diff --git a/ext/fts3/fts3.c b/ext/fts3/fts3.c index ae7e98c26d..0c7db56c52 100644 --- a/ext/fts3/fts3.c +++ b/ext/fts3/fts3.c @@ -285,6 +285,7 @@ #include #include "fts3.h" +#include "fts3_expr.h" #include "fts3_hash.h" #include "fts3_tokenizer.h" #ifndef SQLITE_CORE @@ -312,11 +313,6 @@ # define FTSTRACE(A) #endif -/* -** Default span for NEAR operators. -*/ -#define SQLITE_FTS3_DEFAULT_NEAR_PARAM 10 - /* It is not safe to call isspace(), tolower(), or isalnum() on ** hi-bit-set characters. This is the same solution used in the ** tokenizer. @@ -1788,90 +1784,6 @@ static int sql_prepare(sqlite3 *db, const char *zDb, const char *zName, /* Forward reference */ typedef struct fulltext_vtab fulltext_vtab; -/* A single term in a query is represented by an instances of -** the following structure. Each word which may match against -** document content is a term. Operators, like NEAR or OR, are -** not terms. Query terms are organized as a flat list stored -** in the Query.pTerms array. -** -** If the QueryTerm.nPhrase variable is non-zero, then the QueryTerm -** is the first in a contiguous string of terms that are either part -** of the same phrase, or connected by the NEAR operator. -** -** If the QueryTerm.nNear variable is non-zero, then the token is followed -** by a NEAR operator with span set to (nNear-1). For example, the -** following query: -** -** The QueryTerm.iPhrase variable stores the index of the token within -** its phrase, indexed starting at 1, or 1 if the token is not part -** of any phrase. -** -** For example, the data structure used to represent the following query: -** -** ... MATCH 'sqlite NEAR/5 google NEAR/2 "search engine"' -** -** is: -** -** {nPhrase=4, iPhrase=1, nNear=6, pTerm="sqlite"}, -** {nPhrase=0, iPhrase=1, nNear=3, pTerm="google"}, -** {nPhrase=0, iPhrase=1, nNear=0, pTerm="search"}, -** {nPhrase=0, iPhrase=2, nNear=0, pTerm="engine"}, -** -** compiling the FTS3 syntax to Query structures is done by the parseQuery() -** function. -*/ -typedef struct QueryTerm { - short int nPhrase; /* How many following terms are part of the same phrase */ - short int iPhrase; /* This is the i-th term of a phrase. */ - short int iColumn; /* Column of the index that must match this term */ - short int nNear; /* term followed by a NEAR operator with span=(nNear-1) */ - signed char isOr; /* this term is preceded by "OR" */ - signed char isNot; /* this term is preceded by "-" */ - signed char isPrefix; /* this term is followed by "*" */ - char *pTerm; /* text of the term. '\000' terminated. malloced */ - int nTerm; /* Number of bytes in pTerm[] */ -} QueryTerm; - - -/* A query string is parsed into a Query structure. - * - * We could, in theory, allow query strings to be complicated - * nested expressions with precedence determined by parentheses. - * But none of the major search engines do this. (Perhaps the - * feeling is that an parenthesized expression is two complex of - * an idea for the average user to grasp.) Taking our lead from - * the major search engines, we will allow queries to be a list - * of terms (with an implied AND operator) or phrases in double-quotes, - * with a single optional "-" before each non-phrase term to designate - * negation and an optional OR connector. - * - * OR binds more tightly than the implied AND, which is what the - * major search engines seem to do. So, for example: - * - * [one two OR three] ==> one AND (two OR three) - * [one OR two three] ==> (one OR two) AND three - * - * A "-" before a term matches all entries that lack that term. - * The "-" must occur immediately before the term with in intervening - * space. This is how the search engines do it. - * - * A NOT term cannot be the right-hand operand of an OR. If this - * occurs in the query string, the NOT is ignored: - * - * [one OR -two] ==> one OR two - * - */ -typedef struct Query { - fulltext_vtab *pFts; /* The full text index */ - int nTerms; /* Number of terms in the query */ - QueryTerm *pTerms; /* Array of terms. Space obtained from malloc() */ - int nextIsOr; /* Set the isOr flag on the next inserted term */ - int nextIsNear; /* Set the isOr flag on the next inserted term */ - int nextColumn; /* Next word parsed must be in this column */ - int dfltColumn; /* The default column */ -} Query; - - /* ** An instance of the following structure keeps track of generated ** matching-word offset information and snippets. @@ -2022,14 +1934,14 @@ typedef struct fulltext_cursor { QueryType iCursorType; /* Copy of sqlite3_index_info.idxNum */ sqlite3_stmt *pStmt; /* Prepared statement in use by the cursor */ int eof; /* True if at End Of Results */ - Query q; /* Parsed query string */ + Fts3Expr *pExpr; /* Parsed MATCH query string */ Snippet snippet; /* Cached snippet for the current row */ int iColumn; /* Column being searched */ DataBuffer result; /* Doclist results from fulltextQuery */ DLReader reader; /* Result reader if result not empty */ } fulltext_cursor; -static struct fulltext_vtab *cursor_vtab(fulltext_cursor *c){ +static fulltext_vtab *cursor_vtab(fulltext_cursor *c){ return (fulltext_vtab *) c->base.pVtab; } @@ -3177,18 +3089,6 @@ static int fulltextOpen(sqlite3_vtab *pVTab, sqlite3_vtab_cursor **ppCursor){ } } - -/* Free all of the dynamically allocated memory held by *q -*/ -static void queryClear(Query *q){ - int i; - for(i = 0; i < q->nTerms; ++i){ - sqlite3_free(q->pTerms[i].pTerm); - } - sqlite3_free(q->pTerms); - CLEAR(q); -} - /* Free all of the dynamically allocated memory held by the ** Snippet */ @@ -3198,6 +3098,7 @@ static void snippetClear(Snippet *p){ sqlite3_free(p->zSnippet); CLEAR(p); } + /* ** Append a single entry to the p->aMatch[] log. */ @@ -3233,12 +3134,51 @@ static void snippetAppendMatch( #define FTS3_ROTOR_SZ (32) #define FTS3_ROTOR_MASK (FTS3_ROTOR_SZ-1) +/* +** Function to iterate through the tokens of a compiled expression. +*/ +static int nextExprToken(Fts3Expr **ppExpr, int *piToken){ + Fts3Expr *p = *ppExpr; + int iToken = *piToken; + if( iToken<0 ){ + /* In this case the expression p is the root of an expression tree. + ** Move to the first token in the expression tree. + */ + while( p->pLeft ){ + p = p->pLeft; + } + iToken = 0; + }else{ + assert(p && p->eType==FTSQUERY_PHRASE ); + if( iToken<(p->pPhrase->nToken-1) ){ + iToken++; + }else{ + iToken = 0; + while( p->pParent && p->pParent->pLeft!=p ){ + assert( p->pParent->pRight==p ); + p = p->pParent; + } + p = p->pParent; + if( p ){ + p = p->pRight; + while( p->pLeft ){ + p = p->pLeft; + } + } + } + } + + *ppExpr = p; + *piToken = iToken; + return p?1:0; +} + /* ** Add entries to pSnippet->aMatch[] for every match that occurs against ** document zDoc[0..nDoc-1] which is stored in column iColumn. */ static void snippetOffsetsOfColumn( - Query *pQuery, + fulltext_cursor *pCur, Snippet *pSnippet, int iColumn, const char *zDoc, @@ -3249,8 +3189,6 @@ static void snippetOffsetsOfColumn( sqlite3_tokenizer_cursor *pTCursor; /* Tokenizer cursor */ fulltext_vtab *pVtab; /* The full text index */ int nColumn; /* Number of columns in the index */ - const QueryTerm *aTerm; /* Query string terms */ - int nTerm; /* Number of query string terms */ int i, j; /* Loop counters */ int rc; /* Return code */ unsigned int match, prevMatch; /* Phrase search bitmasks */ @@ -3264,37 +3202,34 @@ static void snippetOffsetsOfColumn( int iRotorBegin[FTS3_ROTOR_SZ]; /* Beginning offset of token */ int iRotorLen[FTS3_ROTOR_SZ]; /* Length of token */ - pVtab = pQuery->pFts; + pVtab = cursor_vtab(pCur); nColumn = pVtab->nColumn; pTokenizer = pVtab->pTokenizer; pTModule = pTokenizer->pModule; rc = pTModule->xOpen(pTokenizer, zDoc, nDoc, &pTCursor); if( rc ) return; pTCursor->pTokenizer = pTokenizer; - aTerm = pQuery->pTerms; - nTerm = pQuery->nTerms; - if( nTerm>=FTS3_ROTOR_SZ ){ - nTerm = FTS3_ROTOR_SZ - 1; - } + prevMatch = 0; - while(1){ - rc = pTModule->xNext(pTCursor, &zToken, &nToken, &iBegin, &iEnd, &iPos); - if( rc ) break; + while( !pTModule->xNext(pTCursor, &zToken, &nToken, &iBegin, &iEnd, &iPos) ){ + Fts3Expr *pIter = pCur->pExpr; + int iIter = -1; iRotorBegin[iRotor&FTS3_ROTOR_MASK] = iBegin; iRotorLen[iRotor&FTS3_ROTOR_MASK] = iEnd-iBegin; match = 0; - for(i=0; ipPhrase->nToken; /* Tokens in current phrase */ + struct PhraseToken *pToken = &pIter->pPhrase->aToken[iIter]; + int iCol = pIter->pPhrase->iColumn; if( iCol>=0 && iColnToken ) continue; - if( !aTerm[i].isPrefix && aTerm[i].nTerm1 && (prevMatch & (1<n>nToken ) continue; + if( !pToken->isPrefix && pToken->nn<=nToken ); + if( memcmp(pToken->z, zToken, pToken->n) ) continue; + if( iIter>0 && (prevMatch & (1<=0; j--){ + if( i==(FTS3_ROTOR_SZ-2) || nPhrase==iIter+1 ){ + for(j=nPhrase-1; j>=0; j--){ int k = (iRotor-j) & FTS3_ROTOR_MASK; snippetAppendMatch(pSnippet, iColumn, i-j, iPos-j, iRotorBegin[k], iRotorLen[k]); @@ -3324,86 +3259,113 @@ static void snippetOffsetsOfColumn( ** then when this function is called the Snippet contains token offsets ** 0, 4 and 5. This function removes the "0" entry (because the first A ** is not near enough to an E). +** +** When this function is called, the value pointed to by parameter piLeft is +** the integer id of the left-most token in the expression tree headed by +** pExpr. This function increments *piLeft by the total number of tokens +** in the expression tree headed by pExpr. */ -static void trimSnippetOffsetsForNear(Query *pQuery, Snippet *pSnippet){ - int ii; - int iDir = 1; +static int trimSnippetOffsets( + Fts3Expr *pExpr, + Snippet *pSnippet, + int *piLeft +){ + if( pExpr ){ + if( trimSnippetOffsets(pExpr->pLeft, pSnippet, piLeft) ){ + return 1; + } - while(iDir>-2) { - assert( iDir==1 || iDir==-1 ); - for(ii=0; iinMatch; ii++){ - int jj; - int nNear; - struct snippetMatch *pMatch = &pSnippet->aMatch[ii]; - QueryTerm *pQueryTerm = &pQuery->pTerms[pMatch->iTerm]; + switch( pExpr->eType ){ + case FTSQUERY_PHRASE: + *piLeft += pExpr->pPhrase->nToken; + break; + case FTSQUERY_NEAR: { + /* The right-hand-side of a NEAR operator is always a phrase. The + ** left-hand-side is either a phrase or an expression tree that is + ** itself headed by a NEAR operator. The following initializations + ** set local variable iLeft to the token number of the left-most + ** token in the right-hand phrase, and iRight to the right most + ** token in the same phrase. For example, if we had: + ** + ** MATCH '"abc def" NEAR/2 "ghi jkl"' + ** + ** then iLeft will be set to 2 (token number of ghi) and nToken will + ** be set to 4. + */ + Fts3Expr *pLeft = pExpr->pLeft; + Fts3Expr *pRight = pExpr->pRight; + int iLeft = *piLeft; + int nNear = pExpr->nNear; + int nToken = pRight->pPhrase->nToken; + int jj, ii; + if( pLeft->eType==FTSQUERY_NEAR ){ + pLeft = pLeft->pRight; + } + assert( pRight->eType==FTSQUERY_PHRASE ); + assert( pLeft->eType==FTSQUERY_PHRASE ); + nToken += pLeft->pPhrase->nToken; - if( (pMatch->iTerm+iDir)<0 - || (pMatch->iTerm+iDir)>=pQuery->nTerms - ){ - continue; - } - - nNear = pQueryTerm->nNear; - if( iDir<0 ){ - nNear = pQueryTerm[-1].nNear; - } - - if( pMatch->iTerm>=0 && nNear ){ - int isOk = 0; - int iNextTerm = pMatch->iTerm+iDir; - int iPrevTerm = iNextTerm; - - int iEndToken; - int iStartToken; - - if( iDir<0 ){ - int nPhrase = 1; - iStartToken = pMatch->iToken; - while( (pMatch->iTerm+nPhrase)nTerms - && pQuery->pTerms[pMatch->iTerm+nPhrase].iPhrase>1 - ){ - nPhrase++; + for(ii=0; iinMatch; ii++){ + struct snippetMatch *p = &pSnippet->aMatch[ii]; + if( p->iTerm==iLeft ){ + int isOk = 0; + /* Snippet ii is an occurence of query term iLeft in the document. + ** It occurs at position (p->iToken) of the document. We now + ** search for an instance of token (iLeft-1) somewhere in the + ** range (p->iToken - nNear)...(p->iToken + nNear + nToken) within + ** the set of snippetMatch structures. If one is found, proceed. + ** If one cannot be found, then remove snippets ii..(ii+N-1) + ** from the matching snippets, where N is the number of tokens + ** in phrase pRight->pPhrase. + */ + for(jj=0; isOk==0 && jjnMatch; jj++){ + struct snippetMatch *p2 = &pSnippet->aMatch[jj]; + if( p2->iTerm==(iLeft-1) ){ + if( p2->iToken>=(p->iToken-nNear-1) + && p2->iToken<(p->iToken+nNear+nToken) + ){ + isOk = 1; + } + } + } + if( !isOk ){ + int kk; + for(kk=0; kkpPhrase->nToken; kk++){ + pSnippet->aMatch[kk+ii].iTerm = -2; + } + return 1; + } } - iEndToken = iStartToken + nPhrase - 1; - }else{ - iEndToken = pMatch->iToken; - iStartToken = pMatch->iToken+1-pQueryTerm->iPhrase; - } - - while( pQuery->pTerms[iNextTerm].iPhrase>1 ){ - iNextTerm--; - } - while( (iPrevTerm+1)nTerms && - pQuery->pTerms[iPrevTerm+1].iPhrase>1 - ){ - iPrevTerm++; - } - - for(jj=0; isOk==0 && jjnMatch; jj++){ - struct snippetMatch *p = &pSnippet->aMatch[jj]; - if( p->iCol==pMatch->iCol && (( - p->iTerm==iNextTerm && - p->iToken>iEndToken && - p->iToken<=iEndToken+nNear - ) || ( - p->iTerm==iPrevTerm && - p->iTokeniToken>=iStartToken-nNear - ))){ - isOk = 1; + if( p->iTerm==(iLeft-1) ){ + int isOk = 0; + for(jj=0; isOk==0 && jjnMatch; jj++){ + struct snippetMatch *p2 = &pSnippet->aMatch[jj]; + if( p2->iTerm==iLeft ){ + if( p2->iToken<=(p->iToken+nNear+1) + && p2->iToken>(p->iToken-nNear-nToken) + ){ + isOk = 1; + } + } + } + if( !isOk ){ + int kk; + for(kk=0; kkpPhrase->nToken; kk++){ + pSnippet->aMatch[ii-kk].iTerm = -2; + } + return 1; + } } } - if( !isOk ){ - for(jj=1-pQueryTerm->iPhrase; jj<=0; jj++){ - pMatch[jj].iTerm = -1; - } - ii = -1; - iDir = 1; - } + break; } } - iDir -= 2; + + if( trimSnippetOffsets(pExpr->pRight, pSnippet, piLeft) ){ + return 1; + } } + return 0; } /* @@ -3414,11 +3376,12 @@ static void snippetAllOffsets(fulltext_cursor *p){ int nColumn; int iColumn, i; int iFirst, iLast; - fulltext_vtab *pFts; + int iTerm = 0; + fulltext_vtab *pFts = cursor_vtab(p); - if( p->snippet.nMatch ) return; - if( p->q.nTerms==0 ) return; - pFts = p->q.pFts; + if( p->snippet.nMatch || p->pExpr==0 ){ + return; + } nColumn = pFts->nColumn; iColumn = (p->iCursorType - QUERY_FULLTEXT); if( iColumn<0 || iColumn>=nColumn ){ @@ -3433,15 +3396,18 @@ static void snippetAllOffsets(fulltext_cursor *p){ int nDoc; zDoc = (const char*)sqlite3_column_text(p->pStmt, i+1); nDoc = sqlite3_column_bytes(p->pStmt, i+1); - snippetOffsetsOfColumn(&p->q, &p->snippet, i, zDoc, nDoc); + snippetOffsetsOfColumn(p, &p->snippet, i, zDoc, nDoc); } - trimSnippetOffsetsForNear(&p->q, &p->snippet); + while( trimSnippetOffsets(p->pExpr, &p->snippet, &iTerm) ){ + iTerm = 0; + } } /* ** Convert the information in the aMatch[] array of the snippet -** into the string zOffset[0..nOffset-1]. +** into the string zOffset[0..nOffset-1]. This string is used as +** the return of the SQL offsets() function. */ static void snippetOffsetText(Snippet *p){ int i; @@ -3556,7 +3522,7 @@ static void snippetText( aMatch[i].snStatus = SNIPPET_IGNORE; } nDesired = 0; - for(i=0; iq.nTerms; i++){ + for(i=0; ipStmt); - queryClear(&c->q); + sqlite3Fts3ExprFree(c->pExpr); snippetClear(&c->snippet); - if( c->result.nData!=0 ) dlrDestroy(&c->reader); + if( c->result.nData!=0 ){ + dlrDestroy(&c->reader); + } dataBufferDestroy(&c->result); sqlite3_free(c); return SQLITE_OK; @@ -3703,255 +3671,127 @@ static int termSelect(fulltext_vtab *v, int iColumn, const char *pTerm, int nTerm, int isPrefix, DocListType iType, DataBuffer *out); -/* Return a DocList corresponding to the query term *pTerm. If *pTerm -** is the first term of a phrase query, go ahead and evaluate the phrase -** query and return the doclist for the entire phrase query. +/* +** Return a DocList corresponding to the phrase *pPhrase. ** ** The resulting DL_DOCIDS doclist is stored in pResult, which is ** overwritten. */ -static int docListOfTerm( - fulltext_vtab *v, /* The full text index */ - int iColumn, /* column to restrict to. No restriction if >=nColumn */ - QueryTerm *pQTerm, /* Term we are looking for, or 1st term of a phrase */ - DataBuffer *pResult /* Write the result here */ +static int docListOfPhrase( + fulltext_vtab *pTab, /* The full text index */ + Fts3Phrase *pPhrase, /* Phrase to return a doclist corresponding to */ + DocListType eListType, /* Either DL_DOCIDS or DL_POSITIONS */ + DataBuffer *pResult /* Write the result here */ ){ - DataBuffer left, right, new; - int i, rc; - - /* No phrase search if no position info. */ - assert( pQTerm->nPhrase==0 || DL_DEFAULT!