MySQL优化器和SemiJoin优化
提示：公众号展示代码会自动折行，建议横屏阅读
1. 背景
1.1. MySQL执行流程
MySQL的执行过程包括多个子阶段：语法分析、语义检查、逻辑优化、物理优化和执行。其中逻辑优化和物理优化统称为查询优化。一个查询优化器的输入是查询树，输出是查询执行计划。
逻辑优化也称为基于规则的查询优化（Rule Based Optimization，简称RBO）。主要是对查询进行逻辑上的等价变换，目的是通过这些变换提高查询的性能。 物理优化也称为基于代价的查询优化（Cost-based Optimization，简称CBO）。主要是通过一些模型，预测一个查询使用某种执行计划时的成本，并选择其中成本最小的一个。
1.2. SemiJoin定义
SemiJoin是表上的一个运算。假定R和S是两个表，R SemiJoin S可记为R ⋉ S，结果是在S中满足条件的元组。例如下面的例子是“雇员”和“部门”和它们的半连接的表：
雇员SemiJoin部门 where 雇员.DeptName=部门.DeptName
雇员Join部门 where 雇员.DeptName=部门.DeptName
从上面的结果可以得出如下结论： ①半连接的结果是左表的子集。 ②增加右表一条重复记录，结果并不随之增多。
2. In子查询转SemiJoin的优化方法
2.1 优化方法
通用的转换格式如下：
SELECT ...
FROM ot1 ... otN
WHERE (oe1, ... oeM) IN (SELECT ie1, ..., ieM
FROM it1 ... itK
[WHERE inner-cond])
[AND outer-cond]
[GROUP BY ...] [HAVING ...] [ORDER BY ...]如上格式的子查询被转换为semijoin的格式如下（ot表示外表，it表示内表）：
SELECT ...
FROM (ot1 ... otN) SJ (it1 ... itK)
ON (oe1, ... oeM) = (ie1, ..., ieM)
[AND inner-cond]
[WHERE outer-cond]
[GROUP BY ...] [HAVING ...] [ORDER BY ...]从上面的结构可以看到，转化是支持多列的。
2.2 为什么可以转化：
当执行如下SQL时，也是先遍历雇员表，然后在部门表里查找是否匹配。实际上和上面SemiJoin的语义是一致的，结果也一样。
select * from 雇员 where 雇员.DeptName in (select 部门.DeptName from 部门)
具体Mysql将In转化成SemiJoin的代码逻辑实现在如下函数里：
bool SELECT_LEX::flatten_subqueries()
2.3 一个子查询能够转换成SemiJoin还需满足如下条件：
1. 必须是In或any子查询
2. 不包含UNION 按照扁平化规则后，UNION会导致结果不一致。比如：t1,t2和t3表定义和数据如下：
则执行如下SQL：select a from t1 where a in (select b from t2 UNION select c from t3); 正确结果应该为：
但是SemiJoin转换后的SQL为：select a from t1 semijoin t2 where t1.a=t2.b UNION select a from t1 semijoin t3 where t1.a=t3.c; 由于UNION的去重，导致结果为：
另外即使是UNION ALL，如果t2与t3存在相同记录，结果也是不正确的。
3. 不包含GROUP BY 或 HAVING 含有groupby的SQL，转换后（即扁平化后），导致与原来语义不一致了。比如 原来是：select a from t1 where b in (select b from t2 group by b) 按照规则转换成semijoin后为：select a from t1 semijoin (t2) where t1.b = t2.b group by t2.b; 由于a1不在group by内，语法错误。
4. 不包含聚集函数。通过测试，确实没有走SemiJoin。
mysql> explain select * from t1 where a in (select count(*) from t2 where t1.b=t2.b group by a);
+----+--------------------+-------+------------+------+---------------+------+---------+------+------+----------+------------------------------+
| id | select_type | table | partitions | type | possible_keys | key | key_len | ref | rows | filtered | Extra |
+----+--------------------+-------+------------+------+---------------+------+---------+------+------+----------+------------------------------+
| 1 | PRIMARY | t1 | NULL | ALL | NULL | NULL | NULL | NULL | 8 | 100.00 | Using where |
| 2 | DEPENDENT SUBQUERY | t2 | NULL | ALL | NULL | NULL | NULL | NULL | 4 | 25.00 | Using where; Using temporary |
+----+--------------------+-------+------------+------+---------------+------+---------+------+------+----------+------------------------------+
2 rows in set, 2 warnings (0.00 sec)
mysql> show warnings;
+-------+------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| Level | Code | Message |
+-------+------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| Note | 1276 | Field or reference 'ldy.t1.b' of SELECT #2 was resolved in SELECT #1 |
| Note | 1003 | /* select#1 */ select `ldy`.`t1`.`a` AS `a`,`ldy`.`t1`.`b` AS `b` from `ldy`.`t1` where <in_optimizer>(`ldy`.`t1`.`a`,<exists>(/* select#2 */ select 1 from `ldy`.`t2` where (`ldy`.`t1`.`b` = `ldy`.`t2`.`b`) group by `ldy`.`t2`.`a` having (<cache>(`ldy`.`t1`.`a`) = <ref_null_helper>(count(0))))) |
+-------+------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
2 rows in set (0.00 sec)5. 不包含LIMIT。通过测试，控制台直接报错。
mysql> select * from t1 where a in (select a from t2 limit 2);
ERROR 1235 (42000): This version of MySQL doesn't yet support 'LIMIT & IN/ALL/ANY/SOME subquery'
mysql>6. 子查询有ORDER BY时，转换后去掉ORDER BY
7. 外部查询没有STRAIGHT_JOIN（错） 关键字STRAIGHT_JOIN表明，该查询的JOIN顺序不需要优化，按照原来SQL的顺序依次进行。 MYSQL官方文档说有限制，经测试没有限制。比如，如下依然能转换为semijoin：
mysql> select * from t3;
mysql> explain select t1.*,t2.* from t2 STRAIGHT_JOIN t1 on t1.b=t2.b where t1.a in (select a from t3);
+----+--------------+-------------+------------+--------+---------------+------------+---------+----------+------+----------+----------------------------------------------------+
| id | select_type | table | partitions | type | possible_keys | key | key_len | ref | rows | filtered | Extra |
+----+--------------+-------------+------------+--------+---------------+------------+---------+----------+------+----------+----------------------------------------------------+
| 1 | SIMPLE | t2 | NULL | ALL | NULL | NULL | NULL | NULL | 4 | 100.00 | NULL |
| 1 | SIMPLE | t1 | NULL | ALL | NULL | NULL | NULL | NULL | 8 | 12.50 | Using where; Using join buffer (Block Nested Loop) |
| 1 | SIMPLE | <subquery2> | NULL | eq_ref | <auto_key> | <auto_key> | 5 | ldy.t1.a | 1 | 100.00 | NULL |
| 2 | MATERIALIZED | t3 | NULL | ALL | NULL | NULL | NULL | NULL | 10 | 100.00 | NULL |
+----+--------------+-------------+------------+--------+---------------+------------+---------+----------+------+----------+----------------------------------------------------+
4 rows in set, 1 warning (0.00 sec)
mysql> show warnings;
+-------+------+------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| Level | Code | Message |
+-------+------+------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| Note | 1003 | /* select#1 */ select `ldy`.`t1`.`a` AS `a`,`ldy`.`t1`.`b` AS `b`,`ldy`.`t2`.`a` AS `a`,`ldy`.`t2`.`b` AS `b` from `ldy`.`t2` semi join (`ldy`.`t3`) straight_join `ldy`.`t1` where ((`ldy`.`t1`.`b` = `ldy`.`t2`.`b`) and (`<subquery2>`.`a` = `ldy`.`t1`.`a`)) |
+-------+------+------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
1 row in set (0.00 sec)8. 是否相关子查询均支持。
注：测试所用的数据参考如下 数据库名为ldy，表如下：
mysql> select * from t1;
+------+------+
| a | b |
+------+------+
| 0 | 1 |
| 1 | 2 |
| 2 | 3 |
| 3 | 4 |
| 4 | 5 |
| 5 | 6 |
| 6 | 7 |
| 7 | 8 |
+------+------+
8 rows in set (0.00 sec)
mysql> select * from t2;
+------+------+
| a | b |
+------+------+
| 0 | 1 |
| 1 | 2 |
| 2 | 3 |
| 3 | 4 |
+------+------+
4 rows in set (0.00 sec)
mysql> select * from t3;
+------+------+
| a | b |
+------+------+
| 0 | 1 |
| 2 | 3 |
| 3 | 4 |
+------+------+
3 rows in set (0.00 sec)2.4 举例
例子1: 二重In子查询嵌套SQL通过两次SemiJoin优化成单层SQL，参考如下：
