From be1c7e50e1e8809ea56f2c9d472eccd8ffd73a97 Mon Sep 17 00:00:00 2001 From: Daniel Baumann Date: Fri, 19 Apr 2024 04:57:58 +0200 Subject: Adding upstream version 1.44.3. Signed-off-by: Daniel Baumann --- libnetdata/libjudy/src/JudyL/JudyLPrevEmpty.c | 1390 +++++++++++++++++++++++++ 1 file changed, 1390 insertions(+) create mode 100644 libnetdata/libjudy/src/JudyL/JudyLPrevEmpty.c (limited to 'libnetdata/libjudy/src/JudyL/JudyLPrevEmpty.c') diff --git a/libnetdata/libjudy/src/JudyL/JudyLPrevEmpty.c b/libnetdata/libjudy/src/JudyL/JudyLPrevEmpty.c new file mode 100644 index 00000000..4da43565 --- /dev/null +++ b/libnetdata/libjudy/src/JudyL/JudyLPrevEmpty.c @@ -0,0 +1,1390 @@ +// Copyright (C) 2000 - 2002 Hewlett-Packard Company +// +// This program is free software; you can redistribute it and/or modify it +// under the term of the GNU Lesser General Public License as published by the +// Free Software Foundation; either version 2 of the License, or (at your +// option) any later version. +// +// This program is distributed in the hope that it will be useful, but WITHOUT +// ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or +// FITNESS FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License +// for more details. +// +// You should have received a copy of the GNU Lesser General Public License +// along with this program; if not, write to the Free Software Foundation, +// Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA +// _________________ + +// @(#) $Revision: 4.32 $ $Source: /judy/src/JudyCommon/JudyPrevNextEmpty.c $ +// +// Judy*PrevEmpty() and Judy*NextEmpty() functions for Judy1 and JudyL. +// Compile with one of -DJUDY1 or -DJUDYL. +// +// Compile with -DJUDYNEXT for the Judy*NextEmpty() function; otherwise +// defaults to Judy*PrevEmpty(). +// +// Compile with -DTRACEJPSE to trace JP traversals. +// +// This file is separate from JudyPrevNext.c because it differs too greatly for +// ifdefs. This might be a bit surprising, but there are two reasons: +// +// - First, down in the details, searching for an empty index (SearchEmpty) is +// remarkably asymmetric with searching for a valid index (SearchValid), +// mainly with respect to: No return of a value area for JudyL; partially- +// full versus totally-full JPs; and handling of narrow pointers. +// +// - Second, we chose to implement SearchEmpty without a backtrack stack or +// backtrack engine, partly as an experiment, and partly because we think +// restarting from the top of the tree is less likely for SearchEmpty than +// for SearchValid, because empty indexes are more likely than valid indexes. +// +// A word about naming: A prior version of this feature (see 4.13) was named +// Judy*Free(), but there were concerns about that being read as a verb rather +// than an adjective. After prolonged debate and based on user input, we +// changed "Free" to "Empty". + +#if (! (defined(JUDY1) || defined(JUDYL))) +#error: One of -DJUDY1 or -DJUDYL must be specified. +#endif + +#ifndef JUDYNEXT +#ifndef JUDYPREV +#define JUDYPREV 1 // neither set => use default. +#endif +#endif + +#ifdef JUDY1 +#include "Judy1.h" +#else +#include "JudyL.h" +#endif + +#include "JudyPrivate1L.h" + +#ifdef TRACEJPSE +#include "JudyPrintJP.c" +#endif + + +// **************************************************************************** +// J U D Y 1 P R E V E M P T Y +// J U D Y 1 N E X T E M P T Y +// J U D Y L P R E V E M P T Y +// J U D Y L N E X T E M P T Y +// +// See the manual entry for the API. +// +// OVERVIEW OF Judy*PrevEmpty() / Judy*NextEmpty(): +// +// See also for comparison the equivalent comments in JudyPrevNext.c. +// +// Take the callers *PIndex and subtract/add 1, but watch out for +// underflow/overflow, which means "no previous/next empty index found." Use a +// reentrant switch statement (state machine, see SMGetRestart and +// SMGetContinue) to decode Index, starting with the JRP (PArray), through a +// JPM and branches, if any, down to an immediate or a leaf. Look for Index in +// that immediate or leaf, and if not found (invalid index), return success +// (Index is empty). +// +// This search can result in a dead end where taking a different path is +// required. There are four kinds of dead ends: +// +// BRANCH PRIMARY dead end: Encountering a fully-populated JP for the +// appropriate digit in Index. Search sideways in the branch for the +// previous/next absent/null/non-full JP, and if one is found, set Index to the +// highest/lowest index possible in that JPs expanse. Then if the JP is an +// absent or null JP, return success; otherwise for a non-full JP, traverse +// through the partially populated JP. +// +// BRANCH SECONDARY dead end: Reaching the end of a branch during a sideways +// search after a branch primary dead end. Set Index to the lowest/highest +// index possible in the whole branchs expanse (one higher/lower than the +// previous/next branchs expanse), then restart at the top of the tree, which +// includes pre-decrementing/incrementing Index (again) and watching for +// underflow/overflow (again). +// +// LEAF PRIMARY dead end: Finding a valid (non-empty) index in an immediate or +// leaf matching Index. Search sideways in the immediate/leaf for the +// previous/next empty index; if found, set *PIndex to match and return success. +// +// LEAF SECONDARY dead end: Reaching the end of an immediate or leaf during a +// sideways search after a leaf primary dead end. Just as for a branch +// secondary dead end, restart at the top of the tree with Index set to the +// lowest/highest index possible in the whole immediate/leafs expanse. +// TBD: If leaf secondary dead end occurs, could shortcut and treat it as a +// branch primary dead end; but this would require remembering the parent +// branchs type and offset (a "one-deep stack"), and also wrestling with +// narrow pointers, at least for leaves (but not for immediates). +// +// Note some ASYMMETRIES between SearchValid and SearchEmpty: +// +// - The SearchValid code, upon descending through a narrow pointer, if Index +// is outside the expanse of the subsidiary node (effectively a secondary +// dead end), must decide whether to backtrack or findlimit. But the +// SearchEmpty code simply returns success (Index is empty). +// +// - Similarly, the SearchValid code, upon finding no previous/next index in +// the expanse of a narrow pointer (again, a secondary dead end), can simply +// start to backtrack at the parent JP. But the SearchEmpty code would have +// to first determine whether or not the parent JPs narrow expanse contains +// a previous/next empty index outside the subexpanse. Rather than keeping a +// parent state stack and backtracking this way, upon a secondary dead end, +// the SearchEmpty code simply restarts at the top of the tree, whether or +// not a narrow pointer is involved. Again, see the equivalent comments in +// JudyPrevNext.c for comparison. +// +// This function is written iteratively for speed, rather than recursively. +// +// TBD: Wed like to enhance this function to make successive searches faster. +// This would require saving some previous state, including the previous Index +// returned, and in which leaf it was found. If the next call is for the same +// Index and the array has not been modified, start at the same leaf. This +// should be much easier to implement since this is iterative rather than +// recursive code. + +#ifdef JUDY1 +#ifdef JUDYPREV +FUNCTION int Judy1PrevEmpty +#else +FUNCTION int Judy1NextEmpty +#endif +#else +#ifdef JUDYPREV +FUNCTION int JudyLPrevEmpty +#else +FUNCTION int JudyLNextEmpty +#endif +#endif + ( + Pcvoid_t PArray, // Judy array to search. + Word_t * PIndex, // starting point and result. + PJError_t PJError // optional, for returning error info. + ) +{ + Word_t Index; // fast copy, in a register. + Pjp_t Pjp; // current JP. + Pjbl_t Pjbl; // Pjp->jp_Addr masked and cast to types: + Pjbb_t Pjbb; + Pjbu_t Pjbu; + Pjlb_t Pjlb; + PWord_t Pword; // alternate name for use by GET* macros. + + Word_t digit; // next digit to decode from Index. + Word_t digits; // current state in SM = digits left to decode. + Word_t pop0; // in a leaf. + Word_t pop0mask; // precalculated to avoid variable shifts. + long offset; // within a branch or leaf (can be large). + int subexp; // subexpanse in a bitmap branch. + BITMAPB_t bitposmaskB; // bit in bitmap for bitmap branch. + BITMAPL_t bitposmaskL; // bit in bitmap for bitmap leaf. + Word_t possfullJP1; // JP types for possibly full subexpanses: + Word_t possfullJP2; + Word_t possfullJP3; + + +// ---------------------------------------------------------------------------- +// M A C R O S +// +// These are intended to make the code a bit more readable and less redundant. + + +// CHECK FOR NULL JP: +// +// TBD: In principle this can be reduced (here and in other *.c files) to just +// the latter clause since no Type should ever be below cJU_JPNULL1, but in +// fact some root pointer types can be lower, so for safety do both checks. + +#define JPNULL(Type) (((Type) >= cJU_JPNULL1) && ((Type) <= cJU_JPNULLMAX)) + + +// CHECK FOR A FULL JP: +// +// Given a JP, indicate if it is fully populated. Use digits, pop0mask, and +// possfullJP1..3 in the context. +// +// This is a difficult problem because it requires checking the Pop0 bits for +// all-ones, but the number of bytes depends on the JP type, which is not +// directly related to the parent branchs type or level -- the JPs child +// could be under a narrow pointer (hence not full). The simple answer +// requires switching on or otherwise calculating the JP type, which could be +// slow. Instead, in SMPREPB* precalculate pop0mask and also record in +// possfullJP1..3 the child JP (branch) types that could possibly be full (one +// level down), and use them here. For level-2 branches (with digits == 2), +// the test for a full child depends on Judy1/JudyL. +// +// Note: This cannot be applied to the JP in a JPM because it doesnt have +// enough pop0 digits. +// +// TBD: JPFULL_BRANCH diligently checks for BranchL or BranchB, where neither +// of those can ever be full as it turns out. Could just check for a BranchU +// at the right level. Also, pop0mask might be overkill, its not used much, +// so perhaps just call cJU_POP0MASK(digits - 1) here? +// +// First, JPFULL_BRANCH checks for a full expanse for a JP whose child can be a +// branch, that is, a JP in a branch at level 3 or higher: + +#define JPFULL_BRANCH(Pjp) \ + ((((JU_JPDCDPOP0(Pjp) ^ cJU_ALLONES) & pop0mask) == 0) \ + && ((JU_JPTYPE(Pjp) == possfullJP1) \ + || (JU_JPTYPE(Pjp) == possfullJP2) \ + || (JU_JPTYPE(Pjp) == possfullJP3))) + +#ifdef JUDY1 +#define JPFULL(Pjp) \ + ((digits == 2) ? \ + (JU_JPTYPE(Pjp) == cJ1_JPFULLPOPU1) : JPFULL_BRANCH(Pjp)) +#else +#define JPFULL(Pjp) \ + ((digits == 2) ? \ + (JU_JPTYPE(Pjp) == cJU_JPLEAF_B1) \ + && (((JU_JPDCDPOP0(Pjp) & cJU_POP0MASK(1)) == cJU_POP0MASK(1))) : \ + JPFULL_BRANCH(Pjp)) +#endif + + +// RETURN SUCCESS: +// +// This hides the need to set *PIndex back to the local value of Index -- use a +// local value for faster operation. Note that the callers *PIndex is ALWAYS +// modified upon success, at least decremented/incremented. + +#define RET_SUCCESS { *PIndex = Index; return(1); } + + +// RETURN A CORRUPTION: + +#define RET_CORRUPT { JU_SET_ERRNO(PJError, JU_ERRNO_CORRUPT); return(JERRI); } + + +// SEARCH A BITMAP BRANCH: +// +// This is a weak analog of j__udySearchLeaf*() for bitmap branches. Return +// the actual or next-left position, base 0, of Digit in a BITMAPB_t bitmap +// (subexpanse of a full bitmap), also given a Bitposmask for Digit. The +// position is the offset within the set bits. +// +// Unlike j__udySearchLeaf*(), the offset is not returned bit-complemented if +// Digits bit is unset, because the caller can check the bitmap themselves to +// determine that. Also, if Digits bit is unset, the returned offset is to +// the next-left JP or index (including -1), not to the "ideal" position for +// the index = next-right JP or index. +// +// Shortcut and skip calling j__udyCountBitsB() if the bitmap is full, in which +// case (Digit % cJU_BITSPERSUBEXPB) itself is the base-0 offset. + +#define SEARCHBITMAPB(Bitmap,Digit,Bitposmask) \ + (((Bitmap) == cJU_FULLBITMAPB) ? (Digit % cJU_BITSPERSUBEXPB) : \ + j__udyCountBitsB((Bitmap) & JU_MASKLOWERINC(Bitposmask)) - 1) + +#ifdef JUDYPREV +// Equivalent to search for the highest offset in Bitmap, that is, one less +// than the number of bits set: + +#define SEARCHBITMAPMAXB(Bitmap) \ + (((Bitmap) == cJU_FULLBITMAPB) ? cJU_BITSPERSUBEXPB - 1 : \ + j__udyCountBitsB(Bitmap) - 1) +#endif + + +// CHECK DECODE BYTES: +// +// Check Decode bytes in a JP against the equivalent portion of Index. If they +// dont match, Index is outside the subexpanse of a narrow pointer, hence is +// empty. + +#define CHECKDCD(cDigits) \ + if (JU_DCDNOTMATCHINDEX(Index, Pjp, cDigits)) RET_SUCCESS + + +// REVISE REMAINDER OF INDEX: +// +// Put one digit in place in Index and clear/set the lower digits, if any, so +// the resulting Index is at the start/end of an expanse, or just clear/set the +// least digits. +// +// Actually, to make simple use of JU_LEASTBYTESMASK, first clear/set all least +// digits of Index including the digit to be overridden, then set the value of +// that one digit. If Digits == 1 the first operation is redundant, but either +// very fast or even removed by the optimizer. + +#define CLEARLEASTDIGITS(Digits) Index &= ~JU_LEASTBYTESMASK(Digits) +#define SETLEASTDIGITS( Digits) Index |= JU_LEASTBYTESMASK(Digits) + +#define CLEARLEASTDIGITS_D(Digit,Digits) \ + { \ + CLEARLEASTDIGITS(Digits); \ + JU_SETDIGIT(Index, Digit, Digits); \ + } + +#define SETLEASTDIGITS_D(Digit,Digits) \ + { \ + SETLEASTDIGITS(Digits); \ + JU_SETDIGIT(Index, Digit, Digits); \ + } + + +// SET REMAINDER OF INDEX AND THEN RETURN OR CONTINUE: + +#define SET_AND_RETURN(OpLeastDigits,Digit,Digits) \ + { \ + OpLeastDigits(Digit, Digits); \ + RET_SUCCESS; \ + } + +#define SET_AND_CONTINUE(OpLeastDigits,Digit,Digits) \ + { \ + OpLeastDigits(Digit, Digits); \ + goto SMGetContinue; \ + } + + +// PREPARE TO HANDLE A LEAFW OR JP BRANCH IN THE STATE MACHINE: +// +// Extract a state-dependent digit from Index in a "constant" way, then jump to +// common code for multiple cases. +// +// TBD: Should this macro do more, such as preparing variable-shift masks for +// use in CLEARLEASTDIGITS and SETLEASTDIGITS? + +#define SMPREPB(cDigits,Next,PossFullJP1,PossFullJP2,PossFullJP3) \ + digits = (cDigits); \ + digit = JU_DIGITATSTATE(Index, cDigits); \ + pop0mask = cJU_POP0MASK((cDigits) - 1); /* for branchs JPs */ \ + possfullJP1 = (PossFullJP1); \ + possfullJP2 = (PossFullJP2); \ + possfullJP3 = (PossFullJP3); \ + goto Next + +// Variations for specific-level branches and for shorthands: +// +// Note: SMPREPB2 need not initialize possfullJP* because JPFULL does not use +// them for digits == 2, but gcc -Wall isnt quite smart enough to see this, so +// waste a bit of time and space to get rid of the warning: + +#define SMPREPB2(Next) \ + digits = 2; \ + digit = JU_DIGITATSTATE(Index, 2); \ + pop0mask = cJU_POP0MASK(1); /* for branchs JPs */ \ + possfullJP1 = possfullJP2 = possfullJP3 = 0; \ + goto Next + +#define SMPREPB3(Next) SMPREPB(3, Next, cJU_JPBRANCH_L2, \ + cJU_JPBRANCH_B2, \ + cJU_JPBRANCH_U2) +#ifndef JU_64BIT +#define SMPREPBL(Next) SMPREPB(cJU_ROOTSTATE, Next, cJU_JPBRANCH_L3, \ + cJU_JPBRANCH_B3, \ + cJU_JPBRANCH_U3) +#else +#define SMPREPB4(Next) SMPREPB(4, Next, cJU_JPBRANCH_L3, \ + cJU_JPBRANCH_B3, \ + cJU_JPBRANCH_U3) +#define SMPREPB5(Next) SMPREPB(5, Next, cJU_JPBRANCH_L4, \ + cJU_JPBRANCH_B4, \ + cJU_JPBRANCH_U4) +#define SMPREPB6(Next) SMPREPB(6, Next, cJU_JPBRANCH_L5, \ + cJU_JPBRANCH_B5, \ + cJU_JPBRANCH_U5) +#define SMPREPB7(Next) SMPREPB(7, Next, cJU_JPBRANCH_L6, \ + cJU_JPBRANCH_B6, \ + cJU_JPBRANCH_U6) +#define SMPREPBL(Next) SMPREPB(cJU_ROOTSTATE, Next, cJU_JPBRANCH_L7, \ + cJU_JPBRANCH_B7, \ + cJU_JPBRANCH_U7) +#endif + + +// RESTART AFTER SECONDARY DEAD END: +// +// Set Index to the first/last index in the branch or leaf subexpanse and start +// over at the top of the tree. + +#ifdef JUDYPREV +#define SMRESTART(Digits) { CLEARLEASTDIGITS(Digits); goto SMGetRestart; } +#else +#define SMRESTART(Digits) { SETLEASTDIGITS( Digits); goto SMGetRestart; } +#endif + + +// CHECK EDGE OF LEAFS EXPANSE: +// +// Given the LSBs of the lowest/highest valid index in a leaf (or equivalently +// in an immediate JP), the level (index size) of the leaf, and the full index +// to return (as Index in the context) already set to the full index matching +// the lowest/highest one, determine if there is an empty index in the leafs +// expanse below/above the lowest/highest index, which is true if the +// lowest/highest index is not at the "edge" of the leafs expanse based on its +// LSBs. If so, return Index decremented/incremented; otherwise restart at the +// top of the tree. +// +// Note: In many cases Index is already at the right spot and calling +// SMRESTART instead of just going directly to SMGetRestart is a bit of +// overkill. +// +// Note: Variable shift occurs if Digits is not a constant. + +#ifdef JUDYPREV +#define LEAF_EDGE(MinIndex,Digits) \ + { \ + if (MinIndex) { --Index; RET_SUCCESS; } \ + SMRESTART(Digits); \ + } +#else +#define LEAF_EDGE(MaxIndex,Digits) \ + { \ + if ((MaxIndex) != JU_LEASTBYTES(cJU_ALLONES, Digits)) \ + { ++Index; RET_SUCCESS; } \ + SMRESTART(Digits); \ + } +#endif + +// Same as above except Index is not already set to match the lowest/highest +// index, so do that before decrementing/incrementing it: + +#ifdef JUDYPREV +#define LEAF_EDGE_SET(MinIndex,Digits) \ + { \ + if (MinIndex) \ + { JU_SETDIGITS(Index, MinIndex, Digits); --Index; RET_SUCCESS; } \ + SMRESTART(Digits); \ + } +#else +#define LEAF_EDGE_SET(MaxIndex,Digits) \ + { \ + if ((MaxIndex) != JU_LEASTBYTES(cJU_ALLONES, Digits)) \ + { JU_SETDIGITS(Index, MaxIndex, Digits); ++Index; RET_SUCCESS; } \ + SMRESTART(Digits); \ + } +#endif + + +// FIND A HOLE (EMPTY INDEX) IN AN IMMEDIATE OR LEAF: +// +// Given an index location in a leaf (or equivalently an immediate JP) known to +// contain a usable hole (an empty index less/greater than Index), and the LSBs +// of a minimum/maximum index to locate, find the previous/next empty index and +// return it. +// +// Note: "Even" index sizes (1,2,4[,8] bytes) have corresponding native C +// types; "odd" index sizes dont, but they are not represented here because +// they are handled completely differently; see elsewhere. + +#ifdef JUDYPREV + +#define LEAF_HOLE_EVEN(cDigits,Pjll,IndexLSB) \ + { \ + while (*(Pjll) > (IndexLSB)) --(Pjll); /* too high */ \ + if (*(Pjll) < (IndexLSB)) RET_SUCCESS /* Index is empty */ \ + while (*(--(Pjll)) == --(IndexLSB)) /* null, find a hole */;\ + JU_SETDIGITS(Index, IndexLSB, cDigits); \ + RET_SUCCESS; \ + } +#else +#define LEAF_HOLE_EVEN(cDigits,Pjll,IndexLSB) \ + { \ + while (*(Pjll) < (IndexLSB)) ++(Pjll); /* too low */ \ + if (*(Pjll) > (IndexLSB)) RET_SUCCESS /* Index is empty */ \ + while (*(++(Pjll)) == ++(IndexLSB)) /* null, find a hole */;\ + JU_SETDIGITS(Index, IndexLSB, cDigits); \ + RET_SUCCESS; \ + } +#endif + + +// SEARCH FOR AN EMPTY INDEX IN AN IMMEDIATE OR LEAF: +// +// Given a pointer to the first index in a leaf (or equivalently an immediate +// JP), the population of the leaf, and a first empty Index to find (inclusive, +// as Index in the context), where Index is known to fall within the expanse of +// the leaf to search, efficiently find the previous/next empty index in the +// leaf, if any. For simplicity the following overview is stated in terms of +// Judy*NextEmpty() only, but the same concepts apply symmetrically for +// Judy*PrevEmpty(). Also, in each case the comparisons are for the LSBs of +// Index and leaf indexes, according to the leafs level. +// +// 1. If Index is GREATER than the last (highest) index in the leaf +// (maxindex), return success, Index is empty. (Remember, Index is known +// to be in the leafs expanse.) +// +// 2. If Index is EQUAL to maxindex: If maxindex is not at the edge of the +// leafs expanse, increment Index and return success, there is an empty +// Index one higher than any in the leaf; otherwise restart with Index +// reset to the upper edge of the leafs expanse. Note: This might cause +// an extra cache line fill, but this is OK for repeatedly-called search +// code, and it saves CPU time. +// +// 3. If Index is LESS than maxindex, check for "dense to end of leaf": +// Subtract Index from maxindex, and back up that many slots in the leaf. +// If the resulting offset is not before the start of the leaf then compare +// the index at this offset (baseindex) with Index: +// +// 3a. If GREATER, the leaf must be corrupt, since indexes are sorted and +// there are no duplicates. +// +// 3b. If EQUAL, the leaf is "dense" from Index to maxindex, meaning there is +// no reason to search it. "Slide right" to the high end of the leaf +// (modify Index to maxindex) and continue with step 2 above. +// +// 3c. If LESS, continue with step 4. +// +// 4. If the offset based on maxindex minus Index falls BEFORE the start of +// the leaf, or if, per 3c above, baseindex is LESS than Index, the leaf is +// guaranteed "not dense to the end" and a usable empty Index must exist. +// This supports a more efficient search loop. Start at the FIRST index in +// the leaf, or one BEYOND baseindex, respectively, and search the leaf as +// follows, comparing each current index (currindex) with Index: +// +// 4a. If LESS, keep going to next index. Note: This is certain to terminate +// because maxindex is known to be greater than Index, hence the loop can +// be small and fast. +// +// 4b. If EQUAL, loop and increment Index until finding currindex greater than +// Index, and return success with the modified Index. +// +// 4c. If GREATER, return success, Index (unmodified) is empty. +// +// Note: These are macros rather than functions for speed. + +#ifdef JUDYPREV + +#define JSLE_EVEN(Addr,Pop0,cDigits,LeafType) \ + { \ + LeafType * PjllLSB = (LeafType *) (Addr); \ + LeafType IndexLSB = Index; /* auto-masking */ \ + \ + /* Index before or at start of leaf: */ \ + \ + if (*PjllLSB >= IndexLSB) /* no need to search */ \ + { \ + if (*PjllLSB > IndexLSB) RET_SUCCESS; /* Index empty */ \ + LEAF_EDGE(*PjllLSB, cDigits); \ + } \ + \ + /* Index in or after leaf: */ \ + \ + offset = IndexLSB - *PjllLSB; /* tentative offset */ \ + if (offset <= (Pop0)) /* can check density */ \ + { \ + PjllLSB += offset; /* move to slot */ \ + \ + if (*PjllLSB <= IndexLSB) /* dense or corrupt */ \ + { \ + if (*PjllLSB == IndexLSB) /* dense, check edge */ \ + LEAF_EDGE_SET(PjllLSB[-offset], cDigits); \ + RET_CORRUPT; \ + } \ + --PjllLSB; /* not dense, start at previous */ \ + } \ + else PjllLSB = ((LeafType *) (Addr)) + (Pop0); /* start at max */ \ + \ + LEAF_HOLE_EVEN(cDigits, PjllLSB, IndexLSB); \ + } + +// JSLE_ODD is completely different from JSLE_EVEN because its important to +// minimize copying odd indexes to compare them (see 4.14). Furthermore, a +// very complex version (4.17, but abandoned before fully debugged) that +// avoided calling j__udySearchLeaf*() ran twice as fast as 4.14, but still +// half as fast as SearchValid. Doug suggested that to minimize complexity and +// share common code we should use j__udySearchLeaf*() for the initial search +// to establish if Index is empty, which should be common. If Index is valid +// in a leaf or immediate indexes, odds are good that an empty Index is nearby, +// so for simplicity just use a *COPY* function to linearly search the +// remainder. +// +// TBD: Pathological case? Average performance should be good, but worst-case +// might suffer. When Search says the initial Index is valid, so a linear +// copy-and-compare is begun, if the caller builds fairly large leaves with +// dense clusters AND frequently does a SearchEmpty at one end of such a +// cluster, performance wont be very good. Might a dense-check help? This +// means checking offset against the index at offset, and then against the +// first/last index in the leaf. We doubt the pathological case will appear +// much in real applications because they will probably alternate SearchValid +// and SearchEmpty calls. + +#define JSLE_ODD(cDigits,Pjll,Pop0,Search,Copy) \ + { \ + Word_t IndexLSB; /* least bytes only */ \ + Word_t IndexFound; /* in leaf */ \ + \ + if ((offset = Search(Pjll, (Pop0) + 1, Index)) < 0) \ + RET_SUCCESS; /* Index is empty */ \ + \ + IndexLSB = JU_LEASTBYTES(Index, cDigits); \ + offset *= (cDigits); \ + \ + while ((offset -= (cDigits)) >= 0) \ + { /* skip until empty or start */ \ + Copy(IndexFound, ((uint8_t *) (Pjll)) + offset); \ + if (IndexFound != (--IndexLSB)) /* found an empty */ \ + { JU_SETDIGITS(Index, IndexLSB, cDigits); RET_SUCCESS; }\ + } \ + LEAF_EDGE_SET(IndexLSB, cDigits); \ + } + +#else // JUDYNEXT + +#define JSLE_EVEN(Addr,Pop0,cDigits,LeafType) \ + { \ + LeafType * PjllLSB = ((LeafType *) (Addr)) + (Pop0); \ + LeafType IndexLSB = Index; /* auto-masking */ \ + \ + /* Index at or after end of leaf: */ \ + \ + if (*PjllLSB <= IndexLSB) /* no need to search */ \ + { \ + if (*PjllLSB < IndexLSB) RET_SUCCESS; /* Index empty */\ + LEAF_EDGE(*PjllLSB, cDigits); \ + } \ + \ + /* Index before or in leaf: */ \ + \ + offset = *PjllLSB - IndexLSB; /* tentative offset */ \ + if (offset <= (Pop0)) /* can check density */ \ + { \ + PjllLSB -= offset; /* move to slot */ \ + \ + if (*PjllLSB >= IndexLSB) /* dense or corrupt */ \ + { \ + if (*PjllLSB == IndexLSB) /* dense, check edge */ \ + LEAF_EDGE_SET(PjllLSB[offset], cDigits); \ + RET_CORRUPT; \ + } \ + ++PjllLSB; /* not dense, start at next */ \ + } \ + else PjllLSB = (LeafType *) (Addr); /* start at minimum */ \ + \ + LEAF_HOLE_EVEN(cDigits, PjllLSB, IndexLSB); \ + } + +#define JSLE_ODD(cDigits,Pjll,Pop0,Search,Copy) \ + { \ + Word_t IndexLSB; /* least bytes only */ \ + Word_t IndexFound; /* in leaf */ \ + int offsetmax; /* in bytes */ \ + \ + if ((offset = Search(Pjll, (Pop0) + 1, Index)) < 0) \ + RET_SUCCESS; /* Index is empty */ \ + \ + IndexLSB = JU_LEASTBYTES(Index, cDigits); \ + offset *= (cDigits); \ + offsetmax = (Pop0) * (cDigits); /* single multiply */ \ + \ + while ((offset += (cDigits)) <= offsetmax) \ + { /* skip until empty or end */ \ + Copy(IndexFound, ((uint8_t *) (Pjll)) + offset); \ + if (IndexFound != (++IndexLSB)) /* found an empty */ \ + { JU_SETDIGITS(Index, IndexLSB, cDigits); RET_SUCCESS; } \ + } \ + LEAF_EDGE_SET(IndexLSB, cDigits); \ + } + +#endif // JUDYNEXT + +// Note: Immediate indexes never fill a single index group, so for odd index +// sizes, save time by calling JSLE_ODD_IMM instead of JSLE_ODD. + +#define j__udySearchLeafEmpty1(Addr,Pop0) \ + JSLE_EVEN(Addr, Pop0, 1, uint8_t) + +#define j__udySearchLeafEmpty2(Addr,Pop0) \ + JSLE_EVEN(Addr, Pop0, 2, uint16_t) + +#define j__udySearchLeafEmpty3(Addr,Pop0) \ + JSLE_ODD(3, Addr, Pop0, j__udySearchLeaf3, JU_COPY3_PINDEX_TO_LONG) + +#ifndef JU_64BIT + +#define j__udySearchLeafEmptyL(Addr,Pop0) \ + JSLE_EVEN(Addr, Pop0, 4, Word_t) + +#else + +#define j__udySearchLeafEmpty4(Addr,Pop0) \ + JSLE_EVEN(Addr, Pop0, 4, uint32_t) + +#define j__udySearchLeafEmpty5(Addr,Pop0) \ + JSLE_ODD(5, Addr, Pop0, j__udySearchLeaf5, JU_COPY5_PINDEX_TO_LONG) + +#define j__udySearchLeafEmpty6(Addr,Pop0) \ + JSLE_ODD(6, Addr, Pop0, j__udySearchLeaf6, JU_COPY6_PINDEX_TO_LONG) + +#define j__udySearchLeafEmpty7(Addr,Pop0) \ + JSLE_ODD(7, Addr, Pop0, j__udySearchLeaf7, JU_COPY7_PINDEX_TO_LONG) + +#define j__udySearchLeafEmptyL(Addr,Pop0) \ + JSLE_EVEN(Addr, Pop0, 8, Word_t) + +#endif // JU_64BIT + + +// ---------------------------------------------------------------------------- +// START OF CODE: +// +// CHECK FOR SHORTCUTS: +// +// Error out if PIndex is null. + + if (PIndex == (PWord_t) NULL) + { + JU_SET_ERRNO(PJError, JU_ERRNO_NULLPINDEX); + return(JERRI); + } + + Index = *PIndex; // fast local copy. + +// Set and pre-decrement/increment Index, watching for underflow/overflow: +// +// An out-of-bounds Index means failure: No previous/next empty index. + +SMGetRestart: // return here with revised Index. + +#ifdef JUDYPREV + if (Index-- == 0) return(0); +#else + if (++Index == 0) return(0); +#endif + +// An empty array with an in-bounds (not underflowed/overflowed) Index means +// success: +// +// Note: This check is redundant after restarting at SMGetRestart, but should +// take insignificant time. + + if (PArray == (Pvoid_t) NULL) RET_SUCCESS; + +// ---------------------------------------------------------------------------- +// ROOT-LEVEL LEAF that starts with a Pop0 word; just look within the leaf: +// +// If Index is not in the leaf, return success; otherwise return the first +// empty Index, if any, below/above where it would belong. + + if (JU_LEAFW_POP0(PArray) < cJU_LEAFW_MAXPOP1) // must be a LEAFW + { + Pjlw_t Pjlw = P_JLW(PArray); // first word of leaf. + pop0 = Pjlw[0]; + +#ifdef JUDY1 + if (pop0 == 0) // special case. + { +#ifdef JUDYPREV + if ((Index != Pjlw[1]) || (Index-- != 0)) RET_SUCCESS; +#else + if ((Index != Pjlw[1]) || (++Index != 0)) RET_SUCCESS; +#endif + return(0); // no previous/next empty index. + } +#endif // JUDY1 + + j__udySearchLeafEmptyL(Pjlw + 1, pop0); + +// No return -- thanks ALAN + + } + else + +// ---------------------------------------------------------------------------- +// HANDLE JRP Branch: +// +// For JRP branches, traverse the JPM; handle LEAFW +// directly; but look for the most common cases first. + + { + Pjpm_t Pjpm = P_JPM(PArray); + Pjp = &(Pjpm->jpm_JP); + +// goto SMGetContinue; + } + + +// ============================================================================ +// STATE MACHINE -- GET INDEX: +// +// Search for Index (already decremented/incremented so as to be an inclusive +// search). If not found (empty index), return success. Otherwise do a +// previous/next search, and if successful modify Index to the empty index +// found. See function header