将JOIN查询重写为子查询，缺少行？

Posted 2021-04-09

我有以下mysql查询返回这些行：

SELECT p.ID AS id, p.post_title AS title, p.post_author AS uid, SUM(v.pageviews) AS pageviews, IFNULL(c.comment_count, 0) AS comment_count 
FROM `views` v 
LEFT JOIN `posts` p ON v.postid = p.ID 
LEFT JOIN (SELECT comment_post_ID, COUNT(comment_post_ID) AS comment_count FROM `comments` WHERE comment_date_gmt > DATE_SUB('2018-08-16 18:34:46', INTERVAL 29 DAY) AND comment_approved = 1 GROUP BY comment_post_ID) c ON p.ID = c.comment_post_ID 
WHERE 1 = 1 AND p.post_type IN('post') AND v.view_datetime > DATE_SUB('2018-08-16 18:34:46', INTERVAL 29 DAY) AND p.post_password = '' AND p.post_status = 'publish' GROUP BY v.postid ORDER BY pageviews DESC LIMIT 5 OFFSET 0;

我目前正在尝试使用子查询而不是JOIN来重写它，以检查性能是否更好。

以下查询返回相同的结果（comment_count列除外，因为我们只查询两个表而不是三个表）：

SELECT p.ID AS id, p.post_title AS title, p.post_author AS uid, v.pageviews AS pageviews 
FROM `posts` p,
    (SELECT SUM(pageviews) AS pageviews, postid FROM `views` WHERE view_datetime > DATE_SUB( '2018-08-16 18:34:46', INTERVAL 29 DAY) GROUP BY postid) v
WHERE 1 = 1 AND p.ID = v.postid AND p.post_type IN('post') AND p.post_password = '' AND p.post_status = 'publish' GROUP BY v.postid ORDER BY pageviews DESC LIMIT 5 OFFSET 0

但是，如果我将comments表添加到混合中，则会自动排除没有注释的帖子（例如ID为2956且包含3100个视图），与原始查询相比返回不同的结果：

SELECT p.ID AS id, p.post_title AS title, p.post_author AS uid, v.pageviews AS pageviews, IFNULL(c.comment_count, 0) AS comment_count 
FROM `posts` p,
    (SELECT SUM(pageviews) AS pageviews, postid FROM `views` WHERE view_datetime > DATE_SUB( '2018-08-16 18:34:46', INTERVAL 29 DAY) GROUP BY postid) v,
    (SELECT COUNT(comment_post_ID) AS comment_count, comment_post_ID FROM `comments` WHERE comment_date_gmt > DATE_SUB('2018-08-16 18:34:46', INTERVAL 29 DAY) AND comment_approved = 1 GROUP BY comment_post_ID) c 
WHERE 1 = 1 AND p.ID = v.postid AND v.postid = c.comment_post_ID AND p.post_type IN('post') AND p.post_password = '' AND p.post_status = 'publish' GROUP BY v.postid ORDER BY pageviews DESC LIMIT 5 OFFSET 0

我对这些东西并不是很有经验（正如你所知道的那样），所以任何有助于理解正在发生的事情的帮助/提示都会非常感激。

答案

正如我在评论中提到的，您的新查询仍然有JOIN。如果你想使用子查询测试性能，你需要将查询更改为这样的（这是我根据你的原始工作查询的最佳猜测;没有看到表结构和数据，很难100％确定）：

SELECT id, post_title AS title, post_author AS uid, 
    (SELECT SUM(pageviews) 
     FROM views v 
     WHERE v.postid = posts.id AND 
           v.view_datetime > DATE_SUB('2018-08-16 18:34:46', INTERVAL 29 DAY)
    ) AS pageviews,
    (SELECT COUNT(comment_post_ID) 
     FROM comments c 
     WHERE c.comment_post_ID = posts.id AND
           c.comment_date_gmt > DATE_SUB('2018-08-16 18:34:46', INTERVAL 29 DAY) AND 
           c.comment_approved = 1
    ) AS comment_count
FROM posts
WHERE p.post_type IN('post') AND p.post_password = '' AND p.post_status = 'publish'
ORDER BY pageviews DESC 
LIMIT 5 OFFSET 0;