Question

我有数百万条记录的大桌子。下面的查询仅执行0.2秒，这很好。

SELECT ch1.*
FROM citizens_history ch1
WHERE ch1.update_id_to = ( 
    SELECT MAX( ch2.update_id_to ) 
    FROM citizens_history ch2
    WHERE ch2.id = ch1.id
)
ORDER BY ch1.experience DESC
LIMIT 100

但是，当我尝试添加INNER JOIN时，就像这样

SELECT ch1.*, upd.*
FROM citizens_history ch1
INNER JOIN updates upd ON upd.id = ch1.update_id_to
WHERE ch1.update_id_to = ( 
    SELECT MAX( ch2.update_id_to ) 
    FROM citizens_history ch2
    WHERE ch2.id = ch1.id
)
ORDER BY ch1.experience DESC
LIMIT 100

执行需要永远。在第一种情况下，我认为MySQL正在这样做：

使用索引经验的订单
适用WHERE
找到100条记录时完成

在第二种情况下，我认为会发生这种情况：

使用索引经验的订单
加入数百万行中的每一条记录＆lt; - slow
适用WHERE
找到100条记录时完成

您有什么建议我如何优化它？

修改：探索快速和慢速查询：

+----+--------------------+-------+-------+---------------+------------+---------+--------------+------+-------------+
| id | select_type        | table | type  | possible_keys | key        | key_len | ref                 | rows | Extra       |
+----+--------------------+-------+-------+---------------+------------+---------+--------------+------+-------------+
|  1 | PRIMARY            | ch1   | index | NULL          | experience | 3       | NULL         |  100 | Using where |
|  2 | DEPENDENT SUBQUERY | ch2   | ref   | id            | id         | 3       | db.ch1.id    |    1 | Using index |
+----+--------------------+-------+-------+---------------+------------+---------+--------------+------+-------------+

+----+--------------------+-------+-------+-------------------------------------+----------+---------+--------------+------+----------------------------------------------+
| id | select_type        | table | type  | possible_keys                       | key      | key_len | ref          | rows | Extra                                        |
+----+--------------------+-------+-------+-------------------------------------+----------+---------+--------------+------+----------------------------------------------+
|  1 | PRIMARY            | upd   | index | PRIMARY                             | datetime | 5       | NULL         |  389 | Using index; Using temporary; Using filesort |
|  1 | PRIMARY            | ch1   | ref   | PRIMARY,update_id_to,update_id_to_2 | PRIMARY  | 4       | db.upd.id    |  112 | Using where                                  |
|  2 | DEPENDENT SUBQUERY | ch2   | ref   | id                                  | id       | 3       | db.ch1.id    |    1 | Using index                                  |
+----+--------------------+-------+-------+-------------------------------------+----------+---------+--------------+------+----------------------------------------------+

Answer 1

重写第一个查询，如下所示，然后根据需要进行修改......

SELECT ch1.*
  FROM citizens_history ch1
  JOIN 
     ( SELECT id
            , MAX(update_id_to) max_update_id_to 
         FROM citizens_history 
        GROUP 
           BY id
     ) ch2
    ON ch2.id = ch1.id
   AND ch2.max_update_id_to = ch1.update_id_to
 ORDER 
    BY ch1.experience DESC
 LIMIT 100

如果您仍在努力（表现性能），请为上述内容提供一个EXPLAIN以及所有相关表格的正确DDLS。

Answer 2

将第一个阶段作为JOIN连接到子查询，然后加入更新表： -

SELECT ch1.*, upd.*
FROM citizens_history ch1
INNER JOIN 
(
    SELECT id, MAX( update_id_to ) AS max_update_id_to
    FROM citizens_history 
    GROUP BY id
) ch2 ON ch1.id = ch2.id AND ch1.update_id_to = ch2.max_update_id_to
INNER JOIN updates upd ON upd.id = ch1.update_id_to
ORDER BY ch1.experience DESC
LIMIT 100

Answer 3

您可以尝试：

  SELECT ch1.*
  FROM citizens_history ch1
  LEFT JOIN citizens_history ch2
    ON ch2.id = ch1.id
    AND ch1.max_update_id_to > ch2.update_id_to
  WHERE ch1.max_update_id_to is null 
  ORDER BY ch1.experience DESC
  LIMIT 100

INNER JOIN导致查询执行时间很长

3 个答案: