Many hyperlinks are disabled.
Use anonymous login
to enable hyperlinks.
Overview
Comment: | Optimizations for the matchinfo() function, particularly the 'y' flag. |
---|---|
Downloads: | Tarball | ZIP archive |
Timelines: | family | ancestors | descendants | both | fts3-matchinfo-y |
Files: | files | file ages | folders |
SHA1: |
dddd7e182943a1d3a9d32830e819a63f |
User & Date: | dan 2015-05-05 19:37:07.819 |
Context
2015-05-05
| ||
20:39 | Add the fts3 matchinfo 'b' flag. (check-in: b9b77972d8 user: dan tags: fts3-matchinfo-y) | |
19:37 | Optimizations for the matchinfo() function, particularly the 'y' flag. (check-in: dddd7e1829 user: dan tags: fts3-matchinfo-y) | |
11:08 | Add #ifdef statements to test_blob.c so that it will build with SQLITE_OMIT_INCRBLOB. (check-in: b8f090e65d user: drh tags: trunk) | |
Changes
Changes to ext/fts3/fts3.c.
︙ | ︙ | |||
1671 1672 1673 1674 1675 1676 1677 | static int fts3CloseMethod(sqlite3_vtab_cursor *pCursor){ Fts3Cursor *pCsr = (Fts3Cursor *)pCursor; assert( ((Fts3Table *)pCsr->base.pVtab)->pSegments==0 ); sqlite3_finalize(pCsr->pStmt); sqlite3Fts3ExprFree(pCsr->pExpr); sqlite3Fts3FreeDeferredTokens(pCsr); sqlite3_free(pCsr->aDoclist); | | | 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 | static int fts3CloseMethod(sqlite3_vtab_cursor *pCursor){ Fts3Cursor *pCsr = (Fts3Cursor *)pCursor; assert( ((Fts3Table *)pCsr->base.pVtab)->pSegments==0 ); sqlite3_finalize(pCsr->pStmt); sqlite3Fts3ExprFree(pCsr->pExpr); sqlite3Fts3FreeDeferredTokens(pCsr); sqlite3_free(pCsr->aDoclist); sqlite3Fts3MIBufferFree(pCsr->pMIBuffer); assert( ((Fts3Table *)pCsr->base.pVtab)->pSegments==0 ); sqlite3_free(pCsr); return SQLITE_OK; } /* ** If pCsr->pStmt has not been prepared (i.e. if pCsr->pStmt==0), then |
︙ | ︙ | |||
3172 3173 3174 3175 3176 3177 3178 | if( idxNum & FTS3_HAVE_DOCID_GE ) pDocidGe = apVal[iIdx++]; if( idxNum & FTS3_HAVE_DOCID_LE ) pDocidLe = apVal[iIdx++]; assert( iIdx==nVal ); /* In case the cursor has been used before, clear it now. */ sqlite3_finalize(pCsr->pStmt); sqlite3_free(pCsr->aDoclist); | | | 3172 3173 3174 3175 3176 3177 3178 3179 3180 3181 3182 3183 3184 3185 3186 | if( idxNum & FTS3_HAVE_DOCID_GE ) pDocidGe = apVal[iIdx++]; if( idxNum & FTS3_HAVE_DOCID_LE ) pDocidLe = apVal[iIdx++]; assert( iIdx==nVal ); /* In case the cursor has been used before, clear it now. */ sqlite3_finalize(pCsr->pStmt); sqlite3_free(pCsr->aDoclist); sqlite3Fts3MIBufferFree(pCsr->pMIBuffer); sqlite3Fts3ExprFree(pCsr->pExpr); memset(&pCursor[1], 0, sizeof(Fts3Cursor)-sizeof(sqlite3_vtab_cursor)); /* Set the lower and upper bounds on docids to return */ pCsr->iMinDocid = fts3DocidRange(pDocidGe, SMALLEST_INT64); pCsr->iMaxDocid = fts3DocidRange(pDocidLe, LARGEST_INT64); |
︙ | ︙ |
Changes to ext/fts3/fts3Int.h.
