int bGe = (flags & FTS5INDEX_QUERY_SCAN);
int bDlidx = 0; /* True if there is a doclist-index */
- static int nCall = 0;
- nCall++;
-
assert( bGe==0 || (flags & FTS5INDEX_QUERY_DESC)==0 );
assert( pTerm && nTerm );
memset(pIter, 0, sizeof(*pIter));
return pNew;
}
-/*
-** Allocate a new Fts5Iter object.
-**
-** The new object will be used to iterate through data in structure pStruct.
-** If iLevel is -ve, then all data in all segments is merged. Or, if iLevel
-** is zero or greater, data from the first nSegment segments on level iLevel
-** is merged.
-**
-** The iterator initially points to the first term/rowid entry in the
-** iterated data.
-*/
-static void fts5MultiIterNew(
- Fts5Index *p, /* FTS5 backend to iterate within */
- Fts5Structure *pStruct, /* Structure of specific index */
- int bSkipEmpty, /* True to ignore delete-keys */
- int flags, /* FTS5INDEX_QUERY_XXX flags */
- const u8 *pTerm, int nTerm, /* Term to seek to (or NULL/0) */
- int iLevel, /* Level to iterate (-1 for all) */
- int nSegment, /* Number of segments to merge (iLevel>=0) */
- Fts5Iter **ppOut /* New object */
+static void fts5PoslistCallback(
+ Fts5Index *p,
+ void *pContext,
+ const u8 *pChunk, int nChunk
){
- int nSeg = 0; /* Number of segment-iters in use */
- int iIter = 0; /* */
- int iSeg; /* Used to iterate through segments */
- Fts5Buffer buf = {0,0,0}; /* Buffer used by fts5SegIterSeekInit() */
- Fts5StructureLevel *pLvl;
- Fts5Iter *pNew;
+ assert_nc( nChunk>=0 );
+ if( nChunk>0 ){
+ fts5BufferSafeAppendBlob((Fts5Buffer*)pContext, pChunk, nChunk);
+ }
+}
- assert( (pTerm==0 && nTerm==0) || iLevel<0 );
+typedef struct PoslistCallbackCtx PoslistCallbackCtx;
+struct PoslistCallbackCtx {
+ Fts5Buffer *pBuf; /* Append to this buffer */
+ Fts5Colset *pColset; /* Restrict matches to this column */
+ int eState; /* See above */
+};
- /* Allocate space for the new multi-seg-iterator. */
- if( p->rc==SQLITE_OK ){
- if( iLevel<0 ){
- assert( pStruct->nSegment==fts5StructureCountSegments(pStruct) );
- nSeg = pStruct->nSegment;
- nSeg += (p->pHash ? 1 : 0);
- }else{
- nSeg = MIN(pStruct->aLevel[iLevel].nSeg, nSegment);
- }
- }
- *ppOut = pNew = fts5MultiIterAlloc(p, nSeg);
- if( pNew==0 ) return;
- pNew->bRev = (0!=(flags & FTS5INDEX_QUERY_DESC));
- pNew->bSkipEmpty = (u8)bSkipEmpty;
- pNew->pStruct = pStruct;
- fts5StructureRef(pStruct);
+typedef struct PoslistOffsetsCtx PoslistOffsetsCtx;
+struct PoslistOffsetsCtx {
+ Fts5Buffer *pBuf; /* Append to this buffer */
+ Fts5Colset *pColset; /* Restrict matches to this column */
+ int iRead;
+ int iWrite;
+};
- /* Initialize each of the component segment iterators. */
- if( iLevel<0 ){
- Fts5StructureLevel *pEnd = &pStruct->aLevel[pStruct->nLevel];
- if( p->pHash ){
- /* Add a segment iterator for the current contents of the hash table. */
- Fts5SegIter *pIter = &pNew->aSeg[iIter++];
- fts5SegIterHashInit(p, pTerm, nTerm, flags, pIter);
- }
- for(pLvl=&pStruct->aLevel[0]; pLvl<pEnd; pLvl++){
- for(iSeg=pLvl->nSeg-1; iSeg>=0; iSeg--){
- Fts5StructureSegment *pSeg = &pLvl->aSeg[iSeg];
- Fts5SegIter *pIter = &pNew->aSeg[iIter++];
- if( pTerm==0 ){
- fts5SegIterInit(p, pSeg, pIter);
- }else{
- fts5SegIterSeekInit(p, &buf, pTerm, nTerm, flags, pSeg, pIter);
- }
- }
- }
- }else{
- pLvl = &pStruct->aLevel[iLevel];
- for(iSeg=nSeg-1; iSeg>=0; iSeg--){
- fts5SegIterInit(p, &pLvl->aSeg[iSeg], &pNew->aSeg[iIter++]);
- }
+/*
+** TODO: Make this more efficient!
+*/
+static int fts5IndexColsetTest(Fts5Colset *pColset, int iCol){
+ int i;
+ for(i=0; i<pColset->nCol; i++){
+ if( pColset->aiCol[i]==iCol ) return 1;
}
- assert( iIter==nSeg );
+ return 0;
+}
- /* If the above was successful, each component iterators now points
- ** to the first entry in its segment. In this case initialize the
- ** aFirst[] array. Or, if an error has occurred, free the iterator
- ** object and set the output variable to NULL. */
- if( p->rc==SQLITE_OK ){
- for(iIter=pNew->nSeg-1; iIter>0; iIter--){
- int iEq;
- if( (iEq = fts5MultiIterDoCompare(pNew, iIter)) ){
- Fts5SegIter *pSeg = &pNew->aSeg[iEq];
- if( p->rc==SQLITE_OK ) pSeg->xNext(p, pSeg, 0);
- fts5MultiIterAdvanced(p, pNew, iEq, iIter);
+static void fts5PoslistOffsetsCallback(
+ Fts5Index *p,
+ void *pContext,
+ const u8 *pChunk, int nChunk
+){
+ PoslistOffsetsCtx *pCtx = (PoslistOffsetsCtx*)pContext;
+ assert_nc( nChunk>=0 );
+ if( nChunk>0 ){
+ int i = 0;
+ while( i<nChunk ){
+ int iVal;
+ i += fts5GetVarint32(&pChunk[i], iVal);
+ iVal += pCtx->iRead - 2;
+ pCtx->iRead = iVal;
+ if( fts5IndexColsetTest(pCtx->pColset, iVal) ){
+ fts5BufferSafeAppendVarint(pCtx->pBuf, iVal + 2 - pCtx->iWrite);
+ pCtx->iWrite = iVal;
}
}
- fts5MultiIterSetEof(pNew);
- fts5AssertMultiIterSetup(p, pNew);
-
- if( pNew->bSkipEmpty && fts5MultiIterIsEmpty(p, pNew) ){
- fts5MultiIterNext(p, pNew, 0, 0);
- }
- }else{
- fts5MultiIterFree(p, pNew);
- *ppOut = 0;
}
- fts5BufferFree(&buf);
}
-/*
-** Create an Fts5Iter that iterates through the doclist provided
-** as the second argument.
