sqlite/test/autoindex1.test

# 2010 April 07
#
# The author disclaims copyright to this source code.  In place of
# a legal notice, here is a blessing:
#
#    May you do good and not evil.
#    May you find forgiveness for yourself and forgive others.
#    May you share freely, never taking more than you give.
#
#*************************************************************************
# This file implements regression tests for SQLite library.  The
# focus of this script is testing automatic index creation logic.
#

set testdir [file dirname $argv0]
source $testdir/tester.tcl

# If the library is not compiled with automatic index support then
# skip all tests in this file.
#
ifcapable {!autoindex} {
  finish_test
  return
}

# With automatic index turned off, we do a full scan of the T2 table
do_test autoindex1-100 {
  db eval {
    CREATE TABLE t1(a,b);
    INSERT INTO t1 VALUES(1,11);
    INSERT INTO t1 VALUES(2,22);
    INSERT INTO t1 SELECT a+2, b+22 FROM t1;
    INSERT INTO t1 SELECT a+4, b+44 FROM t1;
    CREATE TABLE t2(c,d);
    INSERT INTO t2 SELECT a, 900+b FROM t1;
  }
  db eval {
    PRAGMA automatic_index=OFF;
    SELECT b, d FROM t1 JOIN t2 ON a=c ORDER BY b;
  }
} {11 911 22 922 33 933 44 944 55 955 66 966 77 977 88 988}
do_test autoindex1-101 {
  db status step
} {63}
do_test autoindex1-102 {
  db status autoindex
} {0}

# With autoindex turned on, we build an index once and then use that index
# to find T2 values.
do_test autoindex1-110 {
  db eval {
    PRAGMA automatic_index=ON;
    SELECT b, d FROM t1 JOIN t2 ON a=c ORDER BY b;
  }
} {11 911 22 922 33 933 44 944 55 955 66 966 77 977 88 988}
do_test autoindex1-111 {
  db status step
} {7}
do_test autoindex1-112 {
  db status autoindex
} {7}

# The same test as above, but this time the T2 query is a subquery rather
# than a join.
do_test autoindex1-200 {
  db eval {
    PRAGMA automatic_index=OFF;
    SELECT b, (SELECT d FROM t2 WHERE c=a) FROM t1;
  }
} {11 911 22 922 33 933 44 944 55 955 66 966 77 977 88 988}
do_test autoindex1-201 {
  db status step
} {35}
do_test autoindex1-202 {
  db status autoindex
} {0}
do_test autoindex1-210 {
  db eval {
    PRAGMA automatic_index=ON;
    SELECT b, (SELECT d FROM t2 WHERE c=a) FROM t1;
  }
} {11 911 22 922 33 933 44 944 55 955 66 966 77 977 88 988}
do_test autoindex1-211 {
  db status step
} {7}
do_test autoindex1-212 {
  db status autoindex
} {7}


# Modify the second table of the join while the join is in progress
#
do_test autoindex1-300 {
  set r {}
  db eval {SELECT b, d FROM t1 JOIN t2 ON (c=a)} {
    lappend r $b $d
    db eval {UPDATE t2 SET d=d+1}
  }
  set r
} {11 911 22 922 33 933 44 944 55 955 66 966 77 977 88 988}
do_test autoindex1-310 {
  db eval {SELECT d FROM t2 ORDER BY d}
} {919 930 941 952 963 974 985 996}

# The next test does a 10-way join on unindexed tables.  Without
# automatic indices, the join will take a long time to complete.
# With automatic indices, it should only take about a second.
#
do_test autoindex1-400 {
  db eval {
    CREATE TABLE t4(a, b);
    INSERT INTO t4 VALUES(1,2);
    INSERT INTO t4 VALUES(2,3);
  }
  for {set n 2} {$n<4096} {set n [expr {$n+$n}]} {
    db eval {INSERT INTO t4 SELECT a+$n, b+$n FROM t4}
  }
  db eval {
    SELECT count(*) FROM t4;
  }
} {4096}
do_test autoindex1-401 {
  db eval {
    SELECT count(*)
      FROM t4 AS x1
      JOIN t4 AS x2 ON x2.a=x1.b
      JOIN t4 AS x3 ON x3.a=x2.b
      JOIN t4 AS x4 ON x4.a=x3.b
      JOIN t4 AS x5 ON x5.a=x4.b
      JOIN t4 AS x6 ON x6.a=x5.b
      JOIN t4 AS x7 ON x7.a=x6.b
      JOIN t4 AS x8 ON x8.a=x7.b
      JOIN t4 AS x9 ON x9.a=x8.b
      JOIN t4 AS x10 ON x10.a=x9.b;
  }
} {4087}

# Ticket [8011086c85c6c404014c947fcf3eb9f42b184a0d] from 2010-07-08
# Make sure automatic indices are not created for the RHS of an IN expression
# that is not a correlated subquery.
#
do_execsql_test autoindex1-500 {
  CREATE TABLE t501(a INTEGER PRIMARY KEY, b);
  CREATE TABLE t502(x INTEGER PRIMARY KEY, y);
  EXPLAIN QUERY PLAN
  SELECT b FROM t501
   WHERE t501.a IN (SELECT x FROM t502 WHERE y=?);
} {
  0 0 0 {SEARCH TABLE t501 USING INTEGER PRIMARY KEY (rowid=?) (~25 rows)}
  0 0 0 {EXECUTE LIST SUBQUERY 1}
  1 0 0 {SCAN TABLE t502 (~100000 rows)}
}
do_execsql_test autoindex1-501 {
  EXPLAIN QUERY PLAN
  SELECT b FROM t501
   WHERE t501.a IN (SELECT x FROM t502 WHERE y=t501.b);
} {
  0 0 0 {SCAN TABLE t501 (~500000 rows)}
  0 0 0 {EXECUTE CORRELATED LIST SUBQUERY 1}
  1 0 0 {SEARCH TABLE t502 USING AUTOMATIC COVERING INDEX (y=?) (~7 rows)}
}
do_execsql_test autoindex1-502 {
  EXPLAIN QUERY PLAN
  SELECT b FROM t501
   WHERE t501.a=123
     AND t501.a IN (SELECT x FROM t502 WHERE y=t501.b);
} {
  0 0 0 {SEARCH TABLE t501 USING INTEGER PRIMARY KEY (rowid=?) (~1 rows)}
  0 0 0 {EXECUTE CORRELATED LIST SUBQUERY 1}
  1 0 0 {SCAN TABLE t502 (~100000 rows)}
}


# The following code checks a performance regression reported on the
# mailing list on 2010-10-19.  The problem is that the nRowEst field
# of ephermeral tables was not being initialized correctly and so no
# automatic index was being created for the emphemeral table when it was
# used as part of a join.
#
do_execsql_test autoindex1-600 {
  CREATE TABLE flock_owner(
    owner_rec_id INTEGER CONSTRAINT flock_owner_key PRIMARY KEY,
    flock_no VARCHAR(6) NOT NULL REFERENCES flock (flock_no),
    owner_person_id INTEGER NOT NULL REFERENCES person (person_id),
    owner_change_date TEXT, last_changed TEXT NOT NULL,
    CONSTRAINT fo_owner_date UNIQUE (flock_no, owner_change_date)
  );
  CREATE TABLE sheep (
    Sheep_No char(7) NOT NULL,
    Date_of_Birth char(8),
    Sort_DoB text,
    Flock_Book_Vol char(2),
    Breeder_No char(6),
    Breeder_Person integer,
    Originating_Flock char(6),
    Registering_Flock char(6),
    Tag_Prefix char(9),
    Tag_No char(15),
    Sort_Tag_No integer,
    Breeders_Temp_Tag char(15),
    Sex char(1),
    Sheep_Name char(32),
    Sire_No char(7),
    Dam_No char(7),
    Register_Code char(1),
    Colour char(48),
    Colour_Code char(2),
    Pattern_Code char(8),
    Horns char(1),
    Litter_Size char(1),
    Coeff_of_Inbreeding real,
    Date_of_Registration text,
    Date_Last_Changed text,
    UNIQUE(Sheep_No));
  CREATE INDEX fo_flock_no_index
              ON flock_owner (flock_no);
  CREATE INDEX fo_owner_change_date_index
              ON flock_owner (owner_change_date);
  CREATE INDEX fo_owner_person_id_index
              ON flock_owner (owner_person_id);
  CREATE INDEX sheep_org_flock_index
           ON sheep (originating_flock);
  CREATE INDEX sheep_reg_flock_index
           ON sheep (registering_flock);
  EXPLAIN QUERY PLAN
  SELECT x.sheep_no, x.registering_flock, x.date_of_registration
   FROM sheep x LEFT JOIN
       (SELECT s.sheep_no, prev.flock_no, prev.owner_person_id,
       s.date_of_registration, prev.owner_change_date
       FROM sheep s JOIN flock_owner prev ON s.registering_flock =
   prev.flock_no
       AND (prev.owner_change_date <= s.date_of_registration || ' 00:00:00')
       WHERE NOT EXISTS
           (SELECT 'x' FROM flock_owner later
           WHERE prev.flock_no = later.flock_no
           AND later.owner_change_date > prev.owner_change_date
           AND later.owner_change_date <= s.date_of_registration||' 00:00:00')
       ) y ON x.sheep_no = y.sheep_no
   WHERE y.sheep_no IS NULL
   ORDER BY x.registering_flock;
} {
  1 0 0 {SCAN TABLE sheep AS s (~1000000 rows)}
  1 1 1 {SEARCH TABLE flock_owner AS prev USING INDEX sqlite_autoindex_flock_owner_1 (flock_no=? AND owner_change_date<?) (~2 rows)}
  1 0 0 {EXECUTE CORRELATED SCALAR SUBQUERY 2}
  2 0 0 {SEARCH TABLE flock_owner AS later USING COVERING INDEX sqlite_autoindex_flock_owner_1 (flock_no=? AND owner_change_date>? AND owner_change_date<?) (~1 rows)}
  0 0 0 {SCAN TABLE sheep AS x USING INDEX sheep_reg_flock_index (~1000000 rows)}
  0 1 1 {SEARCH SUBQUERY 1 AS y USING AUTOMATIC COVERING INDEX (sheep_no=?) (~8 rows)}
}

finish_test