MySQL性能优化小记：MySQL子查询很慢的问题

今天碰到一个数据库问题，需要用到子查询，但尝试了一下发现它很慢：

mysql> select * from abc_number_prop where number_id in (select number_id from abc_number_phone where phone = '82306839');

为了节省篇幅，省略了输出内容，下同。

67 rows in set (12.00 sec)

只有67行数据返回，却花了12秒，而系统中可能同时会有很多这样的查询，系统肯定扛不住。用desc看一下(注：explain也可)

mysql> desc select * from abc_number_prop where number_id in (select number_id from abc_number_phone where phone = '82306839');

+----+--------------------+------------------+--------+-----------------+-------+---------+------------+---------+--------------------------+

| id | select_type | table | type | possible_keys | key | key_len | ref | rows | Extra |

+----+--------------------+------------------+--------+-----------------+-------+---------+------------+---------+--------------------------+

| 1 | PRIMARY | abc_number_prop | ALL | NULL | NULL | NULL | NULL | 2679838 | Using where |

| 2 | DEPENDENT SUBQUERY | abc_number_phone | eq_ref | phone,number_id | phone | 70 | const,func | 1 | Using where; Using index |

+----+--------------------+------------------+--------+-----------------+-------+---------+------------+---------+--------------------------+

2 rows in set (0.00 sec)

从上面的信息可以看出，在执行此查询时会扫描两百多万行，难道是没有创建索引吗，看一下

mysql> show index from abc_number_phone;

+------------------+------------+-------------+--------------+-----------------+-----------+-------------+----------+--------+------+------------+---------+---------------+

| Table | Non_unique | Key_name | Seq_in_index | Column_name | Collation | Cardinality | Sub_part | Packed | Null | Index_type | Comment | Index_comment |

+------------------+------------+-------------+--------------+-----------------+-----------+-------------+----------+--------+------+------------+---------+---------------+

| abc_number_phone | 0 | PRIMARY | 1 | number_phone_id | A | 36879 | NULL | NULL | | BTREE | | |

| abc_number_phone | 0 | phone | 1 | phone | A | 36879 | NULL | NULL | | BTREE | | |

| abc_number_phone | 0 | phone | 2 | number_id | A | 36879 | NULL | NULL | | BTREE | | |

| abc_number_phone | 1 | number_id | 1 | number_id | A | 36879 | NULL | NULL | | BTREE | | |

| abc_number_phone | 1 | created_by | 1 | created_by | A | 36879 | NULL | NULL | | BTREE | | |

| abc_number_phone | 1 | modified_by | 1 | modified_by | A | 36879 | NULL | NULL | YES | BTREE | | |

+------------------+------------+-------------+--------------+-----------------+-----------+-------------+----------+--------+------+------------+---------+---------------+

6 rows in set (0.06 sec)

mysql> show index from abc_number_prop;

+-----------------+------------+-------------+--------------+----------------+-----------+-------------+----------+--------+------+------------+---------+---------------+

| Table | Non_unique | Key_name | Seq_in_index | Column_name | Collation | Cardinality | Sub_part | Packed | Null | Index_type | Comment | Index_comment |

+-----------------+------------+-------------+--------------+----------------+-----------+-------------+----------+--------+------+------------+---------+---------------+

| abc_number_prop | 0 | PRIMARY | 1 | number_prop_id | A | 311268 | NULL | NULL | | BTREE | | |

| abc_number_prop | 1 | number_id | 1 | number_id | A | 311268 | NULL | NULL | | BTREE | | |

| abc_number_prop | 1 | created_by | 1 | created_by | A | 311268 | NULL | NULL | | BTREE | | |

| abc_number_prop | 1 | modified_by | 1 | modified_by | A | 311268 | NULL | NULL | YES | BTREE | | |

+-----------------+------------+-------------+--------------+----------------+-----------+-------------+----------+--------+------+------------+---------+---------------+

4 rows in set (0.15 sec)

从上面的输出可以看出，这两张表在number_id字段上创建了索引的。

看看子查询本身有没有问题。

mysql> desc select number_id from abc_number_phone where phone = '82306839';

+----+-------------+------------------+------+---------------+-------+---------+-------+------+--------------------------+

| id | select_type | table | type | possible_keys | key | key_len | ref | rows | Extra |

+----+-------------+------------------+------+---------------+-------+---------+-------+------+--------------------------+

| 1 | SIMPLE | abc_number_phone | ref | phone | phone | 66 | const | 6 | Using where; Using index |

+----+-------------+------------------+------+---------------+-------+---------+-------+------+--------------------------+

1 row in set (0.00 sec)

没有问题，只需要扫描几行数据，索引起作用了。查询出来看看

mysql> select number_id from abc_number_phone where phone = '82306839';

+-----------+

| number_id |

+-----------+

| 8585 |

| 10720 |

| 148644 |

| 151307 |

| 170691 |

| 221897 |

+-----------+

6 rows in set (0.00 sec)

