SQLite

#include "fts5Int.h"

typedef struct Fts5Table Fts5Table;
typedef struct Fts5Cursor Fts5Cursor;
typedef struct Fts5Global Fts5Global;
typedef struct Fts5Auxiliary Fts5Auxiliary;
typedef struct Fts5Auxdata Fts5Auxdata;

/*
** A single object of this type is allocated when the FTS5 module is 
** registered with a database handle. It is used to store pointers to
** all registered FTS5 extensions - tokenizers and auxiliary functions.
*/
struct Fts5Global {

  sqlite3_stmt *pStmt;            /* Statement used to read %_content */
  Fts5Expr *pExpr;                /* Expression for MATCH queries */
  int csrflags;                   /* Mask of cursor flags (see below) */
  Fts5Cursor *pNext;              /* Next cursor in Fts5Cursor.pCsr list */

  /* Variables used by auxiliary functions */
  i64 iCsrId;                     /* Cursor id */
  Fts5Auxiliary *pAux;            /* Currently executing extension function */
  Fts5Auxdata *pAuxdata;          /* First in linked list of aux-data */
  int *aColumnSize;               /* Values for xColumnSize() */
};

/*
** Values for Fts5Cursor.csrflags
*/
#define FTS5CSR_REQUIRE_CONTENT   0x01
#define FTS5CSR_REQUIRE_DOCSIZE   0x02
#define FTS5CSR_EOF               0x04

/*
** Macros to Set(), Clear() and Test() cursor flags.
*/
#define CsrFlagSet(pCsr, flag)   ((pCsr)->csrflags |= (flag))
#define CsrFlagClear(pCsr, flag) ((pCsr)->csrflags &= ~(flag))
#define CsrFlagTest(pCsr, flag)  ((pCsr)->csrflags & (flag))

struct Fts5Auxdata {
  Fts5Auxiliary *pAux;            /* Extension to which this belongs */
  void *pPtr;                     /* Pointer value */
  void(*xDelete)(void*);          /* Destructor */
  Fts5Auxdata *pNext;             /* Next object in linked list */
};

/*
** Close a virtual table handle opened by fts5InitVtab(). If the bDestroy
** argument is non-zero, attempt delete the shadow tables from teh database
*/
static int fts5FreeVtab(Fts5Table *pTab, int bDestroy){
  int rc = SQLITE_OK;

** Close the cursor.  For additional information see the documentation
** on the xClose method of the virtual table interface.
*/
static int fts5CloseMethod(sqlite3_vtab_cursor *pCursor){
  Fts5Table *pTab = (Fts5Table*)(pCursor->pVtab);
  Fts5Cursor *pCsr = (Fts5Cursor*)pCursor;
  Fts5Cursor **pp;
  Fts5Auxdata *pData;
  Fts5Auxdata *pNext;

  if( pCsr->pStmt ){
    int eStmt = fts5StmtType(pCsr->idxNum);
    sqlite3Fts5StorageStmtRelease(pTab->pStorage, eStmt, pCsr->pStmt);
  }
  sqlite3Fts5ExprFree(pCsr->pExpr);

  for(pData=pCsr->pAuxdata; pData; pData=pNext){
    pNext = pData->pNext;
    if( pData->xDelete ) pData->xDelete(pData->pPtr);
    sqlite3_free(pData);
  }

  /* Remove the cursor from the Fts5Global.pCsr list */
  for(pp=&pTab->pGlobal->pCsr; (*pp)!=pCsr; pp=&(*pp)->pNext);
  *pp = pCsr->pNext;

  sqlite3_free(pCsr);
  return SQLITE_OK;

      CsrFlagSet(pCsr, FTS5CSR_EOF);
    }
    CsrFlagSet(pCsr, FTS5CSR_REQUIRE_CONTENT | FTS5CSR_REQUIRE_DOCSIZE );
  }
  
  return rc;
}

static int fts5CursorFirst(Fts5Table *pTab, Fts5Cursor *pCsr, int bAsc){
  int rc;
  rc = sqlite3Fts5ExprFirst(pCsr->pExpr, pTab->pIndex, bAsc);
  if( sqlite3Fts5ExprEof(pCsr->pExpr) ){
    CsrFlagSet(pCsr, FTS5CSR_EOF);
  }
  CsrFlagSet(pCsr, FTS5CSR_REQUIRE_CONTENT | FTS5CSR_REQUIRE_DOCSIZE );
  return rc;
}

/*
** This is the xFilter interface for the virtual table.  See
** the virtual table xFilter method documentation for additional
** information.
*/
static int fts5FilterMethod(

  rc = sqlite3Fts5StorageStmt(pTab->pStorage, eStmt, &pCsr->pStmt);
  if( rc==SQLITE_OK ){
    if( ePlan==FTS5_PLAN_MATCH ){
      char **pzErr = &pTab->base.zErrMsg;
      const char *zExpr = (const char*)sqlite3_value_text(apVal[0]);
      rc = sqlite3Fts5ExprNew(pTab->pConfig, zExpr, &pCsr->pExpr, pzErr);
      if( rc==SQLITE_OK ){
        rc = fts5CursorFirst(pTab, pCsr, bAsc);




      }
    }else{
      if( ePlan==FTS5_PLAN_ROWID ){
        sqlite3_bind_value(pCsr->pStmt, 1, apVal[0]);
      }
      rc = fts5NextMethod(pCursor);
    }

