Question

我的数据库架构是这样的：

desc SUB：

Array<$NonMaybeType<T>>

desc Cs：

+-------+-------------+------+-----+---------+----------------+
| Field | Type        | Null | Key | Default | Extra          |
+-------+-------------+------+-----+---------+----------------+
| id    | int(11)     | NO   | PRI | NULL    | auto_increment |
| name  | varchar(30) | NO   |     | NULL    |                |
+-------+-------------+------+-----+---------+----------------+

desc Ap：

+--------------+--------------+------+-----+---------+----------------+
| Field        | Type         | Null | Key | Default | Extra          |
+--------------+--------------+------+-----+---------+----------------+
| id           | int(11)      | NO   | PRI | NULL    | auto_increment |
| other_detail | varchar(255) | NO   |     | NULL    |                |
| created_at   | datetime     | NO   |     | NULL    |                |
| sub_id       | int(11)      | NO   | MUL | NULL    |                |
+--------------+--------------+------+-----+---------+----------------+

desc U：

+--------------+--------------+------+-----+---------+----------------+
| Field        | Type         | Null | Key | Default | Extra          |
+--------------+--------------+------+-----+---------+----------------+
| id           | int(11)      | NO   | PRI | NULL    | auto_increment |
| other_detail | varchar(255) | NO   |     | NULL    |                |
| created_at   | datetime     | NO   |     | NULL    |                |
| sub_id       | int(11)      | NO   | MUL | NULL    |                |
+--------------+--------------+------+-----+---------+----------------+

desc TR：

+------------+-------------+------+-----+---------+-------+
| Field      | Type        | Null | Key | Default | Extra |
+------------+-------------+------+-----+---------+-------+
| id         | int(11)     | NO   | PRI | NULL    |       |
| type       | varchar(30) | NO   |     | NULL    |       |
| created_at | datetime    | NO   |     | NULL    |       |
| sub_id     | int(11)     | NO   | MUL | NULL    |       |
+------------+-------------+------+-----+---------+-------+

desc PR：

+--------------+--------------+------+-----+---------+----------------+
| Field        | Type         | Null | Key | Default | Extra          |
+--------------+--------------+------+-----+---------+----------------+
| id           | int(11)      | NO   | PRI | NULL    | auto_increment |
| other_detail | varchar(255) | NO   |     | NULL    |                |
| created_at   | datetime     | NO   |     | NULL    |                |
| sub_id       | int(11)      | NO   | MUL | NULL    |                |
+--------------+--------------+------+-----+---------+----------------+

desc ID：

+--------------+--------------+------+-----+---------+----------------+
| Field        | Type         | Null | Key | Default | Extra          |
+--------------+--------------+------+-----+---------+----------------+
| id           | int(11)      | NO   | PRI | NULL    | auto_increment |
| other_detail | varchar(255) | NO   |     | NULL    |                |
| created_at   | datetime     | NO   |     | NULL    |                |
| sub_id       | int(11)      | NO   | MUL | NULL    |                |
+--------------+--------------+------+-----+---------+----------------+

我的需求是找到所有+--------------+--------------+------+-----+---------+----------------+ | Field | Type | Null | Key | Default | Extra | +--------------+--------------+------+-----+---------+----------------+ | id | int(11) | NO | PRI | NULL | auto_increment | | other_detail | varchar(255) | NO | | NULL | | | created_at | datetime | NO | | NULL | | | cs_id | int(11) | NO | MUL | NULL | | +--------------+--------------+------+-----+---------+----------------+组合表sub_id（从外键ID获取，Cs有CS_id），sub_id的总数}（PR），sub_id（TR），sub_id（AP）小于N且其sub_id距离一个月内现在和created_at。

我的方法是在u.type = 'fixed_value'的每个表格组中查找计数，在后端找到该组，并查找计数。

sub_id

然后在后端通过SELECT count(*) as action_count, SUB.id FROM Ap INNER JOIN SUB ON Ap.sub_id = SUB.id INNER JOIN U ON U.sub_id = SUB.id WHERE (Ap.created_at >= '2016-07-14') AND (Ap.created_at <= '2016-08-15') AND (U.type = "Customer") AND (U.created_at <= '2016-07-14') GROUP BY SUB.id HAVING action_count > 0 SELECT count(*) as action_count, SUB.id FROM PR INNER JOIN SUB ON PR.sub_id = SUB.id INNER JOIN U ON U.sub_id = SUB.id WHERE (PR.created_at >= '2016-07-14') AND (PR.created_at <= '2016-08-15') AND (U.type = "Customer") AND (U.created_at <= '2016-07-14') GROUP BY SUB.id HAVING action_count > 0 SELECT count(*) as action_count, SUB.id FROM TR INNER JOIN SUB ON TR.sub_id = SUB.id INNER JOIN U ON U.sub_id = SUB.id WHERE (TR.created_at >= '2016-07-14') AND (TR.created_at <= '2016-08-15') AND (U.type = "Customer") AND (U.created_at <= '2016-07-14') GROUP BY SUB.id HAVING action_count > 0 SELECT count(*) as action_count, SUB.id FROM ID INNER JOIN Cs on ID.cs_id = Cs.id inner join SUB ON Cs.sub_id = SUB.id INNER JOIN U ON U.sub_id = SUB.id WHERE (ID.created_at >= '2016-07-14') AND (ID.created_at <= '2016-08-15') AND (U.type = "Customer") AND (U.created_at <= '2016-07-14') GROUP BY SUB.id HAVING action_count > 0再次对获得的结果进行分组，并计算SUB.id的总和，并删除其计数＆lt; action_count的所有SUB.id。 Ñ

如何优化此功能？我们知道，80％的SUB_id属于action_count > N类别，因此在后端获取所有后续过滤是非常糟糕的。

我无法删除所有sub_id＆lt; N在单个查询中，因为在一个查询中它可能有计数＆lt; N，其他＆gt; N，并且休息0.因此，它将被认为具有计数＆lt; N这是错误的。

在这种情况下，联盟所有分组依据DB的计数是否有用？

Select sum(action_count) as count , sub_id from (SELECT count(*) as action_count, SUB.id as sub_id FROM Ap INNER JOIN  SUB ON Ap.sub_id = SUB.id INNER JOIN U  ON U.sub_id = SUB.id WHERE (Ap.created_at >= '2016-07-14') AND (Ap.created_at <= '2016-08-15') AND (U.type = "Customer") AND  (U.created_at <= '2016-07-14') GROUP BY SUB.id HAVING action_count > 0
union all 
SELECT count(*) as action_count, SUB.id  as sub_id  FROM PR INNER JOIN  SUB ON PR.sub_id = SUB.id INNER JOIN U  ON U.sub_id = SUB.id WHERE (PR.created_at >= '2016-07-14') AND (PR.created_at <= '2016-08-15') AND (U.type = "Customer") AND  (U.created_at <= '2016-07-14') GROUP BY SUB.id HAVING action_count > 0
union all 
SELECT count(*) as action_count, SUB.id  as sub_id  FROM TR INNER JOIN  SUB ON TR.sub_id = SUB.id INNER JOIN U  ON U.sub_id = SUB.id WHERE (TR.created_at >= '2016-07-14') AND (TR.created_at <= '2016-08-15') AND (U.type = "Customer") AND  (U.created_at <= '2016-07-14') GROUP BY SUB.id HAVING action_count > 0
union all 
SELECT count(*) as action_count, SUB.id  as sub_id  FROM ID INNER JOIN Cs on ID.cs_id = Cs.id inner join SUB ON Cs.sub_id = SUB.id INNER JOIN U  ON U.sub_id = SUB.id WHERE (ID.created_at >= '2016-07-14') AND (ID.created_at <= '2016-08-15') AND (U.type = "Customer") AND  (U.created_at <= '2016-07-14') GROUP BY SUB.id HAVING action_count > 0) A group by sub_id having count < 30

但是，这也是在最后阶段做过滤。我该如何进一步优化呢？

注意：我已经在https://codereview.stackexchange.com/问了这个问题，但由于我无法得到解决方案，所以也在此重新发布。对不起。

Answer 1

首先，看看是否

SELECT  count(*) as action_count, SUB.id
    FROM  Ap
    INNER JOIN  SUB  ON Ap.sub_id = SUB.id
    INNER JOIN  U  ON U.sub_id = SUB.id
    WHERE  (Ap.created_at >= '2016-07-14')
      AND  (Ap.created_at <= '2016-08-15')
      AND  (U.type = "Customer")
      AND  (U.created_at <= '2016-07-14')
    GROUP BY  SUB.id
    HAVING  action_count > 0

可以这样写，并得到相同的答案：

SELECT  count(*) as action_count, SUB.id
    FROM  SUB
    WHERE  EXISTS(
        SELECT  *
            FROM  Ap
            WHERE  sub_id = SUB.id
              AND  created_at >= '2016-07-14'
              AND  created_at <= '2016-08-15' 
                 )
      AND  EXISTS(
        SELECT  *
            FROM  U
            WHERE  sub_id = SUB.id
              AND  type = "Customer"
              AND  created_at <= '2016-07-14' 
                 )

要获得速度，请添加以下索引：

U: INDEX(sub_id, type, created_at)
AP: INDEX(sub_id, created_at)

请注意，GROUP BY已被删除，速度有所提升。并且HAVING被消除，从而导致较小的结果集。

现在从这样的查询中构建UNION ALL。

具有多个union和group by的MySQL查询

1 个答案: