真正高效的SQLSERVER分页查询
Sqlserver数据库分页查询一直是Sqlserver的短板,闲来无事,想出几种方法,假设有表ARTICLE,字段ID、YEAR...(其他省略),数据53210条(客户真实数据,量不大),分页查询每页30条,查询第1500页(即第45001-45030条数据),字段ID聚集索引,YEAR无索引,Sqlserver版本:2008R2
第一种方案、最简单、普通的方法:
SELECT TOP 30 * FROM ARTICLE WHERE ID NOT IN(SELECT TOP 45000 ID FROM ARTICLE ORDER BY YEAR DESC, ID DESC) ORDER BY YEAR DESC,ID DESC
平均查询100次所需时间:45s
第二种方案:
SELECT * FROM
(
SELECT TOP 30 * FROM (SELECT TOP 45030 * FROM ARTICLE ORDER BY YEAR DESC, ID DESC) f ORDER BY f.YEAR ASC, f.ID DESC
) s ORDER BY s.YEAR DESC,s.ID DESC
平均查询100次所需时间:138S
第三种方案:
SELECT * FROM ARTICLE w1, ( SELECT TOP 30 ID FROM ( SELECT TOP 50030 ID, YEAR FROM ARTICLE ORDER BY YEAR DESC, ID DESC ) w ORDER BY w.YEAR ASC, w.ID ASC ) w2 WHERE w1.ID = w2.ID ORDER BY w1.YEAR DESC, w1.ID DESC
平均查询100次所需时间:21S
第四种方案:
SELECT * FROM ARTICLE w1 WHERE ID in ( SELECT top 30 ID FROM ( SELECT top 45030 ID, YEAR FROM ARTICLE ORDER BY YEAR DESC, ID DESC ) w ORDER BY w.YEAR ASC, w.ID ASC ) ORDER BY w1.YEAR DESC, w1.ID DESC
平均查询100次所需时间:20S
第五种方案:
SELECT w2.n, w1.* FROM ARTICLE w1,
(
SELECT TOP 50030 row_number() OVER (ORDER BY YEAR DESC, ID DESC) n, ID FROM ARTICLE
) w2 WHERE w1.ID = w2.ID AND w2.n > 50000 ORDER BY w2.n ASC
平均查询100次所需时间:15S
查询第1000-1030条记录
第一种方案:
SELECT TOP 30 * FROM ARTICLE WHERE ID NOT IN(SELECT TOP 1000 ID FROM ARTICLE ORDER BY YEAR DESC, ID DESC) ORDER BY YEAR DESC,ID DESC
平均查询100次所需时间:80s
第二种方案:
SELECT * FROM
(
SELECT TOP 30 * FROM (SELECT TOP 1030 * FROM ARTICLE ORDER BY YEAR DESC, ID DESC) f ORDER BY f.YEAR ASC, f.ID DESC
) s ORDER BY s.YEAR DESC,s.ID DESC
平均查询100次所需时间:30S
第三种方案:
SELECT * FROM ARTICLE w1, ( SELECT TOP 30 ID FROM ( SELECT TOP 1030 ID, YEAR FROM ARTICLE ORDER BY YEAR DESC, ID DESC ) w ORDER BY w.YEAR ASC, w.ID ASC ) w2 WHERE w1.ID = w2.ID ORDER BY w1.YEAR DESC, w1.ID DESC
平均查询100次所需时间:12S
第四种方案:
SELECT * FROM ARTICLE w1 WHERE ID in ( SELECT top 30 ID FROM ( SELECT top 1030 ID, YEAR FROM ARTICLE ORDER BY YEAR DESC, ID DESC ) w ORDER BY w.YEAR ASC, w.ID ASC ) ORDER BY w1.YEAR DESC, w1.ID DESC
平均查询100次所需时间:13S
第五种方案:
SELECT w2.n, w1.* FROM ARTICLE w1,
(
SELECT TOP 1030 row_number() OVER (ORDER BY YEAR DESC, ID DESC) n, ID FROM ARTICLE
) w2 WHERE w1.ID = w2.ID AND w2.n > 1000 ORDER BY w2.n ASC
平均查询100次所需时间:14S
由此可见在查询页数靠前时,效率3>4>5>2>1,页码靠后时5>4>3>1>2,再根据用户习惯,一般用户的检索只看最前面几页,因此选择3 4 5方案均可,若综合考虑方案5是最好的选择,但是要注意SQL2000不支持row_number()函数,由于时间和条件的限制没有做更深入、范围更广的测试,有兴趣的可以仔细研究下。
以下是根据第四种方案编写的一个分页存储过程:
if exists (select * from dbo.sysobjects where id = object_id(N'[dbo].[sys_Page_v2]') and OBJECTPROPERTY(id, N'IsProcedure') = 1) drop procedure [dbo].[sys_Page_v2] GO CREATE PROCEDURE [dbo].[sys_Page_v2] @PCount int output, --总页数输出 @RCount int output, --总记录数输出 @sys_Table nvarchar(100), --查询表名 @sys_Key varchar(50), --主键 @sys_Fields nvarchar(500), --查询字段 @sys_Where nvarchar(3000), --查询条件 @sys_Order nvarchar(100), --排序字段 @sys_Begin int, --开始位置 @sys_PageIndex int, --当前页数 @sys_PageSize int --页大小 AS SET NOCOUNT ON SET ANSI_WARNINGS ON IF @sys_PageSize < 0 OR @sys_PageIndex < 0 BEGIN RETURN END DECLARE @new_where1 NVARCHAR(3000) DECLARE @new_order1 NVARCHAR(100) DECLARE @new_order2 NVARCHAR(100) DECLARE @Sql NVARCHAR(4000) DECLARE @SqlCount NVARCHAR(4000) DECLARE @Top int if(@sys_Begin <=0) set @sys_Begin=0 else set @sys_Begin=@sys_Begin-1 IF ISNULL(@sys_Where,'') = '' SET @new_where1 = ' ' ELSE SET @new_where1 = ' WHERE ' + @sys_Where IF ISNULL(@sys_Order,'') <> '' BEGIN SET @new_order1 = ' ORDER BY ' + Replace(@sys_Order,'desc','') SET @new_order1 = Replace(@new_order1,'asc','desc') SET @new_order2 = ' ORDER BY ' + @sys_Order END ELSE BEGIN SET @new_order1 = ' ORDER BY ID DESC' SET @new_order2 = ' ORDER BY ID ASC' END SET @SqlCount = 'SELECT @RCount=COUNT(1),@PCount=CEILING((COUNT(1)+0.0)/' + CAST(@sys_PageSize AS NVARCHAR)+') FROM ' + @sys_Table + @new_where1 EXEC SP_EXECUTESQL @SqlCount,N'@RCount INT OUTPUT,@PCount INT OUTPUT', @RCount OUTPUT,@PCount OUTPUT IF @sys_PageIndex > CEILING((@RCount+0.0)/@sys_PageSize) --如果输入的当前页数大于实际总页数,则把实际总页数赋值给当前页数 BEGIN SET @sys_PageIndex = CEILING((@RCount+0.0)/@sys_PageSize) END set @sql = 'select '+ @sys_fields +' from ' + @sys_Table + ' w1 ' + ' where '+ @sys_Key +' in (' +'select top '+ ltrim(str(@sys_PageSize)) +' ' + @sys_Key + ' from ' +'(' +'select top ' + ltrim(STR(@sys_PageSize * @sys_PageIndex + @sys_Begin)) + ' ' + @sys_Key + ' FROM ' + @sys_Table + @new_where1 + @new_order2 +') w ' + @new_order1 +') ' + @new_order2 print(@sql) Exec(@sql) GO