comments. +// +// ENTRY: Pjp points to next JP to interpret, whose Decode bytes have not yet +// been checked. +// +// Note: Check Decode bytes at the start of each loop, not after looking up a +// new JP, so its easy to do constant shifts/masks. +// +// EXIT: Return, or branch to SMGetRestart with modified Index, or branch to +// SMGetContinue with a modified Pjp, as described elsewhere. +// +// WARNING: For run-time efficiency the following cases replicate code with +// varying constants, rather than using common code with variable values! + +SMGetContinue: // return here for next branch/leaf. + +#ifdef TRACEJPSE + JudyPrintJP(Pjp, "sf", __LINE__); +#endif + + switch (JU_JPTYPE(Pjp)) + { + + +// ---------------------------------------------------------------------------- +// LINEAR BRANCH: +// +// Check Decode bytes, if any, in the current JP, then search for a JP for the +// next digit in Index. + + case cJU_JPBRANCH_L2: CHECKDCD(2); SMPREPB2(SMBranchL); + case cJU_JPBRANCH_L3: CHECKDCD(3); SMPREPB3(SMBranchL); +#ifdef JU_64BIT + case cJU_JPBRANCH_L4: CHECKDCD(4); SMPREPB4(SMBranchL); + case cJU_JPBRANCH_L5: CHECKDCD(5); SMPREPB5(SMBranchL); + case cJU_JPBRANCH_L6: CHECKDCD(6); SMPREPB6(SMBranchL); + case cJU_JPBRANCH_L7: CHECKDCD(7); SMPREPB7(SMBranchL); +#endif + case cJU_JPBRANCH_L: SMPREPBL(SMBranchL); + +// Common code (state-independent) for all cases of linear branches: + +SMBranchL: + Pjbl = P_JBL(Pjp->jp_Addr); + +// First, check if Indexs expanse (digit) is below/above the first/last +// populated expanse in the BranchL, in which case Index is empty; otherwise +// find the offset of the lowest/highest populated expanse at or above/below +// digit, if any: +// +// Note: The for-loop is guaranteed to exit eventually because the first/last +// expanse is known to be a terminator. +// +// Note: Cannot use j__udySearchLeaf*Empty1() here because it only applies to +// leaves and does not know about partial versus full JPs, unlike the use of +// j__udySearchLeaf1() for BranchLs in SearchValid code. Also, since linear +// leaf expanse lists are small, dont waste time calling j__udySearchLeaf1(), +// just scan the expanse list. + +#ifdef JUDYPREV + if ((Pjbl->jbl_Expanse[0]) > digit) RET_SUCCESS; + + for (offset = (Pjbl->jbl_NumJPs) - 1; /* null */; --offset) +#else + if ((Pjbl->jbl_Expanse[(Pjbl->jbl_NumJPs) - 1]) < digit) + RET_SUCCESS; + + for (offset = 0; /* null */; ++offset) +#endif + { + +// Too low/high, keep going; or too high/low, meaning the loop passed a hole +// and the initial Index is empty: + +#ifdef JUDYPREV + if ((Pjbl->jbl_Expanse[offset]) > digit) continue; + if ((Pjbl->jbl_Expanse[offset]) < digit) RET_SUCCESS; +#else + if ((Pjbl->jbl_Expanse[offset]) < digit) continue; + if ((Pjbl->jbl_Expanse[offset]) > digit) RET_SUCCESS; +#endif + +// Found expanse matching digit; if its not full, traverse through it: + + if (! JPFULL((Pjbl->jbl_jp) + offset)) + { + Pjp = (Pjbl->jbl_jp) + offset; + goto SMGetContinue; + } + +// Common code: While searching for a lower/higher hole or a non-full JP, upon +// finding a lower/higher hole, adjust Index using the revised digit and +// return; or upon finding a consecutive lower/higher expanse, if the expanses +// JP is non-full, modify Index and traverse through the JP: + +#define BRANCHL_CHECK(OpIncDec,OpLeastDigits,Digit,Digits) \ + { \ + if ((Pjbl->jbl_Expanse[offset]) != OpIncDec digit) \ + SET_AND_RETURN(OpLeastDigits, Digit, Digits); \ + \ + if (! JPFULL((Pjbl->jbl_jp) + offset)) \ + { \ + Pjp = (Pjbl->jbl_jp) + offset; \ + SET_AND_CONTINUE(OpLeastDigits, Digit, Digits); \ + } \ + } + +// BranchL primary dead end: Expanse matching Index/digit is full (rare except +// for dense/sequential indexes): +// +// Search for a lower/higher hole, a non-full JP, or the end of the expanse +// list, while decrementing/incrementing digit. + +#ifdef JUDYPREV + while (--offset >= 0) + BRANCHL_CHECK(--, SETLEASTDIGITS_D, digit, digits) +#else + while (++offset < Pjbl->jbl_NumJPs) + BRANCHL_CHECK(++, CLEARLEASTDIGITS_D, digit, digits) +#endif + +// Passed end of BranchL expanse list after finding a matching but full +// expanse: +// +// Digit now matches the lowest/highest expanse, which is a full expanse; if +// digit is at the end of BranchLs expanse (no hole before/after), break out +// of the loop; otherwise modify Index to the next lower/higher digit and +// return success: + +#ifdef JUDYPREV + if (digit == 0) break; + --digit; SET_AND_RETURN(SETLEASTDIGITS_D, digit, digits); +#else + if (digit == JU_LEASTBYTES(cJU_ALLONES, 1)) break; + ++digit; SET_AND_RETURN(CLEARLEASTDIGITS_D, digit, digits); +#endif + } // for-loop + +// BranchL secondary dead end, no non-full previous/next JP: + + SMRESTART(digits); + + +// ---------------------------------------------------------------------------- +// BITMAP BRANCH: +// +// Check Decode bytes, if any, in the current JP, then search for a JP for the +// next digit in Index. + + case cJU_JPBRANCH_B2: CHECKDCD(2); SMPREPB2(SMBranchB); + case cJU_JPBRANCH_B3: CHECKDCD(3); SMPREPB3(SMBranchB); +#ifdef JU_64BIT + case cJU_JPBRANCH_B4: CHECKDCD(4); SMPREPB4(SMBranchB); + case cJU_JPBRANCH_B5: CHECKDCD(5); SMPREPB5(SMBranchB); + case cJU_JPBRANCH_B6: CHECKDCD(6); SMPREPB6(SMBranchB); + case cJU_JPBRANCH_B7: CHECKDCD(7); SMPREPB7(SMBranchB); +#endif + case cJU_JPBRANCH_B: SMPREPBL(SMBranchB); + +// Common code (state-independent) for all cases of bitmap branches: + +SMBranchB: + Pjbb = P_JBB(Pjp->jp_Addr); + +// Locate the digits JP in the subexpanse list, if present: + + subexp = digit / cJU_BITSPERSUBEXPB; + assert(subexp < cJU_NUMSUBEXPB); // falls in expected range. + bitposmaskB = JU_BITPOSMASKB(digit); + +// Absent JP = no JP matches current digit in Index: + +// if (! JU_BITMAPTESTB(Pjbb, digit)) // slower. + if (! (JU_JBB_BITMAP(Pjbb, subexp) & bitposmaskB)) // faster. + RET_SUCCESS; + +// Non-full JP matches current digit in Index: +// +// Iterate to the subsidiary non-full JP. + + offset = SEARCHBITMAPB(JU_JBB_BITMAP(Pjbb, subexp), digit, + bitposmaskB); + // not negative since at least one bit is set: + assert(offset >= 0); + assert(offset < (int) cJU_BITSPERSUBEXPB); + +// Watch for null JP subarray pointer with non-null bitmap (a corruption): + + if ((Pjp = P_JP(JU_JBB_PJP(Pjbb, subexp))) + == (Pjp_t) NULL) RET_CORRUPT; + + Pjp += offset; + if (! JPFULL(Pjp)) goto SMGetContinue; + +// BranchB primary dead end: +// +// Upon hitting a full JP in a BranchB for the next digit in Index, search +// sideways for a previous/next absent JP (unset bit) or non-full JP (set bit +// with non-full JP); first in the current bitmap subexpanse, then in +// lower/higher subexpanses. Upon entry, Pjp points to a known-unusable JP, +// ready to decrement/increment. +// +// Note: The preceding code is separate from this loop because Index does not +// need revising (see SET_AND_*()) if the initial index is an empty index. +// +// TBD: For speed, shift bitposmaskB instead of using JU_BITMAPTESTB or +// JU_BITPOSMASKB, but this shift has knowledge of bit order that really should +// be encapsulated in a header file. + +#define BRANCHB_CHECKBIT(OpLeastDigits) \ + if (! (JU_JBB_BITMAP(Pjbb, subexp) & bitposmaskB)) /* absent JP */ \ + SET_AND_RETURN(OpLeastDigits, digit, digits) + +#define BRANCHB_CHECKJPFULL(OpLeastDigits) \ + if (! JPFULL(Pjp)) \ + SET_AND_CONTINUE(OpLeastDigits, digit, digits) + +#define BRANCHB_STARTSUBEXP(OpLeastDigits) \ + if (! JU_JBB_BITMAP(Pjbb, subexp)) /* empty subexpanse, shortcut */ \ + SET_AND_RETURN(OpLeastDigits, digit, digits) \ + if ((Pjp = P_JP(JU_JBB_PJP(Pjbb, subexp))) == (Pjp_t) NULL) RET_CORRUPT + +#ifdef JUDYPREV + + --digit; // skip initial digit. + bitposmaskB >>= 1; // see TBD above. + +BranchBNextSubexp: // return here to check next bitmap subexpanse. + + while (bitposmaskB) // more bits to check in subexp. + { + BRANCHB_CHECKBIT(SETLEASTDIGITS_D); + --Pjp; // previous in subarray. + BRANCHB_CHECKJPFULL(SETLEASTDIGITS_D); + assert(digit >= 0); + --digit; + bitposmaskB >>= 1; + } + + if (subexp-- > 0) // more subexpanses. + { + BRANCHB_STARTSUBEXP(SETLEASTDIGITS_D); + Pjp += SEARCHBITMAPMAXB(JU_JBB_BITMAP(Pjbb, subexp)) + 1; + bitposmaskB = (1U << (cJU_BITSPERSUBEXPB - 1)); + goto BranchBNextSubexp; + } + +#else // JUDYNEXT + + ++digit; // skip initial digit. + bitposmaskB <<= 1; // note: BITMAPB_t. + +BranchBNextSubexp: // return here to check next bitmap subexpanse. + + while (bitposmaskB) // more bits to check in subexp. + { + BRANCHB_CHECKBIT(CLEARLEASTDIGITS_D); + ++Pjp; // previous in subarray. + BRANCHB_CHECKJPFULL(CLEARLEASTDIGITS_D); + assert(digit < cJU_SUBEXPPERSTATE); + ++digit; + bitposmaskB <<= 1; // note: BITMAPB_t. + } + + if (++subexp < cJU_NUMSUBEXPB) // more subexpanses. + { + BRANCHB_STARTSUBEXP(CLEARLEASTDIGITS_D); + --Pjp; // pre-decrement. + bitposmaskB = 1; + goto BranchBNextSubexp; + } + +#endif // JUDYNEXT + +// BranchB secondary dead end, no non-full previous/next JP: + + SMRESTART(digits); + + +// ---------------------------------------------------------------------------- +// UNCOMPRESSED BRANCH: +// +// Check Decode bytes, if any, in the current JP, then search for a JP for the +// next digit in Index. + + case cJU_JPBRANCH_U2: CHECKDCD(2); SMPREPB2(SMBranchU); + case cJU_JPBRANCH_U3: CHECKDCD(3); SMPREPB3(SMBranchU); +#ifdef JU_64BIT + case cJU_JPBRANCH_U4: CHECKDCD(4); SMPREPB4(SMBranchU); + case cJU_JPBRANCH_U5: CHECKDCD(5); SMPREPB5(SMBranchU); + case cJU_JPBRANCH_U6: CHECKDCD(6); SMPREPB6(SMBranchU); + case cJU_JPBRANCH_U7: CHECKDCD(7); SMPREPB7(SMBranchU); +#endif + case cJU_JPBRANCH_U: SMPREPBL(SMBranchU); + +// Common code (state-independent) for all cases of uncompressed branches: + +SMBranchU: + Pjbu = P_JBU(Pjp->jp_Addr); + Pjp = (Pjbu->jbu_jp) + digit; + +// Absent JP = null JP for current digit in Index: + + if (JPNULL(JU_JPTYPE(Pjp))) RET_SUCCESS; + +// Non-full JP matches current digit in Index: +// +// Iterate to the subsidiary JP. + + if (! JPFULL(Pjp)) goto SMGetContinue; + +// BranchU primary dead end: +// +// Upon hitting a full JP in a BranchU for the next digit in Index, search +// sideways for a previous/next null or non-full JP. BRANCHU_CHECKJP() is +// shorthand for common code. +// +// Note: The preceding code is separate from this loop because Index does not +// need revising (see SET_AND_*()) if the initial index is an empty index. + +#define BRANCHU_CHECKJP(OpIncDec,OpLeastDigits) \ + { \ + OpIncDec Pjp; \ + \ + if (JPNULL(JU_JPTYPE(Pjp))) \ + SET_AND_RETURN(OpLeastDigits, digit, digits) \ + \ + if (! JPFULL(Pjp)) \ + SET_AND_CONTINUE(OpLeastDigits, digit, digits) \ + } + +#ifdef JUDYPREV + while (digit-- > 0) + BRANCHU_CHECKJP(--, SETLEASTDIGITS_D); +#else + while (++digit < cJU_BRANCHUNUMJPS) + BRANCHU_CHECKJP(++, CLEARLEASTDIGITS_D); +#endif + +// BranchU secondary dead end, no non-full previous/next JP: + + SMRESTART(digits); + + +// ---------------------------------------------------------------------------- +// LINEAR LEAF: +// +// Check Decode bytes, if any, in the current JP, then search the leaf for the +// previous/next empty index starting at Index. Primary leaf dead end is +// hidden within j__udySearchLeaf*Empty*(). In case of secondary leaf dead +// end, restart at the top of the tree. +// +// Note: Pword is the name known to GET*; think of it as Pjlw. + +#define SMLEAFL(cDigits,Func) \ + Pword = (PWord_t) P_JLW(Pjp->jp_Addr); \ + pop0 = JU_JPLEAF_POP0(Pjp); \ + Func(Pword, pop0) + +#if (defined(JUDYL) || (! defined(JU_64BIT))) + case cJU_JPLEAF1: CHECKDCD(1); SMLEAFL(1, j__udySearchLeafEmpty1); +#endif + case cJU_JPLEAF2: CHECKDCD(2); SMLEAFL(2, j__udySearchLeafEmpty2); + case cJU_JPLEAF3: CHECKDCD(3); SMLEAFL(3, j__udySearchLeafEmpty3); + +#ifdef JU_64BIT + case cJU_JPLEAF4: CHECKDCD(4); SMLEAFL(4, j__udySearchLeafEmpty4); + case cJU_JPLEAF5: CHECKDCD(5); SMLEAFL(5, j__udySearchLeafEmpty5); + case cJU_JPLEAF6: CHECKDCD(6); SMLEAFL(6, j__udySearchLeafEmpty6); + case cJU_JPLEAF7: CHECKDCD(7); SMLEAFL(7, j__udySearchLeafEmpty7); +#endif + + +// ---------------------------------------------------------------------------- +// BITMAP LEAF: +// +// Check Decode bytes, if any, in the current JP, then search the leaf for the +// previous/next empty index starting at Index. + + case cJU_JPLEAF_B1: + + CHECKDCD(1); + + Pjlb = P_JLB(Pjp->jp_Addr); + digit = JU_DIGITATSTATE(Index, 1); + subexp = digit / cJU_BITSPERSUBEXPL; + bitposmaskL = JU_BITPOSMASKL(digit); + assert(subexp < cJU_NUMSUBEXPL); // falls in expected range. + +// Absent index = no index matches current digit in Index: + +// if (! JU_BITMAPTESTL(Pjlb, digit)) // slower. + if (! (JU_JLB_BITMAP(Pjlb, subexp) & bitposmaskL)) // faster. + RET_SUCCESS; + +// LeafB1 primary dead end: +// +// Upon hitting a valid (non-empty) index in a LeafB1 for the last digit in +// Index, search sideways for a previous/next absent index, first in the +// current bitmap subexpanse, then in lower/higher subexpanses. +// LEAFB1_CHECKBIT() is shorthand for common code to handle one bit in one +// bitmap subexpanse. +// +// Note: The preceding code is separate from this loop because Index does not +// need revising (see SET_AND_*()) if the initial index is an empty index. +// +// TBD: For speed, shift bitposmaskL instead of using JU_BITMAPTESTL or +// JU_BITPOSMASKL, but this shift has knowledge of bit order that really should +// be encapsulated in a header file. + +#define LEAFB1_CHECKBIT(OpLeastDigits) \ + if (! (JU_JLB_BITMAP(Pjlb, subexp) & bitposmaskL)) \ + SET_AND_RETURN(OpLeastDigits, digit, 1) + +#define LEAFB1_STARTSUBEXP(OpLeastDigits) \ + if (! JU_JLB_BITMAP(Pjlb, subexp)) /* empty subexp */ \ + SET_AND_RETURN(OpLeastDigits, digit, 1) + +#ifdef JUDYPREV + + --digit; // skip initial digit. + bitposmaskL >>= 1; // see TBD above. + +LeafB1NextSubexp: // return here to check next bitmap subexpanse. + + while (bitposmaskL) // more bits to check in subexp. + { + LEAFB1_CHECKBIT(SETLEASTDIGITS_D); + assert(digit >= 0); + --digit; + bitposmaskL >>= 1; + } + + if (subexp-- > 0) // more subexpanses. + { + LEAFB1_STARTSUBEXP(SETLEASTDIGITS_D); + bitposmaskL = (1UL << (cJU_BITSPERSUBEXPL - 1)); + goto LeafB1NextSubexp; + } + +#else // JUDYNEXT + + ++digit; // skip initial digit. + bitposmaskL <<= 1; // note: BITMAPL_t. + +LeafB1NextSubexp: // return here to check next bitmap subexpanse. + + while (bitposmaskL) // more bits to check in subexp. + { + LEAFB1_CHECKBIT(CLEARLEASTDIGITS_D); + assert(digit < cJU_SUBEXPPERSTATE); + ++digit; + bitposmaskL <<= 1; // note: BITMAPL_t. + } + + if (++subexp < cJU_NUMSUBEXPL) // more subexpanses. + { + LEAFB1_STARTSUBEXP(CLEARLEASTDIGITS_D); + bitposmaskL = 1; + goto LeafB1NextSubexp; + } + +#endif // JUDYNEXT + +// LeafB1 secondary dead end, no empty index: + + SMRESTART(1); + + +#ifdef JUDY1 +// ---------------------------------------------------------------------------- +// FULL POPULATION: +// +// If the Decode bytes do not match, Index is empty (without modification); +// otherwise restart. + + case cJ1_JPFULLPOPU1: + + CHECKDCD(1); + SMRESTART(1); +#endif + + +// ---------------------------------------------------------------------------- +// IMMEDIATE: +// +// Pop1 = 1 Immediate JPs: +// +// If Index is not in the immediate JP, return success; otherwise check if +// there is an empty index below/above the immediate JPs index, and if so, +// return success with modified Index, else restart. +// +// Note: Doug says its fast enough to calculate the index size (digits) in +// the following; no need to set it separately for each case. + + case cJU_JPIMMED_1_01: + case cJU_JPIMMED_2_01: + case cJU_JPIMMED_3_01: +#ifdef JU_64BIT + case cJU_JPIMMED_4_01: + case cJU_JPIMMED_5_01: + case cJU_JPIMMED_6_01: + case cJU_JPIMMED_7_01: +#endif + if (JU_JPDCDPOP0(Pjp) != JU_TRIMTODCDSIZE(Index)) RET_SUCCESS; + digits = JU_JPTYPE(Pjp) - cJU_JPIMMED_1_01 + 1; + LEAF_EDGE(JU_LEASTBYTES(JU_JPDCDPOP0(Pjp), digits), digits); + +// Immediate JPs with Pop1 > 1: + +#define IMM_MULTI(Func,BaseJPType) \ + JUDY1CODE(Pword = (PWord_t) (Pjp->jp_1Index);) \ + JUDYLCODE(Pword = (PWord_t) (Pjp->jp_LIndex);) \ + Func(Pword, JU_JPTYPE(Pjp) - (BaseJPType) + 1) + + case cJU_JPIMMED_1_02: + case cJU_JPIMMED_1_03: +#if (defined(JUDY1) || defined(JU_64BIT)) + case cJU_JPIMMED_1_04: + case cJU_JPIMMED_1_05: + case cJU_JPIMMED_1_06: + case cJU_JPIMMED_1_07: +#endif +#if (defined(JUDY1) && defined(JU_64BIT)) + case cJ1_JPIMMED_1_08: + case cJ1_JPIMMED_1_09: + case cJ1_JPIMMED_1_10: + case cJ1_JPIMMED_1_11: + case cJ1_JPIMMED_1_12: + case cJ1_JPIMMED_1_13: + case cJ1_JPIMMED_1_14: + case cJ1_JPIMMED_1_15: +#endif + IMM_MULTI(j__udySearchLeafEmpty1, cJU_JPIMMED_1_02); + +#if (defined(JUDY1) || defined(JU_64BIT)) + case cJU_JPIMMED_2_02: + case cJU_JPIMMED_2_03: +#endif +#if (defined(JUDY1) && defined(JU_64BIT)) + case cJ1_JPIMMED_2_04: + case cJ1_JPIMMED_2_05: + case cJ1_JPIMMED_2_06: + case cJ1_JPIMMED_2_07: +#endif +#if (defined(JUDY1) || defined(JU_64BIT)) + IMM_MULTI(j__udySearchLeafEmpty2, cJU_JPIMMED_2_02); +#endif + +#if (defined(JUDY1) || defined(JU_64BIT)) + case cJU_JPIMMED_3_02: +#endif +#if (defined(JUDY1) && defined(JU_64BIT)) + case cJ1_JPIMMED_3_03: + case cJ1_JPIMMED_3_04: + case cJ1_JPIMMED_3_05: +#endif +#if (defined(JUDY1) || defined(JU_64BIT)) + IMM_MULTI(j__udySearchLeafEmpty3, cJU_JPIMMED_3_02); +#endif + +#if (defined(JUDY1) && defined(JU_64BIT)) + case cJ1_JPIMMED_4_02: + case cJ1_JPIMMED_4_03: + IMM_MULTI(j__udySearchLeafEmpty4, cJ1_JPIMMED_4_02); + + case cJ1_JPIMMED_5_02: + case cJ1_JPIMMED_5_03: + IMM_MULTI(j__udySearchLeafEmpty5, cJ1_JPIMMED_5_02); + + case cJ1_JPIMMED_6_02: + IMM_MULTI(j__udySearchLeafEmpty6, cJ1_JPIMMED_6_02); + + case cJ1_JPIMMED_7_02: + IMM_MULTI(j__udySearchLeafEmpty7, cJ1_JPIMMED_7_02); +#endif + + +// ---------------------------------------------------------------------------- +// INVALID JP TYPE: + + default: RET_CORRUPT; + + } // SMGet switch. + +} // Judy1PrevEmpty() / Judy1NextEmpty() / JudyLPrevEmpty() / JudyLNextEmpty() -- cgit v1.2.3