#include "fts5Int.h"
typedef struct Fts5VocabTable Fts5VocabTable;
typedef struct Fts5VocabCursor Fts5VocabCursor;
struct Fts5VocabTable {
sqlite3_vtab base;
char *zFts5Tbl;
char *zFts5Db;
sqlite3 *db;
Fts5Global *pGlobal;
int eType;
unsigned bBusy;
};
struct Fts5VocabCursor {
sqlite3_vtab_cursor base;
sqlite3_stmt *pStmt;
Fts5Table *pFts5;
int bEof;
Fts5IndexIter *pIter;
void *pStruct;
int nLeTerm;
char *zLeTerm;
int iCol;
i64 *aCnt;
i64 *aDoc;
i64 rowid;
Fts5Buffer term;
i64 iInstPos;
int iInstOff;
};
#define FTS5_VOCAB_COL 0
#define FTS5_VOCAB_ROW 1
#define FTS5_VOCAB_INSTANCE 2
#define FTS5_VOCAB_COL_SCHEMA "term, col, doc, cnt"
#define FTS5_VOCAB_ROW_SCHEMA "term, doc, cnt"
#define FTS5_VOCAB_INST_SCHEMA "term, doc, col, offset"
#define FTS5_VOCAB_TERM_EQ 0x01
#define FTS5_VOCAB_TERM_GE 0x02
#define FTS5_VOCAB_TERM_LE 0x04
static int fts5VocabTableType(const char *zType, char **pzErr, int *peType){
int rc = SQLITE_OK;
char *zCopy = sqlite3Fts5Strndup(&rc, zType, -1);
if( rc==SQLITE_OK ){
sqlite3Fts5Dequote(zCopy);
if( sqlite3_stricmp(zCopy, "col")==0 ){
*peType = FTS5_VOCAB_COL;
}else
if( sqlite3_stricmp(zCopy, "row")==0 ){
*peType = FTS5_VOCAB_ROW;
}else
if( sqlite3_stricmp(zCopy, "instance")==0 ){
*peType = FTS5_VOCAB_INSTANCE;
}else
{
*pzErr = sqlite3_mprintf("fts5vocab: unknown table type: %Q", zCopy);
rc = SQLITE_ERROR;
}
sqlite3_free(zCopy);
}
return rc;
}
static int fts5VocabDisconnectMethod(sqlite3_vtab *pVtab){
Fts5VocabTable *pTab = (Fts5VocabTable*)pVtab;
sqlite3_free(pTab);
return SQLITE_OK;
}
static int fts5VocabDestroyMethod(sqlite3_vtab *pVtab){
Fts5VocabTable *pTab = (Fts5VocabTable*)pVtab;
sqlite3_free(pTab);
return SQLITE_OK;
}
static int fts5VocabInitVtab(
sqlite3 *db,
void *pAux,
int argc,
const char * const *argv,
sqlite3_vtab **ppVTab,
char **pzErr
){
const char *azSchema[] = {
"CREATE TABlE vocab(" FTS5_VOCAB_COL_SCHEMA ")",
"CREATE TABlE vocab(" FTS5_VOCAB_ROW_SCHEMA ")",
"CREATE TABlE vocab(" FTS5_VOCAB_INST_SCHEMA ")"
};
Fts5VocabTable *pRet = 0;
int rc = SQLITE_OK;
int bDb;
bDb = (argc==6 && strlen(argv[1])==4 && memcmp("temp", argv[1], 4)==0);
if( argc!=5 && bDb==0 ){
*pzErr = sqlite3_mprintf("wrong number of vtable arguments");
rc = SQLITE_ERROR;
}else{
int nByte;
const char *zDb = bDb ? argv[3] : argv[1];
const char *zTab = bDb ? argv[4] : argv[3];
const char *zType = bDb ? argv[5] : argv[4];
int nDb = (int)strlen(zDb)+1;
int nTab = (int)strlen(zTab)+1;
int eType = 0;
rc = fts5VocabTableType(zType, pzErr, &eType);
if( rc==SQLITE_OK ){
assert( eType>=0 && eType<ArraySize(azSchema) );
rc = sqlite3_declare_vtab(db, azSchema[eType]);
}
nByte = sizeof(Fts5VocabTable) + nDb + nTab;
pRet = sqlite3Fts5MallocZero(&rc, nByte);
if( pRet ){
pRet->pGlobal = (Fts5Global*)pAux;
pRet->eType = eType;
pRet->db = db;
pRet->zFts5Tbl = (char*)&pRet[1];
pRet->zFts5Db = &pRet->zFts5Tbl[nTab];
memcpy(pRet->zFts5Tbl, zTab, nTab);
memcpy(pRet->zFts5Db, zDb, nDb);
sqlite3Fts5Dequote(pRet->zFts5Tbl);
sqlite3Fts5Dequote(pRet->zFts5Db);
}
}
*ppVTab = (sqlite3_vtab*)pRet;
return rc;
}
static int fts5VocabConnectMethod(
sqlite3 *db,
void *pAux,
int argc,
const char * const *argv,
sqlite3_vtab **ppVtab,
char **pzErr
){
return fts5VocabInitVtab(db, pAux, argc, argv, ppVtab, pzErr);
}
static int fts5VocabCreateMethod(
sqlite3 *db,
void *pAux,
int argc,
const char * const *argv,
sqlite3_vtab **ppVtab,
char **pzErr
){
return fts5VocabInitVtab(db, pAux, argc, argv, ppVtab, pzErr);
}
static int fts5VocabBestIndexMethod(
sqlite3_vtab *pUnused,
sqlite3_index_info *pInfo
){
int i;
int iTermEq = -1;
int iTermGe = -1;
int iTermLe = -1;
int idxNum = 0;
int nArg = 0;
UNUSED_PARAM(pUnused);
for(i=0; i<pInfo->nConstraint; i++){
struct sqlite3_index_constraint *p = &pInfo->aConstraint[i];
if( p->usable==0 ) continue;
if( p->iColumn==0 ){
if( p->op==SQLITE_INDEX_CONSTRAINT_EQ ) iTermEq = i;
if( p->op==SQLITE_INDEX_CONSTRAINT_LE ) iTermLe = i;
if( p->op==SQLITE_INDEX_CONSTRAINT_LT ) iTermLe = i;
if( p->op==SQLITE_INDEX_CONSTRAINT_GE ) iTermGe = i;
if( p->op==SQLITE_INDEX_CONSTRAINT_GT ) iTermGe = i;
}
}
if( iTermEq>=0 ){
idxNum |= FTS5_VOCAB_TERM_EQ;
pInfo->aConstraintUsage[iTermEq].argvIndex = ++nArg;
pInfo->estimatedCost = 100;
}else{
pInfo->estimatedCost = 1000000;
if( iTermGe>=0 ){
idxNum |= FTS5_VOCAB_TERM_GE;
pInfo->aConstraintUsage[iTermGe].argvIndex = ++nArg;
pInfo->estimatedCost = pInfo->estimatedCost / 2;
}
if( iTermLe>=0 ){
idxNum |= FTS5_VOCAB_TERM_LE;
pInfo->aConstraintUsage[iTermLe].argvIndex = ++nArg;
pInfo->estimatedCost = pInfo->estimatedCost / 2;
}
}
if( pInfo->nOrderBy==1
&& pInfo->aOrderBy[0].iColumn==0
&& pInfo->aOrderBy[0].desc==0
){
pInfo->orderByConsumed = 1;
}
pInfo->idxNum = idxNum;
return SQLITE_OK;
}
static int fts5VocabOpenMethod(
sqlite3_vtab *pVTab,
sqlite3_vtab_cursor **ppCsr
){
Fts5VocabTable *pTab = (Fts5VocabTable*)pVTab;
Fts5Table *pFts5 = 0;
Fts5VocabCursor *pCsr = 0;
int rc = SQLITE_OK;
sqlite3_stmt *pStmt = 0;
char *zSql = 0;
if( pTab->bBusy ){
pVTab->zErrMsg = sqlite3_mprintf(
"recursive definition for %s.%s", pTab->zFts5Db, pTab->zFts5Tbl
);
return SQLITE_ERROR;
}
zSql = sqlite3Fts5Mprintf(&rc,
"SELECT t.%Q FROM %Q.%Q AS t WHERE t.%Q MATCH '*id'",
pTab->zFts5Tbl, pTab->zFts5Db, pTab->zFts5Tbl, pTab->zFts5Tbl
);
if( zSql ){
rc = sqlite3_prepare_v2(pTab->db, zSql, -1, &pStmt, 0);
}
sqlite3_free(zSql);
assert( rc==SQLITE_OK || pStmt==0 );
if( rc==SQLITE_ERROR ) rc = SQLITE_OK;
pTab->bBusy = 1;
if( pStmt && sqlite3_step(pStmt)==SQLITE_ROW ){
i64 iId = sqlite3_column_int64(pStmt, 0);
pFts5 = sqlite3Fts5TableFromCsrid(pTab->pGlobal, iId);
}
pTab->bBusy = 0;
if( rc==SQLITE_OK ){
if( pFts5==0 ){
rc = sqlite3_finalize(pStmt);
pStmt = 0;
if( rc==SQLITE_OK ){
pVTab->zErrMsg = sqlite3_mprintf(
"no such fts5 table: %s.%s", pTab->zFts5Db, pTab->zFts5Tbl
);
rc = SQLITE_ERROR;
}
}else{
rc = sqlite3Fts5FlushToDisk(pFts5);
}
}
if( rc==SQLITE_OK ){
i64 nByte = pFts5->pConfig->nCol * sizeof(i64)*2 + sizeof(Fts5VocabCursor);
pCsr = (Fts5VocabCursor*)sqlite3Fts5MallocZero(&rc, nByte);
}
if( pCsr ){
pCsr->pFts5 = pFts5;
pCsr->pStmt = pStmt;
pCsr->aCnt = (i64*)&pCsr[1];
pCsr->aDoc = &pCsr->aCnt[pFts5->pConfig->nCol];
}else{
sqlite3_finalize(pStmt);
}
*ppCsr = (sqlite3_vtab_cursor*)pCsr;
return rc;
}
static void fts5VocabResetCursor(Fts5VocabCursor *pCsr){
pCsr->rowid = 0;
sqlite3Fts5IterClose(pCsr->pIter);
sqlite3Fts5StructureRelease(pCsr->pStruct);
pCsr->pStruct = 0;
pCsr->pIter = 0;
sqlite3_free(pCsr->zLeTerm);
pCsr->nLeTerm = -1;
pCsr->zLeTerm = 0;
pCsr->bEof = 0;
}
static int fts5VocabCloseMethod(sqlite3_vtab_cursor *pCursor){
Fts5VocabCursor *pCsr = (Fts5VocabCursor*)pCursor;
fts5VocabResetCursor(pCsr);
sqlite3Fts5BufferFree(&pCsr->term);
sqlite3_finalize(pCsr->pStmt);
sqlite3_free(pCsr);
return SQLITE_OK;
}
static int fts5VocabInstanceNewTerm(Fts5VocabCursor *pCsr){
int rc = SQLITE_OK;
if( sqlite3Fts5IterEof(pCsr->pIter) ){
pCsr->bEof = 1;
}else{
const char *zTerm;
int nTerm;
zTerm = sqlite3Fts5IterTerm(pCsr->pIter, &nTerm);
if( pCsr->nLeTerm>=0 ){
int nCmp = MIN(nTerm, pCsr->nLeTerm);
int bCmp = memcmp(pCsr->zLeTerm, zTerm, nCmp);
if( bCmp<0 || (bCmp==0 && pCsr->nLeTerm<nTerm) ){
pCsr->bEof = 1;
}
}
sqlite3Fts5BufferSet(&rc, &pCsr->term, nTerm, (const u8*)zTerm);
}
return rc;
}
static int fts5VocabInstanceNext(Fts5VocabCursor *pCsr){
int eDetail = pCsr->pFts5->pConfig->eDetail;
int rc = SQLITE_OK;
Fts5IndexIter *pIter = pCsr->pIter;
i64 *pp = &pCsr->iInstPos;
int *po = &pCsr->iInstOff;
assert( sqlite3Fts5IterEof(pIter)==0 );
assert( pCsr->bEof==0 );
while( eDetail==FTS5_DETAIL_NONE
|| sqlite3Fts5PoslistNext64(pIter->pData, pIter->nData, po, pp)
){
pCsr->iInstPos = 0;
pCsr->iInstOff = 0;
rc = sqlite3Fts5IterNextScan(pCsr->pIter);
if( rc==SQLITE_OK ){
rc = fts5VocabInstanceNewTerm(pCsr);
if( pCsr->bEof || eDetail==FTS5_DETAIL_NONE ) break;
}
if( rc ){
pCsr->bEof = 1;
break;
}
}
return rc;
}
static int fts5VocabNextMethod(sqlite3_vtab_cursor *pCursor){
Fts5VocabCursor *pCsr = (Fts5VocabCursor*)pCursor;
Fts5VocabTable *pTab = (Fts5VocabTable*)pCursor->pVtab;
int nCol = pCsr->pFts5->pConfig->nCol;
int rc;
rc = sqlite3Fts5StructureTest(pCsr->pFts5->pIndex, pCsr->pStruct);
if( rc!=SQLITE_OK ) return rc;
pCsr->rowid++;
if( pTab->eType==FTS5_VOCAB_INSTANCE ){
return fts5VocabInstanceNext(pCsr);
}
if( pTab->eType==FTS5_VOCAB_COL ){
for(pCsr->iCol++; pCsr->iCol<nCol; pCsr->iCol++){
if( pCsr->aDoc[pCsr->iCol] ) break;
}
}
if( pTab->eType!=FTS5_VOCAB_COL || pCsr->iCol>=nCol ){
if( sqlite3Fts5IterEof(pCsr->pIter) ){
pCsr->bEof = 1;
}else{
const char *zTerm;
int nTerm;
zTerm = sqlite3Fts5IterTerm(pCsr->pIter, &nTerm);
assert( nTerm>=0 );
if( pCsr->nLeTerm>=0 ){
int nCmp = MIN(nTerm, pCsr->nLeTerm);
int bCmp = memcmp(pCsr->zLeTerm, zTerm, nCmp);
if( bCmp<0 || (bCmp==0 && pCsr->nLeTerm<nTerm) ){
pCsr->bEof = 1;
return SQLITE_OK;
}
}
sqlite3Fts5BufferSet(&rc, &pCsr->term, nTerm, (const u8*)zTerm);
memset(pCsr->aCnt, 0, nCol * sizeof(i64));
memset(pCsr->aDoc, 0, nCol * sizeof(i64));
pCsr->iCol = 0;
assert( pTab->eType==FTS5_VOCAB_COL || pTab->eType==FTS5_VOCAB_ROW );
while( rc==SQLITE_OK ){
int eDetail = pCsr->pFts5->pConfig->eDetail;
const u8 *pPos; int nPos;
i64 iPos = 0;
int iOff = 0;
pPos = pCsr->pIter->pData;
nPos = pCsr->pIter->nData;
switch( pTab->eType ){
case FTS5_VOCAB_ROW:
if( eDetail==FTS5_DETAIL_FULL ){
while( 0==sqlite3Fts5PoslistNext64(pPos, nPos, &iOff, &iPos) ){
pCsr->aCnt[0]++;
}
}
pCsr->aDoc[0]++;
break;
case FTS5_VOCAB_COL:
if( eDetail==FTS5_DETAIL_FULL ){
int iCol = -1;
while( 0==sqlite3Fts5PoslistNext64(pPos, nPos, &iOff, &iPos) ){
int ii = FTS5_POS2COLUMN(iPos);
if( iCol!=ii ){
if( ii>=nCol ){
rc = FTS5_CORRUPT;
break;
}
pCsr->aDoc[ii]++;
iCol = ii;
}
pCsr->aCnt[ii]++;
}
}else if( eDetail==FTS5_DETAIL_COLUMNS ){
while( 0==sqlite3Fts5PoslistNext64(pPos, nPos, &iOff,&iPos) ){
assert_nc( iPos>=0 && iPos<nCol );
if( iPos>=nCol ){
rc = FTS5_CORRUPT;
break;
}
pCsr->aDoc[iPos]++;
}
}else{
assert( eDetail==FTS5_DETAIL_NONE );
pCsr->aDoc[0]++;
}
break;
default:
assert( pTab->eType==FTS5_VOCAB_INSTANCE );
break;
}
if( rc==SQLITE_OK ){
rc = sqlite3Fts5IterNextScan(pCsr->pIter);
}
if( pTab->eType==FTS5_VOCAB_INSTANCE ) break;
if( rc==SQLITE_OK ){
zTerm = sqlite3Fts5IterTerm(pCsr->pIter, &nTerm);
if( nTerm!=pCsr->term.n
|| (nTerm>0 && memcmp(zTerm, pCsr->term.p, nTerm))
){
break;
}
if( sqlite3Fts5IterEof(pCsr->pIter) ) break;
}
}
}
}
if( rc==SQLITE_OK && pCsr->bEof==0 && pTab->eType==FTS5_VOCAB_COL ){
for(; pCsr->iCol<nCol && pCsr->aDoc[pCsr->iCol]==0; pCsr->iCol++);
if( pCsr->iCol==nCol ){
rc = FTS5_CORRUPT;
}
}
return rc;
}
static int fts5VocabFilterMethod(
sqlite3_vtab_cursor *pCursor,
int idxNum,
const char *zUnused,
int nUnused,
sqlite3_value **apVal
){
Fts5VocabTable *pTab = (Fts5VocabTable*)pCursor->pVtab;
Fts5VocabCursor *pCsr = (Fts5VocabCursor*)pCursor;
int eType = pTab->eType;
int rc = SQLITE_OK;
int iVal = 0;
int f = FTS5INDEX_QUERY_SCAN;
const char *zTerm = 0;
int nTerm = 0;
sqlite3_value *pEq = 0;
sqlite3_value *pGe = 0;
sqlite3_value *pLe = 0;
UNUSED_PARAM2(zUnused, nUnused);
fts5VocabResetCursor(pCsr);
if( idxNum & FTS5_VOCAB_TERM_EQ ) pEq = apVal[iVal++];
if( idxNum & FTS5_VOCAB_TERM_GE ) pGe = apVal[iVal++];
if( idxNum & FTS5_VOCAB_TERM_LE ) pLe = apVal[iVal++];
if( pEq ){
zTerm = (const char *)sqlite3_value_text(pEq);
nTerm = sqlite3_value_bytes(pEq);
f = 0;
}else{
if( pGe ){
zTerm = (const char *)sqlite3_value_text(pGe);
nTerm = sqlite3_value_bytes(pGe);
}
if( pLe ){
const char *zCopy = (const char *)sqlite3_value_text(pLe);
if( zCopy==0 ) zCopy = "";
pCsr->nLeTerm = sqlite3_value_bytes(pLe);
pCsr->zLeTerm = sqlite3_malloc(pCsr->nLeTerm+1);
if( pCsr->zLeTerm==0 ){
rc = SQLITE_NOMEM;
}else{
memcpy(pCsr->zLeTerm, zCopy, pCsr->nLeTerm+1);
}
}
}
if( rc==SQLITE_OK ){
Fts5Index *pIndex = pCsr->pFts5->pIndex;
rc = sqlite3Fts5IndexQuery(pIndex, zTerm, nTerm, f, 0, &pCsr->pIter);
if( rc==SQLITE_OK ){
pCsr->pStruct = sqlite3Fts5StructureRef(pIndex);
}
}
if( rc==SQLITE_OK && eType==FTS5_VOCAB_INSTANCE ){
rc = fts5VocabInstanceNewTerm(pCsr);
}
if( rc==SQLITE_OK && !pCsr->bEof
&& (eType!=FTS5_VOCAB_INSTANCE