}

static int fts5ApiColumnCount(Fts5Context *pCtx){
  Fts5Cursor *pCsr = (Fts5Cursor*)pCtx;
  return ((Fts5Table*)(pCsr->base.pVtab))->pConfig->nCol;
}

static int fts5ApiColumnTotalSize(
  Fts5Context *pCtx, 
  int iCol, 
  sqlite3_int64 *pnToken
){
  Fts5Cursor *pCsr = (Fts5Cursor*)pCtx;
  Fts5Table *pTab = (Fts5Table*)(pCsr->base.pVtab);
  return sqlite3Fts5StorageSize(pTab->pStorage, iCol, pnToken);
}

static int fts5ApiRowCount(Fts5Context *pCtx, i64 *pnRow){
  Fts5Cursor *pCsr = (Fts5Cursor*)pCtx;
  Fts5Table *pTab = (Fts5Table*)(pCsr->base.pVtab);
  return sqlite3Fts5StorageRowCount(pTab->pStorage, pnRow);
}

static int fts5ApiTokenize(
  Fts5Context *pCtx, 
  const char *pText, int nText, 
  void *pUserData,
  int (*xToken)(void*, const char*, int, int, int, int)

  i64 *piPos 
){
  Fts5Cursor *pCsr = (Fts5Cursor*)pCtx;
  const u8 *a; int n;             /* Poslist for phrase iPhrase */
  n = sqlite3Fts5ExprPoslist(pCsr->pExpr, iPhrase, &a);
  return sqlite3Fts5PoslistNext64(a, n, pi, piPos);
}

static int fts5ApiSetAuxdata(
  Fts5Context *pCtx,              /* Fts5 context */
  void *pPtr,                     /* Pointer to save as auxdata */
  void(*xDelete)(void*)           /* Destructor for pPtr (or NULL) */
){
  Fts5Cursor *pCsr = (Fts5Cursor*)pCtx;
  Fts5Auxdata *pData;

  for(pData=pCsr->pAuxdata; pData; pData=pData->pNext){
    if( pData->pAux==pCsr->pAux ) break;
  }

  if( pData ){
    if( pData->xDelete ){
      pData->xDelete(pData->pPtr);
    }
  }else{
    pData = (Fts5Auxdata*)sqlite3_malloc(sizeof(Fts5Auxdata));
    if( pData==0 ) return SQLITE_NOMEM;
    memset(pData, 0, sizeof(Fts5Auxdata));
    pData->pAux = pCsr->pAux;
    pData->pNext = pCsr->pAuxdata;
    pCsr->pAuxdata = pData;
  }

  pData->xDelete = xDelete;
  pData->pPtr = pPtr;
  return SQLITE_OK;
}

static void *fts5ApiGetAuxdata(Fts5Context *pCtx, int bClear){
  Fts5Cursor *pCsr = (Fts5Cursor*)pCtx;
  Fts5Auxdata *pData;
  void *pRet = 0;

  for(pData=pCsr->pAuxdata; pData; pData=pData->pNext){
    if( pData->pAux==pCsr->pAux ) break;
  }

  if( pData ){
    pRet = pData->pPtr;
    if( bClear ){
      pData->pPtr = 0;
      pData->xDelete = 0;
    }
  }

  return pRet;
}

static int fts5ApiQueryPhrase(Fts5Context*, int, void*, 
    int(*)(const Fts5ExtensionApi*, Fts5Context*, void*)
);

static const Fts5ExtensionApi sFts5Api = {
  1,                            /* iVersion */
  fts5ApiUserData,
  fts5ApiColumnCount,
  fts5ApiRowCount,
  fts5ApiColumnTotalSize,
  fts5ApiTokenize,
  fts5ApiPhraseCount,
  fts5ApiPhraseSize,
  fts5ApiRowid,
  fts5ApiColumnText,
  fts5ApiColumnSize,
  fts5ApiPoslist,
  fts5ApiQueryPhrase,
  fts5ApiSetAuxdata,
  fts5ApiGetAuxdata,
};


/*
** Implementation of API function xQueryPhrase().
*/
static int fts5ApiQueryPhrase(
  Fts5Context *pCtx, 
  int iPhrase, 
  void *pUserData,
  int(*xCallback)(const Fts5ExtensionApi*, Fts5Context*, void*)
){
  Fts5Cursor *pCsr = (Fts5Cursor*)pCtx;
  Fts5Table *pTab = (Fts5Table*)(pCsr->base.pVtab);
  int rc;
  Fts5Cursor *pNew = 0;

  rc = fts5OpenMethod(pCsr->base.pVtab, (sqlite3_vtab_cursor**)&pNew);
  if( rc==SQLITE_OK ){
    Fts5Config *pConf = pTab->pConfig;
    pNew->idxNum = FTS5_PLAN_MATCH;
    pNew->base.pVtab = (sqlite3_vtab*)pTab;
    rc = sqlite3Fts5ExprPhraseExpr(pConf, pCsr->pExpr, iPhrase, &pNew->pExpr);
  }

  if( rc==SQLITE_OK ){
    for(rc = fts5CursorFirst(pTab, pNew, 0);
        rc==SQLITE_OK && CsrFlagTest(pNew, FTS5CSR_EOF)==0;
        rc = fts5NextMethod((sqlite3_vtab_cursor*)pNew)
    ){
      rc = xCallback(&sFts5Api, (Fts5Context*)pNew, pUserData);
      if( rc!=SQLITE_OK ){
        if( rc==SQLITE_DONE ) rc = SQLITE_OK;
        break;
      }
    }
  }

  fts5CloseMethod((sqlite3_vtab_cursor*)pNew);
  return rc;
}

static void fts5ApiCallback(
  sqlite3_context *context,
  int argc,
  sqlite3_value **argv
){














  Fts5Auxiliary *pAux;
  Fts5Cursor *pCsr;
  i64 iCsrId;

  assert( argc>=1 );
  pAux = (Fts5Auxiliary*)sqlite3_user_data(context);
  iCsrId = sqlite3_value_int64(argv[0]);

  for(pCsr=pAux->pGlobal->pCsr; pCsr; pCsr=pCsr->pNext){
    if( pCsr->iCsrId==iCsrId ) break;
  }
  if( pCsr==0 ){
    char *zErr = sqlite3_mprintf("no such cursor: %lld", iCsrId);
    sqlite3_result_error(context, zErr, -1);
  }else{
    assert( pCsr->pAux==0 );
    pCsr->pAux = pAux;
    pAux->xFunc(&sFts5Api, (Fts5Context*)pCsr, context, argc-1, &argv[1]);
    pCsr->pAux = 0;
  }
}


/*
** This routine implements the xFindFunction method for the FTS3

  sqlite3_context *pCtx,          /* Context for returning result/error */
  int nVal,                       /* Number of values in apVal[] array */
  sqlite3_value **apVal           /* Array of trailing arguments */
);

/*
**
** xUserData(pFts):
**
**   Return a copy of the context pointer the extension function was 
**   registered with.
**
**
** xColumnTotalSize(pFts, iCol, pnToken):
**
**   Returns the total number of tokens in column iCol, considering all
**   rows in the FTS5 table.
**
**
** xColumnCount:
**   Returns the number of columns in the FTS5 table.
**
** xColumnSize:
**   Reports the size in tokens of a column value from the current row.
**

** xPoslist:
**   Iterate through instances of phrase iPhrase in the current row. 
**
**   At EOF, a non-zero value is returned and output variable iPos set to -1.
**
** xTokenize:
**   Tokenize text using the tokenizer belonging to the FTS5 table.
**
**
** xQueryPhrase(pFts5, iPhrase, pUserData, xCallback):
**
**   This API function is used to query the FTS table for phrase iPhrase
**   of the current query. Specifically, a query equivalent to:
**
**       ... FROM ftstable WHERE ftstable MATCH $p ORDER BY DESC
**
**   with $p set to a phrase equivalent to the phrase iPhrase of the
**   current query is executed. For each row visited, the callback function
**   passed as the fourth argument is invoked. The context and API objects 
**   passed to the callback function may be used to access the properties of
**   each matched row. Invoking Api.xUserData() returns a copy of the pointer
**   passed as the third argument to pUserData.
**
**   If the callback function returns any value other than SQLITE_OK, the
**   query is abandoned and the xQueryPhrase function returns immediately.
**   If the returned value is SQLITE_DONE, xQueryPhrase returns SQLITE_OK.
**   Otherwise, the error code is propagated upwards.
**
**   If the query runs to completion without incident, SQLITE_OK is returned.
**   Or, if some error occurs before the query completes or is aborted by
**   the callback, an SQLite error code is returned.
**
**
** xSetAuxdata(pFts5, pAux, xDelete)
**
**   Save the pointer passed as the second argument as the extension functions 
**   "auxiliary data". The pointer may then be retrieved by the current or any
**   future invocation of the same fts5 extension function made as part of
**   of the same MATCH query using the xGetAuxdata() API.
**
**   Each extension function is allocated a single auxiliary data slot per
**   query. If the extension function is invoked more than once by the SQL
**   query, then all invocations share a single auxiliary data context.
**
**   If there is already an auxiliary data pointer when this function is
**   invoked, then it is replaced by the new pointer. If an xDelete callback
**   was specified along with the original pointer, it is invoked at this
**   point.
**
**   The xDelete callback, if one is specified, is also invoked on the
**   auxiliary data pointer after the FTS5 query has finished.
**
**
** xGetAuxdata(pFts5, bClear)
**
**   Returns the current auxiliary data pointer for the fts5 extension 
**   function. See the xSetAuxdata() method for details.
**
**   If the bClear argument is non-zero, then the auxiliary data is cleared
**   (set to NULL) before this function returns. In this case the xDelete,
**   if any, is not invoked.
**
**
** xRowCount(pFts5, pnRow)
**
**   This function is used to retrieve the total number of rows in the table.
**   In other words, the same value that would be returned by:
**
**        SELECT count(*) FROM ftstable;
*/
struct Fts5ExtensionApi {
  int iVersion;                   /* Currently always set to 1 */

  void *(*xUserData)(Fts5Context*);

  int (*xColumnCount)(Fts5Context*);
  int (*xRowCount)(Fts5Context*, sqlite3_int64 *pnRow);
  int (*xColumnTotalSize)(Fts5Context*, int iCol, sqlite3_int64 *pnToken);

  int (*xTokenize)(Fts5Context*, 
    const char *pText, int nText, /* Text to tokenize */
    void *pCtx,                   /* Context passed to xToken() */
    int (*xToken)(void*, const char*, int, int, int, int)    /* Callback */
  );

  int (*xPhraseCount)(Fts5Context*);
  int (*xPhraseSize)(Fts5Context*, int iPhrase);

  sqlite3_int64 (*xRowid)(Fts5Context*);
  int (*xColumnText)(Fts5Context*, int iCol, const char **pz, int *pn);
  int (*xColumnSize)(Fts5Context*, int iCol, int *pnToken);
  int (*xPoslist)(Fts5Context*, int iPhrase, int *pi, sqlite3_int64 *piPos);

  int (*xQueryPhrase)(Fts5Context*, int iPhrase, void *pUserData,
    int(*)(const Fts5ExtensionApi*,Fts5Context*,void*)
  );
  int (*xSetAuxdata)(Fts5Context*, void *pAux, void(*xDelete)(void*));
  void *(*xGetAuxdata)(Fts5Context*, int bClear);
};

#define FTS5_POS2COLUMN(iPos) (int)(iPos >> 32)
#define FTS5_POS2OFFSET(iPos) (int)(iPos & 0xFFFFFFFF)

/* 
** CUSTOM AUXILIARY FUNCTIONS
*************************************************************************/
#endif /* _FTS5_H */

int sqlite3Fts5StorageIntegrity(Fts5Storage *p);

int sqlite3Fts5StorageStmt(Fts5Storage *p, int eStmt, sqlite3_stmt **);
void sqlite3Fts5StorageStmtRelease(Fts5Storage *p, int eStmt, sqlite3_stmt*);

int sqlite3Fts5StorageDocsize(Fts5Storage *p, i64 iRowid, int *aCol);
int sqlite3Fts5StorageSize(Fts5Storage *p, int iCol, i64 *pnAvg);
int sqlite3Fts5StorageRowCount(Fts5Storage *p, i64 *pnRow);


/*
** End of interface to code in fts5_storage.c.
**************************************************************************/

/* Called during startup to register a UDF with SQLite */
int sqlite3Fts5ExprInit(sqlite3*);

int sqlite3Fts5ExprPhraseCount(Fts5Expr*);
int sqlite3Fts5ExprPhraseSize(Fts5Expr*, int iPhrase);
int sqlite3Fts5ExprPoslist(Fts5Expr*, int, const u8 **);

int sqlite3Fts5ExprPhraseExpr(Fts5Config*, Fts5Expr*, int, Fts5Expr**);

/*******************************************
** The fts5_expr.c API above this point is used by the other hand-written
** C code in this module. The interfaces below this point are called by
** the parser code in fts5parse.y.  */

void sqlite3Fts5ParseError(Fts5Parse *pParse, const char *zFmt, ...);

**    May you find forgiveness for yourself and forgive others.
**    May you share freely, never taking more than you give.
**
******************************************************************************
*/

#include "fts5Int.h"
#include <math.h>

typedef struct SnippetPhrase SnippetPhrase;
typedef struct SnippetIter SnippetIter;
typedef struct SnippetCtx SnippetCtx;

struct SnippetPhrase {
  u64 mask;                       /* Current mask */

  rc = pApi->xTokenize(pFts, zCol, nCol, (void*)&ctx, fts5SnippetCallback);
  if( rc==SQLITE_OK ){
    int i1;                       /* First token from input to include */
    int i2;                       /* Last token from input to include */

    int iPrint;
    int iMatchto;

    int iLast;

    int *aiStart = ctx.aiStart - ctx.iFirst;
    int *aiEnd = ctx.aiEnd - ctx.iFirst;

    /* Ideally we want to start the snippet with token (ctx.iFirst + nShift).
    ** However, this is only possible if there are sufficient tokens within

  sqlite3_value **apVal           /* Array of trailing arguments */
){
  const char *zStart = "<b>";
  const char *zFinal = "</b>";
  const char *zEllip = "<b>...</b>";
  int nToken = -15;
  int nAbs;

  int rc;
  SnippetIter *pIter = 0;

  if( nVal>=1 ) zStart = (const char*)sqlite3_value_text(apVal[0]);
  if( nVal>=2 ) zFinal = (const char*)sqlite3_value_text(apVal[1]);
  if( nVal>=3 ) zEllip = (const char*)sqlite3_value_text(apVal[2]);
  if( nVal>=4 ){
    nToken = sqlite3_value_int(apVal[3]);
    if( nToken==0 ) nToken = -15;
  }
  nAbs = nToken * (nToken<0 ? -1 : 1);

  rc = fts5SnippetIterNew(pApi, pFts, nAbs, &pIter);
  if( rc==SQLITE_OK ){
    Fts5Buffer buf;               /* Result buffer */
    int nBestScore = 0;           /* Score of best snippet found */



    for(fts5SnippetIterFirst(pIter); 
        pIter->iLast>=0; 
        fts5SnippetIterNext(pIter)
    ){
      if( pIter->nScore>nBestScore ) nBestScore = pIter->nScore;
    }

  }

  fts5SnippetIterFree(pIter);
  if( rc!=SQLITE_OK ){
    sqlite3_result_error_code(pCtx, rc);
  }
}

typedef struct Fts5GatherCtx Fts5GatherCtx;
struct Fts5GatherCtx {
  int nCol;
  int iPhrase;
  int *anVal;
};

static int fts5GatherCallback(
  const Fts5ExtensionApi *pApi, 
  Fts5Context *pFts,
  void *pUserData
){
  Fts5GatherCtx *p = (Fts5GatherCtx*)pUserData;
  int i = 0;
  int iPrev = -1;
  i64 iPos = 0;

  while( 0==pApi->xPoslist(pFts, 0, &i, &iPos) ){
    int iCol = FTS5_POS2COLUMN(iPos);
    if( iCol!=iPrev ){
      p->anVal[p->iPhrase * p->nCol + iCol]++;
      iPrev = iCol;
    }
  }

  return SQLITE_OK;
}

/*
** This function returns a pointer to an array of integers containing entries
** indicating the number of rows in the table for which each phrase features 
** at least once in each column.
**
** If nCol is the number of matchable columns in the table, and nPhrase is
** the number of phrases in the query, the array contains a total of
** (nPhrase*nCol) entries.
**
** For phrase iPhrase and column iCol:
**
**   anVal[iPhrase * nCol + iCol]
**
** is set to the number of rows in the table for which column iCol contains 
** at least one instance of phrase iPhrase.
*/
static int fts5GatherTotals(
  const Fts5ExtensionApi *pApi,   /* API offered by current FTS version */
  Fts5Context *pFts,              /* First arg to pass to pApi functions */
  int **panVal
){
  int rc = SQLITE_OK;
  int *anVal = 0;
  int i;                          /* For iterating through expression phrases */
  int nPhrase = pApi->xPhraseCount(pFts);
  int nCol = pApi->xColumnCount(pFts);
  int nByte = nCol * nPhrase * sizeof(int);
  Fts5GatherCtx sCtx;

  sCtx.nCol = nCol;
  anVal = sCtx.anVal = (int*)sqlite3_malloc(nByte);
  if( anVal==0 ){
    rc = SQLITE_NOMEM;
  }else{
    memset(anVal, 0, nByte);
  }

  for(i=0; i<nPhrase && rc==SQLITE_OK; i++){
    sCtx.iPhrase = i;
    rc = pApi->xQueryPhrase(pFts, i, (void*)&sCtx, fts5GatherCallback);
  }

  if( rc!=SQLITE_OK ){
    sqlite3_free(anVal);
    anVal = 0;
  }

  *panVal = anVal;
  return rc;
}

typedef struct Fts5Bm25Context Fts5Bm25Context;
struct Fts5Bm25Context {
  int nPhrase;
  int nCol;
  double *aIDF;                   /* Array of IDF values */
  double *aAvg;                   /* Average size of each column in tokens */
};

static void fts5Bm25Function(
  const Fts5ExtensionApi *pApi,   /* API offered by current FTS version */
  Fts5Context *pFts,              /* First arg to pass to pApi functions */
  sqlite3_context *pCtx,          /* Context for returning result/error */
  int nVal,                       /* Number of values in apVal[] array */
  sqlite3_value **apVal           /* Array of trailing arguments */
){
  const double k1 = 1.2;
  const double B = 0.75;

  int rc = SQLITE_OK;
  Fts5Bm25Context *p;

  p = pApi->xGetAuxdata(pFts, 0);
  if( p==0 ){
    int *anVal = 0;
    int ic;                       /* For iterating through columns */
    int ip;                       /* For iterating through phrases */
    i64 nRow;                     /* Total number of rows in table */
    int nPhrase = pApi->xPhraseCount(pFts);
    int nCol = pApi->xColumnCount(pFts);
    int nByte = sizeof(Fts5Bm25Context) 
              + sizeof(double) * nPhrase * nCol       /* aIDF[] */
              + sizeof(double) * nCol;                /* aAvg[] */

    p = (Fts5Bm25Context*)sqlite3_malloc(nByte);
    if( p==0 ){
      rc = SQLITE_NOMEM;
    }else{
      memset(p, 0, nByte);
      p->aAvg = (double*)&p[1];
      p->aIDF = (double*)&p->aAvg[nCol];
    }

    if( rc==SQLITE_OK ){
      rc = pApi->xRowCount(pFts, &nRow); 
      assert( nRow>0 || rc!=SQLITE_OK );
    }

    for(ic=0; rc==SQLITE_OK && ic<nCol; ic++){
      i64 nToken = 0;
      rc = pApi->xColumnTotalSize(pFts, ic, &nToken);
      p->aAvg[ic] = (double)nToken / (double)nRow;
    }

    if( rc==SQLITE_OK ){
      rc = fts5GatherTotals(pApi, pFts, &anVal);
    }
    for(ic=0; ic<nCol; ic++){
      for(ip=0; rc==SQLITE_OK && ip<nPhrase; ip++){
        int idx = ip * nCol + ic;
        p->aIDF[idx] = log( (0.5 + nRow - anVal[idx]) / (0.5 + anVal[idx]) );
        if( p->aIDF[idx]<0.0 ) p->aIDF[idx] = 0.0;
      }
    }

    sqlite3_free(anVal);
    if( rc==SQLITE_OK ){
      rc = pApi->xSetAuxdata(pFts, p, sqlite3_free);
    }
    if( rc!=SQLITE_OK ){
      sqlite3_free(p);
    }
  }

  if( rc==SQLITE_OK ){
    int ip;
    double score = 0.0;

    for(ip=0; rc==SQLITE_OK && ip<p->nPhrase; ip++){
      int iPrev = 0;
      int nHit = 0;
      int i = 0;
      i64 iPos = 0;

      while( rc==SQLITE_OK && 0==pApi->xPoslist(pFts, ip, &i, &iPos) ){
        int iCol = FTS5_POS2COLUMN(iPos);
        if( iCol!=iPrev && nHit>0 ){
          int sz = 0;
          int idx = ip * p->nCol + iPrev;
          rc = pApi->xColumnSize(pFts, iPrev, &sz);

          score += p->aIDF[idx] * nHit * (k1+1.0) / 
                  (nHit + k1 * (1.0 - B + B * sz / p->aAvg[iCol]));
          nHit = 0;
        }
        nHit++;
        iPrev = iCol;
      }
    }
    
    if( rc==SQLITE_OK ){
      sqlite3_result_double(pCtx, score);
    }

  }

  if( rc!=SQLITE_OK ){
    sqlite3_result_error_code(pCtx, rc);
  }
}

static int fts5TestCallback(
  void *pContext,                 /* Pointer to Fts5Buffer object */
  const char *pToken,             /* Buffer containing token */
  int nToken,                     /* Size of token in bytes */
  int iStart,                     /* Start offset of token */

    zReq = (const char*)sqlite3_value_text(apVal[0]);
  }

  memset(&s, 0, sizeof(Fts5Buffer));
  nCol = pApi->xColumnCount(pFts);

  if( zReq==0 ){
    sqlite3Fts5BufferAppendPrintf(&rc, &s, "columntotalsize ");
  }
  if( 0==zReq || 0==sqlite3_stricmp(zReq, "columntotalsize") ){
    if( zReq==0 && nCol>1 ) sqlite3Fts5BufferAppendPrintf(&rc, &s, "{");
    for(i=0; rc==SQLITE_OK && i<nCol; i++){
      i64 colsz = 0;
      rc = pApi->xColumnTotalSize(pFts, i, &colsz);
      sqlite3Fts5BufferAppendPrintf(&rc, &s, "%s%d", i==0?"":" ", colsz);
    }
    if( zReq==0 && nCol>1 ) sqlite3Fts5BufferAppendPrintf(&rc, &s, "}");
  }

  if( zReq==0 ){
    sqlite3Fts5BufferAppendPrintf(&rc, &s, "columncount ");

    }else{
      sqlite3Fts5BufferAppendPrintf(&rc, &s, "%s", (const char*)s3.p);
    }
    sqlite3_free(s3.p);
  }

  if( zReq==0 ){
    sqlite3Fts5BufferAppendPrintf(&rc, &s, " queryphrase ");
  }
  if( 0==zReq || 0==sqlite3_stricmp(zReq, "queryphrase") ){
    int ic, ip;
    int *anVal = 0;
    Fts5Buffer buf1;
    memset(&buf1, 0, sizeof(Fts5Buffer));

    if( rc==SQLITE_OK ){
      anVal = (int*)pApi->xGetAuxdata(pFts, 0);
      if( anVal==0 ){
        rc = fts5GatherTotals(pApi, pFts, &anVal);
        if( rc==SQLITE_OK ){
          rc = pApi->xSetAuxdata(pFts, (void*)anVal, sqlite3_free);
        }
      }
    }

    for(ip=0; rc==SQLITE_OK && ip<nPhrase; ip++){
      if( ip>0 ) sqlite3Fts5BufferAppendString(&rc, &buf1, " ");
      if( nCol>1 ) sqlite3Fts5BufferAppendString(&rc, &buf1, "{");
      for(ic=0; ic<nCol; ic++){
        int iVal = anVal[ip * nCol + ic];
        sqlite3Fts5BufferAppendPrintf(&rc, &buf1, "%s%d", ic==0?"":" ", iVal);
      }
      if( nCol>1 ) sqlite3Fts5BufferAppendString(&rc, &buf1, "}");
    }

    if( zReq==0 ){
      sqlite3Fts5BufferAppendListElem(&rc, &s, (const char*)buf1.p, buf1.n);
    }else{
      sqlite3Fts5BufferAppendString(&rc, &s, (const char*)buf1.p);
    }
    sqlite3_free(buf1.p);
  }

  if( zReq==0 ){
    sqlite3Fts5BufferAppendString(&rc, &s, " rowid ");
  }
  if( 0==zReq || 0==sqlite3_stricmp(zReq, "rowid") ){
    iRowid = pApi->xRowid(pFts);
    sqlite3Fts5BufferAppendPrintf(&rc, &s, "%lld", iRowid);
  }

  if( zReq==0 ){
    sqlite3Fts5BufferAppendString(&rc, &s, " rowcount ");
  }
  if( 0==zReq || 0==sqlite3_stricmp(zReq, "rowcount") ){
    i64 nRow;
    rc = pApi->xRowCount(pFts, &nRow);
    sqlite3Fts5BufferAppendPrintf(&rc, &s, "%lld", nRow);
  }

  if( zReq==0 ){
    sqlite3Fts5BufferAppendString(&rc, &s, " tokenize ");
  }
  if( 0==zReq || 0==sqlite3_stricmp(zReq, "tokenize") ){
    Fts5Buffer buf;
    memset(&buf, 0, sizeof(buf));
    for(i=0; rc==SQLITE_OK && i<nCol; i++){
      const char *z;
      int n;

int sqlite3Fts5AuxInit(Fts5Global *pGlobal){
  struct Builtin {
    const char *zFunc;            /* Function name (nul-terminated) */
    void *pUserData;              /* User-data pointer */
    fts5_extension_function xFunc;/* Callback function */
    void (*xDestroy)(void*);      /* Destructor function */
  } aBuiltin [] = {
    { "bm25",      0, fts5Bm25Function,    0 },
    { "snippet",   0, fts5SnippetFunction, 0 },
    { "fts5_test", 0, fts5TestFunction,    0 },
  };

  int rc = SQLITE_OK;             /* Return code */
  int i;                          /* To iterate through builtin functions */

  for(i=0; rc==SQLITE_OK && i<sizeof(aBuiltin)/sizeof(aBuiltin[0]); i++){
    rc = sqlite3Fts5CreateAux(pGlobal, 

