Question

我能做些什么来加速这个查询（使用Postgres 9.3.4）。我正在寻找最大的版本，它是有向图的一部分（在这种情况下版本＆lt; = 300）。

WITH RECURSIVE r(depth, revision_id) AS (
  SELECT count(1) AS depth, r.revision_id AS revision_id 
  FROM revision AS r 
  WHERE r.revision_id = 300 
  GROUP BY r.revision_id 
  UNION ALL 
  SELECT rev1.depth + 1, rev2.parent_revision_id AS parent_revision_id 
  FROM rev1, revision AS rev2 
  WHERE rev2.revision_id != 1 AND rev2.revision_id = rev1.revision_id
)
SELECT t1.id AS id, t1.value as value
FROM (
  SELECT DISTINCT ON (t2.id) t2.id AS id, t2.rev_id AS rev_id 
  FROM product AS t2 
  WHERE t2.rev_id = (
    SELECT max(rev_id)
    FROM product
    WHERE t2.id = product.id 
    AND EXISTS (
      SELECT 1 FROM r 
      WHERE r.revision_id = product.rev_id
    )
  ) AND t2.deleted = false 
) AS t1

我尝试选择select rev_id from product INTERSECT select revision_id from r的最大交点，但这比上面的exists版本略慢。此外，我存储CTE的临时表没有带来任何改进。

上述查询包含以下记录：

revision
--------
revision_id, parent_revision_id
1               null
2               1
3               2
4               2
5               3
6               5

product
-------
id, rev_id, value
1    2       a
2    3       b
3    4       c
1    5       a2
1    6       a3
...
    300      a256

结果

id, value
1   a256
2   b

（注意product.id=3不是结果集的一部分，因为它不是版本300的后代）

如何改进CTE查询？

0 个答案: