Documentation Source Text

Artifact Content
Login

Artifact 40ffcea582cf8aeee2b0af62291168bc9e96d118:



/*
** There are also two SQL user functions registered:
**
**   rank()
**   erank()
**
** rank() interprets the return value of the FTS3 matchinfo() function and
** returns a score for the match (a real number). The higher the score, the
** more relevant the document is considered. This is used to order query
** results when the user searchs the database. The rank() function takes
** (nCol+1) arguments, where nCol is the number of columns in the FTS3
** table. The first argument is the return value of matchinfo(). The
** second argument is the number of tokens in column 0 of the current FTS3 
** table row. The third argument is the number of tokens in column 1, and
** so on.
**
** Function erank() is called in exactly the same way as rank(). Instead
** of returning a score, it returns an HTML formatted table containing
** data that may be used to understand how the score for the current row
** was calculated.
*/

#include <tcl.h>
#include <string.h>
#include <assert.h>
#include <ctype.h>
#include <math.h>

#include "sqlite3.h"

typedef unsigned int u32;
typedef unsigned char u8;
typedef sqlite3_uint64 u64;

/*
** Implementation of search result ranking function.
*/
static void rankfunc(sqlite3_context *pCtx, int nVal, sqlite3_value **apVal){
  u32 *aMatchinfo;
  double score = 0.0;
  int iCol;
  int iPhrase;
  int nCol;
  int nPhrase;

  int isExplain = sqlite3_user_data(pCtx);
  char *zExplain = 0;

  if( nVal==0 ) goto wna;
  aMatchinfo = (u32 *)sqlite3_value_blob(apVal[0]);
  nPhrase = aMatchinfo[0];
  nCol = aMatchinfo[1];
  if( nVal!=nCol+1 ) goto wna;

  if( isExplain ) zExplain = sqlite3_mprintf("<table width=100%%>");

  for(iCol=0; iCol<nCol; iCol++){
    int nToken = sqlite3_value_int(apVal[iCol+1]);
    double colscore = 0.0;
    if( isExplain ){
      zExplain = sqlite3_mprintf("%z<tr><td>%d.<td>( ", zExplain, iCol);
    }
    for(iPhrase=0; iPhrase<nPhrase; iPhrase++){
      u32 nGlobal = aMatchinfo[2 + iPhrase*nCol + iCol];
      u32 nHit = aMatchinfo[2 + nPhrase*nCol + iPhrase*nCol + iCol];

      if( nHit ) colscore += (double)nHit / (double)nGlobal;
      if( isExplain ){
        const char *zDiv = (iPhrase==0 ? "" : "+ ");
        zExplain = sqlite3_mprintf("%z%s%d/%d ", zExplain, zDiv, nHit, nGlobal);
      }
    }
    colscore = colscore / (log(100+nToken)/log(10)); 
    score += colscore;
    if( isExplain ){
      zExplain = sqlite3_mprintf(
          "%z) / log(100+%d)<td> = %.4f", zExplain, nToken, colscore);
    }
  }

  if( isExplain ){
    sqlite3_result_text(pCtx, sqlite3_mprintf(
        "%z<tr><td><td width=100%%><td>= <b>%.4f</b></table>", zExplain, score
    ), -1, sqlite3_free);
  }else{
    sqlite3_result_double(pCtx, score);
  }
  return;

 wna:
  sqlite3_result_error(pCtx,"wrong number of arguments to function rank()",-1);
}

int Sqlite3_Init(Tcl_Interp *interp);

static int initDb(sqlite3 *db, char **pzErr, void *p){
  sqlite3_create_function(db, "rank",-1, SQLITE_UTF8, 0, rankfunc,0,0);
  sqlite3_create_function(db, "erank", -1, SQLITE_UTF8, (void*)1, rankfunc,0,0);
}

static int AppInit(Tcl_Interp *interp) {
  int rc;
  rc = Sqlite3_Init(interp);
  if( rc!=TCL_OK ) return rc;
  sqlite3_auto_extension(initDb);
  return TCL_OK;
}

int main(int argc, char *argv[]) {
  Tcl_Main(argc, argv, AppInit);
  return 0;
}