-*/
-static void fts5MultiIterNew2(
- Fts5Index *p, /* FTS5 backend to iterate within */
- Fts5Data *pData, /* Doclist to iterate through */
- int bDesc, /* True for descending rowid order */
- Fts5Iter **ppOut /* New object */
+static void fts5PoslistFilterCallback(
+ Fts5Index *p,
+ void *pContext,
+ const u8 *pChunk, int nChunk
){
- Fts5Iter *pNew;
- pNew = fts5MultiIterAlloc(p, 2);
- if( pNew ){
- Fts5SegIter *pIter = &pNew->aSeg[1];
+ PoslistCallbackCtx *pCtx = (PoslistCallbackCtx*)pContext;
+ assert_nc( nChunk>=0 );
+ if( nChunk>0 ){
+ /* Search through to find the first varint with value 1. This is the
+ ** start of the next columns hits. */
+ int i = 0;
+ int iStart = 0;
- pNew->bFiltered = 1;
- pIter->flags = FTS5_SEGITER_ONETERM;
- if( pData->szLeaf>0 ){
- pIter->pLeaf = pData;
- pIter->iLeafOffset = fts5GetVarint(pData->p, (u64*)&pIter->iRowid);
- pIter->iEndofDoclist = pData->nn;
- pNew->aFirst[1].iFirst = 1;
- if( bDesc ){
- pNew->bRev = 1;
- pIter->flags |= FTS5_SEGITER_REVERSE;
- fts5SegIterReverseInitPage(p, pIter);
+ if( pCtx->eState==2 ){
+ int iCol;
+ fts5FastGetVarint32(pChunk, i, iCol);
+ if( fts5IndexColsetTest(pCtx->pColset, iCol) ){
+ pCtx->eState = 1;
+ fts5BufferSafeAppendVarint(pCtx->pBuf, 1);
}else{
- fts5SegIterLoadNPos(p, pIter);
+ pCtx->eState = 0;
}
- pData = 0;
- }else{
- pNew->base.bEof = 1;
}
- fts5SegIterSetNext(p, pIter);
-
- *ppOut = pNew;
- }
-
- fts5DataRelease(pData);
-}
-
-/*
-** Return true if the iterator is at EOF or if an error has occurred.
-** False otherwise.
-*/
-static int fts5MultiIterEof(Fts5Index *p, Fts5Iter *pIter){
- assert( p->rc
- || (pIter->aSeg[ pIter->aFirst[1].iFirst ].pLeaf==0)==pIter->base.bEof
- );
- return (p->rc || pIter->base.bEof);
-}
-
-/*
-** Return the rowid of the entry that the iterator currently points
-** to. If the iterator points to EOF when this function is called the
-** results are undefined.
-*/
-static i64 fts5MultiIterRowid(Fts5Iter *pIter){
- assert( pIter->aSeg[ pIter->aFirst[1].iFirst ].pLeaf );
- return pIter->aSeg[ pIter->aFirst[1].iFirst ].iRowid;
-}
-/*
-** Move the iterator to the next entry at or following iMatch.
-*/
-static void fts5MultiIterNextFrom(
- Fts5Index *p,
- Fts5Iter *pIter,
- i64 iMatch
-){
- while( 1 ){
- i64 iRowid;
- fts5MultiIterNext(p, pIter, 1, iMatch);
- if( fts5MultiIterEof(p, pIter) ) break;
- iRowid = fts5MultiIterRowid(pIter);
- if( pIter->bRev==0 && iRowid>=iMatch ) break;
- if( pIter->bRev!=0 && iRowid<=iMatch ) break;
+ do {
+ while( i<nChunk && pChunk[i]!=0x01 ){
+ while( pChunk[i] & 0x80 ) i++;
+ i++;
+ }
+ if( pCtx->eState ){
+ fts5BufferSafeAppendBlob(pCtx->pBuf, &pChunk[iStart], i-iStart);
+ }
+ if( i<nChunk ){
+ int iCol;
+ iStart = i;
+ i++;
+ if( i>=nChunk ){
+ pCtx->eState = 2;
+ }else{
+ fts5FastGetVarint32(pChunk, i, iCol);
+ pCtx->eState = fts5IndexColsetTest(pCtx->pColset, iCol);
+ if( pCtx->eState ){
+ fts5BufferSafeAppendBlob(pCtx->pBuf, &pChunk[iStart], i-iStart);
+ iStart = i;
+ }
+ }
+ }
+ }while( i<nChunk );
}
}
-/*
-** Return a pointer to a buffer containing the term associated with the
-** entry that the iterator currently points to.
-*/
-static const u8 *fts5MultiIterTerm(Fts5Iter *pIter, int *pn){
- Fts5SegIter *p = &pIter->aSeg[ pIter->aFirst[1].iFirst ];
- *pn = p->term.n;
- return p->term.p;
-}
-
static void fts5ChunkIterate(
Fts5Index *p, /* Index object */
Fts5SegIter *pSeg, /* Poslist of this iterator */
}
}
-
-
/*
-** Allocate a new segment-id for the structure pStruct. The new segment
-** id must be between 1 and 65335 inclusive, and must not be used by
-** any currently existing segment. If a free segment id cannot be found,
-** SQLITE_FULL is returned.
-**
-** If an error has already occurred, this function is a no-op. 0 is
-** returned in this case.
+** Iterator pIter currently points to a valid entry (not EOF). This
+** function appends the position list data for the current entry to
+** buffer pBuf. It does not make a copy of the position-list size
+** field.
*/
-static int fts5AllocateSegid(Fts5Index *p, Fts5Structure *pStruct){
+static void fts5SegiterPoslist(
+ Fts5Index *p,
+ Fts5SegIter *pSeg,
+ Fts5Colset *pColset,
+ Fts5Buffer *pBuf
+){
+ if( 0==fts5BufferGrow(&p->rc, pBuf, pSeg->nPos) ){
+ if( pColset==0 ){
+ fts5ChunkIterate(p, pSeg, (void*)pBuf, fts5PoslistCallback);
+ }else{
+ if( p->pConfig->eDetail==FTS5_DETAIL_FULL ){
+ PoslistCallbackCtx sCtx;
+ sCtx.pBuf = pBuf;
+ sCtx.pColset = pColset;
+ sCtx.eState = fts5IndexColsetTest(pColset, 0);
+ assert( sCtx.eState==0 || sCtx.eState==1 );
+ fts5ChunkIterate(p, pSeg, (void*)&sCtx, fts5PoslistFilterCallback);
+ }else{
+ PoslistOffsetsCtx sCtx;
+ memset(&sCtx, 0, sizeof(sCtx));
+ sCtx.pBuf = pBuf;
+ sCtx.pColset = pColset;
+ fts5ChunkIterate(p, pSeg, (void*)&sCtx, fts5PoslistOffsetsCallback);
+ }
+ }
+ }
+}
+
+/*
+** IN/OUT parameter (*pa) points to a position list n bytes in size. If
+** the position list contains entries for column iCol, then (*pa) is set
+** to point to the sub-position-list for that column and the number of
+** bytes in it returned. Or, if the argument position list does not
+** contain any entries for column iCol, return 0.