=DL_DOCIDS ); + int ii; + int rc = SQLITE_OK; + int iCol = pPhrase->iColumn; + DocListType eType = eListType; + assert( eType==DL_POSITIONS || eType==DL_DOCIDS ); + if( pPhrase->nToken>1 ){ + eType = DL_POSITIONS; + } /* This code should never be called with buffered updates. */ - assert( v->nPendingData<0 ); + assert( pTab->nPendingData<0 ); - dataBufferInit(&left, 0); - rc = termSelect(v, iColumn, pQTerm->pTerm, pQTerm->nTerm, pQTerm->isPrefix, - (0nPhrase ? DL_POSITIONS : DL_DOCIDS), &left); - if( rc ) return rc; - for(i=1; i<=pQTerm->nPhrase && left.nData>0; i++){ - /* If this token is connected to the next by a NEAR operator, and - ** the next token is the start of a phrase, then set nPhraseRight - ** to the number of tokens in the phrase. Otherwise leave it at 1. - */ - int nPhraseRight = 1; - while( (i+nPhraseRight)<=pQTerm->nPhrase - && pQTerm[i+nPhraseRight].nNear==0 - ){ - nPhraseRight++; - } - - dataBufferInit(&right, 0); - rc = termSelect(v, iColumn, pQTerm[i].pTerm, pQTerm[i].nTerm, - pQTerm[i].isPrefix, DL_POSITIONS, &right); - if( rc ){ - dataBufferDestroy(&left); - return rc; - } - dataBufferInit(&new, 0); - docListPhraseMerge(left.pData, left.nData, right.pData, right.nData, - pQTerm[i-1].nNear, pQTerm[i-1].iPhrase + nPhraseRight, - ((inPhrase) ? DL_POSITIONS : DL_DOCIDS), - &new); - dataBufferDestroy(&left); - dataBufferDestroy(&right); - left = new; - } - *pResult = left; - return SQLITE_OK; -} - -/* Add a new term pTerm[0..nTerm-1] to the query *q. -*/ -static void queryAdd(Query *q, const char *pTerm, int nTerm){ - QueryTerm *t; - ++q->nTerms; - q->pTerms = sqlite3_realloc(q->pTerms, q->nTerms * sizeof(q->pTerms[0])); - if( q->pTerms==0 ){ - q->nTerms = 0; - return; - } - t = &q->pTerms[q->nTerms - 1]; - CLEAR(t); - t->pTerm = sqlite3_malloc(nTerm+1); - memcpy(t->pTerm, pTerm, nTerm); - t->pTerm[nTerm] = 0; - t->nTerm = nTerm; - t->isOr = q->nextIsOr; - t->isPrefix = 0; - q->nextIsOr = 0; - t->iColumn = q->nextColumn; - q->nextColumn = q->dfltColumn; -} - -/* -** Check to see if the string zToken[0...nToken-1] matches any -** column name in the virtual table. If it does, -** return the zero-indexed column number. If not, return -1. -*/ -static int checkColumnSpecifier( - fulltext_vtab *pVtab, /* The virtual table */ - const char *zToken, /* Text of the token */ - int nToken /* Number of characters in the token */ -){ - int i; - for(i=0; inColumn; i++){ - if( memcmp(pVtab->azColumn[i], zToken, nToken)==0 - && pVtab->azColumn[i][nToken]==0 ){ - return i; - } - } - return -1; -} - -/* -** Parse the text at zSegment[0..nSegment-1]. Add additional terms -** to the query being assemblied in pQuery. -** -** inPhrase is true if zSegment[0..nSegement-1] is contained within -** double-quotes. If inPhrase is true, then the first term -** is marked with the number of terms in the phrase less one and -** OR and "-" syntax is ignored. If inPhrase is false, then every -** term found is marked with nPhrase=0 and OR and "-" syntax is significant. -*/ -static int tokenizeSegment( - sqlite3_tokenizer *pTokenizer, /* The tokenizer to use */ - const char *zSegment, int nSegment, /* Query expression being parsed */ - int inPhrase, /* True if within "..." */ - Query *pQuery /* Append results here */ -){ - const sqlite3_tokenizer_module *pModule = pTokenizer->pModule; - sqlite3_tokenizer_cursor *pCursor; - int firstIndex = pQuery->nTerms; - int iCol; - int nTerm = 1; - - int rc = pModule->xOpen(pTokenizer, zSegment, nSegment, &pCursor); - if( rc!=SQLITE_OK ) return rc; - pCursor->pTokenizer = pTokenizer; - - while( 1 ){ - const char *zToken; - int nToken, iBegin, iEnd, iPos; - - rc = pModule->xNext(pCursor, - &zToken, &nToken, - &iBegin, &iEnd, &iPos); - if( rc!=SQLITE_OK ) break; - if( !inPhrase && - zSegment[iEnd]==':' && - (iCol = checkColumnSpecifier(pQuery->pFts, zToken, nToken))>=0 ){ - pQuery->nextColumn = iCol; - continue; - } - if( !inPhrase && pQuery->nTerms>0 && nToken==2 - && zSegment[iBegin+0]=='O' - && zSegment[iBegin+1]=='R' - ){ - pQuery->nextIsOr = 1; - continue; - } - if( !inPhrase && pQuery->nTerms>0 && !pQuery->nextIsOr && nToken==4 - && memcmp(&zSegment[iBegin], "NEAR", 4)==0 - ){ - QueryTerm *pTerm = &pQuery->pTerms[pQuery->nTerms-1]; - if( (iBegin+6)nNear = 0; - for(k=5; (iBegin+k)<=nSegment && isdigit(zSegment[iBegin+k]); k++){ - pTerm->nNear = pTerm->nNear*10 + (zSegment[iBegin+k] - '0'); + for(ii=0; rc==SQLITE_OK && iinToken; ii++){ + DataBuffer tmp; + struct PhraseToken *p = &pPhrase->aToken[ii]; + rc = termSelect(pTab, iCol, p->z, p->n, p->isPrefix, eType, &tmp); + if( rc==SQLITE_OK ){ + if( ii==0 ){ + *pResult = tmp; + }else{ + DataBuffer res = *pResult; + dataBufferInit(pResult, 0); + if( ii==(pPhrase->nToken-1) ){ + eType = eListType; } - pModule->xNext(pCursor, &zToken, &nToken, &iBegin, &iEnd, &iPos); - } else { - pTerm->nNear = SQLITE_FTS3_DEFAULT_NEAR_PARAM; + docListPhraseMerge( + res.pData, res.nData, tmp.pData, tmp.nData, 0, 0, eType, pResult + ); + dataBufferDestroy(&res); + dataBufferDestroy(&tmp); } - pTerm->nNear++; - continue; - } - - queryAdd(pQuery, zToken, nToken); - if( !inPhrase && iBegin>0 && zSegment[iBegin-1]=='-' ){ - pQuery->pTerms[pQuery->nTerms-1].isNot = 1; - } - if( iEndpTerms[pQuery->nTerms-1].isPrefix = 1; - } - pQuery->pTerms[pQuery->nTerms-1].iPhrase = nTerm; - if( inPhrase ){ - nTerm++; } } - if( inPhrase && pQuery->nTerms>firstIndex ){ - pQuery->pTerms[firstIndex].nPhrase = pQuery->nTerms - firstIndex - 1; - } - - return pModule->xClose(pCursor); + return rc; } -/* Parse a query string, yielding a Query object pQuery. -** -** The calling function will need to queryClear() to clean up -** the dynamically allocated memory held by pQuery. +/* +** Evaluate the full-text expression pExpr against fts3 table pTab. Write +** the results into pRes. */ -static int parseQuery( - fulltext_vtab *v, /* The fulltext index */ - const char *zInput, /* Input text of the query string */ - int nInput, /* Size of the input text */ - int dfltColumn, /* Default column of the index to match against */ - Query *pQuery /* Write the parse results here. */ +static int evalFts3Expr( + fulltext_vtab *pTab, /* Fts3 Virtual table object */ + Fts3Expr *pExpr, /* Parsed fts3 expression */ + DataBuffer *pRes /* OUT: Write results of the expression here */ ){ - int iInput, inPhrase = 0; - int ii; - QueryTerm *aTerm; + int rc = SQLITE_OK; - if( zInput==0 ) nInput = 0; - if( nInput<0 ) nInput = strlen(zInput); - pQuery->nTerms = 0; - pQuery->pTerms = NULL; - pQuery->nextIsOr = 0; - pQuery->nextColumn = dfltColumn; - pQuery->dfltColumn = dfltColumn; - pQuery->pFts = v; - - for(iInput=0; iInputiInput ){ - tokenizeSegment(v->pTokenizer, zInput+iInput, i-iInput, inPhrase, - pQuery); - } - iInput = i; - if( ipTerms; - for(ii=0; iinTerms; ii++){ - if( aTerm[ii].nNear || aTerm[ii].nPhrase ){ - while (aTerm[ii+aTerm[ii].nPhrase].nNear) { - aTerm[ii].nPhrase += (1 + aTerm[ii+aTerm[ii].nPhrase+1].nPhrase); + dataBufferInit(pRes, 0); + + if( pExpr ){ + if( pExpr->eType==FTSQUERY_PHRASE ){ + DocListType eType = DL_DOCIDS; + if( pExpr->pParent && pExpr->pParent->eType==FTSQUERY_NEAR ){ + eType = DL_POSITIONS; } + rc = docListOfPhrase(pTab, pExpr->pPhrase, eType, pRes); + }else{ + DataBuffer lhs; + DataBuffer rhs; + + dataBufferInit(&rhs, 