."l_extendedprice" * ( + SUM( + "lineitem"."l_extendedprice" * ( 1 - "lineitem"."l_discount" ) * ( 1 + "lineitem"."l_tax" - )) AS "sum_charge", + ) + ) AS "sum_charge", AVG("lineitem"."l_quantity") AS "avg_qty", AVG("lineitem"."l_extendedprice") AS "avg_price", AVG("lineitem"."l_discount") AS "avg_disc", @@ -186,7 +188,7 @@ limit SELECT "lineitem"."l_orderkey" AS "l_orderkey", SUM("lineitem"."l_extendedprice" * ( - 1 - "lineitem"."l_discount" + 1 - "lineitem"."l_discount" )) AS "revenue", CAST("orders"."o_orderdate" AS TEXT) AS "o_orderdate", "orders"."o_shippriority" AS "o_shippriority" @@ -286,7 +288,7 @@ order by SELECT "nation"."n_name" AS "n_name", SUM("lineitem"."l_extendedprice" * ( - 1 - "lineitem"."l_discount" + 1 - "lineitem"."l_discount" )) AS "revenue" FROM "customer" AS "customer" JOIN "orders" AS "orders" @@ -377,15 +379,14 @@ WITH "n1" AS ( "nation"."n_name" AS "n_name" FROM "nation" AS "nation" WHERE - "nation"."n_name" = 'FRANCE' - OR "nation"."n_name" = 'GERMANY' + "nation"."n_name" = 'FRANCE' OR "nation"."n_name" = 'GERMANY' ) SELECT "n1"."n_name" AS "supp_nation", "n2"."n_name" AS "cust_nation", EXTRACT(year FROM "lineitem"."l_shipdate") AS "l_year", SUM("lineitem"."l_extendedprice" * ( - 1 - "lineitem"."l_discount" + 1 - "lineitem"."l_discount" )) AS "revenue" FROM "supplier" AS "supplier" JOIN "lineitem" AS "lineitem" @@ -400,12 +401,10 @@ JOIN "n1" AS "n1" JOIN "n1" AS "n2" ON "customer"."c_nationkey" = "n2"."n_nationkey" AND ( - "n1"."n_name" = 'FRANCE' - OR "n2"."n_name" = 'FRANCE' + "n1"."n_name" = 'FRANCE' OR "n2"."n_name" = 'FRANCE' ) AND ( - "n1"."n_name" = 'GERMANY' - OR "n2"."n_name" = 'GERMANY' + "n1"."n_name" = 'GERMANY' OR "n2"."n_name" = 'GERMANY' ) GROUP BY "n1"."n_name", @@ -458,14 +457,16 @@ order by o_year; SELECT EXTRACT(year FROM "orders"."o_orderdate") AS "o_year", - SUM(CASE + SUM( + CASE WHEN "nation_2"."n_name" = 'BRAZIL' THEN "lineitem"."l_extendedprice" * ( - 1 - "lineitem"."l_discount" - ) + 1 - "lineitem"."l_discount" + ) ELSE 0 - END) / SUM("lineitem"."l_extendedprice" * ( - 1 - "lineitem"."l_discount" + END + ) / SUM("lineitem"."l_extendedprice" * ( + 1 - "lineitem"."l_discount" )) AS "mkt_share" FROM "part" AS "part" JOIN "region" AS "region" @@ -529,9 +530,11 @@ order by SELECT "nation"."n_name" AS "nation", EXTRACT(year FROM "orders"."o_orderdate") AS "o_year", - SUM("lineitem"."l_extendedprice" * ( + SUM( + "lineitem"."l_extendedprice" * ( 1 - "lineitem"."l_discount" - ) - "partsupp"."ps_supplycost" * "lineitem"."l_quantity") AS "sum_profit" + ) - "partsupp"."ps_supplycost" * "lineitem"."l_quantity" + ) AS "sum_profit" FROM "part" AS "part" JOIN "lineitem" AS "lineitem" ON "part"."p_partkey" = "lineitem"."l_partkey" @@ -593,7 +596,7 @@ SELECT "customer"."c_custkey" AS "c_custkey", "customer"."c_name" AS "c_name", SUM("lineitem"."l_extendedprice" * ( - 1 - "lineitem"."l_discount" + 1 - "lineitem"."l_discount" )) AS "revenue", "customer"."c_acctbal" AS "c_acctbal", "nation"."n_name" AS "n_name", @@ -606,8 +609,7 @@ JOIN "orders" AS "orders" AND "orders"."o_orderdate" < CAST('1994-01-01' AS DATE) AND "orders"."o_orderdate" >= CAST('1993-10-01' AS DATE) JOIN "lineitem" AS "lineitem" - ON "lineitem"."l_orderkey" = "orders"."o_orderkey" - AND "lineitem"."l_returnflag" = 'R' + ON "lineitem"."l_orderkey" = "orders"."o_orderkey" AND "lineitem"."l_returnflag" = 'R' JOIN "nation" AS "nation" ON "customer"."c_nationkey" = "nation"."n_nationkey" GROUP BY @@ -721,18 +723,20 @@ order by l_shipmode; SELECT "lineitem"."l_shipmode" AS "l_shipmode", - SUM(CASE - WHEN "orders"."o_orderpriority" = '1-URGENT' - OR "orders"."o_orderpriority" = '2-HIGH' + SUM( + CASE + WHEN "orders"."o_orderpriority" = '1-URGENT' OR "orders"."o_orderpriority" = '2-HIGH' THEN 1 ELSE 0 - END) AS "high_line_count", - SUM(CASE - WHEN "orders"."o_orderpriority" <> '1-URGENT' - AND "orders"."o_orderpriority" <> '2-HIGH' + END + ) AS "high_line_count", + SUM( + CASE + WHEN "orders"."o_orderpriority" <> '1-URGENT' AND "orders"."o_orderpriority" <> '2-HIGH' THEN 1 ELSE 0 - END) AS "low_line_count" + END + ) AS "low_line_count" FROM "orders" AS "orders" JOIN "lineitem" AS "lineitem" ON "lineitem"."l_commitdate" < "lineitem"."l_receiptdate" @@ -813,14 +817,16 @@ where and l_shipdate >= date '1995-09-01' and l_shipdate < date '1995-09-01' + interval '1' month; SELECT - 100.00 * SUM(CASE + 100.00 * SUM( + CASE WHEN "part"."p_type" LIKE 'PROMO%' THEN "lineitem"."l_extendedprice" * ( - 1 - "lineitem"."l_discount" - ) + 1 - "lineitem"."l_discount" + ) ELSE 0 - END) / SUM("lineitem"."l_extendedprice" * ( - 1 - "lineitem"."l_discount" + END + ) / SUM("lineitem"."l_extendedprice" * ( + 1 - "lineitem"."l_discount" )) AS "promo_revenue" FROM "lineitem" AS "lineitem" JOIN "part" AS "part" @@ -866,7 +872,7 @@ WITH "revenue" AS ( SELECT "lineitem"."l_suppkey" AS "supplier_no", SUM("lineitem"."l_extendedprice" * ( - 1 - "lineitem"."l_discount" + 1 - "lineitem"."l_discount" )) AS "total_revenue" FROM "lineitem" AS "lineitem" WHERE @@ -997,8 +1003,7 @@ JOIN "part" AS "part" LEFT JOIN "_u_0" AS "_u_0" ON "_u_0"."_u_1" = "part"."p_partkey" WHERE - "lineitem"."l_quantity" < "_u_0"."_col_0" - AND NOT "_u_0"."_u_1" IS NULL; + "lineitem"."l_quantity" < "_u_0"."_col_0" AND NOT "_u_0"."_u_1" IS NULL; -------------------------------------- -- TPC-H 18 @@ -1114,7 +1119,7 @@ where ); SELECT SUM("lineitem"."l_extendedprice" * ( - 1 - "lineitem"."l_discount" + 1 - "lineitem"."l_discount" )) AS "revenue" FROM "lineitem" AS "lineitem" JOIN "part" AS "part" @@ -1233,8 +1238,7 @@ WITH "_u_0" AS ( "partsupp"."ps_suppkey" AS "ps_suppkey" FROM "partsupp" AS "partsupp" LEFT JOIN "_u_0" AS "_u_0" - ON "_u_0"."_u_1" = "partsupp"."ps_partkey" - AND "_u_0"."_u_2" = "partsupp"."ps_suppkey" + ON "_u_0"."_u_1" = "partsupp"."ps_partkey" AND "_u_0"."