︙ | ︙ | |||
193 194 195 196 197 198 199 200 201 202 203 204 205 206 | typedef struct Fts3Doclist Fts3Doclist; typedef struct Fts3SegFilter Fts3SegFilter; typedef struct Fts3DeferredToken Fts3DeferredToken; typedef struct Fts3SegReader Fts3SegReader; typedef struct Fts3MultiSegReader Fts3MultiSegReader; /* ** A connection to a fulltext index is an instance of the following ** structure. The xCreate and xConnect methods create an instance ** of this structure and xDestroy and xDisconnect free that instance. ** All other methods receive a pointer to the structure as one of their ** arguments. */ | > > | 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 | typedef struct Fts3Doclist Fts3Doclist; typedef struct Fts3SegFilter Fts3SegFilter; typedef struct Fts3DeferredToken Fts3DeferredToken; typedef struct Fts3SegReader Fts3SegReader; typedef struct Fts3MultiSegReader Fts3MultiSegReader; typedef struct MatchinfoBuffer MatchinfoBuffer; /* ** A connection to a fulltext index is an instance of the following ** structure. The xCreate and xConnect methods create an instance ** of this structure and xDestroy and xDisconnect free that instance. ** All other methods receive a pointer to the structure as one of their ** arguments. */ |
︙ | ︙ | |||
302 303 304 305 306 307 308 | u8 bDesc; /* True to sort in descending order */ int eEvalmode; /* An FTS3_EVAL_XX constant */ int nRowAvg; /* Average size of database rows, in pages */ sqlite3_int64 nDoc; /* Documents in table */ i64 iMinDocid; /* Minimum docid to return */ i64 iMaxDocid; /* Maximum docid to return */ int isMatchinfoNeeded; /* True when aMatchinfo[] needs filling in */ | < < | | 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 | u8 bDesc; /* True to sort in descending order */ int eEvalmode; /* An FTS3_EVAL_XX constant */ int nRowAvg; /* Average size of database rows, in pages */ sqlite3_int64 nDoc; /* Documents in table */ i64 iMinDocid; /* Minimum docid to return */ i64 iMaxDocid; /* Maximum docid to return */ int isMatchinfoNeeded; /* True when aMatchinfo[] needs filling in */ MatchinfoBuffer *pMIBuffer; /* Buffer for matchinfo data */ }; #define FTS3_EVAL_FILTER 0 #define FTS3_EVAL_NEXT 1 #define FTS3_EVAL_MATCHINFO 2 /* |
︙ | ︙ | |||
560 561 562 563 564 565 566 567 568 569 570 571 572 573 | /* fts3_snippet.c */ void sqlite3Fts3Offsets(sqlite3_context*, Fts3Cursor*); void sqlite3Fts3Snippet(sqlite3_context *, Fts3Cursor *, const char *, const char *, const char *, int, int ); void sqlite3Fts3Matchinfo(sqlite3_context *, Fts3Cursor *, const char *); /* fts3_expr.c */ int sqlite3Fts3ExprParse(sqlite3_tokenizer *, int, char **, int, int, int, const char *, int, Fts3Expr **, char ** ); void sqlite3Fts3ExprFree(Fts3Expr *); #ifdef SQLITE_TEST | > | 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 | /* fts3_snippet.c */ void sqlite3Fts3Offsets(sqlite3_context*, Fts3Cursor*); void sqlite3Fts3Snippet(sqlite3_context *, Fts3Cursor *, const char *, const char *, const char *, int, int ); void sqlite3Fts3Matchinfo(sqlite3_context *, Fts3Cursor *, const char *); void sqlite3Fts3MIBufferFree(MatchinfoBuffer *p); /* fts3_expr.c */ int sqlite3Fts3ExprParse(sqlite3_tokenizer *, int, char **, int, int, int, const char *, int, Fts3Expr **, char ** ); void sqlite3Fts3ExprFree(Fts3Expr *); #ifdef SQLITE_TEST |
︙ | ︙ |
Changes to ext/fts3/fts3_snippet.c.
