Question

我想按给定的日期范围计算每个评分组的数量。我写了下面的查询，它很完美：

SELECT c.day, 
(SELECT COUNT(DISTINCT user_id) FROM ratings r WHERE DATE(r.created_at) = c.day AND r.rating = 1 AND r.campaign_id = 2) AS rating1s,
(SELECT COUNT(DISTINCT user_id) FROM ratings r WHERE DATE(r.created_at) = c.day AND r.rating = 2 AND r.campaign_id = 2) AS rating2s,
(SELECT COUNT(DISTINCT user_id) FROM ratings r WHERE DATE(r.created_at) = c.day AND r.rating = 3 AND r.campaign_id = 2) AS rating3s,
(SELECT COUNT(DISTINCT user_id) FROM ratings r WHERE DATE(r.created_at) = c.day AND r.rating = 4 AND r.campaign_id = 2) AS rating4s,
(SELECT COUNT(DISTINCT user_id) FROM ratings r WHERE DATE(r.created_at) = c.day AND r.rating = 5 AND r.campaign_id = 2) AS rating5s
FROM calendar c
WHERE c.day >= '2018-08-01'
GROUP BY c.day
ORDER BY c.day
LIMIT 0, 31

但是这不是一种优化的方法，因为有5个子查询，而我的本地主机上的查询花费了将近2分钟的时间，如何优化此查询？随附示例输出，我需要相同的输出。

Answer 1

您可以将其重新定义为条件聚合：

SELECT DATE(r.created_at),
       COUNT(DISTINCT CASE WHEN r.rating = 1 THEN r.user_id END) as raging_1,
       COUNT(DISTINCT CASE WHEN r.rating = 2 THEN r.user_id END) as raging_2,
       COUNT(DISTINCT CASE WHEN r.rating = 3 THEN r.user_id END) as raging_3,
       COUNT(DISTINCT CASE WHEN r.rating = 4 THEN r.user_id END) as raging_4,
       COUNT(DISTINCT CASE WHEN r.rating = 5 THEN r.user_id END) as raging_5
FROM ratings r
WHERE r.campaign_id = 2 AND
      r.created_at >= '2018-08-01'
GROUP BY DATE(r.created_at);

COUNT(DISTINCT)可能很昂贵。如果可以，将其删除。否则，一次执行DISTINCT可能会更快：

SELECT dte,
       SUM( r.rating = 1 ) as raging_1,
       SUM( r.rating = 2 ) as raging_2,
       SUM( r.rating = 3 ) as raging_3,
       SUM( r.rating = 4 ) as raging_4,
       SUM( r.rating = 5 ) as raging_5
FROM (SELECT DISTINCT user_id, rating, DATE(r.created_at) as dte
      FROM ratings r
      WHERE r.campaign_id = 2 AND
            r.created_at >= '2018-08-01'
     ) urd
GROUP BY dte;

这将返回每天至少具有一个评分的行。如果有些日子全为零，那么您将需要某种外部联接。这几乎没有增加任何性能，因此，如果上述解决方案之一有效，就可以解决。

Answer 2

这是我使用@Gordon的答案进行的查询：

SELECT  DATE(r.created_at),
       COUNT(
           DISTINCT
           CASE 
             WHEN r.rating = 1 
             THEN  user_id
             ELSE 0
           END
       ) as rating1s,
       COUNT(
           DISTINCT
           CASE 
             WHEN r.rating = 2 
             THEN  user_id
             ELSE 0
           END
       ) as rating2s,
       COUNT(
           DISTINCT
           CASE 
             WHEN r.rating = 3 
             THEN  user_id
             ELSE 0
           END
       ) as rating3s,
       COUNT(
           DISTINCT
           CASE 
             WHEN r.rating = 4
             THEN  user_id
             ELSE 0
           END
       ) as rating4s,
       COUNT(
           DISTINCT
           CASE 
             WHEN r.rating = 5 
             THEN  user_id
             ELSE 0
           END
       ) as rating5s

FROM ratings r
WHERE r.campaign_id = 2 AND
      DATE(r.created_at) >= '2018-08-01'
GROUP BY DATE(r.created_at)

这仍然没有优化，但是比我最初的解决方案要好得多。

计算等级数

2 个答案: