MariaDB社区

标题: 一条经典的SQL语句优化实例 [打印本页]

作者: kider 时间: 2007-9-3 15:38
标题: 一条经典的SQL语句优化实例

1、概述

如下SQL语句发生严重消耗资源的问题，使得OS's load average会在30以上，一条语句需要执行上百秒。

/*
PIXPatient 184176条
DomainPatient 184189条
PersonName 184189条
*/

捕获的SQL语句：

select * from PIXPatient where PIXPatientTID
in (select distinct PIXPatientTID from DomainPatient where DomainPatientTID
in ( select DomainPatientTID from DomainPatient
where PatientBirthday = '1994-01-09' or PatientBirthday = '1994-01-01'
union select DomainPatientTID from PersonName where FamilyName = '倪' or GivenName = '界'));

2、优化

a.优化前执行效率：
mysql> select * from PIXPatient where PIXPatientTID
in (select distinct PIXPatientTID from DomainPatient where DomainPatientTID
in ( select DomainPatientTID from DomainPatient
  where PatientBirthday = '1994-01-09' or PatientBirthday = '1994-01-01'
  union select  DomainPatientTID from PersonName where FamilyName = '倪' or GivenName = '界'));
+---------------+--------------+------------------+
| PIXPatientTID | PIXPatientID | PIXPatientStatus |
+---------------+--------------+------------------+
|       17013 | 17013       | active          |
|       35491 | 35491       | active          |
|       39654 | 39654       | active          |
|       47351 | 47351       | active          |
|       77820 | 77820       | active          |
|       111464 | 111464    | active          |
|       151700 | 151700    | active          |
|       168993 | 168993    | active          |
+---------------+--------------+------------------+
8 rows in set (1 min 58.11 sec)
8 rows in set (23.80 sec) ---- 空闲时

b.加索引
alter table PersonName add index Index_FamilyName (FamilyName), add index Index_GivenName (GivenName);
alter table DomainPatient add index Index_PatientBirthday (PatientBirthday);

效果不明显

c.重构SQL语句（优化）
mysql> select * from PIXPatient inner join (
-> select distinct PIXPatientTID from DomainPatient inner join (
-> select DomainPatientTID from DomainPatient  where PatientBirthday = '1994-01-09'
-> union select DomainPatientTID from DomainPatient  where PatientBirthday = '1994-01-01'
-> union select  DomainPatientTID from PersonName where FamilyName = '倪'
-> union select  DomainPatientTID from PersonName where GivenName = '界' ) a using(DomainPatientTID) ) b using(PIXPatientTID) ;
+---------------+--------------+------------------+
| PIXPatientTID | PIXPatientID | PIXPatientStatus |
+---------------+--------------+------------------+
|       77820 | 77820       | active          |
|       168993 | 168993    | active          |
|       17013 | 17013       | active          |
|       35491 | 35491       | active          |
|       39654 | 39654       | active          |
|       47351 | 47351       | active          |
|       111464 | 111464    | active          |
|       151700 | 151700    | active          |
+---------------+--------------+------------------+
8 rows in set (1.11 sec)
8 rows in set (0.02 sec) ---- 空闲时

效果明显

3、结论

SQL语句中，尽量避免使用or,in关键字，因为执行效率低。

规律：

join > exists > in

union > or

作者: mysqlkumao 时间: 2008-4-9 19:29
高啊

作者: eddy0909 时间: 2010-5-28 10:47
学习了，谢谢！

作者: sissiyam 时间: 2011-4-13 15:45
经典

作者: cndemo 时间: 2011-11-24 15:19
学习了：一条经典的SQL语句优化实例

作者: beingman 时间: 2012-6-18 14:39
优化的效果很明显，但是结论没有切中根本原因，应该是当前版本的MySQL优化器处理模式的问题。

欢迎光临 MariaDB社区 (http://123.56.88.72/)