#include "fts5Int.h"
#include "fts5parse.h"
#ifndef SQLITE_FTS5_MAX_EXPR_DEPTH
# define SQLITE_FTS5_MAX_EXPR_DEPTH 256
#endif
#define FTS5_EOF 0
#define FTS5_LARGEST_INT64 (0xffffffff|(((i64)0x7fffffff)<<32))
typedef struct Fts5ExprTerm Fts5ExprTerm;
void *sqlite3Fts5ParserAlloc(void *(*mallocProc)(u64));
void sqlite3Fts5ParserFree(void*, void (*freeProc)(void*));
void sqlite3Fts5Parser(void*, int, Fts5Token, Fts5Parse*);
#ifndef NDEBUG
#include <stdio.h>
void sqlite3Fts5ParserTrace(FILE*, char*);
#endif
int sqlite3Fts5ParserFallback(int);
struct Fts5Expr {
Fts5Index *pIndex;
Fts5Config *pConfig;
Fts5ExprNode *pRoot;
int bDesc;
int nPhrase;
Fts5ExprPhrase **apExprPhrase;
};
struct Fts5ExprNode {
int eType;
int bEof;
int bNomatch;
int iHeight;
int (*xNext)(Fts5Expr*, Fts5ExprNode*, int, i64);
i64 iRowid;
Fts5ExprNearset *pNear;
int nChild;
Fts5ExprNode *apChild[1];
};
#define Fts5NodeIsString(p) ((p)->eType==FTS5_TERM || (p)->eType==FTS5_STRING)
#define fts5ExprNodeNext(a,b,c,d) (b)->xNext((a), (b), (c), (d))
struct Fts5ExprTerm {
u8 bPrefix;
u8 bFirst;
char *zTerm;
Fts5IndexIter *pIter;
Fts5ExprTerm *pSynonym;
};
struct Fts5ExprPhrase {
Fts5ExprNode *pNode;
Fts5Buffer poslist;
int nTerm;
Fts5ExprTerm aTerm[1];
};
struct Fts5ExprNearset {
int nNear;
Fts5Colset *pColset;
int nPhrase;
Fts5ExprPhrase *apPhrase[1];
};
struct Fts5Parse {
Fts5Config *pConfig;
char *zErr;
int rc;
int nPhrase;
Fts5ExprPhrase **apPhrase;
Fts5ExprNode *pExpr;
int bPhraseToAnd;
};
#ifndef NDEBUG
static void assert_expr_depth_ok(int rc, Fts5ExprNode *p){
if( rc==SQLITE_OK ){
if( p->eType==FTS5_TERM || p->eType==FTS5_STRING || p->eType==0 ){
assert( p->iHeight==0 );
}else{
int ii;
int iMaxChild = 0;
for(ii=0; ii<p->nChild; ii++){
Fts5ExprNode *pChild = p->apChild[ii];
iMaxChild = MAX(iMaxChild, pChild->iHeight);
assert_expr_depth_ok(SQLITE_OK, pChild);
}
assert( p->iHeight==iMaxChild+1 );
}
}
}
#else
# define assert_expr_depth_ok(rc, p)
#endif
void sqlite3Fts5ParseError(Fts5Parse *pParse, const char *zFmt, ...){
va_list ap;
va_start(ap, zFmt);
if( pParse->rc==SQLITE_OK ){
assert( pParse->zErr==0 );
pParse->zErr = sqlite3_vmprintf(zFmt, ap);
pParse->rc = SQLITE_ERROR;
}
va_end(ap);
}
static int fts5ExprIsspace(char t){
return t==' ' || t=='\t' || t=='\n' || t=='\r';
}
static int fts5ExprGetToken(
Fts5Parse *pParse,
const char **pz,
Fts5Token *pToken
){
const char *z = *pz;
int tok;
while( fts5ExprIsspace(*z) ) z++;
pToken->p = z;
pToken->n = 1;
switch( *z ){
case '(': tok = FTS5_LP; break;
case ')': tok = FTS5_RP; break;
case '{': tok = FTS5_LCP; break;
case '}': tok = FTS5_RCP; break;
case ':': tok = FTS5_COLON; break;
case ',': tok = FTS5_COMMA; break;
case '+': tok = FTS5_PLUS; break;
case '*': tok = FTS5_STAR; break;
case '-': tok = FTS5_MINUS; break;
case '^': tok = FTS5_CARET; break;
case '\0': tok = FTS5_EOF; break;
case '"': {
const char *z2;
tok = FTS5_STRING;
for(z2=&z[1]; 1; z2++){
if( z2[0]=='"' ){
z2++;
if( z2[0]!='"' ) break;
}
if( z2[0]=='\0' ){
sqlite3Fts5ParseError(pParse, "unterminated string");
return FTS5_EOF;
}
}
pToken->n = (z2 - z);
break;
}
default: {
const char *z2;
if( sqlite3Fts5IsBareword(z[0])==0 ){
sqlite3Fts5ParseError(pParse, "fts5: syntax error near \"%.1s\"", z);
return FTS5_EOF;
}
tok = FTS5_STRING;
for(z2=&z[1]; sqlite3Fts5IsBareword(*z2); z2++);
pToken->n = (z2 - z);
if( pToken->n==2 && memcmp(pToken->p, "OR", 2)==0 ) tok = FTS5_OR;
if( pToken->n==3 && memcmp(pToken->p, "NOT", 3)==0 ) tok = FTS5_NOT;
if( pToken->n==3 && memcmp(pToken->p, "AND", 3)==0 ) tok = FTS5_AND;
break;
}
}
*pz = &pToken->p[pToken->n];
return tok;
}
static void *fts5ParseAlloc(u64 t){ return sqlite3_malloc64((sqlite3_int64)t);}
static void fts5ParseFree(void *p){ sqlite3_free(p); }
int sqlite3Fts5ExprNew(
Fts5Config *pConfig,
int bPhraseToAnd,
int iCol,
const char *zExpr,
Fts5Expr **ppNew,
char **pzErr
){
Fts5Parse sParse;
Fts5Token token;
const char *z = zExpr;
int t;
void *pEngine;
Fts5Expr *pNew;
*ppNew = 0;
*pzErr = 0;
memset(&sParse, 0, sizeof(sParse));
sParse.bPhraseToAnd = bPhraseToAnd;
pEngine = sqlite3Fts5ParserAlloc(fts5ParseAlloc);
if( pEngine==0 ){ return SQLITE_NOMEM; }
sParse.pConfig = pConfig;
do {
t = fts5ExprGetToken(&sParse, &z, &token);
sqlite3Fts5Parser(pEngine, t, token, &sParse);
}while( sParse.rc==SQLITE_OK && t!=FTS5_EOF );
sqlite3Fts5ParserFree(pEngine, fts5ParseFree);
assert_expr_depth_ok(sParse.rc, sParse.pExpr);
if( iCol<pConfig->nCol && sParse.pExpr && sParse.rc==SQLITE_OK ){
int n = sizeof(Fts5Colset);
Fts5Colset *pColset = (Fts5Colset*)sqlite3Fts5MallocZero(&sParse.rc, n);
if( pColset ){
pColset->nCol = 1;
pColset->aiCol[0] = iCol;
sqlite3Fts5ParseSetColset(&sParse, sParse.pExpr, pColset);
}
}
assert( sParse.rc!=SQLITE_OK || sParse.zErr==0 );
if( sParse.rc==SQLITE_OK ){
*ppNew = pNew = sqlite3_malloc(sizeof(Fts5Expr));
if( pNew==0 ){
sParse.rc = SQLITE_NOMEM;
sqlite3Fts5ParseNodeFree(sParse.pExpr);
}else{
if( !sParse.pExpr ){
const int nByte = sizeof(Fts5ExprNode);
pNew->pRoot = (Fts5ExprNode*)sqlite3Fts5MallocZero(&sParse.rc, nByte);
if( pNew->pRoot ){
pNew->pRoot->bEof = 1;
}
}else{
pNew->pRoot = sParse.pExpr;
}
pNew->pIndex = 0;
pNew->pConfig = pConfig;
pNew->apExprPhrase = sParse.apPhrase;
pNew->nPhrase = sParse.nPhrase;
pNew->bDesc = 0;
sParse.apPhrase = 0;
}
}else{
sqlite3Fts5ParseNodeFree(sParse.pExpr);
}
sqlite3_free(sParse.apPhrase);
*pzErr = sParse.zErr;
return sParse.rc;
}
static int fts5ExprCountChar(const char *z, int nByte){
int nRet = 0;
int ii;
for(ii=0; ii<nByte; ii++){
if( (z[ii] & 0xC0)!=0x80 ) nRet++;
}
return nRet;
}
int sqlite3Fts5ExprPattern(
Fts5Config *pConfig, int bGlob, int iCol, const char *zText, Fts5Expr **pp
){
i64 nText = strlen(zText);
char *zExpr = (char*)sqlite3_malloc64(nText*4 + 1);
int rc = SQLITE_OK;
if( zExpr==0 ){
rc = SQLITE_NOMEM;
}else{
char aSpec[3];
int iOut = 0;
int i = 0;
int iFirst = 0;
if( bGlob==0 ){
aSpec[0] = '_';
aSpec[1] = '%';
aSpec[2] = 0;
}else{
aSpec[0] = '*';
aSpec[1] = '?';
aSpec[2] = '[';
}
while( i<=nText ){
if( i==nText
|| zText[i]==aSpec[0] || zText[i]==aSpec[1] || zText[i]==aSpec[2]
){
if( fts5ExprCountChar(&zText[iFirst], i-iFirst)>=3 ){
int jj;
zExpr[iOut++] = '"';
for(jj=iFirst; jj<i; jj++){
zExpr[iOut++] = zText[jj];
if( zText[jj]=='"' ) zExpr[iOut++] = '"';
}
zExpr[iOut++] = '"';
zExpr[iOut++] = ' ';
}
if( zText[i]==aSpec[2] ){
i += 2;
if( zText[i-1]=='^' ) i++;
while( i<nText && zText[i]!=']' ) i++;
}
iFirst = i+1;
}
i++;
}
if( iOut>0 ){
int bAnd = 0;
if( pConfig->eDetail!=FTS5_DETAIL_FULL ){
bAnd = 1;
if( pConfig->eDetail==FTS5_DETAIL_NONE ){
iCol = pConfig->nCol;
}
}
zExpr[iOut] = '\0';
rc = sqlite3Fts5ExprNew(pConfig, bAnd, iCol, zExpr, pp,pConfig->pzErrmsg);
}else{
*pp = 0;
}
sqlite3_free(zExpr);
}
return rc;
}
void sqlite3Fts5ParseNodeFree(Fts5ExprNode *p){
if( p ){
int i;
for(i=0; i<p->nChild; i++){
sqlite3Fts5ParseNodeFree(p->apChild[i]);
}
sqlite3Fts5ParseNearsetFree(p->pNear);
sqlite3_free(p);
}
}
void sqlite3Fts5ExprFree(Fts5Expr *p){
if( p ){
sqlite3Fts5ParseNodeFree(p->pRoot);
sqlite3_free(p->apExprPhrase);
sqlite3_free(p);
}
}
int sqlite3Fts5ExprAnd(Fts5Expr **pp1, Fts5Expr *p2){
Fts5Parse sParse;
memset(&sParse, 0, sizeof(sParse));
if( *pp1 && p2 ){
Fts5Expr *p1 = *pp1;
int nPhrase = p1->nPhrase + p2->nPhrase;
p1->pRoot = sqlite3Fts5ParseNode(&sParse, FTS5_AND, p1->pRoot, p2->pRoot,0);
p2->pRoot = 0;
if( sParse.rc==SQLITE_OK ){
Fts5ExprPhrase **ap = (Fts5ExprPhrase**)sqlite3_realloc(
p1->apExprPhrase, nPhrase * sizeof(Fts5ExprPhrase*)
);
if( ap==0 ){
sParse.rc = SQLITE_NOMEM;
}else{
int i;
memmove(&ap[p2->nPhrase], ap, p1->nPhrase*sizeof(Fts5ExprPhrase*));
for(i=0; i<p2->nPhrase; i++){
ap[i] = p2->apExprPhrase[i];
}
p1->nPhrase = nPhrase;
p1->apExprPhrase = ap;
}
}
sqlite3_free(p2->apExprPhrase);
sqlite3_free(p2);
}else if( p2 ){
*pp1 = p2;
}
return sParse.rc;
}
static i64 fts5ExprSynonymRowid(Fts5ExprTerm *pTerm, int bDesc, int *pbEof){
i64 iRet = 0;
int bRetValid = 0;
Fts5ExprTerm *p;
assert( pTerm );
assert( pTerm->pSynonym );
assert( bDesc==0 || bDesc==1 );
for(p=pTerm; p; p=p->pSynonym){
if( 0==sqlite3Fts5IterEof(p->pIter) ){
i64 iRowid = p->pIter->iRowid;
if( bRetValid==0 || (bDesc!=(iRowid<iRet)) ){
iRet = iRowid;
bRetValid = 1;
}
}
}
if( pbEof && bRetValid==0 ) *pbEof = 1;
return iRet;
}
static int fts5ExprSynonymList(
Fts5ExprTerm *pTerm,
i64 iRowid,
Fts5Buffer *pBuf,
u8 **pa, int *pn
){
Fts5PoslistReader aStatic[4];
Fts5PoslistReader *aIter = aStatic;
int nIter = 0;
int nAlloc = 4;
int rc = SQLITE_OK;
Fts5ExprTerm *p;
assert( pTerm->pSynonym );
for(p=pTerm; p; p=p->pSynonym){
Fts5IndexIter *pIter = p->pIter;
if( sqlite3Fts5IterEof(pIter)==0 && pIter->iRowid==iRowid ){
if( pIter->nData==0 ) continue;
if( nIter==nAlloc ){
sqlite3_int64 nByte = sizeof(Fts5PoslistReader) * nAlloc * 2;
Fts5PoslistReader *aNew = (Fts5PoslistReader*)sqlite3_malloc64(nByte);
if( aNew==0 ){
rc = SQLITE_NOMEM;
goto synonym_poslist_out;
}
memcpy(aNew, aIter, sizeof(Fts5PoslistReader) * nIter);
nAlloc = nAlloc*2;
if( aIter!=aStatic ) sqlite3_free(aIter);
aIter = aNew;
}
sqlite3Fts5PoslistReaderInit(pIter->pData, pIter->nData, &aIter[nIter]);
assert( aIter[nIter].bEof==0 );
nIter++;
}
}
if( nIter==1 ){
*pa = (u8*)aIter[0].a;
*pn = aIter[0].n;
}else{
Fts5PoslistWriter writer = {0};
i64 iPrev = -1;
fts5BufferZero(pBuf);
while( 1 ){
int i;
i64 iMin = FTS5_LARGEST_INT64;
for(i=0; i<nIter; i++){
if( aIter[i].bEof==0 ){
if( aIter[i].iPos==iPrev ){
if( sqlite3Fts5PoslistReaderNext(&aIter[i]) ) continue;
}
if( aIter[i].iPos<iMin ){
iMin = aIter[i].iPos;
}
}
}
if( iMin==FTS5_LARGEST_INT64 || rc!=SQLITE_OK ) break;
rc = sqlite3Fts5PoslistWriterAppend(pBuf, &writer, iMin);
iPrev = iMin;
}
if( rc==SQLITE_OK ){
*pa = pBuf->p;
*pn = pBuf->n;
}
}
synonym_poslist_out:
if( aIter!=aStatic ) sqlite3_free(aIter);
return rc;
}
static int fts5ExprPhraseIsMatch(
Fts5ExprNode *pNode,
Fts5ExprPhrase *pPhrase,
int *pbMatch
){
Fts5PoslistWriter writer = {0};
Fts5PoslistReader aStatic[4];
Fts5PoslistReader *aIter = aStatic;
int i;
int rc = SQLITE_OK;
int bFirst = pPhrase->aTerm[0].bFirst;
fts5BufferZero(&pPhrase->poslist);
if( pPhrase->nTerm>ArraySize(aStatic) ){
sqlite3_int64 nByte = sizeof(Fts5PoslistReader) * pPhrase->nTerm;
aIter = (Fts5PoslistReader*)sqlite3_malloc64(nByte);
if( !aIter ) return SQLITE_NOMEM;
}
memset(aIter, 0, sizeof(Fts5PoslistReader) * pPhrase->nTerm);
for(i=0; i<pPhrase->nTerm; i++){
Fts5ExprTerm *pTerm = &pPhrase->aTerm[i];
int n = 0;
int bFlag = 0;
u8 *a = 0;
if( pTerm->pSynonym ){
Fts5Buffer buf = {0, 0, 0};
rc = fts5ExprSynonymList(pTerm, pNode->iRowid, &buf, &a, &n);
if( rc ){
sqlite3_free(a);
goto ismatch_out;
}
if( a==buf.p ) bFlag = 1;
}else{
a = (u8*)pTerm->pIter->pData;
n = pTerm->pIter->nData;
}
sqlite3Fts5PoslistReaderInit(a, n, &aIter[i]);
aIter[i].bFlag = (u8)bFlag;
if( aIter[i].bEof ) goto ismatch_out;
}
while( 1 ){
int bMatch;
i64 iPos = aIter[0].iPos;
do {
bMatch = 1;
for(i=0; i<pPhrase->nTerm; i++){
Fts5PoslistReader *pPos = &aIter[i];
i64 iAdj = iPos + i;
if( pPos->iPos!=iAdj ){
bMatch = 0;
while( pPos->iPos<iAdj ){
if( sqlite3Fts5PoslistReaderNext(pPos) ) goto ismatch_out;
}
if( pPos->iPos>iAdj ) iPos = pPos->iPos-i;
}
}
}while( bMatch==0 );
if( bFirst==0 || FTS5_POS2OFFSET(iPos)==0 ){
rc = sqlite3Fts5PoslistWriterAppend(&pPhrase->poslist, &writer, iPos);
if( rc!=SQLITE_OK ) goto ismatch_out;
}
for(i=0; i<pPhrase->nTerm; i++){
if( sqlite3Fts5PoslistReaderNext(&aIter[i]) ) goto ismatch_out;
}
}
ismatch_out:
*pbMatch = (pPhrase->poslist.n>0);
for(i=0; i<pPhrase->nTerm; i++){
if( aIter[i].bFlag ) sqlite3_free((u8*)aIter[i].a);
}
if( aIter!=aStatic ) sqlite3_free(aIter);
return rc;
}
typedef struct Fts5LookaheadReader Fts5LookaheadReader;
struct Fts5LookaheadReader {
const u8 *a;
int n;
int i;
i64 iPos;
i64 iLookahead;
};
#define FTS5_LOOKAHEAD_EOF (((i64)1) << 62)
static int fts5LookaheadReaderNext(Fts5LookaheadReader *p){
p->iPos = p->iLookahead;
if( sqlite3Fts5PoslistNext64(p->a, p->n, &p->i, &p->iLookahead) ){
p->iLookahead = FTS5_LOOKAHEAD_EOF;
}
return (p->iPos==FTS5_LOOKAHEAD_EOF);
}
static int fts5LookaheadReaderInit(
const u8 *a, int n,
Fts5LookaheadReader *p
){
memset(p, 0, sizeof(Fts5LookaheadReader));
p->a = a;
p->n = n;
fts5LookaheadReaderNext(p);
return fts5LookaheadReaderNext(p);
}
typedef struct Fts5NearTrimmer Fts5NearTrimmer;
struct Fts5NearTrimmer {
Fts5LookaheadReader reader;
Fts5PoslistWriter writer;
Fts5Buffer *pOut;
};
static int fts5ExprNearIsMatch(int *pRc, Fts5ExprNearset *pNear){
Fts5NearTrimmer aStatic[4];
Fts5NearTrimmer *a = aStatic;
Fts5ExprPhrase **apPhrase = pNear->apPhrase;
int i;
int rc = *pRc;
int bMatch;
assert( pNear->nPhrase>1 );
if( pNear->nPhrase>ArraySize(aStatic) ){
sqlite3_int64 nByte = sizeof(Fts5NearTrimmer) * pNear->nPhrase;
a = (Fts5NearTrimmer*)sqlite3Fts5MallocZero(&rc, nByte);
}else{
memset(aStatic, 0, sizeof(aStatic));
}
if( rc!=SQLITE_OK ){
*pRc = rc;
return 0;
}
for(i=0; i<pNear->nPhrase; i++){
Fts5Buffer *pPoslist = &apPhrase[i]->poslist;
fts5LookaheadReaderInit(pPoslist->p, pPoslist->n, &a[i].reader);
pPoslist->n = 0;
a[i].pOut = pPoslist;
}
while( 1 ){
int iAdv;
i64 iMin;
i64 iMax;
iMax = a[0].reader.iPos;
do {
bMatch = 1;
for(i=0; i<pNear->nPhrase; i++){
Fts5LookaheadReader *pPos = &a[i].reader;
iMin = iMax - pNear->apPhrase[i]->nTerm - pNear->nNear;
if( pPos->iPos<iMin || pPos->iPos>iMax ){
bMatch = 0;
while( pPos->iPos<iMin ){
if( fts5LookaheadReaderNext(pPos) ) goto ismatch_out;
}
if( pPos->iPos>iMax ) iMax = pPos->iPos;
}
}
}while( bMatch==0 );
for(i=0; i<pNear->nPhrase; i++){
i64 iPos = a[i].reader.iPos;
Fts5PoslistWriter *pWriter = &a[i].writer;
if( a[i].pOut->n==0 || iPos!=pWriter->iPrev ){
sqlite3Fts5PoslistWriterAppend(a[i].pOut, pWriter, iPos);
}
}
iAdv = 0;
iMin = a[0].reader.iLookahead;
for(i=0; i<pNear->nPhrase; i++){
if( a[i].reader.iLookahead < iMin ){
iMin = a[i].reader.iLookahead;
iAdv = i;
}
}
if( fts5LookaheadReaderNext(&a[iAdv].reader) ) goto ismatch_out;
}
ismatch_out: {
int bRet = a[0].pOut->n>0;
*pRc = rc;
if( a!=aStatic ) sqlite3_free(a);
return bRet;
}
}
static int fts5ExprAdvanceto(
Fts5IndexIter *pIter,
int bDesc,
i64 *piLast,
int *pRc,
int *pbEof
){
i64 iLast = *piLast;
i64 iRowid;
iRowid = pIter->iRowid;
if( (bDesc==0 && iLast>iRowid) || (bDesc && iLast<iRowid) ){
int rc = sqlite3Fts5IterNextFrom(pIter, iLast);
if( rc || sqlite3Fts5IterEof(pIter) ){
*pRc = rc;
*pbEof = 1;
return 1;
}
iRowid = pIter->iRowid;
assert( (bDesc==0 && iRowid>=iLast) || (bDesc==1 && iRowid<=iLast) );
}
*piLast = iRowid;
return 0;
}
static int fts5ExprSynonymAdvanceto(
Fts5ExprTerm *pTerm,
int bDesc,
i64 *piLast,
int *pRc
){
int rc = SQLITE_OK;
i64 iLast = *piLast;
Fts5ExprTerm *p;
int bEof = 0;
for(p=pTerm; rc==SQLITE_OK && p; p=p->pSynonym){
if( sqlite3Fts5IterEof(p->pIter)==0 ){
i64 iRowid = p->pIter->iRowid;
if( (bDesc==0 && iLast>iRowid) || (bDesc && iLast<iRowid) ){
rc = sqlite3Fts5IterNextFrom(p->pIter, iLast);
}
}
}
if( rc!=SQLITE_OK ){
*pRc = rc;
bEof = 1;
}else{
*piLast = fts5ExprSynonymRowid(pTerm, bDesc, &bEof);
}
return bEof;
}
static int fts5ExprNearTest(
int *pRc,
Fts5Expr *pExpr,
Fts5ExprNode *pNode
){
Fts5ExprNearset *pNear = pNode->pNear;
int rc = *pRc;
if( pExpr->pConfig->eDetail!=FTS5_DETAIL_FULL ){
Fts5ExprTerm *pTerm;
Fts5ExprPhrase *pPhrase = pNear->apPhrase[0];
pPhrase->poslist.n = 0;
for(pTerm=&pPhrase->aTerm[0]; pTerm; pTerm=pTerm->pSynonym){
Fts5IndexIter *pIter = pTerm->pIter;
if( sqlite3Fts5IterEof(pIter)==0 ){
if( pIter->iRowid==pNode->iRowid && pIter->nData>0 ){
pPhrase->poslist.n = 1;
}
}
}
return pPhrase->poslist.n;
}else{
int i;
for(i=0; rc==SQLITE_OK && i<pNear->nPhrase; i++){
Fts5ExprPhrase *pPhrase = pNear->apPhrase[i];
if( pPhrase->nTerm>1 || pPhrase->aTerm[0].pSynonym
|| pNear->pColset || pPhrase->aTerm[0].bFirst
){
int bMatch = 0;
rc = fts5ExprPhraseIsMatch(pNode, pPhrase, &bMatch);
if( bMatch==0 ) break;
}else{
Fts5IndexIter *pIter = pPhrase->aTerm[0].pIter;
fts5BufferSet(&rc, &pPhrase->poslist, pIter->nData, pIter->pData);
}
}
*pRc = rc;
if( i==pNear->nPhrase && (i==1 || fts5ExprNearIsMatch(pRc, pNear)) ){
return 1;
}
return 0;
}
}
static int fts5ExprNearInitAll(
Fts5Expr *pExpr,
Fts5ExprNode *pNode
){
Fts5ExprNearset *pNear = pNode->pNear;
int i;
assert( pNode->bNomatch==0 );
for(i=0; i<pNear->nPhrase; i++){
Fts5ExprPhrase *pPhrase = pNear->apPhrase[i];
if( pPhrase->nTerm==0 ){
pNode->bEof = 1;
return SQLITE_OK;
}else{
int j;
for(j=0; j<pPhrase->nTerm; j++){
Fts5ExprTerm *pTerm = &pPhrase->aTerm[j];
Fts5ExprTerm *p;
int bHit = 0;
for(p=pTerm; p; p=p->pSynonym){
int rc;
if( p->pIter ){
sqlite3Fts5IterClose(p->pIter);
p->pIter = 0;
}
rc = sqlite3Fts5IndexQuery(
pExpr->pIndex, p->zTerm, (int)strlen(p->zTerm),
(pTerm->bPrefix ? FTS5INDEX_QUERY_PREFIX : 0) |
(pExpr->bDesc ? FTS5INDEX_QUERY_DESC : 0),
pNear->pColset,
&p->pIter
);
assert( (rc==SQLITE_OK)==(p->pIter!=0) );
if( rc!=SQLITE_OK ) return rc;
if( 0==sqlite3Fts5IterEof(p->pIter) ){
bHit = 1;
}
}
if( bHit==0 ){
pNode->bEof = 1;
return SQLITE_OK;
}
}
}
}
pNode->bEof = 0;
return SQLITE_OK;
}
static int fts5RowidCmp(
Fts5Expr *pExpr,
i64 iLhs,
i64 iRhs
){
assert( pExpr->bDesc==0 || pExpr->bDesc==1 );
if( pExpr->bDesc==0 ){
if( iLhs<iRhs ) return -1;
return (iLhs > iRhs);
}else{
if( iLhs>iRhs ) return -1;
return (iLhs < iRhs);
}
}
static void fts5ExprSetEof(Fts5ExprNode *pNode){
int i;
pNode->bEof = 1;
pNode->bNomatch = 0;
for(i=0; i<pNode->nChild; i++){
fts5ExprSetEof(pNode->apChild[i]);
}
}
static void fts5ExprNodeZeroPoslist(Fts5ExprNode *pNode){
if( pNode->eType==FTS5_STRING || pNode->eType==FTS5_TERM ){
Fts5ExprNearset *pNear = pNode->pNear;
int i;
for(i=0; i<pNear->nPhrase; i++){
Fts5ExprPhrase *pPhrase = pNear->apPhrase[i];
pPhrase->poslist.n = 0;
}
}else{
int i;
for(i=0; i<pNode->nChild; i++){
fts5ExprNodeZeroPoslist(pNode->apChild[i]);
}
}
}
static int fts5NodeCompare(
Fts5Expr *pExpr,
Fts5ExprNode *p1,
Fts5ExprNode *p2
){
if( p2->bEof ) return -1;
if( p1->bEof ) return +1;
return fts5RowidCmp(pExpr, p1->iRowid, p2->iRowid);
}
static int fts5ExprNodeTest_STRING(
Fts5Expr *pExpr,
Fts5ExprNode *pNode
){
Fts5ExprNearset *pNear = pNode->pNear;
Fts5ExprPhrase *pLeft = pNear->apPhrase[0];
int rc = SQLITE_OK;
i64 iLast;
int i, j;
int bMatch;
const int bDesc = pExpr->bDesc;
assert( pNear->nPhrase>1
|| pNear->apPhrase[0]->nTerm>1
|| pNear->apPhrase[0]->aTerm[0].pSynonym
|| pNear->apPhrase[0]->aTerm[0].bFirst
);
if( pLeft->aTerm[0].pSynonym ){
iLast = fts5ExprSynonymRowid(&pLeft->aTerm[0], bDesc, 0);
}else{
iLast = pLeft->aTerm[0].pIter->iRowid;
}
do {
bMatch = 1;
for(i=0; i<pNear->nPhrase; i++){
Fts5ExprPhrase *pPhrase = pNear->apPhrase[i];
for(j=0; j<pPhrase->nTerm; j++){
Fts5ExprTerm *pTerm = &pPhrase->aTerm[j];
if( pTerm->pSynonym ){
i64 iRowid = fts5ExprSynonymRowid(pTerm, bDesc, 0);
if( iRowid==iLast ) continue;
bMatch = 0;
if( fts5ExprSynonymAdvanceto(pTerm, bDesc, &iLast, &rc) ){
pNode->bNomatch = 0;
pNode->bEof = 1;
return rc;
}
}else{
Fts5IndexIter *pIter = pPhrase->aTerm[j].pIter;
if( pIter->iRowid==iLast || pIter->bEof ) continue;
bMatch = 0;
if( fts5ExprAdvanceto(pIter, bDesc, &iLast, &rc, &pNode->bEof) ){
return rc;
}
}
}
}
}while( bMatch==0 );
pNode->iRowid = iLast;
pNode->bNomatch = ((0==fts5ExprNearTest(&rc, pExpr, pNode)) && rc==SQLITE_OK);
assert( pNode->bEof==0 || pNode->bNomatch==0 );
return rc;
}
static int fts5ExprNodeNext_STRING(
Fts5Expr *pExpr,
Fts5ExprNode *pNode,
int bFromValid,
i64 iFrom
){
Fts5ExprTerm *pTerm = &pNode->pNear->apPhrase[0]->aTerm[0];
int rc = SQLITE_OK;
pNode->bNomatch = 0;
if( pTerm->pSynonym ){
int bEof = 1;
Fts5ExprTerm *p;
i64 iRowid = fts5ExprSynonymRowid(pTerm, pExpr->bDesc, 0);
for(p=pTerm; p; p=p->pSynonym){
if( sqlite3Fts5IterEof(p->pIter)==0 ){
i64 ii = p->pIter->iRowid;
if( ii==iRowid
|| (bFromValid && ii!=iFrom && (ii>iFrom)==pExpr->bDesc)
){
if( bFromValid ){
rc = sqlite3Fts5IterNextFrom(p->pIter, iFrom);
}else{
rc = sqlite3Fts5IterNext(p->pIter);
}
if( rc!=SQLITE_OK ) break;
if( sqlite3Fts5IterEof(p->pIter)==0 ){
bEof = 0;
}
}else{
bEof = 0;
}
}
}
pNode->bEof = (rc || bEof);
}else{
Fts5IndexIter *pIter = pTerm->pIter;
assert( Fts5NodeIsString(pNode) );
if( bFromValid ){
rc = sqlite3Fts5IterNextFrom(pIter, iFrom);
}else{
rc = sqlite3Fts5IterNext(pIter);
}
pNode->bEof = (rc || sqlite3Fts5IterEof(pIter));
}
if( pNode->bEof==0 ){
assert( rc==SQLITE_OK );
rc = fts5ExprNodeTest_STRING(pExpr, pNode);
}
return rc;
}
static int fts5ExprNodeTest_TERM(
Fts5Expr *pExpr,
Fts5ExprNode *pNode
){
Fts5ExprPhrase *pPhrase = pNode->pNear->apPhrase[0];
Fts5IndexIter *pIter = pPhrase->aTerm[0].pIter;
assert( pNode->eType==FTS5_TERM );
assert( pNode->pNear->nPhrase==1 && pPhrase->nTerm==1 );
assert( pPhrase->aTerm[0].pSynonym==0 );
pPhrase->poslist.n = pIter->nData;
if( pExpr->pConfig->eDetail==FTS5_DETAIL_FULL ){
pPhrase->poslist.p = (u8*)pIter->pData;
}
pNode->iRowid = pIter->iRowid;
pNode->bNomatch = (pPhrase->poslist.n==0);
return SQLITE_OK;
}
static int fts5ExprNodeNext_TERM(
Fts5Expr *pExpr,
Fts5ExprNode *pNode,
int bFromValid,
i64 iFrom
){
int rc;
Fts5IndexIter *pIter = pNode->pNear->apPhrase[0]->aTerm[0].pIter;
assert( pNode->bEof==0 );
if( bFromValid ){
rc = sqlite3Fts5IterNextFrom(pIter, iFrom);
}else{
rc = sqlite3Fts5IterNext(pIter);
}
if( rc==SQLITE_OK && sqlite3Fts5IterEof(pIter)==0 ){
rc = fts5ExprNodeTest_TERM(pExpr, pNode);
}else{
pNode->bEof = 1;
pNode->bNomatch = 0;
}
return rc;
}
static void fts5ExprNodeTest_OR(
Fts5Expr *pExpr,
Fts5ExprNode *pNode
){
