SQLite

  EXPLAIN QUERY PLAN
  SELECT COUNT(*)
    FROM t1
   WHERE bb=21
     AND aa=1
     AND dd BETWEEN 1413833728 and 1413837331;
} {/INDEX good .bb=. AND aa=. AND dd>. AND dd<../}


# Create a table containing 100 rows. Column "a" contains a copy of the
# rowid value - sequentially increasing integers from 1 to 100. Column
# "b" contains the value of (a % 5). Columns "c" and "d" both contain
# constant values (i.e. the same for every row).
#
# Then create a second table t2. t2 is the same as t3 except for the
# order in which the indexes are created.
#
do_execsql_test 3.0 {
  CREATE TABLE t3(a, b, c, d);
  CREATE INDEX t3_ba ON t3(b, a, c);
  CREATE INDEX t3_a ON t3(a);

  WITH d(a, b) AS (
    SELECT 1, 1 
    UNION ALL
    SELECT a+1, (a+1) % 5 FROM d WHERE a<100
  )
  INSERT INTO t3 SELECT a, b, 'c', 'd' FROM d;

  CREATE TABLE t2(a, b, c, d);
  CREATE INDEX t2_a ON t2(a);
  CREATE INDEX t2_ba ON t2(b, a, c);
  INSERT INTO t2 SELECT * FROM t3;

  ANALYZE;
  SELECT * FROM sqlite_stat1;
} {
  t2 t2_ba   {100 20 1 1}
  t2 t2_a    {100 1} 
  t3 t3_a    {100 1} 
  t3 t3_ba   {100 20 1 1}
}

# Use index "t3_a", as (a=?) is expected to match only a single row.
#
do_eqp_test 3.1 {
  SELECT * FROM t3 WHERE a = ? AND c = ?
} {
  0 0 0 {SEARCH TABLE t3 USING INDEX t3_a (a=?)}
}

# The same query on table t2. This should use index "t2_a", for the
# same reason. At one point though, it was mistakenly using a skip-scan.
#
do_eqp_test 3.2 {
  SELECT * FROM t2 WHERE a = ? AND c = ?
} {
  0 0 0 {SEARCH TABLE t2 USING INDEX t2_a (a=?)}
}

finish_test




finish_test