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 assert( iCur>=0 ); 290 do{ 291 for(pTerm=pWC->a+k; k<pWC->nTerm; k++, pTerm++){ 292 assert( (pTerm->eOperator & (WO_OR|WO_AND))==0 || pTerm->leftCursor<0 ); 293 if( pTerm->leftCursor==iCur 294 && pTerm->u.x.leftColumn==iColumn 295 && (iColumn!=XN_EXPR 296 || sqlite3ExprCompareSkip(pTerm->pExpr->pLeft, 297 pScan->pIdxExpr,iCur)==0) 298 && (pScan->iEquiv<=1 || !ExprHasProperty(pTerm->pExpr, EP_FromJoin)) 299 ){ 300 if( (pTerm->eOperator & WO_EQUIV)!=0 301 && pScan->nEquiv<ArraySize(pScan->aiCur) 302 && (pX = whereRightSubexprIsColumn(pTerm->pExpr))!=0 303 ){ 304 int j; 305 for(j=0; j<pScan->nEquiv; j++){ 306 if( pScan->aiCur[j]==pX->iTable 307 && pScan->aiColumn[j]==pX->iColumn ){ 308 break; 309 } 310 } 311 if( j==pScan->nEquiv ){ 312 pScan->aiCur[j] = pX->iTable; 313 pScan->aiColumn[j] = pX->iColumn; 314 pScan->nEquiv++; 315 } 316 } 317 if( (pTerm->eOperator & pScan->opMask)!=0 ){ 318 /* Verify the affinity and collating sequence match */ 319 if( pScan->zCollName && (pTerm->eOperator & WO_ISNULL)==0 ){ 320 CollSeq *pColl; 321 Parse *pParse = pWC->pWInfo->pParse; 322 pX = pTerm->pExpr; 323 if( !sqlite3IndexAffinityOk(pX, pScan->idxaff) ){ 324 continue; 325 } 326 assert(pX->pLeft); 327 pColl = sqlite3ExprCompareCollSeq(pParse, pX); 328 if( pColl==0 ) pColl = pParse->db->pDfltColl; 329 if( sqlite3StrICmp(pColl->zName, pScan->zCollName) ){ 330 continue; 331 } 332 } 333 if( (pTerm->eOperator & (WO_EQ|WO_IS))!=0 334 && (pX = pTerm->pExpr->pRight, ALWAYS(pX!=0)) 335 && pX->op==TK_COLUMN 336 && pX->iTable==pScan->aiCur[0] 337 && pX->iColumn==pScan->aiColumn[0] 338 ){ 339 testcase( pTerm->eOperator & WO_IS ); 340 continue; 341 } 342 pScan->pWC = pWC; 343 pScan->k = k+1; 344 #ifdef WHERETRACE_ENABLED 345 if( sqlite3WhereTrace & 0x20000 ){ 346 int ii; 347 sqlite3DebugPrintf("SCAN-TERM %p: nEquiv=%d", 348 pTerm, pScan->nEquiv); 349 for(ii=0; ii<pScan->nEquiv; ii++){ 350 sqlite3DebugPrintf(" {%d:%d}", 351 pScan->aiCur[ii], pScan->aiColumn[ii]); 352 } 353 sqlite3DebugPrintf("\n"); 354 } 355 #endif 356 return pTerm; 357 } 358 } 359 } 360 pWC = pWC->pOuter; 361 k = 0; 362 }while( pWC!=0 ); 363 if( pScan->iEquiv>=pScan->nEquiv ) break; 364 pWC = pScan->pOrigWC; 365 k = 0; 366 pScan->iEquiv++; 367 } 368 return 0; 369 } 370 371 /* 372 ** This is whereScanInit() for the case of an index on an expression. 373 ** It is factored out into a separate tail-recursion subroutine so that 374 ** the normal whereScanInit() routine, which is a high-runner, does not 375 ** need to push registers onto the stack as part of its prologue. 376 */ 377 static SQLITE_NOINLINE WhereTerm *whereScanInitIndexExpr(WhereScan *pScan){ 378 pScan->idxaff = sqlite3ExprAffinity(pScan->pIdxExpr); 379 return whereScanNext(pScan); 380 } 381 382 /* 383 ** Initialize a WHERE clause scanner object. Return a pointer to the 384 ** first match. Return NULL if there are no matches. 385 ** 386 ** The scanner will be searching the WHERE clause pWC. It will look 387 ** for terms of the form "X <op> <expr>" where X is column iColumn of table 388 ** iCur. Or if pIdx!=0 then X is column iColumn of index pIdx. pIdx 389 ** must be one of the indexes of table iCur. 390 ** 391 ** The <op> must be one of the operators described by opMask. 392 ** 393 ** If the search is for X and the WHERE clause contains terms of the 394 ** form X=Y then this routine might also return terms of the form 395 ** "Y <op> <expr>". The number of levels of transitivity is limited, 396 ** but is enough to handle most commonly occurring SQL statements. 397 ** 398 ** If X is not the INTEGER PRIMARY KEY then X must be compatible with 399 ** index pIdx. 400 */ 401 static WhereTerm *whereScanInit( 402 WhereScan *pScan, /* The WhereScan object being initialized */ 403 WhereClause *pWC, /* The WHERE clause to be scanned */ 404 int iCur, /* Cursor to scan for */ 405 int iColumn, /* Column to scan for */ 406 u32 opMask, /* Operator(s) to scan for */ 407 Index *pIdx /* Must be compatible with this index */ 408 ){ 409 pScan->pOrigWC = pWC; 410 pScan->pWC = pWC; 411 pScan->pIdxExpr = 0; 412 pScan->idxaff = 0; 413 pScan->zCollName = 0; 414 pScan->opMask = opMask; 415 pScan->k = 0; 416 pScan->aiCur[0] = iCur; 417 pScan->nEquiv = 1; 418 pScan->iEquiv = 1; 419 if( pIdx ){ 420 int j = iColumn; 421 iColumn = pIdx->aiColumn[j]; 422 if( iColumn==XN_EXPR ){ 423 pScan->pIdxExpr = pIdx->aColExpr->a[j].pExpr; 424 pScan->zCollName = pIdx->azColl[j]; 425 pScan->aiColumn[0] = XN_EXPR; 426 return whereScanInitIndexExpr(pScan); 427 }else if( iColumn==pIdx->pTable->iPKey ){ 428 iColumn = XN_ROWID; 429 }else if( iColumn>=0 ){ 430 pScan->idxaff = pIdx->pTable->aCol[iColumn].affinity; 431 pScan->zCollName = pIdx->azColl[j]; 432 } 433 }else if( iColumn==XN_EXPR ){ 434 return 0; 435 } 436 pScan->aiColumn[0] = iColumn; 437 return whereScanNext(pScan); 438 } 439 440 /* 441 ** Search for a term in the WHERE clause that is of the form "X <op> <expr>" 442 ** where X is a reference to the iColumn of table iCur or of index pIdx 443 ** if pIdx!=0 and <op> is one of the WO_xx operator codes specified by 444 ** the op parameter. Return a pointer to the term. Return 0 if not found. 445 ** 446 ** If pIdx!=0 then it must be one of the indexes of table iCur. 447 ** Search for terms matching the iColumn-th column of pIdx 448 ** rather than the iColumn-th column of table iCur. 449 ** 450 ** The term returned might by Y=<expr> if there is another constraint in 451 ** the WHERE clause that specifies that X=Y. Any such constraints will be 452 ** identified by the WO_EQUIV bit in the pTerm->eOperator field. The 453 ** aiCur[]/iaColumn[] arrays hold X and all its equivalents. There are 11 454 ** slots in aiCur[]/aiColumn[] so that means we can look for X plus up to 10 455 ** other equivalent values. Hence a search for X will return <expr> if X=A1 456 ** and A1=A2 and A2=A3 and ... and A9=A10 and A10=<expr>. 457 ** 458 ** If there are multiple terms in the WHERE clause of the form "X <op> <expr>" 459 ** then try for the one with no dependencies on <expr> - in other words where 460 ** <expr> is a constant expression of some kind. Only return entries of 461 ** the form "X <op> Y" where Y is a column in another table if no terms of 462 ** the form "X <op> <const-expr>" exist. If no terms with a constant RHS 463 ** exist, try to return a term that does not use WO_EQUIV. 464 */ 465 WhereTerm *sqlite3WhereFindTerm( 466 WhereClause *pWC, /* The WHERE clause to be searched */ 467 int iCur, /* Cursor number of LHS */ 468 int iColumn, /* Column number of LHS */ 469 Bitmask notReady, /* RHS must not overlap with this mask */ 470 u32 op, /* Mask of WO_xx values describing operator */ 471 Index *pIdx /* Must be compatible with this index, if not NULL */ 472 ){ 473 WhereTerm *pResult = 0; 474 WhereTerm *p; 475 WhereScan scan; 476 477 p = whereScanInit(&scan, pWC, iCur, iColumn, op, pIdx); 478 op &= WO_EQ|WO_IS; 479 while( p ){ 480 if( (p->prereqRight & notReady)==0 ){ 481 if( p->prereqRight==0 && (p->eOperator&op)!=0 ){ 482 testcase( p->eOperator & WO_IS ); 483 return p; 484 } 485 if( pResult==0 ) pResult = p; 486 } 487 p = whereScanNext(&scan); 488 } 489 return pResult; 490 } 491 492 /* 493 ** This function searches pList for an entry that matches the iCol-th column 494 ** of index pIdx. 495 ** 496 ** If such an expression is found, its index in pList->a[] is returned. If 497 ** no expression is found, -1 is returned. 498 */ 499 static int findIndexCol( 500 Parse *pParse, /* Parse context */ 501 ExprList *pList, /* Expression list to search */ 502 int iBase, /* Cursor for table associated with pIdx */ 503 Index *pIdx, /* Index to match column of */ 504 int iCol /* Column of index to match */ 505 ){ 506 int i; 507 const char *zColl = pIdx->azColl[iCol]; 508 509 for(i=0; i<pList->nExpr; i++){ 510 Expr *p = sqlite3ExprSkipCollateAndLikely(pList->a[i].pExpr); 511 if( ALWAYS(p!=0) 512 && (p->op==TK_COLUMN || p->op==TK_AGG_COLUMN) 513 && p->iColumn==pIdx->aiColumn[iCol] 514 && p->iTable==iBase 515 ){ 516 CollSeq *pColl = sqlite3ExprNNCollSeq(pParse, pList->a[i].pExpr); 517 if( 0==sqlite3StrICmp(pColl->zName, zColl) ){ 518 return i; 519 } 520 } 521 } 522 523 return -1; 524 } 525 526 /* 527 ** Return TRUE if the iCol-th column of index pIdx is NOT NULL 528 */ 529 static int indexColumnNotNull(Index *pIdx, int iCol){ 530 int j; 531 assert( pIdx!=0 ); 532 assert( iCol>=0 && iCol<pIdx->nColumn ); 533 j = pIdx->aiColumn[iCol]; 534 if( j>=0 ){ 535 return pIdx->pTable->aCol[j].notNull; 536 }else if( j==(-1) ){ 537 return 1; 538 }else{ 539 assert( j==(-2) ); 540 return 0; /* Assume an indexed expression can always yield a NULL */ 541 542 } 543 } 544 545 /* 546 ** Return true if the DISTINCT expression-list passed as the third argument 547 ** is redundant. 548 ** 549 ** A DISTINCT list is redundant if any subset of the columns in the 550 ** DISTINCT list are collectively unique and individually non-null. 551 */ 552 static int isDistinctRedundant( 553 Parse *pParse, /* Parsing context */ 554 SrcList *pTabList, /* The FROM clause */ 555 WhereClause *pWC, /* The WHERE clause */ 556 ExprList *pDistinct /* The result set that needs to be DISTINCT */ 557 ){ 558 Table *pTab; 559 Index *pIdx; 560 int i; 561 int iBase; 562 563 /* If there is more than one table or sub-select in the FROM clause of 564 ** this query, then it will not be possible to show that the DISTINCT 565 ** clause is redundant. */ 566 if( pTabList->nSrc!=1 ) return 0; 567 iBase = pTabList->a[0].iCursor; 568 pTab = pTabList->a[0].pTab; 569 570 /* If any of the expressions is an IPK column on table iBase, then return 571 ** true. Note: The (p->iTable==iBase) part of this test may be false if the 572 ** current SELECT is a correlated sub-query. 573 */ 574 for(i=0; i<pDistinct->nExpr; i++){ 575 Expr *p = sqlite3ExprSkipCollateAndLikely(pDistinct->a[i].pExpr); 576 if( NEVER(p==0) ) continue; 577 if( p->op!=TK_COLUMN && p->op!=TK_AGG_COLUMN ) continue; 578 if( p->iTable==iBase && p->iColumn<0 ) return 1; 579 } 580 581 /* Loop through all indices on the table, checking each to see if it makes 582 ** the DISTINCT qualifier redundant. It does so if: 583 ** 584 ** 1. The index is itself UNIQUE, and 585 ** 586 ** 2. All of the columns in the index are either part of the pDistinct 587 ** list, or else the WHERE clause contains a term of the form "col=X", 588 ** where X is a constant value. The collation sequences of the 589 ** comparison and select-list expressions must match those of the index. 590 ** 591 ** 3. All of those index columns for which the WHERE clause does not 592 ** contain a "col=X" term are subject to a NOT NULL constraint. 593 */ 594 for(pIdx=pTab->pIndex; pIdx; pIdx=pIdx->pNext){ 595 if( !IsUniqueIndex(pIdx) ) continue; 596 if( pIdx->pPartIdxWhere ) continue; 597 for(i=0; i<pIdx->nKeyCol; i++){ 598 if( 0==sqlite3WhereFindTerm(pWC, iBase, i, ~(Bitmask)0, WO_EQ, pIdx) ){ 599 if( findIndexCol(pParse, pDistinct, iBase, pIdx, i)<0 ) break; 600 if( indexColumnNotNull(pIdx, i)==0 ) break; 601 } 602 } 603 if( i==pIdx->nKeyCol ){ 604 /* This index implies that the DISTINCT qualifier is redundant. */ 605 return 1; 606 } 607 } 608 609 return 0; 610 } 611 612 613 /* 614 ** Estimate the logarithm of the input value to base 2. 615 */ 616 static LogEst estLog(LogEst N){ 617 return N<=10 ? 0 : sqlite3LogEst(N) - 33; 618 } 619 620 /* 621 ** Convert OP_Column opcodes to OP_Copy in previously generated code. 622 ** 623 ** This routine runs over generated VDBE code and translates OP_Column 624 ** opcodes into OP_Copy when the table is being accessed via co-routine 625 ** instead of via table lookup. 626 ** 627 ** If the iAutoidxCur is not zero, then any OP_Rowid instructions on 628 ** cursor iTabCur are transformed into OP_Sequence opcode for the 629 ** iAutoidxCur cursor, in order to generate unique rowids for the 630 ** automatic index being generated. 631 */ 632 static void translateColumnToCopy( 633 Parse *pParse, /* Parsing context */ 634 int iStart, /* Translate from this opcode to the end */ 635 int iTabCur, /* OP_Column/OP_Rowid references to this table */ 636 int iRegister, /* The first column is in this register */ 637 int iAutoidxCur /* If non-zero, cursor of autoindex being generated */ 638 ){ 639 Vdbe *v = pParse->pVdbe; 640 VdbeOp *pOp = sqlite3VdbeGetOp(v, iStart); 641 int iEnd = sqlite3VdbeCurrentAddr(v); 642 if( pParse->db->mallocFailed ) return; 643 for(; iStart<iEnd; iStart++, pOp++){ 644 if( pOp->p1!=iTabCur ) continue; 645 if( pOp->opcode==OP_Column ){ 646 pOp->opcode = OP_Copy; 647 pOp->p1 = pOp->p2 + iRegister; 648 pOp->p2 = pOp->p3; 649 pOp->p3 = 0; 650 }else if( pOp->opcode==OP_Rowid ){ 651 pOp->opcode = OP_Sequence; 652 pOp->p1 = iAutoidxCur; 653 #ifdef SQLITE_ALLOW_ROWID_IN_VIEW 654 if( iAutoidxCur==0 ){ 655 pOp->opcode = OP_Null; 656 pOp->p3 = 0; 657 } 658 #endif 659 } 660 } 661 } 662 663 /* 664 ** Two routines for printing the content of an sqlite3_index_info 665 ** structure. Used for testing and debugging only. If neither 666 ** SQLITE_TEST or SQLITE_DEBUG are defined, then these routines 667 ** are no-ops. 668 */ 669 #if !defined(SQLITE_OMIT_VIRTUALTABLE) && defined(WHERETRACE_ENABLED) 670 static void whereTraceIndexInfoInputs(sqlite3_index_info *p){ 671 int i; 672 if( !sqlite3WhereTrace ) return; 673 for(i=0; i<p->nConstraint; i++){ 674 sqlite3DebugPrintf(" constraint[%d]: col=%d termid=%d op=%d usabled=%d\n", 675 i, 676 p->aConstraint[i].iColumn, 677 p->aConstraint[i].iTermOffset, 678 p->aConstraint[i].op, 679 p->aConstraint[i].usable); 680 } 681 for(i=0; i<p->nOrderBy; i++){ 682 sqlite3DebugPrintf(" orderby[%d]: col=%d desc=%d\n", 683 i, 684 p->aOrderBy[i].iColumn, 685 p->aOrderBy[i].desc); 686 } 687 } 688 static void whereTraceIndexInfoOutputs(sqlite3_index_info *p){ 689 int i; 690 if( !sqlite3WhereTrace ) return; 691 for(i=0; i<p->nConstraint; i++){ 692 sqlite3DebugPrintf(" usage[%d]: argvIdx=%d omit=%d\n", 693 i, 694 p->aConstraintUsage[i].argvIndex, 695 p->aConstraintUsage[i].omit); 696 } 697 sqlite3DebugPrintf(" idxNum=%d\n", p->idxNum); 698 sqlite3DebugPrintf(" idxStr=%s\n", p->idxStr); 699 sqlite3DebugPrintf(" orderByConsumed=%d\n", p->orderByConsumed); 700 sqlite3DebugPrintf(" estimatedCost=%g\n", p->estimatedCost); 701 sqlite3DebugPrintf(" estimatedRows=%lld\n", p->estimatedRows); 702 } 703 #else 704 #define whereTraceIndexInfoInputs(A) 705 #define whereTraceIndexInfoOutputs(A) 706 #endif 707 708 #ifndef SQLITE_OMIT_AUTOMATIC_INDEX 709 /* 710 ** Return TRUE if the WHERE clause term pTerm is of a form where it 711 ** could be used with an index to access pSrc, assuming an appropriate 712 ** index existed. 713 */ 714 static int termCanDriveIndex( 715 WhereTerm *pTerm, /* WHERE clause term to check */ 716 SrcItem *pSrc, /* Table we are trying to access */ 717 Bitmask notReady /* Tables in outer loops of the join */ 718 ){ 719 char aff; 720 if( pTerm->leftCursor!=pSrc->iCursor ) return 0; 721 if( (pTerm->eOperator & (WO_EQ|WO_IS))==0 ) return 0; 722 if( (pSrc->fg.jointype & JT_LEFT) 723 && !ExprHasProperty(pTerm->pExpr, EP_FromJoin) 724 && (pTerm->eOperator & WO_IS) 725 ){ 726 /* Cannot use an IS term from the WHERE clause as an index driver for 727 ** the RHS of a LEFT JOIN. Such a term can only be used if it is from 728 ** the ON clause. */ 729 return 0; 730 } 731 if( (pTerm->prereqRight & notReady)!=0 ) return 0; 732 assert( (pTerm->eOperator & (WO_OR|WO_AND))==0 ); 733 if( pTerm->u.x.leftColumn<0 ) return 0; 734 aff = pSrc->pTab->aCol[pTerm->u.x.leftColumn].affinity; 735 if( !sqlite3IndexAffinityOk(pTerm->pExpr, aff) ) return 0; 736 testcase( pTerm->pExpr->op==TK_IS ); 737 return 1; 738 } 739 #endif 740 741 742 #ifndef SQLITE_OMIT_AUTOMATIC_INDEX 743 /* 744 ** Generate code to construct the Index object for an automatic index 745 ** and to set up the WhereLevel object pLevel so that the code generator 746 ** makes use of the automatic index. 747 */ 748 static void constructAutomaticIndex( 749 Parse *pParse, /* The parsing context */ 750 WhereClause *pWC, /* The WHERE clause */ 751 SrcItem *pSrc, /* The FROM clause term to get the next index */ 752 Bitmask notReady, /* Mask of cursors that are not available */ 753 WhereLevel *pLevel /* Write new index here */ 754 ){ 755 int nKeyCol; /* Number of columns in the constructed index */ 756 WhereTerm *pTerm; /* A single term of the WHERE clause */ 757 WhereTerm *pWCEnd; /* End of pWC->a[] */ 758 Index *pIdx; /* Object describing the transient index */ 759 Vdbe *v; /* Prepared statement under construction */ 760 int addrInit; /* Address of the initialization bypass jump */ 761 Table *pTable; /* The table being indexed */ 762 int addrTop; /* Top of the index fill loop */ 763 int regRecord; /* Register holding an index record */ 764 int n; /* Column counter */ 765 int i; /* Loop counter */ 766 int mxBitCol; /* Maximum column in pSrc->colUsed */ 767 CollSeq *pColl; /* Collating sequence to on a column */ 768 WhereLoop *pLoop; /* The Loop object */ 769 char *zNotUsed; /* Extra space on the end of pIdx */ 770 Bitmask idxCols; /* Bitmap of columns used for indexing */ 771 Bitmask extraCols; /* Bitmap of additional columns */ 772 u8 sentWarning = 0; /* True if a warnning has been issued */ 773 Expr *pPartial = 0; /* Partial Index Expression */ 774 int iContinue = 0; /* Jump here to skip excluded rows */ 775 SrcItem *pTabItem; /* FROM clause term being indexed */ 776 int addrCounter = 0; /* Address where integer counter is initialized */ 777 int regBase; /* Array of registers where record is assembled */ 778 779 /* Generate code to skip over the creation and initialization of the 780 ** transient index on 2nd and subsequent iterations of the loop. */ 781 v = pParse->pVdbe; 782 assert( v!=0 ); 783 addrInit = sqlite3VdbeAddOp0(v, OP_Once); VdbeCoverage(v); 784 785 /* Count the number of columns that will be added to the index 786 ** and used to match WHERE clause constraints */ 787 nKeyCol = 0; 788 pTable = pSrc->pTab; 789 pWCEnd = &pWC->a[pWC->nTerm]; 790 pLoop = pLevel->pWLoop; 791 idxCols = 0; 792 for(pTerm=pWC->a; pTerm<pWCEnd; pTerm++){ 793 Expr *pExpr = pTerm->pExpr; 794 assert( !ExprHasProperty(pExpr, EP_FromJoin) /* prereq always non-zero */ 795 || pExpr->iRightJoinTable!=pSrc->iCursor /* for the right-hand */ 796 || pLoop->prereq!=0 ); /* table of a LEFT JOIN */ 797 if( pLoop->prereq==0 798 && (pTerm->wtFlags & TERM_VIRTUAL)==0 799 && !ExprHasProperty(pExpr, EP_FromJoin) 800 && sqlite3ExprIsTableConstant(pExpr, pSrc->iCursor) ){ 801 pPartial = sqlite3ExprAnd(pParse, pPartial, 802 sqlite3ExprDup(pParse->db, pExpr, 0)); 803 } 804 if( termCanDriveIndex(pTerm, pSrc, notReady) ){ 805 int iCol; 806 Bitmask cMask; 807 assert( (pTerm->eOperator & (WO_OR|WO_AND))==0 ); 808 iCol = pTerm->u.x.leftColumn; 809 cMask = iCol>=BMS ? MASKBIT(BMS-1) : MASKBIT(iCol); 810 testcase( iCol==BMS ); 811 testcase( iCol==BMS-1 ); 812 if( !sentWarning ){ 813 sqlite3_log(SQLITE_WARNING_AUTOINDEX, 814 "automatic index on %s(%s)", pTable->zName, 815 pTable->aCol[iCol].zCnName); 816 sentWarning = 1; 817 } 818 if( (idxCols & cMask)==0 ){ 819 if( whereLoopResize(pParse->db, pLoop, nKeyCol+1) ){ 820 goto end_auto_index_create; 821 } 822 pLoop->aLTerm[nKeyCol++] = pTerm; 823 idxCols |= cMask; 824 } 825 } 826 } 827 assert( nKeyCol>0 || pParse->db->mallocFailed ); 828 pLoop->u.btree.nEq = pLoop->nLTerm = nKeyCol; 829 pLoop->wsFlags = WHERE_COLUMN_EQ | WHERE_IDX_ONLY | WHERE_INDEXED 830 | WHERE_AUTO_INDEX; 831 832 /* Count the number of additional columns needed to create a 833 ** covering index. A "covering index" is an index that contains all 834 ** columns that are needed by the query. With a covering index, the 835 ** original table never needs to be accessed. Automatic indices must 836 ** be a covering index because the index will not be updated if the 837 ** original table changes and the index and table cannot both be used 838 ** if they go out of sync. 839 */ 840 extraCols = pSrc->colUsed & (~idxCols | MASKBIT(BMS-1)); 841 mxBitCol = MIN(BMS-1,pTable->nCol); 842 testcase( pTable->nCol==BMS-1 ); 843 testcase( pTable->nCol==BMS-2 ); 844 for(i=0; i<mxBitCol; i++){ 845 if( extraCols & MASKBIT(i) ) nKeyCol++; 846 } 847 if( pSrc->colUsed & MASKBIT(BMS-1) ){ 848 nKeyCol += pTable->nCol - BMS + 1; 849 } 850 851 /* Construct the Index object to describe this index */ 852 pIdx = sqlite3AllocateIndexObject(pParse->db, nKeyCol+1, 0, &zNotUsed); 853 if( pIdx==0 ) goto end_auto_index_create; 854 pLoop->u.btree.pIndex = pIdx; 855 pIdx->zName = "auto-index"; 856 pIdx->pTable = pTable; 857 n = 0; 858 idxCols = 0; 859 for(pTerm=pWC->a; pTerm<pWCEnd; pTerm++){ 860 if( termCanDriveIndex(pTerm, pSrc, notReady) ){ 861 int iCol; 862 Bitmask cMask; 863 assert( (pTerm->eOperator & (WO_OR|WO_AND))==0 ); 864 iCol = pTerm->u.x.leftColumn; 865 cMask = iCol>=BMS ? MASKBIT(BMS-1) : MASKBIT(iCol); 866 testcase( iCol==BMS-1 ); 867 testcase( iCol==BMS ); 868 if( (idxCols & cMask)==0 ){ 869 Expr *pX = pTerm->pExpr; 870 idxCols |= cMask; 871 pIdx->aiColumn[n] = pTerm->u.x.leftColumn; 872 pColl = sqlite3ExprCompareCollSeq(pParse, pX); 873 assert( pColl!=0 || pParse->nErr>0 ); /* TH3 collate01.800 */ 874 pIdx->azColl[n] = pColl ? pColl->zName : sqlite3StrBINARY; 875 n++; 876 } 877 } 878 } 879 assert( (u32)n==pLoop->u.btree.nEq ); 880 881 /* Add additional columns needed to make the automatic index into 882 ** a covering index */ 883 for(i=0; i<mxBitCol; i++){ 884 if( extraCols & MASKBIT(i) ){ 885 pIdx->aiColumn[n] = i; 886 pIdx->azColl[n] = sqlite3StrBINARY; 887 n++; 888 } 889 } 890 if( pSrc->colUsed & MASKBIT(BMS-1) ){ 891 for(i=BMS-1; i<pTable->nCol; i++){ 892 pIdx->aiColumn[n] = i; 893 pIdx->azColl[n] = sqlite3StrBINARY; 894 n++; 895 } 896 } 897 assert( n==nKeyCol ); 898 pIdx->aiColumn[n] = XN_ROWID; 899 pIdx->azColl[n] = sqlite3StrBINARY; 900 901 /* Create the automatic index */ 902 assert( pLevel->iIdxCur>=0 ); 903 pLevel->iIdxCur = pParse->nTab++; 904 sqlite3VdbeAddOp2(v, OP_OpenAutoindex, pLevel->iIdxCur, nKeyCol+1); 905 sqlite3VdbeSetP4KeyInfo(pParse, pIdx); 906 VdbeComment((v, "for %s", pTable->zName)); 907 908 /* Fill the automatic index with content */ 909 pTabItem = &pWC->pWInfo->pTabList->a[pLevel->iFrom]; 910 if( pTabItem->fg.viaCoroutine ){ 911 int regYield = pTabItem->regReturn; 912 addrCounter = sqlite3VdbeAddOp2(v, OP_Integer, 0, 0); 913 sqlite3VdbeAddOp3(v, OP_InitCoroutine, regYield, 0, pTabItem->addrFillSub); 914 addrTop = sqlite3VdbeAddOp1(v, OP_Yield, regYield); 915 VdbeCoverage(v); 916 VdbeComment((v, "next row of %s", pTabItem->pTab->zName)); 917 }else{ 918 addrTop = sqlite3VdbeAddOp1(v, OP_Rewind, pLevel->iTabCur); VdbeCoverage(v); 919 } 920 if( pPartial ){ 921 iContinue = sqlite3VdbeMakeLabel(pParse); 922 sqlite3ExprIfFalse(pParse, pPartial, iContinue, SQLITE_JUMPIFNULL); 923 pLoop->wsFlags |= WHERE_PARTIALIDX; 924 } 925 regRecord = sqlite3GetTempReg(pParse); 926 regBase = sqlite3GenerateIndexKey( 927 pParse, pIdx, pLevel->iTabCur, regRecord, 0, 0, 0, 0 928 ); 929 sqlite3VdbeAddOp2(v, OP_IdxInsert, pLevel->iIdxCur, regRecord); 930 sqlite3VdbeChangeP5(v, OPFLAG_USESEEKRESULT); 931 if( pPartial ) sqlite3VdbeResolveLabel(v, iContinue); 932 if( pTabItem->fg.viaCoroutine ){ 933 sqlite3VdbeChangeP2(v, addrCounter, regBase+n); 934 testcase( pParse->db->mallocFailed ); 935 assert( pLevel->iIdxCur>0 ); 936 translateColumnToCopy(pParse, addrTop, pLevel->iTabCur, 937 pTabItem->regResult, pLevel->iIdxCur); 938 sqlite3VdbeGoto(v, addrTop); 939 pTabItem->fg.viaCoroutine = 0; 940 }else{ 941 sqlite3VdbeAddOp2(v, OP_Next, pLevel->iTabCur, addrTop+1); VdbeCoverage(v); 942 sqlite3VdbeChangeP5(v, SQLITE_STMTSTATUS_AUTOINDEX); 943 } 944 sqlite3VdbeJumpHere(v, addrTop); 945 sqlite3ReleaseTempReg(pParse, regRecord); 946 947 /* Jump here when skipping the initialization */ 948 sqlite3VdbeJumpHere(v, addrInit); 949 950 end_auto_index_create: 951 sqlite3ExprDelete(pParse->db, pPartial); 952 } 953 #endif /* SQLITE_OMIT_AUTOMATIC_INDEX */ 954 955 #ifndef SQLITE_OMIT_VIRTUALTABLE 956 /* 957 ** Allocate and populate an sqlite3_index_info structure. It is the 958 ** responsibility of the caller to eventually release the structure 959 ** by passing the pointer returned by this function to sqlite3_free(). 960 */ 961 static sqlite3_index_info *allocateIndexInfo( 962 Parse *pParse, /* The parsing context */ 963 WhereClause *pWC, /* The WHERE clause being analyzed */ 964 Bitmask mUnusable, /* Ignore terms with these prereqs */ 965 SrcItem *pSrc, /* The FROM clause term that is the vtab */ 966 ExprList *pOrderBy, /* The ORDER BY clause */ 967 u16 *pmNoOmit /* Mask of terms not to omit */ 968 ){ 969 int i, j; 970 int nTerm; 971 struct sqlite3_index_constraint *pIdxCons; 972 struct sqlite3_index_orderby *pIdxOrderBy; 973 struct sqlite3_index_constraint_usage *pUsage; 974 struct HiddenIndexInfo *pHidden; 975 WhereTerm *pTerm; 976 int nOrderBy; 977 sqlite3_index_info *pIdxInfo; 978 u16 mNoOmit = 0; 979 980 /* Count the number of possible WHERE clause constraints referring 981 ** to this virtual table */ 982 for(i=nTerm=0, pTerm=pWC->a; i<pWC->nTerm; i++, pTerm++){ 983 if( pTerm->leftCursor != pSrc->iCursor ) continue; 984 if( pTerm->prereqRight & mUnusable ) continue; 985 assert( IsPowerOfTwo(pTerm->eOperator & ~WO_EQUIV) ); 986 testcase( pTerm->eOperator & WO_IN ); 987 testcase( pTerm->eOperator & WO_ISNULL ); 988 testcase( pTerm->eOperator & WO_IS ); 989 testcase( pTerm->eOperator & WO_ALL ); 990 if( (pTerm->eOperator & ~(WO_EQUIV))==0 ) continue; 991 if( pTerm->wtFlags & TERM_VNULL ) continue; 992 assert( (pTerm->eOperator & (WO_OR|WO_AND))==0 ); 993 assert( pTerm->u.x.leftColumn>=(-1) ); 994 nTerm++; 995 } 996 997 /* If the ORDER BY clause contains only columns in the current 998 ** virtual table then allocate space for the aOrderBy part of 999 ** the sqlite3_index_info structure. 1000 */ 1001 nOrderBy = 0; 1002 if( pOrderBy ){ 1003 int n = pOrderBy->nExpr; 1004 for(i=0; i<n; i++){ 1005 Expr *pExpr = pOrderBy->a[i].pExpr; 1006 if( pExpr->op!=TK_COLUMN || pExpr->iTable!=pSrc->iCursor ) break; 1007 if( pOrderBy->a[i].sortFlags & KEYINFO_ORDER_BIGNULL ) break; 1008 } 1009 if( i==n){ 1010 nOrderBy = n; 1011 } 1012 } 1013 1014 /* Allocate the sqlite3_index_info structure 1015 */ 1016 pIdxInfo = sqlite3DbMallocZero(pParse->db, sizeof(*pIdxInfo) 1017 + (sizeof(*pIdxCons) + sizeof(*pUsage))*nTerm 1018 + sizeof(*pIdxOrderBy)*nOrderBy + sizeof(*pHidden) ); 1019 if( pIdxInfo==0 ){ 1020 sqlite3ErrorMsg(pParse, "out of memory"); 1021 return 0; 1022 } 1023 pHidden = (struct HiddenIndexInfo*)&pIdxInfo[1]; 1024 pIdxCons = (struct sqlite3_index_constraint*)&pHidden[1]; 1025 pIdxOrderBy = (struct sqlite3_index_orderby*)&pIdxCons[nTerm]; 1026 pUsage = (struct sqlite3_index_constraint_usage*)&pIdxOrderBy[nOrderBy]; 1027 pIdxInfo->nOrderBy = nOrderBy; 1028 pIdxInfo->aConstraint = pIdxCons; 1029 pIdxInfo->aOrderBy = pIdxOrderBy; 1030 pIdxInfo->aConstraintUsage = pUsage; 1031 pHidden->pWC = pWC; 1032 pHidden->pParse = pParse; 1033 for(i=j=0, pTerm=pWC->a; i<pWC->nTerm; i++, pTerm++){ 1034 u16 op; 1035 if( pTerm->leftCursor != pSrc->iCursor ) continue; 1036 if( pTerm->prereqRight & mUnusable ) continue; 1037 assert( IsPowerOfTwo(pTerm->eOperator & ~WO_EQUIV) ); 1038 testcase( pTerm->eOperator & WO_IN ); 1039 testcase( pTerm->eOperator & WO_IS ); 1040 testcase( pTerm->eOperator & WO_ISNULL ); 1041 testcase( pTerm->eOperator & WO_ALL ); 1042 if( (pTerm->eOperator & ~(WO_EQUIV))==0 ) continue; 1043 if( pTerm->wtFlags & TERM_VNULL ) continue; 1044 1045 /* tag-20191211-002: WHERE-clause constraints are not useful to the 1046 ** right-hand table of a LEFT JOIN. See tag-20191211-001 for the 1047 ** equivalent restriction for ordinary tables. */ 1048 if( (pSrc->fg.jointype & JT_LEFT)!=0 1049 && !ExprHasProperty(pTerm->pExpr, EP_FromJoin) 1050 ){ 1051 continue; 1052 } 1053 assert( (pTerm->eOperator & (WO_OR|WO_AND))==0 ); 1054 assert( pTerm->u.x.leftColumn>=(-1) ); 1055 pIdxCons[j].iColumn = pTerm->u.x.leftColumn; 1056 pIdxCons[j].iTermOffset = i; 1057 op = pTerm->eOperator & WO_ALL; 1058 if( op==WO_IN ) op = WO_EQ; 1059 if( op==WO_AUX ){ 1060 pIdxCons[j].op = pTerm->eMatchOp; 1061 }else if( op & (WO_ISNULL|WO_IS) ){ 1062 if( op==WO_ISNULL ){ 1063 pIdxCons[j].op = SQLITE_INDEX_CONSTRAINT_ISNULL; 1064 }else{ 1065 pIdxCons[j].op = SQLITE_INDEX_CONSTRAINT_IS; 1066 } 1067 }else{ 1068 pIdxCons[j].op = (u8)op; 1069 /* The direct assignment in the previous line is possible only because 1070 ** the WO_ and SQLITE_INDEX_CONSTRAINT_ codes are identical. The 1071 ** following asserts verify this fact. */ 1072 assert( WO_EQ==SQLITE_INDEX_CONSTRAINT_EQ ); 1073 assert( WO_LT==SQLITE_INDEX_CONSTRAINT_LT ); 1074 assert( WO_LE==SQLITE_INDEX_CONSTRAINT_LE ); 1075 assert( WO_GT==SQLITE_INDEX_CONSTRAINT_GT ); 1076 assert( WO_GE==SQLITE_INDEX_CONSTRAINT_GE ); 1077 assert( pTerm->eOperator&(WO_IN|WO_EQ|WO_LT|WO_LE|WO_GT|WO_GE|WO_AUX) ); 1078 1079 if( op & (WO_LT|WO_LE|WO_GT|WO_GE) 1080 && sqlite3ExprIsVector(pTerm->pExpr->pRight) 1081 ){ 1082 testcase( j!=i ); 1083 if( j<16 ) mNoOmit |= (1 << j); 1084 if( op==WO_LT ) pIdxCons[j].op = WO_LE; 1085 if( op==WO_GT ) pIdxCons[j].op = WO_GE; 1086 } 1087 } 1088 1089 j++; 1090 } 1091 pIdxInfo->nConstraint = j; 1092 for(i=0; i<nOrderBy; i++){ 1093 Expr *pExpr = pOrderBy->a[i].pExpr; 1094 pIdxOrderBy[i].iColumn = pExpr->iColumn; 1095 pIdxOrderBy[i].desc = pOrderBy->a[i].sortFlags & KEYINFO_ORDER_DESC; 1096 } 1097 1098 *pmNoOmit = mNoOmit; 1099 return pIdxInfo; 1100 } 1101 1102 /* 1103 ** The table object reference passed as the second argument to this function 1104 ** must represent a virtual table. This function invokes the xBestIndex() 1105 ** method of the virtual table with the sqlite3_index_info object that 1106 ** comes in as the 3rd argument to this function. 1107 ** 1108 ** If an error occurs, pParse is populated with an error message and an 1109 ** appropriate error code is returned. A return of SQLITE_CONSTRAINT from 1110 ** xBestIndex is not considered an error. SQLITE_CONSTRAINT indicates that 1111 ** the current configuration of "unusable" flags in sqlite3_index_info can 1112 ** not result in a valid plan. 1113 ** 1114 ** Whether or not an error is returned, it is the responsibility of the 1115 ** caller to eventually free p->idxStr if p->needToFreeIdxStr indicates 1116 ** that this is required. 1117 */ 1118 static int vtabBestIndex(Parse *pParse, Table *pTab, sqlite3_index_info *p){ 1119 sqlite3_vtab *pVtab = sqlite3GetVTable(pParse->db, pTab)->pVtab; 1120 int rc; 1121 1122 whereTraceIndexInfoInputs(p); 1123 rc = pVtab->pModule->xBestIndex(pVtab, p); 1124 whereTraceIndexInfoOutputs(p); 1125 1126 if( rc!=SQLITE_OK && rc!=SQLITE_CONSTRAINT ){ 1127 if( rc==SQLITE_NOMEM ){ 1128 sqlite3OomFault(pParse->db); 1129 }else if( !pVtab->zErrMsg ){ 1130 sqlite3ErrorMsg(pParse, "%s", sqlite3ErrStr(rc)); 1131 }else{ 1132 sqlite3ErrorMsg(pParse, "%s", pVtab->zErrMsg); 1133 } 1134 } 1135 sqlite3_free(pVtab->zErrMsg); 1136 pVtab->zErrMsg = 0; 1137 return rc; 1138 } 1139 #endif /* !defined(SQLITE_OMIT_VIRTUALTABLE) */ 1140 1141 #ifdef SQLITE_ENABLE_STAT4 1142 /* 1143 ** Estimate the location of a particular key among all keys in an 1144 ** index. Store the results in aStat as follows: 1145 ** 1146 ** aStat[0] Est. number of rows less than pRec 1147 ** aStat[1] Est. number of rows equal to pRec 1148 ** 1149 ** Return the index of the sample that is the smallest sample that 1150 ** is greater than or equal to pRec. Note that this index is not an index 1151 ** into the aSample[] array - it is an index into a virtual set of samples 1152 ** based on the contents of aSample[] and the number of fields in record 1153 ** pRec. 1154 */ 1155 static int whereKeyStats( 1156 Parse *pParse, /* Database connection */ 1157 Index *pIdx, /* Index to consider domain of */ 1158 UnpackedRecord *pRec, /* Vector of values to consider */ 1159 int roundUp, /* Round up if true. Round down if false */ 1160 tRowcnt *aStat /* OUT: stats written here */ 1161 ){ 1162 IndexSample *aSample = pIdx->aSample; 1163 int iCol; /* Index of required stats in anEq[] etc. */ 1164 int i; /* Index of first sample >= pRec */ 1165 int iSample; /* Smallest sample larger than or equal to pRec */ 1166 int iMin = 0; /* Smallest sample not yet tested */ 1167 int iTest; /* Next sample to test */ 1168 int res; /* Result of comparison operation */ 1169 int nField; /* Number of fields in pRec */ 1170 tRowcnt iLower = 0; /* anLt[] + anEq[] of largest sample pRec is > */ 1171 1172 #ifndef SQLITE_DEBUG 1173 UNUSED_PARAMETER( pParse ); 1174 #endif 1175 assert( pRec!=0 ); 1176 assert( pIdx->nSample>0 ); 1177 assert( pRec->nField>0 && pRec->nField<=pIdx->nSampleCol ); 1178 1179 /* Do a binary search to find the first sample greater than or equal 1180 ** to pRec. If pRec contains a single field, the set of samples to search 1181 ** is simply the aSample[] array. If the samples in aSample[] contain more 1182 ** than one fields, all fields following the first are ignored. 1183 ** 1184 ** If pRec contains N fields, where N is more than one, then as well as the 1185 ** samples in aSample[] (truncated to N fields), the search also has to 1186 ** consider prefixes of those samples. For example, if the set of samples 1187 ** in aSample is: 1188 ** 1189 ** aSample[0] = (a, 5) 1190 ** aSample[1] = (a, 10) 1191 ** aSample[2] = (b, 5) 1192 ** aSample[3] = (c, 100) 1193 ** aSample[4] = (c, 105) 1194 ** 1195 ** Then the search space should ideally be the samples above and the 1196 ** unique prefixes [a], [b] and [c]. But since that is hard to organize, 1197 ** the code actually searches this set: 1198 ** 1199 ** 0: (a) 1200 ** 1: (a, 5) 1201 ** 2: (a, 10) 1202 ** 3: (a, 10) 1203 ** 4: (b) 1204 ** 5: (b, 5) 1205 ** 6: (c) 1206 ** 7: (c, 100) 1207 ** 8: (c, 105) 1208 ** 9: (c, 105) 1209 ** 1210 ** For each sample in the aSample[] array, N samples are present in the 1211 ** effective sample array. In the above, samples 0 and 1 are based on 1212 ** sample aSample[0]. Samples 2 and 3 on aSample[1] etc. 1213 ** 1214 ** Often, sample i of each block of N effective samples has (i+1) fields. 1215 ** Except, each sample may be extended to ensure that it is greater than or 1216 ** equal to the previous sample in the array. For example, in the above, 1217 ** sample 2 is the first sample of a block of N samples, so at first it 1218 ** appears that it should be 1 field in size. However, that would make it 1219 ** smaller than sample 1, so the binary search would not work. As a result, 1220 ** it is extended to two fields. The duplicates that this creates do not 1221 ** cause any problems. 1222 */ 1223 nField = pRec->nField; 1224 iCol = 0; 1225 iSample = pIdx->nSample * nField; 1226 do{ 1227 int iSamp; /* Index in aSample[] of test sample */ 1228 int n; /* Number of fields in test sample */ 1229 1230 iTest = (iMin+iSample)/2; 1231 iSamp = iTest / nField; 1232 if( iSamp>0 ){ 1233 /* The proposed effective sample is a prefix of sample aSample[iSamp]. 1234 ** Specifically, the shortest prefix of at least (1 + iTest%nField) 1235 ** fields that is greater than the previous effective sample. */ 1236 for(n=(iTest % nField) + 1; n<nField; n++){ 1237 if( aSample[iSamp-1].anLt[n-1]!=aSample[iSamp].anLt[n-1] ) break; 1238 } 1239 }else{ 1240 n = iTest + 1; 1241 } 1242 1243 pRec->nField = n; 1244 res = sqlite3VdbeRecordCompare(aSample[iSamp].n, aSample[iSamp].p, pRec); 1245 if( res<0 ){ 1246 iLower = aSample[iSamp].anLt[n-1] + aSample[iSamp].anEq[n-1]; 1247 iMin = iTest+1; 1248 }else if( res==0 && n<nField ){ 1249 iLower = aSample[iSamp].anLt[n-1]; 1250 iMin = iTest+1; 1251 res = -1; 1252 }else{ 1253 iSample = iTest; 1254 iCol = n-1; 1255 } 1256 }while( res && iMin<iSample ); 1257 i = iSample / nField; 1258 1259 #ifdef SQLITE_DEBUG 1260 /* The following assert statements check that the binary search code 1261 ** above found the right answer. This block serves no purpose other 1262 ** than to invoke the asserts. */ 1263 if( pParse->db->mallocFailed==0 ){ 1264 if( res==0 ){ 1265 /* If (res==0) is true, then pRec must be equal to sample i. */ 1266 assert( i<pIdx->nSample ); 1267 assert( iCol==nField-1 ); 1268 pRec->nField = nField; 1269 assert( 0==sqlite3VdbeRecordCompare(aSample[i].n, aSample[i].p, pRec) 1270 || pParse->db->mallocFailed 1271 ); 1272 }else{ 1273 /* Unless i==pIdx->nSample, indicating that pRec is larger than 1274 ** all samples in the aSample[] array, pRec must be smaller than the 1275 ** (iCol+1) field prefix of sample i. */ 1276 assert( i<=pIdx->nSample && i>=0 ); 1277 pRec->nField = iCol+1; 1278 assert( i==pIdx->nSample 1279 || sqlite3VdbeRecordCompare(aSample[i].n, aSample[i].p, pRec)>0 1280 || pParse->db->mallocFailed ); 1281 1282 /* if i==0 and iCol==0, then record pRec is smaller than all samples 1283 ** in the aSample[] array. Otherwise, if (iCol>0) then pRec must 1284 ** be greater than or equal to the (iCol) field prefix of sample i. 1285 ** If (i>0), then pRec must also be greater than sample (i-1). */ 1286 if( iCol>0 ){ 1287 pRec->nField = iCol; 1288 assert( sqlite3VdbeRecordCompare(aSample[i].n, aSample[i].p, pRec)<=0 1289 || pParse->db->mallocFailed ); 1290 } 1291 if( i>0 ){ 1292 pRec->nField = nField; 1293 assert( sqlite3VdbeRecordCompare(aSample[i-1].n, aSample[i-1].p, pRec)<0 1294 || pParse->db->mallocFailed ); 1295 } 1296 } 1297 } 1298 #endif /* ifdef SQLITE_DEBUG */ 1299 1300 if( res==0 ){ 1301 /* Record pRec is equal to sample i */ 1302 assert( iCol==nField-1 ); 1303 aStat[0] = aSample[i].anLt[iCol]; 1304 aStat[1] = aSample[i].anEq[iCol]; 1305 }else{ 1306 /* At this point, the (iCol+1) field prefix of aSample[i] is the first 1307 ** sample that is greater than pRec. Or, if i==pIdx->nSample then pRec 1308 ** is larger than all samples in the array. */ 1309 tRowcnt iUpper, iGap; 1310 if( i>=pIdx->nSample ){ 1311 iUpper = sqlite3LogEstToInt(pIdx->aiRowLogEst[0]); 1312 }else{ 1313 iUpper = aSample[i].anLt[iCol]; 1314 } 1315 1316 if( iLower>=iUpper ){ 1317 iGap = 0; 1318 }else{ 1319 iGap = iUpper - iLower; 1320 } 1321 if( roundUp ){ 1322 iGap = (iGap*2)/3; 1323 }else{ 1324 iGap = iGap/3; 1325 } 1326 aStat[0] = iLower + iGap; 1327 aStat[1] = pIdx->aAvgEq[nField-1]; 1328 } 1329 1330 /* Restore the pRec->nField value before returning. */ 1331 pRec->nField = nField; 1332 return i; 1333 } 1334 #endif /* SQLITE_ENABLE_STAT4 */ 1335 1336 /* 1337 ** If it is not NULL, pTerm is a term that provides an upper or lower 1338 ** bound on a range scan. Without considering pTerm, it is estimated 1339 ** that the scan will visit nNew rows. This function returns the number 1340 ** estimated to be visited after taking pTerm into account. 1341 ** 1342 ** If the user explicitly specified a likelihood() value for this term, 1343 ** then the return value is the likelihood multiplied by the number of 1344 ** input rows. Otherwise, this function assumes that an "IS NOT NULL" term 1345 ** has a likelihood of 0.50, and any other term a likelihood of 0.25. 1346 */ 1347 static LogEst whereRangeAdjust(WhereTerm *pTerm, LogEst nNew){ 1348 LogEst nRet = nNew; 1349 if( pTerm ){ 1350 if( pTerm->truthProb<=0 ){ 1351 nRet += pTerm->truthProb; 1352 }else if( (pTerm->wtFlags & TERM_VNULL)==0 ){ 1353 nRet -= 20; assert( 20==sqlite3LogEst(4) ); 1354 } 1355 } 1356 return nRet; 1357 } 1358 1359 1360 #ifdef SQLITE_ENABLE_STAT4 1361 /* 1362 ** Return the affinity for a single column of an index. 1363 */ 1364 char sqlite3IndexColumnAffinity(sqlite3 *db, Index *pIdx, int iCol){ 1365 assert( iCol>=0 && iCol<pIdx->nColumn ); 1366 if( !pIdx->zColAff ){ 1367 if( sqlite3IndexAffinityStr(db, pIdx)==0 ) return SQLITE_AFF_BLOB; 1368 } 1369 assert( pIdx->zColAff[iCol]!=0 ); 1370 return pIdx->zColAff[iCol]; 1371 } 1372 #endif 1373 1374 1375 #ifdef SQLITE_ENABLE_STAT4 1376 /* 1377 ** This function is called to estimate the number of rows visited by a 1378 ** range-scan on a skip-scan index. For example: 1379 ** 1380 ** CREATE INDEX i1 ON t1(a, b, c); 1381 ** SELECT * FROM t1 WHERE a=? AND c BETWEEN ? AND ?; 1382 ** 1383 ** Value pLoop->nOut is currently set to the estimated number of rows 1384 ** visited for scanning (a=? AND b=?). This function reduces that estimate 1385 ** by some factor to account for the (c BETWEEN ? AND ?) expression based 1386 ** on the stat4 data for the index. this scan will be peformed multiple 1387 ** times (once for each (a,b) combination that matches a=?) is dealt with 1388 ** by the caller. 1389 ** 1390 ** It does this by scanning through all stat4 samples, comparing values 1391 ** extracted from pLower and pUpper with the corresponding column in each 1392 ** sample. If L and U are the number of samples found to be less than or 1393 ** equal to the values extracted from pLower and pUpper respectively, and 1394 ** N is the total number of samples, the pLoop->nOut value is adjusted 1395 ** as follows: 1396 ** 1397 ** nOut = nOut * ( min(U - L, 1) / N ) 1398 ** 1399 ** If pLower is NULL, or a value cannot be extracted from the term, L is 1400 ** set to zero. If pUpper is NULL, or a value cannot be extracted from it, 1401 ** U is set to N. 1402 ** 1403 ** Normally, this function sets *pbDone to 1 before returning. However, 1404 ** if no value can be extracted from either pLower or pUpper (and so the 1405 ** estimate of the number of rows delivered remains unchanged), *pbDone 1406 ** is left as is. 1407 ** 1408 ** If an error occurs, an SQLite error code is returned. Otherwise, 1409 ** SQLITE_OK. 1410 */ 1411 static int whereRangeSkipScanEst( 1412 Parse *pParse, /* Parsing & code generating context */ 1413 WhereTerm *pLower, /* Lower bound on the range. ex: "x>123" Might be NULL */ 1414 WhereTerm *pUpper, /* Upper bound on the range. ex: "x<455" Might be NULL */ 1415 WhereLoop *pLoop, /* Update the .nOut value of this loop */ 1416 int *pbDone /* Set to true if at least one expr. value extracted */ 1417 ){ 1418 Index *p = pLoop->u.btree.pIndex; 1419 int nEq = pLoop->u.btree.nEq; 1420 sqlite3 *db = pParse->db; 1421 int nLower = -1; 1422 int nUpper = p->nSample+1; 1423 int rc = SQLITE_OK; 1424 u8 aff = sqlite3IndexColumnAffinity(db, p, nEq); 1425 CollSeq *pColl; 1426 1427 sqlite3_value *p1 = 0; /* Value extracted from pLower */ 1428 sqlite3_value *p2 = 0; /* Value extracted from pUpper */ 1429 sqlite3_value *pVal = 0; /* Value extracted from record */ 1430 1431 pColl = sqlite3LocateCollSeq(pParse, p->azColl[nEq]); 1432 if( pLower ){ 1433 rc = sqlite3Stat4ValueFromExpr(pParse, pLower->pExpr->pRight, aff, &p1); 1434 nLower = 0; 1435 } 1436 if( pUpper && rc==SQLITE_OK ){ 1437 rc = sqlite3Stat4ValueFromExpr(pParse, pUpper->pExpr->pRight, aff, &p2); 1438 nUpper = p2 ? 0 : p->nSample; 1439 } 1440 1441 if( p1 || p2 ){ 1442 int i; 1443 int nDiff; 1444 for(i=0; rc==SQLITE_OK && i<p->nSample; i++){ 1445 rc = sqlite3Stat4Column(db, p->aSample[i].p, p->aSample[i].n, nEq, &pVal); 1446 if( rc==SQLITE_OK && p1 ){ 1447 int res = sqlite3MemCompare(p1, pVal, pColl); 1448 if( res>=0 ) nLower++; 1449 } 1450 if( rc==SQLITE_OK && p2 ){ 1451 int res = sqlite3MemCompare(p2, pVal, pColl); 1452 if( res>=0 ) nUpper++; 1453 } 1454 } 1455 nDiff = (nUpper - nLower); 1456 if( nDiff<=0 ) nDiff = 1; 1457 1458 /* If there is both an upper and lower bound specified, and the 1459 ** comparisons indicate that they are close together, use the fallback 1460 ** method (assume that the scan visits 1/64 of the rows) for estimating 1461 ** the number of rows visited. Otherwise, estimate the number of rows 1462 ** using the method described in the header comment for this function. */ 1463 if( nDiff!=1 || pUpper==0 || pLower==0 ){ 1464 int nAdjust = (sqlite3LogEst(p->nSample) - sqlite3LogEst(nDiff)); 1465 pLoop->nOut -= nAdjust; 1466 *pbDone = 1; 1467 WHERETRACE(0x10, ("range skip-scan regions: %u..%u adjust=%d est=%d\n", 1468 nLower, nUpper, nAdjust*-1, pLoop->nOut)); 1469 } 1470 1471 }else{ 1472 assert( *pbDone==0 ); 1473 } 1474 1475 sqlite3ValueFree(p1); 1476 sqlite3ValueFree(p2); 1477 sqlite3ValueFree(pVal); 1478 1479 return rc; 1480 } 1481 #endif /* SQLITE_ENABLE_STAT4 */ 1482 1483 /* 1484 ** This function is used to estimate the number of rows that will be visited 1485 ** by scanning an index for a range of values. The range may have an upper 1486 ** bound, a lower bound, or both. The WHERE clause terms that set the upper 1487 ** and lower bounds are represented by pLower and pUpper respectively. For 1488 ** example, assuming that index p is on t1(a): 1489 ** 1490 ** ... FROM t1 WHERE a > ? AND a < ? ... 1491 ** |_____| |_____| 1492 ** | | 1493 ** pLower pUpper 1494 ** 1495 ** If either of the upper or lower bound is not present, then NULL is passed in 1496 ** place of the corresponding WhereTerm. 1497 ** 1498 ** The value in (pBuilder->pNew->u.btree.nEq) is the number of the index 1499 ** column subject to the range constraint. Or, equivalently, the number of 1500 ** equality constraints optimized by the proposed index scan. For example, 1501 ** assuming index p is on t1(a, b), and the SQL query is: 1502 ** 1503 ** ... FROM t1 WHERE a = ? AND b > ? AND b < ? ... 1504 ** 1505 ** then nEq is set to 1 (as the range restricted column, b, is the second 1506 ** left-most column of the index). Or, if the query is: 1507 ** 1508 ** ... FROM t1 WHERE a > ? AND a < ? ... 1509 ** 1510 ** then nEq is set to 0. 1511 ** 1512 ** When this function is called, *pnOut is set to the sqlite3LogEst() of the 1513 ** number of rows that the index scan is expected to visit without 1514 ** considering the range constraints. If nEq is 0, then *pnOut is the number of 1515 ** rows in the index. Assuming no error occurs, *pnOut is adjusted (reduced) 1516 ** to account for the range constraints pLower and pUpper. 1517 ** 1518 ** In the absence of sqlite_stat4 ANALYZE data, or if such data cannot be 1519 ** used, a single range inequality reduces the search space by a factor of 4. 1520 ** and a pair of constraints (x>? AND x<?) reduces the expected number of 1521 ** rows visited by a factor of 64. 1522 */ 1523 static int whereRangeScanEst( 1524 Parse *pParse, /* Parsing & code generating context */ 1525 WhereLoopBuilder *pBuilder, 1526 WhereTerm *pLower, /* Lower bound on the range. ex: "x>123" Might be NULL */ 1527 WhereTerm *pUpper, /* Upper bound on the range. ex: "x<455" Might be NULL */ 1528 WhereLoop *pLoop /* Modify the .nOut and maybe .rRun fields */ 1529 ){ 1530 int rc = SQLITE_OK; 1531 int nOut = pLoop->nOut; 1532 LogEst nNew; 1533 1534 #ifdef SQLITE_ENABLE_STAT4 1535 Index *p = pLoop->u.btree.pIndex; 1536 int nEq = pLoop->u.btree.nEq; 1537 1538 if( p->nSample>0 && ALWAYS(nEq<p->nSampleCol) 1539 && OptimizationEnabled(pParse->db, SQLITE_Stat4) 1540 ){ 1541 if( nEq==pBuilder->nRecValid ){ 1542 UnpackedRecord *pRec = pBuilder->pRec; 1543 tRowcnt a[2]; 1544 int nBtm = pLoop->u.btree.nBtm; 1545 int nTop = pLoop->u.btree.nTop; 1546 1547 /* Variable iLower will be set to the estimate of the number of rows in 1548 ** the index that are less than the lower bound of the range query. The 1549 ** lower bound being the concatenation of $P and $L, where $P is the 1550 ** key-prefix formed by the nEq values matched against the nEq left-most 1551 ** columns of the index, and $L is the value in pLower. 1552 ** 1553 ** Or, if pLower is NULL or $L cannot be extracted from it (because it 1554 ** is not a simple variable or literal value), the lower bound of the 1555 ** range is $P. Due to a quirk in the way whereKeyStats() works, even 1556 ** if $L is available, whereKeyStats() is called for both ($P) and 1557 ** ($P:$L) and the larger of the two returned values is used. 1558 ** 1559 ** Similarly, iUpper is to be set to the estimate of the number of rows 1560 ** less than the upper bound of the range query. Where the upper bound 1561 ** is either ($P) or ($P:$U). Again, even if $U is available, both values 1562 ** of iUpper are requested of whereKeyStats() and the smaller used. 1563 ** 1564 ** The number of rows between the two bounds is then just iUpper-iLower. 1565 */ 1566 tRowcnt iLower; /* Rows less than the lower bound */ 1567 tRowcnt iUpper; /* Rows less than the upper bound */ 1568 int iLwrIdx = -2; /* aSample[] for the lower bound */ 1569 int iUprIdx = -1; /* aSample[] for the upper bound */ 1570 1571 if( pRec ){ 1572 testcase( pRec->nField!=pBuilder->nRecValid ); 1573 pRec->nField = pBuilder->nRecValid; 1574 } 1575 /* Determine iLower and iUpper using ($P) only. */ 1576 if( nEq==0 ){ 1577 iLower = 0; 1578 iUpper = p->nRowEst0; 1579 }else{ 1580 /* Note: this call could be optimized away - since the same values must 1581 ** have been requested when testing key $P in whereEqualScanEst(). */ 1582 whereKeyStats(pParse, p, pRec, 0, a); 1583 iLower = a[0]; 1584 iUpper = a[0] + a[1]; 1585 } 1586 1587 assert( pLower==0 || (pLower->eOperator & (WO_GT|WO_GE))!=0 ); 1588 assert( pUpper==0 || (pUpper->eOperator & (WO_LT|WO_LE))!=0 ); 1589 assert( p->aSortOrder!=0 ); 1590 if( p->aSortOrder[nEq] ){ 1591 /* The roles of pLower and pUpper are swapped for a DESC index */ 1592 SWAP(WhereTerm*, pLower, pUpper); 1593 SWAP(int, nBtm, nTop); 1594 } 1595 1596 /* If possible, improve on the iLower estimate using ($P:$L). */ 1597 if( pLower ){ 1598 int n; /* Values extracted from pExpr */ 1599 Expr *pExpr = pLower->pExpr->pRight; 1600 rc = sqlite3Stat4ProbeSetValue(pParse, p, &pRec, pExpr, nBtm, nEq, &n); 1601 if( rc==SQLITE_OK && n ){ 1602 tRowcnt iNew; 1603 u16 mask = WO_GT|WO_LE; 1604 if( sqlite3ExprVectorSize(pExpr)>n ) mask = (WO_LE|WO_LT); 1605 iLwrIdx = whereKeyStats(pParse, p, pRec, 0, a); 1606 iNew = a[0] + ((pLower->eOperator & mask) ? a[1] : 0); 1607 if( iNew>iLower ) iLower = iNew; 1608 nOut--; 1609 pLower = 0; 1610 } 1611 } 1612 1613 /* If possible, improve on the iUpper estimate using ($P:$U). */ 1614 if( pUpper ){ 1615 int n; /* Values extracted from pExpr */ 1616 Expr *pExpr = pUpper->pExpr->pRight; 1617 rc = sqlite3Stat4ProbeSetValue(pParse, p, &pRec, pExpr, nTop, nEq, &n); 1618 if( rc==SQLITE_OK && n ){ 1619 tRowcnt iNew; 1620 u16 mask = WO_GT|WO_LE; 1621 if( sqlite3ExprVectorSize(pExpr)>n ) mask = (WO_LE|WO_LT); 1622 iUprIdx = whereKeyStats(pParse, p, pRec, 1, a); 1623 iNew = a[0] + ((pUpper->eOperator & mask) ? a[1] : 0); 1624 if( iNew<iUpper ) iUpper = iNew; 1625 nOut--; 1626 pUpper = 0; 1627 } 1628 } 1629 1630 pBuilder->pRec = pRec; 1631 if( rc==SQLITE_OK ){ 1632 if( iUpper>iLower ){ 1633 nNew = sqlite3LogEst(iUpper - iLower); 1634 /* TUNING: If both iUpper and iLower are derived from the same 1635 ** sample, then assume they are 4x more selective. This brings 1636 ** the estimated selectivity more in line with what it would be 1637 ** if estimated without the use of STAT4 tables. */ 1638 if( iLwrIdx==iUprIdx ) nNew -= 20; assert( 20==sqlite3LogEst(4) ); 1639 }else{ 1640 nNew = 10; assert( 10==sqlite3LogEst(2) ); 1641 } 1642 if( nNew<nOut ){ 1643 nOut = nNew; 1644 } 1645 WHERETRACE(0x10, ("STAT4 range scan: %u..%u est=%d\n", 1646 (u32)iLower, (u32)iUpper, nOut)); 1647 } 1648 }else{ 1649 int bDone = 0; 1650 rc = whereRangeSkipScanEst(pParse, pLower, pUpper, pLoop, &bDone); 1651 if( bDone ) return rc; 1652 } 1653 } 1654 #else 1655 UNUSED_PARAMETER(pParse); 1656 UNUSED_PARAMETER(pBuilder); 1657 assert( pLower || pUpper ); 1658 #endif 1659 assert( pUpper==0 || (pUpper->wtFlags & TERM_VNULL)==0 ); 1660 nNew = whereRangeAdjust(pLower, nOut); 1661 nNew = whereRangeAdjust(pUpper, nNew); 1662 1663 /* TUNING: If there is both an upper and lower limit and neither limit 1664 ** has an application-defined likelihood(), assume the range is 1665 ** reduced by an additional 75%. This means that, by default, an open-ended 1666 ** range query (e.g. col > ?) is assumed to match 1/4 of the rows in the 1667 ** index. While a closed range (e.g. col BETWEEN ? AND ?) is estimated to 1668 ** match 1/64 of the index. */ 1669 if( pLower && pLower->truthProb>0 && pUpper && pUpper->truthProb>0 ){ 1670 nNew -= 20; 1671 } 1672 1673 nOut -= (pLower!=0) + (pUpper!=0); 1674 if( nNew<10 ) nNew = 10; 1675 if( nNew<nOut ) nOut = nNew; 1676 #if defined(WHERETRACE_ENABLED) 1677 if( pLoop->nOut>nOut ){ 1678 WHERETRACE(0x10,("Range scan lowers nOut from %d to %d\n", 1679 pLoop->nOut, nOut)); 1680 } 1681 #endif 1682 pLoop->nOut = (LogEst)nOut; 1683 return rc; 1684 } 1685 1686 #ifdef SQLITE_ENABLE_STAT4 1687 /* 1688 ** Estimate the number of rows that will be returned based on 1689 ** an equality constraint x=VALUE and where that VALUE occurs in 1690 ** the histogram data. This only works when x is the left-most 1691 ** column of an index and sqlite_stat4 histogram data is available 1692 ** for that index. When pExpr==NULL that means the constraint is 1693 ** "x IS NULL" instead of "x=VALUE". 1694 ** 1695 ** Write the estimated row count into *pnRow and return SQLITE_OK. 1696 ** If unable to make an estimate, leave *pnRow unchanged and return 1697 ** non-zero. 1698 ** 1699 ** This routine can fail if it is unable to load a collating sequence 1700 ** required for string comparison, or if unable to allocate memory 1701 ** for a UTF conversion required for comparison. The error is stored 1702 ** in the pParse structure. 1703 */ 1704 static int whereEqualScanEst( 1705 Parse *pParse, /* Parsing & code generating context */ 1706 WhereLoopBuilder *pBuilder, 1707 Expr *pExpr, /* Expression for VALUE in the x=VALUE constraint */ 1708 tRowcnt *pnRow /* Write the revised row estimate here */ 1709 ){ 1710 Index *p = pBuilder->pNew->u.btree.pIndex; 1711 int nEq = pBuilder->pNew->u.btree.nEq; 1712 UnpackedRecord *pRec = pBuilder->pRec; 1713 int rc; /* Subfunction return code */ 1714 tRowcnt a[2]; /* Statistics */ 1715 int bOk; 1716 1717 assert( nEq>=1 ); 1718 assert( nEq<=p->nColumn ); 1719 assert( p->aSample!=0 ); 1720 assert( p->nSample>0 ); 1721 assert( pBuilder->nRecValid<nEq ); 1722 1723 /* If values are not available for all fields of the index to the left 1724 ** of this one, no estimate can be made. Return SQLITE_NOTFOUND. */ 1725 if( pBuilder->nRecValid<(nEq-1) ){ 1726 return SQLITE_NOTFOUND; 1727 } 1728 1729 /* This is an optimization only. The call to sqlite3Stat4ProbeSetValue() 1730 ** below would return the same value. */ 1731 if( nEq>=p->nColumn ){ 1732 *pnRow = 1; 1733 return SQLITE_OK; 1734 } 1735 1736 rc = sqlite3Stat4ProbeSetValue(pParse, p, &pRec, pExpr, 1, nEq-1, &bOk); 1737 pBuilder->pRec = pRec; 1738 if( rc!=SQLITE_OK ) return rc; 1739 if( bOk==0 ) return SQLITE_NOTFOUND; 1740 pBuilder->nRecValid = nEq; 1741 1742 whereKeyStats(pParse, p, pRec, 0, a); 1743 WHERETRACE(0x10,("equality scan regions %s(%d): %d\n", 1744 p->zName, nEq-1, (int)a[1])); 1745 *pnRow = a[1]; 1746 1747 return rc; 1748 } 1749 #endif /* SQLITE_ENABLE_STAT4 */ 1750 1751 #ifdef SQLITE_ENABLE_STAT4 1752 /* 1753 ** Estimate the number of rows that will be returned based on 1754 ** an IN constraint where the right-hand side of the IN operator 1755 ** is a list of values. Example: 1756 ** 1757 ** WHERE x IN (1,2,3,4) 1758 ** 1759 ** Write the estimated row count into *pnRow and return SQLITE_OK. 1760 ** If unable to make an estimate, leave *pnRow unchanged and return 1761 ** non-zero. 1762 ** 1763 ** This routine can fail if it is unable to load a collating sequence 1764 ** required for string comparison, or if unable to allocate memory 1765 ** for a UTF conversion required for comparison. The error is stored 1766 ** in the pParse structure. 1767 */ 1768 static int whereInScanEst( 1769 Parse *pParse, /* Parsing & code generating context */ 1770 WhereLoopBuilder *pBuilder, 1771 ExprList *pList, /* The value list on the RHS of "x IN (v1,v2,v3,...)" */ 1772 tRowcnt *pnRow /* Write the revised row estimate here */ 1773 ){ 1774 Index *p = pBuilder->pNew->u.btree.pIndex; 1775 i64 nRow0 = sqlite3LogEstToInt(p->aiRowLogEst[0]); 1776 int nRecValid = pBuilder->nRecValid; 1777 int rc = SQLITE_OK; /* Subfunction return code */ 1778 tRowcnt nEst; /* Number of rows for a single term */ 1779 tRowcnt nRowEst = 0; /* New estimate of the number of rows */ 1780 int i; /* Loop counter */ 1781 1782 assert( p->aSample!=0 ); 1783 for(i=0; rc==SQLITE_OK && i<pList->nExpr; i++){ 1784 nEst = nRow0; 1785 rc = whereEqualScanEst(pParse, pBuilder, pList->a[i].pExpr, &nEst); 1786 nRowEst += nEst; 1787 pBuilder->nRecValid = nRecValid; 1788 } 1789 1790 if( rc==SQLITE_OK ){ 1791 if( nRowEst > nRow0 ) nRowEst = nRow0; 1792 *pnRow = nRowEst; 1793 WHERETRACE(0x10,("IN row estimate: est=%d\n", nRowEst)); 1794 } 1795 assert( pBuilder->nRecValid==nRecValid ); 1796 return rc; 1797 } 1798 #endif /* SQLITE_ENABLE_STAT4 */ 1799 1800 1801 #ifdef WHERETRACE_ENABLED 1802 /* 1803 ** Print the content of a WhereTerm object 1804 */ 1805 void sqlite3WhereTermPrint(WhereTerm *pTerm, int iTerm){ 1806 if( pTerm==0 ){ 1807 sqlite3DebugPrintf("TERM-%-3d NULL\n", iTerm); 1808 }else{ 1809 char zType[8]; 1810 char zLeft[50]; 1811 memcpy(zType, "....", 5); 1812 if( pTerm->wtFlags & TERM_VIRTUAL ) zType[0] = 'V'; 1813 if( pTerm->eOperator & WO_EQUIV ) zType[1] = 'E'; 1814 if( ExprHasProperty(pTerm->pExpr, EP_FromJoin) ) zType[2] = 'L'; 1815 if( pTerm->wtFlags & TERM_CODED ) zType[3] = 'C'; 1816 if( pTerm->eOperator & WO_SINGLE ){ 1817 assert( (pTerm->eOperator & (WO_OR|WO_AND))==0 ); 1818 sqlite3_snprintf(sizeof(zLeft),zLeft,"left={%d:%d}", 1819 pTerm->leftCursor, pTerm->u.x.leftColumn); 1820 }else if( (pTerm->eOperator & WO_OR)!=0 && pTerm->u.pOrInfo!=0 ){ 1821 sqlite3_snprintf(sizeof(zLeft),zLeft,"indexable=0x%llx", 1822 pTerm->u.pOrInfo->indexable); 1823 }else{ 1824 sqlite3_snprintf(sizeof(zLeft),zLeft,"left=%d", pTerm->leftCursor); 1825 } 1826 sqlite3DebugPrintf( 1827 "TERM-%-3d %p %s %-12s op=%03x wtFlags=%04x", 1828 iTerm, pTerm, zType, zLeft, pTerm->eOperator, pTerm->wtFlags); 1829 /* The 0x10000 .wheretrace flag causes extra information to be 1830 ** shown about each Term */ 1831 if( sqlite3WhereTrace & 0x10000 ){ 1832 sqlite3DebugPrintf(" prob=%-3d prereq=%llx,%llx", 1833 pTerm->truthProb, (u64)pTerm->prereqAll, (u64)pTerm->prereqRight); 1834 } 1835 if( (pTerm->eOperator & (WO_OR|WO_AND))==0 && pTerm->u.x.iField ){ 1836 sqlite3DebugPrintf(" iField=%d", pTerm->u.x.iField); 1837 } 1838 if( pTerm->iParent>=0 ){ 1839 sqlite3DebugPrintf(" iParent=%d", pTerm->iParent); 1840 } 1841 sqlite3DebugPrintf("\n"); 1842 sqlite3TreeViewExpr(0, pTerm->pExpr, 0); 1843 } 1844 } 1845 #endif 1846 1847 #ifdef WHERETRACE_ENABLED 1848 /* 1849 ** Show the complete content of a WhereClause 1850 */ 1851 void sqlite3WhereClausePrint(WhereClause *pWC){ 1852 int i; 1853 for(i=0; i<pWC->nTerm; i++){ 1854 sqlite3WhereTermPrint(&pWC->a[i], i); 1855 } 1856 } 1857 #endif 1858 1859 #ifdef WHERETRACE_ENABLED 1860 /* 1861 ** Print a WhereLoop object for debugging purposes 1862 */ 1863 void sqlite3WhereLoopPrint(WhereLoop *p, WhereClause *pWC){ 1864 WhereInfo *pWInfo = pWC->pWInfo; 1865 int nb = 1+(pWInfo->pTabList->nSrc+3)/4; 1866 SrcItem *pItem = pWInfo->pTabList->a + p->iTab; 1867 Table *pTab = pItem->pTab; 1868 Bitmask mAll = (((Bitmask)1)<<(nb*4)) - 1; 1869 sqlite3DebugPrintf("%c%2d.%0*llx.%0*llx", p->cId, 1870 p->iTab, nb, p->maskSelf, nb, p->prereq & mAll); 1871 sqlite3DebugPrintf(" %12s", 1872 pItem->zAlias ? pItem->zAlias : pTab->zName); 1873 if( (p->wsFlags & WHERE_VIRTUALTABLE)==0 ){ 1874 const char *zName; 1875 if( p->u.btree.pIndex && (zName = p->u.btree.pIndex->zName)!=0 ){ 1876 if( strncmp(zName, "sqlite_autoindex_", 17)==0 ){ 1877 int i = sqlite3Strlen30(zName) - 1; 1878 while( zName[i]!='_' ) i--; 1879 zName += i; 1880 } 1881 sqlite3DebugPrintf(".%-16s %2d", zName, p->u.btree.nEq); 1882 }else{ 1883 sqlite3DebugPrintf("%20s",""); 1884 } 1885 }else{ 1886 char *z; 1887 if( p->u.vtab.idxStr ){ 1888 z = sqlite3_mprintf("(%d,\"%s\",%#x)", 1889 p->u.vtab.idxNum, p->u.vtab.idxStr, p->u.vtab.omitMask); 1890 }else{ 1891 z = sqlite3_mprintf("(%d,%x)", p->u.vtab.idxNum, p->u.vtab.omitMask); 1892 } 1893 sqlite3DebugPrintf(" %-19s", z); 1894 sqlite3_free(z); 1895 } 1896 if( p->wsFlags & WHERE_SKIPSCAN ){ 1897 sqlite3DebugPrintf(" f %05x %d-%d", p->wsFlags, p->nLTerm,p->nSkip); 1898 }else{ 1899 sqlite3DebugPrintf(" f %05x N %d", p->wsFlags, p->nLTerm); 1900 } 1901 sqlite3DebugPrintf(" cost %d,%d,%d\n", p->rSetup, p->rRun, p->nOut); 1902 if( p->nLTerm && (sqlite3WhereTrace & 0x100)!=0 ){ 1903 int i; 1904 for(i=0; i<p->nLTerm; i++){ 1905 sqlite3WhereTermPrint(p->aLTerm[i], i); 1906 } 1907 } 1908 } 1909 #endif 1910 1911 /* 1912 ** Convert bulk memory into a valid WhereLoop that can be passed 1913 ** to whereLoopClear harmlessly. 1914 */ 1915 static void whereLoopInit(WhereLoop *p){ 1916 p->aLTerm = p->aLTermSpace; 1917 p->nLTerm = 0; 1918 p->nLSlot = ArraySize(p->aLTermSpace); 1919 p->wsFlags = 0; 1920 } 1921 1922 /* 1923 ** Clear the WhereLoop.u union. Leave WhereLoop.pLTerm intact. 1924 */ 1925 static void whereLoopClearUnion(sqlite3 *db, WhereLoop *p){ 1926 if( p->wsFlags & (WHERE_VIRTUALTABLE|WHERE_AUTO_INDEX) ){ 1927 if( (p->wsFlags & WHERE_VIRTUALTABLE)!=0 && p->u.vtab.needFree ){ 1928 sqlite3_free(p->u.vtab.idxStr); 1929 p->u.vtab.needFree = 0; 1930 p->u.vtab.idxStr = 0; 1931 }else if( (p->wsFlags & WHERE_AUTO_INDEX)!=0 && p->u.btree.pIndex!=0 ){ 1932 sqlite3DbFree(db, p->u.btree.pIndex->zColAff); 1933 sqlite3DbFreeNN(db, p->u.btree.pIndex); 1934 p->u.btree.pIndex = 0; 1935 } 1936 } 1937 } 1938 1939 /* 1940 ** Deallocate internal memory used by a WhereLoop object 1941 */ 1942 static void whereLoopClear(sqlite3 *db, WhereLoop *p){ 1943 if( p->aLTerm!=p->aLTermSpace ) sqlite3DbFreeNN(db, p->aLTerm); 1944 whereLoopClearUnion(db, p); 1945 whereLoopInit(p); 1946 } 1947 1948 /* 1949 ** Increase the memory allocation for pLoop->aLTerm[] to be at least n. 1950 */ 1951 static int whereLoopResize(sqlite3 *db, WhereLoop *p, int n){ 1952 WhereTerm **paNew; 1953 if( p->nLSlot>=n ) return SQLITE_OK; 1954 n = (n+7)&~7; 1955 paNew = sqlite3DbMallocRawNN(db, sizeof(p->aLTerm[0])*n); 1956 if( paNew==0 ) return SQLITE_NOMEM_BKPT; 1957 memcpy(paNew, p->aLTerm, sizeof(p->aLTerm[0])*p->nLSlot); 1958 if( p->aLTerm!=p->aLTermSpace ) sqlite3DbFreeNN(db, p->aLTerm); 1959 p->aLTerm = paNew; 1960 p->nLSlot = n; 1961 return SQLITE_OK; 1962 } 1963 1964 /* 1965 ** Transfer content from the second pLoop into the first. 1966 */ 1967 static int whereLoopXfer(sqlite3 *db, WhereLoop *pTo, WhereLoop *pFrom){ 1968 whereLoopClearUnion(db, pTo); 1969 if( whereLoopResize(db, pTo, pFrom->nLTerm) ){ 1970 memset(pTo, 0, WHERE_LOOP_XFER_SZ); 1971 return SQLITE_NOMEM_BKPT; 1972 } 1973 memcpy(pTo, pFrom, WHERE_LOOP_XFER_SZ); 1974 memcpy(pTo->aLTerm, pFrom->aLTerm, pTo->nLTerm*sizeof(pTo->aLTerm[0])); 1975 if( pFrom->wsFlags & WHERE_VIRTUALTABLE ){ 1976 pFrom->u.vtab.needFree = 0; 1977 }else if( (pFrom->wsFlags & WHERE_AUTO_INDEX)!=0 ){ 1978 pFrom->u.btree.pIndex = 0; 1979 } 1980 return SQLITE_OK; 1981 } 1982 1983 /* 1984 ** Delete a WhereLoop object 1985 */ 1986 static void whereLoopDelete(sqlite3 *db, WhereLoop *p){ 1987 whereLoopClear(db, p); 1988 sqlite3DbFreeNN(db, p); 1989 } 1990 1991 /* 1992 ** Free a WhereInfo structure 1993 */ 1994 static void whereInfoFree(sqlite3 *db, WhereInfo *pWInfo){ 1995 int i; 1996 assert( pWInfo!=0 ); 1997 for(i=0; i<pWInfo->nLevel; i++){ 1998 WhereLevel *pLevel = &pWInfo->a[i]; 1999 if( pLevel->pWLoop && (pLevel->pWLoop->wsFlags & WHERE_IN_ABLE)!=0 ){ 2000 assert( (pLevel->pWLoop->wsFlags & WHERE_MULTI_OR)==0 ); 2001 sqlite3DbFree(db, pLevel->u.in.aInLoop); 2002 } 2003 } 2004 sqlite3WhereClauseClear(&pWInfo->sWC); 2005 while( pWInfo->pLoops ){ 2006 WhereLoop *p = pWInfo->pLoops; 2007 pWInfo->pLoops = p->pNextLoop; 2008 whereLoopDelete(db, p); 2009 } 2010 assert( pWInfo->pExprMods==0 ); 2011 sqlite3DbFreeNN(db, pWInfo); 2012 } 2013 2014 /* Undo all Expr node modifications 2015 */ 2016 static void whereUndoExprMods(WhereInfo *pWInfo){ 2017 while( pWInfo->pExprMods ){ 2018 WhereExprMod *p = pWInfo->pExprMods; 2019 pWInfo->pExprMods = p->pNext; 2020 memcpy(p->pExpr, &p->orig, sizeof(p->orig)); 2021 sqlite3DbFree(pWInfo->pParse->db, p); 2022 } 2023 } 2024 2025 /* 2026 ** Return TRUE if all of the following are true: 2027 ** 2028 ** (1) X has the same or lower cost, or returns the same or fewer rows, 2029 ** than Y. 2030 ** (2) X uses fewer WHERE clause terms than Y 2031 ** (3) Every WHERE clause term used by X is also used by Y 2032 ** (4) X skips at least as many columns as Y 2033 ** (5) If X is a covering index, than Y is too 2034 ** 2035 ** Conditions (2) and (3) mean that X is a "proper subset" of Y. 2036 ** If X is a proper subset of Y then Y is a better choice and ought 2037 ** to have a lower cost. This routine returns TRUE when that cost 2038 ** relationship is inverted and needs to be adjusted. Constraint (4) 2039 ** was added because if X uses skip-scan less than Y it still might 2040 ** deserve a lower cost even if it is a proper subset of Y. Constraint (5) 2041 ** was added because a covering index probably deserves to have a lower cost 2042 ** than a non-covering index even if it is a proper subset. 2043 */ 2044 static int whereLoopCheaperProperSubset( 2045 const WhereLoop *pX, /* First WhereLoop to compare */ 2046 const WhereLoop *pY /* Compare against this WhereLoop */ 2047 ){ 2048 int i, j; 2049 if( pX->nLTerm-pX->nSkip >= pY->nLTerm-pY->nSkip ){ 2050 return 0; /* X is not a subset of Y */ 2051 } 2052 if( pX->rRun>pY->rRun && pX->nOut>pY->nOut ) return 0; 2053 if( pY->nSkip > pX->nSkip ) return 0; 2054 for(i=pX->nLTerm-1; i>=0; i--){ 2055 if( pX->aLTerm[i]==0 ) continue; 2056 for(j=pY->nLTerm-1; j>=0; j--){ 2057 if( pY->aLTerm[j]==pX->aLTerm[i] ) break; 2058 } 2059 if( j<0 ) return 0; /* X not a subset of Y since term X[i] not used by Y */ 2060 } 2061 if( (pX->wsFlags&WHERE_IDX_ONLY)!=0 2062 && (pY->wsFlags&WHERE_IDX_ONLY)==0 ){ 2063 return 0; /* Constraint (5) */ 2064 } 2065 return 1; /* All conditions meet */ 2066 } 2067 2068 /* 2069 ** Try to adjust the cost and number of output rows of WhereLoop pTemplate 2070 ** upwards or downwards so that: 2071 ** 2072 ** (1) pTemplate costs less than any other WhereLoops that are a proper 2073 ** subset of pTemplate 2074 ** 2075 ** (2) pTemplate costs more than any other WhereLoops for which pTemplate 2076 ** is a proper subset. 2077 ** 2078 ** To say "WhereLoop X is a proper subset of Y" means that X uses fewer 2079 ** WHERE clause terms than Y and that every WHERE clause term used by X is 2080 ** also used by Y. 2081 */ 2082 static void whereLoopAdjustCost(const WhereLoop *p, WhereLoop *pTemplate){ 2083 if( (pTemplate->wsFlags & WHERE_INDEXED)==0 ) return; 2084 for(; p; p=p->pNextLoop){ 2085 if( p->iTab!=pTemplate->iTab ) continue; 2086 if( (p->wsFlags & WHERE_INDEXED)==0 ) continue; 2087 if( whereLoopCheaperProperSubset(p, pTemplate) ){ 2088 /* Adjust pTemplate cost downward so that it is cheaper than its 2089 ** subset p. */ 2090 WHERETRACE(0x80,("subset cost adjustment %d,%d to %d,%d\n", 2091 pTemplate->rRun, pTemplate->nOut, 2092 MIN(p->rRun, pTemplate->rRun), 2093 MIN(p->nOut - 1, pTemplate->nOut))); 2094 pTemplate->rRun = MIN(p->rRun, pTemplate->rRun); 2095 pTemplate->nOut = MIN(p->nOut - 1, pTemplate->nOut); 2096 }else if( whereLoopCheaperProperSubset(pTemplate, p) ){ 2097 /* Adjust pTemplate cost upward so that it is costlier than p since 2098 ** pTemplate is a proper subset of p */ 2099 WHERETRACE(0x80,("subset cost adjustment %d,%d to %d,%d\n", 2100 pTemplate->rRun, pTemplate->nOut, 2101 MAX(p->rRun, pTemplate->rRun), 2102 MAX(p->nOut + 1, pTemplate->nOut))); 2103 pTemplate->rRun = MAX(p->rRun, pTemplate->rRun); 2104 pTemplate->nOut = MAX(p->nOut + 1, pTemplate->nOut); 2105 } 2106 } 2107 } 2108 2109 /* 2110 ** Search the list of WhereLoops in *ppPrev looking for one that can be 2111 ** replaced by pTemplate. 2112 ** 2113 ** Return NULL if pTemplate does not belong on the WhereLoop list. 2114 ** In other words if pTemplate ought to be dropped from further consideration. 2115 ** 2116 ** If pX is a WhereLoop that pTemplate can replace, then return the 2117 ** link that points to pX. 2118 ** 2119 ** If pTemplate cannot replace any existing element of the list but needs 2120 ** to be added to the list as a new entry, then return a pointer to the 2121 ** tail of the list. 2122 */ 2123 static WhereLoop **whereLoopFindLesser( 2124 WhereLoop **ppPrev, 2125 const WhereLoop *pTemplate 2126 ){ 2127 WhereLoop *p; 2128 for(p=(*ppPrev); p; ppPrev=&p->pNextLoop, p=*ppPrev){ 2129 if( p->iTab!=pTemplate->iTab || p->iSortIdx!=pTemplate->iSortIdx ){ 2130 /* If either the iTab or iSortIdx values for two WhereLoop are different 2131 ** then those WhereLoops need to be considered separately. Neither is 2132 ** a candidate to replace the other. */ 2133 continue; 2134 } 2135 /* In the current implementation, the rSetup value is either zero 2136 ** or the cost of building an automatic index (NlogN) and the NlogN 2137 ** is the same for compatible WhereLoops. */ 2138 assert( p->rSetup==0 || pTemplate->rSetup==0 2139 || p->rSetup==pTemplate->rSetup ); 2140 2141 /* whereLoopAddBtree() always generates and inserts the automatic index 2142 ** case first. Hence compatible candidate WhereLoops never have a larger 2143 ** rSetup. Call this SETUP-INVARIANT */ 2144 assert( p->rSetup>=pTemplate->rSetup ); 2145 2146 /* Any loop using an appliation-defined index (or PRIMARY KEY or 2147 ** UNIQUE constraint) with one or more == constraints is better 2148 ** than an automatic index. Unless it is a skip-scan. */ 2149 if( (p->wsFlags & WHERE_AUTO_INDEX)!=0 2150 && (pTemplate->nSkip)==0 2151 && (pTemplate->wsFlags & WHERE_INDEXED)!=0 2152 && (pTemplate->wsFlags & WHERE_COLUMN_EQ)!=0 2153 && (p->prereq & pTemplate->prereq)==pTemplate->prereq 2154 ){ 2155 break; 2156 } 2157 2158 /* If existing WhereLoop p is better than pTemplate, pTemplate can be 2159 ** discarded. WhereLoop p is better if: 2160 ** (1) p has no more dependencies than pTemplate, and 2161 ** (2) p has an equal or lower cost than pTemplate 2162 */ 2163 if( (p->prereq & pTemplate->prereq)==p->prereq /* (1) */ 2164 && p->rSetup<=pTemplate->rSetup /* (2a) */ 2165 && p->rRun<=pTemplate->rRun /* (2b) */ 2166 && p->nOut<=pTemplate->nOut /* (2c) */ 2167 ){ 2168 return 0; /* Discard pTemplate */ 2169 } 2170 2171 /* If pTemplate is always better than p, then cause p to be overwritten 2172 ** with pTemplate. pTemplate is better than p if: 2173 ** (1) pTemplate has no more dependences than p, and 2174 ** (2) pTemplate has an equal or lower cost than p. 2175 */ 2176 if( (p->prereq & pTemplate->prereq)==pTemplate->prereq /* (1) */ 2177 && p->rRun>=pTemplate->rRun /* (2a) */ 2178 && p->nOut>=pTemplate->nOut /* (2b) */ 2179 ){ 2180 assert( p->rSetup>=pTemplate->rSetup ); /* SETUP-INVARIANT above */ 2181 break; /* Cause p to be overwritten by pTemplate */ 2182 } 2183 } 2184 return ppPrev; 2185 } 2186 2187 /* 2188 ** Insert or replace a WhereLoop entry using the template supplied. 2189 ** 2190 ** An existing WhereLoop entry might be overwritten if the new template 2191 ** is better and has fewer dependencies. Or the template will be ignored 2192 ** and no insert will occur if an existing WhereLoop is faster and has 2193 ** fewer dependencies than the template. Otherwise a new WhereLoop is 2194 ** added based on the template. 2195 ** 2196 ** If pBuilder->pOrSet is not NULL then we care about only the 2197 ** prerequisites and rRun and nOut costs of the N best loops. That 2198 ** information is gathered in the pBuilder->pOrSet object. This special 2199 ** processing mode is used only for OR clause processing. 2200 ** 2201 ** When accumulating multiple loops (when pBuilder->pOrSet is NULL) we 2202 ** still might overwrite similar loops with the new template if the 2203 ** new template is better. Loops may be overwritten if the following 2204 ** conditions are met: 2205 ** 2206 ** (1) They have the same iTab. 2207 ** (2) They have the same iSortIdx. 2208 ** (3) The template has same or fewer dependencies than the current loop 2209 ** (4) The template has the same or lower cost than the current loop 2210 */ 2211 static int whereLoopInsert(WhereLoopBuilder *pBuilder, WhereLoop *pTemplate){ 2212 WhereLoop **ppPrev, *p; 2213 WhereInfo *pWInfo = pBuilder->pWInfo; 2214 sqlite3 *db = pWInfo->pParse->db; 2215 int rc; 2216 2217 /* Stop the search once we hit the query planner search limit */ 2218 if( pBuilder->iPlanLimit==0 ){ 2219 WHERETRACE(0xffffffff,("=== query planner search limit reached ===\n")); 2220 if( pBuilder->pOrSet ) pBuilder->pOrSet->n = 0; 2221 return SQLITE_DONE; 2222 } 2223 pBuilder->iPlanLimit--; 2224 2225 whereLoopAdjustCost(pWInfo->pLoops, pTemplate); 2226 2227 /* If pBuilder->pOrSet is defined, then only keep track of the costs 2228 ** and prereqs. 2229 */ 2230 if( pBuilder->pOrSet!=0 ){ 2231 if( pTemplate->nLTerm ){ 2232 #if WHERETRACE_ENABLED 2233 u16 n = pBuilder->pOrSet->n; 2234 int x = 2235 #endif 2236 whereOrInsert(pBuilder->pOrSet, pTemplate->prereq, pTemplate->rRun, 2237 pTemplate->nOut); 2238 #if WHERETRACE_ENABLED /* 0x8 */ 2239 if( sqlite3WhereTrace & 0x8 ){ 2240 sqlite3DebugPrintf(x?" or-%d: ":" or-X: ", n); 2241 sqlite3WhereLoopPrint(pTemplate, pBuilder->pWC); 2242 } 2243 #endif 2244 } 2245 return SQLITE_OK; 2246 } 2247 2248 /* Look for an existing WhereLoop to replace with pTemplate 2249 */ 2250 ppPrev = whereLoopFindLesser(&pWInfo->pLoops, pTemplate); 2251 2252 if( ppPrev==0 ){ 2253 /* There already exists a WhereLoop on the list that is better 2254 ** than pTemplate, so just ignore pTemplate */ 2255 #if WHERETRACE_ENABLED /* 0x8 */ 2256 if( sqlite3WhereTrace & 0x8 ){ 2257 sqlite3DebugPrintf(" skip: "); 2258 sqlite3WhereLoopPrint(pTemplate, pBuilder->pWC); 2259 } 2260 #endif 2261 return SQLITE_OK; 2262 }else{ 2263 p = *ppPrev; 2264 } 2265 2266 /* If we reach this point it means that either p[] should be overwritten 2267 ** with pTemplate[] if p[] exists, or if p==NULL then allocate a new 2268 ** WhereLoop and insert it. 2269 */ 2270 #if WHERETRACE_ENABLED /* 0x8 */ 2271 if( sqlite3WhereTrace & 0x8 ){ 2272 if( p!=0 ){ 2273 sqlite3DebugPrintf("replace: "); 2274 sqlite3WhereLoopPrint(p, pBuilder->pWC); 2275 sqlite3DebugPrintf(" with: "); 2276 }else{ 2277 sqlite3DebugPrintf(" add: "); 2278 } 2279 sqlite3WhereLoopPrint(pTemplate, pBuilder->pWC); 2280 } 2281 #endif 2282 if( p==0 ){ 2283 /* Allocate a new WhereLoop to add to the end of the list */ 2284 *ppPrev = p = sqlite3DbMallocRawNN(db, sizeof(WhereLoop)); 2285 if( p==0 ) return SQLITE_NOMEM_BKPT; 2286 whereLoopInit(p); 2287 p->pNextLoop = 0; 2288 }else{ 2289 /* We will be overwriting WhereLoop p[]. But before we do, first 2290 ** go through the rest of the list and delete any other entries besides 2291 ** p[] that are also supplated by pTemplate */ 2292 WhereLoop **ppTail = &p->pNextLoop; 2293 WhereLoop *pToDel; 2294 while( *ppTail ){ 2295 ppTail = whereLoopFindLesser(ppTail, pTemplate); 2296 if( ppTail==0 ) break; 2297 pToDel = *ppTail; 2298 if( pToDel==0 ) break; 2299 *ppTail = pToDel->pNextLoop; 2300 #if WHERETRACE_ENABLED /* 0x8 */ 2301 if( sqlite3WhereTrace & 0x8 ){ 2302 sqlite3DebugPrintf(" delete: "); 2303 sqlite3WhereLoopPrint(pToDel, pBuilder->pWC); 2304 } 2305 #endif 2306 whereLoopDelete(db, pToDel); 2307 } 2308 } 2309 rc = whereLoopXfer(db, p, pTemplate); 2310 if( (p->wsFlags & WHERE_VIRTUALTABLE)==0 ){ 2311 Index *pIndex = p->u.btree.pIndex; 2312 if( pIndex && pIndex->idxType==SQLITE_IDXTYPE_IPK ){ 2313 p->u.btree.pIndex = 0; 2314 } 2315 } 2316 return rc; 2317 } 2318 2319 /* 2320 ** Adjust the WhereLoop.nOut value downward to account for terms of the 2321 ** WHERE clause that reference the loop but which are not used by an 2322 ** index. 2323 * 2324 ** For every WHERE clause term that is not used by the index 2325 ** and which has a truth probability assigned by one of the likelihood(), 2326 ** likely(), or unlikely() SQL functions, reduce the estimated number 2327 ** of output rows by the probability specified. 2328 ** 2329 ** TUNING: For every WHERE clause term that is not used by the index 2330 ** and which does not have an assigned truth probability, heuristics 2331 ** described below are used to try to estimate the truth probability. 2332 ** TODO --> Perhaps this is something that could be improved by better 2333 ** table statistics. 2334 ** 2335 ** Heuristic 1: Estimate the truth probability as 93.75%. The 93.75% 2336 ** value corresponds to -1 in LogEst notation, so this means decrement 2337 ** the WhereLoop.nOut field for every such WHERE clause term. 2338 ** 2339 ** Heuristic 2: If there exists one or more WHERE clause terms of the 2340 ** form "x==EXPR" and EXPR is not a constant 0 or 1, then make sure the 2341 ** final output row estimate is no greater than 1/4 of the total number 2342 ** of rows in the table. In other words, assume that x==EXPR will filter 2343 ** out at least 3 out of 4 rows. If EXPR is -1 or 0 or 1, then maybe the 2344 ** "x" column is boolean or else -1 or 0 or 1 is a common default value 2345 ** on the "x" column and so in that case only cap the output row estimate 2346 ** at 1/2 instead of 1/4. 2347 */ 2348 static void whereLoopOutputAdjust( 2349 WhereClause *pWC, /* The WHERE clause */ 2350 WhereLoop *pLoop, /* The loop to adjust downward */ 2351 LogEst nRow /* Number of rows in the entire table */ 2352 ){ 2353 WhereTerm *pTerm, *pX; 2354 Bitmask notAllowed = ~(pLoop->prereq|pLoop->maskSelf); 2355 int i, j; 2356 LogEst iReduce = 0; /* pLoop->nOut should not exceed nRow-iReduce */ 2357 2358 assert( (pLoop->wsFlags & WHERE_AUTO_INDEX)==0 ); 2359 for(i=pWC->nTerm, pTerm=pWC->a; i>0; i--, pTerm++){ 2360 assert( pTerm!=0 ); 2361 if( (pTerm->wtFlags & TERM_VIRTUAL)!=0 ) break; 2362 if( (pTerm->prereqAll & pLoop->maskSelf)==0 ) continue; 2363 if( (pTerm->prereqAll & notAllowed)!=0 ) continue; 2364 for(j=pLoop->nLTerm-1; j>=0; j--){ 2365 pX = pLoop->aLTerm[j]; 2366 if( pX==0 ) continue; 2367 if( pX==pTerm ) break; 2368 if( pX->iParent>=0 && (&pWC->a[pX->iParent])==pTerm ) break; 2369 } 2370 if( j<0 ){ 2371 if( pTerm->truthProb<=0 ){ 2372 /* If a truth probability is specified using the likelihood() hints, 2373 ** then use the probability provided by the application. */ 2374 pLoop->nOut += pTerm->truthProb; 2375 }else{ 2376 /* In the absence of explicit truth probabilities, use heuristics to 2377 ** guess a reasonable truth probability. */ 2378 pLoop->nOut--; 2379 if( (pTerm->eOperator&(WO_EQ|WO_IS))!=0 2380 && (pTerm->wtFlags & TERM_HIGHTRUTH)==0 /* tag-20200224-1 */ 2381 ){ 2382 Expr *pRight = pTerm->pExpr->pRight; 2383 int k = 0; 2384 testcase( pTerm->pExpr->op==TK_IS ); 2385 if( sqlite3ExprIsInteger(pRight, &k) && k>=(-1) && k<=1 ){ 2386 k = 10; 2387 }else{ 2388 k = 20; 2389 } 2390 if( iReduce<k ){ 2391 pTerm->wtFlags |= TERM_HEURTRUTH; 2392 iReduce = k; 2393 } 2394 } 2395 } 2396 } 2397 } 2398 if( pLoop->nOut > nRow-iReduce ) pLoop->nOut = nRow - iReduce; 2399 } 2400 2401 /* 2402 ** Term pTerm is a vector range comparison operation. The first comparison 2403 ** in the vector can be optimized using column nEq of the index. This 2404 ** function returns the total number of vector elements that can be used 2405 ** as part of the range comparison. 2406 ** 2407 ** For example, if the query is: 2408 ** 2409 ** WHERE a = ? AND (b, c, d) > (?, ?, ?) 2410 ** 2411 ** and the index: 2412 ** 2413 ** CREATE INDEX ... ON (a, b, c, d, e) 2414 ** 2415 ** then this function would be invoked with nEq=1. The value returned in 2416 ** this case is 3. 2417 */ 2418 static int whereRangeVectorLen( 2419 Parse *pParse, /* Parsing context */ 2420 int iCur, /* Cursor open on pIdx */ 2421 Index *pIdx, /* The index to be used for a inequality constraint */ 2422 int nEq, /* Number of prior equality constraints on same index */ 2423 WhereTerm *pTerm /* The vector inequality constraint */ 2424 ){ 2425 int nCmp = sqlite3ExprVectorSize(pTerm->pExpr->pLeft); 2426 int i; 2427 2428 nCmp = MIN(nCmp, (pIdx->nColumn - nEq)); 2429 for(i=1; i<nCmp; i++){ 2430 /* Test if comparison i of pTerm is compatible with column (i+nEq) 2431 ** of the index. If not, exit the loop. */ 2432 char aff; /* Comparison affinity */ 2433 char idxaff = 0; /* Indexed columns affinity */ 2434 CollSeq *pColl; /* Comparison collation sequence */ 2435 Expr *pLhs, *pRhs; 2436 2437 assert( ExprUseXList(pTerm->pExpr->pLeft) ); 2438 pLhs = pTerm->pExpr->pLeft->x.pList->a[i].pExpr; 2439 pRhs = pTerm->pExpr->pRight; 2440 if( ExprUseXSelect(pRhs) ){ 2441 pRhs = pRhs->x.pSelect->pEList->a[i].pExpr; 2442 }else{ 2443 pRhs = pRhs->x.pList->a[i].pExpr; 2444 } 2445 2446 /* Check that the LHS of the comparison is a column reference to 2447 ** the right column of the right source table. And that the sort 2448 ** order of the index column is the same as the sort order of the 2449 ** leftmost index column. */ 2450 if( pLhs->op!=TK_COLUMN 2451 || pLhs->iTable!=iCur 2452 || pLhs->iColumn!=pIdx->aiColumn[i+nEq] 2453 || pIdx->aSortOrder[i+nEq]!=pIdx->aSortOrder[nEq] 2454 ){ 2455 break; 2456 } 2457 2458 testcase( pLhs->iColumn==XN_ROWID ); 2459 aff = sqlite3CompareAffinity(pRhs, sqlite3ExprAffinity(pLhs)); 2460 idxaff = sqlite3TableColumnAffinity(pIdx->pTable, pLhs->iColumn); 2461 if( aff!=idxaff ) break; 2462 2463 pColl = sqlite3BinaryCompareCollSeq(pParse, pLhs, pRhs); 2464 if( pColl==0 ) break; 2465 if( sqlite3StrICmp(pColl->zName, pIdx->azColl[i+nEq]) ) break; 2466 } 2467 return i; 2468 } 2469 2470 /* 2471 ** Adjust the cost C by the costMult facter T. This only occurs if 2472 ** compiled with -DSQLITE_ENABLE_COSTMULT 2473 */ 2474 #ifdef SQLITE_ENABLE_COSTMULT 2475 # define ApplyCostMultiplier(C,T) C += T 2476 #else 2477 # define ApplyCostMultiplier(C,T) 2478 #endif 2479 2480 /* 2481 ** We have so far matched pBuilder->pNew->u.btree.nEq terms of the 2482 ** index pIndex. Try to match one more. 2483 ** 2484 ** When this function is called, pBuilder->pNew->nOut contains the 2485 ** number of rows expected to be visited by filtering using the nEq 2486 ** terms only. If it is modified, this value is restored before this 2487 ** function returns. 2488 ** 2489 ** If pProbe->idxType==SQLITE_IDXTYPE_IPK, that means pIndex is 2490 ** a fake index used for the INTEGER PRIMARY KEY. 2491 */ 2492 static int whereLoopAddBtreeIndex( 2493 WhereLoopBuilder *pBuilder, /* The WhereLoop factory */ 2494 SrcItem *pSrc, /* FROM clause term being analyzed */ 2495 Index *pProbe, /* An index on pSrc */ 2496 LogEst nInMul /* log(Number of iterations due to IN) */ 2497 ){ 2498 WhereInfo *pWInfo = pBuilder->pWInfo; /* WHERE analyse context */ 2499 Parse *pParse = pWInfo->pParse; /* Parsing context */ 2500 sqlite3 *db = pParse->db; /* Database connection malloc context */ 2501 WhereLoop *pNew; /* Template WhereLoop under construction */ 2502 WhereTerm *pTerm; /* A WhereTerm under consideration */ 2503 int opMask; /* Valid operators for constraints */ 2504 WhereScan scan; /* Iterator for WHERE terms */ 2505 Bitmask saved_prereq; /* Original value of pNew->prereq */ 2506 u16 saved_nLTerm; /* Original value of pNew->nLTerm */ 2507 u16 saved_nEq; /* Original value of pNew->u.btree.nEq */ 2508 u16 saved_nBtm; /* Original value of pNew->u.btree.nBtm */ 2509 u16 saved_nTop; /* Original value of pNew->u.btree.nTop */ 2510 u16 saved_nSkip; /* Original value of pNew->nSkip */ 2511 u32 saved_wsFlags; /* Original value of pNew->wsFlags */ 2512 LogEst saved_nOut; /* Original value of pNew->nOut */ 2513 int rc = SQLITE_OK; /* Return code */ 2514 LogEst rSize; /* Number of rows in the table */ 2515 LogEst rLogSize; /* Logarithm of table size */ 2516 WhereTerm *pTop = 0, *pBtm = 0; /* Top and bottom range constraints */ 2517 2518 pNew = pBuilder->pNew; 2519 if( db->mallocFailed ) return SQLITE_NOMEM_BKPT; 2520 WHERETRACE(0x800, ("BEGIN %s.addBtreeIdx(%s), nEq=%d, nSkip=%d, rRun=%d\n", 2521 pProbe->pTable->zName,pProbe->zName, 2522 pNew->u.btree.nEq, pNew->nSkip, pNew->rRun)); 2523 2524 assert( (pNew->wsFlags & WHERE_VIRTUALTABLE)==0 ); 2525 assert( (pNew->wsFlags & WHERE_TOP_LIMIT)==0 ); 2526 if( pNew->wsFlags & WHERE_BTM_LIMIT ){ 2527 opMask = WO_LT|WO_LE; 2528 }else{ 2529 assert( pNew->u.btree.nBtm==0 ); 2530 opMask = WO_EQ|WO_IN|WO_GT|WO_GE|WO_LT|WO_LE|WO_ISNULL|WO_IS; 2531 } 2532 if( pProbe->bUnordered ) opMask &= ~(WO_GT|WO_GE|WO_LT|WO_LE); 2533 2534 assert( pNew->u.btree.nEq<pProbe->nColumn ); 2535 assert( pNew->u.btree.nEq<pProbe->nKeyCol 2536 || pProbe->idxType!=SQLITE_IDXTYPE_PRIMARYKEY ); 2537 2538 saved_nEq = pNew->u.btree.nEq; 2539 saved_nBtm = pNew->u.btree.nBtm; 2540 saved_nTop = pNew->u.btree.nTop; 2541 saved_nSkip = pNew->nSkip; 2542 saved_nLTerm = pNew->nLTerm; 2543 saved_wsFlags = pNew->wsFlags; 2544 saved_prereq = pNew->prereq; 2545 saved_nOut = pNew->nOut; 2546 pTerm = whereScanInit(&scan, pBuilder->pWC, pSrc->iCursor, saved_nEq, 2547 opMask, pProbe); 2548 pNew->rSetup = 0; 2549 rSize = pProbe->aiRowLogEst[0]; 2550 rLogSize = estLog(rSize); 2551 for(; rc==SQLITE_OK && pTerm!=0; pTerm = whereScanNext(&scan)){ 2552 u16 eOp = pTerm->eOperator; /* Shorthand for pTerm->eOperator */ 2553 LogEst rCostIdx; 2554 LogEst nOutUnadjusted; /* nOut before IN() and WHERE adjustments */ 2555 int nIn = 0; 2556 #ifdef SQLITE_ENABLE_STAT4 2557 int nRecValid = pBuilder->nRecValid; 2558 #endif 2559 if( (eOp==WO_ISNULL || (pTerm->wtFlags&TERM_VNULL)!=0) 2560 && indexColumnNotNull(pProbe, saved_nEq) 2561 ){ 2562 continue; /* ignore IS [NOT] NULL constraints on NOT NULL columns */ 2563 } 2564 if( pTerm->prereqRight & pNew->maskSelf ) continue; 2565 2566 /* Do not allow the upper bound of a LIKE optimization range constraint 2567 ** to mix with a lower range bound from some other source */ 2568 if( pTerm->wtFlags & TERM_LIKEOPT && pTerm->eOperator==WO_LT ) continue; 2569 2570 /* tag-20191211-001: Do not allow constraints from the WHERE clause to 2571 ** be used by the right table of a LEFT JOIN. Only constraints in the 2572 ** ON clause are allowed. See tag-20191211-002 for the vtab equivalent. */ 2573 if( (pSrc->fg.jointype & JT_LEFT)!=0 2574 && !ExprHasProperty(pTerm->pExpr, EP_FromJoin) 2575 ){ 2576 continue; 2577 } 2578 2579 if( IsUniqueIndex(pProbe) && saved_nEq==pProbe->nKeyCol-1 ){ 2580 pBuilder->bldFlags1 |= SQLITE_BLDF1_UNIQUE; 2581 }else{ 2582 pBuilder->bldFlags1 |= SQLITE_BLDF1_INDEXED; 2583 } 2584 pNew->wsFlags = saved_wsFlags; 2585 pNew->u.btree.nEq = saved_nEq; 2586 pNew->u.btree.nBtm = saved_nBtm; 2587 pNew->u.btree.nTop = saved_nTop; 2588 pNew->nLTerm = saved_nLTerm; 2589 if( whereLoopResize(db, pNew, pNew->nLTerm+1) ) break; /* OOM */ 2590 pNew->aLTerm[pNew->nLTerm++] = pTerm; 2591 pNew->prereq = (saved_prereq | pTerm->prereqRight) & ~pNew->maskSelf; 2592 2593 assert( nInMul==0 2594 || (pNew->wsFlags & WHERE_COLUMN_NULL)!=0 2595 || (pNew->wsFlags & WHERE_COLUMN_IN)!=0 2596 || (pNew->wsFlags & WHERE_SKIPSCAN)!=0 2597 ); 2598 2599 if( eOp & WO_IN ){ 2600 Expr *pExpr = pTerm->pExpr; 2601 if( ExprUseXSelect(pExpr) ){ 2602 /* "x IN (SELECT ...)": TUNING: the SELECT returns 25 rows */ 2603 int i; 2604 nIn = 46; assert( 46==sqlite3LogEst(25) ); 2605 2606 /* The expression may actually be of the form (x, y) IN (SELECT...). 2607 ** In this case there is a separate term for each of (x) and (y). 2608 ** However, the nIn multiplier should only be applied once, not once 2609 ** for each such term. The following loop checks that pTerm is the 2610 ** first such term in use, and sets nIn back to 0 if it is not. */ 2611 for(i=0; i<pNew->nLTerm-1; i++){ 2612 if( pNew->aLTerm[i] && pNew->aLTerm[i]->pExpr==pExpr ) nIn = 0; 2613 } 2614 }else if( ALWAYS(pExpr->x.pList && pExpr->x.pList->nExpr) ){ 2615 /* "x IN (value, value, ...)" */ 2616 nIn = sqlite3LogEst(pExpr->x.pList->nExpr); 2617 } 2618 if( pProbe->hasStat1 && rLogSize>=10 ){ 2619 LogEst M, logK, x; 2620 /* Let: 2621 ** N = the total number of rows in the table 2622 ** K = the number of entries on the RHS of the IN operator 2623 ** M = the number of rows in the table that match terms to the 2624 ** to the left in the same index. If the IN operator is on 2625 ** the left-most index column, M==N. 2626 ** 2627 ** Given the definitions above, it is better to omit the IN operator 2628 ** from the index lookup and instead do a scan of the M elements, 2629 ** testing each scanned row against the IN operator separately, if: 2630 ** 2631 ** M*log(K) < K*log(N) 2632 ** 2633 ** Our estimates for M, K, and N might be inaccurate, so we build in 2634 ** a safety margin of 2 (LogEst: 10) that favors using the IN operator 2635 ** with the index, as using an index has better worst-case behavior. 2636 ** If we do not have real sqlite_stat1 data, always prefer to use 2637 ** the index. Do not bother with this optimization on very small 2638 ** tables (less than 2 rows) as it is pointless in that case. 2639 */ 2640 M = pProbe->aiRowLogEst[saved_nEq]; 2641 logK = estLog(nIn); 2642 /* TUNING v----- 10 to bias toward indexed IN */ 2643 x = M + logK + 10 - (nIn + rLogSize); 2644 if( x>=0 ){ 2645 WHERETRACE(0x40, 2646 ("IN operator (N=%d M=%d logK=%d nIn=%d rLogSize=%d x=%d) " 2647 "prefers indexed lookup\n", 2648 saved_nEq, M, logK, nIn, rLogSize, x)); 2649 }else if( nInMul<2 && OptimizationEnabled(db, SQLITE_SeekScan) ){ 2650 WHERETRACE(0x40, 2651 ("IN operator (N=%d M=%d logK=%d nIn=%d rLogSize=%d x=%d" 2652 " nInMul=%d) prefers skip-scan\n", 2653 saved_nEq, M, logK, nIn, rLogSize, x, nInMul)); 2654 pNew->wsFlags |= WHERE_IN_SEEKSCAN; 2655 }else{ 2656 WHERETRACE(0x40, 2657 ("IN operator (N=%d M=%d logK=%d nIn=%d rLogSize=%d x=%d" 2658 " nInMul=%d) prefers normal scan\n", 2659 saved_nEq, M, logK, nIn, rLogSize, x, nInMul)); 2660 continue; 2661 } 2662 } 2663 pNew->wsFlags |= WHERE_COLUMN_IN; 2664 }else if( eOp & (WO_EQ|WO_IS) ){ 2665 int iCol = pProbe->aiColumn[saved_nEq]; 2666 pNew->wsFlags |= WHERE_COLUMN_EQ; 2667 assert( saved_nEq==pNew->u.btree.nEq ); 2668 if( iCol==XN_ROWID 2669 || (iCol>=0 && nInMul==0 && saved_nEq==pProbe->nKeyCol-1) 2670 ){ 2671 if( iCol==XN_ROWID || pProbe->uniqNotNull 2672 || (pProbe->nKeyCol==1 && pProbe->onError && eOp==WO_EQ) 2673 ){ 2674 pNew->wsFlags |= WHERE_ONEROW; 2675 }else{ 2676 pNew->wsFlags |= WHERE_UNQ_WANTED; 2677 } 2678 } 2679 if( scan.iEquiv>1 ) pNew->wsFlags |= WHERE_TRANSCONS; 2680 }else if( eOp & WO_ISNULL ){ 2681 pNew->wsFlags |= WHERE_COLUMN_NULL; 2682 }else if( eOp & (WO_GT|WO_GE) ){ 2683 testcase( eOp & WO_GT ); 2684 testcase( eOp & WO_GE ); 2685 pNew->wsFlags |= WHERE_COLUMN_RANGE|WHERE_BTM_LIMIT; 2686 pNew->u.btree.nBtm = whereRangeVectorLen( 2687 pParse, pSrc->iCursor, pProbe, saved_nEq, pTerm 2688 ); 2689 pBtm = pTerm; 2690 pTop = 0; 2691 if( pTerm->wtFlags & TERM_LIKEOPT ){ 2692 /* Range constraints that come from the LIKE optimization are 2693 ** always used in pairs. */ 2694 pTop = &pTerm[1]; 2695 assert( (pTop-(pTerm->pWC->a))<pTerm->pWC->nTerm ); 2696 assert( pTop->wtFlags & TERM_LIKEOPT ); 2697 assert( pTop->eOperator==WO_LT ); 2698 if( whereLoopResize(db, pNew, pNew->nLTerm+1) ) break; /* OOM */ 2699 pNew->aLTerm[pNew->nLTerm++] = pTop; 2700 pNew->wsFlags |= WHERE_TOP_LIMIT; 2701 pNew->u.btree.nTop = 1; 2702 } 2703 }else{ 2704 assert( eOp & (WO_LT|WO_LE) ); 2705 testcase( eOp & WO_LT ); 2706 testcase( eOp & WO_LE ); 2707 pNew->wsFlags |= WHERE_COLUMN_RANGE|WHERE_TOP_LIMIT; 2708 pNew->u.btree.nTop = whereRangeVectorLen( 2709 pParse, pSrc->iCursor, pProbe, saved_nEq, pTerm 2710 ); 2711 pTop = pTerm; 2712 pBtm = (pNew->wsFlags & WHERE_BTM_LIMIT)!=0 ? 2713 pNew->aLTerm[pNew->nLTerm-2] : 0; 2714 } 2715 2716 /* At this point pNew->nOut is set to the number of rows expected to 2717 ** be visited by the index scan before considering term pTerm, or the 2718 ** values of nIn and nInMul. In other words, assuming that all 2719 ** "x IN(...)" terms are replaced with "x = ?". This block updates 2720 ** the value of pNew->nOut to account for pTerm (but not nIn/nInMul). */ 2721 assert( pNew->nOut==saved_nOut ); 2722 if( pNew->wsFlags & WHERE_COLUMN_RANGE ){ 2723 /* Adjust nOut using stat4 data. Or, if there is no stat4 2724 ** data, using some other estimate. */ 2725 whereRangeScanEst(pParse, pBuilder, pBtm, pTop, pNew); 2726 }else{ 2727 int nEq = ++pNew->u.btree.nEq; 2728 assert( eOp & (WO_ISNULL|WO_EQ|WO_IN|WO_IS) ); 2729 2730 assert( pNew->nOut==saved_nOut ); 2731 if( pTerm->truthProb<=0 && pProbe->aiColumn[saved_nEq]>=0 ){ 2732 assert( (eOp & WO_IN) || nIn==0 ); 2733 testcase( eOp & WO_IN ); 2734 pNew->nOut += pTerm->truthProb; 2735 pNew->nOut -= nIn; 2736 }else{ 2737 #ifdef SQLITE_ENABLE_STAT4 2738 tRowcnt nOut = 0; 2739 if( nInMul==0 2740 && pProbe->nSample 2741 && ALWAYS(pNew->u.btree.nEq<=pProbe->nSampleCol) 2742 && ((eOp & WO_IN)==0 || ExprUseXList(pTerm->pExpr)) 2743 && OptimizationEnabled(db, SQLITE_Stat4) 2744 ){ 2745 Expr *pExpr = pTerm->pExpr; 2746 if( (eOp & (WO_EQ|WO_ISNULL|WO_IS))!=0 ){ 2747 testcase( eOp & WO_EQ ); 2748 testcase( eOp & WO_IS ); 2749 testcase( eOp & WO_ISNULL ); 2750 rc = whereEqualScanEst(pParse, pBuilder, pExpr->pRight, &nOut); 2751 }else{ 2752 rc = whereInScanEst(pParse, pBuilder, pExpr->x.pList, &nOut); 2753 } 2754 if( rc==SQLITE_NOTFOUND ) rc = SQLITE_OK; 2755 if( rc!=SQLITE_OK ) break; /* Jump out of the pTerm loop */ 2756 if( nOut ){ 2757 pNew->nOut = sqlite3LogEst(nOut); 2758 if( nEq==1 2759 /* TUNING: Mark terms as "low selectivity" if they seem likely 2760 ** to be true for half or more of the rows in the table. 2761 ** See tag-202002240-1 */ 2762 && pNew->nOut+10 > pProbe->aiRowLogEst[0] 2763 ){ 2764 #if WHERETRACE_ENABLED /* 0x01 */ 2765 if( sqlite3WhereTrace & 0x01 ){ 2766 sqlite3DebugPrintf( 2767 "STAT4 determines term has low selectivity:\n"); 2768 sqlite3WhereTermPrint(pTerm, 999); 2769 } 2770 #endif 2771 pTerm->wtFlags |= TERM_HIGHTRUTH; 2772 if( pTerm->wtFlags & TERM_HEURTRUTH ){ 2773 /* If the term has previously been used with an assumption of 2774 ** higher selectivity, then set the flag to rerun the 2775 ** loop computations. */ 2776 pBuilder->bldFlags2 |= SQLITE_BLDF2_2NDPASS; 2777 } 2778 } 2779 if( pNew->nOut>saved_nOut ) pNew->nOut = saved_nOut; 2780 pNew->nOut -= nIn; 2781 } 2782 } 2783 if( nOut==0 ) 2784 #endif 2785 { 2786 pNew->nOut += (pProbe->aiRowLogEst[nEq] - pProbe->aiRowLogEst[nEq-1]); 2787 if( eOp & WO_ISNULL ){ 2788 /* TUNING: If there is no likelihood() value, assume that a 2789 ** "col IS NULL" expression matches twice as many rows 2790 ** as (col=?). */ 2791 pNew->nOut += 10; 2792 } 2793 } 2794 } 2795 } 2796 2797 /* Set rCostIdx to the cost of visiting selected rows in index. Add 2798 ** it to pNew->rRun, which is currently set to the cost of the index 2799 ** seek only. Then, if this is a non-covering index, add the cost of 2800 ** visiting the rows in the main table. */ 2801 assert( pSrc->pTab->szTabRow>0 ); 2802 rCostIdx = pNew->nOut + 1 + (15*pProbe->szIdxRow)/pSrc->pTab->szTabRow; 2803 pNew->rRun = sqlite3LogEstAdd(rLogSize, rCostIdx); 2804 if( (pNew->wsFlags & (WHERE_IDX_ONLY|WHERE_IPK))==0 ){ 2805 pNew->rRun = sqlite3LogEstAdd(pNew->rRun, pNew->nOut + 16); 2806 } 2807 ApplyCostMultiplier(pNew->rRun, pProbe->pTable->costMult); 2808 2809 nOutUnadjusted = pNew->nOut; 2810 pNew->rRun += nInMul + nIn; 2811 pNew->nOut += nInMul + nIn; 2812 whereLoopOutputAdjust(pBuilder->pWC, pNew, rSize); 2813 rc = whereLoopInsert(pBuilder, pNew); 2814 2815 if( pNew->wsFlags & WHERE_COLUMN_RANGE ){ 2816 pNew->nOut = saved_nOut; 2817 }else{ 2818 pNew->nOut = nOutUnadjusted; 2819 } 2820 2821 if( (pNew->wsFlags & WHERE_TOP_LIMIT)==0 2822 && pNew->u.btree.nEq<pProbe->nColumn 2823 && (pNew->u.btree.nEq<pProbe->nKeyCol || 2824 pProbe->idxType!=SQLITE_IDXTYPE_PRIMARYKEY) 2825 ){ 2826 whereLoopAddBtreeIndex(pBuilder, pSrc, pProbe, nInMul+nIn); 2827 } 2828 pNew->nOut = saved_nOut; 2829 #ifdef SQLITE_ENABLE_STAT4 2830 pBuilder->nRecValid = nRecValid; 2831 #endif 2832 } 2833 pNew->prereq = saved_prereq; 2834 pNew->u.btree.nEq = saved_nEq; 2835 pNew->u.btree.nBtm = saved_nBtm; 2836 pNew->u.btree.nTop = saved_nTop; 2837 pNew->nSkip = saved_nSkip; 2838 pNew->wsFlags = saved_wsFlags; 2839 pNew->nOut = saved_nOut; 2840 pNew->nLTerm = saved_nLTerm; 2841 2842 /* Consider using a skip-scan if there are no WHERE clause constraints 2843 ** available for the left-most terms of the index, and if the average 2844 ** number of repeats in the left-most terms is at least 18. 2845 ** 2846 ** The magic number 18 is selected on the basis that scanning 17 rows 2847 ** is almost always quicker than an index seek (even though if the index 2848 ** contains fewer than 2^17 rows we assume otherwise in other parts of 2849 ** the code). And, even if it is not, it should not be too much slower. 2850 ** On the other hand, the extra seeks could end up being significantly 2851 ** more expensive. */ 2852 assert( 42==sqlite3LogEst(18) ); 2853 if( saved_nEq==saved_nSkip 2854 && saved_nEq+1<pProbe->nKeyCol 2855 && saved_nEq==pNew->nLTerm 2856 && pProbe->noSkipScan==0 2857 && pProbe->hasStat1!=0 2858 && OptimizationEnabled(db, SQLITE_SkipScan) 2859 && pProbe->aiRowLogEst[saved_nEq+1]>=42 /* TUNING: Minimum for skip-scan */ 2860 && (rc = whereLoopResize(db, pNew, pNew->nLTerm+1))==SQLITE_OK 2861 ){ 2862 LogEst nIter; 2863 pNew->u.btree.nEq++; 2864 pNew->nSkip++; 2865 pNew->aLTerm[pNew->nLTerm++] = 0; 2866 pNew->wsFlags |= WHERE_SKIPSCAN; 2867 nIter = pProbe->aiRowLogEst[saved_nEq] - pProbe->aiRowLogEst[saved_nEq+1]; 2868 pNew->nOut -= nIter; 2869 /* TUNING: Because uncertainties in the estimates for skip-scan queries, 2870 ** add a 1.375 fudge factor to make skip-scan slightly less likely. */ 2871 nIter += 5; 2872 whereLoopAddBtreeIndex(pBuilder, pSrc, pProbe, nIter + nInMul); 2873 pNew->nOut = saved_nOut; 2874 pNew->u.btree.nEq = saved_nEq; 2875 pNew->nSkip = saved_nSkip; 2876 pNew->wsFlags = saved_wsFlags; 2877 } 2878 2879 WHERETRACE(0x800, ("END %s.addBtreeIdx(%s), nEq=%d, rc=%d\n", 2880 pProbe->pTable->zName, pProbe->zName, saved_nEq, rc)); 2881 return rc; 2882 } 2883 2884 /* 2885 ** Return True if it is possible that pIndex might be useful in 2886 ** implementing the ORDER BY clause in pBuilder. 2887 ** 2888 ** Return False if pBuilder does not contain an ORDER BY clause or 2889 ** if there is no way for pIndex to be useful in implementing that 2890 ** ORDER BY clause. 2891 */ 2892 static int indexMightHelpWithOrderBy( 2893 WhereLoopBuilder *pBuilder, 2894 Index *pIndex, 2895 int iCursor 2896 ){ 2897 ExprList *pOB; 2898 ExprList *aColExpr; 2899 int ii, jj; 2900 2901 if( pIndex->bUnordered ) return 0; 2902 if( (pOB = pBuilder->pWInfo->pOrderBy)==0 ) return 0; 2903 for(ii=0; ii<pOB->nExpr; ii++){ 2904 Expr *pExpr = sqlite3ExprSkipCollateAndLikely(pOB->a[ii].pExpr); 2905 if( NEVER(pExpr==0) ) continue; 2906 if( pExpr->op==TK_COLUMN && pExpr->iTable==iCursor ){ 2907 if( pExpr->iColumn<0 ) return 1; 2908 for(jj=0; jj<pIndex->nKeyCol; jj++){ 2909 if( pExpr->iColumn==pIndex->aiColumn[jj] ) return 1; 2910 } 2911 }else if( (aColExpr = pIndex->aColExpr)!=0 ){ 2912 for(jj=0; jj<pIndex->nKeyCol; jj++){ 2913 if( pIndex->aiColumn[jj]!=XN_EXPR ) continue; 2914 if( sqlite3ExprCompareSkip(pExpr,aColExpr->a[jj].pExpr,iCursor)==0 ){ 2915 return 1; 2916 } 2917 } 2918 } 2919 } 2920 return 0; 2921 } 2922 2923 /* Check to see if a partial index with pPartIndexWhere can be used 2924 ** in the current query. Return true if it can be and false if not. 2925 */ 2926 static int whereUsablePartialIndex( 2927 int iTab, /* The table for which we want an index */ 2928 int isLeft, /* True if iTab is the right table of a LEFT JOIN */ 2929 WhereClause *pWC, /* The WHERE clause of the query */ 2930 Expr *pWhere /* The WHERE clause from the partial index */ 2931 ){ 2932 int i; 2933 WhereTerm *pTerm; 2934 Parse *pParse = pWC->pWInfo->pParse; 2935 while( pWhere->op==TK_AND ){ 2936 if( !whereUsablePartialIndex(iTab,isLeft,pWC,pWhere->pLeft) ) return 0; 2937 pWhere = pWhere->pRight; 2938 } 2939 if( pParse->db->flags & SQLITE_EnableQPSG ) pParse = 0; 2940 for(i=0, pTerm=pWC->a; i<pWC->nTerm; i++, pTerm++){ 2941 Expr *pExpr; 2942 pExpr = pTerm->pExpr; 2943 if( (!ExprHasProperty(pExpr, EP_FromJoin) || pExpr->iRightJoinTable==iTab) 2944 && (isLeft==0 || ExprHasProperty(pExpr, EP_FromJoin)) 2945 && sqlite3ExprImpliesExpr(pParse, pExpr, pWhere, iTab) 2946 && (pTerm->wtFlags & TERM_VNULL)==0 2947 ){ 2948 return 1; 2949 } 2950 } 2951 return 0; 2952 } 2953 2954 /* 2955 ** Add all WhereLoop objects for a single table of the join where the table 2956 ** is identified by pBuilder->pNew->iTab. That table is guaranteed to be 2957 ** a b-tree table, not a virtual table. 2958 ** 2959 ** The costs (WhereLoop.rRun) of the b-tree loops added by this function 2960 ** are calculated as follows: 2961 ** 2962 ** For a full scan, assuming the table (or index) contains nRow rows: 2963 ** 2964 ** cost = nRow * 3.0 // full-table scan 2965 ** cost = nRow * K // scan of covering index 2966 ** cost = nRow * (K+3.0) // scan of non-covering index 2967 ** 2968 ** where K is a value between 1.1 and 3.0 set based on the relative 2969 ** estimated average size of the index and table records. 2970 ** 2971 ** For an index scan, where nVisit is the number of index rows visited 2972 ** by the scan, and nSeek is the number of seek operations required on 2973 ** the index b-tree: 2974 ** 2975 ** cost = nSeek * (log(nRow) + K * nVisit) // covering index 2976 ** cost = nSeek * (log(nRow) + (K+3.0) * nVisit) // non-covering index 2977 ** 2978 ** Normally, nSeek is 1. nSeek values greater than 1 come about if the 2979 ** WHERE clause includes "x IN (....)" terms used in place of "x=?". Or when 2980 ** implicit "x IN (SELECT x FROM tbl)" terms are added for skip-scans. 2981 ** 2982 ** The estimated values (nRow, nVisit, nSeek) often contain a large amount 2983 ** of uncertainty. For this reason, scoring is designed to pick plans that 2984 ** "do the least harm" if the estimates are inaccurate. For example, a 2985 ** log(nRow) factor is omitted from a non-covering index scan in order to 2986 ** bias the scoring in favor of using an index, since the worst-case 2987 ** performance of using an index is far better than the worst-case performance 2988 ** of a full table scan. 2989 */ 2990 static int whereLoopAddBtree( 2991 WhereLoopBuilder *pBuilder, /* WHERE clause information */ 2992 Bitmask mPrereq /* Extra prerequesites for using this table */ 2993 ){ 2994 WhereInfo *pWInfo; /* WHERE analysis context */ 2995 Index *pProbe; /* An index we are evaluating */ 2996 Index sPk; /* A fake index object for the primary key */ 2997 LogEst aiRowEstPk[2]; /* The aiRowLogEst[] value for the sPk index */ 2998 i16 aiColumnPk = -1; /* The aColumn[] value for the sPk index */ 2999 SrcList *pTabList; /* The FROM clause */ 3000 SrcItem *pSrc; /* The FROM clause btree term to add */ 3001 WhereLoop *pNew; /* Template WhereLoop object */ 3002 int rc = SQLITE_OK; /* Return code */ 3003 int iSortIdx = 1; /* Index number */ 3004 int b; /* A boolean value */ 3005 LogEst rSize; /* number of rows in the table */ 3006 WhereClause *pWC; /* The parsed WHERE clause */ 3007 Table *pTab; /* Table being queried */ 3008 3009 pNew = pBuilder->pNew; 3010 pWInfo = pBuilder->pWInfo; 3011 pTabList = pWInfo->pTabList; 3012 pSrc = pTabList->a + pNew->iTab; 3013 pTab = pSrc->pTab; 3014 pWC = pBuilder->pWC; 3015 assert( !IsVirtual(pSrc->pTab) ); 3016 3017 if( pSrc->fg.isIndexedBy ){ 3018 assert( pSrc->fg.isCte==0 ); 3019 /* An INDEXED BY clause specifies a particular index to use */ 3020 pProbe = pSrc->u2.pIBIndex; 3021 }else if( !HasRowid(pTab) ){ 3022 pProbe = pTab->pIndex; 3023 }else{ 3024 /* There is no INDEXED BY clause. Create a fake Index object in local 3025 ** variable sPk to represent the rowid primary key index. Make this 3026 ** fake index the first in a chain of Index objects with all of the real 3027 ** indices to follow */ 3028 Index *pFirst; /* First of real indices on the table */ 3029 memset(&sPk, 0, sizeof(Index)); 3030 sPk.nKeyCol = 1; 3031 sPk.nColumn = 1; 3032 sPk.aiColumn = &aiColumnPk; 3033 sPk.aiRowLogEst = aiRowEstPk; 3034 sPk.onError = OE_Replace; 3035 sPk.pTable = pTab; 3036 sPk.szIdxRow = pTab->szTabRow; 3037 sPk.idxType = SQLITE_IDXTYPE_IPK; 3038 aiRowEstPk[0] = pTab->nRowLogEst; 3039 aiRowEstPk[1] = 0; 3040 pFirst = pSrc->pTab->pIndex; 3041 if( pSrc->fg.notIndexed==0 ){ 3042 /* The real indices of the table are only considered if the 3043 ** NOT INDEXED qualifier is omitted from the FROM clause */ 3044 sPk.pNext = pFirst; 3045 } 3046 pProbe = &sPk; 3047 } 3048 rSize = pTab->nRowLogEst; 3049 3050 #ifndef SQLITE_OMIT_AUTOMATIC_INDEX 3051 /* Automatic indexes */ 3052 if( !pBuilder->pOrSet /* Not part of an OR optimization */ 3053 && (pWInfo->wctrlFlags & WHERE_OR_SUBCLAUSE)==0 3054 && (pWInfo->pParse->db->flags & SQLITE_AutoIndex)!=0 3055 && !pSrc->fg.isIndexedBy /* Has no INDEXED BY clause */ 3056 && !pSrc->fg.notIndexed /* Has no NOT INDEXED clause */ 3057 && HasRowid(pTab) /* Not WITHOUT ROWID table. (FIXME: Why not?) */ 3058 && !pSrc->fg.isCorrelated /* Not a correlated subquery */ 3059 && !pSrc->fg.isRecursive /* Not a recursive common table expression. */ 3060 ){ 3061 /* Generate auto-index WhereLoops */ 3062 LogEst rLogSize; /* Logarithm of the number of rows in the table */ 3063 WhereTerm *pTerm; 3064 WhereTerm *pWCEnd = pWC->a + pWC->nTerm; 3065 rLogSize = estLog(rSize); 3066 for(pTerm=pWC->a; rc==SQLITE_OK && pTerm<pWCEnd; pTerm++){ 3067 if( pTerm->prereqRight & pNew->maskSelf ) continue; 3068 if( termCanDriveIndex(pTerm, pSrc, 0) ){ 3069 pNew->u.btree.nEq = 1; 3070 pNew->nSkip = 0; 3071 pNew->u.btree.pIndex = 0; 3072 pNew->nLTerm = 1; 3073 pNew->aLTerm[0] = pTerm; 3074 /* TUNING: One-time cost for computing the automatic index is 3075 ** estimated to be X*N*log2(N) where N is the number of rows in 3076 ** the table being indexed and where X is 7 (LogEst=28) for normal 3077 ** tables or 0.5 (LogEst=-10) for views and subqueries. The value 3078 ** of X is smaller for views and subqueries so that the query planner 3079 ** will be more aggressive about generating automatic indexes for 3080 ** those objects, since there is no opportunity to add schema 3081 ** indexes on subqueries and views. */ 3082 pNew->rSetup = rLogSize + rSize; 3083 if( !IsView(pTab) && (pTab->tabFlags & TF_Ephemeral)==0 ){ 3084 pNew->rSetup += 28; 3085 }else{ 3086 pNew->rSetup -= 10; 3087 } 3088 ApplyCostMultiplier(pNew->rSetup, pTab->costMult); 3089 if( pNew->rSetup<0 ) pNew->rSetup = 0; 3090 /* TUNING: Each index lookup yields 20 rows in the table. This 3091 ** is more than the usual guess of 10 rows, since we have no way 3092 ** of knowing how selective the index will ultimately be. It would 3093 ** not be unreasonable to make this value much larger. */ 3094 pNew->nOut = 43; assert( 43==sqlite3LogEst(20) ); 3095 pNew->rRun = sqlite3LogEstAdd(rLogSize,pNew->nOut); 3096 pNew->wsFlags = WHERE_AUTO_INDEX; 3097 pNew->prereq = mPrereq | pTerm->prereqRight; 3098 rc = whereLoopInsert(pBuilder, pNew); 3099 } 3100 } 3101 } 3102 #endif /* SQLITE_OMIT_AUTOMATIC_INDEX */ 3103 3104 /* Loop over all indices. If there was an INDEXED BY clause, then only 3105 ** consider index pProbe. */ 3106 for(; rc==SQLITE_OK && pProbe; 3107 pProbe=(pSrc->fg.isIndexedBy ? 0 : pProbe->pNext), iSortIdx++ 3108 ){ 3109 int isLeft = (pSrc->fg.jointype & JT_OUTER)!=0; 3110 if( pProbe->pPartIdxWhere!=0 3111 && !whereUsablePartialIndex(pSrc->iCursor, isLeft, pWC, 3112 pProbe->pPartIdxWhere) 3113 ){ 3114 testcase( pNew->iTab!=pSrc->iCursor ); /* See ticket [98d973b8f5] */ 3115 continue; /* Partial index inappropriate for this query */ 3116 } 3117 if( pProbe->bNoQuery ) continue; 3118 rSize = pProbe->aiRowLogEst[0]; 3119 pNew->u.btree.nEq = 0; 3120 pNew->u.btree.nBtm = 0; 3121 pNew->u.btree.nTop = 0; 3122 pNew->nSkip = 0; 3123 pNew->nLTerm = 0; 3124 pNew->iSortIdx = 0; 3125 pNew->rSetup = 0; 3126 pNew->prereq = mPrereq; 3127 pNew->nOut = rSize; 3128 pNew->u.btree.pIndex = pProbe; 3129 b = indexMightHelpWithOrderBy(pBuilder, pProbe, pSrc->iCursor); 3130 3131 /* The ONEPASS_DESIRED flags never occurs together with ORDER BY */ 3132 assert( (pWInfo->wctrlFlags & WHERE_ONEPASS_DESIRED)==0 || b==0 ); 3133 if( pProbe->idxType==SQLITE_IDXTYPE_IPK ){ 3134 /* Integer primary key index */ 3135 pNew->wsFlags = WHERE_IPK; 3136 3137 /* Full table scan */ 3138 pNew->iSortIdx = b ? iSortIdx : 0; 3139 /* TUNING: Cost of full table scan is 3.0*N. The 3.0 factor is an 3140 ** extra cost designed to discourage the use of full table scans, 3141 ** since index lookups have better worst-case performance if our 3142 ** stat guesses are wrong. Reduce the 3.0 penalty slightly 3143 ** (to 2.75) if we have valid STAT4 information for the table. 3144 ** At 2.75, a full table scan is preferred over using an index on 3145 ** a column with just two distinct values where each value has about 3146 ** an equal number of appearances. Without STAT4 data, we still want 3147 ** to use an index in that case, since the constraint might be for 3148 ** the scarcer of the two values, and in that case an index lookup is 3149 ** better. 3150 */ 3151 #ifdef SQLITE_ENABLE_STAT4 3152 pNew->rRun = rSize + 16 - 2*((pTab->tabFlags & TF_HasStat4)!=0); 3153 #else 3154 pNew->rRun = rSize + 16; 3155 #endif 3156 ApplyCostMultiplier(pNew->rRun, pTab->costMult); 3157 whereLoopOutputAdjust(pWC, pNew, rSize); 3158 rc = whereLoopInsert(pBuilder, pNew); 3159 pNew->nOut = rSize; 3160 if( rc ) break; 3161 }else{ 3162 Bitmask m; 3163 if( pProbe->isCovering ){ 3164 pNew->wsFlags = WHERE_IDX_ONLY | WHERE_INDEXED; 3165 m = 0; 3166 }else{ 3167 m = pSrc->colUsed & pProbe->colNotIdxed; 3168 pNew->wsFlags = (m==0) ? (WHERE_IDX_ONLY|WHERE_INDEXED) : WHERE_INDEXED; 3169 } 3170 3171 /* Full scan via index */ 3172 if( b 3173 || !HasRowid(pTab) 3174 || pProbe->pPartIdxWhere!=0 3175 || pSrc->fg.isIndexedBy 3176 || ( m==0 3177 && pProbe->bUnordered==0 3178 && (pProbe->szIdxRow<pTab->szTabRow) 3179 && (pWInfo->wctrlFlags & WHERE_ONEPASS_DESIRED)==0 3180 && sqlite3GlobalConfig.bUseCis 3181 && OptimizationEnabled(pWInfo->pParse->db, SQLITE_CoverIdxScan) 3182 ) 3183 ){ 3184 pNew->iSortIdx = b ? iSortIdx : 0; 3185 3186 /* The cost of visiting the index rows is N*K, where K is 3187 ** between 1.1 and 3.0, depending on the relative sizes of the 3188 ** index and table rows. */ 3189 pNew->rRun = rSize + 1 + (15*pProbe->szIdxRow)/pTab->szTabRow; 3190 if( m!=0 ){ 3191 /* If this is a non-covering index scan, add in the cost of 3192 ** doing table lookups. The cost will be 3x the number of 3193 ** lookups. Take into account WHERE clause terms that can be 3194 ** satisfied using just the index, and that do not require a 3195 ** table lookup. */ 3196 LogEst nLookup = rSize + 16; /* Base cost: N*3 */ 3197 int ii; 3198 int iCur = pSrc->iCursor; 3199 WhereClause *pWC2 = &pWInfo->sWC; 3200 for(ii=0; ii<pWC2->nTerm; ii++){ 3201 WhereTerm *pTerm = &pWC2->a[ii]; 3202 if( !sqlite3ExprCoveredByIndex(pTerm->pExpr, iCur, pProbe) ){ 3203 break; 3204 } 3205 /* pTerm can be evaluated using just the index. So reduce 3206 ** the expected number of table lookups accordingly */ 3207 if( pTerm->truthProb<=0 ){ 3208 nLookup += pTerm->truthProb; 3209 }else{ 3210 nLookup--; 3211 if( pTerm->eOperator & (WO_EQ|WO_IS) ) nLookup -= 19; 3212 } 3213 } 3214 3215 pNew->rRun = sqlite3LogEstAdd(pNew->rRun, nLookup); 3216 } 3217 ApplyCostMultiplier(pNew->rRun, pTab->costMult); 3218 whereLoopOutputAdjust(pWC, pNew, rSize); 3219 rc = whereLoopInsert(pBuilder, pNew); 3220 pNew->nOut = rSize; 3221 if( rc ) break; 3222 } 3223 } 3224 3225 pBuilder->bldFlags1 = 0; 3226 rc = whereLoopAddBtreeIndex(pBuilder, pSrc, pProbe, 0); 3227 if( pBuilder->bldFlags1==SQLITE_BLDF1_INDEXED ){ 3228 /* If a non-unique index is used, or if a prefix of the key for 3229 ** unique index is used (making the index functionally non-unique) 3230 ** then the sqlite_stat1 data becomes important for scoring the 3231 ** plan */ 3232 pTab->tabFlags |= TF_StatsUsed; 3233 } 3234 #ifdef SQLITE_ENABLE_STAT4 3235 sqlite3Stat4ProbeFree(pBuilder->pRec); 3236 pBuilder->nRecValid = 0; 3237 pBuilder->pRec = 0; 3238 #endif 3239 } 3240 return rc; 3241 } 3242 3243 #ifndef SQLITE_OMIT_VIRTUALTABLE 3244 3245 /* 3246 ** Argument pIdxInfo is already populated with all constraints that may 3247 ** be used by the virtual table identified by pBuilder->pNew->iTab. This 3248 ** function marks a subset of those constraints usable, invokes the 3249 ** xBestIndex method and adds the returned plan to pBuilder. 3250 ** 3251 ** A constraint is marked usable if: 3252 ** 3253 ** * Argument mUsable indicates that its prerequisites are available, and 3254 ** 3255 ** * It is not one of the operators specified in the mExclude mask passed 3256 ** as the fourth argument (which in practice is either WO_IN or 0). 3257 ** 3258 ** Argument mPrereq is a mask of tables that must be scanned before the 3259 ** virtual table in question. These are added to the plans prerequisites 3260 ** before it is added to pBuilder. 3261 ** 3262 ** Output parameter *pbIn is set to true if the plan added to pBuilder 3263 ** uses one or more WO_IN terms, or false otherwise. 3264 */ 3265 static int whereLoopAddVirtualOne( 3266 WhereLoopBuilder *pBuilder, 3267 Bitmask mPrereq, /* Mask of tables that must be used. */ 3268 Bitmask mUsable, /* Mask of usable tables */ 3269 u16 mExclude, /* Exclude terms using these operators */ 3270 sqlite3_index_info *pIdxInfo, /* Populated object for xBestIndex */ 3271 u16 mNoOmit, /* Do not omit these constraints */ 3272 int *pbIn /* OUT: True if plan uses an IN(...) op */ 3273 ){ 3274 WhereClause *pWC = pBuilder->pWC; 3275 struct sqlite3_index_constraint *pIdxCons; 3276 struct sqlite3_index_constraint_usage *pUsage = pIdxInfo->aConstraintUsage; 3277 int i; 3278 int mxTerm; 3279 int rc = SQLITE_OK; 3280 WhereLoop *pNew = pBuilder->pNew; 3281 Parse *pParse = pBuilder->pWInfo->pParse; 3282 SrcItem *pSrc = &pBuilder->pWInfo->pTabList->a[pNew->iTab]; 3283 int nConstraint = pIdxInfo->nConstraint; 3284 3285 assert( (mUsable & mPrereq)==mPrereq ); 3286 *pbIn = 0; 3287 pNew->prereq = mPrereq; 3288 3289 /* Set the usable flag on the subset of constraints identified by 3290 ** arguments mUsable and mExclude. */ 3291 pIdxCons = *(struct sqlite3_index_constraint**)&pIdxInfo->aConstraint; 3292 for(i=0; i<nConstraint; i++, pIdxCons++){ 3293 WhereTerm *pTerm = &pWC->a[pIdxCons->iTermOffset]; 3294 pIdxCons->usable = 0; 3295 if( (pTerm->prereqRight & mUsable)==pTerm->prereqRight 3296 && (pTerm->eOperator & mExclude)==0 3297 ){ 3298 pIdxCons->usable = 1; 3299 } 3300 } 3301 3302 /* Initialize the output fields of the sqlite3_index_info structure */ 3303 memset(pUsage, 0, sizeof(pUsage[0])*nConstraint); 3304 assert( pIdxInfo->needToFreeIdxStr==0 ); 3305 pIdxInfo->idxStr = 0; 3306 pIdxInfo->idxNum = 0; 3307 pIdxInfo->orderByConsumed = 0; 3308 pIdxInfo->estimatedCost = SQLITE_BIG_DBL / (double)2; 3309 pIdxInfo->estimatedRows = 25; 3310 pIdxInfo->idxFlags = 0; 3311 pIdxInfo->colUsed = (sqlite3_int64)pSrc->colUsed; 3312 3313 /* Invoke the virtual table xBestIndex() method */ 3314 rc = vtabBestIndex(pParse, pSrc->pTab, pIdxInfo); 3315 if( rc ){ 3316 if( rc==SQLITE_CONSTRAINT ){ 3317 /* If the xBestIndex method returns SQLITE_CONSTRAINT, that means 3318 ** that the particular combination of parameters provided is unusable. 3319 ** Make no entries in the loop table. 3320 */ 3321 WHERETRACE(0xffff, (" ^^^^--- non-viable plan rejected!\n")); 3322 return SQLITE_OK; 3323 } 3324 return rc; 3325 } 3326 3327 mxTerm = -1; 3328 assert( pNew->nLSlot>=nConstraint ); 3329 for(i=0; i<nConstraint; i++) pNew->aLTerm[i] = 0; 3330 pNew->u.vtab.omitMask = 0; 3331 pIdxCons = *(struct sqlite3_index_constraint**)&pIdxInfo->aConstraint; 3332 for(i=0; i<nConstraint; i++, pIdxCons++){ 3333 int iTerm; 3334 if( (iTerm = pUsage[i].argvIndex - 1)>=0 ){ 3335 WhereTerm *pTerm; 3336 int j = pIdxCons->iTermOffset; 3337 if( iTerm>=nConstraint 3338 || j<0 3339 || j>=pWC->nTerm 3340 || pNew->aLTerm[iTerm]!=0 3341 || pIdxCons->usable==0 3342 ){ 3343 sqlite3ErrorMsg(pParse,"%s.xBestIndex malfunction",pSrc->pTab->zName); 3344 testcase( pIdxInfo->needToFreeIdxStr ); 3345 return SQLITE_ERROR; 3346 } 3347 testcase( iTerm==nConstraint-1 ); 3348 testcase( j==0 ); 3349 testcase( j==pWC->nTerm-1 ); 3350 pTerm = &pWC->a[j]; 3351 pNew->prereq |= pTerm->prereqRight; 3352 assert( iTerm<pNew->nLSlot ); 3353 pNew->aLTerm[iTerm] = pTerm; 3354 if( iTerm>mxTerm ) mxTerm = iTerm; 3355 testcase( iTerm==15 ); 3356 testcase( iTerm==16 ); 3357 if( pUsage[i].omit ){ 3358 if( i<16 && ((1<<i)&mNoOmit)==0 ){ 3359 testcase( i!=iTerm ); 3360 pNew->u.vtab.omitMask |= 1<<iTerm; 3361 }else{ 3362 testcase( i!=iTerm ); 3363 } 3364 } 3365 if( (pTerm->eOperator & WO_IN)!=0 ){ 3366 /* A virtual table that is constrained by an IN clause may not 3367 ** consume the ORDER BY clause because (1) the order of IN terms 3368 ** is not necessarily related to the order of output terms and 3369 ** (2) Multiple outputs from a single IN value will not merge 3370 ** together. */ 3371 pIdxInfo->orderByConsumed = 0; 3372 pIdxInfo->idxFlags &= ~SQLITE_INDEX_SCAN_UNIQUE; 3373 *pbIn = 1; assert( (mExclude & WO_IN)==0 ); 3374 } 3375 } 3376 } 3377 3378 pNew->nLTerm = mxTerm+1; 3379 for(i=0; i<=mxTerm; i++){ 3380 if( pNew->aLTerm[i]==0 ){ 3381 /* The non-zero argvIdx values must be contiguous. Raise an 3382 ** error if they are not */ 3383 sqlite3ErrorMsg(pParse,"%s.xBestIndex malfunction",pSrc->pTab->zName); 3384 testcase( pIdxInfo->needToFreeIdxStr ); 3385 return SQLITE_ERROR; 3386 } 3387 } 3388 assert( pNew->nLTerm<=pNew->nLSlot ); 3389 pNew->u.vtab.idxNum = pIdxInfo->idxNum; 3390 pNew->u.vtab.needFree = pIdxInfo->needToFreeIdxStr; 3391 pIdxInfo->needToFreeIdxStr = 0; 3392 pNew->u.vtab.idxStr = pIdxInfo->idxStr; 3393 pNew->u.vtab.isOrdered = (i8)(pIdxInfo->orderByConsumed ? 3394 pIdxInfo->nOrderBy : 0); 3395 pNew->rSetup = 0; 3396 pNew->rRun = sqlite3LogEstFromDouble(pIdxInfo->estimatedCost); 3397 pNew->nOut = sqlite3LogEst(pIdxInfo->estimatedRows); 3398 3399 /* Set the WHERE_ONEROW flag if the xBestIndex() method indicated 3400 ** that the scan will visit at most one row. Clear it otherwise. */ 3401 if( pIdxInfo->idxFlags & SQLITE_INDEX_SCAN_UNIQUE ){ 3402 pNew->wsFlags |= WHERE_ONEROW; 3403 }else{ 3404 pNew->wsFlags &= ~WHERE_ONEROW; 3405 } 3406 rc = whereLoopInsert(pBuilder, pNew); 3407 if( pNew->u.vtab.needFree ){ 3408 sqlite3_free(pNew->u.vtab.idxStr); 3409 pNew->u.vtab.needFree = 0; 3410 } 3411 WHERETRACE(0xffff, (" bIn=%d prereqIn=%04llx prereqOut=%04llx\n", 3412 *pbIn, (sqlite3_uint64)mPrereq, 3413 (sqlite3_uint64)(pNew->prereq & ~mPrereq))); 3414 3415 return rc; 3416 } 3417 3418 /* 3419 ** If this function is invoked from within an xBestIndex() callback, it 3420 ** returns a pointer to a buffer containing the name of the collation 3421 ** sequence associated with element iCons of the sqlite3_index_info.aConstraint 3422 ** array. Or, if iCons is out of range or there is no active xBestIndex 3423 ** call, return NULL. 3424 */ 3425 const char *sqlite3_vtab_collation(sqlite3_index_info *pIdxInfo, int iCons){ 3426 HiddenIndexInfo *pHidden = (HiddenIndexInfo*)&pIdxInfo[1]; 3427 const char *zRet = 0; 3428 if( iCons>=0 && iCons<pIdxInfo->nConstraint ){ 3429 CollSeq *pC = 0; 3430 int iTerm = pIdxInfo->aConstraint[iCons].iTermOffset; 3431 Expr *pX = pHidden->pWC->a[iTerm].pExpr; 3432 if( pX->pLeft ){ 3433 pC = sqlite3ExprCompareCollSeq(pHidden->pParse, pX); 3434 } 3435 zRet = (pC ? pC->zName : sqlite3StrBINARY); 3436 } 3437 return zRet; 3438 } 3439 3440 /* 3441 ** Add all WhereLoop objects for a table of the join identified by 3442 ** pBuilder->pNew->iTab. That table is guaranteed to be a virtual table. 3443 ** 3444 ** If there are no LEFT or CROSS JOIN joins in the query, both mPrereq and 3445 ** mUnusable are set to 0. Otherwise, mPrereq is a mask of all FROM clause 3446 ** entries that occur before the virtual table in the FROM clause and are 3447 ** separated from it by at least one LEFT or CROSS JOIN. Similarly, the 3448 ** mUnusable mask contains all FROM clause entries that occur after the 3449 ** virtual table and are separated from it by at least one LEFT or 3450 ** CROSS JOIN. 3451 ** 3452 ** For example, if the query were: 3453 ** 3454 ** ... FROM t1, t2 LEFT JOIN t3, t4, vt CROSS JOIN t5, t6; 3455 ** 3456 ** then mPrereq corresponds to (t1, t2) and mUnusable to (t5, t6). 3457 ** 3458 ** All the tables in mPrereq must be scanned before the current virtual 3459 ** table. So any terms for which all prerequisites are satisfied by 3460 ** mPrereq may be specified as "usable" in all calls to xBestIndex. 3461 ** Conversely, all tables in mUnusable must be scanned after the current 3462 ** virtual table, so any terms for which the prerequisites overlap with 3463 ** mUnusable should always be configured as "not-usable" for xBestIndex. 3464 */ 3465 static int whereLoopAddVirtual( 3466 WhereLoopBuilder *pBuilder, /* WHERE clause information */ 3467 Bitmask mPrereq, /* Tables that must be scanned before this one */ 3468 Bitmask mUnusable /* Tables that must be scanned after this one */ 3469 ){ 3470 int rc = SQLITE_OK; /* Return code */ 3471 WhereInfo *pWInfo; /* WHERE analysis context */ 3472 Parse *pParse; /* The parsing context */ 3473 WhereClause *pWC; /* The WHERE clause */ 3474 SrcItem *pSrc; /* The FROM clause term to search */ 3475 sqlite3_index_info *p; /* Object to pass to xBestIndex() */ 3476 int nConstraint; /* Number of constraints in p */ 3477 int bIn; /* True if plan uses IN(...) operator */ 3478 WhereLoop *pNew; 3479 Bitmask mBest; /* Tables used by best possible plan */ 3480 u16 mNoOmit; 3481 3482 assert( (mPrereq & mUnusable)==0 ); 3483 pWInfo = pBuilder->pWInfo; 3484 pParse = pWInfo->pParse; 3485 pWC = pBuilder->pWC; 3486 pNew = pBuilder->pNew; 3487 pSrc = &pWInfo->pTabList->a[pNew->iTab]; 3488 assert( IsVirtual(pSrc->pTab) ); 3489 p = allocateIndexInfo(pParse, pWC, mUnusable, pSrc, pBuilder->pOrderBy, 3490 &mNoOmit); 3491 if( p==0 ) return SQLITE_NOMEM_BKPT; 3492 pNew->rSetup = 0; 3493 pNew->wsFlags = WHERE_VIRTUALTABLE; 3494 pNew->nLTerm = 0; 3495 pNew->u.vtab.needFree = 0; 3496 nConstraint = p->nConstraint; 3497 if( whereLoopResize(pParse->db, pNew, nConstraint) ){ 3498 sqlite3DbFree(pParse->db, p); 3499 return SQLITE_NOMEM_BKPT; 3500 } 3501 3502 /* First call xBestIndex() with all constraints usable. */ 3503 WHERETRACE(0x800, ("BEGIN %s.addVirtual()\n", pSrc->pTab->zName)); 3504 WHERETRACE(0x40, (" VirtualOne: all usable\n")); 3505 rc = whereLoopAddVirtualOne(pBuilder, mPrereq, ALLBITS, 0, p, mNoOmit, &bIn); 3506 3507 /* If the call to xBestIndex() with all terms enabled produced a plan 3508 ** that does not require any source tables (IOW: a plan with mBest==0) 3509 ** and does not use an IN(...) operator, then there is no point in making 3510 ** any further calls to xBestIndex() since they will all return the same 3511 ** result (if the xBestIndex() implementation is sane). */ 3512 if( rc==SQLITE_OK && ((mBest = (pNew->prereq & ~mPrereq))!=0 || bIn) ){ 3513 int seenZero = 0; /* True if a plan with no prereqs seen */ 3514 int seenZeroNoIN = 0; /* Plan with no prereqs and no IN(...) seen */ 3515 Bitmask mPrev = 0; 3516 Bitmask mBestNoIn = 0; 3517 3518 /* If the plan produced by the earlier call uses an IN(...) term, call 3519 ** xBestIndex again, this time with IN(...) terms disabled. */ 3520 if( bIn ){ 3521 WHERETRACE(0x40, (" VirtualOne: all usable w/o IN\n")); 3522 rc = whereLoopAddVirtualOne( 3523 pBuilder, mPrereq, ALLBITS, WO_IN, p, mNoOmit, &bIn); 3524 assert( bIn==0 ); 3525 mBestNoIn = pNew->prereq & ~mPrereq; 3526 if( mBestNoIn==0 ){ 3527 seenZero = 1; 3528 seenZeroNoIN = 1; 3529 } 3530 } 3531 3532 /* Call xBestIndex once for each distinct value of (prereqRight & ~mPrereq) 3533 ** in the set of terms that apply to the current virtual table. */ 3534 while( rc==SQLITE_OK ){ 3535 int i; 3536 Bitmask mNext = ALLBITS; 3537 assert( mNext>0 ); 3538 for(i=0; i<nConstraint; i++){ 3539 Bitmask mThis = ( 3540 pWC->a[p->aConstraint[i].iTermOffset].prereqRight & ~mPrereq 3541 ); 3542 if( mThis>mPrev && mThis<mNext ) mNext = mThis; 3543 } 3544 mPrev = mNext; 3545 if( mNext==ALLBITS ) break; 3546 if( mNext==mBest || mNext==mBestNoIn ) continue; 3547 WHERETRACE(0x40, (" VirtualOne: mPrev=%04llx mNext=%04llx\n", 3548 (sqlite3_uint64)mPrev, (sqlite3_uint64)mNext)); 3549 rc = whereLoopAddVirtualOne( 3550 pBuilder, mPrereq, mNext|mPrereq, 0, p, mNoOmit, &bIn); 3551 if( pNew->prereq==mPrereq ){ 3552 seenZero = 1; 3553 if( bIn==0 ) seenZeroNoIN = 1; 3554 } 3555 } 3556 3557 /* If the calls to xBestIndex() in the above loop did not find a plan 3558 ** that requires no source tables at all (i.e. one guaranteed to be 3559 ** usable), make a call here with all source tables disabled */ 3560 if( rc==SQLITE_OK && seenZero==0 ){ 3561 WHERETRACE(0x40, (" VirtualOne: all disabled\n")); 3562 rc = whereLoopAddVirtualOne( 3563 pBuilder, mPrereq, mPrereq, 0, p, mNoOmit, &bIn); 3564 if( bIn==0 ) seenZeroNoIN = 1; 3565 } 3566 3567 /* If the calls to xBestIndex() have so far failed to find a plan 3568 ** that requires no source tables at all and does not use an IN(...) 3569 ** operator, make a final call to obtain one here. */ 3570 if( rc==SQLITE_OK && seenZeroNoIN==0 ){ 3571 WHERETRACE(0x40, (" VirtualOne: all disabled and w/o IN\n")); 3572 rc = whereLoopAddVirtualOne( 3573 pBuilder, mPrereq, mPrereq, WO_IN, p, mNoOmit, &bIn); 3574 } 3575 } 3576 3577 if( p->needToFreeIdxStr ) sqlite3_free(p->idxStr); 3578 sqlite3DbFreeNN(pParse->db, p); 3579 WHERETRACE(0x800, ("END %s.addVirtual(), rc=%d\n", pSrc->pTab->zName, rc)); 3580 return rc; 3581 } 3582 #endif /* SQLITE_OMIT_VIRTUALTABLE */ 3583 3584 /* 3585 ** Add WhereLoop entries to handle OR terms. This works for either 3586 ** btrees or virtual tables. 3587 */ 3588 static int whereLoopAddOr( 3589 WhereLoopBuilder *pBuilder, 3590 Bitmask mPrereq, 3591 Bitmask mUnusable 3592 ){ 3593 WhereInfo *pWInfo = pBuilder->pWInfo; 3594 WhereClause *pWC; 3595 WhereLoop *pNew; 3596 WhereTerm *pTerm, *pWCEnd; 3597 int rc = SQLITE_OK; 3598 int iCur; 3599 WhereClause tempWC; 3600 WhereLoopBuilder sSubBuild; 3601 WhereOrSet sSum, sCur; 3602 SrcItem *pItem; 3603 3604 pWC = pBuilder->pWC; 3605 pWCEnd = pWC->a + pWC->nTerm; 3606 pNew = pBuilder->pNew; 3607 memset(&sSum, 0, sizeof(sSum)); 3608 pItem = pWInfo->pTabList->a + pNew->iTab; 3609 iCur = pItem->iCursor; 3610 3611 for(pTerm=pWC->a; pTerm<pWCEnd && rc==SQLITE_OK; pTerm++){ 3612 if( (pTerm->eOperator & WO_OR)!=0 3613 && (pTerm->u.pOrInfo->indexable & pNew->maskSelf)!=0 3614 ){ 3615 WhereClause * const pOrWC = &pTerm->u.pOrInfo->wc; 3616 WhereTerm * const pOrWCEnd = &pOrWC->a[pOrWC->nTerm]; 3617 WhereTerm *pOrTerm; 3618 int once = 1; 3619 int i, j; 3620 3621 sSubBuild = *pBuilder; 3622 sSubBuild.pOrderBy = 0; 3623 sSubBuild.pOrSet = &sCur; 3624 3625 WHERETRACE(0x200, ("Begin processing OR-clause %p\n", pTerm)); 3626 for(pOrTerm=pOrWC->a; pOrTerm<pOrWCEnd; pOrTerm++){ 3627 if( (pOrTerm->eOperator & WO_AND)!=0 ){ 3628 sSubBuild.pWC = &pOrTerm->u.pAndInfo->wc; 3629 }else if( pOrTerm->leftCursor==iCur ){ 3630 tempWC.pWInfo = pWC->pWInfo; 3631 tempWC.pOuter = pWC; 3632 tempWC.op = TK_AND; 3633 tempWC.nTerm = 1; 3634 tempWC.a = pOrTerm; 3635 sSubBuild.pWC = &tempWC; 3636 }else{ 3637 continue; 3638 } 3639 sCur.n = 0; 3640 #ifdef WHERETRACE_ENABLED 3641 WHERETRACE(0x200, ("OR-term %d of %p has %d subterms:\n", 3642 (int)(pOrTerm-pOrWC->a), pTerm, sSubBuild.pWC->nTerm)); 3643 if( sqlite3WhereTrace & 0x400 ){ 3644 sqlite3WhereClausePrint(sSubBuild.pWC); 3645 } 3646 #endif 3647 #ifndef SQLITE_OMIT_VIRTUALTABLE 3648 if( IsVirtual(pItem->pTab) ){ 3649 rc = whereLoopAddVirtual(&sSubBuild, mPrereq, mUnusable); 3650 }else 3651 #endif 3652 { 3653 rc = whereLoopAddBtree(&sSubBuild, mPrereq); 3654 } 3655 if( rc==SQLITE_OK ){ 3656 rc = whereLoopAddOr(&sSubBuild, mPrereq, mUnusable); 3657 } 3658 assert( rc==SQLITE_OK || rc==SQLITE_DONE || sCur.n==0 3659 || rc==SQLITE_NOMEM ); 3660 testcase( rc==SQLITE_NOMEM && sCur.n>0 ); 3661 testcase( rc==SQLITE_DONE ); 3662 if( sCur.n==0 ){ 3663 sSum.n = 0; 3664 break; 3665 }else if( once ){ 3666 whereOrMove(&sSum, &sCur); 3667 once = 0; 3668 }else{ 3669 WhereOrSet sPrev; 3670 whereOrMove(&sPrev, &sSum); 3671 sSum.n = 0; 3672 for(i=0; i<sPrev.n; i++){ 3673 for(j=0; j<sCur.n; j++){ 3674 whereOrInsert(&sSum, sPrev.a[i].prereq | sCur.a[j].prereq, 3675 sqlite3LogEstAdd(sPrev.a[i].rRun, sCur.a[j].rRun), 3676 sqlite3LogEstAdd(sPrev.a[i].nOut, sCur.a[j].nOut)); 3677 } 3678 } 3679 } 3680 } 3681 pNew->nLTerm = 1; 3682 pNew->aLTerm[0] = pTerm; 3683 pNew->wsFlags = WHERE_MULTI_OR; 3684 pNew->rSetup = 0; 3685 pNew->iSortIdx = 0; 3686 memset(&pNew->u, 0, sizeof(pNew->u)); 3687 for(i=0; rc==SQLITE_OK && i<sSum.n; i++){ 3688 /* TUNING: Currently sSum.a[i].rRun is set to the sum of the costs 3689 ** of all sub-scans required by the OR-scan. However, due to rounding 3690 ** errors, it may be that the cost of the OR-scan is equal to its 3691 ** most expensive sub-scan. Add the smallest possible penalty 3692 ** (equivalent to multiplying the cost by 1.07) to ensure that 3693 ** this does not happen. Otherwise, for WHERE clauses such as the 3694 ** following where there is an index on "y": 3695 ** 3696 ** WHERE likelihood(x=?, 0.99) OR y=? 3697 ** 3698 ** the planner may elect to "OR" together a full-table scan and an 3699 ** index lookup. And other similarly odd results. */ 3700 pNew->rRun = sSum.a[i].rRun + 1; 3701 pNew->nOut = sSum.a[i].nOut; 3702 pNew->prereq = sSum.a[i].prereq; 3703 rc = whereLoopInsert(pBuilder, pNew); 3704 } 3705 WHERETRACE(0x200, ("End processing OR-clause %p\n", pTerm)); 3706 } 3707 } 3708 return rc; 3709 } 3710 3711 /* 3712 ** Add all WhereLoop objects for all tables 3713 */ 3714 static int whereLoopAddAll(WhereLoopBuilder *pBuilder){ 3715 WhereInfo *pWInfo = pBuilder->pWInfo; 3716 Bitmask mPrereq = 0; 3717 Bitmask mPrior = 0; 3718 int iTab; 3719 SrcList *pTabList = pWInfo->pTabList; 3720 SrcItem *pItem; 3721 SrcItem *pEnd = &pTabList->a[pWInfo->nLevel]; 3722 sqlite3 *db = pWInfo->pParse->db; 3723 int rc = SQLITE_OK; 3724 WhereLoop *pNew; 3725 3726 /* Loop over the tables in the join, from left to right */ 3727 pNew = pBuilder->pNew; 3728 whereLoopInit(pNew); 3729 pBuilder->iPlanLimit = SQLITE_QUERY_PLANNER_LIMIT; 3730 for(iTab=0, pItem=pTabList->a; pItem<pEnd; iTab++, pItem++){ 3731 Bitmask mUnusable = 0; 3732 pNew->iTab = iTab; 3733 pBuilder->iPlanLimit += SQLITE_QUERY_PLANNER_LIMIT_INCR; 3734 pNew->maskSelf = sqlite3WhereGetMask(&pWInfo->sMaskSet, pItem->iCursor); 3735 if( (pItem->fg.jointype & (JT_LEFT|JT_CROSS))!=0 ){ 3736 /* This condition is true when pItem is the FROM clause term on the 3737 ** right-hand-side of a LEFT or CROSS JOIN. */ 3738 mPrereq = mPrior; 3739 }else{ 3740 mPrereq = 0; 3741 } 3742 #ifndef SQLITE_OMIT_VIRTUALTABLE 3743 if( IsVirtual(pItem->pTab) ){ 3744 SrcItem *p; 3745 for(p=&pItem[1]; p<pEnd; p++){ 3746 if( mUnusable || (p->fg.jointype & (JT_LEFT|JT_CROSS)) ){ 3747 mUnusable |= sqlite3WhereGetMask(&pWInfo->sMaskSet, p->iCursor); 3748 } 3749 } 3750 rc = whereLoopAddVirtual(pBuilder, mPrereq, mUnusable); 3751 }else 3752 #endif /* SQLITE_OMIT_VIRTUALTABLE */ 3753 { 3754 rc = whereLoopAddBtree(pBuilder, mPrereq); 3755 } 3756 if( rc==SQLITE_OK && pBuilder->pWC->hasOr ){ 3757 rc = whereLoopAddOr(pBuilder, mPrereq, mUnusable); 3758 } 3759 mPrior |= pNew->maskSelf; 3760 if( rc || db->mallocFailed ){ 3761 if( rc==SQLITE_DONE ){ 3762 /* We hit the query planner search limit set by iPlanLimit */ 3763 sqlite3_log(SQLITE_WARNING, "abbreviated query algorithm search"); 3764 rc = SQLITE_OK; 3765 }else{ 3766 break; 3767 } 3768 } 3769 } 3770 3771 whereLoopClear(db, pNew); 3772 return rc; 3773 } 3774 3775 /* 3776 ** Examine a WherePath (with the addition of the extra WhereLoop of the 6th 3777 ** parameters) to see if it outputs rows in the requested ORDER BY 3778 ** (or GROUP BY) without requiring a separate sort operation. Return N: 3779 ** 3780 ** N>0: N terms of the ORDER BY clause are satisfied 3781 ** N==0: No terms of the ORDER BY clause are satisfied 3782 ** N<0: Unknown yet how many terms of ORDER BY might be satisfied. 3783 ** 3784 ** Note that processing for WHERE_GROUPBY and WHERE_DISTINCTBY is not as 3785 ** strict. With GROUP BY and DISTINCT the only requirement is that 3786 ** equivalent rows appear immediately adjacent to one another. GROUP BY 3787 ** and DISTINCT do not require rows to appear in any particular order as long 3788 ** as equivalent rows are grouped together. Thus for GROUP BY and DISTINCT 3789 ** the pOrderBy terms can be matched in any order. With ORDER BY, the 3790 ** pOrderBy terms must be matched in strict left-to-right order. 3791 */ 3792 static i8 wherePathSatisfiesOrderBy( 3793 WhereInfo *pWInfo, /* The WHERE clause */ 3794 ExprList *pOrderBy, /* ORDER BY or GROUP BY or DISTINCT clause to check */ 3795 WherePath *pPath, /* The WherePath to check */ 3796 u16 wctrlFlags, /* WHERE_GROUPBY or _DISTINCTBY or _ORDERBY_LIMIT */ 3797 u16 nLoop, /* Number of entries in pPath->aLoop[] */ 3798 WhereLoop *pLast, /* Add this WhereLoop to the end of pPath->aLoop[] */ 3799 Bitmask *pRevMask /* OUT: Mask of WhereLoops to run in reverse order */ 3800 ){ 3801 u8 revSet; /* True if rev is known */ 3802 u8 rev; /* Composite sort order */ 3803 u8 revIdx; /* Index sort order */ 3804 u8 isOrderDistinct; /* All prior WhereLoops are order-distinct */ 3805 u8 distinctColumns; /* True if the loop has UNIQUE NOT NULL columns */ 3806 u8 isMatch; /* iColumn matches a term of the ORDER BY clause */ 3807 u16 eqOpMask; /* Allowed equality operators */ 3808 u16 nKeyCol; /* Number of key columns in pIndex */ 3809 u16 nColumn; /* Total number of ordered columns in the index */ 3810 u16 nOrderBy; /* Number terms in the ORDER BY clause */ 3811 int iLoop; /* Index of WhereLoop in pPath being processed */ 3812 int i, j; /* Loop counters */ 3813 int iCur; /* Cursor number for current WhereLoop */ 3814 int iColumn; /* A column number within table iCur */ 3815 WhereLoop *pLoop = 0; /* Current WhereLoop being processed. */ 3816 WhereTerm *pTerm; /* A single term of the WHERE clause */ 3817 Expr *pOBExpr; /* An expression from the ORDER BY clause */ 3818 CollSeq *pColl; /* COLLATE function from an ORDER BY clause term */ 3819 Index *pIndex; /* The index associated with pLoop */ 3820 sqlite3 *db = pWInfo->pParse->db; /* Database connection */ 3821 Bitmask obSat = 0; /* Mask of ORDER BY terms satisfied so far */ 3822 Bitmask obDone; /* Mask of all ORDER BY terms */ 3823 Bitmask orderDistinctMask; /* Mask of all well-ordered loops */ 3824 Bitmask ready; /* Mask of inner loops */ 3825 3826 /* 3827 ** We say the WhereLoop is "one-row" if it generates no more than one 3828 ** row of output. A WhereLoop is one-row if all of the following are true: 3829 ** (a) All index columns match with WHERE_COLUMN_EQ. 3830 ** (b) The index is unique 3831 ** Any WhereLoop with an WHERE_COLUMN_EQ constraint on the rowid is one-row. 3832 ** Every one-row WhereLoop will have the WHERE_ONEROW bit set in wsFlags. 3833 ** 3834 ** We say the WhereLoop is "order-distinct" if the set of columns from 3835 ** that WhereLoop that are in the ORDER BY clause are different for every 3836 ** row of the WhereLoop. Every one-row WhereLoop is automatically 3837 ** order-distinct. A WhereLoop that has no columns in the ORDER BY clause 3838 ** is not order-distinct. To be order-distinct is not quite the same as being 3839 ** UNIQUE since a UNIQUE column or index can have multiple rows that 3840 ** are NULL and NULL values are equivalent for the purpose of order-distinct. 3841 ** To be order-distinct, the columns must be UNIQUE and NOT NULL. 3842 ** 3843 ** The rowid for a table is always UNIQUE and NOT NULL so whenever the 3844 ** rowid appears in the ORDER BY clause, the corresponding WhereLoop is 3845 ** automatically order-distinct. 3846 */ 3847 3848 assert( pOrderBy!=0 ); 3849 if( nLoop && OptimizationDisabled(db, SQLITE_OrderByIdxJoin) ) return 0; 3850 3851 nOrderBy = pOrderBy->nExpr; 3852 testcase( nOrderBy==BMS-1 ); 3853 if( nOrderBy>BMS-1 ) return 0; /* Cannot optimize overly large ORDER BYs */ 3854 isOrderDistinct = 1; 3855 obDone = MASKBIT(nOrderBy)-1; 3856 orderDistinctMask = 0; 3857 ready = 0; 3858 eqOpMask = WO_EQ | WO_IS | WO_ISNULL; 3859 if( wctrlFlags & (WHERE_ORDERBY_LIMIT|WHERE_ORDERBY_MAX|WHERE_ORDERBY_MIN) ){ 3860 eqOpMask |= WO_IN; 3861 } 3862 for(iLoop=0; isOrderDistinct && obSat<obDone && iLoop<=nLoop; iLoop++){ 3863 if( iLoop>0 ) ready |= pLoop->maskSelf; 3864 if( iLoop<nLoop ){ 3865 pLoop = pPath->aLoop[iLoop]; 3866 if( wctrlFlags & WHERE_ORDERBY_LIMIT ) continue; 3867 }else{ 3868 pLoop = pLast; 3869 } 3870 if( pLoop->wsFlags & WHERE_VIRTUALTABLE ){ 3871 if( pLoop->u.vtab.isOrdered && (wctrlFlags & WHERE_DISTINCTBY)==0 ){ 3872 obSat = obDone; 3873 } 3874 break; 3875 }else if( wctrlFlags & WHERE_DISTINCTBY ){ 3876 pLoop->u.btree.nDistinctCol = 0; 3877 } 3878 iCur = pWInfo->pTabList->a[pLoop->iTab].iCursor; 3879 3880 /* Mark off any ORDER BY term X that is a column in the table of 3881 ** the current loop for which there is term in the WHERE 3882 ** clause of the form X IS NULL or X=? that reference only outer 3883 ** loops. 3884 */ 3885 for(i=0; i<nOrderBy; i++){ 3886 if( MASKBIT(i) & obSat ) continue; 3887 pOBExpr = sqlite3ExprSkipCollateAndLikely(pOrderBy->a[i].pExpr); 3888 if( NEVER(pOBExpr==0) ) continue; 3889 if( pOBExpr->op!=TK_COLUMN && pOBExpr->op!=TK_AGG_COLUMN ) continue; 3890 if( pOBExpr->iTable!=iCur ) continue; 3891 pTerm = sqlite3WhereFindTerm(&pWInfo->sWC, iCur, pOBExpr->iColumn, 3892 ~ready, eqOpMask, 0); 3893 if( pTerm==0 ) continue; 3894 if( pTerm->eOperator==WO_IN ){ 3895 /* IN terms are only valid for sorting in the ORDER BY LIMIT 3896 ** optimization, and then only if they are actually used 3897 ** by the query plan */ 3898 assert( wctrlFlags & 3899 (WHERE_ORDERBY_LIMIT|WHERE_ORDERBY_MIN|WHERE_ORDERBY_MAX) ); 3900 for(j=0; j<pLoop->nLTerm && pTerm!=pLoop->aLTerm[j]; j++){} 3901 if( j>=pLoop->nLTerm ) continue; 3902 } 3903 if( (pTerm->eOperator&(WO_EQ|WO_IS))!=0 && pOBExpr->iColumn>=0 ){ 3904 Parse *pParse = pWInfo->pParse; 3905 CollSeq *pColl1 = sqlite3ExprNNCollSeq(pParse, pOrderBy->a[i].pExpr); 3906 CollSeq *pColl2 = sqlite3ExprCompareCollSeq(pParse, pTerm->pExpr); 3907 assert( pColl1 ); 3908 if( pColl2==0 || sqlite3StrICmp(pColl1->zName, pColl2->zName) ){ 3909 continue; 3910 } 3911 testcase( pTerm->pExpr->op==TK_IS ); 3912 } 3913 obSat |= MASKBIT(i); 3914 } 3915 3916 if( (pLoop->wsFlags & WHERE_ONEROW)==0 ){ 3917 if( pLoop->wsFlags & WHERE_IPK ){ 3918 pIndex = 0; 3919 nKeyCol = 0; 3920 nColumn = 1; 3921 }else if( (pIndex = pLoop->u.btree.pIndex)==0 || pIndex->bUnordered ){ 3922 return 0; 3923 }else{ 3924 nKeyCol = pIndex->nKeyCol; 3925 nColumn = pIndex->nColumn; 3926 assert( nColumn==nKeyCol+1 || !HasRowid(pIndex->pTable) ); 3927 assert( pIndex->aiColumn[nColumn-1]==XN_ROWID 3928 || !HasRowid(pIndex->pTable)); 3929 /* All relevant terms of the index must also be non-NULL in order 3930 ** for isOrderDistinct to be true. So the isOrderDistint value 3931 ** computed here might be a false positive. Corrections will be 3932 ** made at tag-20210426-1 below */ 3933 isOrderDistinct = IsUniqueIndex(pIndex) 3934 && (pLoop->wsFlags & WHERE_SKIPSCAN)==0; 3935 } 3936 3937 /* Loop through all columns of the index and deal with the ones 3938 ** that are not constrained by == or IN. 3939 */ 3940 rev = revSet = 0; 3941 distinctColumns = 0; 3942 for(j=0; j<nColumn; j++){ 3943 u8 bOnce = 1; /* True to run the ORDER BY search loop */ 3944 3945 assert( j>=pLoop->u.btree.nEq 3946 || (pLoop->aLTerm[j]==0)==(j<pLoop->nSkip) 3947 ); 3948 if( j<pLoop->u.btree.nEq && j>=pLoop->nSkip ){ 3949 u16 eOp = pLoop->aLTerm[j]->eOperator; 3950 3951 /* Skip over == and IS and ISNULL terms. (Also skip IN terms when 3952 ** doing WHERE_ORDERBY_LIMIT processing). Except, IS and ISNULL 3953 ** terms imply that the index is not UNIQUE NOT NULL in which case 3954 ** the loop need to be marked as not order-distinct because it can 3955 ** have repeated NULL rows. 3956 ** 3957 ** If the current term is a column of an ((?,?) IN (SELECT...)) 3958 ** expression for which the SELECT returns more than one column, 3959 ** check that it is the only column used by this loop. Otherwise, 3960 ** if it is one of two or more, none of the columns can be 3961 ** considered to match an ORDER BY term. 3962 */ 3963 if( (eOp & eqOpMask)!=0 ){ 3964 if( eOp & (WO_ISNULL|WO_IS) ){ 3965 testcase( eOp & WO_ISNULL ); 3966 testcase( eOp & WO_IS ); 3967 testcase( isOrderDistinct ); 3968 isOrderDistinct = 0; 3969 } 3970 continue; 3971 }else if( ALWAYS(eOp & WO_IN) ){ 3972 /* ALWAYS() justification: eOp is an equality operator due to the 3973 ** j<pLoop->u.btree.nEq constraint above. Any equality other 3974 ** than WO_IN is captured by the previous "if". So this one 3975 ** always has to be WO_IN. */ 3976 Expr *pX = pLoop->aLTerm[j]->pExpr; 3977 for(i=j+1; i<pLoop->u.btree.nEq; i++){ 3978 if( pLoop->aLTerm[i]->pExpr==pX ){ 3979 assert( (pLoop->aLTerm[i]->eOperator & WO_IN) ); 3980 bOnce = 0; 3981 break; 3982 } 3983 } 3984 } 3985 } 3986 3987 /* Get the column number in the table (iColumn) and sort order 3988 ** (revIdx) for the j-th column of the index. 3989 */ 3990 if( pIndex ){ 3991 iColumn = pIndex->aiColumn[j]; 3992 revIdx = pIndex->aSortOrder[j] & KEYINFO_ORDER_DESC; 3993 if( iColumn==pIndex->pTable->iPKey ) iColumn = XN_ROWID; 3994 }else{ 3995 iColumn = XN_ROWID; 3996 revIdx = 0; 3997 } 3998 3999 /* An unconstrained column that might be NULL means that this 4000 ** WhereLoop is not well-ordered. tag-20210426-1 4001 */ 4002 if( isOrderDistinct ){ 4003 if( iColumn>=0 4004 && j>=pLoop->u.btree.nEq 4005 && pIndex->pTable->aCol[iColumn].notNull==0 4006 ){ 4007 isOrderDistinct = 0; 4008 } 4009 if( iColumn==XN_EXPR ){ 4010 isOrderDistinct = 0; 4011 } 4012 } 4013 4014 /* Find the ORDER BY term that corresponds to the j-th column 4015 ** of the index and mark that ORDER BY term off 4016 */ 4017 isMatch = 0; 4018 for(i=0; bOnce && i<nOrderBy; i++){ 4019 if( MASKBIT(i) & obSat ) continue; 4020 pOBExpr = sqlite3ExprSkipCollateAndLikely(pOrderBy->a[i].pExpr); 4021 testcase( wctrlFlags & WHERE_GROUPBY ); 4022 testcase( wctrlFlags & WHERE_DISTINCTBY ); 4023 if( NEVER(pOBExpr==0) ) continue; 4024 if( (wctrlFlags & (WHERE_GROUPBY|WHERE_DISTINCTBY))==0 ) bOnce = 0; 4025 if( iColumn>=XN_ROWID ){ 4026 if( pOBExpr->op!=TK_COLUMN && pOBExpr->op!=TK_AGG_COLUMN ) continue; 4027 if( pOBExpr->iTable!=iCur ) continue; 4028 if( pOBExpr->iColumn!=iColumn ) continue; 4029 }else{ 4030 Expr *pIdxExpr = pIndex->aColExpr->a[j].pExpr; 4031 if( sqlite3ExprCompareSkip(pOBExpr, pIdxExpr, iCur) ){ 4032 continue; 4033 } 4034 } 4035 if( iColumn!=XN_ROWID ){ 4036 pColl = sqlite3ExprNNCollSeq(pWInfo->pParse, pOrderBy->a[i].pExpr); 4037 if( sqlite3StrICmp(pColl->zName, pIndex->azColl[j])!=0 ) continue; 4038 } 4039 if( wctrlFlags & WHERE_DISTINCTBY ){ 4040 pLoop->u.btree.nDistinctCol = j+1; 4041 } 4042 isMatch = 1; 4043 break; 4044 } 4045 if( isMatch && (wctrlFlags & WHERE_GROUPBY)==0 ){ 4046 /* Make sure the sort order is compatible in an ORDER BY clause. 4047 ** Sort order is irrelevant for a GROUP BY clause. */ 4048 if( revSet ){ 4049 if( (rev ^ revIdx)!=(pOrderBy->a[i].sortFlags&KEYINFO_ORDER_DESC) ){ 4050 isMatch = 0; 4051 } 4052 }else{ 4053 rev = revIdx ^ (pOrderBy->a[i].sortFlags & KEYINFO_ORDER_DESC); 4054 if( rev ) *pRevMask |= MASKBIT(iLoop); 4055 revSet = 1; 4056 } 4057 } 4058 if( isMatch && (pOrderBy->a[i].sortFlags & KEYINFO_ORDER_BIGNULL) ){ 4059 if( j==pLoop->u.btree.nEq ){ 4060 pLoop->wsFlags |= WHERE_BIGNULL_SORT; 4061 }else{ 4062 isMatch = 0; 4063 } 4064 } 4065 if( isMatch ){ 4066 if( iColumn==XN_ROWID ){ 4067 testcase( distinctColumns==0 ); 4068 distinctColumns = 1; 4069 } 4070 obSat |= MASKBIT(i); 4071 }else{ 4072 /* No match found */ 4073 if( j==0 || j<nKeyCol ){ 4074 testcase( isOrderDistinct!=0 ); 4075 isOrderDistinct = 0; 4076 } 4077 break; 4078 } 4079 } /* end Loop over all index columns */ 4080 if( distinctColumns ){ 4081 testcase( isOrderDistinct==0 ); 4082 isOrderDistinct = 1; 4083 } 4084 } /* end-if not one-row */ 4085 4086 /* Mark off any other ORDER BY terms that reference pLoop */ 4087 if( isOrderDistinct ){ 4088 orderDistinctMask |= pLoop->maskSelf; 4089 for(i=0; i<nOrderBy; i++){ 4090 Expr *p; 4091 Bitmask mTerm; 4092 if( MASKBIT(i) & obSat ) continue; 4093 p = pOrderBy->a[i].pExpr; 4094 mTerm = sqlite3WhereExprUsage(&pWInfo->sMaskSet,p); 4095 if( mTerm==0 && !sqlite3ExprIsConstant(p) ) continue; 4096 if( (mTerm&~orderDistinctMask)==0 ){ 4097 obSat |= MASKBIT(i); 4098 } 4099 } 4100 } 4101 } /* End the loop over all WhereLoops from outer-most down to inner-most */ 4102 if( obSat==obDone ) return (i8)nOrderBy; 4103 if( !isOrderDistinct ){ 4104 for(i=nOrderBy-1; i>0; i--){ 4105 Bitmask m = ALWAYS(i<BMS) ? MASKBIT(i) - 1 : 0; 4106 if( (obSat&m)==m ) return i; 4107 } 4108 return 0; 4109 } 4110 return -1; 4111 } 4112 4113 4114 /* 4115 ** If the WHERE_GROUPBY flag is set in the mask passed to sqlite3WhereBegin(), 4116 ** the planner assumes that the specified pOrderBy list is actually a GROUP 4117 ** BY clause - and so any order that groups rows as required satisfies the 4118 ** request. 4119 ** 4120 ** Normally, in this case it is not possible for the caller to determine 4121 ** whether or not the rows are really being delivered in sorted order, or 4122 ** just in some other order that provides the required grouping. However, 4123 ** if the WHERE_SORTBYGROUP flag is also passed to sqlite3WhereBegin(), then 4124 ** this function may be called on the returned WhereInfo object. It returns 4125 ** true if the rows really will be sorted in the specified order, or false 4126 ** otherwise. 4127 ** 4128 ** For example, assuming: 4129 ** 4130 ** CREATE INDEX i1 ON t1(x, Y); 4131 ** 4132 ** then 4133 ** 4134 ** SELECT * FROM t1 GROUP BY x,y ORDER BY x,y; -- IsSorted()==1 4135 ** SELECT * FROM t1 GROUP BY y,x ORDER BY y,x; -- IsSorted()==0 4136 */ 4137 int sqlite3WhereIsSorted(WhereInfo *pWInfo){ 4138 assert( pWInfo->wctrlFlags & WHERE_GROUPBY ); 4139 assert( pWInfo->wctrlFlags & WHERE_SORTBYGROUP ); 4140 return pWInfo->sorted; 4141 } 4142 4143 #ifdef WHERETRACE_ENABLED 4144 /* For debugging use only: */ 4145 static const char *wherePathName(WherePath *pPath, int nLoop, WhereLoop *pLast){ 4146 static char zName[65]; 4147 int i; 4148 for(i=0; i<nLoop; i++){ zName[i] = pPath->aLoop[i]->cId; } 4149 if( pLast ) zName[i++] = pLast->cId; 4150 zName[i] = 0; 4151 return zName; 4152 } 4153 #endif 4154 4155 /* 4156 ** Return the cost of sorting nRow rows, assuming that the keys have 4157 ** nOrderby columns and that the first nSorted columns are already in 4158 ** order. 4159 */ 4160 static LogEst whereSortingCost( 4161 WhereInfo *pWInfo, 4162 LogEst nRow, 4163 int nOrderBy, 4164 int nSorted 4165 ){ 4166 /* TUNING: Estimated cost of a full external sort, where N is 4167 ** the number of rows to sort is: 4168 ** 4169 ** cost = (3.0 * N * log(N)). 4170 ** 4171 ** Or, if the order-by clause has X terms but only the last Y 4172 ** terms are out of order, then block-sorting will reduce the 4173 ** sorting cost to: 4174 ** 4175 ** cost = (3.0 * N * log(N)) * (Y/X) 4176 ** 4177 ** The (Y/X) term is implemented using stack variable rScale 4178 ** below. 4179 */ 4180 LogEst rScale, rSortCost; 4181 assert( nOrderBy>0 && 66==sqlite3LogEst(100) ); 4182 rScale = sqlite3LogEst((nOrderBy-nSorted)*100/nOrderBy) - 66; 4183 rSortCost = nRow + rScale + 16; 4184 4185 /* Multiple by log(M) where M is the number of output rows. 4186 ** Use the LIMIT for M if it is smaller. Or if this sort is for 4187 ** a DISTINCT operator, M will be the number of distinct output 4188 ** rows, so fudge it downwards a bit. 4189 */ 4190 if( (pWInfo->wctrlFlags & WHERE_USE_LIMIT)!=0 && pWInfo->iLimit<nRow ){ 4191 nRow = pWInfo->iLimit; 4192 }else if( (pWInfo->wctrlFlags & WHERE_WANT_DISTINCT) ){ 4193 /* TUNING: In the sort for a DISTINCT operator, assume that the DISTINCT 4194 ** reduces the number of output rows by a factor of 2 */ 4195 if( nRow>10 ){ nRow -= 10; assert( 10==sqlite3LogEst(2) ); } 4196 } 4197 rSortCost += estLog(nRow); 4198 return rSortCost; 4199 } 4200 4201 /* 4202 ** Given the list of WhereLoop objects at pWInfo->pLoops, this routine 4203 ** attempts to find the lowest cost path that visits each WhereLoop 4204 ** once. This path is then loaded into the pWInfo->a[].pWLoop fields. 4205 ** 4206 ** Assume that the total number of output rows that will need to be sorted 4207 ** will be nRowEst (in the 10*log2 representation). Or, ignore sorting 4208 ** costs if nRowEst==0. 4209 ** 4210 ** Return SQLITE_OK on success or SQLITE_NOMEM of a memory allocation 4211 ** error occurs. 4212 */ 4213 static int wherePathSolver(WhereInfo *pWInfo, LogEst nRowEst){ 4214 int mxChoice; /* Maximum number of simultaneous paths tracked */ 4215 int nLoop; /* Number of terms in the join */ 4216 Parse *pParse; /* Parsing context */ 4217 sqlite3 *db; /* The database connection */ 4218 int iLoop; /* Loop counter over the terms of the join */ 4219 int ii, jj; /* Loop counters */ 4220 int mxI = 0; /* Index of next entry to replace */ 4221 int nOrderBy; /* Number of ORDER BY clause terms */ 4222 LogEst mxCost = 0; /* Maximum cost of a set of paths */ 4223 LogEst mxUnsorted = 0; /* Maximum unsorted cost of a set of path */ 4224 int nTo, nFrom; /* Number of valid entries in aTo[] and aFrom[] */ 4225 WherePath *aFrom; /* All nFrom paths at the previous level */ 4226 WherePath *aTo; /* The nTo best paths at the current level */ 4227 WherePath *pFrom; /* An element of aFrom[] that we are working on */ 4228 WherePath *pTo; /* An element of aTo[] that we are working on */ 4229 WhereLoop *pWLoop; /* One of the WhereLoop objects */ 4230 WhereLoop **pX; /* Used to divy up the pSpace memory */ 4231 LogEst *aSortCost = 0; /* Sorting and partial sorting costs */ 4232 char *pSpace; /* Temporary memory used by this routine */ 4233 int nSpace; /* Bytes of space allocated at pSpace */ 4234 4235 pParse = pWInfo->pParse; 4236 db = pParse->db; 4237 nLoop = pWInfo->nLevel; 4238 /* TUNING: For simple queries, only the best path is tracked. 4239 ** For 2-way joins, the 5 best paths are followed. 4240 ** For joins of 3 or more tables, track the 10 best paths */ 4241 mxChoice = (nLoop<=1) ? 1 : (nLoop==2 ? 5 : 10); 4242 assert( nLoop<=pWInfo->pTabList->nSrc ); 4243 WHERETRACE(0x002, ("---- begin solver. (nRowEst=%d)\n", nRowEst)); 4244 4245 /* If nRowEst is zero and there is an ORDER BY clause, ignore it. In this 4246 ** case the purpose of this call is to estimate the number of rows returned 4247 ** by the overall query. Once this estimate has been obtained, the caller 4248 ** will invoke this function a second time, passing the estimate as the 4249 ** nRowEst parameter. */ 4250 if( pWInfo->pOrderBy==0 || nRowEst==0 ){ 4251 nOrderBy = 0; 4252 }else{ 4253 nOrderBy = pWInfo->pOrderBy->nExpr; 4254 } 4255 4256 /* Allocate and initialize space for aTo, aFrom and aSortCost[] */ 4257 nSpace = (sizeof(WherePath)+sizeof(WhereLoop*)*nLoop)*mxChoice*2; 4258 nSpace += sizeof(LogEst) * nOrderBy; 4259 pSpace = sqlite3DbMallocRawNN(db, nSpace); 4260 if( pSpace==0 ) return SQLITE_NOMEM_BKPT; 4261 aTo = (WherePath*)pSpace; 4262 aFrom = aTo+mxChoice; 4263 memset(aFrom, 0, sizeof(aFrom[0])); 4264 pX = (WhereLoop**)(aFrom+mxChoice); 4265 for(ii=mxChoice*2, pFrom=aTo; ii>0; ii--, pFrom++, pX += nLoop){ 4266 pFrom->aLoop = pX; 4267 } 4268 if( nOrderBy ){ 4269 /* If there is an ORDER BY clause and it is not being ignored, set up 4270 ** space for the aSortCost[] array. Each element of the aSortCost array 4271 ** is either zero - meaning it has not yet been initialized - or the 4272 ** cost of sorting nRowEst rows of data where the first X terms of 4273 ** the ORDER BY clause are already in order, where X is the array 4274 ** index. */ 4275 aSortCost = (LogEst*)pX; 4276 memset(aSortCost, 0, sizeof(LogEst) * nOrderBy); 4277 } 4278 assert( aSortCost==0 || &pSpace[nSpace]==(char*)&aSortCost[nOrderBy] ); 4279 assert( aSortCost!=0 || &pSpace[nSpace]==(char*)pX ); 4280 4281 /* Seed the search with a single WherePath containing zero WhereLoops. 4282 ** 4283 ** TUNING: Do not let the number of iterations go above 28. If the cost 4284 ** of computing an automatic index is not paid back within the first 28 4285 ** rows, then do not use the automatic index. */ 4286 aFrom[0].nRow = MIN(pParse->nQueryLoop, 48); assert( 48==sqlite3LogEst(28) ); 4287 nFrom = 1; 4288 assert( aFrom[0].isOrdered==0 ); 4289 if( nOrderBy ){ 4290 /* If nLoop is zero, then there are no FROM terms in the query. Since 4291 ** in this case the query may return a maximum of one row, the results 4292 ** are already in the requested order. Set isOrdered to nOrderBy to 4293 ** indicate this. Or, if nLoop is greater than zero, set isOrdered to 4294 ** -1, indicating that the result set may or may not be ordered, 4295 ** depending on the loops added to the current plan. */ 4296 aFrom[0].isOrdered = nLoop>0 ? -1 : nOrderBy; 4297 } 4298 4299 /* Compute successively longer WherePaths using the previous generation 4300 ** of WherePaths as the basis for the next. Keep track of the mxChoice 4301 ** best paths at each generation */ 4302 for(iLoop=0; iLoop<nLoop; iLoop++){ 4303 nTo = 0; 4304 for(ii=0, pFrom=aFrom; ii<nFrom; ii++, pFrom++){ 4305 for(pWLoop=pWInfo->pLoops; pWLoop; pWLoop=pWLoop->pNextLoop){ 4306 LogEst nOut; /* Rows visited by (pFrom+pWLoop) */ 4307 LogEst rCost; /* Cost of path (pFrom+pWLoop) */ 4308 LogEst rUnsorted; /* Unsorted cost of (pFrom+pWLoop) */ 4309 i8 isOrdered = pFrom->isOrdered; /* isOrdered for (pFrom+pWLoop) */ 4310 Bitmask maskNew; /* Mask of src visited by (..) */ 4311 Bitmask revMask = 0; /* Mask of rev-order loops for (..) */ 4312 4313 if( (pWLoop->prereq & ~pFrom->maskLoop)!=0 ) continue; 4314 if( (pWLoop->maskSelf & pFrom->maskLoop)!=0 ) continue; 4315 if( (pWLoop->wsFlags & WHERE_AUTO_INDEX)!=0 && pFrom->nRow<3 ){ 4316 /* Do not use an automatic index if the this loop is expected 4317 ** to run less than 1.25 times. It is tempting to also exclude 4318 ** automatic index usage on an outer loop, but sometimes an automatic 4319 ** index is useful in the outer loop of a correlated subquery. */ 4320 assert( 10==sqlite3LogEst(2) ); 4321 continue; 4322 } 4323 4324 /* At this point, pWLoop is a candidate to be the next loop. 4325 ** Compute its cost */ 4326 rUnsorted = sqlite3LogEstAdd(pWLoop->rSetup,pWLoop->rRun + pFrom->nRow); 4327 rUnsorted = sqlite3LogEstAdd(rUnsorted, pFrom->rUnsorted); 4328 nOut = pFrom->nRow + pWLoop->nOut; 4329 maskNew = pFrom->maskLoop | pWLoop->maskSelf; 4330 if( isOrdered<0 ){ 4331 isOrdered = wherePathSatisfiesOrderBy(pWInfo, 4332 pWInfo->pOrderBy, pFrom, pWInfo->wctrlFlags, 4333 iLoop, pWLoop, &revMask); 4334 }else{ 4335 revMask = pFrom->revLoop; 4336 } 4337 if( isOrdered>=0 && isOrdered<nOrderBy ){ 4338 if( aSortCost[isOrdered]==0 ){ 4339 aSortCost[isOrdered] = whereSortingCost( 4340 pWInfo, nRowEst, nOrderBy, isOrdered 4341 ); 4342 } 4343 /* TUNING: Add a small extra penalty (5) to sorting as an 4344 ** extra encouragment to the query planner to select a plan 4345 ** where the rows emerge in the correct order without any sorting 4346 ** required. */ 4347 rCost = sqlite3LogEstAdd(rUnsorted, aSortCost[isOrdered]) + 5; 4348 4349 WHERETRACE(0x002, 4350 ("---- sort cost=%-3d (%d/%d) increases cost %3d to %-3d\n", 4351 aSortCost[isOrdered], (nOrderBy-isOrdered), nOrderBy, 4352 rUnsorted, rCost)); 4353 }else{ 4354 rCost = rUnsorted; 4355 rUnsorted -= 2; /* TUNING: Slight bias in favor of no-sort plans */ 4356 } 4357 4358 /* Check to see if pWLoop should be added to the set of 4359 ** mxChoice best-so-far paths. 4360 ** 4361 ** First look for an existing path among best-so-far paths 4362 ** that covers the same set of loops and has the same isOrdered 4363 ** setting as the current path candidate. 4364 ** 4365 ** The term "((pTo->isOrdered^isOrdered)&0x80)==0" is equivalent 4366 ** to (pTo->isOrdered==(-1))==(isOrdered==(-1))" for the range 4367 ** of legal values for isOrdered, -1..64. 4368 */ 4369 for(jj=0, pTo=aTo; jj<nTo; jj++, pTo++){ 4370 if( pTo->maskLoop==maskNew 4371 && ((pTo->isOrdered^isOrdered)&0x80)==0 4372 ){ 4373 testcase( jj==nTo-1 ); 4374 break; 4375 } 4376 } 4377 if( jj>=nTo ){ 4378 /* None of the existing best-so-far paths match the candidate. */ 4379 if( nTo>=mxChoice 4380 && (rCost>mxCost || (rCost==mxCost && rUnsorted>=mxUnsorted)) 4381 ){ 4382 /* The current candidate is no better than any of the mxChoice 4383 ** paths currently in the best-so-far buffer. So discard 4384 ** this candidate as not viable. */ 4385 #ifdef WHERETRACE_ENABLED /* 0x4 */ 4386 if( sqlite3WhereTrace&0x4 ){ 4387 sqlite3DebugPrintf("Skip %s cost=%-3d,%3d,%3d order=%c\n", 4388 wherePathName(pFrom, iLoop, pWLoop), rCost, nOut, rUnsorted, 4389 isOrdered>=0 ? isOrdered+'0' : '?'); 4390 } 4391 #endif 4392 continue; 4393 } 4394 /* If we reach this points it means that the new candidate path 4395 ** needs to be added to the set of best-so-far paths. */ 4396 if( nTo<mxChoice ){ 4397 /* Increase the size of the aTo set by one */ 4398 jj = nTo++; 4399 }else{ 4400 /* New path replaces the prior worst to keep count below mxChoice */ 4401 jj = mxI; 4402 } 4403 pTo = &aTo[jj]; 4404 #ifdef WHERETRACE_ENABLED /* 0x4 */ 4405 if( sqlite3WhereTrace&0x4 ){ 4406 sqlite3DebugPrintf("New %s cost=%-3d,%3d,%3d order=%c\n", 4407 wherePathName(pFrom, iLoop, pWLoop), rCost, nOut, rUnsorted, 4408 isOrdered>=0 ? isOrdered+'0' : '?'); 4409 } 4410 #endif 4411 }else{ 4412 /* Control reaches here if best-so-far path pTo=aTo[jj] covers the 4413 ** same set of loops and has the same isOrdered setting as the 4414 ** candidate path. Check to see if the candidate should replace 4415 ** pTo or if the candidate should be skipped. 4416 ** 4417 ** The conditional is an expanded vector comparison equivalent to: 4418 ** (pTo->rCost,pTo->nRow,pTo->rUnsorted) <= (rCost,nOut,rUnsorted) 4419 */ 4420 if( pTo->rCost<rCost 4421 || (pTo->rCost==rCost 4422 && (pTo->nRow<nOut 4423 || (pTo->nRow==nOut && pTo->rUnsorted<=rUnsorted) 4424 ) 4425 ) 4426 ){ 4427 #ifdef WHERETRACE_ENABLED /* 0x4 */ 4428 if( sqlite3WhereTrace&0x4 ){ 4429 sqlite3DebugPrintf( 4430 "Skip %s cost=%-3d,%3d,%3d order=%c", 4431 wherePathName(pFrom, iLoop, pWLoop), rCost, nOut, rUnsorted, 4432 isOrdered>=0 ? isOrdered+'0' : '?'); 4433 sqlite3DebugPrintf(" vs %s cost=%-3d,%3d,%3d order=%c\n", 4434 wherePathName(pTo, iLoop+1, 0), pTo->rCost, pTo->nRow, 4435 pTo->rUnsorted, pTo->isOrdered>=0 ? pTo->isOrdered+'0' : '?'); 4436 } 4437 #endif 4438 /* Discard the candidate path from further consideration */ 4439 testcase( pTo->rCost==rCost ); 4440 continue; 4441 } 4442 testcase( pTo->rCost==rCost+1 ); 4443 /* Control reaches here if the candidate path is better than the 4444 ** pTo path. Replace pTo with the candidate. */ 4445 #ifdef WHERETRACE_ENABLED /* 0x4 */ 4446 if( sqlite3WhereTrace&0x4 ){ 4447 sqlite3DebugPrintf( 4448 "Update %s cost=%-3d,%3d,%3d order=%c", 4449 wherePathName(pFrom, iLoop, pWLoop), rCost, nOut, rUnsorted, 4450 isOrdered>=0 ? isOrdered+'0' : '?'); 4451 sqlite3DebugPrintf(" was %s cost=%-3d,%3d,%3d order=%c\n", 4452 wherePathName(pTo, iLoop+1, 0), pTo->rCost, pTo->nRow, 4453 pTo->rUnsorted, pTo->isOrdered>=0 ? pTo->isOrdered+'0' : '?'); 4454 } 4455 #endif 4456 } 4457 /* pWLoop is a winner. Add it to the set of best so far */ 4458 pTo->maskLoop = pFrom->maskLoop | pWLoop->maskSelf; 4459 pTo->revLoop = revMask; 4460 pTo->nRow = nOut; 4461 pTo->rCost = rCost; 4462 pTo->rUnsorted = rUnsorted; 4463 pTo->isOrdered = isOrdered; 4464 memcpy(pTo->aLoop, pFrom->aLoop, sizeof(WhereLoop*)*iLoop); 4465 pTo->aLoop[iLoop] = pWLoop; 4466 if( nTo>=mxChoice ){ 4467 mxI = 0; 4468 mxCost = aTo[0].rCost; 4469 mxUnsorted = aTo[0].nRow; 4470 for(jj=1, pTo=&aTo[1]; jj<mxChoice; jj++, pTo++){ 4471 if( pTo->rCost>mxCost 4472 || (pTo->rCost==mxCost && pTo->rUnsorted>mxUnsorted) 4473 ){ 4474 mxCost = pTo->rCost; 4475 mxUnsorted = pTo->rUnsorted; 4476 mxI = jj; 4477 } 4478 } 4479 } 4480 } 4481 } 4482 4483 #ifdef WHERETRACE_ENABLED /* >=2 */ 4484 if( sqlite3WhereTrace & 0x02 ){ 4485 sqlite3DebugPrintf("---- after round %d ----\n", iLoop); 4486 for(ii=0, pTo=aTo; ii<nTo; ii++, pTo++){ 4487 sqlite3DebugPrintf(" %s cost=%-3d nrow=%-3d order=%c", 4488 wherePathName(pTo, iLoop+1, 0), pTo->rCost, pTo->nRow, 4489 pTo->isOrdered>=0 ? (pTo->isOrdered+'0') : '?'); 4490 if( pTo->isOrdered>0 ){ 4491 sqlite3DebugPrintf(" rev=0x%llx\n", pTo->revLoop); 4492 }else{ 4493 sqlite3DebugPrintf("\n"); 4494 } 4495 } 4496 } 4497 #endif 4498 4499 /* Swap the roles of aFrom and aTo for the next generation */ 4500 pFrom = aTo; 4501 aTo = aFrom; 4502 aFrom = pFrom; 4503 nFrom = nTo; 4504 } 4505 4506 if( nFrom==0 ){ 4507 sqlite3ErrorMsg(pParse, "no query solution"); 4508 sqlite3DbFreeNN(db, pSpace); 4509 return SQLITE_ERROR; 4510 } 4511 4512 /* Find the lowest cost path. pFrom will be left pointing to that path */ 4513 pFrom = aFrom; 4514 for(ii=1; ii<nFrom; ii++){ 4515 if( pFrom->rCost>aFrom[ii].rCost ) pFrom = &aFrom[ii]; 4516 } 4517 assert( pWInfo->nLevel==nLoop ); 4518 /* Load the lowest cost path into pWInfo */ 4519 for(iLoop=0; iLoop<nLoop; iLoop++){ 4520 WhereLevel *pLevel = pWInfo->a + iLoop; 4521 pLevel->pWLoop = pWLoop = pFrom->aLoop[iLoop]; 4522 pLevel->iFrom = pWLoop->iTab; 4523 pLevel->iTabCur = pWInfo->pTabList->a[pLevel->iFrom].iCursor; 4524 } 4525 if( (pWInfo->wctrlFlags & WHERE_WANT_DISTINCT)!=0 4526 && (pWInfo->wctrlFlags & WHERE_DISTINCTBY)==0 4527 && pWInfo->eDistinct==WHERE_DISTINCT_NOOP 4528 && nRowEst 4529 ){ 4530 Bitmask notUsed; 4531 int rc = wherePathSatisfiesOrderBy(pWInfo, pWInfo->pResultSet, pFrom, 4532 WHERE_DISTINCTBY, nLoop-1, pFrom->aLoop[nLoop-1], ¬Used); 4533 if( rc==pWInfo->pResultSet->nExpr ){ 4534 pWInfo->eDistinct = WHERE_DISTINCT_ORDERED; 4535 } 4536 } 4537 pWInfo->bOrderedInnerLoop = 0; 4538 if( pWInfo->pOrderBy ){ 4539 if( pWInfo->wctrlFlags & WHERE_DISTINCTBY ){ 4540 if( pFrom->isOrdered==pWInfo->pOrderBy->nExpr ){ 4541 pWInfo->eDistinct = WHERE_DISTINCT_ORDERED; 4542 } 4543 }else{ 4544 pWInfo->nOBSat = pFrom->isOrdered; 4545 pWInfo->revMask = pFrom->revLoop; 4546 if( pWInfo->nOBSat<=0 ){ 4547 pWInfo->nOBSat = 0; 4548 if( nLoop>0 ){ 4549 u32 wsFlags = pFrom->aLoop[nLoop-1]->wsFlags; 4550 if( (wsFlags & WHERE_ONEROW)==0 4551 && (wsFlags&(WHERE_IPK|WHERE_COLUMN_IN))!=(WHERE_IPK|WHERE_COLUMN_IN) 4552 ){ 4553 Bitmask m = 0; 4554 int rc = wherePathSatisfiesOrderBy(pWInfo, pWInfo->pOrderBy, pFrom, 4555 WHERE_ORDERBY_LIMIT, nLoop-1, pFrom->aLoop[nLoop-1], &m); 4556 testcase( wsFlags & WHERE_IPK ); 4557 testcase( wsFlags & WHERE_COLUMN_IN ); 4558 if( rc==pWInfo->pOrderBy->nExpr ){ 4559 pWInfo->bOrderedInnerLoop = 1; 4560 pWInfo->revMask = m; 4561 } 4562 } 4563 } 4564 }else if( nLoop 4565 && pWInfo->nOBSat==1 4566 && (pWInfo->wctrlFlags & (WHERE_ORDERBY_MIN|WHERE_ORDERBY_MAX))!=0 4567 ){ 4568 pWInfo->bOrderedInnerLoop = 1; 4569 } 4570 } 4571 if( (pWInfo->wctrlFlags & WHERE_SORTBYGROUP) 4572 && pWInfo->nOBSat==pWInfo->pOrderBy->nExpr && nLoop>0 4573 ){ 4574 Bitmask revMask = 0; 4575 int nOrder = wherePathSatisfiesOrderBy(pWInfo, pWInfo->pOrderBy, 4576 pFrom, 0, nLoop-1, pFrom->aLoop[nLoop-1], &revMask 4577 ); 4578 assert( pWInfo->sorted==0 ); 4579 if( nOrder==pWInfo->pOrderBy->nExpr ){ 4580 pWInfo->sorted = 1; 4581 pWInfo->revMask = revMask; 4582 } 4583 } 4584 } 4585 4586 4587 pWInfo->nRowOut = pFrom->nRow; 4588 4589 /* Free temporary memory and return success */ 4590 sqlite3DbFreeNN(db, pSpace); 4591 return SQLITE_OK; 4592 } 4593 4594 /* 4595 ** Most queries use only a single table (they are not joins) and have 4596 ** simple == constraints against indexed fields. This routine attempts 4597 ** to plan those simple cases using much less ceremony than the 4598 ** general-purpose query planner, and thereby yield faster sqlite3_prepare() 4599 ** times for the common case. 4600 ** 4601 ** Return non-zero on success, if this query can be handled by this 4602 ** no-frills query planner. Return zero if this query needs the 4603 ** general-purpose query planner. 4604 */ 4605 static int whereShortCut(WhereLoopBuilder *pBuilder){ 4606 WhereInfo *pWInfo; 4607 SrcItem *pItem; 4608 WhereClause *pWC; 4609 WhereTerm *pTerm; 4610 WhereLoop *pLoop; 4611 int iCur; 4612 int j; 4613 Table *pTab; 4614 Index *pIdx; 4615 WhereScan scan; 4616 4617 pWInfo = pBuilder->pWInfo; 4618 if( pWInfo->wctrlFlags & WHERE_OR_SUBCLAUSE ) return 0; 4619 assert( pWInfo->pTabList->nSrc>=1 ); 4620 pItem = pWInfo->pTabList->a; 4621 pTab = pItem->pTab; 4622 if( IsVirtual(pTab) ) return 0; 4623 if( pItem->fg.isIndexedBy ) return 0; 4624 iCur = pItem->iCursor; 4625 pWC = &pWInfo->sWC; 4626 pLoop = pBuilder->pNew; 4627 pLoop->wsFlags = 0; 4628 pLoop->nSkip = 0; 4629 pTerm = whereScanInit(&scan, pWC, iCur, -1, WO_EQ|WO_IS, 0); 4630 while( pTerm && pTerm->prereqRight ) pTerm = whereScanNext(&scan); 4631 if( pTerm ){ 4632 testcase( pTerm->eOperator & WO_IS ); 4633 pLoop->wsFlags = WHERE_COLUMN_EQ|WHERE_IPK|WHERE_ONEROW; 4634 pLoop->aLTerm[0] = pTerm; 4635 pLoop->nLTerm = 1; 4636 pLoop->u.btree.nEq = 1; 4637 /* TUNING: Cost of a rowid lookup is 10 */ 4638 pLoop->rRun = 33; /* 33==sqlite3LogEst(10) */ 4639 }else{ 4640 for(pIdx=pTab->pIndex; pIdx; pIdx=pIdx->pNext){ 4641 int opMask; 4642 assert( pLoop->aLTermSpace==pLoop->aLTerm ); 4643 if( !IsUniqueIndex(pIdx) 4644 || pIdx->pPartIdxWhere!=0 4645 || pIdx->nKeyCol>ArraySize(pLoop->aLTermSpace) 4646 ) continue; 4647 opMask = pIdx->uniqNotNull ? (WO_EQ|WO_IS) : WO_EQ; 4648 for(j=0; j<pIdx->nKeyCol; j++){ 4649 pTerm = whereScanInit(&scan, pWC, iCur, j, opMask, pIdx); 4650 while( pTerm && pTerm->prereqRight ) pTerm = whereScanNext(&scan); 4651 if( pTerm==0 ) break; 4652 testcase( pTerm->eOperator & WO_IS ); 4653 pLoop->aLTerm[j] = pTerm; 4654 } 4655 if( j!=pIdx->nKeyCol ) continue; 4656 pLoop->wsFlags = WHERE_COLUMN_EQ|WHERE_ONEROW|WHERE_INDEXED; 4657 if( pIdx->isCovering || (pItem->colUsed & pIdx->colNotIdxed)==0 ){ 4658 pLoop->wsFlags |= WHERE_IDX_ONLY; 4659 } 4660 pLoop->nLTerm = j; 4661 pLoop->u.btree.nEq = j; 4662 pLoop->u.btree.pIndex = pIdx; 4663 /* TUNING: Cost of a unique index lookup is 15 */ 4664 pLoop->rRun = 39; /* 39==sqlite3LogEst(15) */ 4665 break; 4666 } 4667 } 4668 if( pLoop->wsFlags ){ 4669 pLoop->nOut = (LogEst)1; 4670 pWInfo->a[0].pWLoop = pLoop; 4671 assert( pWInfo->sMaskSet.n==1 && iCur==pWInfo->sMaskSet.ix[0] ); 4672 pLoop->maskSelf = 1; /* sqlite3WhereGetMask(&pWInfo->sMaskSet, iCur); */ 4673 pWInfo->a[0].iTabCur = iCur; 4674 pWInfo->nRowOut = 1; 4675 if( pWInfo->pOrderBy ) pWInfo->nOBSat = pWInfo->pOrderBy->nExpr; 4676 if( pWInfo->wctrlFlags & WHERE_WANT_DISTINCT ){ 4677 pWInfo->eDistinct = WHERE_DISTINCT_UNIQUE; 4678 } 4679 if( scan.iEquiv>1 ) pLoop->wsFlags |= WHERE_TRANSCONS; 4680 #ifdef SQLITE_DEBUG 4681 pLoop->cId = '0'; 4682 #endif 4683 #ifdef WHERETRACE_ENABLED 4684 if( sqlite3WhereTrace ){ 4685 sqlite3DebugPrintf("whereShortCut() used to compute solution\n"); 4686 } 4687 #endif 4688 return 1; 4689 } 4690 return 0; 4691 } 4692 4693 /* 4694 ** Helper function for exprIsDeterministic(). 4695 */ 4696 static int exprNodeIsDeterministic(Walker *pWalker, Expr *pExpr){ 4697 if( pExpr->op==TK_FUNCTION && ExprHasProperty(pExpr, EP_ConstFunc)==0 ){ 4698 pWalker->eCode = 0; 4699 return WRC_Abort; 4700 } 4701 return WRC_Continue; 4702 } 4703 4704 /* 4705 ** Return true if the expression contains no non-deterministic SQL 4706 ** functions. Do not consider non-deterministic SQL functions that are 4707 ** part of sub-select statements. 4708 */ 4709 static int exprIsDeterministic(Expr *p){ 4710 Walker w; 4711 memset(&w, 0, sizeof(w)); 4712 w.eCode = 1; 4713 w.xExprCallback = exprNodeIsDeterministic; 4714 w.xSelectCallback = sqlite3SelectWalkFail; 4715 sqlite3WalkExpr(&w, p); 4716 return w.eCode; 4717 } 4718 4719 4720 #ifdef WHERETRACE_ENABLED 4721 /* 4722 ** Display all WhereLoops in pWInfo 4723 */ 4724 static void showAllWhereLoops(WhereInfo *pWInfo, WhereClause *pWC){ 4725 if( sqlite3WhereTrace ){ /* Display all of the WhereLoop objects */ 4726 WhereLoop *p; 4727 int i; 4728 static const char zLabel[] = "0123456789abcdefghijklmnopqrstuvwyxz" 4729 "ABCDEFGHIJKLMNOPQRSTUVWYXZ"; 4730 for(p=pWInfo->pLoops, i=0; p; p=p->pNextLoop, i++){ 4731 p->cId = zLabel[i%(sizeof(zLabel)-1)]; 4732 sqlite3WhereLoopPrint(p, pWC); 4733 } 4734 } 4735 } 4736 # define WHERETRACE_ALL_LOOPS(W,C) showAllWhereLoops(W,C) 4737 #else 4738 # define WHERETRACE_ALL_LOOPS(W,C) 4739 #endif 4740 4741 /* 4742 ** Generate the beginning of the loop used for WHERE clause processing. 4743 ** The return value is a pointer to an opaque structure that contains 4744 ** information needed to terminate the loop. Later, the calling routine 4745 ** should invoke sqlite3WhereEnd() with the return value of this function 4746 ** in order to complete the WHERE clause processing. 4747 ** 4748 ** If an error occurs, this routine returns NULL. 4749 ** 4750 ** The basic idea is to do a nested loop, one loop for each table in 4751 ** the FROM clause of a select. (INSERT and UPDATE statements are the 4752 ** same as a SELECT with only a single table in the FROM clause.) For 4753 ** example, if the SQL is this: 4754 ** 4755 ** SELECT * FROM t1, t2, t3 WHERE ...; 4756 ** 4757 ** Then the code generated is conceptually like the following: 4758 ** 4759 ** foreach row1 in t1 do \ Code generated 4760 ** foreach row2 in t2 do |-- by sqlite3WhereBegin() 4761 ** foreach row3 in t3 do / 4762 ** ... 4763 ** end \ Code generated 4764 ** end |-- by sqlite3WhereEnd() 4765 ** end / 4766 ** 4767 ** Note that the loops might not be nested in the order in which they 4768 ** appear in the FROM clause if a different order is better able to make 4769 ** use of indices. Note also that when the IN operator appears in 4770 ** the WHERE clause, it might result in additional nested loops for 4771 ** scanning through all values on the right-hand side of the IN. 4772 ** 4773 ** There are Btree cursors associated with each table. t1 uses cursor 4774 ** number pTabList->a[0].iCursor. t2 uses the cursor pTabList->a[1].iCursor. 4775 ** And so forth. This routine generates code to open those VDBE cursors 4776 ** and sqlite3WhereEnd() generates the code to close them. 4777 ** 4778 ** The code that sqlite3WhereBegin() generates leaves the cursors named 4779 ** in pTabList pointing at their appropriate entries. The [...] code 4780 ** can use OP_Column and OP_Rowid opcodes on these cursors to extract 4781 ** data from the various tables of the loop. 4782 ** 4783 ** If the WHERE clause is empty, the foreach loops must each scan their 4784 ** entire tables. Thus a three-way join is an O(N^3) operation. But if 4785 ** the tables have indices and there are terms in the WHERE clause that 4786 ** refer to those indices, a complete table scan can be avoided and the 4787 ** code will run much faster. Most of the work of this routine is checking 4788 ** to see if there are indices that can be used to speed up the loop. 4789 ** 4790 ** Terms of the WHERE clause are also used to limit which rows actually 4791 ** make it to the "..." in the middle of the loop. After each "foreach", 4792 ** terms of the WHERE clause that use only terms in that loop and outer 4793 ** loops are evaluated and if false a jump is made around all subsequent 4794 ** inner loops (or around the "..." if the test occurs within the inner- 4795 ** most loop) 4796 ** 4797 ** OUTER JOINS 4798 ** 4799 ** An outer join of tables t1 and t2 is conceptally coded as follows: 4800 ** 4801 ** foreach row1 in t1 do 4802 ** flag = 0 4803 ** foreach row2 in t2 do 4804 ** start: 4805 ** ... 4806 ** flag = 1 4807 ** end 4808 ** if flag==0 then 4809 ** move the row2 cursor to a null row 4810 ** goto start 4811 ** fi 4812 ** end 4813 ** 4814 ** ORDER BY CLAUSE PROCESSING 4815 ** 4816 ** pOrderBy is a pointer to the ORDER BY clause (or the GROUP BY clause 4817 ** if the WHERE_GROUPBY flag is set in wctrlFlags) of a SELECT statement 4818 ** if there is one. If there is no ORDER BY clause or if this routine 4819 ** is called from an UPDATE or DELETE statement, then pOrderBy is NULL. 4820 ** 4821 ** The iIdxCur parameter is the cursor number of an index. If 4822 ** WHERE_OR_SUBCLAUSE is set, iIdxCur is the cursor number of an index 4823 ** to use for OR clause processing. The WHERE clause should use this 4824 ** specific cursor. If WHERE_ONEPASS_DESIRED is set, then iIdxCur is 4825 ** the first cursor in an array of cursors for all indices. iIdxCur should 4826 ** be used to compute the appropriate cursor depending on which index is 4827 ** used. 4828 */ 4829 WhereInfo *sqlite3WhereBegin( 4830 Parse *pParse, /* The parser context */ 4831 SrcList *pTabList, /* FROM clause: A list of all tables to be scanned */ 4832 Expr *pWhere, /* The WHERE clause */ 4833 ExprList *pOrderBy, /* An ORDER BY (or GROUP BY) clause, or NULL */ 4834 ExprList *pResultSet, /* Query result set. Req'd for DISTINCT */ 4835 u16 wctrlFlags, /* The WHERE_* flags defined in sqliteInt.h */ 4836 int iAuxArg /* If WHERE_OR_SUBCLAUSE is set, index cursor number 4837 ** If WHERE_USE_LIMIT, then the limit amount */ 4838 ){ 4839 int nByteWInfo; /* Num. bytes allocated for WhereInfo struct */ 4840 int nTabList; /* Number of elements in pTabList */ 4841 WhereInfo *pWInfo; /* Will become the return value of this function */ 4842 Vdbe *v = pParse->pVdbe; /* The virtual database engine */ 4843 Bitmask notReady; /* Cursors that are not yet positioned */ 4844 WhereLoopBuilder sWLB; /* The WhereLoop builder */ 4845 WhereMaskSet *pMaskSet; /* The expression mask set */ 4846 WhereLevel *pLevel; /* A single level in pWInfo->a[] */ 4847 WhereLoop *pLoop; /* Pointer to a single WhereLoop object */ 4848 int ii; /* Loop counter */ 4849 sqlite3 *db; /* Database connection */ 4850 int rc; /* Return code */ 4851 u8 bFordelete = 0; /* OPFLAG_FORDELETE or zero, as appropriate */ 4852 4853 assert( (wctrlFlags & WHERE_ONEPASS_MULTIROW)==0 || ( 4854 (wctrlFlags & WHERE_ONEPASS_DESIRED)!=0 4855 && (wctrlFlags & WHERE_OR_SUBCLAUSE)==0 4856 )); 4857 4858 /* Only one of WHERE_OR_SUBCLAUSE or WHERE_USE_LIMIT */ 4859 assert( (wctrlFlags & WHERE_OR_SUBCLAUSE)==0 4860 || (wctrlFlags & WHERE_USE_LIMIT)==0 ); 4861 4862 /* Variable initialization */ 4863 db = pParse->db; 4864 memset(&sWLB, 0, sizeof(sWLB)); 4865 4866 /* An ORDER/GROUP BY clause of more than 63 terms cannot be optimized */ 4867 testcase( pOrderBy && pOrderBy->nExpr==BMS-1 ); 4868 if( pOrderBy && pOrderBy->nExpr>=BMS ) pOrderBy = 0; 4869 sWLB.pOrderBy = pOrderBy; 4870 4871 /* Disable the DISTINCT optimization if SQLITE_DistinctOpt is set via 4872 ** sqlite3_test_ctrl(SQLITE_TESTCTRL_OPTIMIZATIONS,...) */ 4873 if( OptimizationDisabled(db, SQLITE_DistinctOpt) ){ 4874 wctrlFlags &= ~WHERE_WANT_DISTINCT; 4875 } 4876 4877 /* The number of tables in the FROM clause is limited by the number of 4878 ** bits in a Bitmask 4879 */ 4880 testcase( pTabList->nSrc==BMS ); 4881 if( pTabList->nSrc>BMS ){ 4882 sqlite3ErrorMsg(pParse, "at most %d tables in a join", BMS); 4883 return 0; 4884 } 4885 4886 /* This function normally generates a nested loop for all tables in 4887 ** pTabList. But if the WHERE_OR_SUBCLAUSE flag is set, then we should 4888 ** only generate code for the first table in pTabList and assume that 4889 ** any cursors associated with subsequent tables are uninitialized. 4890 */ 4891 nTabList = (wctrlFlags & WHERE_OR_SUBCLAUSE) ? 1 : pTabList->nSrc; 4892 4893 /* Allocate and initialize the WhereInfo structure that will become the 4894 ** return value. A single allocation is used to store the WhereInfo 4895 ** struct, the contents of WhereInfo.a[], the WhereClause structure 4896 ** and the WhereMaskSet structure. Since WhereClause contains an 8-byte 4897 ** field (type Bitmask) it must be aligned on an 8-byte boundary on 4898 ** some architectures. Hence the ROUND8() below. 4899 */ 4900 nByteWInfo = ROUND8(sizeof(WhereInfo)+(nTabList-1)*sizeof(WhereLevel)); 4901 pWInfo = sqlite3DbMallocRawNN(db, nByteWInfo + sizeof(WhereLoop)); 4902 if( db->mallocFailed ){ 4903 sqlite3DbFree(db, pWInfo); 4904 pWInfo = 0; 4905 goto whereBeginError; 4906 } 4907 pWInfo->pParse = pParse; 4908 pWInfo->pTabList = pTabList; 4909 pWInfo->pOrderBy = pOrderBy; 4910 pWInfo->pWhere = pWhere; 4911 pWInfo->pResultSet = pResultSet; 4912 pWInfo->aiCurOnePass[0] = pWInfo->aiCurOnePass[1] = -1; 4913 pWInfo->nLevel = nTabList; 4914 pWInfo->iBreak = pWInfo->iContinue = sqlite3VdbeMakeLabel(pParse); 4915 pWInfo->wctrlFlags = wctrlFlags; 4916 pWInfo->iLimit = iAuxArg; 4917 pWInfo->savedNQueryLoop = pParse->nQueryLoop; 4918 memset(&pWInfo->nOBSat, 0, 4919 offsetof(WhereInfo,sWC) - offsetof(WhereInfo,nOBSat)); 4920 memset(&pWInfo->a[0], 0, sizeof(WhereLoop)+nTabList*sizeof(WhereLevel)); 4921 assert( pWInfo->eOnePass==ONEPASS_OFF ); /* ONEPASS defaults to OFF */ 4922 pMaskSet = &pWInfo->sMaskSet; 4923 sWLB.pWInfo = pWInfo; 4924 sWLB.pWC = &pWInfo->sWC; 4925 sWLB.pNew = (WhereLoop*)(((char*)pWInfo)+nByteWInfo); 4926 assert( EIGHT_BYTE_ALIGNMENT(sWLB.pNew) ); 4927 whereLoopInit(sWLB.pNew); 4928 #ifdef SQLITE_DEBUG 4929 sWLB.pNew->cId = '*'; 4930 #endif 4931 4932 /* Split the WHERE clause into separate subexpressions where each 4933 ** subexpression is separated by an AND operator. 4934 */ 4935 initMaskSet(pMaskSet); 4936 sqlite3WhereClauseInit(&pWInfo->sWC, pWInfo); 4937 sqlite3WhereSplit(&pWInfo->sWC, pWhere, TK_AND); 4938 4939 /* Special case: No FROM clause 4940 */ 4941 if( nTabList==0 ){ 4942 if( pOrderBy ) pWInfo->nOBSat = pOrderBy->nExpr; 4943 if( wctrlFlags & WHERE_WANT_DISTINCT ){ 4944 pWInfo->eDistinct = WHERE_DISTINCT_UNIQUE; 4945 } 4946 ExplainQueryPlan((pParse, 0, "SCAN CONSTANT ROW")); 4947 }else{ 4948 /* Assign a bit from the bitmask to every term in the FROM clause. 4949 ** 4950 ** The N-th term of the FROM clause is assigned a bitmask of 1<<N. 4951 ** 4952 ** The rule of the previous sentence ensures thta if X is the bitmask for 4953 ** a table T, then X-1 is the bitmask for all other tables to the left of T. 4954 ** Knowing the bitmask for all tables to the left of a left join is 4955 ** important. Ticket #3015. 4956 ** 4957 ** Note that bitmasks are created for all pTabList->nSrc tables in 4958 ** pTabList, not just the first nTabList tables. nTabList is normally 4959 ** equal to pTabList->nSrc but might be shortened to 1 if the 4960 ** WHERE_OR_SUBCLAUSE flag is set. 4961 */ 4962 ii = 0; 4963 do{ 4964 createMask(pMaskSet, pTabList->a[ii].iCursor); 4965 sqlite3WhereTabFuncArgs(pParse, &pTabList->a[ii], &pWInfo->sWC); 4966 }while( (++ii)<pTabList->nSrc ); 4967 #ifdef SQLITE_DEBUG 4968 { 4969 Bitmask mx = 0; 4970 for(ii=0; ii<pTabList->nSrc; ii++){ 4971 Bitmask m = sqlite3WhereGetMask(pMaskSet, pTabList->a[ii].iCursor); 4972 assert( m>=mx ); 4973 mx = m; 4974 } 4975 } 4976 #endif 4977 } 4978 4979 /* Analyze all of the subexpressions. */ 4980 sqlite3WhereExprAnalyze(pTabList, &pWInfo->sWC); 4981 if( db->mallocFailed ) goto whereBeginError; 4982 4983 /* Special case: WHERE terms that do not refer to any tables in the join 4984 ** (constant expressions). Evaluate each such term, and jump over all the 4985 ** generated code if the result is not true. 4986 ** 4987 ** Do not do this if the expression contains non-deterministic functions 4988 ** that are not within a sub-select. This is not strictly required, but 4989 ** preserves SQLite's legacy behaviour in the following two cases: 4990 ** 4991 ** FROM ... WHERE random()>0; -- eval random() once per row 4992 ** FROM ... WHERE (SELECT random())>0; -- eval random() once overall 4993 */ 4994 for(ii=0; ii<sWLB.pWC->nTerm; ii++){ 4995 WhereTerm *pT = &sWLB.pWC->a[ii]; 4996 if( pT->wtFlags & TERM_VIRTUAL ) continue; 4997 if( pT->prereqAll==0 && (nTabList==0 || exprIsDeterministic(pT->pExpr)) ){ 4998 sqlite3ExprIfFalse(pParse, pT->pExpr, pWInfo->iBreak, SQLITE_JUMPIFNULL); 4999 pT->wtFlags |= TERM_CODED; 5000 } 5001 } 5002 5003 if( wctrlFlags & WHERE_WANT_DISTINCT ){ 5004 if( isDistinctRedundant(pParse, pTabList, &pWInfo->sWC, pResultSet) ){ 5005 /* The DISTINCT marking is pointless. Ignore it. */ 5006 pWInfo->eDistinct = WHERE_DISTINCT_UNIQUE; 5007 }else if( pOrderBy==0 ){ 5008 /* Try to ORDER BY the result set to make distinct processing easier */ 5009 pWInfo->wctrlFlags |= WHERE_DISTINCTBY; 5010 pWInfo->pOrderBy = pResultSet; 5011 } 5012 } 5013 5014 /* Construct the WhereLoop objects */ 5015 #if defined(WHERETRACE_ENABLED) 5016 if( sqlite3WhereTrace & 0xffff ){ 5017 sqlite3DebugPrintf("*** Optimizer Start *** (wctrlFlags: 0x%x",wctrlFlags); 5018 if( wctrlFlags & WHERE_USE_LIMIT ){ 5019 sqlite3DebugPrintf(", limit: %d", iAuxArg); 5020 } 5021 sqlite3DebugPrintf(")\n"); 5022 if( sqlite3WhereTrace & 0x100 ){ 5023 Select sSelect; 5024 memset(&sSelect, 0, sizeof(sSelect)); 5025 sSelect.selFlags = SF_WhereBegin; 5026 sSelect.pSrc = pTabList; 5027 sSelect.pWhere = pWhere; 5028 sSelect.pOrderBy = pOrderBy; 5029 sSelect.pEList = pResultSet; 5030 sqlite3TreeViewSelect(0, &sSelect, 0); 5031 } 5032 } 5033 if( sqlite3WhereTrace & 0x100 ){ /* Display all terms of the WHERE clause */ 5034 sqlite3DebugPrintf("---- WHERE clause at start of analysis:\n"); 5035 sqlite3WhereClausePrint(sWLB.pWC); 5036 } 5037 #endif 5038 5039 if( nTabList!=1 || whereShortCut(&sWLB)==0 ){ 5040 rc = whereLoopAddAll(&sWLB); 5041 if( rc ) goto whereBeginError; 5042 5043 #ifdef SQLITE_ENABLE_STAT4 5044 /* If one or more WhereTerm.truthProb values were used in estimating 5045 ** loop parameters, but then those truthProb values were subsequently 5046 ** changed based on STAT4 information while computing subsequent loops, 5047 ** then we need to rerun the whole loop building process so that all 5048 ** loops will be built using the revised truthProb values. */ 5049 if( sWLB.bldFlags2 & SQLITE_BLDF2_2NDPASS ){ 5050 WHERETRACE_ALL_LOOPS(pWInfo, sWLB.pWC); 5051 WHERETRACE(0xffff, 5052 ("**** Redo all loop computations due to" 5053 " TERM_HIGHTRUTH changes ****\n")); 5054 while( pWInfo->pLoops ){ 5055 WhereLoop *p = pWInfo->pLoops; 5056 pWInfo->pLoops = p->pNextLoop; 5057 whereLoopDelete(db, p); 5058 } 5059 rc = whereLoopAddAll(&sWLB); 5060 if( rc ) goto whereBeginError; 5061 } 5062 #endif 5063 WHERETRACE_ALL_LOOPS(pWInfo, sWLB.pWC); 5064 5065 wherePathSolver(pWInfo, 0); 5066 if( db->mallocFailed ) goto whereBeginError; 5067 if( pWInfo->pOrderBy ){ 5068 wherePathSolver(pWInfo, pWInfo->nRowOut+1); 5069 if( db->mallocFailed ) goto whereBeginError; 5070 } 5071 } 5072 if( pWInfo->pOrderBy==0 && (db->flags & SQLITE_ReverseOrder)!=0 ){ 5073 pWInfo->revMask = ALLBITS; 5074 } 5075 if( pParse->nErr || db->mallocFailed ){ 5076 goto whereBeginError; 5077 } 5078 #ifdef WHERETRACE_ENABLED 5079 if( sqlite3WhereTrace ){ 5080 sqlite3DebugPrintf("---- Solution nRow=%d", pWInfo->nRowOut); 5081 if( pWInfo->nOBSat>0 ){ 5082 sqlite3DebugPrintf(" ORDERBY=%d,0x%llx", pWInfo->nOBSat, pWInfo->revMask); 5083 } 5084 switch( pWInfo->eDistinct ){ 5085 case WHERE_DISTINCT_UNIQUE: { 5086 sqlite3DebugPrintf(" DISTINCT=unique"); 5087 break; 5088 } 5089 case WHERE_DISTINCT_ORDERED: { 5090 sqlite3DebugPrintf(" DISTINCT=ordered"); 5091 break; 5092 } 5093 case WHERE_DISTINCT_UNORDERED: { 5094 sqlite3DebugPrintf(" DISTINCT=unordered"); 5095 break; 5096 } 5097 } 5098 sqlite3DebugPrintf("\n"); 5099 for(ii=0; ii<pWInfo->nLevel; ii++){ 5100 sqlite3WhereLoopPrint(pWInfo->a[ii].pWLoop, sWLB.pWC); 5101 } 5102 } 5103 #endif 5104 5105 /* Attempt to omit tables from the join that do not affect the result. 5106 ** For a table to not affect the result, the following must be true: 5107 ** 5108 ** 1) The query must not be an aggregate. 5109 ** 2) The table must be the RHS of a LEFT JOIN. 5110 ** 3) Either the query must be DISTINCT, or else the ON or USING clause 5111 ** must contain a constraint that limits the scan of the table to 5112 ** at most a single row. 5113 ** 4) The table must not be referenced by any part of the query apart 5114 ** from its own USING or ON clause. 5115 ** 5116 ** For example, given: 5117 ** 5118 ** CREATE TABLE t1(ipk INTEGER PRIMARY KEY, v1); 5119 ** CREATE TABLE t2(ipk INTEGER PRIMARY KEY, v2); 5120 ** CREATE TABLE t3(ipk INTEGER PRIMARY KEY, v3); 5121 ** 5122 ** then table t2 can be omitted from the following: 5123 ** 5124 ** SELECT v1, v3 FROM t1 5125 ** LEFT JOIN t2 ON (t1.ipk=t2.ipk) 5126 ** LEFT JOIN t3 ON (t1.ipk=t3.ipk) 5127 ** 5128 ** or from: 5129 ** 5130 ** SELECT DISTINCT v1, v3 FROM t1 5131 ** LEFT JOIN t2 5132 ** LEFT JOIN t3 ON (t1.ipk=t3.ipk) 5133 */ 5134 notReady = ~(Bitmask)0; 5135 if( pWInfo->nLevel>=2 5136 && pResultSet!=0 /* these two combine to guarantee */ 5137 && 0==(wctrlFlags & WHERE_AGG_DISTINCT) /* condition (1) above */ 5138 && OptimizationEnabled(db, SQLITE_OmitNoopJoin) 5139 ){ 5140 int i; 5141 Bitmask tabUsed = sqlite3WhereExprListUsage(pMaskSet, pResultSet); 5142 if( sWLB.pOrderBy ){ 5143 tabUsed |= sqlite3WhereExprListUsage(pMaskSet, sWLB.pOrderBy); 5144 } 5145 for(i=pWInfo->nLevel-1; i>=1; i--){ 5146 WhereTerm *pTerm, *pEnd; 5147 SrcItem *pItem; 5148 pLoop = pWInfo->a[i].pWLoop; 5149 pItem = &pWInfo->pTabList->a[pLoop->iTab]; 5150 if( (pItem->fg.jointype & JT_LEFT)==0 ) continue; 5151 if( (wctrlFlags & WHERE_WANT_DISTINCT)==0 5152 && (pLoop->wsFlags & WHERE_ONEROW)==0 5153 ){ 5154 continue; 5155 } 5156 if( (tabUsed & pLoop->maskSelf)!=0 ) continue; 5157 pEnd = sWLB.pWC->a + sWLB.pWC->nTerm; 5158 for(pTerm=sWLB.pWC->a; pTerm<pEnd; pTerm++){ 5159 if( (pTerm->prereqAll & pLoop->maskSelf)!=0 ){ 5160 if( !ExprHasProperty(pTerm->pExpr, EP_FromJoin) 5161 || pTerm->pExpr->iRightJoinTable!=pItem->iCursor 5162 ){ 5163 break; 5164 } 5165 } 5166 } 5167 if( pTerm<pEnd ) continue; 5168 WHERETRACE(0xffff, ("-> drop loop %c not used\n", pLoop->cId)); 5169 notReady &= ~pLoop->maskSelf; 5170 for(pTerm=sWLB.pWC->a; pTerm<pEnd; pTerm++){ 5171 if( (pTerm->prereqAll & pLoop->maskSelf)!=0 ){ 5172 pTerm->wtFlags |= TERM_CODED; 5173 } 5174 } 5175 if( i!=pWInfo->nLevel-1 ){ 5176 int nByte = (pWInfo->nLevel-1-i) * sizeof(WhereLevel); 5177 memmove(&pWInfo->a[i], &pWInfo->a[i+1], nByte); 5178 } 5179 pWInfo->nLevel--; 5180 nTabList--; 5181 } 5182 } 5183 #if defined(WHERETRACE_ENABLED) 5184 if( sqlite3WhereTrace & 0x100 ){ /* Display all terms of the WHERE clause */ 5185 sqlite3DebugPrintf("---- WHERE clause at end of analysis:\n"); 5186 sqlite3WhereClausePrint(sWLB.pWC); 5187 } 5188 WHERETRACE(0xffff,("*** Optimizer Finished ***\n")); 5189 #endif 5190 pWInfo->pParse->nQueryLoop += pWInfo->nRowOut; 5191 5192 /* If the caller is an UPDATE or DELETE statement that is requesting 5193 ** to use a one-pass algorithm, determine if this is appropriate. 5194 ** 5195 ** A one-pass approach can be used if the caller has requested one 5196 ** and either (a) the scan visits at most one row or (b) each 5197 ** of the following are true: 5198 ** 5199 ** * the caller has indicated that a one-pass approach can be used 5200 ** with multiple rows (by setting WHERE_ONEPASS_MULTIROW), and 5201 ** * the table is not a virtual table, and 5202 ** * either the scan does not use the OR optimization or the caller 5203 ** is a DELETE operation (WHERE_DUPLICATES_OK is only specified 5204 ** for DELETE). 5205 ** 5206 ** The last qualification is because an UPDATE statement uses 5207 ** WhereInfo.aiCurOnePass[1] to determine whether or not it really can 5208 ** use a one-pass approach, and this is not set accurately for scans 5209 ** that use the OR optimization. 5210 */ 5211 assert( (wctrlFlags & WHERE_ONEPASS_DESIRED)==0 || pWInfo->nLevel==1 ); 5212 if( (wctrlFlags & WHERE_ONEPASS_DESIRED)!=0 ){ 5213 int wsFlags = pWInfo->a[0].pWLoop->wsFlags; 5214 int bOnerow = (wsFlags & WHERE_ONEROW)!=0; 5215 assert( !(wsFlags & WHERE_VIRTUALTABLE) || IsVirtual(pTabList->a[0].pTab) ); 5216 if( bOnerow || ( 5217 0!=(wctrlFlags & WHERE_ONEPASS_MULTIROW) 5218 && !IsVirtual(pTabList->a[0].pTab) 5219 && (0==(wsFlags & WHERE_MULTI_OR) || (wctrlFlags & WHERE_DUPLICATES_OK)) 5220 )){ 5221 pWInfo->eOnePass = bOnerow ? ONEPASS_SINGLE : ONEPASS_MULTI; 5222 if( HasRowid(pTabList->a[0].pTab) && (wsFlags & WHERE_IDX_ONLY) ){ 5223 if( wctrlFlags & WHERE_ONEPASS_MULTIROW ){ 5224 bFordelete = OPFLAG_FORDELETE; 5225 } 5226 pWInfo->a[0].pWLoop->wsFlags = (wsFlags & ~WHERE_IDX_ONLY); 5227 } 5228 } 5229 } 5230 5231 /* Open all tables in the pTabList and any indices selected for 5232 ** searching those tables. 5233 */ 5234 for(ii=0, pLevel=pWInfo->a; ii<nTabList; ii++, pLevel++){ 5235 Table *pTab; /* Table to open */ 5236 int iDb; /* Index of database containing table/index */ 5237 SrcItem *pTabItem; 5238 5239 pTabItem = &pTabList->a[pLevel->iFrom]; 5240 pTab = pTabItem->pTab; 5241 iDb = sqlite3SchemaToIndex(db, pTab->pSchema); 5242 pLoop = pLevel->pWLoop; 5243 if( (pTab->tabFlags & TF_Ephemeral)!=0 || IsView(pTab) ){ 5244 /* Do nothing */ 5245 }else 5246 #ifndef SQLITE_OMIT_VIRTUALTABLE 5247 if( (pLoop->wsFlags & WHERE_VIRTUALTABLE)!=0 ){ 5248 const char *pVTab = (const char *)sqlite3GetVTable(db, pTab); 5249 int iCur = pTabItem->iCursor; 5250 sqlite3VdbeAddOp4(v, OP_VOpen, iCur, 0, 0, pVTab, P4_VTAB); 5251 }else if( IsVirtual(pTab) ){ 5252 /* noop */ 5253 }else 5254 #endif 5255 if( (pLoop->wsFlags & WHERE_IDX_ONLY)==0 5256 && (wctrlFlags & WHERE_OR_SUBCLAUSE)==0 ){ 5257 int op = OP_OpenRead; 5258 if( pWInfo->eOnePass!=ONEPASS_OFF ){ 5259 op = OP_OpenWrite; 5260 pWInfo->aiCurOnePass[0] = pTabItem->iCursor; 5261 }; 5262 sqlite3OpenTable(pParse, pTabItem->iCursor, iDb, pTab, op); 5263 assert( pTabItem->iCursor==pLevel->iTabCur ); 5264 testcase( pWInfo->eOnePass==ONEPASS_OFF && pTab->nCol==BMS-1 ); 5265 testcase( pWInfo->eOnePass==ONEPASS_OFF && pTab->nCol==BMS ); 5266 if( pWInfo->eOnePass==ONEPASS_OFF 5267 && pTab->nCol<BMS 5268 && (pTab->tabFlags & (TF_HasGenerated|TF_WithoutRowid))==0 5269 ){ 5270 /* If we know that only a prefix of the record will be used, 5271 ** it is advantageous to reduce the "column count" field in 5272 ** the P4 operand of the OP_OpenRead/Write opcode. */ 5273 Bitmask b = pTabItem->colUsed; 5274 int n = 0; 5275 for(; b; b=b>>1, n++){} 5276 sqlite3VdbeChangeP4(v, -1, SQLITE_INT_TO_PTR(n), P4_INT32); 5277 assert( n<=pTab->nCol ); 5278 } 5279 #ifdef SQLITE_ENABLE_CURSOR_HINTS 5280 if( pLoop->u.btree.pIndex!=0 ){ 5281 sqlite3VdbeChangeP5(v, OPFLAG_SEEKEQ|bFordelete); 5282 }else 5283 #endif 5284 { 5285 sqlite3VdbeChangeP5(v, bFordelete); 5286 } 5287 #ifdef SQLITE_ENABLE_COLUMN_USED_MASK 5288 sqlite3VdbeAddOp4Dup8(v, OP_ColumnsUsed, pTabItem->iCursor, 0, 0, 5289 (const u8*)&pTabItem->colUsed, P4_INT64); 5290 #endif 5291 }else{ 5292 sqlite3TableLock(pParse, iDb, pTab->tnum, 0, pTab->zName); 5293 } 5294 if( pLoop->wsFlags & WHERE_INDEXED ){ 5295 Index *pIx = pLoop->u.btree.pIndex; 5296 int iIndexCur; 5297 int op = OP_OpenRead; 5298 /* iAuxArg is always set to a positive value if ONEPASS is possible */ 5299 assert( iAuxArg!=0 || (pWInfo->wctrlFlags & WHERE_ONEPASS_DESIRED)==0 ); 5300 if( !HasRowid(pTab) && IsPrimaryKeyIndex(pIx) 5301 && (wctrlFlags & WHERE_OR_SUBCLAUSE)!=0 5302 ){ 5303 /* This is one term of an OR-optimization using the PRIMARY KEY of a 5304 ** WITHOUT ROWID table. No need for a separate index */ 5305 iIndexCur = pLevel->iTabCur; 5306 op = 0; 5307 }else if( pWInfo->eOnePass!=ONEPASS_OFF ){ 5308 Index *pJ = pTabItem->pTab->pIndex; 5309 iIndexCur = iAuxArg; 5310 assert( wctrlFlags & WHERE_ONEPASS_DESIRED ); 5311 while( ALWAYS(pJ) && pJ!=pIx ){ 5312 iIndexCur++; 5313 pJ = pJ->pNext; 5314 } 5315 op = OP_OpenWrite; 5316 pWInfo->aiCurOnePass[1] = iIndexCur; 5317 }else if( iAuxArg && (wctrlFlags & WHERE_OR_SUBCLAUSE)!=0 ){ 5318 iIndexCur = iAuxArg; 5319 op = OP_ReopenIdx; 5320 }else{ 5321 iIndexCur = pParse->nTab++; 5322 } 5323 pLevel->iIdxCur = iIndexCur; 5324 assert( pIx->pSchema==pTab->pSchema ); 5325 assert( iIndexCur>=0 ); 5326 if( op ){ 5327 sqlite3VdbeAddOp3(v, op, iIndexCur, pIx->tnum, iDb); 5328 sqlite3VdbeSetP4KeyInfo(pParse, pIx); 5329 if( (pLoop->wsFlags & WHERE_CONSTRAINT)!=0 5330 && (pLoop->wsFlags & (WHERE_COLUMN_RANGE|WHERE_SKIPSCAN))==0 5331 && (pLoop->wsFlags & WHERE_BIGNULL_SORT)==0 5332 && (pLoop->wsFlags & WHERE_IN_SEEKSCAN)==0 5333 && (pWInfo->wctrlFlags&WHERE_ORDERBY_MIN)==0 5334 && pWInfo->eDistinct!=WHERE_DISTINCT_ORDERED 5335 ){ 5336 sqlite3VdbeChangeP5(v, OPFLAG_SEEKEQ); 5337 } 5338 VdbeComment((v, "%s", pIx->zName)); 5339 #ifdef SQLITE_ENABLE_COLUMN_USED_MASK 5340 { 5341 u64 colUsed = 0; 5342 int ii, jj; 5343 for(ii=0; ii<pIx->nColumn; ii++){ 5344 jj = pIx->aiColumn[ii]; 5345 if( jj<0 ) continue; 5346 if( jj>63 ) jj = 63; 5347 if( (pTabItem->colUsed & MASKBIT(jj))==0 ) continue; 5348 colUsed |= ((u64)1)<<(ii<63 ? ii : 63); 5349 } 5350 sqlite3VdbeAddOp4Dup8(v, OP_ColumnsUsed, iIndexCur, 0, 0, 5351 (u8*)&colUsed, P4_INT64); 5352 } 5353 #endif /* SQLITE_ENABLE_COLUMN_USED_MASK */ 5354 } 5355 } 5356 if( iDb>=0 ) sqlite3CodeVerifySchema(pParse, iDb); 5357 } 5358 pWInfo->iTop = sqlite3VdbeCurrentAddr(v); 5359 if( db->mallocFailed ) goto whereBeginError; 5360 5361 /* Generate the code to do the search. Each iteration of the for 5362 ** loop below generates code for a single nested loop of the VM 5363 ** program. 5364 */ 5365 for(ii=0; ii<nTabList; ii++){ 5366 int addrExplain; 5367 int wsFlags; 5368 if( pParse->nErr ) goto whereBeginError; 5369 pLevel = &pWInfo->a[ii]; 5370 wsFlags = pLevel->pWLoop->wsFlags; 5371 #ifndef SQLITE_OMIT_AUTOMATIC_INDEX 5372 if( (pLevel->pWLoop->wsFlags & WHERE_AUTO_INDEX)!=0 ){ 5373 constructAutomaticIndex(pParse, &pWInfo->sWC, 5374 &pTabList->a[pLevel->iFrom], notReady, pLevel); 5375 if( db->mallocFailed ) goto whereBeginError; 5376 } 5377 #endif 5378 addrExplain = sqlite3WhereExplainOneScan( 5379 pParse, pTabList, pLevel, wctrlFlags 5380 ); 5381 pLevel->addrBody = sqlite3VdbeCurrentAddr(v); 5382 notReady = sqlite3WhereCodeOneLoopStart(pParse,v,pWInfo,ii,pLevel,notReady); 5383 pWInfo->iContinue = pLevel->addrCont; 5384 if( (wsFlags&WHERE_MULTI_OR)==0 && (wctrlFlags&WHERE_OR_SUBCLAUSE)==0 ){ 5385 sqlite3WhereAddScanStatus(v, pTabList, pLevel, addrExplain); 5386 } 5387 } 5388 5389 /* Done. */ 5390 VdbeModuleComment((v, "Begin WHERE-core")); 5391 pWInfo->iEndWhere = sqlite3VdbeCurrentAddr(v); 5392 return pWInfo; 5393 5394 /* Jump here if malloc fails */ 5395 whereBeginError: 5396 if( pWInfo ){ 5397 testcase( pWInfo->pExprMods!=0 ); 5398 whereUndoExprMods(pWInfo); 5399 pParse->nQueryLoop = pWInfo->savedNQueryLoop; 5400 whereInfoFree(db, pWInfo); 5401 } 5402 return 0; 5403 } 5404 5405 /* 5406 ** Part of sqlite3WhereEnd() will rewrite opcodes to reference the 5407 ** index rather than the main table. In SQLITE_DEBUG mode, we want 5408 ** to trace those changes if PRAGMA vdbe_addoptrace=on. This routine 5409 ** does that. 5410 */ 5411 #ifndef SQLITE_DEBUG 5412 # define OpcodeRewriteTrace(D,K,P) /* no-op */ 5413 #else 5414 # define OpcodeRewriteTrace(D,K,P) sqlite3WhereOpcodeRewriteTrace(D,K,P) 5415 static void sqlite3WhereOpcodeRewriteTrace( 5416 sqlite3 *db, 5417 int pc, 5418 VdbeOp *pOp 5419 ){ 5420 if( (db->flags & SQLITE_VdbeAddopTrace)==0 ) return; 5421 sqlite3VdbePrintOp(0, pc, pOp); 5422 } 5423 #endif 5424 5425 /* 5426 ** Generate the end of the WHERE loop. See comments on 5427 ** sqlite3WhereBegin() for additional information. 5428 */ 5429 void sqlite3WhereEnd(WhereInfo *pWInfo){ 5430 Parse *pParse = pWInfo->pParse; 5431 Vdbe *v = pParse->pVdbe; 5432 int i; 5433 WhereLevel *pLevel; 5434 WhereLoop *pLoop; 5435 SrcList *pTabList = pWInfo->pTabList; 5436 sqlite3 *db = pParse->db; 5437 int iEnd = sqlite3VdbeCurrentAddr(v); 5438 5439 /* Generate loop termination code. 5440 */ 5441 VdbeModuleComment((v, "End WHERE-core")); 5442 for(i=pWInfo->nLevel-1; i>=0; i--){ 5443 int addr; 5444 pLevel = &pWInfo->a[i]; 5445 pLoop = pLevel->pWLoop; 5446 if( pLevel->op!=OP_Noop ){ 5447 #ifndef SQLITE_DISABLE_SKIPAHEAD_DISTINCT 5448 int addrSeek = 0; 5449 Index *pIdx; 5450 int n; 5451 if( pWInfo->eDistinct==WHERE_DISTINCT_ORDERED 5452 && i==pWInfo->nLevel-1 /* Ticket [ef9318757b152e3] 2017-10-21 */ 5453 && (pLoop->wsFlags & WHERE_INDEXED)!=0 5454 && (pIdx = pLoop->u.btree.pIndex)->hasStat1 5455 && (n = pLoop->u.btree.nDistinctCol)>0 5456 && pIdx->aiRowLogEst[n]>=36 5457 ){ 5458 int r1 = pParse->nMem+1; 5459 int j, op; 5460 for(j=0; j<n; j++){ 5461 sqlite3VdbeAddOp3(v, OP_Column, pLevel->iIdxCur, j, r1+j); 5462 } 5463 pParse->nMem += n+1; 5464 op = pLevel->op==OP_Prev ? OP_SeekLT : OP_SeekGT; 5465 addrSeek = sqlite3VdbeAddOp4Int(v, op, pLevel->iIdxCur, 0, r1, n); 5466 VdbeCoverageIf(v, op==OP_SeekLT); 5467 VdbeCoverageIf(v, op==OP_SeekGT); 5468 sqlite3VdbeAddOp2(v, OP_Goto, 1, pLevel->p2); 5469 } 5470 #endif /* SQLITE_DISABLE_SKIPAHEAD_DISTINCT */ 5471 /* The common case: Advance to the next row */ 5472 sqlite3VdbeResolveLabel(v, pLevel->addrCont); 5473 sqlite3VdbeAddOp3(v, pLevel->op, pLevel->p1, pLevel->p2, pLevel->p3); 5474 sqlite3VdbeChangeP5(v, pLevel->p5); 5475 VdbeCoverage(v); 5476 VdbeCoverageIf(v, pLevel->op==OP_Next); 5477 VdbeCoverageIf(v, pLevel->op==OP_Prev); 5478 VdbeCoverageIf(v, pLevel->op==OP_VNext); 5479 if( pLevel->regBignull ){ 5480 sqlite3VdbeResolveLabel(v, pLevel->addrBignull); 5481 sqlite3VdbeAddOp2(v, OP_DecrJumpZero, pLevel->regBignull, pLevel->p2-1); 5482 VdbeCoverage(v); 5483 } 5484 #ifndef SQLITE_DISABLE_SKIPAHEAD_DISTINCT 5485 if( addrSeek ) sqlite3VdbeJumpHere(v, addrSeek); 5486 #endif 5487 }else{ 5488 sqlite3VdbeResolveLabel(v, pLevel->addrCont); 5489 } 5490 if( (pLoop->wsFlags & WHERE_IN_ABLE)!=0 && pLevel->u.in.nIn>0 ){ 5491 struct InLoop *pIn; 5492 int j; 5493 sqlite3VdbeResolveLabel(v, pLevel->addrNxt); 5494 for(j=pLevel->u.in.nIn, pIn=&pLevel->u.in.aInLoop[j-1]; j>0; j--, pIn--){ 5495 assert( sqlite3VdbeGetOp(v, pIn->addrInTop+1)->opcode==OP_IsNull 5496 || pParse->db->mallocFailed ); 5497 sqlite3VdbeJumpHere(v, pIn->addrInTop+1); 5498 if( pIn->eEndLoopOp!=OP_Noop ){ 5499 if( pIn->nPrefix ){ 5500 int bEarlyOut = 5501 (pLoop->wsFlags & WHERE_VIRTUALTABLE)==0 5502 && (pLoop->wsFlags & WHERE_IN_EARLYOUT)!=0; 5503 if( pLevel->iLeftJoin ){ 5504 /* For LEFT JOIN queries, cursor pIn->iCur may not have been 5505 ** opened yet. This occurs for WHERE clauses such as 5506 ** "a = ? AND b IN (...)", where the index is on (a, b). If 5507 ** the RHS of the (a=?) is NULL, then the "b IN (...)" may 5508 ** never have been coded, but the body of the loop run to 5509 ** return the null-row. So, if the cursor is not open yet, 5510 ** jump over the OP_Next or OP_Prev instruction about to 5511 ** be coded. */ 5512 sqlite3VdbeAddOp2(v, OP_IfNotOpen, pIn->iCur, 5513 sqlite3VdbeCurrentAddr(v) + 2 + bEarlyOut); 5514 VdbeCoverage(v); 5515 } 5516 if( bEarlyOut ){ 5517 sqlite3VdbeAddOp4Int(v, OP_IfNoHope, pLevel->iIdxCur, 5518 sqlite3VdbeCurrentAddr(v)+2, 5519 pIn->iBase, pIn->nPrefix); 5520 VdbeCoverage(v); 5521 /* Retarget the OP_IsNull against the left operand of IN so 5522 ** it jumps past the OP_IfNoHope. This is because the 5523 ** OP_IsNull also bypasses the OP_Affinity opcode that is 5524 ** required by OP_IfNoHope. */ 5525 sqlite3VdbeJumpHere(v, pIn->addrInTop+1); 5526 } 5527 } 5528 sqlite3VdbeAddOp2(v, pIn->eEndLoopOp, pIn->iCur, pIn->addrInTop); 5529 VdbeCoverage(v); 5530 VdbeCoverageIf(v, pIn->eEndLoopOp==OP_Prev); 5531 VdbeCoverageIf(v, pIn->eEndLoopOp==OP_Next); 5532 } 5533 sqlite3VdbeJumpHere(v, pIn->addrInTop-1); 5534 } 5535 } 5536 sqlite3VdbeResolveLabel(v, pLevel->addrBrk); 5537 if( pLevel->addrSkip ){ 5538 sqlite3VdbeGoto(v, pLevel->addrSkip); 5539 VdbeComment((v, "next skip-scan on %s", pLoop->u.btree.pIndex->zName)); 5540 sqlite3VdbeJumpHere(v, pLevel->addrSkip); 5541 sqlite3VdbeJumpHere(v, pLevel->addrSkip-2); 5542 } 5543 #ifndef SQLITE_LIKE_DOESNT_MATCH_BLOBS 5544 if( pLevel->addrLikeRep ){ 5545 sqlite3VdbeAddOp2(v, OP_DecrJumpZero, (int)(pLevel->iLikeRepCntr>>1), 5546 pLevel->addrLikeRep); 5547 VdbeCoverage(v); 5548 } 5549 #endif 5550 if( pLevel->iLeftJoin ){ 5551 int ws = pLoop->wsFlags; 5552 addr = sqlite3VdbeAddOp1(v, OP_IfPos, pLevel->iLeftJoin); VdbeCoverage(v); 5553 assert( (ws & WHERE_IDX_ONLY)==0 || (ws & WHERE_INDEXED)!=0 ); 5554 if( (ws & WHERE_IDX_ONLY)==0 ){ 5555 assert( pLevel->iTabCur==pTabList->a[pLevel->iFrom].iCursor ); 5556 sqlite3VdbeAddOp1(v, OP_NullRow, pLevel->iTabCur); 5557 } 5558 if( (ws & WHERE_INDEXED) 5559 || ((ws & WHERE_MULTI_OR) && pLevel->u.pCoveringIdx) 5560 ){ 5561 if( ws & WHERE_MULTI_OR ){ 5562 Index *pIx = pLevel->u.pCoveringIdx; 5563 int iDb = sqlite3SchemaToIndex(db, pIx->pSchema); 5564 sqlite3VdbeAddOp3(v, OP_ReopenIdx, pLevel->iIdxCur, pIx->tnum, iDb); 5565 sqlite3VdbeSetP4KeyInfo(pParse, pIx); 5566 } 5567 sqlite3VdbeAddOp1(v, OP_NullRow, pLevel->iIdxCur); 5568 } 5569 if( pLevel->op==OP_Return ){ 5570 sqlite3VdbeAddOp2(v, OP_Gosub, pLevel->p1, pLevel->addrFirst); 5571 }else{ 5572 sqlite3VdbeGoto(v, pLevel->addrFirst); 5573 } 5574 sqlite3VdbeJumpHere(v, addr); 5575 } 5576 VdbeModuleComment((v, "End WHERE-loop%d: %s", i, 5577 pWInfo->pTabList->a[pLevel->iFrom].pTab->zName)); 5578 } 5579 5580 /* The "break" point is here, just past the end of the outer loop. 5581 ** Set it. 5582 */ 5583 sqlite3VdbeResolveLabel(v, pWInfo->iBreak); 5584 5585 assert( pWInfo->nLevel<=pTabList->nSrc ); 5586 for(i=0, pLevel=pWInfo->a; i<pWInfo->nLevel; i++, pLevel++){ 5587 int k, last; 5588 VdbeOp *pOp, *pLastOp; 5589 Index *pIdx = 0; 5590 SrcItem *pTabItem = &pTabList->a[pLevel->iFrom]; 5591 Table *pTab = pTabItem->pTab; 5592 assert( pTab!=0 ); 5593 pLoop = pLevel->pWLoop; 5594 5595 /* For a co-routine, change all OP_Column references to the table of 5596 ** the co-routine into OP_Copy of result contained in a register. 5597 ** OP_Rowid becomes OP_Null. 5598 */ 5599 if( pTabItem->fg.viaCoroutine ){ 5600 testcase( pParse->db->mallocFailed ); 5601 translateColumnToCopy(pParse, pLevel->addrBody, pLevel->iTabCur, 5602 pTabItem->regResult, 0); 5603 continue; 5604 } 5605 5606 #ifdef SQLITE_ENABLE_EARLY_CURSOR_CLOSE 5607 /* Close all of the cursors that were opened by sqlite3WhereBegin. 5608 ** Except, do not close cursors that will be reused by the OR optimization 5609 ** (WHERE_OR_SUBCLAUSE). And do not close the OP_OpenWrite cursors 5610 ** created for the ONEPASS optimization. 5611 */ 5612 if( (pTab->tabFlags & TF_Ephemeral)==0 5613 && !IsView(pTab) 5614 && (pWInfo->wctrlFlags & WHERE_OR_SUBCLAUSE)==0 5615 ){ 5616 int ws = pLoop->wsFlags; 5617 if( pWInfo->eOnePass==ONEPASS_OFF && (ws & WHERE_IDX_ONLY)==0 ){ 5618 sqlite3VdbeAddOp1(v, OP_Close, pTabItem->iCursor); 5619 } 5620 if( (ws & WHERE_INDEXED)!=0 5621 && (ws & (WHERE_IPK|WHERE_AUTO_INDEX))==0 5622 && pLevel->iIdxCur!=pWInfo->aiCurOnePass[1] 5623 ){ 5624 sqlite3VdbeAddOp1(v, OP_Close, pLevel->iIdxCur); 5625 } 5626 } 5627 #endif 5628 5629 /* If this scan uses an index, make VDBE code substitutions to read data 5630 ** from the index instead of from the table where possible. In some cases 5631 ** this optimization prevents the table from ever being read, which can 5632 ** yield a significant performance boost. 5633 ** 5634 ** Calls to the code generator in between sqlite3WhereBegin and 5635 ** sqlite3WhereEnd will have created code that references the table 5636 ** directly. This loop scans all that code looking for opcodes 5637 ** that reference the table and converts them into opcodes that 5638 ** reference the index. 5639 */ 5640 if( pLoop->wsFlags & (WHERE_INDEXED|WHERE_IDX_ONLY) ){ 5641 pIdx = pLoop->u.btree.pIndex; 5642 }else if( pLoop->wsFlags & WHERE_MULTI_OR ){ 5643 pIdx = pLevel->u.pCoveringIdx; 5644 } 5645 if( pIdx 5646 && !db->mallocFailed 5647 ){ 5648 if( pWInfo->eOnePass==ONEPASS_OFF || !HasRowid(pIdx->pTable) ){ 5649 last = iEnd; 5650 }else{ 5651 last = pWInfo->iEndWhere; 5652 } 5653 k = pLevel->addrBody + 1; 5654 #ifdef SQLITE_DEBUG 5655 if( db->flags & SQLITE_VdbeAddopTrace ){ 5656 printf("TRANSLATE opcodes in range %d..%d\n", k, last-1); 5657 } 5658 /* Proof that the "+1" on the k value above is safe */ 5659 pOp = sqlite3VdbeGetOp(v, k - 1); 5660 assert( pOp->opcode!=OP_Column || pOp->p1!=pLevel->iTabCur ); 5661 assert( pOp->opcode!=OP_Rowid || pOp->p1!=pLevel->iTabCur ); 5662 assert( pOp->opcode!=OP_IfNullRow || pOp->p1!=pLevel->iTabCur ); 5663 #endif 5664 pOp = sqlite3VdbeGetOp(v, k); 5665 pLastOp = pOp + (last - k); 5666 assert( pOp<=pLastOp ); 5667 do{ 5668 if( pOp->p1!=pLevel->iTabCur ){ 5669 /* no-op */ 5670 }else if( pOp->opcode==OP_Column 5671 #ifdef SQLITE_ENABLE_OFFSET_SQL_FUNC 5672 || pOp->opcode==OP_Offset 5673 #endif 5674 ){ 5675 int x = pOp->p2; 5676 assert( pIdx->pTable==pTab ); 5677 if( !HasRowid(pTab) ){ 5678 Index *pPk = sqlite3PrimaryKeyIndex(pTab); 5679 x = pPk->aiColumn[x]; 5680 assert( x>=0 ); 5681 }else{ 5682 testcase( x!=sqlite3StorageColumnToTable(pTab,x) ); 5683 x = sqlite3StorageColumnToTable(pTab,x); 5684 } 5685 x = sqlite3TableColumnToIndex(pIdx, x); 5686 if( x>=0 ){ 5687 pOp->p2 = x; 5688 pOp->p1 = pLevel->iIdxCur; 5689 OpcodeRewriteTrace(db, k, pOp); 5690 } 5691 assert( (pLoop->wsFlags & WHERE_IDX_ONLY)==0 || x>=0 5692 || pWInfo->eOnePass ); 5693 }else if( pOp->opcode==OP_Rowid ){ 5694 pOp->p1 = pLevel->iIdxCur; 5695 pOp->opcode = OP_IdxRowid; 5696 OpcodeRewriteTrace(db, k, pOp); 5697 }else if( pOp->opcode==OP_IfNullRow ){ 5698 pOp->p1 = pLevel->iIdxCur; 5699 OpcodeRewriteTrace(db, k, pOp); 5700 } 5701 #ifdef SQLITE_DEBUG 5702 k++; 5703 #endif 5704 }while( (++pOp)<pLastOp ); 5705 #ifdef SQLITE_DEBUG 5706 if( db->flags & SQLITE_VdbeAddopTrace ) printf("TRANSLATE complete\n"); 5707 #endif 5708 } 5709 } 5710 5711 /* Final cleanup 5712 */ 5713 if( pWInfo->pExprMods ) whereUndoExprMods(pWInfo); 5714 pParse->nQueryLoop = pWInfo->savedNQueryLoop; 5715 whereInfoFree(db, pWInfo); 5716 return; 5717 } 5718