SQL Server 性能优化之——T-SQL NOT IN 和 NOT Exists

这次介绍一下T-SQL中“Not IN” 和“Not Exists”的优化。

Not IN 和 Not Exists 命令 :

有些情况下，需要select/update/delete 操作孤立数据。孤立数据：不存在主表中而存在其关联表中。

操作这样的数据，一般第一反应是利用“Not in” 或 “Not Exists”命令。使用Not IN会严重影响性能，因为这个命令会逐一检查每个记录，就会造成资源紧张，尤其是当对大数据进行更新和删除操作时，可能导致资源被这些操作锁住。

选择 NOT IN 还是 NOT Exists

现在SQL Server 中有两个命令可以使用大数据的插入、更新、删除操作，性能方面比NOT IN有很大的提高，语法简单比NOT Exists好很多，写出来的语句看上去很清爽。现在就请它们闪亮登场，Merge 和 Except。

例子:

首先创建两个表

use [MyTest]
create table Test1 ([id] int, [name] varchar(20))
create table Test2 ([id] int, [name] varchar(20), [address] varchar(100))

declare @RowC int
declare @Name varchar(20)
set @RowC = 0
while @RowC < 400000
Begin
set @Name = 'TestName' + CAST(@RowC as varchar(10))
insert into Test1(id, name) values(@RowC, @Name)
set @RowC = @RowC+1
end

declare @RowC int
declare @Name varchar(20)
declare @Address varchar(100)
set @RowC = 0
while @RowC < 500000
Begin
set @Name = 'TestName' + CAST(@RowC as varchar(10))
set @Address = 'TestAddress' + CAST(@RowC as varchar(10))
insert into Test2([id], [name], [address]) values(@RowC, @Name, @Address)
set @RowC = @RowC+1
end

SELECT [name] FROM Test2 where [name] not in (select [name] from Test1)
UPDATE Test2 SET [name] =N'New_Name' where [name] not in (select [name] from Test1)
DELETE Test2 FROM Test2 where [name] not in (select [name] from Test1)

merge Test2 T using (select name from Test2 except select name from Test1 )S on t.name=s.name
when matched then update SET name=N'New_Name' ;
merge Test2 T using (select name from Test2 except select name from Test1 )S on t.name=s.name
when matched then delete ;
SELECT * FROM Test2 S where not exists (select 1 from Test2 inner join Test1 on Test2.name=Test1.name and Test2.name=s.name)

注意，上面还是有一部分使用了Not Exists：

SELECT name FROM Test2 S where not exists (select 1 from Test2 inner join Test1 on Test2.name=Test1.name and Test2.name=s.name)

现在需要使用简洁的Except：

select name from Test1 except select name from Test2