dense_rank()+hash提示改写优化SQL
数据库环境：SQL SERVER 2005
今天看到一条SQL，返回10条数据，执行了50多S。刚好有空，就对它进行了优化，优化后1S出结果。
先看下原始SQL
SELECT t1.line_no ,
MAX(sat100.confrim_date) confrim_date ,
sam63_lq.company_name ,
sat04.c_code ,
ctlm23.corr_name ,
MAX(sat04.l_date_d) l_date_d ,
SUM(sat05.qty_d_order) qty_d_order ,
sat100.tran_no_row
FROM sat100
INNER JOIN sat101 ON sat100.com_id = sat101.com_id
AND sat100.tran_no = sat101.tran_no
AND sat100.tran_row = sat101.tran_row
LEFT JOIN sat05 ON sat101.com_id = sat05.com_id
AND sat101.p_g_order_no = sat05.p_g_order_no
AND sat101.p_g_order_line = sat05.p_g_order_line
LEFT JOIN sat04 ON sat04.com_id = sat05.com_id
AND sat04.p_g_order_no = sat05.p_g_order_no
LEFT JOIN ctlm22 ON sat05.com_id = ctlm22.com_id
AND sat05.item_no = ctlm22.item_no
LEFT JOIN sam63_lq ON sat100.com_id = sam63_lq.com_id
AND sat100.company_id = sam63_lq.company_id
LEFT JOIN sam60_lq ON sat05.com_id = sam60_lq.com_id
AND sat05.cx_item_no = sam60_lq.cx_item_no
LEFT JOIN sam65_lq ON sat100.car_no = sam65_lq.veh_no
AND sat100.company_id = sam65_lq.company_id
LEFT JOIN ctlm17 ON sat05.com_id = ctlm17.com_id
AND sat05.sa_unit = ctlm17.unit
LEFT JOIN salm02 ON sat04.com_id = salm02.com_id
AND sat04.c_code = salm02.client_id
LEFT JOIN ctlm23 ON sat04.com_id = ctlm23.com_id
AND sat04.c_code = ctlm23.corr_id
LEFT JOIN ctlm08 ON salm02.dept_id = ctlm08.dept_id
LEFT JOIN ( SELECT ROW_NUMBER() OVER ( ORDER BY t2.tran_no_row ) line_no ,
t2.tran_no_row
FROM ( SELECT DISTINCT
sat100.tran_no_row
FROM sat100
INNER JOIN sat101 ON sat100.com_id = sat101.com_id
AND sat100.tran_no = sat101.tran_no
AND sat100.tran_row = sat101.tran_row
LEFT JOIN sat05 ON sat101.com_id = sat05.com_id
AND sat101.p_g_order_no = sat05.p_g_order_no
AND sat101.p_g_order_line = sat05.p_g_order_line
LEFT JOIN sat04 ON sat04.com_id = sat05.com_id
AND sat04.p_g_order_no = sat05.p_g_order_no
LEFT JOIN ctlm22 ON sat05.com_id = ctlm22.com_id
AND sat05.item_no = ctlm22.item_no
LEFT JOIN sam63_lq ON sat100.com_id = sam63_lq.com_id
AND sat100.company_id = sam63_lq.company_id
LEFT JOIN sam60_lq ON sat05.com_id = sam60_lq.com_id
AND sat05.cx_item_no = sam60_lq.cx_item_no
LEFT JOIN ctlm17 ON sat05.com_id = ctlm17.com_id
AND sat05.sa_unit = ctlm17.unit
LEFT JOIN salm02 ON sat04.com_id = salm02.com_id
AND sat04.c_code = salm02.client_id
LEFT JOIN ctlm23 ON sat04.com_id = ctlm23.com_id
AND sat04.c_code = ctlm23.corr_id
LEFT JOIN ctlm08 ON salm02.dept_id = ctlm08.dept_id
WHERE salm02.dept_id LIKE '%'
AND sat100.company_id = ''
