SQL Server 不像MySQL那样可以直接简单使用limit 进行分页查询,本文汇总了以下五种分页查询的方法,供参考。假设现有UserInfo表,主键:UserId。设参数:
PageSize = 30
PageNumber = 21

方法一:(最常用的分页代码, top,not in)

    select top 30 UserId from UserInfo where UserId not in (select top 600 UserId from UserInfo order by UserId) order by UserId

备注: 注意前后的order by 一致

方法二:(not exists, not in 的另一种写法而已)

    select top 30 * from UserLog where not exists (select 1 from (select top 600 LogId from UserLog order by LogId) a where a.LogId = UserLog.LogId) order by LogId

备注:EXISTS用于检查子查询是否至少会返回一行数据,该子查询实际上并不返回任何数据,而是返回值True或False。此处的 select 1 from 也可以是select 2 from,select LogId from, select * from 等等,不影响查询。而且select 1 效率最高,不用查字典表。效率值比较:1 > anycol > *

方法三:(top / max, 局限于使用可比较列排序的时候)

    select top 30 * from UserLog where LogId > (select max(LogId) from (select top 600 LogId from UserLog order by LogId) a ) order by LogId

备注:这里max()函数也可以用于文本列,文本列的比较会根据字母顺序排列,数字 < 字母(无视大小写) < 中文字符

方法四:(row_number() over (order by LogId))

    select top 30 * from ( select row_number() over (order by LogId) as rownumber,* from UserLog) a where rownumber > 600 order by LogId

    select * from (select row_number()over(order by LogId) as rownumber,* from UserLog) a
    where rownumber > 600 and rownumber < 630 order by LogId

    select * from (select row_number()over(order by LogId) as rownumber,* from UserLog) a
    where rownumber between 600 and  630 order by LogId

备注: 这里rownumber方法属于排名开窗函数(sum, min, avg等属于聚合开窗函数,ORACLE中叫分析函数)的一种,搭配over关键字使用。

方法五:(offset /fetch next, SQL Server 2012支持)

    select * from UserLog Order by LogId offset 600 rows fetch next 30 rows only

备注: 性能参考文章《SQL Server 2012使用OFFSET/FETCH NEXT分页及性能测试》