+*/
+static int fts5IndexExtractCol(
+ const u8 **pa, /* IN/OUT: Pointer to poslist */
+ int n, /* IN: Size of poslist in bytes */
+ int iCol /* Column to extract from poslist */
+){
+ int iCurrent = 0; /* Anything before the first 0x01 is col 0 */
+ const u8 *p = *pa;
+ const u8 *pEnd = &p[n]; /* One byte past end of position list */
+
+ while( iCol>iCurrent ){
+ /* Advance pointer p until it points to pEnd or an 0x01 byte that is
+ ** not part of a varint. Note that it is not possible for a negative
+ ** or extremely large varint to occur within an uncorrupted position
+ ** list. So the last byte of each varint may be assumed to have a clear
+ ** 0x80 bit. */
+ while( *p!=0x01 ){
+ while( *p++ & 0x80 );
+ if( p>=pEnd ) return 0;
+ }
+ *pa = p++;
+ iCurrent = *p++;
+ if( iCurrent & 0x80 ){
+ p--;
+ p += fts5GetVarint32(p, iCurrent);
+ }
+ }
+ if( iCol!=iCurrent ) return 0;
+
+ /* Advance pointer p until it points to pEnd or an 0x01 byte that is
+ ** not part of a varint */
+ while( p<pEnd && *p!=0x01 ){
+ while( *p++ & 0x80 );
+ }
+
+ return p - (*pa);
+}
+
+static int fts5IndexExtractColset (
+ Fts5Colset *pColset, /* Colset to filter on */
+ const u8 *pPos, int nPos, /* Position list */
+ Fts5Buffer *pBuf /* Output buffer */
+){
+ int rc = SQLITE_OK;
+ int i;
+
+ fts5BufferZero(pBuf);
+ for(i=0; i<pColset->nCol; i++){
+ const u8 *pSub = pPos;
+ int nSub = fts5IndexExtractCol(&pSub, nPos, pColset->aiCol[i]);
+ if( nSub ){
+ fts5BufferAppendBlob(&rc, pBuf, nSub, pSub);
+ }
+ }
+ return rc;
+}
+
+/*
+** xSetOutputs callback used by detail=none tables.
+*/
+static void fts5IterSetOutputs_None(Fts5Iter *pIter, Fts5SegIter *pSeg){
+ assert( pIter->pIndex->pConfig->eDetail==FTS5_DETAIL_NONE );
+ pIter->base.iRowid = pSeg->iRowid;
+ pIter->base.nData = pSeg->nPos;
+}
+
+/*
+** xSetOutputs callback used by detail=full and detail=col tables when no
+** column filters are specified.
+*/
+static void fts5IterSetOutputs_Nocolset(Fts5Iter *pIter, Fts5SegIter *pSeg){
+ pIter->base.iRowid = pSeg->iRowid;
+ pIter->base.nData = pSeg->nPos;
+
+ assert( pIter->pIndex->pConfig->eDetail!=FTS5_DETAIL_NONE );
+ assert( pIter->pColset==0 || pIter->bFiltered );
+
+ if( pSeg->iLeafOffset+pSeg->nPos<=pSeg->pLeaf->szLeaf ){
+ /* All data is stored on the current page. Populate the output
+ ** variables to point into the body of the page object. */
+ pIter->base.pData = &pSeg->pLeaf->p[pSeg->iLeafOffset];
+ }else{
+ /* The data is distributed over two or more pages. Copy it into the
+ ** Fts5Iter.poslist buffer and then set the output pointer to point
+ ** to this buffer. */
+ fts5BufferZero(&pIter->poslist);
+ fts5SegiterPoslist(pIter->pIndex, pSeg, 0, &pIter->poslist);
+ pIter->base.pData = pIter->poslist.p;
+ }
+}
+
+/*
+** xSetOutputs callback used by detail=col when there is a column filter
+** and there are 100 or more columns. Also called as a fallback from
+** fts5IterSetOutputs_Col100 if the column-list spans more than one page.
+*/
+static void fts5IterSetOutputs_Col(Fts5Iter *pIter, Fts5SegIter *pSeg){
+ fts5BufferZero(&pIter->poslist);
+ fts5SegiterPoslist(pIter->pIndex, pSeg, pIter->pColset, &pIter->poslist);
+ pIter->base.iRowid = pSeg->iRowid;
+ pIter->base.pData = pIter->poslist.p;
+ pIter->base.nData = pIter->poslist.n;
+}
+
+/*
+** xSetOutputs callback used when:
+**
+** * detail=col,
+** * there is a column filter, and
+** * the table contains 100 or fewer columns.
+**
+** The last point is to ensure all column numbers are stored as
+** single-byte varints.
+*/
+static void fts5IterSetOutputs_Col100(Fts5Iter *pIter, Fts5SegIter *pSeg){
+
+ assert( pIter->pIndex->pConfig->eDetail==FTS5_DETAIL_COLUMNS );
+ assert( pIter->pColset );
+
+ if( pSeg->iLeafOffset+pSeg->nPos>pSeg->pLeaf->szLeaf ){
+ fts5IterSetOutputs_Col(pIter, pSeg);
+ }else{
+ u8 *a = (u8*)&pSeg->pLeaf->p[pSeg->iLeafOffset];
+ u8 *pEnd = (u8*)&a[pSeg->nPos];
+ int iPrev = 0;
+ int *aiCol = pIter->pColset->aiCol;
+ int *aiColEnd = &aiCol[pIter->pColset->nCol];
+
+ u8 *aOut = pIter->poslist.p;
+ int iPrevOut = 0;
+
+ pIter->base.iRowid = pSeg->iRowid;
+
+ while( a<pEnd ){
+ iPrev += (int)a++[0] - 2;
+ while( *aiCol<iPrev ){
+ aiCol++;
+ if( aiCol==aiColEnd ) goto setoutputs_col_out;
+ }
+ if( *aiCol==iPrev ){
+ *aOut++ = (iPrev - iPrevOut) + 2;
+ iPrevOut = iPrev;
+ }
+ }
+
+setoutputs_col_out:
+ pIter->base.pData = pIter->poslist.p;
+ pIter->base.nData = aOut - pIter->poslist.p;
+ }
+}
+
+/*
+** xSetOutputs callback used by detail=full when there is a column filter.
+*/
+static void fts5IterSetOutputs_Full(Fts5Iter *pIter, Fts5SegIter *pSeg){
+ Fts5Colset *pColset = pIter->pColset;
+ pIter->base.iRowid = pSeg->iRowid;
+
+ assert( pIter->pIndex->pConfig->eDetail==FTS5_DETAIL_FULL );
+ assert( pColset );
+
+ if( pSeg->iLeafOffset+pSeg->nPos<=pSeg->pLeaf->szLeaf ){
+ /* All data is stored on the current page. Populate the output
+ ** variables to point into the body of the page object. */
+ const u8 *a = &pSeg->pLeaf->p[pSeg->iLeafOffset];
+ if( pColset->nCol==1 ){
+ pIter->base.nData = fts5IndexExtractCol(&a, pSeg->nPos,pColset->aiCol[0]);
+ pIter->base.pData = a;
+ }else{
+ fts5BufferZero(&pIter->poslist);
+ fts5IndexExtractColset(pColset, a, pSeg->nPos, &pIter->poslist);
+ pIter->base.pData = pIter->poslist.p;
+ pIter->base.nData = pIter->poslist.n;
+ }
+ }else{
+ /* The data is distributed over two or more pages. Copy it into the
+ ** Fts5Iter.poslist buffer and then set the output pointer to point
+ ** to this buffer. */
+ fts5BufferZero(&pIter->poslist);
+ fts5SegiterPoslist(pIter->pIndex, pSeg, pColset, &pIter->poslist);
+ pIter->base.pData = pIter->poslist.p;
+ pIter->base.nData = pIter->poslist.n;
+ }
+}
+
+static void fts5IterSetOutputCb(int *pRc, Fts5Iter *pIter){
+ if( *pRc==SQLITE_OK ){
+ Fts5Config *pConfig = pIter->pIndex->pConfig;
+ if( pConfig->eDetail==FTS5_DETAIL_NONE ){
+ pIter->xSetOutputs = fts5IterSetOutputs_None;
+ }
+
+ else if( pIter->pColset==0 || pIter->bFiltered ){
+ pIter->xSetOutputs = fts5IterSetOutputs_Nocolset;
+ }
+
+ else if( pConfig->eDetail==FTS5_DETAIL_FULL ){
+ pIter->xSetOutputs = fts5IterSetOutputs_Full;
+ }
+
+ else{
+ assert( pConfig->eDetail==FTS5_DETAIL_COLUMNS );
+ if( pConfig->nCol<=100 ){
+ pIter->xSetOutputs = fts5IterSetOutputs_Col100;
+ sqlite3Fts5BufferSize(pRc, &pIter->poslist, pConfig->nCol);
+ }else{
+ pIter->xSetOutputs = fts5IterSetOutputs_Col;
+ }
+ }
+ }
+}
+
+
+/*
+** Allocate a new Fts5Iter object.
