2000-01-09 04:48:39 +01:00
|
|
|
--
|
|
|
|
-- JOIN
|
2000-09-12 23:07:18 +02:00
|
|
|
-- Test JOIN clauses
|
2000-01-09 04:48:39 +01:00
|
|
|
--
|
2000-02-15 04:31:33 +01:00
|
|
|
CREATE TABLE J1_TBL (
|
1999-02-23 08:27:13 +01:00
|
|
|
i integer,
|
|
|
|
j integer,
|
2000-01-09 04:48:39 +01:00
|
|
|
t text
|
1999-02-23 08:27:13 +01:00
|
|
|
);
|
2000-02-15 04:31:33 +01:00
|
|
|
CREATE TABLE J2_TBL (
|
1999-02-23 08:27:13 +01:00
|
|
|
i integer,
|
|
|
|
k integer
|
|
|
|
);
|
2000-11-06 17:03:47 +01:00
|
|
|
INSERT INTO J1_TBL VALUES (1, 4, 'one');
|
|
|
|
INSERT INTO J1_TBL VALUES (2, 3, 'two');
|
|
|
|
INSERT INTO J1_TBL VALUES (3, 2, 'three');
|
|
|
|
INSERT INTO J1_TBL VALUES (4, 1, 'four');
|
|
|
|
INSERT INTO J1_TBL VALUES (5, 0, 'five');
|
|
|
|
INSERT INTO J1_TBL VALUES (6, 6, 'six');
|
|
|
|
INSERT INTO J1_TBL VALUES (7, 7, 'seven');
|
|
|
|
INSERT INTO J1_TBL VALUES (8, 8, 'eight');
|
|
|
|
INSERT INTO J1_TBL VALUES (0, NULL, 'zero');
|
|
|
|
INSERT INTO J1_TBL VALUES (NULL, NULL, 'null');
|
|
|
|
INSERT INTO J1_TBL VALUES (NULL, 0, 'zero');
|
2000-02-15 04:31:33 +01:00
|
|
|
INSERT INTO J2_TBL VALUES (1, -1);
|
|
|
|
INSERT INTO J2_TBL VALUES (2, 2);
|
|
|
|
INSERT INTO J2_TBL VALUES (3, -3);
|
|
|
|
INSERT INTO J2_TBL VALUES (2, 4);
|
2000-09-12 23:07:18 +02:00
|
|
|
INSERT INTO J2_TBL VALUES (5, -5);
|
2000-11-06 17:03:47 +01:00
|
|
|
INSERT INTO J2_TBL VALUES (5, -5);
|
|
|
|
INSERT INTO J2_TBL VALUES (0, NULL);
|
|
|
|
INSERT INTO J2_TBL VALUES (NULL, NULL);
|
|
|
|
INSERT INTO J2_TBL VALUES (NULL, 0);
|
2000-02-15 04:31:33 +01:00
|
|
|
--
|
|
|
|
-- CORRELATION NAMES
|
|
|
|
-- Make sure that table/column aliases are supported
|
|
|
|
-- before diving into more complex join syntax.
|
|
|
|
--
|
|
|
|
SELECT '' AS "xxx", *
|
|
|
|
FROM J1_TBL AS tx;
|
|
|
|
xxx | i | j | t
|
|
|
|
-----+---+---+-------
|
2000-11-06 17:03:47 +01:00
|
|
|
| 1 | 4 | one
|
|
|
|
| 2 | 3 | two
|
|
|
|
| 3 | 2 | three
|
|
|
|
| 4 | 1 | four
|
|
|
|
| 5 | 0 | five
|
|
|
|
| 6 | 6 | six
|
|
|
|
| 7 | 7 | seven
|
|
|
|
| 8 | 8 | eight
|
|
|
|
| 0 | | zero
|
|
|
|
| | | null
|
|
|
|
| | 0 | zero
|
|
|
|
(11 rows)
|
2000-02-15 04:31:33 +01:00
|
|
|
|
|
|
|
SELECT '' AS "xxx", *
|
|
|
|
FROM J1_TBL tx;
|
|
|
|
xxx | i | j | t
|
|
|
|
-----+---+---+-------
|
2000-11-06 17:03:47 +01:00
|
|
|
| 1 | 4 | one
|
|
|
|
| 2 | 3 | two
|
|
|
|
| 3 | 2 | three
|
|
|
|
| 4 | 1 | four
|
|
|
|
| 5 | 0 | five
|
|
|
|
| 6 | 6 | six
|
|
|
|
| 7 | 7 | seven
|
|
|
|
| 8 | 8 | eight
|
|
|
|
| 0 | | zero
|
|
|
|
| | | null
|
|
|
|
| | 0 | zero
|
|
|
|
(11 rows)
|
2000-02-15 04:31:33 +01:00
|
|
|
|
|
|
|
SELECT '' AS "xxx", *
|
|
|
|
FROM J1_TBL AS t1 (a, b, c);
|
|
|
|
xxx | a | b | c
|
|
|
|
-----+---+---+-------
|
2000-11-06 17:03:47 +01:00
|
|
|
| 1 | 4 | one
|
|
|
|
| 2 | 3 | two
|
|
|
|
| 3 | 2 | three
|
|
|
|
| 4 | 1 | four
|
|
|
|
| 5 | 0 | five
|
|
|
|
| 6 | 6 | six
|
|
|
|
| 7 | 7 | seven
|
|
|
|
| 8 | 8 | eight
|
|
|
|
| 0 | | zero
|
|
|
|
| | | null
|
|
|
|
| | 0 | zero
|
|
|
|
(11 rows)
|
2000-02-15 04:31:33 +01:00
|
|
|
|
|
|
|
SELECT '' AS "xxx", *
|
|
|
|
FROM J1_TBL t1 (a, b, c);
|
|
|
|
xxx | a | b | c
|
|
|
|
-----+---+---+-------
|
2000-11-06 17:03:47 +01:00
|
|
|
| 1 | 4 | one
|
|
|
|
| 2 | 3 | two
|
|
|
|
| 3 | 2 | three
|
|
|
|
| 4 | 1 | four
|
|
|
|
| 5 | 0 | five
|
|
|
|
| 6 | 6 | six
|
|
|
|
| 7 | 7 | seven
|
|
|
|
| 8 | 8 | eight
|
|
|
|
| 0 | | zero
|
|
|
|
| | | null
|
|
|
|
| | 0 | zero
|
|
|
|
(11 rows)
|
2000-02-15 04:31:33 +01:00
|
|
|
|
|
|
|
SELECT '' AS "xxx", *
|
|
|
|
FROM J1_TBL t1 (a, b, c), J2_TBL t2 (d, e);
|
|
|
|
xxx | a | b | c | d | e
|
|
|
|
-----+---+---+-------+---+----
|
2000-11-06 17:03:47 +01:00
|
|
|
| 1 | 4 | one | 1 | -1
|
|
|
|
| 2 | 3 | two | 1 | -1
|
|
|
|
| 3 | 2 | three | 1 | -1
|
|
|
|
| 4 | 1 | four | 1 | -1
|
|
|
|
| 5 | 0 | five | 1 | -1
|
|
|
|
| 6 | 6 | six | 1 | -1
|
|
|
|
| 7 | 7 | seven | 1 | -1
|
|
|
|
| 8 | 8 | eight | 1 | -1
|
|
|
|
| 0 | | zero | 1 | -1
|
|
|
|
| | | null | 1 | -1
|
|
|
|
| | 0 | zero | 1 | -1
|
2004-12-02 02:34:18 +01:00
|
|
|
| 1 | 4 | one | 2 | 2
|
|
|
|
| 2 | 3 | two | 2 | 2
|
|
|
|
| 3 | 2 | three | 2 | 2
|
|
|
|
| 4 | 1 | four | 2 | 2
|
|
|
|
| 5 | 0 | five | 2 | 2
|
|
|
|
| 6 | 6 | six | 2 | 2
|
|
|
|
| 7 | 7 | seven | 2 | 2
|
|
|
|
| 8 | 8 | eight | 2 | 2
|
|
|
|
| 0 | | zero | 2 | 2
|
|
|
|
| | | null | 2 | 2
|
2000-11-06 17:03:47 +01:00
|
|
|
| | 0 | zero | 2 | 2
|
2004-12-02 02:34:18 +01:00
|
|
|
| 1 | 4 | one | 3 | -3
|
|
|
|
| 2 | 3 | two | 3 | -3
|
|
|
|
| 3 | 2 | three | 3 | -3
|
|
|
|
| 4 | 1 | four | 3 | -3
|
|
|
|
| 5 | 0 | five | 3 | -3
|
|
|
|
| 6 | 6 | six | 3 | -3
|
|
|
|
| 7 | 7 | seven | 3 | -3
|
|
|
|
| 8 | 8 | eight | 3 | -3
|
|
|
|
| 0 | | zero | 3 | -3
|
|
|
|
| | | null | 3 | -3
|
2000-11-06 17:03:47 +01:00
|
|
|
| | 0 | zero | 3 | -3
|
2004-12-02 02:34:18 +01:00
|
|
|
| 1 | 4 | one | 2 | 4
|
|
|
|
| 2 | 3 | two | 2 | 4
|
|
|
|
| 3 | 2 | three | 2 | 4
|
|
|
|
| 4 | 1 | four | 2 | 4
|
|
|
|
| 5 | 0 | five | 2 | 4
|
|
|
|
| 6 | 6 | six | 2 | 4
|
|
|
|
| 7 | 7 | seven | 2 | 4
|
|
|
|
| 8 | 8 | eight | 2 | 4
|
|
|
|
| 0 | | zero | 2 | 4
|
|
|
|
| | | null | 2 | 4
|
2000-11-06 17:03:47 +01:00
|
|
|
| | 0 | zero | 2 | 4
|
2004-12-02 02:34:18 +01:00
|
|
|
| 1 | 4 | one | 5 | -5
|
|
|
|
| 2 | 3 | two | 5 | -5
|
|
|
|
| 3 | 2 | three | 5 | -5
|
|
|
|
| 4 | 1 | four | 5 | -5
|
|
|
|
| 5 | 0 | five | 5 | -5
|
|
|
|
| 6 | 6 | six | 5 | -5
|
|
|
|
| 7 | 7 | seven | 5 | -5
|
|
|
|
| 8 | 8 | eight | 5 | -5
|
|
|
|
| 0 | | zero | 5 | -5
|
|
|
|
| | | null | 5 | -5
|
2000-11-06 17:03:47 +01:00
|
|
|
| | 0 | zero | 5 | -5
|
2004-12-02 02:34:18 +01:00
|
|
|
| 1 | 4 | one | 5 | -5
|
|
|
|
| 2 | 3 | two | 5 | -5
|
|
|
|
| 3 | 2 | three | 5 | -5
|
|
|
|
| 4 | 1 | four | 5 | -5
|
|
|
|
| 5 | 0 | five | 5 | -5
|
|
|
|
| 6 | 6 | six | 5 | -5
|
|
|
|
| 7 | 7 | seven | 5 | -5
|
|
|
|
| 8 | 8 | eight | 5 | -5
|
|
|
|
| 0 | | zero | 5 | -5
|
|
|
|
| | | null | 5 | -5
|
2000-11-06 17:03:47 +01:00
|
|
|
| | 0 | zero | 5 | -5
|
2004-12-02 02:34:18 +01:00
|
|
|
| 1 | 4 | one | 0 |
|
|
|
|
| 2 | 3 | two | 0 |
|
|
|
|
| 3 | 2 | three | 0 |
|
|
|
|
| 4 | 1 | four | 0 |
|
|
|
|
| 5 | 0 | five | 0 |
|
|
|
|
| 6 | 6 | six | 0 |
|
|
|
|
| 7 | 7 | seven | 0 |
|
|
|
|
| 8 | 8 | eight | 0 |
|
|
|
|
| 0 | | zero | 0 |
|
|
|
|
| | | null | 0 |
|
2000-11-06 17:03:47 +01:00
|
|
|
| | 0 | zero | 0 |
|
2004-12-02 02:34:18 +01:00
|
|
|
| 1 | 4 | one | |
|
|
|
|
| 2 | 3 | two | |
|
|
|
|
| 3 | 2 | three | |
|
|
|
|
| 4 | 1 | four | |
|
|
|
|
| 5 | 0 | five | |
|
|
|
|
| 6 | 6 | six | |
|
|
|
|
| 7 | 7 | seven | |
|
|
|
|
| 8 | 8 | eight | |
|
|
|
|
| 0 | | zero | |
|
|
|
|
| | | null | |
|
2000-11-06 17:03:47 +01:00
|
|
|
| | 0 | zero | |
|
2004-12-02 02:34:18 +01:00
|
|
|
| 1 | 4 | one | | 0
|
|
|
|
| 2 | 3 | two | | 0
|
|
|
|
| 3 | 2 | three | | 0
|
|
|
|
| 4 | 1 | four | | 0
|
|
|
|
| 5 | 0 | five | | 0
|
|
|
|
| 6 | 6 | six | | 0
|
|
|
|
| 7 | 7 | seven | | 0
|
|
|
|
| 8 | 8 | eight | | 0
|
|
|
|
| 0 | | zero | | 0
|
|
|
|
| | | null | | 0
|
2000-11-06 17:03:47 +01:00
|
|
|
| | 0 | zero | | 0
|
|
|
|
(99 rows)
|
2000-02-15 04:31:33 +01:00
|
|
|
|
|
|
|
SELECT '' AS "xxx", t1.a, t2.e
|
|
|
|
FROM J1_TBL t1 (a, b, c), J2_TBL t2 (d, e)
|
|
|
|
WHERE t1.a = t2.d;
|
|
|
|
xxx | a | e
|
|
|
|
-----+---+----
|
2005-03-24 20:14:49 +01:00
|
|
|
| 0 |
|
2000-02-15 04:31:33 +01:00
|
|
|
| 1 | -1
|
|
|
|
| 2 | 2
|
2004-12-01 20:00:56 +01:00
|
|
|
| 2 | 4
|
2005-03-24 20:14:49 +01:00
|
|
|
| 3 | -3
|
2000-11-06 17:03:47 +01:00
|
|
|
| 5 | -5
|
|
|
|
| 5 | -5
|
|
|
|
(7 rows)
|
2000-02-15 04:31:33 +01:00
|
|
|
|
2000-01-09 04:48:39 +01:00
|
|
|
--
|
|
|
|
-- CROSS JOIN
|
|
|
|
-- Qualifications are not allowed on cross joins,
|
|
|
|
-- which degenerate into a standard unqualified inner join.
|
|
|
|
--
|
|
|
|
SELECT '' AS "xxx", *
|
2000-02-15 04:31:33 +01:00
|
|
|
FROM J1_TBL CROSS JOIN J2_TBL;
|
2000-01-09 04:48:39 +01:00
|
|
|
xxx | i | j | t | i | k
|
|
|
|
-----+---+---+-------+---+----
|
2000-11-06 17:03:47 +01:00
|
|
|
| 1 | 4 | one | 1 | -1
|
|
|
|
| 2 | 3 | two | 1 | -1
|
|
|
|
| 3 | 2 | three | 1 | -1
|
|
|
|
| 4 | 1 | four | 1 | -1
|
|
|
|
| 5 | 0 | five | 1 | -1
|
|
|
|
| 6 | 6 | six | 1 | -1
|
|
|
|
| 7 | 7 | seven | 1 | -1
|
|
|
|
| 8 | 8 | eight | 1 | -1
|
|
|
|
| 0 | | zero | 1 | -1
|
|
|
|
| | | null | 1 | -1
|
|
|
|
| | 0 | zero | 1 | -1
|
2004-12-02 02:34:18 +01:00
|
|
|
| 1 | 4 | one | 2 | 2
|
|
|
|
| 2 | 3 | two | 2 | 2
|
|
|
|
| 3 | 2 | three | 2 | 2
|
|
|
|
| 4 | 1 | four | 2 | 2
|
|
|
|
| 5 | 0 | five | 2 | 2
|
|
|
|
| 6 | 6 | six | 2 | 2
|
|
|
|
| 7 | 7 | seven | 2 | 2
|
|
|
|
| 8 | 8 | eight | 2 | 2
|
|
|
|
| 0 | | zero | 2 | 2
|
|
|
|
| | | null | 2 | 2
|
2000-11-06 17:03:47 +01:00
|
|
|
| | 0 | zero | 2 | 2
|
2004-12-02 02:34:18 +01:00
|
|
|
| 1 | 4 | one | 3 | -3
|
|
|
|
| 2 | 3 | two | 3 | -3
|
|
|
|
| 3 | 2 | three | 3 | -3
|
|
|
|
| 4 | 1 | four | 3 | -3
|
|
|
|
| 5 | 0 | five | 3 | -3
|
|
|
|
| 6 | 6 | six | 3 | -3
|
|
|
|
| 7 | 7 | seven | 3 | -3
|
|
|
|
| 8 | 8 | eight | 3 | -3
|
|
|
|
| 0 | | zero | 3 | -3
|
|
|
|
| | | null | 3 | -3
|
|
|
|
| | 0 | zero | 3 | -3
|
|
|
|
| 1 | 4 | one | 2 | 4
|
|
|
|
| 2 | 3 | two | 2 | 4
|
|
|
|
| 3 | 2 | three | 2 | 4
|
|
|
|
| 4 | 1 | four | 2 | 4
|
|
|
|
| 5 | 0 | five | 2 | 4
|
|
|
|
| 6 | 6 | six | 2 | 4
|
|
|
|
| 7 | 7 | seven | 2 | 4
|
|
|
|
| 8 | 8 | eight | 2 | 4
|
|
|
|
| 0 | | zero | 2 | 4
|
|
|
|
| | | null | 2 | 4
|
|
|
|
| | 0 | zero | 2 | 4
|
|
|
|
| 1 | 4 | one | 5 | -5
|
|
|
|
| 2 | 3 | two | 5 | -5
|
|
|
|
| 3 | 2 | three | 5 | -5
|
|
|
|
| 4 | 1 | four | 5 | -5
|
|
|
|
| 5 | 0 | five | 5 | -5
|
|
|
|
| 6 | 6 | six | 5 | -5
|
|
|
|
| 7 | 7 | seven | 5 | -5
|
|
|
|
| 8 | 8 | eight | 5 | -5
|
|
|
|
| 0 | | zero | 5 | -5
|
|
|
|
| | | null | 5 | -5
|
|
|
|
| | 0 | zero | 5 | -5
|
|
|
|
| 1 | 4 | one | 5 | -5
|
|
|
|
| 2 | 3 | two | 5 | -5
|
|
|
|
| 3 | 2 | three | 5 | -5
|
|
|
|
| 4 | 1 | four | 5 | -5
|
|
|
|
| 5 | 0 | five | 5 | -5
|
|
|
|
| 6 | 6 | six | 5 | -5
|
|
|
|
| 7 | 7 | seven | 5 | -5
|
|
|
|
| 8 | 8 | eight | 5 | -5
|
|
|
|
| 0 | | zero | 5 | -5
|
|
|
|
| | | null | 5 | -5
|
|
|
|
| | 0 | zero | 5 | -5
|
|
|
|
| 1 | 4 | one | 0 |
|
|
|
|
| 2 | 3 | two | 0 |
|
|
|
|
| 3 | 2 | three | 0 |
|
|
|
|
| 4 | 1 | four | 0 |
|
|
|
|
| 5 | 0 | five | 0 |
|
|
|
|
| 6 | 6 | six | 0 |
|
|
|
|
| 7 | 7 | seven | 0 |
|
|
|
|
| 8 | 8 | eight | 0 |
|
|
|
|
| 0 | | zero | 0 |
|
|
|
|
| | | null | 0 |
|
|
|
|
| | 0 | zero | 0 |
|
|
|
|
| 1 | 4 | one | |
|
|
|
|
| 2 | 3 | two | |
|
|
|
|
| 3 | 2 | three | |
|
|
|
|
| 4 | 1 | four | |
|
|
|
|
| 5 | 0 | five | |
|
|
|
|
| 6 | 6 | six | |
|
|
|
|
| 7 | 7 | seven | |
|
|
|
|
| 8 | 8 | eight | |
|
|
|
|
| 0 | | zero | |
|
|
|
|
| | | null | |
|
|
|
|
| | 0 | zero | |
|
|
|
|
| 1 | 4 | one | | 0
|
|
|
|
| 2 | 3 | two | | 0
|
|
|
|
| 3 | 2 | three | | 0
|
|
|
|
| 4 | 1 | four | | 0
|
|
|
|
| 5 | 0 | five | | 0
|
|
|
|
| 6 | 6 | six | | 0
|
|
|
|
| 7 | 7 | seven | | 0
|
|
|
|
| 8 | 8 | eight | | 0
|
|
|
|
| 0 | | zero | | 0
|
|
|
|
| | | null | | 0
|
2000-11-06 17:03:47 +01:00
|
|
|
| | 0 | zero | | 0
|
|
|
|
(99 rows)
|
1999-02-23 08:27:13 +01:00
|
|
|
|
2000-02-15 04:31:33 +01:00
|
|
|
-- ambiguous column
|
2000-01-09 04:48:39 +01:00
|
|
|
SELECT '' AS "xxx", i, k, t
|
2000-02-15 04:31:33 +01:00
|
|
|
FROM J1_TBL CROSS JOIN J2_TBL;
|
2003-07-19 22:20:53 +02:00
|
|
|
ERROR: column reference "i" is ambiguous
|
2006-03-14 23:48:25 +01:00
|
|
|
LINE 1: SELECT '' AS "xxx", i, k, t
|
|
|
|
^
|
2000-02-15 04:31:33 +01:00
|
|
|
-- resolve previous ambiguity by specifying the table name
|
|
|
|
SELECT '' AS "xxx", t1.i, k, t
|
|
|
|
FROM J1_TBL t1 CROSS JOIN J2_TBL t2;
|
|
|
|
xxx | i | k | t
|
|
|
|
-----+---+----+-------
|
|
|
|
| 1 | -1 | one
|
2000-09-12 23:07:18 +02:00
|
|
|
| 2 | -1 | two
|
|
|
|
| 3 | -1 | three
|
|
|
|
| 4 | -1 | four
|
2000-11-06 17:03:47 +01:00
|
|
|
| 5 | -1 | five
|
|
|
|
| 6 | -1 | six
|
|
|
|
| 7 | -1 | seven
|
|
|
|
| 8 | -1 | eight
|
|
|
|
| 0 | -1 | zero
|
|
|
|
| | -1 | null
|
|
|
|
| | -1 | zero
|
2004-12-02 02:34:18 +01:00
|
|
|
| 1 | 2 | one
|
|
|
|
| 2 | 2 | two
|
|
|
|
| 3 | 2 | three
|
|
|
|
| 4 | 2 | four
|
|
|
|
| 5 | 2 | five
|
|
|
|
| 6 | 2 | six
|
|
|
|
| 7 | 2 | seven
|
|
|
|
| 8 | 2 | eight
|
|
|
|
| 0 | 2 | zero
|
|
|
|
| | 2 | null
|
2000-11-06 17:03:47 +01:00
|
|
|
| | 2 | zero
|
2004-12-02 02:34:18 +01:00
|
|
|
| 1 | -3 | one
|
|
|
|
| 2 | -3 | two
|
|
|
|
| 3 | -3 | three
|
|
|
|
| 4 | -3 | four
|
|
|
|
| 5 | -3 | five
|
|
|
|
| 6 | -3 | six
|
|
|
|
| 7 | -3 | seven
|
|
|
|
| 8 | -3 | eight
|
|
|
|
| 0 | -3 | zero
|
|
|
|
| | -3 | null
|
2000-11-06 17:03:47 +01:00
|
|
|
| | -3 | zero
|
2004-12-02 02:34:18 +01:00
|
|
|
| 1 | 4 | one
|
|
|
|
| 2 | 4 | two
|
|
|
|
| 3 | 4 | three
|
|
|
|
| 4 | 4 | four
|
|
|
|
| 5 | 4 | five
|
|
|
|
| 6 | 4 | six
|
|
|
|
| 7 | 4 | seven
|
|
|
|
| 8 | 4 | eight
|
|
|
|
| 0 | 4 | zero
|
|
|
|
| | 4 | null
|
2000-11-06 17:03:47 +01:00
|
|
|
| | 4 | zero
|
2004-12-02 02:34:18 +01:00
|
|
|
| 1 | -5 | one
|
|
|
|
| 2 | -5 | two
|
|
|
|
| 3 | -5 | three
|
|
|
|
| 4 | -5 | four
|
|
|
|
| 5 | -5 | five
|
|
|
|
| 6 | -5 | six
|
|
|
|
| 7 | -5 | seven
|
|
|
|
| 8 | -5 | eight
|
|
|
|
| 0 | -5 | zero
|
|
|
|
| | -5 | null
|
2000-11-06 17:03:47 +01:00
|
|
|
| | -5 | zero
|
2004-12-02 02:34:18 +01:00
|
|
|
| 1 | -5 | one
|
|
|
|
| 2 | -5 | two
|
|
|
|
| 3 | -5 | three
|
|
|
|
| 4 | -5 | four
|
|
|
|
| 5 | -5 | five
|
|
|
|
| 6 | -5 | six
|
|
|
|
| 7 | -5 | seven
|
|
|
|
| 8 | -5 | eight
|
|
|
|
| 0 | -5 | zero
|
|
|
|
| | -5 | null
|
2000-11-06 17:03:47 +01:00
|
|
|
| | -5 | zero
|
2004-12-02 02:34:18 +01:00
|
|
|
| 1 | | one
|
|
|
|
| 2 | | two
|
|
|
|
| 3 | | three
|
|
|
|
| 4 | | four
|
|
|
|
| 5 | | five
|
|
|
|
| 6 | | six
|
|
|
|
| 7 | | seven
|
|
|
|
| 8 | | eight
|
|
|
|
| 0 | | zero
|
|
|
|
| | | null
|
2000-11-06 17:03:47 +01:00
|
|
|
| | | zero
|
2004-12-02 02:34:18 +01:00
|
|
|
| 1 | | one
|
|
|
|
| 2 | | two
|
|
|
|
| 3 | | three
|
|
|
|
| 4 | | four
|
|
|
|
| 5 | | five
|
|
|
|
| 6 | | six
|
|
|
|
| 7 | | seven
|
|
|
|
| 8 | | eight
|
|
|
|
| 0 | | zero
|
|
|
|
| | | null
|
2000-11-06 17:03:47 +01:00
|
|
|
| | | zero
|
2004-12-02 02:34:18 +01:00
|
|
|
| 1 | 0 | one
|
|
|
|
| 2 | 0 | two
|
|
|
|
| 3 | 0 | three
|
|
|
|
| 4 | 0 | four
|
|
|
|
| 5 | 0 | five
|
|
|
|
| 6 | 0 | six
|
|
|
|
| 7 | 0 | seven
|
|
|
|
| 8 | 0 | eight
|
|
|
|
| 0 | 0 | zero
|
|
|
|
| | 0 | null
|
2000-11-06 17:03:47 +01:00
|
|
|
| | 0 | zero
|
|
|
|
(99 rows)
|
2000-02-15 04:31:33 +01:00
|
|
|
|
2000-01-09 04:48:39 +01:00
|
|
|
SELECT '' AS "xxx", ii, tt, kk
|
2000-02-15 04:31:33 +01:00
|
|
|
FROM (J1_TBL CROSS JOIN J2_TBL)
|
|
|
|
AS tx (ii, jj, tt, ii2, kk);
|
2000-09-12 23:07:18 +02:00
|
|
|
xxx | ii | tt | kk
|
|
|
|
-----+----+-------+----
|
|
|
|
| 1 | one | -1
|
|
|
|
| 2 | two | -1
|
|
|
|
| 3 | three | -1
|
|
|
|
| 4 | four | -1
|
2000-11-06 17:03:47 +01:00
|
|
|
| 5 | five | -1
|
|
|
|
| 6 | six | -1
|
|
|
|
| 7 | seven | -1
|
|
|
|
| 8 | eight | -1
|
|
|
|
| 0 | zero | -1
|
|
|
|
| | null | -1
|
|
|
|
| | zero | -1
|
2004-12-02 02:34:18 +01:00
|
|
|
| 1 | one | 2
|
|
|
|
| 2 | two | 2
|
|
|
|
| 3 | three | 2
|
|
|
|
| 4 | four | 2
|
|
|
|
| 5 | five | 2
|
|
|
|
| 6 | six | 2
|
|
|
|
| 7 | seven | 2
|
|
|
|
| 8 | eight | 2
|
|
|
|
| 0 | zero | 2
|
|
|
|
| | null | 2
|
2000-11-06 17:03:47 +01:00
|
|
|
| | zero | 2
|
2004-12-02 02:34:18 +01:00
|
|
|
| 1 | one | -3
|
|
|
|
| 2 | two | -3
|
|
|
|
| 3 | three | -3
|
|
|
|
| 4 | four | -3
|
|
|
|
| 5 | five | -3
|
|
|
|
| 6 | six | -3
|
|
|
|
| 7 | seven | -3
|
|
|
|
| 8 | eight | -3
|
|
|
|
| 0 | zero | -3
|
|
|
|
| | null | -3
|
2000-11-06 17:03:47 +01:00
|
|
|
| | zero | -3
|
2004-12-02 02:34:18 +01:00
|
|
|
| 1 | one | 4
|
|
|
|
| 2 | two | 4
|
|
|
|
| 3 | three | 4
|
|
|
|
| 4 | four | 4
|
|
|
|
| 5 | five | 4
|
|
|
|
| 6 | six | 4
|
|
|
|
| 7 | seven | 4
|
|
|
|
| 8 | eight | 4
|
|
|
|
| 0 | zero | 4
|
|
|
|
| | null | 4
|
2000-11-06 17:03:47 +01:00
|
|
|
| | zero | 4
|
2004-12-02 02:34:18 +01:00
|
|
|
| 1 | one | -5
|
|
|
|
| 2 | two | -5
|
|
|
|
| 3 | three | -5
|
|
|
|
| 4 | four | -5
|
|
|
|
| 5 | five | -5
|
|
|
|
| 6 | six | -5
|
|
|
|
| 7 | seven | -5
|
|
|
|
| 8 | eight | -5
|
|
|
|
| 0 | zero | -5
|
|
|
|
| | null | -5
|
2000-11-06 17:03:47 +01:00
|
|
|
| | zero | -5
|
2004-12-02 02:34:18 +01:00
|
|
|
| 1 | one | -5
|
|
|
|
| 2 | two | -5
|
|
|
|
| 3 | three | -5
|
|
|
|
| 4 | four | -5
|
|
|
|
| 5 | five | -5
|
|
|
|
| 6 | six | -5
|
|
|
|
| 7 | seven | -5
|
|
|
|
| 8 | eight | -5
|
|
|
|
| 0 | zero | -5
|
|
|
|
| | null | -5
|
2000-11-06 17:03:47 +01:00
|
|
|
| | zero | -5
|
2004-12-02 02:34:18 +01:00
|
|
|
| 1 | one |
|
|
|
|
| 2 | two |
|
|
|
|
| 3 | three |
|
|
|
|
| 4 | four |
|
|
|
|
| 5 | five |
|
|
|
|
| 6 | six |
|
|
|
|
| 7 | seven |
|
|
|
|
| 8 | eight |
|
|
|
|
| 0 | zero |
|
|
|
|
| | null |
|
2000-11-06 17:03:47 +01:00
|
|
|
| | zero |
|
2004-12-02 02:34:18 +01:00
|
|
|
| 1 | one |
|
|
|
|
| 2 | two |
|
|
|
|
| 3 | three |
|
|
|
|
| 4 | four |
|
|
|
|
| 5 | five |
|
|
|
|
| 6 | six |
|
|
|
|
| 7 | seven |
|
|
|
|
| 8 | eight |
|
|
|
|
| 0 | zero |
|
|
|
|
| | null |
|
2000-11-06 17:03:47 +01:00
|
|
|
| | zero |
|
2004-12-02 02:34:18 +01:00
|
|
|
| 1 | one | 0
|
|
|
|
| 2 | two | 0
|
|
|
|
| 3 | three | 0
|
|
|
|
| 4 | four | 0
|
|
|
|
| 5 | five | 0
|
|
|
|
| 6 | six | 0
|
|
|
|
| 7 | seven | 0
|
|
|
|
| 8 | eight | 0
|
|
|
|
| 0 | zero | 0
|
|
|
|
| | null | 0
|
2000-11-06 17:03:47 +01:00
|
|
|
| | zero | 0
|
|
|
|
(99 rows)
|
2000-09-12 23:07:18 +02:00
|
|
|
|
2000-02-15 04:31:33 +01:00
|
|
|
SELECT '' AS "xxx", tx.ii, tx.jj, tx.kk
|
|
|
|
FROM (J1_TBL t1 (a, b, c) CROSS JOIN J2_TBL t2 (d, e))
|
|
|
|
AS tx (ii, jj, tt, ii2, kk);
|
2000-09-12 23:07:18 +02:00
|
|
|
xxx | ii | jj | kk
|
|
|
|
-----+----+----+----
|
2000-11-06 17:03:47 +01:00
|
|
|
| 1 | 4 | -1
|
|
|
|
| 2 | 3 | -1
|
|
|
|
| 3 | 2 | -1
|
|
|
|
| 4 | 1 | -1
|
|
|
|
| 5 | 0 | -1
|
|
|
|
| 6 | 6 | -1
|
|
|
|
| 7 | 7 | -1
|
|
|
|
| 8 | 8 | -1
|
|
|
|
| 0 | | -1
|
|
|
|
| | | -1
|
|
|
|
| | 0 | -1
|
2004-12-02 02:34:18 +01:00
|
|
|
| 1 | 4 | 2
|
|
|
|
| 2 | 3 | 2
|
|
|
|
| 3 | 2 | 2
|
|
|
|
| 4 | 1 | 2
|
|
|
|
| 5 | 0 | 2
|
|
|
|
| 6 | 6 | 2
|
|
|
|
| 7 | 7 | 2
|
|
|
|
| 8 | 8 | 2
|
|
|
|
| 0 | | 2
|
|
|
|
| | | 2
|
2000-11-06 17:03:47 +01:00
|
|
|
| | 0 | 2
|
2004-12-02 02:34:18 +01:00
|
|
|
| 1 | 4 | -3
|
|
|
|
| 2 | 3 | -3
|
|
|
|
| 3 | 2 | -3
|
|
|
|
| 4 | 1 | -3
|
|
|
|
| 5 | 0 | -3
|
|
|
|
| 6 | 6 | -3
|
|
|
|
| 7 | 7 | -3
|
|
|
|
| 8 | 8 | -3
|
|
|
|
| 0 | | -3
|
|
|
|
| | | -3
|
2000-11-06 17:03:47 +01:00
|
|
|
| | 0 | -3
|
2004-12-02 02:34:18 +01:00
|
|
|
| 1 | 4 | 4
|
|
|
|
| 2 | 3 | 4
|
|
|
|
| 3 | 2 | 4
|
|
|
|
| 4 | 1 | 4
|
|
|
|
| 5 | 0 | 4
|
|
|
|
| 6 | 6 | 4
|
|
|
|
| 7 | 7 | 4
|
|
|
|
| 8 | 8 | 4
|
|
|
|
| 0 | | 4
|
|
|
|
| | | 4
|
2000-11-06 17:03:47 +01:00
|
|
|
| | 0 | 4
|
2004-12-02 02:34:18 +01:00
|
|
|
| 1 | 4 | -5
|
|
|
|
| 2 | 3 | -5
|
|
|
|
| 3 | 2 | -5
|
|
|
|
| 4 | 1 | -5
|
|
|
|
| 5 | 0 | -5
|
|
|
|
| 6 | 6 | -5
|
|
|
|
| 7 | 7 | -5
|
|
|
|
| 8 | 8 | -5
|
|
|
|
| 0 | | -5
|
|
|
|
| | | -5
|
|
|
|
| | 0 | -5
|
|
|
|
| 1 | 4 | -5
|
|
|
|
| 2 | 3 | -5
|
|
|
|
| 3 | 2 | -5
|
|
|
|
| 4 | 1 | -5
|
|
|
|
| 5 | 0 | -5
|
|
|
|
| 6 | 6 | -5
|
|
|
|
| 7 | 7 | -5
|
|
|
|
| 8 | 8 | -5
|
|
|
|
| 0 | | -5
|
|
|
|
| | | -5
|
|
|
|
| | 0 | -5
|
|
|
|
| 1 | 4 |
|
|
|
|
| 2 | 3 |
|
|
|
|
| 3 | 2 |
|
|
|
|
| 4 | 1 |
|
|
|
|
| 5 | 0 |
|
|
|
|
| 6 | 6 |
|
|
|
|
| 7 | 7 |
|
|
|
|
| 8 | 8 |
|
|
|
|
| 0 | |
|
|
|
|
| | |
|
|
|
|
| | 0 |
|
|
|
|
| 1 | 4 |
|
|
|
|
| 2 | 3 |
|
|
|
|
| 3 | 2 |
|
|
|
|
| 4 | 1 |
|
|
|
|
| 5 | 0 |
|
|
|
|
| 6 | 6 |
|
|
|
|
| 7 | 7 |
|
|
|
|
| 8 | 8 |
|
|
|
|
| 0 | |
|
|
|
|
| | |
|
|
|
|
| | 0 |
|
|
|
|
| 1 | 4 | 0
|
|
|
|
| 2 | 3 | 0
|
|
|
|
| 3 | 2 | 0
|
|
|
|
| 4 | 1 | 0
|
|
|
|
| 5 | 0 | 0
|
|
|
|
| 6 | 6 | 0
|
|
|
|
| 7 | 7 | 0
|
|
|
|
| 8 | 8 | 0
|
|
|
|
| 0 | | 0
|
|
|
|
| | | 0
|
|
|
|
| | 0 | 0
|
|
|
|
(99 rows)
|
2000-09-12 23:07:18 +02:00
|
|
|
|
|
|
|
SELECT '' AS "xxx", *
|
|
|
|
FROM J1_TBL CROSS JOIN J2_TBL a CROSS JOIN J2_TBL b;
|
|
|
|
xxx | i | j | t | i | k | i | k
|
|
|
|
-----+---+---+-------+---+----+---+----
|
2000-11-06 17:03:47 +01:00
|
|
|
| 1 | 4 | one | 1 | -1 | 1 | -1
|
2009-11-28 01:46:19 +01:00
|
|
|
| 1 | 4 | one | 1 | -1 | 2 | 2
|
|
|
|
| 1 | 4 | one | 1 | -1 | 3 | -3
|
|
|
|
| 1 | 4 | one | 1 | -1 | 2 | 4
|
|
|
|
| 1 | 4 | one | 1 | -1 | 5 | -5
|
|
|
|
| 1 | 4 | one | 1 | -1 | 5 | -5
|
|
|
|
| 1 | 4 | one | 1 | -1 | 0 |
|
|
|
|
| 1 | 4 | one | 1 | -1 | |
|
|
|
|
| 1 | 4 | one | 1 | -1 | | 0
|
2009-09-13 00:12:09 +02:00
|
|
|
| 2 | 3 | two | 1 | -1 | 1 | -1
|
2009-11-28 01:46:19 +01:00
|
|
|
| 2 | 3 | two | 1 | -1 | 2 | 2
|
|
|
|
| 2 | 3 | two | 1 | -1 | 3 | -3
|
|
|
|
| 2 | 3 | two | 1 | -1 | 2 | 4
|
|
|
|
| 2 | 3 | two | 1 | -1 | 5 | -5
|
|
|
|
| 2 | 3 | two | 1 | -1 | 5 | -5
|
|
|
|
| 2 | 3 | two | 1 | -1 | 0 |
|
|
|
|
| 2 | 3 | two | 1 | -1 | |
|
|
|
|
| 2 | 3 | two | 1 | -1 | | 0
|
2009-09-13 00:12:09 +02:00
|
|
|
| 3 | 2 | three | 1 | -1 | 1 | -1
|
2009-11-28 01:46:19 +01:00
|
|
|
| 3 | 2 | three | 1 | -1 | 2 | 2
|
|
|
|
| 3 | 2 | three | 1 | -1 | 3 | -3
|
|
|
|
| 3 | 2 | three | 1 | -1 | 2 | 4
|
|
|
|
| 3 | 2 | three | 1 | -1 | 5 | -5
|
|
|
|
| 3 | 2 | three | 1 | -1 | 5 | -5
|
|
|
|
| 3 | 2 | three | 1 | -1 | 0 |
|
|
|
|
| 3 | 2 | three | 1 | -1 | |
|
|
|
|
| 3 | 2 | three | 1 | -1 | | 0
|
2009-09-13 00:12:09 +02:00
|
|
|
| 4 | 1 | four | 1 | -1 | 1 | -1
|
2009-11-28 01:46:19 +01:00
|
|
|
| 4 | 1 | four | 1 | -1 | 2 | 2
|
|
|
|
| 4 | 1 | four | 1 | -1 | 3 | -3
|
|
|
|
| 4 | 1 | four | 1 | -1 | 2 | 4
|
|
|
|
| 4 | 1 | four | 1 | -1 | 5 | -5
|
|
|
|
| 4 | 1 | four | 1 | -1 | 5 | -5
|
|
|
|
| 4 | 1 | four | 1 | -1 | 0 |
|
|
|
|
| 4 | 1 | four | 1 | -1 | |
|
|
|
|
| 4 | 1 | four | 1 | -1 | | 0
|
2009-09-13 00:12:09 +02:00
|
|
|
| 5 | 0 | five | 1 | -1 | 1 | -1
|
2009-11-28 01:46:19 +01:00
|
|
|
| 5 | 0 | five | 1 | -1 | 2 | 2
|
|
|
|
| 5 | 0 | five | 1 | -1 | 3 | -3
|
|
|
|
| 5 | 0 | five | 1 | -1 | 2 | 4
|
|
|
|
| 5 | 0 | five | 1 | -1 | 5 | -5
|
|
|
|
| 5 | 0 | five | 1 | -1 | 5 | -5
|
|
|
|
| 5 | 0 | five | 1 | -1 | 0 |
|
|
|
|
| 5 | 0 | five | 1 | -1 | |
|
|
|
|
| 5 | 0 | five | 1 | -1 | | 0
|
2009-09-13 00:12:09 +02:00
|
|
|
| 6 | 6 | six | 1 | -1 | 1 | -1
|
2009-11-28 01:46:19 +01:00
|
|
|
| 6 | 6 | six | 1 | -1 | 2 | 2
|
|
|
|
| 6 | 6 | six | 1 | -1 | 3 | -3
|
|
|
|
| 6 | 6 | six | 1 | -1 | 2 | 4
|
|
|
|
| 6 | 6 | six | 1 | -1 | 5 | -5
|
|
|
|
| 6 | 6 | six | 1 | -1 | 5 | -5
|
|
|
|
| 6 | 6 | six | 1 | -1 | 0 |
|
|
|
|
| 6 | 6 | six | 1 | -1 | |
|
|
|
|
| 6 | 6 | six | 1 | -1 | | 0
|
2009-09-13 00:12:09 +02:00
|
|
|
| 7 | 7 | seven | 1 | -1 | 1 | -1
|
2009-11-28 01:46:19 +01:00
|
|
|
| 7 | 7 | seven | 1 | -1 | 2 | 2
|
|
|
|
| 7 | 7 | seven | 1 | -1 | 3 | -3
|
|
|
|
| 7 | 7 | seven | 1 | -1 | 2 | 4
|
|
|
|
| 7 | 7 | seven | 1 | -1 | 5 | -5
|
|
|
|
| 7 | 7 | seven | 1 | -1 | 5 | -5
|
|
|
|
| 7 | 7 | seven | 1 | -1 | 0 |
|
|
|
|
| 7 | 7 | seven | 1 | -1 | |
|
|
|
|
| 7 | 7 | seven | 1 | -1 | | 0
|
2009-09-13 00:12:09 +02:00
|
|
|
| 8 | 8 | eight | 1 | -1 | 1 | -1
|
2009-11-28 01:46:19 +01:00
|
|
|
| 8 | 8 | eight | 1 | -1 | 2 | 2
|
|
|
|
| 8 | 8 | eight | 1 | -1 | 3 | -3
|
|
|
|
| 8 | 8 | eight | 1 | -1 | 2 | 4
|
|
|
|
| 8 | 8 | eight | 1 | -1 | 5 | -5
|
|
|
|
| 8 | 8 | eight | 1 | -1 | 5 | -5
|
|
|
|
| 8 | 8 | eight | 1 | -1 | 0 |
|
|
|
|
| 8 | 8 | eight | 1 | -1 | |
|
|
|
|
| 8 | 8 | eight | 1 | -1 | | 0
|
2009-09-13 00:12:09 +02:00
|
|
|
| 0 | | zero | 1 | -1 | 1 | -1
|
2009-11-28 01:46:19 +01:00
|
|
|
| 0 | | zero | 1 | -1 | 2 | 2
|
|
|
|
| 0 | | zero | 1 | -1 | 3 | -3
|
|
|
|
| 0 | | zero | 1 | -1 | 2 | 4
|
|
|
|
| 0 | | zero | 1 | -1 | 5 | -5
|
|
|
|
| 0 | | zero | 1 | -1 | 5 | -5
|
|
|
|
| 0 | | zero | 1 | -1 | 0 |
|
|
|
|
| 0 | | zero | 1 | -1 | |
|
|
|
|
| 0 | | zero | 1 | -1 | | 0
|
2009-09-13 00:12:09 +02:00
|
|
|
| | | null | 1 | -1 | 1 | -1
|
2009-11-28 01:46:19 +01:00
|
|
|
| | | null | 1 | -1 | 2 | 2
|
|
|
|
| | | null | 1 | -1 | 3 | -3
|
|
|
|
| | | null | 1 | -1 | 2 | 4
|
|
|
|
| | | null | 1 | -1 | 5 | -5
|
|
|
|
| | | null | 1 | -1 | 5 | -5
|
|
|
|
| | | null | 1 | -1 | 0 |
|
|
|
|
| | | null | 1 | -1 | |
|
|
|
|
| | | null | 1 | -1 | | 0
|
2009-09-13 00:12:09 +02:00
|
|
|
| | 0 | zero | 1 | -1 | 1 | -1
|
2009-11-28 01:46:19 +01:00
|
|
|
| | 0 | zero | 1 | -1 | 2 | 2
|
|
|
|
| | 0 | zero | 1 | -1 | 3 | -3
|
|
|
|
| | 0 | zero | 1 | -1 | 2 | 4
|
|
|
|
| | 0 | zero | 1 | -1 | 5 | -5
|
|
|
|
| | 0 | zero | 1 | -1 | 5 | -5
|
|
|
|
| | 0 | zero | 1 | -1 | 0 |
|
|
|
|
| | 0 | zero | 1 | -1 | |
|
|
|
|
| | 0 | zero | 1 | -1 | | 0
|
|
|
|
| 1 | 4 | one | 2 | 2 | 1 | -1
|
2009-09-13 00:12:09 +02:00
|
|
|
| 1 | 4 | one | 2 | 2 | 2 | 2
|
2009-11-28 01:46:19 +01:00
|
|
|
| 1 | 4 | one | 2 | 2 | 3 | -3
|
|
|
|
| 1 | 4 | one | 2 | 2 | 2 | 4
|
|
|
|
| 1 | 4 | one | 2 | 2 | 5 | -5
|
|
|
|
| 1 | 4 | one | 2 | 2 | 5 | -5
|
|
|
|
| 1 | 4 | one | 2 | 2 | 0 |
|
|
|
|
| 1 | 4 | one | 2 | 2 | |
|
|
|
|
| 1 | 4 | one | 2 | 2 | | 0
|
|
|
|
| 2 | 3 | two | 2 | 2 | 1 | -1
|
2009-09-13 00:12:09 +02:00
|
|
|
| 2 | 3 | two | 2 | 2 | 2 | 2
|
2009-11-28 01:46:19 +01:00
|
|
|
| 2 | 3 | two | 2 | 2 | 3 | -3
|
|
|
|
| 2 | 3 | two | 2 | 2 | 2 | 4
|
|
|
|
| 2 | 3 | two | 2 | 2 | 5 | -5
|
|
|
|
| 2 | 3 | two | 2 | 2 | 5 | -5
|
|
|
|
| 2 | 3 | two | 2 | 2 | 0 |
|
|
|
|
| 2 | 3 | two | 2 | 2 | |
|
|
|
|
| 2 | 3 | two | 2 | 2 | | 0
|
|
|
|
| 3 | 2 | three | 2 | 2 | 1 | -1
|
2009-09-13 00:12:09 +02:00
|
|
|
| 3 | 2 | three | 2 | 2 | 2 | 2
|
2009-11-28 01:46:19 +01:00
|
|
|
| 3 | 2 | three | 2 | 2 | 3 | -3
|
|
|
|
| 3 | 2 | three | 2 | 2 | 2 | 4
|
|
|
|
| 3 | 2 | three | 2 | 2 | 5 | -5
|
|
|
|
| 3 | 2 | three | 2 | 2 | 5 | -5
|
|
|
|
| 3 | 2 | three | 2 | 2 | 0 |
|
|
|
|
| 3 | 2 | three | 2 | 2 | |
|
|
|
|
| 3 | 2 | three | 2 | 2 | | 0
|
|
|
|
| 4 | 1 | four | 2 | 2 | 1 | -1
|
2009-09-13 00:12:09 +02:00
|
|
|
| 4 | 1 | four | 2 | 2 | 2 | 2
|
2009-11-28 01:46:19 +01:00
|
|
|
| 4 | 1 | four | 2 | 2 | 3 | -3
|
|
|
|
| 4 | 1 | four | 2 | 2 | 2 | 4
|
|
|
|
| 4 | 1 | four | 2 | 2 | 5 | -5
|
|
|
|
| 4 | 1 | four | 2 | 2 | 5 | -5
|
|
|
|
| 4 | 1 | four | 2 | 2 | 0 |
|
|
|
|
| 4 | 1 | four | 2 | 2 | |
|
|
|
|
| 4 | 1 | four | 2 | 2 | | 0
|
|
|
|
| 5 | 0 | five | 2 | 2 | 1 | -1
|
2009-09-13 00:12:09 +02:00
|
|
|
| 5 | 0 | five | 2 | 2 | 2 | 2
|
2009-11-28 01:46:19 +01:00
|
|
|
| 5 | 0 | five | 2 | 2 | 3 | -3
|
|
|
|
| 5 | 0 | five | 2 | 2 | 2 | 4
|
|
|
|
| 5 | 0 | five | 2 | 2 | 5 | -5
|
|
|
|
| 5 | 0 | five | 2 | 2 | 5 | -5
|
|
|
|
| 5 | 0 | five | 2 | 2 | 0 |
|
|
|
|
| 5 | 0 | five | 2 | 2 | |
|
|
|
|
| 5 | 0 | five | 2 | 2 | | 0
|
|
|
|
| 6 | 6 | six | 2 | 2 | 1 | -1
|
2009-09-13 00:12:09 +02:00
|
|
|
| 6 | 6 | six | 2 | 2 | 2 | 2
|
2009-11-28 01:46:19 +01:00
|
|
|
| 6 | 6 | six | 2 | 2 | 3 | -3
|
|
|
|
| 6 | 6 | six | 2 | 2 | 2 | 4
|
|
|
|
| 6 | 6 | six | 2 | 2 | 5 | -5
|
|
|
|
| 6 | 6 | six | 2 | 2 | 5 | -5
|
|
|
|
| 6 | 6 | six | 2 | 2 | 0 |
|
|
|
|
| 6 | 6 | six | 2 | 2 | |
|
|
|
|
| 6 | 6 | six | 2 | 2 | | 0
|
|
|
|
| 7 | 7 | seven | 2 | 2 | 1 | -1
|
2009-09-13 00:12:09 +02:00
|
|
|
| 7 | 7 | seven | 2 | 2 | 2 | 2
|
2009-11-28 01:46:19 +01:00
|
|
|
| 7 | 7 | seven | 2 | 2 | 3 | -3
|
|
|
|
| 7 | 7 | seven | 2 | 2 | 2 | 4
|
|
|
|
| 7 | 7 | seven | 2 | 2 | 5 | -5
|
|
|
|
| 7 | 7 | seven | 2 | 2 | 5 | -5
|
|
|
|
| 7 | 7 | seven | 2 | 2 | 0 |
|
|
|
|
| 7 | 7 | seven | 2 | 2 | |
|
|
|
|
| 7 | 7 | seven | 2 | 2 | | 0
|
|
|
|
| 8 | 8 | eight | 2 | 2 | 1 | -1
|
2009-09-13 00:12:09 +02:00
|
|
|
| 8 | 8 | eight | 2 | 2 | 2 | 2
|
2009-11-28 01:46:19 +01:00
|
|
|
| 8 | 8 | eight | 2 | 2 | 3 | -3
|
|
|
|
| 8 | 8 | eight | 2 | 2 | 2 | 4
|
|
|
|
| 8 | 8 | eight | 2 | 2 | 5 | -5
|
|
|
|
| 8 | 8 | eight | 2 | 2 | 5 | -5
|
|
|
|
| 8 | 8 | eight | 2 | 2 | 0 |
|
|
|
|
| 8 | 8 | eight | 2 | 2 | |
|
|
|
|
| 8 | 8 | eight | 2 | 2 | | 0
|
|
|
|
| 0 | | zero | 2 | 2 | 1 | -1
|
2009-09-13 00:12:09 +02:00
|
|
|
| 0 | | zero | 2 | 2 | 2 | 2
|
2009-11-28 01:46:19 +01:00
|
|
|
| 0 | | zero | 2 | 2 | 3 | -3
|
|
|
|
| 0 | | zero | 2 | 2 | 2 | 4
|
|
|
|
| 0 | | zero | 2 | 2 | 5 | -5
|
|
|
|
| 0 | | zero | 2 | 2 | 5 | -5
|
|
|
|
| 0 | | zero | 2 | 2 | 0 |
|
|
|
|
| 0 | | zero | 2 | 2 | |
|
|
|
|
| 0 | | zero | 2 | 2 | | 0
|
|
|
|
| | | null | 2 | 2 | 1 | -1
|
2009-09-13 00:12:09 +02:00
|
|
|
| | | null | 2 | 2 | 2 | 2
|
2009-11-28 01:46:19 +01:00
|
|
|
| | | null | 2 | 2 | 3 | -3
|
|
|
|
| | | null | 2 | 2 | 2 | 4
|
|
|
|
| | | null | 2 | 2 | 5 | -5
|
|
|
|
| | | null | 2 | 2 | 5 | -5
|
|
|
|
| | | null | 2 | 2 | 0 |
|
|
|
|
| | | null | 2 | 2 | |
|
|
|
|
| | | null | 2 | 2 | | 0
|
|
|
|
| | 0 | zero | 2 | 2 | 1 | -1
|
2009-09-13 00:12:09 +02:00
|
|
|
| | 0 | zero | 2 | 2 | 2 | 2
|
2009-11-28 01:46:19 +01:00
|
|
|
| | 0 | zero | 2 | 2 | 3 | -3
|
|
|
|
| | 0 | zero | 2 | 2 | 2 | 4
|
|
|
|
| | 0 | zero | 2 | 2 | 5 | -5
|
|
|
|
| | 0 | zero | 2 | 2 | 5 | -5
|
|
|
|
| | 0 | zero | 2 | 2 | 0 |
|
|
|
|
| | 0 | zero | 2 | 2 | |
|
|
|
|
| | 0 | zero | 2 | 2 | | 0
|
|
|
|
| 1 | 4 | one | 3 | -3 | 1 | -1
|
|
|
|
| 1 | 4 | one | 3 | -3 | 2 | 2
|
2009-09-13 00:12:09 +02:00
|
|
|
| 1 | 4 | one | 3 | -3 | 3 | -3
|
2009-11-28 01:46:19 +01:00
|
|
|
| 1 | 4 | one | 3 | -3 | 2 | 4
|
|
|
|
| 1 | 4 | one | 3 | -3 | 5 | -5
|
|
|
|
| 1 | 4 | one | 3 | -3 | 5 | -5
|
|
|
|
| 1 | 4 | one | 3 | -3 | 0 |
|
|
|
|
| 1 | 4 | one | 3 | -3 | |
|
|
|
|
| 1 | 4 | one | 3 | -3 | | 0
|
|
|
|
| 2 | 3 | two | 3 | -3 | 1 | -1
|
|
|
|
| 2 | 3 | two | 3 | -3 | 2 | 2
|
2009-09-13 00:12:09 +02:00
|
|
|
| 2 | 3 | two | 3 | -3 | 3 | -3
|
2009-11-28 01:46:19 +01:00
|
|
|
| 2 | 3 | two | 3 | -3 | 2 | 4
|
|
|
|
| 2 | 3 | two | 3 | -3 | 5 | -5
|
|
|
|
| 2 | 3 | two | 3 | -3 | 5 | -5
|
|
|
|
| 2 | 3 | two | 3 | -3 | 0 |
|
|
|
|
| 2 | 3 | two | 3 | -3 | |
|
|
|
|
| 2 | 3 | two | 3 | -3 | | 0
|
|
|
|
| 3 | 2 | three | 3 | -3 | 1 | -1
|
|
|
|
| 3 | 2 | three | 3 | -3 | 2 | 2
|
2009-09-13 00:12:09 +02:00
|
|
|
| 3 | 2 | three | 3 | -3 | 3 | -3
|
2009-11-28 01:46:19 +01:00
|
|
|
| 3 | 2 | three | 3 | -3 | 2 | 4
|
|
|
|
| 3 | 2 | three | 3 | -3 | 5 | -5
|
|
|
|
| 3 | 2 | three | 3 | -3 | 5 | -5
|
|
|
|
| 3 | 2 | three | 3 | -3 | 0 |
|
|
|
|
| 3 | 2 | three | 3 | -3 | |
|
|
|
|
| 3 | 2 | three | 3 | -3 | | 0
|
|
|
|
| 4 | 1 | four | 3 | -3 | 1 | -1
|
|
|
|
| 4 | 1 | four | 3 | -3 | 2 | 2
|
2009-09-13 00:12:09 +02:00
|
|
|
| 4 | 1 | four | 3 | -3 | 3 | -3
|
2009-11-28 01:46:19 +01:00
|
|
|
| 4 | 1 | four | 3 | -3 | 2 | 4
|
|
|
|
| 4 | 1 | four | 3 | -3 | 5 | -5
|
|
|
|
| 4 | 1 | four | 3 | -3 | 5 | -5
|
|
|
|
| 4 | 1 | four | 3 | -3 | 0 |
|
|
|
|
| 4 | 1 | four | 3 | -3 | |
|
|
|
|
| 4 | 1 | four | 3 | -3 | | 0
|
|
|
|
| 5 | 0 | five | 3 | -3 | 1 | -1
|
|
|
|
| 5 | 0 | five | 3 | -3 | 2 | 2
|
2009-09-13 00:12:09 +02:00
|
|
|
| 5 | 0 | five | 3 | -3 | 3 | -3
|
2009-11-28 01:46:19 +01:00
|
|
|
| 5 | 0 | five | 3 | -3 | 2 | 4
|
|
|
|
| 5 | 0 | five | 3 | -3 | 5 | -5
|
|
|
|
| 5 | 0 | five | 3 | -3 | 5 | -5
|
|
|
|
| 5 | 0 | five | 3 | -3 | 0 |
|
|
|
|
| 5 | 0 | five | 3 | -3 | |
|
|
|
|
| 5 | 0 | five | 3 | -3 | | 0
|
|
|
|
| 6 | 6 | six | 3 | -3 | 1 | -1
|
|
|
|
| 6 | 6 | six | 3 | -3 | 2 | 2
|
2009-09-13 00:12:09 +02:00
|
|
|
| 6 | 6 | six | 3 | -3 | 3 | -3
|
2009-11-28 01:46:19 +01:00
|
|
|
| 6 | 6 | six | 3 | -3 | 2 | 4
|
|
|
|
| 6 | 6 | six | 3 | -3 | 5 | -5
|
|
|
|
| 6 | 6 | six | 3 | -3 | 5 | -5
|
|
|
|
| 6 | 6 | six | 3 | -3 | 0 |
|
|
|
|
| 6 | 6 | six | 3 | -3 | |
|
|
|
|
| 6 | 6 | six | 3 | -3 | | 0
|
|
|
|
| 7 | 7 | seven | 3 | -3 | 1 | -1
|
|
|
|
| 7 | 7 | seven | 3 | -3 | 2 | 2
|
2009-09-13 00:12:09 +02:00
|
|
|
| 7 | 7 | seven | 3 | -3 | 3 | -3
|
2009-11-28 01:46:19 +01:00
|
|
|
| 7 | 7 | seven | 3 | -3 | 2 | 4
|
|
|
|
| 7 | 7 | seven | 3 | -3 | 5 | -5
|
|
|
|
| 7 | 7 | seven | 3 | -3 | 5 | -5
|
|
|
|
| 7 | 7 | seven | 3 | -3 | 0 |
|
|
|
|
| 7 | 7 | seven | 3 | -3 | |
|
|
|
|
| 7 | 7 | seven | 3 | -3 | | 0
|
|
|
|
| 8 | 8 | eight | 3 | -3 | 1 | -1
|
|
|
|
| 8 | 8 | eight | 3 | -3 | 2 | 2
|
2009-09-13 00:12:09 +02:00
|
|
|
| 8 | 8 | eight | 3 | -3 | 3 | -3
|
2009-11-28 01:46:19 +01:00
|
|
|
| 8 | 8 | eight | 3 | -3 | 2 | 4
|
|
|
|
| 8 | 8 | eight | 3 | -3 | 5 | -5
|
|
|
|
| 8 | 8 | eight | 3 | -3 | 5 | -5
|
|
|
|
| 8 | 8 | eight | 3 | -3 | 0 |
|
|
|
|
| 8 | 8 | eight | 3 | -3 | |
|
|
|
|
| 8 | 8 | eight | 3 | -3 | | 0
|
|
|
|
| 0 | | zero | 3 | -3 | 1 | -1
|
|
|
|
| 0 | | zero | 3 | -3 | 2 | 2
|
2009-09-13 00:12:09 +02:00
|
|
|
| 0 | | zero | 3 | -3 | 3 | -3
|
2009-11-28 01:46:19 +01:00
|
|
|
| 0 | | zero | 3 | -3 | 2 | 4
|
|
|
|
| 0 | | zero | 3 | -3 | 5 | -5
|
|
|
|
| 0 | | zero | 3 | -3 | 5 | -5
|
|
|
|
| 0 | | zero | 3 | -3 | 0 |
|
|
|
|
| 0 | | zero | 3 | -3 | |
|
|
|
|
| 0 | | zero | 3 | -3 | | 0
|
|
|
|
| | | null | 3 | -3 | 1 | -1
|
|
|
|
| | | null | 3 | -3 | 2 | 2
|
2009-09-13 00:12:09 +02:00
|
|
|
| | | null | 3 | -3 | 3 | -3
|
2009-11-28 01:46:19 +01:00
|
|
|
| | | null | 3 | -3 | 2 | 4
|
|
|
|
| | | null | 3 | -3 | 5 | -5
|
|
|
|
| | | null | 3 | -3 | 5 | -5
|
|
|
|
| | | null | 3 | -3 | 0 |
|
|
|
|
| | | null | 3 | -3 | |
|
|
|
|
| | | null | 3 | -3 | | 0
|
|
|
|
| | 0 | zero | 3 | -3 | 1 | -1
|
|
|
|
| | 0 | zero | 3 | -3 | 2 | 2
|
|
|
|
| | 0 | zero | 3 | -3 | 3 | -3
|
|
|
|
| | 0 | zero | 3 | -3 | 2 | 4
|
|
|
|
| | 0 | zero | 3 | -3 | 5 | -5
|
|
|
|
| | 0 | zero | 3 | -3 | 5 | -5
|
|
|
|
| | 0 | zero | 3 | -3 | 0 |
|
|
|
|
| | 0 | zero | 3 | -3 | |
|
|
|
|
| | 0 | zero | 3 | -3 | | 0
|
|
|
|
| 1 | 4 | one | 2 | 4 | 1 | -1
|
|
|
|
| 1 | 4 | one | 2 | 4 | 2 | 2
|
|
|
|
| 1 | 4 | one | 2 | 4 | 3 | -3
|
2009-09-13 00:12:09 +02:00
|
|
|
| 1 | 4 | one | 2 | 4 | 2 | 4
|
2009-11-28 01:46:19 +01:00
|
|
|
| 1 | 4 | one | 2 | 4 | 5 | -5
|
|
|
|
| 1 | 4 | one | 2 | 4 | 5 | -5
|
|
|
|
| 1 | 4 | one | 2 | 4 | 0 |
|
|
|
|
| 1 | 4 | one | 2 | 4 | |
|
|
|
|
| 1 | 4 | one | 2 | 4 | | 0
|
|
|
|
| 2 | 3 | two | 2 | 4 | 1 | -1
|
|
|
|
| 2 | 3 | two | 2 | 4 | 2 | 2
|
|
|
|
| 2 | 3 | two | 2 | 4 | 3 | -3
|
2009-09-13 00:12:09 +02:00
|
|
|
| 2 | 3 | two | 2 | 4 | 2 | 4
|
2009-11-28 01:46:19 +01:00
|
|
|
| 2 | 3 | two | 2 | 4 | 5 | -5
|
|
|
|
| 2 | 3 | two | 2 | 4 | 5 | -5
|
|
|
|
| 2 | 3 | two | 2 | 4 | 0 |
|
|
|
|
| 2 | 3 | two | 2 | 4 | |
|
|
|
|
| 2 | 3 | two | 2 | 4 | | 0
|
|
|
|
| 3 | 2 | three | 2 | 4 | 1 | -1
|
|
|
|
| 3 | 2 | three | 2 | 4 | 2 | 2
|
|
|
|
| 3 | 2 | three | 2 | 4 | 3 | -3
|
2009-09-13 00:12:09 +02:00
|
|
|
| 3 | 2 | three | 2 | 4 | 2 | 4
|
2009-11-28 01:46:19 +01:00
|
|
|
| 3 | 2 | three | 2 | 4 | 5 | -5
|
|
|
|
| 3 | 2 | three | 2 | 4 | 5 | -5
|
|
|
|
| 3 | 2 | three | 2 | 4 | 0 |
|
|
|
|
| 3 | 2 | three | 2 | 4 | |
|
|
|
|
| 3 | 2 | three | 2 | 4 | | 0
|
|
|
|
| 4 | 1 | four | 2 | 4 | 1 | -1
|
|
|
|
| 4 | 1 | four | 2 | 4 | 2 | 2
|
|
|
|
| 4 | 1 | four | 2 | 4 | 3 | -3
|
2009-09-13 00:12:09 +02:00
|
|
|
| 4 | 1 | four | 2 | 4 | 2 | 4
|
2009-11-28 01:46:19 +01:00
|
|
|
| 4 | 1 | four | 2 | 4 | 5 | -5
|
|
|
|
| 4 | 1 | four | 2 | 4 | 5 | -5
|
|
|
|
| 4 | 1 | four | 2 | 4 | 0 |
|
|
|
|
| 4 | 1 | four | 2 | 4 | |
|
|
|
|
| 4 | 1 | four | 2 | 4 | | 0
|
|
|
|
| 5 | 0 | five | 2 | 4 | 1 | -1
|
|
|
|
| 5 | 0 | five | 2 | 4 | 2 | 2
|
|
|
|
| 5 | 0 | five | 2 | 4 | 3 | -3
|
2009-09-13 00:12:09 +02:00
|
|
|
| 5 | 0 | five | 2 | 4 | 2 | 4
|
2009-11-28 01:46:19 +01:00
|
|
|
| 5 | 0 | five | 2 | 4 | 5 | -5
|
|
|
|
| 5 | 0 | five | 2 | 4 | 5 | -5
|
|
|
|
| 5 | 0 | five | 2 | 4 | 0 |
|
|
|
|
| 5 | 0 | five | 2 | 4 | |
|
|
|
|
| 5 | 0 | five | 2 | 4 | | 0
|
|
|
|
| 6 | 6 | six | 2 | 4 | 1 | -1
|
|
|
|
| 6 | 6 | six | 2 | 4 | 2 | 2
|
|
|
|
| 6 | 6 | six | 2 | 4 | 3 | -3
|
2009-09-13 00:12:09 +02:00
|
|
|
| 6 | 6 | six | 2 | 4 | 2 | 4
|
2009-11-28 01:46:19 +01:00
|
|
|
| 6 | 6 | six | 2 | 4 | 5 | -5
|
|
|
|
| 6 | 6 | six | 2 | 4 | 5 | -5
|
|
|
|
| 6 | 6 | six | 2 | 4 | 0 |
|
|
|
|
| 6 | 6 | six | 2 | 4 | |
|
|
|
|
| 6 | 6 | six | 2 | 4 | | 0
|
|
|
|
| 7 | 7 | seven | 2 | 4 | 1 | -1
|
|
|
|
| 7 | 7 | seven | 2 | 4 | 2 | 2
|
|
|
|
| 7 | 7 | seven | 2 | 4 | 3 | -3
|
2009-09-13 00:12:09 +02:00
|
|
|
| 7 | 7 | seven | 2 | 4 | 2 | 4
|
2009-11-28 01:46:19 +01:00
|
|
|
| 7 | 7 | seven | 2 | 4 | 5 | -5
|
|
|
|
| 7 | 7 | seven | 2 | 4 | 5 | -5
|
|
|
|
| 7 | 7 | seven | 2 | 4 | 0 |
|
|
|
|
| 7 | 7 | seven | 2 | 4 | |
|
|
|
|
| 7 | 7 | seven | 2 | 4 | | 0
|
|
|
|
| 8 | 8 | eight | 2 | 4 | 1 | -1
|
|
|
|
| 8 | 8 | eight | 2 | 4 | 2 | 2
|
|
|
|
| 8 | 8 | eight | 2 | 4 | 3 | -3
|
2009-09-13 00:12:09 +02:00
|
|
|
| 8 | 8 | eight | 2 | 4 | 2 | 4
|
2009-11-28 01:46:19 +01:00
|
|
|
| 8 | 8 | eight | 2 | 4 | 5 | -5
|
|
|
|
| 8 | 8 | eight | 2 | 4 | 5 | -5
|
|
|
|
| 8 | 8 | eight | 2 | 4 | 0 |
|
|
|
|
| 8 | 8 | eight | 2 | 4 | |
|
|
|
|
| 8 | 8 | eight | 2 | 4 | | 0
|
|
|
|
| 0 | | zero | 2 | 4 | 1 | -1
|
|
|
|
| 0 | | zero | 2 | 4 | 2 | 2
|
|
|
|
| 0 | | zero | 2 | 4 | 3 | -3
|
2009-09-13 00:12:09 +02:00
|
|
|
| 0 | | zero | 2 | 4 | 2 | 4
|
2009-11-28 01:46:19 +01:00
|
|
|
| 0 | | zero | 2 | 4 | 5 | -5
|
|
|
|
| 0 | | zero | 2 | 4 | 5 | -5
|
|
|
|
| 0 | | zero | 2 | 4 | 0 |
|
|
|
|
| 0 | | zero | 2 | 4 | |
|
|
|
|
| 0 | | zero | 2 | 4 | | 0
|
|
|
|
| | | null | 2 | 4 | 1 | -1
|
|
|
|
| | | null | 2 | 4 | 2 | 2
|
|
|
|
| | | null | 2 | 4 | 3 | -3
|
2009-09-13 00:12:09 +02:00
|
|
|
| | | null | 2 | 4 | 2 | 4
|
2009-11-28 01:46:19 +01:00
|
|
|
| | | null | 2 | 4 | 5 | -5
|
|
|
|
| | | null | 2 | 4 | 5 | -5
|
|
|
|
| | | null | 2 | 4 | 0 |
|
|
|
|
| | | null | 2 | 4 | |
|
|
|
|
| | | null | 2 | 4 | | 0
|
|
|
|
| | 0 | zero | 2 | 4 | 1 | -1
|
|
|
|
| | 0 | zero | 2 | 4 | 2 | 2
|
|
|
|
| | 0 | zero | 2 | 4 | 3 | -3
|
2009-09-13 00:12:09 +02:00
|
|
|
| | 0 | zero | 2 | 4 | 2 | 4
|
2009-11-28 01:46:19 +01:00
|
|
|
| | 0 | zero | 2 | 4 | 5 | -5
|
|
|
|
| | 0 | zero | 2 | 4 | 5 | -5
|
|
|
|
| | 0 | zero | 2 | 4 | 0 |
|
|
|
|
| | 0 | zero | 2 | 4 | |
|
|
|
|
| | 0 | zero | 2 | 4 | | 0
|
|
|
|
| 1 | 4 | one | 5 | -5 | 1 | -1
|
|
|
|
| 1 | 4 | one | 5 | -5 | 2 | 2
|
|
|
|
| 1 | 4 | one | 5 | -5 | 3 | -3
|
|
|
|
| 1 | 4 | one | 5 | -5 | 2 | 4
|
2009-09-13 00:12:09 +02:00
|
|
|
| 1 | 4 | one | 5 | -5 | 5 | -5
|
|
|
|
| 1 | 4 | one | 5 | -5 | 5 | -5
|
2009-11-28 01:46:19 +01:00
|
|
|
| 1 | 4 | one | 5 | -5 | 0 |
|
|
|
|
| 1 | 4 | one | 5 | -5 | |
|
|
|
|
| 1 | 4 | one | 5 | -5 | | 0
|
|
|
|
| 2 | 3 | two | 5 | -5 | 1 | -1
|
|
|
|
| 2 | 3 | two | 5 | -5 | 2 | 2
|
|
|
|
| 2 | 3 | two | 5 | -5 | 3 | -3
|
|
|
|
| 2 | 3 | two | 5 | -5 | 2 | 4
|
2009-09-13 00:12:09 +02:00
|
|
|
| 2 | 3 | two | 5 | -5 | 5 | -5
|
|
|
|
| 2 | 3 | two | 5 | -5 | 5 | -5
|
2009-11-28 01:46:19 +01:00
|
|
|
| 2 | 3 | two | 5 | -5 | 0 |
|
|
|
|
| 2 | 3 | two | 5 | -5 | |
|
|
|
|
| 2 | 3 | two | 5 | -5 | | 0
|
|
|
|
| 3 | 2 | three | 5 | -5 | 1 | -1
|
|
|
|
| 3 | 2 | three | 5 | -5 | 2 | 2
|
|
|
|
| 3 | 2 | three | 5 | -5 | 3 | -3
|
|
|
|
| 3 | 2 | three | 5 | -5 | 2 | 4
|
2009-09-13 00:12:09 +02:00
|
|
|
| 3 | 2 | three | 5 | -5 | 5 | -5
|
|
|
|
| 3 | 2 | three | 5 | -5 | 5 | -5
|
2009-11-28 01:46:19 +01:00
|
|
|
| 3 | 2 | three | 5 | -5 | 0 |
|
|
|
|
| 3 | 2 | three | 5 | -5 | |
|
|
|
|
| 3 | 2 | three | 5 | -5 | | 0
|
|
|
|
| 4 | 1 | four | 5 | -5 | 1 | -1
|
|
|
|
| 4 | 1 | four | 5 | -5 | 2 | 2
|
|
|
|
| 4 | 1 | four | 5 | -5 | 3 | -3
|
|
|
|
| 4 | 1 | four | 5 | -5 | 2 | 4
|
2009-09-13 00:12:09 +02:00
|
|
|
| 4 | 1 | four | 5 | -5 | 5 | -5
|
|
|
|
| 4 | 1 | four | 5 | -5 | 5 | -5
|
2009-11-28 01:46:19 +01:00
|
|
|
| 4 | 1 | four | 5 | -5 | 0 |
|
|
|
|
| 4 | 1 | four | 5 | -5 | |
|
|
|
|
| 4 | 1 | four | 5 | -5 | | 0
|
|
|
|
| 5 | 0 | five | 5 | -5 | 1 | -1
|
|
|
|
| 5 | 0 | five | 5 | -5 | 2 | 2
|
|
|
|
| 5 | 0 | five | 5 | -5 | 3 | -3
|
|
|
|
| 5 | 0 | five | 5 | -5 | 2 | 4
|
2009-09-13 00:12:09 +02:00
|
|
|
| 5 | 0 | five | 5 | -5 | 5 | -5
|
|
|
|
| 5 | 0 | five | 5 | -5 | 5 | -5
|
2009-11-28 01:46:19 +01:00
|
|
|
| 5 | 0 | five | 5 | -5 | 0 |
|
|
|
|
| 5 | 0 | five | 5 | -5 | |
|
|
|
|
| 5 | 0 | five | 5 | -5 | | 0
|
|
|
|
| 6 | 6 | six | 5 | -5 | 1 | -1
|
|
|
|
| 6 | 6 | six | 5 | -5 | 2 | 2
|
|
|
|
| 6 | 6 | six | 5 | -5 | 3 | -3
|
|
|
|
| 6 | 6 | six | 5 | -5 | 2 | 4
|
2009-09-13 00:12:09 +02:00
|
|
|
| 6 | 6 | six | 5 | -5 | 5 | -5
|
|
|
|
| 6 | 6 | six | 5 | -5 | 5 | -5
|
2009-11-28 01:46:19 +01:00
|
|
|
| 6 | 6 | six | 5 | -5 | 0 |
|
|
|
|
| 6 | 6 | six | 5 | -5 | |
|
|
|
|
| 6 | 6 | six | 5 | -5 | | 0
|
|
|
|
| 7 | 7 | seven | 5 | -5 | 1 | -1
|
|
|
|
| 7 | 7 | seven | 5 | -5 | 2 | 2
|
|
|
|
| 7 | 7 | seven | 5 | -5 | 3 | -3
|
|
|
|
| 7 | 7 | seven | 5 | -5 | 2 | 4
|
2009-09-13 00:12:09 +02:00
|
|
|
| 7 | 7 | seven | 5 | -5 | 5 | -5
|
|
|
|
| 7 | 7 | seven | 5 | -5 | 5 | -5
|
2009-11-28 01:46:19 +01:00
|
|
|
| 7 | 7 | seven | 5 | -5 | 0 |
|
|
|
|
| 7 | 7 | seven | 5 | -5 | |
|
|
|
|
| 7 | 7 | seven | 5 | -5 | | 0
|
|
|
|
| 8 | 8 | eight | 5 | -5 | 1 | -1
|
|
|
|
| 8 | 8 | eight | 5 | -5 | 2 | 2
|
|
|
|
| 8 | 8 | eight | 5 | -5 | 3 | -3
|
|
|
|
| 8 | 8 | eight | 5 | -5 | 2 | 4
|
2009-09-13 00:12:09 +02:00
|
|
|
| 8 | 8 | eight | 5 | -5 | 5 | -5
|
|
|
|
| 8 | 8 | eight | 5 | -5 | 5 | -5
|
2009-11-28 01:46:19 +01:00
|
|
|
| 8 | 8 | eight | 5 | -5 | 0 |