_u_2" = "partsupp"."ps_suppkey" LEFT JOIN "_u_3" AS "_u_3" ON "partsupp"."ps_partkey" = "_u_3"."p_partkey" WHERE @@ -1252,8 +1256,7 @@ FROM "supplier" AS "supplier" LEFT JOIN "_u_4" AS "_u_4" ON "supplier"."s_suppkey" = "_u_4"."ps_suppkey" JOIN "nation" AS "nation" - ON "nation"."n_name" = 'CANADA' - AND "supplier"."s_nationkey" = "nation"."n_nationkey" + ON "nation"."n_name" = 'CANADA' AND "supplier"."s_nationkey" = "nation"."n_nationkey" WHERE NOT "_u_4"."ps_suppkey" IS NULL ORDER BY @@ -1332,8 +1335,7 @@ LEFT JOIN "_u_0" AS "_u_0" LEFT JOIN "_u_2" AS "_u_2" ON "_u_2"."l_orderkey" = "lineitem"."l_orderkey" JOIN "orders" AS "orders" - ON "orders"."o_orderkey" = "lineitem"."l_orderkey" - AND "orders"."o_orderstatus" = 'F' + ON "orders"."o_orderkey" = "lineitem"."l_orderkey" AND "orders"."o_orderstatus" = 'F' JOIN "nation" AS "nation" ON "nation"."n_name" = 'SAUDI ARABIA' AND "supplier"."s_nationkey" = "nation"."n_nationkey" diff --git a/tests/fixtures/optimizer/unnest_subqueries.sql b/tests/fixtures/optimizer/unnest_subqueries.sql index 9c4bd27..f53121a 100644 --- a/tests/fixtures/optimizer/unnest_subqueries.sql +++ b/tests/fixtures/optimizer/unnest_subqueries.sql @@ -108,14 +108,11 @@ LEFT JOIN ( ARRAY_AGG(y.b) AS _u_13 FROM y WHERE - TRUE - AND TRUE - AND TRUE + TRUE AND TRUE AND TRUE GROUP BY y.a ) AS "_u_11" - ON "_u_11"."_u_12" = x.a - AND "_u_11"."_u_12" = x.b + ON "_u_11"."_u_12" = x.a AND "_u_11"."_u_12" = x.b LEFT JOIN ( SELECT y.a AS a @@ -131,38 +128,30 @@ WHERE AND NOT "_u_1"."b" IS NULL AND NOT "_u_2"."a" IS NULL AND ( - x.a = "_u_3".b - AND NOT "_u_3"."_u_4" IS NULL + x.a = "_u_3".b AND NOT "_u_3"."_u_4" IS NULL ) AND ( - x.a > "_u_5".b - AND NOT "_u_5"."_u_6" IS NULL + x.a > "_u_5".b AND NOT "_u_5"."_u_6" IS NULL ) AND ( - None = "_u_7".a - AND NOT "_u_7".a IS NULL + None = "_u_7".a AND NOT "_u_7".a IS NULL ) AND NOT ( - x.a = "_u_8".a - AND NOT "_u_8".a IS NULL + x.a = "_u_8".a AND NOT "_u_8".a IS NULL ) AND ( - ARRAY_ANY("_u_9".a, _x -> _x = x.a) - AND NOT "_u_9"."_u_10" IS NULL + ARRAY_ANY("_u_9".a, _x -> _x = x.a) AND NOT "_u_9"."_u_10" IS NULL ) AND ( ( ( - x.a < "_u_11".a - AND NOT "_u_11"."_u_12" IS NULL - ) - AND NOT "_u_11"."_u_12" IS NULL + x.a < "_u_11".a AND NOT "_u_11"."_u_12" IS NULL + ) AND NOT "_u_11"."_u_12" IS NULL ) AND ARRAY_ANY("_u_11"."_u_13", "_x" -> "_x" <> x.d) ) AND ( - NOT "_u_14".a IS NULL - AND NOT "_u_14".a IS NULL + NOT "_u_14".a IS NULL AND NOT "_u_14".a IS NULL ) AND x.a IN ( SELECT @@ -203,4 +192,3 @@ WHERE y.a = x.a OFFSET 10 ); - diff --git a/tests/fixtures/pretty.sql b/tests/fixtures/pretty.sql index 19a7451..2570650 100644 --- a/tests/fixtures/pretty.sql +++ b/tests/fixtures/pretty.sql @@ -56,14 +56,10 @@ LEFT JOIN ( WITH cte1 AS ( SELECT a, - z - AND e AS b + z AND e AS b FROM cte WHERE - x IN (1, 2, 3) - AND z < -1 - OR z > 1 - AND w = 'AND' + x IN (1, 2, 3) AND z < -1 OR z > 1 AND w = 'AND' ), cte2 AS ( SELECT RANK() OVER (PARTITION BY a, b ORDER BY x DESC) AS a, @@ -77,18 +73,12 @@ WITH cte1 AS ( 2 UNION ALL SELECT - CASE x - AND 1 + 1 = 2 + CASE x AND 1 + 1 = 2 WHEN TRUE - THEN 1 - AND 4 + 3 - AND Z - WHEN x - AND y + THEN 1 AND 4 + 3 AND Z + WHEN x AND y THEN 2 - ELSE 3 - AND 4 - AND g + ELSE 3 AND 4 AND g END UNION ALL SELECT @@ -102,18 +92,8 @@ WITH cte1 AS ( ) AS z UNION ALL SELECT - MAX(COALESCE(x - AND y, a - AND b - AND c, d - AND e)), - FOO(CASE - WHEN a - AND b - THEN c - AND d - ELSE 3 - END) + MAX(COALESCE(x AND y, a AND b AND c, d AND e)), + FOO(CASE WHEN a AND b THEN c AND d ELSE 3 END) GROUP BY x GROUPING SETS ( @@ -154,10 +134,8 @@ LEFT JOIN ( FROM bar WHERE ( - c > 1 - AND d > 1 - ) - OR e > 1 + c > 1 AND d > 1 + ) OR e > 1 GROUP BY a HAVING @@ -165,11 +143,8 @@ LEFT JOIN ( LIMIT 10 ) AS z ) AS y - ON x.a = y.b - AND x.a > 1 - OR ( - x.c = y.d - OR x.c = y.e + ON x.a = y.b AND x.a > 1 OR ( + x.c = y.d OR x.c = y.e ); SELECT myCol1, myCol2 FROM baseTable LATERAL VIEW OUTER explode(col1) myTable1 AS myCol1 LATERAL VIEW explode(col2) myTable2 AS myCol2 @@ -184,9 +159,7 @@ EXPLODE(col1) myTable1 AS myCol1 LATERAL VIEW EXPLODE(col2) myTable2 AS myCol2 WHERE - a > 1 - AND b > 2 - OR c > 3; + a > 1 AND b > 2 OR c > 3; SELECT * FROM (WITH y AS ( SELECT 1 AS z) SELECT z from y) x; SELECT @@ -264,3 +237,53 @@ CREATE TABLE "t_customer_account" ( "account_no" VARCHAR(100) ); + +SELECT +x("aaaaaaaaaaaaaa", "bbbbbbbbbbbbb", "ccccccccc", "ddddddddddddd", "eeeeeeeeeeeee", "fffffff"), +array("aaaaaaaaaaaaaa", "bbbbbbbbbbbbb", "ccccccccc", "ddddddddddddd", "eeeeeeeeeeeee", "fffffff"), +array("aaaaaaaaaaaaaa", "bbbbbbbbbbbbb", "ccccccccc", "ddddddddddddd", "eeeeeeeeeeeee", "fffffff", array("aaaaaaaaaaaaaa", "bbbbbbbbbbbbb", "ccccccccc", "ddddddddddddd", "eeeeeeeeeeeee", "fffffff")), +array(array("aaaaaaaaaaaaaa", "bbbbbbbbbbbbb", "ccccccccc", "ddddddddddddd", "eeeeeeeeeeeee", "fffffff")), +; +SELECT + X( + "aaaaaaaaaaaaaa", + "bbbbbbbbbbbbb", + "ccccccccc", + "ddddddddddddd", + "eeeeeeeeeeeee", + "fffffff" + ), + ARRAY( + "aaaaaaaaaaaaaa", + "bbbbbbbbbbbbb", + "ccccccccc", + "ddddddddddddd", + "eeeeeeeeeeeee", + "fffffff" + ), + ARRAY( + "aaaaaaaaaaaaaa", + "bbbbbbbbbbbbb", + "ccccccccc", + "ddddddddddddd", + "eeeeeeeeeeeee", + "fffffff", + ARRAY( + "aaaaaaaaaaaaaa", + "bbbbbbbbbbbbb", + "ccccccccc", + "ddddddddddddd", + "eeeeeeeeeeeee", + "fffffff" + ) + ), + ARRAY( + ARRAY( + "aaaaaaaaaaaaaa", + "bbbbbbbbbbbbb", + "ccccccccc", + "ddddddddddddd", + "eeeeeeeeeeeee", + "fffffff" + ) + ); diff --git a/tests/test_build.py b/tests/test_build.py index fa9e7f8..a432ef1 100644 --- a/tests/test_build.py +++ b/tests/test_build.py @@ -381,6 +381,11 @@ class TestBuild(unittest.TestCase): lambda: alias(parse_one("LAG(x) OVER ()"), "a"), "LAG(x) OVER () AS a", ), + (lambda: exp.values([("1", 2)]), "VALUES ('1', 2)"), + (lambda: exp.values([("1", 2)], "alias"), "(VALUES ('1', 2)) AS alias"), + (lambda: exp.values([("1", 2), ("2", 3)]), "VALUES ('1', 2), ('2', 3)"), + (lambda: exp.delete("y", where="x > 1"), "DELETE FROM y WHERE x > 1"), + (lambda: exp.delete("y", where=exp.and_("x > 1")), "DELETE FROM y WHERE x > 1"), ]: with self.subTest(sql): self.assertEqual(expression().sql(dialect[0] if dialect else None), sql) diff --git a/tests/test_expressions.py b/tests/test_expressions.py index abc95cb..9ad2bf5 100644 --- a/tests/test_expressions.py +++ b/tests/test_expressions.py @@ -1,3 +1,4 @@ +import datetime import unittest from sqlglot import alias, exp, parse_one @@ -29,6 +30,7 @@ class TestExpressions(unittest.TestCase): self.assertEqual(parse_one("TO_DATE(x)", read="hive"), parse_one("ts_or_ds_to_date(x)")) self.assertEqual(exp.Table(pivots=[]), exp.Table()) self.assertNotEqual(exp.Table(pivots=[None]), exp.Table()) + self.assertEqual(exp.DataType.build("int"), exp.DataType(this=exp.DataType.Type.INT, nested=False)) def test_find(self): expression = parse_one("CREATE TABLE x STORED AS PARQUET AS SELECT * FROM y") @@ -486,6 +488,8 @@ class TestExpressions(unittest.TestCase): ((1, "2", None), "(1, '2', NULL)"), ([1, "2", None], "ARRAY(1, '2', NULL)"), ({"x": None}, "MAP('x', NULL)"), + (datetime.datetime(2022, 10, 1, 1, 1, 1), "TIME_STR_TO_TIME('2022-10-01 01:01:01')"), + (datetime.date(2022, 10, 1), "DATE_STR_TO_DATE('2022-10-01')"), ]: with self.subTest(value): self.assertEqual(exp.convert(value).sql(), expected) @@ -496,3 +500,25 @@ class TestExpressions(unittest.TestCase): [e.alias_or_name for e in expression.expressions], ["a", "B", "c", "D"], ) + + def test_to_table(self): + table_only = exp.to_table("table_name") + self.assertEqual(table_only.name, "table_name") + self.assertIsNone(table_only.args.get("db")) + self.assertIsNone(table_only.args.get("catalog")) + db_and_table = exp.to_table("db.table_name") + self.assertEqual(db_and_table.name, "table_name") + self.assertEqual(db_and_table.args.get("db"), exp.to_identifier("db")) + self.assertIsNone(db_and_table.args.get("catalog")) + catalog_db_and_table = exp.to_table("catalog.db.table_name") + self.assertEqual(catalog_db_and_table.name, "table_name") + self.assertEqual(catalog_db_and_table.args.get("db"), exp.to_identifier("db")) + self.assertEqual(catalog_db_and_table.args.get("catalog"), exp.to_identifier("catalog")) + + def test_union(self): + expression = parse_one("SELECT cola, colb UNION SELECT colx, coly") + self.assertIsInstance(expression, exp.Union) + self.assertEqual(expression.named_selects, ["cola", "colb"]) + self.assertEqual( + expression.selects, [exp.Column(this=exp.to_identifier("cola")), exp.Column(this=exp.to_identifier("colb"))] + ) diff --git a/tests/test_optimizer.py b/tests/test_optimizer.py index 36a7785..a67e9db 100644 --- a/tests/test_optimizer.py +++ b/tests/test_optimizer.py @@ -191,6 +191,20 @@ class TestOptimizer(unittest.TestCase): optimizer.optimize_joins.optimize_joins, ) + def test_eliminate_joins(self): + self.check_file( + "eliminate_joins", + optimizer.eliminate_joins.eliminate_joins, + pretty=True, + ) + + def test_eliminate_ctes(self): + self.check_file( + "eliminate_ctes", + optimizer.eliminate_ctes.eliminate_ctes, + pretty=True, + ) + def test_merge_subqueries(self): optimize = partial( optimizer.optimize, |