OLD | NEW |
| (Empty) |
1 /* | |
2 ** 2001 September 15 | |
3 ** | |
4 ** The author disclaims copyright to this source code. In place of | |
5 ** a legal notice, here is a blessing: | |
6 ** | |
7 ** May you do good and not evil. | |
8 ** May you find forgiveness for yourself and forgive others. | |
9 ** May you share freely, never taking more than you give. | |
10 ** | |
11 ************************************************************************* | |
12 ** This module contains C code that generates VDBE code used to process | |
13 ** the WHERE clause of SQL statements. This module is responsible for | |
14 ** generating the code that loops through a table looking for applicable | |
15 ** rows. Indices are selected and used to speed the search when doing | |
16 ** so is applicable. Because this module is responsible for selecting | |
17 ** indices, you might also think of this module as the "query optimizer". | |
18 */ | |
19 #include "sqliteInt.h" | |
20 #include "whereInt.h" | |
21 | |
22 /* Forward declaration of methods */ | |
23 static int whereLoopResize(sqlite3*, WhereLoop*, int); | |
24 | |
25 /* Test variable that can be set to enable WHERE tracing */ | |
26 #if defined(SQLITE_TEST) || defined(SQLITE_DEBUG) | |
27 /***/ int sqlite3WhereTrace = 0; | |
28 #endif | |
29 | |
30 | |
31 /* | |
32 ** Return the estimated number of output rows from a WHERE clause | |
33 */ | |
34 u64 sqlite3WhereOutputRowCount(WhereInfo *pWInfo){ | |
35 return sqlite3LogEstToInt(pWInfo->nRowOut); | |
36 } | |
37 | |
38 /* | |
39 ** Return one of the WHERE_DISTINCT_xxxxx values to indicate how this | |
40 ** WHERE clause returns outputs for DISTINCT processing. | |
41 */ | |
42 int sqlite3WhereIsDistinct(WhereInfo *pWInfo){ | |
43 return pWInfo->eDistinct; | |
44 } | |
45 | |
46 /* | |
47 ** Return TRUE if the WHERE clause returns rows in ORDER BY order. | |
48 ** Return FALSE if the output needs to be sorted. | |
49 */ | |
50 int sqlite3WhereIsOrdered(WhereInfo *pWInfo){ | |
51 return pWInfo->nOBSat; | |
52 } | |
53 | |
54 /* | |
55 ** Return the VDBE address or label to jump to in order to continue | |
56 ** immediately with the next row of a WHERE clause. | |
57 */ | |
58 int sqlite3WhereContinueLabel(WhereInfo *pWInfo){ | |
59 assert( pWInfo->iContinue!=0 ); | |
60 return pWInfo->iContinue; | |
61 } | |
62 | |
63 /* | |
64 ** Return the VDBE address or label to jump to in order to break | |
65 ** out of a WHERE loop. | |
66 */ | |
67 int sqlite3WhereBreakLabel(WhereInfo *pWInfo){ | |
68 return pWInfo->iBreak; | |
69 } | |
70 | |
71 /* | |
72 ** Return ONEPASS_OFF (0) if an UPDATE or DELETE statement is unable to | |
73 ** operate directly on the rowis returned by a WHERE clause. Return | |
74 ** ONEPASS_SINGLE (1) if the statement can operation directly because only | |
75 ** a single row is to be changed. Return ONEPASS_MULTI (2) if the one-pass | |
76 ** optimization can be used on multiple | |
77 ** | |
78 ** If the ONEPASS optimization is used (if this routine returns true) | |
79 ** then also write the indices of open cursors used by ONEPASS | |
80 ** into aiCur[0] and aiCur[1]. iaCur[0] gets the cursor of the data | |
81 ** table and iaCur[1] gets the cursor used by an auxiliary index. | |
82 ** Either value may be -1, indicating that cursor is not used. | |
83 ** Any cursors returned will have been opened for writing. | |
84 ** | |
85 ** aiCur[0] and aiCur[1] both get -1 if the where-clause logic is | |
86 ** unable to use the ONEPASS optimization. | |
87 */ | |
88 int sqlite3WhereOkOnePass(WhereInfo *pWInfo, int *aiCur){ | |
89 memcpy(aiCur, pWInfo->aiCurOnePass, sizeof(int)*2); | |
90 #ifdef WHERETRACE_ENABLED | |
91 if( sqlite3WhereTrace && pWInfo->eOnePass!=ONEPASS_OFF ){ | |
92 sqlite3DebugPrintf("%s cursors: %d %d\n", | |
93 pWInfo->eOnePass==ONEPASS_SINGLE ? "ONEPASS_SINGLE" : "ONEPASS_MULTI", | |
94 aiCur[0], aiCur[1]); | |
95 } | |
96 #endif | |
97 return pWInfo->eOnePass; | |
98 } | |
99 | |
100 /* | |
101 ** Move the content of pSrc into pDest | |
102 */ | |
103 static void whereOrMove(WhereOrSet *pDest, WhereOrSet *pSrc){ | |
104 pDest->n = pSrc->n; | |
105 memcpy(pDest->a, pSrc->a, pDest->n*sizeof(pDest->a[0])); | |
106 } | |
107 | |
108 /* | |
109 ** Try to insert a new prerequisite/cost entry into the WhereOrSet pSet. | |
110 ** | |
111 ** The new entry might overwrite an existing entry, or it might be | |
112 ** appended, or it might be discarded. Do whatever is the right thing | |
113 ** so that pSet keeps the N_OR_COST best entries seen so far. | |
114 */ | |
115 static int whereOrInsert( | |
116 WhereOrSet *pSet, /* The WhereOrSet to be updated */ | |
117 Bitmask prereq, /* Prerequisites of the new entry */ | |
118 LogEst rRun, /* Run-cost of the new entry */ | |
119 LogEst nOut /* Number of outputs for the new entry */ | |
120 ){ | |
121 u16 i; | |
122 WhereOrCost *p; | |
123 for(i=pSet->n, p=pSet->a; i>0; i--, p++){ | |
124 if( rRun<=p->rRun && (prereq & p->prereq)==prereq ){ | |
125 goto whereOrInsert_done; | |
126 } | |
127 if( p->rRun<=rRun && (p->prereq & prereq)==p->prereq ){ | |
128 return 0; | |
129 } | |
130 } | |
131 if( pSet->n<N_OR_COST ){ | |
132 p = &pSet->a[pSet->n++]; | |
133 p->nOut = nOut; | |
134 }else{ | |
135 p = pSet->a; | |
136 for(i=1; i<pSet->n; i++){ | |
137 if( p->rRun>pSet->a[i].rRun ) p = pSet->a + i; | |
138 } | |
139 if( p->rRun<=rRun ) return 0; | |
140 } | |
141 whereOrInsert_done: | |
142 p->prereq = prereq; | |
143 p->rRun = rRun; | |
144 if( p->nOut>nOut ) p->nOut = nOut; | |
145 return 1; | |
146 } | |
147 | |
148 /* | |
149 ** Return the bitmask for the given cursor number. Return 0 if | |
150 ** iCursor is not in the set. | |
151 */ | |
152 Bitmask sqlite3WhereGetMask(WhereMaskSet *pMaskSet, int iCursor){ | |
153 int i; | |
154 assert( pMaskSet->n<=(int)sizeof(Bitmask)*8 ); | |
155 for(i=0; i<pMaskSet->n; i++){ | |
156 if( pMaskSet->ix[i]==iCursor ){ | |
157 return MASKBIT(i); | |
158 } | |
159 } | |
160 return 0; | |
161 } | |
162 | |
163 /* | |
164 ** Create a new mask for cursor iCursor. | |
165 ** | |
166 ** There is one cursor per table in the FROM clause. The number of | |
167 ** tables in the FROM clause is limited by a test early in the | |
168 ** sqlite3WhereBegin() routine. So we know that the pMaskSet->ix[] | |
169 ** array will never overflow. | |
170 */ | |
171 static void createMask(WhereMaskSet *pMaskSet, int iCursor){ | |
172 assert( pMaskSet->n < ArraySize(pMaskSet->ix) ); | |
173 pMaskSet->ix[pMaskSet->n++] = iCursor; | |
174 } | |
175 | |
176 /* | |
177 ** Advance to the next WhereTerm that matches according to the criteria | |
178 ** established when the pScan object was initialized by whereScanInit(). | |
179 ** Return NULL if there are no more matching WhereTerms. | |
180 */ | |
181 static WhereTerm *whereScanNext(WhereScan *pScan){ | |
182 int iCur; /* The cursor on the LHS of the term */ | |
183 i16 iColumn; /* The column on the LHS of the term. -1 for IPK */ | |
184 Expr *pX; /* An expression being tested */ | |
185 WhereClause *pWC; /* Shorthand for pScan->pWC */ | |
186 WhereTerm *pTerm; /* The term being tested */ | |
187 int k = pScan->k; /* Where to start scanning */ | |
188 | |
189 while( pScan->iEquiv<=pScan->nEquiv ){ | |
190 iCur = pScan->aiCur[pScan->iEquiv-1]; | |
191 iColumn = pScan->aiColumn[pScan->iEquiv-1]; | |
192 if( iColumn==XN_EXPR && pScan->pIdxExpr==0 ) return 0; | |
193 while( (pWC = pScan->pWC)!=0 ){ | |
194 for(pTerm=pWC->a+k; k<pWC->nTerm; k++, pTerm++){ | |
195 if( pTerm->leftCursor==iCur | |
196 && pTerm->u.leftColumn==iColumn | |
197 && (iColumn!=XN_EXPR | |
198 || sqlite3ExprCompare(pTerm->pExpr->pLeft,pScan->pIdxExpr,iCur)==0) | |
199 && (pScan->iEquiv<=1 || !ExprHasProperty(pTerm->pExpr, EP_FromJoin)) | |
200 ){ | |
201 if( (pTerm->eOperator & WO_EQUIV)!=0 | |
202 && pScan->nEquiv<ArraySize(pScan->aiCur) | |
203 && (pX = sqlite3ExprSkipCollate(pTerm->pExpr->pRight))->op==TK_COLUMN | |
204 ){ | |
205 int j; | |
206 for(j=0; j<pScan->nEquiv; j++){ | |
207 if( pScan->aiCur[j]==pX->iTable | |
208 && pScan->aiColumn[j]==pX->iColumn ){ | |
209 break; | |
210 } | |
211 } | |
212 if( j==pScan->nEquiv ){ | |
213 pScan->aiCur[j] = pX->iTable; | |
214 pScan->aiColumn[j] = pX->iColumn; | |
215 pScan->nEquiv++; | |
216 } | |
217 } | |
218 if( (pTerm->eOperator & pScan->opMask)!=0 ){ | |
219 /* Verify the affinity and collating sequence match */ | |
220 if( pScan->zCollName && (pTerm->eOperator & WO_ISNULL)==0 ){ | |
221 CollSeq *pColl; | |
222 Parse *pParse = pWC->pWInfo->pParse; | |
223 pX = pTerm->pExpr; | |
224 if( !sqlite3IndexAffinityOk(pX, pScan->idxaff) ){ | |
225 continue; | |
226 } | |
227 assert(pX->pLeft); | |
228 pColl = sqlite3BinaryCompareCollSeq(pParse, | |
229 pX->pLeft, pX->pRight); | |
230 if( pColl==0 ) pColl = pParse->db->pDfltColl; | |
231 if( sqlite3StrICmp(pColl->zName, pScan->zCollName) ){ | |
232 continue; | |
233 } | |
234 } | |
235 if( (pTerm->eOperator & (WO_EQ|WO_IS))!=0 | |
236 && (pX = pTerm->pExpr->pRight)->op==TK_COLUMN | |
237 && pX->iTable==pScan->aiCur[0] | |
238 && pX->iColumn==pScan->aiColumn[0] | |
239 ){ | |
240 testcase( pTerm->eOperator & WO_IS ); | |
241 continue; | |
242 } | |
243 pScan->k = k+1; | |
244 return pTerm; | |
245 } | |
246 } | |
247 } | |
248 pScan->pWC = pScan->pWC->pOuter; | |
249 k = 0; | |
250 } | |
251 pScan->pWC = pScan->pOrigWC; | |
252 k = 0; | |
253 pScan->iEquiv++; | |
254 } | |
255 return 0; | |
256 } | |
257 | |
258 /* | |
259 ** Initialize a WHERE clause scanner object. Return a pointer to the | |
260 ** first match. Return NULL if there are no matches. | |
261 ** | |
262 ** The scanner will be searching the WHERE clause pWC. It will look | |
263 ** for terms of the form "X <op> <expr>" where X is column iColumn of table | |
264 ** iCur. The <op> must be one of the operators described by opMask. | |
265 ** | |
266 ** If the search is for X and the WHERE clause contains terms of the | |
267 ** form X=Y then this routine might also return terms of the form | |
268 ** "Y <op> <expr>". The number of levels of transitivity is limited, | |
269 ** but is enough to handle most commonly occurring SQL statements. | |
270 ** | |
271 ** If X is not the INTEGER PRIMARY KEY then X must be compatible with | |
272 ** index pIdx. | |
273 */ | |
274 static WhereTerm *whereScanInit( | |
275 WhereScan *pScan, /* The WhereScan object being initialized */ | |
276 WhereClause *pWC, /* The WHERE clause to be scanned */ | |
277 int iCur, /* Cursor to scan for */ | |
278 int iColumn, /* Column to scan for */ | |
279 u32 opMask, /* Operator(s) to scan for */ | |
280 Index *pIdx /* Must be compatible with this index */ | |
281 ){ | |
282 int j = 0; | |
283 | |
284 /* memset(pScan, 0, sizeof(*pScan)); */ | |
285 pScan->pOrigWC = pWC; | |
286 pScan->pWC = pWC; | |
287 pScan->pIdxExpr = 0; | |
288 if( pIdx ){ | |
289 j = iColumn; | |
290 iColumn = pIdx->aiColumn[j]; | |
291 if( iColumn==XN_EXPR ) pScan->pIdxExpr = pIdx->aColExpr->a[j].pExpr; | |
292 } | |
293 if( pIdx && iColumn>=0 ){ | |
294 pScan->idxaff = pIdx->pTable->aCol[iColumn].affinity; | |
295 pScan->zCollName = pIdx->azColl[j]; | |
296 }else{ | |
297 pScan->idxaff = 0; | |
298 pScan->zCollName = 0; | |
299 } | |
300 pScan->opMask = opMask; | |
301 pScan->k = 0; | |
302 pScan->aiCur[0] = iCur; | |
303 pScan->aiColumn[0] = iColumn; | |
304 pScan->nEquiv = 1; | |
305 pScan->iEquiv = 1; | |
306 return whereScanNext(pScan); | |
307 } | |
308 | |
309 /* | |
310 ** Search for a term in the WHERE clause that is of the form "X <op> <expr>" | |
311 ** where X is a reference to the iColumn of table iCur and <op> is one of | |
312 ** the WO_xx operator codes specified by the op parameter. | |
313 ** Return a pointer to the term. Return 0 if not found. | |
314 ** | |
315 ** If pIdx!=0 then search for terms matching the iColumn-th column of pIdx | |
316 ** rather than the iColumn-th column of table iCur. | |
317 ** | |
318 ** The term returned might by Y=<expr> if there is another constraint in | |
319 ** the WHERE clause that specifies that X=Y. Any such constraints will be | |
320 ** identified by the WO_EQUIV bit in the pTerm->eOperator field. The | |
321 ** aiCur[]/iaColumn[] arrays hold X and all its equivalents. There are 11 | |
322 ** slots in aiCur[]/aiColumn[] so that means we can look for X plus up to 10 | |
323 ** other equivalent values. Hence a search for X will return <expr> if X=A1 | |
324 ** and A1=A2 and A2=A3 and ... and A9=A10 and A10=<expr>. | |
325 ** | |
326 ** If there are multiple terms in the WHERE clause of the form "X <op> <expr>" | |
327 ** then try for the one with no dependencies on <expr> - in other words where | |
328 ** <expr> is a constant expression of some kind. Only return entries of | |
329 ** the form "X <op> Y" where Y is a column in another table if no terms of | |
330 ** the form "X <op> <const-expr>" exist. If no terms with a constant RHS | |
331 ** exist, try to return a term that does not use WO_EQUIV. | |
332 */ | |
333 WhereTerm *sqlite3WhereFindTerm( | |
334 WhereClause *pWC, /* The WHERE clause to be searched */ | |
335 int iCur, /* Cursor number of LHS */ | |
336 int iColumn, /* Column number of LHS */ | |
337 Bitmask notReady, /* RHS must not overlap with this mask */ | |
338 u32 op, /* Mask of WO_xx values describing operator */ | |
339 Index *pIdx /* Must be compatible with this index, if not NULL */ | |
340 ){ | |
341 WhereTerm *pResult = 0; | |
342 WhereTerm *p; | |
343 WhereScan scan; | |
344 | |
345 p = whereScanInit(&scan, pWC, iCur, iColumn, op, pIdx); | |
346 op &= WO_EQ|WO_IS; | |
347 while( p ){ | |
348 if( (p->prereqRight & notReady)==0 ){ | |
349 if( p->prereqRight==0 && (p->eOperator&op)!=0 ){ | |
350 testcase( p->eOperator & WO_IS ); | |
351 return p; | |
352 } | |
353 if( pResult==0 ) pResult = p; | |
354 } | |
355 p = whereScanNext(&scan); | |
356 } | |
357 return pResult; | |
358 } | |
359 | |
360 /* | |
361 ** This function searches pList for an entry that matches the iCol-th column | |
362 ** of index pIdx. | |
363 ** | |
364 ** If such an expression is found, its index in pList->a[] is returned. If | |
365 ** no expression is found, -1 is returned. | |
366 */ | |
367 static int findIndexCol( | |
368 Parse *pParse, /* Parse context */ | |
369 ExprList *pList, /* Expression list to search */ | |
370 int iBase, /* Cursor for table associated with pIdx */ | |
371 Index *pIdx, /* Index to match column of */ | |
372 int iCol /* Column of index to match */ | |
373 ){ | |
374 int i; | |
375 const char *zColl = pIdx->azColl[iCol]; | |
376 | |
377 for(i=0; i<pList->nExpr; i++){ | |
378 Expr *p = sqlite3ExprSkipCollate(pList->a[i].pExpr); | |
379 if( p->op==TK_COLUMN | |
380 && p->iColumn==pIdx->aiColumn[iCol] | |
381 && p->iTable==iBase | |
382 ){ | |
383 CollSeq *pColl = sqlite3ExprCollSeq(pParse, pList->a[i].pExpr); | |
384 if( pColl && 0==sqlite3StrICmp(pColl->zName, zColl) ){ | |
385 return i; | |
386 } | |
387 } | |
388 } | |
389 | |
390 return -1; | |
391 } | |
392 | |
393 /* | |
394 ** Return TRUE if the iCol-th column of index pIdx is NOT NULL | |
395 */ | |
396 static int indexColumnNotNull(Index *pIdx, int iCol){ | |
397 int j; | |
398 assert( pIdx!=0 ); | |
399 assert( iCol>=0 && iCol<pIdx->nColumn ); | |
400 j = pIdx->aiColumn[iCol]; | |
401 if( j>=0 ){ | |
402 return pIdx->pTable->aCol[j].notNull; | |
403 }else if( j==(-1) ){ | |
404 return 1; | |
405 }else{ | |
406 assert( j==(-2) ); | |
407 return 0; /* Assume an indexed expression can always yield a NULL */ | |
408 | |
409 } | |
410 } | |
411 | |
412 /* | |
413 ** Return true if the DISTINCT expression-list passed as the third argument | |
414 ** is redundant. | |
415 ** | |
416 ** A DISTINCT list is redundant if any subset of the columns in the | |
417 ** DISTINCT list are collectively unique and individually non-null. | |
418 */ | |
419 static int isDistinctRedundant( | |
420 Parse *pParse, /* Parsing context */ | |
421 SrcList *pTabList, /* The FROM clause */ | |
422 WhereClause *pWC, /* The WHERE clause */ | |
423 ExprList *pDistinct /* The result set that needs to be DISTINCT */ | |
424 ){ | |
425 Table *pTab; | |
426 Index *pIdx; | |
427 int i; | |
428 int iBase; | |
429 | |
430 /* If there is more than one table or sub-select in the FROM clause of | |
431 ** this query, then it will not be possible to show that the DISTINCT | |
432 ** clause is redundant. */ | |
433 if( pTabList->nSrc!=1 ) return 0; | |
434 iBase = pTabList->a[0].iCursor; | |
435 pTab = pTabList->a[0].pTab; | |
436 | |
437 /* If any of the expressions is an IPK column on table iBase, then return | |
438 ** true. Note: The (p->iTable==iBase) part of this test may be false if the | |
439 ** current SELECT is a correlated sub-query. | |
440 */ | |
441 for(i=0; i<pDistinct->nExpr; i++){ | |
442 Expr *p = sqlite3ExprSkipCollate(pDistinct->a[i].pExpr); | |
443 if( p->op==TK_COLUMN && p->iTable==iBase && p->iColumn<0 ) return 1; | |
444 } | |
445 | |
446 /* Loop through all indices on the table, checking each to see if it makes | |
447 ** the DISTINCT qualifier redundant. It does so if: | |
448 ** | |
449 ** 1. The index is itself UNIQUE, and | |
450 ** | |
451 ** 2. All of the columns in the index are either part of the pDistinct | |
452 ** list, or else the WHERE clause contains a term of the form "col=X", | |
453 ** where X is a constant value. The collation sequences of the | |
454 ** comparison and select-list expressions must match those of the index. | |
455 ** | |
456 ** 3. All of those index columns for which the WHERE clause does not | |
457 ** contain a "col=X" term are subject to a NOT NULL constraint. | |
458 */ | |
459 for(pIdx=pTab->pIndex; pIdx; pIdx=pIdx->pNext){ | |
460 if( !IsUniqueIndex(pIdx) ) continue; | |
461 for(i=0; i<pIdx->nKeyCol; i++){ | |
462 if( 0==sqlite3WhereFindTerm(pWC, iBase, i, ~(Bitmask)0, WO_EQ, pIdx) ){ | |
463 if( findIndexCol(pParse, pDistinct, iBase, pIdx, i)<0 ) break; | |
464 if( indexColumnNotNull(pIdx, i)==0 ) break; | |
465 } | |
466 } | |
467 if( i==pIdx->nKeyCol ){ | |
468 /* This index implies that the DISTINCT qualifier is redundant. */ | |
469 return 1; | |
470 } | |
471 } | |
472 | |
473 return 0; | |
474 } | |
475 | |
476 | |
477 /* | |
478 ** Estimate the logarithm of the input value to base 2. | |
479 */ | |
480 static LogEst estLog(LogEst N){ | |
481 return N<=10 ? 0 : sqlite3LogEst(N) - 33; | |
482 } | |
483 | |
484 /* | |
485 ** Convert OP_Column opcodes to OP_Copy in previously generated code. | |
486 ** | |
487 ** This routine runs over generated VDBE code and translates OP_Column | |
488 ** opcodes into OP_Copy when the table is being accessed via co-routine | |
489 ** instead of via table lookup. | |
490 ** | |
491 ** If the bIncrRowid parameter is 0, then any OP_Rowid instructions on | |
492 ** cursor iTabCur are transformed into OP_Null. Or, if bIncrRowid is non-zero, | |
493 ** then each OP_Rowid is transformed into an instruction to increment the | |
494 ** value stored in its output register. | |
495 */ | |
496 static void translateColumnToCopy( | |
497 Vdbe *v, /* The VDBE containing code to translate */ | |
498 int iStart, /* Translate from this opcode to the end */ | |
499 int iTabCur, /* OP_Column/OP_Rowid references to this table */ | |
500 int iRegister, /* The first column is in this register */ | |
501 int bIncrRowid /* If non-zero, transform OP_rowid to OP_AddImm(1) */ | |
502 ){ | |
503 VdbeOp *pOp = sqlite3VdbeGetOp(v, iStart); | |
504 int iEnd = sqlite3VdbeCurrentAddr(v); | |
505 for(; iStart<iEnd; iStart++, pOp++){ | |
506 if( pOp->p1!=iTabCur ) continue; | |
507 if( pOp->opcode==OP_Column ){ | |
508 pOp->opcode = OP_Copy; | |
509 pOp->p1 = pOp->p2 + iRegister; | |
510 pOp->p2 = pOp->p3; | |
511 pOp->p3 = 0; | |
512 }else if( pOp->opcode==OP_Rowid ){ | |
513 if( bIncrRowid ){ | |
514 /* Increment the value stored in the P2 operand of the OP_Rowid. */ | |
515 pOp->opcode = OP_AddImm; | |
516 pOp->p1 = pOp->p2; | |
517 pOp->p2 = 1; | |
518 }else{ | |
519 pOp->opcode = OP_Null; | |
520 pOp->p1 = 0; | |
521 pOp->p3 = 0; | |
522 } | |
523 } | |
524 } | |
525 } | |
526 | |
527 /* | |
528 ** Two routines for printing the content of an sqlite3_index_info | |
529 ** structure. Used for testing and debugging only. If neither | |
530 ** SQLITE_TEST or SQLITE_DEBUG are defined, then these routines | |
531 ** are no-ops. | |
532 */ | |
533 #if !defined(SQLITE_OMIT_VIRTUALTABLE) && defined(WHERETRACE_ENABLED) | |
534 static void TRACE_IDX_INPUTS(sqlite3_index_info *p){ | |
535 int i; | |
536 if( !sqlite3WhereTrace ) return; | |
537 for(i=0; i<p->nConstraint; i++){ | |
538 sqlite3DebugPrintf(" constraint[%d]: col=%d termid=%d op=%d usabled=%d\n", | |
539 i, | |
540 p->aConstraint[i].iColumn, | |
541 p->aConstraint[i].iTermOffset, | |
542 p->aConstraint[i].op, | |
543 p->aConstraint[i].usable); | |
544 } | |
545 for(i=0; i<p->nOrderBy; i++){ | |
546 sqlite3DebugPrintf(" orderby[%d]: col=%d desc=%d\n", | |
547 i, | |
548 p->aOrderBy[i].iColumn, | |
549 p->aOrderBy[i].desc); | |
550 } | |
551 } | |
552 static void TRACE_IDX_OUTPUTS(sqlite3_index_info *p){ | |
553 int i; | |
554 if( !sqlite3WhereTrace ) return; | |
555 for(i=0; i<p->nConstraint; i++){ | |
556 sqlite3DebugPrintf(" usage[%d]: argvIdx=%d omit=%d\n", | |
557 i, | |
558 p->aConstraintUsage[i].argvIndex, | |
559 p->aConstraintUsage[i].omit); | |
560 } | |
561 sqlite3DebugPrintf(" idxNum=%d\n", p->idxNum); | |
562 sqlite3DebugPrintf(" idxStr=%s\n", p->idxStr); | |
563 sqlite3DebugPrintf(" orderByConsumed=%d\n", p->orderByConsumed); | |
564 sqlite3DebugPrintf(" estimatedCost=%g\n", p->estimatedCost); | |
565 sqlite3DebugPrintf(" estimatedRows=%lld\n", p->estimatedRows); | |
566 } | |
567 #else | |
568 #define TRACE_IDX_INPUTS(A) | |
569 #define TRACE_IDX_OUTPUTS(A) | |
570 #endif | |
571 | |
572 #ifndef SQLITE_OMIT_AUTOMATIC_INDEX | |
573 /* | |
574 ** Return TRUE if the WHERE clause term pTerm is of a form where it | |
575 ** could be used with an index to access pSrc, assuming an appropriate | |
576 ** index existed. | |
577 */ | |
578 static int termCanDriveIndex( | |
579 WhereTerm *pTerm, /* WHERE clause term to check */ | |
580 struct SrcList_item *pSrc, /* Table we are trying to access */ | |
581 Bitmask notReady /* Tables in outer loops of the join */ | |
582 ){ | |
583 char aff; | |
584 if( pTerm->leftCursor!=pSrc->iCursor ) return 0; | |
585 if( (pTerm->eOperator & (WO_EQ|WO_IS))==0 ) return 0; | |
586 if( (pTerm->prereqRight & notReady)!=0 ) return 0; | |
587 if( pTerm->u.leftColumn<0 ) return 0; | |
588 aff = pSrc->pTab->aCol[pTerm->u.leftColumn].affinity; | |
589 if( !sqlite3IndexAffinityOk(pTerm->pExpr, aff) ) return 0; | |
590 testcase( pTerm->pExpr->op==TK_IS ); | |
591 return 1; | |
592 } | |
593 #endif | |
594 | |
595 | |
596 #ifndef SQLITE_OMIT_AUTOMATIC_INDEX | |
597 /* | |
598 ** Generate code to construct the Index object for an automatic index | |
599 ** and to set up the WhereLevel object pLevel so that the code generator | |
600 ** makes use of the automatic index. | |
601 */ | |
602 static void constructAutomaticIndex( | |
603 Parse *pParse, /* The parsing context */ | |
604 WhereClause *pWC, /* The WHERE clause */ | |
605 struct SrcList_item *pSrc, /* The FROM clause term to get the next index */ | |
606 Bitmask notReady, /* Mask of cursors that are not available */ | |
607 WhereLevel *pLevel /* Write new index here */ | |
608 ){ | |
609 int nKeyCol; /* Number of columns in the constructed index */ | |
610 WhereTerm *pTerm; /* A single term of the WHERE clause */ | |
611 WhereTerm *pWCEnd; /* End of pWC->a[] */ | |
612 Index *pIdx; /* Object describing the transient index */ | |
613 Vdbe *v; /* Prepared statement under construction */ | |
614 int addrInit; /* Address of the initialization bypass jump */ | |
615 Table *pTable; /* The table being indexed */ | |
616 int addrTop; /* Top of the index fill loop */ | |
617 int regRecord; /* Register holding an index record */ | |
618 int n; /* Column counter */ | |
619 int i; /* Loop counter */ | |
620 int mxBitCol; /* Maximum column in pSrc->colUsed */ | |
621 CollSeq *pColl; /* Collating sequence to on a column */ | |
622 WhereLoop *pLoop; /* The Loop object */ | |
623 char *zNotUsed; /* Extra space on the end of pIdx */ | |
624 Bitmask idxCols; /* Bitmap of columns used for indexing */ | |
625 Bitmask extraCols; /* Bitmap of additional columns */ | |
626 u8 sentWarning = 0; /* True if a warnning has been issued */ | |
627 Expr *pPartial = 0; /* Partial Index Expression */ | |
628 int iContinue = 0; /* Jump here to skip excluded rows */ | |
629 struct SrcList_item *pTabItem; /* FROM clause term being indexed */ | |
630 int addrCounter = 0; /* Address where integer counter is initialized */ | |
631 int regBase; /* Array of registers where record is assembled */ | |
632 | |
633 /* Generate code to skip over the creation and initialization of the | |
634 ** transient index on 2nd and subsequent iterations of the loop. */ | |
635 v = pParse->pVdbe; | |
636 assert( v!=0 ); | |
637 addrInit = sqlite3CodeOnce(pParse); VdbeCoverage(v); | |
638 | |
639 /* Count the number of columns that will be added to the index | |
640 ** and used to match WHERE clause constraints */ | |
641 nKeyCol = 0; | |
642 pTable = pSrc->pTab; | |
643 pWCEnd = &pWC->a[pWC->nTerm]; | |
644 pLoop = pLevel->pWLoop; | |
645 idxCols = 0; | |
646 for(pTerm=pWC->a; pTerm<pWCEnd; pTerm++){ | |
647 Expr *pExpr = pTerm->pExpr; | |
648 assert( !ExprHasProperty(pExpr, EP_FromJoin) /* prereq always non-zero */ | |
649 || pExpr->iRightJoinTable!=pSrc->iCursor /* for the right-hand */ | |
650 || pLoop->prereq!=0 ); /* table of a LEFT JOIN */ | |
651 if( pLoop->prereq==0 | |
652 && (pTerm->wtFlags & TERM_VIRTUAL)==0 | |
653 && !ExprHasProperty(pExpr, EP_FromJoin) | |
654 && sqlite3ExprIsTableConstant(pExpr, pSrc->iCursor) ){ | |
655 pPartial = sqlite3ExprAnd(pParse->db, pPartial, | |
656 sqlite3ExprDup(pParse->db, pExpr, 0)); | |
657 } | |
658 if( termCanDriveIndex(pTerm, pSrc, notReady) ){ | |
659 int iCol = pTerm->u.leftColumn; | |
660 Bitmask cMask = iCol>=BMS ? MASKBIT(BMS-1) : MASKBIT(iCol); | |
661 testcase( iCol==BMS ); | |
662 testcase( iCol==BMS-1 ); | |
663 if( !sentWarning ){ | |
664 sqlite3_log(SQLITE_WARNING_AUTOINDEX, | |
665 "automatic index on %s(%s)", pTable->zName, | |
666 pTable->aCol[iCol].zName); | |
667 sentWarning = 1; | |
668 } | |
669 if( (idxCols & cMask)==0 ){ | |
670 if( whereLoopResize(pParse->db, pLoop, nKeyCol+1) ){ | |
671 goto end_auto_index_create; | |
672 } | |
673 pLoop->aLTerm[nKeyCol++] = pTerm; | |
674 idxCols |= cMask; | |
675 } | |
676 } | |
677 } | |
678 assert( nKeyCol>0 ); | |
679 pLoop->u.btree.nEq = pLoop->nLTerm = nKeyCol; | |
680 pLoop->wsFlags = WHERE_COLUMN_EQ | WHERE_IDX_ONLY | WHERE_INDEXED | |
681 | WHERE_AUTO_INDEX; | |
682 | |
683 /* Count the number of additional columns needed to create a | |
684 ** covering index. A "covering index" is an index that contains all | |
685 ** columns that are needed by the query. With a covering index, the | |
686 ** original table never needs to be accessed. Automatic indices must | |
687 ** be a covering index because the index will not be updated if the | |
688 ** original table changes and the index and table cannot both be used | |
689 ** if they go out of sync. | |
690 */ | |
691 extraCols = pSrc->colUsed & (~idxCols | MASKBIT(BMS-1)); | |
692 mxBitCol = MIN(BMS-1,pTable->nCol); | |
693 testcase( pTable->nCol==BMS-1 ); | |
694 testcase( pTable->nCol==BMS-2 ); | |
695 for(i=0; i<mxBitCol; i++){ | |
696 if( extraCols & MASKBIT(i) ) nKeyCol++; | |
697 } | |
698 if( pSrc->colUsed & MASKBIT(BMS-1) ){ | |
699 nKeyCol += pTable->nCol - BMS + 1; | |
700 } | |
701 | |
702 /* Construct the Index object to describe this index */ | |
703 pIdx = sqlite3AllocateIndexObject(pParse->db, nKeyCol+1, 0, &zNotUsed); | |
704 if( pIdx==0 ) goto end_auto_index_create; | |
705 pLoop->u.btree.pIndex = pIdx; | |
706 pIdx->zName = "auto-index"; | |
707 pIdx->pTable = pTable; | |
708 n = 0; | |
709 idxCols = 0; | |
710 for(pTerm=pWC->a; pTerm<pWCEnd; pTerm++){ | |
711 if( termCanDriveIndex(pTerm, pSrc, notReady) ){ | |
712 int iCol = pTerm->u.leftColumn; | |
713 Bitmask cMask = iCol>=BMS ? MASKBIT(BMS-1) : MASKBIT(iCol); | |
714 testcase( iCol==BMS-1 ); | |
715 testcase( iCol==BMS ); | |
716 if( (idxCols & cMask)==0 ){ | |
717 Expr *pX = pTerm->pExpr; | |
718 idxCols |= cMask; | |
719 pIdx->aiColumn[n] = pTerm->u.leftColumn; | |
720 pColl = sqlite3BinaryCompareCollSeq(pParse, pX->pLeft, pX->pRight); | |
721 pIdx->azColl[n] = pColl ? pColl->zName : sqlite3StrBINARY; | |
722 n++; | |
723 } | |
724 } | |
725 } | |
726 assert( (u32)n==pLoop->u.btree.nEq ); | |
727 | |
728 /* Add additional columns needed to make the automatic index into | |
729 ** a covering index */ | |
730 for(i=0; i<mxBitCol; i++){ | |
731 if( extraCols & MASKBIT(i) ){ | |
732 pIdx->aiColumn[n] = i; | |
733 pIdx->azColl[n] = sqlite3StrBINARY; | |
734 n++; | |
735 } | |
736 } | |
737 if( pSrc->colUsed & MASKBIT(BMS-1) ){ | |
738 for(i=BMS-1; i<pTable->nCol; i++){ | |
739 pIdx->aiColumn[n] = i; | |
740 pIdx->azColl[n] = sqlite3StrBINARY; | |
741 n++; | |
742 } | |
743 } | |
744 assert( n==nKeyCol ); | |
745 pIdx->aiColumn[n] = XN_ROWID; | |
746 pIdx->azColl[n] = sqlite3StrBINARY; | |
747 | |
748 /* Create the automatic index */ | |
749 assert( pLevel->iIdxCur>=0 ); | |
750 pLevel->iIdxCur = pParse->nTab++; | |
751 sqlite3VdbeAddOp2(v, OP_OpenAutoindex, pLevel->iIdxCur, nKeyCol+1); | |
752 sqlite3VdbeSetP4KeyInfo(pParse, pIdx); | |
753 VdbeComment((v, "for %s", pTable->zName)); | |
754 | |
755 /* Fill the automatic index with content */ | |
756 sqlite3ExprCachePush(pParse); | |
757 pTabItem = &pWC->pWInfo->pTabList->a[pLevel->iFrom]; | |
758 if( pTabItem->fg.viaCoroutine ){ | |
759 int regYield = pTabItem->regReturn; | |
760 addrCounter = sqlite3VdbeAddOp2(v, OP_Integer, 0, 0); | |
761 sqlite3VdbeAddOp3(v, OP_InitCoroutine, regYield, 0, pTabItem->addrFillSub); | |
762 addrTop = sqlite3VdbeAddOp1(v, OP_Yield, regYield); | |
763 VdbeCoverage(v); | |
764 VdbeComment((v, "next row of \"%s\"", pTabItem->pTab->zName)); | |
765 }else{ | |
766 addrTop = sqlite3VdbeAddOp1(v, OP_Rewind, pLevel->iTabCur); VdbeCoverage(v); | |
767 } | |
768 if( pPartial ){ | |
769 iContinue = sqlite3VdbeMakeLabel(v); | |
770 sqlite3ExprIfFalse(pParse, pPartial, iContinue, SQLITE_JUMPIFNULL); | |
771 pLoop->wsFlags |= WHERE_PARTIALIDX; | |
772 } | |
773 regRecord = sqlite3GetTempReg(pParse); | |
774 regBase = sqlite3GenerateIndexKey( | |
775 pParse, pIdx, pLevel->iTabCur, regRecord, 0, 0, 0, 0 | |
776 ); | |
777 sqlite3VdbeAddOp2(v, OP_IdxInsert, pLevel->iIdxCur, regRecord); | |
778 sqlite3VdbeChangeP5(v, OPFLAG_USESEEKRESULT); | |
779 if( pPartial ) sqlite3VdbeResolveLabel(v, iContinue); | |
780 if( pTabItem->fg.viaCoroutine ){ | |
781 sqlite3VdbeChangeP2(v, addrCounter, regBase+n); | |
782 translateColumnToCopy(v, addrTop, pLevel->iTabCur, pTabItem->regResult, 1); | |
783 sqlite3VdbeGoto(v, addrTop); | |
784 pTabItem->fg.viaCoroutine = 0; | |
785 }else{ | |
786 sqlite3VdbeAddOp2(v, OP_Next, pLevel->iTabCur, addrTop+1); VdbeCoverage(v); | |
787 } | |
788 sqlite3VdbeChangeP5(v, SQLITE_STMTSTATUS_AUTOINDEX); | |
789 sqlite3VdbeJumpHere(v, addrTop); | |
790 sqlite3ReleaseTempReg(pParse, regRecord); | |
791 sqlite3ExprCachePop(pParse); | |
792 | |
793 /* Jump here when skipping the initialization */ | |
794 sqlite3VdbeJumpHere(v, addrInit); | |
795 | |
796 end_auto_index_create: | |
797 sqlite3ExprDelete(pParse->db, pPartial); | |
798 } | |
799 #endif /* SQLITE_OMIT_AUTOMATIC_INDEX */ | |
800 | |
801 #ifndef SQLITE_OMIT_VIRTUALTABLE | |
802 /* | |
803 ** Allocate and populate an sqlite3_index_info structure. It is the | |
804 ** responsibility of the caller to eventually release the structure | |
805 ** by passing the pointer returned by this function to sqlite3_free(). | |
806 */ | |
807 static sqlite3_index_info *allocateIndexInfo( | |
808 Parse *pParse, | |
809 WhereClause *pWC, | |
810 Bitmask mUnusable, /* Ignore terms with these prereqs */ | |
811 struct SrcList_item *pSrc, | |
812 ExprList *pOrderBy | |
813 ){ | |
814 int i, j; | |
815 int nTerm; | |
816 struct sqlite3_index_constraint *pIdxCons; | |
817 struct sqlite3_index_orderby *pIdxOrderBy; | |
818 struct sqlite3_index_constraint_usage *pUsage; | |
819 WhereTerm *pTerm; | |
820 int nOrderBy; | |
821 sqlite3_index_info *pIdxInfo; | |
822 | |
823 /* Count the number of possible WHERE clause constraints referring | |
824 ** to this virtual table */ | |
825 for(i=nTerm=0, pTerm=pWC->a; i<pWC->nTerm; i++, pTerm++){ | |
826 if( pTerm->leftCursor != pSrc->iCursor ) continue; | |
827 if( pTerm->prereqRight & mUnusable ) continue; | |
828 assert( IsPowerOfTwo(pTerm->eOperator & ~WO_EQUIV) ); | |
829 testcase( pTerm->eOperator & WO_IN ); | |
830 testcase( pTerm->eOperator & WO_ISNULL ); | |
831 testcase( pTerm->eOperator & WO_IS ); | |
832 testcase( pTerm->eOperator & WO_ALL ); | |
833 if( (pTerm->eOperator & ~(WO_ISNULL|WO_EQUIV|WO_IS))==0 ) continue; | |
834 if( pTerm->wtFlags & TERM_VNULL ) continue; | |
835 assert( pTerm->u.leftColumn>=(-1) ); | |
836 nTerm++; | |
837 } | |
838 | |
839 /* If the ORDER BY clause contains only columns in the current | |
840 ** virtual table then allocate space for the aOrderBy part of | |
841 ** the sqlite3_index_info structure. | |
842 */ | |
843 nOrderBy = 0; | |
844 if( pOrderBy ){ | |
845 int n = pOrderBy->nExpr; | |
846 for(i=0; i<n; i++){ | |
847 Expr *pExpr = pOrderBy->a[i].pExpr; | |
848 if( pExpr->op!=TK_COLUMN || pExpr->iTable!=pSrc->iCursor ) break; | |
849 } | |
850 if( i==n){ | |
851 nOrderBy = n; | |
852 } | |
853 } | |
854 | |
855 /* Allocate the sqlite3_index_info structure | |
856 */ | |
857 pIdxInfo = sqlite3DbMallocZero(pParse->db, sizeof(*pIdxInfo) | |
858 + (sizeof(*pIdxCons) + sizeof(*pUsage))*nTerm | |
859 + sizeof(*pIdxOrderBy)*nOrderBy ); | |
860 if( pIdxInfo==0 ){ | |
861 sqlite3ErrorMsg(pParse, "out of memory"); | |
862 return 0; | |
863 } | |
864 | |
865 /* Initialize the structure. The sqlite3_index_info structure contains | |
866 ** many fields that are declared "const" to prevent xBestIndex from | |
867 ** changing them. We have to do some funky casting in order to | |
868 ** initialize those fields. | |
869 */ | |
870 pIdxCons = (struct sqlite3_index_constraint*)&pIdxInfo[1]; | |
871 pIdxOrderBy = (struct sqlite3_index_orderby*)&pIdxCons[nTerm]; | |
872 pUsage = (struct sqlite3_index_constraint_usage*)&pIdxOrderBy[nOrderBy]; | |
873 *(int*)&pIdxInfo->nConstraint = nTerm; | |
874 *(int*)&pIdxInfo->nOrderBy = nOrderBy; | |
875 *(struct sqlite3_index_constraint**)&pIdxInfo->aConstraint = pIdxCons; | |
876 *(struct sqlite3_index_orderby**)&pIdxInfo->aOrderBy = pIdxOrderBy; | |
877 *(struct sqlite3_index_constraint_usage**)&pIdxInfo->aConstraintUsage = | |
878 pUsage; | |
879 | |
880 for(i=j=0, pTerm=pWC->a; i<pWC->nTerm; i++, pTerm++){ | |
881 u8 op; | |
882 if( pTerm->leftCursor != pSrc->iCursor ) continue; | |
883 if( pTerm->prereqRight & mUnusable ) continue; | |
884 assert( IsPowerOfTwo(pTerm->eOperator & ~WO_EQUIV) ); | |
885 testcase( pTerm->eOperator & WO_IN ); | |
886 testcase( pTerm->eOperator & WO_IS ); | |
887 testcase( pTerm->eOperator & WO_ISNULL ); | |
888 testcase( pTerm->eOperator & WO_ALL ); | |
889 if( (pTerm->eOperator & ~(WO_ISNULL|WO_EQUIV|WO_IS))==0 ) continue; | |
890 if( pTerm->wtFlags & TERM_VNULL ) continue; | |
891 assert( pTerm->u.leftColumn>=(-1) ); | |
892 pIdxCons[j].iColumn = pTerm->u.leftColumn; | |
893 pIdxCons[j].iTermOffset = i; | |
894 op = (u8)pTerm->eOperator & WO_ALL; | |
895 if( op==WO_IN ) op = WO_EQ; | |
896 if( op==WO_MATCH ){ | |
897 op = pTerm->eMatchOp; | |
898 } | |
899 pIdxCons[j].op = op; | |
900 /* The direct assignment in the previous line is possible only because | |
901 ** the WO_ and SQLITE_INDEX_CONSTRAINT_ codes are identical. The | |
902 ** following asserts verify this fact. */ | |
903 assert( WO_EQ==SQLITE_INDEX_CONSTRAINT_EQ ); | |
904 assert( WO_LT==SQLITE_INDEX_CONSTRAINT_LT ); | |
905 assert( WO_LE==SQLITE_INDEX_CONSTRAINT_LE ); | |
906 assert( WO_GT==SQLITE_INDEX_CONSTRAINT_GT ); | |
907 assert( WO_GE==SQLITE_INDEX_CONSTRAINT_GE ); | |
908 assert( WO_MATCH==SQLITE_INDEX_CONSTRAINT_MATCH ); | |
909 assert( pTerm->eOperator & (WO_IN|WO_EQ|WO_LT|WO_LE|WO_GT|WO_GE|WO_MATCH) ); | |
910 j++; | |
911 } | |
912 for(i=0; i<nOrderBy; i++){ | |
913 Expr *pExpr = pOrderBy->a[i].pExpr; | |
914 pIdxOrderBy[i].iColumn = pExpr->iColumn; | |
915 pIdxOrderBy[i].desc = pOrderBy->a[i].sortOrder; | |
916 } | |
917 | |
918 return pIdxInfo; | |
919 } | |
920 | |
921 /* | |
922 ** The table object reference passed as the second argument to this function | |
923 ** must represent a virtual table. This function invokes the xBestIndex() | |
924 ** method of the virtual table with the sqlite3_index_info object that | |
925 ** comes in as the 3rd argument to this function. | |
926 ** | |
927 ** If an error occurs, pParse is populated with an error message and a | |
928 ** non-zero value is returned. Otherwise, 0 is returned and the output | |
929 ** part of the sqlite3_index_info structure is left populated. | |
930 ** | |
931 ** Whether or not an error is returned, it is the responsibility of the | |
932 ** caller to eventually free p->idxStr if p->needToFreeIdxStr indicates | |
933 ** that this is required. | |
934 */ | |
935 static int vtabBestIndex(Parse *pParse, Table *pTab, sqlite3_index_info *p){ | |
936 sqlite3_vtab *pVtab = sqlite3GetVTable(pParse->db, pTab)->pVtab; | |
937 int i; | |
938 int rc; | |
939 | |
940 TRACE_IDX_INPUTS(p); | |
941 rc = pVtab->pModule->xBestIndex(pVtab, p); | |
942 TRACE_IDX_OUTPUTS(p); | |
943 | |
944 if( rc!=SQLITE_OK ){ | |
945 if( rc==SQLITE_NOMEM ){ | |
946 pParse->db->mallocFailed = 1; | |
947 }else if( !pVtab->zErrMsg ){ | |
948 sqlite3ErrorMsg(pParse, "%s", sqlite3ErrStr(rc)); | |
949 }else{ | |
950 sqlite3ErrorMsg(pParse, "%s", pVtab->zErrMsg); | |
951 } | |
952 } | |
953 sqlite3_free(pVtab->zErrMsg); | |
954 pVtab->zErrMsg = 0; | |
955 | |
956 for(i=0; i<p->nConstraint; i++){ | |
957 if( !p->aConstraint[i].usable && p->aConstraintUsage[i].argvIndex>0 ){ | |
958 sqlite3ErrorMsg(pParse, | |
959 "table %s: xBestIndex returned an invalid plan", pTab->zName); | |
960 } | |
961 } | |
962 | |
963 return pParse->nErr; | |
964 } | |
965 #endif /* !defined(SQLITE_OMIT_VIRTUALTABLE) */ | |
966 | |
967 #ifdef SQLITE_ENABLE_STAT3_OR_STAT4 | |
968 /* | |
969 ** Estimate the location of a particular key among all keys in an | |
970 ** index. Store the results in aStat as follows: | |
971 ** | |
972 ** aStat[0] Est. number of rows less than pRec | |
973 ** aStat[1] Est. number of rows equal to pRec | |
974 ** | |
975 ** Return the index of the sample that is the smallest sample that | |
976 ** is greater than or equal to pRec. Note that this index is not an index | |
977 ** into the aSample[] array - it is an index into a virtual set of samples | |
978 ** based on the contents of aSample[] and the number of fields in record | |
979 ** pRec. | |
980 */ | |
981 static int whereKeyStats( | |
982 Parse *pParse, /* Database connection */ | |
983 Index *pIdx, /* Index to consider domain of */ | |
984 UnpackedRecord *pRec, /* Vector of values to consider */ | |
985 int roundUp, /* Round up if true. Round down if false */ | |
986 tRowcnt *aStat /* OUT: stats written here */ | |
987 ){ | |
988 IndexSample *aSample = pIdx->aSample; | |
989 int iCol; /* Index of required stats in anEq[] etc. */ | |
990 int i; /* Index of first sample >= pRec */ | |
991 int iSample; /* Smallest sample larger than or equal to pRec */ | |
992 int iMin = 0; /* Smallest sample not yet tested */ | |
993 int iTest; /* Next sample to test */ | |
994 int res; /* Result of comparison operation */ | |
995 int nField; /* Number of fields in pRec */ | |
996 tRowcnt iLower = 0; /* anLt[] + anEq[] of largest sample pRec is > */ | |
997 | |
998 #ifndef SQLITE_DEBUG | |
999 UNUSED_PARAMETER( pParse ); | |
1000 #endif | |
1001 assert( pRec!=0 ); | |
1002 assert( pIdx->nSample>0 ); | |
1003 assert( pRec->nField>0 && pRec->nField<=pIdx->nSampleCol ); | |
1004 | |
1005 /* Do a binary search to find the first sample greater than or equal | |
1006 ** to pRec. If pRec contains a single field, the set of samples to search | |
1007 ** is simply the aSample[] array. If the samples in aSample[] contain more | |
1008 ** than one fields, all fields following the first are ignored. | |
1009 ** | |
1010 ** If pRec contains N fields, where N is more than one, then as well as the | |
1011 ** samples in aSample[] (truncated to N fields), the search also has to | |
1012 ** consider prefixes of those samples. For example, if the set of samples | |
1013 ** in aSample is: | |
1014 ** | |
1015 ** aSample[0] = (a, 5) | |
1016 ** aSample[1] = (a, 10) | |
1017 ** aSample[2] = (b, 5) | |
1018 ** aSample[3] = (c, 100) | |
1019 ** aSample[4] = (c, 105) | |
1020 ** | |
1021 ** Then the search space should ideally be the samples above and the | |
1022 ** unique prefixes [a], [b] and [c]. But since that is hard to organize, | |
1023 ** the code actually searches this set: | |
1024 ** | |
1025 ** 0: (a) | |
1026 ** 1: (a, 5) | |
1027 ** 2: (a, 10) | |
1028 ** 3: (a, 10) | |
1029 ** 4: (b) | |
1030 ** 5: (b, 5) | |
1031 ** 6: (c) | |
1032 ** 7: (c, 100) | |
1033 ** 8: (c, 105) | |
1034 ** 9: (c, 105) | |
1035 ** | |
1036 ** For each sample in the aSample[] array, N samples are present in the | |
1037 ** effective sample array. In the above, samples 0 and 1 are based on | |
1038 ** sample aSample[0]. Samples 2 and 3 on aSample[1] etc. | |
1039 ** | |
1040 ** Often, sample i of each block of N effective samples has (i+1) fields. | |
1041 ** Except, each sample may be extended to ensure that it is greater than or | |
1042 ** equal to the previous sample in the array. For example, in the above, | |
1043 ** sample 2 is the first sample of a block of N samples, so at first it | |
1044 ** appears that it should be 1 field in size. However, that would make it | |
1045 ** smaller than sample 1, so the binary search would not work. As a result, | |
1046 ** it is extended to two fields. The duplicates that this creates do not | |
1047 ** cause any problems. | |
1048 */ | |
1049 nField = pRec->nField; | |
1050 iCol = 0; | |
1051 iSample = pIdx->nSample * nField; | |
1052 do{ | |
1053 int iSamp; /* Index in aSample[] of test sample */ | |
1054 int n; /* Number of fields in test sample */ | |
1055 | |
1056 iTest = (iMin+iSample)/2; | |
1057 iSamp = iTest / nField; | |
1058 if( iSamp>0 ){ | |
1059 /* The proposed effective sample is a prefix of sample aSample[iSamp]. | |
1060 ** Specifically, the shortest prefix of at least (1 + iTest%nField) | |
1061 ** fields that is greater than the previous effective sample. */ | |
1062 for(n=(iTest % nField) + 1; n<nField; n++){ | |
1063 if( aSample[iSamp-1].anLt[n-1]!=aSample[iSamp].anLt[n-1] ) break; | |
1064 } | |
1065 }else{ | |
1066 n = iTest + 1; | |
1067 } | |
1068 | |
1069 pRec->nField = n; | |
1070 res = sqlite3VdbeRecordCompare(aSample[iSamp].n, aSample[iSamp].p, pRec); | |
1071 if( res<0 ){ | |
1072 iLower = aSample[iSamp].anLt[n-1] + aSample[iSamp].anEq[n-1]; | |
1073 iMin = iTest+1; | |
1074 }else if( res==0 && n<nField ){ | |
1075 iLower = aSample[iSamp].anLt[n-1]; | |
1076 iMin = iTest+1; | |
1077 res = -1; | |
1078 }else{ | |
1079 iSample = iTest; | |
1080 iCol = n-1; | |
1081 } | |
1082 }while( res && iMin<iSample ); | |
1083 i = iSample / nField; | |
1084 | |
1085 #ifdef SQLITE_DEBUG | |
1086 /* The following assert statements check that the binary search code | |
1087 ** above found the right answer. This block serves no purpose other | |
1088 ** than to invoke the asserts. */ | |
1089 if( pParse->db->mallocFailed==0 ){ | |
1090 if( res==0 ){ | |
1091 /* If (res==0) is true, then pRec must be equal to sample i. */ | |
1092 assert( i<pIdx->nSample ); | |
1093 assert( iCol==nField-1 ); | |
1094 pRec->nField = nField; | |
1095 assert( 0==sqlite3VdbeRecordCompare(aSample[i].n, aSample[i].p, pRec) | |
1096 || pParse->db->mallocFailed | |
1097 ); | |
1098 }else{ | |
1099 /* Unless i==pIdx->nSample, indicating that pRec is larger than | |
1100 ** all samples in the aSample[] array, pRec must be smaller than the | |
1101 ** (iCol+1) field prefix of sample i. */ | |
1102 assert( i<=pIdx->nSample && i>=0 ); | |
1103 pRec->nField = iCol+1; | |
1104 assert( i==pIdx->nSample | |
1105 || sqlite3VdbeRecordCompare(aSample[i].n, aSample[i].p, pRec)>0 | |
1106 || pParse->db->mallocFailed ); | |
1107 | |
1108 /* if i==0 and iCol==0, then record pRec is smaller than all samples | |
1109 ** in the aSample[] array. Otherwise, if (iCol>0) then pRec must | |
1110 ** be greater than or equal to the (iCol) field prefix of sample i. | |
1111 ** If (i>0), then pRec must also be greater than sample (i-1). */ | |
1112 if( iCol>0 ){ | |
1113 pRec->nField = iCol; | |
1114 assert( sqlite3VdbeRecordCompare(aSample[i].n, aSample[i].p, pRec)<=0 | |
1115 || pParse->db->mallocFailed ); | |
1116 } | |
1117 if( i>0 ){ | |
1118 pRec->nField = nField; | |
1119 assert( sqlite3VdbeRecordCompare(aSample[i-1].n, aSample[i-1].p, pRec)<0 | |
1120 || pParse->db->mallocFailed ); | |
1121 } | |
1122 } | |
1123 } | |
1124 #endif /* ifdef SQLITE_DEBUG */ | |
1125 | |
1126 if( res==0 ){ | |
1127 /* Record pRec is equal to sample i */ | |
1128 assert( iCol==nField-1 ); | |
1129 aStat[0] = aSample[i].anLt[iCol]; | |
1130 aStat[1] = aSample[i].anEq[iCol]; | |
1131 }else{ | |
1132 /* At this point, the (iCol+1) field prefix of aSample[i] is the first | |
1133 ** sample that is greater than pRec. Or, if i==pIdx->nSample then pRec | |
1134 ** is larger than all samples in the array. */ | |
1135 tRowcnt iUpper, iGap; | |
1136 if( i>=pIdx->nSample ){ | |
1137 iUpper = sqlite3LogEstToInt(pIdx->aiRowLogEst[0]); | |
1138 }else{ | |
1139 iUpper = aSample[i].anLt[iCol]; | |
1140 } | |
1141 | |
1142 if( iLower>=iUpper ){ | |
1143 iGap = 0; | |
1144 }else{ | |
1145 iGap = iUpper - iLower; | |
1146 } | |
1147 if( roundUp ){ | |
1148 iGap = (iGap*2)/3; | |
1149 }else{ | |
1150 iGap = iGap/3; | |
1151 } | |
1152 aStat[0] = iLower + iGap; | |
1153 aStat[1] = pIdx->aAvgEq[iCol]; | |
1154 } | |
1155 | |
1156 /* Restore the pRec->nField value before returning. */ | |
1157 pRec->nField = nField; | |
1158 return i; | |
1159 } | |
1160 #endif /* SQLITE_ENABLE_STAT3_OR_STAT4 */ | |
1161 | |
1162 /* | |
1163 ** If it is not NULL, pTerm is a term that provides an upper or lower | |
1164 ** bound on a range scan. Without considering pTerm, it is estimated | |
1165 ** that the scan will visit nNew rows. This function returns the number | |
1166 ** estimated to be visited after taking pTerm into account. | |
1167 ** | |
1168 ** If the user explicitly specified a likelihood() value for this term, | |
1169 ** then the return value is the likelihood multiplied by the number of | |
1170 ** input rows. Otherwise, this function assumes that an "IS NOT NULL" term | |
1171 ** has a likelihood of 0.50, and any other term a likelihood of 0.25. | |
1172 */ | |
1173 static LogEst whereRangeAdjust(WhereTerm *pTerm, LogEst nNew){ | |
1174 LogEst nRet = nNew; | |
1175 if( pTerm ){ | |
1176 if( pTerm->truthProb<=0 ){ | |
1177 nRet += pTerm->truthProb; | |
1178 }else if( (pTerm->wtFlags & TERM_VNULL)==0 ){ | |
1179 nRet -= 20; assert( 20==sqlite3LogEst(4) ); | |
1180 } | |
1181 } | |
1182 return nRet; | |
1183 } | |
1184 | |
1185 | |
1186 #ifdef SQLITE_ENABLE_STAT3_OR_STAT4 | |
1187 /* | |
1188 ** Return the affinity for a single column of an index. | |
1189 */ | |
1190 static char sqlite3IndexColumnAffinity(sqlite3 *db, Index *pIdx, int iCol){ | |
1191 assert( iCol>=0 && iCol<pIdx->nColumn ); | |
1192 if( !pIdx->zColAff ){ | |
1193 if( sqlite3IndexAffinityStr(db, pIdx)==0 ) return SQLITE_AFF_BLOB; | |
1194 } | |
1195 return pIdx->zColAff[iCol]; | |
1196 } | |
1197 #endif | |
1198 | |
1199 | |
1200 #ifdef SQLITE_ENABLE_STAT3_OR_STAT4 | |
1201 /* | |
1202 ** This function is called to estimate the number of rows visited by a | |
1203 ** range-scan on a skip-scan index. For example: | |
1204 ** | |
1205 ** CREATE INDEX i1 ON t1(a, b, c); | |
1206 ** SELECT * FROM t1 WHERE a=? AND c BETWEEN ? AND ?; | |
1207 ** | |
1208 ** Value pLoop->nOut is currently set to the estimated number of rows | |
1209 ** visited for scanning (a=? AND b=?). This function reduces that estimate | |
1210 ** by some factor to account for the (c BETWEEN ? AND ?) expression based | |
1211 ** on the stat4 data for the index. this scan will be peformed multiple | |
1212 ** times (once for each (a,b) combination that matches a=?) is dealt with | |
1213 ** by the caller. | |
1214 ** | |
1215 ** It does this by scanning through all stat4 samples, comparing values | |
1216 ** extracted from pLower and pUpper with the corresponding column in each | |
1217 ** sample. If L and U are the number of samples found to be less than or | |
1218 ** equal to the values extracted from pLower and pUpper respectively, and | |
1219 ** N is the total number of samples, the pLoop->nOut value is adjusted | |
1220 ** as follows: | |
1221 ** | |
1222 ** nOut = nOut * ( min(U - L, 1) / N ) | |
1223 ** | |
1224 ** If pLower is NULL, or a value cannot be extracted from the term, L is | |
1225 ** set to zero. If pUpper is NULL, or a value cannot be extracted from it, | |
1226 ** U is set to N. | |
1227 ** | |
1228 ** Normally, this function sets *pbDone to 1 before returning. However, | |
1229 ** if no value can be extracted from either pLower or pUpper (and so the | |
1230 ** estimate of the number of rows delivered remains unchanged), *pbDone | |
1231 ** is left as is. | |
1232 ** | |
1233 ** If an error occurs, an SQLite error code is returned. Otherwise, | |
1234 ** SQLITE_OK. | |
1235 */ | |
1236 static int whereRangeSkipScanEst( | |
1237 Parse *pParse, /* Parsing & code generating context */ | |
1238 WhereTerm *pLower, /* Lower bound on the range. ex: "x>123" Might be NULL */ | |
1239 WhereTerm *pUpper, /* Upper bound on the range. ex: "x<455" Might be NULL */ | |
1240 WhereLoop *pLoop, /* Update the .nOut value of this loop */ | |
1241 int *pbDone /* Set to true if at least one expr. value extracted */ | |
1242 ){ | |
1243 Index *p = pLoop->u.btree.pIndex; | |
1244 int nEq = pLoop->u.btree.nEq; | |
1245 sqlite3 *db = pParse->db; | |
1246 int nLower = -1; | |
1247 int nUpper = p->nSample+1; | |
1248 int rc = SQLITE_OK; | |
1249 u8 aff = sqlite3IndexColumnAffinity(db, p, nEq); | |
1250 CollSeq *pColl; | |
1251 | |
1252 sqlite3_value *p1 = 0; /* Value extracted from pLower */ | |
1253 sqlite3_value *p2 = 0; /* Value extracted from pUpper */ | |
1254 sqlite3_value *pVal = 0; /* Value extracted from record */ | |
1255 | |
1256 pColl = sqlite3LocateCollSeq(pParse, p->azColl[nEq]); | |
1257 if( pLower ){ | |
1258 rc = sqlite3Stat4ValueFromExpr(pParse, pLower->pExpr->pRight, aff, &p1); | |
1259 nLower = 0; | |
1260 } | |
1261 if( pUpper && rc==SQLITE_OK ){ | |
1262 rc = sqlite3Stat4ValueFromExpr(pParse, pUpper->pExpr->pRight, aff, &p2); | |
1263 nUpper = p2 ? 0 : p->nSample; | |
1264 } | |
1265 | |
1266 if( p1 || p2 ){ | |
1267 int i; | |
1268 int nDiff; | |
1269 for(i=0; rc==SQLITE_OK && i<p->nSample; i++){ | |
1270 rc = sqlite3Stat4Column(db, p->aSample[i].p, p->aSample[i].n, nEq, &pVal); | |
1271 if( rc==SQLITE_OK && p1 ){ | |
1272 int res = sqlite3MemCompare(p1, pVal, pColl); | |
1273 if( res>=0 ) nLower++; | |
1274 } | |
1275 if( rc==SQLITE_OK && p2 ){ | |
1276 int res = sqlite3MemCompare(p2, pVal, pColl); | |
1277 if( res>=0 ) nUpper++; | |
1278 } | |
1279 } | |
1280 nDiff = (nUpper - nLower); | |
1281 if( nDiff<=0 ) nDiff = 1; | |
1282 | |
1283 /* If there is both an upper and lower bound specified, and the | |
1284 ** comparisons indicate that they are close together, use the fallback | |
1285 ** method (assume that the scan visits 1/64 of the rows) for estimating | |
1286 ** the number of rows visited. Otherwise, estimate the number of rows | |
1287 ** using the method described in the header comment for this function. */ | |
1288 if( nDiff!=1 || pUpper==0 || pLower==0 ){ | |
1289 int nAdjust = (sqlite3LogEst(p->nSample) - sqlite3LogEst(nDiff)); | |
1290 pLoop->nOut -= nAdjust; | |
1291 *pbDone = 1; | |
1292 WHERETRACE(0x10, ("range skip-scan regions: %u..%u adjust=%d est=%d\n", | |
1293 nLower, nUpper, nAdjust*-1, pLoop->nOut)); | |
1294 } | |
1295 | |
1296 }else{ | |
1297 assert( *pbDone==0 ); | |
1298 } | |
1299 | |
1300 sqlite3ValueFree(p1); | |
1301 sqlite3ValueFree(p2); | |
1302 sqlite3ValueFree(pVal); | |
1303 | |
1304 return rc; | |
1305 } | |
1306 #endif /* SQLITE_ENABLE_STAT3_OR_STAT4 */ | |
1307 | |
1308 /* | |
1309 ** This function is used to estimate the number of rows that will be visited | |
1310 ** by scanning an index for a range of values. The range may have an upper | |
1311 ** bound, a lower bound, or both. The WHERE clause terms that set the upper | |
1312 ** and lower bounds are represented by pLower and pUpper respectively. For | |
1313 ** example, assuming that index p is on t1(a): | |
1314 ** | |
1315 ** ... FROM t1 WHERE a > ? AND a < ? ... | |
1316 ** |_____| |_____| | |
1317 ** | | | |
1318 ** pLower pUpper | |
1319 ** | |
1320 ** If either of the upper or lower bound is not present, then NULL is passed in | |
1321 ** place of the corresponding WhereTerm. | |
1322 ** | |
1323 ** The value in (pBuilder->pNew->u.btree.nEq) is the number of the index | |
1324 ** column subject to the range constraint. Or, equivalently, the number of | |
1325 ** equality constraints optimized by the proposed index scan. For example, | |
1326 ** assuming index p is on t1(a, b), and the SQL query is: | |
1327 ** | |
1328 ** ... FROM t1 WHERE a = ? AND b > ? AND b < ? ... | |
1329 ** | |
1330 ** then nEq is set to 1 (as the range restricted column, b, is the second | |
1331 ** left-most column of the index). Or, if the query is: | |
1332 ** | |
1333 ** ... FROM t1 WHERE a > ? AND a < ? ... | |
1334 ** | |
1335 ** then nEq is set to 0. | |
1336 ** | |
1337 ** When this function is called, *pnOut is set to the sqlite3LogEst() of the | |
1338 ** number of rows that the index scan is expected to visit without | |
1339 ** considering the range constraints. If nEq is 0, then *pnOut is the number of | |
1340 ** rows in the index. Assuming no error occurs, *pnOut is adjusted (reduced) | |
1341 ** to account for the range constraints pLower and pUpper. | |
1342 ** | |
1343 ** In the absence of sqlite_stat4 ANALYZE data, or if such data cannot be | |
1344 ** used, a single range inequality reduces the search space by a factor of 4. | |
1345 ** and a pair of constraints (x>? AND x<?) reduces the expected number of | |
1346 ** rows visited by a factor of 64. | |
1347 */ | |
1348 static int whereRangeScanEst( | |
1349 Parse *pParse, /* Parsing & code generating context */ | |
1350 WhereLoopBuilder *pBuilder, | |
1351 WhereTerm *pLower, /* Lower bound on the range. ex: "x>123" Might be NULL */ | |
1352 WhereTerm *pUpper, /* Upper bound on the range. ex: "x<455" Might be NULL */ | |
1353 WhereLoop *pLoop /* Modify the .nOut and maybe .rRun fields */ | |
1354 ){ | |
1355 int rc = SQLITE_OK; | |
1356 int nOut = pLoop->nOut; | |
1357 LogEst nNew; | |
1358 | |
1359 #ifdef SQLITE_ENABLE_STAT3_OR_STAT4 | |
1360 Index *p = pLoop->u.btree.pIndex; | |
1361 int nEq = pLoop->u.btree.nEq; | |
1362 | |
1363 if( p->nSample>0 && nEq<p->nSampleCol ){ | |
1364 if( nEq==pBuilder->nRecValid ){ | |
1365 UnpackedRecord *pRec = pBuilder->pRec; | |
1366 tRowcnt a[2]; | |
1367 u8 aff; | |
1368 | |
1369 /* Variable iLower will be set to the estimate of the number of rows in | |
1370 ** the index that are less than the lower bound of the range query. The | |
1371 ** lower bound being the concatenation of $P and $L, where $P is the | |
1372 ** key-prefix formed by the nEq values matched against the nEq left-most | |
1373 ** columns of the index, and $L is the value in pLower. | |
1374 ** | |
1375 ** Or, if pLower is NULL or $L cannot be extracted from it (because it | |
1376 ** is not a simple variable or literal value), the lower bound of the | |
1377 ** range is $P. Due to a quirk in the way whereKeyStats() works, even | |
1378 ** if $L is available, whereKeyStats() is called for both ($P) and | |
1379 ** ($P:$L) and the larger of the two returned values is used. | |
1380 ** | |
1381 ** Similarly, iUpper is to be set to the estimate of the number of rows | |
1382 ** less than the upper bound of the range query. Where the upper bound | |
1383 ** is either ($P) or ($P:$U). Again, even if $U is available, both values | |
1384 ** of iUpper are requested of whereKeyStats() and the smaller used. | |
1385 ** | |
1386 ** The number of rows between the two bounds is then just iUpper-iLower. | |
1387 */ | |
1388 tRowcnt iLower; /* Rows less than the lower bound */ | |
1389 tRowcnt iUpper; /* Rows less than the upper bound */ | |
1390 int iLwrIdx = -2; /* aSample[] for the lower bound */ | |
1391 int iUprIdx = -1; /* aSample[] for the upper bound */ | |
1392 | |
1393 if( pRec ){ | |
1394 testcase( pRec->nField!=pBuilder->nRecValid ); | |
1395 pRec->nField = pBuilder->nRecValid; | |
1396 } | |
1397 aff = sqlite3IndexColumnAffinity(pParse->db, p, nEq); | |
1398 assert( nEq!=p->nKeyCol || aff==SQLITE_AFF_INTEGER ); | |
1399 /* Determine iLower and iUpper using ($P) only. */ | |
1400 if( nEq==0 ){ | |
1401 iLower = 0; | |
1402 iUpper = p->nRowEst0; | |
1403 }else{ | |
1404 /* Note: this call could be optimized away - since the same values must | |
1405 ** have been requested when testing key $P in whereEqualScanEst(). */ | |
1406 whereKeyStats(pParse, p, pRec, 0, a); | |
1407 iLower = a[0]; | |
1408 iUpper = a[0] + a[1]; | |
1409 } | |
1410 | |
1411 assert( pLower==0 || (pLower->eOperator & (WO_GT|WO_GE))!=0 ); | |
1412 assert( pUpper==0 || (pUpper->eOperator & (WO_LT|WO_LE))!=0 ); | |
1413 assert( p->aSortOrder!=0 ); | |
1414 if( p->aSortOrder[nEq] ){ | |
1415 /* The roles of pLower and pUpper are swapped for a DESC index */ | |
1416 SWAP(WhereTerm*, pLower, pUpper); | |
1417 } | |
1418 | |
1419 /* If possible, improve on the iLower estimate using ($P:$L). */ | |
1420 if( pLower ){ | |
1421 int bOk; /* True if value is extracted from pExpr */ | |
1422 Expr *pExpr = pLower->pExpr->pRight; | |
1423 rc = sqlite3Stat4ProbeSetValue(pParse, p, &pRec, pExpr, aff, nEq, &bOk); | |
1424 if( rc==SQLITE_OK && bOk ){ | |
1425 tRowcnt iNew; | |
1426 iLwrIdx = whereKeyStats(pParse, p, pRec, 0, a); | |
1427 iNew = a[0] + ((pLower->eOperator & (WO_GT|WO_LE)) ? a[1] : 0); | |
1428 if( iNew>iLower ) iLower = iNew; | |
1429 nOut--; | |
1430 pLower = 0; | |
1431 } | |
1432 } | |
1433 | |
1434 /* If possible, improve on the iUpper estimate using ($P:$U). */ | |
1435 if( pUpper ){ | |
1436 int bOk; /* True if value is extracted from pExpr */ | |
1437 Expr *pExpr = pUpper->pExpr->pRight; | |
1438 rc = sqlite3Stat4ProbeSetValue(pParse, p, &pRec, pExpr, aff, nEq, &bOk); | |
1439 if( rc==SQLITE_OK && bOk ){ | |
1440 tRowcnt iNew; | |
1441 iUprIdx = whereKeyStats(pParse, p, pRec, 1, a); | |
1442 iNew = a[0] + ((pUpper->eOperator & (WO_GT|WO_LE)) ? a[1] : 0); | |
1443 if( iNew<iUpper ) iUpper = iNew; | |
1444 nOut--; | |
1445 pUpper = 0; | |
1446 } | |
1447 } | |
1448 | |
1449 pBuilder->pRec = pRec; | |
1450 if( rc==SQLITE_OK ){ | |
1451 if( iUpper>iLower ){ | |
1452 nNew = sqlite3LogEst(iUpper - iLower); | |
1453 /* TUNING: If both iUpper and iLower are derived from the same | |
1454 ** sample, then assume they are 4x more selective. This brings | |
1455 ** the estimated selectivity more in line with what it would be | |
1456 ** if estimated without the use of STAT3/4 tables. */ | |
1457 if( iLwrIdx==iUprIdx ) nNew -= 20; assert( 20==sqlite3LogEst(4) ); | |
1458 }else{ | |
1459 nNew = 10; assert( 10==sqlite3LogEst(2) ); | |
1460 } | |
1461 if( nNew<nOut ){ | |
1462 nOut = nNew; | |
1463 } | |
1464 WHERETRACE(0x10, ("STAT4 range scan: %u..%u est=%d\n", | |
1465 (u32)iLower, (u32)iUpper, nOut)); | |
1466 } | |
1467 }else{ | |
1468 int bDone = 0; | |
1469 rc = whereRangeSkipScanEst(pParse, pLower, pUpper, pLoop, &bDone); | |
1470 if( bDone ) return rc; | |
1471 } | |
1472 } | |
1473 #else | |
1474 UNUSED_PARAMETER(pParse); | |
1475 UNUSED_PARAMETER(pBuilder); | |
1476 assert( pLower || pUpper ); | |
1477 #endif | |
1478 assert( pUpper==0 || (pUpper->wtFlags & TERM_VNULL)==0 ); | |
1479 nNew = whereRangeAdjust(pLower, nOut); | |
1480 nNew = whereRangeAdjust(pUpper, nNew); | |
1481 | |
1482 /* TUNING: If there is both an upper and lower limit and neither limit | |
1483 ** has an application-defined likelihood(), assume the range is | |
1484 ** reduced by an additional 75%. This means that, by default, an open-ended | |
1485 ** range query (e.g. col > ?) is assumed to match 1/4 of the rows in the | |
1486 ** index. While a closed range (e.g. col BETWEEN ? AND ?) is estimated to | |
1487 ** match 1/64 of the index. */ | |
1488 if( pLower && pLower->truthProb>0 && pUpper && pUpper->truthProb>0 ){ | |
1489 nNew -= 20; | |
1490 } | |
1491 | |
1492 nOut -= (pLower!=0) + (pUpper!=0); | |
1493 if( nNew<10 ) nNew = 10; | |
1494 if( nNew<nOut ) nOut = nNew; | |
1495 #if defined(WHERETRACE_ENABLED) | |
1496 if( pLoop->nOut>nOut ){ | |
1497 WHERETRACE(0x10,("Range scan lowers nOut from %d to %d\n", | |
1498 pLoop->nOut, nOut)); | |
1499 } | |
1500 #endif | |
1501 pLoop->nOut = (LogEst)nOut; | |
1502 return rc; | |
1503 } | |
1504 | |
1505 #ifdef SQLITE_ENABLE_STAT3_OR_STAT4 | |
1506 /* | |
1507 ** Estimate the number of rows that will be returned based on | |
1508 ** an equality constraint x=VALUE and where that VALUE occurs in | |
1509 ** the histogram data. This only works when x is the left-most | |
1510 ** column of an index and sqlite_stat3 histogram data is available | |
1511 ** for that index. When pExpr==NULL that means the constraint is | |
1512 ** "x IS NULL" instead of "x=VALUE". | |
1513 ** | |
1514 ** Write the estimated row count into *pnRow and return SQLITE_OK. | |
1515 ** If unable to make an estimate, leave *pnRow unchanged and return | |
1516 ** non-zero. | |
1517 ** | |
1518 ** This routine can fail if it is unable to load a collating sequence | |
1519 ** required for string comparison, or if unable to allocate memory | |
1520 ** for a UTF conversion required for comparison. The error is stored | |
1521 ** in the pParse structure. | |
1522 */ | |
1523 static int whereEqualScanEst( | |
1524 Parse *pParse, /* Parsing & code generating context */ | |
1525 WhereLoopBuilder *pBuilder, | |
1526 Expr *pExpr, /* Expression for VALUE in the x=VALUE constraint */ | |
1527 tRowcnt *pnRow /* Write the revised row estimate here */ | |
1528 ){ | |
1529 Index *p = pBuilder->pNew->u.btree.pIndex; | |
1530 int nEq = pBuilder->pNew->u.btree.nEq; | |
1531 UnpackedRecord *pRec = pBuilder->pRec; | |
1532 u8 aff; /* Column affinity */ | |
1533 int rc; /* Subfunction return code */ | |
1534 tRowcnt a[2]; /* Statistics */ | |
1535 int bOk; | |
1536 | |
1537 assert( nEq>=1 ); | |
1538 assert( nEq<=p->nColumn ); | |
1539 assert( p->aSample!=0 ); | |
1540 assert( p->nSample>0 ); | |
1541 assert( pBuilder->nRecValid<nEq ); | |
1542 | |
1543 /* If values are not available for all fields of the index to the left | |
1544 ** of this one, no estimate can be made. Return SQLITE_NOTFOUND. */ | |
1545 if( pBuilder->nRecValid<(nEq-1) ){ | |
1546 return SQLITE_NOTFOUND; | |
1547 } | |
1548 | |
1549 /* This is an optimization only. The call to sqlite3Stat4ProbeSetValue() | |
1550 ** below would return the same value. */ | |
1551 if( nEq>=p->nColumn ){ | |
1552 *pnRow = 1; | |
1553 return SQLITE_OK; | |
1554 } | |
1555 | |
1556 aff = sqlite3IndexColumnAffinity(pParse->db, p, nEq-1); | |
1557 rc = sqlite3Stat4ProbeSetValue(pParse, p, &pRec, pExpr, aff, nEq-1, &bOk); | |
1558 pBuilder->pRec = pRec; | |
1559 if( rc!=SQLITE_OK ) return rc; | |
1560 if( bOk==0 ) return SQLITE_NOTFOUND; | |
1561 pBuilder->nRecValid = nEq; | |
1562 | |
1563 whereKeyStats(pParse, p, pRec, 0, a); | |
1564 WHERETRACE(0x10,("equality scan regions: %d\n", (int)a[1])); | |
1565 *pnRow = a[1]; | |
1566 | |
1567 return rc; | |
1568 } | |
1569 #endif /* SQLITE_ENABLE_STAT3_OR_STAT4 */ | |
1570 | |
1571 #ifdef SQLITE_ENABLE_STAT3_OR_STAT4 | |
1572 /* | |
1573 ** Estimate the number of rows that will be returned based on | |
1574 ** an IN constraint where the right-hand side of the IN operator | |
1575 ** is a list of values. Example: | |
1576 ** | |
1577 ** WHERE x IN (1,2,3,4) | |
1578 ** | |
1579 ** Write the estimated row count into *pnRow and return SQLITE_OK. | |
1580 ** If unable to make an estimate, leave *pnRow unchanged and return | |
1581 ** non-zero. | |
1582 ** | |
1583 ** This routine can fail if it is unable to load a collating sequence | |
1584 ** required for string comparison, or if unable to allocate memory | |
1585 ** for a UTF conversion required for comparison. The error is stored | |
1586 ** in the pParse structure. | |
1587 */ | |
1588 static int whereInScanEst( | |
1589 Parse *pParse, /* Parsing & code generating context */ | |
1590 WhereLoopBuilder *pBuilder, | |
1591 ExprList *pList, /* The value list on the RHS of "x IN (v1,v2,v3,...)" */ | |
1592 tRowcnt *pnRow /* Write the revised row estimate here */ | |
1593 ){ | |
1594 Index *p = pBuilder->pNew->u.btree.pIndex; | |
1595 i64 nRow0 = sqlite3LogEstToInt(p->aiRowLogEst[0]); | |
1596 int nRecValid = pBuilder->nRecValid; | |
1597 int rc = SQLITE_OK; /* Subfunction return code */ | |
1598 tRowcnt nEst; /* Number of rows for a single term */ | |
1599 tRowcnt nRowEst = 0; /* New estimate of the number of rows */ | |
1600 int i; /* Loop counter */ | |
1601 | |
1602 assert( p->aSample!=0 ); | |
1603 for(i=0; rc==SQLITE_OK && i<pList->nExpr; i++){ | |
1604 nEst = nRow0; | |
1605 rc = whereEqualScanEst(pParse, pBuilder, pList->a[i].pExpr, &nEst); | |
1606 nRowEst += nEst; | |
1607 pBuilder->nRecValid = nRecValid; | |
1608 } | |
1609 | |
1610 if( rc==SQLITE_OK ){ | |
1611 if( nRowEst > nRow0 ) nRowEst = nRow0; | |
1612 *pnRow = nRowEst; | |
1613 WHERETRACE(0x10,("IN row estimate: est=%d\n", nRowEst)); | |
1614 } | |
1615 assert( pBuilder->nRecValid==nRecValid ); | |
1616 return rc; | |
1617 } | |
1618 #endif /* SQLITE_ENABLE_STAT3_OR_STAT4 */ | |
1619 | |
1620 | |
1621 #ifdef WHERETRACE_ENABLED | |
1622 /* | |
1623 ** Print the content of a WhereTerm object | |
1624 */ | |
1625 static void whereTermPrint(WhereTerm *pTerm, int iTerm){ | |
1626 if( pTerm==0 ){ | |
1627 sqlite3DebugPrintf("TERM-%-3d NULL\n", iTerm); | |
1628 }else{ | |
1629 char zType[4]; | |
1630 memcpy(zType, "...", 4); | |
1631 if( pTerm->wtFlags & TERM_VIRTUAL ) zType[0] = 'V'; | |
1632 if( pTerm->eOperator & WO_EQUIV ) zType[1] = 'E'; | |
1633 if( ExprHasProperty(pTerm->pExpr, EP_FromJoin) ) zType[2] = 'L'; | |
1634 sqlite3DebugPrintf( | |
1635 "TERM-%-3d %p %s cursor=%-3d prob=%-3d op=0x%03x wtFlags=0x%04x\n", | |
1636 iTerm, pTerm, zType, pTerm->leftCursor, pTerm->truthProb, | |
1637 pTerm->eOperator, pTerm->wtFlags); | |
1638 sqlite3TreeViewExpr(0, pTerm->pExpr, 0); | |
1639 } | |
1640 } | |
1641 #endif | |
1642 | |
1643 #ifdef WHERETRACE_ENABLED | |
1644 /* | |
1645 ** Print a WhereLoop object for debugging purposes | |
1646 */ | |
1647 static void whereLoopPrint(WhereLoop *p, WhereClause *pWC){ | |
1648 WhereInfo *pWInfo = pWC->pWInfo; | |
1649 int nb = 1+(pWInfo->pTabList->nSrc+7)/8; | |
1650 struct SrcList_item *pItem = pWInfo->pTabList->a + p->iTab; | |
1651 Table *pTab = pItem->pTab; | |
1652 sqlite3DebugPrintf("%c%2d.%0*llx.%0*llx", p->cId, | |
1653 p->iTab, nb, p->maskSelf, nb, p->prereq); | |
1654 sqlite3DebugPrintf(" %12s", | |
1655 pItem->zAlias ? pItem->zAlias : pTab->zName); | |
1656 if( (p->wsFlags & WHERE_VIRTUALTABLE)==0 ){ | |
1657 const char *zName; | |
1658 if( p->u.btree.pIndex && (zName = p->u.btree.pIndex->zName)!=0 ){ | |
1659 if( strncmp(zName, "sqlite_autoindex_", 17)==0 ){ | |
1660 int i = sqlite3Strlen30(zName) - 1; | |
1661 while( zName[i]!='_' ) i--; | |
1662 zName += i; | |
1663 } | |
1664 sqlite3DebugPrintf(".%-16s %2d", zName, p->u.btree.nEq); | |
1665 }else{ | |
1666 sqlite3DebugPrintf("%20s",""); | |
1667 } | |
1668 }else{ | |
1669 char *z; | |
1670 if( p->u.vtab.idxStr ){ | |
1671 z = sqlite3_mprintf("(%d,\"%s\",%x)", | |
1672 p->u.vtab.idxNum, p->u.vtab.idxStr, p->u.vtab.omitMask); | |
1673 }else{ | |
1674 z = sqlite3_mprintf("(%d,%x)", p->u.vtab.idxNum, p->u.vtab.omitMask); | |
1675 } | |
1676 sqlite3DebugPrintf(" %-19s", z); | |
1677 sqlite3_free(z); | |
1678 } | |
1679 if( p->wsFlags & WHERE_SKIPSCAN ){ | |
1680 sqlite3DebugPrintf(" f %05x %d-%d", p->wsFlags, p->nLTerm,p->nSkip); | |
1681 }else{ | |
1682 sqlite3DebugPrintf(" f %05x N %d", p->wsFlags, p->nLTerm); | |
1683 } | |
1684 sqlite3DebugPrintf(" cost %d,%d,%d\n", p->rSetup, p->rRun, p->nOut); | |
1685 if( p->nLTerm && (sqlite3WhereTrace & 0x100)!=0 ){ | |
1686 int i; | |
1687 for(i=0; i<p->nLTerm; i++){ | |
1688 whereTermPrint(p->aLTerm[i], i); | |
1689 } | |
1690 } | |
1691 } | |
1692 #endif | |
1693 | |
1694 /* | |
1695 ** Convert bulk memory into a valid WhereLoop that can be passed | |
1696 ** to whereLoopClear harmlessly. | |
1697 */ | |
1698 static void whereLoopInit(WhereLoop *p){ | |
1699 p->aLTerm = p->aLTermSpace; | |
1700 p->nLTerm = 0; | |
1701 p->nLSlot = ArraySize(p->aLTermSpace); | |
1702 p->wsFlags = 0; | |
1703 } | |
1704 | |
1705 /* | |
1706 ** Clear the WhereLoop.u union. Leave WhereLoop.pLTerm intact. | |
1707 */ | |
1708 static void whereLoopClearUnion(sqlite3 *db, WhereLoop *p){ | |
1709 if( p->wsFlags & (WHERE_VIRTUALTABLE|WHERE_AUTO_INDEX) ){ | |
1710 if( (p->wsFlags & WHERE_VIRTUALTABLE)!=0 && p->u.vtab.needFree ){ | |
1711 sqlite3_free(p->u.vtab.idxStr); | |
1712 p->u.vtab.needFree = 0; | |
1713 p->u.vtab.idxStr = 0; | |
1714 }else if( (p->wsFlags & WHERE_AUTO_INDEX)!=0 && p->u.btree.pIndex!=0 ){ | |
1715 sqlite3DbFree(db, p->u.btree.pIndex->zColAff); | |
1716 sqlite3DbFree(db, p->u.btree.pIndex); | |
1717 p->u.btree.pIndex = 0; | |
1718 } | |
1719 } | |
1720 } | |
1721 | |
1722 /* | |
1723 ** Deallocate internal memory used by a WhereLoop object | |
1724 */ | |
1725 static void whereLoopClear(sqlite3 *db, WhereLoop *p){ | |
1726 if( p->aLTerm!=p->aLTermSpace ) sqlite3DbFree(db, p->aLTerm); | |
1727 whereLoopClearUnion(db, p); | |
1728 whereLoopInit(p); | |
1729 } | |
1730 | |
1731 /* | |
1732 ** Increase the memory allocation for pLoop->aLTerm[] to be at least n. | |
1733 */ | |
1734 static int whereLoopResize(sqlite3 *db, WhereLoop *p, int n){ | |
1735 WhereTerm **paNew; | |
1736 if( p->nLSlot>=n ) return SQLITE_OK; | |
1737 n = (n+7)&~7; | |
1738 paNew = sqlite3DbMallocRaw(db, sizeof(p->aLTerm[0])*n); | |
1739 if( paNew==0 ) return SQLITE_NOMEM; | |
1740 memcpy(paNew, p->aLTerm, sizeof(p->aLTerm[0])*p->nLSlot); | |
1741 if( p->aLTerm!=p->aLTermSpace ) sqlite3DbFree(db, p->aLTerm); | |
1742 p->aLTerm = paNew; | |
1743 p->nLSlot = n; | |
1744 return SQLITE_OK; | |
1745 } | |
1746 | |
1747 /* | |
1748 ** Transfer content from the second pLoop into the first. | |
1749 */ | |
1750 static int whereLoopXfer(sqlite3 *db, WhereLoop *pTo, WhereLoop *pFrom){ | |
1751 whereLoopClearUnion(db, pTo); | |
1752 if( whereLoopResize(db, pTo, pFrom->nLTerm) ){ | |
1753 memset(&pTo->u, 0, sizeof(pTo->u)); | |
1754 return SQLITE_NOMEM; | |
1755 } | |
1756 memcpy(pTo, pFrom, WHERE_LOOP_XFER_SZ); | |
1757 memcpy(pTo->aLTerm, pFrom->aLTerm, pTo->nLTerm*sizeof(pTo->aLTerm[0])); | |
1758 if( pFrom->wsFlags & WHERE_VIRTUALTABLE ){ | |
1759 pFrom->u.vtab.needFree = 0; | |
1760 }else if( (pFrom->wsFlags & WHERE_AUTO_INDEX)!=0 ){ | |
1761 pFrom->u.btree.pIndex = 0; | |
1762 } | |
1763 return SQLITE_OK; | |
1764 } | |
1765 | |
1766 /* | |
1767 ** Delete a WhereLoop object | |
1768 */ | |
1769 static void whereLoopDelete(sqlite3 *db, WhereLoop *p){ | |
1770 whereLoopClear(db, p); | |
1771 sqlite3DbFree(db, p); | |
1772 } | |
1773 | |
1774 /* | |
1775 ** Free a WhereInfo structure | |
1776 */ | |
1777 static void whereInfoFree(sqlite3 *db, WhereInfo *pWInfo){ | |
1778 if( ALWAYS(pWInfo) ){ | |
1779 int i; | |
1780 for(i=0; i<pWInfo->nLevel; i++){ | |
1781 WhereLevel *pLevel = &pWInfo->a[i]; | |
1782 if( pLevel->pWLoop && (pLevel->pWLoop->wsFlags & WHERE_IN_ABLE) ){ | |
1783 sqlite3DbFree(db, pLevel->u.in.aInLoop); | |
1784 } | |
1785 } | |
1786 sqlite3WhereClauseClear(&pWInfo->sWC); | |
1787 while( pWInfo->pLoops ){ | |
1788 WhereLoop *p = pWInfo->pLoops; | |
1789 pWInfo->pLoops = p->pNextLoop; | |
1790 whereLoopDelete(db, p); | |
1791 } | |
1792 sqlite3DbFree(db, pWInfo); | |
1793 } | |
1794 } | |
1795 | |
1796 /* | |
1797 ** Return TRUE if all of the following are true: | |
1798 ** | |
1799 ** (1) X has the same or lower cost that Y | |
1800 ** (2) X is a proper subset of Y | |
1801 ** (3) X skips at least as many columns as Y | |
1802 ** | |
1803 ** By "proper subset" we mean that X uses fewer WHERE clause terms | |
1804 ** than Y and that every WHERE clause term used by X is also used | |
1805 ** by Y. | |
1806 ** | |
1807 ** If X is a proper subset of Y then Y is a better choice and ought | |
1808 ** to have a lower cost. This routine returns TRUE when that cost | |
1809 ** relationship is inverted and needs to be adjusted. The third rule | |
1810 ** was added because if X uses skip-scan less than Y it still might | |
1811 ** deserve a lower cost even if it is a proper subset of Y. | |
1812 */ | |
1813 static int whereLoopCheaperProperSubset( | |
1814 const WhereLoop *pX, /* First WhereLoop to compare */ | |
1815 const WhereLoop *pY /* Compare against this WhereLoop */ | |
1816 ){ | |
1817 int i, j; | |
1818 if( pX->nLTerm-pX->nSkip >= pY->nLTerm-pY->nSkip ){ | |
1819 return 0; /* X is not a subset of Y */ | |
1820 } | |
1821 if( pY->nSkip > pX->nSkip ) return 0; | |
1822 if( pX->rRun >= pY->rRun ){ | |
1823 if( pX->rRun > pY->rRun ) return 0; /* X costs more than Y */ | |
1824 if( pX->nOut > pY->nOut ) return 0; /* X costs more than Y */ | |
1825 } | |
1826 for(i=pX->nLTerm-1; i>=0; i--){ | |
1827 if( pX->aLTerm[i]==0 ) continue; | |
1828 for(j=pY->nLTerm-1; j>=0; j--){ | |
1829 if( pY->aLTerm[j]==pX->aLTerm[i] ) break; | |
1830 } | |
1831 if( j<0 ) return 0; /* X not a subset of Y since term X[i] not used by Y */ | |
1832 } | |
1833 return 1; /* All conditions meet */ | |
1834 } | |
1835 | |
1836 /* | |
1837 ** Try to adjust the cost of WhereLoop pTemplate upwards or downwards so | |
1838 ** that: | |
1839 ** | |
1840 ** (1) pTemplate costs less than any other WhereLoops that are a proper | |
1841 ** subset of pTemplate | |
1842 ** | |
1843 ** (2) pTemplate costs more than any other WhereLoops for which pTemplate | |
1844 ** is a proper subset. | |
1845 ** | |
1846 ** To say "WhereLoop X is a proper subset of Y" means that X uses fewer | |
1847 ** WHERE clause terms than Y and that every WHERE clause term used by X is | |
1848 ** also used by Y. | |
1849 */ | |
1850 static void whereLoopAdjustCost(const WhereLoop *p, WhereLoop *pTemplate){ | |
1851 if( (pTemplate->wsFlags & WHERE_INDEXED)==0 ) return; | |
1852 for(; p; p=p->pNextLoop){ | |
1853 if( p->iTab!=pTemplate->iTab ) continue; | |
1854 if( (p->wsFlags & WHERE_INDEXED)==0 ) continue; | |
1855 if( whereLoopCheaperProperSubset(p, pTemplate) ){ | |
1856 /* Adjust pTemplate cost downward so that it is cheaper than its | |
1857 ** subset p. */ | |
1858 WHERETRACE(0x80,("subset cost adjustment %d,%d to %d,%d\n", | |
1859 pTemplate->rRun, pTemplate->nOut, p->rRun, p->nOut-1)); | |
1860 pTemplate->rRun = p->rRun; | |
1861 pTemplate->nOut = p->nOut - 1; | |
1862 }else if( whereLoopCheaperProperSubset(pTemplate, p) ){ | |
1863 /* Adjust pTemplate cost upward so that it is costlier than p since | |
1864 ** pTemplate is a proper subset of p */ | |
1865 WHERETRACE(0x80,("subset cost adjustment %d,%d to %d,%d\n", | |
1866 pTemplate->rRun, pTemplate->nOut, p->rRun, p->nOut+1)); | |
1867 pTemplate->rRun = p->rRun; | |
1868 pTemplate->nOut = p->nOut + 1; | |
1869 } | |
1870 } | |
1871 } | |
1872 | |
1873 /* | |
1874 ** Search the list of WhereLoops in *ppPrev looking for one that can be | |
1875 ** supplanted by pTemplate. | |
1876 ** | |
1877 ** Return NULL if the WhereLoop list contains an entry that can supplant | |
1878 ** pTemplate, in other words if pTemplate does not belong on the list. | |
1879 ** | |
1880 ** If pX is a WhereLoop that pTemplate can supplant, then return the | |
1881 ** link that points to pX. | |
1882 ** | |
1883 ** If pTemplate cannot supplant any existing element of the list but needs | |
1884 ** to be added to the list, then return a pointer to the tail of the list. | |
1885 */ | |
1886 static WhereLoop **whereLoopFindLesser( | |
1887 WhereLoop **ppPrev, | |
1888 const WhereLoop *pTemplate | |
1889 ){ | |
1890 WhereLoop *p; | |
1891 for(p=(*ppPrev); p; ppPrev=&p->pNextLoop, p=*ppPrev){ | |
1892 if( p->iTab!=pTemplate->iTab || p->iSortIdx!=pTemplate->iSortIdx ){ | |
1893 /* If either the iTab or iSortIdx values for two WhereLoop are different | |
1894 ** then those WhereLoops need to be considered separately. Neither is | |
1895 ** a candidate to replace the other. */ | |
1896 continue; | |
1897 } | |
1898 /* In the current implementation, the rSetup value is either zero | |
1899 ** or the cost of building an automatic index (NlogN) and the NlogN | |
1900 ** is the same for compatible WhereLoops. */ | |
1901 assert( p->rSetup==0 || pTemplate->rSetup==0 | |
1902 || p->rSetup==pTemplate->rSetup ); | |
1903 | |
1904 /* whereLoopAddBtree() always generates and inserts the automatic index | |
1905 ** case first. Hence compatible candidate WhereLoops never have a larger | |
1906 ** rSetup. Call this SETUP-INVARIANT */ | |
1907 assert( p->rSetup>=pTemplate->rSetup ); | |
1908 | |
1909 /* Any loop using an appliation-defined index (or PRIMARY KEY or | |
1910 ** UNIQUE constraint) with one or more == constraints is better | |
1911 ** than an automatic index. Unless it is a skip-scan. */ | |
1912 if( (p->wsFlags & WHERE_AUTO_INDEX)!=0 | |
1913 && (pTemplate->nSkip)==0 | |
1914 && (pTemplate->wsFlags & WHERE_INDEXED)!=0 | |
1915 && (pTemplate->wsFlags & WHERE_COLUMN_EQ)!=0 | |
1916 && (p->prereq & pTemplate->prereq)==pTemplate->prereq | |
1917 ){ | |
1918 break; | |
1919 } | |
1920 | |
1921 /* If existing WhereLoop p is better than pTemplate, pTemplate can be | |
1922 ** discarded. WhereLoop p is better if: | |
1923 ** (1) p has no more dependencies than pTemplate, and | |
1924 ** (2) p has an equal or lower cost than pTemplate | |
1925 */ | |
1926 if( (p->prereq & pTemplate->prereq)==p->prereq /* (1) */ | |
1927 && p->rSetup<=pTemplate->rSetup /* (2a) */ | |
1928 && p->rRun<=pTemplate->rRun /* (2b) */ | |
1929 && p->nOut<=pTemplate->nOut /* (2c) */ | |
1930 ){ | |
1931 return 0; /* Discard pTemplate */ | |
1932 } | |
1933 | |
1934 /* If pTemplate is always better than p, then cause p to be overwritten | |
1935 ** with pTemplate. pTemplate is better than p if: | |
1936 ** (1) pTemplate has no more dependences than p, and | |
1937 ** (2) pTemplate has an equal or lower cost than p. | |
1938 */ | |
1939 if( (p->prereq & pTemplate->prereq)==pTemplate->prereq /* (1) */ | |
1940 && p->rRun>=pTemplate->rRun /* (2a) */ | |
1941 && p->nOut>=pTemplate->nOut /* (2b) */ | |
1942 ){ | |
1943 assert( p->rSetup>=pTemplate->rSetup ); /* SETUP-INVARIANT above */ | |
1944 break; /* Cause p to be overwritten by pTemplate */ | |
1945 } | |
1946 } | |
1947 return ppPrev; | |
1948 } | |
1949 | |
1950 /* | |
1951 ** Insert or replace a WhereLoop entry using the template supplied. | |
1952 ** | |
1953 ** An existing WhereLoop entry might be overwritten if the new template | |
1954 ** is better and has fewer dependencies. Or the template will be ignored | |
1955 ** and no insert will occur if an existing WhereLoop is faster and has | |
1956 ** fewer dependencies than the template. Otherwise a new WhereLoop is | |
1957 ** added based on the template. | |
1958 ** | |
1959 ** If pBuilder->pOrSet is not NULL then we care about only the | |
1960 ** prerequisites and rRun and nOut costs of the N best loops. That | |
1961 ** information is gathered in the pBuilder->pOrSet object. This special | |
1962 ** processing mode is used only for OR clause processing. | |
1963 ** | |
1964 ** When accumulating multiple loops (when pBuilder->pOrSet is NULL) we | |
1965 ** still might overwrite similar loops with the new template if the | |
1966 ** new template is better. Loops may be overwritten if the following | |
1967 ** conditions are met: | |
1968 ** | |
1969 ** (1) They have the same iTab. | |
1970 ** (2) They have the same iSortIdx. | |
1971 ** (3) The template has same or fewer dependencies than the current loop | |
1972 ** (4) The template has the same or lower cost than the current loop | |
1973 */ | |
1974 static int whereLoopInsert(WhereLoopBuilder *pBuilder, WhereLoop *pTemplate){ | |
1975 WhereLoop **ppPrev, *p; | |
1976 WhereInfo *pWInfo = pBuilder->pWInfo; | |
1977 sqlite3 *db = pWInfo->pParse->db; | |
1978 | |
1979 /* If pBuilder->pOrSet is defined, then only keep track of the costs | |
1980 ** and prereqs. | |
1981 */ | |
1982 if( pBuilder->pOrSet!=0 ){ | |
1983 if( pTemplate->nLTerm ){ | |
1984 #if WHERETRACE_ENABLED | |
1985 u16 n = pBuilder->pOrSet->n; | |
1986 int x = | |
1987 #endif | |
1988 whereOrInsert(pBuilder->pOrSet, pTemplate->prereq, pTemplate->rRun, | |
1989 pTemplate->nOut); | |
1990 #if WHERETRACE_ENABLED /* 0x8 */ | |
1991 if( sqlite3WhereTrace & 0x8 ){ | |
1992 sqlite3DebugPrintf(x?" or-%d: ":" or-X: ", n); | |
1993 whereLoopPrint(pTemplate, pBuilder->pWC); | |
1994 } | |
1995 #endif | |
1996 } | |
1997 return SQLITE_OK; | |
1998 } | |
1999 | |
2000 /* Look for an existing WhereLoop to replace with pTemplate | |
2001 */ | |
2002 whereLoopAdjustCost(pWInfo->pLoops, pTemplate); | |
2003 ppPrev = whereLoopFindLesser(&pWInfo->pLoops, pTemplate); | |
2004 | |
2005 if( ppPrev==0 ){ | |
2006 /* There already exists a WhereLoop on the list that is better | |
2007 ** than pTemplate, so just ignore pTemplate */ | |
2008 #if WHERETRACE_ENABLED /* 0x8 */ | |
2009 if( sqlite3WhereTrace & 0x8 ){ | |
2010 sqlite3DebugPrintf(" skip: "); | |
2011 whereLoopPrint(pTemplate, pBuilder->pWC); | |
2012 } | |
2013 #endif | |
2014 return SQLITE_OK; | |
2015 }else{ | |
2016 p = *ppPrev; | |
2017 } | |
2018 | |
2019 /* If we reach this point it means that either p[] should be overwritten | |
2020 ** with pTemplate[] if p[] exists, or if p==NULL then allocate a new | |
2021 ** WhereLoop and insert it. | |
2022 */ | |
2023 #if WHERETRACE_ENABLED /* 0x8 */ | |
2024 if( sqlite3WhereTrace & 0x8 ){ | |
2025 if( p!=0 ){ | |
2026 sqlite3DebugPrintf("replace: "); | |
2027 whereLoopPrint(p, pBuilder->pWC); | |
2028 } | |
2029 sqlite3DebugPrintf(" add: "); | |
2030 whereLoopPrint(pTemplate, pBuilder->pWC); | |
2031 } | |
2032 #endif | |
2033 if( p==0 ){ | |
2034 /* Allocate a new WhereLoop to add to the end of the list */ | |
2035 *ppPrev = p = sqlite3DbMallocRaw(db, sizeof(WhereLoop)); | |
2036 if( p==0 ) return SQLITE_NOMEM; | |
2037 whereLoopInit(p); | |
2038 p->pNextLoop = 0; | |
2039 }else{ | |
2040 /* We will be overwriting WhereLoop p[]. But before we do, first | |
2041 ** go through the rest of the list and delete any other entries besides | |
2042 ** p[] that are also supplated by pTemplate */ | |
2043 WhereLoop **ppTail = &p->pNextLoop; | |
2044 WhereLoop *pToDel; | |
2045 while( *ppTail ){ | |
2046 ppTail = whereLoopFindLesser(ppTail, pTemplate); | |
2047 if( ppTail==0 ) break; | |
2048 pToDel = *ppTail; | |
2049 if( pToDel==0 ) break; | |
2050 *ppTail = pToDel->pNextLoop; | |
2051 #if WHERETRACE_ENABLED /* 0x8 */ | |
2052 if( sqlite3WhereTrace & 0x8 ){ | |
2053 sqlite3DebugPrintf(" delete: "); | |
2054 whereLoopPrint(pToDel, pBuilder->pWC); | |
2055 } | |
2056 #endif | |
2057 whereLoopDelete(db, pToDel); | |
2058 } | |
2059 } | |
2060 whereLoopXfer(db, p, pTemplate); | |
2061 if( (p->wsFlags & WHERE_VIRTUALTABLE)==0 ){ | |
2062 Index *pIndex = p->u.btree.pIndex; | |
2063 if( pIndex && pIndex->tnum==0 ){ | |
2064 p->u.btree.pIndex = 0; | |
2065 } | |
2066 } | |
2067 return SQLITE_OK; | |
2068 } | |
2069 | |
2070 /* | |
2071 ** Adjust the WhereLoop.nOut value downward to account for terms of the | |
2072 ** WHERE clause that reference the loop but which are not used by an | |
2073 ** index. | |
2074 * | |
2075 ** For every WHERE clause term that is not used by the index | |
2076 ** and which has a truth probability assigned by one of the likelihood(), | |
2077 ** likely(), or unlikely() SQL functions, reduce the estimated number | |
2078 ** of output rows by the probability specified. | |
2079 ** | |
2080 ** TUNING: For every WHERE clause term that is not used by the index | |
2081 ** and which does not have an assigned truth probability, heuristics | |
2082 ** described below are used to try to estimate the truth probability. | |
2083 ** TODO --> Perhaps this is something that could be improved by better | |
2084 ** table statistics. | |
2085 ** | |
2086 ** Heuristic 1: Estimate the truth probability as 93.75%. The 93.75% | |
2087 ** value corresponds to -1 in LogEst notation, so this means decrement | |
2088 ** the WhereLoop.nOut field for every such WHERE clause term. | |
2089 ** | |
2090 ** Heuristic 2: If there exists one or more WHERE clause terms of the | |
2091 ** form "x==EXPR" and EXPR is not a constant 0 or 1, then make sure the | |
2092 ** final output row estimate is no greater than 1/4 of the total number | |
2093 ** of rows in the table. In other words, assume that x==EXPR will filter | |
2094 ** out at least 3 out of 4 rows. If EXPR is -1 or 0 or 1, then maybe the | |
2095 ** "x" column is boolean or else -1 or 0 or 1 is a common default value | |
2096 ** on the "x" column and so in that case only cap the output row estimate | |
2097 ** at 1/2 instead of 1/4. | |
2098 */ | |
2099 static void whereLoopOutputAdjust( | |
2100 WhereClause *pWC, /* The WHERE clause */ | |
2101 WhereLoop *pLoop, /* The loop to adjust downward */ | |
2102 LogEst nRow /* Number of rows in the entire table */ | |
2103 ){ | |
2104 WhereTerm *pTerm, *pX; | |
2105 Bitmask notAllowed = ~(pLoop->prereq|pLoop->maskSelf); | |
2106 int i, j, k; | |
2107 LogEst iReduce = 0; /* pLoop->nOut should not exceed nRow-iReduce */ | |
2108 | |
2109 assert( (pLoop->wsFlags & WHERE_AUTO_INDEX)==0 ); | |
2110 for(i=pWC->nTerm, pTerm=pWC->a; i>0; i--, pTerm++){ | |
2111 if( (pTerm->wtFlags & TERM_VIRTUAL)!=0 ) break; | |
2112 if( (pTerm->prereqAll & pLoop->maskSelf)==0 ) continue; | |
2113 if( (pTerm->prereqAll & notAllowed)!=0 ) continue; | |
2114 for(j=pLoop->nLTerm-1; j>=0; j--){ | |
2115 pX = pLoop->aLTerm[j]; | |
2116 if( pX==0 ) continue; | |
2117 if( pX==pTerm ) break; | |
2118 if( pX->iParent>=0 && (&pWC->a[pX->iParent])==pTerm ) break; | |
2119 } | |
2120 if( j<0 ){ | |
2121 if( pTerm->truthProb<=0 ){ | |
2122 /* If a truth probability is specified using the likelihood() hints, | |
2123 ** then use the probability provided by the application. */ | |
2124 pLoop->nOut += pTerm->truthProb; | |
2125 }else{ | |
2126 /* In the absence of explicit truth probabilities, use heuristics to | |
2127 ** guess a reasonable truth probability. */ | |
2128 pLoop->nOut--; | |
2129 if( pTerm->eOperator&(WO_EQ|WO_IS) ){ | |
2130 Expr *pRight = pTerm->pExpr->pRight; | |
2131 testcase( pTerm->pExpr->op==TK_IS ); | |
2132 if( sqlite3ExprIsInteger(pRight, &k) && k>=(-1) && k<=1 ){ | |
2133 k = 10; | |
2134 }else{ | |
2135 k = 20; | |
2136 } | |
2137 if( iReduce<k ) iReduce = k; | |
2138 } | |
2139 } | |
2140 } | |
2141 } | |
2142 if( pLoop->nOut > nRow-iReduce ) pLoop->nOut = nRow - iReduce; | |
2143 } | |
2144 | |
2145 /* | |
2146 ** Adjust the cost C by the costMult facter T. This only occurs if | |
2147 ** compiled with -DSQLITE_ENABLE_COSTMULT | |
2148 */ | |
2149 #ifdef SQLITE_ENABLE_COSTMULT | |
2150 # define ApplyCostMultiplier(C,T) C += T | |
2151 #else | |
2152 # define ApplyCostMultiplier(C,T) | |
2153 #endif | |
2154 | |
2155 /* | |
2156 ** We have so far matched pBuilder->pNew->u.btree.nEq terms of the | |
2157 ** index pIndex. Try to match one more. | |
2158 ** | |
2159 ** When this function is called, pBuilder->pNew->nOut contains the | |
2160 ** number of rows expected to be visited by filtering using the nEq | |
2161 ** terms only. If it is modified, this value is restored before this | |
2162 ** function returns. | |
2163 ** | |
2164 ** If pProbe->tnum==0, that means pIndex is a fake index used for the | |
2165 ** INTEGER PRIMARY KEY. | |
2166 */ | |
2167 static int whereLoopAddBtreeIndex( | |
2168 WhereLoopBuilder *pBuilder, /* The WhereLoop factory */ | |
2169 struct SrcList_item *pSrc, /* FROM clause term being analyzed */ | |
2170 Index *pProbe, /* An index on pSrc */ | |
2171 LogEst nInMul /* log(Number of iterations due to IN) */ | |
2172 ){ | |
2173 WhereInfo *pWInfo = pBuilder->pWInfo; /* WHERE analyse context */ | |
2174 Parse *pParse = pWInfo->pParse; /* Parsing context */ | |
2175 sqlite3 *db = pParse->db; /* Database connection malloc context */ | |
2176 WhereLoop *pNew; /* Template WhereLoop under construction */ | |
2177 WhereTerm *pTerm; /* A WhereTerm under consideration */ | |
2178 int opMask; /* Valid operators for constraints */ | |
2179 WhereScan scan; /* Iterator for WHERE terms */ | |
2180 Bitmask saved_prereq; /* Original value of pNew->prereq */ | |
2181 u16 saved_nLTerm; /* Original value of pNew->nLTerm */ | |
2182 u16 saved_nEq; /* Original value of pNew->u.btree.nEq */ | |
2183 u16 saved_nSkip; /* Original value of pNew->nSkip */ | |
2184 u32 saved_wsFlags; /* Original value of pNew->wsFlags */ | |
2185 LogEst saved_nOut; /* Original value of pNew->nOut */ | |
2186 int rc = SQLITE_OK; /* Return code */ | |
2187 LogEst rSize; /* Number of rows in the table */ | |
2188 LogEst rLogSize; /* Logarithm of table size */ | |
2189 WhereTerm *pTop = 0, *pBtm = 0; /* Top and bottom range constraints */ | |
2190 | |
2191 pNew = pBuilder->pNew; | |
2192 if( db->mallocFailed ) return SQLITE_NOMEM; | |
2193 | |
2194 assert( (pNew->wsFlags & WHERE_VIRTUALTABLE)==0 ); | |
2195 assert( (pNew->wsFlags & WHERE_TOP_LIMIT)==0 ); | |
2196 if( pNew->wsFlags & WHERE_BTM_LIMIT ){ | |
2197 opMask = WO_LT|WO_LE; | |
2198 }else if( /*pProbe->tnum<=0 ||*/ (pSrc->fg.jointype & JT_LEFT)!=0 ){ | |
2199 opMask = WO_EQ|WO_IN|WO_GT|WO_GE|WO_LT|WO_LE; | |
2200 }else{ | |
2201 opMask = WO_EQ|WO_IN|WO_GT|WO_GE|WO_LT|WO_LE|WO_ISNULL|WO_IS; | |
2202 } | |
2203 if( pProbe->bUnordered ) opMask &= ~(WO_GT|WO_GE|WO_LT|WO_LE); | |
2204 | |
2205 assert( pNew->u.btree.nEq<pProbe->nColumn ); | |
2206 | |
2207 saved_nEq = pNew->u.btree.nEq; | |
2208 saved_nSkip = pNew->nSkip; | |
2209 saved_nLTerm = pNew->nLTerm; | |
2210 saved_wsFlags = pNew->wsFlags; | |
2211 saved_prereq = pNew->prereq; | |
2212 saved_nOut = pNew->nOut; | |
2213 pTerm = whereScanInit(&scan, pBuilder->pWC, pSrc->iCursor, saved_nEq, | |
2214 opMask, pProbe); | |
2215 pNew->rSetup = 0; | |
2216 rSize = pProbe->aiRowLogEst[0]; | |
2217 rLogSize = estLog(rSize); | |
2218 for(; rc==SQLITE_OK && pTerm!=0; pTerm = whereScanNext(&scan)){ | |
2219 u16 eOp = pTerm->eOperator; /* Shorthand for pTerm->eOperator */ | |
2220 LogEst rCostIdx; | |
2221 LogEst nOutUnadjusted; /* nOut before IN() and WHERE adjustments */ | |
2222 int nIn = 0; | |
2223 #ifdef SQLITE_ENABLE_STAT3_OR_STAT4 | |
2224 int nRecValid = pBuilder->nRecValid; | |
2225 #endif | |
2226 if( (eOp==WO_ISNULL || (pTerm->wtFlags&TERM_VNULL)!=0) | |
2227 && indexColumnNotNull(pProbe, saved_nEq) | |
2228 ){ | |
2229 continue; /* ignore IS [NOT] NULL constraints on NOT NULL columns */ | |
2230 } | |
2231 if( pTerm->prereqRight & pNew->maskSelf ) continue; | |
2232 | |
2233 /* Do not allow the upper bound of a LIKE optimization range constraint | |
2234 ** to mix with a lower range bound from some other source */ | |
2235 if( pTerm->wtFlags & TERM_LIKEOPT && pTerm->eOperator==WO_LT ) continue; | |
2236 | |
2237 pNew->wsFlags = saved_wsFlags; | |
2238 pNew->u.btree.nEq = saved_nEq; | |
2239 pNew->nLTerm = saved_nLTerm; | |
2240 if( whereLoopResize(db, pNew, pNew->nLTerm+1) ) break; /* OOM */ | |
2241 pNew->aLTerm[pNew->nLTerm++] = pTerm; | |
2242 pNew->prereq = (saved_prereq | pTerm->prereqRight) & ~pNew->maskSelf; | |
2243 | |
2244 assert( nInMul==0 | |
2245 || (pNew->wsFlags & WHERE_COLUMN_NULL)!=0 | |
2246 || (pNew->wsFlags & WHERE_COLUMN_IN)!=0 | |
2247 || (pNew->wsFlags & WHERE_SKIPSCAN)!=0 | |
2248 ); | |
2249 | |
2250 if( eOp & WO_IN ){ | |
2251 Expr *pExpr = pTerm->pExpr; | |
2252 pNew->wsFlags |= WHERE_COLUMN_IN; | |
2253 if( ExprHasProperty(pExpr, EP_xIsSelect) ){ | |
2254 /* "x IN (SELECT ...)": TUNING: the SELECT returns 25 rows */ | |
2255 nIn = 46; assert( 46==sqlite3LogEst(25) ); | |
2256 }else if( ALWAYS(pExpr->x.pList && pExpr->x.pList->nExpr) ){ | |
2257 /* "x IN (value, value, ...)" */ | |
2258 nIn = sqlite3LogEst(pExpr->x.pList->nExpr); | |
2259 } | |
2260 assert( nIn>0 ); /* RHS always has 2 or more terms... The parser | |
2261 ** changes "x IN (?)" into "x=?". */ | |
2262 | |
2263 }else if( eOp & (WO_EQ|WO_IS) ){ | |
2264 int iCol = pProbe->aiColumn[saved_nEq]; | |
2265 pNew->wsFlags |= WHERE_COLUMN_EQ; | |
2266 assert( saved_nEq==pNew->u.btree.nEq ); | |
2267 if( iCol==XN_ROWID | |
2268 || (iCol>0 && nInMul==0 && saved_nEq==pProbe->nKeyCol-1) | |
2269 ){ | |
2270 if( iCol>=0 && pProbe->uniqNotNull==0 ){ | |
2271 pNew->wsFlags |= WHERE_UNQ_WANTED; | |
2272 }else{ | |
2273 pNew->wsFlags |= WHERE_ONEROW; | |
2274 } | |
2275 } | |
2276 }else if( eOp & WO_ISNULL ){ | |
2277 pNew->wsFlags |= WHERE_COLUMN_NULL; | |
2278 }else if( eOp & (WO_GT|WO_GE) ){ | |
2279 testcase( eOp & WO_GT ); | |
2280 testcase( eOp & WO_GE ); | |
2281 pNew->wsFlags |= WHERE_COLUMN_RANGE|WHERE_BTM_LIMIT; | |
2282 pBtm = pTerm; | |
2283 pTop = 0; | |
2284 if( pTerm->wtFlags & TERM_LIKEOPT ){ | |
2285 /* Range contraints that come from the LIKE optimization are | |
2286 ** always used in pairs. */ | |
2287 pTop = &pTerm[1]; | |
2288 assert( (pTop-(pTerm->pWC->a))<pTerm->pWC->nTerm ); | |
2289 assert( pTop->wtFlags & TERM_LIKEOPT ); | |
2290 assert( pTop->eOperator==WO_LT ); | |
2291 if( whereLoopResize(db, pNew, pNew->nLTerm+1) ) break; /* OOM */ | |
2292 pNew->aLTerm[pNew->nLTerm++] = pTop; | |
2293 pNew->wsFlags |= WHERE_TOP_LIMIT; | |
2294 } | |
2295 }else{ | |
2296 assert( eOp & (WO_LT|WO_LE) ); | |
2297 testcase( eOp & WO_LT ); | |
2298 testcase( eOp & WO_LE ); | |
2299 pNew->wsFlags |= WHERE_COLUMN_RANGE|WHERE_TOP_LIMIT; | |
2300 pTop = pTerm; | |
2301 pBtm = (pNew->wsFlags & WHERE_BTM_LIMIT)!=0 ? | |
2302 pNew->aLTerm[pNew->nLTerm-2] : 0; | |
2303 } | |
2304 | |
2305 /* At this point pNew->nOut is set to the number of rows expected to | |
2306 ** be visited by the index scan before considering term pTerm, or the | |
2307 ** values of nIn and nInMul. In other words, assuming that all | |
2308 ** "x IN(...)" terms are replaced with "x = ?". This block updates | |
2309 ** the value of pNew->nOut to account for pTerm (but not nIn/nInMul). */ | |
2310 assert( pNew->nOut==saved_nOut ); | |
2311 if( pNew->wsFlags & WHERE_COLUMN_RANGE ){ | |
2312 /* Adjust nOut using stat3/stat4 data. Or, if there is no stat3/stat4 | |
2313 ** data, using some other estimate. */ | |
2314 whereRangeScanEst(pParse, pBuilder, pBtm, pTop, pNew); | |
2315 }else{ | |
2316 int nEq = ++pNew->u.btree.nEq; | |
2317 assert( eOp & (WO_ISNULL|WO_EQ|WO_IN|WO_IS) ); | |
2318 | |
2319 assert( pNew->nOut==saved_nOut ); | |
2320 if( pTerm->truthProb<=0 && pProbe->aiColumn[saved_nEq]>=0 ){ | |
2321 assert( (eOp & WO_IN) || nIn==0 ); | |
2322 testcase( eOp & WO_IN ); | |
2323 pNew->nOut += pTerm->truthProb; | |
2324 pNew->nOut -= nIn; | |
2325 }else{ | |
2326 #ifdef SQLITE_ENABLE_STAT3_OR_STAT4 | |
2327 tRowcnt nOut = 0; | |
2328 if( nInMul==0 | |
2329 && pProbe->nSample | |
2330 && pNew->u.btree.nEq<=pProbe->nSampleCol | |
2331 && ((eOp & WO_IN)==0 || !ExprHasProperty(pTerm->pExpr, EP_xIsSelect)) | |
2332 ){ | |
2333 Expr *pExpr = pTerm->pExpr; | |
2334 if( (eOp & (WO_EQ|WO_ISNULL|WO_IS))!=0 ){ | |
2335 testcase( eOp & WO_EQ ); | |
2336 testcase( eOp & WO_IS ); | |
2337 testcase( eOp & WO_ISNULL ); | |
2338 rc = whereEqualScanEst(pParse, pBuilder, pExpr->pRight, &nOut); | |
2339 }else{ | |
2340 rc = whereInScanEst(pParse, pBuilder, pExpr->x.pList, &nOut); | |
2341 } | |
2342 if( rc==SQLITE_NOTFOUND ) rc = SQLITE_OK; | |
2343 if( rc!=SQLITE_OK ) break; /* Jump out of the pTerm loop */ | |
2344 if( nOut ){ | |
2345 pNew->nOut = sqlite3LogEst(nOut); | |
2346 if( pNew->nOut>saved_nOut ) pNew->nOut = saved_nOut; | |
2347 pNew->nOut -= nIn; | |
2348 } | |
2349 } | |
2350 if( nOut==0 ) | |
2351 #endif | |
2352 { | |
2353 pNew->nOut += (pProbe->aiRowLogEst[nEq] - pProbe->aiRowLogEst[nEq-1]); | |
2354 if( eOp & WO_ISNULL ){ | |
2355 /* TUNING: If there is no likelihood() value, assume that a | |
2356 ** "col IS NULL" expression matches twice as many rows | |
2357 ** as (col=?). */ | |
2358 pNew->nOut += 10; | |
2359 } | |
2360 } | |
2361 } | |
2362 } | |
2363 | |
2364 /* Set rCostIdx to the cost of visiting selected rows in index. Add | |
2365 ** it to pNew->rRun, which is currently set to the cost of the index | |
2366 ** seek only. Then, if this is a non-covering index, add the cost of | |
2367 ** visiting the rows in the main table. */ | |
2368 rCostIdx = pNew->nOut + 1 + (15*pProbe->szIdxRow)/pSrc->pTab->szTabRow; | |
2369 pNew->rRun = sqlite3LogEstAdd(rLogSize, rCostIdx); | |
2370 if( (pNew->wsFlags & (WHERE_IDX_ONLY|WHERE_IPK))==0 ){ | |
2371 pNew->rRun = sqlite3LogEstAdd(pNew->rRun, pNew->nOut + 16); | |
2372 } | |
2373 ApplyCostMultiplier(pNew->rRun, pProbe->pTable->costMult); | |
2374 | |
2375 nOutUnadjusted = pNew->nOut; | |
2376 pNew->rRun += nInMul + nIn; | |
2377 pNew->nOut += nInMul + nIn; | |
2378 whereLoopOutputAdjust(pBuilder->pWC, pNew, rSize); | |
2379 rc = whereLoopInsert(pBuilder, pNew); | |
2380 | |
2381 if( pNew->wsFlags & WHERE_COLUMN_RANGE ){ | |
2382 pNew->nOut = saved_nOut; | |
2383 }else{ | |
2384 pNew->nOut = nOutUnadjusted; | |
2385 } | |
2386 | |
2387 if( (pNew->wsFlags & WHERE_TOP_LIMIT)==0 | |
2388 && pNew->u.btree.nEq<pProbe->nColumn | |
2389 ){ | |
2390 whereLoopAddBtreeIndex(pBuilder, pSrc, pProbe, nInMul+nIn); | |
2391 } | |
2392 pNew->nOut = saved_nOut; | |
2393 #ifdef SQLITE_ENABLE_STAT3_OR_STAT4 | |
2394 pBuilder->nRecValid = nRecValid; | |
2395 #endif | |
2396 } | |
2397 pNew->prereq = saved_prereq; | |
2398 pNew->u.btree.nEq = saved_nEq; | |
2399 pNew->nSkip = saved_nSkip; | |
2400 pNew->wsFlags = saved_wsFlags; | |
2401 pNew->nOut = saved_nOut; | |
2402 pNew->nLTerm = saved_nLTerm; | |
2403 | |
2404 /* Consider using a skip-scan if there are no WHERE clause constraints | |
2405 ** available for the left-most terms of the index, and if the average | |
2406 ** number of repeats in the left-most terms is at least 18. | |
2407 ** | |
2408 ** The magic number 18 is selected on the basis that scanning 17 rows | |
2409 ** is almost always quicker than an index seek (even though if the index | |
2410 ** contains fewer than 2^17 rows we assume otherwise in other parts of | |
2411 ** the code). And, even if it is not, it should not be too much slower. | |
2412 ** On the other hand, the extra seeks could end up being significantly | |
2413 ** more expensive. */ | |
2414 assert( 42==sqlite3LogEst(18) ); | |
2415 if( saved_nEq==saved_nSkip | |
2416 && saved_nEq+1<pProbe->nKeyCol | |
2417 && pProbe->noSkipScan==0 | |
2418 && pProbe->aiRowLogEst[saved_nEq+1]>=42 /* TUNING: Minimum for skip-scan */ | |
2419 && (rc = whereLoopResize(db, pNew, pNew->nLTerm+1))==SQLITE_OK | |
2420 ){ | |
2421 LogEst nIter; | |
2422 pNew->u.btree.nEq++; | |
2423 pNew->nSkip++; | |
2424 pNew->aLTerm[pNew->nLTerm++] = 0; | |
2425 pNew->wsFlags |= WHERE_SKIPSCAN; | |
2426 nIter = pProbe->aiRowLogEst[saved_nEq] - pProbe->aiRowLogEst[saved_nEq+1]; | |
2427 pNew->nOut -= nIter; | |
2428 /* TUNING: Because uncertainties in the estimates for skip-scan queries, | |
2429 ** add a 1.375 fudge factor to make skip-scan slightly less likely. */ | |
2430 nIter += 5; | |
2431 whereLoopAddBtreeIndex(pBuilder, pSrc, pProbe, nIter + nInMul); | |
2432 pNew->nOut = saved_nOut; | |
2433 pNew->u.btree.nEq = saved_nEq; | |
2434 pNew->nSkip = saved_nSkip; | |
2435 pNew->wsFlags = saved_wsFlags; | |
2436 } | |
2437 | |
2438 return rc; | |
2439 } | |
2440 | |
2441 /* | |
2442 ** Return True if it is possible that pIndex might be useful in | |
2443 ** implementing the ORDER BY clause in pBuilder. | |
2444 ** | |
2445 ** Return False if pBuilder does not contain an ORDER BY clause or | |
2446 ** if there is no way for pIndex to be useful in implementing that | |
2447 ** ORDER BY clause. | |
2448 */ | |
2449 static int indexMightHelpWithOrderBy( | |
2450 WhereLoopBuilder *pBuilder, | |
2451 Index *pIndex, | |
2452 int iCursor | |
2453 ){ | |
2454 ExprList *pOB; | |
2455 ExprList *aColExpr; | |
2456 int ii, jj; | |
2457 | |
2458 if( pIndex->bUnordered ) return 0; | |
2459 if( (pOB = pBuilder->pWInfo->pOrderBy)==0 ) return 0; | |
2460 for(ii=0; ii<pOB->nExpr; ii++){ | |
2461 Expr *pExpr = sqlite3ExprSkipCollate(pOB->a[ii].pExpr); | |
2462 if( pExpr->op==TK_COLUMN && pExpr->iTable==iCursor ){ | |
2463 if( pExpr->iColumn<0 ) return 1; | |
2464 for(jj=0; jj<pIndex->nKeyCol; jj++){ | |
2465 if( pExpr->iColumn==pIndex->aiColumn[jj] ) return 1; | |
2466 } | |
2467 }else if( (aColExpr = pIndex->aColExpr)!=0 ){ | |
2468 for(jj=0; jj<pIndex->nKeyCol; jj++){ | |
2469 if( pIndex->aiColumn[jj]!=XN_EXPR ) continue; | |
2470 if( sqlite3ExprCompare(pExpr,aColExpr->a[jj].pExpr,iCursor)==0 ){ | |
2471 return 1; | |
2472 } | |
2473 } | |
2474 } | |
2475 } | |
2476 return 0; | |
2477 } | |
2478 | |
2479 /* | |
2480 ** Return a bitmask where 1s indicate that the corresponding column of | |
2481 ** the table is used by an index. Only the first 63 columns are considered. | |
2482 */ | |
2483 static Bitmask columnsInIndex(Index *pIdx){ | |
2484 Bitmask m = 0; | |
2485 int j; | |
2486 for(j=pIdx->nColumn-1; j>=0; j--){ | |
2487 int x = pIdx->aiColumn[j]; | |
2488 if( x>=0 ){ | |
2489 testcase( x==BMS-1 ); | |
2490 testcase( x==BMS-2 ); | |
2491 if( x<BMS-1 ) m |= MASKBIT(x); | |
2492 } | |
2493 } | |
2494 return m; | |
2495 } | |
2496 | |
2497 /* Check to see if a partial index with pPartIndexWhere can be used | |
2498 ** in the current query. Return true if it can be and false if not. | |
2499 */ | |
2500 static int whereUsablePartialIndex(int iTab, WhereClause *pWC, Expr *pWhere){ | |
2501 int i; | |
2502 WhereTerm *pTerm; | |
2503 while( pWhere->op==TK_AND ){ | |
2504 if( !whereUsablePartialIndex(iTab,pWC,pWhere->pLeft) ) return 0; | |
2505 pWhere = pWhere->pRight; | |
2506 } | |
2507 for(i=0, pTerm=pWC->a; i<pWC->nTerm; i++, pTerm++){ | |
2508 Expr *pExpr = pTerm->pExpr; | |
2509 if( sqlite3ExprImpliesExpr(pExpr, pWhere, iTab) | |
2510 && (!ExprHasProperty(pExpr, EP_FromJoin) || pExpr->iRightJoinTable==iTab) | |
2511 ){ | |
2512 return 1; | |
2513 } | |
2514 } | |
2515 return 0; | |
2516 } | |
2517 | |
2518 /* | |
2519 ** Add all WhereLoop objects for a single table of the join where the table | |
2520 ** is idenfied by pBuilder->pNew->iTab. That table is guaranteed to be | |
2521 ** a b-tree table, not a virtual table. | |
2522 ** | |
2523 ** The costs (WhereLoop.rRun) of the b-tree loops added by this function | |
2524 ** are calculated as follows: | |
2525 ** | |
2526 ** For a full scan, assuming the table (or index) contains nRow rows: | |
2527 ** | |
2528 ** cost = nRow * 3.0 // full-table scan | |
2529 ** cost = nRow * K // scan of covering index | |
2530 ** cost = nRow * (K+3.0) // scan of non-covering index | |
2531 ** | |
2532 ** where K is a value between 1.1 and 3.0 set based on the relative | |
2533 ** estimated average size of the index and table records. | |
2534 ** | |
2535 ** For an index scan, where nVisit is the number of index rows visited | |
2536 ** by the scan, and nSeek is the number of seek operations required on | |
2537 ** the index b-tree: | |
2538 ** | |
2539 ** cost = nSeek * (log(nRow) + K * nVisit) // covering index | |
2540 ** cost = nSeek * (log(nRow) + (K+3.0) * nVisit) // non-covering index | |
2541 ** | |
2542 ** Normally, nSeek is 1. nSeek values greater than 1 come about if the | |
2543 ** WHERE clause includes "x IN (....)" terms used in place of "x=?". Or when | |
2544 ** implicit "x IN (SELECT x FROM tbl)" terms are added for skip-scans. | |
2545 ** | |
2546 ** The estimated values (nRow, nVisit, nSeek) often contain a large amount | |
2547 ** of uncertainty. For this reason, scoring is designed to pick plans that | |
2548 ** "do the least harm" if the estimates are inaccurate. For example, a | |
2549 ** log(nRow) factor is omitted from a non-covering index scan in order to | |
2550 ** bias the scoring in favor of using an index, since the worst-case | |
2551 ** performance of using an index is far better than the worst-case performance | |
2552 ** of a full table scan. | |
2553 */ | |
2554 static int whereLoopAddBtree( | |
2555 WhereLoopBuilder *pBuilder, /* WHERE clause information */ | |
2556 Bitmask mExtra /* Extra prerequesites for using this table */ | |
2557 ){ | |
2558 WhereInfo *pWInfo; /* WHERE analysis context */ | |
2559 Index *pProbe; /* An index we are evaluating */ | |
2560 Index sPk; /* A fake index object for the primary key */ | |
2561 LogEst aiRowEstPk[2]; /* The aiRowLogEst[] value for the sPk index */ | |
2562 i16 aiColumnPk = -1; /* The aColumn[] value for the sPk index */ | |
2563 SrcList *pTabList; /* The FROM clause */ | |
2564 struct SrcList_item *pSrc; /* The FROM clause btree term to add */ | |
2565 WhereLoop *pNew; /* Template WhereLoop object */ | |
2566 int rc = SQLITE_OK; /* Return code */ | |
2567 int iSortIdx = 1; /* Index number */ | |
2568 int b; /* A boolean value */ | |
2569 LogEst rSize; /* number of rows in the table */ | |
2570 LogEst rLogSize; /* Logarithm of the number of rows in the table */ | |
2571 WhereClause *pWC; /* The parsed WHERE clause */ | |
2572 Table *pTab; /* Table being queried */ | |
2573 | |
2574 pNew = pBuilder->pNew; | |
2575 pWInfo = pBuilder->pWInfo; | |
2576 pTabList = pWInfo->pTabList; | |
2577 pSrc = pTabList->a + pNew->iTab; | |
2578 pTab = pSrc->pTab; | |
2579 pWC = pBuilder->pWC; | |
2580 assert( !IsVirtual(pSrc->pTab) ); | |
2581 | |
2582 if( pSrc->pIBIndex ){ | |
2583 /* An INDEXED BY clause specifies a particular index to use */ | |
2584 pProbe = pSrc->pIBIndex; | |
2585 }else if( !HasRowid(pTab) ){ | |
2586 pProbe = pTab->pIndex; | |
2587 }else{ | |
2588 /* There is no INDEXED BY clause. Create a fake Index object in local | |
2589 ** variable sPk to represent the rowid primary key index. Make this | |
2590 ** fake index the first in a chain of Index objects with all of the real | |
2591 ** indices to follow */ | |
2592 Index *pFirst; /* First of real indices on the table */ | |
2593 memset(&sPk, 0, sizeof(Index)); | |
2594 sPk.nKeyCol = 1; | |
2595 sPk.nColumn = 1; | |
2596 sPk.aiColumn = &aiColumnPk; | |
2597 sPk.aiRowLogEst = aiRowEstPk; | |
2598 sPk.onError = OE_Replace; | |
2599 sPk.pTable = pTab; | |
2600 sPk.szIdxRow = pTab->szTabRow; | |
2601 aiRowEstPk[0] = pTab->nRowLogEst; | |
2602 aiRowEstPk[1] = 0; | |
2603 pFirst = pSrc->pTab->pIndex; | |
2604 if( pSrc->fg.notIndexed==0 ){ | |
2605 /* The real indices of the table are only considered if the | |
2606 ** NOT INDEXED qualifier is omitted from the FROM clause */ | |
2607 sPk.pNext = pFirst; | |
2608 } | |
2609 pProbe = &sPk; | |
2610 } | |
2611 rSize = pTab->nRowLogEst; | |
2612 rLogSize = estLog(rSize); | |
2613 | |
2614 #ifndef SQLITE_OMIT_AUTOMATIC_INDEX | |
2615 /* Automatic indexes */ | |
2616 if( !pBuilder->pOrSet /* Not part of an OR optimization */ | |
2617 && (pWInfo->wctrlFlags & WHERE_NO_AUTOINDEX)==0 | |
2618 && (pWInfo->pParse->db->flags & SQLITE_AutoIndex)!=0 | |
2619 && pSrc->pIBIndex==0 /* Has no INDEXED BY clause */ | |
2620 && !pSrc->fg.notIndexed /* Has no NOT INDEXED clause */ | |
2621 && HasRowid(pTab) /* Not WITHOUT ROWID table. (FIXME: Why not?) */ | |
2622 && !pSrc->fg.isCorrelated /* Not a correlated subquery */ | |
2623 && !pSrc->fg.isRecursive /* Not a recursive common table expression. */ | |
2624 ){ | |
2625 /* Generate auto-index WhereLoops */ | |
2626 WhereTerm *pTerm; | |
2627 WhereTerm *pWCEnd = pWC->a + pWC->nTerm; | |
2628 for(pTerm=pWC->a; rc==SQLITE_OK && pTerm<pWCEnd; pTerm++){ | |
2629 if( pTerm->prereqRight & pNew->maskSelf ) continue; | |
2630 if( termCanDriveIndex(pTerm, pSrc, 0) ){ | |
2631 pNew->u.btree.nEq = 1; | |
2632 pNew->nSkip = 0; | |
2633 pNew->u.btree.pIndex = 0; | |
2634 pNew->nLTerm = 1; | |
2635 pNew->aLTerm[0] = pTerm; | |
2636 /* TUNING: One-time cost for computing the automatic index is | |
2637 ** estimated to be X*N*log2(N) where N is the number of rows in | |
2638 ** the table being indexed and where X is 7 (LogEst=28) for normal | |
2639 ** tables or 1.375 (LogEst=4) for views and subqueries. The value | |
2640 ** of X is smaller for views and subqueries so that the query planner | |
2641 ** will be more aggressive about generating automatic indexes for | |
2642 ** those objects, since there is no opportunity to add schema | |
2643 ** indexes on subqueries and views. */ | |
2644 pNew->rSetup = rLogSize + rSize + 4; | |
2645 if( pTab->pSelect==0 && (pTab->tabFlags & TF_Ephemeral)==0 ){ | |
2646 pNew->rSetup += 24; | |
2647 } | |
2648 ApplyCostMultiplier(pNew->rSetup, pTab->costMult); | |
2649 /* TUNING: Each index lookup yields 20 rows in the table. This | |
2650 ** is more than the usual guess of 10 rows, since we have no way | |
2651 ** of knowing how selective the index will ultimately be. It would | |
2652 ** not be unreasonable to make this value much larger. */ | |
2653 pNew->nOut = 43; assert( 43==sqlite3LogEst(20) ); | |
2654 pNew->rRun = sqlite3LogEstAdd(rLogSize,pNew->nOut); | |
2655 pNew->wsFlags = WHERE_AUTO_INDEX; | |
2656 pNew->prereq = mExtra | pTerm->prereqRight; | |
2657 rc = whereLoopInsert(pBuilder, pNew); | |
2658 } | |
2659 } | |
2660 } | |
2661 #endif /* SQLITE_OMIT_AUTOMATIC_INDEX */ | |
2662 | |
2663 /* Loop over all indices | |
2664 */ | |
2665 for(; rc==SQLITE_OK && pProbe; pProbe=pProbe->pNext, iSortIdx++){ | |
2666 if( pProbe->pPartIdxWhere!=0 | |
2667 && !whereUsablePartialIndex(pSrc->iCursor, pWC, pProbe->pPartIdxWhere) ){ | |
2668 testcase( pNew->iTab!=pSrc->iCursor ); /* See ticket [98d973b8f5] */ | |
2669 continue; /* Partial index inappropriate for this query */ | |
2670 } | |
2671 rSize = pProbe->aiRowLogEst[0]; | |
2672 pNew->u.btree.nEq = 0; | |
2673 pNew->nSkip = 0; | |
2674 pNew->nLTerm = 0; | |
2675 pNew->iSortIdx = 0; | |
2676 pNew->rSetup = 0; | |
2677 pNew->prereq = mExtra; | |
2678 pNew->nOut = rSize; | |
2679 pNew->u.btree.pIndex = pProbe; | |
2680 b = indexMightHelpWithOrderBy(pBuilder, pProbe, pSrc->iCursor); | |
2681 /* The ONEPASS_DESIRED flags never occurs together with ORDER BY */ | |
2682 assert( (pWInfo->wctrlFlags & WHERE_ONEPASS_DESIRED)==0 || b==0 ); | |
2683 if( pProbe->tnum<=0 ){ | |
2684 /* Integer primary key index */ | |
2685 pNew->wsFlags = WHERE_IPK; | |
2686 | |
2687 /* Full table scan */ | |
2688 pNew->iSortIdx = b ? iSortIdx : 0; | |
2689 /* TUNING: Cost of full table scan is (N*3.0). */ | |
2690 pNew->rRun = rSize + 16; | |
2691 ApplyCostMultiplier(pNew->rRun, pTab->costMult); | |
2692 whereLoopOutputAdjust(pWC, pNew, rSize); | |
2693 rc = whereLoopInsert(pBuilder, pNew); | |
2694 pNew->nOut = rSize; | |
2695 if( rc ) break; | |
2696 }else{ | |
2697 Bitmask m; | |
2698 if( pProbe->isCovering ){ | |
2699 pNew->wsFlags = WHERE_IDX_ONLY | WHERE_INDEXED; | |
2700 m = 0; | |
2701 }else{ | |
2702 m = pSrc->colUsed & ~columnsInIndex(pProbe); | |
2703 pNew->wsFlags = (m==0) ? (WHERE_IDX_ONLY|WHERE_INDEXED) : WHERE_INDEXED; | |
2704 } | |
2705 | |
2706 /* Full scan via index */ | |
2707 if( b | |
2708 || !HasRowid(pTab) | |
2709 || ( m==0 | |
2710 && pProbe->bUnordered==0 | |
2711 && (pProbe->szIdxRow<pTab->szTabRow) | |
2712 && (pWInfo->wctrlFlags & WHERE_ONEPASS_DESIRED)==0 | |
2713 && sqlite3GlobalConfig.bUseCis | |
2714 && OptimizationEnabled(pWInfo->pParse->db, SQLITE_CoverIdxScan) | |
2715 ) | |
2716 ){ | |
2717 pNew->iSortIdx = b ? iSortIdx : 0; | |
2718 | |
2719 /* The cost of visiting the index rows is N*K, where K is | |
2720 ** between 1.1 and 3.0, depending on the relative sizes of the | |
2721 ** index and table rows. If this is a non-covering index scan, | |
2722 ** also add the cost of visiting table rows (N*3.0). */ | |
2723 pNew->rRun = rSize + 1 + (15*pProbe->szIdxRow)/pTab->szTabRow; | |
2724 if( m!=0 ){ | |
2725 pNew->rRun = sqlite3LogEstAdd(pNew->rRun, rSize+16); | |
2726 } | |
2727 ApplyCostMultiplier(pNew->rRun, pTab->costMult); | |
2728 whereLoopOutputAdjust(pWC, pNew, rSize); | |
2729 rc = whereLoopInsert(pBuilder, pNew); | |
2730 pNew->nOut = rSize; | |
2731 if( rc ) break; | |
2732 } | |
2733 } | |
2734 | |
2735 rc = whereLoopAddBtreeIndex(pBuilder, pSrc, pProbe, 0); | |
2736 #ifdef SQLITE_ENABLE_STAT3_OR_STAT4 | |
2737 sqlite3Stat4ProbeFree(pBuilder->pRec); | |
2738 pBuilder->nRecValid = 0; | |
2739 pBuilder->pRec = 0; | |
2740 #endif | |
2741 | |
2742 /* If there was an INDEXED BY clause, then only that one index is | |
2743 ** considered. */ | |
2744 if( pSrc->pIBIndex ) break; | |
2745 } | |
2746 return rc; | |
2747 } | |
2748 | |
2749 #ifndef SQLITE_OMIT_VIRTUALTABLE | |
2750 /* | |
2751 ** Add all WhereLoop objects for a table of the join identified by | |
2752 ** pBuilder->pNew->iTab. That table is guaranteed to be a virtual table. | |
2753 ** | |
2754 ** If there are no LEFT or CROSS JOIN joins in the query, both mExtra and | |
2755 ** mUnusable are set to 0. Otherwise, mExtra is a mask of all FROM clause | |
2756 ** entries that occur before the virtual table in the FROM clause and are | |
2757 ** separated from it by at least one LEFT or CROSS JOIN. Similarly, the | |
2758 ** mUnusable mask contains all FROM clause entries that occur after the | |
2759 ** virtual table and are separated from it by at least one LEFT or | |
2760 ** CROSS JOIN. | |
2761 ** | |
2762 ** For example, if the query were: | |
2763 ** | |
2764 ** ... FROM t1, t2 LEFT JOIN t3, t4, vt CROSS JOIN t5, t6; | |
2765 ** | |
2766 ** then mExtra corresponds to (t1, t2) and mUnusable to (t5, t6). | |
2767 ** | |
2768 ** All the tables in mExtra must be scanned before the current virtual | |
2769 ** table. So any terms for which all prerequisites are satisfied by | |
2770 ** mExtra may be specified as "usable" in all calls to xBestIndex. | |
2771 ** Conversely, all tables in mUnusable must be scanned after the current | |
2772 ** virtual table, so any terms for which the prerequisites overlap with | |
2773 ** mUnusable should always be configured as "not-usable" for xBestIndex. | |
2774 */ | |
2775 static int whereLoopAddVirtual( | |
2776 WhereLoopBuilder *pBuilder, /* WHERE clause information */ | |
2777 Bitmask mExtra, /* Tables that must be scanned before this one */ | |
2778 Bitmask mUnusable /* Tables that must be scanned after this one */ | |
2779 ){ | |
2780 WhereInfo *pWInfo; /* WHERE analysis context */ | |
2781 Parse *pParse; /* The parsing context */ | |
2782 WhereClause *pWC; /* The WHERE clause */ | |
2783 struct SrcList_item *pSrc; /* The FROM clause term to search */ | |
2784 Table *pTab; | |
2785 sqlite3 *db; | |
2786 sqlite3_index_info *pIdxInfo; | |
2787 struct sqlite3_index_constraint *pIdxCons; | |
2788 struct sqlite3_index_constraint_usage *pUsage; | |
2789 WhereTerm *pTerm; | |
2790 int i, j; | |
2791 int iTerm, mxTerm; | |
2792 int nConstraint; | |
2793 int seenIn = 0; /* True if an IN operator is seen */ | |
2794 int seenVar = 0; /* True if a non-constant constraint is seen */ | |
2795 int iPhase; /* 0: const w/o IN, 1: const, 2: no IN, 2: IN */ | |
2796 WhereLoop *pNew; | |
2797 int rc = SQLITE_OK; | |
2798 | |
2799 assert( (mExtra & mUnusable)==0 ); | |
2800 pWInfo = pBuilder->pWInfo; | |
2801 pParse = pWInfo->pParse; | |
2802 db = pParse->db; | |
2803 pWC = pBuilder->pWC; | |
2804 pNew = pBuilder->pNew; | |
2805 pSrc = &pWInfo->pTabList->a[pNew->iTab]; | |
2806 pTab = pSrc->pTab; | |
2807 assert( IsVirtual(pTab) ); | |
2808 pIdxInfo = allocateIndexInfo(pParse, pWC, mUnusable, pSrc,pBuilder->pOrderBy); | |
2809 if( pIdxInfo==0 ) return SQLITE_NOMEM; | |
2810 pNew->prereq = 0; | |
2811 pNew->rSetup = 0; | |
2812 pNew->wsFlags = WHERE_VIRTUALTABLE; | |
2813 pNew->nLTerm = 0; | |
2814 pNew->u.vtab.needFree = 0; | |
2815 pUsage = pIdxInfo->aConstraintUsage; | |
2816 nConstraint = pIdxInfo->nConstraint; | |
2817 if( whereLoopResize(db, pNew, nConstraint) ){ | |
2818 sqlite3DbFree(db, pIdxInfo); | |
2819 return SQLITE_NOMEM; | |
2820 } | |
2821 | |
2822 for(iPhase=0; iPhase<=3; iPhase++){ | |
2823 if( !seenIn && (iPhase&1)!=0 ){ | |
2824 iPhase++; | |
2825 if( iPhase>3 ) break; | |
2826 } | |
2827 if( !seenVar && iPhase>1 ) break; | |
2828 pIdxCons = *(struct sqlite3_index_constraint**)&pIdxInfo->aConstraint; | |
2829 for(i=0; i<pIdxInfo->nConstraint; i++, pIdxCons++){ | |
2830 j = pIdxCons->iTermOffset; | |
2831 pTerm = &pWC->a[j]; | |
2832 switch( iPhase ){ | |
2833 case 0: /* Constants without IN operator */ | |
2834 pIdxCons->usable = 0; | |
2835 if( (pTerm->eOperator & WO_IN)!=0 ){ | |
2836 seenIn = 1; | |
2837 } | |
2838 if( (pTerm->prereqRight & ~mExtra)!=0 ){ | |
2839 seenVar = 1; | |
2840 }else if( (pTerm->eOperator & WO_IN)==0 ){ | |
2841 pIdxCons->usable = 1; | |
2842 } | |
2843 break; | |
2844 case 1: /* Constants with IN operators */ | |
2845 assert( seenIn ); | |
2846 pIdxCons->usable = (pTerm->prereqRight & ~mExtra)==0; | |
2847 break; | |
2848 case 2: /* Variables without IN */ | |
2849 assert( seenVar ); | |
2850 pIdxCons->usable = (pTerm->eOperator & WO_IN)==0; | |
2851 break; | |
2852 default: /* Variables with IN */ | |
2853 assert( seenVar && seenIn ); | |
2854 pIdxCons->usable = 1; | |
2855 break; | |
2856 } | |
2857 } | |
2858 memset(pUsage, 0, sizeof(pUsage[0])*pIdxInfo->nConstraint); | |
2859 if( pIdxInfo->needToFreeIdxStr ) sqlite3_free(pIdxInfo->idxStr); | |
2860 pIdxInfo->idxStr = 0; | |
2861 pIdxInfo->idxNum = 0; | |
2862 pIdxInfo->needToFreeIdxStr = 0; | |
2863 pIdxInfo->orderByConsumed = 0; | |
2864 pIdxInfo->estimatedCost = SQLITE_BIG_DBL / (double)2; | |
2865 pIdxInfo->estimatedRows = 25; | |
2866 pIdxInfo->idxFlags = 0; | |
2867 pIdxInfo->colUsed = (sqlite3_int64)pSrc->colUsed; | |
2868 rc = vtabBestIndex(pParse, pTab, pIdxInfo); | |
2869 if( rc ) goto whereLoopAddVtab_exit; | |
2870 pIdxCons = *(struct sqlite3_index_constraint**)&pIdxInfo->aConstraint; | |
2871 pNew->prereq = mExtra; | |
2872 mxTerm = -1; | |
2873 assert( pNew->nLSlot>=nConstraint ); | |
2874 for(i=0; i<nConstraint; i++) pNew->aLTerm[i] = 0; | |
2875 pNew->u.vtab.omitMask = 0; | |
2876 for(i=0; i<nConstraint; i++, pIdxCons++){ | |
2877 if( (iTerm = pUsage[i].argvIndex - 1)>=0 ){ | |
2878 j = pIdxCons->iTermOffset; | |
2879 if( iTerm>=nConstraint | |
2880 || j<0 | |
2881 || j>=pWC->nTerm | |
2882 || pNew->aLTerm[iTerm]!=0 | |
2883 ){ | |
2884 rc = SQLITE_ERROR; | |
2885 sqlite3ErrorMsg(pParse, "%s.xBestIndex() malfunction", pTab->zName); | |
2886 goto whereLoopAddVtab_exit; | |
2887 } | |
2888 testcase( iTerm==nConstraint-1 ); | |
2889 testcase( j==0 ); | |
2890 testcase( j==pWC->nTerm-1 ); | |
2891 pTerm = &pWC->a[j]; | |
2892 pNew->prereq |= pTerm->prereqRight; | |
2893 assert( iTerm<pNew->nLSlot ); | |
2894 pNew->aLTerm[iTerm] = pTerm; | |
2895 if( iTerm>mxTerm ) mxTerm = iTerm; | |
2896 testcase( iTerm==15 ); | |
2897 testcase( iTerm==16 ); | |
2898 if( iTerm<16 && pUsage[i].omit ) pNew->u.vtab.omitMask |= 1<<iTerm; | |
2899 if( (pTerm->eOperator & WO_IN)!=0 ){ | |
2900 if( pUsage[i].omit==0 ){ | |
2901 /* Do not attempt to use an IN constraint if the virtual table | |
2902 ** says that the equivalent EQ constraint cannot be safely omitted. | |
2903 ** If we do attempt to use such a constraint, some rows might be | |
2904 ** repeated in the output. */ | |
2905 break; | |
2906 } | |
2907 /* A virtual table that is constrained by an IN clause may not | |
2908 ** consume the ORDER BY clause because (1) the order of IN terms | |
2909 ** is not necessarily related to the order of output terms and | |
2910 ** (2) Multiple outputs from a single IN value will not merge | |
2911 ** together. */ | |
2912 pIdxInfo->orderByConsumed = 0; | |
2913 pIdxInfo->idxFlags &= ~SQLITE_INDEX_SCAN_UNIQUE; | |
2914 } | |
2915 } | |
2916 } | |
2917 if( i>=nConstraint ){ | |
2918 pNew->nLTerm = mxTerm+1; | |
2919 assert( pNew->nLTerm<=pNew->nLSlot ); | |
2920 pNew->u.vtab.idxNum = pIdxInfo->idxNum; | |
2921 pNew->u.vtab.needFree = pIdxInfo->needToFreeIdxStr; | |
2922 pIdxInfo->needToFreeIdxStr = 0; | |
2923 pNew->u.vtab.idxStr = pIdxInfo->idxStr; | |
2924 pNew->u.vtab.isOrdered = (i8)(pIdxInfo->orderByConsumed ? | |
2925 pIdxInfo->nOrderBy : 0); | |
2926 pNew->rSetup = 0; | |
2927 pNew->rRun = sqlite3LogEstFromDouble(pIdxInfo->estimatedCost); | |
2928 pNew->nOut = sqlite3LogEst(pIdxInfo->estimatedRows); | |
2929 | |
2930 /* Set the WHERE_ONEROW flag if the xBestIndex() method indicated | |
2931 ** that the scan will visit at most one row. Clear it otherwise. */ | |
2932 if( pIdxInfo->idxFlags & SQLITE_INDEX_SCAN_UNIQUE ){ | |
2933 pNew->wsFlags |= WHERE_ONEROW; | |
2934 }else{ | |
2935 pNew->wsFlags &= ~WHERE_ONEROW; | |
2936 } | |
2937 whereLoopInsert(pBuilder, pNew); | |
2938 if( pNew->u.vtab.needFree ){ | |
2939 sqlite3_free(pNew->u.vtab.idxStr); | |
2940 pNew->u.vtab.needFree = 0; | |
2941 } | |
2942 } | |
2943 } | |
2944 | |
2945 whereLoopAddVtab_exit: | |
2946 if( pIdxInfo->needToFreeIdxStr ) sqlite3_free(pIdxInfo->idxStr); | |
2947 sqlite3DbFree(db, pIdxInfo); | |
2948 return rc; | |
2949 } | |
2950 #endif /* SQLITE_OMIT_VIRTUALTABLE */ | |
2951 | |
2952 /* | |
2953 ** Add WhereLoop entries to handle OR terms. This works for either | |
2954 ** btrees or virtual tables. | |
2955 */ | |
2956 static int whereLoopAddOr( | |
2957 WhereLoopBuilder *pBuilder, | |
2958 Bitmask mExtra, | |
2959 Bitmask mUnusable | |
2960 ){ | |
2961 WhereInfo *pWInfo = pBuilder->pWInfo; | |
2962 WhereClause *pWC; | |
2963 WhereLoop *pNew; | |
2964 WhereTerm *pTerm, *pWCEnd; | |
2965 int rc = SQLITE_OK; | |
2966 int iCur; | |
2967 WhereClause tempWC; | |
2968 WhereLoopBuilder sSubBuild; | |
2969 WhereOrSet sSum, sCur; | |
2970 struct SrcList_item *pItem; | |
2971 | |
2972 pWC = pBuilder->pWC; | |
2973 pWCEnd = pWC->a + pWC->nTerm; | |
2974 pNew = pBuilder->pNew; | |
2975 memset(&sSum, 0, sizeof(sSum)); | |
2976 pItem = pWInfo->pTabList->a + pNew->iTab; | |
2977 iCur = pItem->iCursor; | |
2978 | |
2979 for(pTerm=pWC->a; pTerm<pWCEnd && rc==SQLITE_OK; pTerm++){ | |
2980 if( (pTerm->eOperator & WO_OR)!=0 | |
2981 && (pTerm->u.pOrInfo->indexable & pNew->maskSelf)!=0 | |
2982 ){ | |
2983 WhereClause * const pOrWC = &pTerm->u.pOrInfo->wc; | |
2984 WhereTerm * const pOrWCEnd = &pOrWC->a[pOrWC->nTerm]; | |
2985 WhereTerm *pOrTerm; | |
2986 int once = 1; | |
2987 int i, j; | |
2988 | |
2989 sSubBuild = *pBuilder; | |
2990 sSubBuild.pOrderBy = 0; | |
2991 sSubBuild.pOrSet = &sCur; | |
2992 | |
2993 WHERETRACE(0x200, ("Begin processing OR-clause %p\n", pTerm)); | |
2994 for(pOrTerm=pOrWC->a; pOrTerm<pOrWCEnd; pOrTerm++){ | |
2995 if( (pOrTerm->eOperator & WO_AND)!=0 ){ | |
2996 sSubBuild.pWC = &pOrTerm->u.pAndInfo->wc; | |
2997 }else if( pOrTerm->leftCursor==iCur ){ | |
2998 tempWC.pWInfo = pWC->pWInfo; | |
2999 tempWC.pOuter = pWC; | |
3000 tempWC.op = TK_AND; | |
3001 tempWC.nTerm = 1; | |
3002 tempWC.a = pOrTerm; | |
3003 sSubBuild.pWC = &tempWC; | |
3004 }else{ | |
3005 continue; | |
3006 } | |
3007 sCur.n = 0; | |
3008 #ifdef WHERETRACE_ENABLED | |
3009 WHERETRACE(0x200, ("OR-term %d of %p has %d subterms:\n", | |
3010 (int)(pOrTerm-pOrWC->a), pTerm, sSubBuild.pWC->nTerm)); | |
3011 if( sqlite3WhereTrace & 0x400 ){ | |
3012 for(i=0; i<sSubBuild.pWC->nTerm; i++){ | |
3013 whereTermPrint(&sSubBuild.pWC->a[i], i); | |
3014 } | |
3015 } | |
3016 #endif | |
3017 #ifndef SQLITE_OMIT_VIRTUALTABLE | |
3018 if( IsVirtual(pItem->pTab) ){ | |
3019 rc = whereLoopAddVirtual(&sSubBuild, mExtra, mUnusable); | |
3020 }else | |
3021 #endif | |
3022 { | |
3023 rc = whereLoopAddBtree(&sSubBuild, mExtra); | |
3024 } | |
3025 if( rc==SQLITE_OK ){ | |
3026 rc = whereLoopAddOr(&sSubBuild, mExtra, mUnusable); | |
3027 } | |
3028 assert( rc==SQLITE_OK || sCur.n==0 ); | |
3029 if( sCur.n==0 ){ | |
3030 sSum.n = 0; | |
3031 break; | |
3032 }else if( once ){ | |
3033 whereOrMove(&sSum, &sCur); | |
3034 once = 0; | |
3035 }else{ | |
3036 WhereOrSet sPrev; | |
3037 whereOrMove(&sPrev, &sSum); | |
3038 sSum.n = 0; | |
3039 for(i=0; i<sPrev.n; i++){ | |
3040 for(j=0; j<sCur.n; j++){ | |
3041 whereOrInsert(&sSum, sPrev.a[i].prereq | sCur.a[j].prereq, | |
3042 sqlite3LogEstAdd(sPrev.a[i].rRun, sCur.a[j].rRun), | |
3043 sqlite3LogEstAdd(sPrev.a[i].nOut, sCur.a[j].nOut)); | |
3044 } | |
3045 } | |
3046 } | |
3047 } | |
3048 pNew->nLTerm = 1; | |
3049 pNew->aLTerm[0] = pTerm; | |
3050 pNew->wsFlags = WHERE_MULTI_OR; | |
3051 pNew->rSetup = 0; | |
3052 pNew->iSortIdx = 0; | |
3053 memset(&pNew->u, 0, sizeof(pNew->u)); | |
3054 for(i=0; rc==SQLITE_OK && i<sSum.n; i++){ | |
3055 /* TUNING: Currently sSum.a[i].rRun is set to the sum of the costs | |
3056 ** of all sub-scans required by the OR-scan. However, due to rounding | |
3057 ** errors, it may be that the cost of the OR-scan is equal to its | |
3058 ** most expensive sub-scan. Add the smallest possible penalty | |
3059 ** (equivalent to multiplying the cost by 1.07) to ensure that | |
3060 ** this does not happen. Otherwise, for WHERE clauses such as the | |
3061 ** following where there is an index on "y": | |
3062 ** | |
3063 ** WHERE likelihood(x=?, 0.99) OR y=? | |
3064 ** | |
3065 ** the planner may elect to "OR" together a full-table scan and an | |
3066 ** index lookup. And other similarly odd results. */ | |
3067 pNew->rRun = sSum.a[i].rRun + 1; | |
3068 pNew->nOut = sSum.a[i].nOut; | |
3069 pNew->prereq = sSum.a[i].prereq; | |
3070 rc = whereLoopInsert(pBuilder, pNew); | |
3071 } | |
3072 WHERETRACE(0x200, ("End processing OR-clause %p\n", pTerm)); | |
3073 } | |
3074 } | |
3075 return rc; | |
3076 } | |
3077 | |
3078 /* | |
3079 ** Add all WhereLoop objects for all tables | |
3080 */ | |
3081 static int whereLoopAddAll(WhereLoopBuilder *pBuilder){ | |
3082 WhereInfo *pWInfo = pBuilder->pWInfo; | |
3083 Bitmask mExtra = 0; | |
3084 Bitmask mPrior = 0; | |
3085 int iTab; | |
3086 SrcList *pTabList = pWInfo->pTabList; | |
3087 struct SrcList_item *pItem; | |
3088 struct SrcList_item *pEnd = &pTabList->a[pWInfo->nLevel]; | |
3089 sqlite3 *db = pWInfo->pParse->db; | |
3090 int rc = SQLITE_OK; | |
3091 WhereLoop *pNew; | |
3092 u8 priorJointype = 0; | |
3093 | |
3094 /* Loop over the tables in the join, from left to right */ | |
3095 pNew = pBuilder->pNew; | |
3096 whereLoopInit(pNew); | |
3097 for(iTab=0, pItem=pTabList->a; pItem<pEnd; iTab++, pItem++){ | |
3098 Bitmask mUnusable = 0; | |
3099 pNew->iTab = iTab; | |
3100 pNew->maskSelf = sqlite3WhereGetMask(&pWInfo->sMaskSet, pItem->iCursor); | |
3101 if( ((pItem->fg.jointype|priorJointype) & (JT_LEFT|JT_CROSS))!=0 ){ | |
3102 /* This condition is true when pItem is the FROM clause term on the | |
3103 ** right-hand-side of a LEFT or CROSS JOIN. */ | |
3104 mExtra = mPrior; | |
3105 } | |
3106 priorJointype = pItem->fg.jointype; | |
3107 if( IsVirtual(pItem->pTab) ){ | |
3108 struct SrcList_item *p; | |
3109 for(p=&pItem[1]; p<pEnd; p++){ | |
3110 if( mUnusable || (p->fg.jointype & (JT_LEFT|JT_CROSS)) ){ | |
3111 mUnusable |= sqlite3WhereGetMask(&pWInfo->sMaskSet, p->iCursor); | |
3112 } | |
3113 } | |
3114 rc = whereLoopAddVirtual(pBuilder, mExtra, mUnusable); | |
3115 }else{ | |
3116 rc = whereLoopAddBtree(pBuilder, mExtra); | |
3117 } | |
3118 if( rc==SQLITE_OK ){ | |
3119 rc = whereLoopAddOr(pBuilder, mExtra, mUnusable); | |
3120 } | |
3121 mPrior |= pNew->maskSelf; | |
3122 if( rc || db->mallocFailed ) break; | |
3123 } | |
3124 | |
3125 whereLoopClear(db, pNew); | |
3126 return rc; | |
3127 } | |
3128 | |
3129 /* | |
3130 ** Examine a WherePath (with the addition of the extra WhereLoop of the 5th | |
3131 ** parameters) to see if it outputs rows in the requested ORDER BY | |
3132 ** (or GROUP BY) without requiring a separate sort operation. Return N: | |
3133 ** | |
3134 ** N>0: N terms of the ORDER BY clause are satisfied | |
3135 ** N==0: No terms of the ORDER BY clause are satisfied | |
3136 ** N<0: Unknown yet how many terms of ORDER BY might be satisfied. | |
3137 ** | |
3138 ** Note that processing for WHERE_GROUPBY and WHERE_DISTINCTBY is not as | |
3139 ** strict. With GROUP BY and DISTINCT the only requirement is that | |
3140 ** equivalent rows appear immediately adjacent to one another. GROUP BY | |
3141 ** and DISTINCT do not require rows to appear in any particular order as long | |
3142 ** as equivalent rows are grouped together. Thus for GROUP BY and DISTINCT | |
3143 ** the pOrderBy terms can be matched in any order. With ORDER BY, the | |
3144 ** pOrderBy terms must be matched in strict left-to-right order. | |
3145 */ | |
3146 static i8 wherePathSatisfiesOrderBy( | |
3147 WhereInfo *pWInfo, /* The WHERE clause */ | |
3148 ExprList *pOrderBy, /* ORDER BY or GROUP BY or DISTINCT clause to check */ | |
3149 WherePath *pPath, /* The WherePath to check */ | |
3150 u16 wctrlFlags, /* Might contain WHERE_GROUPBY or WHERE_DISTINCTBY */ | |
3151 u16 nLoop, /* Number of entries in pPath->aLoop[] */ | |
3152 WhereLoop *pLast, /* Add this WhereLoop to the end of pPath->aLoop[] */ | |
3153 Bitmask *pRevMask /* OUT: Mask of WhereLoops to run in reverse order */ | |
3154 ){ | |
3155 u8 revSet; /* True if rev is known */ | |
3156 u8 rev; /* Composite sort order */ | |
3157 u8 revIdx; /* Index sort order */ | |
3158 u8 isOrderDistinct; /* All prior WhereLoops are order-distinct */ | |
3159 u8 distinctColumns; /* True if the loop has UNIQUE NOT NULL columns */ | |
3160 u8 isMatch; /* iColumn matches a term of the ORDER BY clause */ | |
3161 u16 nKeyCol; /* Number of key columns in pIndex */ | |
3162 u16 nColumn; /* Total number of ordered columns in the index */ | |
3163 u16 nOrderBy; /* Number terms in the ORDER BY clause */ | |
3164 int iLoop; /* Index of WhereLoop in pPath being processed */ | |
3165 int i, j; /* Loop counters */ | |
3166 int iCur; /* Cursor number for current WhereLoop */ | |
3167 int iColumn; /* A column number within table iCur */ | |
3168 WhereLoop *pLoop = 0; /* Current WhereLoop being processed. */ | |
3169 WhereTerm *pTerm; /* A single term of the WHERE clause */ | |
3170 Expr *pOBExpr; /* An expression from the ORDER BY clause */ | |
3171 CollSeq *pColl; /* COLLATE function from an ORDER BY clause term */ | |
3172 Index *pIndex; /* The index associated with pLoop */ | |
3173 sqlite3 *db = pWInfo->pParse->db; /* Database connection */ | |
3174 Bitmask obSat = 0; /* Mask of ORDER BY terms satisfied so far */ | |
3175 Bitmask obDone; /* Mask of all ORDER BY terms */ | |
3176 Bitmask orderDistinctMask; /* Mask of all well-ordered loops */ | |
3177 Bitmask ready; /* Mask of inner loops */ | |
3178 | |
3179 /* | |
3180 ** We say the WhereLoop is "one-row" if it generates no more than one | |
3181 ** row of output. A WhereLoop is one-row if all of the following are true: | |
3182 ** (a) All index columns match with WHERE_COLUMN_EQ. | |
3183 ** (b) The index is unique | |
3184 ** Any WhereLoop with an WHERE_COLUMN_EQ constraint on the rowid is one-row. | |
3185 ** Every one-row WhereLoop will have the WHERE_ONEROW bit set in wsFlags. | |
3186 ** | |
3187 ** We say the WhereLoop is "order-distinct" if the set of columns from | |
3188 ** that WhereLoop that are in the ORDER BY clause are different for every | |
3189 ** row of the WhereLoop. Every one-row WhereLoop is automatically | |
3190 ** order-distinct. A WhereLoop that has no columns in the ORDER BY clause | |
3191 ** is not order-distinct. To be order-distinct is not quite the same as being | |
3192 ** UNIQUE since a UNIQUE column or index can have multiple rows that | |
3193 ** are NULL and NULL values are equivalent for the purpose of order-distinct. | |
3194 ** To be order-distinct, the columns must be UNIQUE and NOT NULL. | |
3195 ** | |
3196 ** The rowid for a table is always UNIQUE and NOT NULL so whenever the | |
3197 ** rowid appears in the ORDER BY clause, the corresponding WhereLoop is | |
3198 ** automatically order-distinct. | |
3199 */ | |
3200 | |
3201 assert( pOrderBy!=0 ); | |
3202 if( nLoop && OptimizationDisabled(db, SQLITE_OrderByIdxJoin) ) return 0; | |
3203 | |
3204 nOrderBy = pOrderBy->nExpr; | |
3205 testcase( nOrderBy==BMS-1 ); | |
3206 if( nOrderBy>BMS-1 ) return 0; /* Cannot optimize overly large ORDER BYs */ | |
3207 isOrderDistinct = 1; | |
3208 obDone = MASKBIT(nOrderBy)-1; | |
3209 orderDistinctMask = 0; | |
3210 ready = 0; | |
3211 for(iLoop=0; isOrderDistinct && obSat<obDone && iLoop<=nLoop; iLoop++){ | |
3212 if( iLoop>0 ) ready |= pLoop->maskSelf; | |
3213 pLoop = iLoop<nLoop ? pPath->aLoop[iLoop] : pLast; | |
3214 if( pLoop->wsFlags & WHERE_VIRTUALTABLE ){ | |
3215 if( pLoop->u.vtab.isOrdered ) obSat = obDone; | |
3216 break; | |
3217 } | |
3218 iCur = pWInfo->pTabList->a[pLoop->iTab].iCursor; | |
3219 | |
3220 /* Mark off any ORDER BY term X that is a column in the table of | |
3221 ** the current loop for which there is term in the WHERE | |
3222 ** clause of the form X IS NULL or X=? that reference only outer | |
3223 ** loops. | |
3224 */ | |
3225 for(i=0; i<nOrderBy; i++){ | |
3226 if( MASKBIT(i) & obSat ) continue; | |
3227 pOBExpr = sqlite3ExprSkipCollate(pOrderBy->a[i].pExpr); | |
3228 if( pOBExpr->op!=TK_COLUMN ) continue; | |
3229 if( pOBExpr->iTable!=iCur ) continue; | |
3230 pTerm = sqlite3WhereFindTerm(&pWInfo->sWC, iCur, pOBExpr->iColumn, | |
3231 ~ready, WO_EQ|WO_ISNULL|WO_IS, 0); | |
3232 if( pTerm==0 ) continue; | |
3233 if( (pTerm->eOperator&(WO_EQ|WO_IS))!=0 && pOBExpr->iColumn>=0 ){ | |
3234 const char *z1, *z2; | |
3235 pColl = sqlite3ExprCollSeq(pWInfo->pParse, pOrderBy->a[i].pExpr); | |
3236 if( !pColl ) pColl = db->pDfltColl; | |
3237 z1 = pColl->zName; | |
3238 pColl = sqlite3ExprCollSeq(pWInfo->pParse, pTerm->pExpr); | |
3239 if( !pColl ) pColl = db->pDfltColl; | |
3240 z2 = pColl->zName; | |
3241 if( sqlite3StrICmp(z1, z2)!=0 ) continue; | |
3242 testcase( pTerm->pExpr->op==TK_IS ); | |
3243 } | |
3244 obSat |= MASKBIT(i); | |
3245 } | |
3246 | |
3247 if( (pLoop->wsFlags & WHERE_ONEROW)==0 ){ | |
3248 if( pLoop->wsFlags & WHERE_IPK ){ | |
3249 pIndex = 0; | |
3250 nKeyCol = 0; | |
3251 nColumn = 1; | |
3252 }else if( (pIndex = pLoop->u.btree.pIndex)==0 || pIndex->bUnordered ){ | |
3253 return 0; | |
3254 }else{ | |
3255 nKeyCol = pIndex->nKeyCol; | |
3256 nColumn = pIndex->nColumn; | |
3257 assert( nColumn==nKeyCol+1 || !HasRowid(pIndex->pTable) ); | |
3258 assert( pIndex->aiColumn[nColumn-1]==XN_ROWID | |
3259 || !HasRowid(pIndex->pTable)); | |
3260 isOrderDistinct = IsUniqueIndex(pIndex); | |
3261 } | |
3262 | |
3263 /* Loop through all columns of the index and deal with the ones | |
3264 ** that are not constrained by == or IN. | |
3265 */ | |
3266 rev = revSet = 0; | |
3267 distinctColumns = 0; | |
3268 for(j=0; j<nColumn; j++){ | |
3269 u8 bOnce; /* True to run the ORDER BY search loop */ | |
3270 | |
3271 /* Skip over == and IS NULL terms */ | |
3272 if( j<pLoop->u.btree.nEq | |
3273 && pLoop->nSkip==0 | |
3274 && ((i = pLoop->aLTerm[j]->eOperator) & (WO_EQ|WO_ISNULL|WO_IS))!=0 | |
3275 ){ | |
3276 if( i & WO_ISNULL ){ | |
3277 testcase( isOrderDistinct ); | |
3278 isOrderDistinct = 0; | |
3279 } | |
3280 continue; | |
3281 } | |
3282 | |
3283 /* Get the column number in the table (iColumn) and sort order | |
3284 ** (revIdx) for the j-th column of the index. | |
3285 */ | |
3286 if( pIndex ){ | |
3287 iColumn = pIndex->aiColumn[j]; | |
3288 revIdx = pIndex->aSortOrder[j]; | |
3289 if( iColumn==pIndex->pTable->iPKey ) iColumn = -1; | |
3290 }else{ | |
3291 iColumn = XN_ROWID; | |
3292 revIdx = 0; | |
3293 } | |
3294 | |
3295 /* An unconstrained column that might be NULL means that this | |
3296 ** WhereLoop is not well-ordered | |
3297 */ | |
3298 if( isOrderDistinct | |
3299 && iColumn>=0 | |
3300 && j>=pLoop->u.btree.nEq | |
3301 && pIndex->pTable->aCol[iColumn].notNull==0 | |
3302 ){ | |
3303 isOrderDistinct = 0; | |
3304 } | |
3305 | |
3306 /* Find the ORDER BY term that corresponds to the j-th column | |
3307 ** of the index and mark that ORDER BY term off | |
3308 */ | |
3309 bOnce = 1; | |
3310 isMatch = 0; | |
3311 for(i=0; bOnce && i<nOrderBy; i++){ | |
3312 if( MASKBIT(i) & obSat ) continue; | |
3313 pOBExpr = sqlite3ExprSkipCollate(pOrderBy->a[i].pExpr); | |
3314 testcase( wctrlFlags & WHERE_GROUPBY ); | |
3315 testcase( wctrlFlags & WHERE_DISTINCTBY ); | |
3316 if( (wctrlFlags & (WHERE_GROUPBY|WHERE_DISTINCTBY))==0 ) bOnce = 0; | |
3317 if( iColumn>=(-1) ){ | |
3318 if( pOBExpr->op!=TK_COLUMN ) continue; | |
3319 if( pOBExpr->iTable!=iCur ) continue; | |
3320 if( pOBExpr->iColumn!=iColumn ) continue; | |
3321 }else{ | |
3322 if( sqlite3ExprCompare(pOBExpr,pIndex->aColExpr->a[j].pExpr,iCur) ){ | |
3323 continue; | |
3324 } | |
3325 } | |
3326 if( iColumn>=0 ){ | |
3327 pColl = sqlite3ExprCollSeq(pWInfo->pParse, pOrderBy->a[i].pExpr); | |
3328 if( !pColl ) pColl = db->pDfltColl; | |
3329 if( sqlite3StrICmp(pColl->zName, pIndex->azColl[j])!=0 ) continue; | |
3330 } | |
3331 isMatch = 1; | |
3332 break; | |
3333 } | |
3334 if( isMatch && (wctrlFlags & WHERE_GROUPBY)==0 ){ | |
3335 /* Make sure the sort order is compatible in an ORDER BY clause. | |
3336 ** Sort order is irrelevant for a GROUP BY clause. */ | |
3337 if( revSet ){ | |
3338 if( (rev ^ revIdx)!=pOrderBy->a[i].sortOrder ) isMatch = 0; | |
3339 }else{ | |
3340 rev = revIdx ^ pOrderBy->a[i].sortOrder; | |
3341 if( rev ) *pRevMask |= MASKBIT(iLoop); | |
3342 revSet = 1; | |
3343 } | |
3344 } | |
3345 if( isMatch ){ | |
3346 if( iColumn<0 ){ | |
3347 testcase( distinctColumns==0 ); | |
3348 distinctColumns = 1; | |
3349 } | |
3350 obSat |= MASKBIT(i); | |
3351 }else{ | |
3352 /* No match found */ | |
3353 if( j==0 || j<nKeyCol ){ | |
3354 testcase( isOrderDistinct!=0 ); | |
3355 isOrderDistinct = 0; | |
3356 } | |
3357 break; | |
3358 } | |
3359 } /* end Loop over all index columns */ | |
3360 if( distinctColumns ){ | |
3361 testcase( isOrderDistinct==0 ); | |
3362 isOrderDistinct = 1; | |
3363 } | |
3364 } /* end-if not one-row */ | |
3365 | |
3366 /* Mark off any other ORDER BY terms that reference pLoop */ | |
3367 if( isOrderDistinct ){ | |
3368 orderDistinctMask |= pLoop->maskSelf; | |
3369 for(i=0; i<nOrderBy; i++){ | |
3370 Expr *p; | |
3371 Bitmask mTerm; | |
3372 if( MASKBIT(i) & obSat ) continue; | |
3373 p = pOrderBy->a[i].pExpr; | |
3374 mTerm = sqlite3WhereExprUsage(&pWInfo->sMaskSet,p); | |
3375 if( mTerm==0 && !sqlite3ExprIsConstant(p) ) continue; | |
3376 if( (mTerm&~orderDistinctMask)==0 ){ | |
3377 obSat |= MASKBIT(i); | |
3378 } | |
3379 } | |
3380 } | |
3381 } /* End the loop over all WhereLoops from outer-most down to inner-most */ | |
3382 if( obSat==obDone ) return (i8)nOrderBy; | |
3383 if( !isOrderDistinct ){ | |
3384 for(i=nOrderBy-1; i>0; i--){ | |
3385 Bitmask m = MASKBIT(i) - 1; | |
3386 if( (obSat&m)==m ) return i; | |
3387 } | |
3388 return 0; | |
3389 } | |
3390 return -1; | |
3391 } | |
3392 | |
3393 | |
3394 /* | |
3395 ** If the WHERE_GROUPBY flag is set in the mask passed to sqlite3WhereBegin(), | |
3396 ** the planner assumes that the specified pOrderBy list is actually a GROUP | |
3397 ** BY clause - and so any order that groups rows as required satisfies the | |
3398 ** request. | |
3399 ** | |
3400 ** Normally, in this case it is not possible for the caller to determine | |
3401 ** whether or not the rows are really being delivered in sorted order, or | |
3402 ** just in some other order that provides the required grouping. However, | |
3403 ** if the WHERE_SORTBYGROUP flag is also passed to sqlite3WhereBegin(), then | |
3404 ** this function may be called on the returned WhereInfo object. It returns | |
3405 ** true if the rows really will be sorted in the specified order, or false | |
3406 ** otherwise. | |
3407 ** | |
3408 ** For example, assuming: | |
3409 ** | |
3410 ** CREATE INDEX i1 ON t1(x, Y); | |
3411 ** | |
3412 ** then | |
3413 ** | |
3414 ** SELECT * FROM t1 GROUP BY x,y ORDER BY x,y; -- IsSorted()==1 | |
3415 ** SELECT * FROM t1 GROUP BY y,x ORDER BY y,x; -- IsSorted()==0 | |
3416 */ | |
3417 int sqlite3WhereIsSorted(WhereInfo *pWInfo){ | |
3418 assert( pWInfo->wctrlFlags & WHERE_GROUPBY ); | |
3419 assert( pWInfo->wctrlFlags & WHERE_SORTBYGROUP ); | |
3420 return pWInfo->sorted; | |
3421 } | |
3422 | |
3423 #ifdef WHERETRACE_ENABLED | |
3424 /* For debugging use only: */ | |
3425 static const char *wherePathName(WherePath *pPath, int nLoop, WhereLoop *pLast){ | |
3426 static char zName[65]; | |
3427 int i; | |
3428 for(i=0; i<nLoop; i++){ zName[i] = pPath->aLoop[i]->cId; } | |
3429 if( pLast ) zName[i++] = pLast->cId; | |
3430 zName[i] = 0; | |
3431 return zName; | |
3432 } | |
3433 #endif | |
3434 | |
3435 /* | |
3436 ** Return the cost of sorting nRow rows, assuming that the keys have | |
3437 ** nOrderby columns and that the first nSorted columns are already in | |
3438 ** order. | |
3439 */ | |
3440 static LogEst whereSortingCost( | |
3441 WhereInfo *pWInfo, | |
3442 LogEst nRow, | |
3443 int nOrderBy, | |
3444 int nSorted | |
3445 ){ | |
3446 /* TUNING: Estimated cost of a full external sort, where N is | |
3447 ** the number of rows to sort is: | |
3448 ** | |
3449 ** cost = (3.0 * N * log(N)). | |
3450 ** | |
3451 ** Or, if the order-by clause has X terms but only the last Y | |
3452 ** terms are out of order, then block-sorting will reduce the | |
3453 ** sorting cost to: | |
3454 ** | |
3455 ** cost = (3.0 * N * log(N)) * (Y/X) | |
3456 ** | |
3457 ** The (Y/X) term is implemented using stack variable rScale | |
3458 ** below. */ | |
3459 LogEst rScale, rSortCost; | |
3460 assert( nOrderBy>0 && 66==sqlite3LogEst(100) ); | |
3461 rScale = sqlite3LogEst((nOrderBy-nSorted)*100/nOrderBy) - 66; | |
3462 rSortCost = nRow + estLog(nRow) + rScale + 16; | |
3463 | |
3464 /* TUNING: The cost of implementing DISTINCT using a B-TREE is | |
3465 ** similar but with a larger constant of proportionality. | |
3466 ** Multiply by an additional factor of 3.0. */ | |
3467 if( pWInfo->wctrlFlags & WHERE_WANT_DISTINCT ){ | |
3468 rSortCost += 16; | |
3469 } | |
3470 | |
3471 return rSortCost; | |
3472 } | |
3473 | |
3474 /* | |
3475 ** Given the list of WhereLoop objects at pWInfo->pLoops, this routine | |
3476 ** attempts to find the lowest cost path that visits each WhereLoop | |
3477 ** once. This path is then loaded into the pWInfo->a[].pWLoop fields. | |
3478 ** | |
3479 ** Assume that the total number of output rows that will need to be sorted | |
3480 ** will be nRowEst (in the 10*log2 representation). Or, ignore sorting | |
3481 ** costs if nRowEst==0. | |
3482 ** | |