mysql> explain select * from t1 where t1.a in (select a from t2 where b in (select b from t3));
+----+-------------+-------+------------+------+---------------+------+---------+------+------+----------+-------------------------------------------------------------------+
| id | select_type | table | partitions | type | possible_keys | key | key_len | ref | rows | filtered | Extra |
+----+-------------+-------+------------+------+---------------+------+---------+------+------+----------+-------------------------------------------------------------------+
| 1 | SIMPLE | t3 | NULL | ALL | NULL | NULL | NULL | NULL | 3 | 100.00 | Start temporary |
| 1 | SIMPLE | t2 | NULL | ALL | NULL | NULL | NULL | NULL | 4 | 25.00 | Using where; Using join buffer (Block Nested Loop) |
| 1 | SIMPLE | t1 | NULL | ALL | NULL | NULL | NULL | NULL | 8 | 12.50 | Using where; End temporary; Using join buffer (Block Nested Loop) |
+----+-------------+-------+------------+------+---------------+------+---------+------+------+----------+-------------------------------------------------------------------+
3 rows in set, 1 warning (0.00 sec)
mysql> show warnings;
+-------+------+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| Level | Code | Message |
+-------+------+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| Note | 1003 | /* select#1 */ select `ldy`.`t1`.`a` AS `a`,`ldy`.`t1`.`b` AS `b` from `ldy`.`t1` semi join (`ldy`.`t3` join `ldy`.`t2`) where ((`ldy`.`t1`.`a` = `ldy`.`t2`.`a`) and (`ldy`.`t2`.`b` = `ldy`.`t3`.`b`)) |
+-------+------+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
1 row in set (0.00 sec)例子2: 并列in查询，and条件下能够转成SemiJoin：
mysql> explain select * from t1 where t1.a in (select a from t2) and t1.b in (select b from t3);
+----+-------------+-------+------------+------+---------------+------+---------+------+------+----------+--------------------------------------------------------------------+
| id | select_type | table | partitions | type | possible_keys | key | key_len | ref | rows | filtered | Extra |
+----+-------------+-------+------------+------+---------------+------+---------+------+------+----------+--------------------------------------------------------------------+
| 1 | SIMPLE | t3 | NULL | ALL | NULL | NULL | NULL | NULL | 3 | 100.00 | Start temporary |
| 1 | SIMPLE | t1 | NULL | ALL | NULL | NULL | NULL | NULL | 8 | 12.50 | Using where; End temporary; Using join buffer (Block Nested Loop) |
| 1 | SIMPLE | t2 | NULL | ALL | NULL | NULL | NULL | NULL | 4 | 25.00 | Using where; FirstMatch(t1); Using join buffer (Block Nested Loop) |
+----+-------------+-------+------------+------+---------------+------+---------+------+------+----------+--------------------------------------------------------------------+
3 rows in set, 1 warning (0.00 sec)
mysql> show warnings;
+-------+------+-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| Level | Code | Message |