直接把子查询得到的数据放到上面的查询中

mysql> select * from abc_number_prop where number_id in (8585, 10720, 148644, 151307, 170691, 221897);

67 rows in set (0.03 sec)

速度也快，看来MySQL在处理子查询的时候是不够好。我在MySQL 5.1.42 和 MySQL 5.5.19 都进行了尝试，都有这个问题。

搜索了一下网络，发现很多人都遇到过这个问题：

参考资料1：使用连接（JOIN）来代替子查询(Sub-Queries) mysql优化系列记录
http://blog.csdn.net/hongsejiaozhu/article/details/1876181
参考资料2：网站开发日记(14)-MYSQL子查询和嵌套查询优化
http://dodomail.iteye.com/blog/250199
根据网上这些资料的建议，改用join来试试。

修改前：select * from abc_number_prop where number_id in (select number_id from abc_number_phone where phone = '82306839');

修改后：select a.* from abc_number_prop a inner join abc_number_phone b on a.number_id = b.number_id where phone = '82306839';

mysql> select a.* from abc_number_prop a inner join abc_number_phone b on a.number_id = b.number_id where phone = '82306839';

67 rows in set (0.00 sec)

效果不错，查询所用时间几乎为0。看一下MySQL是怎么执行这个查询的

mysql> desc select a.* from abc_number_prop a inner join abc_number_phone b on a.number_id = b.number_id where phone = '82306839';

+----+-------------+-------+------+-----------------+-----------+---------+-----------------+------+--------------------------+

| id | select_type | table | type | possible_keys | key | key_len | ref | rows | Extra |

+----+-------------+-------+------+-----------------+-----------+---------+-----------------+------+--------------------------+

| 1 | SIMPLE | b | ref | phone,number_id | phone | 66 | const | 6 | Using where; Using index |

| 1 | SIMPLE | a | ref | number_id | number_id | 4 | eap.b.number_id | 3 | |

+----+-------------+-------+------+-----------------+-----------+---------+-----------------+------+--------------------------+

2 rows in set (0.00 sec)

小结：当子查询速度慢时，可用JOIN来改写一下该查询来进行优化。

网上也有文章说，使用JOIN语句的查询不一定总比使用子查询的语句快。

参考资料3：改变了对Mysql子查询的看法
http://hi.baidu.com/yzx110/blog/item/e694f536f92075360b55a92b.html
本文链接：http://codingstandards.iteye.com/blog/1344833

正好手头有本《高性能MySQL》，翻阅了一下，第4.4节“MySQL查询优化器的限制”4.4.1小节“关联子查询”正好讲到这个问题。

MySQL有时优化子查询很差，特别是在WHERE从句中的IN()子查询。像上面我碰到的情况，其实我的想法是MySQL会把

select * from abc_number_prop where number_id in (select number_id from abc_number_phone where phone = '82306839');

变成下面的样子

select * from abc_number_prop where number_id in (8585, 10720, 148644, 151307, 170691, 221897);

但不幸的是，实际情况正好相反。MySQL试图让它和外面的表产生联系来“帮助”优化查询，它认为下面的exists形式更有效率

select * from abc_number_prop where exists (select * from abc_number_phone where phone = '82306839' and number_id = abc_number_prop.number_id);



mysql> select * from abc_number_prop where exists (select * from abc_number_phone where phone = '82306839' and number_id = abc_number_prop.number_id);

67 rows in set (10.89 sec)

mysql> desc select * from abc_number_prop where exists (select * from abc_number_phone where phone = '82306839' and number_id = abc_number_prop.number_id);

+----+--------------------+------------------+--------+-----------------+-------+---------+-------------------------------------+---------+--------------------------+

| id | select_type | table | type | possible_keys | key | key_len | ref | rows | Extra |

+----+--------------------+------------------+--------+-----------------+-------+---------+-------------------------------------+---------+--------------------------+

| 1 | PRIMARY | abc_number_prop | ALL | NULL | NULL | NULL | NULL | 2660707 | Using where |

| 2 | DEPENDENT SUBQUERY | abc_number_phone | eq_ref | phone,number_id | phone | 70 | const,eap.abc_number_prop.number_id | 1 | Using where; Using index |

+----+--------------------+------------------+--------+-----------------+-------+---------+-------------------------------------+---------+--------------------------+

2 rows in set (0.01 sec)

这种in子查询的形式，在外部表（比如上面的abc_number_prop）数据量较大的时候效率是很差的。（如果对于较小的表，不会造成显著地影响）

文中说到一种优化方式就是，手工将in里面的子查询查询出来，然后再拼装执行 ，这在程序中是可行的。

文中说到：但是总是认为子查询效率很差也是不对的，有时候可能子查询更好些。怎么确定这个事情呢，应该经过评测来决定（执行查询、用desc/explain等来看）。

在网上也能找到《高性能MySQL》的这节内容

参考资料4：MySQL 数据库优化（12）Limitations of the MySQL Query Optimizer
http://www.chenyajun.com/2009/01/03/1657