看代码:
set statistics io on set statistics time on go set statistics profile on go use pubs select * from authors where (au_fname like 'S%' or au_fname like 'a%') and (state like 'CA' or state like 'UT') and (cast(zip as int) > 90000) select * from authors where (au_fname like 'S%' or au_fname like 'a%') and (cast(zip as int) > 90000) and (state like 'CA' or state like 'UT') select * from authors where (cast(zip as int) > 90000) and (au_fname like 'S%' or au_fname like 'a%') and (state like 'CA' or state like 'UT')
目的:想验证where语句的各个条件的查询顺序
环境:SQLServer 2005Express版本
步骤:显示查询计划
结果:无一例外,都被统一成了这样的代码
|--Clustered Index Scan(OBJECT:([pubs].[dbo].[authors].[UPKCL_auidind]), WHERE:(CONVERT(int,[pubs].[dbo].[authors].[zip],0)>(90000) AND ([pubs].[dbo].[authors].[au_fname] like 'S%' OR [pubs].[dbo].[authors].[au_fname] like 'a%') AND ([pubs].[dbo].[authors].[state] like 'CA' OR [pubs].[dbo].[authors].[state] like 'UT')))
结论:貌似在2005之前的版本中,where子句的顺序是从前往后的。但是又貌似在2005的版本中进行了优化,所有的顺序都被统一成了以过滤能力进行排序的语句。
结论2:根据《T-SQL权威指南》,原本上,where子句的顺序就不应该成为影响结果的因素。所以这样推断,where后面的子句肯定都被DBMS给优化了。