SQLite

typedef unsigned char  u8;
typedef unsigned int   u32;
typedef unsigned short u16;
typedef sqlite3_int64 i64;
typedef sqlite3_uint64 u64;

#define ArraySize(x) (sizeof(x) / sizeof(x[0]))

#define testcase(x)
#define ALWAYS(x) 1
#define NEVER(x) 0

#define MIN(x,y) (((x) < (y)) ? (x) : (y))
#define MAX(x,y) (((x) > (y)) ? (x) : (y))

** The various operations on open token or token prefix iterators opened
** using sqlite3Fts5IndexQuery().
*/
int sqlite3Fts5IterEof(Fts5IndexIter*);
int sqlite3Fts5IterNext(Fts5IndexIter*);
int sqlite3Fts5IterNextFrom(Fts5IndexIter*, i64 iMatch);
i64 sqlite3Fts5IterRowid(Fts5IndexIter*);
int sqlite3Fts5IterPoslist(Fts5IndexIter*,Fts5Colset*, const u8**, int*, i64*);
int sqlite3Fts5IterPoslistBuffer(Fts5IndexIter *pIter, Fts5Buffer *pBuf);

/*
** Close an iterator opened by sqlite3Fts5IndexQuery().
*/
void sqlite3Fts5IterClose(Fts5IndexIter*);

/*
** This interface is used by the fts5vocab module.
*/
const char *sqlite3Fts5IterTerm(Fts5IndexIter*, int*);
int sqlite3Fts5IterNextScan(Fts5IndexIter*);

int sqlite3Fts5IterCollist(Fts5IndexIter*, const u8 **, int*);


/*
** Insert or remove data to or from the index. Each time a document is 
** added to or removed from the index, this function is called one or more
** times.
**
** For an insert, it must be called once for each token in the new document.

    { "snippet",   0, fts5SnippetFunction, 0 },
    { "highlight", 0, fts5HighlightFunction, 0 },
    { "bm25",      0, fts5Bm25Function,    0 },
  };
  int rc = SQLITE_OK;             /* Return code */
  int i;                          /* To iterate through builtin functions */

  for(i=0; rc==SQLITE_OK && i<(int)ArraySize(aBuiltin); i++){
    rc = pApi->xCreateFunction(pApi,
        aBuiltin[i].zFunc,
        aBuiltin[i].pUserData,
        aBuiltin[i].xFunc,
        aBuiltin[i].xDestroy
    );
  }

  const char *pTerm, int nTerm, 
  int *pbPresent
){
  int rc = SQLITE_OK;
  *pbPresent = 0;
  if( p ){
    int i;
    int hash = 13;
    Fts5TermsetEntry *pEntry;

    /* Calculate a hash value for this term. This is the same hash checksum
    ** used by the fts5_hash.c module. This is not important for correct
    ** operation of the module, but is necessary to ensure that some tests
    ** designed to produce hash table collisions really do work.  */
    for(i=nTerm-1; i>=0; i--){
      hash = (hash << 3) ^ hash ^ pTerm[i];
    }
    hash = (hash << 3) ^ hash ^ iIdx;
    hash = hash % ArraySize(p->apHash);

    for(pEntry=p->apHash[hash]; pEntry; pEntry=pEntry->pNext){
      if( pEntry->iIdx==iIdx 
          && pEntry->nTerm==nTerm 
          && memcmp(pEntry->pTerm, pTerm, nTerm)==0 
        ){
        *pbPresent = 1;
        break;
      }
    }

    if( pEntry==0 ){
      pEntry = sqlite3Fts5MallocZero(&rc, sizeof(Fts5TermsetEntry) + nTerm);

** Argument pTerm must be a synonym iterator.
*/
static int fts5ExprSynonymList(
  Fts5ExprTerm *pTerm, 
  int bCollist, 
  Fts5Colset *pColset,
  i64 iRowid,
  int *pbDel,                     /* OUT: Caller should sqlite3_free(*pa) */
  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 && sqlite3Fts5IterRowid(pIter)==iRowid ){
      const u8 *a;
      int n;

      if( 0 && bCollist ){
        rc = sqlite3Fts5IterCollist(pIter, &a, &n);
      }else{
        i64 dummy;
#if 0
        rc = sqlite3Fts5IterPoslist(pIter, pColset, &a, &n, &dummy);
#endif
        a = pIter->pData;
        n = pIter->nData;
      }

      if( rc!=SQLITE_OK ) goto synonym_poslist_out;
      if( n==0 ) continue;
      if( nIter==nAlloc ){
        int nByte = sizeof(Fts5PoslistReader) * nAlloc * 2;
        Fts5PoslistReader *aNew = (Fts5PoslistReader*)sqlite3_malloc(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(a, n, &aIter[nIter]);
      assert( aIter[nIter].bEof==0 );
      nIter++;
    }
  }

  assert( *pbDel==0 );
  if( nIter==1 ){
    *pa = (u8*)aIter[0].a;
    *pn = aIter[0].n;
  }else{
    Fts5PoslistWriter writer = {0};
    Fts5Buffer buf = {0,0,0};
    i64 iPrev = -1;

    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(&buf, &writer, iMin);
      iPrev = iMin;
    }
    if( rc ){
      sqlite3_free(buf.p);
    }else{
      *pa = buf.p;
      *pn = buf.n;
      *pbDel = 1;
    }
  }

 synonym_poslist_out:
  if( aIter!=aStatic ) sqlite3_free(aIter);
  return rc;
}

  int i;
  int rc = SQLITE_OK;
  
  fts5BufferZero(&pPhrase->poslist);

  /* If the aStatic[] array is not large enough, allocate a large array
  ** using sqlite3_malloc(). This approach could be improved upon. */
  if( pPhrase->nTerm>(int)ArraySize(aStatic) ){
    int nByte = sizeof(Fts5PoslistReader) * pPhrase->nTerm;
    aIter = (Fts5PoslistReader*)sqlite3_malloc(nByte);
    if( !aIter ) return SQLITE_NOMEM;
  }
  memset(aIter, 0, sizeof(Fts5PoslistReader) * pPhrase->nTerm);

  /* Initialize a term iterator for each term in the phrase */
  for(i=0; i<pPhrase->nTerm; i++){
    Fts5ExprTerm *pTerm = &pPhrase->aTerm[i];
    i64 dummy;
    int n = 0;
    int bFlag = 0;
    const u8 *a = 0;
    if( pTerm->pSynonym ){

      rc = fts5ExprSynonymList(
          pTerm, 0, pColset, pNode->iRowid, &bFlag, (u8**)&a, &n
      );





    }else{
      Fts5IndexIter *pIter = pTerm->pIter;
#if 0
      rc = sqlite3Fts5IterPoslist(pTerm->pIter, pColset, &a, &n, &dummy);
#endif
      a = pIter->pData;
      n = pIter->nData;
    }
    if( rc!=SQLITE_OK ) goto ismatch_out;
    sqlite3Fts5PoslistReaderInit(a, n, &aIter[i]);
    aIter[i].bFlag = (u8)bFlag;
    if( aIter[i].bEof ) goto ismatch_out;
  }

  while( 1 ){
    int bMatch;

  int rc = *pRc;
  int bMatch;

  assert( pNear->nPhrase>1 );

  /* If the aStatic[] array is not large enough, allocate a large array
  ** using sqlite3_malloc(). This approach could be improved upon. */
  if( pNear->nPhrase>(int)ArraySize(aStatic) ){
    int nByte = sizeof(Fts5NearTrimmer) * pNear->nPhrase;
    a = (Fts5NearTrimmer*)sqlite3Fts5MallocZero(&rc, nByte);
  }else{
    memset(aStatic, 0, sizeof(aStatic));
  }
  if( rc!=SQLITE_OK ){
    *pRc = rc;

    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);

        sqlite3_free(pSyn);
      }
    }
    if( pPhrase->poslist.nSpace>0 ) fts5BufferFree(&pPhrase->poslist);
    sqlite3_free(pPhrase);
  }
}

  /* If an error has already occurred, this is a no-op */
  if( pCtx->rc!=SQLITE_OK ) return pCtx->rc;

  assert( pPhrase==0 || pPhrase->nTerm>0 );
  if( pPhrase && (tflags & FTS5_TOKEN_COLOCATED) ){
    Fts5ExprTerm *pSyn;
    int nByte = sizeof(Fts5ExprTerm) + nToken+1;
    pSyn = (Fts5ExprTerm*)sqlite3_malloc(nByte);
    if( pSyn==0 ){
      rc = SQLITE_NOMEM;
    }else{
      memset(pSyn, 0, nByte);
      pSyn->zTerm = (char*)&pSyn[1];
      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 ){

    { "fts5_isalnum",  fts5ExprIsAlnum },
    { "fts5_fold",     fts5ExprFold },
  };
  int i;
  int rc = SQLITE_OK;
  void *pCtx = (void*)pGlobal;

  for(i=0; rc==SQLITE_OK && i<(int)ArraySize(aFunc); i++){
    struct Fts5ExprFunc *p = &aFunc[i];
    rc = sqlite3_create_function(db, p->z, -1, SQLITE_UTF8, pCtx, p->x, 0, 0);
  }

  /* Avoid a warning indicating that sqlite3Fts5ParserTrace() is unused */
#ifndef NDEBUG
  (void)sqlite3Fts5ParserTrace;

  if( pNode->bEof==0 
   && pNode->iRowid==pExpr->pRoot->iRowid 
   && pPhrase->poslist.n>0
  ){
    Fts5ExprTerm *pTerm = &pPhrase->aTerm[0];
    if( pTerm->pSynonym ){
      int bDel = 0;
      u8 *a;
      rc = fts5ExprSynonymList(
          pTerm, 1, 0, pNode->iRowid, &bDel, &a, pnCollist
      );
      if( bDel ){
        sqlite3Fts5BufferSet(&rc, &pPhrase->poslist, *pnCollist, a);
        *ppCollist = pPhrase->poslist.p;
        sqlite3_free(a);
      }else{
        *ppCollist = a;
      }
    }else{
      *ppCollist = pPhrase->aTerm[0].pIter->pData;
      *pnCollist = pPhrase->aTerm[0].pIter->nData;
    }
  }else{
    *ppCollist = 0;
    *pnCollist = 0;
  }

  return rc;
}

const char *sqlite3Fts5IterTerm(Fts5IndexIter *pIndexIter, int *pn){
  int n;
  const char *z = (const char*)fts5MultiIterTerm((Fts5Iter*)pIndexIter, &n);
  *pn = n-1;
  return &z[1];
}


/*
** Return a pointer to a buffer containing a copy of the position list for
** the current entry. Output variable *pn is set to the size of the buffer 
** in bytes before returning.
**
** The returned position list does not include the "number of bytes" varint
** field that starts the position list on disk.
*/
int sqlite3Fts5IterPoslist(
  Fts5IndexIter *pIndexIter, 
  Fts5Colset *pColset,            /* Column filter (or NULL) */
  const u8 **pp,                  /* OUT: Pointer to position-list data */
  int *pn,                        /* OUT: Size of position-list in bytes */
  i64 *piRowid                    /* OUT: Current rowid */
){
  Fts5Iter *pIter = (Fts5Iter*)pIndexIter;
  Fts5SegIter *pSeg = &pIter->aSeg[ pIter->aFirst[1].iFirst ];
  int eDetail = pIter->pIndex->pConfig->eDetail;

  assert( pIter->pIndex->rc==SQLITE_OK );
  *piRowid = pSeg->iRowid;
  if( eDetail==FTS5_DETAIL_NONE ){
    *pn = pSeg->nPos;
  }else
  if( eDetail==FTS5_DETAIL_FULL 
   && pSeg->iLeafOffset+pSeg->nPos<=pSeg->pLeaf->szLeaf 
  ){
    u8 *pPos = &pSeg->pLeaf->p[pSeg->iLeafOffset];
    if( pColset==0 || pIter->bFiltered ){
      *pn = pSeg->nPos;
      *pp = pPos;
    }else if( pColset->nCol==1 ){
      *pp = pPos;
      *pn = fts5IndexExtractCol(pp, pSeg->nPos, pColset->aiCol[0]);
    }else{
      fts5BufferZero(&pIter->poslist);
      fts5IndexExtractColset(pColset, pPos, pSeg->nPos, &pIter->poslist);
      *pp = pIter->poslist.p;
      *pn = pIter->poslist.n;
    }
  }else{
    fts5BufferZero(&pIter->poslist);
    fts5SegiterPoslist(pIter->pIndex, pSeg, pColset, &pIter->poslist);
    if( eDetail==FTS5_DETAIL_FULL ){
      *pp = pIter->poslist.p;
    }
    *pn = pIter->poslist.n;
  }
  return fts5IndexReturn(pIter->pIndex);
}

int sqlite3Fts5IterCollist(
  Fts5IndexIter *pIndexIter, 
  const u8 **pp,                  /* OUT: Pointer to position-list data */
  int *pn                         /* OUT: Size of position-list in bytes */
){
  Fts5Iter *pIter = (Fts5Iter*)pIndexIter;
  assert( pIter->pIndex->pConfig->eDetail==FTS5_DETAIL_COLUMNS );
  *pp = pIter->poslist.p;
  *pn = pIter->poslist.n;
  return SQLITE_OK;
}

/*
** This function is similar to sqlite3Fts5IterPoslist(), except that it
** copies the position list into the buffer supplied as the second 
** argument.
*/
int sqlite3Fts5IterPoslistBuffer(Fts5IndexIter *pIndexIter, Fts5Buffer *pBuf){
  Fts5Iter *pIter = (Fts5Iter*)pIndexIter;
  Fts5Index *p = pIter->pIndex;
  Fts5SegIter *pSeg = &pIter->aSeg[ pIter->aFirst[1].iFirst ];
  assert( p->rc==SQLITE_OK );
  fts5BufferZero(pBuf);
  fts5SegiterPoslist(p, pSeg, 0, pBuf);
  return fts5IndexReturn(p);
}

/*
** Close an iterator opened by an earlier call to sqlite3Fts5IndexQuery().
*/
void sqlite3Fts5IterClose(Fts5IndexIter *pIndexIter){
  if( pIndexIter ){
    Fts5Iter *pIter = (Fts5Iter*)pIndexIter;
    Fts5Index *pIndex = pIter->pIndex;

  aColMap[1] = pConfig->nCol;
  aColMap[2] = pConfig->nCol+1;

  /* Set idxFlags flags for all WHERE clause terms that will be used. */
  for(i=0; i<pInfo->nConstraint; i++){
    struct sqlite3_index_constraint *p = &pInfo->aConstraint[i];
    int j;
    for(j=0; j<(int)ArraySize(aConstraint); j++){
      struct Constraint *pC = &aConstraint[j];
      if( p->iColumn==aColMap[pC->iCol] && p->op & pC->op ){
        if( p->usable ){
          pC->iConsIndex = i;
          idxFlags |= pC->fts5op;
        }else if( j==0 ){
          /* As there exists an unusable MATCH constraint this is an 

    pInfo->estimatedCost = bHasMatch ? 750.0 : 750000.0;
  }else{
    pInfo->estimatedCost = bHasMatch ? 1000.0 : 1000000.0;
  }

  /* Assign argvIndex values to each constraint in use. */
  iNext = 1;
  for(i=0; i<(int)ArraySize(aConstraint); i++){
    struct Constraint *pC = &aConstraint[i];
    if( pC->iConsIndex>=0 ){
      pInfo->aConstraintUsage[pC->iConsIndex].argvIndex = iNext++;
      pInfo->aConstraintUsage[pC->iConsIndex].omit = (unsigned char)pC->omit;
    }
  }

*/
int sqlite3Fts5StorageClose(Fts5Storage *p){
  int rc = SQLITE_OK;
  if( p ){
    int i;

