2

考虑这张表 ( comments)：

         id | post_id |      text
------------+---------|----------------
      79507 |     12  | Lorem Ipsum
      79544 |     12  | Foo, bar
      79545 |     14  | Interesting...

这个聚合查询：

SELECT comment_id, SUM(vote) AS votes
FROM votes 
GROUP BY comment_id;

 comment_id | votes 
------------+-------
      79507 |    3
      79544 |    4
      79545 |    1

我正在寻找加入comments表和聚合查询，但只对数据的一个非常小的子集（只有一个特定的post_id）感兴趣。这种天真的方法使用子查询来正确返回post_id12 的结果：

SELECT comment_id, votes, text FROM comments c LEFT JOIN
  (SELECT comment_id, SUM(votes) AS vote
   FROM votes 
   GROUP BY comment_id) AS v
ON c.id = v.comment_id 
WHERE c.post_id = 12;

 comment_id | votes |      text
------------+-------|----------------
      79507 |    3  | Lorem Ipsum
      79544 |    4  | Foo, bar

然而，这是非常低效的，因为我们正在计算整个表的内部子查询，但我们只对它的一个非常小的子集感兴趣（votes这个应用程序中的表很大）。

Intuitively, it seems we should be filtering the inner query and there we're missing a WHERE comment_id IN (...) in the subselect. However, we don't know which comment_ids we will need at that stage in the computation. Another subselect inside the subselect could be used to retrieve the appropriate comment_ids, but that seems very clumsy.

I'm inexperienced in SQL and not sure if there exists a cleaner solution. Perhaps the subselect approach is the wrong one altogether.

4

1 回答 1

3

Not sure I understood well, don't you need something like that ?

SELECT c.id as comment_id, SUM (v.vote) as votes, c.text
FROM comments c
LEFT JOIN votes v ON c.id = v.comment_id
WHERE c.post_id = 12
GROUP BY c.id, c.text
于 2013-05-29T13:11:27.983 回答