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 /* 23 ** Extra information appended to the end of sqlite3_index_info but not 24 ** visible to the xBestIndex function, at least not directly. The 25 ** sqlite3_vtab_collation() interface knows how to reach it, however. 26 ** 27 ** This object is not an API and can be changed from one release to the 28 ** next. As long as allocateIndexInfo() and sqlite3_vtab_collation() 29 ** agree on the structure, all will be well. 30 */ 31 typedef struct HiddenIndexInfo HiddenIndexInfo; 32 struct HiddenIndexInfo { 33 WhereClause *pWC; /* The Where clause being analyzed */ 34 Parse *pParse; /* The parsing context */ 35 }; 36 37 /* Forward declaration of methods */ 38 static int whereLoopResize(sqlite3*, WhereLoop*, int); 39 40 /* 41 ** Return the estimated number of output rows from a WHERE clause 42 */ 43 LogEst sqlite3WhereOutputRowCount(WhereInfo *pWInfo){ 44 return pWInfo->nRowOut; 45 } 46 47 /* 48 ** Return one of the WHERE_DISTINCT_xxxxx values to indicate how this 49 ** WHERE clause returns outputs for DISTINCT processing. 50 */ 51 int sqlite3WhereIsDistinct(WhereInfo *pWInfo){ 52 return pWInfo->eDistinct; 53 } 54 55 /* 56 ** Return the number of ORDER BY terms that are satisfied by the 57 ** WHERE clause. A return of 0 means that the output must be 58 ** completely sorted. A return equal to the number of ORDER BY 59 ** terms means that no sorting is needed at all. A return that 60 ** is positive but less than the number of ORDER BY terms means that 61 ** block sorting is required. 62 */ 63 int sqlite3WhereIsOrdered(WhereInfo *pWInfo){ 64 return pWInfo->nOBSat; 65 } 66 67 /* 68 ** In the ORDER BY LIMIT optimization, if the inner-most loop is known 69 ** to emit rows in increasing order, and if the last row emitted by the 70 ** inner-most loop did not fit within the sorter, then we can skip all 71 ** subsequent rows for the current iteration of the inner loop (because they 72 ** will not fit in the sorter either) and continue with the second inner 73 ** loop - the loop immediately outside the inner-most. 74 ** 75 ** When a row does not fit in the sorter (because the sorter already 76 ** holds LIMIT+OFFSET rows that are smaller), then a jump is made to the 77 ** label returned by this function. 78 ** 79 ** If the ORDER BY LIMIT optimization applies, the jump destination should 80 ** be the continuation for the second-inner-most loop. If the ORDER BY 81 ** LIMIT optimization does not apply, then the jump destination should 82 ** be the continuation for the inner-most loop. 83 ** 84 ** It is always safe for this routine to return the continuation of the 85 ** inner-most loop, in the sense that a correct answer will result. 86 ** Returning the continuation the second inner loop is an optimization 87 ** that might make the code run a little faster, but should not change 88 ** the final answer. 89 */ 90 int sqlite3WhereOrderByLimitOptLabel(WhereInfo *pWInfo){ 91 WhereLevel *pInner; 92 if( !pWInfo->bOrderedInnerLoop ){ 93 /* The ORDER BY LIMIT optimization does not apply. Jump to the 94 ** continuation of the inner-most loop. */ 95 return pWInfo->iContinue; 96 } 97 pInner = &pWInfo->a[pWInfo->nLevel-1]; 98 assert( pInner->addrNxt!=0 ); 99 return pInner->addrNxt; 100 } 101 102 /* 103 ** While generating code for the min/max optimization, after handling 104 ** the aggregate-step call to min() or max(), check to see if any 105 ** additional looping is required. If the output order is such that 106 ** we are certain that the correct answer has already been found, then 107 ** code an OP_Goto to by pass subsequent processing. 108 ** 109 ** Any extra OP_Goto that is coded here is an optimization. The 110 ** correct answer should be obtained regardless. This OP_Goto just 111 ** makes the answer appear faster. 112 */ 113 void sqlite3WhereMinMaxOptEarlyOut(Vdbe *v, WhereInfo *pWInfo){ 114 WhereLevel *pInner; 115 int i; 116 if( !pWInfo->bOrderedInnerLoop ) return; 117 if( pWInfo->nOBSat==0 ) return; 118 for(i=pWInfo->nLevel-1; i>=0; i--){ 119 pInner = &pWInfo->a[i]; 120 if( (pInner->pWLoop->wsFlags & WHERE_COLUMN_IN)!=0 ){ 121 sqlite3VdbeGoto(v, pInner->addrNxt); 122 return; 123 } 124 } 125 sqlite3VdbeGoto(v, pWInfo->iBreak); 126 } 127 128 /* 129 ** Return the VDBE address or label to jump to in order to continue 130 ** immediately with the next row of a WHERE clause. 131 */ 132 int sqlite3WhereContinueLabel(WhereInfo *pWInfo){ 133 assert( pWInfo->iContinue!=0 ); 134 return pWInfo->iContinue; 135 } 136 137 /* 138 ** Return the VDBE address or label to jump to in order to break 139 ** out of a WHERE loop. 140 */ 141 int sqlite3WhereBreakLabel(WhereInfo *pWInfo){ 142 return pWInfo->iBreak; 143 } 144 145 /* 146 ** Return ONEPASS_OFF (0) if an UPDATE or DELETE statement is unable to 147 ** operate directly on the rowids returned by a WHERE clause. Return 148 ** ONEPASS_SINGLE (1) if the statement can operation directly because only 149 ** a single row is to be changed. Return ONEPASS_MULTI (2) if the one-pass 150 ** optimization can be used on multiple 151 ** 152 ** If the ONEPASS optimization is used (if this routine returns true) 153 ** then also write the indices of open cursors used by ONEPASS 154 ** into aiCur[0] and aiCur[1]. iaCur[0] gets the cursor of the data 155 ** table and iaCur[1] gets the cursor used by an auxiliary index. 156 ** Either value may be -1, indicating that cursor is not used. 157 ** Any cursors returned will have been opened for writing. 158 ** 159 ** aiCur[0] and aiCur[1] both get -1 if the where-clause logic is 160 ** unable to use the ONEPASS optimization. 161 */ 162 int sqlite3WhereOkOnePass(WhereInfo *pWInfo, int *aiCur){ 163 memcpy(aiCur, pWInfo->aiCurOnePass, sizeof(int)*2); 164 #ifdef WHERETRACE_ENABLED 165 if( sqlite3WhereTrace && pWInfo->eOnePass!=ONEPASS_OFF ){ 166 sqlite3DebugPrintf("%s cursors: %d %d\n", 167 pWInfo->eOnePass==ONEPASS_SINGLE ? "ONEPASS_SINGLE" : "ONEPASS_MULTI", 168 aiCur[0], aiCur[1]); 169 } 170 #endif 171 return pWInfo->eOnePass; 172 } 173 174 /* 175 ** Return TRUE if the WHERE loop uses the OP_DeferredSeek opcode to move 176 ** the data cursor to the row selected by the index cursor. 177 */ 178 int sqlite3WhereUsesDeferredSeek(WhereInfo *pWInfo){ 179 return pWInfo->bDeferredSeek; 180 } 181 182 /* 183 ** Move the content of pSrc into pDest 184 */ 185 static void whereOrMove(WhereOrSet *pDest, WhereOrSet *pSrc){ 186 pDest->n = pSrc->n; 187 memcpy(pDest->a, pSrc->a, pDest->n*sizeof(pDest->a[0])); 188 } 189 190 /* 191 ** Try to insert a new prerequisite/cost entry into the WhereOrSet pSet. 192 ** 193 ** The new entry might overwrite an existing entry, or it might be 194 ** appended, or it might be discarded. Do whatever is the right thing 195 ** so that pSet keeps the N_OR_COST best entries seen so far. 196 */ 197 static int whereOrInsert( 198 WhereOrSet *pSet, /* The WhereOrSet to be updated */ 199 Bitmask prereq, /* Prerequisites of the new entry */ 200 LogEst rRun, /* Run-cost of the new entry */ 201 LogEst nOut /* Number of outputs for the new entry */ 202 ){ 203 u16 i; 204 WhereOrCost *p; 205 for(i=pSet->n, p=pSet->a; i>0; i--, p++){ 206 if( rRun<=p->rRun && (prereq & p->prereq)==prereq ){ 207 goto whereOrInsert_done; 208 } 209 if( p->rRun<=rRun && (p->prereq & prereq)==p->prereq ){ 210 return 0; 211 } 212 } 213 if( pSet->n<N_OR_COST ){ 214 p = &pSet->a[pSet->n++]; 215 p->nOut = nOut; 216 }else{ 217 p = pSet->a; 218 for(i=1; i<pSet->n; i++){ 219 if( p->rRun>pSet->a[i].rRun ) p = pSet->a + i; 220 } 221 if( p->rRun<=rRun ) return 0; 222 } 223 whereOrInsert_done: 224 p->prereq = prereq; 225 p->rRun = rRun; 226 if( p->nOut>nOut ) p->nOut = nOut; 227 return 1; 228 } 229 230 /* 231 ** Return the bitmask for the given cursor number. Return 0 if 232 ** iCursor is not in the set. 233 */ 234 Bitmask sqlite3WhereGetMask(WhereMaskSet *pMaskSet, int iCursor){ 235 int i; 236 assert( pMaskSet->n<=(int)sizeof(Bitmask)*8 ); 237 for(i=0; i<pMaskSet->n; i++){ 238 if( pMaskSet->ix[i]==iCursor ){ 239 return MASKBIT(i); 240 } 241 } 242 return 0; 243 } 244 245 /* 246 ** Create a new mask for cursor iCursor. 247 ** 248 ** There is one cursor per table in the FROM clause. The number of 249 ** tables in the FROM clause is limited by a test early in the 250 ** sqlite3WhereBegin() routine. So we know that the pMaskSet->ix[] 251 ** array will never overflow. 252 */ 253 static void createMask(WhereMaskSet *pMaskSet, int iCursor){ 254 assert( pMaskSet->n < ArraySize(pMaskSet->ix) ); 255 pMaskSet->ix[pMaskSet->n++] = iCursor; 256 } 257 258 /* 259 ** If the right-hand branch of the expression is a TK_COLUMN, then return 260 ** a pointer to the right-hand branch. Otherwise, return NULL. 261 */ 262 static Expr *whereRightSubexprIsColumn(Expr *p){ 263 p = sqlite3ExprSkipCollateAndLikely(p->pRight); 264 if( ALWAYS(p!=0) && p->op==TK_COLUMN && !ExprHasProperty(p, EP_FixedCol) ){ 265 return p; 266 } 267 return 0; 268 } 269 270 /* 271 ** Advance to the next WhereTerm that matches according to the criteria 272 ** established when the pScan object was initialized by whereScanInit(). 273 ** Return NULL if there are no more matching WhereTerms. 274 */ 275 static WhereTerm *whereScanNext(WhereScan *pScan){ 276 int iCur; /* The cursor on the LHS of the term */ 277 i16 iColumn; /* The column on the LHS of the term. -1 for IPK */ 278 Expr *pX; /* An expression being tested */ 279 WhereClause *pWC; /* Shorthand for pScan->pWC */ 280 WhereTerm *pTerm; /* The term being tested */ 281 int k = pScan->k; /* Where to start scanning */ 282 283 assert( pScan->iEquiv<=pScan->nEquiv ); 284 pWC = pScan->pWC; 285 while(1){ 286 iColumn = pScan->aiColumn[pScan->iEquiv-1]; 287 iCur = pScan->aiCur[pScan->iEquiv-1]; 288 assert( pWC!=0 ); 289 do{ 290 for(pTerm=pWC->a+k; k<pWC->nTerm; k++, pTerm++){ 291 if( pTerm->leftCursor==iCur 292 && pTerm->u.x.leftColumn==iColumn 293 && (iColumn!=XN_EXPR 294 || sqlite3ExprCompareSkip(pTerm->pExpr->pLeft, 295 pScan->pIdxExpr,iCur)==0) 296 && (pScan->iEquiv<=1 || !ExprHasProperty(pTerm->pExpr, EP_FromJoin)) 297 ){ 298 if( (pTerm->eOperator & WO_EQUIV)!=0 299 && pScan->nEquiv<ArraySize(pScan->aiCur) 300 && (pX = whereRightSubexprIsColumn(pTerm->pExpr))!=0 301 ){ 302 int j; 303 for(j=0; j<pScan->nEquiv; j++){ 304 if( pScan->aiCur[j]==pX->iTable 305 && pScan->aiColumn[j]==pX->iColumn ){ 306 break; 307 } 308 } 309 if( j==pScan->nEquiv ){ 310 pScan->aiCur[j] = pX->iTable; 311 pScan->aiColumn[j] = pX->iColumn; 312 pScan->nEquiv++; 313 } 314 } 315 if( (pTerm->eOperator & pScan->opMask)!=0 ){ 316 /* Verify the affinity and collating sequence match */ 317 if( pScan->zCollName && (pTerm->eOperator & WO_ISNULL)==0 ){ 318 CollSeq *pColl; 319 Parse *pParse = pWC->pWInfo->pParse; 320 pX = pTerm->pExpr; 321 if( !sqlite3IndexAffinityOk(pX, pScan->idxaff) ){ 322 continue; 323 } 324 assert(pX->pLeft); 325 pColl = sqlite3ExprCompareCollSeq(pParse, pX); 326 if( pColl==0 ) pColl = pParse->db->pDfltColl; 327 if( sqlite3StrICmp(pColl->zName, pScan->zCollName) ){ 328 continue; 329 } 330 } 331 if( (pTerm->eOperator & (WO_EQ|WO_IS))!=0 332 && (pX = pTerm->pExpr->pRight)->op==TK_COLUMN 333 && pX->iTable==pScan->aiCur[0] 334 && pX->iColumn==pScan->aiColumn[0] 335 ){ 336 testcase( pTerm->eOperator & WO_IS ); 337 continue; 338 } 339 pScan->pWC = pWC; 340 pScan->k = k+1; 341 #ifdef WHERETRACE_ENABLED 342 if( sqlite3WhereTrace & 0x20000 ){ 343 int ii; 344 sqlite3DebugPrintf("SCAN-TERM %p: nEquiv=%d", 345 pTerm, pScan->nEquiv); 346 for(ii=0; ii<pScan->nEquiv; ii++){ 347 sqlite3DebugPrintf(" {%d:%d}", 348 pScan->aiCur[ii], pScan->aiColumn[ii]); 349 } 350 sqlite3DebugPrintf("\n"); 351 } 352 #endif 353 return pTerm; 354 } 355 } 356 } 357 pWC = pWC->pOuter; 358 k = 0; 359 }while( pWC!=0 ); 360 if( pScan->iEquiv>=pScan->nEquiv ) break; 361 pWC = pScan->pOrigWC; 362 k = 0; 363 pScan->iEquiv++; 364 } 365 return 0; 366 } 367 368 /* 369 ** This is whereScanInit() for the case of an index on an expression. 370 ** It is factored out into a separate tail-recursion subroutine so that 371 ** the normal whereScanInit() routine, which is a high-runner, does not 372 ** need to push registers onto the stack as part of its prologue. 373 */ 374 static SQLITE_NOINLINE WhereTerm *whereScanInitIndexExpr(WhereScan *pScan){ 375 pScan->idxaff = sqlite3ExprAffinity(pScan->pIdxExpr); 376 return whereScanNext(pScan); 377 } 378 379 /* 380 ** Initialize a WHERE clause scanner object. Return a pointer to the 381 ** first match. Return NULL if there are no matches. 382 ** 383 ** The scanner will be searching the WHERE clause pWC. It will look 384 ** for terms of the form "X <op> <expr>" where X is column iColumn of table 385 ** iCur. Or if pIdx!=0 then X is column iColumn of index pIdx. pIdx 386 ** must be one of the indexes of table iCur. 387 ** 388 ** The <op> must be one of the operators described by opMask. 389 ** 390 ** If the search is for X and the WHERE clause contains terms of the 391 ** form X=Y then this routine might also return terms of the form 392 ** "Y <op> <expr>". The number of levels of transitivity is limited, 393 ** but is enough to handle most commonly occurring SQL statements. 394 ** 395 ** If X is not the INTEGER PRIMARY KEY then X must be compatible with 396 ** index pIdx. 397 */ 398 static WhereTerm *whereScanInit( 399 WhereScan *pScan, /* The WhereScan object being initialized */ 400 WhereClause *pWC, /* The WHERE clause to be scanned */ 401 int iCur, /* Cursor to scan for */ 402 int iColumn, /* Column to scan for */ 403 u32 opMask, /* Operator(s) to scan for */ 404 Index *pIdx /* Must be compatible with this index */ 405 ){ 406 pScan->pOrigWC = pWC; 407 pScan->pWC = pWC; 408 pScan->pIdxExpr = 0; 409 pScan->idxaff = 0; 410 pScan->zCollName = 0; 411 pScan->opMask = opMask; 412 pScan->k = 0; 413 pScan->aiCur[0] = iCur; 414 pScan->nEquiv = 1; 415 pScan->iEquiv = 1; 416 if( pIdx ){ 417 int j = iColumn; 418 iColumn = pIdx->aiColumn[j]; 419 if( iColumn==XN_EXPR ){ 420 pScan->pIdxExpr = pIdx->aColExpr->a[j].pExpr; 421 pScan->zCollName = pIdx->azColl[j]; 422 pScan->aiColumn[0] = XN_EXPR; 423 return whereScanInitIndexExpr(pScan); 424 }else if( iColumn==pIdx->pTable->iPKey ){ 425 iColumn = XN_ROWID; 426 }else if( iColumn>=0 ){ 427 pScan->idxaff = pIdx->pTable->aCol[iColumn].affinity; 428 pScan->zCollName = pIdx->azColl[j]; 429 } 430 }else if( iColumn==XN_EXPR ){ 431 return 0; 432 } 433 pScan->aiColumn[0] = iColumn; 434 return whereScanNext(pScan); 435 } 436 437 /* 438 ** Search for a term in the WHERE clause that is of the form "X <op> <expr>" 439 ** where X is a reference to the iColumn of table iCur or of index pIdx 440 ** if pIdx!=0 and <op> is one of the WO_xx operator codes specified by 441 ** the op parameter. Return a pointer to the term. Return 0 if not found. 442 ** 443 ** If pIdx!=0 then it must be one of the indexes of table iCur. 444 ** Search for terms matching the iColumn-th column of pIdx 445 ** rather than the iColumn-th column of table iCur. 446 ** 447 ** The term returned might by Y=<expr> if there is another constraint in 448 ** the WHERE clause that specifies that X=Y. Any such constraints will be 449 ** identified by the WO_EQUIV bit in the pTerm->eOperator field. The 450 ** aiCur[]/iaColumn[] arrays hold X and all its equivalents. There are 11 451 ** slots in aiCur[]/aiColumn[] so that means we can look for X plus up to 10 452 ** other equivalent values. Hence a search for X will return <expr> if X=A1 453 ** and A1=A2 and A2=A3 and ... and A9=A10 and A10=<expr>. 454 ** 455 ** If there are multiple terms in the WHERE clause of the form "X <op> <expr>" 456 ** then try for the one with no dependencies on <expr> - in other words where 457 ** <expr> is a constant expression of some kind. Only return entries of 458 ** the form "X <op> Y" where Y is a column in another table if no terms of 459 ** the form "X <op> <const-expr>" exist. If no terms with a constant RHS 460 ** exist, try to return a term that does not use WO_EQUIV. 461 */ 462 WhereTerm *sqlite3WhereFindTerm( 463 WhereClause *pWC, /* The WHERE clause to be searched */ 464 int iCur, /* Cursor number of LHS */ 465 int iColumn, /* Column number of LHS */ 466 Bitmask notReady, /* RHS must not overlap with this mask */ 467 u32 op, /* Mask of WO_xx values describing operator */ 468 Index *pIdx /* Must be compatible with this index, if not NULL */ 469 ){ 470 WhereTerm *pResult = 0; 471 WhereTerm *p; 472 WhereScan scan; 473 474 p = whereScanInit(&scan, pWC, iCur, iColumn, op, pIdx); 475 op &= WO_EQ|WO_IS; 476 while( p ){ 477 if( (p->prereqRight & notReady)==0 ){ 478 if( p->prereqRight==0 && (p->eOperator&op)!=0 ){ 479 testcase( p->eOperator & WO_IS ); 480 return p; 481 } 482 if( pResult==0 ) pResult = p; 483 } 484 p = whereScanNext(&scan); 485 } 486 return pResult; 487 } 488 489 /* 490 ** This function searches pList for an entry that matches the iCol-th column 491 ** of index pIdx. 492 ** 493 ** If such an expression is found, its index in pList->a[] is returned. If 494 ** no expression is found, -1 is returned. 495 */ 496 static int findIndexCol( 497 Parse *pParse, /* Parse context */ 498 ExprList *pList, /* Expression list to search */ 499 int iBase, /* Cursor for table associated with pIdx */ 500 Index *pIdx, /* Index to match column of */ 501 int iCol /* Column of index to match */ 502 ){ 503 int i; 504 const char *zColl = pIdx->azColl[iCol]; 505 506 for(i=0; i<pList->nExpr; i++){ 507 Expr *p = sqlite3ExprSkipCollateAndLikely(pList->a[i].pExpr); 508 if( ALWAYS(p!=0) 509 && (p->op==TK_COLUMN || p->op==TK_AGG_COLUMN) 510 && p->iColumn==pIdx->aiColumn[iCol] 511 && p->iTable==iBase 512 ){ 513 CollSeq *pColl = sqlite3ExprNNCollSeq(pParse, pList->a[i].pExpr); 514 if( 0==sqlite3StrICmp(pColl->zName, zColl) ){ 515 return i; 516 } 517 } 518 } 519 520 return -1; 521 } 522 523 /* 524 ** Return TRUE if the iCol-th column of index pIdx is NOT NULL 525 */ 526 static int indexColumnNotNull(Index *pIdx, int iCol){ 527 int j; 528 assert( pIdx!=0 ); 529 assert( iCol>=0 && iCol<pIdx->nColumn ); 530 j = pIdx->aiColumn[iCol]; 531 if( j>=0 ){ 532 return pIdx->pTable->aCol[j].notNull; 533 }else if( j==(-1) ){ 534 return 1; 535 }else{ 536 assert( j==(-2) ); 537 return 0; /* Assume an indexed expression can always yield a NULL */ 538 539 } 540 } 541 542 /* 543 ** Return true if the DISTINCT expression-list passed as the third argument 544 ** is redundant. 545 ** 546 ** A DISTINCT list is redundant if any subset of the columns in the 547 ** DISTINCT list are collectively unique and individually non-null. 548 */ 549 static int isDistinctRedundant( 550 Parse *pParse, /* Parsing context */ 551 SrcList *pTabList, /* The FROM clause */ 552 WhereClause *pWC, /* The WHERE clause */ 553 ExprList *pDistinct /* The result set that needs to be DISTINCT */ 554 ){ 555 Table *pTab; 556 Index *pIdx; 557 int i; 558 int iBase; 559 560 /* If there is more than one table or sub-select in the FROM clause of 561 ** this query, then it will not be possible to show that the DISTINCT 562 ** clause is redundant. */ 563 if( pTabList->nSrc!=1 ) return 0; 564 iBase = pTabList->a[0].iCursor; 565 pTab = pTabList->a[0].pTab; 566 567 /* If any of the expressions is an IPK column on table iBase, then return 568 ** true. Note: The (p->iTable==iBase) part of this test may be false if the 569 ** current SELECT is a correlated sub-query. 570 */ 571 for(i=0; i<pDistinct->nExpr; i++){ 572 Expr *p = sqlite3ExprSkipCollateAndLikely(pDistinct->a[i].pExpr); 573 if( NEVER(p==0) ) continue; 574 if( p->op!=TK_COLUMN && p->op!=TK_AGG_COLUMN ) continue; 575 if( p->iTable==iBase && p->iColumn<0 ) return 1; 576 } 577 578 /* Loop through all indices on the table, checking each to see if it makes 579 ** the DISTINCT qualifier redundant. It does so if: 580 ** 581 ** 1. The index is itself UNIQUE, and 582 ** 583 ** 2. All of the columns in the index are either part of the pDistinct 584 ** list, or else the WHERE clause contains a term of the form "col=X", 585 ** where X is a constant value. The collation sequences of the 586 ** comparison and select-list expressions must match those of the index. 587 ** 588 ** 3. All of those index columns for which the WHERE clause does not 589 ** contain a "col=X" term are subject to a NOT NULL constraint. 590 */ 591 for(pIdx=pTab->pIndex; pIdx; pIdx=pIdx->pNext){ 592 if( !IsUniqueIndex(pIdx) ) continue; 593 if( pIdx->pPartIdxWhere ) continue; 594 for(i=0; i<pIdx->nKeyCol; i++){ 595 if( 0==sqlite3WhereFindTerm(pWC, iBase, i, ~(Bitmask)0, WO_EQ, pIdx) ){ 596 if( findIndexCol(pParse, pDistinct, iBase, pIdx, i)<0 ) break; 597 if( indexColumnNotNull(pIdx, i)==0 ) break; 598 } 599 } 600 if( i==pIdx->nKeyCol ){ 601 /* This index implies that the DISTINCT qualifier is redundant. */ 602 return 1; 603 } 604 } 605 606 return 0; 607 } 608 609 610 /* 611 ** Estimate the logarithm of the input value to base 2. 612 */ 613 static LogEst estLog(LogEst N){ 614 return N<=10 ? 0 : sqlite3LogEst(N) - 33; 615 } 616 617 /* 618 ** Convert OP_Column opcodes to OP_Copy in previously generated code. 619 ** 620 ** This routine runs over generated VDBE code and translates OP_Column 621 ** opcodes into OP_Copy when the table is being accessed via co-routine 622 ** instead of via table lookup. 623 ** 624 ** If the iAutoidxCur is not zero, then any OP_Rowid instructions on 625 ** cursor iTabCur are transformed into OP_Sequence opcode for the 626 ** iAutoidxCur cursor, in order to generate unique rowids for the 627 ** automatic index being generated. 628 */ 629 static void translateColumnToCopy( 630 Parse *pParse, /* Parsing context */ 631 int iStart, /* Translate from this opcode to the end */ 632 int iTabCur, /* OP_Column/OP_Rowid references to this table */ 633 int iRegister, /* The first column is in this register */ 634 int iAutoidxCur /* If non-zero, cursor of autoindex being generated */ 635 ){ 636 Vdbe *v = pParse->pVdbe; 637 VdbeOp *pOp = sqlite3VdbeGetOp(v, iStart); 638 int iEnd = sqlite3VdbeCurrentAddr(v); 639 if( pParse->db->mallocFailed ) return; 640 for(; iStart<iEnd; iStart++, pOp++){ 641 if( pOp->p1!=iTabCur ) continue; 642 if( pOp->opcode==OP_Column ){ 643 pOp->opcode = OP_Copy; 644 pOp->p1 = pOp->p2 + iRegister; 645 pOp->p2 = pOp->p3; 646 pOp->p3 = 0; 647 }else if( pOp->opcode==OP_Rowid ){ 648 pOp->opcode = OP_Sequence; 649 pOp->p1 = iAutoidxCur; 650 #ifdef SQLITE_ALLOW_ROWID_IN_VIEW 651 if( iAutoidxCur==0 ){ 652 pOp->opcode = OP_Null; 653 pOp->p3 = 0; 654 } 655 #endif 656 } 657 } 658 } 659 660 /* 661 ** Two routines for printing the content of an sqlite3_index_info 662 ** structure. Used for testing and debugging only. If neither 663 ** SQLITE_TEST or SQLITE_DEBUG are defined, then these routines 664 ** are no-ops. 665 */ 666 #if !defined(SQLITE_OMIT_VIRTUALTABLE) && defined(WHERETRACE_ENABLED) 667 static void whereTraceIndexInfoInputs(sqlite3_index_info *p){ 668 int i; 669 if( !sqlite3WhereTrace ) return; 670 for(i=0; i<p->nConstraint; i++){ 671 sqlite3DebugPrintf(" constraint[%d]: col=%d termid=%d op=%d usabled=%d\n", 672 i, 673 p->aConstraint[i].iColumn, 674 p->aConstraint[i].iTermOffset, 675 p->aConstraint[i].op, 676 p->aConstraint[i].usable); 677 } 678 for(i=0; i<p->nOrderBy; i++){ 679 sqlite3DebugPrintf(" orderby[%d]: col=%d desc=%d\n", 680 i, 681 p->aOrderBy[i].iColumn, 682 p->aOrderBy[i].desc); 683 } 684 } 685 static void whereTraceIndexInfoOutputs(sqlite3_index_info *p){ 686 int i; 687 if( !sqlite3WhereTrace ) return; 688 for(i=0; i<p->nConstraint; i++){ 689 sqlite3DebugPrintf(" usage[%d]: argvIdx=%d omit=%d\n", 690 i, 691 p->aConstraintUsage[i].argvIndex, 692 p->aConstraintUsage[i].omit); 693 } 694 sqlite3DebugPrintf(" idxNum=%d\n", p->idxNum); 695 sqlite3DebugPrintf(" idxStr=%s\n", p->idxStr); 696 sqlite3DebugPrintf(" orderByConsumed=%d\n", p->orderByConsumed); 697 sqlite3DebugPrintf(" estimatedCost=%g\n", p->estimatedCost); 698 sqlite3DebugPrintf(" estimatedRows=%lld\n", p->estimatedRows); 699 } 700 #else 701 #define whereTraceIndexInfoInputs(A) 702 #define whereTraceIndexInfoOutputs(A) 703 #endif 704 705 #ifndef SQLITE_OMIT_AUTOMATIC_INDEX 706 /* 707 ** Return TRUE if the WHERE clause term pTerm is of a form where it 708 ** could be used with an index to access pSrc, assuming an appropriate 709 ** index existed. 710 */ 711 static int termCanDriveIndex( 712 WhereTerm *pTerm, /* WHERE clause term to check */ 713 SrcItem *pSrc, /* Table we are trying to access */ 714 Bitmask notReady /* Tables in outer loops of the join */ 715 ){ 716 char aff; 717 if( pTerm->leftCursor!=pSrc->iCursor ) return 0; 718 if( (pTerm->eOperator & (WO_EQ|WO_IS))==0 ) return 0; 719 if( (pSrc->fg.jointype & JT_LEFT) 720 && !ExprHasProperty(pTerm->pExpr, EP_FromJoin) 721 && (pTerm->eOperator & WO_IS) 722 ){ 723 /* Cannot use an IS term from the WHERE clause as an index driver for 724 ** the RHS of a LEFT JOIN. Such a term can only be used if it is from 725 ** the ON clause. */ 726 return 0; 727 } 728 if( (pTerm->prereqRight & notReady)!=0 ) return 0; 729 if( pTerm->u.x.leftColumn<0 ) return 0; 730 aff = pSrc->pTab->aCol[pTerm->u.x.leftColumn].affinity; 731 if( !sqlite3IndexAffinityOk(pTerm->pExpr, aff) ) return 0; 732 testcase( pTerm->pExpr->op==TK_IS ); 733 return 1; 734 } 735 #endif 736 737 738 #ifndef SQLITE_OMIT_AUTOMATIC_INDEX 739 /* 740 ** Generate code to construct the Index object for an automatic index 741 ** and to set up the WhereLevel object pLevel so that the code generator 742 ** makes use of the automatic index. 743 */ 744 static void constructAutomaticIndex( 745 Parse *pParse, /* The parsing context */ 746 WhereClause *pWC, /* The WHERE clause */ 747 SrcItem *pSrc, /* The FROM clause term to get the next index */ 748 Bitmask notReady, /* Mask of cursors that are not available */ 749 WhereLevel *pLevel /* Write new index here */ 750 ){ 751 int nKeyCol; /* Number of columns in the constructed index */ 752 WhereTerm *pTerm; /* A single term of the WHERE clause */ 753 WhereTerm *pWCEnd; /* End of pWC->a[] */ 754 Index *pIdx; /* Object describing the transient index */ 755 Vdbe *v; /* Prepared statement under construction */ 756 int addrInit; /* Address of the initialization bypass jump */ 757 Table *pTable; /* The table being indexed */ 758 int addrTop; /* Top of the index fill loop */ 759 int regRecord; /* Register holding an index record */ 760 int n; /* Column counter */ 761 int i; /* Loop counter */ 762 int mxBitCol; /* Maximum column in pSrc->colUsed */ 763 CollSeq *pColl; /* Collating sequence to on a column */ 764 WhereLoop *pLoop; /* The Loop object */ 765 char *zNotUsed; /* Extra space on the end of pIdx */ 766 Bitmask idxCols; /* Bitmap of columns used for indexing */ 767 Bitmask extraCols; /* Bitmap of additional columns */ 768 u8 sentWarning = 0; /* True if a warnning has been issued */ 769 Expr *pPartial = 0; /* Partial Index Expression */ 770 int iContinue = 0; /* Jump here to skip excluded rows */ 771 SrcItem *pTabItem; /* FROM clause term being indexed */ 772 int addrCounter = 0; /* Address where integer counter is initialized */ 773 int regBase; /* Array of registers where record is assembled */ 774 775 /* Generate code to skip over the creation and initialization of the 776 ** transient index on 2nd and subsequent iterations of the loop. */ 777 v = pParse->pVdbe; 778 assert( v!=0 ); 779 addrInit = sqlite3VdbeAddOp0(v, OP_Once); VdbeCoverage(v); 780 781 /* Count the number of columns that will be added to the index 782 ** and used to match WHERE clause constraints */ 783 nKeyCol = 0; 784 pTable = pSrc->pTab; 785 pWCEnd = &pWC->a[pWC->nTerm]; 786 pLoop = pLevel->pWLoop; 787 idxCols = 0; 788 for(pTerm=pWC->a; pTerm<pWCEnd; pTerm++){ 789 Expr *pExpr = pTerm->pExpr; 790 assert( !ExprHasProperty(pExpr, EP_FromJoin) /* prereq always non-zero */ 791 || pExpr->iRightJoinTable!=pSrc->iCursor /* for the right-hand */ 792 || pLoop->prereq!=0 ); /* table of a LEFT JOIN */ 793 if( pLoop->prereq==0 794 && (pTerm->wtFlags & TERM_VIRTUAL)==0 795 && !ExprHasProperty(pExpr, EP_FromJoin) 796 && sqlite3ExprIsTableConstant(pExpr, pSrc->iCursor) ){ 797 pPartial = sqlite3ExprAnd(pParse, pPartial, 798 sqlite3ExprDup(pParse->db, pExpr, 0)); 799 } 800 if( termCanDriveIndex(pTerm, pSrc, notReady) ){ 801 int iCol = pTerm->u.x.leftColumn; 802 Bitmask cMask = iCol>=BMS ? MASKBIT(BMS-1) : MASKBIT(iCol); 803 testcase( iCol==BMS ); 804 testcase( iCol==BMS-1 ); 805 if( !sentWarning ){ 806 sqlite3_log(SQLITE_WARNING_AUTOINDEX, 807 "automatic index on %s(%s)", pTable->zName, 808 pTable->aCol[iCol].zName); 809 sentWarning = 1; 810 } 811 if( (idxCols & cMask)==0 ){ 812 if( whereLoopResize(pParse->db, pLoop, nKeyCol+1) ){ 813 goto end_auto_index_create; 814 } 815 pLoop->aLTerm[nKeyCol++] = pTerm; 816 idxCols |= cMask; 817 } 818 } 819 } 820 assert( nKeyCol>0 || pParse->db->mallocFailed ); 821 pLoop->u.btree.nEq = pLoop->nLTerm = nKeyCol; 822 pLoop->wsFlags = WHERE_COLUMN_EQ | WHERE_IDX_ONLY | WHERE_INDEXED 823 | WHERE_AUTO_INDEX; 824 825 /* Count the number of additional columns needed to create a 826 ** covering index. A "covering index" is an index that contains all 827 ** columns that are needed by the query. With a covering index, the 828 ** original table never needs to be accessed. Automatic indices must 829 ** be a covering index because the index will not be updated if the 830 ** original table changes and the index and table cannot both be used 831 ** if they go out of sync. 832 */ 833 extraCols = pSrc->colUsed & (~idxCols | MASKBIT(BMS-1)); 834 mxBitCol = MIN(BMS-1,pTable->nCol); 835 testcase( pTable->nCol==BMS-1 ); 836 testcase( pTable->nCol==BMS-2 ); 837 for(i=0; i<mxBitCol; i++){ 838 if( extraCols & MASKBIT(i) ) nKeyCol++; 839 } 840 if( pSrc->colUsed & MASKBIT(BMS-1) ){ 841 nKeyCol += pTable->nCol - BMS + 1; 842 } 843 844 /* Construct the Index object to describe this index */ 845 pIdx = sqlite3AllocateIndexObject(pParse->db, nKeyCol+1, 0, &zNotUsed); 846 if( pIdx==0 ) goto end_auto_index_create; 847 pLoop->u.btree.pIndex = pIdx; 848 pIdx->zName = "auto-index"; 849 pIdx->pTable = pTable; 850 n = 0; 851 idxCols = 0; 852 for(pTerm=pWC->a; pTerm<pWCEnd; pTerm++){ 853 if( termCanDriveIndex(pTerm, pSrc, notReady) ){ 854 int iCol = pTerm->u.x.leftColumn; 855 Bitmask cMask = iCol>=BMS ? MASKBIT(BMS-1) : MASKBIT(iCol); 856 testcase( iCol==BMS-1 ); 857 testcase( iCol==BMS ); 858 if( (idxCols & cMask)==0 ){ 859 Expr *pX = pTerm->pExpr; 860 idxCols |= cMask; 861 pIdx->aiColumn[n] = pTerm->u.x.leftColumn; 862 pColl = sqlite3ExprCompareCollSeq(pParse, pX); 863 assert( pColl!=0 || pParse->nErr>0 ); /* TH3 collate01.800 */ 864 pIdx->azColl[n] = pColl ? pColl->zName : sqlite3StrBINARY; 865 n++; 866 } 867 } 868 } 869 assert( (u32)n==pLoop->u.btree.nEq ); 870 871 /* Add additional columns needed to make the automatic index into 872 ** a covering index */ 873 for(i=0; i<mxBitCol; i++){ 874 if( extraCols & MASKBIT(i) ){ 875 pIdx->aiColumn[n] = i; 876 pIdx->azColl[n] = sqlite3StrBINARY; 877 n++; 878 } 879 } 880 if( pSrc->colUsed & MASKBIT(BMS-1) ){ 881 for(i=BMS-1; i<pTable->nCol; i++){ 882 pIdx->aiColumn[n] = i; 883 pIdx->azColl[n] = sqlite3StrBINARY; 884 n++; 885 } 886 } 887 assert( n==nKeyCol ); 888 pIdx->aiColumn[n] = XN_ROWID; 889 pIdx->azColl[n] = sqlite3StrBINARY; 890 891 /* Create the automatic index */ 892 assert( pLevel->iIdxCur>=0 ); 893 pLevel->iIdxCur = pParse->nTab++; 894 sqlite3VdbeAddOp2(v, OP_OpenAutoindex, pLevel->iIdxCur, nKeyCol+1); 895 sqlite3VdbeSetP4KeyInfo(pParse, pIdx); 896 VdbeComment((v, "for %s", pTable->zName)); 897 898 /* Fill the automatic index with content */ 899 pTabItem = &pWC->pWInfo->pTabList->a[pLevel->iFrom]; 900 if( pTabItem->fg.viaCoroutine ){ 901 int regYield = pTabItem->regReturn; 902 addrCounter = sqlite3VdbeAddOp2(v, OP_Integer, 0, 0); 903 sqlite3VdbeAddOp3(v, OP_InitCoroutine, regYield, 0, pTabItem->addrFillSub); 904 addrTop = sqlite3VdbeAddOp1(v, OP_Yield, regYield); 905 VdbeCoverage(v); 906 VdbeComment((v, "next row of %s", pTabItem->pTab->zName)); 907 }else{ 908 addrTop = sqlite3VdbeAddOp1(v, OP_Rewind, pLevel->iTabCur); VdbeCoverage(v); 909 } 910 if( pPartial ){ 911 iContinue = sqlite3VdbeMakeLabel(pParse); 912 sqlite3ExprIfFalse(pParse, pPartial, iContinue, SQLITE_JUMPIFNULL); 913 pLoop->wsFlags |= WHERE_PARTIALIDX; 914 } 915 regRecord = sqlite3GetTempReg(pParse); 916 regBase = sqlite3GenerateIndexKey( 917 pParse, pIdx, pLevel->iTabCur, regRecord, 0, 0, 0, 0 918 ); 919 sqlite3VdbeAddOp2(v, OP_IdxInsert, pLevel->iIdxCur, regRecord); 920 sqlite3VdbeChangeP5(v, OPFLAG_USESEEKRESULT); 921 if( pPartial ) sqlite3VdbeResolveLabel(v, iContinue); 922 if( pTabItem->fg.viaCoroutine ){ 923 sqlite3VdbeChangeP2(v, addrCounter, regBase+n); 924 testcase( pParse->db->mallocFailed ); 925 assert( pLevel->iIdxCur>0 ); 926 translateColumnToCopy(pParse, addrTop, pLevel->iTabCur, 927 pTabItem->regResult, pLevel->iIdxCur); 928 sqlite3VdbeGoto(v, addrTop); 929 pTabItem->fg.viaCoroutine = 0; 930 }else{ 931 sqlite3VdbeAddOp2(v, OP_Next, pLevel->iTabCur, addrTop+1); VdbeCoverage(v); 932 sqlite3VdbeChangeP5(v, SQLITE_STMTSTATUS_AUTOINDEX); 933 } 934 sqlite3VdbeJumpHere(v, addrTop); 935 sqlite3ReleaseTempReg(pParse, regRecord); 936 937 /* Jump here when skipping the initialization */ 938 sqlite3VdbeJumpHere(v, addrInit); 939 940 end_auto_index_create: 941 sqlite3ExprDelete(pParse->db, pPartial); 942 } 943 #endif /* SQLITE_OMIT_AUTOMATIC_INDEX */ 944 945 #ifndef SQLITE_OMIT_VIRTUALTABLE 946 /* 947 ** Allocate and populate an sqlite3_index_info structure. It is the 948 ** responsibility of the caller to eventually release the structure 949 ** by passing the pointer returned by this function to sqlite3_free(). 950 */ 951 static sqlite3_index_info *allocateIndexInfo( 952 Parse *pParse, /* The parsing context */ 953 WhereClause *pWC, /* The WHERE clause being analyzed */ 954 Bitmask mUnusable, /* Ignore terms with these prereqs */ 955 SrcItem *pSrc, /* The FROM clause term that is the vtab */ 956 ExprList *pOrderBy, /* The ORDER BY clause */ 957 u16 *pmNoOmit /* Mask of terms not to omit */ 958 ){ 959 int i, j; 960 int nTerm; 961 struct sqlite3_index_constraint *pIdxCons; 962 struct sqlite3_index_orderby *pIdxOrderBy; 963 struct sqlite3_index_constraint_usage *pUsage; 964 struct HiddenIndexInfo *pHidden; 965 WhereTerm *pTerm; 966 int nOrderBy; 967 sqlite3_index_info *pIdxInfo; 968 u16 mNoOmit = 0; 969 970 /* Count the number of possible WHERE clause constraints referring 971 ** to this virtual table */ 972 for(i=nTerm=0, pTerm=pWC->a; i<pWC->nTerm; i++, pTerm++){ 973 if( pTerm->leftCursor != pSrc->iCursor ) continue; 974 if( pTerm->prereqRight & mUnusable ) continue; 975 assert( IsPowerOfTwo(pTerm->eOperator & ~WO_EQUIV) ); 976 testcase( pTerm->eOperator & WO_IN ); 977 testcase( pTerm->eOperator & WO_ISNULL ); 978 testcase( pTerm->eOperator & WO_IS ); 979 testcase( pTerm->eOperator & WO_ALL ); 980 if( (pTerm->eOperator & ~(WO_EQUIV))==0 ) continue; 981 if( pTerm->wtFlags & TERM_VNULL ) continue; 982 assert( pTerm->u.x.leftColumn>=(-1) ); 983 nTerm++; 984 } 985 986 /* If the ORDER BY clause contains only columns in the current 987 ** virtual table then allocate space for the aOrderBy part of 988 ** the sqlite3_index_info structure. 989 */ 990 nOrderBy = 0; 991 if( pOrderBy ){ 992 int n = pOrderBy->nExpr; 993 for(i=0; i<n; i++){ 994 Expr *pExpr = pOrderBy->a[i].pExpr; 995 if( pExpr->op!=TK_COLUMN || pExpr->iTable!=pSrc->iCursor ) break; 996 if( pOrderBy->a[i].sortFlags & KEYINFO_ORDER_BIGNULL ) break; 997 } 998 if( i==n){ 999 nOrderBy = n; 1000 } 1001 } 1002 1003 /* Allocate the sqlite3_index_info structure 1004 */ 1005 pIdxInfo = sqlite3DbMallocZero(pParse->db, sizeof(*pIdxInfo) 1006 + (sizeof(*pIdxCons) + sizeof(*pUsage))*nTerm 1007 + sizeof(*pIdxOrderBy)*nOrderBy + sizeof(*pHidden) ); 1008 if( pIdxInfo==0 ){ 1009 sqlite3ErrorMsg(pParse, "out of memory"); 1010 return 0; 1011 } 1012 pHidden = (struct HiddenIndexInfo*)&pIdxInfo[1]; 1013 pIdxCons = (struct sqlite3_index_constraint*)&pHidden[1]; 1014 pIdxOrderBy = (struct sqlite3_index_orderby*)&pIdxCons[nTerm]; 1015 pUsage = (struct sqlite3_index_constraint_usage*)&pIdxOrderBy[nOrderBy]; 1016 pIdxInfo->nOrderBy = nOrderBy; 1017 pIdxInfo->aConstraint = pIdxCons; 1018 pIdxInfo->aOrderBy = pIdxOrderBy; 1019 pIdxInfo->aConstraintUsage = pUsage; 1020 pHidden->pWC = pWC; 1021 pHidden->pParse = pParse; 1022 for(i=j=0, pTerm=pWC->a; i<pWC->nTerm; i++, pTerm++){ 1023 u16 op; 1024 if( pTerm->leftCursor != pSrc->iCursor ) continue; 1025 if( pTerm->prereqRight & mUnusable ) continue; 1026 assert( IsPowerOfTwo(pTerm->eOperator & ~WO_EQUIV) ); 1027 testcase( pTerm->eOperator & WO_IN ); 1028 testcase( pTerm->eOperator & WO_IS ); 1029 testcase( pTerm->eOperator & WO_ISNULL ); 1030 testcase( pTerm->eOperator & WO_ALL ); 1031 if( (pTerm->eOperator & ~(WO_EQUIV))==0 ) continue; 1032 if( pTerm->wtFlags & TERM_VNULL ) continue; 1033 1034 /* tag-20191211-002: WHERE-clause constraints are not useful to the 1035 ** right-hand table of a LEFT JOIN. See tag-20191211-001 for the 1036 ** equivalent restriction for ordinary tables. */ 1037 if( (pSrc->fg.jointype & JT_LEFT)!=0 1038 && !ExprHasProperty(pTerm->pExpr, EP_FromJoin) 1039 ){ 1040 continue; 1041 } 1042 assert( pTerm->u.x.leftColumn>=(-1) ); 1043 pIdxCons[j].iColumn = pTerm->u.x.leftColumn; 1044 pIdxCons[j].iTermOffset = i; 1045 op = pTerm->eOperator & WO_ALL; 1046 if( op==WO_IN ) op = WO_EQ; 1047 if( op==WO_AUX ){ 1048 pIdxCons[j].op = pTerm->eMatchOp; 1049 }else if( op & (WO_ISNULL|WO_IS) ){ 1050 if( op==WO_ISNULL ){ 1051 pIdxCons[j].op = SQLITE_INDEX_CONSTRAINT_ISNULL; 1052 }else{ 1053 pIdxCons[j].op = SQLITE_INDEX_CONSTRAINT_IS; 1054 } 1055 }else{ 1056 pIdxCons[j].op = (u8)op; 1057 /* The direct assignment in the previous line is possible only because 1058 ** the WO_ and SQLITE_INDEX_CONSTRAINT_ codes are identical. The 1059 ** following asserts verify this fact. */ 1060 assert( WO_EQ==SQLITE_INDEX_CONSTRAINT_EQ ); 1061 assert( WO_LT==SQLITE_INDEX_CONSTRAINT_LT ); 1062 assert( WO_LE==SQLITE_INDEX_CONSTRAINT_LE ); 1063 assert( WO_GT==SQLITE_INDEX_CONSTRAINT_GT ); 1064 assert( WO_GE==SQLITE_INDEX_CONSTRAINT_GE ); 1065 assert( pTerm->eOperator&(WO_IN|WO_EQ|WO_LT|WO_LE|WO_GT|WO_GE|WO_AUX) ); 1066 1067 if( op & (WO_LT|WO_LE|WO_GT|WO_GE) 1068 && sqlite3ExprIsVector(pTerm->pExpr->pRight) 1069 ){ 1070 testcase( j!=i ); 1071 if( j<16 ) mNoOmit |= (1 << j); 1072 if( op==WO_LT ) pIdxCons[j].op = WO_LE; 1073 if( op==WO_GT ) pIdxCons[j].op = WO_GE; 1074 } 1075 } 1076 1077 j++; 1078 } 1079 pIdxInfo->nConstraint = j; 1080 for(i=0; i<nOrderBy; i++){ 1081 Expr *pExpr = pOrderBy->a[i].pExpr; 1082 pIdxOrderBy[i].iColumn = pExpr->iColumn; 1083 pIdxOrderBy[i].desc = pOrderBy->a[i].sortFlags & KEYINFO_ORDER_DESC; 1084 } 1085 1086 *pmNoOmit = mNoOmit; 1087 return pIdxInfo; 1088 } 1089 1090 /* 1091 ** The table object reference passed as the second argument to this function 1092 ** must represent a virtual table. This function invokes the xBestIndex() 1093 ** method of the virtual table with the sqlite3_index_info object that 1094 ** comes in as the 3rd argument to this function. 1095 ** 1096 ** If an error occurs, pParse is populated with an error message and an 1097 ** appropriate error code is returned. A return of SQLITE_CONSTRAINT from 1098 ** xBestIndex is not considered an error. SQLITE_CONSTRAINT indicates that 1099 ** the current configuration of "unusable" flags in sqlite3_index_info can 1100 ** not result in a valid plan. 1101 ** 1102 ** Whether or not an error is returned, it is the responsibility of the 1103 ** caller to eventually free p->idxStr if p->needToFreeIdxStr indicates 1104 ** that this is required. 1105 */ 1106 static int vtabBestIndex(Parse *pParse, Table *pTab, sqlite3_index_info *p){ 1107 sqlite3_vtab *pVtab = sqlite3GetVTable(pParse->db, pTab)->pVtab; 1108 int rc; 1109 1110 whereTraceIndexInfoInputs(p); 1111 rc = pVtab->pModule->xBestIndex(pVtab, p); 1112 whereTraceIndexInfoOutputs(p); 1113 1114 if( rc!=SQLITE_OK && rc!=SQLITE_CONSTRAINT ){ 1115 if( rc==SQLITE_NOMEM ){ 1116 sqlite3OomFault(pParse->db); 1117 }else if( !pVtab->zErrMsg ){ 1118 sqlite3ErrorMsg(pParse, "%s", sqlite3ErrStr(rc)); 1119 }else{ 1120 sqlite3ErrorMsg(pParse, "%s", pVtab->zErrMsg); 1121 } 1122 } 1123 sqlite3_free(pVtab->zErrMsg); 1124 pVtab->zErrMsg = 0; 1125 return rc; 1126 } 1127 #endif /* !defined(SQLITE_OMIT_VIRTUALTABLE) */ 1128 1129 #ifdef SQLITE_ENABLE_STAT4 1130 /* 1131 ** Estimate the location of a particular key among all keys in an 1132 ** index. Store the results in aStat as follows: 1133 ** 1134 ** aStat[0] Est. number of rows less than pRec 1135 ** aStat[1] Est. number of rows equal to pRec 1136 ** 1137 ** Return the index of the sample that is the smallest sample that 1138 ** is greater than or equal to pRec. Note that this index is not an index 1139 ** into the aSample[] array - it is an index into a virtual set of samples 1140 ** based on the contents of aSample[] and the number of fields in record 1141 ** pRec. 1142 */ 1143 static int whereKeyStats( 1144 Parse *pParse, /* Database connection */ 1145 Index *pIdx, /* Index to consider domain of */ 1146 UnpackedRecord *pRec, /* Vector of values to consider */ 1147 int roundUp, /* Round up if true. Round down if false */ 1148 tRowcnt *aStat /* OUT: stats written here */ 1149 ){ 1150 IndexSample *aSample = pIdx->aSample; 1151 int iCol; /* Index of required stats in anEq[] etc. */ 1152 int i; /* Index of first sample >= pRec */ 1153 int iSample; /* Smallest sample larger than or equal to pRec */ 1154 int iMin = 0; /* Smallest sample not yet tested */ 1155 int iTest; /* Next sample to test */ 1156 int res; /* Result of comparison operation */ 1157 int nField; /* Number of fields in pRec */ 1158 tRowcnt iLower = 0; /* anLt[] + anEq[] of largest sample pRec is > */ 1159 1160 #ifndef SQLITE_DEBUG 1161 UNUSED_PARAMETER( pParse ); 1162 #endif 1163 assert( pRec!=0 ); 1164 assert( pIdx->nSample>0 ); 1165 assert( pRec->nField>0 && pRec->nField<=pIdx->nSampleCol ); 1166 1167 /* Do a binary search to find the first sample greater than or equal 1168 ** to pRec. If pRec contains a single field, the set of samples to search 1169 ** is simply the aSample[] array. If the samples in aSample[] contain more 1170 ** than one fields, all fields following the first are ignored. 1171 ** 1172 ** If pRec contains N fields, where N is more than one, then as well as the 1173 ** samples in aSample[] (truncated to N fields), the search also has to 1174 ** consider prefixes of those samples. For example, if the set of samples 1175 ** in aSample is: 1176 ** 1177 ** aSample[0] = (a, 5) 1178 ** aSample[1] = (a, 10) 1179 ** aSample[2] = (b, 5) 1180 ** aSample[3] = (c, 100) 1181 ** aSample[4] = (c, 105) 1182 ** 1183 ** Then the search space should ideally be the samples above and the 1184 ** unique prefixes [a], [b] and [c]. But since that is hard to organize, 1185 ** the code actually searches this set: 1186 ** 1187 ** 0: (a) 1188 ** 1: (a, 5) 1189 ** 2: (a, 10) 1190 ** 3: (a, 10) 1191 ** 4: (b) 1192 ** 5: (b, 5) 1193 ** 6: (c) 1194 ** 7: (c, 100) 1195 ** 8: (c, 105) 1196 ** 9: (c, 105) 1197 ** 1198 ** For each sample in the aSample[] array, N samples are present in the 1199 ** effective sample array. In the above, samples 0 and 1 are based on 1200 ** sample aSample[0]. Samples 2 and 3 on aSample[1] etc. 1201 ** 1202 ** Often, sample i of each block of N effective samples has (i+1) fields. 1203 ** Except, each sample may be extended to ensure that it is greater than or 1204 ** equal to the previous sample in the array. For example, in the above, 1205 ** sample 2 is the first sample of a block of N samples, so at first it 1206 ** appears that it should be 1 field in size. However, that would make it 1207 ** smaller than sample 1, so the binary search would not work. As a result, 1208 ** it is extended to two fields. The duplicates that this creates do not 1209 ** cause any problems. 1210 */ 1211 nField = pRec->nField; 1212 iCol = 0; 1213 iSample = pIdx->nSample * nField; 1214 do{ 1215 int iSamp; /* Index in aSample[] of test sample */ 1216 int n; /* Number of fields in test sample */ 1217 1218 iTest = (iMin+iSample)/2; 1219 iSamp = iTest / nField; 1220 if( iSamp>0 ){ 1221 /* The proposed effective sample is a prefix of sample aSample[iSamp]. 1222 ** Specifically, the shortest prefix of at least (1 + iTest%nField) 1223 ** fields that is greater than the previous effective sample. */ 1224 for(n=(iTest % nField) + 1; n<nField; n++){ 1225 if( aSample[iSamp-1].anLt[n-1]!=aSample[iSamp].anLt[n-1] ) break; 1226 } 1227 }else{ 1228 n = iTest + 1; 1229 } 1230 1231 pRec->nField = n; 1232 res = sqlite3VdbeRecordCompare(aSample[iSamp].n, aSample[iSamp].p, pRec); 1233 if( res<0 ){ 1234 iLower = aSample[iSamp].anLt[n-1] + aSample[iSamp].anEq[n-1]; 1235 iMin = iTest+1; 1236 }else if( res==0 && n<nField ){ 1237 iLower = aSample[iSamp].anLt[n-1]; 1238 iMin = iTest+1; 1239 res = -1; 1240 }else{ 1241 iSample = iTest; 1242 iCol = n-1; 1243 } 1244 }while( res && iMin<iSample ); 1245 i = iSample / nField; 1246 1247 #ifdef SQLITE_DEBUG 1248 /* The following assert statements check that the binary search code 1249 ** above found the right answer. This block serves no purpose other 1250 ** than to invoke the asserts. */ 1251 if( pParse->db->mallocFailed==0 ){ 1252 if( res==0 ){ 1253 /* If (res==0) is true, then pRec must be equal to sample i. */ 1254 assert( i<pIdx->nSample ); 1255 assert( iCol==nField-1 ); 1256 pRec->nField = nField; 1257 assert( 0==sqlite3VdbeRecordCompare(aSample[i].n, aSample[i].p, pRec) 1258 || pParse->db->mallocFailed 1259 ); 1260 }else{ 1261 /* Unless i==pIdx->nSample, indicating that pRec is larger than 1262 ** all samples in the aSample[] array, pRec must be smaller than the 1263 ** (iCol+1) field prefix of sample i. */ 1264 assert( i<=pIdx->nSample && i>=0 ); 1265 pRec->nField = iCol+1; 1266 assert( i==pIdx->nSample 1267 || sqlite3VdbeRecordCompare(aSample[i].n, aSample[i].p, pRec)>0 1268 || pParse->db->mallocFailed ); 1269 1270 /* if i==0 and iCol==0, then record pRec is smaller than all samples 1271 ** in the aSample[] array. Otherwise, if (iCol>0) then pRec must 1272 ** be greater than or equal to the (iCol) field prefix of sample i. 1273 ** If (i>0), then pRec must also be greater than sample (i-1). */ 1274 if( iCol>0 ){ 1275 pRec->nField = iCol; 1276 assert( sqlite3VdbeRecordCompare(aSample[i].n, aSample[i].p, pRec)<=0 1277 || pParse->db->mallocFailed ); 1278 } 1279 if( i>0 ){ 1280 pRec->nField = nField; 1281 assert( sqlite3VdbeRecordCompare(aSample[i-1].n, aSample[i-1].p, pRec)<0 1282 || pParse->db->mallocFailed ); 1283 } 1284 } 1285 } 1286 #endif /* ifdef SQLITE_DEBUG */ 1287 1288 if( res==0 ){ 1289 /* Record pRec is equal to sample i */ 1290 assert( iCol==nField-1 ); 1291 aStat[0] = aSample[i].anLt[iCol]; 1292 aStat[1] = aSample[i].anEq[iCol]; 1293 }else{ 1294 /* At this point, the (iCol+1) field prefix of aSample[i] is the first 1295 ** sample that is greater than pRec. Or, if i==pIdx->nSample then pRec 1296 ** is larger than all samples in the array. */ 1297 tRowcnt iUpper, iGap; 1298 if( i>=pIdx->nSample ){ 1299 iUpper = sqlite3LogEstToInt(pIdx->aiRowLogEst[0]); 1300 }else{ 1301 iUpper = aSample[i].anLt[iCol]; 1302 } 1303 1304 if( iLower>=iUpper ){ 1305 iGap = 0; 1306 }else{ 1307 iGap = iUpper - iLower; 1308 } 1309 if( roundUp ){ 1310 iGap = (iGap*2)/3; 1311 }else{ 1312 iGap = iGap/3; 1313 } 1314 aStat[0] = iLower + iGap; 1315 aStat[1] = pIdx->aAvgEq[nField-1]; 1316 } 1317 1318 /* Restore the pRec->nField value before returning. */ 1319 pRec->nField = nField; 1320 return i; 1321 } 1322 #endif /* SQLITE_ENABLE_STAT4 */ 1323 1324 /* 1325 ** If it is not NULL, pTerm is a term that provides an upper or lower 1326 ** bound on a range scan. Without considering pTerm, it is estimated 1327 ** that the scan will visit nNew rows. This function returns the number 1328 ** estimated to be visited after taking pTerm into account. 1329 ** 1330 ** If the user explicitly specified a likelihood() value for this term, 1331 ** then the return value is the likelihood multiplied by the number of 1332 ** input rows. Otherwise, this function assumes that an "IS NOT NULL" term 1333 ** has a likelihood of 0.50, and any other term a likelihood of 0.25. 1334 */ 1335 static LogEst whereRangeAdjust(WhereTerm *pTerm, LogEst nNew){ 1336 LogEst nRet = nNew; 1337 if( pTerm ){ 1338 if( pTerm->truthProb<=0 ){ 1339 nRet += pTerm->truthProb; 1340 }else if( (pTerm->wtFlags & TERM_VNULL)==0 ){ 1341 nRet -= 20; assert( 20==sqlite3LogEst(4) ); 1342 } 1343 } 1344 return nRet; 1345 } 1346 1347 1348 #ifdef SQLITE_ENABLE_STAT4 1349 /* 1350 ** Return the affinity for a single column of an index. 1351 */ 1352 char sqlite3IndexColumnAffinity(sqlite3 *db, Index *pIdx, int iCol){ 1353 assert( iCol>=0 && iCol<pIdx->nColumn ); 1354 if( !pIdx->zColAff ){ 1355 if( sqlite3IndexAffinityStr(db, pIdx)==0 ) return SQLITE_AFF_BLOB; 1356 } 1357 assert( pIdx->zColAff[iCol]!=0 ); 1358 return pIdx->zColAff[iCol]; 1359 } 1360 #endif 1361 1362 1363 #ifdef SQLITE_ENABLE_STAT4 1364 /* 1365 ** This function is called to estimate the number of rows visited by a 1366 ** range-scan on a skip-scan index. For example: 1367 ** 1368 ** CREATE INDEX i1 ON t1(a, b, c); 1369 ** SELECT * FROM t1 WHERE a=? AND c BETWEEN ? AND ?; 1370 ** 1371 ** Value pLoop->nOut is currently set to the estimated number of rows 1372 ** visited for scanning (a=? AND b=?). This function reduces that estimate 1373 ** by some factor to account for the (c BETWEEN ? AND ?) expression based 1374 ** on the stat4 data for the index. this scan will be peformed multiple 1375 ** times (once for each (a,b) combination that matches a=?) is dealt with 1376 ** by the caller. 1377 ** 1378 ** It does this by scanning through all stat4 samples, comparing values 1379 ** extracted from pLower and pUpper with the corresponding column in each 1380 ** sample. If L and U are the number of samples found to be less than or 1381 ** equal to the values extracted from pLower and pUpper respectively, and 1382 ** N is the total number of samples, the pLoop->nOut value is adjusted 1383 ** as follows: 1384 ** 1385 ** nOut = nOut * ( min(U - L, 1) / N ) 1386 ** 1387 ** If pLower is NULL, or a value cannot be extracted from the term, L is 1388 ** set to zero. If pUpper is NULL, or a value cannot be extracted from it, 1389 ** U is set to N. 1390 ** 1391 ** Normally, this function sets *pbDone to 1 before returning. However, 1392 ** if no value can be extracted from either pLower or pUpper (and so the 1393 ** estimate of the number of rows delivered remains unchanged), *pbDone 1394 ** is left as is. 1395 ** 1396 ** If an error occurs, an SQLite error code is returned. Otherwise, 1397 ** SQLITE_OK. 1398 */ 1399 static int whereRangeSkipScanEst( 1400 Parse *pParse, /* Parsing & code generating context */ 1401 WhereTerm *pLower, /* Lower bound on the range. ex: "x>123" Might be NULL */ 1402 WhereTerm *pUpper, /* Upper bound on the range. ex: "x<455" Might be NULL */ 1403 WhereLoop *pLoop, /* Update the .nOut value of this loop */ 1404 int *pbDone /* Set to true if at least one expr. value extracted */ 1405 ){ 1406 Index *p = pLoop->u.btree.pIndex; 1407 int nEq = pLoop->u.btree.nEq; 1408 sqlite3 *db = pParse->db; 1409 int nLower = -1; 1410 int nUpper = p->nSample+1; 1411 int rc = SQLITE_OK; 1412 u8 aff = sqlite3IndexColumnAffinity(db, p, nEq); 1413 CollSeq *pColl; 1414 1415 sqlite3_value *p1 = 0; /* Value extracted from pLower */ 1416 sqlite3_value *p2 = 0; /* Value extracted from pUpper */ 1417 sqlite3_value *pVal = 0; /* Value extracted from record */ 1418 1419 pColl = sqlite3LocateCollSeq(pParse, p->azColl[nEq]); 1420 if( pLower ){ 1421 rc = sqlite3Stat4ValueFromExpr(pParse, pLower->pExpr->pRight, aff, &p1); 1422 nLower = 0; 1423 } 1424 if( pUpper && rc==SQLITE_OK ){ 1425 rc = sqlite3Stat4ValueFromExpr(pParse, pUpper->pExpr->pRight, aff, &p2); 1426 nUpper = p2 ? 0 : p->nSample; 1427 } 1428 1429 if( p1 || p2 ){ 1430 int i; 1431 int nDiff; 1432 for(i=0; rc==SQLITE_OK && i<p->nSample; i++){ 1433 rc = sqlite3Stat4Column(db, p->aSample[i].p, p->aSample[i].n, nEq, &pVal); 1434 if( rc==SQLITE_OK && p1 ){ 1435 int res = sqlite3MemCompare(p1, pVal, pColl); 1436 if( res>=0 ) nLower++; 1437 } 1438 if( rc==SQLITE_OK && p2 ){ 1439 int res = sqlite3MemCompare(p2, pVal, pColl); 1440 if( res>=0 ) nUpper++; 1441 } 1442 } 1443 nDiff = (nUpper - nLower); 1444 if( nDiff<=0 ) nDiff = 1; 1445 1446 /* If there is both an upper and lower bound specified, and the 1447 ** comparisons indicate that they are close together, use the fallback 1448 ** method (assume that the scan visits 1/64 of the rows) for estimating 1449 ** the number of rows visited. Otherwise, estimate the number of rows 1450 ** using the method described in the header comment for this function. */ 1451 if( nDiff!=1 || pUpper==0 || pLower==0 ){ 1452 int nAdjust = (sqlite3LogEst(p->nSample) - sqlite3LogEst(nDiff)); 1453 pLoop->nOut -= nAdjust; 1454 *pbDone = 1; 1455 WHERETRACE(0x10, ("range skip-scan regions: %u..%u adjust=%d est=%d\n", 1456 nLower, nUpper, nAdjust*-1, pLoop->nOut)); 1457 } 1458 1459 }else{ 1460 assert( *pbDone==0 ); 1461 } 1462 1463 sqlite3ValueFree(p1); 1464 sqlite3ValueFree(p2); 1465 sqlite3ValueFree(pVal); 1466 1467 return rc; 1468 } 1469 #endif /* SQLITE_ENABLE_STAT4 */ 1470 1471 /* 1472 ** This function is used to estimate the number of rows that will be visited 1473 ** by scanning an index for a range of values. The range may have an upper 1474 ** bound, a lower bound, or both. The WHERE clause terms that set the upper 1475 ** and lower bounds are represented by pLower and pUpper respectively. For 1476 ** example, assuming that index p is on t1(a): 1477 ** 1478 ** ... FROM t1 WHERE a > ? AND a < ? ... 1479 ** |_____| |_____| 1480 ** | | 1481 ** pLower pUpper 1482 ** 1483 ** If either of the upper or lower bound is not present, then NULL is passed in 1484 ** place of the corresponding WhereTerm. 1485 ** 1486 ** The value in (pBuilder->pNew->u.btree.nEq) is the number of the index 1487 ** column subject to the range constraint. Or, equivalently, the number of 1488 ** equality constraints optimized by the proposed index scan. For example, 1489 ** assuming index p is on t1(a, b), and the SQL query is: 1490 ** 1491 ** ... FROM t1 WHERE a = ? AND b > ? AND b < ? ... 1492 ** 1493 ** then nEq is set to 1 (as the range restricted column, b, is the second 1494 ** left-most column of the index). Or, if the query is: 1495 ** 1496 ** ... FROM t1 WHERE a > ? AND a < ? ... 1497 ** 1498 ** then nEq is set to 0. 1499 ** 1500 ** When this function is called, *pnOut is set to the sqlite3LogEst() of the 1501 ** number of rows that the index scan is expected to visit without 1502 ** considering the range constraints. If nEq is 0, then *pnOut is the number of 1503 ** rows in the index. Assuming no error occurs, *pnOut is adjusted (reduced) 1504 ** to account for the range constraints pLower and pUpper. 1505 ** 1506 ** In the absence of sqlite_stat4 ANALYZE data, or if such data cannot be 1507 ** used, a single range inequality reduces the search space by a factor of 4. 1508 ** and a pair of constraints (x>? AND x<?) reduces the expected number of 1509 ** rows visited by a factor of 64. 1510 */ 1511 static int whereRangeScanEst( 1512 Parse *pParse, /* Parsing & code generating context */ 1513 WhereLoopBuilder *pBuilder, 1514 WhereTerm *pLower, /* Lower bound on the range. ex: "x>123" Might be NULL */ 1515 WhereTerm *pUpper, /* Upper bound on the range. ex: "x<455" Might be NULL */ 1516 WhereLoop *pLoop /* Modify the .nOut and maybe .rRun fields */ 1517 ){ 1518 int rc = SQLITE_OK; 1519 int nOut = pLoop->nOut; 1520 LogEst nNew; 1521 1522 #ifdef SQLITE_ENABLE_STAT4 1523 Index *p = pLoop->u.btree.pIndex; 1524 int nEq = pLoop->u.btree.nEq; 1525 1526 if( p->nSample>0 && ALWAYS(nEq<p->nSampleCol) 1527 && OptimizationEnabled(pParse->db, SQLITE_Stat4) 1528 ){ 1529 if( nEq==pBuilder->nRecValid ){ 1530 UnpackedRecord *pRec = pBuilder->pRec; 1531 tRowcnt a[2]; 1532 int nBtm = pLoop->u.btree.nBtm; 1533 int nTop = pLoop->u.btree.nTop; 1534 1535 /* Variable iLower will be set to the estimate of the number of rows in 1536 ** the index that are less than the lower bound of the range query. The 1537 ** lower bound being the concatenation of $P and $L, where $P is the 1538 ** key-prefix formed by the nEq values matched against the nEq left-most 1539 ** columns of the index, and $L is the value in pLower. 1540 ** 1541 ** Or, if pLower is NULL or $L cannot be extracted from it (because it 1542 ** is not a simple variable or literal value), the lower bound of the 1543 ** range is $P. Due to a quirk in the way whereKeyStats() works, even 1544 ** if $L is available, whereKeyStats() is called for both ($P) and 1545 ** ($P:$L) and the larger of the two returned values is used. 1546 ** 1547 ** Similarly, iUpper is to be set to the estimate of the number of rows 1548 ** less than the upper bound of the range query. Where the upper bound 1549 ** is either ($P) or ($P:$U). Again, even if $U is available, both values 1550 ** of iUpper are requested of whereKeyStats() and the smaller used. 1551 ** 1552 ** The number of rows between the two bounds is then just iUpper-iLower. 1553 */ 1554 tRowcnt iLower; /* Rows less than the lower bound */ 1555 tRowcnt iUpper; /* Rows less than the upper bound */ 1556 int iLwrIdx = -2; /* aSample[] for the lower bound */ 1557 int iUprIdx = -1; /* aSample[] for the upper bound */ 1558 1559 if( pRec ){ 1560 testcase( pRec->nField!=pBuilder->nRecValid ); 1561 pRec->nField = pBuilder->nRecValid; 1562 } 1563 /* Determine iLower and iUpper using ($P) only. */ 1564 if( nEq==0 ){ 1565 iLower = 0; 1566 iUpper = p->nRowEst0; 1567 }else{ 1568 /* Note: this call could be optimized away - since the same values must 1569 ** have been requested when testing key $P in whereEqualScanEst(). */ 1570 whereKeyStats(pParse, p, pRec, 0, a); 1571 iLower = a[0]; 1572 iUpper = a[0] + a[1]; 1573 } 1574 1575 assert( pLower==0 || (pLower->eOperator & (WO_GT|WO_GE))!=0 ); 1576 assert( pUpper==0 || (pUpper->eOperator & (WO_LT|WO_LE))!=0 ); 1577 assert( p->aSortOrder!=0 ); 1578 if( p->aSortOrder[nEq] ){ 1579 /* The roles of pLower and pUpper are swapped for a DESC index */ 1580 SWAP(WhereTerm*, pLower, pUpper); 1581 SWAP(int, nBtm, nTop); 1582 } 1583 1584 /* If possible, improve on the iLower estimate using ($P:$L). */ 1585 if( pLower ){ 1586 int n; /* Values extracted from pExpr */ 1587 Expr *pExpr = pLower->pExpr->pRight; 1588 rc = sqlite3Stat4ProbeSetValue(pParse, p, &pRec, pExpr, nBtm, nEq, &n); 1589 if( rc==SQLITE_OK && n ){ 1590 tRowcnt iNew; 1591 u16 mask = WO_GT|WO_LE; 1592 if( sqlite3ExprVectorSize(pExpr)>n ) mask = (WO_LE|WO_LT); 1593 iLwrIdx = whereKeyStats(pParse, p, pRec, 0, a); 1594 iNew = a[0] + ((pLower->eOperator & mask) ? a[1] : 0); 1595 if( iNew>iLower ) iLower = iNew; 1596 nOut--; 1597 pLower = 0; 1598 } 1599 } 1600 1601 /* If possible, improve on the iUpper estimate using ($P:$U). */ 1602 if( pUpper ){ 1603 int n; /* Values extracted from pExpr */ 1604 Expr *pExpr = pUpper->pExpr->pRight; 1605 rc = sqlite3Stat4ProbeSetValue(pParse, p, &pRec, pExpr, nTop, nEq, &n); 1606 if( rc==SQLITE_OK && n ){ 1607 tRowcnt iNew; 1608 u16 mask = WO_GT|WO_LE; 1609 if( sqlite3ExprVectorSize(pExpr)>n ) mask = (WO_LE|WO_LT); 1610 iUprIdx = whereKeyStats(pParse, p, pRec, 1, a); 1611 iNew = a[0] + ((pUpper->eOperator & mask) ? a[1] : 0); 1612 if( iNew<iUpper ) iUpper = iNew; 1613 nOut--; 1614 pUpper = 0; 1615 } 1616 } 1617 1618 pBuilder->pRec = pRec; 1619 if( rc==SQLITE_OK ){ 1620 if( iUpper>iLower ){ 1621 nNew = sqlite3LogEst(iUpper - iLower); 1622 /* TUNING: If both iUpper and iLower are derived from the same 1623 ** sample, then assume they are 4x more selective. This brings 1624 ** the estimated selectivity more in line with what it would be 1625 ** if estimated without the use of STAT4 tables. */ 1626 if( iLwrIdx==iUprIdx ) nNew -= 20; assert( 20==sqlite3LogEst(4) ); 1627 }else{ 1628 nNew = 10; assert( 10==sqlite3LogEst(2) ); 1629 } 1630 if( nNew<nOut ){ 1631 nOut = nNew; 1632 } 1633 WHERETRACE(0x10, ("STAT4 range scan: %u..%u est=%d\n", 1634 (u32)iLower, (u32)iUpper, nOut)); 1635 } 1636 }else{ 1637 int bDone = 0; 1638 rc = whereRangeSkipScanEst(pParse, pLower, pUpper, pLoop, &bDone); 1639 if( bDone ) return rc; 1640 } 1641 } 1642 #else 1643 UNUSED_PARAMETER(pParse); 1644 UNUSED_PARAMETER(pBuilder); 1645 assert( pLower || pUpper ); 1646 #endif 1647 assert( pUpper==0 || (pUpper->wtFlags & TERM_VNULL)==0 ); 1648 nNew = whereRangeAdjust(pLower, nOut); 1649 nNew = whereRangeAdjust(pUpper, nNew); 1650 1651 /* TUNING: If there is both an upper and lower limit and neither limit 1652 ** has an application-defined likelihood(), assume the range is 1653 ** reduced by an additional 75%. This means that, by default, an open-ended 1654 ** range query (e.g. col > ?) is assumed to match 1/4 of the rows in the 1655 ** index. While a closed range (e.g. col BETWEEN ? AND ?) is estimated to 1656 ** match 1/64 of the index. */ 1657 if( pLower && pLower->truthProb>0 && pUpper && pUpper->truthProb>0 ){ 1658 nNew -= 20; 1659 } 1660 1661 nOut -= (pLower!=0) + (pUpper!=0); 1662 if( nNew<10 ) nNew = 10; 1663 if( nNew<nOut ) nOut = nNew; 1664 #if defined(WHERETRACE_ENABLED) 1665 if( pLoop->nOut>nOut ){ 1666 WHERETRACE(0x10,("Range scan lowers nOut from %d to %d\n", 1667 pLoop->nOut, nOut)); 1668 } 1669 #endif 1670 pLoop->nOut = (LogEst)nOut; 1671 return rc; 1672 } 1673 1674 #ifdef SQLITE_ENABLE_STAT4 1675 /* 1676 ** Estimate the number of rows that will be returned based on 1677 ** an equality constraint x=VALUE and where that VALUE occurs in 1678 ** the histogram data. This only works when x is the left-most 1679 ** column of an index and sqlite_stat4 histogram data is available 1680 ** for that index. When pExpr==NULL that means the constraint is 1681 ** "x IS NULL" instead of "x=VALUE". 1682 ** 1683 ** Write the estimated row count into *pnRow and return SQLITE_OK. 1684 ** If unable to make an estimate, leave *pnRow unchanged and return 1685 ** non-zero. 1686 ** 1687 ** This routine can fail if it is unable to load a collating sequence 1688 ** required for string comparison, or if unable to allocate memory 1689 ** for a UTF conversion required for comparison. The error is stored 1690 ** in the pParse structure. 1691 */ 1692 static int whereEqualScanEst( 1693 Parse *pParse, /* Parsing & code generating context */ 1694 WhereLoopBuilder *pBuilder, 1695 Expr *pExpr, /* Expression for VALUE in the x=VALUE constraint */ 1696 tRowcnt *pnRow /* Write the revised row estimate here */ 1697 ){ 1698 Index *p = pBuilder->pNew->u.btree.pIndex; 1699 int nEq = pBuilder->pNew->u.btree.nEq; 1700 UnpackedRecord *pRec = pBuilder->pRec; 1701 int rc; /* Subfunction return code */ 1702 tRowcnt a[2]; /* Statistics */ 1703 int bOk; 1704 1705 assert( nEq>=1 ); 1706 assert( nEq<=p->nColumn ); 1707 assert( p->aSample!=0 ); 1708 assert( p->nSample>0 ); 1709 assert( pBuilder->nRecValid<nEq ); 1710 1711 /* If values are not available for all fields of the index to the left 1712 ** of this one, no estimate can be made. Return SQLITE_NOTFOUND. */ 1713 if( pBuilder->nRecValid<(nEq-1) ){ 1714 return SQLITE_NOTFOUND; 1715 } 1716 1717 /* This is an optimization only. The call to sqlite3Stat4ProbeSetValue() 1718 ** below would return the same value. */ 1719 if( nEq>=p->nColumn ){ 1720 *pnRow = 1; 1721 return SQLITE_OK; 1722 } 1723 1724 rc = sqlite3Stat4ProbeSetValue(pParse, p, &pRec, pExpr, 1, nEq-1, &bOk); 1725 pBuilder->pRec = pRec; 1726 if( rc!=SQLITE_OK ) return rc; 1727 if( bOk==0 ) return SQLITE_NOTFOUND; 1728 pBuilder->nRecValid = nEq; 1729 1730 whereKeyStats(pParse, p, pRec, 0, a); 1731 WHERETRACE(0x10,("equality scan regions %s(%d): %d\n", 1732 p->zName, nEq-1, (int)a[1])); 1733 *pnRow = a[1]; 1734 1735 return rc; 1736 } 1737 #endif /* SQLITE_ENABLE_STAT4 */ 1738 1739 #ifdef SQLITE_ENABLE_STAT4 1740 /* 1741 ** Estimate the number of rows that will be returned based on 1742 ** an IN constraint where the right-hand side of the IN operator 1743 ** is a list of values. Example: 1744 ** 1745 ** WHERE x IN (1,2,3,4) 1746 ** 1747 ** Write the estimated row count into *pnRow and return SQLITE_OK. 1748 ** If unable to make an estimate, leave *pnRow unchanged and return 1749 ** non-zero. 1750 ** 1751 ** This routine can fail if it is unable to load a collating sequence 1752 ** required for string comparison, or if unable to allocate memory 1753 ** for a UTF conversion required for comparison. The error is stored 1754 ** in the pParse structure. 1755 */ 1756 static int whereInScanEst( 1757 Parse *pParse, /* Parsing & code generating context */ 1758 WhereLoopBuilder *pBuilder, 1759 ExprList *pList, /* The value list on the RHS of "x IN (v1,v2,v3,...)" */ 1760 tRowcnt *pnRow /* Write the revised row estimate here */ 1761 ){ 1762 Index *p = pBuilder->pNew->u.btree.pIndex; 1763 i64 nRow0 = sqlite3LogEstToInt(p->aiRowLogEst[0]); 1764 int nRecValid = pBuilder->nRecValid; 1765 int rc = SQLITE_OK; /* Subfunction return code */ 1766 tRowcnt nEst; /* Number of rows for a single term */ 1767 tRowcnt nRowEst = 0; /* New estimate of the number of rows */ 1768 int i; /* Loop counter */ 1769 1770 assert( p->aSample!=0 ); 1771 for(i=0; rc==SQLITE_OK && i<pList->nExpr; i++){ 1772 nEst = nRow0; 1773 rc = whereEqualScanEst(pParse, pBuilder, pList->a[i].pExpr, &nEst); 1774 nRowEst += nEst; 1775 pBuilder->nRecValid = nRecValid; 1776 } 1777 1778 if( rc==SQLITE_OK ){ 1779 if( nRowEst > nRow0 ) nRowEst = nRow0; 1780 *pnRow = nRowEst; 1781 WHERETRACE(0x10,("IN row estimate: est=%d\n", nRowEst)); 1782 } 1783 assert( pBuilder->nRecValid==nRecValid ); 1784 return rc; 1785 } 1786 #endif /* SQLITE_ENABLE_STAT4 */ 1787 1788 1789 #ifdef WHERETRACE_ENABLED 1790 /* 1791 ** Print the content of a WhereTerm object 1792 */ 1793 void sqlite3WhereTermPrint(WhereTerm *pTerm, int iTerm){ 1794 if( pTerm==0 ){ 1795 sqlite3DebugPrintf("TERM-%-3d NULL\n", iTerm); 1796 }else{ 1797 char zType[8]; 1798 char zLeft[50]; 1799 memcpy(zType, "....", 5); 1800 if( pTerm->wtFlags & TERM_VIRTUAL ) zType[0] = 'V'; 1801 if( pTerm->eOperator & WO_EQUIV ) zType[1] = 'E'; 1802 if( ExprHasProperty(pTerm->pExpr, EP_FromJoin) ) zType[2] = 'L'; 1803 if( pTerm->wtFlags & TERM_CODED ) zType[3] = 'C'; 1804 if( pTerm->eOperator & WO_SINGLE ){ 1805 sqlite3_snprintf(sizeof(zLeft),zLeft,"left={%d:%d}", 1806 pTerm->leftCursor, pTerm->u.x.leftColumn); 1807 }else if( (pTerm->eOperator & WO_OR)!=0 && pTerm->u.pOrInfo!=0 ){ 1808 sqlite3_snprintf(sizeof(zLeft),zLeft,"indexable=0x%llx", 1809 pTerm->u.pOrInfo->indexable); 1810 }else{ 1811 sqlite3_snprintf(sizeof(zLeft),zLeft,"left=%d", pTerm->leftCursor); 1812 } 1813 sqlite3DebugPrintf( 1814 "TERM-%-3d %p %s %-12s op=%03x wtFlags=%04x", 1815 iTerm, pTerm, zType, zLeft, pTerm->eOperator, pTerm->wtFlags); 1816 /* The 0x10000 .wheretrace flag causes extra information to be 1817 ** shown about each Term */ 1818 if( sqlite3WhereTrace & 0x10000 ){ 1819 sqlite3DebugPrintf(" prob=%-3d prereq=%llx,%llx", 1820 pTerm->truthProb, (u64)pTerm->prereqAll, (u64)pTerm->prereqRight); 1821 } 1822 if( pTerm->u.x.iField ){ 1823 sqlite3DebugPrintf(" iField=%d", pTerm->u.x.iField); 1824 } 1825 if( pTerm->iParent>=0 ){ 1826 sqlite3DebugPrintf(" iParent=%d", pTerm->iParent); 1827 } 1828 sqlite3DebugPrintf("\n"); 1829 sqlite3TreeViewExpr(0, pTerm->pExpr, 0); 1830 } 1831 } 1832 #endif 1833 1834 #ifdef WHERETRACE_ENABLED 1835 /* 1836 ** Show the complete content of a WhereClause 1837 */ 1838 void sqlite3WhereClausePrint(WhereClause *pWC){ 1839 int i; 1840 for(i=0; i<pWC->nTerm; i++){ 1841 sqlite3WhereTermPrint(&pWC->a[i], i); 1842 } 1843 } 1844 #endif 1845 1846 #ifdef WHERETRACE_ENABLED 1847 /* 1848 ** Print a WhereLoop object for debugging purposes 1849 */ 1850 void sqlite3WhereLoopPrint(WhereLoop *p, WhereClause *pWC){ 1851 WhereInfo *pWInfo = pWC->pWInfo; 1852 int nb = 1+(pWInfo->pTabList->nSrc+3)/4; 1853 SrcItem *pItem = pWInfo->pTabList->a + p->iTab; 1854 Table *pTab = pItem->pTab; 1855 Bitmask mAll = (((Bitmask)1)<<(nb*4)) - 1; 1856 sqlite3DebugPrintf("%c%2d.%0*llx.%0*llx", p->cId, 1857 p->iTab, nb, p->maskSelf, nb, p->prereq & mAll); 1858 sqlite3DebugPrintf(" %12s", 1859 pItem->zAlias ? pItem->zAlias : pTab->zName); 1860 if( (p->wsFlags & WHERE_VIRTUALTABLE)==0 ){ 1861 const char *zName; 1862 if( p->u.btree.pIndex && (zName = p->u.btree.pIndex->zName)!=0 ){ 1863 if( strncmp(zName, "sqlite_autoindex_", 17)==0 ){ 1864 int i = sqlite3Strlen30(zName) - 1; 1865 while( zName[i]!='_' ) i--; 1866 zName += i; 1867 } 1868 sqlite3DebugPrintf(".%-16s %2d", zName, p->u.btree.nEq); 1869 }else{ 1870 sqlite3DebugPrintf("%20s",""); 1871 } 1872 }else{ 1873 char *z; 1874 if( p->u.vtab.idxStr ){ 1875 z = sqlite3_mprintf("(%d,\"%s\",%#x)", 1876 p->u.vtab.idxNum, p->u.vtab.idxStr, p->u.vtab.omitMask); 1877 }else{ 1878 z = sqlite3_mprintf("(%d,%x)", p->u.vtab.idxNum, p->u.vtab.omitMask); 1879 } 1880 sqlite3DebugPrintf(" %-19s", z); 1881 sqlite3_free(z); 1882 } 1883 if( p->wsFlags & WHERE_SKIPSCAN ){ 1884 sqlite3DebugPrintf(" f %05x %d-%d", p->wsFlags, p->nLTerm,p->nSkip); 1885 }else{ 1886 sqlite3DebugPrintf(" f %05x N %d", p->wsFlags, p->nLTerm); 1887 } 1888 sqlite3DebugPrintf(" cost %d,%d,%d\n", p->rSetup, p->rRun, p->nOut); 1889 if( p->nLTerm && (sqlite3WhereTrace & 0x100)!=0 ){ 1890 int i; 1891 for(i=0; i<p->nLTerm; i++){ 1892 sqlite3WhereTermPrint(p->aLTerm[i], i); 1893 } 1894 } 1895 } 1896 #endif 1897 1898 /* 1899 ** Convert bulk memory into a valid WhereLoop that can be passed 1900 ** to whereLoopClear harmlessly. 1901 */ 1902 static void whereLoopInit(WhereLoop *p){ 1903 p->aLTerm = p->aLTermSpace; 1904 p->nLTerm = 0; 1905 p->nLSlot = ArraySize(p->aLTermSpace); 1906 p->wsFlags = 0; 1907 } 1908 1909 /* 1910 ** Clear the WhereLoop.u union. Leave WhereLoop.pLTerm intact. 1911 */ 1912 static void whereLoopClearUnion(sqlite3 *db, WhereLoop *p){ 1913 if( p->wsFlags & (WHERE_VIRTUALTABLE|WHERE_AUTO_INDEX) ){ 1914 if( (p->wsFlags & WHERE_VIRTUALTABLE)!=0 && p->u.vtab.needFree ){ 1915 sqlite3_free(p->u.vtab.idxStr); 1916 p->u.vtab.needFree = 0; 1917 p->u.vtab.idxStr = 0; 1918 }else if( (p->wsFlags & WHERE_AUTO_INDEX)!=0 && p->u.btree.pIndex!=0 ){ 1919 sqlite3DbFree(db, p->u.btree.pIndex->zColAff); 1920 sqlite3DbFreeNN(db, p->u.btree.pIndex); 1921 p->u.btree.pIndex = 0; 1922 } 1923 } 1924 } 1925 1926 /* 1927 ** Deallocate internal memory used by a WhereLoop object 1928 */ 1929 static void whereLoopClear(sqlite3 *db, WhereLoop *p){ 1930 if( p->aLTerm!=p->aLTermSpace ) sqlite3DbFreeNN(db, p->aLTerm); 1931 whereLoopClearUnion(db, p); 1932 whereLoopInit(p); 1933 } 1934 1935 /* 1936 ** Increase the memory allocation for pLoop->aLTerm[] to be at least n. 1937 */ 1938 static int whereLoopResize(sqlite3 *db, WhereLoop *p, int n){ 1939 WhereTerm **paNew; 1940 if( p->nLSlot>=n ) return SQLITE_OK; 1941 n = (n+7)&~7; 1942 paNew = sqlite3DbMallocRawNN(db, sizeof(p->aLTerm[0])*n); 1943 if( paNew==0 ) return SQLITE_NOMEM_BKPT; 1944 memcpy(paNew, p->aLTerm, sizeof(p->aLTerm[0])*p->nLSlot); 1945 if( p->aLTerm!=p->aLTermSpace ) sqlite3DbFreeNN(db, p->aLTerm); 1946 p->aLTerm = paNew; 1947 p->nLSlot = n; 1948 return SQLITE_OK; 1949 } 1950 1951 /* 1952 ** Transfer content from the second pLoop into the first. 1953 */ 1954 static int whereLoopXfer(sqlite3 *db, WhereLoop *pTo, WhereLoop *pFrom){ 1955 whereLoopClearUnion(db, pTo); 1956 if( whereLoopResize(db, pTo, pFrom->nLTerm) ){ 1957 memset(pTo, 0, WHERE_LOOP_XFER_SZ); 1958 return SQLITE_NOMEM_BKPT; 1959 } 1960 memcpy(pTo, pFrom, WHERE_LOOP_XFER_SZ); 1961 memcpy(pTo->aLTerm, pFrom->aLTerm, pTo->nLTerm*sizeof(pTo->aLTerm[0])); 1962 if( pFrom->wsFlags & WHERE_VIRTUALTABLE ){ 1963 pFrom->u.vtab.needFree = 0; 1964 }else if( (pFrom->wsFlags & WHERE_AUTO_INDEX)!=0 ){ 1965 pFrom->u.btree.pIndex = 0; 1966 } 1967 return SQLITE_OK; 1968 } 1969 1970 /* 1971 ** Delete a WhereLoop object 1972 */ 1973 static void whereLoopDelete(sqlite3 *db, WhereLoop *p){ 1974 whereLoopClear(db, p); 1975 sqlite3DbFreeNN(db, p); 1976 } 1977 1978 /* 1979 ** Free a WhereInfo structure 1980 */ 1981 static void whereInfoFree(sqlite3 *db, WhereInfo *pWInfo){ 1982 int i; 1983 assert( pWInfo!=0 ); 1984 for(i=0; i<pWInfo->nLevel; i++){ 1985 WhereLevel *pLevel = &pWInfo->a[i]; 1986 if( pLevel->pWLoop && (pLevel->pWLoop->wsFlags & WHERE_IN_ABLE) ){ 1987 sqlite3DbFree(db, pLevel->u.in.aInLoop); 1988 } 1989 } 1990 sqlite3WhereClauseClear(&pWInfo->sWC); 1991 while( pWInfo->pLoops ){ 1992 WhereLoop *p = pWInfo->pLoops; 1993 pWInfo->pLoops = p->pNextLoop; 1994 whereLoopDelete(db, p); 1995 } 1996 assert( pWInfo->pExprMods==0 ); 1997 sqlite3DbFreeNN(db, pWInfo); 1998 } 1999 2000 /* Undo all Expr node modifications 2001 */ 2002 static void whereUndoExprMods(WhereInfo *pWInfo){ 2003 while( pWInfo->pExprMods ){ 2004 WhereExprMod *p = pWInfo->pExprMods; 2005 pWInfo->pExprMods = p->pNext; 2006 memcpy(p->pExpr, &p->orig, sizeof(p->orig)); 2007 sqlite3DbFree(pWInfo->pParse->db, p); 2008 } 2009 } 2010 2011 /* 2012 ** Return TRUE if all of the following are true: 2013 ** 2014 ** (1) X has the same or lower cost that Y 2015 ** (2) X uses fewer WHERE clause terms than Y 2016 ** (3) Every WHERE clause term used by X is also used by Y 2017 ** (4) X skips at least as many columns as Y 2018 ** (5) If X is a covering index, than Y is too 2019 ** 2020 ** Conditions (2) and (3) mean that X is a "proper subset" of Y. 2021 ** If X is a proper subset of Y then Y is a better choice and ought 2022 ** to have a lower cost. This routine returns TRUE when that cost 2023 ** relationship is inverted and needs to be adjusted. Constraint (4) 2024 ** was added because if X uses skip-scan less than Y it still might 2025 ** deserve a lower cost even if it is a proper subset of Y. Constraint (5) 2026 ** was added because a covering index probably deserves to have a lower cost 2027 ** than a non-covering index even if it is a proper subset. 2028 */ 2029 static int whereLoopCheaperProperSubset( 2030 const WhereLoop *pX, /* First WhereLoop to compare */ 2031 const WhereLoop *pY /* Compare against this WhereLoop */ 2032 ){ 2033 int i, j; 2034 if( pX->nLTerm-pX->nSkip >= pY->nLTerm-pY->nSkip ){ 2035 return 0; /* X is not a subset of Y */ 2036 } 2037 if( pY->nSkip > pX->nSkip ) return 0; 2038 if( pX->rRun >= pY->rRun ){ 2039 if( pX->rRun > pY->rRun ) return 0; /* X costs more than Y */ 2040 if( pX->nOut > pY->nOut ) return 0; /* X costs more than Y */ 2041 } 2042 for(i=pX->nLTerm-1; i>=0; i--){ 2043 if( pX->aLTerm[i]==0 ) continue; 2044 for(j=pY->nLTerm-1; j>=0; j--){ 2045 if( pY->aLTerm[j]==pX->aLTerm[i] ) break; 2046 } 2047 if( j<0 ) return 0; /* X not a subset of Y since term X[i] not used by Y */ 2048 } 2049 if( (pX->wsFlags&WHERE_IDX_ONLY)!=0 2050 && (pY->wsFlags&WHERE_IDX_ONLY)==0 ){ 2051 return 0; /* Constraint (5) */ 2052 } 2053 return 1; /* All conditions meet */ 2054 } 2055 2056 /* 2057 ** Try to adjust the cost of WhereLoop pTemplate upwards or downwards so 2058 ** that: 2059 ** 2060 ** (1) pTemplate costs less than any other WhereLoops that are a proper 2061 ** subset of pTemplate 2062 ** 2063 ** (2) pTemplate costs more than any other WhereLoops for which pTemplate 2064 ** is a proper subset. 2065 ** 2066 ** To say "WhereLoop X is a proper subset of Y" means that X uses fewer 2067 ** WHERE clause terms than Y and that every WHERE clause term used by X is 2068 ** also used by Y. 2069 */ 2070 static void whereLoopAdjustCost(const WhereLoop *p, WhereLoop *pTemplate){ 2071 if( (pTemplate->wsFlags & WHERE_INDEXED)==0 ) return; 2072 for(; p; p=p->pNextLoop){ 2073 if( p->iTab!=pTemplate->iTab ) continue; 2074 if( (p->wsFlags & WHERE_INDEXED)==0 ) continue; 2075 if( whereLoopCheaperProperSubset(p, pTemplate) ){ 2076 /* Adjust pTemplate cost downward so that it is cheaper than its 2077 ** subset p. */ 2078 WHERETRACE(0x80,("subset cost adjustment %d,%d to %d,%d\n", 2079 pTemplate->rRun, pTemplate->nOut, p->rRun, p->nOut-1)); 2080 pTemplate->rRun = p->rRun; 2081 pTemplate->nOut = p->nOut - 1; 2082 }else if( whereLoopCheaperProperSubset(pTemplate, p) ){ 2083 /* Adjust pTemplate cost upward so that it is costlier than p since 2084 ** pTemplate is a proper subset of p */ 2085 WHERETRACE(0x80,("subset cost adjustment %d,%d to %d,%d\n", 2086 pTemplate->rRun, pTemplate->nOut, p->rRun, p->nOut+1)); 2087 pTemplate->rRun = p->rRun; 2088 pTemplate->nOut = p->nOut + 1; 2089 } 2090 } 2091 } 2092 2093 /* 2094 ** Search the list of WhereLoops in *ppPrev looking for one that can be 2095 ** replaced by pTemplate. 2096 ** 2097 ** Return NULL if pTemplate does not belong on the WhereLoop list. 2098 ** In other words if pTemplate ought to be dropped from further consideration. 2099 ** 2100 ** If pX is a WhereLoop that pTemplate can replace, then return the 2101 ** link that points to pX. 2102 ** 2103 ** If pTemplate cannot replace any existing element of the list but needs 2104 ** to be added to the list as a new entry, then return a pointer to the 2105 ** tail of the list. 2106 */ 2107 static WhereLoop **whereLoopFindLesser( 2108 WhereLoop **ppPrev, 2109 const WhereLoop *pTemplate 2110 ){ 2111 WhereLoop *p; 2112 for(p=(*ppPrev); p; ppPrev=&p->pNextLoop, p=*ppPrev){ 2113 if( p->iTab!=pTemplate->iTab || p->iSortIdx!=pTemplate->iSortIdx ){ 2114 /* If either the iTab or iSortIdx values for two WhereLoop are different 2115 ** then those WhereLoops need to be considered separately. Neither is 2116 ** a candidate to replace the other. */ 2117 continue; 2118 } 2119 /* In the current implementation, the rSetup value is either zero 2120 ** or the cost of building an automatic index (NlogN) and the NlogN 2121 ** is the same for compatible WhereLoops. */ 2122 assert( p->rSetup==0 || pTemplate->rSetup==0 2123 || p->rSetup==pTemplate->rSetup ); 2124 2125 /* whereLoopAddBtree() always generates and inserts the automatic index 2126 ** case first. Hence compatible candidate WhereLoops never have a larger 2127 ** rSetup. Call this SETUP-INVARIANT */ 2128 assert( p->rSetup>=pTemplate->rSetup ); 2129 2130 /* Any loop using an appliation-defined index (or PRIMARY KEY or 2131 ** UNIQUE constraint) with one or more == constraints is better 2132 ** than an automatic index. Unless it is a skip-scan. */ 2133 if( (p->wsFlags & WHERE_AUTO_INDEX)!=0 2134 && (pTemplate->nSkip)==0 2135 && (pTemplate->wsFlags & WHERE_INDEXED)!=0 2136 && (pTemplate->wsFlags & WHERE_COLUMN_EQ)!=0 2137 && (p->prereq & pTemplate->prereq)==pTemplate->prereq 2138 ){ 2139 break; 2140 } 2141 2142 /* If existing WhereLoop p is better than pTemplate, pTemplate can be 2143 ** discarded. WhereLoop p is better if: 2144 ** (1) p has no more dependencies than pTemplate, and 2145 ** (2) p has an equal or lower cost than pTemplate 2146 */ 2147 if( (p->prereq & pTemplate->prereq)==p->prereq /* (1) */ 2148 && p->rSetup<=pTemplate->rSetup /* (2a) */ 2149 && p->rRun<=pTemplate->rRun /* (2b) */ 2150 && p->nOut<=pTemplate->nOut /* (2c) */ 2151 ){ 2152 return 0; /* Discard pTemplate */ 2153 } 2154 2155 /* If pTemplate is always better than p, then cause p to be overwritten 2156 ** with pTemplate. pTemplate is better than p if: 2157 ** (1) pTemplate has no more dependences than p, and 2158 ** (2) pTemplate has an equal or lower cost than p. 2159 */ 2160 if( (p->prereq & pTemplate->prereq)==pTemplate->prereq /* (1) */ 2161 && p->rRun>=pTemplate->rRun /* (2a) */ 2162 && p->nOut>=pTemplate->nOut /* (2b) */ 2163 ){ 2164 assert( p->rSetup>=pTemplate->rSetup ); /* SETUP-INVARIANT above */ 2165 break; /* Cause p to be overwritten by pTemplate */ 2166 } 2167 } 2168 return ppPrev; 2169 } 2170 2171 /* 2172 ** Insert or replace a WhereLoop entry using the template supplied. 2173 ** 2174 ** An existing WhereLoop entry might be overwritten if the new template 2175 ** is better and has fewer dependencies. Or the template will be ignored 2176 ** and no insert will occur if an existing WhereLoop is faster and has 2177 ** fewer dependencies than the template. Otherwise a new WhereLoop is 2178 ** added based on the template. 2179 ** 2180 ** If pBuilder->pOrSet is not NULL then we care about only the 2181 ** prerequisites and rRun and nOut costs of the N best loops. That 2182 ** information is gathered in the pBuilder->pOrSet object. This special 2183 ** processing mode is used only for OR clause processing. 2184 ** 2185 ** When accumulating multiple loops (when pBuilder->pOrSet is NULL) we 2186 ** still might overwrite similar loops with the new template if the 2187 ** new template is better. Loops may be overwritten if the following 2188 ** conditions are met: 2189 ** 2190 ** (1) They have the same iTab. 2191 ** (2) They have the same iSortIdx. 2192 ** (3) The template has same or fewer dependencies than the current loop 2193 ** (4) The template has the same or lower cost than the current loop 2194 */ 2195 static int whereLoopInsert(WhereLoopBuilder *pBuilder, WhereLoop *pTemplate){ 2196 WhereLoop **ppPrev, *p; 2197 WhereInfo *pWInfo = pBuilder->pWInfo; 2198 sqlite3 *db = pWInfo->pParse->db; 2199 int rc; 2200 2201 /* Stop the search once we hit the query planner search limit */ 2202 if( pBuilder->iPlanLimit==0 ){ 2203 WHERETRACE(0xffffffff,("=== query planner search limit reached ===\n")); 2204 if( pBuilder->pOrSet ) pBuilder->pOrSet->n = 0; 2205 return SQLITE_DONE; 2206 } 2207 pBuilder->iPlanLimit--; 2208 2209 whereLoopAdjustCost(pWInfo->pLoops, pTemplate); 2210 2211 /* If pBuilder->pOrSet is defined, then only keep track of the costs 2212 ** and prereqs. 2213 */ 2214 if( pBuilder->pOrSet!=0 ){ 2215 if( pTemplate->nLTerm ){ 2216 #if WHERETRACE_ENABLED 2217 u16 n = pBuilder->pOrSet->n; 2218 int x = 2219 #endif 2220 whereOrInsert(pBuilder->pOrSet, pTemplate->prereq, pTemplate->rRun, 2221 pTemplate->nOut); 2222 #if WHERETRACE_ENABLED /* 0x8 */ 2223 if( sqlite3WhereTrace & 0x8 ){ 2224 sqlite3DebugPrintf(x?" or-%d: ":" or-X: ", n); 2225 sqlite3WhereLoopPrint(pTemplate, pBuilder->pWC); 2226 } 2227 #endif 2228 } 2229 return SQLITE_OK; 2230 } 2231 2232 /* Look for an existing WhereLoop to replace with pTemplate 2233 */ 2234 ppPrev = whereLoopFindLesser(&pWInfo->pLoops, pTemplate); 2235 2236 if( ppPrev==0 ){ 2237 /* There already exists a WhereLoop on the list that is better 2238 ** than pTemplate, so just ignore pTemplate */ 2239 #if WHERETRACE_ENABLED /* 0x8 */ 2240 if( sqlite3WhereTrace & 0x8 ){ 2241 sqlite3DebugPrintf(" skip: "); 2242 sqlite3WhereLoopPrint(pTemplate, pBuilder->pWC); 2243 } 2244 #endif 2245 return SQLITE_OK; 2246 }else{ 2247 p = *ppPrev; 2248 } 2249 2250 /* If we reach this point it means that either p[] should be overwritten 2251 ** with pTemplate[] if p[] exists, or if p==NULL then allocate a new 2252 ** WhereLoop and insert it. 2253 */ 2254 #if WHERETRACE_ENABLED /* 0x8 */ 2255 if( sqlite3WhereTrace & 0x8 ){ 2256 if( p!=0 ){ 2257 sqlite3DebugPrintf("replace: "); 2258 sqlite3WhereLoopPrint(p, pBuilder->pWC); 2259 sqlite3DebugPrintf(" with: "); 2260 }else{ 2261 sqlite3DebugPrintf(" add: "); 2262 } 2263 sqlite3WhereLoopPrint(pTemplate, pBuilder->pWC); 2264 } 2265 #endif 2266 if( p==0 ){ 2267 /* Allocate a new WhereLoop to add to the end of the list */ 2268 *ppPrev = p = sqlite3DbMallocRawNN(db, sizeof(WhereLoop)); 2269 if( p==0 ) return SQLITE_NOMEM_BKPT; 2270 whereLoopInit(p); 2271 p->pNextLoop = 0; 2272 }else{ 2273 /* We will be overwriting WhereLoop p[]. But before we do, first 2274 ** go through the rest of the list and delete any other entries besides 2275 ** p[] that are also supplated by pTemplate */ 2276 WhereLoop **ppTail = &p->pNextLoop; 2277 WhereLoop *pToDel; 2278 while( *ppTail ){ 2279 ppTail = whereLoopFindLesser(ppTail, pTemplate); 2280 if( ppTail==0 ) break; 2281 pToDel = *ppTail; 2282 if( pToDel==0 ) break; 2283 *ppTail = pToDel->pNextLoop; 2284 #if WHERETRACE_ENABLED /* 0x8 */ 2285 if( sqlite3WhereTrace & 0x8 ){ 2286 sqlite3DebugPrintf(" delete: "); 2287 sqlite3WhereLoopPrint(pToDel, pBuilder->pWC); 2288 } 2289 #endif 2290 whereLoopDelete(db, pToDel); 2291 } 2292 } 2293 rc = whereLoopXfer(db, p, pTemplate); 2294 if( (p->wsFlags & WHERE_VIRTUALTABLE)==0 ){ 2295 Index *pIndex = p->u.btree.pIndex; 2296 if( pIndex && pIndex->idxType==SQLITE_IDXTYPE_IPK ){ 2297 p->u.btree.pIndex = 0; 2298 } 2299 } 2300 return rc; 2301 } 2302 2303 /* 2304 ** Adjust the WhereLoop.nOut value downward to account for terms of the 2305 ** WHERE clause that reference the loop but which are not used by an 2306 ** index. 2307 * 2308 ** For every WHERE clause term that is not used by the index 2309 ** and which has a truth probability assigned by one of the likelihood(), 2310 ** likely(), or unlikely() SQL functions, reduce the estimated number 2311 ** of output rows by the probability specified. 2312 ** 2313 ** TUNING: For every WHERE clause term that is not used by the index 2314 ** and which does not have an assigned truth probability, heuristics 2315 ** described below are used to try to estimate the truth probability. 2316 ** TODO --> Perhaps this is something that could be improved by better 2317 ** table statistics. 2318 ** 2319 ** Heuristic 1: Estimate the truth probability as 93.75%. The 93.75% 2320 ** value corresponds to -1 in LogEst notation, so this means decrement 2321 ** the WhereLoop.nOut field for every such WHERE clause term. 2322 ** 2323 ** Heuristic 2: If there exists one or more WHERE clause terms of the 2324 ** form "x==EXPR" and EXPR is not a constant 0 or 1, then make sure the 2325 ** final output row estimate is no greater than 1/4 of the total number 2326 ** of rows in the table. In other words, assume that x==EXPR will filter 2327 ** out at least 3 out of 4 rows. If EXPR is -1 or 0 or 1, then maybe the 2328 ** "x" column is boolean or else -1 or 0 or 1 is a common default value 2329 ** on the "x" column and so in that case only cap the output row estimate 2330 ** at 1/2 instead of 1/4. 2331 */ 2332 static void whereLoopOutputAdjust( 2333 WhereClause *pWC, /* The WHERE clause */ 2334 WhereLoop *pLoop, /* The loop to adjust downward */ 2335 LogEst nRow /* Number of rows in the entire table */ 2336 ){ 2337 WhereTerm *pTerm, *pX; 2338 Bitmask notAllowed = ~(pLoop->prereq|pLoop->maskSelf); 2339 int i, j; 2340 LogEst iReduce = 0; /* pLoop->nOut should not exceed nRow-iReduce */ 2341 2342 assert( (pLoop->wsFlags & WHERE_AUTO_INDEX)==0 ); 2343 for(i=pWC->nTerm, pTerm=pWC->a; i>0; i--, pTerm++){ 2344 assert( pTerm!=0 ); 2345 if( (pTerm->wtFlags & TERM_VIRTUAL)!=0 ) break; 2346 if( (pTerm->prereqAll & pLoop->maskSelf)==0 ) continue; 2347 if( (pTerm->prereqAll & notAllowed)!=0 ) continue; 2348 for(j=pLoop->nLTerm-1; j>=0; j--){ 2349 pX = pLoop->aLTerm[j]; 2350 if( pX==0 ) continue; 2351 if( pX==pTerm ) break; 2352 if( pX->iParent>=0 && (&pWC->a[pX->iParent])==pTerm ) break; 2353 } 2354 if( j<0 ){ 2355 if( pTerm->truthProb<=0 ){ 2356 /* If a truth probability is specified using the likelihood() hints, 2357 ** then use the probability provided by the application. */ 2358 pLoop->nOut += pTerm->truthProb; 2359 }else{ 2360 /* In the absence of explicit truth probabilities, use heuristics to 2361 ** guess a reasonable truth probability. */ 2362 pLoop->nOut--; 2363 if( (pTerm->eOperator&(WO_EQ|WO_IS))!=0 2364 && (pTerm->wtFlags & TERM_HIGHTRUTH)==0 /* tag-20200224-1 */ 2365 ){ 2366 Expr *pRight = pTerm->pExpr->pRight; 2367 int k = 0; 2368 testcase( pTerm->pExpr->op==TK_IS ); 2369 if( sqlite3ExprIsInteger(pRight, &k) && k>=(-1) && k<=1 ){ 2370 k = 10; 2371 }else{ 2372 k = 20; 2373 } 2374 if( iReduce<k ){ 2375 pTerm->wtFlags |= TERM_HEURTRUTH; 2376 iReduce = k; 2377 } 2378 } 2379 } 2380 } 2381 } 2382 if( pLoop->nOut > nRow-iReduce ) pLoop->nOut = nRow - iReduce; 2383 } 2384 2385 /* 2386 ** Term pTerm is a vector range comparison operation. The first comparison 2387 ** in the vector can be optimized using column nEq of the index. This 2388 ** function returns the total number of vector elements that can be used 2389 ** as part of the range comparison. 2390 ** 2391 ** For example, if the query is: 2392 ** 2393 ** WHERE a = ? AND (b, c, d) > (?, ?, ?) 2394 ** 2395 ** and the index: 2396 ** 2397 ** CREATE INDEX ... ON (a, b, c, d, e) 2398 ** 2399 ** then this function would be invoked with nEq=1. The value returned in 2400 ** this case is 3. 2401 */ 2402 static int whereRangeVectorLen( 2403 Parse *pParse, /* Parsing context */ 2404 int iCur, /* Cursor open on pIdx */ 2405 Index *pIdx, /* The index to be used for a inequality constraint */ 2406 int nEq, /* Number of prior equality constraints on same index */ 2407 WhereTerm *pTerm /* The vector inequality constraint */ 2408 ){ 2409 int nCmp = sqlite3ExprVectorSize(pTerm->pExpr->pLeft); 2410 int i; 2411 2412 nCmp = MIN(nCmp, (pIdx->nColumn - nEq)); 2413 for(i=1; i<nCmp; i++){ 2414 /* Test if comparison i of pTerm is compatible with column (i+nEq) 2415 ** of the index. If not, exit the loop. */ 2416 char aff; /* Comparison affinity */ 2417 char idxaff = 0; /* Indexed columns affinity */ 2418 CollSeq *pColl; /* Comparison collation sequence */ 2419 Expr *pLhs = pTerm->pExpr->pLeft->x.pList->a[i].pExpr; 2420 Expr *pRhs = pTerm->pExpr->pRight; 2421 if( pRhs->flags & EP_xIsSelect ){ 2422 pRhs = pRhs->x.pSelect->pEList->a[i].pExpr; 2423 }else{ 2424 pRhs = pRhs->x.pList->a[i].pExpr; 2425 } 2426 2427 /* Check that the LHS of the comparison is a column reference to 2428 ** the right column of the right source table. And that the sort 2429 ** order of the index column is the same as the sort order of the 2430 ** leftmost index column. */ 2431 if( pLhs->op!=TK_COLUMN 2432 || pLhs->iTable!=iCur 2433 || pLhs->iColumn!=pIdx->aiColumn[i+nEq] 2434 || pIdx->aSortOrder[i+nEq]!=pIdx->aSortOrder[nEq] 2435 ){ 2436 break; 2437 } 2438 2439 testcase( pLhs->iColumn==XN_ROWID ); 2440 aff = sqlite3CompareAffinity(pRhs, sqlite3ExprAffinity(pLhs)); 2441 idxaff = sqlite3TableColumnAffinity(pIdx->pTable, pLhs->iColumn); 2442 if( aff!=idxaff ) break; 2443 2444 pColl = sqlite3BinaryCompareCollSeq(pParse, pLhs, pRhs); 2445 if( pColl==0 ) break; 2446 if( sqlite3StrICmp(pColl->zName, pIdx->azColl[i+nEq]) ) break; 2447 } 2448 return i; 2449 } 2450 2451 /* 2452 ** Adjust the cost C by the costMult facter T. This only occurs if 2453 ** compiled with -DSQLITE_ENABLE_COSTMULT 2454 */ 2455 #ifdef SQLITE_ENABLE_COSTMULT 2456 # define ApplyCostMultiplier(C,T) C += T 2457 #else 2458 # define ApplyCostMultiplier(C,T) 2459 #endif 2460 2461 /* 2462 ** We have so far matched pBuilder->pNew->u.btree.nEq terms of the 2463 ** index pIndex. Try to match one more. 2464 ** 2465 ** When this function is called, pBuilder->pNew->nOut contains the 2466 ** number of rows expected to be visited by filtering using the nEq 2467 ** terms only. If it is modified, this value is restored before this 2468 ** function returns. 2469 ** 2470 ** If pProbe->idxType==SQLITE_IDXTYPE_IPK, that means pIndex is 2471 ** a fake index used for the INTEGER PRIMARY KEY. 2472 */ 2473 static int whereLoopAddBtreeIndex( 2474 WhereLoopBuilder *pBuilder, /* The WhereLoop factory */ 2475 SrcItem *pSrc, /* FROM clause term being analyzed */ 2476 Index *pProbe, /* An index on pSrc */ 2477 LogEst nInMul /* log(Number of iterations due to IN) */ 2478 ){ 2479 WhereInfo *pWInfo = pBuilder->pWInfo; /* WHERE analyse context */ 2480 Parse *pParse = pWInfo->pParse; /* Parsing context */ 2481 sqlite3 *db = pParse->db; /* Database connection malloc context */ 2482 WhereLoop *pNew; /* Template WhereLoop under construction */ 2483 WhereTerm *pTerm; /* A WhereTerm under consideration */ 2484 int opMask; /* Valid operators for constraints */ 2485 WhereScan scan; /* Iterator for WHERE terms */ 2486 Bitmask saved_prereq; /* Original value of pNew->prereq */ 2487 u16 saved_nLTerm; /* Original value of pNew->nLTerm */ 2488 u16 saved_nEq; /* Original value of pNew->u.btree.nEq */ 2489 u16 saved_nBtm; /* Original value of pNew->u.btree.nBtm */ 2490 u16 saved_nTop; /* Original value of pNew->u.btree.nTop */ 2491 u16 saved_nSkip; /* Original value of pNew->nSkip */ 2492 u32 saved_wsFlags; /* Original value of pNew->wsFlags */ 2493 LogEst saved_nOut; /* Original value of pNew->nOut */ 2494 int rc = SQLITE_OK; /* Return code */ 2495 LogEst rSize; /* Number of rows in the table */ 2496 LogEst rLogSize; /* Logarithm of table size */ 2497 WhereTerm *pTop = 0, *pBtm = 0; /* Top and bottom range constraints */ 2498 2499 pNew = pBuilder->pNew; 2500 if( db->mallocFailed ) return SQLITE_NOMEM_BKPT; 2501 WHERETRACE(0x800, ("BEGIN %s.addBtreeIdx(%s), nEq=%d, nSkip=%d, rRun=%d\n", 2502 pProbe->pTable->zName,pProbe->zName, 2503 pNew->u.btree.nEq, pNew->nSkip, pNew->rRun)); 2504 2505 assert( (pNew->wsFlags & WHERE_VIRTUALTABLE)==0 ); 2506 assert( (pNew->wsFlags & WHERE_TOP_LIMIT)==0 ); 2507 if( pNew->wsFlags & WHERE_BTM_LIMIT ){ 2508 opMask = WO_LT|WO_LE; 2509 }else{ 2510 assert( pNew->u.btree.nBtm==0 ); 2511 opMask = WO_EQ|WO_IN|WO_GT|WO_GE|WO_LT|WO_LE|WO_ISNULL|WO_IS; 2512 } 2513 if( pProbe->bUnordered ) opMask &= ~(WO_GT|WO_GE|WO_LT|WO_LE); 2514 2515 assert( pNew->u.btree.nEq<pProbe->nColumn ); 2516 assert( pNew->u.btree.nEq<pProbe->nKeyCol 2517 || pProbe->idxType!=SQLITE_IDXTYPE_PRIMARYKEY ); 2518 2519 saved_nEq = pNew->u.btree.nEq; 2520 saved_nBtm = pNew->u.btree.nBtm; 2521 saved_nTop = pNew->u.btree.nTop; 2522 saved_nSkip = pNew->nSkip; 2523 saved_nLTerm = pNew->nLTerm; 2524 saved_wsFlags = pNew->wsFlags; 2525 saved_prereq = pNew->prereq; 2526 saved_nOut = pNew->nOut; 2527 pTerm = whereScanInit(&scan, pBuilder->pWC, pSrc->iCursor, saved_nEq, 2528 opMask, pProbe); 2529 pNew->rSetup = 0; 2530 rSize = pProbe->aiRowLogEst[0]; 2531 rLogSize = estLog(rSize); 2532 for(; rc==SQLITE_OK && pTerm!=0; pTerm = whereScanNext(&scan)){ 2533 u16 eOp = pTerm->eOperator; /* Shorthand for pTerm->eOperator */ 2534 LogEst rCostIdx; 2535 LogEst nOutUnadjusted; /* nOut before IN() and WHERE adjustments */ 2536 int nIn = 0; 2537 #ifdef SQLITE_ENABLE_STAT4 2538 int nRecValid = pBuilder->nRecValid; 2539 #endif 2540 if( (eOp==WO_ISNULL || (pTerm->wtFlags&TERM_VNULL)!=0) 2541 && indexColumnNotNull(pProbe, saved_nEq) 2542 ){ 2543 continue; /* ignore IS [NOT] NULL constraints on NOT NULL columns */ 2544 } 2545 if( pTerm->prereqRight & pNew->maskSelf ) continue; 2546 2547 /* Do not allow the upper bound of a LIKE optimization range constraint 2548 ** to mix with a lower range bound from some other source */ 2549 if( pTerm->wtFlags & TERM_LIKEOPT && pTerm->eOperator==WO_LT ) continue; 2550 2551 /* tag-20191211-001: Do not allow constraints from the WHERE clause to 2552 ** be used by the right table of a LEFT JOIN. Only constraints in the 2553 ** ON clause are allowed. See tag-20191211-002 for the vtab equivalent. */ 2554 if( (pSrc->fg.jointype & JT_LEFT)!=0 2555 && !ExprHasProperty(pTerm->pExpr, EP_FromJoin) 2556 ){ 2557 continue; 2558 } 2559 2560 if( IsUniqueIndex(pProbe) && saved_nEq==pProbe->nKeyCol-1 ){ 2561 pBuilder->bldFlags1 |= SQLITE_BLDF1_UNIQUE; 2562 }else{ 2563 pBuilder->bldFlags1 |= SQLITE_BLDF1_INDEXED; 2564 } 2565 pNew->wsFlags = saved_wsFlags; 2566 pNew->u.btree.nEq = saved_nEq; 2567 pNew->u.btree.nBtm = saved_nBtm; 2568 pNew->u.btree.nTop = saved_nTop; 2569 pNew->nLTerm = saved_nLTerm; 2570 if( whereLoopResize(db, pNew, pNew->nLTerm+1) ) break; /* OOM */ 2571 pNew->aLTerm[pNew->nLTerm++] = pTerm; 2572 pNew->prereq = (saved_prereq | pTerm->prereqRight) & ~pNew->maskSelf; 2573 2574 assert( nInMul==0 2575 || (pNew->wsFlags & WHERE_COLUMN_NULL)!=0 2576 || (pNew->wsFlags & WHERE_COLUMN_IN)!=0 2577 || (pNew->wsFlags & WHERE_SKIPSCAN)!=0 2578 ); 2579 2580 if( eOp & WO_IN ){ 2581 Expr *pExpr = pTerm->pExpr; 2582 if( ExprHasProperty(pExpr, EP_xIsSelect) ){ 2583 /* "x IN (SELECT ...)": TUNING: the SELECT returns 25 rows */ 2584 int i; 2585 nIn = 46; assert( 46==sqlite3LogEst(25) ); 2586 2587 /* The expression may actually be of the form (x, y) IN (SELECT...). 2588 ** In this case there is a separate term for each of (x) and (y). 2589 ** However, the nIn multiplier should only be applied once, not once 2590 ** for each such term. The following loop checks that pTerm is the 2591 ** first such term in use, and sets nIn back to 0 if it is not. */ 2592 for(i=0; i<pNew->nLTerm-1; i++){ 2593 if( pNew->aLTerm[i] && pNew->aLTerm[i]->pExpr==pExpr ) nIn = 0; 2594 } 2595 }else if( ALWAYS(pExpr->x.pList && pExpr->x.pList->nExpr) ){ 2596 /* "x IN (value, value, ...)" */ 2597 nIn = sqlite3LogEst(pExpr->x.pList->nExpr); 2598 } 2599 if( pProbe->hasStat1 && rLogSize>=10 ){ 2600 LogEst M, logK, x; 2601 /* Let: 2602 ** N = the total number of rows in the table 2603 ** K = the number of entries on the RHS of the IN operator 2604 ** M = the number of rows in the table that match terms to the 2605 ** to the left in the same index. If the IN operator is on 2606 ** the left-most index column, M==N. 2607 ** 2608 ** Given the definitions above, it is better to omit the IN operator 2609 ** from the index lookup and instead do a scan of the M elements, 2610 ** testing each scanned row against the IN operator separately, if: 2611 ** 2612 ** M*log(K) < K*log(N) 2613 ** 2614 ** Our estimates for M, K, and N might be inaccurate, so we build in 2615 ** a safety margin of 2 (LogEst: 10) that favors using the IN operator 2616 ** with the index, as using an index has better worst-case behavior. 2617 ** If we do not have real sqlite_stat1 data, always prefer to use 2618 ** the index. Do not bother with this optimization on very small 2619 ** tables (less than 2 rows) as it is pointless in that case. 2620 */ 2621 M = pProbe->aiRowLogEst[saved_nEq]; 2622 logK = estLog(nIn); 2623 /* TUNING v----- 10 to bias toward indexed IN */ 2624 x = M + logK + 10 - (nIn + rLogSize); 2625 if( x>=0 ){ 2626 WHERETRACE(0x40, 2627 ("IN operator (N=%d M=%d logK=%d nIn=%d rLogSize=%d x=%d) " 2628 "prefers indexed lookup\n", 2629 saved_nEq, M, logK, nIn, rLogSize, x)); 2630 }else if( nInMul<2 && OptimizationEnabled(db, SQLITE_SeekScan) ){ 2631 WHERETRACE(0x40, 2632 ("IN operator (N=%d M=%d logK=%d nIn=%d rLogSize=%d x=%d" 2633 " nInMul=%d) prefers skip-scan\n", 2634 saved_nEq, M, logK, nIn, rLogSize, x, nInMul)); 2635 pNew->wsFlags |= WHERE_IN_SEEKSCAN; 2636 }else{ 2637 WHERETRACE(0x40, 2638 ("IN operator (N=%d M=%d logK=%d nIn=%d rLogSize=%d x=%d" 2639 " nInMul=%d) prefers normal scan\n", 2640 saved_nEq, M, logK, nIn, rLogSize, x, nInMul)); 2641 continue; 2642 } 2643 } 2644 pNew->wsFlags |= WHERE_COLUMN_IN; 2645 }else if( eOp & (WO_EQ|WO_IS) ){ 2646 int iCol = pProbe->aiColumn[saved_nEq]; 2647 pNew->wsFlags |= WHERE_COLUMN_EQ; 2648 assert( saved_nEq==pNew->u.btree.nEq ); 2649 if( iCol==XN_ROWID 2650 || (iCol>=0 && nInMul==0 && saved_nEq==pProbe->nKeyCol-1) 2651 ){ 2652 if( iCol==XN_ROWID || pProbe->uniqNotNull 2653 || (pProbe->nKeyCol==1 && pProbe->onError && eOp==WO_EQ) 2654 ){ 2655 pNew->wsFlags |= WHERE_ONEROW; 2656 }else{ 2657 pNew->wsFlags |= WHERE_UNQ_WANTED; 2658 } 2659 } 2660 if( scan.iEquiv>1 ) pNew->wsFlags |= WHERE_TRANSCONS; 2661 }else if( eOp & WO_ISNULL ){ 2662 pNew->wsFlags |= WHERE_COLUMN_NULL; 2663 }else if( eOp & (WO_GT|WO_GE) ){ 2664 testcase( eOp & WO_GT ); 2665 testcase( eOp & WO_GE ); 2666 pNew->wsFlags |= WHERE_COLUMN_RANGE|WHERE_BTM_LIMIT; 2667 pNew->u.btree.nBtm = whereRangeVectorLen( 2668 pParse, pSrc->iCursor, pProbe, saved_nEq, pTerm 2669 ); 2670 pBtm = pTerm; 2671 pTop = 0; 2672 if( pTerm->wtFlags & TERM_LIKEOPT ){ 2673 /* Range constraints that come from the LIKE optimization are 2674 ** always used in pairs. */ 2675 pTop = &pTerm[1]; 2676 assert( (pTop-(pTerm->pWC->a))<pTerm->pWC->nTerm ); 2677 assert( pTop->wtFlags & TERM_LIKEOPT ); 2678 assert( pTop->eOperator==WO_LT ); 2679 if( whereLoopResize(db, pNew, pNew->nLTerm+1) ) break; /* OOM */ 2680 pNew->aLTerm[pNew->nLTerm++] = pTop; 2681 pNew->wsFlags |= WHERE_TOP_LIMIT; 2682 pNew->u.btree.nTop = 1; 2683 } 2684 }else{ 2685 assert( eOp & (WO_LT|WO_LE) ); 2686 testcase( eOp & WO_LT ); 2687 testcase( eOp & WO_LE ); 2688 pNew->wsFlags |= WHERE_COLUMN_RANGE|WHERE_TOP_LIMIT; 2689 pNew->u.btree.nTop = whereRangeVectorLen( 2690 pParse, pSrc->iCursor, pProbe, saved_nEq, pTerm 2691 ); 2692 pTop = pTerm; 2693 pBtm = (pNew->wsFlags & WHERE_BTM_LIMIT)!=0 ? 2694 pNew->aLTerm[pNew->nLTerm-2] : 0; 2695 } 2696 2697 /* At this point pNew->nOut is set to the number of rows expected to 2698 ** be visited by the index scan before considering term pTerm, or the 2699 ** values of nIn and nInMul. In other words, assuming that all 2700 ** "x IN(...)" terms are replaced with "x = ?". This block updates 2701 ** the value of pNew->nOut to account for pTerm (but not nIn/nInMul). */ 2702 assert( pNew->nOut==saved_nOut ); 2703 if( pNew->wsFlags & WHERE_COLUMN_RANGE ){ 2704 /* Adjust nOut using stat4 data. Or, if there is no stat4 2705 ** data, using some other estimate. */ 2706 whereRangeScanEst(pParse, pBuilder, pBtm, pTop, pNew); 2707 }else{ 2708 int nEq = ++pNew->u.btree.nEq; 2709 assert( eOp & (WO_ISNULL|WO_EQ|WO_IN|WO_IS) ); 2710 2711 assert( pNew->nOut==saved_nOut ); 2712 if( pTerm->truthProb<=0 && pProbe->aiColumn[saved_nEq]>=0 ){ 2713 assert( (eOp & WO_IN) || nIn==0 ); 2714 testcase( eOp & WO_IN ); 2715 pNew->nOut += pTerm->truthProb; 2716 pNew->nOut -= nIn; 2717 }else{ 2718 #ifdef SQLITE_ENABLE_STAT4 2719 tRowcnt nOut = 0; 2720 if( nInMul==0 2721 && pProbe->nSample 2722 && pNew->u.btree.nEq<=pProbe->nSampleCol 2723 && ((eOp & WO_IN)==0 || !ExprHasProperty(pTerm->pExpr, EP_xIsSelect)) 2724 && OptimizationEnabled(db, SQLITE_Stat4) 2725 ){ 2726 Expr *pExpr = pTerm->pExpr; 2727 if( (eOp & (WO_EQ|WO_ISNULL|WO_IS))!=0 ){ 2728 testcase( eOp & WO_EQ ); 2729 testcase( eOp & WO_IS ); 2730 testcase( eOp & WO_ISNULL ); 2731 rc = whereEqualScanEst(pParse, pBuilder, pExpr->pRight, &nOut); 2732 }else{ 2733 rc = whereInScanEst(pParse, pBuilder, pExpr->x.pList, &nOut); 2734 } 2735 if( rc==SQLITE_NOTFOUND ) rc = SQLITE_OK; 2736 if( rc!=SQLITE_OK ) break; /* Jump out of the pTerm loop */ 2737 if( nOut ){ 2738 pNew->nOut = sqlite3LogEst(nOut); 2739 if( nEq==1 2740 /* TUNING: Mark terms as "low selectivity" if they seem likely 2741 ** to be true for half or more of the rows in the table. 2742 ** See tag-202002240-1 */ 2743 && pNew->nOut+10 > pProbe->aiRowLogEst[0] 2744 ){ 2745 #if WHERETRACE_ENABLED /* 0x01 */ 2746 if( sqlite3WhereTrace & 0x01 ){ 2747 sqlite3DebugPrintf( 2748 "STAT4 determines term has low selectivity:\n"); 2749 sqlite3WhereTermPrint(pTerm, 999); 2750 } 2751 #endif 2752 pTerm->wtFlags |= TERM_HIGHTRUTH; 2753 if( pTerm->wtFlags & TERM_HEURTRUTH ){ 2754 /* If the term has previously been used with an assumption of 2755 ** higher selectivity, then set the flag to rerun the 2756 ** loop computations. */ 2757 pBuilder->bldFlags2 |= SQLITE_BLDF2_2NDPASS; 2758 } 2759 } 2760 if( pNew->nOut>saved_nOut ) pNew->nOut = saved_nOut; 2761 pNew->nOut -= nIn; 2762 } 2763 } 2764 if( nOut==0 ) 2765 #endif 2766 { 2767 pNew->nOut += (pProbe->aiRowLogEst[nEq] - pProbe->aiRowLogEst[nEq-1]); 2768 if( eOp & WO_ISNULL ){ 2769 /* TUNING: If there is no likelihood() value, assume that a 2770 ** "col IS NULL" expression matches twice as many rows 2771 ** as (col=?). */ 2772 pNew->nOut += 10; 2773 } 2774 } 2775 } 2776 } 2777 2778 /* Set rCostIdx to the cost of visiting selected rows in index. Add 2779 ** it to pNew->rRun, which is currently set to the cost of the index 2780 ** seek only. Then, if this is a non-covering index, add the cost of 2781 ** visiting the rows in the main table. */ 2782 assert( pSrc->pTab->szTabRow>0 ); 2783 rCostIdx = pNew->nOut + 1 + (15*pProbe->szIdxRow)/pSrc->pTab->szTabRow; 2784 pNew->rRun = sqlite3LogEstAdd(rLogSize, rCostIdx); 2785 if( (pNew->wsFlags & (WHERE_IDX_ONLY|WHERE_IPK))==0 ){ 2786 pNew->rRun = sqlite3LogEstAdd(pNew->rRun, pNew->nOut + 16); 2787 } 2788 ApplyCostMultiplier(pNew->rRun, pProbe->pTable->costMult); 2789 2790 nOutUnadjusted = pNew->nOut; 2791 pNew->rRun += nInMul + nIn; 2792 pNew->nOut += nInMul + nIn; 2793 whereLoopOutputAdjust(pBuilder->pWC, pNew, rSize); 2794 rc = whereLoopInsert(pBuilder, pNew); 2795 2796 if( pNew->wsFlags & WHERE_COLUMN_RANGE ){ 2797 pNew->nOut = saved_nOut; 2798 }else{ 2799 pNew->nOut = nOutUnadjusted; 2800 } 2801 2802 if( (pNew->wsFlags & WHERE_TOP_LIMIT)==0 2803 && pNew->u.btree.nEq<pProbe->nColumn 2804 && (pNew->u.btree.nEq<pProbe->nKeyCol || 2805 pProbe->idxType!=SQLITE_IDXTYPE_PRIMARYKEY) 2806 ){ 2807 whereLoopAddBtreeIndex(pBuilder, pSrc, pProbe, nInMul+nIn); 2808 } 2809 pNew->nOut = saved_nOut; 2810 #ifdef SQLITE_ENABLE_STAT4 2811 pBuilder->nRecValid = nRecValid; 2812 #endif 2813 } 2814 pNew->prereq = saved_prereq; 2815 pNew->u.btree.nEq = saved_nEq; 2816 pNew->u.btree.nBtm = saved_nBtm; 2817 pNew->u.btree.nTop = saved_nTop; 2818 pNew->nSkip = saved_nSkip; 2819 pNew->wsFlags = saved_wsFlags; 2820 pNew->nOut = saved_nOut; 2821 pNew->nLTerm = saved_nLTerm; 2822 2823 /* Consider using a skip-scan if there are no WHERE clause constraints 2824 ** available for the left-most terms of the index, and if the average 2825 ** number of repeats in the left-most terms is at least 18. 2826 ** 2827 ** The magic number 18 is selected on the basis that scanning 17 rows 2828 ** is almost always quicker than an index seek (even though if the index 2829 ** contains fewer than 2^17 rows we assume otherwise in other parts of 2830 ** the code). And, even if it is not, it should not be too much slower. 2831 ** On the other hand, the extra seeks could end up being significantly 2832 ** more expensive. */ 2833 assert( 42==sqlite3LogEst(18) ); 2834 if( saved_nEq==saved_nSkip 2835 && saved_nEq+1<pProbe->nKeyCol 2836 && saved_nEq==pNew->nLTerm 2837 && pProbe->noSkipScan==0 2838 && pProbe->hasStat1!=0 2839 && OptimizationEnabled(db, SQLITE_SkipScan) 2840 && pProbe->aiRowLogEst[saved_nEq+1]>=42 /* TUNING: Minimum for skip-scan */ 2841 && (rc = whereLoopResize(db, pNew, pNew->nLTerm+1))==SQLITE_OK 2842 ){ 2843 LogEst nIter; 2844 pNew->u.btree.nEq++; 2845 pNew->nSkip++; 2846 pNew->aLTerm[pNew->nLTerm++] = 0; 2847 pNew->wsFlags |= WHERE_SKIPSCAN; 2848 nIter = pProbe->aiRowLogEst[saved_nEq] - pProbe->aiRowLogEst[saved_nEq+1]; 2849 pNew->nOut -= nIter; 2850 /* TUNING: Because uncertainties in the estimates for skip-scan queries, 2851 ** add a 1.375 fudge factor to make skip-scan slightly less likely. */ 2852 nIter += 5; 2853 whereLoopAddBtreeIndex(pBuilder, pSrc, pProbe, nIter + nInMul); 2854 pNew->nOut = saved_nOut; 2855 pNew->u.btree.nEq = saved_nEq; 2856 pNew->nSkip = saved_nSkip; 2857 pNew->wsFlags = saved_wsFlags; 2858 } 2859 2860 WHERETRACE(0x800, ("END %s.addBtreeIdx(%s), nEq=%d, rc=%d\n", 2861 pProbe->pTable->zName, pProbe->zName, saved_nEq, rc)); 2862 return rc; 2863 } 2864 2865 /* 2866 ** Return True if it is possible that pIndex might be useful in 2867 ** implementing the ORDER BY clause in pBuilder. 2868 ** 2869 ** Return False if pBuilder does not contain an ORDER BY clause or 2870 ** if there is no way for pIndex to be useful in implementing that 2871 ** ORDER BY clause. 2872 */ 2873 static int indexMightHelpWithOrderBy( 2874 WhereLoopBuilder *pBuilder, 2875 Index *pIndex, 2876 int iCursor 2877 ){ 2878 ExprList *pOB; 2879 ExprList *aColExpr; 2880 int ii, jj; 2881 2882 if( pIndex->bUnordered ) return 0; 2883 if( (pOB = pBuilder->pWInfo->pOrderBy)==0 ) return 0; 2884 for(ii=0; ii<pOB->nExpr; ii++){ 2885 Expr *pExpr = sqlite3ExprSkipCollateAndLikely(pOB->a[ii].pExpr); 2886 if( NEVER(pExpr==0) ) continue; 2887 if( pExpr->op==TK_COLUMN && pExpr->iTable==iCursor ){ 2888 if( pExpr->iColumn<0 ) return 1; 2889 for(jj=0; jj<pIndex->nKeyCol; jj++){ 2890 if( pExpr->iColumn==pIndex->aiColumn[jj] ) return 1; 2891 } 2892 }else if( (aColExpr = pIndex->aColExpr)!=0 ){ 2893 for(jj=0; jj<pIndex->nKeyCol; jj++){ 2894 if( pIndex->aiColumn[jj]!=XN_EXPR ) continue; 2895 if( sqlite3ExprCompareSkip(pExpr,aColExpr->a[jj].pExpr,iCursor)==0 ){ 2896 return 1; 2897 } 2898 } 2899 } 2900 } 2901 return 0; 2902 } 2903 2904 /* Check to see if a partial index with pPartIndexWhere can be used 2905 ** in the current query. Return true if it can be and false if not. 2906 */ 2907 static int whereUsablePartialIndex( 2908 int iTab, /* The table for which we want an index */ 2909 int isLeft, /* True if iTab is the right table of a LEFT JOIN */ 2910 WhereClause *pWC, /* The WHERE clause of the query */ 2911 Expr *pWhere /* The WHERE clause from the partial index */ 2912 ){ 2913 int i; 2914 WhereTerm *pTerm; 2915 Parse *pParse = pWC->pWInfo->pParse; 2916 while( pWhere->op==TK_AND ){ 2917 if( !whereUsablePartialIndex(iTab,isLeft,pWC,pWhere->pLeft) ) return 0; 2918 pWhere = pWhere->pRight; 2919 } 2920 if( pParse->db->flags & SQLITE_EnableQPSG ) pParse = 0; 2921 for(i=0, pTerm=pWC->a; i<pWC->nTerm; i++, pTerm++){ 2922 Expr *pExpr; 2923 pExpr = pTerm->pExpr; 2924 if( (!ExprHasProperty(pExpr, EP_FromJoin) || pExpr->iRightJoinTable==iTab) 2925 && (isLeft==0 || ExprHasProperty(pExpr, EP_FromJoin)) 2926 && sqlite3ExprImpliesExpr(pParse, pExpr, pWhere, iTab) 2927 && (pTerm->wtFlags & TERM_VNULL)==0 2928 ){ 2929 return 1; 2930 } 2931 } 2932 return 0; 2933 } 2934 2935 /* 2936 ** Add all WhereLoop objects for a single table of the join where the table 2937 ** is identified by pBuilder->pNew->iTab. That table is guaranteed to be 2938 ** a b-tree table, not a virtual table. 2939 ** 2940 ** The costs (WhereLoop.rRun) of the b-tree loops added by this function 2941 ** are calculated as follows: 2942 ** 2943 ** For a full scan, assuming the table (or index) contains nRow rows: 2944 ** 2945 ** cost = nRow * 3.0 // full-table scan 2946 ** cost = nRow * K // scan of covering index 2947 ** cost = nRow * (K+3.0) // scan of non-covering index 2948 ** 2949 ** where K is a value between 1.1 and 3.0 set based on the relative 2950 ** estimated average size of the index and table records. 2951 ** 2952 ** For an index scan, where nVisit is the number of index rows visited 2953 ** by the scan, and nSeek is the number of seek operations required on 2954 ** the index b-tree: 2955 ** 2956 ** cost = nSeek * (log(nRow) + K * nVisit) // covering index 2957 ** cost = nSeek * (log(nRow) + (K+3.0) * nVisit) // non-covering index 2958 ** 2959 ** Normally, nSeek is 1. nSeek values greater than 1 come about if the 2960 ** WHERE clause includes "x IN (....)" terms used in place of "x=?". Or when 2961 ** implicit "x IN (SELECT x FROM tbl)" terms are added for skip-scans. 2962 ** 2963 ** The estimated values (nRow, nVisit, nSeek) often contain a large amount 2964 ** of uncertainty. For this reason, scoring is designed to pick plans that 2965 ** "do the least harm" if the estimates are inaccurate. For example, a 2966 ** log(nRow) factor is omitted from a non-covering index scan in order to 2967 ** bias the scoring in favor of using an index, since the worst-case 2968 ** performance of using an index is far better than the worst-case performance 2969 ** of a full table scan. 2970 */ 2971 static int whereLoopAddBtree( 2972 WhereLoopBuilder *pBuilder, /* WHERE clause information */ 2973 Bitmask mPrereq /* Extra prerequesites for using this table */ 2974 ){ 2975 WhereInfo *pWInfo; /* WHERE analysis context */ 2976 Index *pProbe; /* An index we are evaluating */ 2977 Index sPk; /* A fake index object for the primary key */ 2978 LogEst aiRowEstPk[2]; /* The aiRowLogEst[] value for the sPk index */ 2979 i16 aiColumnPk = -1; /* The aColumn[] value for the sPk index */ 2980 SrcList *pTabList; /* The FROM clause */ 2981 SrcItem *pSrc; /* The FROM clause btree term to add */ 2982 WhereLoop *pNew; /* Template WhereLoop object */ 2983 int rc = SQLITE_OK; /* Return code */ 2984 int iSortIdx = 1; /* Index number */ 2985 int b; /* A boolean value */ 2986 LogEst rSize; /* number of rows in the table */ 2987 LogEst rLogSize; /* Logarithm of the number of rows in the table */ 2988 WhereClause *pWC; /* The parsed WHERE clause */ 2989 Table *pTab; /* Table being queried */ 2990 2991 pNew = pBuilder->pNew; 2992 pWInfo = pBuilder->pWInfo; 2993 pTabList = pWInfo->pTabList; 2994 pSrc = pTabList->a + pNew->iTab; 2995 pTab = pSrc->pTab; 2996 pWC = pBuilder->pWC; 2997 assert( !IsVirtual(pSrc->pTab) ); 2998 2999 if( pSrc->fg.isIndexedBy ){ 3000 /* An INDEXED BY clause specifies a particular index to use */ 3001 pProbe = pSrc->u2.pIBIndex; 3002 }else if( !HasRowid(pTab) ){ 3003 pProbe = pTab->pIndex; 3004 }else{ 3005 /* There is no INDEXED BY clause. Create a fake Index object in local 3006 ** variable sPk to represent the rowid primary key index. Make this 3007 ** fake index the first in a chain of Index objects with all of the real 3008 ** indices to follow */ 3009 Index *pFirst; /* First of real indices on the table */ 3010 memset(&sPk, 0, sizeof(Index)); 3011 sPk.nKeyCol = 1; 3012 sPk.nColumn = 1; 3013 sPk.aiColumn = &aiColumnPk; 3014 sPk.aiRowLogEst = aiRowEstPk; 3015 sPk.onError = OE_Replace; 3016 sPk.pTable = pTab; 3017 sPk.szIdxRow = pTab->szTabRow; 3018 sPk.idxType = SQLITE_IDXTYPE_IPK; 3019 aiRowEstPk[0] = pTab->nRowLogEst; 3020 aiRowEstPk[1] = 0; 3021 pFirst = pSrc->pTab->pIndex; 3022 if( pSrc->fg.notIndexed==0 ){ 3023 /* The real indices of the table are only considered if the 3024 ** NOT INDEXED qualifier is omitted from the FROM clause */ 3025 sPk.pNext = pFirst; 3026 } 3027 pProbe = &sPk; 3028 } 3029 rSize = pTab->nRowLogEst; 3030 rLogSize = estLog(rSize); 3031 3032 #ifndef SQLITE_OMIT_AUTOMATIC_INDEX 3033 /* Automatic indexes */ 3034 if( !pBuilder->pOrSet /* Not part of an OR optimization */ 3035 && (pWInfo->wctrlFlags & WHERE_OR_SUBCLAUSE)==0 3036 && (pWInfo->pParse->db->flags & SQLITE_AutoIndex)!=0 3037 && !pSrc->fg.isIndexedBy /* Has no INDEXED BY clause */ 3038 && !pSrc->fg.notIndexed /* Has no NOT INDEXED clause */ 3039 && HasRowid(pTab) /* Not WITHOUT ROWID table. (FIXME: Why not?) */ 3040 && !pSrc->fg.isCorrelated /* Not a correlated subquery */ 3041 && !pSrc->fg.isRecursive /* Not a recursive common table expression. */ 3042 ){ 3043 /* Generate auto-index WhereLoops */ 3044 WhereTerm *pTerm; 3045 WhereTerm *pWCEnd = pWC->a + pWC->nTerm; 3046 for(pTerm=pWC->a; rc==SQLITE_OK && pTerm<pWCEnd; pTerm++){ 3047 if( pTerm->prereqRight & pNew->maskSelf ) continue; 3048 if( termCanDriveIndex(pTerm, pSrc, 0) ){ 3049 pNew->u.btree.nEq = 1; 3050 pNew->nSkip = 0; 3051 pNew->u.btree.pIndex = 0; 3052 pNew->nLTerm = 1; 3053 pNew->aLTerm[0] = pTerm; 3054 /* TUNING: One-time cost for computing the automatic index is 3055 ** estimated to be X*N*log2(N) where N is the number of rows in 3056 ** the table being indexed and where X is 7 (LogEst=28) for normal 3057 ** tables or 0.5 (LogEst=-10) for views and subqueries. The value 3058 ** of X is smaller for views and subqueries so that the query planner 3059 ** will be more aggressive about generating automatic indexes for 3060 ** those objects, since there is no opportunity to add schema 3061 ** indexes on subqueries and views. */ 3062 pNew->rSetup = rLogSize + rSize; 3063 if( pTab->pSelect==0 && (pTab->tabFlags & TF_Ephemeral)==0 ){ 3064 pNew->rSetup += 28; 3065 }else{ 3066 pNew->rSetup -= 10; 3067 } 3068 ApplyCostMultiplier(pNew->rSetup, pTab->costMult); 3069 if( pNew->rSetup<0 ) pNew->rSetup = 0; 3070 /* TUNING: Each index lookup yields 20 rows in the table. This 3071 ** is more than the usual guess of 10 rows, since we have no way 3072 ** of knowing how selective the index will ultimately be. It would 3073 ** not be unreasonable to make this value much larger. */ 3074 pNew->nOut = 43; assert( 43==sqlite3LogEst(20) ); 3075 pNew->rRun = sqlite3LogEstAdd(rLogSize,pNew->nOut); 3076 pNew->wsFlags = WHERE_AUTO_INDEX; 3077 pNew->prereq = mPrereq | pTerm->prereqRight; 3078 rc = whereLoopInsert(pBuilder, pNew); 3079 } 3080 } 3081 } 3082 #endif /* SQLITE_OMIT_AUTOMATIC_INDEX */ 3083 3084 /* Loop over all indices. If there was an INDEXED BY clause, then only 3085 ** consider index pProbe. */ 3086 for(; rc==SQLITE_OK && pProbe; 3087 pProbe=(pSrc->fg.isIndexedBy ? 0 : pProbe->pNext), iSortIdx++ 3088 ){ 3089 int isLeft = (pSrc->fg.jointype & JT_OUTER)!=0; 3090 if( pProbe->pPartIdxWhere!=0 3091 && !whereUsablePartialIndex(pSrc->iCursor, isLeft, pWC, 3092 pProbe->pPartIdxWhere) 3093 ){ 3094 testcase( pNew->iTab!=pSrc->iCursor ); /* See ticket [98d973b8f5] */ 3095 continue; /* Partial index inappropriate for this query */ 3096 } 3097 if( pProbe->bNoQuery ) continue; 3098 rSize = pProbe->aiRowLogEst[0]; 3099 pNew->u.btree.nEq = 0; 3100 pNew->u.btree.nBtm = 0; 3101 pNew->u.btree.nTop = 0; 3102 pNew->nSkip = 0; 3103 pNew->nLTerm = 0; 3104 pNew->iSortIdx = 0; 3105 pNew->rSetup = 0; 3106 pNew->prereq = mPrereq; 3107 pNew->nOut = rSize; 3108 pNew->u.btree.pIndex = pProbe; 3109 b = indexMightHelpWithOrderBy(pBuilder, pProbe, pSrc->iCursor); 3110 3111 /* The ONEPASS_DESIRED flags never occurs together with ORDER BY */ 3112 assert( (pWInfo->wctrlFlags & WHERE_ONEPASS_DESIRED)==0 || b==0 ); 3113 if( pProbe->idxType==SQLITE_IDXTYPE_IPK ){ 3114 /* Integer primary key index */ 3115 pNew->wsFlags = WHERE_IPK; 3116 3117 /* Full table scan */ 3118 pNew->iSortIdx = b ? iSortIdx : 0; 3119 /* TUNING: Cost of full table scan is 3.0*N. The 3.0 factor is an 3120 ** extra cost designed to discourage the use of full table scans, 3121 ** since index lookups have better worst-case performance if our 3122 ** stat guesses are wrong. Reduce the 3.0 penalty slightly 3123 ** (to 2.75) if we have valid STAT4 information for the table. 3124 ** At 2.75, a full table scan is preferred over using an index on 3125 ** a column with just two distinct values where each value has about 3126 ** an equal number of appearances. Without STAT4 data, we still want 3127 ** to use an index in that case, since the constraint might be for 3128 ** the scarcer of the two values, and in that case an index lookup is 3129 ** better. 3130 */ 3131 #ifdef SQLITE_ENABLE_STAT4 3132 pNew->rRun = rSize + 16 - 2*((pTab->tabFlags & TF_HasStat4)!=0); 3133 #else 3134 pNew->rRun = rSize + 16; 3135 #endif 3136 ApplyCostMultiplier(pNew->rRun, pTab->costMult); 3137 whereLoopOutputAdjust(pWC, pNew, rSize); 3138 rc = whereLoopInsert(pBuilder, pNew); 3139 pNew->nOut = rSize; 3140 if( rc ) break; 3141 }else{ 3142 Bitmask m; 3143 if( pProbe->isCovering ){ 3144 pNew->wsFlags = WHERE_IDX_ONLY | WHERE_INDEXED; 3145 m = 0; 3146 }else{ 3147 m = pSrc->colUsed & pProbe->colNotIdxed; 3148 pNew->wsFlags = (m==0) ? (WHERE_IDX_ONLY|WHERE_INDEXED) : WHERE_INDEXED; 3149 } 3150 3151 /* Full scan via index */ 3152 if( b 3153 || !HasRowid(pTab) 3154 || pProbe->pPartIdxWhere!=0 3155 || pSrc->fg.isIndexedBy 3156 || ( m==0 3157 && pProbe->bUnordered==0 3158 && (pProbe->szIdxRow<pTab->szTabRow) 3159 && (pWInfo->wctrlFlags & WHERE_ONEPASS_DESIRED)==0 3160 && sqlite3GlobalConfig.bUseCis 3161 && OptimizationEnabled(pWInfo->pParse->db, SQLITE_CoverIdxScan) 3162 ) 3163 ){ 3164 pNew->iSortIdx = b ? iSortIdx : 0; 3165 3166 /* The cost of visiting the index rows is N*K, where K is 3167 ** between 1.1 and 3.0, depending on the relative sizes of the 3168 ** index and table rows. */ 3169 pNew->rRun = rSize + 1 + (15*pProbe->szIdxRow)/pTab->szTabRow; 3170 if( m!=0 ){ 3171 /* If this is a non-covering index scan, add in the cost of 3172 ** doing table lookups. The cost will be 3x the number of 3173 ** lookups. Take into account WHERE clause terms that can be 3174 ** satisfied using just the index, and that do not require a 3175 ** table lookup. */ 3176 LogEst nLookup = rSize + 16; /* Base cost: N*3 */ 3177 int ii; 3178 int iCur = pSrc->iCursor; 3179 WhereClause *pWC2 = &pWInfo->sWC; 3180 for(ii=0; ii<pWC2->nTerm; ii++){ 3181 WhereTerm *pTerm = &pWC2->a[ii]; 3182 if( !sqlite3ExprCoveredByIndex(pTerm->pExpr, iCur, pProbe) ){ 3183 break; 3184 } 3185 /* pTerm can be evaluated using just the index. So reduce 3186 ** the expected number of table lookups accordingly */ 3187 if( pTerm->truthProb<=0 ){ 3188 nLookup += pTerm->truthProb; 3189 }else{ 3190 nLookup--; 3191 if( pTerm->eOperator & (WO_EQ|WO_IS) ) nLookup -= 19; 3192 } 3193 } 3194 3195 pNew->rRun = sqlite3LogEstAdd(pNew->rRun, nLookup); 3196 } 3197 ApplyCostMultiplier(pNew->rRun, pTab->costMult); 3198 whereLoopOutputAdjust(pWC, pNew, rSize); 3199 rc = whereLoopInsert(pBuilder, pNew); 3200 pNew->nOut = rSize; 3201 if( rc ) break; 3202 } 3203 } 3204 3205 pBuilder->bldFlags1 = 0; 3206 rc = whereLoopAddBtreeIndex(pBuilder, pSrc, pProbe, 0); 3207 if( pBuilder->bldFlags1==SQLITE_BLDF1_INDEXED ){ 3208 /* If a non-unique index is used, or if a prefix of the key for 3209 ** unique index is used (making the index functionally non-unique) 3210 ** then the sqlite_stat1 data becomes important for scoring the 3211 ** plan */ 3212 pTab->tabFlags |= TF_StatsUsed; 3213 } 3214 #ifdef SQLITE_ENABLE_STAT4 3215 sqlite3Stat4ProbeFree(pBuilder->pRec); 3216 pBuilder->nRecValid = 0; 3217 pBuilder->pRec = 0; 3218 #endif 3219 } 3220 return rc; 3221 } 3222 3223 #ifndef SQLITE_OMIT_VIRTUALTABLE 3224 3225 /* 3226 ** Argument pIdxInfo is already populated with all constraints that may 3227 ** be used by the virtual table identified by pBuilder->pNew->iTab. This 3228 ** function marks a subset of those constraints usable, invokes the 3229 ** xBestIndex method and adds the returned plan to pBuilder. 3230 ** 3231 ** A constraint is marked usable if: 3232 ** 3233 ** * Argument mUsable indicates that its prerequisites are available, and 3234 ** 3235 ** * It is not one of the operators specified in the mExclude mask passed 3236 ** as the fourth argument (which in practice is either WO_IN or 0). 3237 ** 3238 ** Argument mPrereq is a mask of tables that must be scanned before the 3239 ** virtual table in question. These are added to the plans prerequisites 3240 ** before it is added to pBuilder. 3241 ** 3242 ** Output parameter *pbIn is set to true if the plan added to pBuilder 3243 ** uses one or more WO_IN terms, or false otherwise. 3244 */ 3245 static int whereLoopAddVirtualOne( 3246 WhereLoopBuilder *pBuilder, 3247 Bitmask mPrereq, /* Mask of tables that must be used. */ 3248 Bitmask mUsable, /* Mask of usable tables */ 3249 u16 mExclude, /* Exclude terms using these operators */ 3250 sqlite3_index_info *pIdxInfo, /* Populated object for xBestIndex */ 3251 u16 mNoOmit, /* Do not omit these constraints */ 3252 int *pbIn /* OUT: True if plan uses an IN(...) op */ 3253 ){ 3254 WhereClause *pWC = pBuilder->pWC; 3255 struct sqlite3_index_constraint *pIdxCons; 3256 struct sqlite3_index_constraint_usage *pUsage = pIdxInfo->aConstraintUsage; 3257 int i; 3258 int mxTerm; 3259 int rc = SQLITE_OK; 3260 WhereLoop *pNew = pBuilder->pNew; 3261 Parse *pParse = pBuilder->pWInfo->pParse; 3262 SrcItem *pSrc = &pBuilder->pWInfo->pTabList->a[pNew->iTab]; 3263 int nConstraint = pIdxInfo->nConstraint; 3264 3265 assert( (mUsable & mPrereq)==mPrereq ); 3266 *pbIn = 0; 3267 pNew->prereq = mPrereq; 3268 3269 /* Set the usable flag on the subset of constraints identified by 3270 ** arguments mUsable and mExclude. */ 3271 pIdxCons = *(struct sqlite3_index_constraint**)&pIdxInfo->aConstraint; 3272 for(i=0; i<nConstraint; i++, pIdxCons++){ 3273 WhereTerm *pTerm = &pWC->a[pIdxCons->iTermOffset]; 3274 pIdxCons->usable = 0; 3275 if( (pTerm->prereqRight & mUsable)==pTerm->prereqRight 3276 && (pTerm->eOperator & mExclude)==0 3277 ){ 3278 pIdxCons->usable = 1; 3279 } 3280 } 3281 3282 /* Initialize the output fields of the sqlite3_index_info structure */ 3283 memset(pUsage, 0, sizeof(pUsage[0])*nConstraint); 3284 assert( pIdxInfo->needToFreeIdxStr==0 ); 3285 pIdxInfo->idxStr = 0; 3286 pIdxInfo->idxNum = 0; 3287 pIdxInfo->orderByConsumed = 0; 3288 pIdxInfo->estimatedCost = SQLITE_BIG_DBL / (double)2; 3289 pIdxInfo->estimatedRows = 25; 3290 pIdxInfo->idxFlags = 0; 3291 pIdxInfo->colUsed = (sqlite3_int64)pSrc->colUsed; 3292 3293 /* Invoke the virtual table xBestIndex() method */ 3294 rc = vtabBestIndex(pParse, pSrc->pTab, pIdxInfo); 3295 if( rc ){ 3296 if( rc==SQLITE_CONSTRAINT ){ 3297 /* If the xBestIndex method returns SQLITE_CONSTRAINT, that means 3298 ** that the particular combination of parameters provided is unusable. 3299 ** Make no entries in the loop table. 3300 */ 3301 WHERETRACE(0xffff, (" ^^^^--- non-viable plan rejected!\n")); 3302 return SQLITE_OK; 3303 } 3304 return rc; 3305 } 3306 3307 mxTerm = -1; 3308 assert( pNew->nLSlot>=nConstraint ); 3309 for(i=0; i<nConstraint; i++) pNew->aLTerm[i] = 0; 3310 pNew->u.vtab.omitMask = 0; 3311 pIdxCons = *(struct sqlite3_index_constraint**)&pIdxInfo->aConstraint; 3312 for(i=0; i<nConstraint; i++, pIdxCons++){ 3313 int iTerm; 3314 if( (iTerm = pUsage[i].argvIndex - 1)>=0 ){ 3315 WhereTerm *pTerm; 3316 int j = pIdxCons->iTermOffset; 3317 if( iTerm>=nConstraint 3318 || j<0 3319 || j>=pWC->nTerm 3320 || pNew->aLTerm[iTerm]!=0 3321 || pIdxCons->usable==0 3322 ){ 3323 sqlite3ErrorMsg(pParse,"%s.xBestIndex malfunction",pSrc->pTab->zName); 3324 testcase( pIdxInfo->needToFreeIdxStr ); 3325 return SQLITE_ERROR; 3326 } 3327 testcase( iTerm==nConstraint-1 ); 3328 testcase( j==0 ); 3329 testcase( j==pWC->nTerm-1 ); 3330 pTerm = &pWC->a[j]; 3331 pNew->prereq |= pTerm->prereqRight; 3332 assert( iTerm<pNew->nLSlot ); 3333 pNew->aLTerm[iTerm] = pTerm; 3334 if( iTerm>mxTerm ) mxTerm = iTerm; 3335 testcase( iTerm==15 ); 3336 testcase( iTerm==16 ); 3337 if( pUsage[i].omit ){ 3338 if( i<16 && ((1<<i)&mNoOmit)==0 ){ 3339 testcase( i!=iTerm ); 3340 pNew->u.vtab.omitMask |= 1<<iTerm; 3341 }else{ 3342 testcase( i!=iTerm ); 3343 } 3344 } 3345 if( (pTerm->eOperator & WO_IN)!=0 ){ 3346 /* A virtual table that is constrained by an IN clause may not 3347 ** consume the ORDER BY clause because (1) the order of IN terms 3348 ** is not necessarily related to the order of output terms and 3349 ** (2) Multiple outputs from a single IN value will not merge 3350 ** together. */ 3351 pIdxInfo->orderByConsumed = 0; 3352 pIdxInfo->idxFlags &= ~SQLITE_INDEX_SCAN_UNIQUE; 3353 *pbIn = 1; assert( (mExclude & WO_IN)==0 ); 3354 } 3355 } 3356 } 3357 3358 pNew->nLTerm = mxTerm+1; 3359 for(i=0; i<=mxTerm; i++){ 3360 if( pNew->aLTerm[i]==0 ){ 3361 /* The non-zero argvIdx values must be contiguous. Raise an 3362 ** error if they are not */ 3363 sqlite3ErrorMsg(pParse,"%s.xBestIndex malfunction",pSrc->pTab->zName); 3364 testcase( pIdxInfo->needToFreeIdxStr ); 3365 return SQLITE_ERROR; 3366 } 3367 } 3368 assert( pNew->nLTerm<=pNew->nLSlot ); 3369 pNew->u.vtab.idxNum = pIdxInfo->idxNum; 3370 pNew->u.vtab.needFree = pIdxInfo->needToFreeIdxStr; 3371 pIdxInfo->needToFreeIdxStr = 0; 3372 pNew->u.vtab.idxStr = pIdxInfo->idxStr; 3373 pNew->u.vtab.isOrdered = (i8)(pIdxInfo->orderByConsumed ? 3374 pIdxInfo->nOrderBy : 0); 3375 pNew->rSetup = 0; 3376 pNew->rRun = sqlite3LogEstFromDouble(pIdxInfo->estimatedCost); 3377 pNew->nOut = sqlite3LogEst(pIdxInfo->estimatedRows); 3378 3379 /* Set the WHERE_ONEROW flag if the xBestIndex() method indicated 3380 ** that the scan will visit at most one row. Clear it otherwise. */ 3381 if( pIdxInfo->idxFlags & SQLITE_INDEX_SCAN_UNIQUE ){ 3382 pNew->wsFlags |= WHERE_ONEROW; 3383 }else{ 3384 pNew->wsFlags &= ~WHERE_ONEROW; 3385 } 3386 rc = whereLoopInsert(pBuilder, pNew); 3387 if( pNew->u.vtab.needFree ){ 3388 sqlite3_free(pNew->u.vtab.idxStr); 3389 pNew->u.vtab.needFree = 0; 3390 } 3391 WHERETRACE(0xffff, (" bIn=%d prereqIn=%04llx prereqOut=%04llx\n", 3392 *pbIn, (sqlite3_uint64)mPrereq, 3393 (sqlite3_uint64)(pNew->prereq & ~mPrereq))); 3394 3395 return rc; 3396 } 3397 3398 /* 3399 ** If this function is invoked from within an xBestIndex() callback, it 3400 ** returns a pointer to a buffer containing the name of the collation 3401 ** sequence associated with element iCons of the sqlite3_index_info.aConstraint 3402 ** array. Or, if iCons is out of range or there is no active xBestIndex 3403 ** call, return NULL. 3404 */ 3405 const char *sqlite3_vtab_collation(sqlite3_index_info *pIdxInfo, int iCons){ 3406 HiddenIndexInfo *pHidden = (HiddenIndexInfo*)&pIdxInfo[1]; 3407 const char *zRet = 0; 3408 if( iCons>=0 && iCons<pIdxInfo->nConstraint ){ 3409 CollSeq *pC = 0; 3410 int iTerm = pIdxInfo->aConstraint[iCons].iTermOffset; 3411 Expr *pX = pHidden->pWC->a[iTerm].pExpr; 3412 if( pX->pLeft ){ 3413 pC = sqlite3ExprCompareCollSeq(pHidden->pParse, pX); 3414 } 3415 zRet = (pC ? pC->zName : sqlite3StrBINARY); 3416 } 3417 return zRet; 3418 } 3419 3420 /* 3421 ** Add all WhereLoop objects for a table of the join identified by 3422 ** pBuilder->pNew->iTab. That table is guaranteed to be a virtual table. 3423 ** 3424 ** If there are no LEFT or CROSS JOIN joins in the query, both mPrereq and 3425 ** mUnusable are set to 0. Otherwise, mPrereq is a mask of all FROM clause 3426 ** entries that occur before the virtual table in the FROM clause and are 3427 ** separated from it by at least one LEFT or CROSS JOIN. Similarly, the 3428 ** mUnusable mask contains all FROM clause entries that occur after the 3429 ** virtual table and are separated from it by at least one LEFT or 3430 ** CROSS JOIN. 3431 ** 3432 ** For example, if the query were: 3433 ** 3434 ** ... FROM t1, t2 LEFT JOIN t3, t4, vt CROSS JOIN t5, t6; 3435 ** 3436 ** then mPrereq corresponds to (t1, t2) and mUnusable to (t5, t6). 3437 ** 3438 ** All the tables in mPrereq must be scanned before the current virtual 3439 ** table. So any terms for which all prerequisites are satisfied by 3440 ** mPrereq may be specified as "usable" in all calls to xBestIndex. 3441 ** Conversely, all tables in mUnusable must be scanned after the current 3442 ** virtual table, so any terms for which the prerequisites overlap with 3443 ** mUnusable should always be configured as "not-usable" for xBestIndex. 3444 */ 3445 static int whereLoopAddVirtual( 3446 WhereLoopBuilder *pBuilder, /* WHERE clause information */ 3447 Bitmask mPrereq, /* Tables that must be scanned before this one */ 3448 Bitmask mUnusable /* Tables that must be scanned after this one */ 3449 ){ 3450 int rc = SQLITE_OK; /* Return code */ 3451 WhereInfo *pWInfo; /* WHERE analysis context */ 3452 Parse *pParse; /* The parsing context */ 3453 WhereClause *pWC; /* The WHERE clause */ 3454 SrcItem *pSrc; /* The FROM clause term to search */ 3455 sqlite3_index_info *p; /* Object to pass to xBestIndex() */ 3456 int nConstraint; /* Number of constraints in p */ 3457 int bIn; /* True if plan uses IN(...) operator */ 3458 WhereLoop *pNew; 3459 Bitmask mBest; /* Tables used by best possible plan */ 3460 u16 mNoOmit; 3461 3462 assert( (mPrereq & mUnusable)==0 ); 3463 pWInfo = pBuilder->pWInfo; 3464 pParse = pWInfo->pParse; 3465 pWC = pBuilder->pWC; 3466 pNew = pBuilder->pNew; 3467 pSrc = &pWInfo->pTabList->a[pNew->iTab]; 3468 assert( IsVirtual(pSrc->pTab) ); 3469 p = allocateIndexInfo(pParse, pWC, mUnusable, pSrc, pBuilder->pOrderBy, 3470 &mNoOmit); 3471 if( p==0 ) return SQLITE_NOMEM_BKPT; 3472 pNew->rSetup = 0; 3473 pNew->wsFlags = WHERE_VIRTUALTABLE; 3474 pNew->nLTerm = 0; 3475 pNew->u.vtab.needFree = 0; 3476 nConstraint = p->nConstraint; 3477 if( whereLoopResize(pParse->db, pNew, nConstraint) ){ 3478 sqlite3DbFree(pParse->db, p); 3479 return SQLITE_NOMEM_BKPT; 3480 } 3481 3482 /* First call xBestIndex() with all constraints usable. */ 3483 WHERETRACE(0x800, ("BEGIN %s.addVirtual()\n", pSrc->pTab->zName)); 3484 WHERETRACE(0x40, (" VirtualOne: all usable\n")); 3485 rc = whereLoopAddVirtualOne(pBuilder, mPrereq, ALLBITS, 0, p, mNoOmit, &bIn); 3486 3487 /* If the call to xBestIndex() with all terms enabled produced a plan 3488 ** that does not require any source tables (IOW: a plan with mBest==0) 3489 ** and does not use an IN(...) operator, then there is no point in making 3490 ** any further calls to xBestIndex() since they will all return the same 3491 ** result (if the xBestIndex() implementation is sane). */ 3492 if( rc==SQLITE_OK && ((mBest = (pNew->prereq & ~mPrereq))!=0 || bIn) ){ 3493 int seenZero = 0; /* True if a plan with no prereqs seen */ 3494 int seenZeroNoIN = 0; /* Plan with no prereqs and no IN(...) seen */ 3495 Bitmask mPrev = 0; 3496 Bitmask mBestNoIn = 0; 3497 3498 /* If the plan produced by the earlier call uses an IN(...) term, call 3499 ** xBestIndex again, this time with IN(...) terms disabled. */ 3500 if( bIn ){ 3501 WHERETRACE(0x40, (" VirtualOne: all usable w/o IN\n")); 3502 rc = whereLoopAddVirtualOne( 3503 pBuilder, mPrereq, ALLBITS, WO_IN, p, mNoOmit, &bIn); 3504 assert( bIn==0 ); 3505 mBestNoIn = pNew->prereq & ~mPrereq; 3506 if( mBestNoIn==0 ){ 3507 seenZero = 1; 3508 seenZeroNoIN = 1; 3509 } 3510 } 3511 3512 /* Call xBestIndex once for each distinct value of (prereqRight & ~mPrereq) 3513 ** in the set of terms that apply to the current virtual table. */ 3514 while( rc==SQLITE_OK ){ 3515 int i; 3516 Bitmask mNext = ALLBITS; 3517 assert( mNext>0 ); 3518 for(i=0; i<nConstraint; i++){ 3519 Bitmask mThis = ( 3520 pWC->a[p->aConstraint[i].iTermOffset].prereqRight & ~mPrereq 3521 ); 3522 if( mThis>mPrev && mThis<mNext ) mNext = mThis; 3523 } 3524 mPrev = mNext; 3525 if( mNext==ALLBITS ) break; 3526 if( mNext==mBest || mNext==mBestNoIn ) continue; 3527 WHERETRACE(0x40, (" VirtualOne: mPrev=%04llx mNext=%04llx\n", 3528 (sqlite3_uint64)mPrev, (sqlite3_uint64)mNext)); 3529 rc = whereLoopAddVirtualOne( 3530 pBuilder, mPrereq, mNext|mPrereq, 0, p, mNoOmit, &bIn); 3531 if( pNew->prereq==mPrereq ){ 3532 seenZero = 1; 3533 if( bIn==0 ) seenZeroNoIN = 1; 3534 } 3535 } 3536 3537 /* If the calls to xBestIndex() in the above loop did not find a plan 3538 ** that requires no source tables at all (i.e. one guaranteed to be 3539 ** usable), make a call here with all source tables disabled */ 3540 if( rc==SQLITE_OK && seenZero==0 ){ 3541 WHERETRACE(0x40, (" VirtualOne: all disabled\n")); 3542 rc = whereLoopAddVirtualOne( 3543 pBuilder, mPrereq, mPrereq, 0, p, mNoOmit, &bIn); 3544 if( bIn==0 ) seenZeroNoIN = 1; 3545 } 3546 3547 /* If the calls to xBestIndex() have so far failed to find a plan 3548 ** that requires no source tables at all and does not use an IN(...) 3549 ** operator, make a final call to obtain one here. */ 3550 if( rc==SQLITE_OK && seenZeroNoIN==0 ){ 3551 WHERETRACE(0x40, (" VirtualOne: all disabled and w/o IN\n")); 3552 rc = whereLoopAddVirtualOne( 3553 pBuilder, mPrereq, mPrereq, WO_IN, p, mNoOmit, &bIn); 3554 } 3555 } 3556 3557 if( p->needToFreeIdxStr ) sqlite3_free(p->idxStr); 3558 sqlite3DbFreeNN(pParse->db, p); 3559 WHERETRACE(0x800, ("END %s.addVirtual(), rc=%d\n", pSrc->pTab->zName, rc)); 3560 return rc; 3561 } 3562 #endif /* SQLITE_OMIT_VIRTUALTABLE */ 3563 3564 /* 3565 ** Add WhereLoop entries to handle OR terms. This works for either 3566 ** btrees or virtual tables. 3567 */ 3568 static int whereLoopAddOr( 3569 WhereLoopBuilder *pBuilder, 3570 Bitmask mPrereq, 3571 Bitmask mUnusable 3572 ){ 3573 WhereInfo *pWInfo = pBuilder->pWInfo; 3574 WhereClause *pWC; 3575 WhereLoop *pNew; 3576 WhereTerm *pTerm, *pWCEnd; 3577 int rc = SQLITE_OK; 3578 int iCur; 3579 WhereClause tempWC; 3580 WhereLoopBuilder sSubBuild; 3581 WhereOrSet sSum, sCur; 3582 SrcItem *pItem; 3583 3584 pWC = pBuilder->pWC; 3585 pWCEnd = pWC->a + pWC->nTerm; 3586 pNew = pBuilder->pNew; 3587 memset(&sSum, 0, sizeof(sSum)); 3588 pItem = pWInfo->pTabList->a + pNew->iTab; 3589 iCur = pItem->iCursor; 3590 3591 for(pTerm=pWC->a; pTerm<pWCEnd && rc==SQLITE_OK; pTerm++){ 3592 if( (pTerm->eOperator & WO_OR)!=0 3593 && (pTerm->u.pOrInfo->indexable & pNew->maskSelf)!=0 3594 ){ 3595 WhereClause * const pOrWC = &pTerm->u.pOrInfo->wc; 3596 WhereTerm * const pOrWCEnd = &pOrWC->a[pOrWC->nTerm]; 3597 WhereTerm *pOrTerm; 3598 int once = 1; 3599 int i, j; 3600 3601 sSubBuild = *pBuilder; 3602 sSubBuild.pOrderBy = 0; 3603 sSubBuild.pOrSet = &sCur; 3604 3605 WHERETRACE(0x200, ("Begin processing OR-clause %p\n", pTerm)); 3606 for(pOrTerm=pOrWC->a; pOrTerm<pOrWCEnd; pOrTerm++){ 3607 if( (pOrTerm->eOperator & WO_AND)!=0 ){ 3608 sSubBuild.pWC = &pOrTerm->u.pAndInfo->wc; 3609 }else if( pOrTerm->leftCursor==iCur ){ 3610 tempWC.pWInfo = pWC->pWInfo; 3611 tempWC.pOuter = pWC; 3612 tempWC.op = TK_AND; 3613 tempWC.nTerm = 1; 3614 tempWC.a = pOrTerm; 3615 sSubBuild.pWC = &tempWC; 3616 }else{ 3617 continue; 3618 } 3619 sCur.n = 0; 3620 #ifdef WHERETRACE_ENABLED 3621 WHERETRACE(0x200, ("OR-term %d of %p has %d subterms:\n", 3622 (int)(pOrTerm-pOrWC->a), pTerm, sSubBuild.pWC->nTerm)); 3623 if( sqlite3WhereTrace & 0x400 ){ 3624 sqlite3WhereClausePrint(sSubBuild.pWC); 3625 } 3626 #endif 3627 #ifndef SQLITE_OMIT_VIRTUALTABLE 3628 if( IsVirtual(pItem->pTab) ){ 3629 rc = whereLoopAddVirtual(&sSubBuild, mPrereq, mUnusable); 3630 }else 3631 #endif 3632 { 3633 rc = whereLoopAddBtree(&sSubBuild, mPrereq); 3634 } 3635 if( rc==SQLITE_OK ){ 3636 rc = whereLoopAddOr(&sSubBuild, mPrereq, mUnusable); 3637 } 3638 assert( rc==SQLITE_OK || rc==SQLITE_DONE || sCur.n==0 3639 || rc==SQLITE_NOMEM ); 3640 testcase( rc==SQLITE_NOMEM && sCur.n>0 ); 3641 testcase( rc==SQLITE_DONE ); 3642 if( sCur.n==0 ){ 3643 sSum.n = 0; 3644 break; 3645 }else if( once ){ 3646 whereOrMove(&sSum, &sCur); 3647 once = 0; 3648 }else{ 3649 WhereOrSet sPrev; 3650 whereOrMove(&sPrev, &sSum); 3651 sSum.n = 0; 3652 for(i=0; i<sPrev.n; i++){ 3653 for(j=0; j<sCur.n; j++){ 3654 whereOrInsert(&sSum, sPrev.a[i].prereq | sCur.a[j].prereq, 3655 sqlite3LogEstAdd(sPrev.a[i].rRun, sCur.a[j].rRun), 3656 sqlite3LogEstAdd(sPrev.a[i].nOut, sCur.a[j].nOut)); 3657 } 3658 } 3659 } 3660 } 3661 pNew->nLTerm = 1; 3662 pNew->aLTerm[0] = pTerm; 3663 pNew->wsFlags = WHERE_MULTI_OR; 3664 pNew->rSetup = 0; 3665 pNew->iSortIdx = 0; 3666 memset(&pNew->u, 0, sizeof(pNew->u)); 3667 for(i=0; rc==SQLITE_OK && i<sSum.n; i++){ 3668 /* TUNING: Currently sSum.a[i].rRun is set to the sum of the costs 3669 ** of all sub-scans required by the OR-scan. However, due to rounding 3670 ** errors, it may be that the cost of the OR-scan is equal to its 3671 ** most expensive sub-scan. Add the smallest possible penalty 3672 ** (equivalent to multiplying the cost by 1.07) to ensure that 3673 ** this does not happen. Otherwise, for WHERE clauses such as the 3674 ** following where there is an index on "y": 3675 ** 3676 ** WHERE likelihood(x=?, 0.99) OR y=? 3677 ** 3678 ** the planner may elect to "OR" together a full-table scan and an 3679 ** index lookup. And other similarly odd results. */ 3680 pNew->rRun = sSum.a[i].rRun + 1; 3681 pNew->nOut = sSum.a[i].nOut; 3682 pNew->prereq = sSum.a[i].prereq; 3683 rc = whereLoopInsert(pBuilder, pNew); 3684 } 3685 WHERETRACE(0x200, ("End processing OR-clause %p\n", pTerm)); 3686 } 3687 } 3688 return rc; 3689 } 3690 3691 /* 3692 ** Add all WhereLoop objects for all tables 3693 */ 3694 static int whereLoopAddAll(WhereLoopBuilder *pBuilder){ 3695 WhereInfo *pWInfo = pBuilder->pWInfo; 3696 Bitmask mPrereq = 0; 3697 Bitmask mPrior = 0; 3698 int iTab; 3699 SrcList *pTabList = pWInfo->pTabList; 3700 SrcItem *pItem; 3701 SrcItem *pEnd = &pTabList->a[pWInfo->nLevel]; 3702 sqlite3 *db = pWInfo->pParse->db; 3703 int rc = SQLITE_OK; 3704 WhereLoop *pNew; 3705 3706 /* Loop over the tables in the join, from left to right */ 3707 pNew = pBuilder->pNew; 3708 whereLoopInit(pNew); 3709 pBuilder->iPlanLimit = SQLITE_QUERY_PLANNER_LIMIT; 3710 for(iTab=0, pItem=pTabList->a; pItem<pEnd; iTab++, pItem++){ 3711 Bitmask mUnusable = 0; 3712 pNew->iTab = iTab; 3713 pBuilder->iPlanLimit += SQLITE_QUERY_PLANNER_LIMIT_INCR; 3714 pNew->maskSelf = sqlite3WhereGetMask(&pWInfo->sMaskSet, pItem->iCursor); 3715 if( (pItem->fg.jointype & (JT_LEFT|JT_CROSS))!=0 ){ 3716 /* This condition is true when pItem is the FROM clause term on the 3717 ** right-hand-side of a LEFT or CROSS JOIN. */ 3718 mPrereq = mPrior; 3719 }else{ 3720 mPrereq = 0; 3721 } 3722 #ifndef SQLITE_OMIT_VIRTUALTABLE 3723 if( IsVirtual(pItem->pTab) ){ 3724 SrcItem *p; 3725 for(p=&pItem[1]; p<pEnd; p++){ 3726 if( mUnusable || (p->fg.jointype & (JT_LEFT|JT_CROSS)) ){ 3727 mUnusable |= sqlite3WhereGetMask(&pWInfo->sMaskSet, p->iCursor); 3728 } 3729 } 3730 rc = whereLoopAddVirtual(pBuilder, mPrereq, mUnusable); 3731 }else 3732 #endif /* SQLITE_OMIT_VIRTUALTABLE */ 3733 { 3734 rc = whereLoopAddBtree(pBuilder, mPrereq); 3735 } 3736 if( rc==SQLITE_OK && pBuilder->pWC->hasOr ){ 3737 rc = whereLoopAddOr(pBuilder, mPrereq, mUnusable); 3738 } 3739 mPrior |= pNew->maskSelf; 3740 if( rc || db->mallocFailed ){ 3741 if( rc==SQLITE_DONE ){ 3742 /* We hit the query planner search limit set by iPlanLimit */ 3743 sqlite3_log(SQLITE_WARNING, "abbreviated query algorithm search"); 3744 rc = SQLITE_OK; 3745 }else{ 3746 break; 3747 } 3748 } 3749 } 3750 3751 whereLoopClear(db, pNew); 3752 return rc; 3753 } 3754 3755 /* 3756 ** Examine a WherePath (with the addition of the extra WhereLoop of the 6th 3757 ** parameters) to see if it outputs rows in the requested ORDER BY 3758 ** (or GROUP BY) without requiring a separate sort operation. Return N: 3759 ** 3760 ** N>0: N terms of the ORDER BY clause are satisfied 3761 ** N==0: No terms of the ORDER BY clause are satisfied 3762 ** N<0: Unknown yet how many terms of ORDER BY might be satisfied. 3763 ** 3764 ** Note that processing for WHERE_GROUPBY and WHERE_DISTINCTBY is not as 3765 ** strict. With GROUP BY and DISTINCT the only requirement is that 3766 ** equivalent rows appear immediately adjacent to one another. GROUP BY 3767 ** and DISTINCT do not require rows to appear in any particular order as long 3768 ** as equivalent rows are grouped together. Thus for GROUP BY and DISTINCT 3769 ** the pOrderBy terms can be matched in any order. With ORDER BY, the 3770 ** pOrderBy terms must be matched in strict left-to-right order. 3771 */ 3772 static i8 wherePathSatisfiesOrderBy( 3773 WhereInfo *pWInfo, /* The WHERE clause */ 3774 ExprList *pOrderBy, /* ORDER BY or GROUP BY or DISTINCT clause to check */ 3775 WherePath *pPath, /* The WherePath to check */ 3776 u16 wctrlFlags, /* WHERE_GROUPBY or _DISTINCTBY or _ORDERBY_LIMIT */ 3777 u16 nLoop, /* Number of entries in pPath->aLoop[] */ 3778 WhereLoop *pLast, /* Add this WhereLoop to the end of pPath->aLoop[] */ 3779 Bitmask *pRevMask /* OUT: Mask of WhereLoops to run in reverse order */ 3780 ){ 3781 u8 revSet; /* True if rev is known */ 3782 u8 rev; /* Composite sort order */ 3783 u8 revIdx; /* Index sort order */ 3784 u8 isOrderDistinct; /* All prior WhereLoops are order-distinct */ 3785 u8 distinctColumns; /* True if the loop has UNIQUE NOT NULL columns */ 3786 u8 isMatch; /* iColumn matches a term of the ORDER BY clause */ 3787 u16 eqOpMask; /* Allowed equality operators */ 3788 u16 nKeyCol; /* Number of key columns in pIndex */ 3789 u16 nColumn; /* Total number of ordered columns in the index */ 3790 u16 nOrderBy; /* Number terms in the ORDER BY clause */ 3791 int iLoop; /* Index of WhereLoop in pPath being processed */ 3792 int i, j; /* Loop counters */ 3793 int iCur; /* Cursor number for current WhereLoop */ 3794 int iColumn; /* A column number within table iCur */ 3795 WhereLoop *pLoop = 0; /* Current WhereLoop being processed. */ 3796 WhereTerm *pTerm; /* A single term of the WHERE clause */ 3797 Expr *pOBExpr; /* An expression from the ORDER BY clause */ 3798 CollSeq *pColl; /* COLLATE function from an ORDER BY clause term */ 3799 Index *pIndex; /* The index associated with pLoop */ 3800 sqlite3 *db = pWInfo->pParse->db; /* Database connection */ 3801 Bitmask obSat = 0; /* Mask of ORDER BY terms satisfied so far */ 3802 Bitmask obDone; /* Mask of all ORDER BY terms */ 3803 Bitmask orderDistinctMask; /* Mask of all well-ordered loops */ 3804 Bitmask ready; /* Mask of inner loops */ 3805 3806 /* 3807 ** We say the WhereLoop is "one-row" if it generates no more than one 3808 ** row of output. A WhereLoop is one-row if all of the following are true: 3809 ** (a) All index columns match with WHERE_COLUMN_EQ. 3810 ** (b) The index is unique 3811 ** Any WhereLoop with an WHERE_COLUMN_EQ constraint on the rowid is one-row. 3812 ** Every one-row WhereLoop will have the WHERE_ONEROW bit set in wsFlags. 3813 ** 3814 ** We say the WhereLoop is "order-distinct" if the set of columns from 3815 ** that WhereLoop that are in the ORDER BY clause are different for every 3816 ** row of the WhereLoop. Every one-row WhereLoop is automatically 3817 ** order-distinct. A WhereLoop that has no columns in the ORDER BY clause 3818 ** is not order-distinct. To be order-distinct is not quite the same as being 3819 ** UNIQUE since a UNIQUE column or index can have multiple rows that 3820 ** are NULL and NULL values are equivalent for the purpose of order-distinct. 3821 ** To be order-distinct, the columns must be UNIQUE and NOT NULL. 3822 ** 3823 ** The rowid for a table is always UNIQUE and NOT NULL so whenever the 3824 ** rowid appears in the ORDER BY clause, the corresponding WhereLoop is 3825 ** automatically order-distinct. 3826 */ 3827 3828 assert( pOrderBy!=0 ); 3829 if( nLoop && OptimizationDisabled(db, SQLITE_OrderByIdxJoin) ) return 0; 3830 3831 nOrderBy = pOrderBy->nExpr; 3832 testcase( nOrderBy==BMS-1 ); 3833 if( nOrderBy>BMS-1 ) return 0; /* Cannot optimize overly large ORDER BYs */ 3834 isOrderDistinct = 1; 3835 obDone = MASKBIT(nOrderBy)-1; 3836 orderDistinctMask = 0; 3837 ready = 0; 3838 eqOpMask = WO_EQ | WO_IS | WO_ISNULL; 3839 if( wctrlFlags & (WHERE_ORDERBY_LIMIT|WHERE_ORDERBY_MAX|WHERE_ORDERBY_MIN) ){ 3840 eqOpMask |= WO_IN; 3841 } 3842 for(iLoop=0; isOrderDistinct && obSat<obDone && iLoop<=nLoop; iLoop++){ 3843 if( iLoop>0 ) ready |= pLoop->maskSelf; 3844 if( iLoop<nLoop ){ 3845 pLoop = pPath->aLoop[iLoop]; 3846 if( wctrlFlags & WHERE_ORDERBY_LIMIT ) continue; 3847 }else{ 3848 pLoop = pLast; 3849 } 3850 if( pLoop->wsFlags & WHERE_VIRTUALTABLE ){ 3851 if( pLoop->u.vtab.isOrdered && (wctrlFlags & WHERE_DISTINCTBY)==0 ){ 3852 obSat = obDone; 3853 } 3854 break; 3855 }else if( wctrlFlags & WHERE_DISTINCTBY ){ 3856 pLoop->u.btree.nDistinctCol = 0; 3857 } 3858 iCur = pWInfo->pTabList->a[pLoop->iTab].iCursor; 3859 3860 /* Mark off any ORDER BY term X that is a column in the table of 3861 ** the current loop for which there is term in the WHERE 3862 ** clause of the form X IS NULL or X=? that reference only outer 3863 ** loops. 3864 */ 3865 for(i=0; i<nOrderBy; i++){ 3866 if( MASKBIT(i) & obSat ) continue; 3867 pOBExpr = sqlite3ExprSkipCollateAndLikely(pOrderBy->a[i].pExpr); 3868 if( NEVER(pOBExpr==0) ) continue; 3869 if( pOBExpr->op!=TK_COLUMN && pOBExpr->op!=TK_AGG_COLUMN ) continue; 3870 if( pOBExpr->iTable!=iCur ) continue; 3871 pTerm = sqlite3WhereFindTerm(&pWInfo->sWC, iCur, pOBExpr->iColumn, 3872 ~ready, eqOpMask, 0); 3873 if( pTerm==0 ) continue; 3874 if( pTerm->eOperator==WO_IN ){ 3875 /* IN terms are only valid for sorting in the ORDER BY LIMIT 3876 ** optimization, and then only if they are actually used 3877 ** by the query plan */ 3878 assert( wctrlFlags & 3879 (WHERE_ORDERBY_LIMIT|WHERE_ORDERBY_MIN|WHERE_ORDERBY_MAX) ); 3880 for(j=0; j<pLoop->nLTerm && pTerm!=pLoop->aLTerm[j]; j++){} 3881 if( j>=pLoop->nLTerm ) continue; 3882 } 3883 if( (pTerm->eOperator&(WO_EQ|WO_IS))!=0 && pOBExpr->iColumn>=0 ){ 3884 Parse *pParse = pWInfo->pParse; 3885 CollSeq *pColl1 = sqlite3ExprNNCollSeq(pParse, pOrderBy->a[i].pExpr); 3886 CollSeq *pColl2 = sqlite3ExprCompareCollSeq(pParse, pTerm->pExpr); 3887 assert( pColl1 ); 3888 if( pColl2==0 || sqlite3StrICmp(pColl1->zName, pColl2->zName) ){ 3889 continue; 3890 } 3891 testcase( pTerm->pExpr->op==TK_IS ); 3892 } 3893 obSat |= MASKBIT(i); 3894 } 3895 3896 if( (pLoop->wsFlags & WHERE_ONEROW)==0 ){ 3897 if( pLoop->wsFlags & WHERE_IPK ){ 3898 pIndex = 0; 3899 nKeyCol = 0; 3900 nColumn = 1; 3901 }else if( (pIndex = pLoop->u.btree.pIndex)==0 || pIndex->bUnordered ){ 3902 return 0; 3903 }else{ 3904 nKeyCol = pIndex->nKeyCol; 3905 nColumn = pIndex->nColumn; 3906 assert( nColumn==nKeyCol+1 || !HasRowid(pIndex->pTable) ); 3907 assert( pIndex->aiColumn[nColumn-1]==XN_ROWID 3908 || !HasRowid(pIndex->pTable)); 3909 /* All relevant terms of the index must also be non-NULL in order 3910 ** for isOrderDistinct to be true. So the isOrderDistint value 3911 ** computed here might be a false positive. Corrections will be 3912 ** made at tag-20210426-1 below */ 3913 isOrderDistinct = IsUniqueIndex(pIndex) 3914 && (pLoop->wsFlags & WHERE_SKIPSCAN)==0; 3915 } 3916 3917 /* Loop through all columns of the index and deal with the ones 3918 ** that are not constrained by == or IN. 3919 */ 3920 rev = revSet = 0; 3921 distinctColumns = 0; 3922 for(j=0; j<nColumn; j++){ 3923 u8 bOnce = 1; /* True to run the ORDER BY search loop */ 3924 3925 assert( j>=pLoop->u.btree.nEq 3926 || (pLoop->aLTerm[j]==0)==(j<pLoop->nSkip) 3927 ); 3928 if( j<pLoop->u.btree.nEq && j>=pLoop->nSkip ){ 3929 u16 eOp = pLoop->aLTerm[j]->eOperator; 3930 3931 /* Skip over == and IS and ISNULL terms. (Also skip IN terms when 3932 ** doing WHERE_ORDERBY_LIMIT processing). Except, IS and ISNULL 3933 ** terms imply that the index is not UNIQUE NOT NULL in which case 3934 ** the loop need to be marked as not order-distinct because it can 3935 ** have repeated NULL rows. 3936 ** 3937 ** If the current term is a column of an ((?,?) IN (SELECT...)) 3938 ** expression for which the SELECT returns more than one column, 3939 ** check that it is the only column used by this loop. Otherwise, 3940 ** if it is one of two or more, none of the columns can be 3941 ** considered to match an ORDER BY term. 3942 */ 3943 if( (eOp & eqOpMask)!=0 ){ 3944 if( eOp & (WO_ISNULL|WO_IS) ){ 3945 testcase( eOp & WO_ISNULL ); 3946 testcase( eOp & WO_IS ); 3947 testcase( isOrderDistinct ); 3948 isOrderDistinct = 0; 3949 } 3950 continue; 3951 }else if( ALWAYS(eOp & WO_IN) ){ 3952 /* ALWAYS() justification: eOp is an equality operator due to the 3953 ** j<pLoop->u.btree.nEq constraint above. Any equality other 3954 ** than WO_IN is captured by the previous "if". So this one 3955 ** always has to be WO_IN. */ 3956 Expr *pX = pLoop->aLTerm[j]->pExpr; 3957 for(i=j+1; i<pLoop->u.btree.nEq; i++){ 3958 if( pLoop->aLTerm[i]->pExpr==pX ){ 3959 assert( (pLoop->aLTerm[i]->eOperator & WO_IN) ); 3960 bOnce = 0; 3961 break; 3962 } 3963 } 3964 } 3965 } 3966 3967 /* Get the column number in the table (iColumn) and sort order 3968 ** (revIdx) for the j-th column of the index. 3969 */ 3970 if( pIndex ){ 3971 iColumn = pIndex->aiColumn[j]; 3972 revIdx = pIndex->aSortOrder[j] & KEYINFO_ORDER_DESC; 3973 if( iColumn==pIndex->pTable->iPKey ) iColumn = XN_ROWID; 3974 }else{ 3975 iColumn = XN_ROWID; 3976 revIdx = 0; 3977 } 3978 3979 /* An unconstrained column that might be NULL means that this 3980 ** WhereLoop is not well-ordered. tag-20210426-1 3981 */ 3982 if( isOrderDistinct ){ 3983 if( iColumn>=0 3984 && j>=pLoop->u.btree.nEq 3985 && pIndex->pTable->aCol[iColumn].notNull==0 3986 ){ 3987 isOrderDistinct = 0; 3988 } 3989 if( iColumn==XN_EXPR ){ 3990 isOrderDistinct = 0; 3991 } 3992 } 3993 3994 /* Find the ORDER BY term that corresponds to the j-th column 3995 ** of the index and mark that ORDER BY term off 3996 */ 3997 isMatch = 0; 3998 for(i=0; bOnce && i<nOrderBy; i++){ 3999 if( MASKBIT(i) & obSat ) continue; 4000 pOBExpr = sqlite3ExprSkipCollateAndLikely(pOrderBy->a[i].pExpr); 4001 testcase( wctrlFlags & WHERE_GROUPBY ); 4002 testcase( wctrlFlags & WHERE_DISTINCTBY ); 4003 if( NEVER(pOBExpr==0) ) continue; 4004 if( (wctrlFlags & (WHERE_GROUPBY|WHERE_DISTINCTBY))==0 ) bOnce = 0; 4005 if( iColumn>=XN_ROWID ){ 4006 if( pOBExpr->op!=TK_COLUMN && pOBExpr->op!=TK_AGG_COLUMN ) continue; 4007 if( pOBExpr->iTable!=iCur ) continue; 4008 if( pOBExpr->iColumn!=iColumn ) continue; 4009 }else{ 4010 Expr *pIdxExpr = pIndex->aColExpr->a[j].pExpr; 4011 if( sqlite3ExprCompareSkip(pOBExpr, pIdxExpr, iCur) ){ 4012 continue; 4013 } 4014 } 4015 if( iColumn!=XN_ROWID ){ 4016 pColl = sqlite3ExprNNCollSeq(pWInfo->pParse, pOrderBy->a[i].pExpr); 4017 if( sqlite3StrICmp(pColl->zName, pIndex->azColl[j])!=0 ) continue; 4018 } 4019 if( wctrlFlags & WHERE_DISTINCTBY ){ 4020 pLoop->u.btree.nDistinctCol = j+1; 4021 } 4022 isMatch = 1; 4023 break; 4024 } 4025 if( isMatch && (wctrlFlags & WHERE_GROUPBY)==0 ){ 4026 /* Make sure the sort order is compatible in an ORDER BY clause. 4027 ** Sort order is irrelevant for a GROUP BY clause. */ 4028 if( revSet ){ 4029 if( (rev ^ revIdx)!=(pOrderBy->a[i].sortFlags&KEYINFO_ORDER_DESC) ){ 4030 isMatch = 0; 4031 } 4032 }else{ 4033 rev = revIdx ^ (pOrderBy->a[i].sortFlags & KEYINFO_ORDER_DESC); 4034 if( rev ) *pRevMask |= MASKBIT(iLoop); 4035 revSet = 1; 4036 } 4037 } 4038 if( isMatch && (pOrderBy->a[i].sortFlags & KEYINFO_ORDER_BIGNULL) ){ 4039 if( j==pLoop->u.btree.nEq ){ 4040 pLoop->wsFlags |= WHERE_BIGNULL_SORT; 4041 }else{ 4042 isMatch = 0; 4043 } 4044 } 4045 if( isMatch ){ 4046 if( iColumn==XN_ROWID ){ 4047 testcase( distinctColumns==0 ); 4048 distinctColumns = 1; 4049 } 4050 obSat |= MASKBIT(i); 4051 }else{ 4052 /* No match found */ 4053 if( j==0 || j<nKeyCol ){ 4054 testcase( isOrderDistinct!=0 ); 4055 isOrderDistinct = 0; 4056 } 4057 break; 4058 } 4059 } /* end Loop over all index columns */ 4060 if( distinctColumns ){ 4061 testcase( isOrderDistinct==0 ); 4062 isOrderDistinct = 1; 4063 } 4064 } /* end-if not one-row */ 4065 4066 /* Mark off any other ORDER BY terms that reference pLoop */ 4067 if( isOrderDistinct ){ 4068 orderDistinctMask |= pLoop->maskSelf; 4069 for(i=0; i<nOrderBy; i++){ 4070 Expr *p; 4071 Bitmask mTerm; 4072 if( MASKBIT(i) & obSat ) continue; 4073 p = pOrderBy->a[i].pExpr; 4074 mTerm = sqlite3WhereExprUsage(&pWInfo->sMaskSet,p); 4075 if( mTerm==0 && !sqlite3ExprIsConstant(p) ) continue; 4076 if( (mTerm&~orderDistinctMask)==0 ){ 4077 obSat |= MASKBIT(i); 4078 } 4079 } 4080 } 4081 } /* End the loop over all WhereLoops from outer-most down to inner-most */ 4082 if( obSat==obDone ) return (i8)nOrderBy; 4083 if( !isOrderDistinct ){ 4084 for(i=nOrderBy-1; i>0; i--){ 4085 Bitmask m = MASKBIT(i) - 1; 4086 if( (obSat&m)==m ) return i; 4087 } 4088 return 0; 4089 } 4090 return -1; 4091 } 4092 4093 4094 /* 4095 ** If the WHERE_GROUPBY flag is set in the mask passed to sqlite3WhereBegin(), 4096 ** the planner assumes that the specified pOrderBy list is actually a GROUP 4097 ** BY clause - and so any order that groups rows as required satisfies the 4098 ** request. 4099 ** 4100 ** Normally, in this case it is not possible for the caller to determine 4101 ** whether or not the rows are really being delivered in sorted order, or 4102 ** just in some other order that provides the required grouping. However, 4103 ** if the WHERE_SORTBYGROUP flag is also passed to sqlite3WhereBegin(), then 4104 ** this function may be called on the returned WhereInfo object. It returns 4105 ** true if the rows really will be sorted in the specified order, or false 4106 ** otherwise. 4107 ** 4108 ** For example, assuming: 4109 ** 4110 ** CREATE INDEX i1 ON t1(x, Y); 4111 ** 4112 ** then 4113 ** 4114 ** SELECT * FROM t1 GROUP BY x,y ORDER BY x,y; -- IsSorted()==1 4115 ** SELECT * FROM t1 GROUP BY y,x ORDER BY y,x; -- IsSorted()==0 4116 */ 4117 int sqlite3WhereIsSorted(WhereInfo *pWInfo){ 4118 assert( pWInfo->wctrlFlags & WHERE_GROUPBY ); 4119 assert( pWInfo->wctrlFlags & WHERE_SORTBYGROUP ); 4120 return pWInfo->sorted; 4121 } 4122 4123 #ifdef WHERETRACE_ENABLED 4124 /* For debugging use only: */ 4125 static const char *wherePathName(WherePath *pPath, int nLoop, WhereLoop *pLast){ 4126 static char zName[65]; 4127 int i; 4128 for(i=0; i<nLoop; i++){ zName[i] = pPath->aLoop[i]->cId; } 4129 if( pLast ) zName[i++] = pLast->cId; 4130 zName[i] = 0; 4131 return zName; 4132 } 4133 #endif 4134 4135 /* 4136 ** Return the cost of sorting nRow rows, assuming that the keys have 4137 ** nOrderby columns and that the first nSorted columns are already in 4138 ** order. 4139 */ 4140 static LogEst whereSortingCost( 4141 WhereInfo *pWInfo, 4142 LogEst nRow, 4143 int nOrderBy, 4144 int nSorted 4145 ){ 4146 /* TUNING: Estimated cost of a full external sort, where N is 4147 ** the number of rows to sort is: 4148 ** 4149 ** cost = (3.0 * N * log(N)). 4150 ** 4151 ** Or, if the order-by clause has X terms but only the last Y 4152 ** terms are out of order, then block-sorting will reduce the 4153 ** sorting cost to: 4154 ** 4155 ** cost = (3.0 * N * log(N)) * (Y/X) 4156 ** 4157 ** The (Y/X) term is implemented using stack variable rScale 4158 ** below. 4159 */ 4160 LogEst rScale, rSortCost; 4161 assert( nOrderBy>0 && 66==sqlite3LogEst(100) ); 4162 rScale = sqlite3LogEst((nOrderBy-nSorted)*100/nOrderBy) - 66; 4163 rSortCost = nRow + rScale + 16; 4164 4165 /* Multiple by log(M) where M is the number of output rows. 4166 ** Use the LIMIT for M if it is smaller. Or if this sort is for 4167 ** a DISTINCT operator, M will be the number of distinct output 4168 ** rows, so fudge it downwards a bit. 4169 */ 4170 if( (pWInfo->wctrlFlags & WHERE_USE_LIMIT)!=0 && pWInfo->iLimit<nRow ){ 4171 nRow = pWInfo->iLimit; 4172 }else if( (pWInfo->wctrlFlags & WHERE_WANT_DISTINCT) ){ 4173 /* TUNING: In the sort for a DISTINCT operator, assume that the DISTINCT 4174 ** reduces the number of output rows by a factor of 2 */ 4175 if( nRow>10 ){ nRow -= 10; assert( 10==sqlite3LogEst(2) ); } 4176 } 4177 rSortCost += estLog(nRow); 4178 return rSortCost; 4179 } 4180 4181 /* 4182 ** Given the list of WhereLoop objects at pWInfo->pLoops, this routine 4183 ** attempts to find the lowest cost path that visits each WhereLoop 4184 ** once. This path is then loaded into the pWInfo->a[].pWLoop fields. 4185 ** 4186 ** Assume that the total number of output rows that will need to be sorted 4187 ** will be nRowEst (in the 10*log2 representation). Or, ignore sorting 4188 ** costs if nRowEst==0. 4189 ** 4190 ** Return SQLITE_OK on success or SQLITE_NOMEM of a memory allocation 4191 ** error occurs. 4192 */ 4193 static int wherePathSolver(WhereInfo *pWInfo, LogEst nRowEst){ 4194 int mxChoice; /* Maximum number of simultaneous paths tracked */ 4195 int nLoop; /* Number of terms in the join */ 4196 Parse *pParse; /* Parsing context */ 4197 sqlite3 *db; /* The database connection */ 4198 int iLoop; /* Loop counter over the terms of the join */ 4199 int ii, jj; /* Loop counters */ 4200 int mxI = 0; /* Index of next entry to replace */ 4201 int nOrderBy; /* Number of ORDER BY clause terms */ 4202 LogEst mxCost = 0; /* Maximum cost of a set of paths */ 4203 LogEst mxUnsorted = 0; /* Maximum unsorted cost of a set of path */ 4204 int nTo, nFrom; /* Number of valid entries in aTo[] and aFrom[] */ 4205 WherePath *aFrom; /* All nFrom paths at the previous level */ 4206 WherePath *aTo; /* The nTo best paths at the current level */ 4207 WherePath *pFrom; /* An element of aFrom[] that we are working on */ 4208 WherePath *pTo; /* An element of aTo[] that we are working on */ 4209 WhereLoop *pWLoop; /* One of the WhereLoop objects */ 4210 WhereLoop **pX; /* Used to divy up the pSpace memory */ 4211 LogEst *aSortCost = 0; /* Sorting and partial sorting costs */ 4212 char *pSpace; /* Temporary memory used by this routine */ 4213 int nSpace; /* Bytes of space allocated at pSpace */ 4214 4215 pParse = pWInfo->pParse; 4216 db = pParse->db; 4217 nLoop = pWInfo->nLevel; 4218 /* TUNING: For simple queries, only the best path is tracked. 4219 ** For 2-way joins, the 5 best paths are followed. 4220 ** For joins of 3 or more tables, track the 10 best paths */ 4221 mxChoice = (nLoop<=1) ? 1 : (nLoop==2 ? 5 : 10); 4222 assert( nLoop<=pWInfo->pTabList->nSrc ); 4223 WHERETRACE(0x002, ("---- begin solver. (nRowEst=%d)\n", nRowEst)); 4224 4225 /* If nRowEst is zero and there is an ORDER BY clause, ignore it. In this 4226 ** case the purpose of this call is to estimate the number of rows returned 4227 ** by the overall query. Once this estimate has been obtained, the caller 4228 ** will invoke this function a second time, passing the estimate as the 4229 ** nRowEst parameter. */ 4230 if( pWInfo->pOrderBy==0 || nRowEst==0 ){ 4231 nOrderBy = 0; 4232 }else{ 4233 nOrderBy = pWInfo->pOrderBy->nExpr; 4234 } 4235 4236 /* Allocate and initialize space for aTo, aFrom and aSortCost[] */ 4237 nSpace = (sizeof(WherePath)+sizeof(WhereLoop*)*nLoop)*mxChoice*2; 4238 nSpace += sizeof(LogEst) * nOrderBy; 4239 pSpace = sqlite3DbMallocRawNN(db, nSpace); 4240 if( pSpace==0 ) return SQLITE_NOMEM_BKPT; 4241 aTo = (WherePath*)pSpace; 4242 aFrom = aTo+mxChoice; 4243 memset(aFrom, 0, sizeof(aFrom[0])); 4244 pX = (WhereLoop**)(aFrom+mxChoice); 4245 for(ii=mxChoice*2, pFrom=aTo; ii>0; ii--, pFrom++, pX += nLoop){ 4246 pFrom->aLoop = pX; 4247 } 4248 if( nOrderBy ){ 4249 /* If there is an ORDER BY clause and it is not being ignored, set up 4250 ** space for the aSortCost[] array. Each element of the aSortCost array 4251 ** is either zero - meaning it has not yet been initialized - or the 4252 ** cost of sorting nRowEst rows of data where the first X terms of 4253 ** the ORDER BY clause are already in order, where X is the array 4254 ** index. */ 4255 aSortCost = (LogEst*)pX; 4256 memset(aSortCost, 0, sizeof(LogEst) * nOrderBy); 4257 } 4258 assert( aSortCost==0 || &pSpace[nSpace]==(char*)&aSortCost[nOrderBy] ); 4259 assert( aSortCost!=0 || &pSpace[nSpace]==(char*)pX ); 4260 4261 /* Seed the search with a single WherePath containing zero WhereLoops. 4262 ** 4263 ** TUNING: Do not let the number of iterations go above 28. If the cost 4264 ** of computing an automatic index is not paid back within the first 28 4265 ** rows, then do not use the automatic index. */ 4266 aFrom[0].nRow = MIN(pParse->nQueryLoop, 48); assert( 48==sqlite3LogEst(28) ); 4267 nFrom = 1; 4268 assert( aFrom[0].isOrdered==0 ); 4269 if( nOrderBy ){ 4270 /* If nLoop is zero, then there are no FROM terms in the query. Since 4271 ** in this case the query may return a maximum of one row, the results 4272 ** are already in the requested order. Set isOrdered to nOrderBy to 4273 ** indicate this. Or, if nLoop is greater than zero, set isOrdered to 4274 ** -1, indicating that the result set may or may not be ordered, 4275 ** depending on the loops added to the current plan. */ 4276 aFrom[0].isOrdered = nLoop>0 ? -1 : nOrderBy; 4277 } 4278 4279 /* Compute successively longer WherePaths using the previous generation 4280 ** of WherePaths as the basis for the next. Keep track of the mxChoice 4281 ** best paths at each generation */ 4282 for(iLoop=0; iLoop<nLoop; iLoop++){ 4283 nTo = 0; 4284 for(ii=0, pFrom=aFrom; ii<nFrom; ii++, pFrom++){ 4285 for(pWLoop=pWInfo->pLoops; pWLoop; pWLoop=pWLoop->pNextLoop){ 4286 LogEst nOut; /* Rows visited by (pFrom+pWLoop) */ 4287 LogEst rCost; /* Cost of path (pFrom+pWLoop) */ 4288 LogEst rUnsorted; /* Unsorted cost of (pFrom+pWLoop) */ 4289 i8 isOrdered = pFrom->isOrdered; /* isOrdered for (pFrom+pWLoop) */ 4290 Bitmask maskNew; /* Mask of src visited by (..) */ 4291 Bitmask revMask = 0; /* Mask of rev-order loops for (..) */ 4292 4293 if( (pWLoop->prereq & ~pFrom->maskLoop)!=0 ) continue; 4294 if( (pWLoop->maskSelf & pFrom->maskLoop)!=0 ) continue; 4295 if( (pWLoop->wsFlags & WHERE_AUTO_INDEX)!=0 && pFrom->nRow<3 ){ 4296 /* Do not use an automatic index if the this loop is expected 4297 ** to run less than 1.25 times. It is tempting to also exclude 4298 ** automatic index usage on an outer loop, but sometimes an automatic 4299 ** index is useful in the outer loop of a correlated subquery. */ 4300 assert( 10==sqlite3LogEst(2) ); 4301 continue; 4302 } 4303 4304 /* At this point, pWLoop is a candidate to be the next loop. 4305 ** Compute its cost */ 4306 rUnsorted = sqlite3LogEstAdd(pWLoop->rSetup,pWLoop->rRun + pFrom->nRow); 4307 rUnsorted = sqlite3LogEstAdd(rUnsorted, pFrom->rUnsorted); 4308 nOut = pFrom->nRow + pWLoop->nOut; 4309 maskNew = pFrom->maskLoop | pWLoop->maskSelf; 4310 if( isOrdered<0 ){ 4311 isOrdered = wherePathSatisfiesOrderBy(pWInfo, 4312 pWInfo->pOrderBy, pFrom, pWInfo->wctrlFlags, 4313 iLoop, pWLoop, &revMask); 4314 }else{ 4315 revMask = pFrom->revLoop; 4316 } 4317 if( isOrdered>=0 && isOrdered<nOrderBy ){ 4318 if( aSortCost[isOrdered]==0 ){ 4319 aSortCost[isOrdered] = whereSortingCost( 4320 pWInfo, nRowEst, nOrderBy, isOrdered 4321 ); 4322 } 4323 /* TUNING: Add a small extra penalty (5) to sorting as an 4324 ** extra encouragment to the query planner to select a plan 4325 ** where the rows emerge in the correct order without any sorting 4326 ** required. */ 4327 rCost = sqlite3LogEstAdd(rUnsorted, aSortCost[isOrdered]) + 5; 4328 4329 WHERETRACE(0x002, 4330 ("---- sort cost=%-3d (%d/%d) increases cost %3d to %-3d\n", 4331 aSortCost[isOrdered], (nOrderBy-isOrdered), nOrderBy, 4332 rUnsorted, rCost)); 4333 }else{ 4334 rCost = rUnsorted; 4335 rUnsorted -= 2; /* TUNING: Slight bias in favor of no-sort plans */ 4336 } 4337 4338 /* Check to see if pWLoop should be added to the set of 4339 ** mxChoice best-so-far paths. 4340 ** 4341 ** First look for an existing path among best-so-far paths 4342 ** that covers the same set of loops and has the same isOrdered 4343 ** setting as the current path candidate. 4344 ** 4345 ** The term "((pTo->isOrdered^isOrdered)&0x80)==0" is equivalent 4346 ** to (pTo->isOrdered==(-1))==(isOrdered==(-1))" for the range 4347 ** of legal values for isOrdered, -1..64. 4348 */ 4349 for(jj=0, pTo=aTo; jj<nTo; jj++, pTo++){ 4350 if( pTo->maskLoop==maskNew 4351 && ((pTo->isOrdered^isOrdered)&0x80)==0 4352 ){ 4353 testcase( jj==nTo-1 ); 4354 break; 4355 } 4356 } 4357 if( jj>=nTo ){ 4358 /* None of the existing best-so-far paths match the candidate. */ 4359 if( nTo>=mxChoice 4360 && (rCost>mxCost || (rCost==mxCost && rUnsorted>=mxUnsorted)) 4361 ){ 4362 /* The current candidate is no better than any of the mxChoice 4363 ** paths currently in the best-so-far buffer. So discard 4364 ** this candidate as not viable. */ 4365 #ifdef WHERETRACE_ENABLED /* 0x4 */ 4366 if( sqlite3WhereTrace&0x4 ){ 4367 sqlite3DebugPrintf("Skip %s cost=%-3d,%3d,%3d order=%c\n", 4368 wherePathName(pFrom, iLoop, pWLoop), rCost, nOut, rUnsorted, 4369 isOrdered>=0 ? isOrdered+'0' : '?'); 4370 } 4371 #endif 4372 continue; 4373 } 4374 /* If we reach this points it means that the new candidate path 4375 ** needs to be added to the set of best-so-far paths. */ 4376 if( nTo<mxChoice ){ 4377 /* Increase the size of the aTo set by one */ 4378 jj = nTo++; 4379 }else{ 4380 /* New path replaces the prior worst to keep count below mxChoice */ 4381 jj = mxI; 4382 } 4383 pTo = &aTo[jj]; 4384 #ifdef WHERETRACE_ENABLED /* 0x4 */ 4385 if( sqlite3WhereTrace&0x4 ){ 4386 sqlite3DebugPrintf("New %s cost=%-3d,%3d,%3d order=%c\n", 4387 wherePathName(pFrom, iLoop, pWLoop), rCost, nOut, rUnsorted, 4388 isOrdered>=0 ? isOrdered+'0' : '?'); 4389 } 4390 #endif 4391 }else{ 4392 /* Control reaches here if best-so-far path pTo=aTo[jj] covers the 4393 ** same set of loops and has the same isOrdered setting as the 4394 ** candidate path. Check to see if the candidate should replace 4395 ** pTo or if the candidate should be skipped. 4396 ** 4397 ** The conditional is an expanded vector comparison equivalent to: 4398 ** (pTo->rCost,pTo->nRow,pTo->rUnsorted) <= (rCost,nOut,rUnsorted) 4399 */ 4400 if( pTo->rCost<rCost 4401 || (pTo->rCost==rCost 4402 && (pTo->nRow<nOut 4403 || (pTo->nRow==nOut && pTo->rUnsorted<=rUnsorted) 4404 ) 4405 ) 4406 ){ 4407 #ifdef WHERETRACE_ENABLED /* 0x4 */ 4408 if( sqlite3WhereTrace&0x4 ){ 4409 sqlite3DebugPrintf( 4410 "Skip %s cost=%-3d,%3d,%3d order=%c", 4411 wherePathName(pFrom, iLoop, pWLoop), rCost, nOut, rUnsorted, 4412 isOrdered>=0 ? isOrdered+'0' : '?'); 4413 sqlite3DebugPrintf(" vs %s cost=%-3d,%3d,%3d order=%c\n", 4414 wherePathName(pTo, iLoop+1, 0), pTo->rCost, pTo->nRow, 4415 pTo->rUnsorted, pTo->isOrdered>=0 ? pTo->isOrdered+'0' : '?'); 4416 } 4417 #endif 4418 /* Discard the candidate path from further consideration */ 4419 testcase( pTo->rCost==rCost ); 4420 continue; 4421 } 4422 testcase( pTo->rCost==rCost+1 ); 4423 /* Control reaches here if the candidate path is better than the 4424 ** pTo path. Replace pTo with the candidate. */ 4425 #ifdef WHERETRACE_ENABLED /* 0x4 */ 4426 if( sqlite3WhereTrace&0x4 ){ 4427 sqlite3DebugPrintf( 4428 "Update %s cost=%-3d,%3d,%3d order=%c", 4429 wherePathName(pFrom, iLoop, pWLoop), rCost, nOut, rUnsorted, 4430 isOrdered>=0 ? isOrdered+'0' : '?'); 4431 sqlite3DebugPrintf(" was %s cost=%-3d,%3d,%3d order=%c\n", 4432 wherePathName(pTo, iLoop+1, 0), pTo->rCost, pTo->nRow, 4433 pTo->rUnsorted, pTo->isOrdered>=0 ? pTo->isOrdered+'0' : '?'); 4434 } 4435 #endif 4436 } 4437 /* pWLoop is a winner. Add it to the set of best so far */ 4438 pTo->maskLoop = pFrom->maskLoop | pWLoop->maskSelf; 4439 pTo->revLoop = revMask; 4440 pTo->nRow = nOut; 4441 pTo->rCost = rCost; 4442 pTo->rUnsorted = rUnsorted; 4443 pTo->isOrdered = isOrdered; 4444 memcpy(pTo->aLoop, pFrom->aLoop, sizeof(WhereLoop*)*iLoop); 4445 pTo->aLoop[iLoop] = pWLoop; 4446 if( nTo>=mxChoice ){ 4447 mxI = 0; 4448 mxCost = aTo[0].rCost; 4449 mxUnsorted = aTo[0].nRow; 4450 for(jj=1, pTo=&aTo[1]; jj<mxChoice; jj++, pTo++){ 4451 if( pTo->rCost>mxCost 4452 || (pTo->rCost==mxCost && pTo->rUnsorted>mxUnsorted) 4453 ){ 4454 mxCost = pTo->rCost; 4455 mxUnsorted = pTo->rUnsorted; 4456 mxI = jj; 4457 } 4458 } 4459 } 4460 } 4461 } 4462 4463 #ifdef WHERETRACE_ENABLED /* >=2 */ 4464 if( sqlite3WhereTrace & 0x02 ){ 4465 sqlite3DebugPrintf("---- after round %d ----\n", iLoop); 4466 for(ii=0, pTo=aTo; ii<nTo; ii++, pTo++){ 4467 sqlite3DebugPrintf(" %s cost=%-3d nrow=%-3d order=%c", 4468 wherePathName(pTo, iLoop+1, 0), pTo->rCost, pTo->nRow, 4469 pTo->isOrdered>=0 ? (pTo->isOrdered+'0') : '?'); 4470 if( pTo->isOrdered>0 ){ 4471 sqlite3DebugPrintf(" rev=0x%llx\n", pTo->revLoop); 4472 }else{ 4473 sqlite3DebugPrintf("\n"); 4474 } 4475 } 4476 } 4477 #endif 4478 4479 /* Swap the roles of aFrom and aTo for the next generation */ 4480 pFrom = aTo; 4481 aTo = aFrom; 4482 aFrom = pFrom; 4483 nFrom = nTo; 4484 } 4485 4486 if( nFrom==0 ){ 4487 sqlite3ErrorMsg(pParse, "no query solution"); 4488 sqlite3DbFreeNN(db, pSpace); 4489 return SQLITE_ERROR; 4490 } 4491 4492 /* Find the lowest cost path. pFrom will be left pointing to that path */ 4493 pFrom = aFrom; 4494 for(ii=1; ii<nFrom; ii++){ 4495 if( pFrom->rCost>aFrom[ii].rCost ) pFrom = &aFrom[ii]; 4496 } 4497 assert( pWInfo->nLevel==nLoop ); 4498 /* Load the lowest cost path into pWInfo */ 4499 for(iLoop=0; iLoop<nLoop; iLoop++){ 4500 WhereLevel *pLevel = pWInfo->a + iLoop; 4501 pLevel->pWLoop = pWLoop = pFrom->aLoop[iLoop]; 4502 pLevel->iFrom = pWLoop->iTab; 4503 pLevel->iTabCur = pWInfo->pTabList->a[pLevel->iFrom].iCursor; 4504 } 4505 if( (pWInfo->wctrlFlags & WHERE_WANT_DISTINCT)!=0 4506 && (pWInfo->wctrlFlags & WHERE_DISTINCTBY)==0 4507 && pWInfo->eDistinct==WHERE_DISTINCT_NOOP 4508 && nRowEst 4509 ){ 4510 Bitmask notUsed; 4511 int rc = wherePathSatisfiesOrderBy(pWInfo, pWInfo->pResultSet, pFrom, 4512 WHERE_DISTINCTBY, nLoop-1, pFrom->aLoop[nLoop-1], ¬Used); 4513 if( rc==pWInfo->pResultSet->nExpr ){ 4514 pWInfo->eDistinct = WHERE_DISTINCT_ORDERED; 4515 } 4516 } 4517 pWInfo->bOrderedInnerLoop = 0; 4518 if( pWInfo->pOrderBy ){ 4519 if( pWInfo->wctrlFlags & WHERE_DISTINCTBY ){ 4520 if( pFrom->isOrdered==pWInfo->pOrderBy->nExpr ){ 4521 pWInfo->eDistinct = WHERE_DISTINCT_ORDERED; 4522 } 4523 }else{ 4524 pWInfo->nOBSat = pFrom->isOrdered; 4525 pWInfo->revMask = pFrom->revLoop; 4526 if( pWInfo->nOBSat<=0 ){ 4527 pWInfo->nOBSat = 0; 4528 if( nLoop>0 ){ 4529 u32 wsFlags = pFrom->aLoop[nLoop-1]->wsFlags; 4530 if( (wsFlags & WHERE_ONEROW)==0 4531 && (wsFlags&(WHERE_IPK|WHERE_COLUMN_IN))!=(WHERE_IPK|WHERE_COLUMN_IN) 4532 ){ 4533 Bitmask m = 0; 4534 int rc = wherePathSatisfiesOrderBy(pWInfo, pWInfo->pOrderBy, pFrom, 4535 WHERE_ORDERBY_LIMIT, nLoop-1, pFrom->aLoop[nLoop-1], &m); 4536 testcase( wsFlags & WHERE_IPK ); 4537 testcase( wsFlags & WHERE_COLUMN_IN ); 4538 if( rc==pWInfo->pOrderBy->nExpr ){ 4539 pWInfo->bOrderedInnerLoop = 1; 4540 pWInfo->revMask = m; 4541 } 4542 } 4543 } 4544 }else if( nLoop 4545 && pWInfo->nOBSat==1 4546 && (pWInfo->wctrlFlags & (WHERE_ORDERBY_MIN|WHERE_ORDERBY_MAX))!=0 4547 ){ 4548 pWInfo->bOrderedInnerLoop = 1; 4549 } 4550 } 4551 if( (pWInfo->wctrlFlags & WHERE_SORTBYGROUP) 4552 && pWInfo->nOBSat==pWInfo->pOrderBy->nExpr && nLoop>0 4553 ){ 4554 Bitmask revMask = 0; 4555 int nOrder = wherePathSatisfiesOrderBy(pWInfo, pWInfo->pOrderBy, 4556 pFrom, 0, nLoop-1, pFrom->aLoop[nLoop-1], &revMask 4557 ); 4558 assert( pWInfo->sorted==0 ); 4559 if( nOrder==pWInfo->pOrderBy->nExpr ){ 4560 pWInfo->sorted = 1; 4561 pWInfo->revMask = revMask; 4562 } 4563 } 4564 } 4565 4566 4567 pWInfo->nRowOut = pFrom->nRow; 4568 4569 /* Free temporary memory and return success */ 4570 sqlite3DbFreeNN(db, pSpace); 4571 return SQLITE_OK; 4572 } 4573 4574 /* 4575 ** Most queries use only a single table (they are not joins) and have 4576 ** simple == constraints against indexed fields. This routine attempts 4577 ** to plan those simple cases using much less ceremony than the 4578 ** general-purpose query planner, and thereby yield faster sqlite3_prepare() 4579 ** times for the common case. 4580 ** 4581 ** Return non-zero on success, if this query can be handled by this 4582 ** no-frills query planner. Return zero if this query needs the 4583 ** general-purpose query planner. 4584 */ 4585 static int whereShortCut(WhereLoopBuilder *pBuilder){ 4586 WhereInfo *pWInfo; 4587 SrcItem *pItem; 4588 WhereClause *pWC; 4589 WhereTerm *pTerm; 4590 WhereLoop *pLoop; 4591 int iCur; 4592 int j; 4593 Table *pTab; 4594 Index *pIdx; 4595 4596 pWInfo = pBuilder->pWInfo; 4597 if( pWInfo->wctrlFlags & WHERE_OR_SUBCLAUSE ) return 0; 4598 assert( pWInfo->pTabList->nSrc>=1 ); 4599 pItem = pWInfo->pTabList->a; 4600 pTab = pItem->pTab; 4601 if( IsVirtual(pTab) ) return 0; 4602 if( pItem->fg.isIndexedBy ) return 0; 4603 iCur = pItem->iCursor; 4604 pWC = &pWInfo->sWC; 4605 pLoop = pBuilder->pNew; 4606 pLoop->wsFlags = 0; 4607 pLoop->nSkip = 0; 4608 pTerm = sqlite3WhereFindTerm(pWC, iCur, -1, 0, WO_EQ|WO_IS, 0); 4609 if( pTerm ){ 4610 testcase( pTerm->eOperator & WO_IS ); 4611 pLoop->wsFlags = WHERE_COLUMN_EQ|WHERE_IPK|WHERE_ONEROW; 4612 pLoop->aLTerm[0] = pTerm; 4613 pLoop->nLTerm = 1; 4614 pLoop->u.btree.nEq = 1; 4615 /* TUNING: Cost of a rowid lookup is 10 */ 4616 pLoop->rRun = 33; /* 33==sqlite3LogEst(10) */ 4617 }else{ 4618 for(pIdx=pTab->pIndex; pIdx; pIdx=pIdx->pNext){ 4619 int opMask; 4620 assert( pLoop->aLTermSpace==pLoop->aLTerm ); 4621 if( !IsUniqueIndex(pIdx) 4622 || pIdx->pPartIdxWhere!=0 4623 || pIdx->nKeyCol>ArraySize(pLoop->aLTermSpace) 4624 ) continue; 4625 opMask = pIdx->uniqNotNull ? (WO_EQ|WO_IS) : WO_EQ; 4626 for(j=0; j<pIdx->nKeyCol; j++){ 4627 pTerm = sqlite3WhereFindTerm(pWC, iCur, j, 0, opMask, pIdx); 4628 if( pTerm==0 ) break; 4629 testcase( pTerm->eOperator & WO_IS ); 4630 pLoop->aLTerm[j] = pTerm; 4631 } 4632 if( j!=pIdx->nKeyCol ) continue; 4633 pLoop->wsFlags = WHERE_COLUMN_EQ|WHERE_ONEROW|WHERE_INDEXED; 4634 if( pIdx->isCovering || (pItem->colUsed & pIdx->colNotIdxed)==0 ){ 4635 pLoop->wsFlags |= WHERE_IDX_ONLY; 4636 } 4637 pLoop->nLTerm = j; 4638 pLoop->u.btree.nEq = j; 4639 pLoop->u.btree.pIndex = pIdx; 4640 /* TUNING: Cost of a unique index lookup is 15 */ 4641 pLoop->rRun = 39; /* 39==sqlite3LogEst(15) */ 4642 break; 4643 } 4644 } 4645 if( pLoop->wsFlags ){ 4646 pLoop->nOut = (LogEst)1; 4647 pWInfo->a[0].pWLoop = pLoop; 4648 assert( pWInfo->sMaskSet.n==1 && iCur==pWInfo->sMaskSet.ix[0] ); 4649 pLoop->maskSelf = 1; /* sqlite3WhereGetMask(&pWInfo->sMaskSet, iCur); */ 4650 pWInfo->a[0].iTabCur = iCur; 4651 pWInfo->nRowOut = 1; 4652 if( pWInfo->pOrderBy ) pWInfo->nOBSat = pWInfo->pOrderBy->nExpr; 4653 if( pWInfo->wctrlFlags & WHERE_WANT_DISTINCT ){ 4654 pWInfo->eDistinct = WHERE_DISTINCT_UNIQUE; 4655 } 4656 #ifdef SQLITE_DEBUG 4657 pLoop->cId = '0'; 4658 #endif 4659 return 1; 4660 } 4661 return 0; 4662 } 4663 4664 /* 4665 ** Helper function for exprIsDeterministic(). 4666 */ 4667 static int exprNodeIsDeterministic(Walker *pWalker, Expr *pExpr){ 4668 if( pExpr->op==TK_FUNCTION && ExprHasProperty(pExpr, EP_ConstFunc)==0 ){ 4669 pWalker->eCode = 0; 4670 return WRC_Abort; 4671 } 4672 return WRC_Continue; 4673 } 4674 4675 /* 4676 ** Return true if the expression contains no non-deterministic SQL 4677 ** functions. Do not consider non-deterministic SQL functions that are 4678 ** part of sub-select statements. 4679 */ 4680 static int exprIsDeterministic(Expr *p){ 4681 Walker w; 4682 memset(&w, 0, sizeof(w)); 4683 w.eCode = 1; 4684 w.xExprCallback = exprNodeIsDeterministic; 4685 w.xSelectCallback = sqlite3SelectWalkFail; 4686 sqlite3WalkExpr(&w, p); 4687 return w.eCode; 4688 } 4689 4690 4691 #ifdef WHERETRACE_ENABLED 4692 /* 4693 ** Display all WhereLoops in pWInfo 4694 */ 4695 static void showAllWhereLoops(WhereInfo *pWInfo, WhereClause *pWC){ 4696 if( sqlite3WhereTrace ){ /* Display all of the WhereLoop objects */ 4697 WhereLoop *p; 4698 int i; 4699 static const char zLabel[] = "0123456789abcdefghijklmnopqrstuvwyxz" 4700 "ABCDEFGHIJKLMNOPQRSTUVWYXZ"; 4701 for(p=pWInfo->pLoops, i=0; p; p=p->pNextLoop, i++){ 4702 p->cId = zLabel[i%(sizeof(zLabel)-1)]; 4703 sqlite3WhereLoopPrint(p, pWC); 4704 } 4705 } 4706 } 4707 # define WHERETRACE_ALL_LOOPS(W,C) showAllWhereLoops(W,C) 4708 #else 4709 # define WHERETRACE_ALL_LOOPS(W,C) 4710 #endif 4711 4712 /* 4713 ** Generate the beginning of the loop used for WHERE clause processing. 4714 ** The return value is a pointer to an opaque structure that contains 4715 ** information needed to terminate the loop. Later, the calling routine 4716 ** should invoke sqlite3WhereEnd() with the return value of this function 4717 ** in order to complete the WHERE clause processing. 4718 ** 4719 ** If an error occurs, this routine returns NULL. 4720 ** 4721 ** The basic idea is to do a nested loop, one loop for each table in 4722 ** the FROM clause of a select. (INSERT and UPDATE statements are the 4723 ** same as a SELECT with only a single table in the FROM clause.) For 4724 ** example, if the SQL is this: 4725 ** 4726 ** SELECT * FROM t1, t2, t3 WHERE ...; 4727 ** 4728 ** Then the code generated is conceptually like the following: 4729 ** 4730 ** foreach row1 in t1 do \ Code generated 4731 ** foreach row2 in t2 do |-- by sqlite3WhereBegin() 4732 ** foreach row3 in t3 do / 4733 ** ... 4734 ** end \ Code generated 4735 ** end |-- by sqlite3WhereEnd() 4736 ** end / 4737 ** 4738 ** Note that the loops might not be nested in the order in which they 4739 ** appear in the FROM clause if a different order is better able to make 4740 ** use of indices. Note also that when the IN operator appears in 4741 ** the WHERE clause, it might result in additional nested loops for 4742 ** scanning through all values on the right-hand side of the IN. 4743 ** 4744 ** There are Btree cursors associated with each table. t1 uses cursor 4745 ** number pTabList->a[0].iCursor. t2 uses the cursor pTabList->a[1].iCursor. 4746 ** And so forth. This routine generates code to open those VDBE cursors 4747 ** and sqlite3WhereEnd() generates the code to close them. 4748 ** 4749 ** The code that sqlite3WhereBegin() generates leaves the cursors named 4750 ** in pTabList pointing at their appropriate entries. The [...] code 4751 ** can use OP_Column and OP_Rowid opcodes on these cursors to extract 4752 ** data from the various tables of the loop. 4753 ** 4754 ** If the WHERE clause is empty, the foreach loops must each scan their 4755 ** entire tables. Thus a three-way join is an O(N^3) operation. But if 4756 ** the tables have indices and there are terms in the WHERE clause that 4757 ** refer to those indices, a complete table scan can be avoided and the 4758 ** code will run much faster. Most of the work of this routine is checking 4759 ** to see if there are indices that can be used to speed up the loop. 4760 ** 4761 ** Terms of the WHERE clause are also used to limit which rows actually 4762 ** make it to the "..." in the middle of the loop. After each "foreach", 4763 ** terms of the WHERE clause that use only terms in that loop and outer 4764 ** loops are evaluated and if false a jump is made around all subsequent 4765 ** inner loops (or around the "..." if the test occurs within the inner- 4766 ** most loop) 4767 ** 4768 ** OUTER JOINS 4769 ** 4770 ** An outer join of tables t1 and t2 is conceptally coded as follows: 4771 ** 4772 ** foreach row1 in t1 do 4773 ** flag = 0 4774 ** foreach row2 in t2 do 4775 ** start: 4776 ** ... 4777 ** flag = 1 4778 ** end 4779 ** if flag==0 then 4780 ** move the row2 cursor to a null row 4781 ** goto start 4782 ** fi 4783 ** end 4784 ** 4785 ** ORDER BY CLAUSE PROCESSING 4786 ** 4787 ** pOrderBy is a pointer to the ORDER BY clause (or the GROUP BY clause 4788 ** if the WHERE_GROUPBY flag is set in wctrlFlags) of a SELECT statement 4789 ** if there is one. If there is no ORDER BY clause or if this routine 4790 ** is called from an UPDATE or DELETE statement, then pOrderBy is NULL. 4791 ** 4792 ** The iIdxCur parameter is the cursor number of an index. If 4793 ** WHERE_OR_SUBCLAUSE is set, iIdxCur is the cursor number of an index 4794 ** to use for OR clause processing. The WHERE clause should use this 4795 ** specific cursor. If WHERE_ONEPASS_DESIRED is set, then iIdxCur is 4796 ** the first cursor in an array of cursors for all indices. iIdxCur should 4797 ** be used to compute the appropriate cursor depending on which index is 4798 ** used. 4799 */ 4800 WhereInfo *sqlite3WhereBegin( 4801 Parse *pParse, /* The parser context */ 4802 SrcList *pTabList, /* FROM clause: A list of all tables to be scanned */ 4803 Expr *pWhere, /* The WHERE clause */ 4804 ExprList *pOrderBy, /* An ORDER BY (or GROUP BY) clause, or NULL */ 4805 ExprList *pResultSet, /* Query result set. Req'd for DISTINCT */ 4806 u16 wctrlFlags, /* The WHERE_* flags defined in sqliteInt.h */ 4807 int iAuxArg /* If WHERE_OR_SUBCLAUSE is set, index cursor number 4808 ** If WHERE_USE_LIMIT, then the limit amount */ 4809 ){ 4810 int nByteWInfo; /* Num. bytes allocated for WhereInfo struct */ 4811 int nTabList; /* Number of elements in pTabList */ 4812 WhereInfo *pWInfo; /* Will become the return value of this function */ 4813 Vdbe *v = pParse->pVdbe; /* The virtual database engine */ 4814 Bitmask notReady; /* Cursors that are not yet positioned */ 4815 WhereLoopBuilder sWLB; /* The WhereLoop builder */ 4816 WhereMaskSet *pMaskSet; /* The expression mask set */ 4817 WhereLevel *pLevel; /* A single level in pWInfo->a[] */ 4818 WhereLoop *pLoop; /* Pointer to a single WhereLoop object */ 4819 int ii; /* Loop counter */ 4820 sqlite3 *db; /* Database connection */ 4821 int rc; /* Return code */ 4822 u8 bFordelete = 0; /* OPFLAG_FORDELETE or zero, as appropriate */ 4823 4824 assert( (wctrlFlags & WHERE_ONEPASS_MULTIROW)==0 || ( 4825 (wctrlFlags & WHERE_ONEPASS_DESIRED)!=0 4826 && (wctrlFlags & WHERE_OR_SUBCLAUSE)==0 4827 )); 4828 4829 /* Only one of WHERE_OR_SUBCLAUSE or WHERE_USE_LIMIT */ 4830 assert( (wctrlFlags & WHERE_OR_SUBCLAUSE)==0 4831 || (wctrlFlags & WHERE_USE_LIMIT)==0 ); 4832 4833 /* Variable initialization */ 4834 db = pParse->db; 4835 memset(&sWLB, 0, sizeof(sWLB)); 4836 4837 /* An ORDER/GROUP BY clause of more than 63 terms cannot be optimized */ 4838 testcase( pOrderBy && pOrderBy->nExpr==BMS-1 ); 4839 if( pOrderBy && pOrderBy->nExpr>=BMS ) pOrderBy = 0; 4840 sWLB.pOrderBy = pOrderBy; 4841 4842 /* Disable the DISTINCT optimization if SQLITE_DistinctOpt is set via 4843 ** sqlite3_test_ctrl(SQLITE_TESTCTRL_OPTIMIZATIONS,...) */ 4844 if( OptimizationDisabled(db, SQLITE_DistinctOpt) ){ 4845 wctrlFlags &= ~WHERE_WANT_DISTINCT; 4846 } 4847 4848 /* The number of tables in the FROM clause is limited by the number of 4849 ** bits in a Bitmask 4850 */ 4851 testcase( pTabList->nSrc==BMS ); 4852 if( pTabList->nSrc>BMS ){ 4853 sqlite3ErrorMsg(pParse, "at most %d tables in a join", BMS); 4854 return 0; 4855 } 4856 4857 /* This function normally generates a nested loop for all tables in 4858 ** pTabList. But if the WHERE_OR_SUBCLAUSE flag is set, then we should 4859 ** only generate code for the first table in pTabList and assume that 4860 ** any cursors associated with subsequent tables are uninitialized. 4861 */ 4862 nTabList = (wctrlFlags & WHERE_OR_SUBCLAUSE) ? 1 : pTabList->nSrc; 4863 4864 /* Allocate and initialize the WhereInfo structure that will become the 4865 ** return value. A single allocation is used to store the WhereInfo 4866 ** struct, the contents of WhereInfo.a[], the WhereClause structure 4867 ** and the WhereMaskSet structure. Since WhereClause contains an 8-byte 4868 ** field (type Bitmask) it must be aligned on an 8-byte boundary on 4869 ** some architectures. Hence the ROUND8() below. 4870 */ 4871 nByteWInfo = ROUND8(sizeof(WhereInfo)+(nTabList-1)*sizeof(WhereLevel)); 4872 pWInfo = sqlite3DbMallocRawNN(db, nByteWInfo + sizeof(WhereLoop)); 4873 if( db->mallocFailed ){ 4874 sqlite3DbFree(db, pWInfo); 4875 pWInfo = 0; 4876 goto whereBeginError; 4877 } 4878 pWInfo->pParse = pParse; 4879 pWInfo->pTabList = pTabList; 4880 pWInfo->pOrderBy = pOrderBy; 4881 pWInfo->pWhere = pWhere; 4882 pWInfo->pResultSet = pResultSet; 4883 pWInfo->aiCurOnePass[0] = pWInfo->aiCurOnePass[1] = -1; 4884 pWInfo->nLevel = nTabList; 4885 pWInfo->iBreak = pWInfo->iContinue = sqlite3VdbeMakeLabel(pParse); 4886 pWInfo->wctrlFlags = wctrlFlags; 4887 pWInfo->iLimit = iAuxArg; 4888 pWInfo->savedNQueryLoop = pParse->nQueryLoop; 4889 memset(&pWInfo->nOBSat, 0, 4890 offsetof(WhereInfo,sWC) - offsetof(WhereInfo,nOBSat)); 4891 memset(&pWInfo->a[0], 0, sizeof(WhereLoop)+nTabList*sizeof(WhereLevel)); 4892 assert( pWInfo->eOnePass==ONEPASS_OFF ); /* ONEPASS defaults to OFF */ 4893 pMaskSet = &pWInfo->sMaskSet; 4894 sWLB.pWInfo = pWInfo; 4895 sWLB.pWC = &pWInfo->sWC; 4896 sWLB.pNew = (WhereLoop*)(((char*)pWInfo)+nByteWInfo); 4897 assert( EIGHT_BYTE_ALIGNMENT(sWLB.pNew) ); 4898 whereLoopInit(sWLB.pNew); 4899 #ifdef SQLITE_DEBUG 4900 sWLB.pNew->cId = '*'; 4901 #endif 4902 4903 /* Split the WHERE clause into separate subexpressions where each 4904 ** subexpression is separated by an AND operator. 4905 */ 4906 initMaskSet(pMaskSet); 4907 sqlite3WhereClauseInit(&pWInfo->sWC, pWInfo); 4908 sqlite3WhereSplit(&pWInfo->sWC, pWhere, TK_AND); 4909 4910 /* Special case: No FROM clause 4911 */ 4912 if( nTabList==0 ){ 4913 if( pOrderBy ) pWInfo->nOBSat = pOrderBy->nExpr; 4914 if( wctrlFlags & WHERE_WANT_DISTINCT ){ 4915 pWInfo->eDistinct = WHERE_DISTINCT_UNIQUE; 4916 } 4917 ExplainQueryPlan((pParse, 0, "SCAN CONSTANT ROW")); 4918 }else{ 4919 /* Assign a bit from the bitmask to every term in the FROM clause. 4920 ** 4921 ** The N-th term of the FROM clause is assigned a bitmask of 1<<N. 4922 ** 4923 ** The rule of the previous sentence ensures thta if X is the bitmask for 4924 ** a table T, then X-1 is the bitmask for all other tables to the left of T. 4925 ** Knowing the bitmask for all tables to the left of a left join is 4926 ** important. Ticket #3015. 4927 ** 4928 ** Note that bitmasks are created for all pTabList->nSrc tables in 4929 ** pTabList, not just the first nTabList tables. nTabList is normally 4930 ** equal to pTabList->nSrc but might be shortened to 1 if the 4931 ** WHERE_OR_SUBCLAUSE flag is set. 4932 */ 4933 ii = 0; 4934 do{ 4935 createMask(pMaskSet, pTabList->a[ii].iCursor); 4936 sqlite3WhereTabFuncArgs(pParse, &pTabList->a[ii], &pWInfo->sWC); 4937 }while( (++ii)<pTabList->nSrc ); 4938 #ifdef SQLITE_DEBUG 4939 { 4940 Bitmask mx = 0; 4941 for(ii=0; ii<pTabList->nSrc; ii++){ 4942 Bitmask m = sqlite3WhereGetMask(pMaskSet, pTabList->a[ii].iCursor); 4943 assert( m>=mx ); 4944 mx = m; 4945 } 4946 } 4947 #endif 4948 } 4949 4950 /* Analyze all of the subexpressions. */ 4951 sqlite3WhereExprAnalyze(pTabList, &pWInfo->sWC); 4952 if( db->mallocFailed ) goto whereBeginError; 4953 4954 /* Special case: WHERE terms that do not refer to any tables in the join 4955 ** (constant expressions). Evaluate each such term, and jump over all the 4956 ** generated code if the result is not true. 4957 ** 4958 ** Do not do this if the expression contains non-deterministic functions 4959 ** that are not within a sub-select. This is not strictly required, but 4960 ** preserves SQLite's legacy behaviour in the following two cases: 4961 ** 4962 ** FROM ... WHERE random()>0; -- eval random() once per row 4963 ** FROM ... WHERE (SELECT random())>0; -- eval random() once overall 4964 */ 4965 for(ii=0; ii<sWLB.pWC->nTerm; ii++){ 4966 WhereTerm *pT = &sWLB.pWC->a[ii]; 4967 if( pT->wtFlags & TERM_VIRTUAL ) continue; 4968 if( pT->prereqAll==0 && (nTabList==0 || exprIsDeterministic(pT->pExpr)) ){ 4969 sqlite3ExprIfFalse(pParse, pT->pExpr, pWInfo->iBreak, SQLITE_JUMPIFNULL); 4970 pT->wtFlags |= TERM_CODED; 4971 } 4972 } 4973 4974 if( wctrlFlags & WHERE_WANT_DISTINCT ){ 4975 if( isDistinctRedundant(pParse, pTabList, &pWInfo->sWC, pResultSet) ){ 4976 /* The DISTINCT marking is pointless. Ignore it. */ 4977 pWInfo->eDistinct = WHERE_DISTINCT_UNIQUE; 4978 }else if( pOrderBy==0 ){ 4979 /* Try to ORDER BY the result set to make distinct processing easier */ 4980 pWInfo->wctrlFlags |= WHERE_DISTINCTBY; 4981 pWInfo->pOrderBy = pResultSet; 4982 } 4983 } 4984 4985 /* Construct the WhereLoop objects */ 4986 #if defined(WHERETRACE_ENABLED) 4987 if( sqlite3WhereTrace & 0xffff ){ 4988 sqlite3DebugPrintf("*** Optimizer Start *** (wctrlFlags: 0x%x",wctrlFlags); 4989 if( wctrlFlags & WHERE_USE_LIMIT ){ 4990 sqlite3DebugPrintf(", limit: %d", iAuxArg); 4991 } 4992 sqlite3DebugPrintf(")\n"); 4993 if( sqlite3WhereTrace & 0x100 ){ 4994 Select sSelect; 4995 memset(&sSelect, 0, sizeof(sSelect)); 4996 sSelect.selFlags = SF_WhereBegin; 4997 sSelect.pSrc = pTabList; 4998 sSelect.pWhere = pWhere; 4999 sSelect.pOrderBy = pOrderBy; 5000 sSelect.pEList = pResultSet; 5001 sqlite3TreeViewSelect(0, &sSelect, 0); 5002 } 5003 } 5004 if( sqlite3WhereTrace & 0x100 ){ /* Display all terms of the WHERE clause */ 5005 sqlite3DebugPrintf("---- WHERE clause at start of analysis:\n"); 5006 sqlite3WhereClausePrint(sWLB.pWC); 5007 } 5008 #endif 5009 5010 if( nTabList!=1 || whereShortCut(&sWLB)==0 ){ 5011 rc = whereLoopAddAll(&sWLB); 5012 if( rc ) goto whereBeginError; 5013 5014 #ifdef SQLITE_ENABLE_STAT4 5015 /* If one or more WhereTerm.truthProb values were used in estimating 5016 ** loop parameters, but then those truthProb values were subsequently 5017 ** changed based on STAT4 information while computing subsequent loops, 5018 ** then we need to rerun the whole loop building process so that all 5019 ** loops will be built using the revised truthProb values. */ 5020 if( sWLB.bldFlags2 & SQLITE_BLDF2_2NDPASS ){ 5021 WHERETRACE_ALL_LOOPS(pWInfo, sWLB.pWC); 5022 WHERETRACE(0xffff, 5023 ("**** Redo all loop computations due to" 5024 " TERM_HIGHTRUTH changes ****\n")); 5025 while( pWInfo->pLoops ){ 5026 WhereLoop *p = pWInfo->pLoops; 5027 pWInfo->pLoops = p->pNextLoop; 5028 whereLoopDelete(db, p); 5029 } 5030 rc = whereLoopAddAll(&sWLB); 5031 if( rc ) goto whereBeginError; 5032 } 5033 #endif 5034 WHERETRACE_ALL_LOOPS(pWInfo, sWLB.pWC); 5035 5036 wherePathSolver(pWInfo, 0); 5037 if( db->mallocFailed ) goto whereBeginError; 5038 if( pWInfo->pOrderBy ){ 5039 wherePathSolver(pWInfo, pWInfo->nRowOut+1); 5040 if( db->mallocFailed ) goto whereBeginError; 5041 } 5042 } 5043 if( pWInfo->pOrderBy==0 && (db->flags & SQLITE_ReverseOrder)!=0 ){ 5044 pWInfo->revMask = ALLBITS; 5045 } 5046 if( pParse->nErr || db->mallocFailed ){ 5047 goto whereBeginError; 5048 } 5049 #ifdef WHERETRACE_ENABLED 5050 if( sqlite3WhereTrace ){ 5051 sqlite3DebugPrintf("---- Solution nRow=%d", pWInfo->nRowOut); 5052 if( pWInfo->nOBSat>0 ){ 5053 sqlite3DebugPrintf(" ORDERBY=%d,0x%llx", pWInfo->nOBSat, pWInfo->revMask); 5054 } 5055 switch( pWInfo->eDistinct ){ 5056 case WHERE_DISTINCT_UNIQUE: { 5057 sqlite3DebugPrintf(" DISTINCT=unique"); 5058 break; 5059 } 5060 case WHERE_DISTINCT_ORDERED: { 5061 sqlite3DebugPrintf(" DISTINCT=ordered"); 5062 break; 5063 } 5064 case WHERE_DISTINCT_UNORDERED: { 5065 sqlite3DebugPrintf(" DISTINCT=unordered"); 5066 break; 5067 } 5068 } 5069 sqlite3DebugPrintf("\n"); 5070 for(ii=0; ii<pWInfo->nLevel; ii++){ 5071 sqlite3WhereLoopPrint(pWInfo->a[ii].pWLoop, sWLB.pWC); 5072 } 5073 } 5074 #endif 5075 5076 /* Attempt to omit tables from the join that do not affect the result. 5077 ** For a table to not affect the result, the following must be true: 5078 ** 5079 ** 1) The query must not be an aggregate. 5080 ** 2) The table must be the RHS of a LEFT JOIN. 5081 ** 3) Either the query must be DISTINCT, or else the ON or USING clause 5082 ** must contain a constraint that limits the scan of the table to 5083 ** at most a single row. 5084 ** 4) The table must not be referenced by any part of the query apart 5085 ** from its own USING or ON clause. 5086 ** 5087 ** For example, given: 5088 ** 5089 ** CREATE TABLE t1(ipk INTEGER PRIMARY KEY, v1); 5090 ** CREATE TABLE t2(ipk INTEGER PRIMARY KEY, v2); 5091 ** CREATE TABLE t3(ipk INTEGER PRIMARY KEY, v3); 5092 ** 5093 ** then table t2 can be omitted from the following: 5094 ** 5095 ** SELECT v1, v3 FROM t1 5096 ** LEFT JOIN t2 ON (t1.ipk=t2.ipk) 5097 ** LEFT JOIN t3 ON (t1.ipk=t3.ipk) 5098 ** 5099 ** or from: 5100 ** 5101 ** SELECT DISTINCT v1, v3 FROM t1 5102 ** LEFT JOIN t2 5103 ** LEFT JOIN t3 ON (t1.ipk=t3.ipk) 5104 */ 5105 notReady = ~(Bitmask)0; 5106 if( pWInfo->nLevel>=2 5107 && pResultSet!=0 /* these two combine to guarantee */ 5108 && 0==(wctrlFlags & WHERE_AGG_DISTINCT) /* condition (1) above */ 5109 && OptimizationEnabled(db, SQLITE_OmitNoopJoin) 5110 ){ 5111 int i; 5112 Bitmask tabUsed = sqlite3WhereExprListUsage(pMaskSet, pResultSet); 5113 if( sWLB.pOrderBy ){ 5114 tabUsed |= sqlite3WhereExprListUsage(pMaskSet, sWLB.pOrderBy); 5115 } 5116 for(i=pWInfo->nLevel-1; i>=1; i--){ 5117 WhereTerm *pTerm, *pEnd; 5118 SrcItem *pItem; 5119 pLoop = pWInfo->a[i].pWLoop; 5120 pItem = &pWInfo->pTabList->a[pLoop->iTab]; 5121 if( (pItem->fg.jointype & JT_LEFT)==0 ) continue; 5122 if( (wctrlFlags & WHERE_WANT_DISTINCT)==0 5123 && (pLoop->wsFlags & WHERE_ONEROW)==0 5124 ){ 5125 continue; 5126 } 5127 if( (tabUsed & pLoop->maskSelf)!=0 ) continue; 5128 pEnd = sWLB.pWC->a + sWLB.pWC->nTerm; 5129 for(pTerm=sWLB.pWC->a; pTerm<pEnd; pTerm++){ 5130 if( (pTerm->prereqAll & pLoop->maskSelf)!=0 ){ 5131 if( !ExprHasProperty(pTerm->pExpr, EP_FromJoin) 5132 || pTerm->pExpr->iRightJoinTable!=pItem->iCursor 5133 ){ 5134 break; 5135 } 5136 } 5137 } 5138 if( pTerm<pEnd ) continue; 5139 WHERETRACE(0xffff, ("-> drop loop %c not used\n", pLoop->cId)); 5140 notReady &= ~pLoop->maskSelf; 5141 for(pTerm=sWLB.pWC->a; pTerm<pEnd; pTerm++){ 5142 if( (pTerm->prereqAll & pLoop->maskSelf)!=0 ){ 5143 pTerm->wtFlags |= TERM_CODED; 5144 } 5145 } 5146 if( i!=pWInfo->nLevel-1 ){ 5147 int nByte = (pWInfo->nLevel-1-i) * sizeof(WhereLevel); 5148 memmove(&pWInfo->a[i], &pWInfo->a[i+1], nByte); 5149 } 5150 pWInfo->nLevel--; 5151 nTabList--; 5152 } 5153 } 5154 #if defined(WHERETRACE_ENABLED) 5155 if( sqlite3WhereTrace & 0x100 ){ /* Display all terms of the WHERE clause */ 5156 sqlite3DebugPrintf("---- WHERE clause at end of analysis:\n"); 5157 sqlite3WhereClausePrint(sWLB.pWC); 5158 } 5159 WHERETRACE(0xffff,("*** Optimizer Finished ***\n")); 5160 #endif 5161 pWInfo->pParse->nQueryLoop += pWInfo->nRowOut; 5162 5163 /* If the caller is an UPDATE or DELETE statement that is requesting 5164 ** to use a one-pass algorithm, determine if this is appropriate. 5165 ** 5166 ** A one-pass approach can be used if the caller has requested one 5167 ** and either (a) the scan visits at most one row or (b) each 5168 ** of the following are true: 5169 ** 5170 ** * the caller has indicated that a one-pass approach can be used 5171 ** with multiple rows (by setting WHERE_ONEPASS_MULTIROW), and 5172 ** * the table is not a virtual table, and 5173 ** * either the scan does not use the OR optimization or the caller 5174 ** is a DELETE operation (WHERE_DUPLICATES_OK is only specified 5175 ** for DELETE). 5176 ** 5177 ** The last qualification is because an UPDATE statement uses 5178 ** WhereInfo.aiCurOnePass[1] to determine whether or not it really can 5179 ** use a one-pass approach, and this is not set accurately for scans 5180 ** that use the OR optimization. 5181 */ 5182 assert( (wctrlFlags & WHERE_ONEPASS_DESIRED)==0 || pWInfo->nLevel==1 ); 5183 if( (wctrlFlags & WHERE_ONEPASS_DESIRED)!=0 ){ 5184 int wsFlags = pWInfo->a[0].pWLoop->wsFlags; 5185 int bOnerow = (wsFlags & WHERE_ONEROW)!=0; 5186 assert( !(wsFlags & WHERE_VIRTUALTABLE) || IsVirtual(pTabList->a[0].pTab) ); 5187 if( bOnerow || ( 5188 0!=(wctrlFlags & WHERE_ONEPASS_MULTIROW) 5189 && !IsVirtual(pTabList->a[0].pTab) 5190 && (0==(wsFlags & WHERE_MULTI_OR) || (wctrlFlags & WHERE_DUPLICATES_OK)) 5191 )){ 5192 pWInfo->eOnePass = bOnerow ? ONEPASS_SINGLE : ONEPASS_MULTI; 5193 if( HasRowid(pTabList->a[0].pTab) && (wsFlags & WHERE_IDX_ONLY) ){ 5194 if( wctrlFlags & WHERE_ONEPASS_MULTIROW ){ 5195 bFordelete = OPFLAG_FORDELETE; 5196 } 5197 pWInfo->a[0].pWLoop->wsFlags = (wsFlags & ~WHERE_IDX_ONLY); 5198 } 5199 } 5200 } 5201 5202 /* Open all tables in the pTabList and any indices selected for 5203 ** searching those tables. 5204 */ 5205 for(ii=0, pLevel=pWInfo->a; ii<nTabList; ii++, pLevel++){ 5206 Table *pTab; /* Table to open */ 5207 int iDb; /* Index of database containing table/index */ 5208 SrcItem *pTabItem; 5209 5210 pTabItem = &pTabList->a[pLevel->iFrom]; 5211 pTab = pTabItem->pTab; 5212 iDb = sqlite3SchemaToIndex(db, pTab->pSchema); 5213 pLoop = pLevel->pWLoop; 5214 if( (pTab->tabFlags & TF_Ephemeral)!=0 || pTab->pSelect ){ 5215 /* Do nothing */ 5216 }else 5217 #ifndef SQLITE_OMIT_VIRTUALTABLE 5218 if( (pLoop->wsFlags & WHERE_VIRTUALTABLE)!=0 ){ 5219 const char *pVTab = (const char *)sqlite3GetVTable(db, pTab); 5220 int iCur = pTabItem->iCursor; 5221 sqlite3VdbeAddOp4(v, OP_VOpen, iCur, 0, 0, pVTab, P4_VTAB); 5222 }else if( IsVirtual(pTab) ){ 5223 /* noop */ 5224 }else 5225 #endif 5226 if( (pLoop->wsFlags & WHERE_IDX_ONLY)==0 5227 && (wctrlFlags & WHERE_OR_SUBCLAUSE)==0 ){ 5228 int op = OP_OpenRead; 5229 if( pWInfo->eOnePass!=ONEPASS_OFF ){ 5230 op = OP_OpenWrite; 5231 pWInfo->aiCurOnePass[0] = pTabItem->iCursor; 5232 }; 5233 sqlite3OpenTable(pParse, pTabItem->iCursor, iDb, pTab, op); 5234 assert( pTabItem->iCursor==pLevel->iTabCur ); 5235 testcase( pWInfo->eOnePass==ONEPASS_OFF && pTab->nCol==BMS-1 ); 5236 testcase( pWInfo->eOnePass==ONEPASS_OFF && pTab->nCol==BMS ); 5237 if( pWInfo->eOnePass==ONEPASS_OFF 5238 && pTab->nCol<BMS 5239 && (pTab->tabFlags & (TF_HasGenerated|TF_WithoutRowid))==0 5240 ){ 5241 /* If we know that only a prefix of the record will be used, 5242 ** it is advantageous to reduce the "column count" field in 5243 ** the P4 operand of the OP_OpenRead/Write opcode. */ 5244 Bitmask b = pTabItem->colUsed; 5245 int n = 0; 5246 for(; b; b=b>>1, n++){} 5247 sqlite3VdbeChangeP4(v, -1, SQLITE_INT_TO_PTR(n), P4_INT32); 5248 assert( n<=pTab->nCol ); 5249 } 5250 #ifdef SQLITE_ENABLE_CURSOR_HINTS 5251 if( pLoop->u.btree.pIndex!=0 ){ 5252 sqlite3VdbeChangeP5(v, OPFLAG_SEEKEQ|bFordelete); 5253 }else 5254 #endif 5255 { 5256 sqlite3VdbeChangeP5(v, bFordelete); 5257 } 5258 #ifdef SQLITE_ENABLE_COLUMN_USED_MASK 5259 sqlite3VdbeAddOp4Dup8(v, OP_ColumnsUsed, pTabItem->iCursor, 0, 0, 5260 (const u8*)&pTabItem->colUsed, P4_INT64); 5261 #endif 5262 }else{ 5263 sqlite3TableLock(pParse, iDb, pTab->tnum, 0, pTab->zName); 5264 } 5265 if( pLoop->wsFlags & WHERE_INDEXED ){ 5266 Index *pIx = pLoop->u.btree.pIndex; 5267 int iIndexCur; 5268 int op = OP_OpenRead; 5269 /* iAuxArg is always set to a positive value if ONEPASS is possible */ 5270 assert( iAuxArg!=0 || (pWInfo->wctrlFlags & WHERE_ONEPASS_DESIRED)==0 ); 5271 if( !HasRowid(pTab) && IsPrimaryKeyIndex(pIx) 5272 && (wctrlFlags & WHERE_OR_SUBCLAUSE)!=0 5273 ){ 5274 /* This is one term of an OR-optimization using the PRIMARY KEY of a 5275 ** WITHOUT ROWID table. No need for a separate index */ 5276 iIndexCur = pLevel->iTabCur; 5277 op = 0; 5278 }else if( pWInfo->eOnePass!=ONEPASS_OFF ){ 5279 Index *pJ = pTabItem->pTab->pIndex; 5280 iIndexCur = iAuxArg; 5281 assert( wctrlFlags & WHERE_ONEPASS_DESIRED ); 5282 while( ALWAYS(pJ) && pJ!=pIx ){ 5283 iIndexCur++; 5284 pJ = pJ->pNext; 5285 } 5286 op = OP_OpenWrite; 5287 pWInfo->aiCurOnePass[1] = iIndexCur; 5288 }else if( iAuxArg && (wctrlFlags & WHERE_OR_SUBCLAUSE)!=0 ){ 5289 iIndexCur = iAuxArg; 5290 op = OP_ReopenIdx; 5291 }else{ 5292 iIndexCur = pParse->nTab++; 5293 } 5294 pLevel->iIdxCur = iIndexCur; 5295 assert( pIx->pSchema==pTab->pSchema ); 5296 assert( iIndexCur>=0 ); 5297 if( op ){ 5298 sqlite3VdbeAddOp3(v, op, iIndexCur, pIx->tnum, iDb); 5299 sqlite3VdbeSetP4KeyInfo(pParse, pIx); 5300 if( (pLoop->wsFlags & WHERE_CONSTRAINT)!=0 5301 && (pLoop->wsFlags & (WHERE_COLUMN_RANGE|WHERE_SKIPSCAN))==0 5302 && (pLoop->wsFlags & WHERE_BIGNULL_SORT)==0 5303 && (pLoop->wsFlags & WHERE_IN_SEEKSCAN)==0 5304 && (pWInfo->wctrlFlags&WHERE_ORDERBY_MIN)==0 5305 && pWInfo->eDistinct!=WHERE_DISTINCT_ORDERED 5306 ){ 5307 sqlite3VdbeChangeP5(v, OPFLAG_SEEKEQ); 5308 } 5309 VdbeComment((v, "%s", pIx->zName)); 5310 #ifdef SQLITE_ENABLE_COLUMN_USED_MASK 5311 { 5312 u64 colUsed = 0; 5313 int ii, jj; 5314 for(ii=0; ii<pIx->nColumn; ii++){ 5315 jj = pIx->aiColumn[ii]; 5316 if( jj<0 ) continue; 5317 if( jj>63 ) jj = 63; 5318 if( (pTabItem->colUsed & MASKBIT(jj))==0 ) continue; 5319 colUsed |= ((u64)1)<<(ii<63 ? ii : 63); 5320 } 5321 sqlite3VdbeAddOp4Dup8(v, OP_ColumnsUsed, iIndexCur, 0, 0, 5322 (u8*)&colUsed, P4_INT64); 5323 } 5324 #endif /* SQLITE_ENABLE_COLUMN_USED_MASK */ 5325 } 5326 } 5327 if( iDb>=0 ) sqlite3CodeVerifySchema(pParse, iDb); 5328 } 5329 pWInfo->iTop = sqlite3VdbeCurrentAddr(v); 5330 if( db->mallocFailed ) goto whereBeginError; 5331 5332 /* Generate the code to do the search. Each iteration of the for 5333 ** loop below generates code for a single nested loop of the VM 5334 ** program. 5335 */ 5336 for(ii=0; ii<nTabList; ii++){ 5337 int addrExplain; 5338 int wsFlags; 5339 pLevel = &pWInfo->a[ii]; 5340 wsFlags = pLevel->pWLoop->wsFlags; 5341 #ifndef SQLITE_OMIT_AUTOMATIC_INDEX 5342 if( (pLevel->pWLoop->wsFlags & WHERE_AUTO_INDEX)!=0 ){ 5343 constructAutomaticIndex(pParse, &pWInfo->sWC, 5344 &pTabList->a[pLevel->iFrom], notReady, pLevel); 5345 if( db->mallocFailed ) goto whereBeginError; 5346 } 5347 #endif 5348 addrExplain = sqlite3WhereExplainOneScan( 5349 pParse, pTabList, pLevel, wctrlFlags 5350 ); 5351 pLevel->addrBody = sqlite3VdbeCurrentAddr(v); 5352 notReady = sqlite3WhereCodeOneLoopStart(pParse,v,pWInfo,ii,pLevel,notReady); 5353 pWInfo->iContinue = pLevel->addrCont; 5354 if( (wsFlags&WHERE_MULTI_OR)==0 && (wctrlFlags&WHERE_OR_SUBCLAUSE)==0 ){ 5355 sqlite3WhereAddScanStatus(v, pTabList, pLevel, addrExplain); 5356 } 5357 } 5358 5359 /* Done. */ 5360 VdbeModuleComment((v, "Begin WHERE-core")); 5361 pWInfo->iEndWhere = sqlite3VdbeCurrentAddr(v); 5362 return pWInfo; 5363 5364 /* Jump here if malloc fails */ 5365 whereBeginError: 5366 if( pWInfo ){ 5367 testcase( pWInfo->pExprMods!=0 ); 5368 whereUndoExprMods(pWInfo); 5369 pParse->nQueryLoop = pWInfo->savedNQueryLoop; 5370 whereInfoFree(db, pWInfo); 5371 } 5372 return 0; 5373 } 5374 5375 /* 5376 ** Part of sqlite3WhereEnd() will rewrite opcodes to reference the 5377 ** index rather than the main table. In SQLITE_DEBUG mode, we want 5378 ** to trace those changes if PRAGMA vdbe_addoptrace=on. This routine 5379 ** does that. 5380 */ 5381 #ifndef SQLITE_DEBUG 5382 # define OpcodeRewriteTrace(D,K,P) /* no-op */ 5383 #else 5384 # define OpcodeRewriteTrace(D,K,P) sqlite3WhereOpcodeRewriteTrace(D,K,P) 5385 static void sqlite3WhereOpcodeRewriteTrace( 5386 sqlite3 *db, 5387 int pc, 5388 VdbeOp *pOp 5389 ){ 5390 if( (db->flags & SQLITE_VdbeAddopTrace)==0 ) return; 5391 sqlite3VdbePrintOp(0, pc, pOp); 5392 } 5393 #endif 5394 5395 /* 5396 ** Generate the end of the WHERE loop. See comments on 5397 ** sqlite3WhereBegin() for additional information. 5398 */ 5399 void sqlite3WhereEnd(WhereInfo *pWInfo){ 5400 Parse *pParse = pWInfo->pParse; 5401 Vdbe *v = pParse->pVdbe; 5402 int i; 5403 WhereLevel *pLevel; 5404 WhereLoop *pLoop; 5405 SrcList *pTabList = pWInfo->pTabList; 5406 sqlite3 *db = pParse->db; 5407 int iEnd = sqlite3VdbeCurrentAddr(v); 5408 5409 /* Generate loop termination code. 5410 */ 5411 VdbeModuleComment((v, "End WHERE-core")); 5412 for(i=pWInfo->nLevel-1; i>=0; i--){ 5413 int addr; 5414 pLevel = &pWInfo->a[i]; 5415 pLoop = pLevel->pWLoop; 5416 if( pLevel->op!=OP_Noop ){ 5417 #ifndef SQLITE_DISABLE_SKIPAHEAD_DISTINCT 5418 int addrSeek = 0; 5419 Index *pIdx; 5420 int n; 5421 if( pWInfo->eDistinct==WHERE_DISTINCT_ORDERED 5422 && i==pWInfo->nLevel-1 /* Ticket [ef9318757b152e3] 2017-10-21 */ 5423 && (pLoop->wsFlags & WHERE_INDEXED)!=0 5424 && (pIdx = pLoop->u.btree.pIndex)->hasStat1 5425 && (n = pLoop->u.btree.nDistinctCol)>0 5426 && pIdx->aiRowLogEst[n]>=36 5427 ){ 5428 int r1 = pParse->nMem+1; 5429 int j, op; 5430 for(j=0; j<n; j++){ 5431 sqlite3VdbeAddOp3(v, OP_Column, pLevel->iIdxCur, j, r1+j); 5432 } 5433 pParse->nMem += n+1; 5434 op = pLevel->op==OP_Prev ? OP_SeekLT : OP_SeekGT; 5435 addrSeek = sqlite3VdbeAddOp4Int(v, op, pLevel->iIdxCur, 0, r1, n); 5436 VdbeCoverageIf(v, op==OP_SeekLT); 5437 VdbeCoverageIf(v, op==OP_SeekGT); 5438 sqlite3VdbeAddOp2(v, OP_Goto, 1, pLevel->p2); 5439 } 5440 #endif /* SQLITE_DISABLE_SKIPAHEAD_DISTINCT */ 5441 /* The common case: Advance to the next row */ 5442 sqlite3VdbeResolveLabel(v, pLevel->addrCont); 5443 sqlite3VdbeAddOp3(v, pLevel->op, pLevel->p1, pLevel->p2, pLevel->p3); 5444 sqlite3VdbeChangeP5(v, pLevel->p5); 5445 VdbeCoverage(v); 5446 VdbeCoverageIf(v, pLevel->op==OP_Next); 5447 VdbeCoverageIf(v, pLevel->op==OP_Prev); 5448 VdbeCoverageIf(v, pLevel->op==OP_VNext); 5449 if( pLevel->regBignull ){ 5450 sqlite3VdbeResolveLabel(v, pLevel->addrBignull); 5451 sqlite3VdbeAddOp2(v, OP_DecrJumpZero, pLevel->regBignull, pLevel->p2-1); 5452 VdbeCoverage(v); 5453 } 5454 #ifndef SQLITE_DISABLE_SKIPAHEAD_DISTINCT 5455 if( addrSeek ) sqlite3VdbeJumpHere(v, addrSeek); 5456 #endif 5457 }else{ 5458 sqlite3VdbeResolveLabel(v, pLevel->addrCont); 5459 } 5460 if( pLoop->wsFlags & WHERE_IN_ABLE && pLevel->u.in.nIn>0 ){ 5461 struct InLoop *pIn; 5462 int j; 5463 sqlite3VdbeResolveLabel(v, pLevel->addrNxt); 5464 for(j=pLevel->u.in.nIn, pIn=&pLevel->u.in.aInLoop[j-1]; j>0; j--, pIn--){ 5465 assert( sqlite3VdbeGetOp(v, pIn->addrInTop+1)->opcode==OP_IsNull 5466 || pParse->db->mallocFailed ); 5467 sqlite3VdbeJumpHere(v, pIn->addrInTop+1); 5468 if( pIn->eEndLoopOp!=OP_Noop ){ 5469 if( pIn->nPrefix ){ 5470 int bEarlyOut = 5471 (pLoop->wsFlags & WHERE_VIRTUALTABLE)==0 5472 && (pLoop->wsFlags & WHERE_IN_EARLYOUT)!=0; 5473 if( pLevel->iLeftJoin ){ 5474 /* For LEFT JOIN queries, cursor pIn->iCur may not have been 5475 ** opened yet. This occurs for WHERE clauses such as 5476 ** "a = ? AND b IN (...)", where the index is on (a, b). If 5477 ** the RHS of the (a=?) is NULL, then the "b IN (...)" may 5478 ** never have been coded, but the body of the loop run to 5479 ** return the null-row. So, if the cursor is not open yet, 5480 ** jump over the OP_Next or OP_Prev instruction about to 5481 ** be coded. */ 5482 sqlite3VdbeAddOp2(v, OP_IfNotOpen, pIn->iCur, 5483 sqlite3VdbeCurrentAddr(v) + 2 + bEarlyOut); 5484 VdbeCoverage(v); 5485 } 5486 if( bEarlyOut ){ 5487 sqlite3VdbeAddOp4Int(v, OP_IfNoHope, pLevel->iIdxCur, 5488 sqlite3VdbeCurrentAddr(v)+2, 5489 pIn->iBase, pIn->nPrefix); 5490 VdbeCoverage(v); 5491 /* Retarget the OP_IsNull against the left operand of IN so 5492 ** it jumps past the OP_IfNoHope. This is because the 5493 ** OP_IsNull also bypasses the OP_Affinity opcode that is 5494 ** required by OP_IfNoHope. */ 5495 sqlite3VdbeJumpHere(v, pIn->addrInTop+1); 5496 } 5497 } 5498 sqlite3VdbeAddOp2(v, pIn->eEndLoopOp, pIn->iCur, pIn->addrInTop); 5499 VdbeCoverage(v); 5500 VdbeCoverageIf(v, pIn->eEndLoopOp==OP_Prev); 5501 VdbeCoverageIf(v, pIn->eEndLoopOp==OP_Next); 5502 } 5503 sqlite3VdbeJumpHere(v, pIn->addrInTop-1); 5504 } 5505 } 5506 sqlite3VdbeResolveLabel(v, pLevel->addrBrk); 5507 if( pLevel->addrSkip ){ 5508 sqlite3VdbeGoto(v, pLevel->addrSkip); 5509 VdbeComment((v, "next skip-scan on %s", pLoop->u.btree.pIndex->zName)); 5510 sqlite3VdbeJumpHere(v, pLevel->addrSkip); 5511 sqlite3VdbeJumpHere(v, pLevel->addrSkip-2); 5512 } 5513 #ifndef SQLITE_LIKE_DOESNT_MATCH_BLOBS 5514 if( pLevel->addrLikeRep ){ 5515 sqlite3VdbeAddOp2(v, OP_DecrJumpZero, (int)(pLevel->iLikeRepCntr>>1), 5516 pLevel->addrLikeRep); 5517 VdbeCoverage(v); 5518 } 5519 #endif 5520 if( pLevel->iLeftJoin ){ 5521 int ws = pLoop->wsFlags; 5522 addr = sqlite3VdbeAddOp1(v, OP_IfPos, pLevel->iLeftJoin); VdbeCoverage(v); 5523 assert( (ws & WHERE_IDX_ONLY)==0 || (ws & WHERE_INDEXED)!=0 ); 5524 if( (ws & WHERE_IDX_ONLY)==0 ){ 5525 assert( pLevel->iTabCur==pTabList->a[pLevel->iFrom].iCursor ); 5526 sqlite3VdbeAddOp1(v, OP_NullRow, pLevel->iTabCur); 5527 } 5528 if( (ws & WHERE_INDEXED) 5529 || ((ws & WHERE_MULTI_OR) && pLevel->u.pCovidx) 5530 ){ 5531 sqlite3VdbeAddOp1(v, OP_NullRow, pLevel->iIdxCur); 5532 } 5533 if( pLevel->op==OP_Return ){ 5534 sqlite3VdbeAddOp2(v, OP_Gosub, pLevel->p1, pLevel->addrFirst); 5535 }else{ 5536 sqlite3VdbeGoto(v, pLevel->addrFirst); 5537 } 5538 sqlite3VdbeJumpHere(v, addr); 5539 } 5540 VdbeModuleComment((v, "End WHERE-loop%d: %s", i, 5541 pWInfo->pTabList->a[pLevel->iFrom].pTab->zName)); 5542 } 5543 5544 /* The "break" point is here, just past the end of the outer loop. 5545 ** Set it. 5546 */ 5547 sqlite3VdbeResolveLabel(v, pWInfo->iBreak); 5548 5549 assert( pWInfo->nLevel<=pTabList->nSrc ); 5550 for(i=0, pLevel=pWInfo->a; i<pWInfo->nLevel; i++, pLevel++){ 5551 int k, last; 5552 VdbeOp *pOp, *pLastOp; 5553 Index *pIdx = 0; 5554 SrcItem *pTabItem = &pTabList->a[pLevel->iFrom]; 5555 Table *pTab = pTabItem->pTab; 5556 assert( pTab!=0 ); 5557 pLoop = pLevel->pWLoop; 5558 5559 /* For a co-routine, change all OP_Column references to the table of 5560 ** the co-routine into OP_Copy of result contained in a register. 5561 ** OP_Rowid becomes OP_Null. 5562 */ 5563 if( pTabItem->fg.viaCoroutine ){ 5564 testcase( pParse->db->mallocFailed ); 5565 translateColumnToCopy(pParse, pLevel->addrBody, pLevel->iTabCur, 5566 pTabItem->regResult, 0); 5567 continue; 5568 } 5569 5570 #ifdef SQLITE_ENABLE_EARLY_CURSOR_CLOSE 5571 /* Close all of the cursors that were opened by sqlite3WhereBegin. 5572 ** Except, do not close cursors that will be reused by the OR optimization 5573 ** (WHERE_OR_SUBCLAUSE). And do not close the OP_OpenWrite cursors 5574 ** created for the ONEPASS optimization. 5575 */ 5576 if( (pTab->tabFlags & TF_Ephemeral)==0 5577 && pTab->pSelect==0 5578 && (pWInfo->wctrlFlags & WHERE_OR_SUBCLAUSE)==0 5579 ){ 5580 int ws = pLoop->wsFlags; 5581 if( pWInfo->eOnePass==ONEPASS_OFF && (ws & WHERE_IDX_ONLY)==0 ){ 5582 sqlite3VdbeAddOp1(v, OP_Close, pTabItem->iCursor); 5583 } 5584 if( (ws & WHERE_INDEXED)!=0 5585 && (ws & (WHERE_IPK|WHERE_AUTO_INDEX))==0 5586 && pLevel->iIdxCur!=pWInfo->aiCurOnePass[1] 5587 ){ 5588 sqlite3VdbeAddOp1(v, OP_Close, pLevel->iIdxCur); 5589 } 5590 } 5591 #endif 5592 5593 /* If this scan uses an index, make VDBE code substitutions to read data 5594 ** from the index instead of from the table where possible. In some cases 5595 ** this optimization prevents the table from ever being read, which can 5596 ** yield a significant performance boost. 5597 ** 5598 ** Calls to the code generator in between sqlite3WhereBegin and 5599 ** sqlite3WhereEnd will have created code that references the table 5600 ** directly. This loop scans all that code looking for opcodes 5601 ** that reference the table and converts them into opcodes that 5602 ** reference the index. 5603 */ 5604 if( pLoop->wsFlags & (WHERE_INDEXED|WHERE_IDX_ONLY) ){ 5605 pIdx = pLoop->u.btree.pIndex; 5606 }else if( pLoop->wsFlags & WHERE_MULTI_OR ){ 5607 pIdx = pLevel->u.pCovidx; 5608 } 5609 if( pIdx 5610 && !db->mallocFailed 5611 ){ 5612 if( pWInfo->eOnePass==ONEPASS_OFF || !HasRowid(pIdx->pTable) ){ 5613 last = iEnd; 5614 }else{ 5615 last = pWInfo->iEndWhere; 5616 } 5617 k = pLevel->addrBody + 1; 5618 #ifdef SQLITE_DEBUG 5619 if( db->flags & SQLITE_VdbeAddopTrace ){ 5620 printf("TRANSLATE opcodes in range %d..%d\n", k, last-1); 5621 } 5622 /* Proof that the "+1" on the k value above is safe */ 5623 pOp = sqlite3VdbeGetOp(v, k - 1); 5624 assert( pOp->opcode!=OP_Column || pOp->p1!=pLevel->iTabCur ); 5625 assert( pOp->opcode!=OP_Rowid || pOp->p1!=pLevel->iTabCur ); 5626 assert( pOp->opcode!=OP_IfNullRow || pOp->p1!=pLevel->iTabCur ); 5627 #endif 5628 pOp = sqlite3VdbeGetOp(v, k); 5629 pLastOp = pOp + (last - k); 5630 assert( pOp<=pLastOp ); 5631 do{ 5632 if( pOp->p1!=pLevel->iTabCur ){ 5633 /* no-op */ 5634 }else if( pOp->opcode==OP_Column 5635 #ifdef SQLITE_ENABLE_OFFSET_SQL_FUNC 5636 || pOp->opcode==OP_Offset 5637 #endif 5638 ){ 5639 int x = pOp->p2; 5640 assert( pIdx->pTable==pTab ); 5641 if( !HasRowid(pTab) ){ 5642 Index *pPk = sqlite3PrimaryKeyIndex(pTab); 5643 x = pPk->aiColumn[x]; 5644 assert( x>=0 ); 5645 }else{ 5646 testcase( x!=sqlite3StorageColumnToTable(pTab,x) ); 5647 x = sqlite3StorageColumnToTable(pTab,x); 5648 } 5649 x = sqlite3TableColumnToIndex(pIdx, x); 5650 if( x>=0 ){ 5651 pOp->p2 = x; 5652 pOp->p1 = pLevel->iIdxCur; 5653 OpcodeRewriteTrace(db, k, pOp); 5654 } 5655 assert( (pLoop->wsFlags & WHERE_IDX_ONLY)==0 || x>=0 5656 || pWInfo->eOnePass ); 5657 }else if( pOp->opcode==OP_Rowid ){ 5658 pOp->p1 = pLevel->iIdxCur; 5659 pOp->opcode = OP_IdxRowid; 5660 OpcodeRewriteTrace(db, k, pOp); 5661 }else if( pOp->opcode==OP_IfNullRow ){ 5662 pOp->p1 = pLevel->iIdxCur; 5663 OpcodeRewriteTrace(db, k, pOp); 5664 } 5665 #ifdef SQLITE_DEBUG 5666 k++; 5667 #endif 5668 }while( (++pOp)<pLastOp ); 5669 #ifdef SQLITE_DEBUG 5670 if( db->flags & SQLITE_VdbeAddopTrace ) printf("TRANSLATE complete\n"); 5671 #endif 5672 } 5673 } 5674 5675 /* Final cleanup 5676 */ 5677 if( pWInfo->pExprMods ) whereUndoExprMods(pWInfo); 5678 pParse->nQueryLoop = pWInfo->savedNQueryLoop; 5679 whereInfoFree(db, pWInfo); 5680 return; 5681 } 5682