Fts5ExprNode *pNext = pNode->apChild[0];
int i;
for(i=1; i<pNode->nChild; i++){
Fts5ExprNode *pChild = pNode->apChild[i];
int cmp = fts5NodeCompare(pExpr, pNext, pChild);
if( cmp>0 || (cmp==0 && pChild->bNomatch==0) ){
pNext = pChild;
}
}
pNode->iRowid = pNext->iRowid;
pNode->bEof = pNext->bEof;
pNode->bNomatch = pNext->bNomatch;
}
static int fts5ExprNodeNext_OR(
Fts5Expr *pExpr,
Fts5ExprNode *pNode,
int bFromValid,
i64 iFrom
){
int i;
i64 iLast = pNode->iRowid;
for(i=0; i<pNode->nChild; i++){
Fts5ExprNode *p1 = pNode->apChild[i];
assert( p1->bEof || fts5RowidCmp(pExpr, p1->iRowid, iLast)>=0 );
if( p1->bEof==0 ){
if( (p1->iRowid==iLast)
|| (bFromValid && fts5RowidCmp(pExpr, p1->iRowid, iFrom)<0)
){
int rc = fts5ExprNodeNext(pExpr, p1, bFromValid, iFrom);
if( rc!=SQLITE_OK ){
pNode->bNomatch = 0;
return rc;
}
}
}
}
fts5ExprNodeTest_OR(pExpr, pNode);
return SQLITE_OK;
}
static int fts5ExprNodeTest_AND(
Fts5Expr *pExpr,
Fts5ExprNode *pAnd
){
int iChild;
i64 iLast = pAnd->iRowid;
int rc = SQLITE_OK;
int bMatch;
assert( pAnd->bEof==0 );
do {
pAnd->bNomatch = 0;
bMatch = 1;
for(iChild=0; iChild<pAnd->nChild; iChild++){
Fts5ExprNode *pChild = pAnd->apChild[iChild];
int cmp = fts5RowidCmp(pExpr, iLast, pChild->iRowid);
if( cmp>0 ){
rc = fts5ExprNodeNext(pExpr, pChild, 1, iLast);
if( rc!=SQLITE_OK ){
pAnd->bNomatch = 0;
return rc;
}
}
assert( pChild->bEof || fts5RowidCmp(pExpr, iLast, pChild->iRowid)<=0 );
if( pChild->bEof ){
fts5ExprSetEof(pAnd);
bMatch = 1;
break;
}else if( iLast!=pChild->iRowid ){
bMatch = 0;
iLast = pChild->iRowid;
}
if( pChild->bNomatch ){
pAnd->bNomatch = 1;
}
}
}while( bMatch==0 );
if( pAnd->bNomatch && pAnd!=pExpr->pRoot ){
fts5ExprNodeZeroPoslist(pAnd);
}
pAnd->iRowid = iLast;
return SQLITE_OK;
}
static int fts5ExprNodeNext_AND(
Fts5Expr *pExpr,
Fts5ExprNode *pNode,
int bFromValid,
i64 iFrom
){
int rc = fts5ExprNodeNext(pExpr, pNode->apChild[0], bFromValid, iFrom);
if( rc==SQLITE_OK ){
rc = fts5ExprNodeTest_AND(pExpr, pNode);
}else{
pNode->bNomatch = 0;
}
return rc;
}
static int fts5ExprNodeTest_NOT(
Fts5Expr *pExpr,
Fts5ExprNode *pNode
){
int rc = SQLITE_OK;
Fts5ExprNode *p1 = pNode->apChild[0];
Fts5ExprNode *p2 = pNode->apChild[1];
assert( pNode->nChild==2 );
while( rc==SQLITE_OK && p1->bEof==0 ){
int cmp = fts5NodeCompare(pExpr, p1, p2);
if( cmp>0 ){
rc = fts5ExprNodeNext(pExpr, p2, 1, p1->iRowid);
cmp = fts5NodeCompare(pExpr, p1, p2);
}
assert( rc!=SQLITE_OK || cmp<=0 );
if( cmp || p2->bNomatch ) break;
rc = fts5ExprNodeNext(pExpr, p1, 0, 0);
}
pNode->bEof = p1->bEof;
pNode->bNomatch = p1->bNomatch;
pNode->iRowid = p1->iRowid;
if( p1->bEof ){
fts5ExprNodeZeroPoslist(p2);
}
return rc;
}
static int fts5ExprNodeNext_NOT(
Fts5Expr *pExpr,
Fts5ExprNode *pNode,
int bFromValid,
i64 iFrom
){
int rc = fts5ExprNodeNext(pExpr, pNode->apChild[0], bFromValid, iFrom);
if( rc==SQLITE_OK ){
rc = fts5ExprNodeTest_NOT(pExpr, pNode);
}
if( rc!=SQLITE_OK ){
pNode->bNomatch = 0;
}
return rc;
}
static int fts5ExprNodeTest(
Fts5Expr *pExpr,
Fts5ExprNode *pNode
){
int rc = SQLITE_OK;
if( pNode->bEof==0 ){
switch( pNode->eType ){
case FTS5_STRING: {
rc = fts5ExprNodeTest_STRING(pExpr, pNode);
break;
}
case FTS5_TERM: {
rc = fts5ExprNodeTest_TERM(pExpr, pNode);
break;
}
case FTS5_AND: {
rc = fts5ExprNodeTest_AND(pExpr, pNode);
break;
}
case FTS5_OR: {
fts5ExprNodeTest_OR(pExpr, pNode);
break;
}
default: assert( pNode->eType==FTS5_NOT ); {
rc = fts5ExprNodeTest_NOT(pExpr, pNode);
break;
}
}
}
return rc;
}
static int fts5ExprNodeFirst(Fts5Expr *pExpr, Fts5ExprNode *pNode){
int rc = SQLITE_OK;
pNode->bEof = 0;
pNode->bNomatch = 0;
if( Fts5NodeIsString(pNode) ){
rc = fts5ExprNearInitAll(pExpr, pNode);
}else if( pNode->xNext==0 ){
pNode->bEof = 1;
}else{
int i;
int nEof = 0;
for(i=0; i<pNode->nChild && rc==SQLITE_OK; i++){
Fts5ExprNode *pChild = pNode->apChild[i];
rc = fts5ExprNodeFirst(pExpr, pNode->apChild[i]);
assert( pChild->bEof==0 || pChild->bEof==1 );
nEof += pChild->bEof;
}
pNode->iRowid = pNode->apChild[0]->iRowid;
switch( pNode->eType ){
case FTS5_AND:
if( nEof>0 ) fts5ExprSetEof(pNode);
break;
case FTS5_OR:
if( pNode->nChild==nEof ) fts5ExprSetEof(pNode);
break;
default:
assert( pNode->eType==FTS5_NOT );
pNode->bEof = pNode->apChild[0]->bEof;
break;
}
}
if( rc==SQLITE_OK ){
rc = fts5ExprNodeTest(pExpr, pNode);
}
return rc;
}
int sqlite3Fts5ExprFirst(Fts5Expr *p, Fts5Index *pIdx, i64 iFirst, int bDesc){
Fts5ExprNode *pRoot = p->pRoot;
int rc;
p->pIndex = pIdx;
p->bDesc = bDesc;
rc = fts5ExprNodeFirst(p, pRoot);
if( rc==SQLITE_OK
&& 0==pRoot->bEof
&& fts5RowidCmp(p, pRoot->iRowid, iFirst)<0
){
rc = fts5ExprNodeNext(p, pRoot, 1, iFirst);
}
while( pRoot->bNomatch && rc==SQLITE_OK ){
assert( pRoot->bEof==0 );
rc = fts5ExprNodeNext(p, pRoot, 0, 0);
}
return rc;
}
int sqlite3Fts5ExprNext(Fts5Expr *p, i64 iLast){
int rc;
Fts5ExprNode *pRoot = p->pRoot;
assert( pRoot->bEof==0 && pRoot->bNomatch==0 );
do {
rc = fts5ExprNodeNext(p, pRoot, 0, 0);
assert( pRoot->bNomatch==0 || (rc==SQLITE_OK && pRoot->bEof==0) );
}while( pRoot->bNomatch );
if( fts5RowidCmp(p, pRoot->iRowid, iLast)>0 ){
pRoot->bEof = 1;
}
return rc;
}
int sqlite3Fts5ExprEof(Fts5Expr *p){
return p->pRoot->bEof;
}
i64 sqlite3Fts5ExprRowid(Fts5Expr *p){
return p->pRoot->iRowid;
}
static int fts5ParseStringFromToken(Fts5Token *pToken, char **pz){
int rc = SQLITE_OK;
*pz = sqlite3Fts5Strndup(&rc, pToken->p, pToken->n);
return rc;
}
static void fts5ExprPhraseFree(Fts5ExprPhrase *pPhrase){
if( pPhrase ){
int i;
for(i=0; i<pPhrase->nTerm; i++){
Fts5ExprTerm *pSyn;
Fts5ExprTerm *pNext;
Fts5ExprTerm *pTerm = &pPhrase->aTerm[i];
sqlite3_free(pTerm->zTerm);
sqlite3Fts5IterClose(pTerm->pIter);
for(pSyn=pTerm->pSynonym; pSyn; pSyn=pNext){
pNext = pSyn->pSynonym;
sqlite3Fts5IterClose(pSyn->pIter);
fts5BufferFree((Fts5Buffer*)&pSyn[1]);
sqlite3_free(pSyn);
}
}
if( pPhrase->poslist.nSpace>0 ) fts5BufferFree(&pPhrase->poslist);
sqlite3_free(pPhrase);
}
}
void sqlite3Fts5ParseSetCaret(Fts5ExprPhrase *pPhrase){
if( pPhrase && pPhrase->nTerm ){
pPhrase->aTerm[0].bFirst = 1;
}
}
Fts5ExprNearset *sqlite3Fts5ParseNearset(
Fts5Parse *pParse,
Fts5ExprNearset *pNear,
Fts5ExprPhrase *pPhrase
){
const int SZALLOC = 8;
Fts5ExprNearset *pRet = 0;
if( pParse->rc==SQLITE_OK ){
if( pPhrase==0 ){
return pNear;
}
if( pNear==0 ){
sqlite3_int64 nByte;
nByte = sizeof(Fts5ExprNearset) + SZALLOC * sizeof(Fts5ExprPhrase*);
pRet = sqlite3_malloc64(nByte);
if( pRet==0 ){
pParse->rc = SQLITE_NOMEM;
}else{
memset(pRet, 0, (size_t)nByte);
}
}else if( (pNear->nPhrase % SZALLOC)==0 ){
int nNew = pNear->nPhrase + SZALLOC;
sqlite3_int64 nByte;
nByte = sizeof(Fts5ExprNearset) + nNew * sizeof(Fts5ExprPhrase*);
pRet = (Fts5ExprNearset*)sqlite3_realloc64(pNear, nByte);
if( pRet==0 ){
pParse->rc = SQLITE_NOMEM;
}
}else{
pRet = pNear;
}
}
if( pRet==0 ){
assert( pParse->rc!=SQLITE_OK );
sqlite3Fts5ParseNearsetFree(pNear);
sqlite3Fts5ParsePhraseFree(pPhrase);
}else{
if( pRet->nPhrase>0 ){
Fts5ExprPhrase *pLast = pRet->apPhrase[pRet->nPhrase-1];
assert( pParse!=0 );
assert( pParse->apPhrase!=0 );
assert( pParse->nPhrase>=2 );
assert( pLast==pParse->apPhrase[pParse->nPhrase-2] );
if( pPhrase->nTerm==0 ){
fts5ExprPhraseFree(pPhrase);
pRet->nPhrase--;
pParse->nPhrase--;
pPhrase = pLast;
}else if( pLast->nTerm==0 ){
fts5ExprPhraseFree(pLast);
pParse->apPhrase[pParse->nPhrase-2] = pPhrase;
pParse->nPhrase--;
pRet->nPhrase--;
}
}
pRet->apPhrase[pRet->nPhrase++] = pPhrase;
}
return pRet;
}
typedef struct TokenCtx TokenCtx;
struct TokenCtx {
Fts5ExprPhrase *pPhrase;
int rc;
};
static int fts5ParseTokenize(
void *pContext,
int tflags,
const char *pToken,
int nToken,
int iUnused1,
int iUnused2
){
int rc = SQLITE_OK;
const int SZALLOC = 8;
TokenCtx *pCtx = (TokenCtx*)pContext;
Fts5ExprPhrase *pPhrase = pCtx->pPhrase;
UNUSED_PARAM2(iUnused1, iUnused2);
if( pCtx->rc!=SQLITE_OK ) return pCtx->rc;
if( nToken>FTS5_MAX_TOKEN_SIZE ) nToken = FTS5_MAX_TOKEN_SIZE;
if( pPhrase && pPhrase->nTerm>0 && (tflags & FTS5_TOKEN_COLOCATED) ){
Fts5ExprTerm *pSyn;
sqlite3_int64 nByte = sizeof(Fts5ExprTerm) + sizeof(Fts5Buffer) + nToken+1;
pSyn = (Fts5ExprTerm*)sqlite3_malloc64(nByte);
if( pSyn==0 ){
rc = SQLITE_NOMEM;
}else{
memset(pSyn, 0, (size_t)nByte);
pSyn->zTerm = ((char*)pSyn) + sizeof(Fts5ExprTerm) + sizeof(Fts5Buffer);
memcpy(pSyn->zTerm, pToken, nToken);
pSyn->pSynonym = pPhrase->aTerm[pPhrase->nTerm-1].pSynonym;
pPhrase->aTerm[pPhrase->nTerm-1].pSynonym = pSyn;
}
}else{
Fts5ExprTerm *pTerm;
if( pPhrase==0 || (pPhrase->nTerm % SZALLOC)==0 ){
Fts5ExprPhrase *pNew;
int nNew = SZALLOC + (pPhrase ? pPhrase->nTerm : 0);
pNew = (Fts5ExprPhrase*)sqlite3_realloc64(pPhrase,
sizeof(Fts5ExprPhrase) + sizeof(Fts5ExprTerm) * nNew
);
if( pNew==0 ){
rc = SQLITE_NOMEM;
}else{
if( pPhrase==0 ) memset(pNew, 0, sizeof(Fts5ExprPhrase));
pCtx->pPhrase = pPhrase = pNew;
pNew->nTerm = nNew - SZALLOC;
}
}
if( rc==SQLITE_OK ){
pTerm = &pPhrase->aTerm[pPhrase->nTerm++];
memset(pTerm, 0, sizeof(Fts5ExprTerm));
pTerm->zTerm = sqlite3Fts5Strndup(&rc, pToken, nToken);
}
}
pCtx->rc = rc;
return rc;
}
void sqlite3Fts5ParsePhraseFree(Fts5ExprPhrase *pPhrase){
fts5ExprPhraseFree(pPhrase);
}
void sqlite3Fts5ParseNearsetFree(Fts5ExprNearset *pNear){
if( pNear ){
int i;
for(i=0; i<pNear->nPhrase; i++){
fts5ExprPhraseFree(pNear->apPhrase[i]);
}
sqlite3_free(pNear->pColset);
sqlite3_free(pNear);
}
}
void sqlite3Fts5ParseFinished(Fts5Parse *pParse, Fts5ExprNode *p){
assert( pParse->pExpr==0 );
pParse->pExpr = p;
}
static int parseGrowPhraseArray(Fts5Parse *pParse){
if( (pParse->nPhrase % 8)==0 ){
sqlite3_int64 nByte = sizeof(Fts5ExprPhrase*) * (pParse->nPhrase + 8);
Fts5ExprPhrase **apNew;
apNew = (Fts5ExprPhrase**)sqlite3_realloc64(pParse->apPhrase, nByte);
if( apNew==0 ){
pParse->rc = SQLITE_NOMEM;
return SQLITE_NOMEM;
}
pParse->apPhrase = apNew;
}
return SQLITE_OK;
}
Fts5ExprPhrase *sqlite3Fts5ParseTerm(
Fts5Parse *pParse,
Fts5ExprPhrase *pAppend,
Fts5Token *pToken,
int bPrefix
){
Fts5Config *pConfig = pParse->pConfig;
TokenCtx sCtx;
int rc;
char *z = 0;
memset(&sCtx, 0, sizeof(TokenCtx));
sCtx.pPhrase = pAppend;
rc = fts5ParseStringFromToken(pToken, &z);
if( rc==SQLITE_OK ){
int flags = FTS5_TOKENIZE_QUERY | (bPrefix ? FTS5_TOKENIZE_PREFIX : 0);
int n;
sqlite3Fts5Dequote(z);
n = (int)strlen(z);
rc = sqlite3Fts5Tokenize(pConfig, flags, z, n, &sCtx, fts5ParseTokenize);
}
sqlite3_free(z);
if( rc || (rc = sCtx.rc) ){
pParse->rc = rc;
fts5ExprPhraseFree(sCtx.pPhrase);
sCtx.pPhrase = 0;
}else{
if( pAppend==0 ){
if( parseGrowPhraseArray(pParse) ){
fts5ExprPhraseFree(sCtx.pPhrase);
return 0;
}
pParse->nPhrase++;
}
if( sCtx.pPhrase==0 ){
sCtx.pPhrase = sqlite3Fts5MallocZero(&pParse->rc, sizeof(Fts5ExprPhrase));
}else if( sCtx.pPhrase->nTerm ){
sCtx.pPhrase->aTerm[sCtx.pPhrase->nTerm-1].bPrefix = (u8)bPrefix;
}
pParse->apPhrase[pParse->nPhrase-1] = sCtx.pPhrase;
}
return sCtx.pPhrase;
}
int sqlite3Fts5ExprClonePhrase(
Fts5Expr *pExpr,
int iPhrase,
Fts5Expr **ppNew
){
int rc = SQLITE_OK;
Fts5ExprPhrase *pOrig;
Fts5Expr *pNew = 0;
TokenCtx sCtx = {0,0};
pOrig = pExpr->apExprPhrase[iPhrase];
pNew = (Fts5Expr*)sqlite3Fts5MallocZero(&rc, sizeof(Fts5Expr));
if( rc==SQLITE_OK ){
pNew->apExprPhrase = (Fts5ExprPhrase**)sqlite3Fts5MallocZero(&rc,
sizeof(Fts5ExprPhrase*));
}
if( rc==SQLITE_OK ){
pNew->pRoot = (Fts5ExprNode*)sqlite3Fts5MallocZero(&rc,
sizeof(Fts5ExprNode));
}
if( rc==SQLITE_OK ){
pNew->pRoot->pNear = (Fts5ExprNearset*)sqlite3Fts5MallocZero(&rc,
sizeof(Fts5ExprNearset) + sizeof(Fts5ExprPhrase*));
}
if( rc==SQLITE_OK ){
Fts5Colset *pColsetOrig = pOrig->pNode->pNear->pColset;
if( pColsetOrig ){
sqlite3_int64 nByte;
Fts5Colset *pColset;
nByte = sizeof(Fts5Colset) + (pColsetOrig->nCol-1) * sizeof(int);
pColset = (Fts5Colset*)sqlite3Fts5MallocZero(&rc, nByte);
if( pColset ){
memcpy(pColset, pColsetOrig, (size_t)nByte);
}
pNew->pRoot->pNear->pColset = pColset;
}
}
if( pOrig->nTerm ){
int i;
for(i=0; rc==SQLITE_OK && i<pOrig->nTerm; i++){
int tflags = 0;
Fts5ExprTerm *p;
for(p=&pOrig->aTerm[i]; p && rc==SQLITE_OK; p=p->pSynonym){
const char *zTerm = p->zTerm;
rc = fts5ParseTokenize((void*)&sCtx, tflags, zTerm, (int)strlen(zTerm),
0, 0);
tflags = FTS5_TOKEN_COLOCATED;
}
if( rc==SQLITE_OK ){
sCtx.pPhrase->aTerm[i].bPrefix = pOrig->aTerm[i].bPrefix;
sCtx.pPhrase->aTerm[i].bFirst = pOrig->aTerm[i].bFirst;
}
}
}else{
sCtx.pPhrase = sqlite3Fts5MallocZero(&rc, sizeof(Fts5ExprPhrase));
}
if( rc==SQLITE_OK && ALWAYS(sCtx.pPhrase) ){
pNew->pIndex = pExpr->pIndex;
pNew->pConfig = pExpr->pConfig;
pNew->nPhrase = 1;
pNew->apExprPhrase[0] = sCtx.pPhrase;
pNew->pRoot->pNear->apPhrase[0] = sCtx.pPhrase;
pNew->pRoot->pNear->nPhrase = 1;
sCtx.pPhrase->pNode = pNew->pRoot;
if( pOrig->nTerm==1
&& pOrig->aTerm[0].pSynonym==0
&& pOrig->aTerm[0].bFirst==0
){
pNew->pRoot->eType = FTS5_TERM;
pNew->pRoot->xNext = fts5ExprNodeNext_TERM;
}else{
pNew->pRoot->eType = FTS5_STRING;
pNew->pRoot->xNext = fts5ExprNodeNext_STRING;
}
}else{
sqlite3Fts5ExprFree(pNew);
fts5ExprPhraseFree(sCtx.pPhrase);
pNew = 0;
}
*ppNew = pNew;
return rc;
}
void sqlite3Fts5ParseNear(Fts5Parse *pParse, Fts5Token *pTok){
if( pTok->n!=4 || memcmp("NEAR", pTok->p, 4) ){
sqlite3Fts5ParseError(
pParse, "fts5: syntax error near \"%.*s\"", pTok->n, pTok->p
);
}
}
void sqlite3Fts5ParseSetDistance(
Fts5Parse *pParse,
Fts5ExprNearset *pNear,
Fts5Token *p
){
if( pNear ){
int nNear = 0;
int i;
if( p->n ){
for(i=0; i<p->n; i++){
char c = (char)p->p[i];
if( c<'0' || c>'9' ){
sqlite3Fts5ParseError(
pParse, "expected integer, got \"%.*s\"", p->n, p->p
);
return;
}
nNear = nNear * 10 + (p->p[i] - '0');
}
}else{
nNear = FTS5_DEFAULT_NEARDIST;
}
pNear->nNear = nNear;
}
}
static Fts5Colset *fts5ParseColset(
Fts5Parse *pParse,
Fts5Colset *p,
int iCol
){
int nCol = p ? p->nCol : 0;
Fts5Colset *pNew;
assert( pParse->rc==SQLITE_OK );
assert( iCol>=0 && iCol<pParse->pConfig->nCol );
pNew = sqlite3_realloc64(p, sizeof(Fts5Colset) + sizeof(int)*nCol);
if( pNew==0 ){
pParse->rc = SQLITE_NOMEM;
}else{
int *aiCol = pNew->aiCol;
int i, j;
for(i=0; i<nCol; i++){
if( aiCol[i]==iCol ) return pNew;
if( aiCol[i]>iCol ) break;
}
for(j=nCol; j>i; j--){
aiCol[j] = aiCol[j-1];
}
aiCol[i] = iCol;
pNew->nCol = nCol+1;
#ifndef NDEBUG
for(i=1; i<pNew->nCol; i++) assert( pNew->aiCol[i]>pNew->aiCol[i-1] );
#endif
}
return pNew;
}
Fts5Colset *sqlite3Fts5ParseColsetInvert(Fts5Parse *pParse, Fts5Colset *p){
Fts5Colset *pRet;
int nCol = pParse->pConfig->nCol;
pRet = (Fts5Colset*)sqlite3Fts5MallocZero(&pParse->rc,
sizeof(Fts5Colset) + sizeof(int)*nCol
);
if( pRet ){
int i;
int iOld = 0;
for(i=0; i<nCol; i++){
if( iOld>=p->nCol || p->aiCol[iOld]!=i ){
pRet->aiCol[pRet->nCol++] = i;
}else{
iOld++;
}
}
}
sqlite3_free(p);
return pRet;
}
Fts5Colset *sqlite3Fts5ParseColset(
Fts5Parse *pParse,
Fts5Colset *pColset,
Fts5Token *p
){
Fts5Colset *pRet = 0;
int iCol;
char *z;
z = sqlite3Fts5Strndup(&pParse->rc, p->p, p->n);
if( pParse->rc==SQLITE_OK ){
Fts5Config *pConfig = pParse->pConfig;
sqlite3Fts5Dequote(z);
for(iCol=0; iCol<pConfig->nCol; iCol++){
if( 0==sqlite3_stricmp(pConfig->azCol[iCol], z) ) break;
}
if( iCol==pConfig->nCol ){
sqlite3Fts5ParseError(pParse, "no such column: %s", z);
}else{
pRet = fts5ParseColset(pParse, pColset, iCol);
}
sqlite3_free(z);
}
if( pRet==0 ){
assert( pParse->rc!=SQLITE_OK );
sqlite3_free(pColset);
}
return pRet;
}
static Fts5Colset *fts5CloneColset(int *pRc, Fts5Colset *pOrig){
Fts5Colset *pRet;
if( pOrig ){
sqlite3_int64 nByte = sizeof(Fts5Colset) + (pOrig->nCol-1) * sizeof(int);
pRet = (Fts5Colset*)sqlite3Fts5MallocZero(pRc, nByte);
if( pRet ){
memcpy(pRet, pOrig, (size_t)nByte);
}
}else{
pRet = 0;
}
return pRet;
}
static void fts5MergeColset(Fts5Colset *pColset, Fts5Colset *pMerge){
int iIn = 0;
int iMerge = 0;
int iOut = 0;
while( iIn<pColset->nCol && iMerge<pMerge->nCol ){
int iDiff = pColset->aiCol[iIn] - pMerge->aiCol[iMerge];
if( iDiff==0 ){
pColset->aiCol[iOut++] = pMerge->aiCol[iMerge];
iMerge++;
iIn++;
}else if( iDiff>0 ){
iMerge++;
}else{
iIn++;
}
}
pColset->nCol = iOut;
}
static void fts5ParseSetColset(
Fts5Parse *pParse,
Fts5ExprNode *pNode,
Fts5Colset *pColset,
Fts5Colset **ppFree
){
if( pParse->rc==SQLITE_OK ){
assert( pNode->eType==FTS5_TERM || pNode->eType==FTS5_STRING
|| pNode->eType==FTS5_AND || pNode->eType==FTS5_OR
|| pNode->eType==FTS5_NOT || pNode->eType==FTS5_EOF
);
if( pNode->eType==FTS5_STRING || pNode->eType==FTS5_TERM ){
Fts5ExprNearset *pNear = pNode->pNear;
if( pNear->pColset ){
fts5MergeColset(pNear->pColset, pColset);
if( pNear->pColset->nCol==0 ){
pNode->eType = FTS5_EOF;
pNode->xNext = 0;
}
}else if( *ppFree ){
pNear->pColset = pColset;
*ppFree = 0;
}else{
pNear->pColset = fts5CloneColset(&pParse->rc, pColset);
}
}else{
int i;
assert( pNode->eType!=FTS5_EOF || pNode->nChild==0 );
for(i=0; i<pNode->nChild; i++){
fts5ParseSetColset(pParse, pNode->apChild[i], pColset, ppFree);
}
}
}
}
void sqlite3Fts5ParseSetColset(
Fts5Parse *pParse,
Fts5ExprNode *pExpr,
Fts5Colset *pColset
){
Fts5Colset *pFree = pColset;
if( pParse->pConfig->eDetail==FTS5_DETAIL_NONE ){
sqlite3Fts5ParseError(pParse,
"fts5: column queries are not supported (detail=none)"
);
}else{
fts5ParseSetColset(pParse, pExpr, pColset, &pFree);
}
sqlite3_free(pFree);
}
static void fts5ExprAssignXNext(Fts5ExprNode *pNode){
switch( pNode->eType ){
case FTS5_STRING: {
Fts5ExprNearset *pNear = pNode->pNear;
if( pNear->nPhrase==1 && pNear->apPhrase[0]->nTerm==1
&& pNear->apPhrase[0]->aTerm[0].pSynonym==0
&& pNear->apPhrase[0]->aTerm[0].bFirst==0
){
pNode->eType = FTS5_TERM;
pNode->xNext = fts5ExprNodeNext_TERM;
}else{
pNode->xNext = fts5ExprNodeNext_STRING;
}
break;
};
case FTS5_OR: {
pNode->xNext = fts5ExprNodeNext_OR;
break;
};
case FTS5_AND: {
pNode->xNext = fts5ExprNodeNext_AND;
break;
};
default: assert( pNode->eType==FTS5_NOT ); {
pNode->xNext = fts5ExprNodeNext_NOT;
break;
};
}
}
static void fts5ExprAddChildren(Fts5ExprNode *p, Fts5ExprNode *pSub){
int ii = p->nChild;
if( p->eType!=FTS5_NOT && pSub->eType==p->eType ){
int nByte = sizeof(Fts5ExprNode*) * pSub->nChild;
memcpy(&p->apChild[p->nChild], pSub->apChild, nByte);
p->nChild += pSub->nChild;
sqlite3_free(pSub);
}else{
p->apChild[p->nChild++] = pSub;
}
for( ; ii<p->nChild; ii++){
p->iHeight = MAX(p->iHeight, p->apChild[ii]->iHeight + 1);
}
}
static Fts5ExprNode *fts5ParsePhraseToAnd(
Fts5Parse *pParse,
Fts5ExprNearset *pNear
){
int nTerm = pNear->apPhrase[0]->nTerm;
int ii;
int nByte;
Fts5ExprNode *pRet;
assert( pNear->nPhrase==1 );
assert( pParse->bPhraseToAnd );
nByte = sizeof(Fts5ExprNode) + nTerm*sizeof(Fts5ExprNode*);
pRet = (Fts5ExprNode*)sqlite3Fts5MallocZero(&pParse->rc, nByte);
if( pRet ){
pRet->eType = FTS5_AND;
pRet->nChild = nTerm;
pRet->iHeight = 1;
fts5ExprAssignXNext(pRet);
pParse->nPhrase--;
for(ii=0; ii<nTerm; ii++){
Fts5ExprPhrase *pPhrase = (Fts5ExprPhrase*)sqlite3Fts5MallocZero(
&pParse->rc, sizeof(Fts5ExprPhrase)
);
if( pPhrase ){
if( parseGrowPhraseArray(pParse) ){
fts5ExprPhraseFree(pPhrase);
}else{
pParse->apPhrase[pParse->nPhrase++] = pPhrase;
pPhrase->nTerm = 1;
pPhrase->aTerm[0].zTerm = sqlite3Fts5Strndup(
&pParse->rc, pNear->apPhrase[0]->aTerm[ii].zTerm, -1
);
pRet->apChild[ii] = sqlite3Fts5ParseNode(pParse, FTS5_STRING,
0, 0, sqlite3Fts5ParseNearset(pParse, 0, pPhrase)
);
}
}
}
if( pParse->rc ){
sqlite3Fts5ParseNodeFree(pRet);
pRet = 0;
}else{
sqlite3Fts5ParseNearsetFree(pNear);
}
}
return pRet;
}
Fts5ExprNode *sqlite3Fts5ParseNode(
Fts5Parse *pParse,
int eType,
Fts5ExprNode *pLeft,
Fts5ExprNode *pRight,
Fts5ExprNearset *pNear
){
Fts5ExprNode *pRet = 0;
if( pParse->rc==SQLITE_OK ){
int nChild = 0;
sqlite3_int64 nByte;
assert( (eType!=FTS5_STRING && !pNear)
|| (eType==FTS5_STRING && !pLeft && !pRight)
);
if( eType==FTS5_STRING && pNear==0 ) return 0;
if( eType!=FTS5_STRING && pLeft==0 ) return pRight;
if( eType!=FTS5_STRING && pRight==0 ) return pLeft;
if( eType==FTS5_STRING
&& pParse->bPhraseToAnd
&& pNear->apPhrase[0]->nTerm>1
){
pRet = fts5ParsePhraseToAnd(pParse, pNear);
}else{
if( eType==FTS5_NOT ){
nChild = 2;
}else if( eType==FTS5_AND || eType==FTS5_OR ){
nChild = 2;
if( pLeft->eType==eType ) nChild += pLeft->nChild-1;
if( pRight->eType==eType ) nChild += pRight->nChild-1;
}
nByte = sizeof(Fts5ExprNode) + sizeof(Fts5ExprNode*)*(nChild-1);
pRet = (Fts5ExprNode*)sqlite3Fts5MallocZero(&pParse->rc, nByte);
if( pRet ){
pRet->eType = eType;
pRet->pNear = pNear;
fts5ExprAssignXNext(pRet);
if( eType==FTS5_STRING ){
int iPhrase;
for(iPhrase=0; iPhrase<pNear->nPhrase; iPhrase++){
pNear->apPhrase[iPhrase]->pNode = pRet;
if( pNear->apPhrase[iPhrase]->nTerm==0 ){
pRet->xNext = 0;
pRet->eType = FTS5_EOF;
}
}
if( pParse->pConfig->eDetail!=FTS5_DETAIL_FULL ){
Fts5ExprPhrase *pPhrase = pNear->apPhrase[0];
if( pNear->nPhrase!=1
|| pPhrase->nTerm>1
|| (pPhrase->nTerm>0 && pPhrase->aTerm[0].bFirst)
){
sqlite3Fts5ParseError(pParse,
"fts5: %s queries are not supported (detail!=full)",
pNear->nPhrase==1 ? "phrase": "NEAR"
);
sqlite3_free(pRet);
pRet = 0;
}
}
}else{
fts5ExprAddChildren(pRet, pLeft);
fts5ExprAddChildren(pRet, pRight);
if( pRet->iHeight>SQLITE_FTS5_MAX_EXPR_DEPTH ){
sqlite3Fts5ParseError(pParse,
"fts5 expression tree is too large (maximum depth %d)",
SQLITE_FTS5_MAX_EXPR_DEPTH
);
sqlite3_free(pRet);
pRet = 0;
}
}
}
}
}
if( pRet==0 ){
assert( pParse->rc!=SQLITE_OK );
sqlite3Fts5ParseNodeFree(pLeft);
sqlite3Fts5ParseNodeFree(pRight);
sqlite3Fts5ParseNearsetFree(pNear);
}
return pRet;
}
Fts5ExprNode *sqlite3Fts5ParseImplicitAnd(
Fts5Parse *pParse,
Fts5ExprNode *pLeft,
Fts5ExprNode *pRight
){
Fts5ExprNode *pRet = 0;
Fts5ExprNode *pPrev;
if( pParse->rc ){
sqlite3Fts5ParseNodeFree(pLeft);
sqlite3Fts5ParseNodeFree(pRight);
}else{
assert( pLeft->eType==FTS5_STRING
|| pLeft->eType==FTS5_TERM
|| pLeft->eType==FTS5_EOF
|| pLeft->eType==FTS5_AND
);
assert( pRight->eType==FTS5_STRING
|| pRight->eType==FTS5_TERM
|| pRight->eType==FTS5_EOF
);
if( pLeft->eType==FTS5_AND ){
pPrev = pLeft->apChild[pLeft->nChild-1];
}else{
pPrev = pLeft;
}
assert( pPrev->eType==FTS5_STRING
|| pPrev->eType==FTS5_TERM
|| pPrev->eType==FTS5_EOF
);
if( pRight->eType==FTS5_EOF ){
assert( pParse->apPhrase[pParse->nPhrase-1]==pRight->pNear->apPhrase[0] );
sqlite3Fts5ParseNodeFree(pRight);
pRet = pLeft;
pParse->nPhrase--;
}
else if( pPrev->eType==FTS5_EOF ){
Fts5ExprPhrase **ap;
if( pPrev==pLeft ){
pRet = pRight;
}else{
pLeft->apChild[pLeft->nChild-1] = pRight;
pRet = pLeft;
}
ap = &pParse->apPhrase[pParse->nPhrase-1-pRight->pNear->nPhrase];
assert( ap[0]==pPrev->pNear->apPhrase[0] );
memmove(ap, &ap[1], sizeof(Fts5ExprPhrase*)*pRight->pNear->nPhrase);
pParse->nPhrase--;
sqlite3Fts5ParseNodeFree(pPrev);
}
else{
pRet = sqlite3Fts5ParseNode(pParse, FTS5_AND, pLeft, pRight, 0);
}
}
return pRet;
}
#ifdef SQLITE_TEST
static char *fts5ExprTermPrint(Fts5ExprTerm *pTerm){
sqlite3_int64 nByte = 0;
Fts5ExprTerm *p;
char *zQuoted;
for(p=pTerm; p; p=p->pSynonym){
nByte += (int)strlen(pTerm->zTerm) * 2 + 3 + 2;
}
zQuoted = sqlite3_malloc64(nByte);
if( zQuoted ){
int i = 0;
for(p=pTerm; p; p=p->pSynonym){
char *zIn = p->zTerm;
zQuoted[i++] = '"';
while( *zIn ){
if( *zIn=='"' ) zQuoted[i++] = '"';
zQuoted[i++] = *zIn++;
}
zQuoted[i++] = '"';
if( p->pSynonym ) zQuoted[i++] = '|';
}
if( pTerm->bPrefix ){
zQuoted[i++] = ' ';
zQuoted[i++] = '*';
}
zQuoted[i++] = '\0';
}
return zQuoted;
}
static char *fts5PrintfAppend(char *zApp, const char *zFmt, ...){
char *zNew;
va_list ap;
va_start(ap, zFmt);
zNew = sqlite3_vmprintf(zFmt, ap);
va_end(ap);
if( zApp && zNew ){
char *zNew2 = sqlite3_mprintf("%s%s", zApp, zNew);
sqlite3_free(zNew);
zNew = zNew2;
}
sqlite3_free(zApp);
return zNew;
}
static char *fts5ExprPrintTcl(
Fts5Config *pConfig,
const char *zNearsetCmd,
Fts5ExprNode *pExpr
){
char *zRet = 0;
if( pExpr->eType==FTS5_STRING || pExpr->eType==FTS5_TERM ){
Fts5ExprNearset *pNear = pExpr->pNear;
int i;
int iTerm;
zRet = fts5PrintfAppend(zRet, "%s ", zNearsetCmd);
if( zRet==0 ) return 0;
if( pNear->pColset ){
int *aiCol = pNear->pColset->aiCol;
int nCol = pNear->pColset->nCol;
if( nCol==1 ){
zRet = fts5PrintfAppend(zRet, "-col %d ", aiCol[0]);
}else{
zRet = fts5PrintfAppend(zRet, "-col {%d", aiCol[0]);
for(i=1; i<pNear->pColset->nCol; i++){
zRet = fts5PrintfAppend(zRet, " %d", aiCol[i]);
}
zRet = fts5PrintfAppend(zRet, "} ");
}
if( zRet==0 ) return 0;
}
if( pNear->nPhrase>1 ){
zRet = fts5PrintfAppend(zRet, "-near %d ", pNear->nNear);
if( zRet==0 ) return 0;
}
zRet = fts5PrintfAppend(zRet, "--");
if( zRet==0 ) return 0;
for(i=0; i<pNear->nPhrase; i++){
Fts5ExprPhrase *pPhrase = pNear->apPhrase[i];
zRet = fts5PrintfAppend(zRet, " {");
for(iTerm=0; zRet && iTerm<pPhrase->nTerm; iTerm++){
char *zTerm = pPhrase->aTerm[iTerm].zTerm;
zRet = fts5PrintfAppend(zRet, "%s%s", iTerm==0?"":" ", zTerm);
if( pPhrase->aTerm[iTerm].bPrefix ){
zRet = fts5PrintfAppend(zRet, "*");
}
}
if( zRet ) zRet = fts5PrintfAppend(zRet, "}");
if( zRet==0 ) return 0;
}
}else{
char const *zOp = 0;
int i;
switch( pExpr->eType ){
case FTS5_AND: zOp = "AND"; break;
case FTS5_NOT: zOp = "NOT"; break;
default:
assert( pExpr->eType==FTS5_OR );
zOp = "OR";
break;
}
zRet = sqlite3_mprintf("%s", zOp);
for(i=0; zRet && i<pExpr->nChild; i++){
char *z = fts5ExprPrintTcl(pConfig, zNearsetCmd, pExpr->apChild[i]);
if( !z ){
sqlite3_free(zRet);
zRet = 0;
}else{
zRet = fts5PrintfAppend(zRet, " [%z]", z);
}
}
}
return zRet;
}
static char *fts5ExprPrint(Fts5Config *pConfig, Fts5ExprNode *pExpr){
char *zRet = 0;
if( pExpr->eType==0 ){
return sqlite3_mprintf("\"\"");
}else
if( pExpr->eType==FTS5_STRING || pExpr->eType==FTS5_TERM ){
Fts5ExprNearset *pNear = pExpr->pNear;
int i;
int iTerm;
if( pNear->pColset ){
int ii;
Fts5Colset *pColset = pNear->pColset;
if( pColset->nCol>1 ) zRet = fts5PrintfAppend(zRet, "{");
for(ii=0; ii<pColset->nCol; ii++){
zRet = fts5PrintfAppend(zRet, "%s%s",
pConfig->azCol[pColset->aiCol[ii]], ii==pColset->nCol-1 ? "" : " "
);
}
if( zRet ){
zRet = fts5PrintfAppend(zRet, "%s : ", pColset->nCol>1 ? "}" : "");
}
if( zRet==0 ) return 0;
}
if( pNear->nPhrase>1 ){
zRet = fts5PrintfAppend(zRet, "NEAR(");
if( zRet==0 ) return 0;
}
for(i=0; i<pNear->nPhrase; i++){
Fts5ExprPhrase *pPhrase = pNear->apPhrase[i];
if( i!=0 ){
zRet = fts5PrintfAppend(zRet, " ");
if( zRet==0 ) return 0;
}
for(iTerm=0; iTerm<pPhrase->nTerm; iTerm++){
char *zTerm = fts5ExprTermPrint(&pPhrase->aTerm[iTerm]);
if( zTerm ){
zRet = fts5PrintfAppend(zRet, "%s%s", iTerm==0?"":" + ", zTerm);
sqlite3_free(zTerm);
}
if( zTerm==0 || zRet==0 ){
sqlite3_free(zRet);
return 0;
}
}
}
if( pNear->nPhrase>1 ){
zRet = fts5PrintfAppend(zRet, ", %d)", pNear->nNear);
if( zRet==0 ) return 0;
}
}else{
char const *zOp = 0;
int i;
switch( pExpr->eType ){
case FTS5_AND: zOp = " AND "; break;
case FTS5_NOT: zOp = " NOT "; break;
default:
assert( pExpr->eType==FTS5_OR );
zOp = " OR ";
break;
}
for(i=0; i<pExpr->nChild; i++){
char *z = fts5ExprPrint(pConfig, pExpr->apChild[i]);
if( z==0 ){
sqlite3_free(zRet);
zRet = 0;
}else{
int e = pExpr->apChild[i]->eType;
int b = (e!=FTS5_STRING && e!=FTS5_TERM && e!=FTS5_EOF);
zRet = fts5PrintfAppend(zRet, "%s%s%z%s",
(i==0 ? "" : zOp),
(b?"(":""), z, (b?")":"")
);
}
if( zRet==0 ) break;
}
}
return zRet;
}
static void fts5ExprFunction(
sqlite3_context *pCtx,
int nArg,
sqlite3_value **apVal,
int bTcl
){
Fts5Global *pGlobal = (Fts5Global*)sqlite3_user_data(pCtx);
sqlite3 *db = sqlite3_context_db_handle(pCtx);
const char *zExpr = 0;
char *zErr = 0;
Fts5Expr *pExpr = 0;
int rc;
int i;
const char **azConfig;
const char *zNearsetCmd = "nearset";
int nConfig;
Fts5Config *pConfig = 0;
int iArg = 1;
if( nArg<1 ){
zErr = sqlite3_mprintf("wrong number of arguments to function %s",
bTcl ? "fts5_expr_tcl" : "fts5_expr"
);
sqlite3_result_error(pCtx, zErr, -1);
sqlite3_free(zErr);
return;
}
if( bTcl && nArg>1 ){
zNearsetCmd = (const char*)sqlite3_value_text(apVal[1]);
iArg = 2;
}
nConfig = 3 + (nArg-iArg);
azConfig = (const char**)sqlite3_malloc64(sizeof(char*) * nConfig);
if( azConfig==0 ){
sqlite3_result_error_nomem(pCtx);
return;
}
azConfig[0] = 0;
azConfig[1] = "main";
azConfig[2] = "tbl";
for(i=3; iArg<nArg; iArg++){
const char *z = (const char*)sqlite3_value_text(apVal[iArg]);
azConfig[i++] = (z ? z : "");
}
zExpr = (const char*)sqlite3_value_text(apVal[0]);
if( zExpr==0 ) zExpr = "";
rc = sqlite3Fts5ConfigParse(pGlobal, db, nConfig, azConfig, &pConfig, &zErr);
if( rc==SQLITE_OK ){
rc = sqlite3Fts5ExprNew(pConfig, 0, pConfig->nCol, zExpr, &pExpr, &zErr);
}
if( rc==SQLITE_OK ){
char *zText;
if( pExpr->pRoot->xNext==0 ){
zText = sqlite3_mprintf("");
}else if( bTcl ){
zText = fts5ExprPrintTcl(pConfig, zNearsetCmd, pExpr->pRoot);
}else{
zText = fts5ExprPrint(pConfig, pExpr->pRoot);
}
if( zText==0 ){
rc = SQLITE_NOMEM;
}else{
sqlite3_result_text(pCtx, zText, -1, SQLITE_TRANSIENT);
sqlite3_free(zText);
}
}
if( rc!=SQLITE_OK ){
if( zErr ){
sqlite3_result_error(pCtx, zErr, -1);
sqlite3_free(zErr);
}else{
sqlite3_result_error_code(pCtx, rc);
}
}
sqlite3_free((void *)azConfig);
sqlite3Fts5ConfigFree(pConfig);
sqlite3Fts5ExprFree(pExpr);
}
static void fts5ExprFunctionHr(
sqlite3_context *pCtx,
int nArg,
sqlite3_value **apVal
){
fts5ExprFunction(pCtx, nArg, apVal, 0);
}
static void fts5ExprFunctionTcl(
sqlite3_context *pCtx,
int nArg,
sqlite3_value **apVal
){
fts5ExprFunction(pCtx, nArg, apVal, 1);
}
static void fts5ExprIsAlnum(
sqlite3_context *pCtx,
int nArg,
sqlite3_value **apVal
){
int iCode;
u8 aArr[32];
if( nArg!=1 ){
sqlite3_result_error(pCtx,
"wrong number of arguments to function fts5_isalnum", -1
);
return;
}
memset(aArr, 0, sizeof(aArr));
sqlite3Fts5UnicodeCatParse("L*", aArr);
sqlite3Fts5UnicodeCatParse("N*", aArr);
sqlite3Fts5UnicodeCatParse("Co", aArr);
iCode = sqlite3_value_int(apVal[0]);
sqlite3_result_int(pCtx, aArr[sqlite3Fts5UnicodeCategory((u32)iCode)]);
}
static void fts5ExprFold(
sqlite3_context *pCtx,
int nArg,
sqlite3_value **apVal
){
if( nArg!=1 && nArg!=2 ){
sqlite3_result_error(pCtx,
"wrong number of arguments to function fts5_fold", -1
);
}else{
int iCode;
int bRemoveDiacritics = 0;
iCode = sqlite3_value_int(apVal[0]);
if( nArg==2 ) bRemoveDiacritics = sqlite3_value_int(apVal[1]);
sqlite3_result_int(pCtx, sqlite3Fts5UnicodeFold(iCode, bRemoveDiacritics));
}
}
#endif
int sqlite3Fts5ExprInit(Fts5Global *pGlobal, sqlite3 *db){
#ifdef SQLITE_TEST
struct Fts5ExprFunc {
const char *z;
void (*x)(sqlite3_context*,int,sqlite3_value**);
} aFunc[] = {
{ "fts5_expr", fts5ExprFunctionHr },
{ "fts5_expr_tcl", fts5ExprFunctionTcl },
{ "fts5_isalnum", fts5ExprIsAlnum },
{ "fts5_fold", fts5ExprFold },
};
int i;
int rc = SQLITE_OK;
void *pCtx = (void*)pGlobal;
for(i=0; rc==SQLITE_OK && i<ArraySize(aFunc); i++){
struct Fts5ExprFunc *p = &aFunc[i];
rc = sqlite3_create_function(db, p->z, -1, SQLITE_UTF8, pCtx, p->x, 0, 0);
}
#else
int rc = SQLITE_OK;
UNUSED_PARAM2(pGlobal,db);
#endif
#ifndef NDEBUG
(void)sqlite3Fts5ParserTrace;
#endif
(void)sqlite3Fts5ParserFallback;
return rc;
}
int sqlite3Fts5ExprPhraseCount(Fts5Expr *pExpr){
return (pExpr ? pExpr->nPhrase : 0);
}
int sqlite3Fts5ExprPhraseSize(Fts5Expr *pExpr, int iPhrase){
if( iPhrase<0 || iPhrase>=pExpr->nPhrase ) return 0;
return pExpr->apExprPhrase[iPhrase]->nTerm;
}
int sqlite3Fts5ExprPoslist(Fts5Expr *pExpr, int iPhrase, const u8 **pa){
int nRet;
Fts5ExprPhrase *pPhrase = pExpr->apExprPhrase[iPhrase];
Fts5ExprNode *pNode = pPhrase->pNode;
if( pNode->bEof==0 && pNode->iRowid==pExpr->pRoot->iRowid ){
*pa = pPhrase->poslist.p;
nRet = pPhrase->poslist.n;
}else{
*pa = 0;
nRet = 0;
}
return nRet;
}
struct Fts5PoslistPopulator {
Fts5PoslistWriter writer;
int bOk;
int bMiss;
};
Fts5PoslistPopulator *sqlite3Fts5ExprClearPoslists(Fts5Expr *pExpr, int bLive){
Fts5PoslistPopulator *pRet;
pRet = sqlite3_malloc64(sizeof(Fts5PoslistPopulator)*pExpr->nPhrase);
if( pRet ){
int i;
memset(pRet, 0, sizeof(Fts5PoslistPopulator)*pExpr->nPhrase);
for(i=0; i<pExpr->nPhrase; i++){
Fts5Buffer *pBuf = &pExpr->apExprPhrase[i]->poslist;
Fts5ExprNode *pNode = pExpr->apExprPhrase[i]->pNode;
assert( pExpr->apExprPhrase[i]->nTerm<=1 );
if( bLive &&
(pBuf->n==0 || pNode->iRowid!=pExpr->pRoot->iRowid || pNode->bEof)
){
pRet[i].bMiss = 1;
}else{
pBuf->n = 0;
}
}
}
return pRet;
}
struct Fts5ExprCtx {
Fts5Expr *pExpr;
Fts5PoslistPopulator *aPopulator;
i64 iOff;
};
typedef struct Fts5ExprCtx Fts5ExprCtx;
static int fts5ExprColsetTest(Fts5Colset *pColset, int iCol){
int i;
for(i=0; i<pColset->nCol; i++){
if( pColset->aiCol[i]==iCol ) return 1;
}
return 0;
}
static int fts5ExprPopulatePoslistsCb(
void *pCtx,
int tflags,
const char *pToken,
int nToken,
int iUnused1,
int iUnused2
){
Fts5ExprCtx *p = (Fts5ExprCtx*)pCtx;
Fts5Expr *pExpr = p->pExpr;
int i;
UNUSED_PARAM2(iUnused1, iUnused2);
if( nToken>FTS5_MAX_TOKEN_SIZE ) nToken = FTS5_MAX_TOKEN_SIZE;
if( (tflags & FTS5_TOKEN_COLOCATED)==0 ) p->iOff++;
for(i=0; i<pExpr->nPhrase; i++){
Fts5ExprTerm *pTerm;
if( p->aPopulator[i].bOk==0 ) continue;
for(pTerm=&pExpr->apExprPhrase[i]->aTerm[0]; pTerm; pTerm=pTerm->pSynonym){
int nTerm = (int)strlen(pTerm->zTerm);
if( (nTerm==nToken || (nTerm<nToken && pTerm->bPrefix))
&& memcmp(pTerm->zTerm, pToken, nTerm)==0
){
int rc = sqlite3Fts5PoslistWriterAppend(
&pExpr->apExprPhrase[i]->poslist, &p->aPopulator[i].writer, p->iOff
);
if( rc ) return rc;
break;
}
}
}
return SQLITE_OK;
}
int sqlite3Fts5ExprPopulatePoslists(
Fts5Config *pConfig,
Fts5Expr *pExpr,
Fts5PoslistPopulator *aPopulator,
int iCol,
const char *z, int n
){
int i;
Fts5ExprCtx sCtx;
sCtx.pExpr = pExpr;
sCtx.aPopulator = aPopulator;
sCtx.iOff = (((i64)iCol) << 32) - 1;
for(i=0; i<pExpr->nPhrase; i++){
Fts5ExprNode *pNode = pExpr->apExprPhrase[i]->pNode;
Fts5Colset *pColset = pNode->pNear->pColset;
if( (pColset && 0==fts5ExprColsetTest(pColset, iCol))
|| aPopulator[i].bMiss
){
aPopulator[i].bOk = 0;
}else{
aPopulator[i].bOk = 1;
}
}
return sqlite3Fts5Tokenize(pConfig,
FTS5_TOKENIZE_DOCUMENT, z, n, (void*)&sCtx, fts5ExprPopulatePoslistsCb
);
}
static void fts5ExprClearPoslists(Fts5ExprNode *pNode){
if( pNode->eType==FTS5_TERM || pNode->eType==FTS5_STRING ){
pNode->pNear->apPhrase[0]->poslist.n = 0;
}else{
int i;
for(i=0; i<pNode->nChild; i++){
fts5ExprClearPoslists(pNode->apChild[i]);
}
}
}
static int fts5ExprCheckPoslists(Fts5ExprNode *pNode, i64 iRowid){
pNode->iRowid = iRowid;
pNode->bEof = 0;
switch( pNode->eType ){
case FTS5_TERM:
case FTS5_STRING:
return (pNode->pNear->apPhrase[0]->poslist.n>0);
case FTS5_AND: {
int i;
for(i=0; i<pNode->nChild; i++){
if( fts5ExprCheckPoslists(pNode->apChild[i], iRowid)==0 ){
fts5ExprClearPoslists(pNode);
return 0;
}
}
break;
}
case FTS5_OR: {
int i;
int bRet = 0;
for(i=0; i<pNode->nChild; i++){
if( fts5ExprCheckPoslists(pNode->apChild[i], iRowid) ){
bRet = 1;
}
}
return bRet;
}
default: {
assert( pNode->eType==FTS5_NOT );
if( 0==fts5ExprCheckPoslists(pNode->apChild[0], iRowid)
|| 0!=fts5ExprCheckPoslists(pNode->apChild[1], iRowid)
){
fts5ExprClearPoslists(pNode);
return 0;
}
break;
}
}
return 1;
}
void sqlite3Fts5ExprCheckPoslists(Fts5Expr *pExpr, i64 iRowid){
fts5ExprCheckPoslists(pExpr->pRoot, iRowid);
}
int sqlite3Fts5ExprPhraseCollist(
Fts5Expr *pExpr,
int iPhrase,
const u8 **ppCollist,
int *pnCollist
){
Fts5ExprPhrase *pPhrase = pExpr->apExprPhrase[iPhrase];
Fts5ExprNode *pNode = pPhrase->pNode;
int rc = SQLITE_OK;
assert( iPhrase>=0 && iPhrase<pExpr->nPhrase );
assert( pExpr->pConfig->eDetail==FTS5_DETAIL_COLUMNS );
if( pNode->bEof==0
&& pNode->iRowid==pExpr->pRoot->iRowid
&& pPhrase->poslist.n>0
){
Fts5ExprTerm *pTerm = &pPhrase->aTerm[0];
if( pTerm->pSynonym ){
Fts5Buffer *pBuf = (Fts5Buffer*)&pTerm->pSynonym[1];
rc = fts5ExprSynonymList(
pTerm, pNode->iRowid, pBuf, (u8**)ppCollist, pnCollist
);
}else{
*ppCollist = pPhrase->aTerm[0].pIter->pData;
*pnCollist = pPhrase->aTerm[0].pIter->nData;
}
}else{
*ppCollist = 0;
*pnCollist = 0;
}
return rc;
}