3483 ** Return SQLITE_OK on success or SQLITE_NOMEM of a memory allocation | |
3484 ** error occurs. | |
3485 */ | |
3486 static int wherePathSolver(WhereInfo *pWInfo, LogEst nRowEst){ | |
3487 int mxChoice; /* Maximum number of simultaneous paths tracked */ | |
3488 int nLoop; /* Number of terms in the join */ | |
3489 Parse *pParse; /* Parsing context */ | |
3490 sqlite3 *db; /* The database connection */ | |
3491 int iLoop; /* Loop counter over the terms of the join */ | |
3492 int ii, jj; /* Loop counters */ | |
3493 int mxI = 0; /* Index of next entry to replace */ | |
3494 int nOrderBy; /* Number of ORDER BY clause terms */ | |
3495 LogEst mxCost = 0; /* Maximum cost of a set of paths */ | |
3496 LogEst mxUnsorted = 0; /* Maximum unsorted cost of a set of path */ | |
3497 int nTo, nFrom; /* Number of valid entries in aTo[] and aFrom[] */ | |
3498 WherePath *aFrom; /* All nFrom paths at the previous level */ | |
3499 WherePath *aTo; /* The nTo best paths at the current level */ | |
3500 WherePath *pFrom; /* An element of aFrom[] that we are working on */ | |
3501 WherePath *pTo; /* An element of aTo[] that we are working on */ | |
3502 WhereLoop *pWLoop; /* One of the WhereLoop objects */ | |
3503 WhereLoop **pX; /* Used to divy up the pSpace memory */ | |
3504 LogEst *aSortCost = 0; /* Sorting and partial sorting costs */ | |
3505 char *pSpace; /* Temporary memory used by this routine */ | |
3506 int nSpace; /* Bytes of space allocated at pSpace */ | |
3507 | |
3508 pParse = pWInfo->pParse; | |
3509 db = pParse->db; | |
3510 nLoop = pWInfo->nLevel; | |
3511 /* TUNING: For simple queries, only the best path is tracked. | |
3512 ** For 2-way joins, the 5 best paths are followed. | |
3513 ** For joins of 3 or more tables, track the 10 best paths */ | |
3514 mxChoice = (nLoop<=1) ? 1 : (nLoop==2 ? 5 : 10); | |
3515 assert( nLoop<=pWInfo->pTabList->nSrc ); | |
3516 WHERETRACE(0x002, ("---- begin solver. (nRowEst=%d)\n", nRowEst)); | |
3517 | |
3518 /* If nRowEst is zero and there is an ORDER BY clause, ignore it. In this | |
3519 ** case the purpose of this call is to estimate the number of rows returned | |
3520 ** by the overall query. Once this estimate has been obtained, the caller | |
3521 ** will invoke this function a second time, passing the estimate as the | |
3522 ** nRowEst parameter. */ | |
3523 if( pWInfo->pOrderBy==0 || nRowEst==0 ){ | |
3524 nOrderBy = 0; | |
3525 }else{ | |
3526 nOrderBy = pWInfo->pOrderBy->nExpr; | |
3527 } | |
3528 | |
3529 /* Allocate and initialize space for aTo, aFrom and aSortCost[] */ | |
3530 nSpace = (sizeof(WherePath)+sizeof(WhereLoop*)*nLoop)*mxChoice*2; | |
3531 nSpace += sizeof(LogEst) * nOrderBy; | |
3532 pSpace = sqlite3DbMallocRaw(db, nSpace); | |
3533 if( pSpace==0 ) return SQLITE_NOMEM; | |
3534 aTo = (WherePath*)pSpace; | |
3535 aFrom = aTo+mxChoice; | |
3536 memset(aFrom, 0, sizeof(aFrom[0])); | |
3537 pX = (WhereLoop**)(aFrom+mxChoice); | |
3538 for(ii=mxChoice*2, pFrom=aTo; ii>0; ii--, pFrom++, pX += nLoop){ | |
3539 pFrom->aLoop = pX; | |
3540 } | |
3541 if( nOrderBy ){ | |
3542 /* If there is an ORDER BY clause and it is not being ignored, set up | |
3543 ** space for the aSortCost[] array. Each element of the aSortCost array | |
3544 ** is either zero - meaning it has not yet been initialized - or the | |
3545 ** cost of sorting nRowEst rows of data where the first X terms of | |
3546 ** the ORDER BY clause are already in order, where X is the array | |
3547 ** index. */ | |
3548 aSortCost = (LogEst*)pX; | |
3549 memset(aSortCost, 0, sizeof(LogEst) * nOrderBy); | |
3550 } | |
3551 assert( aSortCost==0 || &pSpace[nSpace]==(char*)&aSortCost[nOrderBy] ); | |
3552 assert( aSortCost!=0 || &pSpace[nSpace]==(char*)pX ); | |
3553 | |
3554 /* Seed the search with a single WherePath containing zero WhereLoops. | |
3555 ** | |
3556 ** TUNING: Do not let the number of iterations go above 28. If the cost | |
3557 ** of computing an automatic index is not paid back within the first 28 | |
3558 ** rows, then do not use the automatic index. */ | |
3559 aFrom[0].nRow = MIN(pParse->nQueryLoop, 48); assert( 48==sqlite3LogEst(28) ); | |
3560 nFrom = 1; | |
3561 assert( aFrom[0].isOrdered==0 ); | |
3562 if( nOrderBy ){ | |
3563 /* If nLoop is zero, then there are no FROM terms in the query. Since | |
3564 ** in this case the query may return a maximum of one row, the results | |
3565 ** are already in the requested order. Set isOrdered to nOrderBy to | |
3566 ** indicate this. Or, if nLoop is greater than zero, set isOrdered to | |
3567 ** -1, indicating that the result set may or may not be ordered, | |
3568 ** depending on the loops added to the current plan. */ | |
3569 aFrom[0].isOrdered = nLoop>0 ? -1 : nOrderBy; | |
3570 } | |
3571 | |
3572 /* Compute successively longer WherePaths using the previous generation | |
3573 ** of WherePaths as the basis for the next. Keep track of the mxChoice | |
3574 ** best paths at each generation */ | |
3575 for(iLoop=0; iLoop<nLoop; iLoop++){ | |
3576 nTo = 0; | |
3577 for(ii=0, pFrom=aFrom; ii<nFrom; ii++, pFrom++){ | |
3578 for(pWLoop=pWInfo->pLoops; pWLoop; pWLoop=pWLoop->pNextLoop){ | |
3579 LogEst nOut; /* Rows visited by (pFrom+pWLoop) */ | |
3580 LogEst rCost; /* Cost of path (pFrom+pWLoop) */ | |
3581 LogEst rUnsorted; /* Unsorted cost of (pFrom+pWLoop) */ | |
3582 i8 isOrdered = pFrom->isOrdered; /* isOrdered for (pFrom+pWLoop) */ | |
3583 Bitmask maskNew; /* Mask of src visited by (..) */ | |
3584 Bitmask revMask = 0; /* Mask of rev-order loops for (..) */ | |
3585 | |
3586 if( (pWLoop->prereq & ~pFrom->maskLoop)!=0 ) continue; | |
3587 if( (pWLoop->maskSelf & pFrom->maskLoop)!=0 ) continue; | |
3588 /* At this point, pWLoop is a candidate to be the next loop. | |
3589 ** Compute its cost */ | |
3590 rUnsorted = sqlite3LogEstAdd(pWLoop->rSetup,pWLoop->rRun + pFrom->nRow); | |
3591 rUnsorted = sqlite3LogEstAdd(rUnsorted, pFrom->rUnsorted); | |
3592 nOut = pFrom->nRow + pWLoop->nOut; | |
3593 maskNew = pFrom->maskLoop | pWLoop->maskSelf; | |
3594 if( isOrdered<0 ){ | |
3595 isOrdered = wherePathSatisfiesOrderBy(pWInfo, | |
3596 pWInfo->pOrderBy, pFrom, pWInfo->wctrlFlags, | |
3597 iLoop, pWLoop, &revMask); | |
3598 }else{ | |
3599 revMask = pFrom->revLoop; | |
3600 } | |
3601 if( isOrdered>=0 && isOrdered<nOrderBy ){ | |
3602 if( aSortCost[isOrdered]==0 ){ | |
3603 aSortCost[isOrdered] = whereSortingCost( | |
3604 pWInfo, nRowEst, nOrderBy, isOrdered | |
3605 ); | |
3606 } | |
3607 rCost = sqlite3LogEstAdd(rUnsorted, aSortCost[isOrdered]); | |
3608 | |
3609 WHERETRACE(0x002, | |
3610 ("---- sort cost=%-3d (%d/%d) increases cost %3d to %-3d\n", | |
3611 aSortCost[isOrdered], (nOrderBy-isOrdered), nOrderBy, | |
3612 rUnsorted, rCost)); | |
3613 }else{ | |
3614 rCost = rUnsorted; | |
3615 } | |
3616 | |
3617 /* Check to see if pWLoop should be added to the set of | |
3618 ** mxChoice best-so-far paths. | |
3619 ** | |
3620 ** First look for an existing path among best-so-far paths | |
3621 ** that covers the same set of loops and has the same isOrdered | |
3622 ** setting as the current path candidate. | |
3623 ** | |
3624 ** The term "((pTo->isOrdered^isOrdered)&0x80)==0" is equivalent | |
3625 ** to (pTo->isOrdered==(-1))==(isOrdered==(-1))" for the range | |
3626 ** of legal values for isOrdered, -1..64. | |
3627 */ | |
3628 for(jj=0, pTo=aTo; jj<nTo; jj++, pTo++){ | |
3629 if( pTo->maskLoop==maskNew | |
3630 && ((pTo->isOrdered^isOrdered)&0x80)==0 | |
3631 ){ | |
3632 testcase( jj==nTo-1 ); | |
3633 break; | |
3634 } | |
3635 } | |
3636 if( jj>=nTo ){ | |
3637 /* None of the existing best-so-far paths match the candidate. */ | |
3638 if( nTo>=mxChoice | |
3639 && (rCost>mxCost || (rCost==mxCost && rUnsorted>=mxUnsorted)) | |
3640 ){ | |
3641 /* The current candidate is no better than any of the mxChoice | |
3642 ** paths currently in the best-so-far buffer. So discard | |
3643 ** this candidate as not viable. */ | |
3644 #ifdef WHERETRACE_ENABLED /* 0x4 */ | |
3645 if( sqlite3WhereTrace&0x4 ){ | |
3646 sqlite3DebugPrintf("Skip %s cost=%-3d,%3d order=%c\n", | |
3647 wherePathName(pFrom, iLoop, pWLoop), rCost, nOut, | |
3648 isOrdered>=0 ? isOrdered+'0' : '?'); | |
3649 } | |
3650 #endif | |
3651 continue; | |
3652 } | |
3653 /* If we reach this points it means that the new candidate path | |
3654 ** needs to be added to the set of best-so-far paths. */ | |
3655 if( nTo<mxChoice ){ | |
3656 /* Increase the size of the aTo set by one */ | |
3657 jj = nTo++; | |
3658 }else{ | |
3659 /* New path replaces the prior worst to keep count below mxChoice */ | |
3660 jj = mxI; | |
3661 } | |
3662 pTo = &aTo[jj]; | |
3663 #ifdef WHERETRACE_ENABLED /* 0x4 */ | |
3664 if( sqlite3WhereTrace&0x4 ){ | |
3665 sqlite3DebugPrintf("New %s cost=%-3d,%3d order=%c\n", | |
3666 wherePathName(pFrom, iLoop, pWLoop), rCost, nOut, | |
3667 isOrdered>=0 ? isOrdered+'0' : '?'); | |
3668 } | |
3669 #endif | |
3670 }else{ | |
3671 /* Control reaches here if best-so-far path pTo=aTo[jj] covers the | |
3672 ** same set of loops and has the sam isOrdered setting as the | |
3673 ** candidate path. Check to see if the candidate should replace | |
3674 ** pTo or if the candidate should be skipped */ | |
3675 if( pTo->rCost<rCost || (pTo->rCost==rCost && pTo->nRow<=nOut) ){ | |
3676 #ifdef WHERETRACE_ENABLED /* 0x4 */ | |
3677 if( sqlite3WhereTrace&0x4 ){ | |
3678 sqlite3DebugPrintf( | |
3679 "Skip %s cost=%-3d,%3d order=%c", | |
3680 wherePathName(pFrom, iLoop, pWLoop), rCost, nOut, | |
3681 isOrdered>=0 ? isOrdered+'0' : '?'); | |
3682 sqlite3DebugPrintf(" vs %s cost=%-3d,%d order=%c\n", | |
3683 wherePathName(pTo, iLoop+1, 0), pTo->rCost, pTo->nRow, | |
3684 pTo->isOrdered>=0 ? pTo->isOrdered+'0' : '?'); | |
3685 } | |
3686 #endif | |
3687 /* Discard the candidate path from further consideration */ | |
3688 testcase( pTo->rCost==rCost ); | |
3689 continue; | |
3690 } | |
3691 testcase( pTo->rCost==rCost+1 ); | |
3692 /* Control reaches here if the candidate path is better than the | |
3693 ** pTo path. Replace pTo with the candidate. */ | |
3694 #ifdef WHERETRACE_ENABLED /* 0x4 */ | |
3695 if( sqlite3WhereTrace&0x4 ){ | |
3696 sqlite3DebugPrintf( | |
3697 "Update %s cost=%-3d,%3d order=%c", | |
3698 wherePathName(pFrom, iLoop, pWLoop), rCost, nOut, | |
3699 isOrdered>=0 ? isOrdered+'0' : '?'); | |
3700 sqlite3DebugPrintf(" was %s cost=%-3d,%3d order=%c\n", | |
3701 wherePathName(pTo, iLoop+1, 0), pTo->rCost, pTo->nRow, | |
3702 pTo->isOrdered>=0 ? pTo->isOrdered+'0' : '?'); | |
3703 } | |
3704 #endif | |
3705 } | |
3706 /* pWLoop is a winner. Add it to the set of best so far */ | |
3707 pTo->maskLoop = pFrom->maskLoop | pWLoop->maskSelf; | |
3708 pTo->revLoop = revMask; | |
3709 pTo->nRow = nOut; | |
3710 pTo->rCost = rCost; | |
3711 pTo->rUnsorted = rUnsorted; | |
3712 pTo->isOrdered = isOrdered; | |
3713 memcpy(pTo->aLoop, pFrom->aLoop, sizeof(WhereLoop*)*iLoop); | |
3714 pTo->aLoop[iLoop] = pWLoop; | |
3715 if( nTo>=mxChoice ){ | |
3716 mxI = 0; | |
3717 mxCost = aTo[0].rCost; | |
3718 mxUnsorted = aTo[0].nRow; | |
3719 for(jj=1, pTo=&aTo[1]; jj<mxChoice; jj++, pTo++){ | |
3720 if( pTo->rCost>mxCost | |
3721 || (pTo->rCost==mxCost && pTo->rUnsorted>mxUnsorted) | |
3722 ){ | |
3723 mxCost = pTo->rCost; | |
3724 mxUnsorted = pTo->rUnsorted; | |
3725 mxI = jj; | |
3726 } | |
3727 } | |
3728 } | |
3729 } | |
3730 } | |
3731 | |
3732 #ifdef WHERETRACE_ENABLED /* >=2 */ | |
3733 if( sqlite3WhereTrace & 0x02 ){ | |
3734 sqlite3DebugPrintf("---- after round %d ----\n", iLoop); | |
3735 for(ii=0, pTo=aTo; ii<nTo; ii++, pTo++){ | |
3736 sqlite3DebugPrintf(" %s cost=%-3d nrow=%-3d order=%c", | |
3737 wherePathName(pTo, iLoop+1, 0), pTo->rCost, pTo->nRow, | |
3738 pTo->isOrdered>=0 ? (pTo->isOrdered+'0') : '?'); | |
3739 if( pTo->isOrdered>0 ){ | |
3740 sqlite3DebugPrintf(" rev=0x%llx\n", pTo->revLoop); | |
3741 }else{ | |
3742 sqlite3DebugPrintf("\n"); | |
3743 } | |
3744 } | |
3745 } | |
3746 #endif | |
3747 | |
3748 /* Swap the roles of aFrom and aTo for the next generation */ | |
3749 pFrom = aTo; | |
3750 aTo = aFrom; | |
3751 aFrom = pFrom; | |
3752 nFrom = nTo; | |
3753 } | |
3754 | |
3755 if( nFrom==0 ){ | |
3756 sqlite3ErrorMsg(pParse, "no query solution"); | |
3757 sqlite3DbFree(db, pSpace); | |
3758 return SQLITE_ERROR; | |
3759 } | |
3760 | |
3761 /* Find the lowest cost path. pFrom will be left pointing to that path */ | |
3762 pFrom = aFrom; | |
3763 for(ii=1; ii<nFrom; ii++){ | |
3764 if( pFrom->rCost>aFrom[ii].rCost ) pFrom = &aFrom[ii]; | |
3765 } | |
3766 assert( pWInfo->nLevel==nLoop ); | |
3767 /* Load the lowest cost path into pWInfo */ | |
3768 for(iLoop=0; iLoop<nLoop; iLoop++){ | |
3769 WhereLevel *pLevel = pWInfo->a + iLoop; | |
3770 pLevel->pWLoop = pWLoop = pFrom->aLoop[iLoop]; | |
3771 pLevel->iFrom = pWLoop->iTab; | |
3772 pLevel->iTabCur = pWInfo->pTabList->a[pLevel->iFrom].iCursor; | |
3773 } | |
3774 if( (pWInfo->wctrlFlags & WHERE_WANT_DISTINCT)!=0 | |
3775 && (pWInfo->wctrlFlags & WHERE_DISTINCTBY)==0 | |
3776 && pWInfo->eDistinct==WHERE_DISTINCT_NOOP | |
3777 && nRowEst | |
3778 ){ | |
3779 Bitmask notUsed; | |
3780 int rc = wherePathSatisfiesOrderBy(pWInfo, pWInfo->pResultSet, pFrom, | |
3781 WHERE_DISTINCTBY, nLoop-1, pFrom->aLoop[nLoop-1], ¬Used); | |
3782 if( rc==pWInfo->pResultSet->nExpr ){ | |
3783 pWInfo->eDistinct = WHERE_DISTINCT_ORDERED; | |
3784 } | |
3785 } | |
3786 if( pWInfo->pOrderBy ){ | |
3787 if( pWInfo->wctrlFlags & WHERE_DISTINCTBY ){ | |
3788 if( pFrom->isOrdered==pWInfo->pOrderBy->nExpr ){ | |
3789 pWInfo->eDistinct = WHERE_DISTINCT_ORDERED; | |
3790 } | |
3791 }else{ | |
3792 pWInfo->nOBSat = pFrom->isOrdered; | |
3793 if( pWInfo->nOBSat<0 ) pWInfo->nOBSat = 0; | |
3794 pWInfo->revMask = pFrom->revLoop; | |
3795 } | |
3796 if( (pWInfo->wctrlFlags & WHERE_SORTBYGROUP) | |
3797 && pWInfo->nOBSat==pWInfo->pOrderBy->nExpr && nLoop>0 | |
3798 ){ | |
3799 Bitmask revMask = 0; | |
3800 int nOrder = wherePathSatisfiesOrderBy(pWInfo, pWInfo->pOrderBy, | |
3801 pFrom, 0, nLoop-1, pFrom->aLoop[nLoop-1], &revMask | |
3802 ); | |
3803 assert( pWInfo->sorted==0 ); | |
3804 if( nOrder==pWInfo->pOrderBy->nExpr ){ | |
3805 pWInfo->sorted = 1; | |
3806 pWInfo->revMask = revMask; | |
3807 } | |
3808 } | |
3809 } | |
3810 | |
3811 | |
3812 pWInfo->nRowOut = pFrom->nRow; | |
3813 | |
3814 /* Free temporary memory and return success */ | |
3815 sqlite3DbFree(db, pSpace); | |
3816 return SQLITE_OK; | |
3817 } | |
3818 | |
3819 /* | |
3820 ** Most queries use only a single table (they are not joins) and have | |
3821 ** simple == constraints against indexed fields. This routine attempts | |
3822 ** to plan those simple cases using much less ceremony than the | |
3823 ** general-purpose query planner, and thereby yield faster sqlite3_prepare() | |
3824 ** times for the common case. | |
3825 ** | |
3826 ** Return non-zero on success, if this query can be handled by this | |
3827 ** no-frills query planner. Return zero if this query needs the | |
3828 ** general-purpose query planner. | |
3829 */ | |
3830 static int whereShortCut(WhereLoopBuilder *pBuilder){ | |
3831 WhereInfo *pWInfo; | |
3832 struct SrcList_item *pItem; | |
3833 WhereClause *pWC; | |
3834 WhereTerm *pTerm; | |
3835 WhereLoop *pLoop; | |
3836 int iCur; | |
3837 int j; | |
3838 Table *pTab; | |
3839 Index *pIdx; | |
3840 | |
3841 pWInfo = pBuilder->pWInfo; | |
3842 if( pWInfo->wctrlFlags & WHERE_FORCE_TABLE ) return 0; | |
3843 assert( pWInfo->pTabList->nSrc>=1 ); | |
3844 pItem = pWInfo->pTabList->a; | |
3845 pTab = pItem->pTab; | |
3846 if( IsVirtual(pTab) ) return 0; | |
3847 if( pItem->fg.isIndexedBy ) return 0; | |
3848 iCur = pItem->iCursor; | |
3849 pWC = &pWInfo->sWC; | |
3850 pLoop = pBuilder->pNew; | |
3851 pLoop->wsFlags = 0; | |
3852 pLoop->nSkip = 0; | |
3853 pTerm = sqlite3WhereFindTerm(pWC, iCur, -1, 0, WO_EQ|WO_IS, 0); | |
3854 if( pTerm ){ | |
3855 testcase( pTerm->eOperator & WO_IS ); | |
3856 pLoop->wsFlags = WHERE_COLUMN_EQ|WHERE_IPK|WHERE_ONEROW; | |
3857 pLoop->aLTerm[0] = pTerm; | |
3858 pLoop->nLTerm = 1; | |
3859 pLoop->u.btree.nEq = 1; | |
3860 /* TUNING: Cost of a rowid lookup is 10 */ | |
3861 pLoop->rRun = 33; /* 33==sqlite3LogEst(10) */ | |
3862 }else{ | |
3863 for(pIdx=pTab->pIndex; pIdx; pIdx=pIdx->pNext){ | |
3864 int opMask; | |
3865 assert( pLoop->aLTermSpace==pLoop->aLTerm ); | |
3866 if( !IsUniqueIndex(pIdx) | |
3867 || pIdx->pPartIdxWhere!=0 | |
3868 || pIdx->nKeyCol>ArraySize(pLoop->aLTermSpace) | |
3869 ) continue; | |
3870 opMask = pIdx->uniqNotNull ? (WO_EQ|WO_IS) : WO_EQ; | |
3871 for(j=0; j<pIdx->nKeyCol; j++){ | |
3872 pTerm = sqlite3WhereFindTerm(pWC, iCur, j, 0, opMask, pIdx); | |
3873 if( pTerm==0 ) break; | |
3874 testcase( pTerm->eOperator & WO_IS ); | |
3875 pLoop->aLTerm[j] = pTerm; | |
3876 } | |
3877 if( j!=pIdx->nKeyCol ) continue; | |
3878 pLoop->wsFlags = WHERE_COLUMN_EQ|WHERE_ONEROW|WHERE_INDEXED; | |
3879 if( pIdx->isCovering || (pItem->colUsed & ~columnsInIndex(pIdx))==0 ){ | |
3880 pLoop->wsFlags |= WHERE_IDX_ONLY; | |
3881 } | |
3882 pLoop->nLTerm = j; | |
3883 pLoop->u.btree.nEq = j; | |
3884 pLoop->u.btree.pIndex = pIdx; | |
3885 /* TUNING: Cost of a unique index lookup is 15 */ | |
3886 pLoop->rRun = 39; /* 39==sqlite3LogEst(15) */ | |
3887 break; | |
3888 } | |
3889 } | |
3890 if( pLoop->wsFlags ){ | |
3891 pLoop->nOut = (LogEst)1; | |
3892 pWInfo->a[0].pWLoop = pLoop; | |
3893 pLoop->maskSelf = sqlite3WhereGetMask(&pWInfo->sMaskSet, iCur); | |
3894 pWInfo->a[0].iTabCur = iCur; | |
3895 pWInfo->nRowOut = 1; | |
3896 if( pWInfo->pOrderBy ) pWInfo->nOBSat = pWInfo->pOrderBy->nExpr; | |
3897 if( pWInfo->wctrlFlags & WHERE_WANT_DISTINCT ){ | |
3898 pWInfo->eDistinct = WHERE_DISTINCT_UNIQUE; | |
3899 } | |
3900 #ifdef SQLITE_DEBUG | |
3901 pLoop->cId = '0'; | |
3902 #endif | |
3903 return 1; | |
3904 } | |
3905 return 0; | |
3906 } | |
3907 | |
3908 /* | |
3909 ** Generate the beginning of the loop used for WHERE clause processing. | |
3910 ** The return value is a pointer to an opaque structure that contains | |
3911 ** information needed to terminate the loop. Later, the calling routine | |
3912 ** should invoke sqlite3WhereEnd() with the return value of this function | |
3913 ** in order to complete the WHERE clause processing. | |
3914 ** | |
3915 ** If an error occurs, this routine returns NULL. | |
3916 ** | |
3917 ** The basic idea is to do a nested loop, one loop for each table in | |
3918 ** the FROM clause of a select. (INSERT and UPDATE statements are the | |
3919 ** same as a SELECT with only a single table in the FROM clause.) For | |
3920 ** example, if the SQL is this: | |
3921 ** | |
3922 ** SELECT * FROM t1, t2, t3 WHERE ...; | |
3923 ** | |
3924 ** Then the code generated is conceptually like the following: | |
3925 ** | |
3926 ** foreach row1 in t1 do \ Code generated | |
3927 ** foreach row2 in t2 do |-- by sqlite3WhereBegin() | |
3928 ** foreach row3 in t3 do / | |
3929 ** ... | |
3930 ** end \ Code generated | |
3931 ** end |-- by sqlite3WhereEnd() | |
3932 ** end / | |
3933 ** | |
3934 ** Note that the loops might not be nested in the order in which they | |
3935 ** appear in the FROM clause if a different order is better able to make | |
3936 ** use of indices. Note also that when the IN operator appears in | |
3937 ** the WHERE clause, it might result in additional nested loops for | |
3938 ** scanning through all values on the right-hand side of the IN. | |
3939 ** | |
3940 ** There are Btree cursors associated with each table. t1 uses cursor | |
3941 ** number pTabList->a[0].iCursor. t2 uses the cursor pTabList->a[1].iCursor. | |
3942 ** And so forth. This routine generates code to open those VDBE cursors | |
3943 ** and sqlite3WhereEnd() generates the code to close them. | |
3944 ** | |
3945 ** The code that sqlite3WhereBegin() generates leaves the cursors named | |
3946 ** in pTabList pointing at their appropriate entries. The [...] code | |
3947 ** can use OP_Column and OP_Rowid opcodes on these cursors to extract | |
3948 ** data from the various tables of the loop. | |
3949 ** | |
3950 ** If the WHERE clause is empty, the foreach loops must each scan their | |
3951 ** entire tables. Thus a three-way join is an O(N^3) operation. But if | |
3952 ** the tables have indices and there are terms in the WHERE clause that | |
3953 ** refer to those indices, a complete table scan can be avoided and the | |
3954 ** code will run much faster. Most of the work of this routine is checking | |
3955 ** to see if there are indices that can be used to speed up the loop. | |
3956 ** | |
3957 ** Terms of the WHERE clause are also used to limit which rows actually | |
3958 ** make it to the "..." in the middle of the loop. After each "foreach", | |
3959 ** terms of the WHERE clause that use only terms in that loop and outer | |
3960 ** loops are evaluated and if false a jump is made around all subsequent | |
3961 ** inner loops (or around the "..." if the test occurs within the inner- | |
3962 ** most loop) | |
3963 ** | |
3964 ** OUTER JOINS | |
3965 ** | |
3966 ** An outer join of tables t1 and t2 is conceptally coded as follows: | |
3967 ** | |
3968 ** foreach row1 in t1 do | |
3969 ** flag = 0 | |
3970 ** foreach row2 in t2 do | |
3971 ** start: | |
3972 ** ... | |
3973 ** flag = 1 | |
3974 ** end | |
3975 ** if flag==0 then | |
3976 ** move the row2 cursor to a null row | |
3977 ** goto start | |
3978 ** fi | |
3979 ** end | |
3980 ** | |
3981 ** ORDER BY CLAUSE PROCESSING | |
3982 ** | |
3983 ** pOrderBy is a pointer to the ORDER BY clause (or the GROUP BY clause | |
3984 ** if the WHERE_GROUPBY flag is set in wctrlFlags) of a SELECT statement | |
3985 ** if there is one. If there is no ORDER BY clause or if this routine | |
3986 ** is called from an UPDATE or DELETE statement, then pOrderBy is NULL. | |
3987 ** | |
3988 ** The iIdxCur parameter is the cursor number of an index. If | |
3989 ** WHERE_ONETABLE_ONLY is set, iIdxCur is the cursor number of an index | |
3990 ** to use for OR clause processing. The WHERE clause should use this | |
3991 ** specific cursor. If WHERE_ONEPASS_DESIRED is set, then iIdxCur is | |
3992 ** the first cursor in an array of cursors for all indices. iIdxCur should | |
3993 ** be used to compute the appropriate cursor depending on which index is | |
3994 ** used. | |
3995 */ | |
3996 WhereInfo *sqlite3WhereBegin( | |
3997 Parse *pParse, /* The parser context */ | |
3998 SrcList *pTabList, /* FROM clause: A list of all tables to be scanned */ | |
3999 Expr *pWhere, /* The WHERE clause */ | |
4000 ExprList *pOrderBy, /* An ORDER BY (or GROUP BY) clause, or NULL */ | |
4001 ExprList *pResultSet, /* Result set of the query */ | |
4002 u16 wctrlFlags, /* One of the WHERE_* flags defined in sqliteInt.h */ | |
4003 int iIdxCur /* If WHERE_ONETABLE_ONLY is set, index cursor number */ | |
4004 ){ | |
4005 int nByteWInfo; /* Num. bytes allocated for WhereInfo struct */ | |
4006 int nTabList; /* Number of elements in pTabList */ | |
4007 WhereInfo *pWInfo; /* Will become the return value of this function */ | |
4008 Vdbe *v = pParse->pVdbe; /* The virtual database engine */ | |
4009 Bitmask notReady; /* Cursors that are not yet positioned */ | |
4010 WhereLoopBuilder sWLB; /* The WhereLoop builder */ | |
4011 WhereMaskSet *pMaskSet; /* The expression mask set */ | |
4012 WhereLevel *pLevel; /* A single level in pWInfo->a[] */ | |
4013 WhereLoop *pLoop; /* Pointer to a single WhereLoop object */ | |
4014 int ii; /* Loop counter */ | |
4015 sqlite3 *db; /* Database connection */ | |
4016 int rc; /* Return code */ | |
4017 u8 bFordelete = 0; | |
4018 | |
4019 assert( (wctrlFlags & WHERE_ONEPASS_MULTIROW)==0 || ( | |
4020 (wctrlFlags & WHERE_ONEPASS_DESIRED)!=0 | |
4021 && (wctrlFlags & WHERE_OMIT_OPEN_CLOSE)==0 | |
4022 )); | |
4023 | |
4024 /* Variable initialization */ | |
4025 db = pParse->db; | |
4026 memset(&sWLB, 0, sizeof(sWLB)); | |
4027 | |
4028 /* An ORDER/GROUP BY clause of more than 63 terms cannot be optimized */ | |
4029 testcase( pOrderBy && pOrderBy->nExpr==BMS-1 ); | |
4030 if( pOrderBy && pOrderBy->nExpr>=BMS ) pOrderBy = 0; | |
4031 sWLB.pOrderBy = pOrderBy; | |
4032 | |
4033 /* Disable the DISTINCT optimization if SQLITE_DistinctOpt is set via | |
4034 ** sqlite3_test_ctrl(SQLITE_TESTCTRL_OPTIMIZATIONS,...) */ | |
4035 if( OptimizationDisabled(db, SQLITE_DistinctOpt) ){ | |
4036 wctrlFlags &= ~WHERE_WANT_DISTINCT; | |
4037 } | |
4038 | |
4039 /* The number of tables in the FROM clause is limited by the number of | |
4040 ** bits in a Bitmask | |
4041 */ | |
4042 testcase( pTabList->nSrc==BMS ); | |
4043 if( pTabList->nSrc>BMS ){ | |
4044 sqlite3ErrorMsg(pParse, "at most %d tables in a join", BMS); | |
4045 return 0; | |
4046 } | |
4047 | |
4048 /* This function normally generates a nested loop for all tables in | |
4049 ** pTabList. But if the WHERE_ONETABLE_ONLY flag is set, then we should | |
4050 ** only generate code for the first table in pTabList and assume that | |
4051 ** any cursors associated with subsequent tables are uninitialized. | |
4052 */ | |
4053 nTabList = (wctrlFlags & WHERE_ONETABLE_ONLY) ? 1 : pTabList->nSrc; | |
4054 | |
4055 /* Allocate and initialize the WhereInfo structure that will become the | |
4056 ** return value. A single allocation is used to store the WhereInfo | |
4057 ** struct, the contents of WhereInfo.a[], the WhereClause structure | |
4058 ** and the WhereMaskSet structure. Since WhereClause contains an 8-byte | |
4059 ** field (type Bitmask) it must be aligned on an 8-byte boundary on | |
4060 ** some architectures. Hence the ROUND8() below. | |
4061 */ | |
4062 nByteWInfo = ROUND8(sizeof(WhereInfo)+(nTabList-1)*sizeof(WhereLevel)); | |
4063 pWInfo = sqlite3DbMallocZero(db, nByteWInfo + sizeof(WhereLoop)); | |
4064 if( db->mallocFailed ){ | |
4065 sqlite3DbFree(db, pWInfo); | |
4066 pWInfo = 0; | |
4067 goto whereBeginError; | |
4068 } | |
4069 pWInfo->aiCurOnePass[0] = pWInfo->aiCurOnePass[1] = -1; | |
4070 pWInfo->nLevel = nTabList; | |
4071 pWInfo->pParse = pParse; | |
4072 pWInfo->pTabList = pTabList; | |
4073 pWInfo->pOrderBy = pOrderBy; | |
4074 pWInfo->pResultSet = pResultSet; | |
4075 pWInfo->iBreak = pWInfo->iContinue = sqlite3VdbeMakeLabel(v); | |
4076 pWInfo->wctrlFlags = wctrlFlags; | |
4077 pWInfo->savedNQueryLoop = pParse->nQueryLoop; | |
4078 assert( pWInfo->eOnePass==ONEPASS_OFF ); /* ONEPASS defaults to OFF */ | |
4079 pMaskSet = &pWInfo->sMaskSet; | |
4080 sWLB.pWInfo = pWInfo; | |
4081 sWLB.pWC = &pWInfo->sWC; | |
4082 sWLB.pNew = (WhereLoop*)(((char*)pWInfo)+nByteWInfo); | |
4083 assert( EIGHT_BYTE_ALIGNMENT(sWLB.pNew) ); | |
4084 whereLoopInit(sWLB.pNew); | |
4085 #ifdef SQLITE_DEBUG | |
4086 sWLB.pNew->cId = '*'; | |
4087 #endif | |
4088 | |
4089 /* Split the WHERE clause into separate subexpressions where each | |
4090 ** subexpression is separated by an AND operator. | |
4091 */ | |
4092 initMaskSet(pMaskSet); | |
4093 sqlite3WhereClauseInit(&pWInfo->sWC, pWInfo); | |
4094 sqlite3WhereSplit(&pWInfo->sWC, pWhere, TK_AND); | |
4095 | |
4096 /* Special case: a WHERE clause that is constant. Evaluate the | |
4097 ** expression and either jump over all of the code or fall thru. | |
4098 */ | |
4099 for(ii=0; ii<sWLB.pWC->nTerm; ii++){ | |
4100 if( nTabList==0 || sqlite3ExprIsConstantNotJoin(sWLB.pWC->a[ii].pExpr) ){ | |
4101 sqlite3ExprIfFalse(pParse, sWLB.pWC->a[ii].pExpr, pWInfo->iBreak, | |
4102 SQLITE_JUMPIFNULL); | |
4103 sWLB.pWC->a[ii].wtFlags |= TERM_CODED; | |
4104 } | |
4105 } | |
4106 | |
4107 /* Special case: No FROM clause | |
4108 */ | |
4109 if( nTabList==0 ){ | |
4110 if( pOrderBy ) pWInfo->nOBSat = pOrderBy->nExpr; | |
4111 if( wctrlFlags & WHERE_WANT_DISTINCT ){ | |
4112 pWInfo->eDistinct = WHERE_DISTINCT_UNIQUE; | |
4113 } | |
4114 } | |
4115 | |
4116 /* Assign a bit from the bitmask to every term in the FROM clause. | |
4117 ** | |
4118 ** The N-th term of the FROM clause is assigned a bitmask of 1<<N. | |
4119 ** | |
4120 ** The rule of the previous sentence ensures thta if X is the bitmask for | |
4121 ** a table T, then X-1 is the bitmask for all other tables to the left of T. | |
4122 ** Knowing the bitmask for all tables to the left of a left join is | |
4123 ** important. Ticket #3015. | |
4124 ** | |
4125 ** Note that bitmasks are created for all pTabList->nSrc tables in | |
4126 ** pTabList, not just the first nTabList tables. nTabList is normally | |
4127 ** equal to pTabList->nSrc but might be shortened to 1 if the | |
4128 ** WHERE_ONETABLE_ONLY flag is set. | |
4129 */ | |
4130 for(ii=0; ii<pTabList->nSrc; ii++){ | |
4131 createMask(pMaskSet, pTabList->a[ii].iCursor); | |
4132 sqlite3WhereTabFuncArgs(pParse, &pTabList->a[ii], &pWInfo->sWC); | |
4133 } | |
4134 #ifdef SQLITE_DEBUG | |
4135 for(ii=0; ii<pTabList->nSrc; ii++){ | |
4136 Bitmask m = sqlite3WhereGetMask(pMaskSet, pTabList->a[ii].iCursor); | |
4137 assert( m==MASKBIT(ii) ); | |
4138 } | |
4139 #endif | |
4140 | |
4141 /* Analyze all of the subexpressions. */ | |
4142 sqlite3WhereExprAnalyze(pTabList, &pWInfo->sWC); | |
4143 if( db->mallocFailed ) goto whereBeginError; | |
4144 | |
4145 if( wctrlFlags & WHERE_WANT_DISTINCT ){ | |
4146 if( isDistinctRedundant(pParse, pTabList, &pWInfo->sWC, pResultSet) ){ | |
4147 /* The DISTINCT marking is pointless. Ignore it. */ | |
4148 pWInfo->eDistinct = WHERE_DISTINCT_UNIQUE; | |
4149 }else if( pOrderBy==0 ){ | |
4150 /* Try to ORDER BY the result set to make distinct processing easier */ | |
4151 pWInfo->wctrlFlags |= WHERE_DISTINCTBY; | |
4152 pWInfo->pOrderBy = pResultSet; | |
4153 } | |
4154 } | |
4155 | |
4156 /* Construct the WhereLoop objects */ | |
4157 WHERETRACE(0xffff,("*** Optimizer Start *** (wctrlFlags: 0x%x)\n", | |
4158 wctrlFlags)); | |
4159 #if defined(WHERETRACE_ENABLED) | |
4160 if( sqlite3WhereTrace & 0x100 ){ /* Display all terms of the WHERE clause */ | |
4161 int i; | |
4162 for(i=0; i<sWLB.pWC->nTerm; i++){ | |
4163 whereTermPrint(&sWLB.pWC->a[i], i); | |
4164 } | |
4165 } | |
4166 #endif | |
4167 | |
4168 if( nTabList!=1 || whereShortCut(&sWLB)==0 ){ | |
4169 rc = whereLoopAddAll(&sWLB); | |
4170 if( rc ) goto whereBeginError; | |
4171 | |
4172 #ifdef WHERETRACE_ENABLED | |
4173 if( sqlite3WhereTrace ){ /* Display all of the WhereLoop objects */ | |
4174 WhereLoop *p; | |
4175 int i; | |
4176 static const char zLabel[] = "0123456789abcdefghijklmnopqrstuvwyxz" | |
4177 "ABCDEFGHIJKLMNOPQRSTUVWYXZ"; | |
4178 for(p=pWInfo->pLoops, i=0; p; p=p->pNextLoop, i++){ | |
4179 p->cId = zLabel[i%sizeof(zLabel)]; | |
4180 whereLoopPrint(p, sWLB.pWC); | |
4181 } | |
4182 } | |
4183 #endif | |
4184 | |
4185 wherePathSolver(pWInfo, 0); | |
4186 if( db->mallocFailed ) goto whereBeginError; | |
4187 if( pWInfo->pOrderBy ){ | |
4188 wherePathSolver(pWInfo, pWInfo->nRowOut+1); | |
4189 if( db->mallocFailed ) goto whereBeginError; | |
4190 } | |
4191 } | |
4192 if( pWInfo->pOrderBy==0 && (db->flags & SQLITE_ReverseOrder)!=0 ){ | |
4193 pWInfo->revMask = (Bitmask)(-1); | |
4194 } | |
4195 if( pParse->nErr || NEVER(db->mallocFailed) ){ | |
4196 goto whereBeginError; | |
4197 } | |
4198 #ifdef WHERETRACE_ENABLED | |
4199 if( sqlite3WhereTrace ){ | |
4200 sqlite3DebugPrintf("---- Solution nRow=%d", pWInfo->nRowOut); | |
4201 if( pWInfo->nOBSat>0 ){ | |
4202 sqlite3DebugPrintf(" ORDERBY=%d,0x%llx", pWInfo->nOBSat, pWInfo->revMask); | |
4203 } | |
4204 switch( pWInfo->eDistinct ){ | |
4205 case WHERE_DISTINCT_UNIQUE: { | |
4206 sqlite3DebugPrintf(" DISTINCT=unique"); | |
4207 break; | |
4208 } | |
4209 case WHERE_DISTINCT_ORDERED: { | |
4210 sqlite3DebugPrintf(" DISTINCT=ordered"); | |
4211 break; | |
4212 } | |
4213 case WHERE_DISTINCT_UNORDERED: { | |
4214 sqlite3DebugPrintf(" DISTINCT=unordered"); | |
4215 break; | |
4216 } | |
4217 } | |
4218 sqlite3DebugPrintf("\n"); | |
4219 for(ii=0; ii<pWInfo->nLevel; ii++){ | |
4220 whereLoopPrint(pWInfo->a[ii].pWLoop, sWLB.pWC); | |
4221 } | |
4222 } | |
4223 #endif | |
4224 /* Attempt to omit tables from the join that do not effect the result */ | |
4225 if( pWInfo->nLevel>=2 | |
4226 && pResultSet!=0 | |
4227 && OptimizationEnabled(db, SQLITE_OmitNoopJoin) | |
4228 ){ | |
4229 Bitmask tabUsed = sqlite3WhereExprListUsage(pMaskSet, pResultSet); | |
4230 if( sWLB.pOrderBy ){ | |
4231 tabUsed |= sqlite3WhereExprListUsage(pMaskSet, sWLB.pOrderBy); | |
4232 } | |
4233 while( pWInfo->nLevel>=2 ){ | |
4234 WhereTerm *pTerm, *pEnd; | |
4235 pLoop = pWInfo->a[pWInfo->nLevel-1].pWLoop; | |
4236 if( (pWInfo->pTabList->a[pLoop->iTab].fg.jointype & JT_LEFT)==0 ) break; | |
4237 if( (wctrlFlags & WHERE_WANT_DISTINCT)==0 | |
4238 && (pLoop->wsFlags & WHERE_ONEROW)==0 | |
4239 ){ | |
4240 break; | |
4241 } | |
4242 if( (tabUsed & pLoop->maskSelf)!=0 ) break; | |
4243 pEnd = sWLB.pWC->a + sWLB.pWC->nTerm; | |
4244 for(pTerm=sWLB.pWC->a; pTerm<pEnd; pTerm++){ | |
4245 if( (pTerm->prereqAll & pLoop->maskSelf)!=0 | |
4246 && !ExprHasProperty(pTerm->pExpr, EP_FromJoin) | |
4247 ){ | |
4248 break; | |
4249 } | |
4250 } | |
4251 if( pTerm<pEnd ) break; | |
4252 WHERETRACE(0xffff, ("-> drop loop %c not used\n", pLoop->cId)); | |
4253 pWInfo->nLevel--; | |
4254 nTabList--; | |
4255 } | |
4256 } | |
4257 WHERETRACE(0xffff,("*** Optimizer Finished ***\n")); | |
4258 pWInfo->pParse->nQueryLoop += pWInfo->nRowOut; | |
4259 | |
4260 /* If the caller is an UPDATE or DELETE statement that is requesting | |
4261 ** to use a one-pass algorithm, determine if this is appropriate. | |
4262 ** The one-pass algorithm only works if the WHERE clause constrains | |
4263 ** the statement to update or delete a single row. | |
4264 */ | |
4265 assert( (wctrlFlags & WHERE_ONEPASS_DESIRED)==0 || pWInfo->nLevel==1 ); | |
4266 if( (wctrlFlags & WHERE_ONEPASS_DESIRED)!=0 ){ | |
4267 int wsFlags = pWInfo->a[0].pWLoop->wsFlags; | |
4268 int bOnerow = (wsFlags & WHERE_ONEROW)!=0; | |
4269 if( bOnerow || ( (wctrlFlags & WHERE_ONEPASS_MULTIROW) | |
4270 && 0==(wsFlags & WHERE_VIRTUALTABLE) | |
4271 )){ | |
4272 pWInfo->eOnePass = bOnerow ? ONEPASS_SINGLE : ONEPASS_MULTI; | |
4273 if( HasRowid(pTabList->a[0].pTab) && (wsFlags & WHERE_IDX_ONLY) ){ | |
4274 if( wctrlFlags & WHERE_ONEPASS_MULTIROW ){ | |
4275 bFordelete = OPFLAG_FORDELETE; | |
4276 } | |
4277 pWInfo->a[0].pWLoop->wsFlags = (wsFlags & ~WHERE_IDX_ONLY); | |
4278 } | |
4279 } | |
4280 } | |
4281 | |
4282 /* Open all tables in the pTabList and any indices selected for | |
4283 ** searching those tables. | |
4284 */ | |
4285 for(ii=0, pLevel=pWInfo->a; ii<nTabList; ii++, pLevel++){ | |
4286 Table *pTab; /* Table to open */ | |
4287 int iDb; /* Index of database containing table/index */ | |
4288 struct SrcList_item *pTabItem; | |
4289 | |
4290 pTabItem = &pTabList->a[pLevel->iFrom]; | |
4291 pTab = pTabItem->pTab; | |
4292 iDb = sqlite3SchemaToIndex(db, pTab->pSchema); | |
4293 pLoop = pLevel->pWLoop; | |
4294 if( (pTab->tabFlags & TF_Ephemeral)!=0 || pTab->pSelect ){ | |
4295 /* Do nothing */ | |
4296 }else | |
4297 #ifndef SQLITE_OMIT_VIRTUALTABLE | |
4298 if( (pLoop->wsFlags & WHERE_VIRTUALTABLE)!=0 ){ | |
4299 const char *pVTab = (const char *)sqlite3GetVTable(db, pTab); | |
4300 int iCur = pTabItem->iCursor; | |
4301 sqlite3VdbeAddOp4(v, OP_VOpen, iCur, 0, 0, pVTab, P4_VTAB); | |
4302 }else if( IsVirtual(pTab) ){ | |
4303 /* noop */ | |
4304 }else | |
4305 #endif | |
4306 if( (pLoop->wsFlags & WHERE_IDX_ONLY)==0 | |
4307 && (wctrlFlags & WHERE_OMIT_OPEN_CLOSE)==0 ){ | |
4308 int op = OP_OpenRead; | |
4309 if( pWInfo->eOnePass!=ONEPASS_OFF ){ | |
4310 op = OP_OpenWrite; | |
4311 pWInfo->aiCurOnePass[0] = pTabItem->iCursor; | |
4312 }; | |
4313 sqlite3OpenTable(pParse, pTabItem->iCursor, iDb, pTab, op); | |
4314 assert( pTabItem->iCursor==pLevel->iTabCur ); | |
4315 testcase( pWInfo->eOnePass==ONEPASS_OFF && pTab->nCol==BMS-1 ); | |
4316 testcase( pWInfo->eOnePass==ONEPASS_OFF && pTab->nCol==BMS ); | |
4317 if( pWInfo->eOnePass==ONEPASS_OFF && pTab->nCol<BMS && HasRowid(pTab) ){ | |
4318 Bitmask b = pTabItem->colUsed; | |
4319 int n = 0; | |
4320 for(; b; b=b>>1, n++){} | |
4321 sqlite3VdbeChangeP4(v, sqlite3VdbeCurrentAddr(v)-1, | |
4322 SQLITE_INT_TO_PTR(n), P4_INT32); | |
4323 assert( n<=pTab->nCol ); | |
4324 } | |
4325 #ifdef SQLITE_ENABLE_CURSOR_HINTS | |
4326 if( pLoop->u.btree.pIndex!=0 ){ | |
4327 sqlite3VdbeChangeP5(v, OPFLAG_SEEKEQ|bFordelete); | |
4328 }else | |
4329 #endif | |
4330 { | |
4331 sqlite3VdbeChangeP5(v, bFordelete); | |
4332 } | |
4333 #ifdef SQLITE_ENABLE_COLUMN_USED_MASK | |
4334 sqlite3VdbeAddOp4Dup8(v, OP_ColumnsUsed, pTabItem->iCursor, 0, 0, | |
4335 (const u8*)&pTabItem->colUsed, P4_INT64); | |
4336 #endif | |
4337 }else{ | |
4338 sqlite3TableLock(pParse, iDb, pTab->tnum, 0, pTab->zName); | |
4339 } | |
4340 if( pLoop->wsFlags & WHERE_INDEXED ){ | |
4341 Index *pIx = pLoop->u.btree.pIndex; | |
4342 int iIndexCur; | |
4343 int op = OP_OpenRead; | |
4344 /* iIdxCur is always set if to a positive value if ONEPASS is possible */ | |
4345 assert( iIdxCur!=0 || (pWInfo->wctrlFlags & WHERE_ONEPASS_DESIRED)==0 ); | |
4346 if( !HasRowid(pTab) && IsPrimaryKeyIndex(pIx) | |
4347 && (wctrlFlags & WHERE_ONETABLE_ONLY)!=0 | |
4348 ){ | |
4349 /* This is one term of an OR-optimization using the PRIMARY KEY of a | |
4350 ** WITHOUT ROWID table. No need for a separate index */ | |
4351 iIndexCur = pLevel->iTabCur; | |
4352 op = 0; | |
4353 }else if( pWInfo->eOnePass!=ONEPASS_OFF ){ | |
4354 Index *pJ = pTabItem->pTab->pIndex; | |
4355 iIndexCur = iIdxCur; | |
4356 assert( wctrlFlags & WHERE_ONEPASS_DESIRED ); | |
4357 while( ALWAYS(pJ) && pJ!=pIx ){ | |
4358 iIndexCur++; | |
4359 pJ = pJ->pNext; | |
4360 } | |
4361 op = OP_OpenWrite; | |
4362 pWInfo->aiCurOnePass[1] = iIndexCur; | |
4363 }else if( iIdxCur && (wctrlFlags & WHERE_ONETABLE_ONLY)!=0 ){ | |
4364 iIndexCur = iIdxCur; | |
4365 if( wctrlFlags & WHERE_REOPEN_IDX ) op = OP_ReopenIdx; | |
4366 }else{ | |
4367 iIndexCur = pParse->nTab++; | |
4368 } | |
4369 pLevel->iIdxCur = iIndexCur; | |
4370 assert( pIx->pSchema==pTab->pSchema ); | |
4371 assert( iIndexCur>=0 ); | |
4372 if( op ){ | |
4373 sqlite3VdbeAddOp3(v, op, iIndexCur, pIx->tnum, iDb); | |
4374 sqlite3VdbeSetP4KeyInfo(pParse, pIx); | |
4375 if( (pLoop->wsFlags & WHERE_CONSTRAINT)!=0 | |
4376 && (pLoop->wsFlags & (WHERE_COLUMN_RANGE|WHERE_SKIPSCAN))==0 | |
4377 && (pWInfo->wctrlFlags&WHERE_ORDERBY_MIN)==0 | |
4378 ){ | |
4379 sqlite3VdbeChangeP5(v, OPFLAG_SEEKEQ); /* Hint to COMDB2 */ | |
4380 } | |
4381 VdbeComment((v, "%s", pIx->zName)); | |
4382 #ifdef SQLITE_ENABLE_COLUMN_USED_MASK | |
4383 { | |
4384 u64 colUsed = 0; | |
4385 int ii, jj; | |
4386 for(ii=0; ii<pIx->nColumn; ii++){ | |
4387 jj = pIx->aiColumn[ii]; | |
4388 if( jj<0 ) continue; | |
4389 if( jj>63 ) jj = 63; | |
4390 if( (pTabItem->colUsed & MASKBIT(jj))==0 ) continue; | |
4391 colUsed |= ((u64)1)<<(ii<63 ? ii : 63); | |
4392 } | |
4393 sqlite3VdbeAddOp4Dup8(v, OP_ColumnsUsed, iIndexCur, 0, 0, | |
4394 (u8*)&colUsed, P4_INT64); | |
4395 } | |
4396 #endif /* SQLITE_ENABLE_COLUMN_USED_MASK */ | |
4397 } | |
4398 } | |
4399 if( iDb>=0 ) sqlite3CodeVerifySchema(pParse, iDb); | |
4400 } | |
4401 pWInfo->iTop = sqlite3VdbeCurrentAddr(v); | |
4402 if( db->mallocFailed ) goto whereBeginError; | |
4403 | |
4404 /* Generate the code to do the search. Each iteration of the for | |
4405 ** loop below generates code for a single nested loop of the VM | |
4406 ** program. | |
4407 */ | |
4408 notReady = ~(Bitmask)0; | |
4409 for(ii=0; ii<nTabList; ii++){ | |
4410 int addrExplain; | |
4411 int wsFlags; | |
4412 pLevel = &pWInfo->a[ii]; | |
4413 wsFlags = pLevel->pWLoop->wsFlags; | |
4414 #ifndef SQLITE_OMIT_AUTOMATIC_INDEX | |
4415 if( (pLevel->pWLoop->wsFlags & WHERE_AUTO_INDEX)!=0 ){ | |
4416 constructAutomaticIndex(pParse, &pWInfo->sWC, | |
4417 &pTabList->a[pLevel->iFrom], notReady, pLevel); | |
4418 if( db->mallocFailed ) goto whereBeginError; | |
4419 } | |
4420 #endif | |
4421 addrExplain = sqlite3WhereExplainOneScan( | |
4422 pParse, pTabList, pLevel, ii, pLevel->iFrom, wctrlFlags | |
4423 ); | |
4424 pLevel->addrBody = sqlite3VdbeCurrentAddr(v); | |
4425 notReady = sqlite3WhereCodeOneLoopStart(pWInfo, ii, notReady); | |
4426 pWInfo->iContinue = pLevel->addrCont; | |
4427 if( (wsFlags&WHERE_MULTI_OR)==0 && (wctrlFlags&WHERE_ONETABLE_ONLY)==0 ){ | |
4428 sqlite3WhereAddScanStatus(v, pTabList, pLevel, addrExplain); | |
4429 } | |
4430 } | |
4431 | |
4432 /* Done. */ | |
4433 VdbeModuleComment((v, "Begin WHERE-core")); | |
4434 return pWInfo; | |
4435 | |
4436 /* Jump here if malloc fails */ | |
4437 whereBeginError: | |
4438 if( pWInfo ){ | |
4439 pParse->nQueryLoop = pWInfo->savedNQueryLoop; | |
4440 whereInfoFree(db, pWInfo); | |
4441 } | |
4442 return 0; | |
4443 } | |
4444 | |
4445 /* | |
4446 ** Generate the end of the WHERE loop. See comments on | |
4447 ** sqlite3WhereBegin() for additional information. | |
4448 */ | |
4449 void sqlite3WhereEnd(WhereInfo *pWInfo){ | |
4450 Parse *pParse = pWInfo->pParse; | |
4451 Vdbe *v = pParse->pVdbe; | |
4452 int i; | |
4453 WhereLevel *pLevel; | |
4454 WhereLoop *pLoop; | |
4455 SrcList *pTabList = pWInfo->pTabList; | |
4456 sqlite3 *db = pParse->db; | |
4457 | |
4458 /* Generate loop termination code. | |
4459 */ | |
4460 VdbeModuleComment((v, "End WHERE-core")); | |
4461 sqlite3ExprCacheClear(pParse); | |
4462 for(i=pWInfo->nLevel-1; i>=0; i--){ | |
4463 int addr; | |
4464 pLevel = &pWInfo->a[i]; | |
4465 pLoop = pLevel->pWLoop; | |
4466 sqlite3VdbeResolveLabel(v, pLevel->addrCont); | |
4467 if( pLevel->op!=OP_Noop ){ | |
4468 sqlite3VdbeAddOp3(v, pLevel->op, pLevel->p1, pLevel->p2, pLevel->p3); | |
4469 sqlite3VdbeChangeP5(v, pLevel->p5); | |
4470 VdbeCoverage(v); | |
4471 VdbeCoverageIf(v, pLevel->op==OP_Next); | |
4472 VdbeCoverageIf(v, pLevel->op==OP_Prev); | |
4473 VdbeCoverageIf(v, pLevel->op==OP_VNext); | |
4474 } | |
4475 if( pLoop->wsFlags & WHERE_IN_ABLE && pLevel->u.in.nIn>0 ){ | |
4476 struct InLoop *pIn; | |
4477 int j; | |
4478 sqlite3VdbeResolveLabel(v, pLevel->addrNxt); | |
4479 for(j=pLevel->u.in.nIn, pIn=&pLevel->u.in.aInLoop[j-1]; j>0; j--, pIn--){ | |
4480 sqlite3VdbeJumpHere(v, pIn->addrInTop+1); | |
4481 sqlite3VdbeAddOp2(v, pIn->eEndLoopOp, pIn->iCur, pIn->addrInTop); | |
4482 VdbeCoverage(v); | |
4483 VdbeCoverageIf(v, pIn->eEndLoopOp==OP_PrevIfOpen); | |
4484 VdbeCoverageIf(v, pIn->eEndLoopOp==OP_NextIfOpen); | |
4485 sqlite3VdbeJumpHere(v, pIn->addrInTop-1); | |
4486 } | |
4487 } | |
4488 sqlite3VdbeResolveLabel(v, pLevel->addrBrk); | |
4489 if( pLevel->addrSkip ){ | |
4490 sqlite3VdbeGoto(v, pLevel->addrSkip); | |
4491 VdbeComment((v, "next skip-scan on %s", pLoop->u.btree.pIndex->zName)); | |
4492 sqlite3VdbeJumpHere(v, pLevel->addrSkip); | |
4493 sqlite3VdbeJumpHere(v, pLevel->addrSkip-2); | |
4494 } | |
4495 #ifndef SQLITE_LIKE_DOESNT_MATCH_BLOBS | |
4496 if( pLevel->addrLikeRep ){ | |
4497 int op; | |
4498 if( sqlite3VdbeGetOp(v, pLevel->addrLikeRep-1)->p1 ){ | |
4499 op = OP_DecrJumpZero; | |
4500 }else{ | |
4501 op = OP_JumpZeroIncr; | |
4502 } | |
4503 sqlite3VdbeAddOp2(v, op, pLevel->iLikeRepCntr, pLevel->addrLikeRep); | |
4504 VdbeCoverage(v); | |
4505 } | |
4506 #endif | |
4507 if( pLevel->iLeftJoin ){ | |
4508 addr = sqlite3VdbeAddOp1(v, OP_IfPos, pLevel->iLeftJoin); VdbeCoverage(v); | |
4509 assert( (pLoop->wsFlags & WHERE_IDX_ONLY)==0 | |
4510 || (pLoop->wsFlags & WHERE_INDEXED)!=0 ); | |
4511 if( (pLoop->wsFlags & WHERE_IDX_ONLY)==0 ){ | |
4512 sqlite3VdbeAddOp1(v, OP_NullRow, pTabList->a[i].iCursor); | |
4513 } | |
4514 if( pLoop->wsFlags & WHERE_INDEXED ){ | |
4515 sqlite3VdbeAddOp1(v, OP_NullRow, pLevel->iIdxCur); | |
4516 } | |
4517 if( pLevel->op==OP_Return ){ | |
4518 sqlite3VdbeAddOp2(v, OP_Gosub, pLevel->p1, pLevel->addrFirst); | |
4519 }else{ | |
4520 sqlite3VdbeGoto(v, pLevel->addrFirst); | |
4521 } | |
4522 sqlite3VdbeJumpHere(v, addr); | |
4523 } | |
4524 VdbeModuleComment((v, "End WHERE-loop%d: %s", i, | |
4525 pWInfo->pTabList->a[pLevel->iFrom].pTab->zName)); | |
4526 } | |
4527 | |
4528 /* The "break" point is here, just past the end of the outer loop. | |
4529 ** Set it. | |
4530 */ | |
4531 sqlite3VdbeResolveLabel(v, pWInfo->iBreak); | |
4532 | |
4533 assert( pWInfo->nLevel<=pTabList->nSrc ); | |
4534 for(i=0, pLevel=pWInfo->a; i<pWInfo->nLevel; i++, pLevel++){ | |
4535 int k, last; | |
4536 VdbeOp *pOp; | |
4537 Index *pIdx = 0; | |
4538 struct SrcList_item *pTabItem = &pTabList->a[pLevel->iFrom]; | |
4539 Table *pTab = pTabItem->pTab; | |
4540 assert( pTab!=0 ); | |
4541 pLoop = pLevel->pWLoop; | |
4542 | |
4543 /* For a co-routine, change all OP_Column references to the table of | |
4544 ** the co-routine into OP_Copy of result contained in a register. | |
4545 ** OP_Rowid becomes OP_Null. | |
4546 */ | |
4547 if( pTabItem->fg.viaCoroutine && !db->mallocFailed ){ | |
4548 translateColumnToCopy(v, pLevel->addrBody, pLevel->iTabCur, | |
4549 pTabItem->regResult, 0); | |
4550 continue; | |
4551 } | |
4552 | |
4553 /* Close all of the cursors that were opened by sqlite3WhereBegin. | |
4554 ** Except, do not close cursors that will be reused by the OR optimization | |
4555 ** (WHERE_OMIT_OPEN_CLOSE). And do not close the OP_OpenWrite cursors | |
4556 ** created for the ONEPASS optimization. | |
4557 */ | |
4558 if( (pTab->tabFlags & TF_Ephemeral)==0 | |
4559 && pTab->pSelect==0 | |
4560 && (pWInfo->wctrlFlags & WHERE_OMIT_OPEN_CLOSE)==0 | |
4561 ){ | |
4562 int ws = pLoop->wsFlags; | |
4563 if( pWInfo->eOnePass==ONEPASS_OFF && (ws & WHERE_IDX_ONLY)==0 ){ | |
4564 sqlite3VdbeAddOp1(v, OP_Close, pTabItem->iCursor); | |
4565 } | |
4566 if( (ws & WHERE_INDEXED)!=0 | |
4567 && (ws & (WHERE_IPK|WHERE_AUTO_INDEX))==0 | |
4568 && pLevel->iIdxCur!=pWInfo->aiCurOnePass[1] | |
4569 ){ | |
4570 sqlite3VdbeAddOp1(v, OP_Close, pLevel->iIdxCur); | |
4571 } | |
4572 } | |
4573 | |
4574 /* If this scan uses an index, make VDBE code substitutions to read data | |
4575 ** from the index instead of from the table where possible. In some cases | |
4576 ** this optimization prevents the table from ever being read, which can | |
4577 ** yield a significant performance boost. | |
4578 ** | |
4579 ** Calls to the code generator in between sqlite3WhereBegin and | |
4580 ** sqlite3WhereEnd will have created code that references the table | |
4581 ** directly. This loop scans all that code looking for opcodes | |
4582 ** that reference the table and converts them into opcodes that | |
4583 ** reference the index. | |
4584 */ | |
4585 if( pLoop->wsFlags & (WHERE_INDEXED|WHERE_IDX_ONLY) ){ | |
4586 pIdx = pLoop->u.btree.pIndex; | |
4587 }else if( pLoop->wsFlags & WHERE_MULTI_OR ){ | |
4588 pIdx = pLevel->u.pCovidx; | |
4589 } | |
4590 if( pIdx | |
4591 && (pWInfo->eOnePass==ONEPASS_OFF || !HasRowid(pIdx->pTable)) | |
4592 && !db->mallocFailed | |
4593 ){ | |
4594 last = sqlite3VdbeCurrentAddr(v); | |
4595 k = pLevel->addrBody; | |
4596 pOp = sqlite3VdbeGetOp(v, k); | |
4597 for(; k<last; k++, pOp++){ | |
4598 if( pOp->p1!=pLevel->iTabCur ) continue; | |
4599 if( pOp->opcode==OP_Column ){ | |
4600 int x = pOp->p2; | |
4601 assert( pIdx->pTable==pTab ); | |
4602 if( !HasRowid(pTab) ){ | |
4603 Index *pPk = sqlite3PrimaryKeyIndex(pTab); | |
4604 x = pPk->aiColumn[x]; | |
4605 assert( x>=0 ); | |
4606 } | |
4607 x = sqlite3ColumnOfIndex(pIdx, x); | |
4608 if( x>=0 ){ | |
4609 pOp->p2 = x; | |
4610 pOp->p1 = pLevel->iIdxCur; | |
4611 } | |
4612 assert( (pLoop->wsFlags & WHERE_IDX_ONLY)==0 || x>=0 ); | |
4613 }else if( pOp->opcode==OP_Rowid ){ | |
4614 pOp->p1 = pLevel->iIdxCur; | |
4615 pOp->opcode = OP_IdxRowid; | |
4616 } | |
4617 } | |
4618 } | |
4619 } | |
4620 | |
4621 /* Final cleanup | |
4622 */ | |
4623 pParse->nQueryLoop = pWInfo->savedNQueryLoop; | |
4624 whereInfoFree(db, pWInfo); | |
4625 return; | |
4626 } | |
OLD | NEW |