+**
+** The new object will be used to iterate through data in structure pStruct.
+** If iLevel is -ve, then all data in all segments is merged. Or, if iLevel
+** is zero or greater, data from the first nSegment segments on level iLevel
+** is merged.
+**
+** The iterator initially points to the first term/rowid entry in the
+** iterated data.
+*/
+static void fts5MultiIterNew(
+ Fts5Index *p, /* FTS5 backend to iterate within */
+ Fts5Structure *pStruct, /* Structure of specific index */
+ int flags, /* FTS5INDEX_QUERY_XXX flags */
+ Fts5Colset *pColset, /* Colset to filter on (or NULL) */
+ const u8 *pTerm, int nTerm, /* Term to seek to (or NULL/0) */
+ int iLevel, /* Level to iterate (-1 for all) */
+ int nSegment, /* Number of segments to merge (iLevel>=0) */
+ Fts5Iter **ppOut /* New object */
+){
+ int nSeg = 0; /* Number of segment-iters in use */
+ int iIter = 0; /* */
+ int iSeg; /* Used to iterate through segments */
+ Fts5Buffer buf = {0,0,0}; /* Buffer used by fts5SegIterSeekInit() */
+ Fts5StructureLevel *pLvl;
+ Fts5Iter *pNew;
+
+ assert( (pTerm==0 && nTerm==0) || iLevel<0 );
+
+ /* Allocate space for the new multi-seg-iterator. */
+ if( p->rc==SQLITE_OK ){
+ if( iLevel<0 ){
+ assert( pStruct->nSegment==fts5StructureCountSegments(pStruct) );
+ nSeg = pStruct->nSegment;
+ nSeg += (p->pHash ? 1 : 0);
+ }else{
+ nSeg = MIN(pStruct->aLevel[iLevel].nSeg, nSegment);
+ }
+ }
+ *ppOut = pNew = fts5MultiIterAlloc(p, nSeg);
+ if( pNew==0 ) return;
+ pNew->bRev = (0!=(flags & FTS5INDEX_QUERY_DESC));
+ pNew->bSkipEmpty = (0!=(flags & FTS5INDEX_QUERY_SKIPEMPTY));
+ pNew->pStruct = pStruct;
+ pNew->pColset = pColset;
+ fts5StructureRef(pStruct);
+ if( (flags & FTS5INDEX_QUERY_NOOUTPUT)==0 ){
+ fts5IterSetOutputCb(&p->rc, pNew);
+ }
+
+ /* Initialize each of the component segment iterators. */
+ if( p->rc==SQLITE_OK ){
+ if( iLevel<0 ){
+ Fts5StructureLevel *pEnd = &pStruct->aLevel[pStruct->nLevel];
+ if( p->pHash ){
+ /* Add a segment iterator for the current contents of the hash table. */
+ Fts5SegIter *pIter = &pNew->aSeg[iIter++];
+ fts5SegIterHashInit(p, pTerm, nTerm, flags, pIter);
+ }
+ for(pLvl=&pStruct->aLevel[0]; pLvl<pEnd; pLvl++){
+ for(iSeg=pLvl->nSeg-1; iSeg>=0; iSeg--){
+ Fts5StructureSegment *pSeg = &pLvl->aSeg[iSeg];
+ Fts5SegIter *pIter = &pNew->aSeg[iIter++];
+ if( pTerm==0 ){
+ fts5SegIterInit(p, pSeg, pIter);
+ }else{
+ fts5SegIterSeekInit(p, &buf, pTerm, nTerm, flags, pSeg, pIter);
+ }
+ }
+ }
+ }else{
+ pLvl = &pStruct->aLevel[iLevel];
+ for(iSeg=nSeg-1; iSeg>=0; iSeg--){
+ fts5SegIterInit(p, &pLvl->aSeg[iSeg], &pNew->aSeg[iIter++]);
+ }
+ }
+ assert( iIter==nSeg );
+ }
+
+ /* If the above was successful, each component iterators now points
+ ** to the first entry in its segment. In this case initialize the
+ ** aFirst[] array. Or, if an error has occurred, free the iterator
+ ** object and set the output variable to NULL. */
+ if( p->rc==SQLITE_OK ){
+ for(iIter=pNew->nSeg-1; iIter>0; iIter--){
+ int iEq;
+ if( (iEq = fts5MultiIterDoCompare(pNew, iIter)) ){
+ Fts5SegIter *pSeg = &pNew->aSeg[iEq];
+ if( p->rc==SQLITE_OK ) pSeg->xNext(p, pSeg, 0);
+ fts5MultiIterAdvanced(p, pNew, iEq, iIter);
+ }
+ }
+ fts5MultiIterSetEof(pNew);
+ fts5AssertMultiIterSetup(p, pNew);
+
+ if( pNew->bSkipEmpty && fts5MultiIterIsEmpty(p, pNew) ){
+ fts5MultiIterNext(p, pNew, 0, 0);
+ }else if( pNew->base.bEof==0 ){
+ Fts5SegIter *pSeg = &pNew->aSeg[pNew->aFirst[1].iFirst];
+ pNew->xSetOutputs(pNew, pSeg);
+ }
+
+ }else{
+ fts5MultiIterFree(p, pNew);
+ *ppOut = 0;
+ }
+ fts5BufferFree(&buf);
+
+}
+
+/*
+** Create an Fts5Iter that iterates through the doclist provided
+** as the second argument.