|| pCsr->pFts5->pConfig->eDetail!=FTS5_DETAIL_NONE)
){
rc = fts5VocabNextMethod(pCursor);
}
return rc;
}
static int fts5VocabEofMethod(sqlite3_vtab_cursor *pCursor){
Fts5VocabCursor *pCsr = (Fts5VocabCursor*)pCursor;
return pCsr->bEof;
}
static int fts5VocabColumnMethod(
sqlite3_vtab_cursor *pCursor,
sqlite3_context *pCtx,
int iCol
){
Fts5VocabCursor *pCsr = (Fts5VocabCursor*)pCursor;
int eDetail = pCsr->pFts5->pConfig->eDetail;
int eType = ((Fts5VocabTable*)(pCursor->pVtab))->eType;
i64 iVal = 0;
if( iCol==0 ){
sqlite3_result_text(
pCtx, (const char*)pCsr->term.p, pCsr->term.n, SQLITE_TRANSIENT
);
}else if( eType==FTS5_VOCAB_COL ){
assert( iCol==1 || iCol==2 || iCol==3 );
if( iCol==1 ){
if( eDetail!=FTS5_DETAIL_NONE ){
const char *z = pCsr->pFts5->pConfig->azCol[pCsr->iCol];
sqlite3_result_text(pCtx, z, -1, SQLITE_STATIC);
}
}else if( iCol==2 ){
iVal = pCsr->aDoc[pCsr->iCol];
}else{
iVal = pCsr->aCnt[pCsr->iCol];
}
}else if( eType==FTS5_VOCAB_ROW ){
assert( iCol==1 || iCol==2 );
if( iCol==1 ){
iVal = pCsr->aDoc[0];
}else{
iVal = pCsr->aCnt[0];
}
}else{
assert( eType==FTS5_VOCAB_INSTANCE );
switch( iCol ){
case 1:
sqlite3_result_int64(pCtx, pCsr->pIter->iRowid);
break;
case 2: {
int ii = -1;
if( eDetail==FTS5_DETAIL_FULL ){
ii = FTS5_POS2COLUMN(pCsr->iInstPos);
}else if( eDetail==FTS5_DETAIL_COLUMNS ){
ii = (int)pCsr->iInstPos;
}
if( ii>=0 && ii<pCsr->pFts5->pConfig->nCol ){
const char *z = pCsr->pFts5->pConfig->azCol[ii];
sqlite3_result_text(pCtx, z, -1, SQLITE_STATIC);
}
break;
}
default: {
assert( iCol==3 );
if( eDetail==FTS5_DETAIL_FULL ){
int ii = FTS5_POS2OFFSET(pCsr->iInstPos);
sqlite3_result_int(pCtx, ii);
}
break;
}
}
}
if( iVal>0 ) sqlite3_result_int64(pCtx, iVal);
return SQLITE_OK;
}
static int fts5VocabRowidMethod(
sqlite3_vtab_cursor *pCursor,
sqlite_int64 *pRowid
){
Fts5VocabCursor *pCsr = (Fts5VocabCursor*)pCursor;
*pRowid = pCsr->rowid;
return SQLITE_OK;
}
int sqlite3Fts5VocabInit(Fts5Global *pGlobal, sqlite3 *db){
static const sqlite3_module fts5Vocab = {
2,
fts5VocabCreateMethod,
fts5VocabConnectMethod,
fts5VocabBestIndexMethod,
fts5VocabDisconnectMethod,
fts5VocabDestroyMethod,
fts5VocabOpenMethod,
fts5VocabCloseMethod,
fts5VocabFilterMethod,
fts5VocabNextMethod,
fts5VocabEofMethod,
fts5VocabColumnMethod,
fts5VocabRowidMethod,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0
};
void *p = (void*)pGlobal;
return sqlite3_create_module_v2(db, "fts5vocab", &fts5Vocab, p, 0);
}