AND sat100.corr_id LIKE 'A010131%'
AND sat04.l_date_d >= '2015/06/01 00:00:00'
AND sat04.l_date_d <= '2015/06/30 23:59:59'
) t2
) t1 ON sat100.tran_no_row = t1.tran_no_row
WHERE salm02.dept_id LIKE '%'
AND sat100.company_id = ''
AND sat100.corr_id LIKE 'A010131%'
AND sat04.l_date_d >= '2015/06/01 00:00:00'
AND sat04.l_date_d <= '2015/06/30 23:59:59'
GROUP BY t1.line_no ,
sam63_lq.company_name ,
sat04.c_code ,
ctlm23.corr_name ,
sat100.tran_no_row
ORDER BY t1.line_no ,
sat100.tran_no_row
下面，我来说下我的优化思路：
1.检查SQL的写法是否有问题
先看下子查询部分，发现和外部访问的表及过滤的条件都差不多，用BeyondCompare工具检查外部查询和子查询的差别，相对于外部查询，
子查询少访问了一个表sam65_lq，即少了这部分内容“LEFT JOIN sam65_lq ON sat100.car_no = sam65_lq.veh_no AND sat100.company_id
= sam65_lq.company_id”，恰巧字段veh_no和字段company_id是sam65_lq的联合主键，因此，这部分并没有影响查询的数据。
再看下子查询，它要实现的功能就是根据不重复的tran_no_row生成一个序号，因此，可以用dense_rank()替代子查询实现相同的功能。同时，
sat04有过滤条件，因而可以将left join sat04改成inner join sat04。
改写后的SQL如下：
SELECT line_no ,
MAX(confrim_date) confrim_date ,
company_name ,
c_code ,
corr_name ,
MAX(l_date_d) l_date_d ,
SUM(qty_d_order) qty_d_order ,
tran_no_row
FROM ( SELECT DENSE_RANK() OVER ( ORDER BY sat100.tran_no_row ) AS line_no ,
sat100.confrim_date ,
sam63_lq.company_name ,
sat04.c_code ,
ctlm23.corr_name ,
sat04.l_date_d ,
sat05.qty_d_order ,
sat100.tran_no_row
FROM sat100
INNER JOIN sat101 ON sat100.com_id = sat101.com_id
AND sat100.tran_no = sat101.tran_no
AND sat100.tran_row = sat101.tran_row
LEFT JOIN sat05 ON sat101.com_id = sat05.com_id
AND sat101.p_g_order_no = sat05.p_g_order_no
AND sat101.p_g_order_line = sat05.p_g_order_line
INNER JOIN sat04 ON sat04.com_id = sat05.com_id
AND sat04.p_g_order_no = sat05.p_g_order_no
LEFT JOIN ctlm22 ON sat05.com_id = ctlm22.com_id
AND sat05.item_no = ctlm22.item_no
LEFT JOIN sam63_lq ON sat100.com_id = sam63_lq.com_id
AND sat100.company_id = sam63_lq.company_id
LEFT JOIN sam60_lq ON sat05.com_id = sam60_lq.com_id
AND sat05.cx_item_no = sam60_lq.cx_item_no
LEFT JOIN sam65_lq ON sat100.car_no = sam65_lq.veh_no
AND sat100.company_id = sam65_lq.company_id
LEFT JOIN ctlm17 ON sat05.com_id = ctlm17.com_id
AND sat05.sa_unit = ctlm17.unit
LEFT JOIN salm02 ON sat04.com_id = salm02.com_id
AND sat04.c_code = salm02.client_id
LEFT JOIN ctlm23 ON sat04.com_id = ctlm23.com_id
AND sat04.c_code = ctlm23.corr_id
LEFT JOIN ctlm08 ON salm02.dept_id = ctlm08.dept_id
WHERE salm02.dept_id IS NOT NULL
AND sat100.company_id = ''
AND sat100.corr_id LIKE 'A010131%'
AND sat04.l_date_d >= '2015/06/01 00:00:00'
AND sat04.l_date_d <= '2015/06/30 23:59:59'
) t
GROUP BY line_no ,
company_name ,
c_code ,
corr_name ,
tran_no_row
ORDER BY line_no ,
tran_no_row
执行后，检查数据，发现和原始SQL的数据一致，说明改写的语法没有问题。但是，执行的时间没多大改观。
2.检查是否有缺失索引
看了下执行计划，发现在表sat100和表sat04上都走了索引扫描，在sat100表上没有针对company_id和corr_id建和组合索引，sat04表上也没有
针对l_date_d字段建的索引，而恰巧这些列的选择性都和高，适合建索引。
建了索引后，再执行SQL，发现最快的还是要10S。不行，还得继续优化。
3.检查执行计划，看执行计划是否合理
sat05走了索引查找，它的查询开销达到88%。sat05表作为嵌套循环的内部表，外部表有5W多行记录，sat05表总共有50多W条记录，
且在sat05表上没有任何过滤条件。因此，可以判定，这里不应该走嵌套循环，应该使用哈希连接。所以，直接强制走哈希即可。
改写优化后的完整SQL如下：
SELECT line_no ,
MAX(confrim_date) confrim_date ,
company_name ,
c_code ,
corr_name ,
MAX(l_date_d) l_date_d ,
SUM(qty_d_order) qty_d_order ,
tran_no_row
FROM ( SELECT DENSE_RANK() OVER ( ORDER BY sat100.tran_no_row ) AS line_no ,
sat100.confrim_date ,
sam63_lq.company_name ,
sat04.c_code ,
ctlm23.corr_name ,
sat04.l_date_d ,
sat05.qty_d_order ,
sat100.tran_no_row
FROM sat100
INNER JOIN sat101 ON sat100.com_id = sat101.com_id
AND sat100.tran_no = sat101.tran_no
AND sat100.tran_row = sat101.tran_row
LEFT HASH JOIN sat05 ON sat101.com_id = sat05.com_id
AND sat101.p_g_order_no = sat05.p_g_order_no
AND sat101.p_g_order_line = sat05.p_g_order_line
INNER JOIN sat04 ON sat04.com_id = sat05.com_id
AND sat04.p_g_order_no = sat05.p_g_order_no
LEFT JOIN ctlm22 ON sat05.com_id = ctlm22.com_id
AND sat05.item_no = ctlm22.item_no
LEFT JOIN sam63_lq ON sat100.com_id = sam63_lq.com_id
AND sat100.company_id = sam63_lq.company_id
LEFT JOIN sam60_lq ON sat05.com_id = sam60_lq.com_id
AND sat05.cx_item_no = sam60_lq.cx_item_no
LEFT JOIN sam65_lq ON sat100.car_no = sam65_lq.veh_no
AND sat100.company_id = sam65_lq.company_id
LEFT JOIN ctlm17 ON sat05.com_id = ctlm17.com_id
AND sat05.sa_unit = ctlm17.unit
LEFT JOIN salm02 ON sat04.com_id = salm02.com_id
AND sat04.c_code = salm02.client_id
LEFT JOIN ctlm23 ON sat04.com_id = ctlm23.com_id
AND sat04.c_code = ctlm23.corr_id
LEFT JOIN ctlm08 ON salm02.dept_id = ctlm08.dept_id
WHERE salm02.dept_id IS NOT NULL
AND sat100.company_id = ''
AND sat100.corr_id LIKE 'A010131%'
AND sat04.l_date_d >= '2015/06/01 00:00:00'
AND sat04.l_date_d <= '2015/06/30 23:59:59'
) t
GROUP BY line_no ,
company_name ,
c_code ,
corr_name ,
tran_no_row
ORDER BY line_no ,
tran_no_row
dense_rank()+hash提示改写优化SQL的更多相关文章
- 利用查询提示优化SQL
数据库环境:SQL SERVER 2005 我们来看一下SQL语句及对应的数据量 SELECT COUNT(*) FROM cinvout_02 a WHERE ( a.dept_id IN ( SE ...
- 优化sql，返回行数少情况下，NL比hash快好多
sql如下 select t.id, t.value, tt.sort as sortno from ENGINEERING_TYPE t left join ENGINEERING_TYPE tt ...