+*/
+static void fts5MultiIterNew2(
+ Fts5Index *p, /* FTS5 backend to iterate within */
+ Fts5Data *pData, /* Doclist to iterate through */
+ int bDesc, /* True for descending rowid order */
+ Fts5Iter **ppOut /* New object */
+){
+ Fts5Iter *pNew;
+ pNew = fts5MultiIterAlloc(p, 2);
+ if( pNew ){
+ Fts5SegIter *pIter = &pNew->aSeg[1];
+
+ pNew->bFiltered = 1;
+ pIter->flags = FTS5_SEGITER_ONETERM;
+ if( pData->szLeaf>0 ){
+ pIter->pLeaf = pData;
+ pIter->iLeafOffset = fts5GetVarint(pData->p, (u64*)&pIter->iRowid);
+ pIter->iEndofDoclist = pData->nn;
+ pNew->aFirst[1].iFirst = 1;
+ if( bDesc ){
+ pNew->bRev = 1;
+ pIter->flags |= FTS5_SEGITER_REVERSE;
+ fts5SegIterReverseInitPage(p, pIter);
+ }else{
+ fts5SegIterLoadNPos(p, pIter);
+ }
+ pData = 0;
+ }else{
+ pNew->base.bEof = 1;
+ }
+ fts5SegIterSetNext(p, pIter);
+
+ *ppOut = pNew;
+ }
+
+ fts5DataRelease(pData);
+}
+
+/*
+** Return true if the iterator is at EOF or if an error has occurred.
+** False otherwise.
+*/
+static int fts5MultiIterEof(Fts5Index *p, Fts5Iter *pIter){
+ assert( p->rc
+ || (pIter->aSeg[ pIter->aFirst[1].iFirst ].pLeaf==0)==pIter->base.bEof
+ );
+ return (p->rc || pIter->base.bEof);
+}
+
+/*
+** Return the rowid of the entry that the iterator currently points
+** to. If the iterator points to EOF when this function is called the
+** results are undefined.
+*/
+static i64 fts5MultiIterRowid(Fts5Iter *pIter){
+ assert( pIter->aSeg[ pIter->aFirst[1].iFirst ].pLeaf );
+ return pIter->aSeg[ pIter->aFirst[1].iFirst ].iRowid;
+}
+
+/*
+** Move the iterator to the next entry at or following iMatch.
+*/
+static void fts5MultiIterNextFrom(
+ Fts5Index *p,
+ Fts5Iter *pIter,
+ i64 iMatch
+){
+ while( 1 ){
+ i64 iRowid;
+ fts5MultiIterNext(p, pIter, 1, iMatch);
+ if( fts5MultiIterEof(p, pIter) ) break;
+ iRowid = fts5MultiIterRowid(pIter);
+ if( pIter->bRev==0 && iRowid>=iMatch ) break;
+ if( pIter->bRev!=0 && iRowid<=iMatch ) break;
+ }
+}
+
+/*
+** Return a pointer to a buffer containing the term associated with the
+** entry that the iterator currently points to.
+*/
+static const u8 *fts5MultiIterTerm(Fts5Iter *pIter, int *pn){
+ Fts5SegIter *p = &pIter->aSeg[ pIter->aFirst[1].iFirst ];
+ *pn = p->term.n;
+ return p->term.p;
+}
+
+/*
+** Allocate a new segment-id for the structure pStruct. The new segment
+** id must be between 1 and 65335 inclusive, and must not be used by
+** any currently existing segment. If a free segment id cannot be found,
+** SQLITE_FULL is returned.
+**
+** If an error has already occurred, this function is a no-op. 0 is
+** returned in this case.
+*/
+static int fts5AllocateSegid(Fts5Index *p, Fts5Structure *pStruct){
int iSegid = 0;
if( p->rc==SQLITE_OK ){
Fts5Buffer term;
int bOldest; /* True if the output segment is the oldest */
int eDetail = p->pConfig->eDetail;
+ const int flags = FTS5INDEX_QUERY_NOOUTPUT;
assert( iLvl<pStruct->nLevel );
assert( pLvl->nMerge<=pLvl->nSeg );
bOldest = (pLvlOut->nSeg==1 && pStruct->nLevel==iLvl+2);
assert( iLvl>=0 );
- for(fts5MultiIterNew(p, pStruct, 0, 0, 0, 0, iLvl, nInput, &pIter);
+ for(fts5MultiIterNew(p, pStruct, flags, 0, 0, 0, iLvl, nInput, &pIter);
fts5MultiIterEof(p, pIter)==0;
fts5MultiIterNext(p, pIter, 0, 0)
){
}
pNew->nSegment = pLvl->nSeg = nSeg;
}else{
- sqlite3_free(pNew);
- pNew = 0;
- }
- }
-
- if( pNew ){
- int iLvl = pNew->nLevel-1;
- while( p->rc==SQLITE_OK && pNew->aLevel[iLvl].nSeg>0 ){
- int nRem = FTS5_OPT_WORK_UNIT;
- fts5IndexMergeLevel(p, &pNew, iLvl, &nRem);
- }
-
- fts5StructureWrite(p, pNew);
- fts5StructureRelease(pNew);
- }
-
- fts5StructureRelease(pStruct);
- return fts5IndexReturn(p);
-}
-
-int sqlite3Fts5IndexMerge(Fts5Index *p, int nMerge){
- Fts5Structure *pStruct;
-
- pStruct = fts5StructureRead(p);
- if( pStruct && pStruct->nLevel ){
- fts5IndexMerge(p, &pStruct, nMerge);
- fts5StructureWrite(p, pStruct);
- }
- fts5StructureRelease(pStruct);
-
- return fts5IndexReturn(p);
-}
-
-static void fts5PoslistCallback(
- Fts5Index *p,
- void *pContext,
- const u8 *pChunk, int nChunk
-){
- assert_nc( nChunk>=0 );
- if( nChunk>0 ){
- fts5BufferSafeAppendBlob((Fts5Buffer*)pContext, pChunk, nChunk);
- }
-}
-
-typedef struct PoslistCallbackCtx PoslistCallbackCtx;
-struct PoslistCallbackCtx {
- Fts5Buffer *pBuf; /* Append to this buffer */
- Fts5Colset *pColset; /* Restrict matches to this column */
- int eState; /* See above */
-};
-
-typedef struct PoslistOffsetsCtx PoslistOffsetsCtx;
-struct PoslistOffsetsCtx {
- Fts5Buffer *pBuf; /* Append to this buffer */
- Fts5Colset *pColset; /* Restrict matches to this column */
- int iRead;
- int iWrite;
-};
-
-/*
-** TODO: Make this more efficient!
-*/
-static int fts5IndexColsetTest(Fts5Colset *pColset, int iCol){
- int i;
- for(i=0; i<pColset->nCol; i++){
- if( pColset->aiCol[i]==iCol ) return 1;
- }
- return 0;
-}
-
-static void fts5PoslistOffsetsCallback(
- Fts5Index *p,
- void *pContext,
- const u8 *pChunk, int nChunk
-){
- PoslistOffsetsCtx *pCtx = (PoslistOffsetsCtx*)pContext;
- assert_nc( nChunk>=0 );
- if( nChunk>0 ){
- int i = 0;
- while( i<nChunk ){
- int iVal;
- i += fts5GetVarint32(&pChunk[i], iVal);
- iVal += pCtx->iRead - 2;
- pCtx->iRead = iVal;
- if( fts5IndexColsetTest(pCtx->pColset, iVal) ){
- fts5BufferSafeAppendVarint(pCtx->pBuf, iVal + 2 - pCtx->iWrite);
- pCtx->iWrite = iVal;
- }
- }
- }
-}
-
-static void fts5PoslistFilterCallback(
- Fts5Index *p,
- void *pContext,
- const u8 *pChunk, int nChunk
-){
- PoslistCallbackCtx *pCtx = (PoslistCallbackCtx*)pContext;
- assert_nc( nChunk>=0 );
- if( nChunk>0 ){
- /* Search through to find the first varint with value 1. This is the
- ** start of the next columns hits. */
- int i = 0;
- int iStart = 0;
-
- if( pCtx->eState==2 ){
- int iCol;
- fts5FastGetVarint32(pChunk, i, iCol);
- if( fts5IndexColsetTest(pCtx->pColset, iCol) ){
- pCtx->eState = 1;
- fts5BufferSafeAppendVarint(pCtx->pBuf, 1);
- }else{
- pCtx->eState = 0;
- }
- }
-
- do {
- while( i<nChunk && pChunk[i]!=0x01 ){
- while( pChunk[i] & 0x80 ) i++;
- i++;
- }
- if( pCtx->eState ){
- fts5BufferSafeAppendBlob(pCtx->pBuf, &pChunk[iStart], i-iStart);
- }
- if( i<nChunk ){
- int iCol;
- iStart = i;
- i++;
- if( i>=nChunk ){
- pCtx->eState = 2;
- }else{
- fts5FastGetVarint32(pChunk, i, iCol);
- pCtx->eState = fts5IndexColsetTest(pCtx->pColset, iCol);
- if( pCtx->eState ){
- fts5BufferSafeAppendBlob(pCtx->pBuf, &pChunk[iStart], i-iStart);
- iStart = i;
- }
- }
- }
- }while( i<nChunk );
- }
-}
-
-/*
-** Iterator pIter currently points to a valid entry (not EOF). This
-** function appends the position list data for the current entry to
-** buffer pBuf. It does not make a copy of the position-list size
-** field.
-*/
-static void fts5SegiterPoslist(
- Fts5Index *p,
- Fts5SegIter *pSeg,
- Fts5Colset *pColset,
- Fts5Buffer *pBuf
-){
- if( 0==fts5BufferGrow(&p->rc, pBuf, pSeg->nPos) ){
- if( pColset==0 ){
- fts5ChunkIterate(p, pSeg, (void*)pBuf, fts5PoslistCallback);
- }else{
- if( p->pConfig->eDetail==FTS5_DETAIL_FULL ){
- PoslistCallbackCtx sCtx;
- sCtx.pBuf = pBuf;
- sCtx.pColset = pColset;
- sCtx.eState = fts5IndexColsetTest(pColset, 0);
- assert( sCtx.eState==0 || sCtx.eState==1 );
- fts5ChunkIterate(p, pSeg, (void*)&sCtx, fts5PoslistFilterCallback);
- }else{
- PoslistOffsetsCtx sCtx;
- memset(&sCtx, 0, sizeof(sCtx));
- sCtx.pBuf = pBuf;
- sCtx.pColset = pColset;
- fts5ChunkIterate(p, pSeg, (void*)&sCtx, fts5PoslistOffsetsCallback);
- }
+ sqlite3_free(pNew);
+ pNew = 0;
}
}
-}
-
-/*
-** IN/OUT parameter (*pa) points to a position list n bytes in size. If
-** the position list contains entries for column iCol, then (*pa) is set
-** to point to the sub-position-list for that column and the number of
-** bytes in it returned. Or, if the argument position list does not
-** contain any entries for column iCol, return 0.
-*/
-static int fts5IndexExtractCol(
- const u8 **pa, /* IN/OUT: Pointer to poslist */
- int n, /* IN: Size of poslist in bytes */
- int iCol /* Column to extract from poslist */
-){
- int iCurrent = 0; /* Anything before the first 0x01 is col 0 */
- const u8 *p = *pa;
- const u8 *pEnd = &p[n]; /* One byte past end of position list */
- while( iCol>iCurrent ){
- /* Advance pointer p until it points to pEnd or an 0x01 byte that is
- ** not part of a varint. Note that it is not possible for a negative
- ** or extremely large varint to occur within an uncorrupted position
- ** list. So the last byte of each varint may be assumed to have a clear
- ** 0x80 bit. */
- while( *p!=0x01 ){
- while( *p++ & 0x80 );
- if( p>=pEnd ) return 0;
- }
- *pa = p++;
- iCurrent = *p++;
- if( iCurrent & 0x80 ){
- p--;
- p += fts5GetVarint32(p, iCurrent);
+ if( pNew ){
+ int iLvl = pNew->nLevel-1;
+ while( p->rc==SQLITE_OK && pNew->aLevel[iLvl].nSeg>0 ){
+ int nRem = FTS5_OPT_WORK_UNIT;
+ fts5IndexMergeLevel(p, &pNew, iLvl, &nRem);
}
+
+ fts5StructureWrite(p, pNew);
+ fts5StructureRelease(pNew);
}
- if( iCol!=iCurrent ) return 0;
- /* Advance pointer p until it points to pEnd or an 0x01 byte that is
- ** not part of a varint */
- while( p<pEnd && *p!=0x01 ){
- while( *p++ & 0x80 );
+ fts5StructureRelease(pStruct);
+ return fts5IndexReturn(p);
+}
+
+int sqlite3Fts5IndexMerge(Fts5Index *p, int nMerge){
+ Fts5Structure *pStruct;
+
+ pStruct = fts5StructureRead(p);
+ if( pStruct && pStruct->nLevel ){
+ fts5IndexMerge(p, &pStruct, nMerge);
+ fts5StructureWrite(p, pStruct);
}
+ fts5StructureRelease(pStruct);
- return p - (*pa);
+ return fts5IndexReturn(p);
}
-static int fts5AppendRowid(
+static void fts5AppendRowid(
Fts5Index *p,
i64 iDelta,
Fts5Iter *pMulti,
- Fts5Colset *pColset,
Fts5Buffer *pBuf
){
fts5BufferAppendVarint(&p->rc, pBuf, iDelta);
- return 0;
}
-/*
-** Iterator pMulti currently points to a valid entry (not EOF). This
-** function appends the following to buffer pBuf:
-**
-** * The varint iDelta, and
-** * the position list that currently points to, including the size field.
-**
-** If argument pColset is NULL, then the position list is filtered according
-** to pColset before being appended to the buffer. If this means there are
-** no entries in the position list, nothing is appended to the buffer (not
-** even iDelta).
-**
-** If an error occurs, an error code is left in p->rc.