    }
  }

  sqlite3_free(sParse.apPhrase);
  *pzErr = sParse.zErr;
  return sParse.rc;
}

static char *fts5ExprStrdup(int *pRc, const char *zIn){
  char *zRet = 0;
  if( *pRc==SQLITE_OK ){
    int nByte = strlen(zIn) + 1;
    zRet = sqlite3_malloc(nByte);
    if( zRet ){
      memcpy(zRet, zIn, nByte);
    }else{
      *pRc = SQLITE_NOMEM;
    }
  }
  return zRet;
}

static void *fts5ExprMalloc(int *pRc, int nByte){
  void *pRet = 0;
  if( *pRc==SQLITE_OK ){
    pRet = sqlite3_malloc(nByte);
    if( pRet ){
      memset(pRet, 0, nByte);
    }else{
      *pRc = SQLITE_NOMEM;
    }
  }
  return pRet;
}

/*
** Create a new FTS5 expression by cloning phrase iPhrase of the
** expression passed as the second argument.
*/
int sqlite3Fts5ExprPhraseExpr(
  Fts5Config *pConfig,
  Fts5Expr *pExpr, 
  int iPhrase, 
  Fts5Expr **ppNew
){
  int rc = SQLITE_OK;             /* Return code */
  Fts5ExprPhrase *pOrig = 0;      /* The phrase extracted from pExpr */
  int i;                          /* Used to iterate through phrase terms */

  /* Components of the new expression object */
  Fts5Expr *pNew;
  Fts5ExprPhrase **apPhrase;
  Fts5ExprNode *pNode;
  Fts5ExprNearset *pNear;
  Fts5ExprPhrase *pCopy;

  pOrig = pExpr->apPhrase[iPhrase];
  pNew = (Fts5Expr*)fts5ExprMalloc(&rc, sizeof(Fts5Expr));
  apPhrase = (Fts5ExprPhrase**)fts5ExprMalloc(&rc, sizeof(Fts5ExprPhrase*));
  pNode = (Fts5ExprNode*)fts5ExprMalloc(&rc, sizeof(Fts5ExprNode));
  pNear = (Fts5ExprNearset*)fts5ExprMalloc(&rc, 
      sizeof(Fts5ExprNearset) + sizeof(Fts5ExprPhrase*)
  );
  pCopy = (Fts5ExprPhrase*)fts5ExprMalloc(&rc, 
      sizeof(Fts5ExprPhrase) + sizeof(Fts5ExprTerm) * pOrig->nTerm
  );

  for(i=0; rc==SQLITE_OK && i<pOrig->nTerm; i++){
    pCopy->aTerm[i].zTerm = fts5ExprStrdup(&rc, pOrig->aTerm[i].zTerm);
    pCopy->aTerm[i].bPrefix = pOrig->aTerm[i].bPrefix;
  }

  if( rc==SQLITE_OK ){
    /* All the allocations succeeded. Put the expression object together. */
    pNew->pIndex = pExpr->pIndex;
    pNew->pRoot = pNode;
    pNew->nPhrase = 1;
    pNew->apPhrase = apPhrase;
    pNew->apPhrase[0] = pCopy;

    pNode->eType = FTS5_STRING;
    pNode->pNear = pNear;

    pNear->iCol = -1;
    pNear->nPhrase = 1;
    pNear->apPhrase[0] = pCopy;

    pCopy->nTerm = pOrig->nTerm;
    pCopy->pNode = pNode;
  }else{
    /* At least one allocation failed. Free them all. */
    if( pCopy ){
      for(i=0; i<pOrig->nTerm; i++){
        sqlite3_free(pCopy->aTerm[i].zTerm);
      }
      sqlite3_free(pCopy);
      sqlite3_free(pNear);
      sqlite3_free(pNode);
      sqlite3_free(apPhrase);
      sqlite3_free(pNew);
      pNew = 0;
    }
  }

  *ppNew = pNew;
  return rc;
}

/*
** Free the expression node object passed as the only argument.
*/
void sqlite3Fts5ParseNodeFree(Fts5ExprNode *p){
  if( p ){
    sqlite3Fts5ParseNodeFree(p->pLeft);

    if( bCorrupt && rc==SQLITE_OK ){
      rc = SQLITE_CORRUPT_VTAB;
    }
  }
  return rc;
}

int sqlite3Fts5StorageSize(Fts5Storage *p, int iCol, i64 *pnToken){
  int rc = fts5StorageLoadTotals(p);
  if( rc==SQLITE_OK ){


    *pnToken = p->aTotalSize[iCol];


  }
  return rc;
}

int sqlite3Fts5StorageRowCount(Fts5Storage *p, i64 *pnRow){
  int rc = fts5StorageLoadTotals(p);
  if( rc==SQLITE_OK ){
    *pnRow = p->nTotalRow;
  }
  return rc;
}

} {
  3 {a {}}
  2 {{} a}
  1 {{a b c d} {e f g h i j}}
}

do_execsql_test 5.3 {
  SELECT rowid, fts5_test(t5, 'columntotalsize') FROM t5 WHERE t5 MATCH 'a'
  ORDER BY rowid DESC;
} {
  3 {5 7}
  2 {5 7}
  1 {5 7}
}

do_execsql_test 5.4 {
  INSERT INTO t5 VALUES('x y z', 'v w x y z');
  SELECT rowid, fts5_test(t5, 'columntotalsize') FROM t5 WHERE t5 MATCH 'a'
  ORDER BY rowid DESC;
} {
  3 {8 12}
  2 {8 12}
  1 {8 12}
}

#-------------------------------------------------------------------------
# Test the xTokenize() API
#
reset_db
do_execsql_test 6.1 {
  CREATE VIRTUAL TABLE t6 USING fts5(x, y);
  INSERT INTO t6 VALUES('There are more', 'things in heaven and earth');
  INSERT INTO t6 VALUES(', Horatio, Than are', 'dreamt of in your philosophy.');
}

do_execsql_test 6.2 {
  SELECT rowid, fts5_test(t6, 'tokenize') FROM t6 WHERE t6 MATCH 't*'
} {
  2 {{horatio than are} {dreamt of in your philosophy}}
  1 {{there are more} {things in heaven and earth}}
}

#-------------------------------------------------------------------------
# Test the xQueryPhrase() API
#
reset_db
do_execsql_test 7.1 {
  CREATE VIRTUAL TABLE t7 USING fts5(x, y);
}
do_test 7.2 {
  foreach {x y} {
    {q i b w s a a e l o} {i b z a l f p t e u}
    {b a z t a l o x d i} {b p a d b f h d w y}
    {z m h n p p u i e g} {v h d v b x j j c z}
    {a g i m v a u c b i} {p k s o t l r t b m}
    {v v c j o d a s c p} {f f v o k p o f o g}
  } {
    execsql {INSERT INTO t7 VALUES($x, $y)}
  }
  execsql { SELECT count(*) FROM t7 }
} {5}

foreach {tn q res} {
  1 a {{4 2}}
  2 b {{3 4}}
  3 c {{2 1}}
  4 d {{2 2}}
  5 {a AND b} {{4 2} {3 4}}
  6 {a OR b OR c OR d} {{4 2} {3 4} {2 1} {2 2}}
} {
  do_execsql_test 7.3.$tn { 
    SELECT fts5_test(t7, 'queryphrase') FROM t7 WHERE t7 MATCH $q LIMIT 1
  } [list $res]
}

do_execsql_test 7.4 {
  SELECT fts5_test(t7, 'rowcount') FROM t7 WHERE t7 MATCH 'a';
} {5 5 5 5}



finish_test

}


do_execsql_test 1.0 {
  CREATE VIRTUAL TABLE t1 USING fts5(x, y);
}

proc do_snippet_test {tn doc match res} {

  uplevel #0 [list set v1 $doc]
  uplevel #0 [list set v2 $match]

  do_execsql_test $tn.1 {
    DELETE FROM t1;
    INSERT INTO t1 VALUES($v1, NULL);
    SELECT snippet(t1, '[', ']', '...', 7) FROM t1 WHERE t1 MATCH $v2;
  } [list $res]

  do_execsql_test $tn.2 {
    DELETE FROM t1;
    INSERT INTO t1 VALUES(NULL, $v1);
    SELECT snippet(t1, '[', ']', '...', 7) FROM t1 WHERE t1 MATCH $v2;
  } [list $res]

}


foreach {tn doc res} {

  1.1 {X o o o o o o} {[X] o o o o o o}
  1.2 {o X o o o o o} {o [X] o o o o o}
  1.3 {o o X o o o o} {o o [X] o o o o}
  1.4 {o o o X o o o} {o o o [X] o o o}

  7.1 {X o o X o o o o o} {[X] o o [X] o o o...}
  7.2 {o X o o X o o o o} {o [X] o o [X] o o...}
  7.3 {o o X o o X o o o} {...o [X] o o [X] o o...}
  7.4 {o o o X o o X o o} {...o [X] o o [X] o o}
  7.5 {o o o o X o o X o} {...o o [X] o o [X] o}
  7.6 {o o o o o X o o X} {...o o o [X] o o [X]}
} {





  do_snippet_test 1.$tn $doc X $res





}

foreach {tn doc res} {
  1.1 {X Y o o o o o} {[X Y] o o o o o}
  1.2 {o X Y o o o o} {o [X Y] o o o o}
  1.3 {o o X Y o o o} {o o [X Y] o o o}
  1.4 {o o o X Y o o} {o o o [X Y] o o}

  3.4 {o o o X Y o o o o} {...o o [X Y] o o o...}
  3.5 {o o o o X Y o o o} {...o o [X Y] o o o}
  3.6 {o o o o o X Y o o} {...o o o [X Y] o o}
  3.7 {o o o o o o X Y o} {...o o o o [X Y] o}
  3.8 {o o o o o o o X Y} {...o o o o o [X Y]}

} {





  do_snippet_test 1.$tn $doc "X + Y" $res





}

finish_test