+-------+------+-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| Note | 1003 | /* select#1 */ select `ldy`.`t1`.`a` AS `a`,`ldy`.`t1`.`b` AS `b` from `ldy`.`t1` semi join (`ldy`.`t2`) semi join (`ldy`.`t3`) where ((`ldy`.`t1`.`b` = `ldy`.`t3`.`b`) and (`ldy`.`t2`.`a` = `ldy`.`t1`.`a`)) |
+-------+------+-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
1 row in set (0.00 sec)例子3，并列in查询，or条件下不能转成semijoin：
mysql> explain select * from t1 where t1.a in (select a from t2) or t1.b in (select b from t3);
+----+-------------+-------+------------+------+---------------+------+---------+------+------+----------+-------------+
| id | select_type | table | partitions | type | possible_keys | key | key_len | ref | rows | filtered | Extra |
+----+-------------+-------+------------+------+---------------+------+---------+------+------+----------+-------------+
| 1 | PRIMARY | t1 | NULL | ALL | NULL | NULL | NULL | NULL | 8 | 100.00 | Using where |
| 3 | SUBQUERY | t3 | NULL | ALL | NULL | NULL | NULL | NULL | 3 | 100.00 | NULL |
| 2 | SUBQUERY | t2 | NULL | ALL | NULL | NULL | NULL | NULL | 4 | 100.00 | NULL |
+----+-------------+-------+------------+------+---------------+------+---------+------+------+----------+-------------+
3 rows in set, 1 warning (0.00 sec)
mysql> show warnings;
+-------+------+--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| Level | Code | Message |
+-------+------+--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| Note | 1003 | /* select#1 */ select `ldy`.`t1`.`a` AS `a`,`ldy`.`t1`.`b` AS `b` from `ldy`.`t1` where (<in_optimizer>(`ldy`.`t1`.`a`,`ldy`.`t1`.`a` in ( <materialize> (/* select#2 */ select `ldy`.`t2`.`a` from `ldy`.`t2` where 1 ), <primary_index_lookup>(`ldy`.`t1`.`a` in <temporary table> on <auto_key> where ((`ldy`.`t1`.`a` = `materialized-subquery`.`a`))))) or <in_optimizer>(`ldy`.`t1`.`b`,`ldy`.`t1`.`b` in ( <materialize> (/* select#3 */ select `ldy`.`t3`.`b` from `ldy`.`t3` where 1 ), <primary_index_lookup>(`ldy`.`t1`.`b` in <temporary table> on <auto_key> where ((`ldy`.`t1`.`b` = `materialized-subquery`.`b`)))))) |
+-------+------+--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
1 row in set (0.00 sec)例子4: 支持多列SemiJoin：
mysql> explain select * from t1 where (a, b) in (select a, b from t2);
+----+-------------+-------+------------+------+---------------+------+---------+------+------+----------+-------------------------------------------------------------------+
| id | select_type | table | partitions | type | possible_keys | key | key_len | ref | rows | filtered | Extra |
+----+-------------+-------+------------+------+---------------+------+---------+------+------+----------+-------------------------------------------------------------------+
| 1 | SIMPLE | t2 | NULL | ALL | NULL | NULL | NULL | NULL | 4 | 100.00 | Start temporary |
| 1 | SIMPLE | t1 | NULL | ALL | NULL | NULL | NULL | NULL | 8 | 12.50 | Using where; End temporary; Using join buffer (Block Nested Loop) |
+----+-------------+-------+------------+------+---------------+------+---------+------+------+----------+-------------------------------------------------------------------+
2 rows in set, 1 warning (0.00 sec)