- sql语句优化SQL Server
MS SQL Server查询优化方法查询速度慢的原因很多,常见如下几种 1.没有索引或者没有用到索引(这是查询慢最常见的问题,是程序设计的缺陷) 2.I/O吞吐量小,形成了 ...
- 转载 50种方法优化SQL Server数据库查询
原文地址 http://www.cnblogs.com/zhycyq/articles/2636748.html 50种方法优化SQL Server数据库查询 查询速度慢的原因很多,常见如下几种: 1 ...
- [20181220]使用提示OR_EXPAND优化.txt
[20181220]使用提示OR_EXPAND优化.txt --//链接http://www.itpub.net/thread-2107240-2-1.html,http://www.itpub.ne ...
- merge into优化sql（转）
使用Merge INTO优化SQL,性能提升巨大 分类: Oracle 2017-04-13 10:55:07 说说背景:开发有个需求,需要对新加的一个字段根据特定的业务逻辑更新数据.TPS_TR ...
- 如何用 SQL Tuning Advisor (STA) 优化SQL语句
在Oracle10g之前,优化SQL是个比较费力的技术活,不停的分析执行计划,加hint,分析统计信息等等.在10g中,Oracle推出了自己的SQL优化辅助工具: SQL优化器(SQL Tuning ...
- mysql优化SQL语句的一般步骤及常用方法
一.优化SQL语句的一般步骤 1. 通过show status命令了解各种SQL的执行频率 mysqladmin extended-status 或: show [session|global]sta ...
- 利用ordered hints优化SQL
SQL_ID 4g70n3k9bqc5v, child number 0 ------------------------------------- MERGE INTO YJBZH_GRXDFHZ ...
随机推荐
- 数学之美 zt
数学是美丽的,哪里有数哪里就有美. 数学的定义是研究数量关系和空间形式的一门科学.但有句名言说:数学比科学大得多,因为它是科学的语言.数学不仅用来写科学,而且可用来写人生.所以说数学是一切学科的基础, ...
- HDOJ --- 2577
How to Type Time Limit: 2000/1000 MS (Java/Others) Memory Limit: 32768/32768 K (Java/Others) Total S ...
- windows下面配置jdk环境变量
在环境变量中添加如下: Path D:\Program Files\Java\jdk1.6.0_26\binJAVA_HOME D:\Program Files\Java\jdk1.6.0_26CLA ...
- Yii PHP 框架分析（四）
作者:wdy http://hi.baidu.com/delphiss/blog/item/c15b314f05f9dfc0d0c86a26.html Yii应用的入口脚本最后一句启动了WebAppl ...
- LigerUi中的Grid中不分页显示（local）！
LigerUi中的Grid中不分页显示! grid为local usePager: true, //是否分页
- 04、生成 HTMLTestRunner 测试报告
1.HTMLTestRunner 是 Python 标准库的 unittest 模块的一个扩展.它生成易于使用的 HTML 测试报告 1>下载HTMLTestRunner.py文件,地址为: h ...
- leetcode第一刷_Triangle
非常easy的一道DP,看到空间限制是O(N)的,不要习惯性的以为是要保存每一行的最小值,不难想到是要保存一行其中各个数为路径终点时的和的大小.当算到最后一行时,就是从顶部究竟部以这个底部位置为终点的 ...
- HDU 4389——X mod f(x)（数位DP）
X mod f(x) Time Limit: 4000/2000 MS (Java/Others) Memory Limit: 32768/32768 K (Java/Others) Probl ...
- 像C++一样写JavaScript
像C++一样写JavaScript C/C++/Java的include或import可以引用第3方文件和包. 这个功能在Html/Js里没有默认的实现. 假设我们有这样一个HTML文件index.h ...
- web开发性能优化---扩展性能篇
1.实现代码分离 一个成熟的软件开发团队一般都不会全然手写代码.这里讲的代码分离仅仅要是开发中用到的小技巧,通过底层框架+手工代码方式结合实现高速开发和高速扩展. Code目录内文件不同意改动,目录主 ...