SQLite

Check-in [c69c3e21db]
Login

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

Overview
Comment:Add an option to generate stat1 data based on a subset of the user database table contents to sqlite3_expert.
Downloads: Tarball | ZIP archive
Timelines: family | ancestors | descendants | both | schemalint
Files: files | file ages | folders
SHA3-256: c69c3e21db6e141f7e24226c6432f2ed31fe5f177bd23781915871f8600ee56a
User & Date: dan 2017-04-20 09:54:04.700
Context
2017-04-20
16:18
Avoid adding INTEGER PRIMARY KEY columns to candidate indexes. (check-in: 4577fea5cd user: dan tags: schemalint)
16:08
Avoid creating a temp table in the user database in the sqlite3_expert code. Trouble is, this makes sampling for stat1 data much slower. (check-in: c62e358243 user: dan tags: schemalint-failure)
09:54
Add an option to generate stat1 data based on a subset of the user database table contents to sqlite3_expert. (check-in: c69c3e21db user: dan tags: schemalint)
2017-04-18
20:10
Have sqlite3_expert_analyze() populate the sqlite_stat1 table before running queries through the planner for the second time. (check-in: a157fcfde5 user: dan tags: schemalint)
Changes
Side-by-Side Diff Ignore Whitespace Patch
Changes to ext/expert/expert.c.
32
33
34
35
36
37
38

39
40
41
42
43
44
45
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46







+







  fprintf(stderr, "\n");
  fprintf(stderr, "Usage %s ?OPTIONS? DATABASE\n", argv[0]);
  fprintf(stderr, "\n");
  fprintf(stderr, "Options are:\n");
  fprintf(stderr, "  -sql SQL   (analyze SQL statements passed as argument)\n");
  fprintf(stderr, "  -file FILE (read SQL statements from file FILE)\n");
  fprintf(stderr, "  -verbose LEVEL (integer verbosity level. default 1)\n");
  fprintf(stderr, "  -sample PERCENT (percent of db to sample. default 100)\n");
  exit(-1);
}

static int readSqlFromFile(sqlite3expert *p, const char *zFile, char **pzErr){
  FILE *in = fopen(zFile, "rb");
  long nIn;
  size_t nRead;
93
94
95
96
97
98
99
100

101
102
103







104
105
106
107
108
109
110
94
95
96
97
98
99
100

101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118







-
+



+
+
+
+
+
+
+







      char *zArg = argv[i];
      int nArg = strlen(zArg);
      if( nArg>=2 && 0==sqlite3_strnicmp(zArg, "-file", nArg) ){
        if( ++i==(argc-1) ) option_requires_argument("-file");
        rc = readSqlFromFile(p, argv[i], &zErr);
      }

      else if( nArg>=2 && 0==sqlite3_strnicmp(zArg, "-sql", nArg) ){
      else if( nArg>=3 && 0==sqlite3_strnicmp(zArg, "-sql", nArg) ){
        if( ++i==(argc-1) ) option_requires_argument("-sql");
        rc = sqlite3_expert_sql(p, argv[i], &zErr);
      }

      else if( nArg>=3 && 0==sqlite3_strnicmp(zArg, "-sample", nArg) ){
        int iSample;
        if( ++i==(argc-1) ) option_requires_argument("-sample");
        iSample = option_integer_arg(argv[i]);
        sqlite3_expert_config(p, EXPERT_CONFIG_SAMPLE, iSample);
      }

      else if( nArg>=2 && 0==sqlite3_strnicmp(zArg, "-verbose", nArg) ){
        if( ++i==(argc-1) ) option_requires_argument("-verbose");
        iVerbose = option_integer_arg(argv[i]);
      }

      else{
Changes to ext/expert/sqlite3expert.c.
23
24
25
26
27
28
29



30
31
32
33
34
35
36
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39







+
+
+








typedef struct IdxColumn IdxColumn;
typedef struct IdxConstraint IdxConstraint;
typedef struct IdxScan IdxScan;
typedef struct IdxStatement IdxStatement;
typedef struct IdxTable IdxTable;
typedef struct IdxWrite IdxWrite;

#define UNIQUE_TABLE_NAME "t592690916721053953805701627921227776"


/*
** A single constraint. Equivalent to either "col = ?" or "col < ?" (or
** any other type of single-ended range constraint on a column).
**
** pLink:
**   Used to temporarily link IdxConstraint objects into lists while
123
124
125
126
127
128
129

130
131
132
133
134
135
136
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140







+







  IdxHashEntry *aHash[IDX_HASH_SIZE];
};

/*
** sqlite3expert object.
*/
struct sqlite3expert {
  int iSample;                    /* Percentage of tables to sample for stat1 */
  sqlite3 *db;                    /* User database */
  sqlite3 *dbm;                   /* In-memory db for this analysis */
  sqlite3 *dbv;                   /* Vtab schema for this analysis */
  IdxTable *pTable;               /* List of all IdxTable objects */
  IdxScan *pScan;                 /* List of scan objects */
  IdxWrite *pWrite;               /* List of write objects */
  IdxStatement *pStatement;       /* List of IdxStatement objects */
1076
1077
1078
1079
1080
1081
1082
1083
1084


1085
1086
1087
1088
1089
1090
1091
1080
1081
1082
1083
1084
1085
1086


1087
1088
1089
1090
1091
1092
1093
1094
1095







-
-
+
+







}

static int idxProcessOneTrigger(
  sqlite3expert *p, 
  IdxWrite *pWrite, 
  char **pzErr
){
  static const char *zInt = "t592690916721053953805701627921227776";
  static const char *zDrop = "DROP TABLE t592690916721053953805701627921227776";
  static const char *zInt = UNIQUE_TABLE_NAME;
  static const char *zDrop = "DROP TABLE " UNIQUE_TABLE_NAME;
  IdxTable *pTab = pWrite->pTab;
  const char *zTab = pTab->zName;
  const char *zSql = 
    "SELECT 'CREATE TEMP' || substr(sql, 7) FROM sqlite_master "
    "WHERE tbl_name = %Q AND type IN ('table', 'trigger') "
    "ORDER BY type;";
  sqlite3_stmt *pSelect = 0;
1230
1231
1232
1233
1234
1235
1236
































1237
1238
1239
1240
1241
1242
1243
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279







+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+







        sqlite3_free(zOuter);
      }
    }
  }
  idxFinalize(&rc, pSchema);
  return rc;
}

struct IdxSampleCtx {
  int iTarget;
  double target;                  /* Target nRet/nRow value */
  double nRow;                    /* Number of rows seen */
  double nRet;                    /* Number of rows returned */
};

static void idxSampleFunc(
  sqlite3_context *pCtx,
  int argc,
  sqlite3_value **argv
){
  struct IdxSampleCtx *p = (struct IdxSampleCtx*)sqlite3_user_data(pCtx);
  int bRet;

  assert( argc==0 );
  if( p->nRow==0.0 ){
    bRet = 1;
  }else{
    bRet = (p->nRet / p->nRow) <= p->target;
    if( bRet==0 ){
      unsigned short rnd;
      sqlite3_randomness(2, (void*)&rnd);
      bRet = ((int)rnd % 100) <= p->iTarget;
    }
  }

  sqlite3_result_int(pCtx, bRet);
  p->nRow += 1.0;
  p->nRet += (double)bRet;
}

struct IdxRemCtx {
  int nSlot;
  struct IdxRemSlot {
    int eType;                    /* SQLITE_NULL, INTEGER, REAL, TEXT, BLOB */
    i64 iVal;                     /* SQLITE_INTEGER value */
    double rVal;                  /* SQLITE_FLOAT value */
1355
1356
1357
1358
1359
1360
1361


1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374

1375
1376
1377








1378
1379
1380
1381
1382
1383
1384
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413



1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428







+
+













+
-
-
-
+
+
+
+
+
+
+
+







  char *zOrder = 0;
  char *zQuery = 0;
  int nCol = 0;
  int i;
  sqlite3_stmt *pQuery = 0;
  int *aStat = 0;
  int rc = SQLITE_OK;

  assert( p->iSample>0 );

  /* Formulate the query text */
  sqlite3_bind_text(pIndexXInfo, 1, zIdx, -1, SQLITE_STATIC);
  while( SQLITE_OK==rc && SQLITE_ROW==sqlite3_step(pIndexXInfo) ){
    const char *zComma = zCols==0 ? "" : ", ";
    const char *zName = (const char*)sqlite3_column_text(pIndexXInfo, 0);
    const char *zColl = (const char*)sqlite3_column_text(pIndexXInfo, 1);
    zCols = idxAppendText(&rc, zCols, 
        "%sx.%Q IS rem(%d, x.%Q) COLLATE %s", zComma, zName, nCol, zName, zColl
    );
    zOrder = idxAppendText(&rc, zOrder, "%s%d", zComma, ++nCol);
  }
  if( rc==SQLITE_OK ){
    if( p->iSample==100 ){
    zQuery = sqlite3_mprintf(
        "SELECT %s FROM %Q x ORDER BY %s", zCols, zTab, zOrder
    );
      zQuery = sqlite3_mprintf(
          "SELECT %s FROM %Q x ORDER BY %s", zCols, zTab, zOrder
      );
    }else{
      zQuery = sqlite3_mprintf(
          "SELECT %s FROM temp."UNIQUE_TABLE_NAME" x ORDER BY %s", zCols, zOrder
      );
    }
  }
  sqlite3_free(zCols);
  sqlite3_free(zOrder);

  /* Formulate the query text */
  if( rc==SQLITE_OK ){
    rc = idxPrepareStmt(p->db, &pQuery, pzErr, zQuery);
1428
1429
1430
1431
1432
1433
1434



















1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446

1447

1448
1449
1450
1451
1452
1453


1454
1455
1456
1457
1458
1459



1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475





1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488

1489
1490










1491






1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505


1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525

1526
1527
1528
1529
1530
1531
1532
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516


1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562


1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623







+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+












+

+




-
-
+
+






+
+
+
















+
+
+
+
+













+
-
-
+
+
+
+
+
+
+
+
+
+

+
+
+
+
+
+














+
+




















+







    }
  }
  sqlite3_free(aStat);
  idxFinalize(&rc, pQuery);

  return rc;
}

static int idxBuildSampleTable(sqlite3expert *p, const char *zTab){
  int rc;
  char *zSql;

  rc = sqlite3_exec(p->db, "DROP TABLE IF EXISTS temp."UNIQUE_TABLE_NAME,0,0,0);
  if( rc!=SQLITE_OK ) return rc;

  zSql = sqlite3_mprintf(
      "CREATE TABLE temp." UNIQUE_TABLE_NAME 
      "  AS SELECT * FROM %Q WHERE sample()"
      , zTab
  );
  if( zSql==0 ) return SQLITE_NOMEM;
  rc = sqlite3_exec(p->db, zSql, 0, 0, 0);
  sqlite3_free(zSql);

  return rc;
}

/*
** This function is called as part of sqlite3_expert_analyze(). Candidate
** indexes have already been created in database sqlite3expert.dbm, this
** function populates sqlite_stat1 table in the same database.
**
** The stat1 data is generated by querying the 
*/
static int idxPopulateStat1(sqlite3expert *p, char **pzErr){
  int rc = SQLITE_OK;
  int nMax =0;
  struct IdxRemCtx *pCtx = 0;
  struct IdxSampleCtx samplectx; 
  int i;
  i64 iPrev = -100000;
  sqlite3_stmt *pAllIndex = 0;
  sqlite3_stmt *pIndexXInfo = 0;
  sqlite3_stmt *pWrite = 0;

  const char *zAllIndex = 
    "SELECT s.name, l.name FROM "
  const char *zAllIndex =
    "SELECT s.rowid, s.name, l.name FROM "
    "  sqlite_master AS s, "
    "  pragma_index_list(s.name) AS l "
    "WHERE s.type = 'table'";
  const char *zIndexXInfo = 
    "SELECT name, coll FROM pragma_index_xinfo(?) WHERE key";
  const char *zWrite = "INSERT INTO sqlite_stat1 VALUES(?, ?, ?)";

  /* If iSample==0, no sqlite_stat1 data is required. */
  if( p->iSample==0 ) return SQLITE_OK;

  rc = idxLargestIndex(p->dbm, &nMax, pzErr);
  if( nMax<=0 || rc!=SQLITE_OK ) return rc;

  rc = sqlite3_exec(p->dbm, "ANALYZE; PRAGMA writable_schema=1", 0, 0, 0);

  if( rc==SQLITE_OK ){
    int nByte = sizeof(struct IdxRemCtx) + (sizeof(struct IdxRemSlot) * nMax);
    pCtx = (struct IdxRemCtx*)idxMalloc(&rc, nByte);
  }

  if( rc==SQLITE_OK ){
    rc = sqlite3_create_function(
        p->db, "rem", 2, SQLITE_UTF8, (void*)pCtx, idxRemFunc, 0, 0
    );
  }
  if( rc==SQLITE_OK ){
    rc = sqlite3_create_function(
        p->db, "sample", 0, SQLITE_UTF8, (void*)&samplectx, idxSampleFunc, 0, 0
    );
  }

  if( rc==SQLITE_OK ){
    pCtx->nSlot = nMax+1;
    rc = idxPrepareStmt(p->dbm, &pAllIndex, pzErr, zAllIndex);
  }
  if( rc==SQLITE_OK ){
    rc = idxPrepareStmt(p->dbm, &pIndexXInfo, pzErr, zIndexXInfo);
  }
  if( rc==SQLITE_OK ){
    rc = idxPrepareStmt(p->dbm, &pWrite, pzErr, zWrite);
  }

  while( rc==SQLITE_OK && SQLITE_ROW==sqlite3_step(pAllIndex) ){
    i64 iRowid = sqlite3_column_int64(pAllIndex, 0);
    const char *zTab = (const char*)sqlite3_column_text(pAllIndex, 0);
    const char *zIdx = (const char*)sqlite3_column_text(pAllIndex, 1);
    const char *zTab = (const char*)sqlite3_column_text(pAllIndex, 1);
    const char *zIdx = (const char*)sqlite3_column_text(pAllIndex, 2);
    if( p->iSample<100 && iPrev!=iRowid ){
      samplectx.target = (double)p->iSample / 100.0;
      samplectx.iTarget = p->iSample;
      samplectx.nRow = 0.0;
      samplectx.nRet = 0.0;
      rc = idxBuildSampleTable(p, zTab);
      if( rc!=SQLITE_OK ) break;
    }
    rc = idxPopulateOneStat1(p, pIndexXInfo, pWrite, zTab, zIdx, pzErr);
    iPrev = iRowid;
  }
  if( p->iSample<100 ){
    rc = sqlite3_exec(p->db, "DROP TABLE IF EXISTS temp." UNIQUE_TABLE_NAME,
        0,0,0
    );
  }

  idxFinalize(&rc, pAllIndex);
  idxFinalize(&rc, pIndexXInfo);
  idxFinalize(&rc, pWrite);

  for(i=0; i<pCtx->nSlot; i++){
    sqlite3_free(pCtx->aSlot[i].z);
  }
  sqlite3_free(pCtx);

  if( rc==SQLITE_OK ){
    rc = sqlite3_exec(p->dbm, "ANALYZE sqlite_master", 0, 0, 0);
  }

  sqlite3_exec(p->db, "DROP TABLE IF EXISTS temp."UNIQUE_TABLE_NAME,0,0,0);
  return rc;
}

/*
** Allocate a new sqlite3expert object.
*/
sqlite3expert *sqlite3_expert_new(sqlite3 *db, char **pzErrmsg){
  int rc = SQLITE_OK;
  sqlite3expert *pNew;

  pNew = (sqlite3expert*)idxMalloc(&rc, sizeof(sqlite3expert));

  /* Open two in-memory databases to work with. The "vtab database" (dbv)
  ** will contain a virtual table corresponding to each real table in
  ** the user database schema, and a copy of each view. It is used to
  ** collect information regarding the WHERE, ORDER BY and other clauses
  ** of the user's query.
  */
  if( rc==SQLITE_OK ){
    pNew->db = db;
    pNew->iSample = 100;
    rc = sqlite3_open(":memory:", &pNew->dbv);
  }
  if( rc==SQLITE_OK ){
    rc = sqlite3_open(":memory:", &pNew->dbm);
    if( rc==SQLITE_OK ){
      sqlite3_db_config(pNew->dbm, SQLITE_DBCONFIG_FULL_EQP, 1, (int*)0);
    }
1560
1561
1562
1563
1564
1565
1566
























1567
1568
1569
1570
1571
1572
1573
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688







+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+







  ** return the new sqlite3expert handle.  */
  if( rc!=SQLITE_OK ){
    sqlite3_expert_destroy(pNew);
    pNew = 0;
  }
  return pNew;
}

/*
** Configure an sqlite3expert object.
*/
int sqlite3_expert_config(sqlite3expert *p, int op, ...){
  int rc = SQLITE_OK;
  va_list ap;
  va_start(ap, op);
  switch( op ){
    case EXPERT_CONFIG_SAMPLE: {
      int iVal = va_arg(ap, int);
      if( iVal<0 ) iVal = 0;
      if( iVal>100 ) iVal = 100;
      p->iSample = iVal;
      break;
    }
    default:
      rc = SQLITE_NOTFOUND;
      break;
  }

  va_end(ap);
  return rc;
}

/*
** Add an SQL statement to the analysis.
*/
int sqlite3_expert_sql(
  sqlite3expert *p,               /* From sqlite3_expert_new() */
  const char *zSql,               /* SQL statement to add */
Changes to ext/expert/sqlite3expert.h.
23
24
25
26
27
28
29
































30
31
32
33
34
35
36
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68







+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+







** to NULL. Or, if an error occurs, NULL is returned and (*pzErr) set to
** an English-language error message. In this case it is the responsibility
** of the caller to eventually free the error message buffer using
** sqlite3_free().
*/
sqlite3expert *sqlite3_expert_new(sqlite3 *db, char **pzErr);

/*
** Configure an sqlite3expert object.
**
** EXPERT_CONFIG_SAMPLE:
**   By default, sqlite3_expert_analyze() generates sqlite_stat1 data for
**   each candidate index. This involves scanning and sorting the entire
**   contents of each user database table once for each candidate index
**   associated with the table. For large databases, this can be 
**   prohibitively slow. This option allows the sqlite3expert object to
**   be configured so that sqlite_stat1 data is instead generated based on a
**   subset of each table, or so that no sqlite_stat1 data is used at all.
**
**   A single integer argument is passed to this option. If the value is less
**   than or equal to zero, then no sqlite_stat1 data is generated or used by
**   the analysis - indexes are recommended based on the database schema only.
**   Or, if the value is 100 or greater, complete sqlite_stat1 data is
**   generated for each candidate index (this is the default). Finally, if the
**   value falls between 0 and 100, then it represents the percentage of user
**   table rows that should be considered when generating sqlite_stat1 data.
**
**   Examples:
**
**     // Do not generate any sqlite_stat1 data
**     sqlite3_expert_config(pExpert, EXPERT_CONFIG_SAMPLE, 0);
**
**     // Generate sqlite_stat1 data based on 10% of the rows in each table.
**     sqlite3_expert_config(pExpert, EXPERT_CONFIG_SAMPLE, 10);
*/
int sqlite3_expert_config(sqlite3expert *p, int op, ...);

#define EXPERT_CONFIG_SAMPLE 1    /* int */

/*
** Specify zero or more SQL statements to be included in the analysis.
**
** Buffer zSql must contain zero or more complete SQL statements. This
** function parses all statements contained in the buffer and adds them
** to the internal list of statements to analyze. If successful, SQLITE_OK
** is returned and (*pzErr) set to NULL. Or, if an error occurs - for example
49
50
51
52
53
54
55

56
57
58
59
60
61
62
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95







+







** immediately and no statements are added to the analysis.
*/
int sqlite3_expert_sql(
  sqlite3expert *p,               /* From a successful sqlite3_expert_new() */
  const char *zSql,               /* SQL statement(s) to add */
  char **pzErr                    /* OUT: Error message (if any) */
);


/*
** This function is called after the sqlite3expert object has been configured
** with all SQL statements using sqlite3_expert_sql() to actually perform
** the analysis. Once this function has been called, it is not possible to
** add further SQL statements to the analysis.
**