0); + if( SQLITE_OK==(rc = evalFts3Expr(pTab, pExpr->pLeft, &lhs)) + && SQLITE_OK==(rc = evalFts3Expr(pTab, pExpr->pRight, &rhs)) + ){ + switch( pExpr->eType ){ + case FTSQUERY_NEAR: { + int nToken; + Fts3Expr *pLeft; + DocListType eType = DL_DOCIDS; + if( pExpr->pParent && pExpr->pParent->eType==FTSQUERY_NEAR ){ + eType = DL_POSITIONS; + } + pLeft = pExpr->pLeft; + while( pLeft->eType==FTSQUERY_NEAR ){ + pLeft=pLeft->pRight; + } + assert( pExpr->pRight->eType==FTSQUERY_PHRASE ); + assert( pLeft->eType==FTSQUERY_PHRASE ); + nToken = pLeft->pPhrase->nToken + pExpr->pRight->pPhrase->nToken; + docListPhraseMerge(lhs.pData, lhs.nData, rhs.pData, rhs.nData, + pExpr->nNear+1, nToken, eType, pRes + ); + break; + } + case FTSQUERY_NOT: { + docListExceptMerge(lhs.pData, lhs.nData, rhs.pData, rhs.nData,pRes); + break; + } + case FTSQUERY_AND: { + docListAndMerge(lhs.pData, lhs.nData, rhs.pData, rhs.nData, pRes); + break; + } + case FTSQUERY_OR: { + docListOrMerge(lhs.pData, lhs.nData, rhs.pData, rhs.nData, pRes); + break; + } + } + } + dataBufferDestroy(&lhs); + dataBufferDestroy(&rhs); } } - return SQLITE_OK; + return rc; } /* TODO(shess) Refactor the code to remove this forward decl. */ @@ -3970,12 +3810,9 @@ static int fulltextQuery( const char *zInput, /* The query string */ int nInput, /* Number of bytes in zInput[] */ DataBuffer *pResult, /* Write the result doclist here */ - Query *pQuery /* Put parsed query string here */ + Fts3Expr **ppExpr /* Put parsed query string here */ ){ - int i, iNext, rc; - DataBuffer left, right, or, new; - int nNot = 0; - QueryTerm *aTerm; + int rc; /* TODO(shess) Instead of flushing pendingTerms, we could query for ** the relevant term and merge the doclist into what we receive from @@ -3987,86 +3824,20 @@ static int fulltextQuery( /* Flush any buffered updates before executing the query. */ rc = flushPendingTerms(v); - if( rc!=SQLITE_OK ) return rc; - - /* TODO(shess) I think that the queryClear() calls below are not - ** necessary, because fulltextClose() already clears the query. - */ - rc = parseQuery(v, zInput, nInput, iColumn, pQuery); - if( rc!=SQLITE_OK ) return rc; - - /* Empty or NULL queries return no results. */ - if( pQuery->nTerms==0 ){ - dataBufferInit(pResult, 0); - return SQLITE_OK; + if( rc!=SQLITE_OK ){ + return rc; } - /* Merge AND terms. */ - /* TODO(shess) I think we can early-exit if( i>nNot && left.nData==0 ). */ - aTerm = pQuery->pTerms; - for(i = 0; inTerms; i=iNext){ - if( aTerm[i].isNot ){ - /* Handle all NOT terms in a separate pass */ - nNot++; - iNext = i + aTerm[i].nPhrase+1; - continue; - } - iNext = i + aTerm[i].nPhrase + 1; - rc = docListOfTerm(v, aTerm[i].iColumn, &aTerm[i], &right); - if( rc ){ - if( i!=nNot ) dataBufferDestroy(&left); - queryClear(pQuery); - return rc; - } - while( iNextnTerms && aTerm[iNext].isOr ){ - rc = docListOfTerm(v, aTerm[iNext].iColumn, &aTerm[iNext], &or); - iNext += aTerm[iNext].nPhrase + 1; - if( rc ){ - if( i!=nNot ) dataBufferDestroy(&left); - dataBufferDestroy(&right); - queryClear(pQuery); - return rc; - } - dataBufferInit(&new, 0); - docListOrMerge(right.pData, right.nData, or.pData, or.nData, &new); - dataBufferDestroy(&right); - dataBufferDestroy(&or); - right = new; - } - if( i==nNot ){ /* first term processed. */ - left = right; - }else{ - dataBufferInit(&new, 0); - docListAndMerge(left.pData, left.nData, right.pData, right.nData, &new); - dataBufferDestroy(&right); - dataBufferDestroy(&left); - left = new; - } + /* Parse the query passed to the MATCH operator. */ + rc = sqlite3Fts3ExprParse(v->pTokenizer, + v->azColumn, v->nColumn, iColumn, zInput, nInput, ppExpr + ); + if( rc!=SQLITE_OK ){ + assert( 0==(*ppExpr) ); + return rc; } - if( nNot==pQuery->nTerms ){ - /* We do not yet know how to handle a query of only NOT terms */ - return SQLITE_ERROR; - } - - /* Do the EXCEPT terms */ - for(i=0; inTerms; i += aTerm[i].nPhrase + 1){ - if( !aTerm[i].isNot ) continue; - rc = docListOfTerm(v, aTerm[i].iColumn, &aTerm[i], &right); - if( rc ){ - queryClear(pQuery); - dataBufferDestroy(&left); - return rc; - } - dataBufferInit(&new, 0); - docListExceptMerge(left.pData, left.nData, right.pData, right.nData, &new); - dataBufferDestroy(&right); - dataBufferDestroy(&left); - left = new; - } - - *pResult = left; - return rc; + return evalFts3Expr(v, *ppExpr, pResult); } /* @@ -4146,10 +3917,10 @@ static int fulltextFilter( default: /* full-text search */ { + int iCol = idxNum-QUERY_FULLTEXT; const char *zQuery = (const char *)sqlite3_value_text(argv[0]); assert( idxNum<=QUERY_FULLTEXT+v->nColumn); assert( argc==1 ); - queryClear(&c->q); if( c->result.nData!=0 ){ /* This case happens if the same cursor is used repeatedly. */ dlrDestroy(&c->reader); @@ -4157,7 +3928,7 @@ static int fulltextFilter( }else{ dataBufferInit(&c->result, 0); } - rc = fulltextQuery(v, idxNum-QUERY_FULLTEXT, zQuery, -1, &c->result, &c->q); + rc = fulltextQuery(v, iCol, zQuery, -1, &c->result, &c->pExpr); if( rc!=SQLITE_OK ) return rc; if( c->result.nData!=0 ){ dlrInit(&c->reader, DL_DOCIDS, c->result.pData, c->result.nData); @@ -6041,9 +5812,14 @@ static int loadSegment(fulltext_vtab *v, const char *pData, int nData, /* Scan the database and merge together the posting lists for the term ** into *out. */ -static int termSelect(fulltext_vtab *v, int iColumn, - const char *pTerm, int nTerm, int isPrefix, - DocListType iType, DataBuffer *out){ +static int termSelect( + fulltext_vtab *v, + int iColumn, + const char *pTerm, int nTerm, /* Term to query for */ + int isPrefix, /* True for a prefix search */ + DocListType iType, + DataBuffer *out /* Write results here */ +){ DataBuffer doclist; sqlite3_stmt *s; int rc = sql_get_statement(v, SEGDIR_SELECT_ALL_STMT, &s); @@ -6053,6 +5829,7 @@ static int termSelect(fulltext_vtab *v, int iColumn, assert( v->nPendingData<0 ); dataBufferInit(&doclist, 0); + dataBufferInit(out, 0); /* Traverse the segments from oldest to newest so that newer doclist ** elements for given docids overwrite older elements. @@ -7170,6 +6947,10 @@ int sqlite3Fts3Init(sqlite3 *db){ } } +#ifdef SQLITE_TEST + sqlite3Fts3ExprInitTestInterface(db); +#endif + /* Create the virtual table wrapper around the hash-table and overload ** the two scalar functions. If this is successful, register the ** module with sqlite. diff --git a/ext/fts3/fts3_expr.c b/ext/fts3/fts3_expr.c new file mode 100644 index 0000000000..546dcba9fa --- /dev/null +++ b/ext/fts3/fts3_expr.c @@ -0,0 +1,864 @@ +/* +** 2008 Nov 28 +** +** The author disclaims copyright to this source code. In place of +** a legal notice, here is a blessing: +** +** May you do good and not evil. +** May you find forgiveness for yourself and forgive others. +** May you share freely, never taking more than you give. +** +****************************************************************************** +** +** This module contains code that implements a parser for fts3 query strings +** (the right-hand argument to the MATCH operator). Because the supported +** syntax is relatively simple, the whole tokenizer/parser system is +** hand-coded. The public interface to this module is declared in source +** code file "fts3_expr.h". +*/ + +/* +** By default, this module parses the legacy syntax that has been +** traditionally used by fts3. Or, if SQLITE_ENABLE_FTS3_PARENTHESIS +** is defined, then it uses the new syntax. The differences between +** the new and the old syntaxes are: +** +** a) The new syntax supports parenthesis. The old does not. +** +** b) The new syntax supports the AND and NOT operators. The old does not. +** +** c) The old syntax supports the "-" token qualifier. This is not +** supported by the new syntax (it is replaced by the NOT operator). +** +** d) When using the old syntax, the OR operator has a greater precedence +** than an implicit AND. When using the new, both implicity and explicit +** AND operators have a higher precedence than OR. +** +** If compiled with SQLITE_TEST defined, then this module exports the +** symbol "int sqlite3_fts3_enable_parentheses". Setting this variable +** to zero causes the module to use the old syntax. If it is set to +** non-zero the new syntax is activated. This is so both syntaxes can +** be tested using a single build of testfixture. +*/ +#ifdef SQLITE_TEST +int sqlite3_fts3_enable_parentheses = 0; +#else +# ifdef SQLITE_ENABLE_FTS3_PARENTHESIS +# define sqlite3_fts3_enable_parentheses 1 +# else +# define sqlite3_fts3_enable_parentheses 0 +# endif +#endif + +/* +** Default span for NEAR operators. +*/ +#define SQLITE_FTS3_DEFAULT_NEAR_PARAM 10 + +#include "fts3_expr.h" +#include "sqlite3.h" +#include +#include +#include + +typedef struct ParseContext ParseContext; +struct ParseContext { + sqlite3_tokenizer *pTokenizer; /* Tokenizer module */ + const char **azCol; /* Array of column names for fts3 table */ + int nCol; /* Number of entries in azCol[] */ + int iDefaultCol; /* Default column to query */ + sqlite3_context *pCtx; /* Write error message here */ + int nNest; /* Number of nested brackets */ +}; + +/* +** This function is equivalent to the standard isspace() function. +** +** The standard isspace() can be awkward to use safely, because although it +** is defined to accept an argument of type int, its behaviour when passed +** an integer that falls outside of the range of the unsigned char type +** is undefined (and sometimes, "undefined" means segfault). This wrapper +** is defined to accept an argument of type char, and always returns 0 for +** any values that fall outside of the range of the unsigned char type (i.e. +** negative values). +*/ +static int safe_isspace(char c){ + return (c&0x80)==0 ? isspace(c) : 0; +} + +/* +** Extract the next token from buffer z (length n) using the tokenizer +** and other information (column names etc.) in pParse. Create an Fts3Expr +** structure of type FTSQUERY_PHRASE containing a phrase consisting of this +** single token and set *ppExpr to point to it. If the end of the buffer is +** reached before a token is found, set *ppExpr to zero. It is the +** responsibility of the caller to eventually deallocate the allocated +** Fts3Expr structure (if any) by passing it to sqlite3_free(). +** +** Return SQLITE_OK if successful, or SQLITE_NOMEM if a memory allocation +** fails. +*/ +static int getNextToken( + ParseContext *pParse, /* fts3 query parse context */ + int iCol, /* Value for Fts3Phrase.iColumn */ + const char *z, int n, /* Input string */ + Fts3Expr **ppExpr, /* OUT: expression */ + int *pnConsumed /* OUT: Number of bytes consumed */ +){ + sqlite3_tokenizer *pTokenizer = pParse->pTokenizer; + sqlite3_tokenizer_module const *pModule = pTokenizer->pModule; + int rc; + sqlite3_tokenizer_cursor *pCursor; + Fts3Expr *pRet = 0; + int nConsumed = 0; + + rc = pModule->xOpen(pTokenizer, z, n, &pCursor); + if( rc==SQLITE_OK ){ + const char *zToken; + int nToken, iStart, iEnd, iPosition; + int nByte; /* total space to allocate */ + + pCursor->pTokenizer = pTokenizer; + rc = pModule->xNext(pCursor, &zToken, &nToken, &iStart, &iEnd, &iPosition); + + if( rc==SQLITE_OK ){ + nByte = sizeof(Fts3Expr) + sizeof(Fts3Phrase) + nToken; + pRet = (Fts3Expr *)sqlite3_malloc(nByte); + if( !pRet ){ + rc = SQLITE_NOMEM; + }else{ + memset(pRet, 0, nByte); + pRet->eType = FTSQUERY_PHRASE; + pRet->pPhrase = (Fts3Phrase *)&pRet[1]; + pRet->pPhrase->nToken = 1; + pRet->pPhrase->iColumn = iCol; + pRet->pPhrase->aToken[0].n = nToken; + pRet->pPhrase->aToken[0].z = (char *)&pRet->pPhrase[1]; + memcpy(pRet->pPhrase->aToken[0].z, zToken, nToken); + + if( iEndpPhrase->aToken[0].isPrefix = 1; + iEnd++; + } + if( !sqlite3_fts3_enable_parentheses && iStart>0 && z[iStart-1]=='-' ){ + pRet->pPhrase->isNot = 1; + } + } + } + nConsumed = iEnd; + + pModule->xClose(pCursor); + } + + *pnConsumed = nConsumed; + *ppExpr = pRet; + return rc; +} + +void realloc_or_free(void **ppOrig, int nNew){ + void *pRet = sqlite3_realloc(*ppOrig, nNew); + if( !pRet ){ + sqlite3_free(*ppOrig); + } + *ppOrig = pRet; +} + +/* +** Buffer zInput, length nInput, contains the contents of a quoted string +** that appeared as part of an fts3 query expression. Neither quote character +** is included in the buffer. This function attempts to tokenize the entire +** input buffer and create an Fts3Expr structure of type FTSQUERY_PHRASE +** containing the results. +** +** If successful, SQLITE_OK is returned and *ppExpr set to point at the +** allocated Fts3Expr structure. Otherwise, either SQLITE_NOMEM (out of memory +** error) or SQLITE_ERROR (tokenization error) is returned and *ppExpr set +** to 0. +*/ +static int getNextString( + ParseContext *pParse, /* fts3 query parse context */ + const char *zInput, int nInput, /* Input string */ + Fts3Expr **ppExpr /* OUT: expression */ +){ + sqlite3_tokenizer *pTokenizer = pParse->pTokenizer; + sqlite3_tokenizer_module const *pModule = pTokenizer->pModule; + int rc; + Fts3Expr *p = 0; + sqlite3_tokenizer_cursor *pCursor = 0; + char *zTemp = 0; + int nTemp = 0; + + rc = pModule->xOpen(pTokenizer, zInput, nInput, &pCursor); + if( rc==SQLITE_OK ){ + int ii; + pCursor->pTokenizer = pTokenizer; + for(ii=0; rc==SQLITE_OK; ii++){ + const char *zToken; + int nToken, iBegin, iEnd, iPos; + rc = pModule->xNext(pCursor, &zToken, &nToken, &iBegin, &iEnd, &iPos); + if( rc==SQLITE_OK ){ + int nByte = sizeof(Fts3Expr) + sizeof(Fts3Phrase); + realloc_or_free((void **)&p, nByte+ii*sizeof(struct PhraseToken)); + realloc_or_free((void **)&zTemp, nTemp + nToken); + if( !p || !zTemp ){ + goto no_mem; + } + if( ii==0 ){ + memset(p, 0, nByte); + p->pPhrase = (Fts3Phrase *)&p[1]; + p->eType = FTSQUERY_PHRASE; + p->pPhrase->iColumn = pParse->iDefaultCol; + } + p->pPhrase = (Fts3Phrase *)&p[1]; + p->pPhrase->nToken = ii+1; + p->pPhrase->aToken[ii].n = nToken; + memcpy(&zTemp[nTemp], zToken, nToken); + nTemp += nToken; + if( iEndpPhrase->aToken[ii].isPrefix = 1; + }else{ + p->pPhrase->aToken[ii].isPrefix = 0; + } + } + } + + pModule->xClose(pCursor); + pCursor = 0; + } + + if( rc==SQLITE_DONE ){ + int jj; + char *zNew; + int nNew = 0; + int nByte = sizeof(Fts3Expr) + sizeof(Fts3Phrase); + nByte += (p->pPhrase->nToken-1) * sizeof(struct PhraseToken); + realloc_or_free((void **)&p, nByte + nTemp); + if( !p ){ + goto no_mem; + } + p->pPhrase = (Fts3Phrase *)&p[1]; + zNew = &(((char *)p)[nByte]); + memcpy(zNew, zTemp, nTemp); + for(jj=0; jjpPhrase->nToken; jj++){ + p->pPhrase->aToken[jj].z = &zNew[nNew]; + nNew += p->pPhrase->aToken[jj].n; + } + sqlite3_free(zTemp); + rc = SQLITE_OK; + } + + *ppExpr = p; + return rc; +no_mem: + + if( pCursor ){ + pModule->xClose(pCursor); + } + sqlite3_free(zTemp); + sqlite3_free(p); + *ppExpr = 0; + return SQLITE_NOMEM; +} + +/* +** Function getNextNode(), which is called by fts3ExprParse(), may itself +** call fts3ExprParse(). So this forward declaration is required. +*/ +static int fts3ExprParse(ParseContext *, const char *, int, Fts3Expr **, int *); + +/* +** The output variable *ppExpr is populated with an allocated Fts3Expr +** structure, or set to 0 if the end of the input buffer is reached. +** +** Returns an SQLite error code. SQLITE_OK if everything works, SQLITE_NOMEM +** if a malloc failure occurs, or SQLITE_ERROR if a parse error is encountered. +** If SQLITE_ERROR is returned, pContext is populated with an error message. +*/ +static int getNextNode( + ParseContext *pParse, /* fts3 query parse context */ + const char *z, int n, /* Input string */ + Fts3Expr **ppExpr, /* OUT: expression */ + int *pnConsumed /* OUT: Number of bytes consumed */ +){ + struct Fts3Keyword { + char *z; + int n; + int eType; + } aKeyword[] = { + { "OR" , 2, FTSQUERY_OR }, + { "AND", 3, FTSQUERY_AND }, + { "NOT", 3, FTSQUERY_NOT }, + { "NEAR", 4, FTSQUERY_NEAR } + }; + int ii; + int iCol; + int iColLen; + int rc; + Fts3Expr *pRet = 0; + + const char *zInput = z; + int nInput = n; + + /* Skip over any whitespace before checking for a keyword, an open or + ** close bracket, or a quoted string. + */ + while( nInput>0 && safe_isspace(*zInput) ){ + nInput--; + zInput++; + } + + /* See if we are dealing with a keyword. */ + for(ii=0; iieType==FTSQUERY_AND || pKey->eType==FTSQUERY_NOT) + ){ + continue; + } + + if( nInput>=pKey->n && 0==memcmp(zInput, pKey->z, pKey->n) ){ + int nNear = SQLITE_FTS3_DEFAULT_NEAR_PARAM; + int nKey = pKey->n; + char cNext; + + /* If this is a "NEAR" keyword, check for an explicit nearness. */ + if( pKey->eType==FTSQUERY_NEAR ){ + assert( nKey==4 ); + if( zInput[4]=='/' && zInput[5]>='0' && zInput[5]<='9' ){ + nNear = 0; + for(nKey=5; zInput[nKey]>='0' && zInput[nKey]<='9'; nKey++){ + nNear = nNear * 10 + (zInput[nKey] - '0'); + } + } + } + + /* At this point this is probably a keyword. But for that to be true, + ** the next byte must contain either whitespace, an open or close + ** bracket, a quote character, or EOF. + */ + cNext = zInput[nKey]; + if( safe_isspace(cNext) + || cNext=='"' || cNext=='(' || cNext==')' || cNext==0 + ){ + pRet = (Fts3Expr *)sqlite3_malloc(sizeof(Fts3Expr)); + memset(pRet, 0, sizeof(Fts3Expr)); + pRet->eType = pKey->eType; + pRet->nNear = nNear; + *ppExpr = pRet; + *pnConsumed = (zInput - z) + nKey; + return SQLITE_OK; + } + + /* Turns out that wasn't a keyword after all. This happens if the + ** user has supplied a token such as "ORacle". Continue. + */ + } + } + + /* Check for an open bracket. */ + if( sqlite3_fts3_enable_parentheses ){ + if( *zInput=='(' ){ + int nConsumed; + int rc; + pParse->nNest++; + rc = fts3ExprParse(pParse, &zInput[1], nInput-1, ppExpr, &nConsumed); + *pnConsumed = (zInput - z) + 1 + nConsumed; + return rc; + } + + /* Check for a close bracket. */ + if( *zInput==')' ){ + pParse->nNest--; + *pnConsumed = (zInput - z) + 1; + return SQLITE_DONE; + } + } + + /* See if we are dealing with a quoted phrase. If this is the case, then + ** search for the closing quote and pass the whole string to getNextString() + ** for processing. This is easy to do, as fts3 has no syntax for escaping + ** a quote character embedded in a string. + */ + if( *zInput=='"' ){ + for(ii=1; iiiDefaultCol; + iColLen = 0; + for(ii=0; iinCol; ii++){ + const char *zStr = pParse->azCol[ii]; + int nStr = strlen(zStr); + if( nInput>nStr && zInput[nStr]==':' && memcmp(zStr, zInput, nStr)==0 ){ + iCol = ii; + iColLen = ((zInput - z) + nStr + 1); + break; + } + } + rc = getNextToken(pParse, iCol, &z[iColLen], n-iColLen, ppExpr, pnConsumed); + *pnConsumed += iColLen; + return rc; +} + +/* +** The argument is an Fts3Expr structure for a binary operator (any type +** except an FTSQUERY_PHRASE). Return an integer value representing the +** precedence of the operator. Lower values have a higher precedence (i.e. +** group more tightly). For example, in the C language, the == operator +** groups more tightly than ||, and would therefore have a higher precedence. +** +** When using the new fts3 query syntax (when SQLITE_ENABLE_FTS3_PARENTHESIS +** is defined), the order of the operators in precedence from highest to +** lowest is: +** +** NEAR +** NOT +** AND (including implicit ANDs) +** OR +** +** Note that when using the old query syntax, the OR operator has a higher +** precedence than the AND operator. +*/ +static int opPrecedence(Fts3Expr *p){ + assert( p->eType!=FTSQUERY_PHRASE ); + if( sqlite3_fts3_enable_parentheses ){ + return p->eType; + }else if( p->eType==FTSQUERY_NEAR ){ + return 1; + }else if( p->eType==FTSQUERY_OR ){ + return 2; + } + assert( p->eType==FTSQUERY_AND ); + return 3; +} + +/* +** Argument ppHead contains a pointer to the current head of a query +** expression tree being parsed. pPrev is the expression node most recently +** inserted into the tree. This function adds pNew, which is always a binary +** operator node, into the expression tree based on the relative precedence +** of pNew and the existing nodes of the tree. This may result in the head +** of the tree changing, in which case *ppHead is set to the new root node. +*/ +static void insertBinaryOperator( + Fts3Expr **ppHead, /* Pointer to the root node of a tree */ + Fts3Expr *pPrev, /* Node most recently inserted into the tree */ + Fts3Expr *pNew /* New binary node to insert into expression tree */ +){ + Fts3Expr *pSplit = pPrev; + while( pSplit->pParent && opPrecedence(pSplit->pParent)<=opPrecedence(pNew) ){ + pSplit = pSplit->pParent; + } + + if( pSplit->pParent ){ + assert( pSplit->pParent->pRight==pSplit ); + pSplit->pParent->pRight = pNew; + pNew->pParent = pSplit->pParent; + }else{ + *ppHead = pNew; + } + pNew->pLeft = pSplit; + pSplit->pParent = pNew; +} + +/* +** Parse the fts3 query expression found in buffer z, length n. This function +** returns either when the end of the buffer is reached or an unmatched +** closing bracket - ')' - is encountered. +** +** If successful, SQLITE_OK is returned, *ppExpr is set to point to the +** parsed form of the expression and *pnConsumed is set to the number of +** bytes read from buffer z. Otherwise, *ppExpr is set to 0 and SQLITE_NOMEM +** (out of memory error) or SQLITE_ERROR (parse error) is returned. +*/ +static int fts3ExprParse( + ParseContext *pParse, /* fts3 query parse context */ + const char *z, int n, /* Text of MATCH query */ + Fts3Expr **ppExpr, /* OUT: Parsed query structure */ + int *pnConsumed /* OUT: Number of bytes consumed */ +){ + Fts3Expr *pRet = 0; + Fts3Expr *pPrev = 0; + Fts3Expr *pNotBranch = 0; /* Only used in legacy parse mode */ + int nIn = n; + const char *zIn = z; + int rc = SQLITE_OK; + int isRequirePhrase = 1; + + while( rc==SQLITE_OK ){ + Fts3Expr *p = 0; + int nByte; + rc = getNextNode(pParse, zIn, nIn, &p, &nByte); + if( rc==SQLITE_OK ){ + int isPhrase; + + if( !sqlite3_fts3_enable_parentheses + && p->eType==FTSQUERY_PHRASE && p->pPhrase->isNot + ){ + /* Create an implicit NOT operator. */ + Fts3Expr *pNot = sqlite3_malloc(sizeof(Fts3Expr)); + if( !pNot ){ + sqlite3Fts3ExprFree(p); + rc = SQLITE_NOMEM; + goto exprparse_out; + } + memset(pNot, 0, sizeof(Fts3Expr)); + pNot->eType = FTSQUERY_NOT; + pNot->pRight = p; + if( pNotBranch ){ + pNotBranch->pLeft = p; + pNot->pRight = pNotBranch; + } + pNotBranch = pNot; + }else{ + assert( p->eType!=FTSQUERY_PHRASE || !p->pPhrase->isNot ); + isPhrase = (p->eType==FTSQUERY_PHRASE || p->pLeft); + if( !isPhrase && isRequirePhrase ){ + sqlite3Fts3ExprFree(p); + rc = SQLITE_ERROR; + goto exprparse_out; + } + + if( isPhrase && !isRequirePhrase ){ + /* Insert an implicit AND operator. */ + Fts3Expr *pAnd; + assert( pRet && pPrev ); + pAnd = sqlite3_malloc(sizeof(Fts3Expr)); + if( !pAnd ){ + sqlite3Fts3ExprFree(p); + rc = SQLITE_NOMEM; + goto exprparse_out; + } + memset(pAnd, 0, sizeof(Fts3Expr)); + pAnd->eType = FTSQUERY_AND; + insertBinaryOperator(&pRet, pPrev, pAnd); + pPrev = pAnd; + } + + if( pPrev && ( + (pPrev->eType==FTSQUERY_NEAR && p->eType!=FTSQUERY_PHRASE) + || (p->eType==FTSQUERY_NEAR && pPrev->eType!=FTSQUERY_PHRASE) + )){ + /* This is an attempt to do "phrase NEAR (bracketed expression)" + ** or "(bracketed expression) NEAR phrase", both of which are + ** illegal. Return an error. + */ + sqlite3Fts3ExprFree(p); + rc = SQLITE_ERROR; + goto exprparse_out; + } + + if( isPhrase ){ + if( pRet ){ + assert( pPrev && pPrev->pLeft && pPrev->pRight==0 ); + pPrev->pRight = p; + p->pParent = pPrev; + }else{ + pRet = p; + } + }else{ + insertBinaryOperator(&pRet, pPrev, p); + } + isRequirePhrase = !isPhrase; + } + assert( nByte>0 ); + } + nIn -= nByte; + zIn += nByte; + pPrev = p; + } + + if( rc==SQLITE_DONE && pRet && isRequirePhrase ){ + rc = SQLITE_ERROR; + } + + if( rc==SQLITE_DONE ){ + rc = SQLITE_OK; + if( !sqlite3_fts3_enable_parentheses && pNotBranch ){ + if( !pRet ){ + rc = SQLITE_ERROR; + }else{ + pNotBranch->pLeft = pRet; + pRet = pNotBranch; + } + } + } + *pnConsumed = n - nIn; + +exprparse_out: + if( rc!=SQLITE_OK ){ + sqlite3Fts3ExprFree(pRet); + sqlite3Fts3ExprFree(pNotBranch); + pRet = 0; + } + *ppExpr = pRet; + return rc; +} + +/* +** Parameters z and n contain a pointer to and length of a buffer containing +** an fts3 query expression, respectively. This function attempts to parse the +** query expression and create a tree of Fts3Expr structures representing the +** parsed expression. If successful, *ppExpr is set to point to the head +** of the parsed expression tree and SQLITE_OK is returned. If an error +** occurs, either SQLITE_NOMEM (out-of-memory error) or SQLITE_ERROR (parse +** error) is returned and *ppExpr is set to 0. +** +** If parameter n is a negative number, then z is assumed to point to a +** nul-terminated string and the length is determined using strlen(). +** +** The first parameter, pTokenizer, is passed the fts3 tokenizer module to +** use to normalize query tokens while parsing the expression. The azCol[] +** array, which is assumed to contain nCol entries, should contain the names +** of each column in the target fts3 table, in order from left to right. +** Column names must be nul-terminated strings. +** +** The iDefaultCol parameter should be passed the index of the table column +** that appears on the left-hand-side of the MATCH operator (the default +** column to match against for tokens for which a column name is not explicitly +** specified as part of the query string), or -1 if tokens may by default +** match any table column. +*/ +int sqlite3Fts3ExprParse( + sqlite3_tokenizer *pTokenizer, /* Tokenizer module */ + char **azCol, /* Array of column names for fts3 table */ + int nCol, /* Number of entries in azCol[] */ + int iDefaultCol, /* Default column to query */ + const char *z, int n, /* Text of MATCH query */ + Fts3Expr **ppExpr /* OUT: Parsed query structure */ +){ + int nParsed; + int rc; + ParseContext sParse; + sParse.pTokenizer = pTokenizer; + sParse.azCol = (const char **)azCol; + sParse.nCol = nCol; + sParse.iDefaultCol = iDefaultCol; + sParse.nNest = 0; + if( z==0 ){ + *ppExpr = 0; + return SQLITE_OK; + } + if( n<0 ){ + n = strlen(z); + } + rc = fts3ExprParse(&sParse, z, n, ppExpr, &nParsed); + + /* Check for mismatched parenthesis */ + if( rc==SQLITE_OK && sParse.nNest ){ + rc = SQLITE_ERROR; + sqlite3Fts3ExprFree(*ppExpr); + *ppExpr = 0; + } + + return rc; +} + +/* +** Free a parsed fts3 query expression allocated by sqlite3Fts3ExprParse(). +*/ +void sqlite3Fts3ExprFree(Fts3Expr *p){ + if( p ){ + sqlite3Fts3ExprFree(p->pLeft); + sqlite3Fts3ExprFree(p->pRight); + sqlite3_free(p); + } +} + +/**************************************************************************** +***************************************************************************** +** Everything after this point is just test code. +*/ + +#ifdef SQLITE_TEST + +#include + +/* +** Function to query the hash-table of tokenizers (see README.tokenizers). +*/ +static int queryTokenizer( + sqlite3 *db, + const char *zName, + const sqlite3_tokenizer_module **pp +){ + int rc; + sqlite3_stmt *pStmt; + const char zSql[] = "SELECT fts3_tokenizer(?)"; + + *pp = 0; + rc = sqlite3_prepare_v2(db, zSql, -1, &pStmt, 0); + if( rc!=SQLITE_OK ){ + return rc; + } + + sqlite3_bind_text(pStmt, 1, zName, -1, SQLITE_STATIC); + if( SQLITE_ROW==sqlite3_step(pStmt) ){ + if( sqlite3_column_type(pStmt, 0)==SQLITE_BLOB ){ + memcpy(pp, sqlite3_column_blob(pStmt, 0), sizeof(*pp)); + } + } + + return sqlite3_finalize(pStmt); +} + +/* +** This function is part of the test interface for the query parser. It +** writes a text representation of the query expression pExpr into the +** buffer pointed to by argument zBuf. It is assumed that zBuf is large +** enough to store the required text representation. +*/ +static void exprToString(Fts3Expr *pExpr, char *zBuf){ + switch( pExpr->eType ){ + case FTSQUERY_PHRASE: { + Fts3Phrase *pPhrase = pExpr->pPhrase; + int i; + zBuf += sprintf(zBuf, "PHRASE %d %d", pPhrase->iColumn, pPhrase->isNot); + for(i=0; inToken; i++){ + zBuf += sprintf(zBuf," %.*s",pPhrase->aToken[i].n,pPhrase->aToken[i].z); + zBuf += sprintf(zBuf,"%s", (pPhrase->aToken[i].isPrefix?"+":"")); + } + return; + } + + case FTSQUERY_NEAR: + zBuf += sprintf(zBuf, "NEAR/%d ", pExpr->nNear); + break; + case FTSQUERY_NOT: + zBuf += sprintf(zBuf, "NOT "); + break; + case FTSQUERY_AND: + zBuf += sprintf(zBuf, "AND "); + break; + case FTSQUERY_OR: + zBuf += sprintf(zBuf, "OR "); + break; + } + + zBuf += sprintf(zBuf, "{"); + exprToString(pExpr->pLeft, zBuf); + zBuf += strlen(zBuf); + zBuf += sprintf(zBuf, "} "); + + zBuf += sprintf(zBuf, "{"); + exprToString(pExpr->pRight, zBuf); + zBuf += strlen(zBuf); + zBuf += sprintf(zBuf, "}"); +} + +/* +** This is the implementation of a scalar SQL function used to test the +** expression parser. It should be called as follows: +** +** fts3_exprtest(, , , ...); +** +** The first argument, , is the name of the fts3 tokenizer used +** to parse the query expression (see README.tokenizers). The second argument +** is the query expression to parse. Each subsequent argument is the name +** of a column of the fts3 table that the query expression may refer to. +** For example: +** +** SELECT fts3_exprtest('simple', 'Bill col2:Bloggs', 'col1', 'col2'); +*/ +static void fts3ExprTest( + sqlite3_context *context, + int argc, + sqlite3_value **argv +){ + sqlite3_tokenizer_module const *pModule = 0; + sqlite3_tokenizer *pTokenizer; + int rc; + char **azCol = 0; + const char *zExpr; + int nExpr; + int nCol; + int ii; + Fts3Expr *pExpr; + sqlite3 *db = sqlite3_context_db_handle(context); + + if( argc<3 ){ + sqlite3_result_error(context, + "Usage: fts3_exprtest(tokenizer, expr, col1, ...", -1 + ); + return; + } + + rc = queryTokenizer(db, (const char *)sqlite3_value_text(argv[0]), &pModule); + if( rc==SQLITE_NOMEM ){ + sqlite3_result_error_nomem(context); + goto exprtest_out; + }else if( !pModule ){ + sqlite3_result_error(context, "No such tokenizer module", -1); + goto exprtest_out; + } + + rc = pModule->xCreate(0, 0, &pTokenizer); + assert( rc==SQLITE_NOMEM || rc==SQLITE_OK ); + if( rc==SQLITE_NOMEM ){ + sqlite3_result_error_nomem(context); + goto exprtest_out; + } + pTokenizer->pModule = pModule; + + zExpr = (const char *)sqlite3_value_text(argv[1]); + nExpr = sqlite3_value_bytes(argv[1]); + nCol = argc-2; + azCol = (char **)sqlite3_malloc(nCol*sizeof(char *)); + if( !azCol ){ + sqlite3_result_error_nomem(context); + goto exprtest_out; + } + for(ii=0; iixDestroy(pTokenizer); + } + sqlite3_free(azCol); +} + +/* +** Register the query expression parser test function fts3_exprtest() +** with database connection db. +*/ +void sqlite3Fts3ExprInitTestInterface(sqlite3* db){ + sqlite3_create_function( + db, "fts3_exprtest", -1, SQLITE_UTF8, 0, fts3ExprTest, 0, 0 + ); +} + +#endif diff --git a/ext/fts3/fts3_expr.h b/ext/fts3/fts3_expr.h new file mode 100644 index 0000000000..df6015042f --- /dev/null +++ b/ext/fts3/fts3_expr.h @@ -0,0 +1,87 @@ +/* +** 2008 Nov 28 +** +** The author disclaims copyright to this source code. In place of +** a legal notice, here is a blessing: +** +** May you do good and not evil. +** May you find forgiveness for yourself and forgive others. +** May you share freely, never taking more than you give. +** +****************************************************************************** +** +*/ + +#include "fts3_tokenizer.h" +#include "sqlite3.h" + +/* +** The following describes the syntax supported by the fts3 MATCH +** operator in a similar format to that used by the lemon parser +** generator. This module does not use actually lemon, it uses a +** custom parser. +** +** phrase ::= TOKEN. +** phrase ::= TOKEN:COLUMN. +** phrase ::= "TOKEN TOKEN TOKEN...". +** phrase ::= phrase near phrase. +** +** near ::= NEAR. +** near ::= NEAR / INTEGER. +** +** query ::= -TOKEN. +** query ::= phrase. +** query ::= LP query RP. +** query ::= query NOT query. +** query ::= query OR query. +** query ::= query AND query. +*/ + +typedef struct Fts3Expr Fts3Expr; +typedef struct Fts3Phrase Fts3Phrase; + +struct Fts3Phrase { + int nToken; /* Number of entries in aToken[] */ + int iColumn; /* Index of column this phrase must match */ + int isNot; /* Phrase prefixed by unary not (-) operator */ + struct PhraseToken { + char *z; + int n; /* Number of bytes in buffer pointed to by z */ + int isPrefix; /* True if token ends in with a "*" character */ + } aToken[1]; +}; + +struct Fts3Expr { + int eType; /* One of the FTSQUERY_XXX values defined below */ + int nNear; /* Valid if eType==FTSQUERY_NEAR */ + Fts3Expr *pParent; + Fts3Expr *pLeft; + Fts3Expr *pRight; + Fts3Phrase *pPhrase; /* Valid if eType==FTSQUERY_PHRASE */ +}; + +int sqlite3Fts3ExprParse(sqlite3_tokenizer *, char **, int, int, + const char *, int, Fts3Expr **); +void sqlite3Fts3ExprFree(Fts3Expr *); + +/* +** Candidate values for Fts3Query.eType. Note that the order of the first +** four values is in order of precedence when parsing expressions. For +** example, the following: +** +** "a OR b AND c NOT d NEAR e" +** +** is equivalent to: +** +** "a OR (b AND (c NOT (d NEAR e)))" +*/ +#define FTSQUERY_NEAR 1 +#define FTSQUERY_NOT 2 +#define FTSQUERY_AND 3 +#define FTSQUERY_OR 4 +#define FTSQUERY_PHRASE 5 + +#ifdef SQLITE_TEST +void sqlite3Fts3ExprInitTestInterface(sqlite3 *db); +#endif + diff --git a/main.mk b/main.mk index d3c33517ac..b693a6839a 100644 --- a/main.mk +++ b/main.mk @@ -52,7 +52,7 @@ TCCX += -I$(TOP)/ext/rtree -I$(TOP)/ext/icu -I$(TOP)/ext/fts3 LIBOBJ+= alter.o analyze.o attach.o auth.o \ bitvec.o btmutex.o btree.o build.o \ callback.o complete.o date.o delete.o expr.o fault.o \ - fts3.o fts3_hash.o fts3_icu.o fts3_porter.o \ + fts3.o fts3_expr.o fts3_hash.o fts3_icu.o fts3_porter.o \ fts3_tokenizer.o fts3_tokenizer1.o \ func.o global.o hash.o \ icu.o insert.o journal.o legacy.o loadext.o \ @@ -178,6 +178,8 @@ SRC += \ SRC += \ $(TOP)/ext/fts3/fts3.c \ $(TOP)/ext/fts3/fts3.h \ + $(TOP)/ext/fts3/fts3_expr.c \ + $(TOP)/ext/fts3/fts3_expr.h \ $(TOP)/ext/fts3/fts3_hash.c \ $(TOP)/ext/fts3/fts3_hash.h \ $(TOP)/ext/fts3/fts3_icu.c \ @@ -247,7 +249,8 @@ TESTSRC2 = \ $(TOP)/src/pcache1.c $(TOP)/src/select.c $(TOP)/src/tokenize.c \ $(TOP)/src/utf.c $(TOP)/src/util.c $(TOP)/src/vdbeapi.c $(TOP)/src/vdbeaux.c \ $(TOP)/src/vdbe.c $(TOP)/src/vdbemem.c $(TOP)/src/where.c parse.c \ - $(TOP)/ext/fts3/fts3.c $(TOP)/ext/fts3/fts3_tokenizer.c + $(TOP)/ext/fts3/fts3.c $(TOP)/ext/fts3/fts3_expr.c \ + $(TOP)/ext/fts3/fts3_tokenizer.c # Header files used by all library source files. # @@ -283,6 +286,7 @@ EXTHDR += \ $(TOP)/ext/fts2/fts2_tokenizer.h EXTHDR += \ $(TOP)/ext/fts3/fts3.h \ + $(TOP)/ext/fts3/fts3_expr.h \ $(TOP)/ext/fts3/fts3_hash.h \ $(TOP)/ext/fts3/fts3_tokenizer.h EXTHDR += \ @@ -412,6 +416,9 @@ fts2_tokenizer1.o: $(TOP)/ext/fts2/fts2_tokenizer1.c $(HDR) $(EXTHDR) fts3.o: $(TOP)/ext/fts3/fts3.c $(HDR) $(EXTHDR) $(TCCX) -DSQLITE_CORE -c $(TOP)/ext/fts3/fts3.c +fts3_expr.o: $(TOP)/ext/fts3/fts3_expr.c $(HDR) $(EXTHDR) + $(TCCX) -DSQLITE_CORE -c $(TOP)/ext/fts3/fts3_expr.c + fts3_hash.o: $(TOP)/ext/fts3/fts3_hash.c $(HDR) $(EXTHDR) $(TCCX) -DSQLITE_CORE -c $(TOP)/ext/fts3/fts3_hash.c diff --git a/manifest b/manifest index ab531a9faa..770a260a3c 100644 --- a/manifest +++ b/manifest @@ -1,5 +1,5 @@ -C Version\s3.6.7\s(CVS\s6033) -D 2008-12-16T18:15:59 +C Modify\sfts3\sto\ssupport\sa\smore\scomplex\sexpression\ssyntax\sthat\sallows\sparenthesis.\sThe\snew\ssyntax\sis\snot\sentirely\sbackwards\scompatible,\sso\sis\sdisabled\sby\sdefault.\sUse\s-DSQLITE_ENABLE_FTS3_PARENTHESIS\sto\senable\sit.\s(CVS\s6034) +D 2008-12-17T15:18:18 F Makefile.arm-wince-mingw32ce-gcc fcd5e9cd67fe88836360bb4f9ef4cb7f8e2fb5a0 F Makefile.in f7e4c81c347b04f7b0f1c1b081a168645d7b8af7 F Makefile.linux-gcc d53183f4aa6a9192d249731c90dbdffbd2c68654 @@ -52,8 +52,10 @@ F ext/fts2/fts2_tokenizer1.c 8545ce12b41922004da46e91a7b023b92b76f94e F ext/fts2/mkfts2amal.tcl 974d5d438cb3f7c4a652639262f82418c1e4cff0 F ext/fts3/README.tokenizers 226644a0eab97724e8de83061912e8bb248461b6 F ext/fts3/README.txt 8c18f41574404623b76917b9da66fcb0ab38328d -F ext/fts3/fts3.c e67453b6ac421b79e600385491ed7f038b3bb271 +F ext/fts3/fts3.c 3aa6aef1eadc44606f6ed3c841062735a5210077 F ext/fts3/fts3.h 3a10a0af180d502cecc50df77b1b22df142817fe +F ext/fts3/fts3_expr.c b141145197cc749accb03d2b970813443b723edd +F ext/fts3/fts3_expr.h 4dad4d87cf5d41ea924a815fe89a6f87dc76f277 F ext/fts3/fts3_hash.c e15e84d18f8df149ab290029872d4559c4c7c15a F ext/fts3/fts3_hash.h 004b759e1602ff16dfa02fea3ca1c77336ad6798 F ext/fts3/fts3_icu.c ac494aed69835008185299315403044664bda295 @@ -80,7 +82,7 @@ F ext/rtree/tkt3363.test 6662237ea75bb431cd5d262dfc9535e1023315fc F ext/rtree/viewrtree.tcl 09526398dae87a5a87c5aac2b3854dbaf8376869 F install-sh 9d4de14ab9fb0facae2f48780b874848cbf2f895 F ltmain.sh 09fe5815427dc7d0abb188bbcdf0e34896577210 -F main.mk 5923e75b5ac4b265f322597c3953dda7175f4405 +F main.mk f6eb58a66f942bf672ab58e74e30e72cad39b93f F mkdll.sh 7d09b23c05d56532e9d44a50868eb4b12ff4f74a F mkextu.sh 416f9b7089d80e5590a29692c9d9280a10dbad9f F mkextw.sh 4123480947681d9b434a5e7b1ee08135abe409ac @@ -160,7 +162,7 @@ F src/sqliteLimit.h f435e728c6b620ef7312814d660a81f9356eb5c8 F src/status.c 237b193efae0cf6ac3f0817a208de6c6c6ef6d76 F src/table.c 23db1e5f27c03160987c122a078b4bb51ef0b2f8 F src/tclsqlite.c 23afb60549af943e135ded441a631f4745be6040 -F src/test1.c 9c0502c3627162f969099e57932782057d9139b6 +F src/test1.c b193b8b80617bdb8297b25a87d00ee8d5a125d0d F src/test2.c 897528183edf2839c2a3c991d415905db56f1240 F src/test3.c 88a246b56b824275300e6c899634fbac1dc94b14 F src/test4.c f79ab52d27ff49b784b631a42e2ccd52cfd5c84c @@ -364,6 +366,7 @@ F test/fts3b.test b3a25180a633873d37d86e1ccd00ed690d37237a F test/fts3c.test 4c7ef29b37aca3e8ebb6a39b57910caa6506034e F test/fts3d.test d92a47fe8ed59c9e53d2d8e6d2685bb380aadadc F test/fts3e.test 1f6c6ac9cc8b772ca256e6b22aaeed50c9350851 +F test/fts3expr.test 18143e61503845b940fd7caacce53bce4307426f F test/fts3near.test e8a9b4e16c63a795918b334b74d4aec14815bf8b F test/func.test a50f0a4b69ac251debe1dce3ba29da7476dc8c52 F test/fuzz.test 62fc19dd36a427777fd671b569df07166548628a @@ -675,7 +678,7 @@ F tool/speedtest16.c c8a9c793df96db7e4933f0852abb7a03d48f2e81 F tool/speedtest2.tcl ee2149167303ba8e95af97873c575c3e0fab58ff F tool/speedtest8.c 2902c46588c40b55661e471d7a86e4dd71a18224 F tool/speedtest8inst1.c 293327bc76823f473684d589a8160bde1f52c14e -P 8b8f6a6ab597e06e60557ab56c6ee7f8522ed570 -R a6ff965a91307b50874e830368f8fcd6 -U drh -Z 76064440713dce47dc110c16fc71382b +P f4f40370fb83d677df3fbf2c51c4bb4a3e5ccc7a +R 0170cc023fb1827148ff15e7bc02335a +U danielk1977 +Z 3401f60bb1566cbc5da4c344a91c4fb9 diff --git a/manifest.uuid b/manifest.uuid index a4df069db2..d2c9a8b4cf 100644 --- a/manifest.uuid +++ b/manifest.uuid @@ -1 +1 @@ -f4f40370fb83d677df3fbf2c51c4bb4a3e5ccc7a \ No newline at end of file +7389b9ecb80294569845c40a23e0c832d07f7a45 \ No newline at end of file diff --git a/src/test1.c b/src/test1.c index 37f5778932..d73dec9367 100644 --- a/src/test1.c +++ b/src/test1.c @@ -13,7 +13,7 @@ ** is not included in the SQLite library. It is used for automated ** testing of the SQLite library. ** -** $Id: test1.c,v 1.337 2008/12/11 02:56:07 drh Exp $ +** $Id: test1.c,v 1.338 2008/12/17 15:18:18 danielk1977 Exp $ */ #include "sqliteInt.h" #include "tcl.h" @@ -4930,6 +4930,9 @@ int Sqlitetest1_Init(Tcl_Interp *interp){ extern int sqlite3_enable_in_opt; extern char sqlite3_query_plan[]; static char *query_plan = sqlite3_query_plan; +#ifdef SQLITE_ENABLE_FTS3 + extern int sqlite3_fts3_enable_parentheses; +#endif #endif for(i=0; i