|
|
|
|
| 8 | 8 | eight | 5 | -5 | |
|
|
|
|
| 8 | 8 | eight | 5 | -5 | | 0
|
|
|
|
| 0 | | zero | 5 | -5 | 1 | -1
|
|
|
|
| 0 | | zero | 5 | -5 | 2 | 2
|
|
|
|
| 0 | | zero | 5 | -5 | 3 | -3
|
|
|
|
| 0 | | zero | 5 | -5 | 2 | 4
|
2009-09-13 00:12:09 +02:00
|
|
|
| 0 | | zero | 5 | -5 | 5 | -5
|
|
|
|
| 0 | | zero | 5 | -5 | 5 | -5
|
2009-11-28 01:46:19 +01:00
|
|
|
| 0 | | zero | 5 | -5 | 0 |
|
|
|
|
| 0 | | zero | 5 | -5 | |
|
|
|
|
| 0 | | zero | 5 | -5 | | 0
|
|
|
|
| | | null | 5 | -5 | 1 | -1
|
|
|
|
| | | null | 5 | -5 | 2 | 2
|
|
|
|
| | | null | 5 | -5 | 3 | -3
|
|
|
|
| | | null | 5 | -5 | 2 | 4
|
2009-09-13 00:12:09 +02:00
|
|
|
| | | null | 5 | -5 | 5 | -5
|
|
|
|
| | | null | 5 | -5 | 5 | -5
|
2009-11-28 01:46:19 +01:00
|
|
|
| | | null | 5 | -5 | 0 |
|
|
|
|
| | | null | 5 | -5 | |
|
|
|
|
| | | null | 5 | -5 | | 0
|
|
|
|
| | 0 | zero | 5 | -5 | 1 | -1
|
|
|
|
| | 0 | zero | 5 | -5 | 2 | 2
|
|
|
|
| | 0 | zero | 5 | -5 | 3 | -3
|
|
|
|
| | 0 | zero | 5 | -5 | 2 | 4
|
2009-09-13 00:12:09 +02:00
|
|
|
| | 0 | zero | 5 | -5 | 5 | -5
|
|
|
|
| | 0 | zero | 5 | -5 | 5 | -5
|
2009-11-28 01:46:19 +01:00
|
|
|
| | 0 | zero | 5 | -5 | 0 |
|
|
|
|
| | 0 | zero | 5 | -5 | |
|
|
|
|
| | 0 | zero | 5 | -5 | | 0
|
|
|
|
| 1 | 4 | one | 5 | -5 | 1 | -1
|
|
|
|
| 1 | 4 | one | 5 | -5 | 2 | 2
|
|
|
|
| 1 | 4 | one | 5 | -5 | 3 | -3
|
|
|
|
| 1 | 4 | one | 5 | -5 | 2 | 4
|
2009-09-13 00:12:09 +02:00
|
|
|
| 1 | 4 | one | 5 | -5 | 5 | -5
|
|
|
|
| 1 | 4 | one | 5 | -5 | 5 | -5
|
2009-11-28 01:46:19 +01:00
|
|
|
| 1 | 4 | one | 5 | -5 | 0 |
|
|
|
|
| 1 | 4 | one | 5 | -5 | |
|
|
|
|
| 1 | 4 | one | 5 | -5 | | 0
|
|
|
|
| 2 | 3 | two | 5 | -5 | 1 | -1
|
|
|
|
| 2 | 3 | two | 5 | -5 | 2 | 2
|
|
|
|
| 2 | 3 | two | 5 | -5 | 3 | -3
|
|
|
|
| 2 | 3 | two | 5 | -5 | 2 | 4
|
2009-09-13 00:12:09 +02:00
|
|
|
| 2 | 3 | two | 5 | -5 | 5 | -5
|
|
|
|
| 2 | 3 | two | 5 | -5 | 5 | -5
|
2009-11-28 01:46:19 +01:00
|
|
|
| 2 | 3 | two | 5 | -5 | 0 |
|
|
|
|
| 2 | 3 | two | 5 | -5 | |
|
|
|
|
| 2 | 3 | two | 5 | -5 | | 0
|
|
|
|
| 3 | 2 | three | 5 | -5 | 1 | -1
|
|
|
|
| 3 | 2 | three | 5 | -5 | 2 | 2
|
|
|
|
| 3 | 2 | three | 5 | -5 | 3 | -3
|
|
|
|
| 3 | 2 | three | 5 | -5 | 2 | 4
|
2009-09-13 00:12:09 +02:00
|
|
|
| 3 | 2 | three | 5 | -5 | 5 | -5
|
|
|
|
| 3 | 2 | three | 5 | -5 | 5 | -5
|
2009-11-28 01:46:19 +01:00
|
|
|
| 3 | 2 | three | 5 | -5 | 0 |
|
|
|
|
| 3 | 2 | three | 5 | -5 | |
|
|
|
|
| 3 | 2 | three | 5 | -5 | | 0
|
|
|
|
| 4 | 1 | four | 5 | -5 | 1 | -1
|
|
|
|
| 4 | 1 | four | 5 | -5 | 2 | 2
|
|
|
|
| 4 | 1 | four | 5 | -5 | 3 | -3
|
|
|
|
| 4 | 1 | four | 5 | -5 | 2 | 4
|
2009-09-13 00:12:09 +02:00
|
|
|
| 4 | 1 | four | 5 | -5 | 5 | -5
|
|
|
|
| 4 | 1 | four | 5 | -5 | 5 | -5
|
2009-11-28 01:46:19 +01:00
|
|
|
| 4 | 1 | four | 5 | -5 | 0 |
|
|
|
|
| 4 | 1 | four | 5 | -5 | |
|
|
|
|
| 4 | 1 | four | 5 | -5 | | 0
|
|
|
|
| 5 | 0 | five | 5 | -5 | 1 | -1
|
|
|
|
| 5 | 0 | five | 5 | -5 | 2 | 2
|
|
|
|
| 5 | 0 | five | 5 | -5 | 3 | -3
|
|
|
|
| 5 | 0 | five | 5 | -5 | 2 | 4
|
2009-09-13 00:12:09 +02:00
|
|
|
| 5 | 0 | five | 5 | -5 | 5 | -5
|
|
|
|
| 5 | 0 | five | 5 | -5 | 5 | -5
|
2009-11-28 01:46:19 +01:00
|
|
|
| 5 | 0 | five | 5 | -5 | 0 |
|
|
|
|
| 5 | 0 | five | 5 | -5 | |
|
|
|
|
| 5 | 0 | five | 5 | -5 | | 0
|
|
|
|
| 6 | 6 | six | 5 | -5 | 1 | -1
|
|
|
|
| 6 | 6 | six | 5 | -5 | 2 | 2
|
|
|
|
| 6 | 6 | six | 5 | -5 | 3 | -3
|
|
|
|
| 6 | 6 | six | 5 | -5 | 2 | 4
|
2009-09-13 00:12:09 +02:00
|
|
|
| 6 | 6 | six | 5 | -5 | 5 | -5
|
|
|
|
| 6 | 6 | six | 5 | -5 | 5 | -5
|
2009-11-28 01:46:19 +01:00
|
|
|
| 6 | 6 | six | 5 | -5 | 0 |
|
|
|
|
| 6 | 6 | six | 5 | -5 | |
|
|
|
|
| 6 | 6 | six | 5 | -5 | | 0
|
|
|
|
| 7 | 7 | seven | 5 | -5 | 1 | -1
|
|
|
|
| 7 | 7 | seven | 5 | -5 | 2 | 2
|
|
|
|
| 7 | 7 | seven | 5 | -5 | 3 | -3
|
|
|
|
| 7 | 7 | seven | 5 | -5 | 2 | 4
|
2009-09-13 00:12:09 +02:00
|
|
|
| 7 | 7 | seven | 5 | -5 | 5 | -5
|
|
|
|
| 7 | 7 | seven | 5 | -5 | 5 | -5
|
2009-11-28 01:46:19 +01:00
|
|
|
| 7 | 7 | seven | 5 | -5 | 0 |
|
|
|
|
| 7 | 7 | seven | 5 | -5 | |
|
|
|
|
| 7 | 7 | seven | 5 | -5 | | 0
|
|
|
|
| 8 | 8 | eight | 5 | -5 | 1 | -1
|
|
|
|
| 8 | 8 | eight | 5 | -5 | 2 | 2
|
|
|
|
| 8 | 8 | eight | 5 | -5 | 3 | -3
|
|
|
|
| 8 | 8 | eight | 5 | -5 | 2 | 4
|
2009-09-13 00:12:09 +02:00
|
|
|
| 8 | 8 | eight | 5 | -5 | 5 | -5
|
|
|
|
| 8 | 8 | eight | 5 | -5 | 5 | -5
|
2009-11-28 01:46:19 +01:00
|
|
|
| 8 | 8 | eight | 5 | -5 | 0 |
|
|
|
|
| 8 | 8 | eight | 5 | -5 | |
|
|
|
|
| 8 | 8 | eight | 5 | -5 | | 0
|
|
|
|
| 0 | | zero | 5 | -5 | 1 | -1
|
|
|
|
| 0 | | zero | 5 | -5 | 2 | 2
|
|
|
|
| 0 | | zero | 5 | -5 | 3 | -3
|
|
|
|
| 0 | | zero | 5 | -5 | 2 | 4
|
2009-09-13 00:12:09 +02:00
|
|
|
| 0 | | zero | 5 | -5 | 5 | -5
|
|
|
|
| 0 | | zero | 5 | -5 | 5 | -5
|
2009-11-28 01:46:19 +01:00
|
|
|
| 0 | | zero | 5 | -5 | 0 |
|
|
|
|
| 0 | | zero | 5 | -5 | |
|
|
|
|
| 0 | | zero | 5 | -5 | | 0
|
|
|
|
| | | null | 5 | -5 | 1 | -1
|
|
|
|
| | | null | 5 | -5 | 2 | 2
|
|
|
|
| | | null | 5 | -5 | 3 | -3
|
|
|
|
| | | null | 5 | -5 | 2 | 4
|
2009-09-13 00:12:09 +02:00
|
|
|
| | | null | 5 | -5 | 5 | -5
|
|
|
|
| | | null | 5 | -5 | 5 | -5
|
2009-11-28 01:46:19 +01:00
|
|
|
| | | null | 5 | -5 | 0 |
|
|
|
|
| | | null | 5 | -5 | |
|
|
|
|
| | | null | 5 | -5 | | 0
|
|
|
|
| | 0 | zero | 5 | -5 | 1 | -1
|
|
|
|
| | 0 | zero | 5 | -5 | 2 | 2
|
|
|
|
| | 0 | zero | 5 | -5 | 3 | -3
|
|
|
|
| | 0 | zero | 5 | -5 | 2 | 4
|
2009-09-13 00:12:09 +02:00
|
|
|
| | 0 | zero | 5 | -5 | 5 | -5
|
|
|
|
| | 0 | zero | 5 | -5 | 5 | -5
|
2009-11-28 01:46:19 +01:00
|
|
|
| | 0 | zero | 5 | -5 | 0 |
|
|
|
|
| | 0 | zero | 5 | -5 | |
|
|
|
|
| | 0 | zero | 5 | -5 | | 0
|
|
|
|
| 1 | 4 | one | 0 | | 1 | -1
|
|
|
|
| 1 | 4 | one | 0 | | 2 | 2
|
|
|
|
| 1 | 4 | one | 0 | | 3 | -3
|
|
|
|
| 1 | 4 | one | 0 | | 2 | 4
|
|
|
|
| 1 | 4 | one | 0 | | 5 | -5
|
|
|
|
| 1 | 4 | one | 0 | | 5 | -5
|
|
|
|
| 1 | 4 | one | 0 | | 0 |
|
|
|
|
| 1 | 4 | one | 0 | | |
|
|
|
|
| 1 | 4 | one | 0 | | | 0
|
|
|
|
| 2 | 3 | two | 0 | | 1 | -1
|
|
|
|
| 2 | 3 | two | 0 | | 2 | 2
|
|
|
|
| 2 | 3 | two | 0 | | 3 | -3
|
|
|
|
| 2 | 3 | two | 0 | | 2 | 4
|
|
|
|
| 2 | 3 | two | 0 | | 5 | -5
|
|
|
|
| 2 | 3 | two | 0 | | 5 | -5
|
2009-09-13 00:12:09 +02:00
|
|
|
| 2 | 3 | two | 0 | | 0 |
|
2009-11-28 01:46:19 +01:00
|
|
|
| 2 | 3 | two | 0 | | |
|
|
|
|
| 2 | 3 | two | 0 | | | 0
|
|
|
|
| 3 | 2 | three | 0 | | 1 | -1
|
|
|
|
| 3 | 2 | three | 0 | | 2 | 2
|
|
|
|
| 3 | 2 | three | 0 | | 3 | -3
|
|
|
|
| 3 | 2 | three | 0 | | 2 | 4
|
|
|
|
| 3 | 2 | three | 0 | | 5 | -5
|
|
|
|
| 3 | 2 | three | 0 | | 5 | -5
|
2009-09-13 00:12:09 +02:00
|
|
|
| 3 | 2 | three | 0 | | 0 |
|
2009-11-28 01:46:19 +01:00
|
|
|
| 3 | 2 | three | 0 | | |
|
|
|
|
| 3 | 2 | three | 0 | | | 0
|
|
|
|
| 4 | 1 | four | 0 | | 1 | -1
|
|
|
|
| 4 | 1 | four | 0 | | 2 | 2
|
|
|
|
| 4 | 1 | four | 0 | | 3 | -3
|
|
|
|
| 4 | 1 | four | 0 | | 2 | 4
|
|
|
|
| 4 | 1 | four | 0 | | 5 | -5
|
|
|
|
| 4 | 1 | four | 0 | | 5 | -5
|
2009-09-13 00:12:09 +02:00
|
|
|
| 4 | 1 | four | 0 | | 0 |
|
2009-11-28 01:46:19 +01:00
|
|
|
| 4 | 1 | four | 0 | | |
|
|
|
|
| 4 | 1 | four | 0 | | | 0
|
|
|
|
| 5 | 0 | five | 0 | | 1 | -1
|
|
|
|
| 5 | 0 | five | 0 | | 2 | 2
|
|
|
|
| 5 | 0 | five | 0 | | 3 | -3
|
|
|
|
| 5 | 0 | five | 0 | | 2 | 4
|
|
|
|
| 5 | 0 | five | 0 | | 5 | -5
|
|
|
|
| 5 | 0 | five | 0 | | 5 | -5
|
2009-09-13 00:12:09 +02:00
|
|
|
| 5 | 0 | five | 0 | | 0 |
|
2009-11-28 01:46:19 +01:00
|
|
|
| 5 | 0 | five | 0 | | |
|
|
|
|
| 5 | 0 | five | 0 | | | 0
|
|
|
|
| 6 | 6 | six | 0 | | 1 | -1
|
|
|
|
| 6 | 6 | six | 0 | | 2 | 2
|
|
|
|
| 6 | 6 | six | 0 | | 3 | -3
|
|
|
|
| 6 | 6 | six | 0 | | 2 | 4
|
|
|
|
| 6 | 6 | six | 0 | | 5 | -5
|
|
|
|
| 6 | 6 | six | 0 | | 5 | -5
|
2009-09-13 00:12:09 +02:00
|
|
|
| 6 | 6 | six | 0 | | 0 |
|
2009-11-28 01:46:19 +01:00
|
|
|
| 6 | 6 | six | 0 | | |
|
|
|
|
| 6 | 6 | six | 0 | | | 0
|
|
|
|
| 7 | 7 | seven | 0 | | 1 | -1
|
|
|
|
| 7 | 7 | seven | 0 | | 2 | 2
|
|
|
|
| 7 | 7 | seven | 0 | | 3 | -3
|
|
|
|
| 7 | 7 | seven | 0 | | 2 | 4
|
|
|
|
| 7 | 7 | seven | 0 | | 5 | -5
|
|
|
|
| 7 | 7 | seven | 0 | | 5 | -5
|
2009-09-13 00:12:09 +02:00
|
|
|
| 7 | 7 | seven | 0 | | 0 |
|
2009-11-28 01:46:19 +01:00
|
|
|
| 7 | 7 | seven | 0 | | |
|
|
|
|
| 7 | 7 | seven | 0 | | | 0
|
|
|
|
| 8 | 8 | eight | 0 | | 1 | -1
|
|
|
|
| 8 | 8 | eight | 0 | | 2 | 2
|
|
|
|
| 8 | 8 | eight | 0 | | 3 | -3
|
|
|
|
| 8 | 8 | eight | 0 | | 2 | 4
|
|
|
|
| 8 | 8 | eight | 0 | | 5 | -5
|
|
|
|
| 8 | 8 | eight | 0 | | 5 | -5
|
2009-09-13 00:12:09 +02:00
|
|
|
| 8 | 8 | eight | 0 | | 0 |
|
2009-11-28 01:46:19 +01:00
|
|
|
| 8 | 8 | eight | 0 | | |
|
|
|
|
| 8 | 8 | eight | 0 | | | 0
|
|
|
|
| 0 | | zero | 0 | | 1 | -1
|
|
|
|
| 0 | | zero | 0 | | 2 | 2
|
|
|
|
| 0 | | zero | 0 | | 3 | -3
|
|
|
|
| 0 | | zero | 0 | | 2 | 4
|
|
|
|
| 0 | | zero | 0 | | 5 | -5
|
|
|
|
| 0 | | zero | 0 | | 5 | -5
|
2009-09-13 00:12:09 +02:00
|
|
|
| 0 | | zero | 0 | | 0 |
|
2009-11-28 01:46:19 +01:00
|
|
|
| 0 | | zero | 0 | | |
|
|
|
|
| 0 | | zero | 0 | | | 0
|
|
|
|
| | | null | 0 | | 1 | -1
|
|
|
|
| | | null | 0 | | 2 | 2
|
|
|
|
| | | null | 0 | | 3 | -3
|
|
|
|
| | | null | 0 | | 2 | 4
|
|
|
|
| | | null | 0 | | 5 | -5
|
|
|
|
| | | null | 0 | | 5 | -5
|
2009-09-13 00:12:09 +02:00
|
|
|
| | | null | 0 | | 0 |
|
2009-11-28 01:46:19 +01:00
|
|
|
| | | null | 0 | | |
|
|
|
|
| | | null | 0 | | | 0
|
|
|
|
| | 0 | zero | 0 | | 1 | -1
|
|
|
|
| | 0 | zero | 0 | | 2 | 2
|
|
|
|
| | 0 | zero | 0 | | 3 | -3
|
|
|
|
| | 0 | zero | 0 | | 2 | 4
|
|
|
|
| | 0 | zero | 0 | | 5 | -5
|
|
|
|
| | 0 | zero | 0 | | 5 | -5
|
2009-09-13 00:12:09 +02:00
|
|
|
| | 0 | zero | 0 | | 0 |
|
2009-11-28 01:46:19 +01:00
|
|
|
| | 0 | zero | 0 | | |
|
|
|
|
| | 0 | zero | 0 | | | 0
|
|
|
|
| 1 | 4 | one | | | 1 | -1
|
|
|
|
| 1 | 4 | one | | | 2 | 2
|
|
|
|
| 1 | 4 | one | | | 3 | -3
|
|
|
|
| 1 | 4 | one | | | 2 | 4
|
|
|
|
| 1 | 4 | one | | | 5 | -5
|
|
|
|
| 1 | 4 | one | | | 5 | -5
|
|
|
|
| 1 | 4 | one | | | 0 |
|
2009-09-13 00:12:09 +02:00
|
|
|
| 1 | 4 | one | | | |
|
2009-11-28 01:46:19 +01:00
|
|
|
| 1 | 4 | one | | | | 0
|
|
|
|
| 2 | 3 | two | | | 1 | -1
|
|
|
|
| 2 | 3 | two | | | 2 | 2
|
|
|
|
| 2 | 3 | two | | | 3 | -3
|
|
|
|
| 2 | 3 | two | | | 2 | 4
|
|
|
|
| 2 | 3 | two | | | 5 | -5
|
|
|
|
| 2 | 3 | two | | | 5 | -5
|
|
|
|
| 2 | 3 | two | | | 0 |
|
2009-09-13 00:12:09 +02:00
|
|
|
| 2 | 3 | two | | | |
|
2009-11-28 01:46:19 +01:00
|
|
|
| 2 | 3 | two | | | | 0
|
|
|
|
| 3 | 2 | three | | | 1 | -1
|
|
|
|
| 3 | 2 | three | | | 2 | 2
|
|
|
|
| 3 | 2 | three | | | 3 | -3
|
|
|
|
| 3 | 2 | three | | | 2 | 4
|
|
|
|
| 3 | 2 | three | | | 5 | -5
|
|
|
|
| 3 | 2 | three | | | 5 | -5
|
|
|
|
| 3 | 2 | three | | | 0 |
|
2009-09-13 00:12:09 +02:00
|
|
|
| 3 | 2 | three | | | |
|
2009-11-28 01:46:19 +01:00
|
|
|
| 3 | 2 | three | | | | 0
|
|
|
|
| 4 | 1 | four | | | 1 | -1
|
|
|
|
| 4 | 1 | four | | | 2 | 2
|
|
|
|
| 4 | 1 | four | | | 3 | -3
|
|
|
|
| 4 | 1 | four | | | 2 | 4
|
|
|
|
| 4 | 1 | four | | | 5 | -5
|
|
|
|
| 4 | 1 | four | | | 5 | -5
|
|
|
|
| 4 | 1 | four | | | 0 |
|
2009-09-13 00:12:09 +02:00
|
|
|
| 4 | 1 | four | | | |
|
2009-11-28 01:46:19 +01:00
|
|
|
| 4 | 1 | four | | | | 0
|
|
|
|
| 5 | 0 | five | | | 1 | -1
|
|
|
|
| 5 | 0 | five | | | 2 | 2
|
|
|
|
| 5 | 0 | five | | | 3 | -3
|
|
|
|
| 5 | 0 | five | | | 2 | 4
|
|
|
|
| 5 | 0 | five | | | 5 | -5
|
|
|
|
| 5 | 0 | five | | | 5 | -5
|
|
|
|
| 5 | 0 | five | | | 0 |
|
2009-09-13 00:12:09 +02:00
|
|
|
| 5 | 0 | five | | | |
|
2009-11-28 01:46:19 +01:00
|
|
|
| 5 | 0 | five | | | | 0
|
|
|
|
| 6 | 6 | six | | | 1 | -1
|
|
|
|
| 6 | 6 | six | | | 2 | 2
|
|
|
|
| 6 | 6 | six | | | 3 | -3
|
|
|
|
| 6 | 6 | six | | | 2 | 4
|
|
|
|
| 6 | 6 | six | | | 5 | -5
|
|
|
|
| 6 | 6 | six | | | 5 | -5
|
|
|
|
| 6 | 6 | six | | | 0 |
|
|
|
|
| 6 | 6 | six | | | |
|
|
|
|
| 6 | 6 | six | | | | 0
|
|
|
|
| 7 | 7 | seven | | | 1 | -1
|
|
|
|
| 7 | 7 | seven | | | 2 | 2
|
|
|
|
| 7 | 7 | seven | | | 3 | -3
|
|
|
|
| 7 | 7 | seven | | | 2 | 4
|
|
|
|
| 7 | 7 | seven | | | 5 | -5
|
|
|
|
| 7 | 7 | seven | | | 5 | -5
|
|
|
|
| 7 | 7 | seven | | | 0 |
|
2009-09-13 00:12:09 +02:00
|
|
|
| 7 | 7 | seven | | | |
|
2009-11-28 01:46:19 +01:00
|
|
|
| 7 | 7 | seven | | | | 0
|
|
|
|
| 8 | 8 | eight | | | 1 | -1
|
|
|
|
| 8 | 8 | eight | | | 2 | 2
|
|
|
|
| 8 | 8 | eight | | | 3 | -3
|
|
|
|
| 8 | 8 | eight | | | 2 | 4
|
|
|
|
| 8 | 8 | eight | | | 5 | -5
|
|
|
|
| 8 | 8 | eight | | | 5 | -5
|
|
|
|
| 8 | 8 | eight | | | 0 |
|
2009-09-13 00:12:09 +02:00
|
|
|
| 8 | 8 | eight | | | |
|
2009-11-28 01:46:19 +01:00
|
|
|
| 8 | 8 | eight | | | | 0
|
|
|
|
| 0 | | zero | | | 1 | -1
|
|
|
|
| 0 | | zero | | | 2 | 2
|
|
|
|
| 0 | | zero | | | 3 | -3
|
|
|
|
| 0 | | zero | | | 2 | 4
|
|
|
|
| 0 | | zero | | | 5 | -5
|
|
|
|
| 0 | | zero | | | 5 | -5
|
|
|
|
| 0 | | zero | | | 0 |
|
2009-09-13 00:12:09 +02:00
|
|
|
| 0 | | zero | | | |
|
2009-11-28 01:46:19 +01:00
|
|
|
| 0 | | zero | | | | 0
|
|
|
|
| | | null | | | 1 | -1
|
|
|
|
| | | null | | | 2 | 2
|
|
|
|
| | | null | | | 3 | -3
|
|
|
|
| | | null | | | 2 | 4
|
|
|
|
| | | null | | | 5 | -5
|
|
|
|
| | | null | | | 5 | -5
|
|
|
|
| | | null | | | 0 |
|
2009-09-13 00:12:09 +02:00
|
|
|
| | | null | | | |
|
2009-11-28 01:46:19 +01:00
|
|
|
| | | null | | | | 0
|
|
|
|
| | 0 | zero | | | 1 | -1
|
|
|
|
| | 0 | zero | | | 2 | 2
|
|
|
|
| | 0 | zero | | | 3 | -3
|
|
|
|
| | 0 | zero | | | 2 | 4
|
|
|
|
| | 0 | zero | | | 5 | -5
|
|
|
|
| | 0 | zero | | | 5 | -5
|
|
|
|
| | 0 | zero | | | 0 |
|
2009-09-13 00:12:09 +02:00
|
|
|
| | 0 | zero | | | |
|
2009-11-28 01:46:19 +01:00
|
|
|
| | 0 | zero | | | | 0
|
|
|
|
| 1 | 4 | one | | 0 | 1 | -1
|
|
|
|
| 1 | 4 | one | | 0 | 2 | 2
|
|
|
|
| 1 | 4 | one | | 0 | 3 | -3
|
|
|
|
| 1 | 4 | one | | 0 | 2 | 4
|
|
|
|
| 1 | 4 | one | | 0 | 5 | -5
|
|
|
|
| 1 | 4 | one | | 0 | 5 | -5
|
|
|
|
| 1 | 4 | one | | 0 | 0 |
|
|
|
|
| 1 | 4 | one | | 0 | |
|
2005-07-22 21:12:02 +02:00
|
|
|
| 1 | 4 | one | | 0 | | 0
|
2009-11-28 01:46:19 +01:00
|
|
|
| 2 | 3 | two | | 0 | 1 | -1
|
|
|
|
| 2 | 3 | two | | 0 | 2 | 2
|
|
|
|
| 2 | 3 | two | | 0 | 3 | -3
|
|
|
|
| 2 | 3 | two | | 0 | 2 | 4
|
|
|
|
| 2 | 3 | two | | 0 | 5 | -5
|
|
|
|
| 2 | 3 | two | | 0 | 5 | -5
|
|
|
|
| 2 | 3 | two | | 0 | 0 |
|
|
|
|
| 2 | 3 | two | | 0 | |
|
2005-07-22 21:12:02 +02:00
|
|
|
| 2 | 3 | two | | 0 | | 0
|
2009-11-28 01:46:19 +01:00
|
|
|
| 3 | 2 | three | | 0 | 1 | -1
|
|
|
|
| 3 | 2 | three | | 0 | 2 | 2
|
|
|
|
| 3 | 2 | three | | 0 | 3 | -3
|
|
|
|
| 3 | 2 | three | | 0 | 2 | 4
|
|
|
|
| 3 | 2 | three | | 0 | 5 | -5
|
|
|
|
| 3 | 2 | three | | 0 | 5 | -5
|
|
|
|
| 3 | 2 | three | | 0 | 0 |
|
|
|
|
| 3 | 2 | three | | 0 | |
|
2005-07-22 21:12:02 +02:00
|
|
|
| 3 | 2 | three | | 0 | | 0
|
2009-11-28 01:46:19 +01:00
|
|
|
| 4 | 1 | four | | 0 | 1 | -1
|
|
|
|
| 4 | 1 | four | | 0 | 2 | 2
|
|
|
|
| 4 | 1 | four | | 0 | 3 | -3
|
|
|
|
| 4 | 1 | four | | 0 | 2 | 4
|
|
|
|
| 4 | 1 | four | | 0 | 5 | -5
|
|
|
|
| 4 | 1 | four | | 0 | 5 | -5
|
|
|
|
| 4 | 1 | four | | 0 | 0 |
|
|
|
|
| 4 | 1 | four | | 0 | |
|
2005-07-22 21:12:02 +02:00
|
|
|
| 4 | 1 | four | | 0 | | 0
|
2009-11-28 01:46:19 +01:00
|
|
|
| 5 | 0 | five | | 0 | 1 | -1
|
|
|
|
| 5 | 0 | five | | 0 | 2 | 2
|
|
|
|
| 5 | 0 | five | | 0 | 3 | -3
|
|
|
|
| 5 | 0 | five | | 0 | 2 | 4
|
|
|
|
| 5 | 0 | five | | 0 | 5 | -5
|
|
|
|
| 5 | 0 | five | | 0 | 5 | -5
|
|
|
|
| 5 | 0 | five | | 0 | 0 |
|
|
|
|
| 5 | 0 | five | | 0 | |
|
2005-07-22 21:12:02 +02:00
|
|
|
| 5 | 0 | five | | 0 | | 0
|
2009-11-28 01:46:19 +01:00
|
|
|
| 6 | 6 | six | | 0 | 1 | -1
|
|
|
|
| 6 | 6 | six | | 0 | 2 | 2
|
|
|
|
| 6 | 6 | six | | 0 | 3 | -3
|
|
|
|
| 6 | 6 | six | | 0 | 2 | 4
|
|
|
|
| 6 | 6 | six | | 0 | 5 | -5
|
|
|
|
| 6 | 6 | six | | 0 | 5 | -5
|
|
|
|
| 6 | 6 | six | | 0 | 0 |
|
|
|
|
| 6 | 6 | six | | 0 | |
|
2005-07-22 21:12:02 +02:00
|
|
|
| 6 | 6 | six | | 0 | | 0
|
2009-11-28 01:46:19 +01:00
|
|
|
| 7 | 7 | seven | | 0 | 1 | -1
|
|
|
|
| 7 | 7 | seven | | 0 | 2 | 2
|
|
|
|
| 7 | 7 | seven | | 0 | 3 | -3
|
|
|
|
| 7 | 7 | seven | | 0 | 2 | 4
|
|
|
|
| 7 | 7 | seven | | 0 | 5 | -5
|
|
|
|
| 7 | 7 | seven | | 0 | 5 | -5
|
|
|
|
| 7 | 7 | seven | | 0 | 0 |
|
|
|
|
| 7 | 7 | seven | | 0 | |
|
2005-07-22 21:12:02 +02:00
|
|
|
| 7 | 7 | seven | | 0 | | 0
|
2009-11-28 01:46:19 +01:00
|
|
|
| 8 | 8 | eight | | 0 | 1 | -1
|
|
|
|
| 8 | 8 | eight | | 0 | 2 | 2
|
|
|
|
| 8 | 8 | eight | | 0 | 3 | -3
|
|
|
|
| 8 | 8 | eight | | 0 | 2 | 4
|
|
|
|
| 8 | 8 | eight | | 0 | 5 | -5
|
|
|
|
| 8 | 8 | eight | | 0 | 5 | -5
|
|
|
|
| 8 | 8 | eight | | 0 | 0 |
|
|
|
|
| 8 | 8 | eight | | 0 | |
|
2005-07-22 21:12:02 +02:00
|
|
|
| 8 | 8 | eight | | 0 | | 0
|
2009-11-28 01:46:19 +01:00
|
|
|
| 0 | | zero | | 0 | 1 | -1
|
|
|
|
| 0 | | zero | | 0 | 2 | 2
|
|
|
|
| 0 | | zero | | 0 | 3 | -3
|
|
|
|
| 0 | | zero | | 0 | 2 | 4
|
|
|
|
| 0 | | zero | | 0 | 5 | -5
|
|
|
|
| 0 | | zero | | 0 | 5 | -5
|
|
|
|
| 0 | | zero | | 0 | 0 |
|
|
|
|
| 0 | | zero | | 0 | |
|
2005-07-22 21:12:02 +02:00
|
|
|
| 0 | | zero | | 0 | | 0
|
2009-11-28 01:46:19 +01:00
|
|
|
| | | null | | 0 | 1 | -1
|
|
|
|
| | | null | | 0 | 2 | 2
|
|
|
|
| | | null | | 0 | 3 | -3
|
|
|
|
| | | null | | 0 | 2 | 4
|
|
|
|
| | | null | | 0 | 5 | -5
|
|
|
|
| | | null | | 0 | 5 | -5
|
|
|
|
| | | null | | 0 | 0 |
|
|
|
|
| | | null | | 0 | |
|
2005-07-22 21:12:02 +02:00
|
|
|
| | | null | | 0 | | 0
|
2009-11-28 01:46:19 +01:00
|
|
|
| | 0 | zero | | 0 | 1 | -1
|
|
|
|
| | 0 | zero | | 0 | 2 | 2
|
|
|
|
| | 0 | zero | | 0 | 3 | -3
|
|
|
|
| | 0 | zero | | 0 | 2 | 4
|
|
|
|
| | 0 | zero | | 0 | 5 | -5
|
|
|
|
| | 0 | zero | | 0 | 5 | -5
|
|
|
|
| | 0 | zero | | 0 | 0 |
|
|
|
|
| | 0 | zero | | 0 | |
|
2000-11-06 17:03:47 +01:00
|
|
|
| | 0 | zero | | 0 | | 0
|
|
|
|
(891 rows)
|
2000-09-12 23:07:18 +02:00
|
|
|
|
2000-01-09 04:48:39 +01:00
|
|
|
--
|
|
|
|
--
|
|
|
|
-- Inner joins (equi-joins)
|
|
|
|
--
|
|
|
|
--
|
|
|
|
--
|
|
|
|
-- Inner joins (equi-joins) with USING clause
|
|
|
|
-- The USING syntax changes the shape of the resulting table
|
|
|
|
-- by including a column in the USING clause only once in the result.
|
|
|
|
--
|
|
|
|
-- Inner equi-join on specified column
|
|
|
|
SELECT '' AS "xxx", *
|
2000-02-15 04:31:33 +01:00
|
|
|
FROM J1_TBL INNER JOIN J2_TBL USING (i);
|
|
|
|
xxx | i | j | t | k
|
|
|
|
-----+---+---+-------+----
|
2005-03-24 20:14:49 +01:00
|
|
|
| 0 | | zero |
|
2000-11-06 17:03:47 +01:00
|
|
|
| 1 | 4 | one | -1
|
2004-12-01 20:00:56 +01:00
|
|
|
| 2 | 3 | two | 2
|
2005-03-24 20:14:49 +01:00
|
|
|
| 2 | 3 | two | 4
|
2000-11-06 17:03:47 +01:00
|
|
|
| 3 | 2 | three | -3
|
|
|
|
| 5 | 0 | five | -5
|
|
|
|
| 5 | 0 | five | -5
|
|
|
|
(7 rows)
|
2000-02-15 04:31:33 +01:00
|
|
|
|
2000-01-09 04:48:39 +01:00
|
|
|
-- Same as above, slightly different syntax
|
|
|
|
SELECT '' AS "xxx", *
|
2000-02-15 04:31:33 +01:00
|
|
|
FROM J1_TBL JOIN J2_TBL USING (i);
|
|
|
|
xxx | i | j | t | k
|
|
|
|
-----+---+---+-------+----
|
2005-03-24 20:14:49 +01:00
|
|
|
| 0 | | zero |
|
2000-11-06 17:03:47 +01:00
|
|
|
| 1 | 4 | one | -1
|
2004-12-01 20:00:56 +01:00
|
|
|
| 2 | 3 | two | 2
|
2005-03-24 20:14:49 +01:00
|
|
|
| 2 | 3 | two | 4
|
2000-11-06 17:03:47 +01:00
|
|
|
| 3 | 2 | three | -3
|
|
|
|
| 5 | 0 | five | -5
|
|
|
|
| 5 | 0 | five | -5
|
|
|
|
(7 rows)
|
2000-02-15 04:31:33 +01:00
|
|
|
|
|
|
|
SELECT '' AS "xxx", *
|
2000-11-06 19:11:46 +01:00
|
|
|
FROM J1_TBL t1 (a, b, c) JOIN J2_TBL t2 (a, d) USING (a)
|
|
|
|
ORDER BY a, d;
|
2000-02-15 04:31:33 +01:00
|
|
|
xxx | a | b | c | d
|
|
|
|
-----+---+---+-------+----
|
2000-11-06 17:03:47 +01:00
|
|
|
| 0 | | zero |
|
|
|
|
| 1 | 4 | one | -1
|
|
|
|
| 2 | 3 | two | 2
|
|
|
|
| 2 | 3 | two | 4
|
|
|
|
| 3 | 2 | three | -3
|
|
|
|
| 5 | 0 | five | -5
|
|
|
|
| 5 | 0 | five | -5
|
|
|
|
(7 rows)
|
2000-02-15 04:31:33 +01:00
|
|
|
|
|
|
|
SELECT '' AS "xxx", *
|
2000-11-06 19:11:46 +01:00
|
|
|
FROM J1_TBL t1 (a, b, c) JOIN J2_TBL t2 (a, b) USING (b)
|
|
|
|
ORDER BY b, t1.a;
|
2000-11-06 17:03:47 +01:00
|
|
|
xxx | b | a | c | a
|
|
|
|
-----+---+---+-------+---
|
|
|
|
| 0 | 5 | five |
|
|
|
|
| 0 | | zero |
|
|
|
|
| 2 | 3 | three | 2
|
|
|
|
| 4 | 1 | one | 2
|
|
|
|
(4 rows)
|
2000-02-15 04:31:33 +01:00
|
|
|
|
|
|
|
--
|
|
|
|
-- NATURAL JOIN
|
|
|
|
-- Inner equi-join on all columns with the same name
|
|
|
|
--
|
|
|
|
SELECT '' AS "xxx", *
|
|
|
|
FROM J1_TBL NATURAL JOIN J2_TBL;
|
|
|
|
xxx | i | j | t | k
|
|
|
|
-----+---+---+-------+----
|
2005-03-24 20:14:49 +01:00
|
|
|
| 0 | | zero |
|
2000-11-06 17:03:47 +01:00
|
|
|
| 1 | 4 | one | -1
|
2004-12-01 20:00:56 +01:00
|
|
|
| 2 | 3 | two | 2
|
2005-03-24 20:14:49 +01:00
|
|
|
| 2 | 3 | two | 4
|
2000-11-06 17:03:47 +01:00
|
|
|
| 3 | 2 | three | -3
|
|
|
|
| 5 | 0 | five | -5
|
|
|
|
| 5 | 0 | five | -5
|
|
|
|
(7 rows)
|
2000-02-15 04:31:33 +01:00
|
|
|
|
|
|
|
SELECT '' AS "xxx", *
|
|
|
|
FROM J1_TBL t1 (a, b, c) NATURAL JOIN J2_TBL t2 (a, d);
|
|
|
|
xxx | a | b | c | d
|
|
|
|
-----+---+---+-------+----
|
2005-03-24 20:14:49 +01:00
|
|
|
| 0 | | zero |
|
2000-11-06 17:03:47 +01:00
|
|
|
| 1 | 4 | one | -1
|
2004-12-01 20:00:56 +01:00
|
|
|
| 2 | 3 | two | 2
|
2005-03-24 20:14:49 +01:00
|
|
|
| 2 | 3 | two | 4
|
2000-11-06 17:03:47 +01:00
|
|
|
| 3 | 2 | three | -3
|
|
|
|
| 5 | 0 | five | -5
|
|
|
|
| 5 | 0 | five | -5
|
|
|
|
(7 rows)
|
2000-02-15 04:31:33 +01:00
|
|
|
|
|
|
|
SELECT '' AS "xxx", *
|
|
|
|
FROM J1_TBL t1 (a, b, c) NATURAL JOIN J2_TBL t2 (d, a);
|
|
|
|
xxx | a | b | c | d
|
|
|
|
-----+---+---+------+---
|
2005-03-24 20:14:49 +01:00
|
|
|
| 0 | | zero |
|
2000-11-06 17:03:47 +01:00
|
|
|
| 2 | 3 | two | 2
|
|
|
|
| 4 | 1 | four | 2
|
|
|
|
(3 rows)
|
2000-02-15 04:31:33 +01:00
|
|
|
|
|
|
|
-- mismatch number of columns
|
|
|
|
-- currently, Postgres will fill in with underlying names
|
|
|
|
SELECT '' AS "xxx", *
|
|
|
|
FROM J1_TBL t1 (a, b) NATURAL JOIN J2_TBL t2 (a);
|
|
|
|
xxx | a | b | t | k
|
|
|
|
-----+---+---+-------+----
|
2005-03-24 20:14:49 +01:00
|
|
|
| 0 | | zero |
|
2000-11-06 17:03:47 +01:00
|
|
|
| 1 | 4 | one | -1
|
2004-12-01 20:00:56 +01:00
|
|
|
| 2 | 3 | two | 2
|
2005-03-24 20:14:49 +01:00
|
|
|
| 2 | 3 | two | 4
|
2000-11-06 17:03:47 +01:00
|
|
|
| 3 | 2 | three | -3
|
|
|
|
| 5 | 0 | five | -5
|
|
|
|
| 5 | 0 | five | -5
|
|
|
|
(7 rows)
|
2000-02-15 04:31:33 +01:00
|
|
|
|
2000-01-09 04:48:39 +01:00
|
|
|
--
|
|
|
|
-- Inner joins (equi-joins)
|
|
|
|
--
|
|
|
|
SELECT '' AS "xxx", *
|
2000-02-15 04:31:33 +01:00
|
|
|
FROM J1_TBL JOIN J2_TBL ON (J1_TBL.i = J2_TBL.i);
|
2000-05-12 03:33:56 +02:00
|
|
|
xxx | i | j | t | i | k
|
|
|
|
-----+---+---+-------+---+----
|
2005-03-24 20:14:49 +01:00
|
|
|
| 0 | | zero | 0 |
|
2000-11-06 17:03:47 +01:00
|
|
|
| 1 | 4 | one | 1 | -1
|
2004-12-01 20:00:56 +01:00
|
|
|
| 2 | 3 | two | 2 | 2
|
2005-03-24 20:14:49 +01:00
|
|
|
| 2 | 3 | two | 2 | 4
|
2000-11-06 17:03:47 +01:00
|
|
|
| 3 | 2 | three | 3 | -3
|
|
|
|
| 5 | 0 | five | 5 | -5
|
|
|
|
| 5 | 0 | five | 5 | -5
|
|
|
|
(7 rows)
|
2000-05-12 03:33:56 +02:00
|
|
|
|
2000-01-09 04:48:39 +01:00
|
|
|
SELECT '' AS "xxx", *
|
2000-02-15 04:31:33 +01:00
|
|
|
FROM J1_TBL JOIN J2_TBL ON (J1_TBL.i = J2_TBL.k);
|
2000-05-12 03:33:56 +02:00
|
|
|
xxx | i | j | t | i | k
|
|
|
|
-----+---+---+------+---+---
|
2005-03-24 20:14:49 +01:00
|
|
|
| 0 | | zero | | 0
|
2000-11-06 17:03:47 +01:00
|
|
|
| 2 | 3 | two | 2 | 2
|
|
|
|
| 4 | 1 | four | 2 | 4
|
|
|
|
(3 rows)
|
2000-05-12 03:33:56 +02:00
|
|
|
|
2000-01-09 04:48:39 +01:00
|
|
|
--
|
|
|
|
-- Non-equi-joins
|
|
|
|
--
|
|
|
|
SELECT '' AS "xxx", *
|
2000-02-15 04:31:33 +01:00
|
|
|
FROM J1_TBL JOIN J2_TBL ON (J1_TBL.i <= J2_TBL.k);
|
2000-05-12 03:33:56 +02:00
|
|
|
xxx | i | j | t | i | k
|
|
|
|
-----+---+---+-------+---+---
|
2000-11-06 17:03:47 +01:00
|
|
|
| 1 | 4 | one | 2 | 2
|
|
|
|
| 2 | 3 | two | 2 | 2
|
2004-12-02 02:34:18 +01:00
|
|
|
| 0 | | zero | 2 | 2
|
|
|
|
| 1 | 4 | one | 2 | 4
|
2000-11-06 17:03:47 +01:00
|
|
|
| 2 | 3 | two | 2 | 4
|
|
|
|
| 3 | 2 | three | 2 | 4
|
|
|
|
| 4 | 1 | four | 2 | 4
|
|
|
|
| 0 | | zero | 2 | 4
|
|
|
|
| 0 | | zero | | 0
|
|
|
|
(9 rows)
|
2000-05-12 03:33:56 +02:00
|
|
|
|
2000-01-09 04:48:39 +01:00
|
|
|
--
|
|
|
|
-- Outer joins
|
2000-11-06 17:03:47 +01:00
|
|
|
-- Note that OUTER is a noise word
|
2000-01-09 04:48:39 +01:00
|
|
|
--
|
|
|
|
SELECT '' AS "xxx", *
|
2002-10-28 23:54:45 +01:00
|
|
|
FROM J1_TBL LEFT OUTER JOIN J2_TBL USING (i)
|
2004-12-03 23:19:28 +01:00
|
|
|
ORDER BY i, k, t;
|
2000-09-12 23:07:18 +02:00
|
|
|
xxx | i | j | t | k
|
|
|
|
-----+---+---+-------+----
|
2000-11-06 17:03:47 +01:00
|
|
|
| 0 | | zero |
|
|
|
|
| 1 | 4 | one | -1
|
|
|
|
| 2 | 3 | two | 2
|
|
|
|
| 2 | 3 | two | 4
|
|
|
|
| 3 | 2 | three | -3
|
|
|
|
| 4 | 1 | four |
|
|
|
|
| 5 | 0 | five | -5
|
|
|
|
| 5 | 0 | five | -5
|
|
|
|
| 6 | 6 | six |
|
|
|
|
| 7 | 7 | seven |
|
|
|
|
| 8 | 8 | eight |
|
|
|
|
| | | null |
|
|
|
|
| | 0 | zero |
|
|
|
|
(13 rows)
|
|
|
|
|
|
|
|
SELECT '' AS "xxx", *
|
2002-10-28 23:54:45 +01:00
|
|
|
FROM J1_TBL LEFT JOIN J2_TBL USING (i)
|
2004-12-03 23:19:28 +01:00
|
|
|
ORDER BY i, k, t;
|
2000-11-06 17:03:47 +01:00
|
|
|
xxx | i | j | t | k
|
|
|
|
-----+---+---+-------+----
|
|
|
|
| 0 | | zero |
|
|
|
|
| 1 | 4 | one | -1
|
|
|
|
| 2 | 3 | two | 2
|
|
|
|
| 2 | 3 | two | 4
|
|
|
|
| 3 | 2 | three | -3
|
|
|
|
| 4 | 1 | four |
|
|
|
|
| 5 | 0 | five | -5
|
|
|
|
| 5 | 0 | five | -5
|
|
|
|
| 6 | 6 | six |
|
|
|
|
| 7 | 7 | seven |
|
|
|
|
| 8 | 8 | eight |
|
|
|
|
| | | null |
|
|
|
|
| | 0 | zero |
|
|
|
|
(13 rows)
|
2000-09-12 23:07:18 +02:00
|
|
|
|
2000-01-09 04:48:39 +01:00
|
|
|
SELECT '' AS "xxx", *
|
2000-02-15 04:31:33 +01:00
|
|
|
FROM J1_TBL RIGHT OUTER JOIN J2_TBL USING (i);
|
2000-09-12 23:07:18 +02:00
|
|
|
xxx | i | j | t | k
|
|
|
|
-----+---+---+-------+----
|
2005-03-24 20:14:49 +01:00
|
|
|
| 0 | | zero |
|
2000-11-06 17:03:47 +01:00
|
|
|
| 1 | 4 | one | -1
|
|
|
|
| 2 | 3 | two | 2
|
2004-12-01 20:00:56 +01:00
|
|
|
| 2 | 3 | two | 4
|
2005-03-24 20:14:49 +01:00
|
|
|
| 3 | 2 | three | -3
|
2000-11-06 17:03:47 +01:00
|
|
|
| 5 | 0 | five | -5
|
|
|
|
| 5 | 0 | five | -5
|
|
|
|
| | | |
|
|
|
|
| | | | 0
|
|
|
|
(9 rows)
|
|
|
|
|
|
|
|
SELECT '' AS "xxx", *
|
|
|
|
FROM J1_TBL RIGHT JOIN J2_TBL USING (i);
|
|
|
|
xxx | i | j | t | k
|
|
|
|
-----+---+---+-------+----
|
2005-03-24 20:14:49 +01:00
|
|
|
| 0 | | zero |
|
2000-11-06 17:03:47 +01:00
|
|
|
| 1 | 4 | one | -1
|
|
|
|
| 2 | 3 | two | 2
|
2004-12-01 20:00:56 +01:00
|
|
|
| 2 | 3 | two | 4
|
2005-03-24 20:14:49 +01:00
|
|
|
| 3 | 2 | three | -3
|
2000-11-06 17:03:47 +01:00
|
|
|
| 5 | 0 | five | -5
|
|
|
|
| 5 | 0 | five | -5
|
|
|
|
| | | |
|
|
|
|
| | | | 0
|
|
|
|
(9 rows)
|
|
|
|
|
|
|
|
SELECT '' AS "xxx", *
|
2002-10-28 23:54:45 +01:00
|
|
|
FROM J1_TBL FULL OUTER JOIN J2_TBL USING (i)
|
2004-12-03 23:19:28 +01:00
|
|
|
ORDER BY i, k, t;
|
2000-11-06 17:03:47 +01:00
|
|
|
xxx | i | j | t | k
|
|
|
|
-----+---+---+-------+----
|
|
|
|
| 0 | | zero |
|
|
|
|
| 1 | 4 | one | -1
|
|
|
|
| 2 | 3 | two | 2
|
|
|
|
| 2 | 3 | two | 4
|
|
|
|
| 3 | 2 | three | -3
|
|
|
|
| 4 | 1 | four |
|
|
|
|
| 5 | 0 | five | -5
|
|
|
|
| 5 | 0 | five | -5
|
|
|
|
| 6 | 6 | six |
|
|
|
|
| 7 | 7 | seven |
|
|
|
|
| 8 | 8 | eight |
|
2004-12-02 02:34:18 +01:00
|
|
|
| | | | 0
|
2000-11-06 17:03:47 +01:00
|
|
|
| | | null |
|
|
|
|
| | 0 | zero |
|
2000-12-14 23:30:45 +01:00
|
|
|
| | | |
|
2000-11-06 17:03:47 +01:00
|
|
|
(15 rows)
|
2000-09-12 23:07:18 +02:00
|
|
|
|
2000-01-09 04:48:39 +01:00
|
|
|
SELECT '' AS "xxx", *
|
2002-10-28 23:54:45 +01:00
|
|
|
FROM J1_TBL FULL JOIN J2_TBL USING (i)
|
2004-12-03 23:19:28 +01:00
|
|
|
ORDER BY i, k, t;
|
2000-09-12 23:07:18 +02:00
|
|
|
xxx | i | j | t | k
|
|
|
|
-----+---+---+-------+----
|
2000-11-06 17:03:47 +01:00
|
|
|
| 0 | | zero |
|
|
|
|
| 1 | 4 | one | -1
|
|
|
|
| 2 | 3 | two | 2
|
|
|
|
| 2 | 3 | two | 4
|
|
|
|
| 3 | 2 | three | -3
|
|
|
|
| 4 | 1 | four |
|
|
|
|
| 5 | 0 | five | -5
|
|
|
|
| 5 | 0 | five | -5
|
|
|
|
| 6 | 6 | six |
|
|
|
|
| 7 | 7 | seven |
|
|
|
|
| 8 | 8 | eight |
|
2004-12-02 02:34:18 +01:00
|
|
|
| | | | 0
|
2000-11-06 17:03:47 +01:00
|
|
|
| | | null |
|
|
|
|
| | 0 | zero |
|
2000-12-14 23:30:45 +01:00
|
|
|
| | | |
|
2000-11-06 17:03:47 +01:00
|
|
|
(15 rows)
|
|
|
|
|
|
|
|
SELECT '' AS "xxx", *
|
|
|
|
FROM J1_TBL LEFT JOIN J2_TBL USING (i) WHERE (k = 1);
|
|
|
|
xxx | i | j | t | k
|
|
|
|
-----+---+---+---+---
|
|
|
|
(0 rows)
|
|
|
|
|
|
|
|
SELECT '' AS "xxx", *
|
|
|
|
FROM J1_TBL LEFT JOIN J2_TBL USING (i) WHERE (i = 1);
|
|
|
|
xxx | i | j | t | k
|
|
|
|
-----+---+---+-----+----
|
|
|
|
| 1 | 4 | one | -1
|
|
|
|
(1 row)
|
2000-09-12 23:07:18 +02:00
|
|
|
|
2017-11-30 04:00:29 +01:00
|
|
|
--
|
|
|
|
-- semijoin selectivity for <>
|
|
|
|
--
|
|
|
|
explain (costs off)
|
|
|
|
select * from int4_tbl i4, tenk1 a
|
|
|
|
where exists(select * from tenk1 b
|
|
|
|
where a.twothousand = b.twothousand and a.fivethous <> b.fivethous)
|
|
|
|
and i4.f1 = a.tenthous;
|
|
|
|
QUERY PLAN
|
|
|
|
----------------------------------------------
|
|
|
|
Hash Semi Join
|
|
|
|
Hash Cond: (a.twothousand = b.twothousand)
|
|
|
|
Join Filter: (a.fivethous <> b.fivethous)
|
|
|
|
-> Hash Join
|
|
|
|
Hash Cond: (a.tenthous = i4.f1)
|
|
|
|
-> Seq Scan on tenk1 a
|
|
|
|
-> Hash
|
|
|
|
-> Seq Scan on int4_tbl i4
|
|
|
|
-> Hash
|
|
|
|
-> Seq Scan on tenk1 b
|
|
|
|
(10 rows)
|
|
|
|
|
2000-01-09 04:48:39 +01:00
|
|
|
--
|
|
|
|
-- More complicated constructs
|
|
|
|
--
|
2002-03-12 01:52:10 +01:00
|
|
|
--
|
|
|
|
-- Multiway full join
|
|
|
|
--
|
|
|
|
CREATE TABLE t1 (name TEXT, n INTEGER);
|
|
|
|
CREATE TABLE t2 (name TEXT, n INTEGER);
|
|
|
|
CREATE TABLE t3 (name TEXT, n INTEGER);
|
2009-01-19 14:38:47 +01:00
|
|
|
INSERT INTO t1 VALUES ( 'bb', 11 );
|
|
|
|
INSERT INTO t2 VALUES ( 'bb', 12 );
|
|
|
|
INSERT INTO t2 VALUES ( 'cc', 22 );
|
|
|
|
INSERT INTO t2 VALUES ( 'ee', 42 );
|
|
|
|
INSERT INTO t3 VALUES ( 'bb', 13 );
|
|
|
|
INSERT INTO t3 VALUES ( 'cc', 23 );
|
|
|
|
INSERT INTO t3 VALUES ( 'dd', 33 );
|
2002-03-12 01:52:10 +01:00
|
|
|
SELECT * FROM t1 FULL JOIN t2 USING (name) FULL JOIN t3 USING (name);
|
|
|
|
name | n | n | n
|
|
|
|
------+----+----+----
|
2009-01-19 14:38:47 +01:00
|
|
|
bb | 11 | 12 | 13
|
|
|
|
cc | | 22 | 23
|
|
|
|
dd | | | 33
|
|
|
|
ee | | 42 |
|
2002-03-12 01:52:10 +01:00
|
|
|
(4 rows)
|
|
|
|
|
2002-04-28 21:54:29 +02:00
|
|
|
--
|
|
|
|
-- Test interactions of join syntax and subqueries
|
|
|
|
--
|
|
|
|
-- Basic cases (we expect planner to pull up the subquery here)
|
|
|
|
SELECT * FROM
|
|
|
|
(SELECT * FROM t2) as s2
|
|
|
|
INNER JOIN
|
|
|
|
(SELECT * FROM t3) s3
|
|
|
|
USING (name);
|
|
|
|
name | n | n
|
|
|
|
------+----+----
|
2009-01-19 14:38:47 +01:00
|
|
|
bb | 12 | 13
|
|
|
|
cc | 22 | 23
|
2002-04-28 21:54:29 +02:00
|
|
|
(2 rows)
|
|
|
|
|
|
|
|
SELECT * FROM
|
|
|
|
(SELECT * FROM t2) as s2
|
|
|
|
LEFT JOIN
|
|
|
|
(SELECT * FROM t3) s3
|
|
|
|
USING (name);
|
|
|
|
name | n | n
|
|
|
|
------+----+----
|
2009-01-19 14:38:47 +01:00
|
|
|
bb | 12 | 13
|
|
|
|
cc | 22 | 23
|
|
|
|
ee | 42 |
|
2002-04-28 21:54:29 +02:00
|
|
|
(3 rows)
|
|
|
|
|
|
|
|
SELECT * FROM
|
|
|
|
(SELECT * FROM t2) as s2
|
|
|
|
FULL JOIN
|
|
|
|
(SELECT * FROM t3) s3
|
|
|
|
USING (name);
|
|
|
|
name | n | n
|
|
|
|
------+----+----
|
2009-01-19 14:38:47 +01:00
|
|
|
bb | 12 | 13
|
|
|
|
cc | 22 | 23
|
|
|
|
dd | | 33
|
|
|
|
ee | 42 |
|
2002-04-28 21:54:29 +02:00
|
|
|
(4 rows)
|
|
|
|
|
|
|
|
-- Cases with non-nullable expressions in subquery results;
|
|
|
|
-- make sure these go to null as expected
|
|
|
|
SELECT * FROM
|
|
|
|
(SELECT name, n as s2_n, 2 as s2_2 FROM t2) as s2
|
|
|
|
NATURAL INNER JOIN
|
|
|
|
(SELECT name, n as s3_n, 3 as s3_2 FROM t3) s3;
|
|
|
|
name | s2_n | s2_2 | s3_n | s3_2
|
|
|
|
------+------+------+------+------
|
2009-01-19 14:38:47 +01:00
|
|
|
bb | 12 | 2 | 13 | 3
|
|
|
|
cc | 22 | 2 | 23 | 3
|
2002-04-28 21:54:29 +02:00
|
|
|
(2 rows)
|
|
|
|
|
|
|
|
SELECT * FROM
|
|
|
|
(SELECT name, n as s2_n, 2 as s2_2 FROM t2) as s2
|
|
|
|
NATURAL LEFT JOIN
|
|
|
|
(SELECT name, n as s3_n, 3 as s3_2 FROM t3) s3;
|
|
|
|
name | s2_n | s2_2 | s3_n | s3_2
|
|
|
|
------+------+------+------+------
|
2009-01-19 14:38:47 +01:00
|
|
|
bb | 12 | 2 | 13 | 3
|
|
|
|
cc | 22 | 2 | 23 | 3
|
|
|
|
ee | 42 | 2 | |
|
2002-04-28 21:54:29 +02:00
|
|
|
(3 rows)
|
|
|
|
|
|
|
|
SELECT * FROM
|
|
|
|
(SELECT name, n as s2_n, 2 as s2_2 FROM t2) as s2
|
|
|
|
NATURAL FULL JOIN
|
|
|
|
(SELECT name, n as s3_n, 3 as s3_2 FROM t3) s3;
|
|
|
|
name | s2_n | s2_2 | s3_n | s3_2
|
|
|
|
------+------+------+------+------
|
2009-01-19 14:38:47 +01:00
|
|
|
bb | 12 | 2 | 13 | 3
|
|
|
|
cc | 22 | 2 | 23 | 3
|
|
|
|
dd | | | 33 | 3
|
|
|
|
ee | 42 | 2 | |
|
2002-04-28 21:54:29 +02:00
|
|
|
(4 rows)
|
|
|
|
|
|
|
|
SELECT * FROM
|
|
|
|
(SELECT name, n as s1_n, 1 as s1_1 FROM t1) as s1
|
|
|
|
NATURAL INNER JOIN
|
|
|
|
(SELECT name, n as s2_n, 2 as s2_2 FROM t2) as s2
|
|
|
|
NATURAL INNER JOIN
|
|
|
|
(SELECT name, n as s3_n, 3 as s3_2 FROM t3) s3;
|
|
|
|
name | s1_n | s1_1 | s2_n | s2_2 | s3_n | s3_2
|
|
|
|
------+------+------+------+------+------+------
|
2009-01-19 14:38:47 +01:00
|
|
|
bb | 11 | 1 | 12 | 2 | 13 | 3
|
2002-04-28 21:54:29 +02:00
|
|
|
(1 row)
|
|
|
|
|
|
|
|
SELECT * FROM
|
|
|
|
(SELECT name, n as s1_n, 1 as s1_1 FROM t1) as s1
|
|
|
|
NATURAL FULL JOIN
|
|
|
|
(SELECT name, n as s2_n, 2 as s2_2 FROM t2) as s2
|
|
|
|
NATURAL FULL JOIN
|
|
|
|
(SELECT name, n as s3_n, 3 as s3_2 FROM t3) s3;
|
|
|
|
name | s1_n | s1_1 | s2_n | s2_2 | s3_n | s3_2
|
|
|
|
------+------+------+------+------+------+------
|
2009-01-19 14:38:47 +01:00
|
|
|
bb | 11 | 1 | 12 | 2 | 13 | 3
|
|
|
|
cc | | | 22 | 2 | 23 | 3
|
|
|
|
dd | | | | | 33 | 3
|
|
|
|
ee | | | 42 | 2 | |
|
2002-04-28 21:54:29 +02:00
|
|
|
(4 rows)
|
|
|
|
|
|
|
|
SELECT * FROM
|
|
|
|
(SELECT name, n as s1_n FROM t1) as s1
|
|
|
|
NATURAL FULL JOIN
|
|
|
|
(SELECT * FROM
|
|
|
|
(SELECT name, n as s2_n FROM t2) as s2
|
|
|
|
NATURAL FULL JOIN
|
|
|
|
(SELECT name, n as s3_n FROM t3) as s3
|
|
|
|
) ss2;
|
|
|
|
name | s1_n | s2_n | s3_n
|
|
|
|
------+------+------+------
|
2009-01-19 14:38:47 +01:00
|
|
|
bb | 11 | 12 | 13
|
|
|
|
cc | | 22 | 23
|
|
|
|
dd | | | 33
|
|
|
|
ee | | 42 |
|
2002-04-28 21:54:29 +02:00
|
|
|
(4 rows)
|
|
|
|
|
|
|
|
SELECT * FROM
|
|
|
|
(SELECT name, n as s1_n FROM t1) as s1
|
|
|
|
NATURAL FULL JOIN
|
|
|
|
(SELECT * FROM
|
|
|
|
(SELECT name, n as s2_n, 2 as s2_2 FROM t2) as s2
|
|
|
|
NATURAL FULL JOIN
|
|
|
|
(SELECT name, n as s3_n FROM t3) as s3
|
|
|
|
) ss2;
|
|
|
|
name | s1_n | s2_n | s2_2 | s3_n
|
|
|
|
------+------+------+------+------
|
2009-01-19 14:38:47 +01:00
|
|
|
bb | 11 | 12 | 2 | 13
|
|
|
|
cc | | 22 | 2 | 23
|
|
|
|
dd | | | | 33
|
|
|
|
ee | | 42 | 2 |
|
2002-04-28 21:54:29 +02:00
|
|
|
(4 rows)
|
|
|
|
|
2003-02-10 18:08:50 +01:00
|
|
|
-- Test for propagation of nullability constraints into sub-joins
|
|
|
|
create temp table x (x1 int, x2 int);
|
|
|
|
insert into x values (1,11);
|
|
|
|
insert into x values (2,22);
|
|
|
|
insert into x values (3,null);
|
|
|
|
insert into x values (4,44);
|
|
|
|
insert into x values (5,null);
|
|
|
|
create temp table y (y1 int, y2 int);
|
|
|
|
insert into y values (1,111);
|
|
|
|
insert into y values (2,222);
|
|
|
|
insert into y values (3,333);
|
|
|
|
insert into y values (4,null);
|
|
|
|
select * from x;
|
|
|
|
x1 | x2
|
|
|
|
----+----
|
|
|
|
1 | 11
|
|
|
|
2 | 22
|
|
|
|
3 |
|
|
|
|
4 | 44
|
|
|
|
5 |
|
|
|
|
(5 rows)
|
|
|
|
|
|
|
|
select * from y;
|
|
|
|
y1 | y2
|
|
|
|
----+-----
|
|
|
|
1 | 111
|
|
|
|
2 | 222
|
|
|
|
3 | 333
|
|
|
|
4 |
|
|
|
|
(4 rows)
|
|
|
|
|
|
|
|
select * from x left join y on (x1 = y1 and x2 is not null);
|
|
|
|
x1 | x2 | y1 | y2
|
|
|
|
----+----+----+-----
|
|
|
|
1 | 11 | 1 | 111
|
|
|
|
2 | 22 | 2 | 222
|
|
|
|
3 | | |
|
|
|
|
4 | 44 | 4 |
|
|
|
|
5 | | |
|
|
|
|
(5 rows)
|
|
|
|
|
|
|
|
select * from x left join y on (x1 = y1 and y2 is not null);
|
|
|
|
x1 | x2 | y1 | y2
|
|
|
|
----+----+----+-----
|
|
|
|
1 | 11 | 1 | 111
|
|
|
|
2 | 22 | 2 | 222
|
|
|
|
3 | | 3 | 333
|
|
|
|
4 | 44 | |
|
|
|
|
5 | | |
|
|
|
|
(5 rows)
|
|
|
|
|
|
|
|
select * from (x left join y on (x1 = y1)) left join x xx(xx1,xx2)
|
|
|
|
on (x1 = xx1);
|
|
|
|
x1 | x2 | y1 | y2 | xx1 | xx2
|
|
|
|
----+----+----+-----+-----+-----
|
|
|
|
1 | 11 | 1 | 111 | 1 | 11
|
|
|
|
2 | 22 | 2 | 222 | 2 | 22
|
|
|
|
3 | | 3 | 333 | 3 |
|
|
|
|
4 | 44 | 4 | | 4 | 44
|
|
|
|
5 | | | | 5 |
|
|
|
|
(5 rows)
|
|
|
|
|
|
|
|
select * from (x left join y on (x1 = y1)) left join x xx(xx1,xx2)
|
|
|
|
on (x1 = xx1 and x2 is not null);
|
|
|
|
x1 | x2 | y1 | y2 | xx1 | xx2
|
|
|
|
----+----+----+-----+-----+-----
|
|
|
|
1 | 11 | 1 | 111 | 1 | 11
|
|
|
|
2 | 22 | 2 | 222 | 2 | 22
|
|
|
|
3 | | 3 | 333 | |
|
|
|
|
4 | 44 | 4 | | 4 | 44
|
|
|
|
5 | | | | |
|
|
|
|
(5 rows)
|
|
|
|
|
|
|
|
select * from (x left join y on (x1 = y1)) left join x xx(xx1,xx2)
|
|
|
|
on (x1 = xx1 and y2 is not null);
|
|
|
|
x1 | x2 | y1 | y2 | xx1 | xx2
|
|
|
|
----+----+----+-----+-----+-----
|
|
|
|
1 | 11 | 1 | 111 | 1 | 11
|
|
|
|
2 | 22 | 2 | 222 | 2 | 22
|
|
|
|
3 | | 3 | 333 | 3 |
|
|
|
|
4 | 44 | 4 | | |
|
|
|
|
5 | | | | |
|
|
|
|
(5 rows)
|
|
|
|
|
|
|
|
select * from (x left join y on (x1 = y1)) left join x xx(xx1,xx2)
|
|
|
|
on (x1 = xx1 and xx2 is not null);
|
|
|
|
x1 | x2 | y1 | y2 | xx1 | xx2
|
|
|
|
----+----+----+-----+-----+-----
|
|
|
|
1 | 11 | 1 | 111 | 1 | 11
|
|
|
|
2 | 22 | 2 | 222 | 2 | 22
|
|
|
|
3 | | 3 | 333 | |
|
|
|
|
4 | 44 | 4 | | 4 | 44
|
|
|
|
5 | | | | |
|
|
|
|
(5 rows)
|
|
|
|
|
|
|
|
-- these should NOT give the same answers as above
|
|
|
|
select * from (x left join y on (x1 = y1)) left join x xx(xx1,xx2)
|
|
|
|
on (x1 = xx1) where (x2 is not null);
|
|
|
|
x1 | x2 | y1 | y2 | xx1 | xx2
|
|
|
|
----+----+----+-----+-----+-----
|
|
|
|
1 | 11 | 1 | 111 | 1 | 11
|
|
|
|
2 | 22 | 2 | 222 | 2 | 22
|
|
|
|
4 | 44 | 4 | | 4 | 44
|
|
|
|
(3 rows)
|
|
|
|
|
|
|
|
select * from (x left join y on (x1 = y1)) left join x xx(xx1,xx2)
|
|
|
|
on (x1 = xx1) where (y2 is not null);
|
|
|
|
x1 | x2 | y1 | y2 | xx1 | xx2
|
|
|
|
----+----+----+-----+-----+-----
|
|
|
|
1 | 11 | 1 | 111 | 1 | 11
|
|
|
|
2 | 22 | 2 | 222 | 2 | 22
|
|
|
|
3 | | 3 | 333 | 3 |
|
|
|
|
(3 rows)
|
|
|
|
|
|
|
|
select * from (x left join y on (x1 = y1)) left join x xx(xx1,xx2)
|
|
|
|
on (x1 = xx1) where (xx2 is not null);
|
|
|
|
x1 | x2 | y1 | y2 | xx1 | xx2
|
|
|
|
----+----+----+-----+-----+-----
|
|
|
|
1 | 11 | 1 | 111 | 1 | 11
|
|
|
|
2 | 22 | 2 | 222 | 2 | 22
|
|
|
|
4 | 44 | 4 | | 4 | 44
|
|
|
|
(3 rows)
|
|
|
|
|
2003-08-07 21:20:24 +02:00
|
|
|
--
|
|
|
|
-- regression test: check for bug with propagation of implied equality
|
|
|
|
-- to outside an IN
|
|
|
|
--
|
|
|
|
select count(*) from tenk1 a where unique1 in
|
|
|
|
(select unique1 from tenk1 b join tenk1 c using (unique1)
|
|
|
|
where b.unique2 = 42);
|
|
|
|
count
|
|
|
|
-------
|
|
|
|
1
|
|
|
|
(1 row)
|
|
|
|
|
Restructure code that is responsible for ensuring that clauseless joins are
considered when it is necessary to do so because of a join-order restriction
(that is, an outer-join or IN-subselect construct). The former coding was a
bit ad-hoc and inconsistent, and it missed some cases, as exposed by Mario
Weilguni's recent bug report. His specific problem was that an IN could be
turned into a "clauseless" join due to constant-propagation removing the IN's
joinclause, and if the IN's subselect involved more than one relation and
there was more than one such IN linking to the same upper relation, then the
only valid join orders involve "bushy" plans but we would fail to consider the
specific paths needed to get there. (See the example case added to the join
regression test.) On examining the code I wonder if there weren't some other
problem cases too; in particular it seems that GEQO was defending against a
different set of corner cases than the main planner was. There was also an
efficiency problem, in that when we did realize we needed a clauseless join
because of an IN, we'd consider clauseless joins against every other relation
whether this was sensible or not. It seems a better design is to use the
outer-join and in-clause lists as a backup heuristic, just as the rule of
joining only where there are joinclauses is a heuristic: we'll join two
relations if they have a usable joinclause *or* this might be necessary to
satisfy an outer-join or IN-clause join order restriction. I refactored the
code to have just one place considering this instead of three, and made sure
that it covered all the cases that any of them had been considering.
