/ Check-in [21981e35]
Login

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

Overview
Comment:Ensure that all fields are loaded from the stat4 table for records that correspond to indexes on WITHOUT ROWID tables with composite primary keys.
Downloads: Tarball | ZIP archive | SQL archive
Timelines: family | ancestors | descendants | both | trunk
Files: files | file ages | folders
SHA1:21981e35062cc6b30e9576786cbf55265a7a4d41
User & Date: dan 2014-07-01 11:54:02
Context
2014-07-01
15:22
Add another test to verify that SQLite is using stat4 data for composite primary keys on WITHOUT ROWID tables. check-in: 0df1fe72 user: dan tags: trunk
11:54
Ensure that all fields are loaded from the stat4 table for records that correspond to indexes on WITHOUT ROWID tables with composite primary keys. check-in: 21981e35 user: dan tags: trunk
2014-06-30
19:28
Bump the version number to 3.8.6. check-in: f925e9ba user: drh tags: trunk
Changes
Hide Diffs Unified Diffs Ignore Whitespace Patch

Changes to src/analyze.c.

1515
1516
1517
1518
1519
1520
1521









1522
1523
1524
1525
1526
1527
1528
1529
....
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
....
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
....
1615
1616
1617
1618
1619
1620
1621


1622

1623

1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
** stored in pIdx->aSample[]. 
*/
static void initAvgEq(Index *pIdx){
  if( pIdx ){
    IndexSample *aSample = pIdx->aSample;
    IndexSample *pFinal = &aSample[pIdx->nSample-1];
    int iCol;









    for(iCol=0; iCol<pIdx->nKeyCol; iCol++){
      int i;                    /* Used to iterate through samples */
      tRowcnt sumEq = 0;        /* Sum of the nEq values */
      tRowcnt nSum = 0;         /* Number of terms contributing to sumEq */
      tRowcnt avgEq = 0;
      tRowcnt nDLt = pFinal->anDLt[iCol];

      /* Set nSum to the number of distinct (iCol+1) field prefixes that
................................................................................
        }
      }
      if( nDLt>nSum ){
        avgEq = (pFinal->anLt[iCol] - sumEq)/(nDLt - nSum);
      }
      if( avgEq==0 ) avgEq = 1;
      pIdx->aAvgEq[iCol] = avgEq;
      if( pIdx->nSampleCol==1 ) break;
    }
  }
}

/*
** Look up an index by name.  Or, if the name of a WITHOUT ROWID table
** is supplied instead, find the PRIMARY KEY index for that table.
................................................................................
  }
  rc = sqlite3_prepare(db, zSql, -1, &pStmt, 0);
  sqlite3DbFree(db, zSql);
  if( rc ) return rc;

  while( sqlite3_step(pStmt)==SQLITE_ROW ){
    int nIdxCol = 1;              /* Number of columns in stat4 records */
    int nAvgCol = 1;              /* Number of entries in Index.aAvgEq */

    char *zIndex;   /* Index name */
    Index *pIdx;    /* Pointer to the index object */
    int nSample;    /* Number of samples */
    int nByte;      /* Bytes of space required */
    int i;          /* Bytes of space required */
    tRowcnt *pSpace;
................................................................................
    nSample = sqlite3_column_int(pStmt, 1);
    pIdx = findIndexOrPrimaryKey(db, zIndex, zDb);
    assert( pIdx==0 || bStat3 || pIdx->nSample==0 );
    /* Index.nSample is non-zero at this point if data has already been
    ** loaded from the stat4 table. In this case ignore stat3 data.  */
    if( pIdx==0 || pIdx->nSample ) continue;
    if( bStat3==0 ){


      nIdxCol = pIdx->nKeyCol+1;

      nAvgCol = pIdx->nKeyCol;

    }
    pIdx->nSampleCol = nIdxCol;
    nByte = sizeof(IndexSample) * nSample;
    nByte += sizeof(tRowcnt) * nIdxCol * 3 * nSample;
    nByte += nAvgCol * sizeof(tRowcnt);     /* Space for Index.aAvgEq[] */

    pIdx->aSample = sqlite3DbMallocZero(db, nByte);
    if( pIdx->aSample==0 ){
      sqlite3_finalize(pStmt);
      return SQLITE_NOMEM;
    }
    pSpace = (tRowcnt*)&pIdx->aSample[nSample];
    pIdx->aAvgEq = pSpace; pSpace += nAvgCol;
    for(i=0; i<nSample; i++){
      pIdx->aSample[i].anEq = pSpace; pSpace += nIdxCol;
      pIdx->aSample[i].anLt = pSpace; pSpace += nIdxCol;
      pIdx->aSample[i].anDLt = pSpace; pSpace += nIdxCol;
    }
    assert( ((u8*)pSpace)-nByte==(u8*)(pIdx->aSample) );
  }







>
>
>
>
>
>
>
>
>
|







 







<







 







<







 







>
>
|
>
|
>




|







|







1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
....
1547
1548
1549
1550
1551
1552
1553

1554
1555
1556
1557
1558
1559
1560
....
1605
1606
1607
1608
1609
1610
1611

1612
1613
1614
1615
1616
1617
1618
....
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
** stored in pIdx->aSample[]. 
*/
static void initAvgEq(Index *pIdx){
  if( pIdx ){
    IndexSample *aSample = pIdx->aSample;
    IndexSample *pFinal = &aSample[pIdx->nSample-1];
    int iCol;
    int nCol = 1;
    if( pIdx->nSampleCol>1 ){
      /* If this is stat4 data, then calculate aAvgEq[] values for all
      ** sample columns except the last. The last is always set to 1, as
      ** once the trailing PK fields are considered all index keys are
      ** unique.  */
      nCol = pIdx->nSampleCol-1;
      pIdx->aAvgEq[nCol] = 1;
    }
    for(iCol=0; iCol<nCol; iCol++){
      int i;                    /* Used to iterate through samples */
      tRowcnt sumEq = 0;        /* Sum of the nEq values */
      tRowcnt nSum = 0;         /* Number of terms contributing to sumEq */
      tRowcnt avgEq = 0;
      tRowcnt nDLt = pFinal->anDLt[iCol];

      /* Set nSum to the number of distinct (iCol+1) field prefixes that
................................................................................
        }
      }
      if( nDLt>nSum ){
        avgEq = (pFinal->anLt[iCol] - sumEq)/(nDLt - nSum);
      }
      if( avgEq==0 ) avgEq = 1;
      pIdx->aAvgEq[iCol] = avgEq;

    }
  }
}

/*
** Look up an index by name.  Or, if the name of a WITHOUT ROWID table
** is supplied instead, find the PRIMARY KEY index for that table.
................................................................................
  }
  rc = sqlite3_prepare(db, zSql, -1, &pStmt, 0);
  sqlite3DbFree(db, zSql);
  if( rc ) return rc;

  while( sqlite3_step(pStmt)==SQLITE_ROW ){
    int nIdxCol = 1;              /* Number of columns in stat4 records */


    char *zIndex;   /* Index name */
    Index *pIdx;    /* Pointer to the index object */
    int nSample;    /* Number of samples */
    int nByte;      /* Bytes of space required */
    int i;          /* Bytes of space required */
    tRowcnt *pSpace;
................................................................................
    nSample = sqlite3_column_int(pStmt, 1);
    pIdx = findIndexOrPrimaryKey(db, zIndex, zDb);
    assert( pIdx==0 || bStat3 || pIdx->nSample==0 );
    /* Index.nSample is non-zero at this point if data has already been
    ** loaded from the stat4 table. In this case ignore stat3 data.  */
    if( pIdx==0 || pIdx->nSample ) continue;
    if( bStat3==0 ){
      assert( !HasRowid(pIdx->pTable) || pIdx->nColumn==pIdx->nKeyCol+1 );
      if( !HasRowid(pIdx->pTable) && IsPrimaryKeyIndex(pIdx) ){
        nIdxCol = pIdx->nKeyCol;
      }else{
        nIdxCol = pIdx->nColumn;
      }
    }
    pIdx->nSampleCol = nIdxCol;
    nByte = sizeof(IndexSample) * nSample;
    nByte += sizeof(tRowcnt) * nIdxCol * 3 * nSample;
    nByte += nIdxCol * sizeof(tRowcnt);     /* Space for Index.aAvgEq[] */

    pIdx->aSample = sqlite3DbMallocZero(db, nByte);
    if( pIdx->aSample==0 ){
      sqlite3_finalize(pStmt);
      return SQLITE_NOMEM;
    }
    pSpace = (tRowcnt*)&pIdx->aSample[nSample];
    pIdx->aAvgEq = pSpace; pSpace += nIdxCol;
    for(i=0; i<nSample; i++){
      pIdx->aSample[i].anEq = pSpace; pSpace += nIdxCol;
      pIdx->aSample[i].anLt = pSpace; pSpace += nIdxCol;
      pIdx->aSample[i].anDLt = pSpace; pSpace += nIdxCol;
    }
    assert( ((u8*)pSpace)-nByte==(u8*)(pIdx->aSample) );
  }

Changes to src/where.c.

1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
      iLower = 0;
      iUpper = aSample[0].anLt[iCol];
    }else{
      i64 nRow0 = sqlite3LogEstToInt(pIdx->aiRowLogEst[0]);
      iUpper = i>=pIdx->nSample ? nRow0 : aSample[i].anLt[iCol];
      iLower = aSample[i-1].anEq[iCol] + aSample[i-1].anLt[iCol];
    }
    aStat[1] = (pIdx->nKeyCol>iCol ? pIdx->aAvgEq[iCol] : 1);
    if( iLower>=iUpper ){
      iGap = 0;
    }else{
      iGap = iUpper - iLower;
    }
    if( roundUp ){
      iGap = (iGap*2)/3;







|







1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
      iLower = 0;
      iUpper = aSample[0].anLt[iCol];
    }else{
      i64 nRow0 = sqlite3LogEstToInt(pIdx->aiRowLogEst[0]);
      iUpper = i>=pIdx->nSample ? nRow0 : aSample[i].anLt[iCol];
      iLower = aSample[i-1].anEq[iCol] + aSample[i-1].anLt[iCol];
    }
    aStat[1] = pIdx->aAvgEq[iCol];
    if( iLower>=iUpper ){
      iGap = 0;
    }else{
      iGap = iUpper - iLower;
    }
    if( roundUp ){
      iGap = (iGap*2)/3;

Changes to test/analyze9.test.

1014
1015
1016
1017
1018
1019
1020
1021












































1022
1023
1024
# and (d<20) over (c='one' AND a='A').
foreach {tn where res} {
  1 "c='one' AND a='B' AND d < 20"   {/*INDEX i3 (c=? AND a=?)*/}
  2 "c='one' AND a='A' AND d < 20"   {/*INDEX i4 (d<?)*/}
} {
  do_eqp_test 22.2.$tn "SELECT * FROM t3 WHERE $where" $res
}













































finish_test










>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>



1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
# and (d<20) over (c='one' AND a='A').
foreach {tn where res} {
  1 "c='one' AND a='B' AND d < 20"   {/*INDEX i3 (c=? AND a=?)*/}
  2 "c='one' AND a='A' AND d < 20"   {/*INDEX i4 (d<?)*/}
} {
  do_eqp_test 22.2.$tn "SELECT * FROM t3 WHERE $where" $res
}

proc int_to_char {i} {
  set ret ""
  set char [list a b c d e f g h i j]
  foreach {div} {1000 100 10 1} {
    append ret [lindex $char [expr ($i / $div) % 10]]
  }
  set ret
}
db func int_to_char int_to_char

do_execsql_test 23.0 {
  CREATE TABLE t4(
    a COLLATE nocase, b, c, 
    d, e, f, 
    PRIMARY KEY(c, b, a)
  ) WITHOUT ROWID;
  CREATE INDEX i41 ON t4(e);
  CREATE INDEX i42 ON t4(f);

  WITH data(a, b, c, d, e, f) AS (
    SELECT int_to_char(0), 'xyz', 'zyx', '*', 0, 0
    UNION ALL
    SELECT 
      int_to_char(f+1), b, c, d, (e+1) % 2, f+1
    FROM data WHERE f<1024
  )
  INSERT INTO t4 SELECT a, b, c, d, e, f FROM data;
  ANALYZE;
} {}

do_eqp_test 23.1 {
  SELECT * FROM t4 WHERE 
    (e=1 AND b='xyz' AND c='zyx' AND a<'AEA') AND f<300
} {
  0 0 0 {SEARCH TABLE t4 USING INDEX i41 (e=? AND c=? AND b=? AND a<?)}
}

do_eqp_test 23.2 {
  SELECT * FROM t4 WHERE 
    (e=1 AND b='xyz' AND c='zyx' AND a<'JJJ') AND f<300
} {
  0 0 0 {SEARCH TABLE t4 USING INDEX i42 (f<?)}
}

finish_test