SQLite: Check-in [111ca616]

Many hyperlinks are disabled.
Use anonymous login to enable hyperlinks.

Overview

Comment:	When gathering a doclist for querying, don't discard empty position lists until the end; this allows empty position lists to override non-empty lists encountered later in the gathering process. This fixes #1982, which was caused by the fact that for all-column queries we weren't discarding empty position lists at all. (CVS 3433)
Downloads:	Tarball \| ZIP archive \| SQL archive
Timelines:	family \| ancestors \| descendants \| both \| trunk
Files:	files \| file ages \| folders
SHA1:	111ca616713dd89b5d1e114de29c83256731c482
User & Date:	adamd 2006-09-21 20:56:52

Context

2006-09-22
00:06		Implemented UPDATE for full-text tables. We handle an UPDATE to a row by performing an UPDATE on the content table and by building new position lists for each term which appears in either the old or new versions of the row. We write these position lists all at once; this is presumably more efficient than a delete followed by an insert (which would first write empty position lists, then new position lists). (CVS 3434) check-in: 757fa224 user: adamd tags: trunk
2006-09-21
20:56		When gathering a doclist for querying, don't discard empty position lists until the end; this allows empty position lists to override non-empty lists encountered later in the gathering process. This fixes #1982, which was caused by the fact that for all-column queries we weren't discarding empty position lists at all. (CVS 3433) check-in: 111ca616 user: adamd tags: trunk
11:02		Be more aggressive with the SQLITE_OMIT_VACUUM macro. Saves about 150 bytes of code space. (CVS 3432) check-in: 7e618db4 user: drh tags: trunk

Changes

Hide Diffs Unified Diffs Ignore Whitespace Patch

Changes to ext/fts1/fts1.c.

  sqlite_int64 ret;
  assert( !atEnd(pReader) );
  assert( pReader->iLastPos==-1 );
  getVarint(pReader->p, &ret);
  return ret;
}

/* Read the next docid.   See also nextValidDocid().
*/
static sqlite_int64 readDocid(DocListReader *pReader){
  sqlite_int64 ret;
  assert( !atEnd(pReader) );
  assert( pReader->iLastPos==-1 );
  pReader->p += getVarint(pReader->p, &ret);
  if( pReader->pDoclist->iType>=DL_POSITIONS ){
................................................................................
    }
  }
  printf("\n");
  fflush(stdout);
}
#endif /* SQLITE_DEBUG */

/* Trim the given doclist to contain only positions in column [iRestrictColumn],
 * discarding any docids without any remaining positions. */
static void docListRestrictColumn(DocList *in, int iRestrictColumn){
  DocListReader r;
  DocList out;

  assert( in->iType>=DL_POSITIONS );
  readerInit(&r, in);
  docListInit(&out, DL_POSITIONS, NULL, 0);

  while( !atEnd(&r) ){
    sqlite_int64 iDocid = readDocid(&r);
    int match = 0;
    int iPos, iColumn;


    while( (iPos = readPosition(&r, &iColumn)) != -1 ){
      if( iColumn==iRestrictColumn ){


























        if( !match ){
          docListAddDocid(&out, iDocid);
          match = 1;
        }
        docListAddPos(&out, iColumn, iPos);
      }
    }
  }

  docListDestroy(in);
  *in = out;
}

................................................................................
    sqlite_int64 iDocid = readDocid(&updateReader);
    skipPositionList(&updateReader);
    docListSpliceElement(&accReader, iDocid, pSource, updateReader.p-pSource);
  }
}

/*
** Read the next non-deleted docid off of pIn.  Return
** 0 if we reach the end of pDoclist.
*/
static sqlite_int64 nextValidDocid(DocListReader *pIn){
  sqlite_int64 docid = 0;
  skipPositionList(pIn);
  while( !atEnd(pIn) && (docid = readDocid(pIn))==0 ){
    skipPositionList(pIn);
  }
  return docid;
}

/*
** pLeft and pRight are two DocListReaders that are pointing to
** positions lists of the same document: iDocid. 
**
** If there are no instances in pLeft or pRight where the position
................................................................................
  DocList *pOut      /* Write the combined doclist here */
){
  DocListReader left, right;
  sqlite_int64 docidLeft, docidRight;

  readerInit(&left, pLeft);
  readerInit(&right, pRight);
  docidLeft = nextValidDocid(&left);
  docidRight = nextValidDocid(&right);

  while( docidLeft>0 && docidRight>0 ){
    if( docidLeft<docidRight ){
      docidLeft = nextValidDocid(&left);
    }else if( docidRight<docidLeft ){
      docidRight = nextValidDocid(&right);
    }else{
      mergePosList(&left, &right, docidLeft, pOut);
      docidLeft = nextValidDocid(&left);
      docidRight = nextValidDocid(&right);
    }
  }
}

/* We have two doclists:  pLeft and pRight.
** Write the intersection of these two doclists into pOut.
** Only docids are matched.  Position information is ignored.
................................................................................
  DocListReader left, right;
  sqlite_int64 docidLeft, docidRight;

  assert( pOut->iType<DL_POSITIONS );

  readerInit(&left, pLeft);
  readerInit(&right, pRight);
  docidLeft = nextValidDocid(&left);
  docidRight = nextValidDocid(&right);

  while( docidLeft>0 && docidRight>0 ){
    if( docidLeft<docidRight ){
      docidLeft = nextValidDocid(&left);
    }else if( docidRight<docidLeft ){
      docidRight = nextValidDocid(&right);
    }else{
      docListAddDocid(pOut, docidLeft);
      docidLeft = nextValidDocid(&left);
      docidRight = nextValidDocid(&right);
    }
  }
}

/* We have two doclists:  pLeft and pRight.
** Write the union of these two doclists into pOut.
** Only docids are matched.  Position information is ignored.
................................................................................
  DocList *pOut      /* Write the combined doclist here */
){
  DocListReader left, right;
  sqlite_int64 docidLeft, docidRight, priorLeft;

  readerInit(&left, pLeft);
  readerInit(&right, pRight);
  docidLeft = nextValidDocid(&left);
  docidRight = nextValidDocid(&right);

  while( docidLeft>0 && docidRight>0 ){
    if( docidLeft<=docidRight ){
      docListAddDocid(pOut, docidLeft);
    }else{
      docListAddDocid(pOut, docidRight);
    }
    priorLeft = docidLeft;
    if( docidLeft<=docidRight ){
      docidLeft = nextValidDocid(&left);
    }
    if( docidRight>0 && docidRight<=priorLeft ){
      docidRight = nextValidDocid(&right);
    }
  }
  while( docidLeft>0 ){
    docListAddDocid(pOut, docidLeft);
    docidLeft = nextValidDocid(&left);
  }
  while( docidRight>0 ){
    docListAddDocid(pOut, docidRight);
    docidRight = nextValidDocid(&right);
  }
}

/* We have two doclists:  pLeft and pRight.
** Write into pOut all documents that occur in pLeft but not
** in pRight.
**
................................................................................
  DocList *pOut      /* Write the combined doclist here */
){
  DocListReader left, right;
  sqlite_int64 docidLeft, docidRight, priorLeft;

  readerInit(&left, pLeft);
  readerInit(&right, pRight);
  docidLeft = nextValidDocid(&left);
  docidRight = nextValidDocid(&right);

  while( docidLeft>0 && docidRight>0 ){
    priorLeft = docidLeft;
    if( docidLeft<docidRight ){
      docListAddDocid(pOut, docidLeft);
    }
    if( docidLeft<=docidRight ){
      docidLeft = nextValidDocid(&left);
    }
    if( docidRight>0 && docidRight<=priorLeft ){
      docidRight = nextValidDocid(&right);
    }
  }
  while( docidLeft>0 ){
    docListAddDocid(pOut, docidLeft);
    docidLeft = nextValidDocid(&left);
  }
}

static char *string_dup_n(const char *s, int n){
  char *str = malloc(n + 1);
  memcpy(str, s, n);
  str[n] = '\0';
................................................................................
    doclist = old;
  }
  if( rc!=SQLITE_DONE ){
    docListDestroy(&doclist);
    return rc;
  }


  *out = doclist;
  return SQLITE_OK;
}

/* insert into %_term (rowid, term, segment, doclist)
               values ([piRowid], [pTerm], [iSegment], [doclist])
** Lets sqlite select rowid if piRowid is NULL, else uses *piRowid.
................................................................................
        c->eof = 1;
        return rc;
    }
  } else {  /* full-text query */
    rc = sqlite3_reset(c->pStmt);
    if( rc!=SQLITE_OK ) return rc;

    iDocid = nextValidDocid(&c->result);
    if( iDocid==0 ){
      c->eof = 1;
      return SQLITE_OK;
    }
    rc = sqlite3_bind_int64(c->pStmt, 1, iDocid);
    if( rc!=SQLITE_OK ) return rc;
    /* TODO(shess) Handle SQLITE_SCHEMA AND SQLITE_BUSY. */








|







 







|
|










<

>
>


>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
|
|
|
|
|
<







 







|
|
|
|
|
|
|
|
<
<







 







|
|



|

|


|
|







 







|
|



|

|


|
|







 







|
|









|


|




|



|







 







|
|







|


|




|







 







>







 







|

  sqlite_int64 ret;
  assert( !atEnd(pReader) );
  assert( pReader->iLastPos==-1 );
  getVarint(pReader->p, &ret);
  return ret;
}

/* Read the next docid.   See also nextDocid().
*/
static sqlite_int64 readDocid(DocListReader *pReader){
  sqlite_int64 ret;
  assert( !atEnd(pReader) );
  assert( pReader->iLastPos==-1 );
  pReader->p += getVarint(pReader->p, &ret);
  if( pReader->pDoclist->iType>=DL_POSITIONS ){
................................................................................
    }
  }
  printf("\n");
  fflush(stdout);
}
#endif /* SQLITE_DEBUG */

/* Trim the given doclist to contain only positions in column
 * [iRestrictColumn]. */
static void docListRestrictColumn(DocList *in, int iRestrictColumn){
  DocListReader r;
  DocList out;

  assert( in->iType>=DL_POSITIONS );
  readerInit(&r, in);
  docListInit(&out, DL_POSITIONS, NULL, 0);

  while( !atEnd(&r) ){
    sqlite_int64 iDocid = readDocid(&r);

    int iPos, iColumn;

    docListAddDocid(&out, iDocid);
    while( (iPos = readPosition(&r, &iColumn)) != -1 ){
      if( iColumn==iRestrictColumn ){
        docListAddPos(&out, iColumn, iPos);
      }
    }
  }

  docListDestroy(in);
  *in = out;
}

/* Trim the given doclist by discarding any docids without any remaining
 * positions. */
static void docListDiscardEmpty(DocList *in) {
  DocListReader r;
  DocList out;

  /* TODO: It would be nice to implement this operation in place; that
   * could save a significant amount of memory in queries with long doclists. */
  assert( in->iType>=DL_POSITIONS );
  readerInit(&r, in);
  docListInit(&out, DL_POSITIONS, NULL, 0);

  while( !atEnd(&r) ){
    sqlite_int64 iDocid = readDocid(&r);
    int match = 0;
    int iPos, iColumn;
    while( (iPos = readPosition(&r, &iColumn)) != -1 ){
      if( !match ){
        docListAddDocid(&out, iDocid);
        match = 1;
      }
      docListAddPos(&out, iColumn, iPos);

    }
  }

  docListDestroy(in);
  *in = out;
}

................................................................................
    sqlite_int64 iDocid = readDocid(&updateReader);
    skipPositionList(&updateReader);
    docListSpliceElement(&accReader, iDocid, pSource, updateReader.p-pSource);
  }
}

/*
** Read the next docid off of pIn.  Return 0 if we reach the end.
*
* TODO: This assumes that docids are never 0, but they may actually be 0 since
* users can choose docids when inserting into a full-text table.  Fix this.
*/
static sqlite_int64 nextDocid(DocListReader *pIn){
  skipPositionList(pIn);
  return atEnd(pIn) ? 0 : readDocid(pIn);


}

/*
** pLeft and pRight are two DocListReaders that are pointing to
** positions lists of the same document: iDocid. 
**
** If there are no instances in pLeft or pRight where the position
................................................................................
  DocList *pOut      /* Write the combined doclist here */
){
  DocListReader left, right;
  sqlite_int64 docidLeft, docidRight;

  readerInit(&left, pLeft);
  readerInit(&right, pRight);
  docidLeft = nextDocid(&left);
  docidRight = nextDocid(&right);

  while( docidLeft>0 && docidRight>0 ){
    if( docidLeft<docidRight ){
      docidLeft = nextDocid(&left);
    }else if( docidRight<docidLeft ){
      docidRight = nextDocid(&right);
    }else{
      mergePosList(&left, &right, docidLeft, pOut);
      docidLeft = nextDocid(&left);
      docidRight = nextDocid(&right);
    }
  }
}

/* We have two doclists:  pLeft and pRight.
** Write the intersection of these two doclists into pOut.
** Only docids are matched.  Position information is ignored.
................................................................................
  DocListReader left, right;
  sqlite_int64 docidLeft, docidRight;

  assert( pOut->iType<DL_POSITIONS );

  readerInit(&left, pLeft);
  readerInit(&right, pRight);
  docidLeft = nextDocid(&left);
  docidRight = nextDocid(&right);

  while( docidLeft>0 && docidRight>0 ){
    if( docidLeft<docidRight ){
      docidLeft = nextDocid(&left);
    }else if( docidRight<docidLeft ){
      docidRight = nextDocid(&right);
    }else{
      docListAddDocid(pOut, docidLeft);
      docidLeft = nextDocid(&left);
      docidRight = nextDocid(&right);
    }
  }
}

/* We have two doclists:  pLeft and pRight.
** Write the union of these two doclists into pOut.
** Only docids are matched.  Position information is ignored.
................................................................................
  DocList *pOut      /* Write the combined doclist here */
){
  DocListReader left, right;
  sqlite_int64 docidLeft, docidRight, priorLeft;

  readerInit(&left, pLeft);
  readerInit(&right, pRight);
  docidLeft = nextDocid(&left);
  docidRight = nextDocid(&right);

  while( docidLeft>0 && docidRight>0 ){
    if( docidLeft<=docidRight ){
      docListAddDocid(pOut, docidLeft);
    }else{
      docListAddDocid(pOut, docidRight);
    }
    priorLeft = docidLeft;
    if( docidLeft<=docidRight ){
      docidLeft = nextDocid(&left);
    }
    if( docidRight>0 && docidRight<=priorLeft ){
      docidRight = nextDocid(&right);
    }
  }
  while( docidLeft>0 ){
    docListAddDocid(pOut, docidLeft);
    docidLeft = nextDocid(&left);
  }
  while( docidRight>0 ){
    docListAddDocid(pOut, docidRight);
    docidRight = nextDocid(&right);
  }
}

/* We have two doclists:  pLeft and pRight.
** Write into pOut all documents that occur in pLeft but not
** in pRight.
**
................................................................................
  DocList *pOut      /* Write the combined doclist here */
){
  DocListReader left, right;
  sqlite_int64 docidLeft, docidRight, priorLeft;

  readerInit(&left, pLeft);
  readerInit(&right, pRight);
  docidLeft = nextDocid(&left);
  docidRight = nextDocid(&right);

  while( docidLeft>0 && docidRight>0 ){
    priorLeft = docidLeft;
    if( docidLeft<docidRight ){
      docListAddDocid(pOut, docidLeft);
    }
    if( docidLeft<=docidRight ){
      docidLeft = nextDocid(&left);
    }
    if( docidRight>0 && docidRight<=priorLeft ){
      docidRight = nextDocid(&right);
    }
  }
  while( docidLeft>0 ){
    docListAddDocid(pOut, docidLeft);
    docidLeft = nextDocid(&left);
  }
}

static char *string_dup_n(const char *s, int n){
  char *str = malloc(n + 1);
  memcpy(str, s, n);
  str[n] = '\0';
................................................................................
    doclist = old;
  }
  if( rc!=SQLITE_DONE ){
    docListDestroy(&doclist);
    return rc;
  }

  docListDiscardEmpty(&doclist);
  *out = doclist;
  return SQLITE_OK;
}

/* insert into %_term (rowid, term, segment, doclist)
               values ([piRowid], [pTerm], [iSegment], [doclist])
** Lets sqlite select rowid if piRowid is NULL, else uses *piRowid.
................................................................................
        c->eof = 1;
        return rc;
    }
  } else {  /* full-text query */
    rc = sqlite3_reset(c->pStmt);
    if( rc!=SQLITE_OK ) return rc;

    iDocid = nextDocid(&c->result);
    if( iDocid==0 ){
      c->eof = 1;
      return SQLITE_OK;
    }
    rc = sqlite3_bind_int64(c->pStmt, 1, iDocid);
    if( rc!=SQLITE_OK ) return rc;
    /* TODO(shess) Handle SQLITE_SCHEMA AND SQLITE_BUSY. */