Backpatch as far as 8.1 (which has only the IN-clause form of the disease).
By rights 8.0 and 7.4 should have the bug too, but they accidentally fail
to fail, because the joininfo structure used in those releases preserves some
memory of there having once been a joinclause between the inner and outer
sides of an IN, and so it leads the code in the right direction anyway.
I'll be conservative and not touch them.
2007-02-16 01:14:01 +01:00
|
|
|
--
|
|
|
|
-- regression test: check for failure to generate a plan with multiple
|
|
|
|
-- degenerate IN clauses
|
|
|
|
--
|
|
|
|
select count(*) from tenk1 x where
|
|
|
|
x.unique1 in (select a.f1 from int4_tbl a,float8_tbl b where a.f1=b.f1) and
|
|
|
|
x.unique1 = 0 and
|
|
|
|
x.unique1 in (select aa.f1 from int4_tbl aa,float8_tbl bb where aa.f1=bb.f1);
|
|
|
|
count
|
|
|
|
-------
|
|
|
|
1
|
|
|
|
(1 row)
|
|
|
|
|
2009-07-19 22:32:48 +02:00
|
|
|
-- try that with GEQO too
|
|
|
|
begin;
|
|
|
|
set geqo = on;
|
|
|
|
set geqo_threshold = 2;
|
|
|
|
select count(*) from tenk1 x where
|
|
|
|
x.unique1 in (select a.f1 from int4_tbl a,float8_tbl b where a.f1=b.f1) and
|
|
|
|
x.unique1 = 0 and
|
|
|
|
x.unique1 in (select aa.f1 from int4_tbl aa,float8_tbl bb where aa.f1=bb.f1);
|
|
|
|
count
|
|
|
|
-------
|
|
|
|
1
|
|
|
|
(1 row)
|
|
|
|
|
|
|
|
rollback;
|
Make entirely-dummy appendrels get marked as such in set_append_rel_size.
The planner generally expects that the estimated rowcount of any relation
is at least one row, *unless* it has been proven empty by constraint
exclusion or similar mechanisms, which is marked by installing a dummy path
as the rel's cheapest path (cf. IS_DUMMY_REL). When I split up
allpaths.c's processing of base rels into separate set_base_rel_sizes and
set_base_rel_pathlists steps, the intention was that dummy rels would get
marked as such during the "set size" step; this is what justifies an Assert
in indxpath.c's get_loop_count that other relations should either be dummy
or have positive rowcount. Unfortunately I didn't get that quite right
for append relations: if all the child rels have been proven empty then
set_append_rel_size would come up with a rowcount of zero, which is
correct, but it didn't then do set_dummy_rel_pathlist. (We would have
ended up with the right state after set_append_rel_pathlist, but that's
too late, if we generate indexpaths for some other rel first.)
In addition to fixing the actual bug, I installed an Assert enforcing this
convention in set_rel_size; that then allows simplification of a couple
of now-redundant tests for zero rowcount in set_append_rel_size.
Also, to cover the possibility that third-party FDWs have been careless
about not returning a zero rowcount estimate, apply clamp_row_est to
whatever an FDW comes up with as the rows estimate.
Per report from Andreas Seltenreich. Back-patch to 9.2. Earlier branches
did not have the separation between set_base_rel_sizes and
set_base_rel_pathlists steps, so there was no intermediate state where an
appendrel would have had inconsistent rowcount and pathlist. It's possible
that adding the Assert to set_rel_size would be a good idea in older
branches too; but since they're not under development any more, it's likely
not worth the trouble.
2015-07-26 22:19:08 +02:00
|
|
|
--
|
|
|
|
-- regression test: be sure we cope with proven-dummy append rels
|
|
|
|
--
|
|
|
|
explain (costs off)
|
|
|
|
select aa, bb, unique1, unique1
|
|
|
|
from tenk1 right join b on aa = unique1
|
|
|
|
where bb < bb and bb is null;
|
|
|
|
QUERY PLAN
|
|
|
|
--------------------------
|
|
|
|
Result
|
|
|
|
One-Time Filter: false
|
|
|
|
(2 rows)
|
|
|
|
|
|
|
|
select aa, bb, unique1, unique1
|
|
|
|
from tenk1 right join b on aa = unique1
|
|
|
|
where bb < bb and bb is null;
|
|
|
|
aa | bb | unique1 | unique1
|
|
|
|
----+----+---------+---------
|
|
|
|
(0 rows)
|
|
|
|
|
2015-07-26 23:44:27 +02:00
|
|
|
--
|
|
|
|
-- regression test: check handling of empty-FROM subquery underneath outer join
|
|
|
|
--
|
|
|
|
explain (costs off)
|
|
|
|
select * from int8_tbl i1 left join (int8_tbl i2 join
|
|
|
|
(select 123 as x) ss on i2.q1 = x) on i1.q2 = i2.q2
|
|
|
|
order by 1, 2;
|
|
|
|
QUERY PLAN
|
|
|
|
-------------------------------------------------
|
|
|
|
Sort
|
|
|
|
Sort Key: i1.q1, i1.q2
|
|
|
|
-> Hash Left Join
|
|
|
|
Hash Cond: (i1.q2 = i2.q2)
|
|
|
|
-> Seq Scan on int8_tbl i1
|
|
|
|
-> Hash
|
|
|
|
-> Hash Join
|
|
|
|
Hash Cond: (i2.q1 = (123))
|
|
|
|
-> Seq Scan on int8_tbl i2
|
|
|
|
-> Hash
|
|
|
|
-> Result
|
|
|
|
(11 rows)
|
|
|
|
|
|
|
|
select * from int8_tbl i1 left join (int8_tbl i2 join
|
|
|
|
(select 123 as x) ss on i2.q1 = x) on i1.q2 = i2.q2
|
|
|
|
order by 1, 2;
|
|
|
|
q1 | q2 | q1 | q2 | x
|
|
|
|
------------------+-------------------+-----+------------------+-----
|
|
|
|
123 | 456 | 123 | 456 | 123
|
|
|
|
123 | 4567890123456789 | 123 | 4567890123456789 | 123
|
|
|
|
4567890123456789 | -4567890123456789 | | |
|
|
|
|
4567890123456789 | 123 | | |
|
|
|
|
4567890123456789 | 4567890123456789 | 123 | 4567890123456789 | 123
|
|
|
|
(5 rows)
|
|
|
|
|
2015-07-28 19:20:39 +02:00
|
|
|
--
|
|
|
|
-- regression test: check a case where join_clause_is_movable_into() gives
|
2015-08-13 19:25:01 +02:00
|
|
|
-- an imprecise result, causing an assertion failure
|
2015-07-28 19:20:39 +02:00
|
|
|
--
|
2015-08-13 19:25:01 +02:00
|
|
|
select count(*)
|
2015-07-28 19:20:39 +02:00
|
|
|
from
|
2015-08-13 19:25:01 +02:00
|
|
|
(select t3.tenthous as x1, coalesce(t1.stringu1, t2.stringu1) as x2
|
|
|
|
from tenk1 t1
|
|
|
|
left join tenk1 t2 on t1.unique1 = t2.unique1
|
|
|
|
join tenk1 t3 on t1.unique2 = t3.unique2) ss,
|
|
|
|
tenk1 t4,
|
|
|
|
tenk1 t5
|
|
|
|
where t4.thousand = t5.unique1 and ss.x1 = t4.tenthous and ss.x2 = t5.stringu1;
|
|
|
|
count
|
|
|
|
-------
|
|
|
|
1000
|
|
|
|
(1 row)
|
2015-07-28 19:20:39 +02:00
|
|
|
|
Fix mishandling of equivalence-class tests in parameterized plans.
Given a three-or-more-way equivalence class, such as X.Y = Y.Y = Z.Z,
it was possible for the planner to omit one of the quals needed to
enforce that all members of the equivalence class are actually equal.
This only happened in the case of a parameterized join node for two
of the relations, that is a plan tree like
Nested Loop
-> Scan X
-> Nested Loop
-> Scan Y
-> Scan Z
Filter: Z.Z = X.X
The eclass machinery normally expects to apply X.X = Y.Y when those
two relations are joined, but in this shape of plan tree they aren't
joined until the top node --- and, if the lower nested loop is marked
as parameterized by X, the top node will assume that the relevant eclass
condition(s) got pushed down into the lower node. On the other hand,
the scan of Z assumes that it's only responsible for constraining Z.Z
to match any one of the other eclass members. So one or another of
the required quals sometimes fell between the cracks, depending on
whether consideration of the eclass in get_joinrel_parampathinfo()
for the lower nested loop chanced to generate X.X = Y.Y or X.X = Z.Z
as the appropriate constraint there. If it generated the latter,
it'd erroneously suppose that the Z scan would take care of matters.
To fix, force X.X = Y.Y to be generated and applied at that join node
when this case occurs.
This is *extremely* hard to hit in practice, because various planner
behaviors conspire to mask the problem; starting with the fact that the
planner doesn't really like to generate a parameterized plan of the
above shape. (It might have been impossible to hit it before we
tweaked things to allow this plan shape for star-schema cases.) Many
thanks to Alexander Kirkouski for submitting a reproducible test case.
The bug can be demonstrated in all branches back to 9.2 where parameterized
paths were introduced, so back-patch that far.
2016-04-30 02:19:38 +02:00
|
|
|
--
|
|
|
|
-- regression test: check a case where we formerly missed including an EC
|
|
|
|
-- enforcement clause because it was expected to be handled at scan level
|
|
|
|
--
|
|
|
|
explain (costs off)
|
|
|
|
select a.f1, b.f1, t.thousand, t.tenthous from
|
|
|
|
tenk1 t,
|
|
|
|
(select sum(f1)+1 as f1 from int4_tbl i4a) a,
|
|
|
|
(select sum(f1) as f1 from int4_tbl i4b) b
|
|
|
|
where b.f1 = t.thousand and a.f1 = b.f1 and (a.f1+b.f1+999) = t.tenthous;
|
|
|
|
QUERY PLAN
|
|
|
|
-----------------------------------------------------------------------------------------------------------------------
|
|
|
|
Nested Loop
|
|
|
|
-> Aggregate
|
|
|
|
-> Seq Scan on int4_tbl i4b
|
|
|
|
-> Nested Loop
|
|
|
|
Join Filter: ((sum(i4b.f1)) = ((sum(i4a.f1) + 1)))
|
|
|
|
-> Aggregate
|
|
|
|
-> Seq Scan on int4_tbl i4a
|
|
|
|
-> Index Only Scan using tenk1_thous_tenthous on tenk1 t
|
|
|
|
Index Cond: ((thousand = (sum(i4b.f1))) AND (tenthous = ((((sum(i4a.f1) + 1)) + (sum(i4b.f1))) + 999)))
|
|
|
|
(9 rows)
|
|
|
|
|
|
|
|
select a.f1, b.f1, t.thousand, t.tenthous from
|
|
|
|
tenk1 t,
|
|
|
|
(select sum(f1)+1 as f1 from int4_tbl i4a) a,
|
|
|
|
(select sum(f1) as f1 from int4_tbl i4b) b
|
|
|
|
where b.f1 = t.thousand and a.f1 = b.f1 and (a.f1+b.f1+999) = t.tenthous;
|
|
|
|
f1 | f1 | thousand | tenthous
|
|
|
|
----+----+----------+----------
|
|
|
|
(0 rows)
|
|
|
|
|
Fix planner failures with overlapping mergejoin clauses in an outer join.
Given overlapping or partially redundant join clauses, for example
t1 JOIN t2 ON t1.a = t2.x AND t1.b = t2.x
the planner's EquivalenceClass machinery will ordinarily refactor the
clauses as "t1.a = t1.b AND t1.a = t2.x", so that join processing doesn't
see multiple references to the same EquivalenceClass in a list of join
equality clauses. However, if the join is outer, it's incorrect to derive
a restriction clause on the outer side from the join conditions, so the
clause refactoring does not happen and we end up with overlapping join
conditions. The code that attempted to deal with such cases had several
subtle bugs, which could result in "left and right pathkeys do not match in
mergejoin" or "outer pathkeys do not match mergeclauses" planner errors,
if the selected join plan type was a mergejoin. (It does not appear that
any actually incorrect plan could have been emitted.)
The core of the problem really was failure to recognize that the outer and
inner relations' pathkeys have different relationships to the mergeclause
list. A join's mergeclause list is constructed by reference to the outer
pathkeys, so it will always be ordered the same as the outer pathkeys, but
this cannot be presumed true for the inner pathkeys. If the inner sides of
the mergeclauses contain multiple references to the same EquivalenceClass
({t2.x} in the above example) then a simplistic rendering of the required
inner sort order is like "ORDER BY t2.x, t2.x", but the pathkey machinery
recognizes that the second sort column is redundant and throws it away.
The mergejoin planning code failed to account for that behavior properly.
One error was to try to generate cut-down versions of the mergeclause list
from cut-down versions of the inner pathkeys in the same way as the initial
construction of the mergeclause list from the outer pathkeys was done; this
could lead to choosing a mergeclause list that fails to match the outer
pathkeys. The other problem was that the pathkey cross-checking code in
create_mergejoin_plan treated the inner and outer pathkey lists
identically, whereas actually the expectations for them must be different.
That led to false "pathkeys do not match" failures in some cases, and in
principle could have led to failure to detect bogus plans in other cases,
though there is no indication that such bogus plans could be generated.
Reported by Alexander Kuzmenkov, who also reviewed this patch. This has
been broken for years (back to around 8.3 according to my testing), so
back-patch to all supported branches.
Discussion: https://postgr.es/m/5dad9160-4632-0e47-e120-8e2082000c01@postgrespro.ru
2018-02-23 19:47:33 +01:00
|
|
|
--
|
|
|
|
-- check a case where we formerly got confused by conflicting sort orders
|
|
|
|
-- in redundant merge join path keys
|
|
|
|
--
|
|
|
|
explain (costs off)
|
|
|
|
select * from
|
|
|
|
j1_tbl full join
|
|
|
|
(select * from j2_tbl order by j2_tbl.i desc, j2_tbl.k asc) j2_tbl
|
|
|
|
on j1_tbl.i = j2_tbl.i and j1_tbl.i = j2_tbl.k;
|
|
|
|
QUERY PLAN
|
|
|
|
-----------------------------------------------------------------
|
|
|
|
Merge Full Join
|
|
|
|
Merge Cond: ((j2_tbl.i = j1_tbl.i) AND (j2_tbl.k = j1_tbl.i))
|
|
|
|
-> Sort
|
|
|
|
Sort Key: j2_tbl.i DESC, j2_tbl.k
|
|
|
|
-> Seq Scan on j2_tbl
|
|
|
|
-> Sort
|
|
|
|
Sort Key: j1_tbl.i DESC
|
|
|
|
-> Seq Scan on j1_tbl
|
|
|
|
(8 rows)
|
|
|
|
|
|
|
|
select * from
|
|
|
|
j1_tbl full join
|
|
|
|
(select * from j2_tbl order by j2_tbl.i desc, j2_tbl.k asc) j2_tbl
|
|
|
|
on j1_tbl.i = j2_tbl.i and j1_tbl.i = j2_tbl.k;
|
|
|
|
i | j | t | i | k
|
|
|
|
---+---+-------+---+----
|
|
|
|
| | | | 0
|
|
|
|
| | | |
|
|
|
|
| 0 | zero | |
|
|
|
|
| | null | |
|
|
|
|
8 | 8 | eight | |
|
|
|
|
7 | 7 | seven | |
|
|
|
|
6 | 6 | six | |
|
|
|
|
| | | 5 | -5
|
|
|
|
| | | 5 | -5
|
|
|
|
5 | 0 | five | |
|
|
|
|
4 | 1 | four | |
|
|
|
|
| | | 3 | -3
|
|
|
|
3 | 2 | three | |
|
|
|
|
2 | 3 | two | 2 | 2
|
|
|
|
| | | 2 | 4
|
|
|
|
| | | 1 | -1
|
|
|
|
| | | 0 |
|
|
|
|
1 | 4 | one | |
|
|
|
|
0 | | zero | |
|
|
|
|
(19 rows)
|
|
|
|
|
|
|
|
--
|
|
|
|
-- a different check for handling of redundant sort keys in merge joins
|
|
|
|
--
|
|
|
|
explain (costs off)
|
|
|
|
select count(*) from
|
|
|
|
(select * from tenk1 x order by x.thousand, x.twothousand, x.fivethous) x
|
|
|
|
left join
|
|
|
|
(select * from tenk1 y order by y.unique2) y
|
|
|
|
on x.thousand = y.unique2 and x.twothousand = y.hundred and x.fivethous = y.unique2;
|
|
|
|
QUERY PLAN
|
|
|
|
----------------------------------------------------------------------------------
|
|
|
|
Aggregate
|
|
|
|
-> Merge Left Join
|
|
|
|
Merge Cond: (x.thousand = y.unique2)
|
|
|
|
Join Filter: ((x.twothousand = y.hundred) AND (x.fivethous = y.unique2))
|
|
|
|
-> Sort
|
|
|
|
Sort Key: x.thousand, x.twothousand, x.fivethous
|
|
|
|
-> Seq Scan on tenk1 x
|
|
|
|
-> Materialize
|
|
|
|
-> Index Scan using tenk1_unique2 on tenk1 y
|
|
|
|
(9 rows)
|
|
|
|
|
|
|
|
select count(*) from
|
|
|
|
(select * from tenk1 x order by x.thousand, x.twothousand, x.fivethous) x
|
|
|
|
left join
|
|
|
|
(select * from tenk1 y order by y.unique2) y
|
|
|
|
on x.thousand = y.unique2 and x.twothousand = y.hundred and x.fivethous = y.unique2;
|
|
|
|
count
|
|
|
|
-------
|
|
|
|
10000
|
|
|
|
(1 row)
|
|
|
|
|
2000-01-09 04:48:39 +01:00
|
|
|
--
|
|
|
|
-- Clean up
|
|
|
|
--
|
2002-03-12 01:52:10 +01:00
|
|
|
DROP TABLE t1;
|
|
|
|
DROP TABLE t2;
|
|
|
|
DROP TABLE t3;
|
2000-02-15 04:31:33 +01:00
|
|
|
DROP TABLE J1_TBL;
|
|
|
|
DROP TABLE J2_TBL;
|
2005-04-07 03:51:41 +02:00
|
|
|
-- Both DELETE and UPDATE allow the specification of additional tables
|
|
|
|
-- to "join" against to determine which rows should be modified.
|
|
|
|
CREATE TEMP TABLE t1 (a int, b int);
|
|
|
|
CREATE TEMP TABLE t2 (a int, b int);
|
|
|
|
CREATE TEMP TABLE t3 (x int, y int);
|
|
|
|
INSERT INTO t1 VALUES (5, 10);
|
|
|
|
INSERT INTO t1 VALUES (15, 20);
|
|
|
|
INSERT INTO t1 VALUES (100, 100);
|
|
|
|
INSERT INTO t1 VALUES (200, 1000);
|
|
|
|
INSERT INTO t2 VALUES (200, 2000);
|
|
|
|
INSERT INTO t3 VALUES (5, 20);
|
|
|
|
INSERT INTO t3 VALUES (6, 7);
|
|
|
|
INSERT INTO t3 VALUES (7, 8);
|
|
|
|
INSERT INTO t3 VALUES (500, 100);
|
|
|
|
DELETE FROM t3 USING t1 table1 WHERE t3.x = table1.a;
|
|
|
|
SELECT * FROM t3;
|
|
|
|
x | y
|
|
|
|
-----+-----
|
|
|
|
6 | 7
|
|
|
|
7 | 8
|
|
|
|
500 | 100
|
|
|
|
(3 rows)
|
|
|
|
|
|
|
|
DELETE FROM t3 USING t1 JOIN t2 USING (a) WHERE t3.x > t1.a;
|
|
|
|
SELECT * FROM t3;
|
|
|
|
x | y
|
|
|
|
---+---
|
|
|
|
6 | 7
|
|
|
|
7 | 8
|
|
|
|
(2 rows)
|
|
|
|
|
|
|
|
DELETE FROM t3 USING t3 t3_other WHERE t3.x = t3_other.x AND t3.y = t3_other.y;
|
|
|
|
SELECT * FROM t3;
|
|
|
|
x | y
|
|
|
|
---+---
|
|
|
|
(0 rows)
|
|
|
|
|
2009-08-13 19:14:38 +02:00
|
|
|
-- Test join against inheritance tree
|
|
|
|
create temp table t2a () inherits (t2);
|
|
|
|
insert into t2a values (200, 2001);
|
|
|
|
select * from t1 left join t2 on (t1.a = t2.a);
|
|
|
|
a | b | a | b
|
|
|
|
-----+------+-----+------
|
|
|
|
5 | 10 | |
|
|
|
|
15 | 20 | |
|
|
|
|
100 | 100 | |
|
|
|
|
200 | 1000 | 200 | 2000
|
|
|
|
200 | 1000 | 200 | 2001
|
|
|
|
(5 rows)
|
|
|
|
|
2015-03-11 15:44:04 +01:00
|
|
|
-- Test matching of column name with wrong alias
|
|
|
|
select t1.x from t1 join t3 on (t1.a = t3.x);
|
|
|
|
ERROR: column t1.x does not exist
|
|
|
|
LINE 1: select t1.x from t1 join t3 on (t1.a = t3.x);
|
|
|
|
^
|
2015-11-17 03:16:42 +01:00
|
|
|
HINT: Perhaps you meant to reference the column "t3.x".
|
2006-03-17 20:38:12 +01:00
|
|
|
--
|
|
|
|
-- regression test for 8.1 merge right join bug
|
|
|
|
--
|
|
|
|
CREATE TEMP TABLE tt1 ( tt1_id int4, joincol int4 );
|
|
|
|
INSERT INTO tt1 VALUES (1, 11);
|
|
|
|
INSERT INTO tt1 VALUES (2, NULL);
|
|
|
|
CREATE TEMP TABLE tt2 ( tt2_id int4, joincol int4 );
|
|
|
|
INSERT INTO tt2 VALUES (21, 11);
|
|
|
|
INSERT INTO tt2 VALUES (22, 11);
|
|
|
|
set enable_hashjoin to off;
|
|
|
|
set enable_nestloop to off;
|
|
|
|
-- these should give the same results
|
|
|
|
select tt1.*, tt2.* from tt1 left join tt2 on tt1.joincol = tt2.joincol;
|
|
|
|
tt1_id | joincol | tt2_id | joincol
|
|
|
|
--------+---------+--------+---------
|
|
|
|
1 | 11 | 21 | 11
|
|
|
|
1 | 11 | 22 | 11
|
|
|
|
2 | | |
|
|
|
|
(3 rows)
|
|
|
|
|
|
|
|
select tt1.*, tt2.* from tt2 right join tt1 on tt1.joincol = tt2.joincol;
|
|
|
|
tt1_id | joincol | tt2_id | joincol
|
|
|
|
--------+---------+--------+---------
|
|
|
|
1 | 11 | 21 | 11
|
|
|
|
1 | 11 | 22 | 11
|
|
|
|
2 | | |
|
|
|
|
(3 rows)
|
|
|
|
|
2007-05-23 01:23:58 +02:00
|
|
|
reset enable_hashjoin;
|
|
|
|
reset enable_nestloop;
|
|
|
|
--
|
2016-02-07 18:29:17 +01:00
|
|
|
-- regression test for bug #13908 (hash join with skew tuples & nbatch increase)
|
|
|
|
--
|
|
|
|
set work_mem to '64kB';
|
|
|
|
set enable_mergejoin to off;
|
|
|
|
explain (costs off)
|
|
|
|
select count(*) from tenk1 a, tenk1 b
|
|
|
|
where a.hundred = b.thousand and (b.fivethous % 10) < 10;
|
|
|
|
QUERY PLAN
|
|
|
|
------------------------------------------------------------
|
|
|
|
Aggregate
|
|
|
|
-> Hash Join
|
|
|
|
Hash Cond: (a.hundred = b.thousand)
|
|
|
|
-> Index Only Scan using tenk1_hundred on tenk1 a
|
|
|
|
-> Hash
|
|
|
|
-> Seq Scan on tenk1 b
|
|
|
|
Filter: ((fivethous % 10) < 10)
|
|
|
|
(7 rows)
|
|
|
|
|
|
|
|
select count(*) from tenk1 a, tenk1 b
|
|
|
|
where a.hundred = b.thousand and (b.fivethous % 10) < 10;
|
|
|
|
count
|
|
|
|
--------
|
|
|
|
100000
|
|
|
|
(1 row)
|
|
|
|
|
|
|
|
reset work_mem;
|
|
|
|
reset enable_mergejoin;
|
|
|
|
--
|
2007-05-23 01:23:58 +02:00
|
|
|
-- regression test for 8.2 bug with improper re-ordering of left joins
|
|
|
|
--
|
|
|
|
create temp table tt3(f1 int, f2 text);
|
|
|
|
insert into tt3 select x, repeat('xyzzy', 100) from generate_series(1,10000) x;
|
|
|
|
create index tt3i on tt3(f1);
|
|
|
|
analyze tt3;
|
|
|
|
create temp table tt4(f1 int);
|
|
|
|
insert into tt4 values (0),(1),(9999);
|
|
|
|
analyze tt4;
|
|
|
|
SELECT a.f1
|
|
|
|
FROM tt4 a
|
|
|
|
LEFT JOIN (
|
|
|
|
SELECT b.f1
|
|
|
|
FROM tt3 b LEFT JOIN tt3 c ON (b.f1 = c.f1)
|
|
|
|
WHERE c.f1 IS NULL
|
|
|
|
) AS d ON (a.f1 = d.f1)
|
|
|
|
WHERE d.f1 IS NULL;
|
|
|
|
f1
|
|
|
|
------
|
|
|
|
0
|
|
|
|
1
|
|
|
|
9999
|
|
|
|
(3 rows)
|
|
|
|
|
2015-04-25 22:44:27 +02:00
|
|
|
--
|
|
|
|
-- regression test for proper handling of outer joins within antijoins
|
|
|
|
--
|
|
|
|
create temp table tt4x(c1 int, c2 int, c3 int);
|
|
|
|
explain (costs off)
|
|
|
|
select * from tt4x t1
|
|
|
|
where not exists (
|
|
|
|
select 1 from tt4x t2
|
|
|
|
left join tt4x t3 on t2.c3 = t3.c1
|
|
|
|
left join ( select t5.c1 as c1
|
|
|
|
from tt4x t4 left join tt4x t5 on t4.c2 = t5.c1
|
|
|
|
) a1 on t3.c2 = a1.c1
|
|
|
|
where t1.c1 = t2.c2
|
|
|
|
);
|
|
|
|
QUERY PLAN
|
|
|
|
---------------------------------------------------------
|
|
|
|
Hash Anti Join
|
|
|
|
Hash Cond: (t1.c1 = t2.c2)
|
|
|
|
-> Seq Scan on tt4x t1
|
|
|
|
-> Hash
|
|
|
|
-> Merge Right Join
|
|
|
|
Merge Cond: (t5.c1 = t3.c2)
|
|
|
|
-> Merge Join
|
|
|
|
Merge Cond: (t4.c2 = t5.c1)
|
|
|
|
-> Sort
|
|
|
|
Sort Key: t4.c2
|
|
|
|
-> Seq Scan on tt4x t4
|
|
|
|
-> Sort
|
|
|
|
Sort Key: t5.c1
|
|
|
|
-> Seq Scan on tt4x t5
|
|
|
|
-> Sort
|
|
|
|
Sort Key: t3.c2
|
|
|
|
-> Merge Left Join
|
|
|
|
Merge Cond: (t2.c3 = t3.c1)
|
|
|
|
-> Sort
|
|
|
|
Sort Key: t2.c3
|
|
|
|
-> Seq Scan on tt4x t2
|
|
|
|
-> Sort
|
|
|
|
Sort Key: t3.c1
|
|
|
|
-> Seq Scan on tt4x t3
|
|
|
|
(24 rows)
|
|
|
|
|
2007-07-31 21:53:37 +02:00
|
|
|
--
|
|
|
|
-- regression test for problems of the sort depicted in bug #3494
|
|
|
|
--
|
|
|
|
create temp table tt5(f1 int, f2 int);
|
|
|
|
create temp table tt6(f1 int, f2 int);
|
|
|
|
insert into tt5 values(1, 10);
|
|
|
|
insert into tt5 values(1, 11);
|
|
|
|
insert into tt6 values(1, 9);
|
|
|
|
insert into tt6 values(1, 2);
|
|
|
|
insert into tt6 values(2, 9);
|
|
|
|
select * from tt5,tt6 where tt5.f1 = tt6.f1 and tt5.f1 = tt5.f2 - tt6.f2;
|
|
|
|
f1 | f2 | f1 | f2
|
|
|
|
----+----+----+----
|
|
|
|
1 | 10 | 1 | 9
|
|
|
|
(1 row)
|
|
|
|
|
2007-08-31 03:44:06 +02:00
|
|
|
--
|
|
|
|
-- regression test for problems of the sort depicted in bug #3588
|
|
|
|
--
|
|
|
|
create temp table xx (pkxx int);
|
|
|
|
create temp table yy (pkyy int, pkxx int);
|
|
|
|
insert into xx values (1);
|
|
|
|
insert into xx values (2);
|
|
|
|
insert into xx values (3);
|
|
|
|
insert into yy values (101, 1);
|
|
|
|
insert into yy values (201, 2);
|
|
|
|
insert into yy values (301, NULL);
|
|
|
|
select yy.pkyy as yy_pkyy, yy.pkxx as yy_pkxx, yya.pkyy as yya_pkyy,
|
|
|
|
xxa.pkxx as xxa_pkxx, xxb.pkxx as xxb_pkxx
|
|
|
|
from yy
|
|
|
|
left join (SELECT * FROM yy where pkyy = 101) as yya ON yy.pkyy = yya.pkyy
|
|
|
|
left join xx xxa on yya.pkxx = xxa.pkxx
|
|
|
|
left join xx xxb on coalesce (xxa.pkxx, 1) = xxb.pkxx;
|
|
|
|
yy_pkyy | yy_pkxx | yya_pkyy | xxa_pkxx | xxb_pkxx
|
|
|
|
---------+---------+----------+----------+----------
|
|
|
|
101 | 1 | 101 | 1 | 1
|
|
|
|
201 | 2 | | | 1
|
|
|
|
301 | | | | 1
|
|
|
|
(3 rows)
|
|
|
|
|
Fix some planner issues found while investigating Kevin Grittner's report
of poorer planning in 8.3 than 8.2:
1. After pushing a constant across an outer join --- ie, given
"a LEFT JOIN b ON (a.x = b.y) WHERE a.x = 42", we can deduce that b.y is
sort of equal to 42, in the sense that we needn't fetch any b rows where
it isn't 42 --- loop to see if any additional deductions can be made.
Previous releases did that by recursing, but I had mistakenly thought that
this was no longer necessary given the EquivalenceClass machinery.
2. Allow pushing constants across outer join conditions even if the
condition is outerjoin_delayed due to a lower outer join. This is safe
as long as the condition is strict and we re-test it at the upper join.
3. Keep the outer-join clause even if we successfully push a constant
across it. This is *necessary* in the outerjoin_delayed case, but
even in the simple case, it seems better to do this to ensure that the
join search order heuristics will consider the join as reasonable to
make. Mark such a clause as having selectivity 1.0, though, since it's
not going to eliminate very many rows after application of the constant
condition.
4. Tweak have_relevant_eclass_joinclause to report that two relations
are joinable when they have vars that are equated to the same constant.
We won't actually generate any joinclause from such an EquivalenceClass,
but again it seems that in such a case it's a good idea to consider
the join as worth costing out.
5. Fix a bug in select_mergejoin_clauses that was exposed by these
changes: we have to reject candidate mergejoin clauses if either side was
equated to a constant, because we can't construct a canonical pathkey list
for such a clause. This is an implementation restriction that might be
worth fixing someday, but it doesn't seem critical to get it done for 8.3.
2008-01-09 21:42:29 +01:00
|
|
|
--
|
|
|
|
-- regression test for improper pushing of constants across outer-join clauses
|
|
|
|
-- (as seen in early 8.2.x releases)
|
|
|
|
--
|
|
|
|
create temp table zt1 (f1 int primary key);
|
|
|
|
create temp table zt2 (f2 int primary key);
|
|
|
|
create temp table zt3 (f3 int primary key);
|
|
|
|
insert into zt1 values(53);
|
|
|
|
insert into zt2 values(53);
|
|
|
|
select * from
|
|
|
|
zt2 left join zt3 on (f2 = f3)
|
|
|
|
left join zt1 on (f3 = f1)
|
|
|
|
where f2 = 53;
|
|
|
|
f2 | f3 | f1
|
|
|
|
----+----+----
|
|
|
|
53 | |
|
|
|
|
(1 row)
|
|
|
|
|
|
|
|
create temp view zv1 as select *,'dummy'::text AS junk from zt1;
|
|
|
|
select * from
|
|
|
|
zt2 left join zt3 on (f2 = f3)
|
|
|
|
left join zv1 on (f3 = f1)
|
|
|
|
where f2 = 53;
|
|
|
|
f2 | f3 | f1 | junk
|
|
|
|
----+----+----+------
|
|
|
|
53 | | |
|
|
|
|
(1 row)
|
|
|
|
|
2008-06-27 22:54:37 +02:00
|
|
|
--
|
|
|
|
-- regression test for improper extraction of OR indexqual conditions
|
|
|
|
-- (as seen in early 8.3.x releases)
|
|
|
|
--
|
|
|
|
select a.unique2, a.ten, b.tenthous, b.unique2, b.hundred
|
|
|
|
from tenk1 a left join tenk1 b on a.unique2 = b.tenthous
|
|
|
|
where a.unique1 = 42 and
|
|
|
|
((b.unique2 is null and a.ten = 2) or b.hundred = 3);
|
|
|
|
unique2 | ten | tenthous | unique2 | hundred
|
|
|
|
---------+-----+----------+---------+---------
|
|
|
|
(0 rows)
|
|
|
|
|
2009-02-25 04:30:38 +01:00
|
|
|
--
|
|
|
|
-- test proper positioning of one-time quals in EXISTS (8.4devel bug)
|
|
|
|
--
|
|
|
|
prepare foo(bool) as
|
|
|
|
select count(*) from tenk1 a left join tenk1 b
|
|
|
|
on (a.unique2 = b.unique1 and exists
|
|
|
|
(select 1 from tenk1 c where c.thousand = b.unique2 and $1));
|
|
|
|
execute foo(true);
|
|
|
|
count
|
|
|
|
-------
|
|
|
|
10000
|
|
|
|
(1 row)
|
|
|
|
|
|
|
|
execute foo(false);
|
|
|
|
count
|
|
|
|
-------
|
|
|
|
10000
|
|
|
|
(1 row)
|
|
|
|
|
2009-07-18 01:19:34 +02:00
|
|
|
--
|
|
|
|
-- test for sane behavior with noncanonical merge clauses, per bug #4926
|
|
|
|
--
|
|
|
|
begin;
|
|
|
|
set enable_mergejoin = 1;
|
|
|
|
set enable_hashjoin = 0;
|
|
|
|
set enable_nestloop = 0;
|
|
|
|
create temp table a (i integer);
|
|
|
|
create temp table b (x integer, y integer);
|
|
|
|
select * from a left join b on i = x and i = y and x = i;
|
|
|
|
i | x | y
|
|
|
|
---+---+---
|
|
|
|
(0 rows)
|
|
|
|
|
|
|
|
rollback;
|
Fix subquery pullup to wrap a PlaceHolderVar around the entire RowExpr
that's generated for a whole-row Var referencing the subquery, when the
subquery is in the nullable side of an outer join. The previous coding
instead put PlaceHolderVars around the elements of the RowExpr. The effect
was that when the outer join made the subquery outputs go to null, the
whole-row Var produced ROW(NULL,NULL,...) rather than just NULL. There
are arguments afoot about whether those things ought to be semantically
indistinguishable, but for the moment they are not entirely so, and the
planner needs to take care that its machinations preserve the difference.
Per bug #5025.
Making this feasible required refactoring ResolveNew() to allow more caller
control over what is substituted for a Var. I chose to make ResolveNew()
a wrapper around a new general-purpose function replace_rte_variables().
I also fixed the ancient bogosity that ResolveNew might fail to set
a query's hasSubLinks field after inserting a SubLink in it. Although
all current callers make sure that happens anyway, we've had bugs of that
sort before, and it seemed like a good time to install a proper solution.
Back-patch to 8.4. The problem can be demonstrated clear back to 8.0,
but the fix would be too invasive in earlier branches; not to mention
that people may be depending on the subtly-incorrect behavior. The
8.4 series is new enough that fixing this probably won't cause complaints,
but it might in older branches. Also, 8.4 shows the incorrect behavior
in more cases than older branches do, because it is able to flatten
subqueries in more cases.
2009-09-02 19:52:24 +02:00
|
|
|
--
|
|
|
|
-- test NULL behavior of whole-row Vars, per bug #5025
|
|
|
|
--
|
|
|
|
select t1.q2, count(t2.*)
|
|
|
|
from int8_tbl t1 left join int8_tbl t2 on (t1.q2 = t2.q1)
|
|
|
|
group by t1.q2 order by 1;
|
|
|
|
q2 | count
|
|
|
|
-------------------+-------
|
|
|
|
-4567890123456789 | 0
|
|
|
|
123 | 2
|
|
|
|
456 | 0
|
|
|
|
4567890123456789 | 6
|
|
|
|
(4 rows)
|
|
|
|
|
|
|
|
select t1.q2, count(t2.*)
|
|
|
|
from int8_tbl t1 left join (select * from int8_tbl) t2 on (t1.q2 = t2.q1)
|
|
|
|
group by t1.q2 order by 1;
|
|
|
|
q2 | count
|
|
|
|
-------------------+-------
|
|
|
|
-4567890123456789 | 0
|
|
|
|
123 | 2
|
|
|
|
456 | 0
|
|
|
|
4567890123456789 | 6
|
|
|
|
(4 rows)
|
|
|
|
|
|
|
|
select t1.q2, count(t2.*)
|
|
|
|
from int8_tbl t1 left join (select * from int8_tbl offset 0) t2 on (t1.q2 = t2.q1)
|
|
|
|
group by t1.q2 order by 1;
|
|
|
|
q2 | count
|
|
|
|
-------------------+-------
|
|
|
|
-4567890123456789 | 0
|
|
|
|
123 | 2
|
|
|
|
456 | 0
|
|
|
|
4567890123456789 | 6
|
|
|
|
(4 rows)
|
|
|
|
|
|
|
|
select t1.q2, count(t2.*)
|
|
|
|
from int8_tbl t1 left join
|
|
|
|
(select q1, case when q2=1 then 1 else q2 end as q2 from int8_tbl) t2
|
|
|
|
on (t1.q2 = t2.q1)
|
|
|
|
group by t1.q2 order by 1;
|
|
|
|
q2 | count
|
|
|
|
-------------------+-------
|
|
|
|
-4567890123456789 | 0
|
|
|
|
123 | 2
|
|
|
|
456 | 0
|
|
|
|
4567890123456789 | 6
|
|
|
|
(4 rows)
|
|
|
|
|
2010-09-28 18:08:56 +02:00
|
|
|
--
|
|
|
|
-- test incorrect failure to NULL pulled-up subexpressions
|
|
|
|
--
|
|
|
|
begin;
|
|
|
|
create temp table a (
|
|
|
|
code char not null,
|
|
|
|
constraint a_pk primary key (code)
|
|
|
|
);
|
|
|
|
create temp table b (
|
|
|
|
a char not null,
|
|
|
|
num integer not null,
|
|
|
|
constraint b_pk primary key (a, num)
|
|
|
|
);
|
|
|
|
create temp table c (
|
|
|
|
name char not null,
|
|
|
|
a char,
|
|
|
|
constraint c_pk primary key (name)
|
|
|
|
);
|
|
|
|
insert into a (code) values ('p');
|
|
|
|
insert into a (code) values ('q');
|
|
|
|
insert into b (a, num) values ('p', 1);
|
|
|
|
insert into b (a, num) values ('p', 2);
|
|
|
|
insert into c (name, a) values ('A', 'p');
|
|
|
|
insert into c (name, a) values ('B', 'q');
|
|
|
|
insert into c (name, a) values ('C', null);
|
|
|
|
select c.name, ss.code, ss.b_cnt, ss.const
|
|
|
|
from c left join
|
|
|
|
(select a.code, coalesce(b_grp.cnt, 0) as b_cnt, -1 as const
|
|
|
|
from a left join
|
|
|
|
(select count(1) as cnt, b.a from b group by b.a) as b_grp
|
|
|
|
on a.code = b_grp.a
|
|
|
|
) as ss
|
|
|
|
on (c.a = ss.code)
|
|
|
|
order by c.name;
|
|
|
|
name | code | b_cnt | const
|
|
|
|
------+------+-------+-------
|
|
|
|
A | p | 2 | -1
|
|
|
|
B | q | 0 | -1
|
|
|
|
C | | |
|
|
|
|
(3 rows)
|
|
|
|
|
|
|
|
rollback;
|
2011-08-09 06:48:51 +02:00
|
|
|
--
|
|
|
|
-- test incorrect handling of placeholders that only appear in targetlists,
|
|
|
|
-- per bug #6154
|
|
|
|
--
|
|
|
|
SELECT * FROM
|
|
|
|
( SELECT 1 as key1 ) sub1
|
|
|
|
LEFT JOIN
|
|
|
|
( SELECT sub3.key3, sub4.value2, COALESCE(sub4.value2, 66) as value3 FROM
|
|
|
|
( SELECT 1 as key3 ) sub3
|
|
|
|
LEFT JOIN
|
|
|
|
( SELECT sub5.key5, COALESCE(sub6.value1, 1) as value2 FROM
|
|
|
|
( SELECT 1 as key5 ) sub5
|
|
|
|
LEFT JOIN
|
|
|
|
( SELECT 2 as key6, 42 as value1 ) sub6
|
|
|
|
ON sub5.key5 = sub6.key6
|
|
|
|
) sub4
|
|
|
|
ON sub4.key5 = sub3.key3
|
|
|
|
) sub2
|
|
|
|
ON sub1.key1 = sub2.key3;
|
|
|
|
key1 | key3 | value2 | value3
|
|
|
|
------+------+--------+--------
|
|
|
|
1 | 1 | 1 | 1
|
|
|
|
(1 row)
|
|
|
|
|
|
|
|
-- test the path using join aliases, too
|
|
|
|
SELECT * FROM
|
|
|
|
( SELECT 1 as key1 ) sub1
|
|
|
|
LEFT JOIN
|
|
|
|
( SELECT sub3.key3, value2, COALESCE(value2, 66) as value3 FROM
|
|
|
|
( SELECT 1 as key3 ) sub3
|
|
|
|
LEFT JOIN
|
|
|
|
( SELECT sub5.key5, COALESCE(sub6.value1, 1) as value2 FROM
|
|
|
|
( SELECT 1 as key5 ) sub5
|
|
|
|
LEFT JOIN
|
|
|
|
( SELECT 2 as key6, 42 as value1 ) sub6
|
|
|
|
ON sub5.key5 = sub6.key6
|
|
|
|
) sub4
|
|
|
|
ON sub4.key5 = sub3.key3
|
|
|
|
) sub2
|
|
|
|
ON sub1.key1 = sub2.key3;
|
|
|
|
key1 | key3 | value2 | value3
|
|
|
|
------+------+--------+--------
|
|
|
|
1 | 1 | 1 | 1
|
|
|
|
(1 row)
|
|
|
|
|
2011-11-03 05:50:58 +01:00
|
|
|
--
|
|
|
|
-- test case where a PlaceHolderVar is used as a nestloop parameter
|
|
|
|
--
|
|
|
|
EXPLAIN (COSTS OFF)
|
|
|
|
SELECT qq, unique1
|
|
|
|
FROM
|
|
|
|
( SELECT COALESCE(q1, 0) AS qq FROM int8_tbl a ) AS ss1
|
|
|
|
FULL OUTER JOIN
|
|
|
|
( SELECT COALESCE(q2, -1) AS qq FROM int8_tbl b ) AS ss2
|
|
|
|
USING (qq)
|
|
|
|
INNER JOIN tenk1 c ON qq = unique2;
|
2015-03-30 20:59:49 +02:00
|
|
|
QUERY PLAN
|
|
|
|
---------------------------------------------------------------------------------------------------------
|
2011-11-03 05:50:58 +01:00
|
|
|
Nested Loop
|
|
|
|
-> Hash Full Join
|
2015-03-30 20:59:49 +02:00
|
|
|
Hash Cond: (COALESCE(a.q1, '0'::bigint) = COALESCE(b.q2, '-1'::bigint))
|
2011-11-03 05:50:58 +01:00
|
|
|
-> Seq Scan on int8_tbl a
|
|
|
|
-> Hash
|
|
|
|
-> Seq Scan on int8_tbl b
|
|
|
|
-> Index Scan using tenk1_unique2 on tenk1 c
|
2015-03-30 20:59:49 +02:00
|
|
|
Index Cond: (unique2 = COALESCE((COALESCE(a.q1, '0'::bigint)), (COALESCE(b.q2, '-1'::bigint))))
|
2011-11-03 05:50:58 +01:00
|
|
|
(8 rows)
|
|
|
|
|
|
|
|
SELECT qq, unique1
|
|
|
|
FROM
|
|
|
|
( SELECT COALESCE(q1, 0) AS qq FROM int8_tbl a ) AS ss1
|
|
|
|
FULL OUTER JOIN
|
|
|
|
( SELECT COALESCE(q2, -1) AS qq FROM int8_tbl b ) AS ss2
|
|
|
|
USING (qq)
|
|
|
|
INNER JOIN tenk1 c ON qq = unique2;
|
|
|
|
qq | unique1
|
|
|
|
-----+---------
|
|
|
|
123 | 4596
|
|
|
|
123 | 4596
|
|
|
|
456 | 7318
|
|
|
|
(3 rows)
|
|
|
|
|
2014-06-10 03:37:18 +02:00
|
|
|
--
|
|
|
|
-- nested nestloops can require nested PlaceHolderVars
|
|
|
|
--
|
|
|
|
create temp table nt1 (
|
|
|
|
id int primary key,
|
|
|
|
a1 boolean,
|
|
|
|
a2 boolean
|
|
|
|
);
|
|
|
|
create temp table nt2 (
|
|
|
|
id int primary key,
|
|
|
|
nt1_id int,
|
|
|
|
b1 boolean,
|
|
|
|
b2 boolean,
|
|
|
|
foreign key (nt1_id) references nt1(id)
|
|
|
|
);
|
|
|
|
create temp table nt3 (
|
|
|
|
id int primary key,
|
|
|
|
nt2_id int,
|
|
|
|
c1 boolean,
|
|
|
|
foreign key (nt2_id) references nt2(id)
|
|
|
|
);
|
|
|
|
insert into nt1 values (1,true,true);
|
|
|
|
insert into nt1 values (2,true,false);
|
|
|
|
insert into nt1 values (3,false,false);
|
|
|
|
insert into nt2 values (1,1,true,true);
|
|
|
|
insert into nt2 values (2,2,true,false);
|
|
|
|
insert into nt2 values (3,3,false,false);
|
|
|
|
insert into nt3 values (1,1,true);
|
|
|
|
insert into nt3 values (2,2,false);
|
|
|
|
insert into nt3 values (3,3,true);
|
|
|
|
explain (costs off)
|
|
|
|
select nt3.id
|
|
|
|
from nt3 as nt3
|
|
|
|
left join
|
|
|
|
(select nt2.*, (nt2.b1 and ss1.a3) AS b3
|
|
|
|
from nt2 as nt2
|
|
|
|
left join
|
|
|
|
(select nt1.*, (nt1.id is not null) as a3 from nt1) as ss1
|
|
|
|
on ss1.id = nt2.nt1_id
|
|
|
|
) as ss2
|
|
|
|
on ss2.id = nt3.nt2_id
|
|
|
|
where nt3.id = 1 and ss2.b3;
|
|
|
|
QUERY PLAN
|
|
|
|
-----------------------------------------------
|
|
|
|
Nested Loop
|
|
|
|
-> Nested Loop
|
|
|
|
-> Index Scan using nt3_pkey on nt3
|
|
|
|
Index Cond: (id = 1)
|
|
|
|
-> Index Scan using nt2_pkey on nt2
|
|
|
|
Index Cond: (id = nt3.nt2_id)
|
|
|
|
-> Index Only Scan using nt1_pkey on nt1
|
|
|
|
Index Cond: (id = nt2.nt1_id)
|
|
|
|
Filter: (nt2.b1 AND (id IS NOT NULL))
|
|
|
|
(9 rows)
|
|
|
|
|
|
|
|
select nt3.id
|
|
|
|
from nt3 as nt3
|
|
|
|
left join
|
|
|
|
(select nt2.*, (nt2.b1 and ss1.a3) AS b3
|
|
|
|
from nt2 as nt2
|
|
|
|
left join
|
|
|
|
(select nt1.*, (nt1.id is not null) as a3 from nt1) as ss1
|
|
|
|
on ss1.id = nt2.nt1_id
|
|
|
|
) as ss2
|
|
|
|
on ss2.id = nt3.nt2_id
|
|
|
|
where nt3.id = 1 and ss2.b3;
|
|
|
|
id
|
|
|
|
----
|
|
|
|
1
|
|
|
|
(1 row)
|
|
|
|
|
2012-03-24 21:21:39 +01:00
|
|
|
--
|
|
|
|
-- test case where a PlaceHolderVar is propagated into a subquery
|
|
|
|
--
|
|
|
|
explain (costs off)
|
|
|
|
select * from
|
|
|
|
int8_tbl t1 left join
|
|
|
|
(select q1 as x, 42 as y from int8_tbl t2) ss
|
|
|
|
on t1.q2 = ss.x
|
|
|
|
where
|
|
|
|
1 = (select 1 from int8_tbl t3 where ss.y is not null limit 1)
|
|
|
|
order by 1,2;
|
|
|
|
QUERY PLAN
|
|
|
|
-----------------------------------------------------------
|
|
|
|
Sort
|
|
|
|
Sort Key: t1.q1, t1.q2
|
|
|
|
-> Hash Left Join
|
|
|
|
Hash Cond: (t1.q2 = t2.q1)
|
|
|
|
Filter: (1 = (SubPlan 1))
|
|
|
|
-> Seq Scan on int8_tbl t1
|
|
|
|
-> Hash
|
|
|
|
-> Seq Scan on int8_tbl t2
|
|
|
|
SubPlan 1
|
|
|
|
-> Limit
|
|
|
|
-> Result
|
|
|
|
One-Time Filter: ((42) IS NOT NULL)
|
|
|
|
-> Seq Scan on int8_tbl t3
|
|
|
|
(13 rows)
|
|
|
|
|
|
|
|
select * from
|
|
|
|
int8_tbl t1 left join
|
|
|
|
(select q1 as x, 42 as y from int8_tbl t2) ss
|
|
|
|
on t1.q2 = ss.x
|
|
|
|
where
|
|
|
|
1 = (select 1 from int8_tbl t3 where ss.y is not null limit 1)
|
|
|
|
order by 1,2;
|
|
|
|
q1 | q2 | x | y
|
|
|
|
------------------+------------------+------------------+----
|
|
|
|
123 | 4567890123456789 | 4567890123456789 | 42
|
|
|
|
123 | 4567890123456789 | 4567890123456789 | 42
|
|
|
|
123 | 4567890123456789 | 4567890123456789 | 42
|
|
|
|
4567890123456789 | 123 | 123 | 42
|
|
|
|
4567890123456789 | 123 | 123 | 42
|
|
|
|
4567890123456789 | 4567890123456789 | 4567890123456789 | 42
|
|
|
|
4567890123456789 | 4567890123456789 | 4567890123456789 | 42
|
|
|
|
4567890123456789 | 4567890123456789 | 4567890123456789 | 42
|
|
|
|
(8 rows)
|
|
|
|
|
2010-01-06 00:25:36 +01:00
|
|
|
--
|
|
|
|
-- test the corner cases FULL JOIN ON TRUE and FULL JOIN ON FALSE
|
|
|
|
--
|
|
|
|
select * from int4_tbl a full join int4_tbl b on true;
|
|
|
|
f1 | f1
|
|
|
|
-------------+-------------
|
|
|
|
0 | 0
|
|
|
|
0 | 123456
|
|
|
|
0 | -123456
|
|
|
|
0 | 2147483647
|
|
|
|
0 | -2147483647
|
|
|
|
123456 | 0
|
|
|
|
123456 | 123456
|
|
|
|
123456 | -123456
|
|
|
|
123456 | 2147483647
|
|
|
|
123456 | -2147483647
|
|
|
|
-123456 | 0
|
|
|
|
-123456 | 123456
|
|
|
|
-123456 | -123456
|
|
|
|
-123456 | 2147483647
|
|
|
|
-123456 | -2147483647
|
|
|
|
2147483647 | 0
|
|
|
|
2147483647 | 123456
|
|
|
|
2147483647 | -123456
|
|
|
|
2147483647 | 2147483647
|
|
|
|
2147483647 | -2147483647
|
|
|
|
-2147483647 | 0
|
|
|
|
-2147483647 | 123456
|
|
|
|
-2147483647 | -123456
|
|
|
|
-2147483647 | 2147483647
|
|
|
|
-2147483647 | -2147483647
|
|
|
|
(25 rows)
|
|
|
|
|
|
|
|
select * from int4_tbl a full join int4_tbl b on false;
|
|
|
|
f1 | f1
|
|
|
|
-------------+-------------
|
|
|
|
| 0
|
|
|
|
| 123456
|
|
|
|
| -123456
|
|
|
|
| 2147483647
|
|
|
|
| -2147483647
|
|
|
|
0 |
|
|
|
|
123456 |
|
|
|
|
-123456 |
|
|
|
|
2147483647 |
|
|
|
|
-2147483647 |
|
|
|
|
(10 rows)
|
|
|
|
|
2012-04-13 21:32:34 +02:00
|
|
|
--
|
|
|
|
-- test for ability to use a cartesian join when necessary
|
|
|
|
--
|
|
|
|
explain (costs off)
|
|
|
|
select * from
|
|
|
|
tenk1 join int4_tbl on f1 = twothousand,
|
|
|
|
int4(sin(1)) q1,
|
|
|
|
int4(sin(0)) q2
|
|
|
|
where q1 = thousand or q2 = thousand;
|
Revise parameterized-path mechanism to fix assorted issues.
This patch adjusts the treatment of parameterized paths so that all paths
with the same parameterization (same set of required outer rels) for the
same relation will have the same rowcount estimate. We cache the rowcount
estimates to ensure that property, and hopefully save a few cycles too.
Doing this makes it practical for add_path_precheck to operate without
a rowcount estimate: it need only assume that paths with different
parameterizations never dominate each other, which is close enough to
true anyway for coarse filtering, because normally a more-parameterized
path should yield fewer rows thanks to having more join clauses to apply.
In add_path, we do the full nine yards of comparing rowcount estimates
along with everything else, so that we can discard parameterized paths that
don't actually have an advantage. This fixes some issues I'd found with
add_path rejecting parameterized paths on the grounds that they were more
expensive than not-parameterized ones, even though they yielded many fewer
rows and hence would be cheaper once subsequent joining was considered.
To make the same-rowcounts assumption valid, we have to require that any
parameterized path enforce *all* join clauses that could be obtained from
the particular set of outer rels, even if not all of them are useful for
indexing. This is required at both base scans and joins. It's a good
thing anyway since the net impact is that join quals are checked at the
lowest practical level in the join tree. Hence, discard the original
rather ad-hoc mechanism for choosing parameterization joinquals, and build
a better one that has a more principled rule for when clauses can be moved.
The original rule was actually buggy anyway for lack of knowledge about
which relations are part of an outer join's outer side; getting this right
requires adding an outer_relids field to RestrictInfo.
2012-04-19 21:52:46 +02:00
|
|
|
QUERY PLAN
|
|
|
|
------------------------------------------------------------------------
|
2012-04-13 21:32:34 +02:00
|
|
|
Hash Join
|
|
|
|
Hash Cond: (tenk1.twothousand = int4_tbl.f1)
|
|
|
|
-> Nested Loop
|
|
|
|
-> Nested Loop
|
|
|
|
-> Function Scan on q1
|
|
|
|
-> Function Scan on q2
|
|
|
|
-> Bitmap Heap Scan on tenk1
|
|
|
|
Recheck Cond: ((q1.q1 = thousand) OR (q2.q2 = thousand))
|
|
|
|
-> BitmapOr
|
|
|
|
-> Bitmap Index Scan on tenk1_thous_tenthous
|
|
|
|
Index Cond: (q1.q1 = thousand)
|
|
|
|
-> Bitmap Index Scan on tenk1_thous_tenthous
|
|
|
|
Index Cond: (q2.q2 = thousand)
|
|
|
|
-> Hash
|
|
|
|
-> Seq Scan on int4_tbl
|
Revise parameterized-path mechanism to fix assorted issues.
This patch adjusts the treatment of parameterized paths so that all paths
with the same parameterization (same set of required outer rels) for the
same relation will have the same rowcount estimate. We cache the rowcount
estimates to ensure that property, and hopefully save a few cycles too.
Doing this makes it practical for add_path_precheck to operate without
a rowcount estimate: it need only assume that paths with different
parameterizations never dominate each other, which is close enough to
true anyway for coarse filtering, because normally a more-parameterized
path should yield fewer rows thanks to having more join clauses to apply.
In add_path, we do the full nine yards of comparing rowcount estimates
along with everything else, so that we can discard parameterized paths that
don't actually have an advantage. This fixes some issues I'd found with
add_path rejecting parameterized paths on the grounds that they were more
expensive than not-parameterized ones, even though they yielded many fewer
rows and hence would be cheaper once subsequent joining was considered.
To make the same-rowcounts assumption valid, we have to require that any
parameterized path enforce *all* join clauses that could be obtained from
the particular set of outer rels, even if not all of them are useful for
indexing. This is required at both base scans and joins. It's a good
thing anyway since the net impact is that join quals are checked at the
lowest practical level in the join tree. Hence, discard the original
rather ad-hoc mechanism for choosing parameterization joinquals, and build
a better one that has a more principled rule for when clauses can be moved.
The original rule was actually buggy anyway for lack of knowledge about
which relations are part of an outer join's outer side; getting this right
requires adding an outer_relids field to RestrictInfo.
2012-04-19 21:52:46 +02:00
|
|
|
(15 rows)
|
2012-04-13 21:32:34 +02:00
|
|
|
|
|
|
|
explain (costs off)
|
|
|
|
select * from
|
|
|
|
tenk1 join int4_tbl on f1 = twothousand,
|
|
|
|
int4(sin(1)) q1,
|
|
|
|
int4(sin(0)) q2
|
|
|
|
where thousand = (q1 + q2);
|
|
|
|
QUERY PLAN
|
|
|
|
--------------------------------------------------------------
|
|
|
|
Hash Join
|
|
|
|
Hash Cond: (tenk1.twothousand = int4_tbl.f1)
|
|
|
|
-> Nested Loop
|
|
|
|
-> Nested Loop
|
|
|
|
-> Function Scan on q1
|
|
|
|
-> Function Scan on q2
|
|
|
|
-> Bitmap Heap Scan on tenk1
|
|
|
|
Recheck Cond: (thousand = (q1.q1 + q2.q2))
|
|
|
|
-> Bitmap Index Scan on tenk1_thous_tenthous
|
|
|
|
Index Cond: (thousand = (q1.q1 + q2.q2))
|
|
|
|
-> Hash
|
|
|
|
-> Seq Scan on int4_tbl
|
|
|
|
(12 rows)
|
|
|
|
|
2015-02-28 18:43:04 +01:00
|
|
|
--
|
|
|
|
-- test ability to generate a suitable plan for a star-schema query
|
|
|
|
--
|
|
|
|
explain (costs off)
|
|
|
|
select * from
|
|
|
|
tenk1, int8_tbl a, int8_tbl b
|
|
|
|
where thousand = a.q1 and tenthous = b.q1 and a.q2 = 1 and b.q2 = 2;
|
|
|
|
QUERY PLAN
|
|
|
|
---------------------------------------------------------------------
|
|
|
|
Nested Loop
|
|
|
|
-> Seq Scan on int8_tbl b
|
|
|
|
Filter: (q2 = 2)
|
|
|
|
-> Nested Loop
|
|
|
|
-> Seq Scan on int8_tbl a
|
|
|
|
Filter: (q2 = 1)
|
|
|
|
-> Index Scan using tenk1_thous_tenthous on tenk1
|
|
|
|
Index Cond: ((thousand = a.q1) AND (tenthous = b.q1))
|
|
|
|
(8 rows)
|
|
|
|
|
2015-08-04 20:55:32 +02:00
|
|
|
--
|
|
|
|
-- test a corner case in which we shouldn't apply the star-schema optimization
|
|
|
|
--
|
|
|
|
explain (costs off)
|
|
|
|
select t1.unique2, t1.stringu1, t2.unique1, t2.stringu2 from
|
|
|
|
tenk1 t1
|
|
|
|
inner join int4_tbl i1
|
|
|
|
left join (select v1.x2, v2.y1, 11 AS d1
|
|
|
|
from (values(1,0)) v1(x1,x2)
|
|
|
|
left join (values(3,1)) v2(y1,y2)
|
|
|
|
on v1.x1 = v2.y2) subq1
|
|
|
|
on (i1.f1 = subq1.x2)
|
|
|
|
on (t1.unique2 = subq1.d1)
|
|
|
|
left join tenk1 t2
|
|
|
|
on (subq1.y1 = t2.unique1)
|
|
|
|
where t1.unique2 < 42 and t1.stringu1 > t2.stringu2;
|
|
|
|
QUERY PLAN
|
|
|
|
-----------------------------------------------------------------------
|
|
|
|
Nested Loop
|
|
|
|
Join Filter: (t1.stringu1 > t2.stringu2)
|
|
|
|
-> Nested Loop
|
|
|
|
Join Filter: ((0) = i1.f1)
|
|
|
|
-> Nested Loop
|
|
|
|
-> Nested Loop
|
|
|
|
Join Filter: ((1) = (1))
|
|
|
|
-> Result
|
|
|
|
-> Result
|
|
|
|
-> Index Scan using tenk1_unique2 on tenk1 t1
|
|
|
|
Index Cond: ((unique2 = (11)) AND (unique2 < 42))
|
|
|
|
-> Seq Scan on int4_tbl i1
|
|
|
|
-> Index Scan using tenk1_unique1 on tenk1 t2
|
|
|
|
Index Cond: (unique1 = (3))
|
|
|
|
(14 rows)
|
|
|
|
|
|
|
|
select t1.unique2, t1.stringu1, t2.unique1, t2.stringu2 from
|
|
|
|
tenk1 t1
|
|
|
|
inner join int4_tbl i1
|
|
|
|
left join (select v1.x2, v2.y1, 11 AS d1
|
|
|
|
from (values(1,0)) v1(x1,x2)
|
|
|
|
left join (values(3,1)) v2(y1,y2)
|
|
|
|
on v1.x1 = v2.y2) subq1
|
|
|
|
on (i1.f1 = subq1.x2)
|
|
|
|
on (t1.unique2 = subq1.d1)
|
|
|
|
left join tenk1 t2
|
|
|
|
on (subq1.y1 = t2.unique1)
|
|
|
|
where t1.unique2 < 42 and t1.stringu1 > t2.stringu2;
|
|
|
|
unique2 | stringu1 | unique1 | stringu2
|
|
|
|
---------+----------+---------+----------
|
|
|
|
11 | WFAAAA | 3 | LKIAAA
|
|
|
|
(1 row)
|
|
|
|
|
2015-08-10 23:18:17 +02:00
|
|
|
-- variant that isn't quite a star-schema case
|
|
|
|
select ss1.d1 from
|
|
|
|
tenk1 as t1
|
|
|
|
inner join tenk1 as t2
|
|
|
|
on t1.tenthous = t2.ten
|
|
|
|
inner join
|
|
|
|
int8_tbl as i8
|
|
|
|
left join int4_tbl as i4
|
|
|
|
inner join (select 64::information_schema.cardinal_number as d1
|
|
|
|
from tenk1 t3,
|
|
|
|
lateral (select abs(t3.unique1) + random()) ss0(x)
|
|
|
|
where t3.fivethous < 0) as ss1
|
|
|
|
on i4.f1 = ss1.d1
|
|
|
|
on i8.q1 = i4.f1
|
|
|
|
on t1.tenthous = ss1.d1
|
|
|
|
where t1.unique1 < i4.f1;
|
|
|
|
d1
|
|
|
|
----
|
|
|
|
(0 rows)
|
|
|
|
|
Extract restriction OR clauses whether or not they are indexable.
It's possible to extract a restriction OR clause from a join clause that
has the form of an OR-of-ANDs, if each sub-AND includes a clause that
mentions only one specific relation. While PG has been aware of that idea
for many years, the code previously only did it if it could extract an
indexable OR clause. On reflection, though, that seems a silly limitation:
adding a restriction clause can be a win by reducing the number of rows
that have to be filtered at the join step, even if we have to test the
clause as a plain filter clause during the scan. This should be especially
useful for foreign tables, where the change can cut the number of rows that
have to be retrieved from the foreign server; but testing shows it can win
even on local tables. Per a suggestion from Robert Haas.
As a heuristic, I made the code accept an extracted restriction clause
if its estimated selectivity is less than 0.9, which will probably result
in accepting extracted clauses just about always. We might need to tweak
that later based on experience.
Since the code no longer has even a weak connection to Path creation,
remove orindxpath.c and create a new file optimizer/util/orclauses.c.
There's some additional janitorial cleanup of now-dead code that needs
to happen, but it seems like that's a fit subject for a separate commit.
2013-12-30 18:24:37 +01:00
|
|
|
--
|
|
|
|
-- test extraction of restriction OR clauses from join OR clause
|
|
|
|
-- (we used to only do this for indexable clauses)
|
|
|
|
--
|
|
|
|
explain (costs off)
|
|
|
|
select * from tenk1 a join tenk1 b on
|
|
|
|
(a.unique1 = 1 and b.unique1 = 2) or (a.unique2 = 3 and b.hundred = 4);
|
|
|
|
QUERY PLAN
|
|
|
|
-------------------------------------------------------------------------------------------------
|
|
|
|
Nested Loop
|
|
|
|
Join Filter: (((a.unique1 = 1) AND (b.unique1 = 2)) OR ((a.unique2 = 3) AND (b.hundred = 4)))
|
|
|
|
-> Bitmap Heap Scan on tenk1 b
|
|
|
|
Recheck Cond: ((unique1 = 2) OR (hundred = 4))
|
|
|
|
-> BitmapOr
|
|
|
|
-> Bitmap Index Scan on tenk1_unique1
|
|
|
|
Index Cond: (unique1 = 2)
|
|
|
|
-> Bitmap Index Scan on tenk1_hundred
|
|
|
|
Index Cond: (hundred = 4)
|
|
|
|
-> Materialize
|
|
|
|
-> Bitmap Heap Scan on tenk1 a
|
|
|
|
Recheck Cond: ((unique1 = 1) OR (unique2 = 3))
|
|
|
|
-> BitmapOr
|
|
|
|
-> Bitmap Index Scan on tenk1_unique1
|
|
|
|
Index Cond: (unique1 = 1)
|
|
|
|
-> Bitmap Index Scan on tenk1_unique2
|
|
|
|
Index Cond: (unique2 = 3)
|
|
|
|
(17 rows)
|
|
|
|
|
|
|
|
explain (costs off)
|
|
|
|
select * from tenk1 a join tenk1 b on
|
|
|
|
(a.unique1 = 1 and b.unique1 = 2) or (a.unique2 = 3 and b.ten = 4);
|
|
|
|
QUERY PLAN
|
|
|
|
---------------------------------------------------------------------------------------------
|
|
|
|
Nested Loop
|
|
|
|
Join Filter: (((a.unique1 = 1) AND (b.unique1 = 2)) OR ((a.unique2 = 3) AND (b.ten = 4)))
|
|
|
|
-> Seq Scan on tenk1 b
|
|
|
|
Filter: ((unique1 = 2) OR (ten = 4))
|
|
|
|
-> Materialize
|
|
|
|
-> Bitmap Heap Scan on tenk1 a
|
|
|
|
Recheck Cond: ((unique1 = 1) OR (unique2 = 3))
|
|
|
|
-> BitmapOr
|
|
|
|
-> Bitmap Index Scan on tenk1_unique1
|
|
|
|
Index Cond: (unique1 = 1)
|
|
|
|
-> Bitmap Index Scan on tenk1_unique2
|
|
|
|
Index Cond: (unique2 = 3)
|
|
|
|
(12 rows)
|
|
|
|
|
2014-09-10 00:35:14 +02:00
|
|
|
explain (costs off)
|
|
|
|
select * from tenk1 a join tenk1 b on
|
|
|
|
(a.unique1 = 1 and b.unique1 = 2) or
|
|
|
|
((a.unique2 = 3 or a.unique2 = 7) and b.hundred = 4);
|
|
|
|
QUERY PLAN
|
|
|
|
----------------------------------------------------------------------------------------------------------------------
|
|
|
|
Nested Loop
|
|
|
|
Join Filter: (((a.unique1 = 1) AND (b.unique1 = 2)) OR (((a.unique2 = 3) OR (a.unique2 = 7)) AND (b.hundred = 4)))
|
|
|
|
-> Bitmap Heap Scan on tenk1 b
|
|
|
|
Recheck Cond: ((unique1 = 2) OR (hundred = 4))
|
|
|
|
-> BitmapOr
|
|
|
|
-> Bitmap Index Scan on tenk1_unique1
|
|
|
|
Index Cond: (unique1 = 2)
|
|
|
|
-> Bitmap Index Scan on tenk1_hundred
|
|
|
|
Index Cond: (hundred = 4)
|
|
|
|
-> Materialize
|
|
|
|
-> Bitmap Heap Scan on tenk1 a
|
|
|
|
Recheck Cond: ((unique1 = 1) OR (unique2 = 3) OR (unique2 = 7))
|
|
|
|
-> BitmapOr
|
|
|
|
-> Bitmap Index Scan on tenk1_unique1
|
|
|
|
Index Cond: (unique1 = 1)
|
|
|
|
-> Bitmap Index Scan on tenk1_unique2
|
|
|
|
Index Cond: (unique2 = 3)
|
|
|
|
-> Bitmap Index Scan on tenk1_unique2
|
|
|
|
Index Cond: (unique2 = 7)
|
|
|
|
(19 rows)
|
|
|
|
|
Revise parameterized-path mechanism to fix assorted issues.
This patch adjusts the treatment of parameterized paths so that all paths
with the same parameterization (same set of required outer rels) for the
same relation will have the same rowcount estimate. We cache the rowcount
estimates to ensure that property, and hopefully save a few cycles too.
Doing this makes it practical for add_path_precheck to operate without
a rowcount estimate: it need only assume that paths with different
parameterizations never dominate each other, which is close enough to
true anyway for coarse filtering, because normally a more-parameterized
path should yield fewer rows thanks to having more join clauses to apply.
In add_path, we do the full nine yards of comparing rowcount estimates
along with everything else, so that we can discard parameterized paths that
don't actually have an advantage. This fixes some issues I'd found with
add_path rejecting parameterized paths on the grounds that they were more
expensive than not-parameterized ones, even though they yielded many fewer
rows and hence would be cheaper once subsequent joining was considered.
To make the same-rowcounts assumption valid, we have to require that any
parameterized path enforce *all* join clauses that could be obtained from
the particular set of outer rels, even if not all of them are useful for
indexing. This is required at both base scans and joins. It's a good
thing anyway since the net impact is that join quals are checked at the
lowest practical level in the join tree. Hence, discard the original
rather ad-hoc mechanism for choosing parameterization joinquals, and build
a better one that has a more principled rule for when clauses can be moved.
The original rule was actually buggy anyway for lack of knowledge about
which relations are part of an outer join's outer side; getting this right
requires adding an outer_relids field to RestrictInfo.
2012-04-19 21:52:46 +02:00
|
|
|
--
|
|
|
|
-- test placement of movable quals in a parameterized join tree
|
|
|
|
--
|
|
|
|
explain (costs off)
|
|
|
|
select * from tenk1 t1 left join
|
|
|
|
(tenk1 t2 join tenk1 t3 on t2.thousand = t3.unique2)
|
|
|
|
on t1.hundred = t2.hundred and t1.ten = t3.ten
|
|
|
|
where t1.unique1 = 1;
|
|
|
|
QUERY PLAN
|
|
|
|
--------------------------------------------------------
|
|
|
|
Nested Loop Left Join
|
|
|
|
-> Index Scan using tenk1_unique1 on tenk1 t1
|
|
|
|
Index Cond: (unique1 = 1)
|
|
|
|
-> Nested Loop
|
|
|
|
Join Filter: (t1.ten = t3.ten)
|
2012-08-16 19:03:54 +02:00
|
|
|
-> Bitmap Heap Scan on tenk1 t2
|
|
|
|
Recheck Cond: (t1.hundred = hundred)
|
|
|
|
-> Bitmap Index Scan on tenk1_hundred
|
|
|
|
Index Cond: (t1.hundred = hundred)
|
Revise parameterized-path mechanism to fix assorted issues.
This patch adjusts the treatment of parameterized paths so that all paths
with the same parameterization (same set of required outer rels) for the
same relation will have the same rowcount estimate. We cache the rowcount
estimates to ensure that property, and hopefully save a few cycles too.
Doing this makes it practical for add_path_precheck to operate without
a rowcount estimate: it need only assume that paths with different
parameterizations never dominate each other, which is close enough to
true anyway for coarse filtering, because normally a more-parameterized
path should yield fewer rows thanks to having more join clauses to apply.
In add_path, we do the full nine yards of comparing rowcount estimates
along with everything else, so that we can discard parameterized paths that
don't actually have an advantage. This fixes some issues I'd found with
add_path rejecting parameterized paths on the grounds that they were more
expensive than not-parameterized ones, even though they yielded many fewer
rows and hence would be cheaper once subsequent joining was considered.
To make the same-rowcounts assumption valid, we have to require that any
parameterized path enforce *all* join clauses that could be obtained from
the particular set of outer rels, even if not all of them are useful for
indexing. This is required at both base scans and joins. It's a good
thing anyway since the net impact is that join quals are checked at the
lowest practical level in the join tree. Hence, discard the original
rather ad-hoc mechanism for choosing parameterization joinquals, and build
a better one that has a more principled rule for when clauses can be moved.
The original rule was actually buggy anyway for lack of knowledge about
which relations are part of an outer join's outer side; getting this right
requires adding an outer_relids field to RestrictInfo.
2012-04-19 21:52:46 +02:00
|
|
|
-> Index Scan using tenk1_unique2 on tenk1 t3
|
|
|
|
Index Cond: (unique2 = t2.thousand)
|
2012-08-16 19:03:54 +02:00
|
|
|
(11 rows)
|
Revise parameterized-path mechanism to fix assorted issues.
This patch adjusts the treatment of parameterized paths so that all paths
with the same parameterization (same set of required outer rels) for the
same relation will have the same rowcount estimate. We cache the rowcount
estimates to ensure that property, and hopefully save a few cycles too.
Doing this makes it practical for add_path_precheck to operate without
a rowcount estimate: it need only assume that paths with different
parameterizations never dominate each other, which is close enough to
true anyway for coarse filtering, because normally a more-parameterized
path should yield fewer rows thanks to having more join clauses to apply.
In add_path, we do the full nine yards of comparing rowcount estimates
along with everything else, so that we can discard parameterized paths that
don't actually have an advantage. This fixes some issues I'd found with
add_path rejecting parameterized paths on the grounds that they were more
expensive than not-parameterized ones, even though they yielded many fewer
rows and hence would be cheaper once subsequent joining was considered.
To make the same-rowcounts assumption valid, we have to require that any
parameterized path enforce *all* join clauses that could be obtained from
the particular set of outer rels, even if not all of them are useful for
indexing. This is required at both base scans and joins. It's a good
thing anyway since the net impact is that join quals are checked at the
lowest practical level in the join tree. Hence, discard the original
rather ad-hoc mechanism for choosing parameterization joinquals, and build
a better one that has a more principled rule for when clauses can be moved.
The original rule was actually buggy anyway for lack of knowledge about
which relations are part of an outer join's outer side; getting this right
requires adding an outer_relids field to RestrictInfo.
2012-04-19 21:52:46 +02:00
|
|
|
|
|
|
|
explain (costs off)
|
|
|
|
select * from tenk1 t1 left join
|
|
|
|
(tenk1 t2 join tenk1 t3 on t2.thousand = t3.unique2)
|
|
|
|
on t1.hundred = t2.hundred and t1.ten + t2.ten = t3.ten
|
|
|
|
where t1.unique1 = 1;
|
|
|
|
QUERY PLAN
|
|
|
|
--------------------------------------------------------
|
|
|
|
Nested Loop Left Join
|
|
|
|
-> Index Scan using tenk1_unique1 on tenk1 t1
|
|
|
|
Index Cond: (unique1 = 1)
|
|
|
|
-> Nested Loop
|
|
|
|
Join Filter: ((t1.ten + t2.ten) = t3.ten)
|
2012-08-16 19:03:54 +02:00
|
|
|
-> Bitmap Heap Scan on tenk1 t2
|
|
|
|
Recheck Cond: (t1.hundred = hundred)
|
|
|
|
-> Bitmap Index Scan on tenk1_hundred
|
|
|
|
Index Cond: (t1.hundred = hundred)
|
Revise parameterized-path mechanism to fix assorted issues.
This patch adjusts the treatment of parameterized paths so that all paths
with the same parameterization (same set of required outer rels) for the
same relation will have the same rowcount estimate. We cache the rowcount
estimates to ensure that property, and hopefully save a few cycles too.
Doing this makes it practical for add_path_precheck to operate without
a rowcount estimate: it need only assume that paths with different
parameterizations never dominate each other, which is close enough to
true anyway for coarse filtering, because normally a more-parameterized
path should yield fewer rows thanks to having more join clauses to apply.
In add_path, we do the full nine yards of comparing rowcount estimates
along with everything else, so that we can discard parameterized paths that
don't actually have an advantage. This fixes some issues I'd found with
add_path rejecting parameterized paths on the grounds that they were more
expensive than not-parameterized ones, even though they yielded many fewer
rows and hence would be cheaper once subsequent joining was considered.
To make the same-rowcounts assumption valid, we have to require that any
parameterized path enforce *all* join clauses that could be obtained from
the particular set of outer rels, even if not all of them are useful for
indexing. This is required at both base scans and joins. It's a good
thing anyway since the net impact is that join quals are checked at the
lowest practical level in the join tree. Hence, discard the original
rather ad-hoc mechanism for choosing parameterization joinquals, and build
a better one that has a more principled rule for when clauses can be moved.
The original rule was actually buggy anyway for lack of knowledge about
which relations are part of an outer join's outer side; getting this right
requires adding an outer_relids field to RestrictInfo.
2012-04-19 21:52:46 +02:00
|
|
|
-> Index Scan using tenk1_unique2 on tenk1 t3
|
|
|
|
Index Cond: (unique2 = t2.thousand)
|
2012-08-16 19:03:54 +02:00
|
|
|
(11 rows)
|
Revise parameterized-path mechanism to fix assorted issues.
This patch adjusts the treatment of parameterized paths so that all paths
with the same parameterization (same set of required outer rels) for the
same relation will have the same rowcount estimate. We cache the rowcount
estimates to ensure that property, and hopefully save a few cycles too.
Doing this makes it practical for add_path_precheck to operate without
a rowcount estimate: it need only assume that paths with different
parameterizations never dominate each other, which is close enough to
true anyway for coarse filtering, because normally a more-parameterized
path should yield fewer rows thanks to having more join clauses to apply.
In add_path, we do the full nine yards of comparing rowcount estimates
along with everything else, so that we can discard parameterized paths that
don't actually have an advantage. This fixes some issues I'd found with
add_path rejecting parameterized paths on the grounds that they were more
expensive than not-parameterized ones, even though they yielded many fewer
rows and hence would be cheaper once subsequent joining was considered.
To make the same-rowcounts assumption valid, we have to require that any
parameterized path enforce *all* join clauses that could be obtained from
the particular set of outer rels, even if not all of them are useful for
indexing. This is required at both base scans and joins. It's a good
thing anyway since the net impact is that join quals are checked at the
lowest practical level in the join tree. Hence, discard the original
rather ad-hoc mechanism for choosing parameterization joinquals, and build
a better one that has a more principled rule for when clauses can be moved.
The original rule was actually buggy anyway for lack of knowledge about
which relations are part of an outer join's outer side; getting this right
requires adding an outer_relids field to RestrictInfo.
2012-04-19 21:52:46 +02:00
|
|
|
|
|
|
|
explain (costs off)
|
|
|
|
select count(*) from
|
|
|
|
tenk1 a join tenk1 b on a.unique1 = b.unique2
|
|
|
|
left join tenk1 c on a.unique2 = b.unique1 and c.thousand = a.thousand
|
|
|
|
join int4_tbl on b.thousand = f1;
|
2012-08-16 19:03:54 +02:00
|
|
|
QUERY PLAN
|
|
|
|
-------------------------------------------------------------------------
|
Revise parameterized-path mechanism to fix assorted issues.
This patch adjusts the treatment of parameterized paths so that all paths
with the same parameterization (same set of required outer rels) for the
same relation will have the same rowcount estimate. We cache the rowcount
estimates to ensure that property, and hopefully save a few cycles too.
Doing this makes it practical for add_path_precheck to operate without
a rowcount estimate: it need only assume that paths with different
parameterizations never dominate each other, which is close enough to
true anyway for coarse filtering, because normally a more-parameterized
path should yield fewer rows thanks to having more join clauses to apply.
In add_path, we do the full nine yards of comparing rowcount estimates
along with everything else, so that we can discard parameterized paths that
don't actually have an advantage. This fixes some issues I'd found with
add_path rejecting parameterized paths on the grounds that they were more
expensive than not-parameterized ones, even though they yielded many fewer
rows and hence would be cheaper once subsequent joining was considered.
To make the same-rowcounts assumption valid, we have to require that any
parameterized path enforce *all* join clauses that could be obtained from
the particular set of outer rels, even if not all of them are useful for
indexing. This is required at both base scans and joins. It's a good
thing anyway since the net impact is that join quals are checked at the
lowest practical level in the join tree. Hence, discard the original
rather ad-hoc mechanism for choosing parameterization joinquals, and build
a better one that has a more principled rule for when clauses can be moved.
The original rule was actually buggy anyway for lack of knowledge about
which relations are part of an outer join's outer side; getting this right
requires adding an outer_relids field to RestrictInfo.
2012-04-19 21:52:46 +02:00
|
|
|
Aggregate
|
|
|
|
-> Nested Loop Left Join
|
|
|
|
Join Filter: (a.unique2 = b.unique1)
|
|
|
|
-> Nested Loop
|
|
|
|
-> Nested Loop
|
|
|
|
-> Seq Scan on int4_tbl
|
2012-08-16 19:03:54 +02:00
|
|
|
-> Bitmap Heap Scan on tenk1 b
|
|
|
|
Recheck Cond: (thousand = int4_tbl.f1)
|
|
|
|
-> Bitmap Index Scan on tenk1_thous_tenthous
|
|
|
|
Index Cond: (thousand = int4_tbl.f1)
|
Revise parameterized-path mechanism to fix assorted issues.
This patch adjusts the treatment of parameterized paths so that all paths
with the same parameterization (same set of required outer rels) for the
same relation will have the same rowcount estimate. We cache the rowcount
estimates to ensure that property, and hopefully save a few cycles too.
Doing this makes it practical for add_path_precheck to operate without
a rowcount estimate: it need only assume that paths with different
parameterizations never dominate each other, which is close enough to
true anyway for coarse filtering, because normally a more-parameterized
path should yield fewer rows thanks to having more join clauses to apply.
In add_path, we do the full nine yards of comparing rowcount estimates
along with everything else, so that we can discard parameterized paths that
don't actually have an advantage. This fixes some issues I'd found with
add_path rejecting parameterized paths on the grounds that they were more
expensive than not-parameterized ones, even though they yielded many fewer
rows and hence would be cheaper once subsequent joining was considered.
To make the same-rowcounts assumption valid, we have to require that any
parameterized path enforce *all* join clauses that could be obtained from
the particular set of outer rels, even if not all of them are useful for
indexing. This is required at both base scans and joins. It's a good
thing anyway since the net impact is that join quals are checked at the
lowest practical level in the join tree. Hence, discard the original
rather ad-hoc mechanism for choosing parameterization joinquals, and build
a better one that has a more principled rule for when clauses can be moved.
The original rule was actually buggy anyway for lack of knowledge about
which relations are part of an outer join's outer side; getting this right
requires adding an outer_relids field to RestrictInfo.
2012-04-19 21:52:46 +02:00
|
|
|
-> Index Scan using tenk1_unique1 on tenk1 a
|
|
|
|
Index Cond: (unique1 = b.unique2)
|
|
|
|
-> Index Only Scan using tenk1_thous_tenthous on tenk1 c
|
|
|
|
Index Cond: (thousand = a.thousand)
|
2012-08-16 19:03:54 +02:00
|
|
|
(14 rows)
|
Revise parameterized-path mechanism to fix assorted issues.
This patch adjusts the treatment of parameterized paths so that all paths
with the same parameterization (same set of required outer rels) for the
same relation will have the same rowcount estimate. We cache the rowcount
estimates to ensure that property, and hopefully save a few cycles too.
Doing this makes it practical for add_path_precheck to operate without
a rowcount estimate: it need only assume that paths with different
parameterizations never dominate each other, which is close enough to
true anyway for coarse filtering, because normally a more-parameterized
path should yield fewer rows thanks to having more join clauses to apply.
In add_path, we do the full nine yards of comparing rowcount estimates
along with everything else, so that we can discard parameterized paths that
don't actually have an advantage. This fixes some issues I'd found with
add_path rejecting parameterized paths on the grounds that they were more
expensive than not-parameterized ones, even though they yielded many fewer
rows and hence would be cheaper once subsequent joining was considered.
To make the same-rowcounts assumption valid, we have to require that any
parameterized path enforce *all* join clauses that could be obtained from
the particular set of outer rels, even if not all of them are useful for
indexing. This is required at both base scans and joins. It's a good
thing anyway since the net impact is that join quals are checked at the
lowest practical level in the join tree. Hence, discard the original
rather ad-hoc mechanism for choosing parameterization joinquals, and build
a better one that has a more principled rule for when clauses can be moved.
The original rule was actually buggy anyway for lack of knowledge about
which relations are part of an outer join's outer side; getting this right
requires adding an outer_relids field to RestrictInfo.
2012-04-19 21:52:46 +02:00
|
|
|
|
|
|
|
select count(*) from
|
|
|
|
tenk1 a join tenk1 b on a.unique1 = b.unique2
|
|
|
|
left join tenk1 c on a.unique2 = b.unique1 and c.thousand = a.thousand
|
|
|
|
join int4_tbl on b.thousand = f1;
|
|
|
|
count
|
|
|
|
-------
|
|
|
|
10
|
|
|
|
(1 row)
|
|
|
|
|
|
|
|
explain (costs off)
|
|
|
|
select b.unique1 from
|
|
|
|
tenk1 a join tenk1 b on a.unique1 = b.unique2
|
|
|
|
left join tenk1 c on b.unique1 = 42 and c.thousand = a.thousand
|
|
|
|
join int4_tbl i1 on b.thousand = f1
|
|
|
|
right join int4_tbl i2 on i2.f1 = b.tenthous
|
|
|
|
order by 1;
|
|
|
|
QUERY PLAN
|
|
|
|
-----------------------------------------------------------------------------------------
|
|
|
|
Sort
|
|
|
|
Sort Key: b.unique1
|
|
|
|
-> Nested Loop Left Join
|
|
|
|
-> Seq Scan on int4_tbl i2
|
2012-09-16 23:57:18 +02:00
|
|
|
-> Nested Loop Left Join
|
|
|
|
Join Filter: (b.unique1 = 42)
|
|
|
|
-> Nested Loop
|
Revise parameterized-path mechanism to fix assorted issues.
This patch adjusts the treatment of parameterized paths so that all paths
with the same parameterization (same set of required outer rels) for the
same relation will have the same rowcount estimate. We cache the rowcount
estimates to ensure that property, and hopefully save a few cycles too.
Doing this makes it practical for add_path_precheck to operate without
a rowcount estimate: it need only assume that paths with different
parameterizations never dominate each other, which is close enough to
true anyway for coarse filtering, because normally a more-parameterized
path should yield fewer rows thanks to having more join clauses to apply.
In add_path, we do the full nine yards of comparing rowcount estimates
along with everything else, so that we can discard parameterized paths that
don't actually have an advantage. This fixes some issues I'd found with
add_path rejecting parameterized paths on the grounds that they were more
expensive than not-parameterized ones, even though they yielded many fewer
rows and hence would be cheaper once subsequent joining was considered.
To make the same-rowcounts assumption valid, we have to require that any
parameterized path enforce *all* join clauses that could be obtained from
the particular set of outer rels, even if not all of them are useful for
indexing. This is required at both base scans and joins. It's a good
thing anyway since the net impact is that join quals are checked at the
lowest practical level in the join tree. Hence, discard the original
rather ad-hoc mechanism for choosing parameterization joinquals, and build
a better one that has a more principled rule for when clauses can be moved.
The original rule was actually buggy anyway for lack of knowledge about
which relations are part of an outer join's outer side; getting this right
requires adding an outer_relids field to RestrictInfo.
2012-04-19 21:52:46 +02:00
|
|
|
-> Nested Loop
|
2012-09-16 23:57:18 +02:00
|
|
|
-> Seq Scan on int4_tbl i1
|
Revise parameterized-path mechanism to fix assorted issues.
This patch adjusts the treatment of parameterized paths so that all paths
with the same parameterization (same set of required outer rels) for the
same relation will have the same rowcount estimate. We cache the rowcount
estimates to ensure that property, and hopefully save a few cycles too.
Doing this makes it practical for add_path_precheck to operate without
a rowcount estimate: it need only assume that paths with different
parameterizations never dominate each other, which is close enough to
true anyway for coarse filtering, because normally a more-parameterized
path should yield fewer rows thanks to having more join clauses to apply.
In add_path, we do the full nine yards of comparing rowcount estimates
along with everything else, so that we can discard parameterized paths that
don't actually have an advantage. This fixes some issues I'd found with
add_path rejecting parameterized paths on the grounds that they were more
expensive than not-parameterized ones, even though they yielded many fewer
rows and hence would be cheaper once subsequent joining was considered.
To make the same-rowcounts assumption valid, we have to require that any
parameterized path enforce *all* join clauses that could be obtained from
the particular set of outer rels, even if not all of them are useful for
indexing. This is required at both base scans and joins. It's a good
thing anyway since the net impact is that join quals are checked at the
lowest practical level in the join tree. Hence, discard the original
rather ad-hoc mechanism for choosing parameterization joinquals, and build
a better one that has a more principled rule for when clauses can be moved.
The original rule was actually buggy anyway for lack of knowledge about
which relations are part of an outer join's outer side; getting this right
requires adding an outer_relids field to RestrictInfo.
2012-04-19 21:52:46 +02:00
|
|
|
-> Index Scan using tenk1_thous_tenthous on tenk1 b
|
|
|
|
Index Cond: ((thousand = i1.f1) AND (i2.f1 = tenthous))
|
2012-09-16 23:57:18 +02:00
|
|
|
-> Index Scan using tenk1_unique1 on tenk1 a
|
|
|
|
Index Cond: (unique1 = b.unique2)
|
|
|
|
-> Index Only Scan using tenk1_thous_tenthous on tenk1 c
|
|
|
|
Index Cond: (thousand = a.thousand)
|
Revise parameterized-path mechanism to fix assorted issues.
This patch adjusts the treatment of parameterized paths so that all paths
with the same parameterization (same set of required outer rels) for the
same relation will have the same rowcount estimate. We cache the rowcount
estimates to ensure that property, and hopefully save a few cycles too.
Doing this makes it practical for add_path_precheck to operate without
a rowcount estimate: it need only assume that paths with different
parameterizations never dominate each other, which is close enough to
true anyway for coarse filtering, because normally a more-parameterized
path should yield fewer rows thanks to having more join clauses to apply.
In add_path, we do the full nine yards of comparing rowcount estimates
along with everything else, so that we can discard parameterized paths that
don't actually have an advantage. This fixes some issues I'd found with
add_path rejecting parameterized paths on the grounds that they were more
expensive than not-parameterized ones, even though they yielded many fewer
rows and hence would be cheaper once subsequent joining was considered.
To make the same-rowcounts assumption valid, we have to require that any
parameterized path enforce *all* join clauses that could be obtained from
the particular set of outer rels, even if not all of them are useful for
indexing. This is required at both base scans and joins. It's a good
thing anyway since the net impact is that join quals are checked at the
lowest practical level in the join tree. Hence, discard the original
rather ad-hoc mechanism for choosing parameterization joinquals, and build
a better one that has a more principled rule for when clauses can be moved.
The original rule was actually buggy anyway for lack of knowledge about
which relations are part of an outer join's outer side; getting this right
requires adding an outer_relids field to RestrictInfo.
2012-04-19 21:52:46 +02:00
|
|
|
(15 rows)
|
|
|
|
|
|
|
|
select b.unique1 from
|
|
|
|
tenk1 a join tenk1 b on a.unique1 = b.unique2
|
|
|
|
left join tenk1 c on b.unique1 = 42 and c.thousand = a.thousand
|
|
|
|
join int4_tbl i1 on b.thousand = f1
|
|
|
|
right join int4_tbl i2 on i2.f1 = b.tenthous
|
|
|
|
order by 1;
|
|
|
|
unique1
|
|
|
|
---------
|
|
|
|
0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
(5 rows)
|
|
|
|
|
Postpone creation of pathkeys lists to fix bug #8049.
This patch gets rid of the concept of, and infrastructure for,
non-canonical PathKeys; we now only ever create canonical pathkey lists.
The need for non-canonical pathkeys came from the desire to have
grouping_planner initialize query_pathkeys and related pathkey lists before
calling query_planner. However, since query_planner didn't actually *do*
anything with those lists before they'd been made canonical, we can get rid
of the whole mess by just not creating the lists at all until the point
where we formerly canonicalized them.
There are several ways in which we could implement that without making
query_planner itself deal with grouping/sorting features (which are
supposed to be the province of grouping_planner). I chose to add a
callback function to query_planner's API; other alternatives would have
required adding more fields to PlannerInfo, which while not bad in itself
would create an ABI break for planner-related plugins in the 9.2 release
series. This still breaks ABI for anything that calls query_planner
directly, but it seems somewhat unlikely that there are any such plugins.
I had originally conceived of this change as merely a step on the way to
fixing bug #8049 from Teun Hoogendoorn; but it turns out that this fixes
that bug all by itself, as per the added regression test. The reason is
that now get_eclass_for_sort_expr is adding the ORDER BY expression at the
end of EquivalenceClass creation not the start, and so anything that is in
a multi-member EquivalenceClass has already been created with correct
em_nullable_relids. I am suspicious that there are related scenarios in
which we still need to teach get_eclass_for_sort_expr to compute correct
nullable_relids, but am not eager to risk destabilizing either 9.2 or 9.3
to fix bugs that are only hypothetical. So for the moment, do this and
stop here.
Back-patch to 9.2 but not to earlier branches, since they don't exhibit
this bug for lack of join-clause-movement logic that depends on
em_nullable_relids being correct. (We might have to revisit that choice
if any related bugs turn up.) In 9.2, don't change the signature of
make_pathkeys_for_sortclauses nor remove canonicalize_pathkeys, so as
not to risk more plugin breakage than we have to.
2013-04-29 20:49:01 +02:00
|
|
|
explain (costs off)
|
|
|
|
select * from
|
|
|
|
(
|
|
|
|
select unique1, q1, coalesce(unique1, -1) + q1 as fault
|
|
|
|
from int8_tbl left join tenk1 on (q2 = unique2)
|
|
|
|
) ss
|
|
|
|
where fault = 122
|
|
|
|
order by fault;
|
2015-03-30 20:59:49 +02:00
|
|
|
QUERY PLAN
|
|
|
|
--------------------------------------------------------------------------
|
Postpone creation of pathkeys lists to fix bug #8049.
This patch gets rid of the concept of, and infrastructure for,
non-canonical PathKeys; we now only ever create canonical pathkey lists.
The need for non-canonical pathkeys came from the desire to have
grouping_planner initialize query_pathkeys and related pathkey lists before
calling query_planner. However, since query_planner didn't actually *do*
anything with those lists before they'd been made canonical, we can get rid
of the whole mess by just not creating the lists at all until the point
where we formerly canonicalized them.
There are several ways in which we could implement that without making
query_planner itself deal with grouping/sorting features (which are
supposed to be the province of grouping_planner). I chose to add a
callback function to query_planner's API; other alternatives would have
required adding more fields to PlannerInfo, which while not bad in itself
would create an ABI break for planner-related plugins in the 9.2 release
series. This still breaks ABI for anything that calls query_planner
directly, but it seems somewhat unlikely that there are any such plugins.
I had originally conceived of this change as merely a step on the way to
fixing bug #8049 from Teun Hoogendoorn; but it turns out that this fixes
that bug all by itself, as per the added regression test. The reason is
that now get_eclass_for_sort_expr is adding the ORDER BY expression at the
end of EquivalenceClass creation not the start, and so anything that is in
a multi-member EquivalenceClass has already been created with correct
em_nullable_relids. I am suspicious that there are related scenarios in
which we still need to teach get_eclass_for_sort_expr to compute correct
nullable_relids, but am not eager to risk destabilizing either 9.2 or 9.3
to fix bugs that are only hypothetical. So for the moment, do this and
stop here.
Back-patch to 9.2 but not to earlier branches, since they don't exhibit
this bug for lack of join-clause-movement logic that depends on
em_nullable_relids being correct. (We might have to revisit that choice
if any related bugs turn up.) In 9.2, don't change the signature of
make_pathkeys_for_sortclauses nor remove canonicalize_pathkeys, so as
not to risk more plugin breakage than we have to.
2013-04-29 20:49:01 +02:00
|
|
|
Nested Loop Left Join
|
2015-03-30 20:59:49 +02:00
|
|
|
Filter: ((COALESCE(tenk1.unique1, '-1'::integer) + int8_tbl.q1) = 122)
|
Postpone creation of pathkeys lists to fix bug #8049.
This patch gets rid of the concept of, and infrastructure for,
non-canonical PathKeys; we now only ever create canonical pathkey lists.
The need for non-canonical pathkeys came from the desire to have
grouping_planner initialize query_pathkeys and related pathkey lists before
calling query_planner. However, since query_planner didn't actually *do*
anything with those lists before they'd been made canonical, we can get rid
of the whole mess by just not creating the lists at all until the point
where we formerly canonicalized them.
There are several ways in which we could implement that without making
query_planner itself deal with grouping/sorting features (which are
supposed to be the province of grouping_planner). I chose to add a
callback function to query_planner's API; other alternatives would have
required adding more fields to PlannerInfo, which while not bad in itself
would create an ABI break for planner-related plugins in the 9.2 release
series. This still breaks ABI for anything that calls query_planner
directly, but it seems somewhat unlikely that there are any such plugins.
I had originally conceived of this change as merely a step on the way to
fixing bug #8049 from Teun Hoogendoorn; but it turns out that this fixes
that bug all by itself, as per the added regression test. The reason is
that now get_eclass_for_sort_expr is adding the ORDER BY expression at the
end of EquivalenceClass creation not the start, and so anything that is in
a multi-member EquivalenceClass has already been created with correct
em_nullable_relids. I am suspicious that there are related scenarios in
which we still need to teach get_eclass_for_sort_expr to compute correct
nullable_relids, but am not eager to risk destabilizing either 9.2 or 9.3
to fix bugs that are only hypothetical. So for the moment, do this and
stop here.
Back-patch to 9.2 but not to earlier branches, since they don't exhibit
this bug for lack of join-clause-movement logic that depends on
em_nullable_relids being correct. (We might have to revisit that choice
if any related bugs turn up.) In 9.2, don't change the signature of
make_pathkeys_for_sortclauses nor remove canonicalize_pathkeys, so as
not to risk more plugin breakage than we have to.
2013-04-29 20:49:01 +02:00
|
|
|
-> Seq Scan on int8_tbl
|
|
|
|
-> Index Scan using tenk1_unique2 on tenk1
|
|
|
|
Index Cond: (int8_tbl.q2 = unique2)
|
|
|
|
(5 rows)
|
|
|
|
|
|
|
|
select * from
|
|
|
|
(
|
|
|
|
select unique1, q1, coalesce(unique1, -1) + q1 as fault
|
|
|
|
from int8_tbl left join tenk1 on (q2 = unique2)
|
|
|
|
) ss
|
|
|
|
where fault = 122
|
|
|
|
order by fault;
|
|
|
|
unique1 | q1 | fault
|
|
|
|
---------+-----+-------
|
|
|
|
| 123 | 122
|
|
|
|
(1 row)
|
|
|
|
|
Fix planning of non-strict equivalence clauses above outer joins.
If a potential equivalence clause references a variable from the nullable
side of an outer join, the planner needs to take care that derived clauses
are not pushed to below the outer join; else they may use the wrong value
for the variable. (The problem arises only with non-strict clauses, since
if an upper clause can be proven strict then the outer join will get
simplified to a plain join.) The planner attempted to prevent this type
of error by checking that potential equivalence clauses aren't
outerjoin-delayed as a whole, but actually we have to check each side
separately, since the two sides of the clause will get moved around
separately if it's treated as an equivalence. Bugs of this type can be
demonstrated as far back as 7.4, even though releases before 8.3 had only
a very ad-hoc notion of equivalence clauses.
In addition, we neglected to account for the possibility that such clauses
might have nonempty nullable_relids even when not outerjoin-delayed; so the
equivalence-class machinery lacked logic to compute correct nullable_relids
values for clauses it constructs. This oversight was harmless before 9.2
because we were only using RestrictInfo.nullable_relids for OR clauses;
but as of 9.2 it could result in pushing constructed equivalence clauses
to incorrect places. (This accounts for bug #7604 from Bill MacArthur.)
Fix the first problem by adding a new test check_equivalence_delay() in
distribute_qual_to_rels, and fix the second one by adding code in
equivclass.c and called functions to set correct nullable_relids for
generated clauses. Although I believe the second part of this is not
currently necessary before 9.2, I chose to back-patch it anyway, partly to
keep the logic similar across branches and partly because it seems possible
we might find other reasons why we need valid values of nullable_relids in
the older branches.
Add regression tests illustrating these problems. In 9.0 and up, also
add test cases checking that we can push constants through outer joins,
since we've broken that optimization before and I nearly broke it again
with an overly simplistic patch for this problem.
2012-10-18 18:28:45 +02:00
|
|
|
--
|
|
|
|
-- test handling of potential equivalence clauses above outer joins
|
|
|
|
--
|
|
|
|
explain (costs off)
|
|
|
|
select q1, unique2, thousand, hundred
|
|
|
|
from int8_tbl a left join tenk1 b on q1 = unique2
|
|
|
|
where coalesce(thousand,123) = q1 and q1 = coalesce(hundred,123);
|
|
|
|
QUERY PLAN
|
|
|
|
--------------------------------------------------------------------------------------
|
|
|
|
Nested Loop Left Join
|
|
|
|
Filter: ((COALESCE(b.thousand, 123) = a.q1) AND (a.q1 = COALESCE(b.hundred, 123)))
|
|
|
|
-> Seq Scan on int8_tbl a
|
|
|
|
-> Index Scan using tenk1_unique2 on tenk1 b
|
|
|
|
Index Cond: (a.q1 = unique2)
|
|
|
|
(5 rows)
|
|
|
|
|
|
|
|
select q1, unique2, thousand, hundred
|
|
|
|
from int8_tbl a left join tenk1 b on q1 = unique2
|
|
|
|
where coalesce(thousand,123) = q1 and q1 = coalesce(hundred,123);
|
|
|
|
q1 | unique2 | thousand | hundred
|
|
|
|
----+---------+----------+---------
|
|
|
|
(0 rows)
|
|
|
|
|
|
|
|
explain (costs off)
|
|
|
|
select f1, unique2, case when unique2 is null then f1 else 0 end
|
|
|
|
from int4_tbl a left join tenk1 b on f1 = unique2
|
|
|
|
where (case when unique2 is null then f1 else 0 end) = 0;
|
|
|
|
QUERY PLAN
|
|
|
|
--------------------------------------------------------------------
|
|
|
|
Nested Loop Left Join
|
|
|
|
Filter: (CASE WHEN (b.unique2 IS NULL) THEN a.f1 ELSE 0 END = 0)
|
|
|
|
-> Seq Scan on int4_tbl a
|
|
|
|
-> Index Only Scan using tenk1_unique2 on tenk1 b
|
|
|
|
Index Cond: (unique2 = a.f1)
|
|
|
|
(5 rows)
|
|
|
|
|
|
|
|
select f1, unique2, case when unique2 is null then f1 else 0 end
|
|
|
|
from int4_tbl a left join tenk1 b on f1 = unique2
|
|
|
|
where (case when unique2 is null then f1 else 0 end) = 0;
|
|
|
|
f1 | unique2 | case
|
|
|
|
----+---------+------
|
|
|
|
0 | 0 | 0
|
|
|
|
(1 row)
|
|
|
|
|
Compute correct em_nullable_relids in get_eclass_for_sort_expr().
Bug #8591 from Claudio Freire demonstrates that get_eclass_for_sort_expr
must be able to compute valid em_nullable_relids for any new equivalence
class members it creates. I'd worried about this in the commit message
for db9f0e1d9a4a0842c814a464cdc9758c3f20b96c, but claimed that it wasn't a
problem because multi-member ECs should already exist when it runs. That
is transparently wrong, though, because this function is also called by
initialize_mergeclause_eclasses, which runs during deconstruct_jointree.
The example given in the bug report (which the new regression test item
is based upon) fails because the COALESCE() expression is first seen by
initialize_mergeclause_eclasses rather than process_equivalence.
Fixing this requires passing the appropriate nullable_relids set to
get_eclass_for_sort_expr, and it requires new code to compute that set
for top-level expressions such as ORDER BY, GROUP BY, etc. We store
the top-level nullable_relids in a new field in PlannerInfo to avoid
computing it many times. In the back branches, I've added the new
field at the end of the struct to minimize ABI breakage for planner
plugins. There doesn't seem to be a good alternative to changing
get_eclass_for_sort_expr's API signature, though. There probably aren't
any third-party extensions calling that function directly; moreover,
if there are, they probably need to think about what to pass for
nullable_relids anyway.
Back-patch to 9.2, like the previous patch in this area.
2013-11-15 22:46:18 +01:00
|
|
|
--
|
|
|
|
-- another case with equivalence clauses above outer joins (bug #8591)
|
|
|
|
--
|
|
|
|
explain (costs off)
|
|
|
|
select a.unique1, b.unique1, c.unique1, coalesce(b.twothousand, a.twothousand)
|
|
|
|
from tenk1 a left join tenk1 b on b.thousand = a.unique1 left join tenk1 c on c.unique2 = coalesce(b.twothousand, a.twothousand)
|
2015-03-12 03:53:32 +01:00
|
|
|
where a.unique2 < 10 and coalesce(b.twothousand, a.twothousand) = 44;
|
Compute correct em_nullable_relids in get_eclass_for_sort_expr().
Bug #8591 from Claudio Freire demonstrates that get_eclass_for_sort_expr
must be able to compute valid em_nullable_relids for any new equivalence
class members it creates. I'd worried about this in the commit message
for db9f0e1d9a4a0842c814a464cdc9758c3f20b96c, but claimed that it wasn't a
problem because multi-member ECs should already exist when it runs. That
is transparently wrong, though, because this function is also called by
initialize_mergeclause_eclasses, which runs during deconstruct_jointree.
The example given in the bug report (which the new regression test item
is based upon) fails because the COALESCE() expression is first seen by
initialize_mergeclause_eclasses rather than process_equivalence.
Fixing this requires passing the appropriate nullable_relids set to
get_eclass_for_sort_expr, and it requires new code to compute that set
for top-level expressions such as ORDER BY, GROUP BY, etc. We store
the top-level nullable_relids in a new field in PlannerInfo to avoid
computing it many times. In the back branches, I've added the new
field at the end of the struct to minimize ABI breakage for planner
plugins. There doesn't seem to be a good alternative to changing
get_eclass_for_sort_expr's API signature, though. There probably aren't
any third-party extensions calling that function directly; moreover,
if there are, they probably need to think about what to pass for
nullable_relids anyway.
Back-patch to 9.2, like the previous patch in this area.
2013-11-15 22:46:18 +01:00
|
|
|
QUERY PLAN
|
|
|
|
---------------------------------------------------------------------------------------------
|
|
|
|
Nested Loop Left Join
|
|
|
|
-> Nested Loop Left Join
|
|
|
|
Filter: (COALESCE(b.twothousand, a.twothousand) = 44)
|
|
|
|
-> Index Scan using tenk1_unique2 on tenk1 a
|
2015-03-12 03:53:32 +01:00
|
|
|
Index Cond: (unique2 < 10)
|
Compute correct em_nullable_relids in get_eclass_for_sort_expr().
Bug #8591 from Claudio Freire demonstrates that get_eclass_for_sort_expr
must be able to compute valid em_nullable_relids for any new equivalence
class members it creates. I'd worried about this in the commit message
for db9f0e1d9a4a0842c814a464cdc9758c3f20b96c, but claimed that it wasn't a
problem because multi-member ECs should already exist when it runs. That
is transparently wrong, though, because this function is also called by
initialize_mergeclause_eclasses, which runs during deconstruct_jointree.
The example given in the bug report (which the new regression test item
is based upon) fails because the COALESCE() expression is first seen by
initialize_mergeclause_eclasses rather than process_equivalence.
Fixing this requires passing the appropriate nullable_relids set to
get_eclass_for_sort_expr, and it requires new code to compute that set
for top-level expressions such as ORDER BY, GROUP BY, etc. We store
the top-level nullable_relids in a new field in PlannerInfo to avoid
computing it many times. In the back branches, I've added the new
field at the end of the struct to minimize ABI breakage for planner
plugins. There doesn't seem to be a good alternative to changing
get_eclass_for_sort_expr's API signature, though. There probably aren't
any third-party extensions calling that function directly; moreover,
if there are, they probably need to think about what to pass for
nullable_relids anyway.
Back-patch to 9.2, like the previous patch in this area.
2013-11-15 22:46:18 +01:00
|
|
|
-> Bitmap Heap Scan on tenk1 b
|
|
|
|
Recheck Cond: (thousand = a.unique1)
|
|
|
|
-> Bitmap Index Scan on tenk1_thous_tenthous
|
|
|
|
Index Cond: (thousand = a.unique1)
|
|
|
|
-> Index Scan using tenk1_unique2 on tenk1 c
|
|
|
|
Index Cond: ((unique2 = COALESCE(b.twothousand, a.twothousand)) AND (unique2 = 44))
|
|
|
|
(11 rows)
|
|
|
|
|
|
|
|
select a.unique1, b.unique1, c.unique1, coalesce(b.twothousand, a.twothousand)
|
|
|
|
from tenk1 a left join tenk1 b on b.thousand = a.unique1 left join tenk1 c on c.unique2 = coalesce(b.twothousand, a.twothousand)
|
2015-03-12 03:53:32 +01:00
|
|
|
where a.unique2 < 10 and coalesce(b.twothousand, a.twothousand) = 44;
|
Compute correct em_nullable_relids in get_eclass_for_sort_expr().
Bug #8591 from Claudio Freire demonstrates that get_eclass_for_sort_expr
must be able to compute valid em_nullable_relids for any new equivalence
class members it creates. I'd worried about this in the commit message
for db9f0e1d9a4a0842c814a464cdc9758c3f20b96c, but claimed that it wasn't a
problem because multi-member ECs should already exist when it runs. That
is transparently wrong, though, because this function is also called by
initialize_mergeclause_eclasses, which runs during deconstruct_jointree.
The example given in the bug report (which the new regression test item
is based upon) fails because the COALESCE() expression is first seen by
initialize_mergeclause_eclasses rather than process_equivalence.
Fixing this requires passing the appropriate nullable_relids set to
get_eclass_for_sort_expr, and it requires new code to compute that set
for top-level expressions such as ORDER BY, GROUP BY, etc. We store
the top-level nullable_relids in a new field in PlannerInfo to avoid
computing it many times. In the back branches, I've added the new
field at the end of the struct to minimize ABI breakage for planner
plugins. There doesn't seem to be a good alternative to changing
get_eclass_for_sort_expr's API signature, though. There probably aren't
any third-party extensions calling that function directly; moreover,
if there are, they probably need to think about what to pass for
nullable_relids anyway.
Back-patch to 9.2, like the previous patch in this area.
2013-11-15 22:46:18 +01:00
|
|
|
unique1 | unique1 | unique1 | coalesce
|
|
|
|
---------+---------+---------+----------
|
|
|
|
(0 rows)
|
|
|
|
|
Flatten join alias Vars before pulling up targetlist items from a subquery.
pullup_replace_vars()'s decisions about whether a pulled-up replacement
expression needs to be wrapped in a PlaceHolderVar depend on the assumption
that what looks like a Var behaves like a Var. However, if the Var is a
join alias reference, later flattening of join aliases might replace the
Var with something that's not a Var at all, and should have been wrapped.
To fix, do a forcible pass of flatten_join_alias_vars() on the subquery
targetlist before we start to copy items out of it. We'll re-run that
processing on the pulled-up expressions later, but that's harmless.
Per report from Ken Tanzer; the added regression test case is based on his
example. This bug has been there since the PlaceHolderVar mechanism was
invented, but has escaped detection because the circumstances that trigger
it are fairly narrow. You need a flattenable query underneath an outer
join, which contains another flattenable query inside a join of its own,
with a dangerous expression (a constant or something else non-strict)
in that one's targetlist.
Having seen this, I'm wondering if it wouldn't be prudent to do all
alias-variable flattening earlier, perhaps even in the rewriter.
But that would probably not be a back-patchable change.
2013-11-22 20:37:21 +01:00
|
|
|
--
|
|
|
|
-- check handling of join aliases when flattening multiple levels of subquery
|
|
|
|
--
|
|
|
|
explain (verbose, costs off)
|
|
|
|
select foo1.join_key as foo1_id, foo3.join_key AS foo3_id, bug_field from
|
|
|
|
(values (0),(1)) foo1(join_key)
|
|
|
|
left join
|
|
|
|
(select join_key, bug_field from
|
|
|
|
(select ss1.join_key, ss1.bug_field from
|
|
|
|
(select f1 as join_key, 666 as bug_field from int4_tbl i1) ss1
|
|
|
|
) foo2
|
|
|
|
left join
|
|
|
|
(select unique2 as join_key from tenk1 i2) ss2
|
|
|
|
using (join_key)
|
|
|
|
) foo3
|
|
|
|
using (join_key);
|
|
|
|
QUERY PLAN
|
|
|
|
--------------------------------------------------------------------------
|
|
|
|
Nested Loop Left Join
|
|
|
|
Output: "*VALUES*".column1, i1.f1, (666)
|
|
|
|
Join Filter: ("*VALUES*".column1 = i1.f1)
|
|
|
|
-> Values Scan on "*VALUES*"
|
|
|
|
Output: "*VALUES*".column1
|
|
|
|
-> Materialize
|
|
|
|
Output: i1.f1, (666)
|
|
|
|
-> Nested Loop Left Join
|
|
|
|
Output: i1.f1, 666
|
|
|
|
-> Seq Scan on public.int4_tbl i1
|
|
|
|
Output: i1.f1
|
|
|
|
-> Index Only Scan using tenk1_unique2 on public.tenk1 i2
|
|
|
|
Output: i2.unique2
|
|
|
|
Index Cond: (i2.unique2 = i1.f1)
|
|
|
|
(14 rows)
|
|
|
|
|
|
|
|
select foo1.join_key as foo1_id, foo3.join_key AS foo3_id, bug_field from
|
|
|
|
(values (0),(1)) foo1(join_key)
|
|
|
|
left join
|
|
|
|
(select join_key, bug_field from
|
|
|
|
(select ss1.join_key, ss1.bug_field from
|
|
|
|
(select f1 as join_key, 666 as bug_field from int4_tbl i1) ss1
|
|
|
|
) foo2
|
|
|
|
left join
|
|
|
|
(select unique2 as join_key from tenk1 i2) ss2
|
|
|
|
using (join_key)
|
|
|
|
) foo3
|
|
|
|
using (join_key);
|
|
|
|
foo1_id | foo3_id | bug_field
|
|
|
|
---------+---------+-----------
|
|
|
|
0 | 0 | 666
|
|
|
|
1 | |
|
|
|
|
(2 rows)
|
|
|
|
|
2015-08-02 02:57:41 +02:00
|
|
|
--
|
|
|
|
-- test successful handling of nested outer joins with degenerate join quals
|
|
|
|
--
|
|
|
|
explain (verbose, costs off)
|
|
|
|
select t1.* from
|
|
|
|
text_tbl t1
|
|
|
|
left join (select *, '***'::text as d1 from int8_tbl i8b1) b1
|
|
|
|
left join int8_tbl i8
|
|
|
|
left join (select *, null::int as d2 from int8_tbl i8b2) b2
|
|
|
|
on (i8.q1 = b2.q1)
|
|
|
|
on (b2.d2 = b1.q2)
|
|
|
|
on (t1.f1 = b1.d1)
|
|
|
|
left join int4_tbl i4
|
|
|
|
on (i8.q2 = i4.f1);
|
|
|
|
QUERY PLAN
|
|
|
|
----------------------------------------------------------------------
|
|
|
|
Hash Left Join
|
|
|
|
Output: t1.f1
|
|
|
|
Hash Cond: (i8.q2 = i4.f1)
|
|
|
|
-> Nested Loop Left Join
|
|
|
|
Output: t1.f1, i8.q2
|
|
|
|
Join Filter: (t1.f1 = '***'::text)
|
|
|
|
-> Seq Scan on public.text_tbl t1
|
|
|
|
Output: t1.f1
|
|
|
|
-> Materialize
|
|
|
|
Output: i8.q2
|
|
|
|
-> Hash Right Join
|
|
|
|
Output: i8.q2
|
|
|
|
Hash Cond: ((NULL::integer) = i8b1.q2)
|
|
|
|
-> Hash Left Join
|
|
|
|
Output: i8.q2, (NULL::integer)
|
|
|
|
Hash Cond: (i8.q1 = i8b2.q1)
|
|
|
|
-> Seq Scan on public.int8_tbl i8
|
|
|
|
Output: i8.q1, i8.q2
|
|
|
|
-> Hash
|
|
|
|
Output: i8b2.q1, (NULL::integer)
|
|
|
|
-> Seq Scan on public.int8_tbl i8b2
|
|
|
|
Output: i8b2.q1, NULL::integer
|
|
|
|
-> Hash
|
|
|
|
Output: i8b1.q2
|
|
|
|
-> Seq Scan on public.int8_tbl i8b1
|
|
|
|
Output: i8b1.q2
|
|
|
|
-> Hash
|
|
|
|
Output: i4.f1
|
|
|
|
-> Seq Scan on public.int4_tbl i4
|
|
|
|
Output: i4.f1
|
|
|
|
(30 rows)
|
|
|
|
|
|
|
|
select t1.* from
|
|
|
|
text_tbl t1
|
|
|
|
left join (select *, '***'::text as d1 from int8_tbl i8b1) b1
|
|
|
|
left join int8_tbl i8
|
|
|
|
left join (select *, null::int as d2 from int8_tbl i8b2) b2
|
|
|
|
on (i8.q1 = b2.q1)
|
|
|
|
on (b2.d2 = b1.q2)
|
|
|
|
on (t1.f1 = b1.d1)
|
|
|
|
left join int4_tbl i4
|
|
|
|
on (i8.q2 = i4.f1);
|
|
|
|
f1
|
|
|
|
-------------------
|
|
|
|
doh!
|
|
|
|
hi de ho neighbor
|
|
|
|
(2 rows)
|
|
|
|
|
|
|
|
explain (verbose, costs off)
|
|
|
|
select t1.* from
|
|
|
|
text_tbl t1
|
|
|
|
left join (select *, '***'::text as d1 from int8_tbl i8b1) b1
|
|
|
|
left join int8_tbl i8
|
|
|
|
left join (select *, null::int as d2 from int8_tbl i8b2, int4_tbl i4b2) b2
|
|
|
|
on (i8.q1 = b2.q1)
|
|
|
|
on (b2.d2 = b1.q2)
|
|
|
|
on (t1.f1 = b1.d1)
|
|
|
|
left join int4_tbl i4
|
|
|
|
on (i8.q2 = i4.f1);
|
|
|
|
QUERY PLAN
|
|
|
|
----------------------------------------------------------------------------
|
|
|
|
Hash Left Join
|
|
|
|
Output: t1.f1
|
|
|
|
Hash Cond: (i8.q2 = i4.f1)
|
|
|
|
-> Nested Loop Left Join
|
2015-08-06 21:35:27 +02:00
|
|
|
Output: t1.f1, i8.q2
|
2015-08-02 02:57:41 +02:00
|
|
|
Join Filter: (t1.f1 = '***'::text)
|
|
|
|
-> Seq Scan on public.text_tbl t1
|
|
|
|
Output: t1.f1
|
|
|
|
-> Materialize
|
|
|
|
Output: i8.q2
|
|
|
|
-> Hash Right Join
|
|
|
|
Output: i8.q2
|
|
|
|
Hash Cond: ((NULL::integer) = i8b1.q2)
|
|
|
|
-> Hash Right Join
|
|
|
|
Output: i8.q2, (NULL::integer)
|
|
|
|
Hash Cond: (i8b2.q1 = i8.q1)
|
|
|
|
-> Nested Loop
|
|
|
|
Output: i8b2.q1, NULL::integer
|
|
|
|
-> Seq Scan on public.int8_tbl i8b2
|
|
|
|
Output: i8b2.q1, i8b2.q2
|
|
|
|
-> Materialize
|
|
|
|
-> Seq Scan on public.int4_tbl i4b2
|
|
|
|
-> Hash
|
|
|
|
Output: i8.q1, i8.q2
|
|
|
|
-> Seq Scan on public.int8_tbl i8
|
|
|
|
Output: i8.q1, i8.q2
|
|
|
|
-> Hash
|
|
|
|
Output: i8b1.q2
|
|
|
|
-> Seq Scan on public.int8_tbl i8b1
|
|
|
|
Output: i8b1.q2
|
|
|
|
-> Hash
|
|
|
|
Output: i4.f1
|
|
|
|
-> Seq Scan on public.int4_tbl i4
|
|
|
|
Output: i4.f1
|
|
|
|
(34 rows)
|
|
|
|
|
|
|
|
select t1.* from
|
|
|
|
text_tbl t1
|
|
|
|
left join (select *, '***'::text as d1 from int8_tbl i8b1) b1
|
|
|
|
left join int8_tbl i8
|
|
|
|
left join (select *, null::int as d2 from int8_tbl i8b2, int4_tbl i4b2) b2
|
|
|
|
on (i8.q1 = b2.q1)
|
|
|
|
on (b2.d2 = b1.q2)
|
|
|
|
on (t1.f1 = b1.d1)
|
|
|
|
left join int4_tbl i4
|
|
|
|
on (i8.q2 = i4.f1);
|
|
|
|
f1
|
|
|
|
-------------------
|
2015-08-06 21:35:27 +02:00
|
|
|
doh!
|
|
|
|
hi de ho neighbor
|
|
|
|
(2 rows)
|
|
|
|
|
|
|
|
explain (verbose, costs off)
|
|
|
|
select t1.* from
|
|
|
|
text_tbl t1
|
|
|
|
left join (select *, '***'::text as d1 from int8_tbl i8b1) b1
|
|
|
|
left join int8_tbl i8
|
|
|
|
left join (select *, null::int as d2 from int8_tbl i8b2, int4_tbl i4b2
|
|
|
|
where q1 = f1) b2
|
|
|
|
on (i8.q1 = b2.q1)
|
|
|
|
on (b2.d2 = b1.q2)
|
|
|
|
on (t1.f1 = b1.d1)
|
|
|
|
left join int4_tbl i4
|
|
|
|
on (i8.q2 = i4.f1);
|
|
|
|
QUERY PLAN
|
|
|
|
----------------------------------------------------------------------------
|
|
|
|
Hash Left Join
|
|
|
|
Output: t1.f1
|
|
|
|
Hash Cond: (i8.q2 = i4.f1)
|
|
|
|
-> Nested Loop Left Join
|
|
|
|
Output: t1.f1, i8.q2
|
|
|
|
Join Filter: (t1.f1 = '***'::text)
|
|
|
|
-> Seq Scan on public.text_tbl t1
|
|
|
|
Output: t1.f1
|
|
|
|
-> Materialize
|
|
|
|
Output: i8.q2
|
|
|
|
-> Hash Right Join
|
|
|
|
Output: i8.q2
|
|
|
|
Hash Cond: ((NULL::integer) = i8b1.q2)
|
|
|
|
-> Hash Right Join
|
|
|
|
Output: i8.q2, (NULL::integer)
|
|
|
|
Hash Cond: (i8b2.q1 = i8.q1)
|
|
|
|
-> Hash Join
|
|
|
|
Output: i8b2.q1, NULL::integer
|
|
|
|
Hash Cond: (i8b2.q1 = i4b2.f1)
|
|
|
|
-> Seq Scan on public.int8_tbl i8b2
|
|
|
|
Output: i8b2.q1, i8b2.q2
|
|
|
|
-> Hash
|
|
|
|
Output: i4b2.f1
|
|
|
|
-> Seq Scan on public.int4_tbl i4b2
|
|
|
|
Output: i4b2.f1
|
|
|
|
-> Hash
|
|
|
|
Output: i8.q1, i8.q2
|
|
|
|
-> Seq Scan on public.int8_tbl i8
|
|
|
|
Output: i8.q1, i8.q2
|
|
|
|
-> Hash
|
|
|
|
Output: i8b1.q2
|
|
|
|
-> Seq Scan on public.int8_tbl i8b1
|
|
|
|
Output: i8b1.q2
|
|
|
|
-> Hash
|
|
|
|
Output: i4.f1
|
|
|
|
-> Seq Scan on public.int4_tbl i4
|
|
|
|
Output: i4.f1
|
|
|
|
(37 rows)
|
|
|
|
|
|
|
|
select t1.* from
|
|
|
|
text_tbl t1
|
|
|
|
left join (select *, '***'::text as d1 from int8_tbl i8b1) b1
|
|
|
|
left join int8_tbl i8
|
|
|
|
left join (select *, null::int as d2 from int8_tbl i8b2, int4_tbl i4b2
|
|
|
|
where q1 = f1) b2
|
|
|
|
on (i8.q1 = b2.q1)
|
|
|
|
on (b2.d2 = b1.q2)
|
|
|
|
on (t1.f1 = b1.d1)
|
|
|
|
left join int4_tbl i4
|
|
|
|
on (i8.q2 = i4.f1);
|
|
|
|
f1
|
|
|
|
-------------------
|
2015-08-02 02:57:41 +02:00
|
|
|
doh!
|
|
|
|
hi de ho neighbor
|
|
|
|
(2 rows)
|
|
|
|
|
2015-08-13 03:18:45 +02:00
|
|
|
explain (verbose, costs off)
|
|
|
|
select * from
|
|
|
|
text_tbl t1
|
|
|
|
inner join int8_tbl i8
|
|
|
|
on i8.q2 = 456
|
|
|
|
right join text_tbl t2
|
|
|
|
on t1.f1 = 'doh!'
|
|
|
|
left join int4_tbl i4
|
|
|
|
on i8.q1 = i4.f1;
|
|
|
|
QUERY PLAN
|
|
|
|
--------------------------------------------------------
|
|
|
|
Nested Loop Left Join
|
|
|
|
Output: t1.f1, i8.q1, i8.q2, t2.f1, i4.f1
|
|
|
|
-> Seq Scan on public.text_tbl t2
|
|
|
|
Output: t2.f1
|
|
|
|
-> Materialize
|
|
|
|
Output: i8.q1, i8.q2, i4.f1, t1.f1
|
|
|
|
-> Nested Loop
|
|
|
|
Output: i8.q1, i8.q2, i4.f1, t1.f1
|
|
|
|
-> Nested Loop Left Join
|
|
|
|
Output: i8.q1, i8.q2, i4.f1
|
|
|
|
Join Filter: (i8.q1 = i4.f1)
|
|
|
|
-> Seq Scan on public.int8_tbl i8
|
|
|
|
Output: i8.q1, i8.q2
|
|
|
|
Filter: (i8.q2 = 456)
|
|
|
|
-> Seq Scan on public.int4_tbl i4
|
|
|
|
Output: i4.f1
|
|
|
|
-> Seq Scan on public.text_tbl t1
|
|
|
|
Output: t1.f1
|
|
|
|
Filter: (t1.f1 = 'doh!'::text)
|
|
|
|
(19 rows)
|
|
|
|
|
|
|
|
select * from
|
|
|
|
text_tbl t1
|
|
|
|
inner join int8_tbl i8
|
|
|
|
on i8.q2 = 456
|
|
|
|
right join text_tbl t2
|
|
|
|
on t1.f1 = 'doh!'
|
|
|
|
left join int4_tbl i4
|
|
|
|
on i8.q1 = i4.f1;
|
|
|
|
f1 | q1 | q2 | f1 | f1
|
|
|
|
------+-----+-----+-------------------+----
|
|
|
|
doh! | 123 | 456 | doh! |
|
|
|
|
doh! | 123 | 456 | hi de ho neighbor |
|
|
|
|
(2 rows)
|
|
|
|
|
2015-12-07 23:41:45 +01:00
|
|
|
--
|
|
|
|
-- test for appropriate join order in the presence of lateral references
|
|
|
|
--
|
|
|
|
explain (verbose, costs off)
|
|
|
|
select * from
|
|
|
|
text_tbl t1
|
|
|
|
left join int8_tbl i8
|
|
|
|
on i8.q2 = 123,
|
|
|
|
lateral (select i8.q1, t2.f1 from text_tbl t2 limit 1) as ss
|
|
|
|
where t1.f1 = ss.f1;
|
|
|
|
QUERY PLAN
|
|
|
|
--------------------------------------------------
|
|
|
|
Nested Loop
|
|
|
|
Output: t1.f1, i8.q1, i8.q2, (i8.q1), t2.f1
|
|
|
|
Join Filter: (t1.f1 = t2.f1)
|
|
|
|
-> Nested Loop Left Join
|
|
|
|
Output: t1.f1, i8.q1, i8.q2
|
|
|
|
-> Seq Scan on public.text_tbl t1
|
|
|
|
Output: t1.f1
|
|
|
|
-> Materialize
|
|
|
|
Output: i8.q1, i8.q2
|
|
|
|
-> Seq Scan on public.int8_tbl i8
|
|
|
|
Output: i8.q1, i8.q2
|
|
|
|
Filter: (i8.q2 = 123)
|
|
|
|
-> Limit
|
|
|
|
Output: (i8.q1), t2.f1
|
|
|
|
-> Seq Scan on public.text_tbl t2
|
|
|
|
Output: i8.q1, t2.f1
|
|
|
|
(16 rows)
|
|
|
|
|
|
|
|
select * from
|
|
|
|
text_tbl t1
|
|
|
|
left join int8_tbl i8
|
|
|
|
on i8.q2 = 123,
|
|
|
|
lateral (select i8.q1, t2.f1 from text_tbl t2 limit 1) as ss
|
|
|
|
where t1.f1 = ss.f1;
|
|
|
|
f1 | q1 | q2 | q1 | f1
|
|
|
|
------+------------------+-----+------------------+------
|
|
|
|
doh! | 4567890123456789 | 123 | 4567890123456789 | doh!
|
|
|
|
(1 row)
|
|
|
|
|
Still more fixes for planner's handling of LATERAL references.
More fuzz testing by Andreas Seltenreich exposed that the planner did not
cope well with chains of lateral references. If relation X references Y
laterally, and Y references Z laterally, then we will have to scan X on the
inside of a nestloop with Z, so for all intents and purposes X is laterally
dependent on Z too. The planner did not understand this and would generate
intermediate joins that could not be used. While that was usually harmless
except for wasting some planning cycles, under the right circumstances it
would lead to "failed to build any N-way joins" or "could not devise a
query plan" planner failures.
To fix that, convert the existing per-relation lateral_relids and
lateral_referencers relid sets into their transitive closures; that is,
they now show all relations on which a rel is directly or indirectly
laterally dependent. This not only fixes the chained-reference problem
but allows some of the relevant tests to be made substantially simpler
and faster, since they can be reduced to simple bitmap manipulations
instead of searches of the LateralJoinInfo list.
Also, when a PlaceHolderVar that is due to be evaluated at a join contains
lateral references, we should treat those references as indirect lateral
dependencies of each of the join's base relations. This prevents us from
trying to join any individual base relations to the lateral reference
source before the join is formed, which again cannot work.
Andreas' testing also exposed another oversight in the "dangerous
PlaceHolderVar" test added in commit 85e5e222b1dd02f1. Simply rejecting
unsafe join paths in joinpath.c is insufficient, because in some cases
we will end up rejecting *all* possible paths for a particular join, again
leading to "could not devise a query plan" failures. The restriction has
to be known also to join_is_legal and its cohort functions, so that they
will not select a join for which that will happen. I chose to move the
supporting logic into joinrels.c where the latter functions are.
Back-patch to 9.3 where LATERAL support was introduced.
2015-12-11 20:22:20 +01:00
|
|
|
explain (verbose, costs off)
|
|
|
|
select * from
|
|
|
|
text_tbl t1
|
|
|
|
left join int8_tbl i8
|
|
|
|
on i8.q2 = 123,
|
|
|
|
lateral (select i8.q1, t2.f1 from text_tbl t2 limit 1) as ss1,
|
|
|
|
lateral (select ss1.* from text_tbl t3 limit 1) as ss2
|
|
|
|
where t1.f1 = ss2.f1;
|
|
|
|
QUERY PLAN
|
|
|
|
-------------------------------------------------------------------
|
|
|
|
Nested Loop
|
|
|
|
Output: t1.f1, i8.q1, i8.q2, (i8.q1), t2.f1, ((i8.q1)), (t2.f1)
|
|
|
|
Join Filter: (t1.f1 = (t2.f1))
|
|
|
|
-> Nested Loop
|
|
|
|
Output: t1.f1, i8.q1, i8.q2, (i8.q1), t2.f1
|
|
|
|
-> Nested Loop Left Join
|
|
|
|
Output: t1.f1, i8.q1, i8.q2
|
|
|
|
-> Seq Scan on public.text_tbl t1
|
|
|
|
Output: t1.f1
|
|
|
|
-> Materialize
|
|
|
|
Output: i8.q1, i8.q2
|
|
|
|
-> Seq Scan on public.int8_tbl i8
|
|
|
|
Output: i8.q1, i8.q2
|
|
|
|
Filter: (i8.q2 = 123)
|
|
|
|
-> Limit
|
|
|
|
Output: (i8.q1), t2.f1
|
|
|
|
-> Seq Scan on public.text_tbl t2
|
|
|
|
Output: i8.q1, t2.f1
|
|
|
|
-> Limit
|
|
|
|
Output: ((i8.q1)), (t2.f1)
|
|
|
|
-> Seq Scan on public.text_tbl t3
|
|
|
|
Output: (i8.q1), t2.f1
|
|
|
|
(22 rows)
|
|
|
|
|
|
|
|
select * from
|
|
|
|
text_tbl t1
|
|
|
|
left join int8_tbl i8
|
|
|
|
on i8.q2 = 123,
|
|
|
|
lateral (select i8.q1, t2.f1 from text_tbl t2 limit 1) as ss1,
|
|
|
|
lateral (select ss1.* from text_tbl t3 limit 1) as ss2
|
|
|
|
where t1.f1 = ss2.f1;
|
|
|
|
f1 | q1 | q2 | q1 | f1 | q1 | f1
|
|
|
|
------+------------------+-----+------------------+------+------------------+------
|
|
|
|
doh! | 4567890123456789 | 123 | 4567890123456789 | doh! | 4567890123456789 | doh!
|
|
|
|
(1 row)
|
|
|
|
|
|
|
|
explain (verbose, costs off)
|
|
|
|
select 1 from
|
|
|
|
text_tbl as tt1
|
|
|
|
inner join text_tbl as tt2 on (tt1.f1 = 'foo')
|
|
|
|
left join text_tbl as tt3 on (tt3.f1 = 'foo')
|
|
|
|
left join text_tbl as tt4 on (tt3.f1 = tt4.f1),
|
|
|
|
lateral (select tt4.f1 as c0 from text_tbl as tt5 limit 1) as ss1
|
|
|
|
where tt1.f1 = ss1.c0;
|
|
|
|
QUERY PLAN
|
|
|
|
----------------------------------------------------------
|
|
|
|
Nested Loop
|
|
|
|
Output: 1
|
|
|
|
-> Nested Loop Left Join
|
|
|
|
Output: tt1.f1, tt4.f1
|
|
|
|
-> Nested Loop
|
|
|
|
Output: tt1.f1
|
|
|
|
-> Seq Scan on public.text_tbl tt1
|
|
|
|
Output: tt1.f1
|
|
|
|
Filter: (tt1.f1 = 'foo'::text)
|
|
|
|
-> Seq Scan on public.text_tbl tt2
|
|
|
|
Output: tt2.f1
|
|
|
|
-> Materialize
|
|
|
|
Output: tt4.f1
|
|
|
|
-> Nested Loop Left Join
|
|
|
|
Output: tt4.f1
|
|
|
|
Join Filter: (tt3.f1 = tt4.f1)
|
|
|
|
-> Seq Scan on public.text_tbl tt3
|
|
|
|
Output: tt3.f1
|
|
|
|
Filter: (tt3.f1 = 'foo'::text)
|
|
|
|
-> Seq Scan on public.text_tbl tt4
|
|
|
|
Output: tt4.f1
|
|
|
|
Filter: (tt4.f1 = 'foo'::text)
|
|
|
|
-> Subquery Scan on ss1
|
|
|
|
Output: ss1.c0
|
|
|
|
Filter: (ss1.c0 = 'foo'::text)
|
|
|
|
-> Limit
|
|
|
|
Output: (tt4.f1)
|
|
|
|
-> Seq Scan on public.text_tbl tt5
|
|
|
|
Output: tt4.f1
|
|
|
|
(29 rows)
|
|
|
|
|
|
|
|
select 1 from
|
|
|
|
text_tbl as tt1
|
|
|
|
inner join text_tbl as tt2 on (tt1.f1 = 'foo')
|
|
|
|
left join text_tbl as tt3 on (tt3.f1 = 'foo')
|
|
|
|
left join text_tbl as tt4 on (tt3.f1 = tt4.f1),
|
|
|
|
lateral (select tt4.f1 as c0 from text_tbl as tt5 limit 1) as ss1
|
|
|
|
where tt1.f1 = ss1.c0;
|
|
|
|
?column?
|
|
|
|
----------
|
|
|
|
(0 rows)
|
|
|
|
|
|
|
|
--
|
|
|
|
-- check a case in which a PlaceHolderVar forces join order
|
|
|
|
--
|
|
|
|
explain (verbose, costs off)
|
|
|
|
select ss2.* from
|
|
|
|
int4_tbl i41
|
|
|
|
left join int8_tbl i8
|
|
|
|
join (select i42.f1 as c1, i43.f1 as c2, 42 as c3
|
|
|
|
from int4_tbl i42, int4_tbl i43) ss1
|
|
|
|
on i8.q1 = ss1.c2
|
|
|
|
on i41.f1 = ss1.c1,
|
|
|
|
lateral (select i41.*, i8.*, ss1.* from text_tbl limit 1) ss2
|
|
|
|
where ss1.c2 = 0;
|
|
|
|
QUERY PLAN
|
|
|
|
------------------------------------------------------------------------
|
|
|
|
Nested Loop
|
|
|
|
Output: (i41.f1), (i8.q1), (i8.q2), (i42.f1), (i43.f1), ((42))
|
|
|
|
-> Hash Join
|
|
|
|
Output: i41.f1, i42.f1, i8.q1, i8.q2, i43.f1, 42
|
|
|
|
Hash Cond: (i41.f1 = i42.f1)
|
|
|
|
-> Nested Loop
|
|
|
|
Output: i8.q1, i8.q2, i43.f1, i41.f1
|
|
|
|
-> Nested Loop
|
|
|
|
Output: i8.q1, i8.q2, i43.f1
|
|
|
|
-> Seq Scan on public.int8_tbl i8
|
|
|
|
Output: i8.q1, i8.q2
|
|
|
|
Filter: (i8.q1 = 0)
|
|
|
|
-> Seq Scan on public.int4_tbl i43
|
|
|
|
Output: i43.f1
|
|
|
|
Filter: (i43.f1 = 0)
|
|
|
|
-> Seq Scan on public.int4_tbl i41
|
|
|
|
Output: i41.f1
|
|
|
|
-> Hash
|
|
|
|
Output: i42.f1
|
|
|
|
-> Seq Scan on public.int4_tbl i42
|
|
|
|
Output: i42.f1
|
|
|
|
-> Limit
|
|
|
|
Output: (i41.f1), (i8.q1), (i8.q2), (i42.f1), (i43.f1), ((42))
|
|
|
|
-> Seq Scan on public.text_tbl
|
|
|
|
Output: i41.f1, i8.q1, i8.q2, i42.f1, i43.f1, (42)
|
|
|
|
(25 rows)
|
|
|
|
|
|
|
|
select ss2.* from
|
|
|
|
int4_tbl i41
|
|
|
|
left join int8_tbl i8
|
|
|
|
join (select i42.f1 as c1, i43.f1 as c2, 42 as c3
|
|
|
|
from int4_tbl i42, int4_tbl i43) ss1
|
|
|
|
on i8.q1 = ss1.c2
|
|
|
|
on i41.f1 = ss1.c1,
|
|
|
|
lateral (select i41.*, i8.*, ss1.* from text_tbl limit 1) ss2
|
|
|
|
where ss1.c2 = 0;
|
|
|
|
f1 | q1 | q2 | c1 | c2 | c3
|
|
|
|
----+----+----+----+----+----
|
|
|
|
(0 rows)
|
|
|
|
|
Fix planner failure with full join in RHS of left join.
Given a left join containing a full join in its righthand side, with
the left join's joinclause referencing only one side of the full join
(in a non-strict fashion, so that the full join doesn't get simplified),
the planner could fail with "failed to build any N-way joins" or related
errors. This happened because the full join was seen as overlapping the
left join's RHS, and then recent changes within join_is_legal() caused
that function to conclude that the full join couldn't validly be formed.
Rather than try to rejigger join_is_legal() yet more to allow this,
I think it's better to fix initsplan.c so that the required join order
is explicit in the SpecialJoinInfo data structure. The previous coding
there essentially ignored full joins, relying on the fact that we don't
flatten them in the joinlist data structure to preserve their ordering.
That's sufficient to prevent a wrong plan from being formed, but as this
example shows, it's not sufficient to ensure that the right plan will
be formed. We need to work a bit harder to ensure that the right plan
looks sane according to the SpecialJoinInfos.
Per bug #14105 from Vojtech Rylko. This was apparently induced by
commit 8703059c6 (though now that I've seen it, I wonder whether there
are related cases that could have failed before that); so back-patch
to all active branches. Unfortunately, that patch also went into 9.0,
so this bug is a regression that won't be fixed in that branch.
2016-04-22 02:05:58 +02:00
|
|
|
--
|
|
|
|
-- test successful handling of full join underneath left join (bug #14105)
|
|
|
|
--
|
|
|
|
explain (costs off)
|
|
|
|
select * from
|
|
|
|
(select 1 as id) as xx
|
|
|
|
left join
|
|
|
|
(tenk1 as a1 full join (select 1 as id) as yy on (a1.unique1 = yy.id))
|
|
|
|
on (xx.id = coalesce(yy.id));
|
|
|
|
QUERY PLAN
|
|
|
|
---------------------------------------
|
|
|
|
Nested Loop Left Join
|
|
|
|
Join Filter: ((1) = COALESCE((1)))
|
|
|
|
-> Result
|
|
|
|
-> Hash Full Join
|
|
|
|
Hash Cond: (a1.unique1 = (1))
|
|
|
|
-> Seq Scan on tenk1 a1
|
|
|
|
-> Hash
|
|
|
|
-> Result
|
|
|
|
(8 rows)
|
|
|
|
|
|
|
|
select * from
|
|
|
|
(select 1 as id) as xx
|
|
|
|
left join
|
|
|
|
(tenk1 as a1 full join (select 1 as id) as yy on (a1.unique1 = yy.id))
|
|
|
|
on (xx.id = coalesce(yy.id));
|
|
|
|
id | unique1 | unique2 | two | four | ten | twenty | hundred | thousand | twothousand | fivethous | tenthous | odd | even | stringu1 | stringu2 | string4 | id
|
|
|
|
----+---------+---------+-----+------+-----+--------+---------+----------+-------------+-----------+----------+-----+------+----------+----------+---------+----
|
|
|
|
1 | 1 | 2838 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 2 | 3 | BAAAAA | EFEAAA | OOOOxx | 1
|
|
|
|
(1 row)
|
|
|
|
|
Fix planning of non-strict equivalence clauses above outer joins.