    /* Finalize all SQL statements */
    for(i=0; i<(int)ArraySize(p->aStmt); i++){
      sqlite3_finalize(p->aStmt[i]);
    }

    sqlite3_free(p);
  }
  return rc;
}

    { "ascii",     {fts5AsciiCreate, fts5AsciiDelete, fts5AsciiTokenize }},
    { "porter",    {fts5PorterCreate, fts5PorterDelete, fts5PorterTokenize }},
  };
  
  int rc = SQLITE_OK;             /* Return code */
  int i;                          /* To iterate through builtin functions */

  for(i=0; rc==SQLITE_OK && i<(int)ArraySize(aBuiltin); i++){
    rc = pApi->xCreateTokenizer(pApi,
        aBuiltin[i].zName,
        (void*)pApi,
        &aBuiltin[i].x,
        0
    );
  }

    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<sizeof(azSchema)/sizeof(azSchema[0]) );
      rc = sqlite3_declare_vtab(db, azSchema[eType]);
    }

    nByte = sizeof(Fts5VocabTable) + nDb + nTab;
    pRet = sqlite3Fts5MallocZero(&rc, nByte);
    if( pRet ){
      pRet->pGlobal = (Fts5Global*)pAux;

      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 ){
        i64 dummy;
        const u8 *pPos; int nPos;   /* Position list */
        i64 iPos = 0;               /* 64-bit position read from poslist */
        int iOff = 0;               /* Current offset within position list */

        pPos = pCsr->pIter->pData;
        nPos = pCsr->pIter->nData;
        switch( pCsr->pConfig->eDetail ){

  for {set i 0} {$i < [$cmd xPhraseCount]} {incr i} {
    $cmd xPhraseForeach $i c o {
      lappend res $i.$c.$o
    }
  }

  set res

}

proc fts5_test_collist {cmd} {
  set res [list]

  for {set i 0} {$i < [$cmd xPhraseCount]} {incr i} {
    $cmd xPhraseColumnForeach $i c { lappend res $i.$c }

foreach tok {query document} {
foreach_detail_mode $testprefix {

fts5_tclnum_register db
fts5_aux_test_functions db
















proc fts5_rowid {cmd} { expr [$cmd xColumnText -1] }
sqlite3_fts5_create_function db fts5_rowid fts5_rowid

do_execsql_test 1.$tok.0.1 "
  CREATE VIRTUAL TABLE ss USING fts5(a, b, 
       tokenize='tclnum $tok', detail=%DETAIL%);
  INSERT INTO ss(ss, rank) VALUES('rank', 'fts5_rowid()');

  INSERT INTO ss VALUES('eight vii eight six 3', 'i vii 1 six 9 vii');
  INSERT INTO ss VALUES('9 0 viii viii five', 'i 1 viii ix 3 4');
  INSERT INTO ss VALUES('three nine 5 nine viii four zero', 'ii i 1 5 2 viii');
  INSERT INTO ss VALUES('5 vii three 9 four', 'three five one 7 2 eight one');
}

foreach {tn expr} {


  1.1 "one"   1.2 "two"   1.3 "three"   1.4 "four"
  1.5 "v"     1.6 "vi"    1.7 "vii"     1.8 "viii"
  1.9 "9"    1.10 "0"    1.11 "1"      1.12 "2"

  2.1 "one OR two OR three OR four"
  2.2 "(one AND two) OR (three AND four)"
  2.3 "(one AND two) OR (three AND four) NOT five"

  if {[fts5_expr_ok $expr ss]==0} {
    do_test 1.$tok.$tn.OMITTED { list } [list]
    continue
  }

  set res [fts5_query_data $expr ss ASC ::tclnum_syn]
  do_execsql_test 1.$tok.$tn.[llength $res].asc.1 {
    SELECT rowid, fts5_test_poslist(ss), fts5_test_collist(ss) FROM ss($expr)
  } $res

  do_execsql_test 1.$tok.$tn.[llength $res].asc.2 {
    SELECT rowid, fts5_test_poslist(ss), fts5_test_collist(ss) FROM ss($expr)




    ORDER BY rank ASC
  } $res














}

}
}

finish_test