︙ | ︙ | |||
88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 | Fts3Cursor *pCursor; /* FTS3 Cursor */ int nCol; /* Number of columns in table */ int nPhrase; /* Number of matchable phrases in query */ sqlite3_int64 nDoc; /* Number of docs in database */ u32 *aMatchinfo; /* Pre-allocated buffer */ }; /* ** The snippet() and offsets() functions both return text values. An instance ** of the following structure is used to accumulate those values while the ** functions are running. See fts3StringAppend() for details. */ typedef struct StrBuffer StrBuffer; struct StrBuffer { char *z; /* Pointer to buffer containing string */ int n; /* Length of z in bytes (excl. nul-term) */ int nAlloc; /* Allocated size of buffer z in bytes */ }; /* ** This function is used to help iterate through a position-list. A position ** list is a list of unique integers, sorted from smallest to largest. Each ** element of the list is represented by an FTS3 varint that takes the value ** of the difference between the current element and the previous one plus ** two. For example, to store the position-list: | > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > | 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 | Fts3Cursor *pCursor; /* FTS3 Cursor */ int nCol; /* Number of columns in table */ int nPhrase; /* Number of matchable phrases in query */ sqlite3_int64 nDoc; /* Number of docs in database */ u32 *aMatchinfo; /* Pre-allocated buffer */ }; /* ** An instance of this structure is used to manage a pair of buffers, each ** (nElem * sizeof(u32)) bytes in size. See the MatchinfoBuffer code below ** for details. */ struct MatchinfoBuffer { u8 aRef[3]; int nElem; int bGlobal; /* Set if global data is loaded */ char *zMatchinfo; u32 aMatchinfo[0]; }; /* ** The snippet() and offsets() functions both return text values. An instance ** of the following structure is used to accumulate those values while the ** functions are running. See fts3StringAppend() for details. */ typedef struct StrBuffer StrBuffer; struct StrBuffer { char *z; /* Pointer to buffer containing string */ int n; /* Length of z in bytes (excl. nul-term) */ int nAlloc; /* Allocated size of buffer z in bytes */ }; /************************************************************************* ** Start of MatchinfoBuffer code. */ /* ** Allocate a two-slot MatchinfoBuffer object. */ static MatchinfoBuffer *fts3MIBufferNew(int nElem, const char *zMatchinfo){ MatchinfoBuffer *pRet; int nByte = sizeof(u32) * (2*nElem + 2) + sizeof(MatchinfoBuffer); int nStr = strlen(zMatchinfo); pRet = sqlite3_malloc(nByte + nStr+1); if( pRet ){ memset(pRet, 0, nByte); pRet->aMatchinfo[0] = (u8*)(&pRet->aMatchinfo[1]) - (u8*)pRet; pRet->aMatchinfo[1+nElem] = pRet->aMatchinfo[0] + sizeof(u32)*(nElem+1); pRet->nElem = nElem; pRet->zMatchinfo = ((char*)pRet) + nByte; memcpy(pRet->zMatchinfo, zMatchinfo, nStr+1); pRet->aRef[0] = 1; } return pRet; } static void fts3MIBufferFree(void *p){ MatchinfoBuffer *pBuf = (MatchinfoBuffer*)((u8*)p - ((u32*)p)[-1]); assert( (u32*)p==&pBuf->aMatchinfo[1] || (u32*)p==&pBuf->aMatchinfo[pBuf->nElem+2] ); if( (u32*)p==&pBuf->aMatchinfo[1] ){ pBuf->aRef[1] = 0; }else{ pBuf->aRef[2] = 0; } if( pBuf->aRef[0]==0 && pBuf->aRef[1]==0 && pBuf->aRef[2]==0 ){ sqlite3_free(pBuf); } } static void (*fts3MIBufferAlloc(MatchinfoBuffer *p, u32 **paOut))(void*){ void (*xRet)(void*) = 0; u32 *aOut = 0; if( p->aRef[1]==0 ){ p->aRef[1] = 1; aOut = &p->aMatchinfo[1]; xRet = fts3MIBufferFree; } else if( p->aRef[2]==0 ){ p->aRef[2] = 1; aOut = &p->aMatchinfo[p->nElem+2]; xRet = fts3MIBufferFree; }else{ aOut = (u32*)sqlite3_malloc(p->nElem * sizeof(u32)); if( aOut ){ xRet = sqlite3_free; if( p->bGlobal ) memcpy(aOut, &p->aMatchinfo[1], p->nElem*sizeof(u32)); } } *paOut = aOut; return xRet; } static void fts3MIBufferSetGlobal(MatchinfoBuffer *p){ p->bGlobal = 1; memcpy(&p->aMatchinfo[2+p->nElem], &p->aMatchinfo[1], p->nElem*sizeof(u32)); } /* ** Free a MatchinfoBuffer object allocated using fts3MIBufferNew() */ void sqlite3Fts3MIBufferFree(MatchinfoBuffer *p){ if( p ){ assert( p->aRef[0]==1 ); p->aRef[0] = 0; if( p->aRef[0]==0 && p->aRef[1]==0 && p->aRef[2]==0 ){ sqlite3_free(p); } } } /* ** End of MatchinfoBuffer code. *************************************************************************/ /* ** This function is used to help iterate through a position-list. A position ** list is a list of unique integers, sorted from smallest to largest. Each ** element of the list is represented by an FTS3 varint that takes the value ** of the difference between the current element and the previous one plus ** two. For example, to store the position-list: |
︙ | ︙ | |||
129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 | */ static void fts3GetDeltaPosition(char **pp, int *piPos){ int iVal; *pp += fts3GetVarint32(*pp, &iVal); *piPos += (iVal-2); } /* ** Helper function for fts3ExprIterate() (see below). */ static int fts3ExprIterate2( Fts3Expr *pExpr, /* Expression to iterate phrases of */ int *piPhrase, /* Pointer to phrase counter */ int (*x)(Fts3Expr*,int,void*), /* Callback function to invoke for phrases */ void *pCtx /* Second argument to pass to callback */ ){ int rc; /* Return code */ | > > < > > > > | | | | | | | | | > | 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 | */ static void fts3GetDeltaPosition(char **pp, int *piPos){ int iVal; *pp += fts3GetVarint32(*pp, &iVal); *piPos += (iVal-2); } static int fts3ExprLHitsCb(Fts3Expr*,int,void*); /* ** Helper function for fts3ExprIterate() (see below). */ static int fts3ExprIterate2( Fts3Expr *pExpr, /* Expression to iterate phrases of */ int *piPhrase, /* Pointer to phrase counter */ int (*x)(Fts3Expr*,int,void*), /* Callback function to invoke for phrases */ void *pCtx /* Second argument to pass to callback */ ){ int rc; /* Return code */ if( x==fts3ExprLHitsCb && pExpr->bEof ){ rc = SQLITE_OK; }else{ int eType = pExpr->eType; /* Type of expression node pExpr */ if( eType!=FTSQUERY_PHRASE ){ assert( pExpr->pLeft && pExpr->pRight ); rc = fts3ExprIterate2(pExpr->pLeft, piPhrase, x, pCtx); if( rc==SQLITE_OK && eType!=FTSQUERY_NOT ){ rc = fts3ExprIterate2(pExpr->pRight, piPhrase, x, pCtx); } }else{ rc = x(pExpr, *piPhrase, pCtx); (*piPhrase)++; } } return rc; } /* ** Iterate through all phrase nodes in an FTS3 query, except those that ** are part of a sub-tree that is the right-hand-side of a NOT operator. |
︙ | ︙ | |||
815 816 817 818 819 820 821 | ** directive 'y'. */ static int fts3ExprLHitsCb( Fts3Expr *pExpr, /* Phrase expression node */ int iPhrase, /* Phrase number */ void *pCtx /* Pointer to MatchInfo structure */ ){ | < < < < > < < < < < < < | > > | 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 | ** directive 'y'. */ static int fts3ExprLHitsCb( Fts3Expr *pExpr, /* Phrase expression node */ int iPhrase, /* Phrase number */ void *pCtx /* Pointer to MatchInfo structure */ ){ int rc = SQLITE_OK; MatchInfo *p = (MatchInfo *)pCtx; /* This must be a phrase */ assert( pExpr->pPhrase ); if( pExpr->iDocid==p->pCursor->iPrevId ){ Fts3Table *pTab = (Fts3Table *)p->pCursor->base.pVtab; int iStart = iPhrase * p->nCol; Fts3Phrase *pPhrase = pExpr->pPhrase; char *pIter = pPhrase->doclist.pList; int iCol = 0; while( 1 ){ int nHit = fts3ColumnlistCount(&pIter); if( (pPhrase->iColumn>=pTab->nColumn || pPhrase->iColumn==iCol) ){ |
︙ | ︙ | |||
1145 1146 1147 1148 1149 1150 1151 | case FTS3_MATCHINFO_LCS: rc = fts3ExprLoadDoclists(pCsr, 0, 0); if( rc==SQLITE_OK ){ rc = fts3MatchinfoLcs(pCsr, pInfo); } break; | | > > > | 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 | case FTS3_MATCHINFO_LCS: rc = fts3ExprLoadDoclists(pCsr, 0, 0); if( rc==SQLITE_OK ){ rc = fts3MatchinfoLcs(pCsr, pInfo); } break; case FTS3_MATCHINFO_LHITS: { int nZero = fts3MatchinfoSize(pInfo, FTS3_MATCHINFO_LHITS)*sizeof(u32); memset(pInfo->aMatchinfo, 0, nZero); (void)fts3ExprIterate(pCsr->pExpr, fts3ExprLHitsCb, (void*)pInfo); break; } default: { Fts3Expr *pExpr; assert( zArg[i]==FTS3_MATCHINFO_HITS ); pExpr = pCsr->pExpr; rc = fts3ExprLoadDoclists(pCsr, 0, 0); if( rc!=SQLITE_OK ) break; |
︙ | ︙ | |||
1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 | /* ** Populate pCsr->aMatchinfo[] with data for the current row. The ** 'matchinfo' data is an array of 32-bit unsigned integers (C type u32). */ static int fts3GetMatchinfo( Fts3Cursor *pCsr, /* FTS3 Cursor object */ const char *zArg /* Second argument to matchinfo() function */ ){ MatchInfo sInfo; Fts3Table *pTab = (Fts3Table *)pCsr->base.pVtab; int rc = SQLITE_OK; int bGlobal = 0; /* Collect 'global' stats as well as local */ memset(&sInfo, 0, sizeof(MatchInfo)); sInfo.pCursor = pCsr; sInfo.nCol = pTab->nColumn; /* If there is cached matchinfo() data, but the format string for the ** cache does not match the format string for this request, discard ** the cached data. */ | > > > > | < | | < | | | < | < < < < > > > > > > > > | | < > > > > | > > > > > > | 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 | /* ** Populate pCsr->aMatchinfo[] with data for the current row. The ** 'matchinfo' data is an array of 32-bit unsigned integers (C type u32). */ static int fts3GetMatchinfo( sqlite3_context *pCtx, /* Return results here */ Fts3Cursor *pCsr, /* FTS3 Cursor object */ const char *zArg /* Second argument to matchinfo() function */ ){ MatchInfo sInfo; Fts3Table *pTab = (Fts3Table *)pCsr->base.pVtab; int rc = SQLITE_OK; int bGlobal = 0; /* Collect 'global' stats as well as local */ u32 *aOut = 0; void (*xDestroyOut)(void*) = 0; memset(&sInfo, 0, sizeof(MatchInfo)); sInfo.pCursor = pCsr; sInfo.nCol = pTab->nColumn; /* If there is cached matchinfo() data, but the format string for the ** cache does not match the format string for this request, discard ** the cached data. */ if( pCsr->pMIBuffer && strcmp(pCsr->pMIBuffer->zMatchinfo, zArg) ){ sqlite3Fts3MIBufferFree(pCsr->pMIBuffer); pCsr->pMIBuffer = 0; } /* If Fts3Cursor.pMIBuffer is NULL, then this is the first time the ** matchinfo function has been called for this query. In this case ** allocate the array used to accumulate the matchinfo data and ** initialize those elements that are constant for every row. */ if( pCsr->pMIBuffer==0 ){ int nMatchinfo = 0; /* Number of u32 elements in match-info */ int nArg; /* Bytes in zArg */ int i; /* Used to iterate through zArg */ /* Determine the number of phrases in the query */ pCsr->nPhrase = fts3ExprPhraseCount(pCsr->pExpr); sInfo.nPhrase = pCsr->nPhrase; /* Determine the number of integers in the buffer returned by this call. */ for(i=0; zArg[i]; i++){ nMatchinfo += fts3MatchinfoSize(&sInfo, zArg[i]); } /* Allocate space for Fts3Cursor.aMatchinfo[] and Fts3Cursor.zMatchinfo. */ pCsr->pMIBuffer = fts3MIBufferNew(nMatchinfo, zArg); if( !pCsr->pMIBuffer ) rc = SQLITE_NOMEM; pCsr->isMatchinfoNeeded = 1; bGlobal = 1; } if( rc==SQLITE_OK ){ xDestroyOut = fts3MIBufferAlloc(pCsr->pMIBuffer, &aOut); if( xDestroyOut==0 ){ rc = SQLITE_NOMEM; } } if( rc==SQLITE_OK ){ sInfo.aMatchinfo = aOut; sInfo.nPhrase = pCsr->nPhrase; rc = fts3MatchinfoValues(pCsr, bGlobal, &sInfo, zArg); if( bGlobal ){ fts3MIBufferSetGlobal(pCsr->pMIBuffer); } } if( rc!=SQLITE_OK ){ sqlite3_result_error_code(pCtx, rc); if( xDestroyOut ) xDestroyOut(aOut); }else{ int n = pCsr->pMIBuffer->nElem * sizeof(u32); sqlite3_result_blob(pCtx, aOut, n, xDestroyOut); } return rc; } /* ** Implementation of snippet() function. |
︙ | ︙ | |||
1564 1565 1566 1567 1568 1569 1570 | if( !pCsr->pExpr ){ sqlite3_result_blob(pContext, "", 0, SQLITE_STATIC); return; } /* Retrieve matchinfo() data. */ | | < < < < < < < | 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 | if( !pCsr->pExpr ){ sqlite3_result_blob(pContext, "", 0, SQLITE_STATIC); return; } /* Retrieve matchinfo() data. */ rc = fts3GetMatchinfo(pContext, pCsr, zFormat); sqlite3Fts3SegmentsClose(pTab); } #endif |