SQL：选择最新话题和最新帖子，按论坛分组，按最新帖子排序

Question

我正在尝试获取

• 最新线程（id、主题、时间戳、author_id）和
• 最新帖子（id、thread_id、timestamp、author_id）
• 每个论坛（ID、名称）
• 按最新帖子排序，与帖子的创建日期无关。

为什么？

我希望能够显示如下详细信息：

"The latest Answer of forum $forum_id was given on Question $thread_id. Here it is: $post_id"

---

SELECT  f.id AS forum_id,
        f.name AS forum_name,
        t.id AS thread_id,
        t.topic AS thread_topic,
        t.ts AS thread_timestamp,
        p.id AS post_id,
        p.content AS post_content,
        p.ts AS post_timestamp

 FROM   forums f,
        threads t,
        posts p

WHERE   f.id = t.forum_id 
  AND   t.id = p.thread_id

GROUP BY f.id
ORDER BY p.ts

---

任何建议，如何更改 SQL 以获得尽可能高的性能？我试图避免子查询，但我很开放！

提前致谢！

Answer 1

由于 MySQL 不支持窗口函数，我认为没有子查询就没有办法做到这一点：

SELECT  f.id AS forum_id,
    f.name AS forum_name,
    t.id AS thread_id,
    t.topic AS thread_topic,
    t.ts AS thread_timestamp,
    p.id AS post_id,
    p.content AS post_content,
    p.ts AS post_timestamp

FROM   forums f
JOIN (SELECT t2.forum_id, max(p2.ts) as ts
      FROM posts p2
      JOIN threads t2 ON p2.thread_id = t2.id
      GROUP BY t2.forum_id) max_p ON f.id = max_p.forum_id
JOIN   posts p ON max_p.ts = p.ts
JOIN   threads t ON f.id = t.forum_id AND p.thread_id = t.id
ORDER BY p.ts

当然，缓存最新结果可以让您在没有调用 MAX() 的性能损失的情况下执行此操作，但是使用正确的索引，这应该不是什么大问题...

更新

包含没有帖子的主题和没有主题的论坛的最简洁方法是使用 LEFT JOIN 而不是 INNER JOIN：

SELECT  f.id AS forum_id,
    f.name AS forum_name,
    t.id AS thread_id,
    t.topic AS thread_topic,
    t.ts AS thread_timestamp,
    p.id AS post_id,
    p.content AS post_content,
    p.ts AS post_timestamp

FROM   forums f
LEFT JOIN (SELECT t2.forum_id, max(COALESCE(p2.ts, t2.ts)) as ts, COUNT(p2.ts) as post_count
      FROM threads t2 
      LEFT JOIN posts p2 ON p2.thread_id = t2.id
      GROUP BY t2.forum_id) max_p ON f.id = max_p.forum_id
LEFT JOIN   posts p ON max_p.ts = p.ts
LEFT JOIN   threads t ON f.id = t.forum_id AND (max_p.post_count = 0 OR p.thread_id = t.id)
ORDER BY p.ts

Answer 2

我可以想到两种“正确”的方式来做到这一点。第一种是使用连接和子查询：

SELECT  f.id AS forum_id,
        f.name AS forum_name,
        t.id AS thread_id,
        t.topic AS thread_topic,
        t.ts AS thread_timestamp,
        p.id AS post_id,
        p.content AS post_content,
        p.ts AS post_timestamp
 FROM   forums f join
        threads t
        on f.id = t.forum_id join
        posts p
        on t.id = p.thread_id
WHERE   t.ts = (select ts from threads t2 where t2.forum_id = t.forum_id order by ts desc limit 1) and
        p.ts = (select ts from posts p2 where p2.thread_id = p.thread_id order by ts desc limit 1)
GROUP BY f.id
ORDER BY max(p.ts)

这种方法的问题在于它返回了最新的线程和该线程上的最新帖子。解决这个问题很麻烦（这可能是你真正想要的。）

子查询获取threads 和posts 的最新日期。性能取决于您拥有的索引。这可能是可以接受的。这是标准 SQL。

另一个是 substring_index()/group_concat() 的技巧，这是 MySQL 特有的：

SELECT  f.id AS forum_id,
        f.name AS forum_name,
        substring_index(group_concat(t.id order by t.ts desc separator '|'), '|', 1) AS thread_id,
        substring_index(group_concat(t.topic order by t.ts desc separator '|'), '|', 1)  AS thread_topic,
        substring_index(group_concat(t.ts order by p.ts desc separator '|'), '|', 1)  AS thread_timestamp,
        substring_index(group_concat(p.id order by p.ts desc separator '|'), '|', 1)  AS post_id,
        substring_index(group_concat(p.content order by p.ts desc separator '|'), '|', 1)  AS post_content,
        substring_index(group_concat(p.ts order by p.ts desc separator '|'), '|', 1)  AS post_timestamp
 FROM   forums f join
        threads t
        on f.id = t.forum_id join
        posts p
        on t.id = p.thread_id
GROUP BY f.id
ORDER BY max(p.ts);

此版本的性能可能会更好（因为您已经承担了group by 的开销）。必须选择分隔符，因此它不在任何值中。否则，只会出现分隔符之前的部分。

一个优点是线程和帖子是独立处理的，因此您可以获得最新的线程，并分别获得最新的帖子。您可以通过更改group_concat() 中的order by 条件来获取给定线程上的最新帖子。

此外，要获得您想要的订购，您需要通过max(p.ts) 订购，而不仅仅是p.ts。后者将通过论坛上的任意时间戳进行排序；不能保证它是最新的。