mysql> show warnings;
+-------+------+------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| Level | Code | Message |
+-------+------+------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| Note | 1003 | /* select#1 */ select `ldy`.`t1`.`a` AS `a`,`ldy`.`t1`.`b` AS `b` from `ldy`.`t1` semi join (`ldy`.`t2`) where ((`ldy`.`t1`.`b` = `ldy`.`t2`.`b`) and (`ldy`.`t1`.`a` = `ldy`.`t2`.`a`)) |
+-------+------+------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
1 row in set (0.00 sec)3. SemiJoin的几种执行策略
转换成SemiJoin后，MySQL将要根据权重，采用如下执行策略（以下策略也能通过Mysql的修改系统变量optimizer_switch，将某些策略关闭或打开）。假设有两个表：国家表（Country）和城市表（City），查找存在某一城市人口大于该国人口33%的所有国家。查询语句如下：
select *
from Country
where
Country.code IN (select City.Country
from City
where
City.Population > 0.33 * Country.Population and
City.Population > 1*1000*1000);1. Duplicate Weedout：将semi-join当做普通innerJoin，之后使用临时表将重复项去除。 举例：Country SemiJoin City的执行过程如下：
Country与City内连接后有German有三条记录，通过建立临时表，由于临时表的主键是Country表的RowId，重复项将不能插入。具体可以看到，去重过程如下：
2. FirstMatch：执行join只挑选第一次匹配的项，非常类似于In语义。 举例，Country InnerJoin City的嵌套循环流程如下，先遍历Country再遍历City：
当是SemiJoin时，在City里找到一个，就不需要继续找了。
3. LooseScan：sj-inner表将重复项分组，选出一个即返回。 查询拥有人造卫星的国家，数据库有国家表（Country）和人造卫星表（Satellite）。查询SQL如下：
select * from Country
where
Country.code in (select country_code from Satellite)假如Satellite.country_code是索引，通过按索引顺序遍历，发现已经是分组的了。
利用索引很容易从一个组调到另一组，而一组内只需要判断第一条记录即可，具体如下：
4. Materialize：
假如不是相关子查询，可以先物化内表为一个临时表，由于该临时表在条件字段上采用了索引，保证了唯一性（即消除了重复字段），SemiJoin的结果就和innerJoin相同了。该临时表与外表执行JOIN。
根据临时表与外表Join过程，采用类似NestedLoop形式，根据先遍历临时表还是外表，进一步包括Lookup和Scan两种。Lookup为物化inner表（该表包含一个索引，该索引既可以删除重复元素，也可用于后面执行Join运算），先循环遍历outer表，并在物化表查找。
Scan为物化inner表（该表包含一个索引，该索引既可以删除重复元素，也可用于后面执行Join运算），先循环遍历物化表，并在outer表查找。
4. 参考资料
https://mariadb.com/kb/en/library/optimization-strategies/ https://dev.mysql.com/doc/refman/8.0/en/semi-joins.html
腾讯数据库技术团队对内支持微信红包，彩票、数据银行等集团内部业务，对外为腾讯云提供各种数据库产品，如CDB、CTSDB、CKV、CMongo， 腾讯数据库技术团队专注于增强数据库内核功能，提升数据库性能，保证系统稳定性并解决用户在生产过程中遇到的问题，并对生产环境中遇到的问题及知识进行分享。
var first_sceen__time = (+new Date());if ("" == 1 && document.getElementById('js_content')) { document.getElementById('js_content').addEventListener("selectstart",function(e){ e.preventDefault(); }); } (function(){ if (navigator.userAgent.indexOf("WindowsWechat") != -1){ var link = document.createElement('link'); var head = document.getElementsByTagName('head')[0]; link.rel = 'stylesheet'; link.type = 'text/css'; link.href = "//res.wx.qq.com/mmbizwap/zh_CN/htmledition/style/page/appmsg_new/winwx45ba31.css"; head.appendChild(link); } })();
腾讯数据库技术
赞赏
长按二维码向我转账
受苹果公司新规定影响，微信 iOS 版的赞赏功能被关闭，可通过二维码转账支持公众号。
阅读
分享 在看
已同步到看一看
取消 发送
我知道了
朋友会在“发现-看一看”看到你“在看”的内容
确定
已同步到看一看写下你的想法
最多200字，当前共字 发送
已发送
朋友将在看一看看到
确定
写下你的想法...
取消
发布到看一看
确定
最多200字，当前共字
发送中
微信扫一扫 关注该公众号
微信扫一扫 使用小程序
即将打开""小程序
取消 打开
本文参与 腾讯云自媒体同步曝光计划，分享自微信公众号。
原始发表：2019-04-18，如有侵权请联系 cloudcommunity@tencent.com 删除
评论
登录后参与评论
推荐阅读
目录
腾讯云开发者
Copyright © 2013 - 2026 Tencent Cloud. All Rights Reserved. 腾讯云 版权所有
深圳市腾讯计算机系统有限公司 ICP备案/许可证号：粤B2-20090059 
粤公网安备44030502008569号
腾讯云计算（北京）有限责任公司 京ICP证150476号 | 京ICP备11018762号