If a potential equivalence clause references a variable from the nullable
side of an outer join, the planner needs to take care that derived clauses
are not pushed to below the outer join; else they may use the wrong value
for the variable. (The problem arises only with non-strict clauses, since
if an upper clause can be proven strict then the outer join will get
simplified to a plain join.) The planner attempted to prevent this type
of error by checking that potential equivalence clauses aren't
outerjoin-delayed as a whole, but actually we have to check each side
separately, since the two sides of the clause will get moved around
separately if it's treated as an equivalence. Bugs of this type can be
demonstrated as far back as 7.4, even though releases before 8.3 had only
a very ad-hoc notion of equivalence clauses.
In addition, we neglected to account for the possibility that such clauses
might have nonempty nullable_relids even when not outerjoin-delayed; so the
equivalence-class machinery lacked logic to compute correct nullable_relids
values for clauses it constructs. This oversight was harmless before 9.2
because we were only using RestrictInfo.nullable_relids for OR clauses;
but as of 9.2 it could result in pushing constructed equivalence clauses
to incorrect places. (This accounts for bug #7604 from Bill MacArthur.)
Fix the first problem by adding a new test check_equivalence_delay() in
distribute_qual_to_rels, and fix the second one by adding code in
equivclass.c and called functions to set correct nullable_relids for
generated clauses. Although I believe the second part of this is not
currently necessary before 9.2, I chose to back-patch it anyway, partly to
keep the logic similar across branches and partly because it seems possible
we might find other reasons why we need valid values of nullable_relids in
the older branches.
Add regression tests illustrating these problems. In 9.0 and up, also
add test cases checking that we can push constants through outer joins,
since we've broken that optimization before and I nearly broke it again
with an overly simplistic patch for this problem.
2012-10-18 18:28:45 +02:00
|
|
|
--
|
|
|
|
-- test ability to push constants through outer join clauses
|
|
|
|
--
|
|
|
|
explain (costs off)
|
|
|
|
select * from int4_tbl a left join tenk1 b on f1 = unique2 where f1 = 0;
|
|
|
|
QUERY PLAN
|
|
|
|
-------------------------------------------------
|
|
|
|
Nested Loop Left Join
|
|
|
|
Join Filter: (a.f1 = b.unique2)
|
|
|
|
-> Seq Scan on int4_tbl a
|
|
|
|
Filter: (f1 = 0)
|
|
|
|
-> Index Scan using tenk1_unique2 on tenk1 b
|
|
|
|
Index Cond: (unique2 = 0)
|
|
|
|
(6 rows)
|
|
|
|
|
|
|
|
explain (costs off)
|
|
|
|
select * from tenk1 a full join tenk1 b using(unique2) where unique2 = 42;
|
|
|
|
QUERY PLAN
|
|
|
|
-------------------------------------------------
|
|
|
|
Merge Full Join
|
|
|
|
Merge Cond: (a.unique2 = b.unique2)
|
|
|
|
-> Index Scan using tenk1_unique2 on tenk1 a
|
|
|
|
Index Cond: (unique2 = 42)
|
|
|
|
-> Index Scan using tenk1_unique2 on tenk1 b
|
|
|
|
Index Cond: (unique2 = 42)
|
|
|
|
(6 rows)
|
|
|
|
|
Fix incorrect matching of subexpressions in outer-join plan nodes.
Previously we would re-use input subexpressions in all expression trees
attached to a Join plan node. However, if it's an outer join and the
subexpression appears in the nullable-side input, this is potentially
incorrect for apparently-matching subexpressions that came from above
the outer join (ie, targetlist and qpqual expressions), because the
executor will treat the subexpression value as NULL when maybe it should
not be.
The case is fairly hard to hit because (a) you need a non-strict
subexpression (else NULL is correct), and (b) we don't usually compute
expressions in the outputs of non-toplevel plan nodes. But we might do
so if the expressions are sort keys for a mergejoin, for example.
Probably in the long run we should make a more explicit distinction between
Vars appearing above and below an outer join, but that will be a major
planner redesign and not at all back-patchable. For the moment, just hack
set_join_references so that it will not match any non-Var expressions
coming from nullable inputs to expressions that came from above the join.
(This is somewhat overkill, in that a strict expression could still be
matched, but it doesn't seem worth the effort to check that.)
Per report from Qingqing Zhou. The added regression test case is based
on his example.
This has been broken for a very long time, so back-patch to all active
branches.
2015-04-05 01:55:15 +02:00
|
|
|
--
|
|
|
|
-- test that quals attached to an outer join have correct semantics,
|
|
|
|
-- specifically that they don't re-use expressions computed below the join;
|
|
|
|
-- we force a mergejoin so that coalesce(b.q1, 1) appears as a join input
|
|
|
|
--
|
|
|
|
set enable_hashjoin to off;
|
|
|
|
set enable_nestloop to off;
|
|
|
|
explain (verbose, costs off)
|
|
|
|
select a.q2, b.q1
|
|
|
|
from int8_tbl a left join int8_tbl b on a.q2 = coalesce(b.q1, 1)
|
|
|
|
where coalesce(b.q1, 1) > 0;
|
|
|
|
QUERY PLAN
|
|
|
|
---------------------------------------------------------
|
|
|
|
Merge Left Join
|
|
|
|
Output: a.q2, b.q1
|
|
|
|
Merge Cond: (a.q2 = (COALESCE(b.q1, '1'::bigint)))
|
|
|
|
Filter: (COALESCE(b.q1, '1'::bigint) > 0)
|
|
|
|
-> Sort
|
|
|
|
Output: a.q2
|
|
|
|
Sort Key: a.q2
|
|
|
|
-> Seq Scan on public.int8_tbl a
|
|
|
|
Output: a.q2
|
|
|
|
-> Sort
|
|
|
|
Output: b.q1, (COALESCE(b.q1, '1'::bigint))
|
|
|
|
Sort Key: (COALESCE(b.q1, '1'::bigint))
|
|
|
|
-> Seq Scan on public.int8_tbl b
|
|
|
|
Output: b.q1, COALESCE(b.q1, '1'::bigint)
|
|
|
|
(14 rows)
|
|
|
|
|
|
|
|
select a.q2, b.q1
|
|
|
|
from int8_tbl a left join int8_tbl b on a.q2 = coalesce(b.q1, 1)
|
|
|
|
where coalesce(b.q1, 1) > 0;
|
|
|
|
q2 | q1
|
|
|
|
-------------------+------------------
|
|
|
|
-4567890123456789 |
|
|
|
|
123 | 123
|
|
|
|
123 | 123
|
|
|
|
456 |
|
|
|
|
4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | 4567890123456789
|
|
|
|
(10 rows)
|
|
|
|
|
|
|
|
reset enable_hashjoin;
|
|
|
|
reset enable_nestloop;
|
2010-03-22 14:57:16 +01:00
|
|
|
--
|
|
|
|
-- test join removal
|
|
|
|
--
|
2010-03-29 00:59:34 +02:00
|
|
|
begin;
|
|
|
|
CREATE TEMP TABLE a (id int PRIMARY KEY, b_id int);
|
|
|
|
CREATE TEMP TABLE b (id int PRIMARY KEY, c_id int);
|
|
|
|
CREATE TEMP TABLE c (id int PRIMARY KEY);
|
2014-07-16 03:12:43 +02:00
|
|
|
CREATE TEMP TABLE d (a int, b int);
|
2010-03-29 00:59:34 +02:00
|
|
|
INSERT INTO a VALUES (0, 0), (1, NULL);
|
|
|
|
INSERT INTO b VALUES (0, 0), (1, NULL);
|
|
|
|
INSERT INTO c VALUES (0), (1);
|
2014-07-16 03:12:43 +02:00
|
|
|
INSERT INTO d VALUES (1,3), (2,2), (3,1);
|
2010-03-29 00:59:34 +02:00
|
|
|
-- all three cases should be optimizable into a simple seqscan
|
|
|
|
explain (costs off) SELECT a.* FROM a LEFT JOIN b ON a.b_id = b.id;
|
|
|
|
QUERY PLAN
|
|
|
|
---------------
|
|
|
|
Seq Scan on a
|
|
|
|
(1 row)
|
|
|
|
|
|
|
|
explain (costs off) SELECT b.* FROM b LEFT JOIN c ON b.c_id = c.id;
|
|
|
|
QUERY PLAN
|
|
|
|
---------------
|
|
|
|
Seq Scan on b
|
|
|
|
(1 row)
|
|
|
|
|
|
|
|
explain (costs off)
|
|
|
|
SELECT a.* FROM a LEFT JOIN (b left join c on b.c_id = c.id)
|
|
|
|
ON (a.b_id = b.id);
|
|
|
|
QUERY PLAN
|
|
|
|
---------------
|
|
|
|
Seq Scan on a
|
|
|
|
(1 row)
|
|
|
|
|
2010-05-23 18:34:38 +02:00
|
|
|
-- check optimization of outer join within another special join
|
|
|
|
explain (costs off)
|
|
|
|
select id from a where id in (
|
|
|
|
select b.id from b left join c on b.id = c.id
|
|
|
|
);
|
|
|
|
QUERY PLAN
|
|
|
|
----------------------------
|
2017-04-08 04:20:03 +02:00
|
|
|
Hash Join
|
2010-05-23 18:34:38 +02:00
|
|
|
Hash Cond: (a.id = b.id)
|
|
|
|
-> Seq Scan on a
|
|
|
|
-> Hash
|
|
|
|
-> Seq Scan on b
|
|
|
|
(5 rows)
|
|
|
|
|
2014-07-16 03:12:43 +02:00
|
|
|
-- check that join removal works for a left join when joining a subquery
|
|
|
|
-- that is guaranteed to be unique by its GROUP BY clause
|
|
|
|
explain (costs off)
|
|
|
|
select d.* from d left join (select * from b group by b.id, b.c_id) s
|
|
|
|
on d.a = s.id and d.b = s.c_id;
|
|
|
|
QUERY PLAN
|
|
|
|
---------------
|
|
|
|
Seq Scan on d
|
|
|
|
(1 row)
|
|
|
|
|
|
|
|
-- similarly, but keying off a DISTINCT clause
|
|
|
|
explain (costs off)
|
|
|
|
select d.* from d left join (select distinct * from b) s
|
|
|
|
on d.a = s.id and d.b = s.c_id;
|
|
|
|
QUERY PLAN
|
|
|
|
---------------
|
|
|
|
Seq Scan on d
|
|
|
|
(1 row)
|
|
|
|
|
|
|
|
-- join removal is not possible when the GROUP BY contains a column that is
|
2016-02-11 23:34:59 +01:00
|
|
|
-- not in the join condition. (Note: as of 9.6, we notice that b.id is a
|
|
|
|
-- primary key and so drop b.c_id from the GROUP BY of the resulting plan;
|
|
|
|
-- but this happens too late for join removal in the outer plan level.)
|
2014-07-16 03:12:43 +02:00
|
|
|
explain (costs off)
|
|
|
|
select d.* from d left join (select * from b group by b.id, b.c_id) s
|
|
|
|
on d.a = s.id;
|
Make the upper part of the planner work by generating and comparing Paths.
I've been saying we needed to do this for more than five years, and here it
finally is. This patch removes the ever-growing tangle of spaghetti logic
that grouping_planner() used to use to try to identify the best plan for
post-scan/join query steps. Now, there is (nearly) independent
consideration of each execution step, and entirely separate construction of
Paths to represent each of the possible ways to do that step. We choose
the best Path or set of Paths using the same add_path() logic that's been
used inside query_planner() for years.
In addition, this patch removes the old restriction that subquery_planner()
could return only a single Plan. It now returns a RelOptInfo containing a
set of Paths, just as query_planner() does, and the parent query level can
use each of those Paths as the basis of a SubqueryScanPath at its level.
This allows finding some optimizations that we missed before, wherein a
subquery was capable of returning presorted data and thereby avoiding a
sort in the parent level, making the overall cost cheaper even though
delivering sorted output was not the cheapest plan for the subquery in
isolation. (A couple of regression test outputs change in consequence of
that. However, there is very little change in visible planner behavior
overall, because the point of this patch is not to get immediate planning
benefits but to create the infrastructure for future improvements.)
There is a great deal left to do here. This patch unblocks a lot of
planner work that was basically impractical in the old code structure,
such as allowing FDWs to implement remote aggregation, or rewriting
plan_set_operations() to allow consideration of multiple implementation
orders for set operations. (The latter will likely require a full
rewrite of plan_set_operations(); what I've done here is only to fix it
to return Paths not Plans.) I have also left unfinished some localized
refactoring in createplan.c and planner.c, because it was not necessary
to get this patch to a working state.
Thanks to Robert Haas, David Rowley, and Amit Kapila for review.
2016-03-07 21:58:22 +01:00
|
|
|
QUERY PLAN
|
|
|
|
------------------------------------------
|
|
|
|
Merge Right Join
|
|
|
|
Merge Cond: (b.id = d.a)
|
|
|
|
-> Group
|
|
|
|
Group Key: b.id
|
|
|
|
-> Index Scan using b_pkey on b
|
2014-07-16 03:12:43 +02:00
|
|
|
-> Sort
|
|
|
|
Sort Key: d.a
|
|
|
|
-> Seq Scan on d
|
Make the upper part of the planner work by generating and comparing Paths.
I've been saying we needed to do this for more than five years, and here it
finally is. This patch removes the ever-growing tangle of spaghetti logic
that grouping_planner() used to use to try to identify the best plan for
post-scan/join query steps. Now, there is (nearly) independent
consideration of each execution step, and entirely separate construction of
Paths to represent each of the possible ways to do that step. We choose
the best Path or set of Paths using the same add_path() logic that's been
used inside query_planner() for years.
In addition, this patch removes the old restriction that subquery_planner()
could return only a single Plan. It now returns a RelOptInfo containing a
set of Paths, just as query_planner() does, and the parent query level can
use each of those Paths as the basis of a SubqueryScanPath at its level.
This allows finding some optimizations that we missed before, wherein a
subquery was capable of returning presorted data and thereby avoiding a
sort in the parent level, making the overall cost cheaper even though
delivering sorted output was not the cheapest plan for the subquery in
isolation. (A couple of regression test outputs change in consequence of
that. However, there is very little change in visible planner behavior
overall, because the point of this patch is not to get immediate planning
benefits but to create the infrastructure for future improvements.)
There is a great deal left to do here. This patch unblocks a lot of
planner work that was basically impractical in the old code structure,
such as allowing FDWs to implement remote aggregation, or rewriting
plan_set_operations() to allow consideration of multiple implementation
orders for set operations. (The latter will likely require a full
rewrite of plan_set_operations(); what I've done here is only to fix it
to return Paths not Plans.) I have also left unfinished some localized
refactoring in createplan.c and planner.c, because it was not necessary
to get this patch to a working state.
Thanks to Robert Haas, David Rowley, and Amit Kapila for review.
2016-03-07 21:58:22 +01:00
|
|
|
(8 rows)
|
2014-07-16 03:12:43 +02:00
|
|
|
|
|
|
|
-- similarly, but keying off a DISTINCT clause
|
|
|
|
explain (costs off)
|
|
|
|
select d.* from d left join (select distinct * from b) s
|
|
|
|
on d.a = s.id;
|
Make the upper part of the planner work by generating and comparing Paths.
I've been saying we needed to do this for more than five years, and here it
finally is. This patch removes the ever-growing tangle of spaghetti logic
that grouping_planner() used to use to try to identify the best plan for
post-scan/join query steps. Now, there is (nearly) independent
consideration of each execution step, and entirely separate construction of
Paths to represent each of the possible ways to do that step. We choose
the best Path or set of Paths using the same add_path() logic that's been
used inside query_planner() for years.
In addition, this patch removes the old restriction that subquery_planner()
could return only a single Plan. It now returns a RelOptInfo containing a
set of Paths, just as query_planner() does, and the parent query level can
use each of those Paths as the basis of a SubqueryScanPath at its level.
This allows finding some optimizations that we missed before, wherein a
subquery was capable of returning presorted data and thereby avoiding a
sort in the parent level, making the overall cost cheaper even though
delivering sorted output was not the cheapest plan for the subquery in
isolation. (A couple of regression test outputs change in consequence of
that. However, there is very little change in visible planner behavior
overall, because the point of this patch is not to get immediate planning
benefits but to create the infrastructure for future improvements.)
There is a great deal left to do here. This patch unblocks a lot of
planner work that was basically impractical in the old code structure,
such as allowing FDWs to implement remote aggregation, or rewriting
plan_set_operations() to allow consideration of multiple implementation
orders for set operations. (The latter will likely require a full
rewrite of plan_set_operations(); what I've done here is only to fix it
to return Paths not Plans.) I have also left unfinished some localized
refactoring in createplan.c and planner.c, because it was not necessary
to get this patch to a working state.
Thanks to Robert Haas, David Rowley, and Amit Kapila for review.
2016-03-07 21:58:22 +01:00
|
|
|
QUERY PLAN
|
|
|
|
--------------------------------------
|
|
|
|
Merge Right Join
|
|
|
|
Merge Cond: (b.id = d.a)
|
|
|
|
-> Unique
|
|
|
|
-> Sort
|
|
|
|
Sort Key: b.id, b.c_id
|
|
|
|
-> Seq Scan on b
|
2014-07-16 03:12:43 +02:00
|
|
|
-> Sort
|
|
|
|
Sort Key: d.a
|
|
|
|
-> Seq Scan on d
|
Make the upper part of the planner work by generating and comparing Paths.
I've been saying we needed to do this for more than five years, and here it
finally is. This patch removes the ever-growing tangle of spaghetti logic
that grouping_planner() used to use to try to identify the best plan for
post-scan/join query steps. Now, there is (nearly) independent
consideration of each execution step, and entirely separate construction of
Paths to represent each of the possible ways to do that step. We choose
the best Path or set of Paths using the same add_path() logic that's been
used inside query_planner() for years.
In addition, this patch removes the old restriction that subquery_planner()
could return only a single Plan. It now returns a RelOptInfo containing a
set of Paths, just as query_planner() does, and the parent query level can
use each of those Paths as the basis of a SubqueryScanPath at its level.
This allows finding some optimizations that we missed before, wherein a
subquery was capable of returning presorted data and thereby avoiding a
sort in the parent level, making the overall cost cheaper even though
delivering sorted output was not the cheapest plan for the subquery in
isolation. (A couple of regression test outputs change in consequence of
that. However, there is very little change in visible planner behavior
overall, because the point of this patch is not to get immediate planning
benefits but to create the infrastructure for future improvements.)
There is a great deal left to do here. This patch unblocks a lot of
planner work that was basically impractical in the old code structure,
such as allowing FDWs to implement remote aggregation, or rewriting
plan_set_operations() to allow consideration of multiple implementation
orders for set operations. (The latter will likely require a full
rewrite of plan_set_operations(); what I've done here is only to fix it
to return Paths not Plans.) I have also left unfinished some localized
refactoring in createplan.c and planner.c, because it was not necessary
to get this patch to a working state.
Thanks to Robert Haas, David Rowley, and Amit Kapila for review.
2016-03-07 21:58:22 +01:00
|
|
|
(9 rows)
|
2014-07-16 03:12:43 +02:00
|
|
|
|
|
|
|
-- check join removal works when uniqueness of the join condition is enforced
|
|
|
|
-- by a UNION
|
|
|
|
explain (costs off)
|
|
|
|
select d.* from d left join (select id from a union select id from b) s
|
|
|
|
on d.a = s.id;
|
|
|
|
QUERY PLAN
|
|
|
|
---------------
|
|
|
|
Seq Scan on d
|
|
|
|
(1 row)
|
|
|
|
|
|
|
|
-- check join removal with a cross-type comparison operator
|
|
|
|
explain (costs off)
|
|
|
|
select i8.* from int8_tbl i8 left join (select f1 from int4_tbl group by f1) i4
|
|
|
|
on i8.q1 = i4.f1;
|
|
|
|
QUERY PLAN
|
|
|
|
-------------------------
|
|
|
|
Seq Scan on int8_tbl i8
|
|
|
|
(1 row)
|
|
|
|
|
2017-09-20 16:20:10 +02:00
|
|
|
-- check join removal with lateral references
|
|
|
|
explain (costs off)
|
|
|
|
select 1 from (select a.id FROM a left join b on a.b_id = b.id) q,
|
|
|
|
lateral generate_series(1, q.id) gs(i) where q.id = gs.i;
|
|
|
|
QUERY PLAN
|
|
|
|
-------------------------------------------
|
|
|
|
Nested Loop
|
|
|
|
-> Seq Scan on a
|
|
|
|
-> Function Scan on generate_series gs
|
|
|
|
Filter: (a.id = i)
|
|
|
|
(4 rows)
|
|
|
|
|
2010-03-29 00:59:34 +02:00
|
|
|
rollback;
|
2010-03-22 14:57:16 +01:00
|
|
|
create temp table parent (k int primary key, pd int);
|
|
|
|
create temp table child (k int unique, cd int);
|
|
|
|
insert into parent values (1, 10), (2, 20), (3, 30);
|
|
|
|
insert into child values (1, 100), (4, 400);
|
|
|
|
-- this case is optimizable
|
|
|
|
select p.* from parent p left join child c on (p.k = c.k);
|
|
|
|
k | pd
|
|
|
|
---+----
|
|
|
|
1 | 10
|
|
|
|
2 | 20
|
|
|
|
3 | 30
|
|
|
|
(3 rows)
|
|
|
|
|
|
|
|
explain (costs off)
|
|
|
|
select p.* from parent p left join child c on (p.k = c.k);
|
|
|
|
QUERY PLAN
|
|
|
|
----------------------
|
|
|
|
Seq Scan on parent p
|
|
|
|
(1 row)
|
|
|
|
|
|
|
|
-- this case is not
|
|
|
|
select p.*, linked from parent p
|
|
|
|
left join (select c.*, true as linked from child c) as ss
|
|
|
|
on (p.k = ss.k);
|
|
|
|
k | pd | linked
|
|
|
|
---+----+--------
|
|
|
|
1 | 10 | t
|
|
|
|
2 | 20 |
|
|
|
|
3 | 30 |
|
|
|
|
(3 rows)
|
|
|
|
|
|
|
|
explain (costs off)
|
|
|
|
select p.*, linked from parent p
|
|
|
|
left join (select c.*, true as linked from child c) as ss
|
|
|
|
on (p.k = ss.k);
|
|
|
|
QUERY PLAN
|
|
|
|
---------------------------------
|
|
|
|
Hash Left Join
|
|
|
|
Hash Cond: (p.k = c.k)
|
|
|
|
-> Seq Scan on parent p
|
|
|
|
-> Hash
|
|
|
|
-> Seq Scan on child c
|
|
|
|
(5 rows)
|
|
|
|
|
2010-09-15 01:15:29 +02:00
|
|
|
-- check for a 9.0rc1 bug: join removal breaks pseudoconstant qual handling
|
|
|
|
select p.* from
|
|
|
|
parent p left join child c on (p.k = c.k)
|
|
|
|
where p.k = 1 and p.k = 2;
|
|
|
|
k | pd
|
|
|
|
---+----
|
|
|
|
(0 rows)
|
|
|
|
|
|
|
|
explain (costs off)
|
|
|
|
select p.* from
|
|
|
|
parent p left join child c on (p.k = c.k)
|
|
|
|
where p.k = 1 and p.k = 2;
|
|
|
|
QUERY PLAN
|
|
|
|
------------------------------------------------
|
|
|
|
Result
|
|
|
|
One-Time Filter: false
|
|
|
|
-> Index Scan using parent_pkey on parent p
|
|
|
|
Index Cond: (k = 1)
|
|
|
|
(4 rows)
|
|
|
|
|
|
|
|
select p.* from
|
|
|
|
(parent p left join child c on (p.k = c.k)) join parent x on p.k = x.k
|
|
|
|
where p.k = 1 and p.k = 2;
|
|
|
|
k | pd
|
|
|
|
---+----
|
|
|
|
(0 rows)
|
|
|
|
|
|
|
|
explain (costs off)
|
|
|
|
select p.* from
|
|
|
|
(parent p left join child c on (p.k = c.k)) join parent x on p.k = x.k
|
|
|
|
where p.k = 1 and p.k = 2;
|
|
|
|
QUERY PLAN
|
|
|
|
--------------------------
|
|
|
|
Result
|
|
|
|
One-Time Filter: false
|
|
|
|
(2 rows)
|
|
|
|
|
2010-03-29 00:59:34 +02:00
|
|
|
-- bug 5255: this is not optimizable by join removal
|
|
|
|
begin;
|
|
|
|
CREATE TEMP TABLE a (id int PRIMARY KEY);
|
|
|
|
CREATE TEMP TABLE b (id int PRIMARY KEY, a_id int);
|
|
|
|
INSERT INTO a VALUES (0), (1);
|
|
|
|
INSERT INTO b VALUES (0, 0), (1, NULL);
|
|
|
|
SELECT * FROM b LEFT JOIN a ON (b.a_id = a.id) WHERE (a.id IS NULL OR a.id > 0);
|
|
|
|
id | a_id | id
|
|
|
|
----+------+----
|
|
|
|
1 | |
|
|
|
|
(1 row)
|
|
|
|
|
|
|
|
SELECT b.* FROM b LEFT JOIN a ON (b.a_id = a.id) WHERE (a.id IS NULL OR a.id > 0);
|
|
|
|
id | a_id
|
|
|
|
----+------
|
|
|
|
1 |
|
|
|
|
(1 row)
|
|
|
|
|
|
|
|
rollback;
|
2010-09-26 01:03:50 +02:00
|
|
|
-- another join removal bug: this is not optimizable, either
|
|
|
|
begin;
|
|
|
|
create temp table innertab (id int8 primary key, dat1 int8);
|
|
|
|
insert into innertab values(123, 42);
|
|
|
|
SELECT * FROM
|
|
|
|
(SELECT 1 AS x) ss1
|
|
|
|
LEFT JOIN
|
|
|
|
(SELECT q1, q2, COALESCE(dat1, q1) AS y
|
|
|
|
FROM int8_tbl LEFT JOIN innertab ON q2 = id) ss2
|
|
|
|
ON true;
|
|
|
|
x | q1 | q2 | y
|
|
|
|
---+------------------+-------------------+------------------
|
|
|
|
1 | 123 | 456 | 123
|
|
|
|
1 | 123 | 4567890123456789 | 123
|
|
|
|
1 | 4567890123456789 | 123 | 42
|
|
|
|
1 | 4567890123456789 | 4567890123456789 | 4567890123456789
|
|
|
|
1 | 4567890123456789 | -4567890123456789 | 4567890123456789
|
|
|
|
(5 rows)
|
|
|
|
|
2015-08-07 04:14:07 +02:00
|
|
|
rollback;
|
|
|
|
-- another join removal bug: we must clean up correctly when removing a PHV
|
|
|
|
begin;
|
|
|
|
create temp table uniquetbl (f1 text unique);
|
|
|
|
explain (costs off)
|
|
|
|
select t1.* from
|
|
|
|
uniquetbl as t1
|
|
|
|
left join (select *, '***'::text as d1 from uniquetbl) t2
|
|
|
|
on t1.f1 = t2.f1
|
|
|
|
left join uniquetbl t3
|
|
|
|
on t2.d1 = t3.f1;
|
|
|
|
QUERY PLAN
|
|
|
|
--------------------------
|
|
|
|
Seq Scan on uniquetbl t1
|
|
|
|
(1 row)
|
|
|
|
|
2015-08-07 20:13:38 +02:00
|
|
|
explain (costs off)
|
|
|
|
select t0.*
|
|
|
|
from
|
|
|
|
text_tbl t0
|
|
|
|
left join
|
|
|
|
(select case t1.ten when 0 then 'doh!'::text else null::text end as case1,
|
|
|
|
t1.stringu2
|
|
|
|
from tenk1 t1
|
|
|
|
join int4_tbl i4 ON i4.f1 = t1.unique2
|
|
|
|
left join uniquetbl u1 ON u1.f1 = t1.string4) ss
|
|
|
|
on t0.f1 = ss.case1
|
|
|
|
where ss.stringu2 !~* ss.case1;
|
|
|
|
QUERY PLAN
|
|
|
|
--------------------------------------------------------------------------------------------
|
|
|
|
Nested Loop
|
|
|
|
Join Filter: (CASE t1.ten WHEN 0 THEN 'doh!'::text ELSE NULL::text END = t0.f1)
|
|
|
|
-> Nested Loop
|
|
|
|
-> Seq Scan on int4_tbl i4
|
|
|
|
-> Index Scan using tenk1_unique2 on tenk1 t1
|
|
|
|
Index Cond: (unique2 = i4.f1)
|
|
|
|
Filter: (stringu2 !~* CASE ten WHEN 0 THEN 'doh!'::text ELSE NULL::text END)
|
|
|
|
-> Materialize
|
|
|
|
-> Seq Scan on text_tbl t0
|
|
|
|
(9 rows)
|
|
|
|
|
|
|
|
select t0.*
|
|
|
|
from
|
|
|
|
text_tbl t0
|
|
|
|
left join
|
|
|
|
(select case t1.ten when 0 then 'doh!'::text else null::text end as case1,
|
|
|
|
t1.stringu2
|
|
|
|
from tenk1 t1
|
|
|
|
join int4_tbl i4 ON i4.f1 = t1.unique2
|
|
|
|
left join uniquetbl u1 ON u1.f1 = t1.string4) ss
|
|
|
|
on t0.f1 = ss.case1
|
|
|
|
where ss.stringu2 !~* ss.case1;
|
|
|
|
f1
|
|
|
|
------
|
|
|
|
doh!
|
|
|
|
(1 row)
|
|
|
|
|
2010-09-26 01:03:50 +02:00
|
|
|
rollback;
|
2013-11-11 16:42:57 +01:00
|
|
|
-- bug #8444: we've historically allowed duplicate aliases within aliased JOINs
|
|
|
|
select * from
|
|
|
|
int8_tbl x join (int4_tbl x cross join int4_tbl y) j on q1 = f1; -- error
|
|
|
|
ERROR: column reference "f1" is ambiguous
|
|
|
|
LINE 2: ..._tbl x join (int4_tbl x cross join int4_tbl y) j on q1 = f1;
|
|
|
|
^
|
|
|
|
select * from
|
|
|
|
int8_tbl x join (int4_tbl x cross join int4_tbl y) j on q1 = y.f1; -- error
|
|
|
|
ERROR: invalid reference to FROM-clause entry for table "y"
|
|
|
|
LINE 2: ...bl x join (int4_tbl x cross join int4_tbl y) j on q1 = y.f1;
|
|
|
|
^
|
|
|
|
HINT: There is an entry for table "y", but it cannot be referenced from this part of the query.
|
|
|
|
select * from
|
|
|
|
int8_tbl x join (int4_tbl x cross join int4_tbl y(ff)) j on q1 = f1; -- ok
|
|
|
|
q1 | q2 | f1 | ff
|
|
|
|
----+----+----+----
|
|
|
|
(0 rows)
|
|
|
|
|
2015-03-11 15:44:04 +01:00
|
|
|
--
|
|
|
|
-- Test hints given on incorrect column references are useful
|
|
|
|
--
|
|
|
|
select t1.uunique1 from
|
2017-02-06 10:33:58 +01:00
|
|
|
tenk1 t1 join tenk2 t2 on t1.two = t2.two; -- error, prefer "t1" suggestion
|
2015-03-11 15:44:04 +01:00
|
|
|
ERROR: column t1.uunique1 does not exist
|
|
|
|
LINE 1: select t1.uunique1 from
|
|
|
|
^
|
2015-11-17 03:16:42 +01:00
|
|
|
HINT: Perhaps you meant to reference the column "t1.unique1".
|
2015-03-11 15:44:04 +01:00
|
|
|
select t2.uunique1 from
|
|
|
|
tenk1 t1 join tenk2 t2 on t1.two = t2.two; -- error, prefer "t2" suggestion
|
|
|
|
ERROR: column t2.uunique1 does not exist
|
|
|
|
LINE 1: select t2.uunique1 from
|
|
|
|
^
|
2015-11-17 03:16:42 +01:00
|
|
|
HINT: Perhaps you meant to reference the column "t2.unique1".
|
2015-03-11 15:44:04 +01:00
|
|
|
select uunique1 from
|
|
|
|
tenk1 t1 join tenk2 t2 on t1.two = t2.two; -- error, suggest both at once
|
|
|
|
ERROR: column "uunique1" does not exist
|
|
|
|
LINE 1: select uunique1 from
|
|
|
|
^
|
2015-11-17 03:16:42 +01:00
|
|
|
HINT: Perhaps you meant to reference the column "t1.unique1" or the column "t2.unique1".
|
2015-03-11 15:44:04 +01:00
|
|
|
--
|
|
|
|
-- Take care to reference the correct RTE
|
|
|
|
--
|
|
|
|
select atts.relid::regclass, s.* from pg_stats s join
|
|
|
|
pg_attribute a on s.attname = a.attname and s.tablename =
|
|
|
|
a.attrelid::regclass::text join (select unnest(indkey) attnum,
|
|
|
|
indexrelid from pg_index i) atts on atts.attnum = a.attnum where
|
|
|
|
schemaname != 'pg_catalog';
|
|
|
|
ERROR: column atts.relid does not exist
|
|
|
|
LINE 1: select atts.relid::regclass, s.* from pg_stats s join
|
|
|
|
^
|
2012-08-08 01:02:54 +02:00
|
|
|
--
|
|
|
|
-- Test LATERAL
|
|
|
|
--
|
|
|
|
select unique2, x.*
|
|
|
|
from tenk1 a, lateral (select * from int4_tbl b where f1 = a.unique1) x;
|
|
|
|
unique2 | f1
|
|
|
|
---------+----
|
|
|
|
9998 | 0
|
|
|
|
(1 row)
|
|
|
|
|
|
|
|
explain (costs off)
|
|
|
|
select unique2, x.*
|
|
|
|
from tenk1 a, lateral (select * from int4_tbl b where f1 = a.unique1) x;
|
2012-08-12 22:01:26 +02:00
|
|
|
QUERY PLAN
|
|
|
|
-------------------------------------------------
|
2012-08-08 01:02:54 +02:00
|
|
|
Nested Loop
|
|
|
|
-> Seq Scan on int4_tbl b
|
2012-08-12 22:01:26 +02:00
|
|
|
-> Index Scan using tenk1_unique1 on tenk1 a
|
|
|
|
Index Cond: (unique1 = b.f1)
|
2012-08-08 01:02:54 +02:00
|
|
|
(4 rows)
|
|
|
|
|
|
|
|
select unique2, x.*
|
|
|
|
from int4_tbl x, lateral (select unique2 from tenk1 where f1 = unique1) ss;
|
|
|
|
unique2 | f1
|
|
|
|
---------+----
|
|
|
|
9998 | 0
|
|
|
|
(1 row)
|
|
|
|
|
|
|
|
explain (costs off)
|
|
|
|
select unique2, x.*
|
|
|
|
from int4_tbl x, lateral (select unique2 from tenk1 where f1 = unique1) ss;
|
|
|
|
QUERY PLAN
|
|
|
|
-----------------------------------------------
|
|
|
|
Nested Loop
|
|
|
|
-> Seq Scan on int4_tbl x
|
|
|
|
-> Index Scan using tenk1_unique1 on tenk1
|
2012-08-12 22:01:26 +02:00
|
|
|
Index Cond: (unique1 = x.f1)
|
2012-08-08 01:02:54 +02:00
|
|
|
(4 rows)
|
|
|
|
|
|
|
|
explain (costs off)
|
|
|
|
select unique2, x.*
|
|
|
|
from int4_tbl x cross join lateral (select unique2 from tenk1 where f1 = unique1) ss;
|
|
|
|
QUERY PLAN
|
|
|
|
-----------------------------------------------
|
|
|
|
Nested Loop
|
|
|
|
-> Seq Scan on int4_tbl x
|
|
|
|
-> Index Scan using tenk1_unique1 on tenk1
|
2012-08-12 22:01:26 +02:00
|
|
|
Index Cond: (unique1 = x.f1)
|
2012-08-08 01:02:54 +02:00
|
|
|
(4 rows)
|
|
|
|
|
|
|
|
select unique2, x.*
|
2012-09-16 23:57:18 +02:00
|
|
|
from int4_tbl x left join lateral (select unique1, unique2 from tenk1 where f1 = unique1) ss on true;
|
2012-08-08 01:02:54 +02:00
|
|
|
unique2 | f1
|
|
|
|
---------+-------------
|
|
|
|
9998 | 0
|
|
|
|
| 123456
|
|
|
|
| -123456
|
|
|
|
| 2147483647
|
|
|
|
| -2147483647
|
|
|
|
(5 rows)
|
|
|
|
|
|
|
|
explain (costs off)
|
|
|
|
select unique2, x.*
|
2012-09-16 23:57:18 +02:00
|
|
|
from int4_tbl x left join lateral (select unique1, unique2 from tenk1 where f1 = unique1) ss on true;
|
2012-08-12 22:01:26 +02:00
|
|
|
QUERY PLAN
|
|
|
|
-----------------------------------------------
|
2012-08-08 01:02:54 +02:00
|
|
|
Nested Loop Left Join
|
|
|
|
-> Seq Scan on int4_tbl x
|
2012-08-12 22:01:26 +02:00
|
|
|
-> Index Scan using tenk1_unique1 on tenk1
|
2013-08-19 19:19:25 +02:00
|
|
|
Index Cond: (x.f1 = unique1)
|
2012-08-12 22:01:26 +02:00
|
|
|
(4 rows)
|
2012-08-08 01:02:54 +02:00
|
|
|
|
|
|
|
-- check scoping of lateral versus parent references
|
|
|
|
-- the first of these should return int8_tbl.q2, the second int8_tbl.q1
|
|
|
|
select *, (select r from (select q1 as q2) x, (select q2 as r) y) from int8_tbl;
|
|
|
|
q1 | q2 | r
|
|
|
|
------------------+-------------------+-------------------
|
|
|
|
123 | 456 | 456
|
|
|
|
123 | 4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | 123 | 123
|
|
|
|
4567890123456789 | 4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | -4567890123456789 | -4567890123456789
|
|
|
|
(5 rows)
|
|
|
|
|
|
|
|
select *, (select r from (select q1 as q2) x, lateral (select q2 as r) y) from int8_tbl;
|
|
|
|
q1 | q2 | r
|
|
|
|
------------------+-------------------+------------------
|
|
|
|
123 | 456 | 123
|
|
|
|
123 | 4567890123456789 | 123
|
|
|
|
4567890123456789 | 123 | 4567890123456789
|
|
|
|
4567890123456789 | 4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | -4567890123456789 | 4567890123456789
|
|
|
|
(5 rows)
|
|
|
|
|
2013-01-26 22:18:42 +01:00
|
|
|
-- lateral with function in FROM
|
2012-08-08 01:02:54 +02:00
|
|
|
select count(*) from tenk1 a, lateral generate_series(1,two) g;
|
|
|
|
count
|
|
|
|
-------
|
|
|
|
5000
|
|
|
|
(1 row)
|
|
|
|
|
|
|
|
explain (costs off)
|
|
|
|
select count(*) from tenk1 a, lateral generate_series(1,two) g;
|
|
|
|
QUERY PLAN
|
|
|
|
------------------------------------------------
|
|
|
|
Aggregate
|
|
|
|
-> Nested Loop
|
|
|
|
-> Seq Scan on tenk1 a
|
|
|
|
-> Function Scan on generate_series g
|
|
|
|
(4 rows)
|
|
|
|
|
|
|
|
explain (costs off)
|
|
|
|
select count(*) from tenk1 a cross join lateral generate_series(1,two) g;
|
|
|
|
QUERY PLAN
|
|
|
|
------------------------------------------------
|
|
|
|
Aggregate
|
|
|
|
-> Nested Loop
|
|
|
|
-> Seq Scan on tenk1 a
|
|
|
|
-> Function Scan on generate_series g
|
|
|
|
(4 rows)
|
|
|
|
|
2013-01-26 22:18:42 +01:00
|
|
|
-- don't need the explicit LATERAL keyword for functions
|
|
|
|
explain (costs off)
|
|
|
|
select count(*) from tenk1 a, generate_series(1,two) g;
|
|
|
|
QUERY PLAN
|
|
|
|
------------------------------------------------
|
|
|
|
Aggregate
|
|
|
|
-> Nested Loop
|
|
|
|
-> Seq Scan on tenk1 a
|
|
|
|
-> Function Scan on generate_series g
|
|
|
|
(4 rows)
|
|
|
|
|
2012-08-12 00:42:20 +02:00
|
|
|
-- lateral with UNION ALL subselect
|
|
|
|
explain (costs off)
|
|
|
|
select * from generate_series(100,200) g,
|
|
|
|
lateral (select * from int8_tbl a where g = q1 union all
|
|
|
|
select * from int8_tbl b where g = q2) ss;
|
|
|
|
QUERY PLAN
|
|
|
|
------------------------------------------
|
|
|
|
Nested Loop
|
|
|
|
-> Function Scan on generate_series g
|
|
|
|
-> Append
|
|
|
|
-> Seq Scan on int8_tbl a
|
|
|
|
Filter: (g.g = q1)
|
|
|
|
-> Seq Scan on int8_tbl b
|
|
|
|
Filter: (g.g = q2)
|
|
|
|
(7 rows)
|
|
|
|
|
|
|
|
select * from generate_series(100,200) g,
|
|
|
|
lateral (select * from int8_tbl a where g = q1 union all
|
|
|
|
select * from int8_tbl b where g = q2) ss;
|
|
|
|
g | q1 | q2
|
|
|
|
-----+------------------+------------------
|
|
|
|
123 | 123 | 456
|
|
|
|
123 | 123 | 4567890123456789
|
|
|
|
123 | 4567890123456789 | 123
|
|
|
|
(3 rows)
|
|
|
|
|
2012-08-12 22:01:26 +02:00
|
|
|
-- lateral with VALUES
|
|
|
|
explain (costs off)
|
|
|
|
select count(*) from tenk1 a,
|
|
|
|
tenk1 b join lateral (values(a.unique1)) ss(x) on b.unique2 = ss.x;
|
2015-03-12 04:18:03 +01:00
|
|
|
QUERY PLAN
|
|
|
|
------------------------------------------------------------
|
|
|
|
Aggregate
|
|
|
|
-> Merge Join
|
|
|
|
Merge Cond: (a.unique1 = b.unique2)
|
|
|
|
-> Index Only Scan using tenk1_unique1 on tenk1 a
|
|
|
|
-> Index Only Scan using tenk1_unique2 on tenk1 b
|
|
|
|
(5 rows)
|
|
|
|
|
|
|
|
select count(*) from tenk1 a,
|
|
|
|
tenk1 b join lateral (values(a.unique1)) ss(x) on b.unique2 = ss.x;
|
|
|
|
count
|
|
|
|
-------
|
|
|
|
10000
|
|
|
|
(1 row)
|
|
|
|
|
|
|
|
-- lateral with VALUES, no flattening possible
|
|
|
|
explain (costs off)
|
|
|
|
select count(*) from tenk1 a,
|
|
|
|
tenk1 b join lateral (values(a.unique1),(-1)) ss(x) on b.unique2 = ss.x;
|
2012-08-12 22:01:26 +02:00
|
|
|
QUERY PLAN
|
|
|
|
------------------------------------------------------------------
|
|
|
|
Aggregate
|
|
|
|
-> Hash Join
|
|
|
|
Hash Cond: ("*VALUES*".column1 = b.unique2)
|
|
|
|
-> Nested Loop
|
|
|
|
-> Index Only Scan using tenk1_unique1 on tenk1 a
|
|
|
|
-> Values Scan on "*VALUES*"
|
|
|
|
-> Hash
|
|
|
|
-> Index Only Scan using tenk1_unique2 on tenk1 b
|
|
|
|
(8 rows)
|
|
|
|
|
|
|
|
select count(*) from tenk1 a,
|
2015-03-12 04:18:03 +01:00
|
|
|
tenk1 b join lateral (values(a.unique1),(-1)) ss(x) on b.unique2 = ss.x;
|
2012-08-12 22:01:26 +02:00
|
|
|
count
|
|
|
|
-------
|
|
|
|
10000
|
|
|
|
(1 row)
|
|
|
|
|
|
|
|
-- lateral injecting a strange outer join condition
|
|
|
|
explain (costs off)
|
|
|
|
select * from int8_tbl a,
|
|
|
|
int8_tbl x left join lateral (select a.q1 from int4_tbl y) ss(z)
|
2017-02-09 01:58:21 +01:00
|
|
|
on x.q2 = ss.z
|
|
|
|
order by a.q1, a.q2, x.q1, x.q2, ss.z;
|
|
|
|
QUERY PLAN
|
|
|
|
------------------------------------------------
|
|
|
|
Sort
|
|
|
|
Sort Key: a.q1, a.q2, x.q1, x.q2, (a.q1)
|
|
|
|
-> Nested Loop
|
|
|
|
-> Seq Scan on int8_tbl a
|
|
|
|
-> Hash Right Join
|
|
|
|
Hash Cond: ((a.q1) = x.q2)
|
|
|
|
-> Seq Scan on int4_tbl y
|
|
|
|
-> Hash
|
|
|
|
-> Seq Scan on int8_tbl x
|
|
|
|
(9 rows)
|
2012-08-12 22:01:26 +02:00
|
|
|
|
|
|
|
select * from int8_tbl a,
|
|
|
|
int8_tbl x left join lateral (select a.q1 from int4_tbl y) ss(z)
|
2017-02-09 01:58:21 +01:00
|
|
|
on x.q2 = ss.z
|
|
|
|
order by a.q1, a.q2, x.q1, x.q2, ss.z;
|
2012-08-12 22:01:26 +02:00
|
|
|
q1 | q2 | q1 | q2 | z
|
|
|
|
------------------+-------------------+------------------+-------------------+------------------
|
2017-02-09 01:58:21 +01:00
|
|
|
123 | 456 | 123 | 456 |
|
|
|
|
123 | 456 | 123 | 4567890123456789 |
|
|
|
|
123 | 456 | 4567890123456789 | -4567890123456789 |
|
2012-08-12 22:01:26 +02:00
|
|
|
123 | 456 | 4567890123456789 | 123 | 123
|
|
|
|
123 | 456 | 4567890123456789 | 123 | 123
|
|
|
|
123 | 456 | 4567890123456789 | 123 | 123
|
|
|
|
123 | 456 | 4567890123456789 | 123 | 123
|
|
|
|
123 | 456 | 4567890123456789 | 123 | 123
|
|
|
|
123 | 456 | 4567890123456789 | 4567890123456789 |
|
2017-02-09 01:58:21 +01:00
|
|
|
123 | 4567890123456789 | 123 | 456 |
|
|
|
|
123 | 4567890123456789 | 123 | 4567890123456789 |
|
|
|
|
123 | 4567890123456789 | 4567890123456789 | -4567890123456789 |
|
2012-08-12 22:01:26 +02:00
|
|
|
123 | 4567890123456789 | 4567890123456789 | 123 | 123
|
|
|
|
123 | 4567890123456789 | 4567890123456789 | 123 | 123
|
|
|
|
123 | 4567890123456789 | 4567890123456789 | 123 | 123
|
|
|
|
123 | 4567890123456789 | 4567890123456789 | 123 | 123
|
|
|
|
123 | 4567890123456789 | 4567890123456789 | 123 | 123
|
|
|
|
123 | 4567890123456789 | 4567890123456789 | 4567890123456789 |
|
2017-02-09 01:58:21 +01:00
|
|
|
4567890123456789 | -4567890123456789 | 123 | 456 |
|
|
|
|
4567890123456789 | -4567890123456789 | 123 | 4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | -4567890123456789 | 123 | 4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | -4567890123456789 | 123 | 4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | -4567890123456789 | 123 | 4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | -4567890123456789 | 123 | 4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | -4567890123456789 | 4567890123456789 | -4567890123456789 |
|
|
|
|
4567890123456789 | -4567890123456789 | 4567890123456789 | 123 |
|
|
|
|
4567890123456789 | -4567890123456789 | 4567890123456789 | 4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | -4567890123456789 | 4567890123456789 | 4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | -4567890123456789 | 4567890123456789 | 4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | -4567890123456789 | 4567890123456789 | 4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | -4567890123456789 | 4567890123456789 | 4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | 123 | 123 | 456 |
|
2012-08-12 22:01:26 +02:00
|
|
|
4567890123456789 | 123 | 123 | 4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | 123 | 123 | 4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | 123 | 123 | 4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | 123 | 123 | 4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | 123 | 123 | 4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | 123 | 4567890123456789 | -4567890123456789 |
|
2017-02-09 01:58:21 +01:00
|
|
|
4567890123456789 | 123 | 4567890123456789 | 123 |
|
|
|
|
4567890123456789 | 123 | 4567890123456789 | 4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | 123 | 4567890123456789 | 4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | 123 | 4567890123456789 | 4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | 123 | 4567890123456789 | 4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | 123 | 4567890123456789 | 4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | 4567890123456789 | 123 | 456 |
|
2012-08-12 22:01:26 +02:00
|
|
|
4567890123456789 | 4567890123456789 | 123 | 4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | 4567890123456789 | 123 | 4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | 4567890123456789 | 123 | 4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | 4567890123456789 | 123 | 4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | 4567890123456789 | 123 | 4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | 4567890123456789 | 4567890123456789 | -4567890123456789 |
|
2017-02-09 01:58:21 +01:00
|
|
|
4567890123456789 | 4567890123456789 | 4567890123456789 | 123 |
|
|
|
|
4567890123456789 | 4567890123456789 | 4567890123456789 | 4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | 4567890123456789 | 4567890123456789 | 4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | 4567890123456789 | 4567890123456789 | 4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | 4567890123456789 | 4567890123456789 | 4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | 4567890123456789 | 4567890123456789 | 4567890123456789 | 4567890123456789
|
2012-08-12 22:01:26 +02:00
|
|
|
(57 rows)
|
|
|
|
|
2012-08-31 23:44:01 +02:00
|
|
|
-- lateral reference to a join alias variable
|
|
|
|
select * from (select f1/2 as x from int4_tbl) ss1 join int4_tbl i4 on x = f1,
|
|
|
|
lateral (select x) ss2(y);
|
|
|
|
x | f1 | y
|
|
|
|
---+----+---
|
|
|
|
0 | 0 | 0
|
|
|
|
(1 row)
|
|
|
|
|
|
|
|
select * from (select f1 as x from int4_tbl) ss1 join int4_tbl i4 on x = f1,
|
|
|
|
lateral (values(x)) ss2(y);
|
|
|
|
x | f1 | y
|
|
|
|
-------------+-------------+-------------
|
|
|
|
0 | 0 | 0
|
|
|
|
123456 | 123456 | 123456
|
|
|
|
-123456 | -123456 | -123456
|
|
|
|
2147483647 | 2147483647 | 2147483647
|
|
|
|
-2147483647 | -2147483647 | -2147483647
|
|
|
|
(5 rows)
|
|
|
|
|
|
|
|
select * from ((select f1/2 as x from int4_tbl) ss1 join int4_tbl i4 on x = f1) j,
|
|
|
|
lateral (select x) ss2(y);
|
|
|
|
x | f1 | y
|
|
|
|
---+----+---
|
|
|
|
0 | 0 | 0
|
|
|
|
(1 row)
|
|
|
|
|
2012-08-18 20:10:17 +02:00
|
|
|
-- lateral references requiring pullup
|
|
|
|
select * from (values(1)) x(lb),
|
|
|
|
lateral generate_series(lb,4) x4;
|
|
|
|
lb | x4
|
|
|
|
----+----
|
|
|
|
1 | 1
|
|
|
|
1 | 2
|
|
|
|
1 | 3
|
|
|
|
1 | 4
|
|
|
|
(4 rows)
|
|
|
|
|
|
|
|
select * from (select f1/1000000000 from int4_tbl) x(lb),
|
|
|
|
lateral generate_series(lb,4) x4;
|
|
|
|
lb | x4
|
|
|
|
----+----
|
|
|
|
0 | 0
|
|
|
|
0 | 1
|
|
|
|
0 | 2
|
|
|
|
0 | 3
|
|
|
|
0 | 4
|
|
|
|
0 | 0
|
|
|
|
0 | 1
|
|
|
|
0 | 2
|
|
|
|
0 | 3
|
|
|
|
0 | 4
|
|
|
|
0 | 0
|
|
|
|
0 | 1
|
|
|
|
0 | 2
|
|
|
|
0 | 3
|
|
|
|
0 | 4
|
|
|
|
2 | 2
|
|
|
|
2 | 3
|
|
|
|
2 | 4
|
|
|
|
-2 | -2
|
|
|
|
-2 | -1
|
|
|
|
-2 | 0
|
|
|
|
-2 | 1
|
|
|
|
-2 | 2
|
|
|
|
-2 | 3
|
|
|
|
-2 | 4
|
|
|
|
(25 rows)
|
|
|
|
|
|
|
|
select * from (values(1)) x(lb),
|
|
|
|
lateral (values(lb)) y(lbcopy);
|
|
|
|
lb | lbcopy
|
|
|
|
----+--------
|
|
|
|
1 | 1
|
|
|
|
(1 row)
|
|
|
|
|
|
|
|
select * from (values(1)) x(lb),
|
|
|
|
lateral (select lb from int4_tbl) y(lbcopy);
|
|
|
|
lb | lbcopy
|
|
|
|
----+--------
|
|
|
|
1 | 1
|
|
|
|
1 | 1
|
|
|
|
1 | 1
|
|
|
|
1 | 1
|
|
|
|
1 | 1
|
|
|
|
(5 rows)
|
|
|
|
|
|
|
|
select * from
|
|
|
|
int8_tbl x left join (select q1,coalesce(q2,0) q2 from int8_tbl) y on x.q2 = y.q1,
|
|
|
|
lateral (values(x.q1,y.q1,y.q2)) v(xq1,yq1,yq2);
|
|
|
|
q1 | q2 | q1 | q2 | xq1 | yq1 | yq2
|
|
|
|
------------------+-------------------+------------------+-------------------+------------------+------------------+-------------------
|
|
|
|
123 | 456 | | | 123 | |
|
|
|
|
123 | 4567890123456789 | 4567890123456789 | -4567890123456789 | 123 | 4567890123456789 | -4567890123456789
|
|
|
|
123 | 4567890123456789 | 4567890123456789 | 4567890123456789 | 123 | 4567890123456789 | 4567890123456789
|
|
|
|
123 | 4567890123456789 | 4567890123456789 | 123 | 123 | 4567890123456789 | 123
|
|
|
|
4567890123456789 | 123 | 123 | 4567890123456789 | 4567890123456789 | 123 | 4567890123456789
|
|
|
|
4567890123456789 | 123 | 123 | 456 | 4567890123456789 | 123 | 456
|
|
|
|
4567890123456789 | 4567890123456789 | 4567890123456789 | -4567890123456789 | 4567890123456789 | 4567890123456789 | -4567890123456789
|
|
|
|
4567890123456789 | 4567890123456789 | 4567890123456789 | 4567890123456789 | 4567890123456789 | 4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | 4567890123456789 | 4567890123456789 | 123 | 4567890123456789 | 4567890123456789 | 123
|
|
|
|
4567890123456789 | -4567890123456789 | | | 4567890123456789 | |
|
|
|
|
(10 rows)
|
|
|
|
|
|
|
|
select * from
|
|
|
|
int8_tbl x left join (select q1,coalesce(q2,0) q2 from int8_tbl) y on x.q2 = y.q1,
|
|
|
|
lateral (select x.q1,y.q1,y.q2) v(xq1,yq1,yq2);
|
|
|
|
q1 | q2 | q1 | q2 | xq1 | yq1 | yq2
|
|
|
|
------------------+-------------------+------------------+-------------------+------------------+------------------+-------------------
|
|
|
|
123 | 456 | | | 123 | |
|
|
|
|
123 | 4567890123456789 | 4567890123456789 | -4567890123456789 | 123 | 4567890123456789 | -4567890123456789
|
|
|
|
123 | 4567890123456789 | 4567890123456789 | 4567890123456789 | 123 | 4567890123456789 | 4567890123456789
|
|
|
|
123 | 4567890123456789 | 4567890123456789 | 123 | 123 | 4567890123456789 | 123
|
|
|
|
4567890123456789 | 123 | 123 | 4567890123456789 | 4567890123456789 | 123 | 4567890123456789
|
|
|
|
4567890123456789 | 123 | 123 | 456 | 4567890123456789 | 123 | 456
|
|
|
|
4567890123456789 | 4567890123456789 | 4567890123456789 | -4567890123456789 | 4567890123456789 | 4567890123456789 | -4567890123456789
|
|
|
|
4567890123456789 | 4567890123456789 | 4567890123456789 | 4567890123456789 | 4567890123456789 | 4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | 4567890123456789 | 4567890123456789 | 123 | 4567890123456789 | 4567890123456789 | 123
|
|
|
|
4567890123456789 | -4567890123456789 | | | 4567890123456789 | |
|
|
|
|
(10 rows)
|
|
|
|
|
2012-08-27 04:48:55 +02:00
|
|
|
select x.* from
|
|
|
|
int8_tbl x left join (select q1,coalesce(q2,0) q2 from int8_tbl) y on x.q2 = y.q1,
|
|
|
|
lateral (select x.q1,y.q1,y.q2) v(xq1,yq1,yq2);
|
|
|
|
q1 | q2
|
|
|
|
------------------+-------------------
|
|
|
|
123 | 456
|
|
|
|
123 | 4567890123456789
|
|
|
|
123 | 4567890123456789
|
|
|
|
123 | 4567890123456789
|
|
|
|
4567890123456789 | 123
|
|
|
|
4567890123456789 | 123
|
|
|
|
4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | -4567890123456789
|
|
|
|
(10 rows)
|
|
|
|
|
|
|
|
select v.* from
|
|
|
|
(int8_tbl x left join (select q1,coalesce(q2,0) q2 from int8_tbl) y on x.q2 = y.q1)
|
|
|
|
left join int4_tbl z on z.f1 = x.q2,
|
|
|
|
lateral (select x.q1,y.q1 union all select x.q2,y.q2) v(vx,vy);
|
|
|
|
vx | vy
|
|
|
|
-------------------+-------------------
|
|
|
|
123 |
|
|
|
|
456 |
|
|
|
|
123 | 4567890123456789
|
|
|
|
4567890123456789 | -4567890123456789
|
|
|
|
123 | 4567890123456789
|
|
|
|
4567890123456789 | 4567890123456789
|
|
|
|
123 | 4567890123456789
|
|
|
|
4567890123456789 | 123
|
|
|
|
4567890123456789 | 123
|
|
|
|
123 | 4567890123456789
|
|
|
|
4567890123456789 | 123
|
|
|
|
123 | 456
|
|
|
|
4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | -4567890123456789
|
|
|
|
4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | 123
|
|
|
|
4567890123456789 |
|
|
|
|
-4567890123456789 |
|
|
|
|
(20 rows)
|
|
|
|
|
|
|
|
select v.* from
|
|
|
|
(int8_tbl x left join (select q1,(select coalesce(q2,0)) q2 from int8_tbl) y on x.q2 = y.q1)
|
|
|
|
left join int4_tbl z on z.f1 = x.q2,
|
|
|
|
lateral (select x.q1,y.q1 union all select x.q2,y.q2) v(vx,vy);
|
|
|
|
vx | vy
|
|
|
|
-------------------+-------------------
|
Add an explicit representation of the output targetlist to Paths.
Up to now, there's been an assumption that all Paths for a given relation
compute the same output column set (targetlist). However, there are good
reasons to remove that assumption. For example, an indexscan on an
expression index might be able to return the value of an expensive function
"for free". While we have the ability to generate such a plan today in
simple cases, we don't have a way to model that it's cheaper than a plan
that computes the function from scratch, nor a way to create such a plan
in join cases (where the function computation would normally happen at
the topmost join node). Also, we need this so that we can have Paths
representing post-scan/join steps, where the targetlist may well change
from one step to the next. Therefore, invent a "struct PathTarget"
representing the columns we expect a plan step to emit. It's convenient
to include the output tuple width and tlist evaluation cost in this struct,
and there will likely be additional fields in future.
While Path nodes that actually do have custom outputs will need their own
PathTargets, it will still be true that most Paths for a given relation
will compute the same tlist. To reduce the overhead added by this patch,
keep a "default PathTarget" in RelOptInfo, and allow Paths that compute
that column set to just point to their parent RelOptInfo's reltarget.
(In the patch as committed, actually every Path is like that, since we
do not yet have any cases of custom PathTargets.)
I took this opportunity to provide some more-honest costing of
PlaceHolderVar evaluation. Up to now, the assumption that "scan/join
reltargetlists have cost zero" was applied not only to Vars, where it's
reasonable, but also PlaceHolderVars where it isn't. Now, we add the eval
cost of a PlaceHolderVar's expression to the first plan level where it can
be computed, by including it in the PathTarget cost field and adding that
to the cost estimates for Paths. This isn't perfect yet but it's much
better than before, and there is a way forward to improve it more. This
costing change affects the join order chosen for a couple of the regression
tests, changing expected row ordering.
2016-02-19 02:01:49 +01:00
|
|
|
4567890123456789 | 123
|
|
|
|
123 | 456
|
|
|
|
4567890123456789 | 123
|
2012-08-27 04:48:55 +02:00
|
|
|
123 | 4567890123456789
|
|
|
|
4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | 123
|
|
|
|
123 | 4567890123456789
|
|
|
|
4567890123456789 | 123
|
|
|
|
4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | 4567890123456789
|
Add an explicit representation of the output targetlist to Paths.
Up to now, there's been an assumption that all Paths for a given relation
compute the same output column set (targetlist). However, there are good
reasons to remove that assumption. For example, an indexscan on an
expression index might be able to return the value of an expensive function
"for free". While we have the ability to generate such a plan today in
simple cases, we don't have a way to model that it's cheaper than a plan
that computes the function from scratch, nor a way to create such a plan
in join cases (where the function computation would normally happen at
the topmost join node). Also, we need this so that we can have Paths
representing post-scan/join steps, where the targetlist may well change
from one step to the next. Therefore, invent a "struct PathTarget"
representing the columns we expect a plan step to emit. It's convenient
to include the output tuple width and tlist evaluation cost in this struct,
and there will likely be additional fields in future.
While Path nodes that actually do have custom outputs will need their own
PathTargets, it will still be true that most Paths for a given relation
will compute the same tlist. To reduce the overhead added by this patch,
keep a "default PathTarget" in RelOptInfo, and allow Paths that compute
that column set to just point to their parent RelOptInfo's reltarget.
(In the patch as committed, actually every Path is like that, since we
do not yet have any cases of custom PathTargets.)
I took this opportunity to provide some more-honest costing of
PlaceHolderVar evaluation. Up to now, the assumption that "scan/join
reltargetlists have cost zero" was applied not only to Vars, where it's
reasonable, but also PlaceHolderVars where it isn't. Now, we add the eval
cost of a PlaceHolderVar's expression to the first plan level where it can
be computed, by including it in the PathTarget cost field and adding that
to the cost estimates for Paths. This isn't perfect yet but it's much
better than before, and there is a way forward to improve it more. This
costing change affects the join order chosen for a couple of the regression
tests, changing expected row ordering.
2016-02-19 02:01:49 +01:00
|
|
|
123 | 4567890123456789
|
2012-08-27 04:48:55 +02:00
|
|
|
4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | 4567890123456789
|
Add an explicit representation of the output targetlist to Paths.
Up to now, there's been an assumption that all Paths for a given relation
compute the same output column set (targetlist). However, there are good
reasons to remove that assumption. For example, an indexscan on an
expression index might be able to return the value of an expensive function
"for free". While we have the ability to generate such a plan today in
simple cases, we don't have a way to model that it's cheaper than a plan
that computes the function from scratch, nor a way to create such a plan
in join cases (where the function computation would normally happen at
the topmost join node). Also, we need this so that we can have Paths
representing post-scan/join steps, where the targetlist may well change
from one step to the next. Therefore, invent a "struct PathTarget"
representing the columns we expect a plan step to emit. It's convenient
to include the output tuple width and tlist evaluation cost in this struct,
and there will likely be additional fields in future.
While Path nodes that actually do have custom outputs will need their own
PathTargets, it will still be true that most Paths for a given relation
will compute the same tlist. To reduce the overhead added by this patch,
keep a "default PathTarget" in RelOptInfo, and allow Paths that compute
that column set to just point to their parent RelOptInfo's reltarget.
(In the patch as committed, actually every Path is like that, since we
do not yet have any cases of custom PathTargets.)
I took this opportunity to provide some more-honest costing of
PlaceHolderVar evaluation. Up to now, the assumption that "scan/join
reltargetlists have cost zero" was applied not only to Vars, where it's
reasonable, but also PlaceHolderVars where it isn't. Now, we add the eval
cost of a PlaceHolderVar's expression to the first plan level where it can
be computed, by including it in the PathTarget cost field and adding that
to the cost estimates for Paths. This isn't perfect yet but it's much
better than before, and there is a way forward to improve it more. This
costing change affects the join order chosen for a couple of the regression
tests, changing expected row ordering.
2016-02-19 02:01:49 +01:00
|
|
|
4567890123456789 | -4567890123456789
|
|
|
|
123 | 4567890123456789
|
|
|
|
4567890123456789 | -4567890123456789
|
|
|
|
123 |
|
|
|
|
456 |
|
2012-08-27 04:48:55 +02:00
|
|
|
4567890123456789 |
|
|
|
|
-4567890123456789 |
|
|
|
|
(20 rows)
|
|
|
|
|
|
|
|
create temp table dual();
|
|
|
|
insert into dual default values;
|
2012-09-02 00:16:24 +02:00
|
|
|
analyze dual;
|
2012-08-27 04:48:55 +02:00
|
|
|
select v.* from
|
|
|
|
(int8_tbl x left join (select q1,(select coalesce(q2,0)) q2 from int8_tbl) y on x.q2 = y.q1)
|
|
|
|
left join int4_tbl z on z.f1 = x.q2,
|
|
|
|
lateral (select x.q1,y.q1 from dual union all select x.q2,y.q2 from dual) v(vx,vy);
|
|
|
|
vx | vy
|
|
|
|
-------------------+-------------------
|
Add an explicit representation of the output targetlist to Paths.
Up to now, there's been an assumption that all Paths for a given relation
compute the same output column set (targetlist). However, there are good
reasons to remove that assumption. For example, an indexscan on an
expression index might be able to return the value of an expensive function
"for free". While we have the ability to generate such a plan today in
simple cases, we don't have a way to model that it's cheaper than a plan
that computes the function from scratch, nor a way to create such a plan
in join cases (where the function computation would normally happen at
the topmost join node). Also, we need this so that we can have Paths
representing post-scan/join steps, where the targetlist may well change
from one step to the next. Therefore, invent a "struct PathTarget"
representing the columns we expect a plan step to emit. It's convenient
to include the output tuple width and tlist evaluation cost in this struct,
and there will likely be additional fields in future.
While Path nodes that actually do have custom outputs will need their own
PathTargets, it will still be true that most Paths for a given relation
will compute the same tlist. To reduce the overhead added by this patch,
keep a "default PathTarget" in RelOptInfo, and allow Paths that compute
that column set to just point to their parent RelOptInfo's reltarget.
(In the patch as committed, actually every Path is like that, since we
do not yet have any cases of custom PathTargets.)
I took this opportunity to provide some more-honest costing of
PlaceHolderVar evaluation. Up to now, the assumption that "scan/join
reltargetlists have cost zero" was applied not only to Vars, where it's
reasonable, but also PlaceHolderVars where it isn't. Now, we add the eval
cost of a PlaceHolderVar's expression to the first plan level where it can
be computed, by including it in the PathTarget cost field and adding that
to the cost estimates for Paths. This isn't perfect yet but it's much
better than before, and there is a way forward to improve it more. This
costing change affects the join order chosen for a couple of the regression
tests, changing expected row ordering.
2016-02-19 02:01:49 +01:00
|
|
|
4567890123456789 | 123
|
|
|
|
123 | 456
|
|
|
|
4567890123456789 | 123
|
2012-08-27 04:48:55 +02:00
|
|
|
123 | 4567890123456789
|
|
|
|
4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | 123
|
|
|
|
123 | 4567890123456789
|
|
|
|
4567890123456789 | 123
|
|
|
|
4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | 4567890123456789
|
Add an explicit representation of the output targetlist to Paths.
Up to now, there's been an assumption that all Paths for a given relation
compute the same output column set (targetlist). However, there are good
reasons to remove that assumption. For example, an indexscan on an
expression index might be able to return the value of an expensive function
"for free". While we have the ability to generate such a plan today in
simple cases, we don't have a way to model that it's cheaper than a plan
that computes the function from scratch, nor a way to create such a plan
in join cases (where the function computation would normally happen at
the topmost join node). Also, we need this so that we can have Paths
representing post-scan/join steps, where the targetlist may well change
from one step to the next. Therefore, invent a "struct PathTarget"
representing the columns we expect a plan step to emit. It's convenient
to include the output tuple width and tlist evaluation cost in this struct,
and there will likely be additional fields in future.
While Path nodes that actually do have custom outputs will need their own
PathTargets, it will still be true that most Paths for a given relation
will compute the same tlist. To reduce the overhead added by this patch,
keep a "default PathTarget" in RelOptInfo, and allow Paths that compute
that column set to just point to their parent RelOptInfo's reltarget.
(In the patch as committed, actually every Path is like that, since we
do not yet have any cases of custom PathTargets.)
I took this opportunity to provide some more-honest costing of
PlaceHolderVar evaluation. Up to now, the assumption that "scan/join
reltargetlists have cost zero" was applied not only to Vars, where it's
reasonable, but also PlaceHolderVars where it isn't. Now, we add the eval
cost of a PlaceHolderVar's expression to the first plan level where it can
be computed, by including it in the PathTarget cost field and adding that
to the cost estimates for Paths. This isn't perfect yet but it's much
better than before, and there is a way forward to improve it more. This
costing change affects the join order chosen for a couple of the regression
tests, changing expected row ordering.
2016-02-19 02:01:49 +01:00
|
|
|
123 | 4567890123456789
|
2012-08-27 04:48:55 +02:00
|
|
|
4567890123456789 | 4567890123456789
|
2012-09-02 00:16:24 +02:00
|
|
|
4567890123456789 | 4567890123456789
|
Add an explicit representation of the output targetlist to Paths.
Up to now, there's been an assumption that all Paths for a given relation
compute the same output column set (targetlist). However, there are good
reasons to remove that assumption. For example, an indexscan on an
expression index might be able to return the value of an expensive function
"for free". While we have the ability to generate such a plan today in
simple cases, we don't have a way to model that it's cheaper than a plan
that computes the function from scratch, nor a way to create such a plan
in join cases (where the function computation would normally happen at
the topmost join node). Also, we need this so that we can have Paths
representing post-scan/join steps, where the targetlist may well change
from one step to the next. Therefore, invent a "struct PathTarget"
representing the columns we expect a plan step to emit. It's convenient
to include the output tuple width and tlist evaluation cost in this struct,
and there will likely be additional fields in future.
While Path nodes that actually do have custom outputs will need their own
PathTargets, it will still be true that most Paths for a given relation
will compute the same tlist. To reduce the overhead added by this patch,
keep a "default PathTarget" in RelOptInfo, and allow Paths that compute
that column set to just point to their parent RelOptInfo's reltarget.
(In the patch as committed, actually every Path is like that, since we
do not yet have any cases of custom PathTargets.)
I took this opportunity to provide some more-honest costing of
PlaceHolderVar evaluation. Up to now, the assumption that "scan/join
reltargetlists have cost zero" was applied not only to Vars, where it's
reasonable, but also PlaceHolderVars where it isn't. Now, we add the eval
cost of a PlaceHolderVar's expression to the first plan level where it can
be computed, by including it in the PathTarget cost field and adding that
to the cost estimates for Paths. This isn't perfect yet but it's much
better than before, and there is a way forward to improve it more. This
costing change affects the join order chosen for a couple of the regression
tests, changing expected row ordering.
2016-02-19 02:01:49 +01:00
|
|
|
4567890123456789 | -4567890123456789
|
|
|
|
123 | 4567890123456789
|
|
|
|
4567890123456789 | -4567890123456789
|
|
|
|
123 |
|
|
|
|
456 |
|
2012-08-27 04:48:55 +02:00
|
|
|
4567890123456789 |
|
|
|
|
-4567890123456789 |
|
|
|
|
(20 rows)
|
|
|
|
|
2013-08-18 02:22:37 +02:00
|
|
|
explain (verbose, costs off)
|
|
|
|
select * from
|
|
|
|
int8_tbl a left join
|
|
|
|
lateral (select *, a.q2 as x from int8_tbl b) ss on a.q2 = ss.q1;
|
|
|
|
QUERY PLAN
|
|
|
|
------------------------------------------
|
|
|
|
Nested Loop Left Join
|
|
|
|
Output: a.q1, a.q2, b.q1, b.q2, (a.q2)
|
|
|
|
-> Seq Scan on public.int8_tbl a
|
|
|
|
Output: a.q1, a.q2
|
|
|
|
-> Seq Scan on public.int8_tbl b
|
|
|
|
Output: b.q1, b.q2, a.q2
|
|
|
|
Filter: (a.q2 = b.q1)
|
|
|
|
(7 rows)
|
|
|
|
|
|
|
|
select * from
|
|
|
|
int8_tbl a left join
|
|
|
|
lateral (select *, a.q2 as x from int8_tbl b) ss on a.q2 = ss.q1;
|
|
|
|
q1 | q2 | q1 | q2 | x
|
|
|
|
------------------+-------------------+------------------+-------------------+------------------
|
|
|
|
123 | 456 | | |
|
|
|
|
123 | 4567890123456789 | 4567890123456789 | 123 | 4567890123456789
|
|
|
|
123 | 4567890123456789 | 4567890123456789 | 4567890123456789 | 4567890123456789
|
|
|
|
123 | 4567890123456789 | 4567890123456789 | -4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | 123 | 123 | 456 | 123
|
|
|
|
4567890123456789 | 123 | 123 | 4567890123456789 | 123
|
|
|
|
4567890123456789 | 4567890123456789 | 4567890123456789 | 123 | 4567890123456789
|
|
|
|
4567890123456789 | 4567890123456789 | 4567890123456789 | 4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | 4567890123456789 | 4567890123456789 | -4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | -4567890123456789 | | |
|
|
|
|
(10 rows)
|
|
|
|
|
|
|
|
explain (verbose, costs off)
|
|
|
|
select * from
|
|
|
|
int8_tbl a left join
|
|
|
|
lateral (select *, coalesce(a.q2, 42) as x from int8_tbl b) ss on a.q2 = ss.q1;
|
2015-03-30 20:59:49 +02:00
|
|
|
QUERY PLAN
|
|
|
|
------------------------------------------------------------------
|
2013-08-18 02:22:37 +02:00
|
|
|
Nested Loop Left Join
|
2015-03-30 20:59:49 +02:00
|
|
|
Output: a.q1, a.q2, b.q1, b.q2, (COALESCE(a.q2, '42'::bigint))
|
2013-08-18 02:22:37 +02:00
|
|
|
-> Seq Scan on public.int8_tbl a
|
|
|
|
Output: a.q1, a.q2
|
|
|
|
-> Seq Scan on public.int8_tbl b
|
2015-03-30 20:59:49 +02:00
|
|
|
Output: b.q1, b.q2, COALESCE(a.q2, '42'::bigint)
|
2013-08-18 02:22:37 +02:00
|
|
|
Filter: (a.q2 = b.q1)
|
|
|
|
(7 rows)
|
|
|
|
|
|
|
|
select * from
|
|
|
|
int8_tbl a left join
|
|
|
|
lateral (select *, coalesce(a.q2, 42) as x from int8_tbl b) ss on a.q2 = ss.q1;
|
|
|
|
q1 | q2 | q1 | q2 | x
|
|
|
|
------------------+-------------------+------------------+-------------------+------------------
|
|
|
|
123 | 456 | | |
|
|
|
|
123 | 4567890123456789 | 4567890123456789 | 123 | 4567890123456789
|
|
|
|
123 | 4567890123456789 | 4567890123456789 | 4567890123456789 | 4567890123456789
|
|
|
|
123 | 4567890123456789 | 4567890123456789 | -4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | 123 | 123 | 456 | 123
|
|
|
|
4567890123456789 | 123 | 123 | 4567890123456789 | 123
|
|
|
|
4567890123456789 | 4567890123456789 | 4567890123456789 | 123 | 4567890123456789
|
|
|
|
4567890123456789 | 4567890123456789 | 4567890123456789 | 4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | 4567890123456789 | 4567890123456789 | -4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | -4567890123456789 | | |
|
|
|
|
(10 rows)
|
|
|
|
|
|
|
|
-- lateral can result in join conditions appearing below their
|
|
|
|
-- real semantic level
|
|
|
|
explain (verbose, costs off)
|
|
|
|
select * from int4_tbl i left join
|
|
|
|
lateral (select * from int2_tbl j where i.f1 = j.f1) k on true;
|
|
|
|
QUERY PLAN
|
|
|
|
-------------------------------------------
|
2013-08-19 19:19:25 +02:00
|
|
|
Hash Left Join
|
2013-08-18 02:22:37 +02:00
|
|
|
Output: i.f1, j.f1
|
2013-08-19 19:19:25 +02:00
|
|
|
Hash Cond: (i.f1 = j.f1)
|
2013-08-18 02:22:37 +02:00
|
|
|
-> Seq Scan on public.int4_tbl i
|
|
|
|
Output: i.f1
|
2013-08-19 19:19:25 +02:00
|
|
|
-> Hash
|
2013-08-18 02:22:37 +02:00
|
|
|
Output: j.f1
|
|
|
|
-> Seq Scan on public.int2_tbl j
|
|
|
|
Output: j.f1
|
|
|
|
(9 rows)
|
|
|
|
|
|
|
|
select * from int4_tbl i left join
|
|
|
|
lateral (select * from int2_tbl j where i.f1 = j.f1) k on true;
|
2013-08-19 19:19:25 +02:00
|
|
|
f1 | f1
|
|
|
|
-------------+----
|
|
|
|
0 | 0
|
|
|
|
123456 |
|
|
|
|
-123456 |
|
|
|
|
2147483647 |
|
|
|
|
-2147483647 |
|
|
|
|
(5 rows)
|
2013-08-18 02:22:37 +02:00
|
|
|
|
|
|
|
explain (verbose, costs off)
|
|
|
|
select * from int4_tbl i left join
|
|
|
|
lateral (select coalesce(i) from int2_tbl j where i.f1 = j.f1) k on true;
|
|
|
|
QUERY PLAN
|
|
|
|
-------------------------------------
|
|
|
|
Nested Loop Left Join
|
|
|
|
Output: i.f1, (COALESCE(i.*))
|
|
|
|
-> Seq Scan on public.int4_tbl i
|
|
|
|
Output: i.f1, i.*
|
|
|
|
-> Seq Scan on public.int2_tbl j
|
|
|
|
Output: j.f1, COALESCE(i.*)
|
|
|
|
Filter: (i.f1 = j.f1)
|
|
|
|
(7 rows)
|
|
|
|
|
|
|
|
select * from int4_tbl i left join
|
|
|
|
lateral (select coalesce(i) from int2_tbl j where i.f1 = j.f1) k on true;
|
|
|
|
f1 | coalesce
|
|
|
|
-------------+----------
|
|
|
|
0 | (0)
|
|
|
|
123456 |
|
|
|
|
-123456 |
|
|
|
|
2147483647 |
|
|
|
|
-2147483647 |
|
|
|
|
(5 rows)
|
|
|
|
|
2013-08-19 19:19:25 +02:00
|
|
|
explain (verbose, costs off)
|
|
|
|
select * from int4_tbl a,
|
|
|
|
lateral (
|
|
|
|
select * from int4_tbl b left join int8_tbl c on (b.f1 = q1 and a.f1 = q2)
|
|
|
|
) ss;
|
|
|
|
QUERY PLAN
|
|
|
|
-------------------------------------------------
|
|
|
|
Nested Loop
|
|
|
|
Output: a.f1, b.f1, c.q1, c.q2
|
|
|
|
-> Seq Scan on public.int4_tbl a
|
|
|
|
Output: a.f1
|
|
|
|
-> Hash Left Join
|
|
|
|
Output: b.f1, c.q1, c.q2
|
|
|
|
Hash Cond: (b.f1 = c.q1)
|
|
|
|
-> Seq Scan on public.int4_tbl b
|
|
|
|
Output: b.f1
|
|
|
|
-> Hash
|
|
|
|
Output: c.q1, c.q2
|
|
|
|
-> Seq Scan on public.int8_tbl c
|
|
|
|
Output: c.q1, c.q2
|
|
|
|
Filter: (a.f1 = c.q2)
|
|
|
|
(14 rows)
|
|
|
|
|
|
|
|
select * from int4_tbl a,
|
|
|
|
lateral (
|
|
|
|
select * from int4_tbl b left join int8_tbl c on (b.f1 = q1 and a.f1 = q2)
|
|
|
|
) ss;
|
|
|
|
f1 | f1 | q1 | q2
|
|
|
|
-------------+-------------+----+----
|
|
|
|
0 | 0 | |
|
|
|
|
0 | 123456 | |
|
|
|
|
0 | -123456 | |
|
|
|
|
0 | 2147483647 | |
|
|
|
|
0 | -2147483647 | |
|
|
|
|
123456 | 0 | |
|
|
|
|
123456 | 123456 | |
|
|
|
|
123456 | -123456 | |
|
|
|
|
123456 | 2147483647 | |
|
|
|
|
123456 | -2147483647 | |
|
|
|
|
-123456 | 0 | |
|
|
|
|
-123456 | 123456 | |
|
|
|
|
-123456 | -123456 | |
|
|
|
|
-123456 | 2147483647 | |
|
|
|
|
-123456 | -2147483647 | |
|
|
|
|
2147483647 | 0 | |
|
|
|
|
2147483647 | 123456 | |
|
|
|
|
2147483647 | -123456 | |
|
|
|
|
2147483647 | 2147483647 | |
|
|
|
|
2147483647 | -2147483647 | |
|
|
|
|
-2147483647 | 0 | |
|
|
|
|
-2147483647 | 123456 | |
|
|
|
|
-2147483647 | -123456 | |
|
|
|
|
-2147483647 | 2147483647 | |
|
|
|
|
-2147483647 | -2147483647 | |
|
|
|
|
(25 rows)
|
|
|
|
|
2013-08-18 02:22:37 +02:00
|
|
|
-- lateral reference in a PlaceHolderVar evaluated at join level
|
|
|
|
explain (verbose, costs off)
|
|
|
|
select * from
|
|
|
|
int8_tbl a left join lateral
|
|
|
|
(select b.q1 as bq1, c.q1 as cq1, least(a.q1,b.q1,c.q1) from
|
|
|
|
int8_tbl b cross join int8_tbl c) ss
|
|
|
|
on a.q2 = ss.bq1;
|
|
|
|
QUERY PLAN
|
|
|
|
-------------------------------------------------------------
|
|
|
|
Nested Loop Left Join
|
|
|
|
Output: a.q1, a.q2, b.q1, c.q1, (LEAST(a.q1, b.q1, c.q1))
|
|
|
|
-> Seq Scan on public.int8_tbl a
|
|
|
|
Output: a.q1, a.q2
|
|
|
|
-> Nested Loop
|
|
|
|
Output: b.q1, c.q1, LEAST(a.q1, b.q1, c.q1)
|
|
|
|
-> Seq Scan on public.int8_tbl b
|
|
|
|
Output: b.q1, b.q2
|
Still more fixes for planner's handling of LATERAL references.
More fuzz testing by Andreas Seltenreich exposed that the planner did not
cope well with chains of lateral references. If relation X references Y
laterally, and Y references Z laterally, then we will have to scan X on the
inside of a nestloop with Z, so for all intents and purposes X is laterally
dependent on Z too. The planner did not understand this and would generate
intermediate joins that could not be used. While that was usually harmless
except for wasting some planning cycles, under the right circumstances it
would lead to "failed to build any N-way joins" or "could not devise a
query plan" planner failures.
To fix that, convert the existing per-relation lateral_relids and
lateral_referencers relid sets into their transitive closures; that is,
they now show all relations on which a rel is directly or indirectly
laterally dependent. This not only fixes the chained-reference problem
but allows some of the relevant tests to be made substantially simpler
and faster, since they can be reduced to simple bitmap manipulations
instead of searches of the LateralJoinInfo list.
Also, when a PlaceHolderVar that is due to be evaluated at a join contains
lateral references, we should treat those references as indirect lateral
dependencies of each of the join's base relations. This prevents us from
trying to join any individual base relations to the lateral reference
source before the join is formed, which again cannot work.
Andreas' testing also exposed another oversight in the "dangerous
PlaceHolderVar" test added in commit 85e5e222b1dd02f1. Simply rejecting
unsafe join paths in joinpath.c is insufficient, because in some cases
we will end up rejecting *all* possible paths for a particular join, again
leading to "could not devise a query plan" failures. The restriction has
to be known also to join_is_legal and its cohort functions, so that they
will not select a join for which that will happen. I chose to move the
supporting logic into joinrels.c where the latter functions are.
Back-patch to 9.3 where LATERAL support was introduced.
2015-12-11 20:22:20 +01:00
|
|
|
Filter: (a.q2 = b.q1)
|
|
|
|
-> Seq Scan on public.int8_tbl c
|
|
|
|
Output: c.q1, c.q2
|
|
|
|
(11 rows)
|
2013-08-18 02:22:37 +02:00
|
|
|
|
|
|
|
select * from
|
|
|
|
int8_tbl a left join lateral
|
|
|
|
(select b.q1 as bq1, c.q1 as cq1, least(a.q1,b.q1,c.q1) from
|
|
|
|
int8_tbl b cross join int8_tbl c) ss
|
|
|
|
on a.q2 = ss.bq1;
|
|
|
|
q1 | q2 | bq1 | cq1 | least
|
|
|
|
------------------+-------------------+------------------+------------------+------------------
|
|
|
|
123 | 456 | | |
|
|
|
|
123 | 4567890123456789 | 4567890123456789 | 123 | 123
|
|
|
|
123 | 4567890123456789 | 4567890123456789 | 123 | 123
|
|
|
|
123 | 4567890123456789 | 4567890123456789 | 4567890123456789 | 123
|
|
|
|
123 | 4567890123456789 | 4567890123456789 | 4567890123456789 | 123
|
|
|
|
123 | 4567890123456789 | 4567890123456789 | 4567890123456789 | 123
|
|
|
|
123 | 4567890123456789 | 4567890123456789 | 123 | 123
|
|
|
|
123 | 4567890123456789 | 4567890123456789 | 123 | 123
|
|
|
|
123 | 4567890123456789 | 4567890123456789 | 4567890123456789 | 123
|
|
|
|
123 | 4567890123456789 | 4567890123456789 | 4567890123456789 | 123
|
|
|
|
123 | 4567890123456789 | 4567890123456789 | 4567890123456789 | 123
|
|
|
|
123 | 4567890123456789 | 4567890123456789 | 123 | 123
|
|
|
|
123 | 4567890123456789 | 4567890123456789 | 123 | 123
|
|
|
|
123 | 4567890123456789 | 4567890123456789 | 4567890123456789 | 123
|
|
|
|
123 | 4567890123456789 | 4567890123456789 | 4567890123456789 | 123
|
|
|
|
123 | 4567890123456789 | 4567890123456789 | 4567890123456789 | 123
|
|
|
|
4567890123456789 | 123 | 123 | 123 | 123
|
|
|
|
4567890123456789 | 123 | 123 | 123 | 123
|
|
|
|
4567890123456789 | 123 | 123 | 4567890123456789 | 123
|
|
|
|
4567890123456789 | 123 | 123 | 4567890123456789 | 123
|
|
|
|
4567890123456789 | 123 | 123 | 4567890123456789 | 123
|
|
|
|
4567890123456789 | 123 | 123 | 123 | 123
|
|
|
|
4567890123456789 | 123 | 123 | 123 | 123
|
|
|
|
4567890123456789 | 123 | 123 | 4567890123456789 | 123
|
|
|
|
4567890123456789 | 123 | 123 | 4567890123456789 | 123
|
|
|
|
4567890123456789 | 123 | 123 | 4567890123456789 | 123
|
|
|
|
4567890123456789 | 4567890123456789 | 4567890123456789 | 123 | 123
|
|
|
|
4567890123456789 | 4567890123456789 | 4567890123456789 | 123 | 123
|
|
|
|
4567890123456789 | 4567890123456789 | 4567890123456789 | 4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | 4567890123456789 | 4567890123456789 | 4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | 4567890123456789 | 4567890123456789 | 4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | 4567890123456789 | 4567890123456789 | 123 | 123
|
|
|
|
4567890123456789 | 4567890123456789 | 4567890123456789 | 123 | 123
|
|
|
|
4567890123456789 | 4567890123456789 | 4567890123456789 | 4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | 4567890123456789 | 4567890123456789 | 4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | 4567890123456789 | 4567890123456789 | 4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | 4567890123456789 | 4567890123456789 | 123 | 123
|
|
|
|
4567890123456789 | 4567890123456789 | 4567890123456789 | 123 | 123
|
|
|
|
4567890123456789 | 4567890123456789 | 4567890123456789 | 4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | 4567890123456789 | 4567890123456789 | 4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | 4567890123456789 | 4567890123456789 | 4567890123456789 | 4567890123456789
|
|
|
|
4567890123456789 | -4567890123456789 | | |
|
|
|
|
(42 rows)
|
|
|
|
|
2012-09-01 19:56:14 +02:00
|
|
|
-- case requiring nested PlaceHolderVars
|
|
|
|
explain (verbose, costs off)
|
|
|
|
select * from
|
|
|
|
int8_tbl c left join (
|
|
|
|
int8_tbl a left join (select q1, coalesce(q2,42) as x from int8_tbl b) ss1
|
|
|
|
on a.q2 = ss1.q1
|
|
|
|
cross join
|
|
|
|
lateral (select q1, coalesce(ss1.x,q2) as y from int8_tbl d) ss2
|
|
|
|
) on c.q2 = ss2.q1,
|
2015-03-12 04:18:03 +01:00
|
|
|
lateral (select ss2.y offset 0) ss3;
|
2015-03-30 20:59:49 +02:00
|
|
|
QUERY PLAN
|
|
|
|
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
|
2012-09-01 19:56:14 +02:00
|
|
|
Nested Loop
|
2015-03-30 20:59:49 +02:00
|
|
|
Output: c.q1, c.q2, a.q1, a.q2, b.q1, (COALESCE(b.q2, '42'::bigint)), d.q1, (COALESCE((COALESCE(b.q2, '42'::bigint)), d.q2)), ((COALESCE((COALESCE(b.q2, '42'::bigint)), d.q2)))
|
2012-09-01 19:56:14 +02:00
|
|
|
-> Hash Right Join
|
2015-03-30 20:59:49 +02:00
|
|
|
Output: c.q1, c.q2, a.q1, a.q2, b.q1, d.q1, (COALESCE(b.q2, '42'::bigint)), (COALESCE((COALESCE(b.q2, '42'::bigint)), d.q2))
|
2012-09-01 19:56:14 +02:00
|
|
|
Hash Cond: (d.q1 = c.q2)
|
|
|
|
-> Nested Loop
|
2015-03-30 20:59:49 +02:00
|
|
|
Output: a.q1, a.q2, b.q1, d.q1, (COALESCE(b.q2, '42'::bigint)), (COALESCE((COALESCE(b.q2, '42'::bigint)), d.q2))
|
2012-09-01 19:56:14 +02:00
|
|
|
-> Hash Left Join
|
2015-03-30 20:59:49 +02:00
|
|
|
Output: a.q1, a.q2, b.q1, (COALESCE(b.q2, '42'::bigint))
|
2012-09-01 19:56:14 +02:00
|
|
|
Hash Cond: (a.q2 = b.q1)
|
|
|
|
-> Seq Scan on public.int8_tbl a
|
|
|
|
Output: a.q1, a.q2
|
|
|
|
-> Hash
|
2015-03-30 20:59:49 +02:00
|
|
|
Output: b.q1, (COALESCE(b.q2, '42'::bigint))
|
2012-09-01 19:56:14 +02:00
|
|
|
-> Seq Scan on public.int8_tbl b
|
2015-03-30 20:59:49 +02:00
|
|
|
Output: b.q1, COALESCE(b.q2, '42'::bigint)
|
2013-08-18 02:22:37 +02:00
|
|
|
-> Seq Scan on public.int8_tbl d
|
2015-03-30 20:59:49 +02:00
|
|
|
Output: d.q1, COALESCE((COALESCE(b.q2, '42'::bigint)), d.q2)
|
2012-09-01 19:56:14 +02:00
|
|
|
-> Hash
|
|
|
|
Output: c.q1, c.q2
|
|
|
|
-> Seq Scan on public.int8_tbl c
|
|
|
|
Output: c.q1, c.q2
|
|
|
|
-> Result
|
2015-03-30 20:59:49 +02:00
|
|
|
Output: (COALESCE((COALESCE(b.q2, '42'::bigint)), d.q2))
|
2013-08-18 02:22:37 +02:00
|
|
|
(24 rows)
|
2012-09-01 19:56:14 +02:00
|
|
|
|
2013-08-15 00:38:32 +02:00
|
|
|
-- case that breaks the old ph_may_need optimization
|
|
|
|
explain (verbose, costs off)
|
|
|
|
select c.*,a.*,ss1.q1,ss2.q1,ss3.* from
|
|
|
|
int8_tbl c left join (
|
|
|
|
int8_tbl a left join
|
|
|
|
(select q1, coalesce(q2,f1) as x from int8_tbl b, int4_tbl b2
|
|
|
|
where q1 < f1) ss1
|
|
|
|
on a.q2 = ss1.q1
|
|
|
|
cross join
|
|
|
|
lateral (select q1, coalesce(ss1.x,q2) as y from int8_tbl d) ss2
|
|
|
|
) on c.q2 = ss2.q1,
|
|
|
|
lateral (select * from int4_tbl i where ss2.y > f1) ss3;
|
2013-08-18 02:22:37 +02:00
|
|
|
QUERY PLAN
|
|
|
|
---------------------------------------------------------------------------------------------------------
|
|
|
|
Nested Loop
|
2013-08-15 00:38:32 +02:00
|
|
|
Output: c.q1, c.q2, a.q1, a.q2, b.q1, d.q1, i.f1
|
2013-08-18 02:22:37 +02:00
|
|
|
Join Filter: ((COALESCE((COALESCE(b.q2, (b2.f1)::bigint)), d.q2)) > i.f1)
|
|
|
|
-> Hash Right Join
|
|
|
|
Output: c.q1, c.q2, a.q1, a.q2, b.q1, d.q1, (COALESCE((COALESCE(b.q2, (b2.f1)::bigint)), d.q2))
|
|
|
|
Hash Cond: (d.q1 = c.q2)
|
|
|
|
-> Nested Loop
|
|
|
|
Output: a.q1, a.q2, b.q1, d.q1, (COALESCE((COALESCE(b.q2, (b2.f1)::bigint)), d.q2))
|
|
|
|
-> Hash Right Join
|
|
|
|
Output: a.q1, a.q2, b.q1, (COALESCE(b.q2, (b2.f1)::bigint))
|
|
|
|
Hash Cond: (b.q1 = a.q2)
|
|
|
|
-> Nested Loop
|
|
|
|
Output: b.q1, COALESCE(b.q2, (b2.f1)::bigint)
|
|
|
|
Join Filter: (b.q1 < b2.f1)
|
|
|
|
-> Seq Scan on public.int8_tbl b
|
|
|
|
Output: b.q1, b.q2
|
|
|
|
-> Materialize
|
2013-08-15 00:38:32 +02:00
|
|
|
Output: b2.f1
|
2013-08-18 02:22:37 +02:00
|
|
|
-> Seq Scan on public.int4_tbl b2
|
|
|
|
Output: b2.f1
|
|
|
|
-> Hash
|
2013-08-15 00:38:32 +02:00
|
|
|
Output: a.q1, a.q2
|
2013-08-18 02:22:37 +02:00
|
|
|
-> Seq Scan on public.int8_tbl a
|
|
|
|
Output: a.q1, a.q2
|
2013-08-15 00:38:32 +02:00
|
|
|
-> Seq Scan on public.int8_tbl d
|
2013-08-18 02:22:37 +02:00
|
|
|
Output: d.q1, COALESCE((COALESCE(b.q2, (b2.f1)::bigint)), d.q2)
|
|
|
|
-> Hash
|
|
|
|
Output: c.q1, c.q2
|
2013-08-15 00:38:32 +02:00
|
|
|
-> Seq Scan on public.int8_tbl c
|
|
|
|
Output: c.q1, c.q2
|
2013-08-18 02:22:37 +02:00
|
|
|
-> Materialize
|
|
|
|
Output: i.f1
|
|
|
|
-> Seq Scan on public.int4_tbl i
|
|
|
|
Output: i.f1
|
|
|
|
(34 rows)
|
2013-08-15 00:38:32 +02:00
|
|
|
|
2014-01-30 20:51:16 +01:00
|
|
|
-- check processing of postponed quals (bug #9041)
|
|
|
|
explain (verbose, costs off)
|
|
|
|
select * from
|
2015-03-12 04:18:03 +01:00
|
|
|
(select 1 as x offset 0) x cross join (select 2 as y offset 0) y
|
2014-01-30 20:51:16 +01:00
|
|
|
left join lateral (
|
2015-03-12 04:18:03 +01:00
|
|
|
select * from (select 3 as z offset 0) z where z.z = x.x
|
2014-01-30 20:51:16 +01:00
|
|
|
) zz on zz.z = y.y;
|
|
|
|
QUERY PLAN
|
|
|
|
----------------------------------------------
|
|
|
|
Nested Loop Left Join
|
|
|
|
Output: (1), (2), (3)
|
|
|
|
Join Filter: (((3) = (1)) AND ((3) = (2)))
|
|
|
|
-> Nested Loop
|
|
|
|
Output: (1), (2)
|
|
|
|
-> Result
|
|
|
|
Output: 1
|
|
|
|
-> Result
|
|
|
|
Output: 2
|
|
|
|
-> Result
|
|
|
|
Output: 3
|
|
|
|
(11 rows)
|
|
|
|
|
2018-01-23 22:50:34 +01:00
|
|
|
-- check handling of nested appendrels inside LATERAL
|
|
|
|
select * from
|
|
|
|
((select 2 as v) union all (select 3 as v)) as q1
|
|
|
|
cross join lateral
|
|
|
|
((select * from
|
|
|
|
((select 4 as v) union all (select 5 as v)) as q3)
|
|
|
|
union all
|
|
|
|
(select q1.v)
|
|
|
|
) as q2;
|
|
|
|
v | v
|
|
|
|
---+---
|
|
|
|
2 | 4
|
|
|
|
2 | 5
|
|
|
|
2 | 2
|
|
|
|
3 | 4
|
|
|
|
3 | 5
|
|
|
|
3 | 3
|
|
|
|
(6 rows)
|
|
|
|
|
2015-08-01 01:26:33 +02:00
|
|
|
-- check we don't try to do a unique-ified semijoin with LATERAL
|
|
|
|
explain (verbose, costs off)
|
|
|
|
select * from
|
|
|
|
(values (0,9998), (1,1000)) v(id,x),
|
|
|
|
lateral (select f1 from int4_tbl
|
|
|
|
where f1 = any (select unique1 from tenk1
|
|
|
|
where unique2 = v.x offset 0)) ss;
|
|
|
|
QUERY PLAN
|
|
|
|
----------------------------------------------------------------------
|
|
|
|
Nested Loop
|
|
|
|
Output: "*VALUES*".column1, "*VALUES*".column2, int4_tbl.f1
|
|
|
|
-> Values Scan on "*VALUES*"
|
|
|
|
Output: "*VALUES*".column1, "*VALUES*".column2
|
|
|
|
-> Nested Loop Semi Join
|
|
|
|
Output: int4_tbl.f1
|
|
|
|
Join Filter: (int4_tbl.f1 = tenk1.unique1)
|
|
|
|
-> Seq Scan on public.int4_tbl
|
|
|
|
Output: int4_tbl.f1
|
|
|
|
-> Materialize
|
|
|
|
Output: tenk1.unique1
|
|
|
|
-> Index Scan using tenk1_unique2 on public.tenk1
|
|
|
|
Output: tenk1.unique1
|
|
|
|
Index Cond: (tenk1.unique2 = "*VALUES*".column2)
|
|
|
|
(14 rows)
|
|
|
|
|
|
|
|
select * from
|
|
|
|
(values (0,9998), (1,1000)) v(id,x),
|
|
|
|
lateral (select f1 from int4_tbl
|
|
|
|
where f1 = any (select unique1 from tenk1
|
|
|
|
where unique2 = v.x offset 0)) ss;
|
|
|
|
id | x | f1
|
|
|
|
----+------+----
|
|
|
|
0 | 9998 | 0
|
|
|
|
(1 row)
|
|
|
|
|
2015-08-12 05:48:37 +02:00
|
|
|
-- check proper extParam/allParam handling (this isn't exactly a LATERAL issue,
|
|
|
|
-- but we can make the test case much more compact with LATERAL)
|
|
|
|
explain (verbose, costs off)
|
|
|
|
select * from (values (0), (1)) v(id),
|
|
|
|
lateral (select * from int8_tbl t1,
|
|
|
|
lateral (select * from
|
|
|
|
(select * from int8_tbl t2
|
|
|
|
where q1 = any (select q2 from int8_tbl t3
|
|
|
|
where q2 = (select greatest(t1.q1,t2.q2))
|
|
|
|
and (select v.id=0)) offset 0) ss2) ss
|
|
|
|
where t1.q1 = ss.q2) ss0;
|
|
|
|
QUERY PLAN
|
|
|
|
-----------------------------------------------------------------
|
|
|
|
Nested Loop
|
|
|
|
Output: "*VALUES*".column1, t1.q1, t1.q2, ss2.q1, ss2.q2
|
|
|
|
-> Seq Scan on public.int8_tbl t1
|
|
|
|
Output: t1.q1, t1.q2
|
|
|
|
-> Nested Loop
|
|
|
|
Output: "*VALUES*".column1, ss2.q1, ss2.q2
|
|
|
|
-> Values Scan on "*VALUES*"
|
|
|
|
Output: "*VALUES*".column1
|
|
|
|
-> Subquery Scan on ss2
|
|
|
|
Output: ss2.q1, ss2.q2
|
|
|
|
Filter: (t1.q1 = ss2.q2)
|
|
|
|
-> Seq Scan on public.int8_tbl t2
|
|
|
|
Output: t2.q1, t2.q2
|
|
|
|
Filter: (SubPlan 3)
|
|
|
|
SubPlan 3
|
|
|
|
-> Result
|
|
|
|
Output: t3.q2
|
|
|
|
One-Time Filter: $4
|
|
|
|
InitPlan 1 (returns $2)
|
|
|
|
-> Result
|
|
|
|
Output: GREATEST($0, t2.q2)
|
|
|
|
InitPlan 2 (returns $4)
|
|
|
|
-> Result
|
|
|
|
Output: ($3 = 0)
|
|
|
|
-> Seq Scan on public.int8_tbl t3
|
|
|
|
Output: t3.q1, t3.q2
|
|
|
|
Filter: (t3.q2 = $2)
|
|
|
|
(27 rows)
|
|
|
|
|
|
|
|
select * from (values (0), (1)) v(id),
|
|
|
|
lateral (select * from int8_tbl t1,
|
|
|
|
lateral (select * from
|
|
|
|
(select * from int8_tbl t2
|
|
|
|
where q1 = any (select q2 from int8_tbl t3
|
|
|
|
where q2 = (select greatest(t1.q1,t2.q2))
|
|
|
|
and (select v.id=0)) offset 0) ss2) ss
|
|
|
|
where t1.q1 = ss.q2) ss0;
|
|
|
|
id | q1 | q2 | q1 | q2
|
|
|
|
----+------------------+-------------------+------------------+------------------
|
|
|
|
0 | 4567890123456789 | 123 | 4567890123456789 | 4567890123456789
|
|
|
|
0 | 4567890123456789 | 4567890123456789 | 4567890123456789 | 4567890123456789
|
|
|
|
0 | 4567890123456789 | -4567890123456789 | 4567890123456789 | 4567890123456789
|
|
|
|
(3 rows)
|
|
|
|
|
2012-08-08 01:02:54 +02:00
|
|
|
-- test some error cases where LATERAL should have been used but wasn't
|
2013-01-26 22:18:42 +01:00
|
|
|
select f1,g from int4_tbl a, (select f1 as g) ss;
|
2012-08-08 01:02:54 +02:00
|
|
|
ERROR: column "f1" does not exist
|
2013-01-26 22:18:42 +01:00
|
|
|
LINE 1: select f1,g from int4_tbl a, (select f1 as g) ss;
|
|
|
|
^
|
2012-08-08 01:02:54 +02:00
|
|
|
HINT: There is a column named "f1" in table "a", but it cannot be referenced from this part of the query.
|
2013-01-26 22:18:42 +01:00
|
|
|
select f1,g from int4_tbl a, (select a.f1 as g) ss;
|
2012-08-08 01:02:54 +02:00
|
|
|
ERROR: invalid reference to FROM-clause entry for table "a"
|
2013-01-26 22:18:42 +01:00
|
|
|
LINE 1: select f1,g from int4_tbl a, (select a.f1 as g) ss;
|
|
|
|
^
|
2012-08-08 01:02:54 +02:00
|
|
|
HINT: There is an entry for table "a", but it cannot be referenced from this part of the query.
|
2013-01-26 22:18:42 +01:00
|
|
|
select f1,g from int4_tbl a cross join (select f1 as g) ss;
|
2012-08-08 01:02:54 +02:00
|
|
|
ERROR: column "f1" does not exist
|
2013-01-26 22:18:42 +01:00
|
|
|
LINE 1: select f1,g from int4_tbl a cross join (select f1 as g) ss;
|
|
|
|
^
|
2012-08-08 01:02:54 +02:00
|
|
|
HINT: There is a column named "f1" in table "a", but it cannot be referenced from this part of the query.
|
2013-01-26 22:18:42 +01:00
|
|
|
select f1,g from int4_tbl a cross join (select a.f1 as g) ss;
|
2012-08-08 01:02:54 +02:00
|
|
|
ERROR: invalid reference to FROM-clause entry for table "a"
|
2013-01-26 22:18:42 +01:00
|
|
|
LINE 1: select f1,g from int4_tbl a cross join (select a.f1 as g) ss...
|
|
|
|
^
|
2012-08-08 01:02:54 +02:00
|
|
|
HINT: There is an entry for table "a", but it cannot be referenced from this part of the query.
|
|
|
|
-- SQL:2008 says the left table is in scope but illegal to access here
|
|
|
|
select f1,g from int4_tbl a right join lateral generate_series(0, a.f1) g on true;
|
|
|
|
ERROR: invalid reference to FROM-clause entry for table "a"
|
|
|
|
LINE 1: ... int4_tbl a right join lateral generate_series(0, a.f1) g on...
|
|
|
|
^
|
|
|
|
DETAIL: The combining JOIN type must be INNER or LEFT for a LATERAL reference.
|
|
|
|
select f1,g from int4_tbl a full join lateral generate_series(0, a.f1) g on true;
|
|
|
|
ERROR: invalid reference to FROM-clause entry for table "a"
|
|
|
|
LINE 1: ...m int4_tbl a full join lateral generate_series(0, a.f1) g on...
|
|
|
|
^
|
|
|
|
DETAIL: The combining JOIN type must be INNER or LEFT for a LATERAL reference.
|
2013-11-11 16:42:57 +01:00
|
|
|
-- check we complain about ambiguous table references
|
|
|
|
select * from
|
|
|
|
int8_tbl x cross join (int4_tbl x cross join lateral (select x.f1) ss);
|
|
|
|
ERROR: table reference "x" is ambiguous
|
|
|
|
LINE 2: ...cross join (int4_tbl x cross join lateral (select x.f1) ss);
|
|
|
|
^
|
2012-08-08 01:02:54 +02:00
|
|
|
-- LATERAL can be used to put an aggregate into the FROM clause of its query
|
|
|
|
select 1 from tenk1 a, lateral (select max(a.unique1) from int4_tbl b) ss;
|
Centralize the logic for detecting misplaced aggregates, window funcs, etc.
Formerly we relied on checking after-the-fact to see if an expression
contained aggregates, window functions, or sub-selects when it shouldn't.
This is grotty, easily forgotten (indeed, we had forgotten to teach
DefineIndex about rejecting window functions), and none too efficient
since it requires extra traversals of the parse tree. To improve matters,
define an enum type that classifies all SQL sub-expressions, store it in
ParseState to show what kind of expression we are currently parsing, and
make transformAggregateCall, transformWindowFuncCall, and transformSubLink
check the expression type and throw error if the type indicates the
construct is disallowed. This allows removal of a large number of ad-hoc
checks scattered around the code base. The enum type is sufficiently
fine-grained that we can still produce error messages of at least the
same specificity as before.
Bringing these error checks together revealed that we'd been none too
consistent about phrasing of the error messages, so standardize the wording
a bit.
Also, rewrite checking of aggregate arguments so that it requires only one
traversal of the arguments, rather than up to three as before.
In passing, clean up some more comments left over from add_missing_from
support, and annotate some tests that I think are dead code now that that's
gone. (I didn't risk actually removing said dead code, though.)
2012-08-10 17:35:33 +02:00
|
|
|
ERROR: aggregate functions are not allowed in FROM clause of their own query level
|
2012-08-08 01:02:54 +02:00
|
|
|
LINE 1: select 1 from tenk1 a, lateral (select max(a.unique1) from i...
|
|
|
|
^
|
2014-01-07 21:25:16 +01:00
|
|
|
-- check behavior of LATERAL in UPDATE/DELETE
|
|
|
|
create temp table xx1 as select f1 as x1, -f1 as x2 from int4_tbl;
|
2014-01-12 01:03:12 +01:00
|
|
|
-- error, can't do this:
|
2014-01-07 21:25:16 +01:00
|
|
|
update xx1 set x2 = f1 from (select * from int4_tbl where f1 = x1) ss;
|
|
|
|
ERROR: column "x1" does not exist
|
|
|
|
LINE 1: ... set x2 = f1 from (select * from int4_tbl where f1 = x1) ss;
|
|
|
|
^
|
|
|
|
HINT: There is a column named "x1" in table "xx1", but it cannot be referenced from this part of the query.
|
|
|
|
update xx1 set x2 = f1 from (select * from int4_tbl where f1 = xx1.x1) ss;
|
|
|
|
ERROR: invalid reference to FROM-clause entry for table "xx1"
|
|
|
|
LINE 1: ...t x2 = f1 from (select * from int4_tbl where f1 = xx1.x1) ss...
|
|
|
|
^
|
|
|
|
HINT: There is an entry for table "xx1", but it cannot be referenced from this part of the query.
|
2014-01-12 01:03:12 +01:00
|
|
|
-- can't do it even with LATERAL:
|
2014-01-07 21:25:16 +01:00
|
|
|
update xx1 set x2 = f1 from lateral (select * from int4_tbl where f1 = x1) ss;
|
2014-01-12 01:03:12 +01:00
|
|
|
ERROR: invalid reference to FROM-clause entry for table "xx1"
|
|
|
|
LINE 1: ...= f1 from lateral (select * from int4_tbl where f1 = x1) ss;
|
|
|
|
^
|
|
|
|
HINT: There is an entry for table "xx1", but it cannot be referenced from this part of the query.
|
|
|
|
-- we might in future allow something like this, but for now it's an error:
|
|
|
|
update xx1 set x2 = f1 from xx1, lateral (select * from int4_tbl where f1 = x1) ss;
|
|
|
|
ERROR: table name "xx1" specified more than once
|
|
|
|
-- also errors:
|
2014-01-07 21:25:16 +01:00
|
|
|
delete from xx1 using (select * from int4_tbl where f1 = x1) ss;
|
|
|
|
ERROR: column "x1" does not exist
|
|
|
|
LINE 1: ...te from xx1 using (select * from int4_tbl where f1 = x1) ss;
|
|
|
|
^
|
|
|
|
HINT: There is a column named "x1" in table "xx1", but it cannot be referenced from this part of the query.
|
2014-01-12 01:03:12 +01:00
|
|
|
delete from xx1 using (select * from int4_tbl where f1 = xx1.x1) ss;
|
|
|
|
ERROR: invalid reference to FROM-clause entry for table "xx1"
|
|
|
|
LINE 1: ...from xx1 using (select * from int4_tbl where f1 = xx1.x1) ss...
|
|
|
|
^
|
|
|
|
HINT: There is an entry for table "xx1", but it cannot be referenced from this part of the query.
|
2014-01-07 21:25:16 +01:00
|
|
|
delete from xx1 using lateral (select * from int4_tbl where f1 = x1) ss;
|
2014-01-12 01:03:12 +01:00
|
|
|
ERROR: invalid reference to FROM-clause entry for table "xx1"
|
|
|
|
LINE 1: ...xx1 using lateral (select * from int4_tbl where f1 = x1) ss;
|
|
|
|
^
|
|
|
|
HINT: There is an entry for table "xx1", but it cannot be referenced from this part of the query.
|
2017-04-08 04:20:03 +02:00
|
|
|
--
|
2017-09-14 21:41:08 +02:00
|
|
|
-- test LATERAL reference propagation down a multi-level inheritance hierarchy
|
|
|
|
-- produced for a multi-level partitioned table hierarchy.
|
|
|
|
--
|
Clean up duplicate table and function names in regression tests.
Many of the objects we create during the regression tests are put in the
public schema, so that using the same names in different regression tests
creates a hazard of test failures if any two such scripts run concurrently.
This patch cleans up a bunch of latent hazards of that sort, as well as two
live hazards.
The current situation in this regard is far worse than it was a year or two
back, because practically all of the partitioning-related test cases have
reused table names with enthusiasm. I despaired of cleaning up that mess
within the five most-affected tests (create_table, alter_table, insert,
update, inherit); fortunately those don't run concurrently.
Other than partitioning problems, most of the issues boil down to using
names like "foo", "bar", "tmp", etc, without thought for the fact that
other test scripts might use similar names concurrently. I've made an
effort to make all such names more specific.
One of the live hazards was that commit 7421f4b8 caused with.sql to
create a table named "test", conflicting with a similarly-named table
in alter_table.sql; this was exposed in the buildfarm recently.
The other one was that join.sql and transactions.sql both create tables
named "foo" and "bar"; but join.sql's uses of those names date back
only to December or so.
Since commit 7421f4b8 was back-patched to v10, back-patch a minimal
fix for that problem. The rest of this is just future-proofing.
Discussion: https://postgr.es/m/4627.1521070268@sss.pgh.pa.us
2018-03-15 22:08:51 +01:00
|
|
|
create table join_pt1 (a int, b int, c varchar) partition by range(a);
|
|
|
|
create table join_pt1p1 partition of join_pt1 for values from (0) to (100) partition by range(b);
|
|
|
|
create table join_pt1p2 partition of join_pt1 for values from (100) to (200);
|
|
|
|
create table join_pt1p1p1 partition of join_pt1p1 for values from (0) to (100);
|
|
|
|
insert into join_pt1 values (1, 1, 'x'), (101, 101, 'y');
|
|
|
|
create table join_ut1 (a int, b int, c varchar);
|
|
|
|
insert into join_ut1 values (101, 101, 'y'), (2, 2, 'z');
|
2017-09-14 21:41:08 +02:00
|
|
|
explain (verbose, costs off)
|
Clean up duplicate table and function names in regression tests.
Many of the objects we create during the regression tests are put in the
public schema, so that using the same names in different regression tests
creates a hazard of test failures if any two such scripts run concurrently.
This patch cleans up a bunch of latent hazards of that sort, as well as two
live hazards.
The current situation in this regard is far worse than it was a year or two
back, because practically all of the partitioning-related test cases have
reused table names with enthusiasm. I despaired of cleaning up that mess
within the five most-affected tests (create_table, alter_table, insert,
update, inherit); fortunately those don't run concurrently.
Other than partitioning problems, most of the issues boil down to using
names like "foo", "bar", "tmp", etc, without thought for the fact that
other test scripts might use similar names concurrently. I've made an
effort to make all such names more specific.
One of the live hazards was that commit 7421f4b8 caused with.sql to
create a table named "test", conflicting with a similarly-named table
in alter_table.sql; this was exposed in the buildfarm recently.
The other one was that join.sql and transactions.sql both create tables
named "foo" and "bar"; but join.sql's uses of those names date back
only to December or so.
Since commit 7421f4b8 was back-patched to v10, back-patch a minimal
fix for that problem. The rest of this is just future-proofing.
Discussion: https://postgr.es/m/4627.1521070268@sss.pgh.pa.us
2018-03-15 22:08:51 +01:00
|
|
|
select t1.b, ss.phv from join_ut1 t1 left join lateral
|
2017-09-14 21:41:08 +02:00
|
|
|
(select t2.a as t2a, t3.a t3a, least(t1.a, t2.a, t3.a) phv
|
Clean up duplicate table and function names in regression tests.
Many of the objects we create during the regression tests are put in the
public schema, so that using the same names in different regression tests
creates a hazard of test failures if any two such scripts run concurrently.
This patch cleans up a bunch of latent hazards of that sort, as well as two
live hazards.
The current situation in this regard is far worse than it was a year or two
back, because practically all of the partitioning-related test cases have
reused table names with enthusiasm. I despaired of cleaning up that mess
within the five most-affected tests (create_table, alter_table, insert,
update, inherit); fortunately those don't run concurrently.
Other than partitioning problems, most of the issues boil down to using
names like "foo", "bar", "tmp", etc, without thought for the fact that
other test scripts might use similar names concurrently. I've made an
effort to make all such names more specific.
One of the live hazards was that commit 7421f4b8 caused with.sql to
create a table named "test", conflicting with a similarly-named table
in alter_table.sql; this was exposed in the buildfarm recently.
The other one was that join.sql and transactions.sql both create tables
named "foo" and "bar"; but join.sql's uses of those names date back
only to December or so.
Since commit 7421f4b8 was back-patched to v10, back-patch a minimal
fix for that problem. The rest of this is just future-proofing.
Discussion: https://postgr.es/m/4627.1521070268@sss.pgh.pa.us
2018-03-15 22:08:51 +01:00
|
|
|
from join_pt1 t2 join join_ut1 t3 on t2.a = t3.b) ss
|
2017-09-14 21:41:08 +02:00
|
|
|
on t1.a = ss.t2a order by t1.a;
|
Clean up duplicate table and function names in regression tests.
Many of the objects we create during the regression tests are put in the
public schema, so that using the same names in different regression tests
creates a hazard of test failures if any two such scripts run concurrently.
This patch cleans up a bunch of latent hazards of that sort, as well as two
live hazards.
The current situation in this regard is far worse than it was a year or two
back, because practically all of the partitioning-related test cases have
reused table names with enthusiasm. I despaired of cleaning up that mess
within the five most-affected tests (create_table, alter_table, insert,
update, inherit); fortunately those don't run concurrently.
Other than partitioning problems, most of the issues boil down to using
names like "foo", "bar", "tmp", etc, without thought for the fact that
other test scripts might use similar names concurrently. I've made an
effort to make all such names more specific.
One of the live hazards was that commit 7421f4b8 caused with.sql to
create a table named "test", conflicting with a similarly-named table
in alter_table.sql; this was exposed in the buildfarm recently.
The other one was that join.sql and transactions.sql both create tables
named "foo" and "bar"; but join.sql's uses of those names date back
only to December or so.
Since commit 7421f4b8 was back-patched to v10, back-patch a minimal
fix for that problem. The rest of this is just future-proofing.
Discussion: https://postgr.es/m/4627.1521070268@sss.pgh.pa.us
2018-03-15 22:08:51 +01:00
|
|
|
QUERY PLAN
|
|
|
|
------------------------------------------------------------------
|
2017-09-14 21:41:08 +02:00
|
|
|
Sort
|
|
|
|
Output: t1.b, (LEAST(t1.a, t2.a, t3.a)), t1.a
|
|
|
|
Sort Key: t1.a
|
|
|
|
-> Nested Loop Left Join
|
|
|
|
Output: t1.b, (LEAST(t1.a, t2.a, t3.a)), t1.a
|
Clean up duplicate table and function names in regression tests.
Many of the objects we create during the regression tests are put in the
public schema, so that using the same names in different regression tests
creates a hazard of test failures if any two such scripts run concurrently.
This patch cleans up a bunch of latent hazards of that sort, as well as two
live hazards.
The current situation in this regard is far worse than it was a year or two
back, because practically all of the partitioning-related test cases have
reused table names with enthusiasm. I despaired of cleaning up that mess
within the five most-affected tests (create_table, alter_table, insert,
update, inherit); fortunately those don't run concurrently.
Other than partitioning problems, most of the issues boil down to using
names like "foo", "bar", "tmp", etc, without thought for the fact that
other test scripts might use similar names concurrently. I've made an
effort to make all such names more specific.
One of the live hazards was that commit 7421f4b8 caused with.sql to
create a table named "test", conflicting with a similarly-named table
in alter_table.sql; this was exposed in the buildfarm recently.
The other one was that join.sql and transactions.sql both create tables
named "foo" and "bar"; but join.sql's uses of those names date back
only to December or so.
Since commit 7421f4b8 was back-patched to v10, back-patch a minimal
fix for that problem. The rest of this is just future-proofing.
Discussion: https://postgr.es/m/4627.1521070268@sss.pgh.pa.us
2018-03-15 22:08:51 +01:00
|
|
|
-> Seq Scan on public.join_ut1 t1
|
2017-09-14 21:41:08 +02:00
|
|
|
Output: t1.a, t1.b, t1.c
|
|
|
|
-> Hash Join
|
|
|
|
Output: t2.a, LEAST(t1.a, t2.a, t3.a)
|
|
|
|
Hash Cond: (t3.b = t2.a)
|
Clean up duplicate table and function names in regression tests.
Many of the objects we create during the regression tests are put in the
public schema, so that using the same names in different regression tests
creates a hazard of test failures if any two such scripts run concurrently.
This patch cleans up a bunch of latent hazards of that sort, as well as two
live hazards.
The current situation in this regard is far worse than it was a year or two
back, because practically all of the partitioning-related test cases have
reused table names with enthusiasm. I despaired of cleaning up that mess
within the five most-affected tests (create_table, alter_table, insert,
update, inherit); fortunately those don't run concurrently.
Other than partitioning problems, most of the issues boil down to using
names like "foo", "bar", "tmp", etc, without thought for the fact that
other test scripts might use similar names concurrently. I've made an
effort to make all such names more specific.
One of the live hazards was that commit 7421f4b8 caused with.sql to
create a table named "test", conflicting with a similarly-named table
in alter_table.sql; this was exposed in the buildfarm recently.
The other one was that join.sql and transactions.sql both create tables
named "foo" and "bar"; but join.sql's uses of those names date back
only to December or so.
Since commit 7421f4b8 was back-patched to v10, back-patch a minimal
fix for that problem. The rest of this is just future-proofing.
Discussion: https://postgr.es/m/4627.1521070268@sss.pgh.pa.us
2018-03-15 22:08:51 +01:00
|
|
|
-> Seq Scan on public.join_ut1 t3
|
2017-09-14 21:41:08 +02:00
|
|
|
Output: t3.a, t3.b, t3.c
|
|
|
|
-> Hash
|
|
|
|
Output: t2.a
|
|
|
|
-> Append
|
Clean up duplicate table and function names in regression tests.
Many of the objects we create during the regression tests are put in the
public schema, so that using the same names in different regression tests
creates a hazard of test failures if any two such scripts run concurrently.
This patch cleans up a bunch of latent hazards of that sort, as well as two
live hazards.
The current situation in this regard is far worse than it was a year or two
back, because practically all of the partitioning-related test cases have
reused table names with enthusiasm. I despaired of cleaning up that mess
within the five most-affected tests (create_table, alter_table, insert,
update, inherit); fortunately those don't run concurrently.
Other than partitioning problems, most of the issues boil down to using
names like "foo", "bar", "tmp", etc, without thought for the fact that
other test scripts might use similar names concurrently. I've made an
effort to make all such names more specific.
One of the live hazards was that commit 7421f4b8 caused with.sql to
create a table named "test", conflicting with a similarly-named table
in alter_table.sql; this was exposed in the buildfarm recently.
The other one was that join.sql and transactions.sql both create tables
named "foo" and "bar"; but join.sql's uses of those names date back
only to December or so.
Since commit 7421f4b8 was back-patched to v10, back-patch a minimal
fix for that problem. The rest of this is just future-proofing.
Discussion: https://postgr.es/m/4627.1521070268@sss.pgh.pa.us
2018-03-15 22:08:51 +01:00
|
|
|
-> Seq Scan on public.join_pt1p1p1 t2
|
2017-09-14 21:41:08 +02:00
|
|
|
Output: t2.a
|
|
|
|
Filter: (t1.a = t2.a)
|
Clean up duplicate table and function names in regression tests.
Many of the objects we create during the regression tests are put in the
public schema, so that using the same names in different regression tests
creates a hazard of test failures if any two such scripts run concurrently.
This patch cleans up a bunch of latent hazards of that sort, as well as two
live hazards.
The current situation in this regard is far worse than it was a year or two
back, because practically all of the partitioning-related test cases have
reused table names with enthusiasm. I despaired of cleaning up that mess
within the five most-affected tests (create_table, alter_table, insert,
update, inherit); fortunately those don't run concurrently.
Other than partitioning problems, most of the issues boil down to using
names like "foo", "bar", "tmp", etc, without thought for the fact that
other test scripts might use similar names concurrently. I've made an
effort to make all such names more specific.
One of the live hazards was that commit 7421f4b8 caused with.sql to
create a table named "test", conflicting with a similarly-named table
in alter_table.sql; this was exposed in the buildfarm recently.
The other one was that join.sql and transactions.sql both create tables
named "foo" and "bar"; but join.sql's uses of those names date back
only to December or so.
Since commit 7421f4b8 was back-patched to v10, back-patch a minimal
fix for that problem. The rest of this is just future-proofing.
Discussion: https://postgr.es/m/4627.1521070268@sss.pgh.pa.us
2018-03-15 22:08:51 +01:00
|
|
|
-> Seq Scan on public.join_pt1p2 t2_1
|
2017-09-14 21:41:08 +02:00
|
|
|
Output: t2_1.a
|
|
|
|
Filter: (t1.a = t2_1.a)
|
|
|
|
(21 rows)
|
|
|
|
|
Clean up duplicate table and function names in regression tests.
Many of the objects we create during the regression tests are put in the
public schema, so that using the same names in different regression tests
creates a hazard of test failures if any two such scripts run concurrently.
This patch cleans up a bunch of latent hazards of that sort, as well as two
live hazards.
The current situation in this regard is far worse than it was a year or two
back, because practically all of the partitioning-related test cases have
reused table names with enthusiasm. I despaired of cleaning up that mess
within the five most-affected tests (create_table, alter_table, insert,
update, inherit); fortunately those don't run concurrently.
Other than partitioning problems, most of the issues boil down to using
names like "foo", "bar", "tmp", etc, without thought for the fact that
other test scripts might use similar names concurrently. I've made an
effort to make all such names more specific.
One of the live hazards was that commit 7421f4b8 caused with.sql to
create a table named "test", conflicting with a similarly-named table
in alter_table.sql; this was exposed in the buildfarm recently.
The other one was that join.sql and transactions.sql both create tables
named "foo" and "bar"; but join.sql's uses of those names date back
only to December or so.
Since commit 7421f4b8 was back-patched to v10, back-patch a minimal
fix for that problem. The rest of this is just future-proofing.
Discussion: https://postgr.es/m/4627.1521070268@sss.pgh.pa.us
2018-03-15 22:08:51 +01:00
|
|
|
select t1.b, ss.phv from join_ut1 t1 left join lateral
|
2017-09-14 21:41:08 +02:00
|
|
|
(select t2.a as t2a, t3.a t3a, least(t1.a, t2.a, t3.a) phv
|
Clean up duplicate table and function names in regression tests.
Many of the objects we create during the regression tests are put in the
public schema, so that using the same names in different regression tests
creates a hazard of test failures if any two such scripts run concurrently.
This patch cleans up a bunch of latent hazards of that sort, as well as two
live hazards.
The current situation in this regard is far worse than it was a year or two
back, because practically all of the partitioning-related test cases have
reused table names with enthusiasm. I despaired of cleaning up that mess
within the five most-affected tests (create_table, alter_table, insert,
update, inherit); fortunately those don't run concurrently.
Other than partitioning problems, most of the issues boil down to using
names like "foo", "bar", "tmp", etc, without thought for the fact that
other test scripts might use similar names concurrently. I've made an
effort to make all such names more specific.
One of the live hazards was that commit 7421f4b8 caused with.sql to
create a table named "test", conflicting with a similarly-named table
in alter_table.sql; this was exposed in the buildfarm recently.
The other one was that join.sql and transactions.sql both create tables
named "foo" and "bar"; but join.sql's uses of those names date back
only to December or so.
Since commit 7421f4b8 was back-patched to v10, back-patch a minimal
fix for that problem. The rest of this is just future-proofing.
Discussion: https://postgr.es/m/4627.1521070268@sss.pgh.pa.us
2018-03-15 22:08:51 +01:00
|
|
|
from join_pt1 t2 join join_ut1 t3 on t2.a = t3.b) ss
|
2017-09-14 21:41:08 +02:00
|
|
|
on t1.a = ss.t2a order by t1.a;
|
|
|
|
b | phv
|
|
|
|
-----+-----
|
|
|
|
2 |
|
|
|
|
101 | 101
|
|
|
|
(2 rows)
|
|
|
|
|
Clean up duplicate table and function names in regression tests.
Many of the objects we create during the regression tests are put in the
public schema, so that using the same names in different regression tests
creates a hazard of test failures if any two such scripts run concurrently.
This patch cleans up a bunch of latent hazards of that sort, as well as two
live hazards.
The current situation in this regard is far worse than it was a year or two
back, because practically all of the partitioning-related test cases have
reused table names with enthusiasm. I despaired of cleaning up that mess
within the five most-affected tests (create_table, alter_table, insert,
update, inherit); fortunately those don't run concurrently.
Other than partitioning problems, most of the issues boil down to using
names like "foo", "bar", "tmp", etc, without thought for the fact that
other test scripts might use similar names concurrently. I've made an
effort to make all such names more specific.
One of the live hazards was that commit 7421f4b8 caused with.sql to
create a table named "test", conflicting with a similarly-named table
in alter_table.sql; this was exposed in the buildfarm recently.
The other one was that join.sql and transactions.sql both create tables
named "foo" and "bar"; but join.sql's uses of those names date back
only to December or so.
Since commit 7421f4b8 was back-patched to v10, back-patch a minimal
fix for that problem. The rest of this is just future-proofing.
Discussion: https://postgr.es/m/4627.1521070268@sss.pgh.pa.us
2018-03-15 22:08:51 +01:00
|
|
|
drop table join_pt1;
|
|
|
|
drop table join_ut1;
|
2017-09-14 21:41:08 +02:00
|
|
|
--
|
2017-06-19 21:33:41 +02:00
|
|
|
-- test that foreign key join estimation performs sanely for outer joins
|
|
|
|
--
|
|
|
|
begin;
|
|
|
|
create table fkest (a int, b int, c int unique, primary key(a,b));
|
|
|
|
create table fkest1 (a int, b int, primary key(a,b));
|
|
|
|
insert into fkest select x/10, x%10, x from generate_series(1,1000) x;
|
|
|
|
insert into fkest1 select x/10, x%10 from generate_series(1,1000) x;
|
|
|
|
alter table fkest1
|
|
|
|
add constraint fkest1_a_b_fkey foreign key (a,b) references fkest;
|
|
|
|
analyze fkest;
|
|
|
|
analyze fkest1;
|
|
|
|
explain (costs off)
|
|
|
|
select *
|
|
|
|
from fkest f
|
|
|
|
left join fkest1 f1 on f.a = f1.a and f.b = f1.b
|
|
|
|
left join fkest1 f2 on f.a = f2.a and f.b = f2.b
|
|
|
|
left join fkest1 f3 on f.a = f3.a and f.b = f3.b
|
|
|
|
where f.c = 1;
|
|
|
|
QUERY PLAN
|
|
|
|
------------------------------------------------------------------
|
|
|
|
Nested Loop Left Join
|
|
|
|
-> Nested Loop Left Join
|
|
|
|
-> Nested Loop Left Join
|
|
|
|
-> Index Scan using fkest_c_key on fkest f
|
|
|
|
Index Cond: (c = 1)
|
|
|
|
-> Index Only Scan using fkest1_pkey on fkest1 f1
|
|
|
|
Index Cond: ((a = f.a) AND (b = f.b))
|
|
|
|
-> Index Only Scan using fkest1_pkey on fkest1 f2
|
|
|
|
Index Cond: ((a = f.a) AND (b = f.b))
|
|
|
|
-> Index Only Scan using fkest1_pkey on fkest1 f3
|
|
|
|
Index Cond: ((a = f.a) AND (b = f.b))
|
|
|
|
(11 rows)
|
|
|
|
|
|
|
|
rollback;
|
|
|
|
--
|
2017-04-08 04:20:03 +02:00
|
|
|
-- test planner's ability to mark joins as unique
|
|
|
|
--
|
|
|
|
create table j1 (id int primary key);
|
|
|
|
create table j2 (id int primary key);
|
|
|
|
create table j3 (id int);
|
|
|
|
insert into j1 values(1),(2),(3);
|
|
|
|
insert into j2 values(1),(2),(3);
|
|
|
|
insert into j3 values(1),(1);
|
|
|
|
analyze j1;
|
|
|
|
analyze j2;
|
|
|
|
analyze j3;
|
|
|
|
-- ensure join is properly marked as unique
|
|
|
|
explain (verbose, costs off)
|
|
|
|
select * from j1 inner join j2 on j1.id = j2.id;
|
|
|
|
QUERY PLAN
|
|
|
|
-----------------------------------
|
|
|
|
Hash Join
|
|
|
|
Output: j1.id, j2.id
|
|
|
|
Inner Unique: true
|
|
|
|
Hash Cond: (j1.id = j2.id)
|
|
|
|
-> Seq Scan on public.j1
|
|
|
|
Output: j1.id
|
|
|
|
-> Hash
|
|
|
|
Output: j2.id
|
|
|
|
-> Seq Scan on public.j2
|
|
|
|
Output: j2.id
|
|
|
|
(10 rows)
|
|
|
|
|
|
|
|
-- ensure join is not unique when not an equi-join
|
|
|
|
explain (verbose, costs off)
|
|
|
|
select * from j1 inner join j2 on j1.id > j2.id;
|
|
|
|
QUERY PLAN
|
|
|
|
-----------------------------------
|
|
|
|
Nested Loop
|
|
|
|
Output: j1.id, j2.id
|
|
|
|
Join Filter: (j1.id > j2.id)
|
|
|
|
-> Seq Scan on public.j1
|
|
|
|
Output: j1.id
|
|
|
|
-> Materialize
|
|
|
|
Output: j2.id
|
|
|
|
-> Seq Scan on public.j2
|
|
|
|
Output: j2.id
|
|
|
|
(9 rows)
|
|
|
|
|
|
|
|
-- ensure non-unique rel is not chosen as inner
|
|
|
|
explain (verbose, costs off)
|
|
|
|
select * from j1 inner join j3 on j1.id = j3.id;
|
|
|
|
QUERY PLAN
|
|
|
|
-----------------------------------
|
|
|
|
Hash Join
|
|
|
|
Output: j1.id, j3.id
|
|
|
|
Inner Unique: true
|
|
|
|
Hash Cond: (j3.id = j1.id)
|
|
|
|
-> Seq Scan on public.j3
|
|
|
|
Output: j3.id
|
|
|
|
-> Hash
|
|
|
|
Output: j1.id
|
|
|
|
-> Seq Scan on public.j1
|
|
|
|
Output: j1.id
|
|
|
|
(10 rows)
|
|
|
|
|
|
|
|
-- ensure left join is marked as unique
|
|
|
|
explain (verbose, costs off)
|
|
|
|
select * from j1 left join j2 on j1.id = j2.id;
|
|
|
|
QUERY PLAN
|
|
|
|
-----------------------------------
|
|
|
|
Hash Left Join
|
|
|
|
Output: j1.id, j2.id
|
|
|
|
Inner Unique: true
|
|
|
|
Hash Cond: (j1.id = j2.id)
|
|
|
|
-> Seq Scan on public.j1
|
|
|
|
Output: j1.id
|
|
|
|
-> Hash
|
|
|
|
Output: j2.id
|
|
|
|
-> Seq Scan on public.j2
|
|
|
|
Output: j2.id
|
|
|
|
(10 rows)
|
|
|
|
|
|
|
|
-- ensure right join is marked as unique
|
|
|
|
explain (verbose, costs off)
|
|
|
|
select * from j1 right join j2 on j1.id = j2.id;
|
|
|
|
QUERY PLAN
|
|
|
|
-----------------------------------
|
|
|
|
Hash Left Join
|
|
|
|
Output: j1.id, j2.id
|
|
|
|
Inner Unique: true
|
|
|
|
Hash Cond: (j2.id = j1.id)
|
|
|
|
-> Seq Scan on public.j2
|
|
|
|
Output: j2.id
|
|
|
|
-> Hash
|
|
|
|
Output: j1.id
|
|
|
|
-> Seq Scan on public.j1
|
|
|
|
Output: j1.id
|
|
|
|
(10 rows)
|
|
|
|
|
|
|
|
-- ensure full join is marked as unique
|
|
|
|
explain (verbose, costs off)
|
|
|
|
select * from j1 full join j2 on j1.id = j2.id;
|
|
|
|
QUERY PLAN
|
|
|
|
-----------------------------------
|
|
|
|
Hash Full Join
|
|
|
|
Output: j1.id, j2.id
|
|
|
|
Inner Unique: true
|
|
|
|
Hash Cond: (j1.id = j2.id)
|
|
|
|
-> Seq Scan on public.j1
|
|
|
|
Output: j1.id
|
|
|
|
-> Hash
|
|
|
|
Output: j2.id
|
|
|
|
-> Seq Scan on public.j2
|
|
|
|
Output: j2.id
|
|
|
|
(10 rows)
|
|
|
|
|
|
|
|
-- a clauseless (cross) join can't be unique
|
|
|
|
explain (verbose, costs off)
|
|
|
|
select * from j1 cross join j2;
|
|
|
|
QUERY PLAN
|
|
|
|
-----------------------------------
|
|
|
|
Nested Loop
|
|
|
|
Output: j1.id, j2.id
|
|
|
|
-> Seq Scan on public.j1
|
|
|
|
Output: j1.id
|
|
|
|
-> Materialize
|
|
|
|
Output: j2.id
|
|
|
|
-> Seq Scan on public.j2
|
|
|
|
Output: j2.id
|
|
|
|
(8 rows)
|
|
|
|
|
|
|
|
-- ensure a natural join is marked as unique
|
|
|
|
explain (verbose, costs off)
|
|
|
|
select * from j1 natural join j2;
|
|
|
|
QUERY PLAN
|
|
|
|
-----------------------------------
|
|
|
|
Hash Join
|
|
|
|
Output: j1.id
|
|
|
|
Inner Unique: true
|
|
|
|
Hash Cond: (j1.id = j2.id)
|
|
|
|
-> Seq Scan on public.j1
|
|
|
|
Output: j1.id
|
|
|
|
-> Hash
|
|
|
|
Output: j2.id
|
|
|
|
-> Seq Scan on public.j2
|
|
|
|
Output: j2.id
|
|
|
|
(10 rows)
|
|
|
|
|
|
|
|
-- ensure a distinct clause allows the inner to become unique
|
|
|
|
explain (verbose, costs off)
|
|
|
|
select * from j1
|
|
|
|
inner join (select distinct id from j3) j3 on j1.id = j3.id;
|
Fix old corner-case logic error in final_cost_nestloop().
When costing a nestloop with stop-at-first-inner-match semantics, and a
non-indexscan inner path, final_cost_nestloop() wants to charge the full
scan cost of the inner rel at least once, with additional scans charged
at inner_rescan_run_cost which might be less. However the logic for
doing this effectively assumed that outer_matched_rows is at least 1.
If it's zero, which is not unlikely for a small outer rel, we ended up
charging inner_run_cost plus N times inner_rescan_run_cost, as much as
double the correct charge for an outer rel with only one row that
we're betting won't be matched. (Unless the inner rel is materialized,
in which case it has very small inner_rescan_run_cost and the cost
is not so far off what it should have been.)
The upshot of this was that the planner had a tendency to select plans
that failed to make effective use of the stop-at-first-inner-match
semantics, and that might have Materialize nodes in them even when the
predicted number of executions of the Materialize subplan was only 1.
This was not so obvious before commit 9c7f5229a, because the case only
arose in connection with semi/anti joins where there's not freedom to
reverse the join order. But with the addition of unique-inner joins,
it could result in some fairly bad planning choices, as reported by
Teodor Sigaev. Indeed, some of the test cases added by that commit
have plans that look dubious on closer inspection, and are changed
by this patch.
Fix the logic to ensure that we don't charge for too many inner scans.
I chose to adjust it so that the full-freight scan cost is associated
with an unmatched outer row if possible, not a matched one, since that
seems like a better model of what would happen at runtime.
This is a longstanding bug, but given the lesser impact in back branches,
and the lack of field complaints, I won't risk a back-patch.
Discussion: https://postgr.es/m/CAKJS1f-LzkUsFxdJ_-Luy38orQ+AdEXM5o+vANR+-pHAWPSecg@mail.gmail.com
2017-06-03 19:48:15 +02:00
|
|
|
QUERY PLAN
|
|
|
|
-----------------------------------------
|
2017-04-08 04:20:03 +02:00
|
|
|
Nested Loop
|
|
|
|
Output: j1.id, j3.id
|
|
|
|
Inner Unique: true
|
|
|
|
Join Filter: (j1.id = j3.id)
|
Fix old corner-case logic error in final_cost_nestloop().
When costing a nestloop with stop-at-first-inner-match semantics, and a
non-indexscan inner path, final_cost_nestloop() wants to charge the full
scan cost of the inner rel at least once, with additional scans charged
at inner_rescan_run_cost which might be less. However the logic for
doing this effectively assumed that outer_matched_rows is at least 1.
If it's zero, which is not unlikely for a small outer rel, we ended up
charging inner_run_cost plus N times inner_rescan_run_cost, as much as
double the correct charge for an outer rel with only one row that
we're betting won't be matched. (Unless the inner rel is materialized,
in which case it has very small inner_rescan_run_cost and the cost
is not so far off what it should have been.)
The upshot of this was that the planner had a tendency to select plans
that failed to make effective use of the stop-at-first-inner-match
semantics, and that might have Materialize nodes in them even when the
predicted number of executions of the Materialize subplan was only 1.
This was not so obvious before commit 9c7f5229a, because the case only
arose in connection with semi/anti joins where there's not freedom to
reverse the join order. But with the addition of unique-inner joins,
it could result in some fairly bad planning choices, as reported by
Teodor Sigaev. Indeed, some of the test cases added by that commit
have plans that look dubious on closer inspection, and are changed
by this patch.
Fix the logic to ensure that we don't charge for too many inner scans.
I chose to adjust it so that the full-freight scan cost is associated
with an unmatched outer row if possible, not a matched one, since that
seems like a better model of what would happen at runtime.
This is a longstanding bug, but given the lesser impact in back branches,
and the lack of field complaints, I won't risk a back-patch.
Discussion: https://postgr.es/m/CAKJS1f-LzkUsFxdJ_-Luy38orQ+AdEXM5o+vANR+-pHAWPSecg@mail.gmail.com
2017-06-03 19:48:15 +02:00
|
|
|
-> Unique
|
2017-04-08 04:20:03 +02:00
|
|
|
Output: j3.id
|
Fix old corner-case logic error in final_cost_nestloop().
When costing a nestloop with stop-at-first-inner-match semantics, and a
non-indexscan inner path, final_cost_nestloop() wants to charge the full
scan cost of the inner rel at least once, with additional scans charged
at inner_rescan_run_cost which might be less. However the logic for
doing this effectively assumed that outer_matched_rows is at least 1.
If it's zero, which is not unlikely for a small outer rel, we ended up
charging inner_run_cost plus N times inner_rescan_run_cost, as much as
double the correct charge for an outer rel with only one row that
we're betting won't be matched. (Unless the inner rel is materialized,
in which case it has very small inner_rescan_run_cost and the cost
is not so far off what it should have been.)
The upshot of this was that the planner had a tendency to select plans
that failed to make effective use of the stop-at-first-inner-match
semantics, and that might have Materialize nodes in them even when the
predicted number of executions of the Materialize subplan was only 1.
This was not so obvious before commit 9c7f5229a, because the case only
arose in connection with semi/anti joins where there's not freedom to
reverse the join order. But with the addition of unique-inner joins,
it could result in some fairly bad planning choices, as reported by
Teodor Sigaev. Indeed, some of the test cases added by that commit
have plans that look dubious on closer inspection, and are changed
by this patch.
Fix the logic to ensure that we don't charge for too many inner scans.
I chose to adjust it so that the full-freight scan cost is associated
with an unmatched outer row if possible, not a matched one, since that
seems like a better model of what would happen at runtime.
This is a longstanding bug, but given the lesser impact in back branches,
and the lack of field complaints, I won't risk a back-patch.
Discussion: https://postgr.es/m/CAKJS1f-LzkUsFxdJ_-Luy38orQ+AdEXM5o+vANR+-pHAWPSecg@mail.gmail.com
2017-06-03 19:48:15 +02:00
|
|
|
-> Sort
|
2017-04-08 04:20:03 +02:00
|
|
|
Output: j3.id
|
Fix old corner-case logic error in final_cost_nestloop().
When costing a nestloop with stop-at-first-inner-match semantics, and a
non-indexscan inner path, final_cost_nestloop() wants to charge the full
scan cost of the inner rel at least once, with additional scans charged
at inner_rescan_run_cost which might be less. However the logic for
doing this effectively assumed that outer_matched_rows is at least 1.
If it's zero, which is not unlikely for a small outer rel, we ended up
charging inner_run_cost plus N times inner_rescan_run_cost, as much as
double the correct charge for an outer rel with only one row that
we're betting won't be matched. (Unless the inner rel is materialized,
in which case it has very small inner_rescan_run_cost and the cost
is not so far off what it should have been.)
The upshot of this was that the planner had a tendency to select plans
that failed to make effective use of the stop-at-first-inner-match
semantics, and that might have Materialize nodes in them even when the
predicted number of executions of the Materialize subplan was only 1.
This was not so obvious before commit 9c7f5229a, because the case only
arose in connection with semi/anti joins where there's not freedom to
reverse the join order. But with the addition of unique-inner joins,
it could result in some fairly bad planning choices, as reported by
Teodor Sigaev. Indeed, some of the test cases added by that commit
have plans that look dubious on closer inspection, and are changed
by this patch.
Fix the logic to ensure that we don't charge for too many inner scans.
I chose to adjust it so that the full-freight scan cost is associated
with an unmatched outer row if possible, not a matched one, since that
seems like a better model of what would happen at runtime.
This is a longstanding bug, but given the lesser impact in back branches,
and the lack of field complaints, I won't risk a back-patch.
Discussion: https://postgr.es/m/CAKJS1f-LzkUsFxdJ_-Luy38orQ+AdEXM5o+vANR+-pHAWPSecg@mail.gmail.com
2017-06-03 19:48:15 +02:00
|
|
|
Sort Key: j3.id
|
|
|
|
-> Seq Scan on public.j3
|
2017-04-08 04:20:03 +02:00
|
|
|
Output: j3.id
|
Fix old corner-case logic error in final_cost_nestloop().
When costing a nestloop with stop-at-first-inner-match semantics, and a
non-indexscan inner path, final_cost_nestloop() wants to charge the full
scan cost of the inner rel at least once, with additional scans charged
at inner_rescan_run_cost which might be less. However the logic for
doing this effectively assumed that outer_matched_rows is at least 1.
If it's zero, which is not unlikely for a small outer rel, we ended up
charging inner_run_cost plus N times inner_rescan_run_cost, as much as
double the correct charge for an outer rel with only one row that
we're betting won't be matched. (Unless the inner rel is materialized,
in which case it has very small inner_rescan_run_cost and the cost
is not so far off what it should have been.)
The upshot of this was that the planner had a tendency to select plans
that failed to make effective use of the stop-at-first-inner-match
semantics, and that might have Materialize nodes in them even when the
predicted number of executions of the Materialize subplan was only 1.
This was not so obvious before commit 9c7f5229a, because the case only
arose in connection with semi/anti joins where there's not freedom to
reverse the join order. But with the addition of unique-inner joins,
it could result in some fairly bad planning choices, as reported by
Teodor Sigaev. Indeed, some of the test cases added by that commit
have plans that look dubious on closer inspection, and are changed
by this patch.
Fix the logic to ensure that we don't charge for too many inner scans.
I chose to adjust it so that the full-freight scan cost is associated
with an unmatched outer row if possible, not a matched one, since that
seems like a better model of what would happen at runtime.
This is a longstanding bug, but given the lesser impact in back branches,
and the lack of field complaints, I won't risk a back-patch.
Discussion: https://postgr.es/m/CAKJS1f-LzkUsFxdJ_-Luy38orQ+AdEXM5o+vANR+-pHAWPSecg@mail.gmail.com
2017-06-03 19:48:15 +02:00
|
|
|
-> Seq Scan on public.j1
|
|
|
|
Output: j1.id
|
|
|
|
(13 rows)
|
2017-04-08 04:20:03 +02:00
|
|
|
|
|
|
|
-- ensure group by clause allows the inner to become unique
|
|
|
|
explain (verbose, costs off)
|
|
|
|
select * from j1
|
|
|
|
inner join (select id from j3 group by id) j3 on j1.id = j3.id;
|
Fix old corner-case logic error in final_cost_nestloop().
When costing a nestloop with stop-at-first-inner-match semantics, and a
non-indexscan inner path, final_cost_nestloop() wants to charge the full
scan cost of the inner rel at least once, with additional scans charged
at inner_rescan_run_cost which might be less. However the logic for
doing this effectively assumed that outer_matched_rows is at least 1.
If it's zero, which is not unlikely for a small outer rel, we ended up
charging inner_run_cost plus N times inner_rescan_run_cost, as much as
double the correct charge for an outer rel with only one row that
we're betting won't be matched. (Unless the inner rel is materialized,
in which case it has very small inner_rescan_run_cost and the cost
is not so far off what it should have been.)
The upshot of this was that the planner had a tendency to select plans
that failed to make effective use of the stop-at-first-inner-match
semantics, and that might have Materialize nodes in them even when the
predicted number of executions of the Materialize subplan was only 1.
This was not so obvious before commit 9c7f5229a, because the case only
arose in connection with semi/anti joins where there's not freedom to
reverse the join order. But with the addition of unique-inner joins,
it could result in some fairly bad planning choices, as reported by
Teodor Sigaev. Indeed, some of the test cases added by that commit
have plans that look dubious on closer inspection, and are changed
by this patch.
Fix the logic to ensure that we don't charge for too many inner scans.
I chose to adjust it so that the full-freight scan cost is associated
with an unmatched outer row if possible, not a matched one, since that
seems like a better model of what would happen at runtime.
This is a longstanding bug, but given the lesser impact in back branches,
and the lack of field complaints, I won't risk a back-patch.
Discussion: https://postgr.es/m/CAKJS1f-LzkUsFxdJ_-Luy38orQ+AdEXM5o+vANR+-pHAWPSecg@mail.gmail.com
2017-06-03 19:48:15 +02:00
|
|
|
QUERY PLAN
|
|
|
|
-----------------------------------------
|
2017-04-08 04:20:03 +02:00
|
|
|
Nested Loop
|
|
|
|
Output: j1.id, j3.id
|
|
|
|
Inner Unique: true
|
|
|
|
Join Filter: (j1.id = j3.id)
|
Fix old corner-case logic error in final_cost_nestloop().
When costing a nestloop with stop-at-first-inner-match semantics, and a
non-indexscan inner path, final_cost_nestloop() wants to charge the full
scan cost of the inner rel at least once, with additional scans charged
at inner_rescan_run_cost which might be less. However the logic for
doing this effectively assumed that outer_matched_rows is at least 1.
If it's zero, which is not unlikely for a small outer rel, we ended up
charging inner_run_cost plus N times inner_rescan_run_cost, as much as
double the correct charge for an outer rel with only one row that
we're betting won't be matched. (Unless the inner rel is materialized,
in which case it has very small inner_rescan_run_cost and the cost
is not so far off what it should have been.)
The upshot of this was that the planner had a tendency to select plans
that failed to make effective use of the stop-at-first-inner-match
semantics, and that might have Materialize nodes in them even when the
predicted number of executions of the Materialize subplan was only 1.
This was not so obvious before commit 9c7f5229a, because the case only
arose in connection with semi/anti joins where there's not freedom to
reverse the join order. But with the addition of unique-inner joins,
it could result in some fairly bad planning choices, as reported by
Teodor Sigaev. Indeed, some of the test cases added by that commit
have plans that look dubious on closer inspection, and are changed
by this patch.
Fix the logic to ensure that we don't charge for too many inner scans.
I chose to adjust it so that the full-freight scan cost is associated
with an unmatched outer row if possible, not a matched one, since that
seems like a better model of what would happen at runtime.
This is a longstanding bug, but given the lesser impact in back branches,
and the lack of field complaints, I won't risk a back-patch.
Discussion: https://postgr.es/m/CAKJS1f-LzkUsFxdJ_-Luy38orQ+AdEXM5o+vANR+-pHAWPSecg@mail.gmail.com
2017-06-03 19:48:15 +02:00
|
|
|
-> Group
|
2017-04-08 04:20:03 +02:00
|
|
|
Output: j3.id
|
Fix old corner-case logic error in final_cost_nestloop().
When costing a nestloop with stop-at-first-inner-match semantics, and a
non-indexscan inner path, final_cost_nestloop() wants to charge the full
scan cost of the inner rel at least once, with additional scans charged
at inner_rescan_run_cost which might be less. However the logic for
doing this effectively assumed that outer_matched_rows is at least 1.
If it's zero, which is not unlikely for a small outer rel, we ended up
charging inner_run_cost plus N times inner_rescan_run_cost, as much as
double the correct charge for an outer rel with only one row that
we're betting won't be matched. (Unless the inner rel is materialized,
in which case it has very small inner_rescan_run_cost and the cost
is not so far off what it should have been.)
The upshot of this was that the planner had a tendency to select plans
that failed to make effective use of the stop-at-first-inner-match
semantics, and that might have Materialize nodes in them even when the
predicted number of executions of the Materialize subplan was only 1.
This was not so obvious before commit 9c7f5229a, because the case only
arose in connection with semi/anti joins where there's not freedom to
reverse the join order. But with the addition of unique-inner joins,
it could result in some fairly bad planning choices, as reported by
Teodor Sigaev. Indeed, some of the test cases added by that commit
have plans that look dubious on closer inspection, and are changed
by this patch.
Fix the logic to ensure that we don't charge for too many inner scans.
I chose to adjust it so that the full-freight scan cost is associated
with an unmatched outer row if possible, not a matched one, since that
seems like a better model of what would happen at runtime.
This is a longstanding bug, but given the lesser impact in back branches,
and the lack of field complaints, I won't risk a back-patch.
Discussion: https://postgr.es/m/CAKJS1f-LzkUsFxdJ_-Luy38orQ+AdEXM5o+vANR+-pHAWPSecg@mail.gmail.com
2017-06-03 19:48:15 +02:00
|
|
|
Group Key: j3.id
|
|
|
|
-> Sort
|
2017-04-08 04:20:03 +02:00
|
|
|
Output: j3.id
|
Fix old corner-case logic error in final_cost_nestloop().
When costing a nestloop with stop-at-first-inner-match semantics, and a
non-indexscan inner path, final_cost_nestloop() wants to charge the full
scan cost of the inner rel at least once, with additional scans charged
at inner_rescan_run_cost which might be less. However the logic for
doing this effectively assumed that outer_matched_rows is at least 1.
If it's zero, which is not unlikely for a small outer rel, we ended up
charging inner_run_cost plus N times inner_rescan_run_cost, as much as
double the correct charge for an outer rel with only one row that
we're betting won't be matched. (Unless the inner rel is materialized,
in which case it has very small inner_rescan_run_cost and the cost
is not so far off what it should have been.)
The upshot of this was that the planner had a tendency to select plans
that failed to make effective use of the stop-at-first-inner-match
semantics, and that might have Materialize nodes in them even when the
predicted number of executions of the Materialize subplan was only 1.
This was not so obvious before commit 9c7f5229a, because the case only
arose in connection with semi/anti joins where there's not freedom to
reverse the join order. But with the addition of unique-inner joins,
it could result in some fairly bad planning choices, as reported by
Teodor Sigaev. Indeed, some of the test cases added by that commit
have plans that look dubious on closer inspection, and are changed
by this patch.
Fix the logic to ensure that we don't charge for too many inner scans.
I chose to adjust it so that the full-freight scan cost is associated
with an unmatched outer row if possible, not a matched one, since that
seems like a better model of what would happen at runtime.
This is a longstanding bug, but given the lesser impact in back branches,
and the lack of field complaints, I won't risk a back-patch.
Discussion: https://postgr.es/m/CAKJS1f-LzkUsFxdJ_-Luy38orQ+AdEXM5o+vANR+-pHAWPSecg@mail.gmail.com
2017-06-03 19:48:15 +02:00
|
|
|
Sort Key: j3.id
|
|
|
|
-> Seq Scan on public.j3
|
2017-04-08 04:20:03 +02:00
|
|
|
Output: j3.id
|
Fix old corner-case logic error in final_cost_nestloop().
When costing a nestloop with stop-at-first-inner-match semantics, and a
non-indexscan inner path, final_cost_nestloop() wants to charge the full
scan cost of the inner rel at least once, with additional scans charged
at inner_rescan_run_cost which might be less. However the logic for
doing this effectively assumed that outer_matched_rows is at least 1.
If it's zero, which is not unlikely for a small outer rel, we ended up
charging inner_run_cost plus N times inner_rescan_run_cost, as much as
double the correct charge for an outer rel with only one row that
we're betting won't be matched. (Unless the inner rel is materialized,
in which case it has very small inner_rescan_run_cost and the cost
is not so far off what it should have been.)
The upshot of this was that the planner had a tendency to select plans
that failed to make effective use of the stop-at-first-inner-match
semantics, and that might have Materialize nodes in them even when the
predicted number of executions of the Materialize subplan was only 1.
This was not so obvious before commit 9c7f5229a, because the case only
arose in connection with semi/anti joins where there's not freedom to
reverse the join order. But with the addition of unique-inner joins,
it could result in some fairly bad planning choices, as reported by
Teodor Sigaev. Indeed, some of the test cases added by that commit
have plans that look dubious on closer inspection, and are changed
by this patch.
Fix the logic to ensure that we don't charge for too many inner scans.
I chose to adjust it so that the full-freight scan cost is associated
with an unmatched outer row if possible, not a matched one, since that
seems like a better model of what would happen at runtime.
This is a longstanding bug, but given the lesser impact in back branches,
and the lack of field complaints, I won't risk a back-patch.
Discussion: https://postgr.es/m/CAKJS1f-LzkUsFxdJ_-Luy38orQ+AdEXM5o+vANR+-pHAWPSecg@mail.gmail.com
2017-06-03 19:48:15 +02:00
|
|
|
-> Seq Scan on public.j1
|
|
|
|
Output: j1.id
|
|
|
|
(14 rows)
|
2017-04-08 04:20:03 +02:00
|
|
|
|
|
|
|
drop table j1;
|
|
|
|
drop table j2;
|
|
|
|
drop table j3;
|
|
|
|
-- test more complex permutations of unique joins
|
|
|
|
create table j1 (id1 int, id2 int, primary key(id1,id2));
|
|
|
|
create table j2 (id1 int, id2 int, primary key(id1,id2));
|
|
|
|
create table j3 (id1 int, id2 int, primary key(id1,id2));
|
|
|
|
insert into j1 values(1,1),(1,2);
|
|
|
|
insert into j2 values(1,1);
|
|
|
|
insert into j3 values(1,1);
|
|
|
|
analyze j1;
|
|
|
|
analyze j2;
|
|
|
|
analyze j3;
|
|
|
|
-- ensure there's no unique join when not all columns which are part of the
|
|
|
|
-- unique index are seen in the join clause
|
|
|
|
explain (verbose, costs off)
|
|
|
|
select * from j1
|
|
|
|
inner join j2 on j1.id1 = j2.id1;
|
|
|
|
QUERY PLAN
|
|
|
|
------------------------------------------
|
|
|
|
Nested Loop
|
|
|
|
Output: j1.id1, j1.id2, j2.id1, j2.id2
|
|
|
|
Join Filter: (j1.id1 = j2.id1)
|
|
|
|
-> Seq Scan on public.j2
|
|
|
|
Output: j2.id1, j2.id2
|
|
|
|
-> Seq Scan on public.j1
|
|
|
|
Output: j1.id1, j1.id2
|
|
|
|
(7 rows)
|
|
|
|
|
|
|
|
-- ensure proper unique detection with multiple join quals
|
|
|
|
explain (verbose, costs off)
|
|
|
|
select * from j1
|
|
|
|
inner join j2 on j1.id1 = j2.id1 and j1.id2 = j2.id2;
|
|
|
|
QUERY PLAN
|
|
|
|
----------------------------------------------------------
|
|
|
|
Nested Loop
|
|
|
|
Output: j1.id1, j1.id2, j2.id1, j2.id2
|
|
|
|
Inner Unique: true
|
|
|
|
Join Filter: ((j1.id1 = j2.id1) AND (j1.id2 = j2.id2))
|
Fix old corner-case logic error in final_cost_nestloop().
When costing a nestloop with stop-at-first-inner-match semantics, and a
non-indexscan inner path, final_cost_nestloop() wants to charge the full
scan cost of the inner rel at least once, with additional scans charged
at inner_rescan_run_cost which might be less. However the logic for
doing this effectively assumed that outer_matched_rows is at least 1.
If it's zero, which is not unlikely for a small outer rel, we ended up
charging inner_run_cost plus N times inner_rescan_run_cost, as much as
double the correct charge for an outer rel with only one row that
we're betting won't be matched. (Unless the inner rel is materialized,
in which case it has very small inner_rescan_run_cost and the cost
is not so far off what it should have been.)
The upshot of this was that the planner had a tendency to select plans
that failed to make effective use of the stop-at-first-inner-match
semantics, and that might have Materialize nodes in them even when the
predicted number of executions of the Materialize subplan was only 1.
This was not so obvious before commit 9c7f5229a, because the case only
arose in connection with semi/anti joins where there's not freedom to
reverse the join order. But with the addition of unique-inner joins,
it could result in some fairly bad planning choices, as reported by
Teodor Sigaev. Indeed, some of the test cases added by that commit
have plans that look dubious on closer inspection, and are changed
by this patch.
Fix the logic to ensure that we don't charge for too many inner scans.
I chose to adjust it so that the full-freight scan cost is associated
with an unmatched outer row if possible, not a matched one, since that
seems like a better model of what would happen at runtime.
This is a longstanding bug, but given the lesser impact in back branches,
and the lack of field complaints, I won't risk a back-patch.
Discussion: https://postgr.es/m/CAKJS1f-LzkUsFxdJ_-Luy38orQ+AdEXM5o+vANR+-pHAWPSecg@mail.gmail.com
2017-06-03 19:48:15 +02:00
|
|
|
-> Seq Scan on public.j2
|
|
|
|
Output: j2.id1, j2.id2
|
2017-04-08 04:20:03 +02:00
|
|
|
-> Seq Scan on public.j1
|
|
|
|
Output: j1.id1, j1.id2
|
Fix old corner-case logic error in final_cost_nestloop().
When costing a nestloop with stop-at-first-inner-match semantics, and a
non-indexscan inner path, final_cost_nestloop() wants to charge the full
scan cost of the inner rel at least once, with additional scans charged
at inner_rescan_run_cost which might be less. However the logic for
doing this effectively assumed that outer_matched_rows is at least 1.
If it's zero, which is not unlikely for a small outer rel, we ended up
charging inner_run_cost plus N times inner_rescan_run_cost, as much as
double the correct charge for an outer rel with only one row that
we're betting won't be matched. (Unless the inner rel is materialized,
in which case it has very small inner_rescan_run_cost and the cost
is not so far off what it should have been.)
The upshot of this was that the planner had a tendency to select plans
that failed to make effective use of the stop-at-first-inner-match
semantics, and that might have Materialize nodes in them even when the
predicted number of executions of the Materialize subplan was only 1.
This was not so obvious before commit 9c7f5229a, because the case only
arose in connection with semi/anti joins where there's not freedom to
reverse the join order. But with the addition of unique-inner joins,
it could result in some fairly bad planning choices, as reported by
Teodor Sigaev. Indeed, some of the test cases added by that commit
have plans that look dubious on closer inspection, and are changed
by this patch.
Fix the logic to ensure that we don't charge for too many inner scans.
I chose to adjust it so that the full-freight scan cost is associated
with an unmatched outer row if possible, not a matched one, since that
seems like a better model of what would happen at runtime.
This is a longstanding bug, but given the lesser impact in back branches,
and the lack of field complaints, I won't risk a back-patch.
Discussion: https://postgr.es/m/CAKJS1f-LzkUsFxdJ_-Luy38orQ+AdEXM5o+vANR+-pHAWPSecg@mail.gmail.com
2017-06-03 19:48:15 +02:00
|
|
|
(8 rows)
|
2017-04-08 04:20:03 +02:00
|
|
|
|
|
|
|
-- ensure we don't detect the join to be unique when quals are not part of the
|
|
|
|
-- join condition
|
|
|
|
explain (verbose, costs off)
|
|
|
|
select * from j1
|
|
|
|
inner join j2 on j1.id1 = j2.id1 where j1.id2 = 1;
|
|
|
|
QUERY PLAN
|
|
|
|
------------------------------------------
|
|
|
|
Nested Loop
|
|
|
|
Output: j1.id1, j1.id2, j2.id1, j2.id2
|
|
|
|
Join Filter: (j1.id1 = j2.id1)
|
|
|
|
-> Seq Scan on public.j1
|
|
|
|
Output: j1.id1, j1.id2
|
|
|
|
Filter: (j1.id2 = 1)
|
|
|
|
-> Seq Scan on public.j2
|
|
|
|
Output: j2.id1, j2.id2
|
|
|
|
(8 rows)
|
|
|
|
|
|
|
|
-- as above, but for left joins.
|
|
|
|
explain (verbose, costs off)
|
|
|
|
select * from j1
|
|
|
|
left join j2 on j1.id1 = j2.id1 where j1.id2 = 1;
|
|
|
|
QUERY PLAN
|
|
|
|
------------------------------------------
|
|
|
|
Nested Loop Left Join
|
|
|
|
Output: j1.id1, j1.id2, j2.id1, j2.id2
|
|
|
|
Join Filter: (j1.id1 = j2.id1)
|
|
|
|
-> Seq Scan on public.j1
|
|
|
|
Output: j1.id1, j1.id2
|
|
|
|
Filter: (j1.id2 = 1)
|
|
|
|
-> Seq Scan on public.j2
|
|
|
|
Output: j2.id1, j2.id2
|
|
|
|
(8 rows)
|
|
|
|
|
|
|
|
-- validate logic in merge joins which skips mark and restore.
|
|
|
|
-- it should only do this if all quals which were used to detect the unique
|
|
|
|
-- are present as join quals, and not plain quals.
|
|
|
|
set enable_nestloop to 0;
|
|
|
|
set enable_hashjoin to 0;
|
|
|
|
set enable_sort to 0;
|
|
|
|
-- create an index that will be preferred over the PK to perform the join
|
|
|
|
create index j1_id1_idx on j1 (id1) where id1 % 1000 = 1;
|
|
|
|
explain (costs off) select * from j1 j1
|
|
|
|
inner join j1 j2 on j1.id1 = j2.id1 and j1.id2 = j2.id2
|
|
|
|
where j1.id1 % 1000 = 1 and j2.id1 % 1000 = 1;
|
|
|
|
QUERY PLAN
|
|
|
|
--------------------------------------------
|
|
|
|
Merge Join
|
|
|
|
Merge Cond: (j1.id1 = j2.id1)
|
|
|
|
Join Filter: (j1.id2 = j2.id2)
|
|
|
|
-> Index Scan using j1_id1_idx on j1
|
|
|
|
-> Index Scan using j1_id1_idx on j1 j2
|
|
|
|
(5 rows)
|
|
|
|
|
|
|
|
select * from j1 j1
|
|
|
|
inner join j1 j2 on j1.id1 = j2.id1 and j1.id2 = j2.id2
|
|
|
|
where j1.id1 % 1000 = 1 and j2.id1 % 1000 = 1;
|
|
|
|
id1 | id2 | id1 | id2
|
|
|
|
-----+-----+-----+-----
|
|
|
|
1 | 1 | 1 | 1
|
|
|
|
1 | 2 | 1 | 2
|
|
|
|
(2 rows)
|
|
|
|
|
|
|
|
reset enable_nestloop;
|
|
|
|
reset enable_hashjoin;
|
|
|
|
reset enable_sort;
|
|
|
|
drop table j1;
|
|
|
|
drop table j2;
|
|
|
|
drop table j3;
|
2017-05-01 20:39:11 +02:00
|
|
|
-- check that semijoin inner is not seen as unique for a portion of the outerrel
|
|
|
|
explain (verbose, costs off)
|
|
|
|
select t1.unique1, t2.hundred
|
|
|
|
from onek t1, tenk1 t2
|
|
|
|
where exists (select 1 from tenk1 t3
|
|
|
|
where t3.thousand = t1.unique1 and t3.tenthous = t2.hundred)
|
|
|
|
and t1.unique1 < 1;
|
|
|
|
QUERY PLAN
|
|
|
|
---------------------------------------------------------------------------------
|
|
|
|
Nested Loop
|
|
|
|
Output: t1.unique1, t2.hundred
|
|
|
|
-> Hash Join
|
|
|
|
Output: t1.unique1, t3.tenthous
|
|
|
|
Hash Cond: (t3.thousand = t1.unique1)
|
|
|
|
-> HashAggregate
|
|
|
|
Output: t3.thousand, t3.tenthous
|
|
|
|
Group Key: t3.thousand, t3.tenthous
|
|
|
|
-> Index Only Scan using tenk1_thous_tenthous on public.tenk1 t3
|
|
|
|
Output: t3.thousand, t3.tenthous
|
|
|
|
-> Hash
|
|
|
|
Output: t1.unique1
|
|
|
|
-> Index Only Scan using onek_unique1 on public.onek t1
|
|
|
|
Output: t1.unique1
|
|
|
|
Index Cond: (t1.unique1 < 1)
|
|
|
|
-> Index Only Scan using tenk1_hundred on public.tenk1 t2
|
|
|
|
Output: t2.hundred
|
|
|
|
Index Cond: (t2.hundred = t3.tenthous)
|
|
|
|
(18 rows)
|
|
|
|
|
2017-05-01 20:53:42 +02:00
|
|
|
-- ... unless it actually is unique
|
|
|
|
create table j3 as select unique1, tenthous from onek;
|
|
|
|
vacuum analyze j3;
|
|
|
|
create unique index on j3(unique1, tenthous);
|
|
|
|
explain (verbose, costs off)
|
|
|
|
select t1.unique1, t2.hundred
|
|
|
|
from onek t1, tenk1 t2
|
|
|
|
where exists (select 1 from j3
|
|
|
|
where j3.unique1 = t1.unique1 and j3.tenthous = t2.hundred)
|
|
|
|
and t1.unique1 < 1;
|
|
|
|
QUERY PLAN
|
|
|
|
------------------------------------------------------------------------
|
|
|
|
Nested Loop
|
|
|
|
Output: t1.unique1, t2.hundred
|
|
|
|
-> Nested Loop
|
|
|
|
Output: t1.unique1, j3.tenthous
|
|
|
|
-> Index Only Scan using onek_unique1 on public.onek t1
|
|
|
|
Output: t1.unique1
|
|
|
|
Index Cond: (t1.unique1 < 1)
|
|
|
|
-> Index Only Scan using j3_unique1_tenthous_idx on public.j3
|
|
|
|
Output: j3.unique1, j3.tenthous
|
|
|
|
Index Cond: (j3.unique1 = t1.unique1)
|
|
|
|
-> Index Only Scan using tenk1_hundred on public.tenk1 t2
|
|
|
|
Output: t2.hundred
|
|
|
|
Index Cond: (t2.hundred = j3.tenthous)
|
|
|
|
(13 rows)
|
|
|
|
|
|
|
|
drop table j3;
|
2017-11-30 01:06:50 +01:00
|
|
|
--
|
|
|
|
-- exercises for the hash join code
|
|
|
|
--
|
|
|
|
begin;
|
|
|
|
set local min_parallel_table_scan_size = 0;
|
|
|
|
set local parallel_setup_cost = 0;
|
|
|
|
-- Extract bucket and batch counts from an explain analyze plan. In
|
|
|
|
-- general we can't make assertions about how many batches (or
|
|
|
|
-- buckets) will be required because it can vary, but we can in some
|
|
|
|
-- special cases and we can check for growth.
|
|
|
|
create or replace function find_hash(node json)
|
|
|
|
returns json language plpgsql
|
|
|
|
as
|
|
|
|
$$
|
|
|
|
declare
|
|
|
|
x json;
|
|
|
|
child json;
|
|
|
|
begin
|
|
|
|
if node->>'Node Type' = 'Hash' then
|
|
|
|
return node;
|
|
|
|
else
|
|
|
|
for child in select json_array_elements(node->'Plans')
|
|
|
|
loop
|
|
|
|
x := find_hash(child);
|
|
|
|
if x is not null then
|
|
|
|
return x;
|
|
|
|
end if;
|
|
|
|
end loop;
|
|
|
|
return null;
|
|
|
|
end if;
|
|
|
|
end;
|
|
|
|
$$;
|
|
|
|
create or replace function hash_join_batches(query text)
|
|
|
|
returns table (original int, final int) language plpgsql
|
|
|
|
as
|
|
|
|
$$
|
|
|
|
declare
|
|
|
|
whole_plan json;
|
|
|
|
hash_node json;
|
|
|
|
begin
|
|
|
|
for whole_plan in
|
|
|
|
execute 'explain (analyze, format ''json'') ' || query
|
|
|
|
loop
|
|
|
|
hash_node := find_hash(json_extract_path(whole_plan, '0', 'Plan'));
|
|
|
|
original := hash_node->>'Original Hash Batches';
|
|
|
|
final := hash_node->>'Hash Batches';
|
|
|
|
return next;
|
|
|
|
end loop;
|
|
|
|
end;
|
|
|
|
$$;
|
|
|
|
-- Make a simple relation with well distributed keys and correctly
|
|
|
|
-- estimated size.
|
|
|
|
create table simple as
|
|
|
|
select generate_series(1, 20000) AS id, 'aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa';
|
|
|
|
alter table simple set (parallel_workers = 2);
|
|
|
|
analyze simple;
|
|
|
|
-- Make a relation whose size we will under-estimate. We want stats
|
|
|
|
-- to say 1000 rows, but actually there are 20,000 rows.
|
|
|
|
create table bigger_than_it_looks as
|
|
|
|
select generate_series(1, 20000) as id, 'aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa';
|
|
|
|
alter table bigger_than_it_looks set (autovacuum_enabled = 'false');
|
|
|
|
alter table bigger_than_it_looks set (parallel_workers = 2);
|
|
|
|
analyze bigger_than_it_looks;
|
|
|
|
update pg_class set reltuples = 1000 where relname = 'bigger_than_it_looks';
|
|
|
|
-- Make a relation whose size we underestimate and that also has a
|
|
|
|
-- kind of skew that breaks our batching scheme. We want stats to say
|
|
|
|
-- 2 rows, but actually there are 20,000 rows with the same key.
|
|
|
|
create table extremely_skewed (id int, t text);
|
|
|
|
alter table extremely_skewed set (autovacuum_enabled = 'false');
|
|
|
|
alter table extremely_skewed set (parallel_workers = 2);
|
|
|
|
analyze extremely_skewed;
|
|
|
|
insert into extremely_skewed
|
|
|
|
select 42 as id, 'aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa'
|
|
|
|
from generate_series(1, 20000);
|
|
|
|
update pg_class
|
|
|
|
set reltuples = 2, relpages = pg_relation_size('extremely_skewed') / 8192
|
|
|
|
where relname = 'extremely_skewed';
|
Add parallel-aware hash joins.
Introduce parallel-aware hash joins that appear in EXPLAIN plans as Parallel
Hash Join with Parallel Hash. While hash joins could already appear in
parallel queries, they were previously always parallel-oblivious and had a
partial subplan only on the outer side, meaning that the work of the inner
subplan was duplicated in every worker.
After this commit, the planner will consider using a partial subplan on the
inner side too, using the Parallel Hash node to divide the work over the
available CPU cores and combine its results in shared memory. If the join
needs to be split into multiple batches in order to respect work_mem, then
workers process different batches as much as possible and then work together
on the remaining batches.
The advantages of a parallel-aware hash join over a parallel-oblivious hash
join used in a parallel query are that it:
* avoids wasting memory on duplicated hash tables
* avoids wasting disk space on duplicated batch files
* divides the work of building the hash table over the CPUs
One disadvantage is that there is some communication between the participating
CPUs which might outweigh the benefits of parallelism in the case of small
hash tables. This is avoided by the planner's existing reluctance to supply
partial plans for small scans, but it may be necessary to estimate
synchronization costs in future if that situation changes. Another is that
outer batch 0 must be written to disk if multiple batches are required.
A potential future advantage of parallel-aware hash joins is that right and
full outer joins could be supported, since there is a single set of matched
bits for each hashtable, but that is not yet implemented.
A new GUC enable_parallel_hash is defined to control the feature, defaulting
to on.
Author: Thomas Munro
Reviewed-By: Andres Freund, Robert Haas
Tested-By: Rafia Sabih, Prabhat Sahu
Discussion:
https://postgr.es/m/CAEepm=2W=cOkiZxcg6qiFQP-dHUe09aqTrEMM7yJDrHMhDv_RA@mail.gmail.com
https://postgr.es/m/CAEepm=37HKyJ4U6XOLi=JgfSHM3o6B-GaeO-6hkOmneTDkH+Uw@mail.gmail.com
2017-12-21 08:39:21 +01:00
|
|
|
-- Make a relation with a couple of enormous tuples.
|
|
|
|
create table wide as select generate_series(1, 2) as id, rpad('', 320000, 'x') as t;
|
|
|
|
alter table wide set (parallel_workers = 2);
|
2017-11-30 01:06:50 +01:00
|
|
|
-- The "optimal" case: the hash table fits in memory; we plan for 1
|
|
|
|
-- batch, we stick to that number, and peak memory usage stays within
|
|
|
|
-- our work_mem budget
|
|
|
|
-- non-parallel
|
|
|
|
savepoint settings;
|
|
|
|
set local max_parallel_workers_per_gather = 0;
|
|
|
|
set local work_mem = '4MB';
|
|
|
|
explain (costs off)
|
|
|
|
select count(*) from simple r join simple s using (id);
|
|
|
|
QUERY PLAN
|
|
|
|
----------------------------------------
|
|
|
|
Aggregate
|
|
|
|
-> Hash Join
|
|
|
|
Hash Cond: (r.id = s.id)
|
|
|
|
-> Seq Scan on simple r
|
|
|
|
-> Hash
|
|
|
|
-> Seq Scan on simple s
|
|
|
|
(6 rows)
|
|
|
|
|
|
|
|
select count(*) from simple r join simple s using (id);
|
|
|
|
count
|
|
|
|
-------
|
|
|
|
20000
|
|
|
|
(1 row)
|
|
|
|
|
|
|
|
select original > 1 as initially_multibatch, final > original as increased_batches
|
|
|
|
from hash_join_batches(
|
|
|
|
$$
|
|
|
|
select count(*) from simple r join simple s using (id);
|
|
|
|
$$);
|
|
|
|
initially_multibatch | increased_batches
|
|
|
|
----------------------+-------------------
|
|
|
|
f | f
|
|
|
|
(1 row)
|
|
|
|
|
|
|
|
rollback to settings;
|
|
|
|
-- parallel with parallel-oblivious hash join
|
|
|
|
savepoint settings;
|
|
|
|
set local max_parallel_workers_per_gather = 2;
|
|
|
|
set local work_mem = '4MB';
|
Add parallel-aware hash joins.
Introduce parallel-aware hash joins that appear in EXPLAIN plans as Parallel
Hash Join with Parallel Hash. While hash joins could already appear in
parallel queries, they were previously always parallel-oblivious and had a
partial subplan only on the outer side, meaning that the work of the inner
subplan was duplicated in every worker.
After this commit, the planner will consider using a partial subplan on the
inner side too, using the Parallel Hash node to divide the work over the
available CPU cores and combine its results in shared memory. If the join
needs to be split into multiple batches in order to respect work_mem, then
workers process different batches as much as possible and then work together
on the remaining batches.
The advantages of a parallel-aware hash join over a parallel-oblivious hash
join used in a parallel query are that it:
* avoids wasting memory on duplicated hash tables
* avoids wasting disk space on duplicated batch files
* divides the work of building the hash table over the CPUs
One disadvantage is that there is some communication between the participating
CPUs which might outweigh the benefits of parallelism in the case of small
hash tables. This is avoided by the planner's existing reluctance to supply
partial plans for small scans, but it may be necessary to estimate
synchronization costs in future if that situation changes. Another is that
outer batch 0 must be written to disk if multiple batches are required.
A potential future advantage of parallel-aware hash joins is that right and
full outer joins could be supported, since there is a single set of matched
bits for each hashtable, but that is not yet implemented.
A new GUC enable_parallel_hash is defined to control the feature, defaulting
to on.
Author: Thomas Munro
Reviewed-By: Andres Freund, Robert Haas
Tested-By: Rafia Sabih, Prabhat Sahu
Discussion:
https://postgr.es/m/CAEepm=2W=cOkiZxcg6qiFQP-dHUe09aqTrEMM7yJDrHMhDv_RA@mail.gmail.com
https://postgr.es/m/CAEepm=37HKyJ4U6XOLi=JgfSHM3o6B-GaeO-6hkOmneTDkH+Uw@mail.gmail.com
2017-12-21 08:39:21 +01:00
|
|
|
set local enable_parallel_hash = off;
|
2017-11-30 01:06:50 +01:00
|
|
|
explain (costs off)
|
|
|
|
select count(*) from simple r join simple s using (id);
|
|
|
|
QUERY PLAN
|
|
|
|
-------------------------------------------------------
|
|
|
|
Finalize Aggregate
|
|
|
|
-> Gather
|
|
|
|
Workers Planned: 2
|
|
|
|
-> Partial Aggregate
|
|
|
|
-> Hash Join
|
|
|
|
Hash Cond: (r.id = s.id)
|
|
|
|
-> Parallel Seq Scan on simple r
|
|
|
|
-> Hash
|
|
|
|
-> Seq Scan on simple s
|
|
|
|
(9 rows)
|
|
|
|
|
|
|
|
select count(*) from simple r join simple s using (id);
|
|
|
|
count
|
|
|
|
-------
|
|
|
|
20000
|
|
|
|
(1 row)
|
|
|
|
|
|
|
|
select original > 1 as initially_multibatch, final > original as increased_batches
|
|
|
|
from hash_join_batches(
|
|
|
|
$$
|
|
|
|
select count(*) from simple r join simple s using (id);
|
|
|
|
$$);
|
|
|
|
initially_multibatch | increased_batches
|
|
|
|
----------------------+-------------------
|
|
|
|
f | f
|
|
|
|
(1 row)
|
|
|
|
|
Add parallel-aware hash joins.
Introduce parallel-aware hash joins that appear in EXPLAIN plans as Parallel
Hash Join with Parallel Hash. While hash joins could already appear in
parallel queries, they were previously always parallel-oblivious and had a
partial subplan only on the outer side, meaning that the work of the inner
subplan was duplicated in every worker.
After this commit, the planner will consider using a partial subplan on the
inner side too, using the Parallel Hash node to divide the work over the
available CPU cores and combine its results in shared memory. If the join
needs to be split into multiple batches in order to respect work_mem, then
workers process different batches as much as possible and then work together
on the remaining batches.
The advantages of a parallel-aware hash join over a parallel-oblivious hash
join used in a parallel query are that it:
* avoids wasting memory on duplicated hash tables
* avoids wasting disk space on duplicated batch files
* divides the work of building the hash table over the CPUs
One disadvantage is that there is some communication between the participating
CPUs which might outweigh the benefits of parallelism in the case of small
hash tables. This is avoided by the planner's existing reluctance to supply
partial plans for small scans, but it may be necessary to estimate
synchronization costs in future if that situation changes. Another is that
outer batch 0 must be written to disk if multiple batches are required.
A potential future advantage of parallel-aware hash joins is that right and
full outer joins could be supported, since there is a single set of matched
bits for each hashtable, but that is not yet implemented.
A new GUC enable_parallel_hash is defined to control the feature, defaulting
to on.
Author: Thomas Munro
Reviewed-By: Andres Freund, Robert Haas
Tested-By: Rafia Sabih, Prabhat Sahu
Discussion:
https://postgr.es/m/CAEepm=2W=cOkiZxcg6qiFQP-dHUe09aqTrEMM7yJDrHMhDv_RA@mail.gmail.com
https://postgr.es/m/CAEepm=37HKyJ4U6XOLi=JgfSHM3o6B-GaeO-6hkOmneTDkH+Uw@mail.gmail.com
2017-12-21 08:39:21 +01:00
|
|
|
rollback to settings;
|
|
|
|
-- parallel with parallel-aware hash join
|
|
|
|
savepoint settings;
|
|
|
|
set local max_parallel_workers_per_gather = 2;
|
|
|
|
set local work_mem = '4MB';
|
|
|
|
set local enable_parallel_hash = on;
|
|
|
|
explain (costs off)
|
|
|
|
select count(*) from simple r join simple s using (id);
|
|
|
|
QUERY PLAN
|
|
|
|
-------------------------------------------------------------
|
|
|
|
Finalize Aggregate
|
|
|
|
-> Gather
|
|
|
|
Workers Planned: 2
|
|
|
|
-> Partial Aggregate
|
|
|
|
-> Parallel Hash Join
|
|
|
|
Hash Cond: (r.id = s.id)
|
|
|
|
-> Parallel Seq Scan on simple r
|
|
|
|
-> Parallel Hash
|
|
|
|
-> Parallel Seq Scan on simple s
|
|
|
|
(9 rows)
|
|
|
|
|
|
|
|
select count(*) from simple r join simple s using (id);
|
|
|
|
count
|
|
|
|
-------
|
|
|
|
20000
|
|
|
|
(1 row)
|
|
|
|
|
|
|
|
select original > 1 as initially_multibatch, final > original as increased_batches
|
|
|
|
from hash_join_batches(
|
|
|
|
$$
|
|
|
|
select count(*) from simple r join simple s using (id);
|
|
|
|
$$);
|
|
|
|
initially_multibatch | increased_batches
|
|
|
|
----------------------+-------------------
|
|
|
|
f | f
|
|
|
|
(1 row)
|
|
|
|
|
2017-11-30 01:06:50 +01:00
|
|
|
rollback to settings;
|
|
|
|
-- The "good" case: batches required, but we plan the right number; we
|
|
|
|
-- plan for some number of batches, and we stick to that number, and
|
|
|
|
-- peak memory usage says within our work_mem budget
|
|
|
|
-- non-parallel
|
|
|
|
savepoint settings;
|
|
|
|
set local max_parallel_workers_per_gather = 0;
|
|
|
|
set local work_mem = '128kB';
|
|
|
|
explain (costs off)
|
|
|
|
select count(*) from simple r join simple s using (id);
|
|
|
|
QUERY PLAN
|
|
|
|
----------------------------------------
|
|
|
|
Aggregate
|
|
|
|
-> Hash Join
|
|
|
|
Hash Cond: (r.id = s.id)
|
|
|
|
-> Seq Scan on simple r
|
|
|
|
-> Hash
|
|
|
|
-> Seq Scan on simple s
|
|
|
|
(6 rows)
|
|
|
|
|
|
|
|
select count(*) from simple r join simple s using (id);
|
|
|
|
count
|
|
|
|
-------
|
|
|
|
20000
|
|
|
|
(1 row)
|
|
|
|
|
|
|
|
select original > 1 as initially_multibatch, final > original as increased_batches
|
|
|
|
from hash_join_batches(
|
|
|
|
$$
|
|
|
|
select count(*) from simple r join simple s using (id);
|
|
|
|
$$);
|
|
|
|
initially_multibatch | increased_batches
|
|
|
|
----------------------+-------------------
|
|
|
|
t | f
|
|
|
|
(1 row)
|
|
|
|
|
|
|
|
rollback to settings;
|
|
|
|
-- parallel with parallel-oblivious hash join
|
|
|
|
savepoint settings;
|
|
|
|
set local max_parallel_workers_per_gather = 2;
|
|
|
|
set local work_mem = '128kB';
|
Add parallel-aware hash joins.
Introduce parallel-aware hash joins that appear in EXPLAIN plans as Parallel
Hash Join with Parallel Hash. While hash joins could already appear in
parallel queries, they were previously always parallel-oblivious and had a
partial subplan only on the outer side, meaning that the work of the inner
subplan was duplicated in every worker.
After this commit, the planner will consider using a partial subplan on the
inner side too, using the Parallel Hash node to divide the work over the
available CPU cores and combine its results in shared memory. If the join
needs to be split into multiple batches in order to respect work_mem, then
workers process different batches as much as possible and then work together
on the remaining batches.
The advantages of a parallel-aware hash join over a parallel-oblivious hash
join used in a parallel query are that it:
* avoids wasting memory on duplicated hash tables
* avoids wasting disk space on duplicated batch files
* divides the work of building the hash table over the CPUs
One disadvantage is that there is some communication between the participating
CPUs which might outweigh the benefits of parallelism in the case of small
hash tables. This is avoided by the planner's existing reluctance to supply
partial plans for small scans, but it may be necessary to estimate
synchronization costs in future if that situation changes. Another is that
outer batch 0 must be written to disk if multiple batches are required.
A potential future advantage of parallel-aware hash joins is that right and
full outer joins could be supported, since there is a single set of matched
bits for each hashtable, but that is not yet implemented.
A new GUC enable_parallel_hash is defined to control the feature, defaulting
to on.
Author: Thomas Munro
Reviewed-By: Andres Freund, Robert Haas
Tested-By: Rafia Sabih, Prabhat Sahu
Discussion:
https://postgr.es/m/CAEepm=2W=cOkiZxcg6qiFQP-dHUe09aqTrEMM7yJDrHMhDv_RA@mail.gmail.com
https://postgr.es/m/CAEepm=37HKyJ4U6XOLi=JgfSHM3o6B-GaeO-6hkOmneTDkH+Uw@mail.gmail.com
2017-12-21 08:39:21 +01:00
|
|
|
set local enable_parallel_hash = off;
|
2017-11-30 01:06:50 +01:00
|
|
|
explain (costs off)
|
|
|
|
select count(*) from simple r join simple s using (id);
|
|
|
|
QUERY PLAN
|
|
|
|
-------------------------------------------------------
|
|
|
|
Finalize Aggregate
|
|
|
|
-> Gather
|
|
|
|
Workers Planned: 2
|
|
|
|
-> Partial Aggregate
|
|
|
|
-> Hash Join
|
|
|
|
Hash Cond: (r.id = s.id)
|
|
|
|
-> Parallel Seq Scan on simple r
|
|
|
|
-> Hash
|
|
|
|
-> Seq Scan on simple s
|
|
|
|
(9 rows)
|
|
|
|
|
|
|
|
select count(*) from simple r join simple s using (id);
|
|
|
|
count
|
|
|
|
-------
|
|
|
|
20000
|
|
|
|
(1 row)
|
|
|
|
|
|
|
|
select original > 1 as initially_multibatch, final > original as increased_batches
|
|
|
|
from hash_join_batches(
|
|
|
|
$$
|
|
|
|
select count(*) from simple r join simple s using (id);
|
|
|
|
$$);
|
|
|
|
initially_multibatch | increased_batches
|
|
|
|
----------------------+-------------------
|
|
|
|
t | f
|
|
|
|
(1 row)
|
|
|
|
|
Add parallel-aware hash joins.
Introduce parallel-aware hash joins that appear in EXPLAIN plans as Parallel
Hash Join with Parallel Hash. While hash joins could already appear in
parallel queries, they were previously always parallel-oblivious and had a
partial subplan only on the outer side, meaning that the work of the inner
subplan was duplicated in every worker.
After this commit, the planner will consider using a partial subplan on the
inner side too, using the Parallel Hash node to divide the work over the
available CPU cores and combine its results in shared memory. If the join
needs to be split into multiple batches in order to respect work_mem, then
workers process different batches as much as possible and then work together
on the remaining batches.
The advantages of a parallel-aware hash join over a parallel-oblivious hash
join used in a parallel query are that it:
* avoids wasting memory on duplicated hash tables
* avoids wasting disk space on duplicated batch files
* divides the work of building the hash table over the CPUs
One disadvantage is that there is some communication between the participating
CPUs which might outweigh the benefits of parallelism in the case of small
hash tables. This is avoided by the planner's existing reluctance to supply
partial plans for small scans, but it may be necessary to estimate
synchronization costs in future if that situation changes. Another is that
outer batch 0 must be written to disk if multiple batches are required.
A potential future advantage of parallel-aware hash joins is that right and
full outer joins could be supported, since there is a single set of matched
bits for each hashtable, but that is not yet implemented.
A new GUC enable_parallel_hash is defined to control the feature, defaulting
to on.
Author: Thomas Munro
Reviewed-By: Andres Freund, Robert Haas
Tested-By: Rafia Sabih, Prabhat Sahu
Discussion:
https://postgr.es/m/CAEepm=2W=cOkiZxcg6qiFQP-dHUe09aqTrEMM7yJDrHMhDv_RA@mail.gmail.com
https://postgr.es/m/CAEepm=37HKyJ4U6XOLi=JgfSHM3o6B-GaeO-6hkOmneTDkH+Uw@mail.gmail.com
2017-12-21 08:39:21 +01:00
|
|
|
rollback to settings;
|
|
|
|
-- parallel with parallel-aware hash join
|
|
|
|
savepoint settings;
|
|
|
|
set local max_parallel_workers_per_gather = 2;
|
2018-01-04 07:06:58 +01:00
|
|
|
set local work_mem = '192kB';
|
Add parallel-aware hash joins.
Introduce parallel-aware hash joins that appear in EXPLAIN plans as Parallel
Hash Join with Parallel Hash. While hash joins could already appear in
parallel queries, they were previously always parallel-oblivious and had a
partial subplan only on the outer side, meaning that the work of the inner
subplan was duplicated in every worker.
After this commit, the planner will consider using a partial subplan on the
inner side too, using the Parallel Hash node to divide the work over the
available CPU cores and combine its results in shared memory. If the join
needs to be split into multiple batches in order to respect work_mem, then
workers process different batches as much as possible and then work together
on the remaining batches.
The advantages of a parallel-aware hash join over a parallel-oblivious hash
join used in a parallel query are that it:
* avoids wasting memory on duplicated hash tables
* avoids wasting disk space on duplicated batch files
* divides the work of building the hash table over the CPUs
One disadvantage is that there is some communication between the participating
CPUs which might outweigh the benefits of parallelism in the case of small
hash tables. This is avoided by the planner's existing reluctance to supply
partial plans for small scans, but it may be necessary to estimate
synchronization costs in future if that situation changes. Another is that
outer batch 0 must be written to disk if multiple batches are required.
A potential future advantage of parallel-aware hash joins is that right and
full outer joins could be supported, since there is a single set of matched
bits for each hashtable, but that is not yet implemented.
A new GUC enable_parallel_hash is defined to control the feature, defaulting
to on.
Author: Thomas Munro
Reviewed-By: Andres Freund, Robert Haas
Tested-By: Rafia Sabih, Prabhat Sahu
Discussion:
https://postgr.es/m/CAEepm=2W=cOkiZxcg6qiFQP-dHUe09aqTrEMM7yJDrHMhDv_RA@mail.gmail.com
https://postgr.es/m/CAEepm=37HKyJ4U6XOLi=JgfSHM3o6B-GaeO-6hkOmneTDkH+Uw@mail.gmail.com
2017-12-21 08:39:21 +01:00
|
|
|
set local enable_parallel_hash = on;
|
|
|
|
explain (costs off)
|
|
|
|
select count(*) from simple r join simple s using (id);
|
|
|
|
QUERY PLAN
|
|
|
|
-------------------------------------------------------------
|
|
|
|
Finalize Aggregate
|
|
|
|
-> Gather
|
|
|
|
Workers Planned: 2
|
|
|
|
-> Partial Aggregate
|
|
|
|
-> Parallel Hash Join
|
|
|
|
Hash Cond: (r.id = s.id)
|
|
|
|
-> Parallel Seq Scan on simple r
|
|
|
|
-> Parallel Hash
|
|
|
|
-> Parallel Seq Scan on simple s
|
|
|
|
(9 rows)
|
|
|
|
|
|
|
|
select count(*) from simple r join simple s using (id);
|
|
|
|
count
|
|
|
|
-------
|
|
|
|
20000
|
|
|
|
(1 row)
|
|
|
|
|
|
|
|
select original > 1 as initially_multibatch, final > original as increased_batches
|
|
|
|
from hash_join_batches(
|
|
|
|
$$
|
|
|
|
select count(*) from simple r join simple s using (id);
|
|
|
|
$$);
|
|
|
|
initially_multibatch | increased_batches
|
|
|
|
----------------------+-------------------
|
|
|
|
t | f
|
|
|
|
(1 row)
|
|
|
|
|
2017-11-30 01:06:50 +01:00
|
|
|
rollback to settings;
|
|
|
|
-- The "bad" case: during execution we need to increase number of
|
|
|
|
-- batches; in this case we plan for 1 batch, and increase at least a
|
|
|
|
-- couple of times, and peak memory usage stays within our work_mem
|
|
|
|
-- budget
|
|
|
|
-- non-parallel
|
|
|
|
savepoint settings;
|
|
|
|
set local max_parallel_workers_per_gather = 0;
|
|
|
|
set local work_mem = '128kB';
|
|
|
|
explain (costs off)
|
|
|
|
select count(*) FROM simple r JOIN bigger_than_it_looks s USING (id);
|
|
|
|
QUERY PLAN
|
|
|
|
------------------------------------------------------
|
|
|
|
Aggregate
|
|
|
|
-> Hash Join
|
|
|
|
Hash Cond: (r.id = s.id)
|
|
|
|
-> Seq Scan on simple r
|
|
|
|
-> Hash
|
|
|
|
-> Seq Scan on bigger_than_it_looks s
|
|
|
|
(6 rows)
|
|
|
|
|
|
|
|
select count(*) FROM simple r JOIN bigger_than_it_looks s USING (id);
|
|
|
|
count
|
|
|
|
-------
|
|
|
|
20000
|
|
|
|
(1 row)
|
|
|
|
|
|
|
|
select original > 1 as initially_multibatch, final > original as increased_batches
|
|
|
|
from hash_join_batches(
|
|
|
|
$$
|
|
|
|
select count(*) FROM simple r JOIN bigger_than_it_looks s USING (id);
|
|
|
|
$$);
|
|
|
|
initially_multibatch | increased_batches
|
|
|
|
----------------------+-------------------
|
|
|
|
f | t
|
|
|
|
(1 row)
|
|
|
|
|
|
|
|
rollback to settings;
|
|
|
|
-- parallel with parallel-oblivious hash join
|
|
|
|
savepoint settings;
|
|
|
|
set local max_parallel_workers_per_gather = 2;
|
|
|
|
set local work_mem = '128kB';
|
Add parallel-aware hash joins.
Introduce parallel-aware hash joins that appear in EXPLAIN plans as Parallel
Hash Join with Parallel Hash. While hash joins could already appear in
parallel queries, they were previously always parallel-oblivious and had a
partial subplan only on the outer side, meaning that the work of the inner
subplan was duplicated in every worker.
After this commit, the planner will consider using a partial subplan on the
inner side too, using the Parallel Hash node to divide the work over the
available CPU cores and combine its results in shared memory. If the join
needs to be split into multiple batches in order to respect work_mem, then
workers process different batches as much as possible and then work together
on the remaining batches.
The advantages of a parallel-aware hash join over a parallel-oblivious hash
join used in a parallel query are that it:
* avoids wasting memory on duplicated hash tables
* avoids wasting disk space on duplicated batch files
* divides the work of building the hash table over the CPUs
One disadvantage is that there is some communication between the participating
CPUs which might outweigh the benefits of parallelism in the case of small
hash tables. This is avoided by the planner's existing reluctance to supply
partial plans for small scans, but it may be necessary to estimate
synchronization costs in future if that situation changes. Another is that
outer batch 0 must be written to disk if multiple batches are required.
A potential future advantage of parallel-aware hash joins is that right and
full outer joins could be supported, since there is a single set of matched
bits for each hashtable, but that is not yet implemented.
A new GUC enable_parallel_hash is defined to control the feature, defaulting
to on.
Author: Thomas Munro
Reviewed-By: Andres Freund, Robert Haas
Tested-By: Rafia Sabih, Prabhat Sahu
Discussion:
https://postgr.es/m/CAEepm=2W=cOkiZxcg6qiFQP-dHUe09aqTrEMM7yJDrHMhDv_RA@mail.gmail.com
https://postgr.es/m/CAEepm=37HKyJ4U6XOLi=JgfSHM3o6B-GaeO-6hkOmneTDkH+Uw@mail.gmail.com
2017-12-21 08:39:21 +01:00
|
|
|
set local enable_parallel_hash = off;
|
2017-11-30 01:06:50 +01:00
|
|
|
explain (costs off)
|
|
|
|
select count(*) from simple r join bigger_than_it_looks s using (id);
|
|
|
|
QUERY PLAN
|
|
|
|
------------------------------------------------------------------
|
|
|
|
Finalize Aggregate
|
|
|
|
-> Gather
|
|
|
|
Workers Planned: 2
|
|
|
|
-> Partial Aggregate
|
|
|
|
-> Hash Join
|
|
|
|
Hash Cond: (r.id = s.id)
|
|
|
|
-> Parallel Seq Scan on simple r
|
|
|
|
-> Hash
|
|
|
|
-> Seq Scan on bigger_than_it_looks s
|
|
|
|
(9 rows)
|
|
|
|
|
|
|
|
select count(*) from simple r join bigger_than_it_looks s using (id);
|
|
|
|
count
|
|
|
|
-------
|
|
|
|
20000
|
|
|
|
(1 row)
|
|
|
|
|
|
|
|
select original > 1 as initially_multibatch, final > original as increased_batches
|
|
|
|
from hash_join_batches(
|
|
|
|
$$
|
|
|
|
select count(*) from simple r join bigger_than_it_looks s using (id);
|
|
|
|
$$);
|
|
|
|
initially_multibatch | increased_batches
|
|
|
|
----------------------+-------------------
|
|
|
|
f | t
|
|
|
|
(1 row)
|
|
|
|
|
Add parallel-aware hash joins.
Introduce parallel-aware hash joins that appear in EXPLAIN plans as Parallel
Hash Join with Parallel Hash. While hash joins could already appear in
parallel queries, they were previously always parallel-oblivious and had a
partial subplan only on the outer side, meaning that the work of the inner
subplan was duplicated in every worker.
After this commit, the planner will consider using a partial subplan on the
inner side too, using the Parallel Hash node to divide the work over the
available CPU cores and combine its results in shared memory. If the join
needs to be split into multiple batches in order to respect work_mem, then
workers process different batches as much as possible and then work together
on the remaining batches.
The advantages of a parallel-aware hash join over a parallel-oblivious hash
join used in a parallel query are that it:
* avoids wasting memory on duplicated hash tables
* avoids wasting disk space on duplicated batch files
* divides the work of building the hash table over the CPUs
One disadvantage is that there is some communication between the participating
CPUs which might outweigh the benefits of parallelism in the case of small
hash tables. This is avoided by the planner's existing reluctance to supply
partial plans for small scans, but it may be necessary to estimate
synchronization costs in future if that situation changes. Another is that
outer batch 0 must be written to disk if multiple batches are required.
A potential future advantage of parallel-aware hash joins is that right and
full outer joins could be supported, since there is a single set of matched
bits for each hashtable, but that is not yet implemented.
A new GUC enable_parallel_hash is defined to control the feature, defaulting
to on.
Author: Thomas Munro
Reviewed-By: Andres Freund, Robert Haas
Tested-By: Rafia Sabih, Prabhat Sahu
Discussion:
https://postgr.es/m/CAEepm=2W=cOkiZxcg6qiFQP-dHUe09aqTrEMM7yJDrHMhDv_RA@mail.gmail.com
https://postgr.es/m/CAEepm=37HKyJ4U6XOLi=JgfSHM3o6B-GaeO-6hkOmneTDkH+Uw@mail.gmail.com
2017-12-21 08:39:21 +01:00
|
|
|
rollback to settings;
|
|
|
|
-- parallel with parallel-aware hash join
|
|
|
|
savepoint settings;
|
|
|
|
set local max_parallel_workers_per_gather = 1;
|
|
|
|
set local work_mem = '192kB';
|
|
|
|
set local enable_parallel_hash = on;
|
|
|
|
explain (costs off)
|
|
|
|
select count(*) from simple r join bigger_than_it_looks s using (id);
|
|
|
|
QUERY PLAN
|
|
|
|
---------------------------------------------------------------------------
|
|
|
|
Finalize Aggregate
|
|
|
|
-> Gather
|
|
|
|
Workers Planned: 1
|
|
|
|
-> Partial Aggregate
|
|
|
|
-> Parallel Hash Join
|
|
|
|
Hash Cond: (r.id = s.id)
|
|
|
|
-> Parallel Seq Scan on simple r
|
|
|
|
-> Parallel Hash
|
|
|
|
-> Parallel Seq Scan on bigger_than_it_looks s
|
|
|
|
(9 rows)
|
|
|
|
|
|
|
|
select count(*) from simple r join bigger_than_it_looks s using (id);
|
|
|
|
count
|
|
|
|
-------
|
|
|
|
20000
|
|
|
|
(1 row)
|
|
|
|
|
|
|
|
select original > 1 as initially_multibatch, final > original as increased_batches
|
|
|
|
from hash_join_batches(
|
|
|
|
$$
|
|
|
|
select count(*) from simple r join bigger_than_it_looks s using (id);
|
|
|
|
$$);
|
|
|
|
initially_multibatch | increased_batches
|
|
|
|
----------------------+-------------------
|
|
|
|
f | t
|
|
|
|
(1 row)
|
|
|
|
|
2017-11-30 01:06:50 +01:00
|
|
|
rollback to settings;
|
|
|
|
-- The "ugly" case: increasing the number of batches during execution
|
|
|
|
-- doesn't help, so stop trying to fit in work_mem and hope for the
|
|
|
|
-- best; in this case we plan for 1 batch, increases just once and
|
|
|
|
-- then stop increasing because that didn't help at all, so we blow
|
|
|
|
-- right through the work_mem budget and hope for the best...
|
|
|
|
-- non-parallel
|
|
|
|
savepoint settings;
|
|
|
|
set local max_parallel_workers_per_gather = 0;
|
|
|
|
set local work_mem = '128kB';
|
|
|
|
explain (costs off)
|
|
|
|
select count(*) from simple r join extremely_skewed s using (id);
|
|
|
|
QUERY PLAN
|
|
|
|
--------------------------------------------------
|
|
|
|
Aggregate
|
|
|
|
-> Hash Join
|
|
|
|
Hash Cond: (r.id = s.id)
|
|
|
|
-> Seq Scan on simple r
|
|
|
|
-> Hash
|
|
|
|
-> Seq Scan on extremely_skewed s
|
|
|
|
(6 rows)
|
|
|
|
|
|
|
|
select count(*) from simple r join extremely_skewed s using (id);
|
|
|
|
count
|
|
|
|
-------
|
|
|
|
20000
|
|
|
|
(1 row)
|
|
|
|
|
|
|
|
select * from hash_join_batches(
|
|
|
|
$$
|
|
|
|
select count(*) from simple r join extremely_skewed s using (id);
|
|
|
|
$$);
|
|
|
|
original | final
|
|
|
|
----------+-------
|
|
|
|
1 | 2
|
|
|
|
(1 row)
|
|
|
|
|
|
|
|
rollback to settings;
|
|
|
|
-- parallel with parallel-oblivious hash join
|
|
|
|
savepoint settings;
|
|
|
|
set local max_parallel_workers_per_gather = 2;
|
|
|
|
set local work_mem = '128kB';
|
Add parallel-aware hash joins.
Introduce parallel-aware hash joins that appear in EXPLAIN plans as Parallel
Hash Join with Parallel Hash. While hash joins could already appear in
parallel queries, they were previously always parallel-oblivious and had a
partial subplan only on the outer side, meaning that the work of the inner
subplan was duplicated in every worker.
After this commit, the planner will consider using a partial subplan on the
inner side too, using the Parallel Hash node to divide the work over the
available CPU cores and combine its results in shared memory. If the join
needs to be split into multiple batches in order to respect work_mem, then
workers process different batches as much as possible and then work together
on the remaining batches.
The advantages of a parallel-aware hash join over a parallel-oblivious hash
join used in a parallel query are that it:
* avoids wasting memory on duplicated hash tables
* avoids wasting disk space on duplicated batch files
* divides the work of building the hash table over the CPUs
One disadvantage is that there is some communication between the participating
CPUs which might outweigh the benefits of parallelism in the case of small
hash tables. This is avoided by the planner's existing reluctance to supply
partial plans for small scans, but it may be necessary to estimate
synchronization costs in future if that situation changes. Another is that
outer batch 0 must be written to disk if multiple batches are required.
A potential future advantage of parallel-aware hash joins is that right and
full outer joins could be supported, since there is a single set of matched
bits for each hashtable, but that is not yet implemented.
A new GUC enable_parallel_hash is defined to control the feature, defaulting
to on.
Author: Thomas Munro
Reviewed-By: Andres Freund, Robert Haas
Tested-By: Rafia Sabih, Prabhat Sahu
Discussion:
https://postgr.es/m/CAEepm=2W=cOkiZxcg6qiFQP-dHUe09aqTrEMM7yJDrHMhDv_RA@mail.gmail.com
https://postgr.es/m/CAEepm=37HKyJ4U6XOLi=JgfSHM3o6B-GaeO-6hkOmneTDkH+Uw@mail.gmail.com
2017-12-21 08:39:21 +01:00
|
|
|
set local enable_parallel_hash = off;
|
2017-11-30 01:06:50 +01:00
|
|
|
explain (costs off)
|
|
|
|
select count(*) from simple r join extremely_skewed s using (id);
|
|
|
|
QUERY PLAN
|
|
|
|
--------------------------------------------------------
|
|
|
|
Aggregate
|
|
|
|
-> Gather
|
|
|
|
Workers Planned: 2
|
|
|
|
-> Hash Join
|
|
|
|
Hash Cond: (r.id = s.id)
|
|
|
|
-> Parallel Seq Scan on simple r
|
|
|
|
-> Hash
|
|
|
|
-> Seq Scan on extremely_skewed s
|
|
|
|
(8 rows)
|
|
|
|
|
|
|
|
select count(*) from simple r join extremely_skewed s using (id);
|
|
|
|
count
|
|
|
|
-------
|
|
|
|
20000
|
|
|
|
(1 row)
|
|
|
|
|
|
|
|
select * from hash_join_batches(
|
|
|
|
$$
|
|
|
|
select count(*) from simple r join extremely_skewed s using (id);
|
|
|
|
$$);
|
|
|
|
original | final
|
|
|
|
----------+-------
|
|
|
|
1 | 2
|
|
|
|
(1 row)
|
|
|
|
|
Add parallel-aware hash joins.
Introduce parallel-aware hash joins that appear in EXPLAIN plans as Parallel
Hash Join with Parallel Hash. While hash joins could already appear in
parallel queries, they were previously always parallel-oblivious and had a
partial subplan only on the outer side, meaning that the work of the inner
subplan was duplicated in every worker.
After this commit, the planner will consider using a partial subplan on the
inner side too, using the Parallel Hash node to divide the work over the
available CPU cores and combine its results in shared memory. If the join
needs to be split into multiple batches in order to respect work_mem, then
workers process different batches as much as possible and then work together
on the remaining batches.
The advantages of a parallel-aware hash join over a parallel-oblivious hash
join used in a parallel query are that it:
* avoids wasting memory on duplicated hash tables
* avoids wasting disk space on duplicated batch files
* divides the work of building the hash table over the CPUs
One disadvantage is that there is some communication between the participating
CPUs which might outweigh the benefits of parallelism in the case of small
hash tables. This is avoided by the planner's existing reluctance to supply
partial plans for small scans, but it may be necessary to estimate
synchronization costs in future if that situation changes. Another is that
outer batch 0 must be written to disk if multiple batches are required.
A potential future advantage of parallel-aware hash joins is that right and
full outer joins could be supported, since there is a single set of matched
bits for each hashtable, but that is not yet implemented.
A new GUC enable_parallel_hash is defined to control the feature, defaulting
to on.
Author: Thomas Munro
Reviewed-By: Andres Freund, Robert Haas
Tested-By: Rafia Sabih, Prabhat Sahu
Discussion:
https://postgr.es/m/CAEepm=2W=cOkiZxcg6qiFQP-dHUe09aqTrEMM7yJDrHMhDv_RA@mail.gmail.com
https://postgr.es/m/CAEepm=37HKyJ4U6XOLi=JgfSHM3o6B-GaeO-6hkOmneTDkH+Uw@mail.gmail.com
2017-12-21 08:39:21 +01:00
|
|
|
rollback to settings;
|
|
|
|
-- parallel with parallel-aware hash join
|
|
|
|
savepoint settings;
|
|
|
|
set local max_parallel_workers_per_gather = 1;
|
|
|
|
set local work_mem = '128kB';
|
|
|
|
set local enable_parallel_hash = on;
|
|
|
|
explain (costs off)
|
|
|
|
select count(*) from simple r join extremely_skewed s using (id);
|
|
|
|
QUERY PLAN
|
|
|
|
-----------------------------------------------------------------------
|
|
|
|
Finalize Aggregate
|
|
|
|
-> Gather
|
|
|
|
Workers Planned: 1
|
|
|
|
-> Partial Aggregate
|
|
|
|
-> Parallel Hash Join
|
|
|
|
Hash Cond: (r.id = s.id)
|
|
|
|
-> Parallel Seq Scan on simple r
|
|
|
|
-> Parallel Hash
|
|
|
|
-> Parallel Seq Scan on extremely_skewed s
|
|
|
|
(9 rows)
|
|
|
|
|
|
|
|
select count(*) from simple r join extremely_skewed s using (id);
|
|
|
|
count
|
|
|
|
-------
|
|
|
|
20000
|
|
|
|
(1 row)
|
|
|
|
|
|
|
|
select * from hash_join_batches(
|
|
|
|
$$
|
|
|
|
select count(*) from simple r join extremely_skewed s using (id);
|
|
|
|
$$);
|
|
|
|
original | final
|
|
|
|
----------+-------
|
|
|
|
1 | 4
|
|
|
|
(1 row)
|
|
|
|
|
2017-11-30 01:06:50 +01:00
|
|
|
rollback to settings;
|
|
|
|
-- A couple of other hash join tests unrelated to work_mem management.
|
2017-12-05 19:55:56 +01:00
|
|
|
-- Check that EXPLAIN ANALYZE has data even if the leader doesn't participate
|
|
|
|
savepoint settings;
|
|
|
|
set local max_parallel_workers_per_gather = 2;
|
|
|
|
set local work_mem = '4MB';
|
|
|
|
set local parallel_leader_participation = off;
|
|
|
|
select * from hash_join_batches(
|
|
|
|
$$
|
|
|
|
select count(*) from simple r join simple s using (id);
|
|
|
|
$$);
|
|
|
|
original | final
|
|
|
|
----------+-------
|
|
|
|
1 | 1
|
|
|
|
(1 row)
|
|
|
|
|
2017-12-19 21:26:09 +01:00
|
|
|
rollback to settings;
|
|
|
|
-- Exercise rescans. We'll turn off parallel_leader_participation so
|
|
|
|
-- that we can check that instrumentation comes back correctly.
|
Clean up duplicate table and function names in regression tests.
Many of the objects we create during the regression tests are put in the
public schema, so that using the same names in different regression tests
creates a hazard of test failures if any two such scripts run concurrently.
This patch cleans up a bunch of latent hazards of that sort, as well as two
live hazards.
The current situation in this regard is far worse than it was a year or two
back, because practically all of the partitioning-related test cases have
reused table names with enthusiasm. I despaired of cleaning up that mess
within the five most-affected tests (create_table, alter_table, insert,
update, inherit); fortunately those don't run concurrently.
Other than partitioning problems, most of the issues boil down to using
names like "foo", "bar", "tmp", etc, without thought for the fact that
other test scripts might use similar names concurrently. I've made an
effort to make all such names more specific.
One of the live hazards was that commit 7421f4b8 caused with.sql to
create a table named "test", conflicting with a similarly-named table
in alter_table.sql; this was exposed in the buildfarm recently.
The other one was that join.sql and transactions.sql both create tables
named "foo" and "bar"; but join.sql's uses of those names date back
only to December or so.
Since commit 7421f4b8 was back-patched to v10, back-patch a minimal
fix for that problem. The rest of this is just future-proofing.
Discussion: https://postgr.es/m/4627.1521070268@sss.pgh.pa.us
2018-03-15 22:08:51 +01:00
|
|
|
create table join_foo as select generate_series(1, 3) as id, 'xxxxx'::text as t;
|
|
|
|
alter table join_foo set (parallel_workers = 0);
|
|
|
|
create table join_bar as select generate_series(1, 10000) as id, 'xxxxx'::text as t;
|
|
|
|
alter table join_bar set (parallel_workers = 2);
|
2017-12-19 21:26:09 +01:00
|
|
|
-- multi-batch with rescan, parallel-oblivious
|
|
|
|
savepoint settings;
|
Add parallel-aware hash joins.
Introduce parallel-aware hash joins that appear in EXPLAIN plans as Parallel
Hash Join with Parallel Hash. While hash joins could already appear in
parallel queries, they were previously always parallel-oblivious and had a
partial subplan only on the outer side, meaning that the work of the inner
subplan was duplicated in every worker.
After this commit, the planner will consider using a partial subplan on the
inner side too, using the Parallel Hash node to divide the work over the
available CPU cores and combine its results in shared memory. If the join
needs to be split into multiple batches in order to respect work_mem, then
workers process different batches as much as possible and then work together
on the remaining batches.
The advantages of a parallel-aware hash join over a parallel-oblivious hash
join used in a parallel query are that it:
* avoids wasting memory on duplicated hash tables
* avoids wasting disk space on duplicated batch files
* divides the work of building the hash table over the CPUs
One disadvantage is that there is some communication between the participating
CPUs which might outweigh the benefits of parallelism in the case of small
hash tables. This is avoided by the planner's existing reluctance to supply
partial plans for small scans, but it may be necessary to estimate
synchronization costs in future if that situation changes. Another is that
outer batch 0 must be written to disk if multiple batches are required.
A potential future advantage of parallel-aware hash joins is that right and
full outer joins could be supported, since there is a single set of matched
bits for each hashtable, but that is not yet implemented.
A new GUC enable_parallel_hash is defined to control the feature, defaulting
to on.
Author: Thomas Munro
Reviewed-By: Andres Freund, Robert Haas
Tested-By: Rafia Sabih, Prabhat Sahu
Discussion:
https://postgr.es/m/CAEepm=2W=cOkiZxcg6qiFQP-dHUe09aqTrEMM7yJDrHMhDv_RA@mail.gmail.com
https://postgr.es/m/CAEepm=37HKyJ4U6XOLi=JgfSHM3o6B-GaeO-6hkOmneTDkH+Uw@mail.gmail.com
2017-12-21 08:39:21 +01:00
|
|
|
set enable_parallel_hash = off;
|
2017-12-19 21:26:09 +01:00
|
|
|
set parallel_leader_participation = off;
|
|
|
|
set min_parallel_table_scan_size = 0;
|
|
|
|
set parallel_setup_cost = 0;
|
|
|
|
set parallel_tuple_cost = 0;
|
|
|
|
set max_parallel_workers_per_gather = 2;
|
|
|
|
set enable_material = off;
|
|
|
|
set enable_mergejoin = off;
|
|
|
|
set work_mem = '64kB';
|
|
|
|
explain (costs off)
|
Clean up duplicate table and function names in regression tests.
Many of the objects we create during the regression tests are put in the
public schema, so that using the same names in different regression tests
creates a hazard of test failures if any two such scripts run concurrently.
This patch cleans up a bunch of latent hazards of that sort, as well as two
live hazards.
The current situation in this regard is far worse than it was a year or two
back, because practically all of the partitioning-related test cases have
reused table names with enthusiasm. I despaired of cleaning up that mess
within the five most-affected tests (create_table, alter_table, insert,
update, inherit); fortunately those don't run concurrently.
Other than partitioning problems, most of the issues boil down to using
names like "foo", "bar", "tmp", etc, without thought for the fact that
other test scripts might use similar names concurrently. I've made an
effort to make all such names more specific.
One of the live hazards was that commit 7421f4b8 caused with.sql to
create a table named "test", conflicting with a similarly-named table
in alter_table.sql; this was exposed in the buildfarm recently.
The other one was that join.sql and transactions.sql both create tables
named "foo" and "bar"; but join.sql's uses of those names date back
only to December or so.
Since commit 7421f4b8 was back-patched to v10, back-patch a minimal
fix for that problem. The rest of this is just future-proofing.
Discussion: https://postgr.es/m/4627.1521070268@sss.pgh.pa.us
2018-03-15 22:08:51 +01:00
|
|
|
select count(*) from join_foo
|
|
|
|
left join (select b1.id, b1.t from join_bar b1 join join_bar b2 using (id)) ss
|
|
|
|
on join_foo.id < ss.id + 1 and join_foo.id > ss.id - 1;
|
|
|
|
QUERY PLAN
|
|
|
|
------------------------------------------------------------------------------------
|
2017-12-19 21:26:09 +01:00
|
|
|
Aggregate
|
|
|
|
-> Nested Loop Left Join
|
Clean up duplicate table and function names in regression tests.
Many of the objects we create during the regression tests are put in the
public schema, so that using the same names in different regression tests
creates a hazard of test failures if any two such scripts run concurrently.
This patch cleans up a bunch of latent hazards of that sort, as well as two
live hazards.
The current situation in this regard is far worse than it was a year or two
back, because practically all of the partitioning-related test cases have
reused table names with enthusiasm. I despaired of cleaning up that mess
within the five most-affected tests (create_table, alter_table, insert,
update, inherit); fortunately those don't run concurrently.
Other than partitioning problems, most of the issues boil down to using
names like "foo", "bar", "tmp", etc, without thought for the fact that
other test scripts might use similar names concurrently. I've made an
effort to make all such names more specific.
One of the live hazards was that commit 7421f4b8 caused with.sql to
create a table named "test", conflicting with a similarly-named table
in alter_table.sql; this was exposed in the buildfarm recently.
The other one was that join.sql and transactions.sql both create tables
named "foo" and "bar"; but join.sql's uses of those names date back
only to December or so.
Since commit 7421f4b8 was back-patched to v10, back-patch a minimal
fix for that problem. The rest of this is just future-proofing.
Discussion: https://postgr.es/m/4627.1521070268@sss.pgh.pa.us
2018-03-15 22:08:51 +01:00
|
|
|
Join Filter: ((join_foo.id < (b1.id + 1)) AND (join_foo.id > (b1.id - 1)))
|
|
|
|
-> Seq Scan on join_foo
|
2017-12-19 21:26:09 +01:00
|
|
|
-> Gather
|
|
|
|
Workers Planned: 2
|
|
|
|
-> Hash Join
|
|
|
|
Hash Cond: (b1.id = b2.id)
|
Clean up duplicate table and function names in regression tests.
Many of the objects we create during the regression tests are put in the
public schema, so that using the same names in different regression tests
creates a hazard of test failures if any two such scripts run concurrently.
This patch cleans up a bunch of latent hazards of that sort, as well as two
live hazards.
The current situation in this regard is far worse than it was a year or two
back, because practically all of the partitioning-related test cases have
reused table names with enthusiasm. I despaired of cleaning up that mess
within the five most-affected tests (create_table, alter_table, insert,
update, inherit); fortunately those don't run concurrently.
Other than partitioning problems, most of the issues boil down to using
names like "foo", "bar", "tmp", etc, without thought for the fact that
other test scripts might use similar names concurrently. I've made an
effort to make all such names more specific.
One of the live hazards was that commit 7421f4b8 caused with.sql to
create a table named "test", conflicting with a similarly-named table
in alter_table.sql; this was exposed in the buildfarm recently.
The other one was that join.sql and transactions.sql both create tables
named "foo" and "bar"; but join.sql's uses of those names date back
only to December or so.
Since commit 7421f4b8 was back-patched to v10, back-patch a minimal
fix for that problem. The rest of this is just future-proofing.
Discussion: https://postgr.es/m/4627.1521070268@sss.pgh.pa.us
2018-03-15 22:08:51 +01:00
|
|
|
-> Parallel Seq Scan on join_bar b1
|
2017-12-19 21:26:09 +01:00
|
|
|
-> Hash
|
Clean up duplicate table and function names in regression tests.
Many of the objects we create during the regression tests are put in the
public schema, so that using the same names in different regression tests
creates a hazard of test failures if any two such scripts run concurrently.
This patch cleans up a bunch of latent hazards of that sort, as well as two
live hazards.
The current situation in this regard is far worse than it was a year or two
back, because practically all of the partitioning-related test cases have
reused table names with enthusiasm. I despaired of cleaning up that mess
within the five most-affected tests (create_table, alter_table, insert,
update, inherit); fortunately those don't run concurrently.
Other than partitioning problems, most of the issues boil down to using
names like "foo", "bar", "tmp", etc, without thought for the fact that
other test scripts might use similar names concurrently. I've made an
effort to make all such names more specific.
One of the live hazards was that commit 7421f4b8 caused with.sql to
create a table named "test", conflicting with a similarly-named table
in alter_table.sql; this was exposed in the buildfarm recently.
The other one was that join.sql and transactions.sql both create tables
named "foo" and "bar"; but join.sql's uses of those names date back
only to December or so.
Since commit 7421f4b8 was back-patched to v10, back-patch a minimal
fix for that problem. The rest of this is just future-proofing.
Discussion: https://postgr.es/m/4627.1521070268@sss.pgh.pa.us
2018-03-15 22:08:51 +01:00
|
|
|
-> Seq Scan on join_bar b2
|
2017-12-19 21:26:09 +01:00
|
|
|
(11 rows)
|
|
|
|
|
Clean up duplicate table and function names in regression tests.
Many of the objects we create during the regression tests are put in the
public schema, so that using the same names in different regression tests
creates a hazard of test failures if any two such scripts run concurrently.
This patch cleans up a bunch of latent hazards of that sort, as well as two
live hazards.
The current situation in this regard is far worse than it was a year or two
back, because practically all of the partitioning-related test cases have
reused table names with enthusiasm. I despaired of cleaning up that mess
within the five most-affected tests (create_table, alter_table, insert,
update, inherit); fortunately those don't run concurrently.
Other than partitioning problems, most of the issues boil down to using
names like "foo", "bar", "tmp", etc, without thought for the fact that
other test scripts might use similar names concurrently. I've made an
effort to make all such names more specific.
One of the live hazards was that commit 7421f4b8 caused with.sql to
create a table named "test", conflicting with a similarly-named table
in alter_table.sql; this was exposed in the buildfarm recently.
The other one was that join.sql and transactions.sql both create tables
named "foo" and "bar"; but join.sql's uses of those names date back
only to December or so.
Since commit 7421f4b8 was back-patched to v10, back-patch a minimal
fix for that problem. The rest of this is just future-proofing.
Discussion: https://postgr.es/m/4627.1521070268@sss.pgh.pa.us
2018-03-15 22:08:51 +01:00
|
|
|
select count(*) from join_foo
|
|
|
|
left join (select b1.id, b1.t from join_bar b1 join join_bar b2 using (id)) ss
|
|
|
|
on join_foo.id < ss.id + 1 and join_foo.id > ss.id - 1;
|
2017-12-19 21:26:09 +01:00
|
|
|
count
|
|
|
|
-------
|
|
|
|
3
|
|
|
|
(1 row)
|
|
|
|
|
|
|
|
select final > 1 as multibatch
|
|
|
|
from hash_join_batches(
|
|
|
|
$$
|
Clean up duplicate table and function names in regression tests.
Many of the objects we create during the regression tests are put in the
public schema, so that using the same names in different regression tests
creates a hazard of test failures if any two such scripts run concurrently.
This patch cleans up a bunch of latent hazards of that sort, as well as two
live hazards.
The current situation in this regard is far worse than it was a year or two
back, because practically all of the partitioning-related test cases have
reused table names with enthusiasm. I despaired of cleaning up that mess
within the five most-affected tests (create_table, alter_table, insert,
update, inherit); fortunately those don't run concurrently.
Other than partitioning problems, most of the issues boil down to using
names like "foo", "bar", "tmp", etc, without thought for the fact that
other test scripts might use similar names concurrently. I've made an
effort to make all such names more specific.
One of the live hazards was that commit 7421f4b8 caused with.sql to
create a table named "test", conflicting with a similarly-named table
in alter_table.sql; this was exposed in the buildfarm recently.
The other one was that join.sql and transactions.sql both create tables
named "foo" and "bar"; but join.sql's uses of those names date back
only to December or so.
Since commit 7421f4b8 was back-patched to v10, back-patch a minimal
fix for that problem. The rest of this is just future-proofing.
Discussion: https://postgr.es/m/4627.1521070268@sss.pgh.pa.us
2018-03-15 22:08:51 +01:00
|
|
|
select count(*) from join_foo
|
|
|
|
left join (select b1.id, b1.t from join_bar b1 join join_bar b2 using (id)) ss
|
|
|
|
on join_foo.id < ss.id + 1 and join_foo.id > ss.id - 1;
|
2017-12-19 21:26:09 +01:00
|
|
|
$$);
|
|
|
|
multibatch
|
|
|
|
------------
|
|
|
|
t
|
|
|
|
(1 row)
|
|
|
|
|
|
|
|
rollback to settings;
|
|
|
|
-- single-batch with rescan, parallel-oblivious
|
|
|
|
savepoint settings;
|
Add parallel-aware hash joins.
Introduce parallel-aware hash joins that appear in EXPLAIN plans as Parallel
Hash Join with Parallel Hash. While hash joins could already appear in
parallel queries, they were previously always parallel-oblivious and had a
partial subplan only on the outer side, meaning that the work of the inner
subplan was duplicated in every worker.
After this commit, the planner will consider using a partial subplan on the
inner side too, using the Parallel Hash node to divide the work over the
available CPU cores and combine its results in shared memory. If the join
needs to be split into multiple batches in order to respect work_mem, then
workers process different batches as much as possible and then work together
on the remaining batches.
The advantages of a parallel-aware hash join over a parallel-oblivious hash
join used in a parallel query are that it:
* avoids wasting memory on duplicated hash tables
* avoids wasting disk space on duplicated batch files
* divides the work of building the hash table over the CPUs
One disadvantage is that there is some communication between the participating
CPUs which might outweigh the benefits of parallelism in the case of small
hash tables. This is avoided by the planner's existing reluctance to supply
partial plans for small scans, but it may be necessary to estimate
synchronization costs in future if that situation changes. Another is that
outer batch 0 must be written to disk if multiple batches are required.
A potential future advantage of parallel-aware hash joins is that right and
full outer joins could be supported, since there is a single set of matched
bits for each hashtable, but that is not yet implemented.
A new GUC enable_parallel_hash is defined to control the feature, defaulting
to on.
Author: Thomas Munro
Reviewed-By: Andres Freund, Robert Haas
Tested-By: Rafia Sabih, Prabhat Sahu
Discussion:
https://postgr.es/m/CAEepm=2W=cOkiZxcg6qiFQP-dHUe09aqTrEMM7yJDrHMhDv_RA@mail.gmail.com
https://postgr.es/m/CAEepm=37HKyJ4U6XOLi=JgfSHM3o6B-GaeO-6hkOmneTDkH+Uw@mail.gmail.com
2017-12-21 08:39:21 +01:00
|
|
|
set enable_parallel_hash = off;
|
2017-12-19 21:26:09 +01:00
|
|
|
set parallel_leader_participation = off;
|
|
|
|
set min_parallel_table_scan_size = 0;
|
|
|
|
set parallel_setup_cost = 0;
|
|
|
|
set parallel_tuple_cost = 0;
|
|
|
|
set max_parallel_workers_per_gather = 2;
|
|
|
|
set enable_material = off;
|
|
|
|
set enable_mergejoin = off;
|
|
|
|
set work_mem = '4MB';
|
|
|
|
explain (costs off)
|
Clean up duplicate table and function names in regression tests.
Many of the objects we create during the regression tests are put in the
public schema, so that using the same names in different regression tests
creates a hazard of test failures if any two such scripts run concurrently.
This patch cleans up a bunch of latent hazards of that sort, as well as two
live hazards.
The current situation in this regard is far worse than it was a year or two
back, because practically all of the partitioning-related test cases have
reused table names with enthusiasm. I despaired of cleaning up that mess
within the five most-affected tests (create_table, alter_table, insert,
update, inherit); fortunately those don't run concurrently.
Other than partitioning problems, most of the issues boil down to using
names like "foo", "bar", "tmp", etc, without thought for the fact that
other test scripts might use similar names concurrently. I've made an
effort to make all such names more specific.
One of the live hazards was that commit 7421f4b8 caused with.sql to
create a table named "test", conflicting with a similarly-named table
in alter_table.sql; this was exposed in the buildfarm recently.
The other one was that join.sql and transactions.sql both create tables
named "foo" and "bar"; but join.sql's uses of those names date back
only to December or so.
Since commit 7421f4b8 was back-patched to v10, back-patch a minimal
fix for that problem. The rest of this is just future-proofing.
Discussion: https://postgr.es/m/4627.1521070268@sss.pgh.pa.us
2018-03-15 22:08:51 +01:00
|
|
|
select count(*) from join_foo
|
|
|
|
left join (select b1.id, b1.t from join_bar b1 join join_bar b2 using (id)) ss
|
|
|
|
on join_foo.id < ss.id + 1 and join_foo.id > ss.id - 1;
|
|
|
|
QUERY PLAN
|
|
|
|
------------------------------------------------------------------------------------
|
2017-12-19 21:26:09 +01:00
|
|
|
Aggregate
|
|
|
|
-> Nested Loop Left Join
|
Clean up duplicate table and function names in regression tests.
Many of the objects we create during the regression tests are put in the
public schema, so that using the same names in different regression tests
creates a hazard of test failures if any two such scripts run concurrently.
This patch cleans up a bunch of latent hazards of that sort, as well as two
live hazards.
The current situation in this regard is far worse than it was a year or two
back, because practically all of the partitioning-related test cases have
reused table names with enthusiasm. I despaired of cleaning up that mess
within the five most-affected tests (create_table, alter_table, insert,
update, inherit); fortunately those don't run concurrently.
Other than partitioning problems, most of the issues boil down to using
names like "foo", "bar", "tmp", etc, without thought for the fact that
other test scripts might use similar names concurrently. I've made an
effort to make all such names more specific.
One of the live hazards was that commit 7421f4b8 caused with.sql to
create a table named "test", conflicting with a similarly-named table
in alter_table.sql; this was exposed in the buildfarm recently.
The other one was that join.sql and transactions.sql both create tables
named "foo" and "bar"; but join.sql's uses of those names date back
only to December or so.
Since commit 7421f4b8 was back-patched to v10, back-patch a minimal
fix for that problem. The rest of this is just future-proofing.
Discussion: https://postgr.es/m/4627.1521070268@sss.pgh.pa.us
2018-03-15 22:08:51 +01:00
|
|
|
Join Filter: ((join_foo.id < (b1.id + 1)) AND (join_foo.id > (b1.id - 1)))
|
|
|
|
-> Seq Scan on join_foo
|
2017-12-19 21:26:09 +01:00
|
|
|
-> Gather
|
|
|
|
Workers Planned: 2
|
|
|
|
-> Hash Join
|
|
|
|
Hash Cond: (b1.id = b2.id)
|
Clean up duplicate table and function names in regression tests.
Many of the objects we create during the regression tests are put in the
public schema, so that using the same names in different regression tests
creates a hazard of test failures if any two such scripts run concurrently.
This patch cleans up a bunch of latent hazards of that sort, as well as two
live hazards.
The current situation in this regard is far worse than it was a year or two
back, because practically all of the partitioning-related test cases have
reused table names with enthusiasm. I despaired of cleaning up that mess
within the five most-affected tests (create_table, alter_table, insert,
update, inherit); fortunately those don't run concurrently.
Other than partitioning problems, most of the issues boil down to using
names like "foo", "bar", "tmp", etc, without thought for the fact that
other test scripts might use similar names concurrently. I've made an
effort to make all such names more specific.
One of the live hazards was that commit 7421f4b8 caused with.sql to
create a table named "test", conflicting with a similarly-named table
in alter_table.sql; this was exposed in the buildfarm recently.
The other one was that join.sql and transactions.sql both create tables
named "foo" and "bar"; but join.sql's uses of those names date back
only to December or so.
Since commit 7421f4b8 was back-patched to v10, back-patch a minimal
fix for that problem. The rest of this is just future-proofing.
Discussion: https://postgr.es/m/4627.1521070268@sss.pgh.pa.us
2018-03-15 22:08:51 +01:00
|
|
|
-> Parallel Seq Scan on join_bar b1
|
2017-12-19 21:26:09 +01:00
|
|
|
-> Hash
|
Clean up duplicate table and function names in regression tests.
Many of the objects we create during the regression tests are put in the
public schema, so that using the same names in different regression tests
creates a hazard of test failures if any two such scripts run concurrently.
This patch cleans up a bunch of latent hazards of that sort, as well as two
live hazards.
The current situation in this regard is far worse than it was a year or two
back, because practically all of the partitioning-related test cases have
reused table names with enthusiasm. I despaired of cleaning up that mess
within the five most-affected tests (create_table, alter_table, insert,
update, inherit); fortunately those don't run concurrently.
Other than partitioning problems, most of the issues boil down to using
names like "foo", "bar", "tmp", etc, without thought for the fact that
other test scripts might use similar names concurrently. I've made an
effort to make all such names more specific.
One of the live hazards was that commit 7421f4b8 caused with.sql to
create a table named "test", conflicting with a similarly-named table
in alter_table.sql; this was exposed in the buildfarm recently.
The other one was that join.sql and transactions.sql both create tables
named "foo" and "bar"; but join.sql's uses of those names date back
only to December or so.
Since commit 7421f4b8 was back-patched to v10, back-patch a minimal
fix for that problem. The rest of this is just future-proofing.
Discussion: https://postgr.es/m/4627.1521070268@sss.pgh.pa.us
2018-03-15 22:08:51 +01:00
|
|
|
-> Seq Scan on join_bar b2
|
2017-12-19 21:26:09 +01:00
|
|
|
(11 rows)
|
|
|
|
|
Clean up duplicate table and function names in regression tests.
Many of the objects we create during the regression tests are put in the
public schema, so that using the same names in different regression tests
creates a hazard of test failures if any two such scripts run concurrently.
This patch cleans up a bunch of latent hazards of that sort, as well as two
live hazards.
The current situation in this regard is far worse than it was a year or two
back, because practically all of the partitioning-related test cases have
reused table names with enthusiasm. I despaired of cleaning up that mess
within the five most-affected tests (create_table, alter_table, insert,
update, inherit); fortunately those don't run concurrently.
Other than partitioning problems, most of the issues boil down to using
names like "foo", "bar", "tmp", etc, without thought for the fact that
other test scripts might use similar names concurrently. I've made an
effort to make all such names more specific.
One of the live hazards was that commit 7421f4b8 caused with.sql to
create a table named "test", conflicting with a similarly-named table
in alter_table.sql; this was exposed in the buildfarm recently.
The other one was that join.sql and transactions.sql both create tables
named "foo" and "bar"; but join.sql's uses of those names date back
only to December or so.
Since commit 7421f4b8 was back-patched to v10, back-patch a minimal
fix for that problem. The rest of this is just future-proofing.
Discussion: https://postgr.es/m/4627.1521070268@sss.pgh.pa.us
2018-03-15 22:08:51 +01:00
|
|
|
select count(*) from join_foo
|
|
|
|
left join (select b1.id, b1.t from join_bar b1 join join_bar b2 using (id)) ss
|
|
|
|
on join_foo.id < ss.id + 1 and join_foo.id > ss.id - 1;
|
2017-12-19 21:26:09 +01:00
|
|
|
count
|
|
|
|
-------
|
|
|
|
3
|
|
|
|
(1 row)
|
|
|
|
|
|
|
|
select final > 1 as multibatch
|
|
|
|
from hash_join_batches(
|
|
|
|
$$
|
Clean up duplicate table and function names in regression tests.
Many of the objects we create during the regression tests are put in the
public schema, so that using the same names in different regression tests
creates a hazard of test failures if any two such scripts run concurrently.
This patch cleans up a bunch of latent hazards of that sort, as well as two
live hazards.
The current situation in this regard is far worse than it was a year or two
back, because practically all of the partitioning-related test cases have
reused table names with enthusiasm. I despaired of cleaning up that mess
within the five most-affected tests (create_table, alter_table, insert,
update, inherit); fortunately those don't run concurrently.
Other than partitioning problems, most of the issues boil down to using
names like "foo", "bar", "tmp", etc, without thought for the fact that
other test scripts might use similar names concurrently. I've made an
effort to make all such names more specific.
One of the live hazards was that commit 7421f4b8 caused with.sql to
create a table named "test", conflicting with a similarly-named table
in alter_table.sql; this was exposed in the buildfarm recently.
The other one was that join.sql and transactions.sql both create tables
named "foo" and "bar"; but join.sql's uses of those names date back
only to December or so.
Since commit 7421f4b8 was back-patched to v10, back-patch a minimal
fix for that problem. The rest of this is just future-proofing.
Discussion: https://postgr.es/m/4627.1521070268@sss.pgh.pa.us
2018-03-15 22:08:51 +01:00
|
|
|
select count(*) from join_foo
|
|
|
|
left join (select b1.id, b1.t from join_bar b1 join join_bar b2 using (id)) ss
|
|
|
|
on join_foo.id < ss.id + 1 and join_foo.id > ss.id - 1;
|
2017-12-19 21:26:09 +01:00
|
|
|
$$);
|
|
|
|
multibatch
|
|
|
|
------------
|
|
|
|
f
|
|
|
|
(1 row)
|
|
|
|
|
Add parallel-aware hash joins.
Introduce parallel-aware hash joins that appear in EXPLAIN plans as Parallel
Hash Join with Parallel Hash. While hash joins could already appear in
parallel queries, they were previously always parallel-oblivious and had a
partial subplan only on the outer side, meaning that the work of the inner
subplan was duplicated in every worker.
After this commit, the planner will consider using a partial subplan on the
inner side too, using the Parallel Hash node to divide the work over the
available CPU cores and combine its results in shared memory. If the join
needs to be split into multiple batches in order to respect work_mem, then
workers process different batches as much as possible and then work together
on the remaining batches.
The advantages of a parallel-aware hash join over a parallel-oblivious hash
join used in a parallel query are that it:
* avoids wasting memory on duplicated hash tables
* avoids wasting disk space on duplicated batch files
* divides the work of building the hash table over the CPUs
One disadvantage is that there is some communication between the participating
CPUs which might outweigh the benefits of parallelism in the case of small
hash tables. This is avoided by the planner's existing reluctance to supply
partial plans for small scans, but it may be necessary to estimate
synchronization costs in future if that situation changes. Another is that
outer batch 0 must be written to disk if multiple batches are required.
A potential future advantage of parallel-aware hash joins is that right and
full outer joins could be supported, since there is a single set of matched
bits for each hashtable, but that is not yet implemented.
A new GUC enable_parallel_hash is defined to control the feature, defaulting
to on.
Author: Thomas Munro
Reviewed-By: Andres Freund, Robert Haas
Tested-By: Rafia Sabih, Prabhat Sahu
Discussion:
https://postgr.es/m/CAEepm=2W=cOkiZxcg6qiFQP-dHUe09aqTrEMM7yJDrHMhDv_RA@mail.gmail.com
https://postgr.es/m/CAEepm=37HKyJ4U6XOLi=JgfSHM3o6B-GaeO-6hkOmneTDkH+Uw@mail.gmail.com
2017-12-21 08:39:21 +01:00
|
|
|
rollback to settings;
|
|
|
|
-- multi-batch with rescan, parallel-aware
|
|
|
|
savepoint settings;
|
|
|
|
set enable_parallel_hash = on;
|
|
|
|
set parallel_leader_participation = off;
|
|
|
|
set min_parallel_table_scan_size = 0;
|
|
|
|
set parallel_setup_cost = 0;
|
|
|
|
set parallel_tuple_cost = 0;
|
|
|
|
set max_parallel_workers_per_gather = 2;
|
|
|
|
set enable_material = off;
|
|
|
|
set enable_mergejoin = off;
|
|
|
|
set work_mem = '64kB';
|
|
|
|
explain (costs off)
|
Clean up duplicate table and function names in regression tests.
Many of the objects we create during the regression tests are put in the
public schema, so that using the same names in different regression tests
creates a hazard of test failures if any two such scripts run concurrently.
This patch cleans up a bunch of latent hazards of that sort, as well as two
live hazards.
The current situation in this regard is far worse than it was a year or two
back, because practically all of the partitioning-related test cases have
reused table names with enthusiasm. I despaired of cleaning up that mess
within the five most-affected tests (create_table, alter_table, insert,
update, inherit); fortunately those don't run concurrently.
Other than partitioning problems, most of the issues boil down to using
names like "foo", "bar", "tmp", etc, without thought for the fact that
other test scripts might use similar names concurrently. I've made an
effort to make all such names more specific.
One of the live hazards was that commit 7421f4b8 caused with.sql to
create a table named "test", conflicting with a similarly-named table
in alter_table.sql; this was exposed in the buildfarm recently.
The other one was that join.sql and transactions.sql both create tables
named "foo" and "bar"; but join.sql's uses of those names date back
only to December or so.
Since commit 7421f4b8 was back-patched to v10, back-patch a minimal
fix for that problem. The rest of this is just future-proofing.
Discussion: https://postgr.es/m/4627.1521070268@sss.pgh.pa.us
2018-03-15 22:08:51 +01:00
|
|
|
select count(*) from join_foo
|
|
|
|
left join (select b1.id, b1.t from join_bar b1 join join_bar b2 using (id)) ss
|
|
|
|
on join_foo.id < ss.id + 1 and join_foo.id > ss.id - 1;
|
|
|
|
QUERY PLAN
|
|
|
|
------------------------------------------------------------------------------------
|
Add parallel-aware hash joins.
Introduce parallel-aware hash joins that appear in EXPLAIN plans as Parallel
Hash Join with Parallel Hash. While hash joins could already appear in
parallel queries, they were previously always parallel-oblivious and had a
partial subplan only on the outer side, meaning that the work of the inner
subplan was duplicated in every worker.
After this commit, the planner will consider using a partial subplan on the
inner side too, using the Parallel Hash node to divide the work over the
available CPU cores and combine its results in shared memory. If the join
needs to be split into multiple batches in order to respect work_mem, then
workers process different batches as much as possible and then work together
on the remaining batches.
The advantages of a parallel-aware hash join over a parallel-oblivious hash
join used in a parallel query are that it:
* avoids wasting memory on duplicated hash tables
* avoids wasting disk space on duplicated batch files
* divides the work of building the hash table over the CPUs
One disadvantage is that there is some communication between the participating
CPUs which might outweigh the benefits of parallelism in the case of small
hash tables. This is avoided by the planner's existing reluctance to supply
partial plans for small scans, but it may be necessary to estimate
synchronization costs in future if that situation changes. Another is that
outer batch 0 must be written to disk if multiple batches are required.
A potential future advantage of parallel-aware hash joins is that right and
full outer joins could be supported, since there is a single set of matched
bits for each hashtable, but that is not yet implemented.
A new GUC enable_parallel_hash is defined to control the feature, defaulting
to on.
Author: Thomas Munro
Reviewed-By: Andres Freund, Robert Haas
Tested-By: Rafia Sabih, Prabhat Sahu
Discussion:
https://postgr.es/m/CAEepm=2W=cOkiZxcg6qiFQP-dHUe09aqTrEMM7yJDrHMhDv_RA@mail.gmail.com
https://postgr.es/m/CAEepm=37HKyJ4U6XOLi=JgfSHM3o6B-GaeO-6hkOmneTDkH+Uw@mail.gmail.com
2017-12-21 08:39:21 +01:00
|
|
|
Aggregate
|
|
|
|
-> Nested Loop Left Join
|
Clean up duplicate table and function names in regression tests.
Many of the objects we create during the regression tests are put in the
public schema, so that using the same names in different regression tests
creates a hazard of test failures if any two such scripts run concurrently.
This patch cleans up a bunch of latent hazards of that sort, as well as two
live hazards.
The current situation in this regard is far worse than it was a year or two
back, because practically all of the partitioning-related test cases have
reused table names with enthusiasm. I despaired of cleaning up that mess
within the five most-affected tests (create_table, alter_table, insert,
update, inherit); fortunately those don't run concurrently.
Other than partitioning problems, most of the issues boil down to using
names like "foo", "bar", "tmp", etc, without thought for the fact that
other test scripts might use similar names concurrently. I've made an
effort to make all such names more specific.
One of the live hazards was that commit 7421f4b8 caused with.sql to
create a table named "test", conflicting with a similarly-named table
in alter_table.sql; this was exposed in the buildfarm recently.
The other one was that join.sql and transactions.sql both create tables
named "foo" and "bar"; but join.sql's uses of those names date back
only to December or so.
Since commit 7421f4b8 was back-patched to v10, back-patch a minimal
fix for that problem. The rest of this is just future-proofing.
Discussion: https://postgr.es/m/4627.1521070268@sss.pgh.pa.us
2018-03-15 22:08:51 +01:00
|
|
|
Join Filter: ((join_foo.id < (b1.id + 1)) AND (join_foo.id > (b1.id - 1)))
|
|
|
|
-> Seq Scan on join_foo
|
Add parallel-aware hash joins.
Introduce parallel-aware hash joins that appear in EXPLAIN plans as Parallel
Hash Join with Parallel Hash. While hash joins could already appear in
parallel queries, they were previously always parallel-oblivious and had a
partial subplan only on the outer side, meaning that the work of the inner
subplan was duplicated in every worker.
After this commit, the planner will consider using a partial subplan on the
inner side too, using the Parallel Hash node to divide the work over the
available CPU cores and combine its results in shared memory. If the join
needs to be split into multiple batches in order to respect work_mem, then
workers process different batches as much as possible and then work together
on the remaining batches.
The advantages of a parallel-aware hash join over a parallel-oblivious hash
join used in a parallel query are that it:
* avoids wasting memory on duplicated hash tables
* avoids wasting disk space on duplicated batch files
* divides the work of building the hash table over the CPUs
One disadvantage is that there is some communication between the participating
CPUs which might outweigh the benefits of parallelism in the case of small
hash tables. This is avoided by the planner's existing reluctance to supply
partial plans for small scans, but it may be necessary to estimate
synchronization costs in future if that situation changes. Another is that
outer batch 0 must be written to disk if multiple batches are required.
A potential future advantage of parallel-aware hash joins is that right and
full outer joins could be supported, since there is a single set of matched
bits for each hashtable, but that is not yet implemented.
A new GUC enable_parallel_hash is defined to control the feature, defaulting
to on.
Author: Thomas Munro
Reviewed-By: Andres Freund, Robert Haas
Tested-By: Rafia Sabih, Prabhat Sahu
Discussion:
https://postgr.es/m/CAEepm=2W=cOkiZxcg6qiFQP-dHUe09aqTrEMM7yJDrHMhDv_RA@mail.gmail.com
https://postgr.es/m/CAEepm=37HKyJ4U6XOLi=JgfSHM3o6B-GaeO-6hkOmneTDkH+Uw@mail.gmail.com
2017-12-21 08:39:21 +01:00
|
|
|
-> Gather
|
|
|
|
Workers Planned: 2
|
|
|
|
-> Parallel Hash Join
|
|
|
|
Hash Cond: (b1.id = b2.id)
|
Clean up duplicate table and function names in regression tests.
Many of the objects we create during the regression tests are put in the
public schema, so that using the same names in different regression tests
creates a hazard of test failures if any two such scripts run concurrently.
This patch cleans up a bunch of latent hazards of that sort, as well as two
live hazards.
The current situation in this regard is far worse than it was a year or two
back, because practically all of the partitioning-related test cases have
reused table names with enthusiasm. I despaired of cleaning up that mess
within the five most-affected tests (create_table, alter_table, insert,
update, inherit); fortunately those don't run concurrently.
Other than partitioning problems, most of the issues boil down to using
names like "foo", "bar", "tmp", etc, without thought for the fact that
other test scripts might use similar names concurrently. I've made an
effort to make all such names more specific.
One of the live hazards was that commit 7421f4b8 caused with.sql to
create a table named "test", conflicting with a similarly-named table
in alter_table.sql; this was exposed in the buildfarm recently.
The other one was that join.sql and transactions.sql both create tables
named "foo" and "bar"; but join.sql's uses of those names date back
only to December or so.
Since commit 7421f4b8 was back-patched to v10, back-patch a minimal
fix for that problem. The rest of this is just future-proofing.
Discussion: https://postgr.es/m/4627.1521070268@sss.pgh.pa.us
2018-03-15 22:08:51 +01:00
|
|
|
-> Parallel Seq Scan on join_bar b1
|
Add parallel-aware hash joins.
Introduce parallel-aware hash joins that appear in EXPLAIN plans as Parallel
Hash Join with Parallel Hash. While hash joins could already appear in
parallel queries, they were previously always parallel-oblivious and had a
partial subplan only on the outer side, meaning that the work of the inner
subplan was duplicated in every worker.
After this commit, the planner will consider using a partial subplan on the
inner side too, using the Parallel Hash node to divide the work over the
available CPU cores and combine its results in shared memory. If the join
needs to be split into multiple batches in order to respect work_mem, then
workers process different batches as much as possible and then work together
on the remaining batches.
The advantages of a parallel-aware hash join over a parallel-oblivious hash
join used in a parallel query are that it:
* avoids wasting memory on duplicated hash tables
* avoids wasting disk space on duplicated batch files
* divides the work of building the hash table over the CPUs
One disadvantage is that there is some communication between the participating
CPUs which might outweigh the benefits of parallelism in the case of small
hash tables. This is avoided by the planner's existing reluctance to supply
partial plans for small scans, but it may be necessary to estimate
synchronization costs in future if that situation changes. Another is that
outer batch 0 must be written to disk if multiple batches are required.
A potential future advantage of parallel-aware hash joins is that right and
full outer joins could be supported, since there is a single set of matched
bits for each hashtable, but that is not yet implemented.
A new GUC enable_parallel_hash is defined to control the feature, defaulting
to on.
Author: Thomas Munro
Reviewed-By: Andres Freund, Robert Haas
Tested-By: Rafia Sabih, Prabhat Sahu
Discussion:
https://postgr.es/m/CAEepm=2W=cOkiZxcg6qiFQP-dHUe09aqTrEMM7yJDrHMhDv_RA@mail.gmail.com
https://postgr.es/m/CAEepm=37HKyJ4U6XOLi=JgfSHM3o6B-GaeO-6hkOmneTDkH+Uw@mail.gmail.com
2017-12-21 08:39:21 +01:00
|
|
|
-> Parallel Hash
|
Clean up duplicate table and function names in regression tests.
Many of the objects we create during the regression tests are put in the
public schema, so that using the same names in different regression tests
creates a hazard of test failures if any two such scripts run concurrently.
This patch cleans up a bunch of latent hazards of that sort, as well as two
live hazards.
The current situation in this regard is far worse than it was a year or two
back, because practically all of the partitioning-related test cases have
reused table names with enthusiasm. I despaired of cleaning up that mess
within the five most-affected tests (create_table, alter_table, insert,
update, inherit); fortunately those don't run concurrently.
Other than partitioning problems, most of the issues boil down to using
names like "foo", "bar", "tmp", etc, without thought for the fact that
other test scripts might use similar names concurrently. I've made an
effort to make all such names more specific.
One of the live hazards was that commit 7421f4b8 caused with.sql to
create a table named "test", conflicting with a similarly-named table
in alter_table.sql; this was exposed in the buildfarm recently.
The other one was that join.sql and transactions.sql both create tables
named "foo" and "bar"; but join.sql's uses of those names date back
only to December or so.
Since commit 7421f4b8 was back-patched to v10, back-patch a minimal
fix for that problem. The rest of this is just future-proofing.
Discussion: https://postgr.es/m/4627.1521070268@sss.pgh.pa.us
2018-03-15 22:08:51 +01:00
|
|
|
-> Parallel Seq Scan on join_bar b2
|
Add parallel-aware hash joins.
Introduce parallel-aware hash joins that appear in EXPLAIN plans as Parallel
Hash Join with Parallel Hash. While hash joins could already appear in
parallel queries, they were previously always parallel-oblivious and had a
partial subplan only on the outer side, meaning that the work of the inner
subplan was duplicated in every worker.
After this commit, the planner will consider using a partial subplan on the
inner side too, using the Parallel Hash node to divide the work over the
available CPU cores and combine its results in shared memory. If the join
needs to be split into multiple batches in order to respect work_mem, then
workers process different batches as much as possible and then work together
on the remaining batches.
The advantages of a parallel-aware hash join over a parallel-oblivious hash
join used in a parallel query are that it:
* avoids wasting memory on duplicated hash tables
* avoids wasting disk space on duplicated batch files
* divides the work of building the hash table over the CPUs
One disadvantage is that there is some communication between the participating
CPUs which might outweigh the benefits of parallelism in the case of small
hash tables. This is avoided by the planner's existing reluctance to supply
partial plans for small scans, but it may be necessary to estimate
synchronization costs in future if that situation changes. Another is that
outer batch 0 must be written to disk if multiple batches are required.
A potential future advantage of parallel-aware hash joins is that right and
full outer joins could be supported, since there is a single set of matched
bits for each hashtable, but that is not yet implemented.
A new GUC enable_parallel_hash is defined to control the feature, defaulting
to on.
Author: Thomas Munro
Reviewed-By: Andres Freund, Robert Haas
Tested-By: Rafia Sabih, Prabhat Sahu
Discussion:
https://postgr.es/m/CAEepm=2W=cOkiZxcg6qiFQP-dHUe09aqTrEMM7yJDrHMhDv_RA@mail.gmail.com
https://postgr.es/m/CAEepm=37HKyJ4U6XOLi=JgfSHM3o6B-GaeO-6hkOmneTDkH+Uw@mail.gmail.com
2017-12-21 08:39:21 +01:00
|
|
|
(11 rows)
|
|
|
|
|
Clean up duplicate table and function names in regression tests.
Many of the objects we create during the regression tests are put in the
public schema, so that using the same names in different regression tests
creates a hazard of test failures if any two such scripts run concurrently.
This patch cleans up a bunch of latent hazards of that sort, as well as two
live hazards.
The current situation in this regard is far worse than it was a year or two
back, because practically all of the partitioning-related test cases have
reused table names with enthusiasm. I despaired of cleaning up that mess
within the five most-affected tests (create_table, alter_table, insert,
update, inherit); fortunately those don't run concurrently.
Other than partitioning problems, most of the issues boil down to using
names like "foo", "bar", "tmp", etc, without thought for the fact that
other test scripts might use similar names concurrently. I've made an
effort to make all such names more specific.
One of the live hazards was that commit 7421f4b8 caused with.sql to
create a table named "test", conflicting with a similarly-named table
in alter_table.sql; this was exposed in the buildfarm recently.
The other one was that join.sql and transactions.sql both create tables
named "foo" and "bar"; but join.sql's uses of those names date back
only to December or so.
Since commit 7421f4b8 was back-patched to v10, back-patch a minimal
fix for that problem. The rest of this is just future-proofing.
Discussion: https://postgr.es/m/4627.1521070268@sss.pgh.pa.us
2018-03-15 22:08:51 +01:00
|
|
|
select count(*) from join_foo
|
|
|
|
left join (select b1.id, b1.t from join_bar b1 join join_bar b2 using (id)) ss
|
|
|
|
on join_foo.id < ss.id + 1 and join_foo.id > ss.id - 1;
|
Add parallel-aware hash joins.
Introduce parallel-aware hash joins that appear in EXPLAIN plans as Parallel
Hash Join with Parallel Hash. While hash joins could already appear in
parallel queries, they were previously always parallel-oblivious and had a
partial subplan only on the outer side, meaning that the work of the inner
subplan was duplicated in every worker.
After this commit, the planner will consider using a partial subplan on the
inner side too, using the Parallel Hash node to divide the work over the
available CPU cores and combine its results in shared memory. If the join
needs to be split into multiple batches in order to respect work_mem, then
workers process different batches as much as possible and then work together
on the remaining batches.
The advantages of a parallel-aware hash join over a parallel-oblivious hash
join used in a parallel query are that it:
* avoids wasting memory on duplicated hash tables
* avoids wasting disk space on duplicated batch files
* divides the work of building the hash table over the CPUs
One disadvantage is that there is some communication between the participating
CPUs which might outweigh the benefits of parallelism in the case of small
hash tables. This is avoided by the planner's existing reluctance to supply
partial plans for small scans, but it may be necessary to estimate
synchronization costs in future if that situation changes. Another is that
outer batch 0 must be written to disk if multiple batches are required.
A potential future advantage of parallel-aware hash joins is that right and
full outer joins could be supported, since there is a single set of matched
bits for each hashtable, but that is not yet implemented.
A new GUC enable_parallel_hash is defined to control the feature, defaulting
to on.
Author: Thomas Munro
Reviewed-By: Andres Freund, Robert Haas
Tested-By: Rafia Sabih, Prabhat Sahu
Discussion:
https://postgr.es/m/CAEepm=2W=cOkiZxcg6qiFQP-dHUe09aqTrEMM7yJDrHMhDv_RA@mail.gmail.com
https://postgr.es/m/CAEepm=37HKyJ4U6XOLi=JgfSHM3o6B-GaeO-6hkOmneTDkH+Uw@mail.gmail.com
2017-12-21 08:39:21 +01:00
|
|
|
count
|
|
|
|
-------
|
|
|
|
3
|
|
|
|
(1 row)
|
|
|
|
|
|
|
|
select final > 1 as multibatch
|
|
|
|
from hash_join_batches(
|
|
|
|
$$
|
Clean up duplicate table and function names in regression tests.
Many of the objects we create during the regression tests are put in the
public schema, so that using the same names in different regression tests
creates a hazard of test failures if any two such scripts run concurrently.
This patch cleans up a bunch of latent hazards of that sort, as well as two
live hazards.
The current situation in this regard is far worse than it was a year or two
back, because practically all of the partitioning-related test cases have
reused table names with enthusiasm. I despaired of cleaning up that mess
within the five most-affected tests (create_table, alter_table, insert,
update, inherit); fortunately those don't run concurrently.
Other than partitioning problems, most of the issues boil down to using
names like "foo", "bar", "tmp", etc, without thought for the fact that
other test scripts might use similar names concurrently. I've made an
effort to make all such names more specific.
One of the live hazards was that commit 7421f4b8 caused with.sql to
create a table named "test", conflicting with a similarly-named table
in alter_table.sql; this was exposed in the buildfarm recently.
The other one was that join.sql and transactions.sql both create tables
named "foo" and "bar"; but join.sql's uses of those names date back
only to December or so.
Since commit 7421f4b8 was back-patched to v10, back-patch a minimal
fix for that problem. The rest of this is just future-proofing.
Discussion: https://postgr.es/m/4627.1521070268@sss.pgh.pa.us
2018-03-15 22:08:51 +01:00
|
|
|
select count(*) from join_foo
|
|
|
|
left join (select b1.id, b1.t from join_bar b1 join join_bar b2 using (id)) ss
|
|
|
|
on join_foo.id < ss.id + 1 and join_foo.id > ss.id - 1;
|
Add parallel-aware hash joins.
Introduce parallel-aware hash joins that appear in EXPLAIN plans as Parallel
Hash Join with Parallel Hash. While hash joins could already appear in
parallel queries, they were previously always parallel-oblivious and had a
partial subplan only on the outer side, meaning that the work of the inner
subplan was duplicated in every worker.
After this commit, the planner will consider using a partial subplan on the
inner side too, using the Parallel Hash node to divide the work over the
available CPU cores and combine its results in shared memory. If the join
needs to be split into multiple batches in order to respect work_mem, then
workers process different batches as much as possible and then work together
on the remaining batches.
The advantages of a parallel-aware hash join over a parallel-oblivious hash
join used in a parallel query are that it:
* avoids wasting memory on duplicated hash tables
* avoids wasting disk space on duplicated batch files
* divides the work of building the hash table over the CPUs
One disadvantage is that there is some communication between the participating
CPUs which might outweigh the benefits of parallelism in the case of small
hash tables. This is avoided by the planner's existing reluctance to supply
partial plans for small scans, but it may be necessary to estimate
synchronization costs in future if that situation changes. Another is that
outer batch 0 must be written to disk if multiple batches are required.
A potential future advantage of parallel-aware hash joins is that right and
full outer joins could be supported, since there is a single set of matched
bits for each hashtable, but that is not yet implemented.
A new GUC enable_parallel_hash is defined to control the feature, defaulting
to on.
Author: Thomas Munro
Reviewed-By: Andres Freund, Robert Haas
Tested-By: Rafia Sabih, Prabhat Sahu
Discussion:
https://postgr.es/m/CAEepm=2W=cOkiZxcg6qiFQP-dHUe09aqTrEMM7yJDrHMhDv_RA@mail.gmail.com
https://postgr.es/m/CAEepm=37HKyJ4U6XOLi=JgfSHM3o6B-GaeO-6hkOmneTDkH+Uw@mail.gmail.com
2017-12-21 08:39:21 +01:00
|
|
|
$$);
|
|
|
|
multibatch
|
|
|
|
------------
|
|
|
|
t
|
|
|
|
(1 row)
|
|
|
|
|
|
|
|
rollback to settings;
|
|
|
|
-- single-batch with rescan, parallel-aware
|
|
|
|
savepoint settings;
|
|
|
|
set enable_parallel_hash = on;
|
|
|
|
set parallel_leader_participation = off;
|
|
|
|
set min_parallel_table_scan_size = 0;
|
|
|
|
set parallel_setup_cost = 0;
|
|
|
|
set parallel_tuple_cost = 0;
|
|
|
|
set max_parallel_workers_per_gather = 2;
|
|
|
|
set enable_material = off;
|
|
|
|
set enable_mergejoin = off;
|
|
|
|
set work_mem = '4MB';
|
|
|
|
explain (costs off)
|
Clean up duplicate table and function names in regression tests.
Many of the objects we create during the regression tests are put in the
public schema, so that using the same names in different regression tests
creates a hazard of test failures if any two such scripts run concurrently.
This patch cleans up a bunch of latent hazards of that sort, as well as two
live hazards.
The current situation in this regard is far worse than it was a year or two
back, because practically all of the partitioning-related test cases have
reused table names with enthusiasm. I despaired of cleaning up that mess
within the five most-affected tests (create_table, alter_table, insert,
update, inherit); fortunately those don't run concurrently.
Other than partitioning problems, most of the issues boil down to using
names like "foo", "bar", "tmp", etc, without thought for the fact that
other test scripts might use similar names concurrently. I've made an
effort to make all such names more specific.
One of the live hazards was that commit 7421f4b8 caused with.sql to
create a table named "test", conflicting with a similarly-named table
in alter_table.sql; this was exposed in the buildfarm recently.
The other one was that join.sql and transactions.sql both create tables
named "foo" and "bar"; but join.sql's uses of those names date back
only to December or so.
Since commit 7421f4b8 was back-patched to v10, back-patch a minimal
fix for that problem. The rest of this is just future-proofing.
Discussion: https://postgr.es/m/4627.1521070268@sss.pgh.pa.us
2018-03-15 22:08:51 +01:00
|
|
|
select count(*) from join_foo
|
|
|
|
left join (select b1.id, b1.t from join_bar b1 join join_bar b2 using (id)) ss
|
|
|
|
on join_foo.id < ss.id + 1 and join_foo.id > ss.id - 1;
|
|
|
|
QUERY PLAN
|
|
|
|
------------------------------------------------------------------------------------
|
Add parallel-aware hash joins.
Introduce parallel-aware hash joins that appear in EXPLAIN plans as Parallel
Hash Join with Parallel Hash. While hash joins could already appear in
parallel queries, they were previously always parallel-oblivious and had a
partial subplan only on the outer side, meaning that the work of the inner
subplan was duplicated in every worker.
After this commit, the planner will consider using a partial subplan on the
inner side too, using the Parallel Hash node to divide the work over the
available CPU cores and combine its results in shared memory. If the join
needs to be split into multiple batches in order to respect work_mem, then
workers process different batches as much as possible and then work together
on the remaining batches.
The advantages of a parallel-aware hash join over a parallel-oblivious hash
join used in a parallel query are that it:
* avoids wasting memory on duplicated hash tables
* avoids wasting disk space on duplicated batch files
* divides the work of building the hash table over the CPUs
One disadvantage is that there is some communication between the participating
CPUs which might outweigh the benefits of parallelism in the case of small
hash tables. This is avoided by the planner's existing reluctance to supply
partial plans for small scans, but it may be necessary to estimate
synchronization costs in future if that situation changes. Another is that
outer batch 0 must be written to disk if multiple batches are required.
A potential future advantage of parallel-aware hash joins is that right and
full outer joins could be supported, since there is a single set of matched
bits for each hashtable, but that is not yet implemented.
A new GUC enable_parallel_hash is defined to control the feature, defaulting
to on.
Author: Thomas Munro
Reviewed-By: Andres Freund, Robert Haas
Tested-By: Rafia Sabih, Prabhat Sahu
Discussion:
https://postgr.es/m/CAEepm=2W=cOkiZxcg6qiFQP-dHUe09aqTrEMM7yJDrHMhDv_RA@mail.gmail.com
https://postgr.es/m/CAEepm=37HKyJ4U6XOLi=JgfSHM3o6B-GaeO-6hkOmneTDkH+Uw@mail.gmail.com
2017-12-21 08:39:21 +01:00
|
|
|
Aggregate
|
|
|
|
-> Nested Loop Left Join
|
Clean up duplicate table and function names in regression tests.
Many of the objects we create during the regression tests are put in the
public schema, so that using the same names in different regression tests
creates a hazard of test failures if any two such scripts run concurrently.
This patch cleans up a bunch of latent hazards of that sort, as well as two
live hazards.
The current situation in this regard is far worse than it was a year or two
back, because practically all of the partitioning-related test cases have
reused table names with enthusiasm. I despaired of cleaning up that mess
within the five most-affected tests (create_table, alter_table, insert,
update, inherit); fortunately those don't run concurrently.
Other than partitioning problems, most of the issues boil down to using
names like "foo", "bar", "tmp", etc, without thought for the fact that
other test scripts might use similar names concurrently. I've made an
effort to make all such names more specific.
One of the live hazards was that commit 7421f4b8 caused with.sql to
create a table named "test", conflicting with a similarly-named table
in alter_table.sql; this was exposed in the buildfarm recently.
The other one was that join.sql and transactions.sql both create tables
named "foo" and "bar"; but join.sql's uses of those names date back
only to December or so.
Since commit 7421f4b8 was back-patched to v10, back-patch a minimal
fix for that problem. The rest of this is just future-proofing.
Discussion: https://postgr.es/m/4627.1521070268@sss.pgh.pa.us
2018-03-15 22:08:51 +01:00
|
|
|
Join Filter: ((join_foo.id < (b1.id + 1)) AND (join_foo.id > (b1.id - 1)))
|
|
|
|
-> Seq Scan on join_foo
|
Add parallel-aware hash joins.
Introduce parallel-aware hash joins that appear in EXPLAIN plans as Parallel
Hash Join with Parallel Hash. While hash joins could already appear in
parallel queries, they were previously always parallel-oblivious and had a
partial subplan only on the outer side, meaning that the work of the inner
subplan was duplicated in every worker.
After this commit, the planner will consider using a partial subplan on the
inner side too, using the Parallel Hash node to divide the work over the
available CPU cores and combine its results in shared memory. If the join
needs to be split into multiple batches in order to respect work_mem, then
workers process different batches as much as possible and then work together
on the remaining batches.
The advantages of a parallel-aware hash join over a parallel-oblivious hash
join used in a parallel query are that it:
* avoids wasting memory on duplicated hash tables
* avoids wasting disk space on duplicated batch files
* divides the work of building the hash table over the CPUs
One disadvantage is that there is some communication between the participating
CPUs which might outweigh the benefits of parallelism in the case of small
hash tables. This is avoided by the planner's existing reluctance to supply
partial plans for small scans, but it may be necessary to estimate
synchronization costs in future if that situation changes. Another is that
outer batch 0 must be written to disk if multiple batches are required.
A potential future advantage of parallel-aware hash joins is that right and
full outer joins could be supported, since there is a single set of matched
bits for each hashtable, but that is not yet implemented.
A new GUC enable_parallel_hash is defined to control the feature, defaulting
to on.
Author: Thomas Munro
Reviewed-By: Andres Freund, Robert Haas
Tested-By: Rafia Sabih, Prabhat Sahu
Discussion:
https://postgr.es/m/CAEepm=2W=cOkiZxcg6qiFQP-dHUe09aqTrEMM7yJDrHMhDv_RA@mail.gmail.com
https://postgr.es/m/CAEepm=37HKyJ4U6XOLi=JgfSHM3o6B-GaeO-6hkOmneTDkH+Uw@mail.gmail.com
2017-12-21 08:39:21 +01:00
|
|
|
-> Gather
|
|
|
|
Workers Planned: 2
|
|
|
|
-> Parallel Hash Join
|
|
|
|
Hash Cond: (b1.id = b2.id)
|
Clean up duplicate table and function names in regression tests.
Many of the objects we create during the regression tests are put in the
public schema, so that using the same names in different regression tests
creates a hazard of test failures if any two such scripts run concurrently.
This patch cleans up a bunch of latent hazards of that sort, as well as two
live hazards.
The current situation in this regard is far worse than it was a year or two
back, because practically all of the partitioning-related test cases have
reused table names with enthusiasm. I despaired of cleaning up that mess
within the five most-affected tests (create_table, alter_table, insert,
update, inherit); fortunately those don't run concurrently.
Other than partitioning problems, most of the issues boil down to using
names like "foo", "bar", "tmp", etc, without thought for the fact that
other test scripts might use similar names concurrently. I've made an
effort to make all such names more specific.
One of the live hazards was that commit 7421f4b8 caused with.sql to
create a table named "test", conflicting with a similarly-named table
in alter_table.sql; this was exposed in the buildfarm recently.
The other one was that join.sql and transactions.sql both create tables
named "foo" and "bar"; but join.sql's uses of those names date back
only to December or so.
Since commit 7421f4b8 was back-patched to v10, back-patch a minimal
fix for that problem. The rest of this is just future-proofing.
Discussion: https://postgr.es/m/4627.1521070268@sss.pgh.pa.us
2018-03-15 22:08:51 +01:00
|
|
|
-> Parallel Seq Scan on join_bar b1
|
Add parallel-aware hash joins.
Introduce parallel-aware hash joins that appear in EXPLAIN plans as Parallel
Hash Join with Parallel Hash. While hash joins could already appear in
parallel queries, they were previously always parallel-oblivious and had a
partial subplan only on the outer side, meaning that the work of the inner
subplan was duplicated in every worker.
After this commit, the planner will consider using a partial subplan on the
inner side too, using the Parallel Hash node to divide the work over the
available CPU cores and combine its results in shared memory. If the join
needs to be split into multiple batches in order to respect work_mem, then
workers process different batches as much as possible and then work together
on the remaining batches.
The advantages of a parallel-aware hash join over a parallel-oblivious hash
join used in a parallel query are that it:
* avoids wasting memory on duplicated hash tables
* avoids wasting disk space on duplicated batch files
* divides the work of building the hash table over the CPUs
One disadvantage is that there is some communication between the participating
CPUs which might outweigh the benefits of parallelism in the case of small
hash tables. This is avoided by the planner's existing reluctance to supply
partial plans for small scans, but it may be necessary to estimate
synchronization costs in future if that situation changes. Another is that
outer batch 0 must be written to disk if multiple batches are required.
A potential future advantage of parallel-aware hash joins is that right and
full outer joins could be supported, since there is a single set of matched
bits for each hashtable, but that is not yet implemented.
A new GUC enable_parallel_hash is defined to control the feature, defaulting
to on.
Author: Thomas Munro
Reviewed-By: Andres Freund, Robert Haas
Tested-By: Rafia Sabih, Prabhat Sahu
Discussion:
https://postgr.es/m/CAEepm=2W=cOkiZxcg6qiFQP-dHUe09aqTrEMM7yJDrHMhDv_RA@mail.gmail.com
https://postgr.es/m/CAEepm=37HKyJ4U6XOLi=JgfSHM3o6B-GaeO-6hkOmneTDkH+Uw@mail.gmail.com
2017-12-21 08:39:21 +01:00
|
|
|
-> Parallel Hash
|
Clean up duplicate table and function names in regression tests.
Many of the objects we create during the regression tests are put in the
public schema, so that using the same names in different regression tests
creates a hazard of test failures if any two such scripts run concurrently.
This patch cleans up a bunch of latent hazards of that sort, as well as two
live hazards.
The current situation in this regard is far worse than it was a year or two
back, because practically all of the partitioning-related test cases have
reused table names with enthusiasm. I despaired of cleaning up that mess
within the five most-affected tests (create_table, alter_table, insert,
update, inherit); fortunately those don't run concurrently.
Other than partitioning problems, most of the issues boil down to using
names like "foo", "bar", "tmp", etc, without thought for the fact that
other test scripts might use similar names concurrently. I've made an
effort to make all such names more specific.
One of the live hazards was that commit 7421f4b8 caused with.sql to
create a table named "test", conflicting with a similarly-named table
in alter_table.sql; this was exposed in the buildfarm recently.
The other one was that join.sql and transactions.sql both create tables
named "foo" and "bar"; but join.sql's uses of those names date back
only to December or so.
Since commit 7421f4b8 was back-patched to v10, back-patch a minimal
fix for that problem. The rest of this is just future-proofing.
Discussion: https://postgr.es/m/4627.1521070268@sss.pgh.pa.us
2018-03-15 22:08:51 +01:00
|
|
|
-> Parallel Seq Scan on join_bar b2
|
Add parallel-aware hash joins.
Introduce parallel-aware hash joins that appear in EXPLAIN plans as Parallel
Hash Join with Parallel Hash. While hash joins could already appear in
parallel queries, they were previously always parallel-oblivious and had a
partial subplan only on the outer side, meaning that the work of the inner
subplan was duplicated in every worker.
After this commit, the planner will consider using a partial subplan on the
inner side too, using the Parallel Hash node to divide the work over the
available CPU cores and combine its results in shared memory. If the join
needs to be split into multiple batches in order to respect work_mem, then
workers process different batches as much as possible and then work together
on the remaining batches.
The advantages of a parallel-aware hash join over a parallel-oblivious hash
join used in a parallel query are that it:
* avoids wasting memory on duplicated hash tables
* avoids wasting disk space on duplicated batch files
* divides the work of building the hash table over the CPUs
One disadvantage is that there is some communication between the participating
CPUs which might outweigh the benefits of parallelism in the case of small
hash tables. This is avoided by the planner's existing reluctance to supply
partial plans for small scans, but it may be necessary to estimate
synchronization costs in future if that situation changes. Another is that
outer batch 0 must be written to disk if multiple batches are required.
A potential future advantage of parallel-aware hash joins is that right and
full outer joins could be supported, since there is a single set of matched
bits for each hashtable, but that is not yet implemented.
A new GUC enable_parallel_hash is defined to control the feature, defaulting
to on.
Author: Thomas Munro
Reviewed-By: Andres Freund, Robert Haas
Tested-By: Rafia Sabih, Prabhat Sahu
Discussion:
https://postgr.es/m/CAEepm=2W=cOkiZxcg6qiFQP-dHUe09aqTrEMM7yJDrHMhDv_RA@mail.gmail.com
https://postgr.es/m/CAEepm=37HKyJ4U6XOLi=JgfSHM3o6B-GaeO-6hkOmneTDkH+Uw@mail.gmail.com
2017-12-21 08:39:21 +01:00
|
|
|
(11 rows)
|
|
|
|
|
Clean up duplicate table and function names in regression tests.
Many of the objects we create during the regression tests are put in the
public schema, so that using the same names in different regression tests
creates a hazard of test failures if any two such scripts run concurrently.
This patch cleans up a bunch of latent hazards of that sort, as well as two
live hazards.
The current situation in this regard is far worse than it was a year or two
back, because practically all of the partitioning-related test cases have
reused table names with enthusiasm. I despaired of cleaning up that mess
within the five most-affected tests (create_table, alter_table, insert,
update, inherit); fortunately those don't run concurrently.
Other than partitioning problems, most of the issues boil down to using
names like "foo", "bar", "tmp", etc, without thought for the fact that
other test scripts might use similar names concurrently. I've made an
effort to make all such names more specific.
One of the live hazards was that commit 7421f4b8 caused with.sql to
create a table named "test", conflicting with a similarly-named table
in alter_table.sql; this was exposed in the buildfarm recently.
The other one was that join.sql and transactions.sql both create tables
named "foo" and "bar"; but join.sql's uses of those names date back
only to December or so.
Since commit 7421f4b8 was back-patched to v10, back-patch a minimal
fix for that problem. The rest of this is just future-proofing.
Discussion: https://postgr.es/m/4627.1521070268@sss.pgh.pa.us
2018-03-15 22:08:51 +01:00
|
|
|
select count(*) from join_foo
|
|
|
|
left join (select b1.id, b1.t from join_bar b1 join join_bar b2 using (id)) ss
|
|
|
|
on join_foo.id < ss.id + 1 and join_foo.id > ss.id - 1;
|
Add parallel-aware hash joins.
Introduce parallel-aware hash joins that appear in EXPLAIN plans as Parallel
Hash Join with Parallel Hash. While hash joins could already appear in
parallel queries, they were previously always parallel-oblivious and had a
partial subplan only on the outer side, meaning that the work of the inner
subplan was duplicated in every worker.
After this commit, the planner will consider using a partial subplan on the
inner side too, using the Parallel Hash node to divide the work over the
available CPU cores and combine its results in shared memory. If the join
needs to be split into multiple batches in order to respect work_mem, then
workers process different batches as much as possible and then work together
on the remaining batches.
The advantages of a parallel-aware hash join over a parallel-oblivious hash
join used in a parallel query are that it:
* avoids wasting memory on duplicated hash tables
* avoids wasting disk space on duplicated batch files
* divides the work of building the hash table over the CPUs
One disadvantage is that there is some communication between the participating
CPUs which might outweigh the benefits of parallelism in the case of small
hash tables. This is avoided by the planner's existing reluctance to supply
partial plans for small scans, but it may be necessary to estimate
synchronization costs in future if that situation changes. Another is that
outer batch 0 must be written to disk if multiple batches are required.
A potential future advantage of parallel-aware hash joins is that right and
full outer joins could be supported, since there is a single set of matched
bits for each hashtable, but that is not yet implemented.
A new GUC enable_parallel_hash is defined to control the feature, defaulting
to on.
Author: Thomas Munro
Reviewed-By: Andres Freund, Robert Haas
Tested-By: Rafia Sabih, Prabhat Sahu
Discussion:
https://postgr.es/m/CAEepm=2W=cOkiZxcg6qiFQP-dHUe09aqTrEMM7yJDrHMhDv_RA@mail.gmail.com
https://postgr.es/m/CAEepm=37HKyJ4U6XOLi=JgfSHM3o6B-GaeO-6hkOmneTDkH+Uw@mail.gmail.com
2017-12-21 08:39:21 +01:00
|
|
|
count
|
|
|
|
-------
|
|
|
|
3
|
|
|
|
(1 row)
|
|
|
|
|
|
|
|
select final > 1 as multibatch
|
|
|
|
from hash_join_batches(
|
|
|
|
$$
|
Clean up duplicate table and function names in regression tests.
Many of the objects we create during the regression tests are put in the
public schema, so that using the same names in different regression tests
creates a hazard of test failures if any two such scripts run concurrently.
This patch cleans up a bunch of latent hazards of that sort, as well as two
live hazards.
The current situation in this regard is far worse than it was a year or two
back, because practically all of the partitioning-related test cases have
reused table names with enthusiasm. I despaired of cleaning up that mess
within the five most-affected tests (create_table, alter_table, insert,
update, inherit); fortunately those don't run concurrently.
Other than partitioning problems, most of the issues boil down to using
names like "foo", "bar", "tmp", etc, without thought for the fact that
other test scripts might use similar names concurrently. I've made an
effort to make all such names more specific.
One of the live hazards was that commit 7421f4b8 caused with.sql to
create a table named "test", conflicting with a similarly-named table
in alter_table.sql; this was exposed in the buildfarm recently.
The other one was that join.sql and transactions.sql both create tables
named "foo" and "bar"; but join.sql's uses of those names date back
only to December or so.
Since commit 7421f4b8 was back-patched to v10, back-patch a minimal
fix for that problem. The rest of this is just future-proofing.
Discussion: https://postgr.es/m/4627.1521070268@sss.pgh.pa.us
2018-03-15 22:08:51 +01:00
|
|
|
select count(*) from join_foo
|
|
|
|
left join (select b1.id, b1.t from join_bar b1 join join_bar b2 using (id)) ss
|
|
|
|
on join_foo.id < ss.id + 1 and join_foo.id > ss.id - 1;
|
Add parallel-aware hash joins.
Introduce parallel-aware hash joins that appear in EXPLAIN plans as Parallel
Hash Join with Parallel Hash. While hash joins could already appear in
parallel queries, they were previously always parallel-oblivious and had a
partial subplan only on the outer side, meaning that the work of the inner
subplan was duplicated in every worker.
After this commit, the planner will consider using a partial subplan on the
inner side too, using the Parallel Hash node to divide the work over the
available CPU cores and combine its results in shared memory. If the join
needs to be split into multiple batches in order to respect work_mem, then
workers process different batches as much as possible and then work together
on the remaining batches.
The advantages of a parallel-aware hash join over a parallel-oblivious hash
join used in a parallel query are that it:
* avoids wasting memory on duplicated hash tables
* avoids wasting disk space on duplicated batch files
* divides the work of building the hash table over the CPUs
One disadvantage is that there is some communication between the participating
CPUs which might outweigh the benefits of parallelism in the case of small
hash tables. This is avoided by the planner's existing reluctance to supply
partial plans for small scans, but it may be necessary to estimate
synchronization costs in future if that situation changes. Another is that
outer batch 0 must be written to disk if multiple batches are required.
A potential future advantage of parallel-aware hash joins is that right and
full outer joins could be supported, since there is a single set of matched
bits for each hashtable, but that is not yet implemented.
A new GUC enable_parallel_hash is defined to control the feature, defaulting
to on.
Author: Thomas Munro
Reviewed-By: Andres Freund, Robert Haas
Tested-By: Rafia Sabih, Prabhat Sahu
Discussion:
https://postgr.es/m/CAEepm=2W=cOkiZxcg6qiFQP-dHUe09aqTrEMM7yJDrHMhDv_RA@mail.gmail.com
https://postgr.es/m/CAEepm=37HKyJ4U6XOLi=JgfSHM3o6B-GaeO-6hkOmneTDkH+Uw@mail.gmail.com
2017-12-21 08:39:21 +01:00
|
|
|
$$);
|
|
|
|
multibatch
|
|
|
|
------------
|
|
|
|
f
|
|
|
|
(1 row)
|
|
|
|
|
2017-12-05 19:55:56 +01:00
|
|
|
rollback to settings;
|
2017-11-30 01:06:50 +01:00
|
|
|
-- A full outer join where every record is matched.
|
|
|
|
-- non-parallel
|
|
|
|
savepoint settings;
|
|
|
|
set local max_parallel_workers_per_gather = 0;
|
|
|
|
explain (costs off)
|
|
|
|
select count(*) from simple r full outer join simple s using (id);
|
|
|
|
QUERY PLAN
|
|
|
|
----------------------------------------
|
|
|
|
Aggregate
|
|
|
|
-> Hash Full Join
|
|
|
|
Hash Cond: (r.id = s.id)
|
|
|
|
-> Seq Scan on simple r
|
|
|
|
-> Hash
|
|
|
|
-> Seq Scan on simple s
|
|
|
|
(6 rows)
|
|
|
|
|
|
|
|
select count(*) from simple r full outer join simple s using (id);
|
|
|
|
count
|
|
|
|
-------
|
|
|
|
20000
|
|
|
|
(1 row)
|
|
|
|
|
|
|
|
rollback to settings;
|
|
|
|
-- parallelism not possible with parallel-oblivious outer hash join
|
|
|
|
savepoint settings;
|
|
|
|
set local max_parallel_workers_per_gather = 2;
|
|
|
|
explain (costs off)
|
|
|
|
select count(*) from simple r full outer join simple s using (id);
|
|
|
|
QUERY PLAN
|
|
|
|
----------------------------------------
|
|
|
|
Aggregate
|
|
|
|
-> Hash Full Join
|
|
|
|
Hash Cond: (r.id = s.id)
|
|
|
|
-> Seq Scan on simple r
|
|
|
|
-> Hash
|
|
|
|
-> Seq Scan on simple s
|
|
|
|
(6 rows)
|
|
|
|
|
|
|
|
select count(*) from simple r full outer join simple s using (id);
|
|
|
|
count
|
|
|
|
-------
|
|
|
|
20000
|
|
|
|
(1 row)
|
|
|
|
|
|
|
|
rollback to settings;
|
|
|
|
-- An full outer join where every record is not matched.
|
|
|
|
-- non-parallel
|
|
|
|
savepoint settings;
|
|
|
|
set local max_parallel_workers_per_gather = 0;
|
|
|
|
explain (costs off)
|
|
|
|
select count(*) from simple r full outer join simple s on (r.id = 0 - s.id);
|
|
|
|
QUERY PLAN
|
|
|
|
----------------------------------------
|
|
|
|
Aggregate
|
|
|
|
-> Hash Full Join
|
|
|
|
Hash Cond: ((0 - s.id) = r.id)
|
|
|
|
-> Seq Scan on simple s
|
|
|
|
-> Hash
|
|
|
|
-> Seq Scan on simple r
|
|
|
|
(6 rows)
|
|
|
|
|
|
|
|
select count(*) from simple r full outer join simple s on (r.id = 0 - s.id);
|
|
|
|
count
|
|
|
|
-------
|
|
|
|
40000
|
|
|
|
(1 row)
|
|
|
|
|
|
|
|
rollback to settings;
|
|
|
|
-- parallelism not possible with parallel-oblivious outer hash join
|
|
|
|
savepoint settings;
|
|
|
|
set local max_parallel_workers_per_gather = 2;
|
|
|
|
explain (costs off)
|
|
|
|
select count(*) from simple r full outer join simple s on (r.id = 0 - s.id);
|
|
|
|
QUERY PLAN
|
|
|
|
----------------------------------------
|
|
|
|
Aggregate
|
|
|
|
-> Hash Full Join
|
|
|
|
Hash Cond: ((0 - s.id) = r.id)
|
|
|
|
-> Seq Scan on simple s
|
|
|
|
-> Hash
|
|
|
|
-> Seq Scan on simple r
|
|
|
|
(6 rows)
|
|
|
|
|
|
|
|
select count(*) from simple r full outer join simple s on (r.id = 0 - s.id);
|
|
|
|
count
|
|
|
|
-------
|
|
|
|
40000
|
|
|
|
(1 row)
|
|
|
|
|
Add parallel-aware hash joins.
Introduce parallel-aware hash joins that appear in EXPLAIN plans as Parallel
Hash Join with Parallel Hash. While hash joins could already appear in
parallel queries, they were previously always parallel-oblivious and had a
partial subplan only on the outer side, meaning that the work of the inner
subplan was duplicated in every worker.
After this commit, the planner will consider using a partial subplan on the
inner side too, using the Parallel Hash node to divide the work over the
available CPU cores and combine its results in shared memory. If the join
needs to be split into multiple batches in order to respect work_mem, then
workers process different batches as much as possible and then work together
on the remaining batches.
The advantages of a parallel-aware hash join over a parallel-oblivious hash
join used in a parallel query are that it:
* avoids wasting memory on duplicated hash tables
* avoids wasting disk space on duplicated batch files
* divides the work of building the hash table over the CPUs
One disadvantage is that there is some communication between the participating
CPUs which might outweigh the benefits of parallelism in the case of small
hash tables. This is avoided by the planner's existing reluctance to supply
partial plans for small scans, but it may be necessary to estimate
synchronization costs in future if that situation changes. Another is that
outer batch 0 must be written to disk if multiple batches are required.
A potential future advantage of parallel-aware hash joins is that right and
full outer joins could be supported, since there is a single set of matched
bits for each hashtable, but that is not yet implemented.
A new GUC enable_parallel_hash is defined to control the feature, defaulting
to on.
Author: Thomas Munro
Reviewed-By: Andres Freund, Robert Haas
Tested-By: Rafia Sabih, Prabhat Sahu
Discussion:
https://postgr.es/m/CAEepm=2W=cOkiZxcg6qiFQP-dHUe09aqTrEMM7yJDrHMhDv_RA@mail.gmail.com
https://postgr.es/m/CAEepm=37HKyJ4U6XOLi=JgfSHM3o6B-GaeO-6hkOmneTDkH+Uw@mail.gmail.com
2017-12-21 08:39:21 +01:00
|
|
|
rollback to settings;
|
|
|
|
-- exercise special code paths for huge tuples (note use of non-strict
|
|
|
|
-- expression and left join required to get the detoasted tuple into
|
|
|
|
-- the hash table)
|
|
|
|
-- parallel with parallel-aware hash join (hits ExecParallelHashLoadTuple and
|
|
|
|
-- sts_puttuple oversized tuple cases because it's multi-batch)
|
|
|
|
savepoint settings;
|
|
|
|
set max_parallel_workers_per_gather = 2;
|
|
|
|
set enable_parallel_hash = on;
|
|
|
|
set work_mem = '128kB';
|
|
|
|
explain (costs off)
|
|
|
|
select length(max(s.t))
|
|
|
|
from wide left join (select id, coalesce(t, '') || '' as t from wide) s using (id);
|
|
|
|
QUERY PLAN
|
|
|
|
----------------------------------------------------------------
|
|
|
|
Finalize Aggregate
|
|
|
|
-> Gather
|
|
|
|
Workers Planned: 2
|
|
|
|
-> Partial Aggregate
|
|
|
|
-> Parallel Hash Left Join
|
|
|
|
Hash Cond: (wide.id = wide_1.id)
|
|
|
|
-> Parallel Seq Scan on wide
|
|
|
|
-> Parallel Hash
|
|
|
|
-> Parallel Seq Scan on wide wide_1
|
|
|
|
(9 rows)
|
|
|
|
|
|
|
|
select length(max(s.t))
|
|
|
|
from wide left join (select id, coalesce(t, '') || '' as t from wide) s using (id);
|
|
|
|
length
|
|
|
|
--------
|
|
|
|
320000
|
|
|
|
(1 row)
|
|
|
|
|
|
|
|
select final > 1 as multibatch
|
|
|
|
from hash_join_batches(
|
|
|
|
$$
|
|
|
|
select length(max(s.t))
|
|
|
|
from wide left join (select id, coalesce(t, '') || '' as t from wide) s using (id);
|
|
|
|
$$);
|
|
|
|
multibatch
|
|
|
|
------------
|
|
|
|
t
|
|
|
|
(1 row)
|
|
|
|
|
2017-11-30 01:06:50 +01:00
|
|
|
rollback to settings;
|
|
|
|
rollback;
|