-*/
-static int fts5AppendPoslist(
+static void fts5AppendPoslist(
Fts5Index *p,
i64 iDelta,
Fts5Iter *pMulti,
- Fts5Colset *pColset,
Fts5Buffer *pBuf
){
- if( p->rc==SQLITE_OK ){
- Fts5SegIter *pSeg = &pMulti->aSeg[ pMulti->aFirst[1].iFirst ];
- assert( fts5MultiIterEof(p, pMulti)==0 );
- assert( pSeg->nPos>0 );
- if( 0==fts5BufferGrow(&p->rc, pBuf, pSeg->nPos+9+9) ){
- if( p->pConfig->eDetail==FTS5_DETAIL_FULL
- && pSeg->iLeafOffset+pSeg->nPos<=pSeg->pLeaf->szLeaf
- && (pColset==0 || pColset->nCol==1)
- ){
- const u8 *pPos = &pSeg->pLeaf->p[pSeg->iLeafOffset];
- int nPos;
- if( pColset ){
- nPos = fts5IndexExtractCol(&pPos, pSeg->nPos, pColset->aiCol[0]);
- if( nPos==0 ) return 1;
- }else{
- nPos = pSeg->nPos;
- }
- assert( nPos>0 );
- fts5BufferSafeAppendVarint(pBuf, iDelta);
- fts5BufferSafeAppendVarint(pBuf, nPos*2);
- fts5BufferSafeAppendBlob(pBuf, pPos, nPos);
- }else{
- int iSv1;
- int iSv2;
- int iData;
-
- /* Append iDelta */
- iSv1 = pBuf->n;
- fts5BufferSafeAppendVarint(pBuf, iDelta);
-
- /* WRITEPOSLISTSIZE */
- iSv2 = pBuf->n;
- fts5BufferSafeAppendVarint(pBuf, pSeg->nPos*2);
- iData = pBuf->n;
-
- fts5SegiterPoslist(p, pSeg, pColset, pBuf);
-
- if( pColset ){
- int nActual = pBuf->n - iData;
- if( nActual!=pSeg->nPos ){
- if( nActual==0 ){
- pBuf->n = iSv1;
- return 1;
- }else{
- int nReq = sqlite3Fts5GetVarintLen((u32)(nActual*2));
- while( iSv2<(iData-nReq) ){ pBuf->p[iSv2++] = 0x80; }
- sqlite3Fts5PutVarint(&pBuf->p[iSv2], nActual*2);
- }
- }
- }
- }
- }
+ int nData = pMulti->base.nData;
+ assert( nData>0 );
+ if( p->rc==SQLITE_OK && 0==fts5BufferGrow(&p->rc, pBuf, nData+9+9) ){
+ fts5BufferSafeAppendVarint(pBuf, iDelta);
+ fts5BufferSafeAppendVarint(pBuf, nData*2);
+ fts5BufferSafeAppendBlob(pBuf, pMulti->base.pData, nData);
}
-
- return 0;
}
const int nBuf = 32;
void (*xMerge)(Fts5Index*, Fts5Buffer*, Fts5Buffer*);
- int (*xAppend)(Fts5Index*, i64, Fts5Iter*, Fts5Colset*, Fts5Buffer*);
+ void (*xAppend)(Fts5Index*, i64, Fts5Iter*, Fts5Buffer*);
if( p->pConfig->eDetail==FTS5_DETAIL_NONE ){
xMerge = fts5MergeRowidLists;
xAppend = fts5AppendRowid;
pStruct = fts5StructureRead(p);
if( aBuf && pStruct ){
- const int flags = FTS5INDEX_QUERY_SCAN;
+ const int flags = FTS5INDEX_QUERY_SCAN
+ | FTS5INDEX_QUERY_SKIPEMPTY
+ | FTS5INDEX_QUERY_NOOUTPUT;
int i;
i64 iLastRowid = 0;
Fts5Iter *p1 = 0; /* Iterator used to gather data from index */
int bNewTerm = 1;
memset(&doclist, 0, sizeof(doclist));
- for(fts5MultiIterNew(p, pStruct, 1, flags, pToken, nToken, -1, 0, &p1);
+ fts5MultiIterNew(p, pStruct, flags, pColset, pToken, nToken, -1, 0, &p1);
+ fts5IterSetOutputCb(&p->rc, p1);
+ for( /* no-op */ ;
fts5MultiIterEof(p, p1)==0;
fts5MultiIterNext2(p, p1, &bNewTerm)
){
- i64 iRowid = fts5MultiIterRowid(p1);
- int nTerm;
- const u8 *pTerm = fts5MultiIterTerm(p1, &nTerm);
+ Fts5SegIter *pSeg = &p1->aSeg[ p1->aFirst[1].iFirst ];
+ int nTerm = pSeg->term.n;
+ const u8 *pTerm = pSeg->term.p;
+ p1->xSetOutputs(p1, pSeg);
+
assert_nc( memcmp(pToken, pTerm, MIN(nToken, nTerm))<=0 );
if( bNewTerm ){
if( nTerm<nToken || memcmp(pToken, pTerm, nToken) ) break;
}
- if( doclist.n>0 && iRowid<=iLastRowid ){
+ if( p1->base.nData==0 ) continue;
+
+ if( p1->base.iRowid<=iLastRowid && doclist.n>0 ){
for(i=0; p->rc==SQLITE_OK && doclist.n; i++){
assert( i<nBuf );
if( aBuf[i].n==0 ){
iLastRowid = 0;
}
- if( !xAppend(p, iRowid-iLastRowid, p1, pColset, &doclist) ){
- iLastRowid = iRowid;
- }
+ xAppend(p, p1->base.iRowid-iLastRowid, p1, &doclist);
+ iLastRowid = p1->base.iRowid;
}
for(i=0; i<nBuf; i++){
return rc;
}
-
-static int fts5IndexExtractColset (
- Fts5Colset *pColset, /* Colset to filter on */
- const u8 *pPos, int nPos, /* Position list */
- Fts5Buffer *pBuf /* Output buffer */
-){
- int rc = SQLITE_OK;
- int i;
-
- fts5BufferZero(pBuf);
- for(i=0; i<pColset->nCol; i++){
- const u8 *pSub = pPos;
- int nSub = fts5IndexExtractCol(&pSub, nPos, pColset->aiCol[i]);
- if( nSub ){
- fts5BufferAppendBlob(&rc, pBuf, nSub, pSub);
- }
- }
- return rc;
-}
-
-/*
-** xSetOutputs callback used by detail=none tables.
-*/
-static void fts5IterSetOutputs_None(Fts5Iter *pIter, Fts5SegIter *pSeg){
- assert( pIter->pIndex->pConfig->eDetail==FTS5_DETAIL_NONE );
- pIter->base.iRowid = pSeg->iRowid;
- pIter->base.nData = pSeg->nPos;
-}
-
-/*
-** xSetOutputs callback used by detail=full and detail=col tables when no
-** column filters are specified.
-*/
-static void fts5IterSetOutputs_Nocolset(Fts5Iter *pIter, Fts5SegIter *pSeg){
- pIter->base.iRowid = pSeg->iRowid;
- pIter->base.nData = pSeg->nPos;
-
- assert( pIter->pIndex->pConfig->eDetail!=FTS5_DETAIL_NONE );
- assert( pIter->pColset==0 || pIter->bFiltered );
-
- if( pSeg->iLeafOffset+pSeg->nPos<=pSeg->pLeaf->szLeaf ){
- /* All data is stored on the current page. Populate the output
- ** variables to point into the body of the page object. */
- pIter->base.pData = &pSeg->pLeaf->p[pSeg->iLeafOffset];
- }else{
- /* The data is distributed over two or more pages. Copy it into the
- ** Fts5Iter.poslist buffer and then set the output pointer to point
- ** to this buffer. */
- fts5BufferZero(&pIter->poslist);
- fts5SegiterPoslist(pIter->pIndex, pSeg, 0, &pIter->poslist);
- pIter->base.pData = pIter->poslist.p;
- }
-}
-
-/*
-** xSetOutputs callback used by detail=col when there is a column filter
-** and there are 100 or more columns. Also called as a fallback from
-** fts5IterSetOutputs_Col100 if the column-list spans more than one page.
-*/
-static void fts5IterSetOutputs_Col(Fts5Iter *pIter, Fts5SegIter *pSeg){
- fts5BufferZero(&pIter->poslist);
- fts5SegiterPoslist(pIter->pIndex, pSeg, pIter->pColset, &pIter->poslist);
- pIter->base.iRowid = pSeg->iRowid;
- pIter->base.pData = pIter->poslist.p;
- pIter->base.nData = pIter->poslist.n;
-}
-
-/*
-** xSetOutputs callback used when:
-**
-** * detail=col,
-** * there is a column filter, and
-** * the table contains 100 or fewer columns.
-**
-** The last point is to ensure all column numbers are stored as
-** single-byte varints.
-*/
-static void fts5IterSetOutputs_Col100(Fts5Iter *pIter, Fts5SegIter *pSeg){
-
- assert( pIter->pIndex->pConfig->eDetail==FTS5_DETAIL_COLUMNS );
- assert( pIter->pColset );
-
- if( pSeg->iLeafOffset+pSeg->nPos>pSeg->pLeaf->szLeaf ){
- fts5IterSetOutputs_Col(pIter, pSeg);
- }else{
- u8 *a = (u8*)&pSeg->pLeaf->p[pSeg->iLeafOffset];
- u8 *pEnd = (u8*)&a[pSeg->nPos];
- int iPrev = 0;
- int *aiCol = pIter->pColset->aiCol;
- int *aiColEnd = &aiCol[pIter->pColset->nCol];
-
- u8 *aOut = pIter->poslist.p;
- int iPrevOut = 0;
-
- pIter->base.iRowid = pSeg->iRowid;
-
- while( a<pEnd ){
- iPrev += (int)a++[0] - 2;
- while( *aiCol<iPrev ){
- aiCol++;
- if( aiCol==aiColEnd ) goto setoutputs_col_out;
- }
- if( *aiCol==iPrev ){
- *aOut++ = (iPrev - iPrevOut) + 2;
- iPrevOut = iPrev;
- }
- }
-
-setoutputs_col_out:
- pIter->base.pData = pIter->poslist.p;
- pIter->base.nData = aOut - pIter->poslist.p;
- }
-}
-
-/*
-** xSetOutputs callback used by detail=full when there is a column filter.
-*/
-static void fts5IterSetOutputs_Full(Fts5Iter *pIter, Fts5SegIter *pSeg){
- Fts5Colset *pColset = pIter->pColset;
- pIter->base.iRowid = pSeg->iRowid;
-
- assert( pIter->pIndex->pConfig->eDetail==FTS5_DETAIL_FULL );
- assert( pColset );
-
- if( pSeg->iLeafOffset+pSeg->nPos<=pSeg->pLeaf->szLeaf ){
- /* All data is stored on the current page. Populate the output
- ** variables to point into the body of the page object. */
- const u8 *a = &pSeg->pLeaf->p[pSeg->iLeafOffset];
- if( pColset->nCol==1 ){
- pIter->base.nData = fts5IndexExtractCol(&a, pSeg->nPos,pColset->aiCol[0]);
- pIter->base.pData = a;
- }else{
- fts5BufferZero(&pIter->poslist);
- fts5IndexExtractColset(pColset, a, pSeg->nPos, &pIter->poslist);
- pIter->base.pData = pIter->poslist.p;
- pIter->base.nData = pIter->poslist.n;
- }
- }else{
- /* The data is distributed over two or more pages. Copy it into the
- ** Fts5Iter.poslist buffer and then set the output pointer to point
- ** to this buffer. */
- fts5BufferZero(&pIter->poslist);
- fts5SegiterPoslist(pIter->pIndex, pSeg, pColset, &pIter->poslist);
- pIter->base.pData = pIter->poslist.p;
- pIter->base.nData = pIter->poslist.n;
- }
-}
-
-static void fts5IterSetOutputCb(int *pRc, Fts5Iter *pIter){
- Fts5Config *pConfig = pIter->pIndex->pConfig;
- if( pConfig->eDetail==FTS5_DETAIL_NONE ){
- pIter->xSetOutputs = fts5IterSetOutputs_None;
- }
-
- else if( pIter->pColset==0 || pIter->bFiltered ){
- pIter->xSetOutputs = fts5IterSetOutputs_Nocolset;
- }
-
- else if( pConfig->eDetail==FTS5_DETAIL_FULL ){
- pIter->xSetOutputs = fts5IterSetOutputs_Full;
- }
-
- else{
- assert( pConfig->eDetail==FTS5_DETAIL_COLUMNS );
- if( pConfig->nCol<=100 ){
- pIter->xSetOutputs = fts5IterSetOutputs_Col100;
- sqlite3Fts5BufferSize(pRc, &pIter->poslist, pConfig->nCol);
- }else{
- pIter->xSetOutputs = fts5IterSetOutputs_Col;
- }
- }
-}
-
/*
** Open a new iterator to iterate though all rowid that match the
** specified token or token prefix.
Fts5Structure *pStruct = fts5StructureRead(p);
buf.p[0] = (u8)(FTS5_MAIN_PREFIX + iIdx);
if( pStruct ){
- fts5MultiIterNew(p, pStruct, 1, flags, buf.p, nToken+1, -1, 0, &pRet);
+ fts5MultiIterNew(p, pStruct, flags | FTS5INDEX_QUERY_SKIPEMPTY,
+ pColset, buf.p, nToken+1, -1, 0, &pRet
+ );
fts5StructureRelease(pStruct);
}
}else{
int bDesc = (flags & FTS5INDEX_QUERY_DESC)!=0;
buf.p[0] = FTS5_MAIN_PREFIX;
fts5SetupPrefixIter(p, bDesc, buf.p, nToken+1, pColset, &pRet);
- }
-
- if( p->rc==SQLITE_OK ){
- Fts5SegIter *pSeg = &pRet->aSeg[pRet->aFirst[1].iFirst];
- pRet->pColset = pColset;
fts5IterSetOutputCb(&p->rc, pRet);
- if( p->rc==SQLITE_OK && pSeg->pLeaf ) pRet->xSetOutputs(pRet, pSeg);
+ if( p->rc==SQLITE_OK ){
+ Fts5SegIter *pSeg = &pRet->aSeg[pRet->aFirst[1].iFirst];
+ if( p->rc==SQLITE_OK && pSeg->pLeaf ) pRet->xSetOutputs(pRet, pSeg);
+ }
}
+
if( p->rc ){
sqlite3Fts5IterClose(&pRet->base);
pRet = 0;
u64 cksum3 = 0; /* Checksum based on contents of indexes */
Fts5Buffer term = {0,0,0}; /* Buffer used to hold most recent term */
#endif
+ const int flags = FTS5INDEX_QUERY_NOOUTPUT;
/* Load the FTS index structure */
pStruct = fts5StructureRead(p);
** same term is performed. cksum3 is calculated based on the entries
** extracted by these queries.
*/
- for(fts5MultiIterNew(p, pStruct, 0, 0, 0, 0, -1, 0, &pIter);
+ for(fts5MultiIterNew(p, pStruct, flags, 0, 0, 0, -1, 0, &pIter);
fts5MultiIterEof(p, pIter)==0;
fts5MultiIterNext(p, pIter, 0, 0)
){
projects / thirdparty / sqlite.git / commitdiff
