MySQL派生表(derived)优化一例
1、什么是派生表derived
关键字：子查询–>在From后where前的子查询
mysql> explain select * from (select * from t) a where id=2;
+----+-------------+------------+------+---------------+-------------+---------+-------+------+-------+
| id | select_type | table | type | possible_keys | key | key_len | ref | rows | Extra |
+----+-------------+------------+------+---------------+-------------+---------+-------+------+-------+
| 1 | PRIMARY | <derived2> | ref | <auto_key0> | <auto_key0> | 4 | const | 0 | NULL |
| 2 | DERIVED | t | ALL | NULL | NULL | NULL | NULL | 4 | NULL |
+----+-------------+------------+------+---------------+-------------+---------+-------+------+-------+
MySQL 5.7开始优化器引入derived_merge，可以理解为Oracle的子查询展开，有优化器参数optimizer_switch='derived_merge=ON’来控制，默认为打开。
但是仍然有很多限制，当派生子查询存在以下操作时该特性无法生效：UNION 、GROUP BY、DISTINCT、LIMIT/OFFSET以及聚合操作
2、派生表SQL优化一例
mysql> select @@version;
+------------+
| @@version |
+------------+
| 5.6.16-log |
+------------+ SELECT A.*, b.vendor_name, c.chinese_name, c.english_name, d.name AS category_name, e.value, a.season_code
FROM PR_XXX A
LEFT JOIN VXX B
ON A.VENDOR_ID = B.VENDOR_ID
LEFT JOIN BXX C
ON A.BRAND_ID = C.BRAND_ID
LEFT JOIN CXX D
ON A.CATEGORY_ID = D.CATEGORY_ID
LEFT JOIN PXX E
ON A.PRODUCT_ID = E.PRODUCT_ID
INNER JOIN (SELECT SS.PRODUCT_ID
FROM SKU_XXX SS
WHERE SS.ENABLED = 1
GROUP BY SS.PRODUCT_ID
HAVING SUM(SS.STORE) >= 1) ST
ON ST.PRODUCT_ID = A.PRODUCT_ID
WHERE A.vendor_id = 91011
AND A.enabled = 1
AND A.status = ''
AND a.season_code = ''
AND 1 = 1
AND a.brand_id = ''
AND E.KEY_NAME = 'XXX'
ORDER BY product_id LIMIT 0, 10
执行计划如下
++----+-------------+------------+-------------+----------------------------------------------------------------+-----------------------------------+---------+--------------------------+---------+---------------------------------------------------------------------------------+----+--
| id | select_type | table | type | possible_keys | key | key_len | ref | rows | Extra |
+----+-------------+------------+-------------+----------------------------------------------------------------+-----------------------------------+---------+--------------------------+---------+---------------------------------------------------------------------------------+
| 1 | PRIMARY | A | index_merge | PRIMARY,vendor_id,IDX_VENDOR_ID_STATUS,IDX_BRAND_ID,IDX_STATUS | IDX_VENDOR_ID_STATUS,IDX_BRAND_ID | 9,8 | NULL | 5 | Using intersect(IDX_VENDOR_ID_STATUS,IDX_BRAND_ID); Using where; Using filesort |
| 1 | PRIMARY | B | const | PRIMARY | PRIMARY | 8 | const | 1 | NULL |
| 1 | PRIMARY | C | const | PRIMARY | PRIMARY | 8 | const | 1 | Using where |
| 1 | PRIMARY | D | eq_ref | PRIMARY | PRIMARY | 8 | ger-prd-db.A.category_id | 1 | NULL |
| 1 | PRIMARY | E | ref | IDX_PRODUCT_PROPERTY_PRODUCT_ID | IDX_PRODUCT_PROPERTY_PRODUCT_ID | 8 | ger-prd-db.A.product_id | 3 | Using where |
| 1 | PRIMARY | <derived2> | ref | <auto_key0> | <auto_key0> | 9 | ger-prd-db.A.product_id | 10 | Using index |
| 2 | DERIVED | SS | index | IDX_PRODUCT_ID | IDX_PRODUCT_ID | 9 | NULL | 1715829 | Using where |
+----+-------------+------------+-------------+----------------------------------------------------------------+-----------------------------------+---------+--------------------------+---------+---------------------------------------------------------------------------------+
关键表索引信息
mysql> show index from PR_XXX ;
+---------+------------+-------------------------+--------------+--------------+-----------+-------------+----------+--------+------+------------+---------+---------------+
| Table | Non_unique | Key_name | Seq_in_index | Column_name | Collation | Cardinality | Sub_part | Packed | Null | Index_type | Comment | Index_comment |
+---------+------------+-------------------------+--------------+--------------+-----------+-------------+----------+--------+------+------------+---------+---------------+
| PR_XXX | 0 | PRIMARY | 1 | product_id | A | 283783 | NULL | NULL | | BTREE | | |
| PR_XXX | 0 | vendor_id | 1 | vendor_id | A | 92 | NULL | NULL | | BTREE | | |
| PR_XXX | 0 | vendor_id | 2 | product_code | A | 283783 | NULL | NULL | | BTREE | | |
| PR_XXX | 1 | IDX_VENDOR_ID_STATUS | 1 | vendor_id | A | 82 | NULL | NULL | | BTREE | | |
| PR_XXX | 1 | IDX_VENDOR_ID_STATUS | 2 | status | A | 392 | NULL | NULL | | BTREE | | |
| PR_XXX | 1 | IDX_BRAND_ID | 1 | brand_id | A | 2866 | NULL | NULL | | BTREE | | |
| PR_XXX | 1 | IDX_STATUS | 1 | status | A | 10 | NULL | NULL | | BTREE | | |
| PR_XXX | 1 | IDX_PRODUCT_CATEGORY_ID | 1 | category_id | A | 276 | NULL | NULL | | BTREE | | |
| PR_XXX | 1 | IDX_SPU_ID | 1 | spu_id | A | 283783 | NULL | NULL | YES | BTREE | | |
+---------+------------+-------------------------+--------------+--------------+-----------+-------------+----------+--------+------+------------+---------+---------------+
9 rows in set (0.00 sec)
mysql> show index from SKU_XXX;
+-----------+------------+----------------+--------------+--------------+-----------+-------------+----------+--------+------+------------+---------+---------------+
| Table | Non_unique | Key_name | Seq_in_index | Column_name | Collation | Cardinality | Sub_part | Packed | Null | Index_type | Comment | Index_comment |
+-----------+------------+----------------+--------------+--------------+-----------+-------------+----------+--------+------+------------+---------+---------------+
| SKU_XXX | 0 | PRIMARY | 1 | sku_store_id | A | 1715829 | NULL | NULL | | BTREE | | |
| SKU_XXX | 1 | IDX_SKU_ID | 1 | sku_id | A | 1715829 | NULL | NULL | YES | BTREE | | |
| SKU_XXX | 1 | IDX_PRODUCT_ID | 1 | product_id | A | 857914 | NULL | NULL | YES | BTREE | | |
+-----------+------------+----------------+--------------+--------------+-----------+-------------+----------+--------+------+------------+---------+---------------+
建议进行SQL改写
SELECT A.*, b.vendor_name, c.chinese_name, c.english_name, d.name as category_name, e.value, a.season_code
FROM PR_XXX A
LEFT JOIN VXX B
ON A.VENDOR_ID = B.VENDOR_ID
LEFT JOIN BXX C
ON A.BRAND_ID = C.BRAND_ID
LEFT JOIN CXX D
ON A.CATEGORY_ID = D.CATEGORY_ID
LEFT JOIN PXX E
ON A.PRODUCT_ID = E.PRODUCT_ID
INNER JOIN SKU_XXX SS
ON SS.ENABLED = 1
and SS.PRODUCT_ID = A.PRODUCT_ID
where A.vendor_id = 20
AND A.enabled = 1
AND A.status = ''
AND a.season_code = '18SS'
AND 1 = 1
AND a.brand_id = ''
AND E.KEY_NAME = 'BrandID'
and SS.ENABLED = 1
GROUP BY a.PRODUCT_ID
having SUM(SS.STORE) >= 1
ORDER BY a.product_id limit 0, 10;
改写后的执行计划
+----+-------------+-------+-------------+----------------------------------------+-----------------------------------+---------+--------------------------+------+--------------------------------------------------------------------------------------------------+
| id | select_type | table | type | possible_keys | key | key_len | ref | rows | Extra |
+----+-------------+-------+-------------+----------------------------------------+-----------------------------------+---------+--------------------------+------+--------------------------------------------------------------------------------------------------+
| 1 | SIMPLE | A | index_merge | PRIMARY,vendor_id,....................,| IDX_VENDOR_ID_STATUS,IDX_BRAND_ID | 9,8 | NULL | 5 | Using intersect(IDX_VENDOR_ID_STATUS,IDX_BRAND_ID); Using where; Using temporary; Using filesort |
| 1 | SIMPLE | B | const | PRIMARY | PRIMARY | 8 | const | 1 | NULL |
| 1 | SIMPLE | C | const | PRIMARY | PRIMARY | 8 | const | 1 | Using where |
| 1 | SIMPLE | D | eq_ref | PRIMARY | PRIMARY | 8 | ger-prd-db.A.category_id | 1 | NULL |
| 1 | SIMPLE | SS | ref | IDX_PRODUCT_ID | IDX_PRODUCT_ID | 9 | ger-prd-db.A.product_id | 2 | Using where |
| 1 | SIMPLE | E | ref | IDX_PRODUCT_PROPERTY_PRODUCT_ID | IDX_PRODUCT_PROPERTY_PRODUCT_ID | 8 | ger-prd-db.A.product_id | 3 | Using where |
+----+-------------+-------+-------------+----------------------------------------+-----------------------------------+---------+--------------------------+------+--------------------------------------------------------------------------------------------------+
优化前：2 rows in set (2.12 sec)
优化后：2 rows in set (0.00 sec)
MySQL派生表(derived)优化一例的更多相关文章
- MySQL 派生表（Derived Table） Merge Optimization
本文将通过演示告诉你:MySQL中派生表(Derived Table)是什么?以及MySQL对它的优化. Background 有如下一张表: mysql> desc city; +------ ...
- 数据库~Mysql派生表注意的几点~关于百万数据的慢查询问题
基础概念 派生表是从SELECT语句返回的虚拟表.派生表类似于临时表,但是在SELECT语句中使用派生表比临时表简单得多,因为它不需要创建临时表的步骤. 术语:*派生表*和子查询通常可互换使用.当SE ...
- MySQL的表的优化和列类型的选择
列选择原则: 1:字段类型优先级 整型 > date,time > enum,char>varchar > blob 列的特点分析: 整型: 定长,没有国家/地区之分,没有 ...
- mysql服务器iowait高优化一例完整深入解析
我们有一服务器,上面运行着两个mysql实例,这几天iowait一直很高,在20-30%,下午特地专门排查和解决了下,相关过程整理如下. 该服务器有两个挂载盘,服务器在阿里云上,一个系统盘,一个数据盘 ...
- mysql大表如何优化
作者:哈哈链接:http://www.zhihu.com/question/19719997/answer/81930332来源:知乎著作权归作者所有.商业转载请联系作者获得授权,非商业转载请注明出处 ...
- Mysql学习总结（17）——MySQL数据库表设计优化
1.选择优化的数据类型 MySQL支持很多种不同的数据类型,并且选择正确的数据类型对于获得高性能至关重要.不管选择何种类型,下面的简单原则都会有助于做出更好的选择: (1).更小通常更好 一般来说,要 ...
- mysql 两表索引优化
建表语句 CREATE TABLE IF NOT EXISTS `class`( `id` INT(10) UNSIGNED NOT NULL AUTO_INCREMENT, `card` INT(1 ...
- mysql 单表索引优化
建表语句 CREATE TABLE IF NOT EXISTS `article` ( `id` INT(10) UNSIGNED NOT NULL PRIMARY KEY AUTO_INCREMEN ...
- mysql 修复表和优化表
REPAIR TABLE `table_name` 修复表 OPTIMIZE TABLE `table_name` 优化表
随机推荐
- Kubernetes基本概念之Name和NameSpace
在Kubernetes中,所有对象都会被指定一个唯一的Name和UID. 用户还可以指定一些不要求唯一性的数据附加到对象上,例如Label和Annotation. 1. Name Name是创建一个K ...
- Eclipse中新建Maven Web项目报错：The superclass "javax.servlet.http.HttpServlet" was not found on the Java Build Path
在maven web项目中的index.jsp中的错误信息如下: The superclass "javax.servlet.http.HttpServlet" was not f ...
- springboot 简单自定义starter - dubbo
首先需要引入pom 这里使用nacos注册中心 所以引入了nacos-client 使用zookeeper注册中心的话需要引入其相应的client <dependency> <gro ...
- 调试JMETER脚本的5种方法
如果你曾经设计过JMeter脚本,我敢打赌你至少有一次弄清楚Json Extractor无法正常工作的原因.你猜怎么着?我去过那儿! 你知道为什么最好的JMeter Performance Engin ...
- js千分位处理
一.去掉千分位 function removeThousands(num) { var x = num.split(','); return parseFloat(x.join("" ...
- Node.js实现TCP和HTTP并作简单的比较
TCP和Node 传输控制协议是一个面向连接的协议,换句话说,它是一个传输层的协议,它主要的职务呢,就是确保信息传输的正确性. 我们使用的很多如HTTP协议都是基于TCP的,为什么呢?因为我们不希望传 ...
- rgb值转换成16进制
由于jQuery获取css中的background有时候是rgb值,所以往往需要一个转换函数. 以前觉得难,还写个博客记录,现在觉得好容易. let testColor = "rgb(20, ...
- little w and Soda（思维题）
链接:https://ac.nowcoder.com/acm/contest/297/A 来源:牛客网 时间限制:C/C++ 1秒,其他语言2秒 空间限制:C/C++ 262144K,其他语言5242 ...
- Android Studio、adb
Android Studio:一个超级全的教程,mark~ https://www.cnblogs.com/laughingQing/p/5848425.html adb命令: https://blo ...
- eclipse ctrl+h
之前我的eclipse使用ctrl + h 出现的总是Plug-in Search ,总是需要点那个小三角才能找到File Search ,深感使用不便, 修改方法:Window->Genera ...