本文的三种分页方案来自于:
http://blog.csdn.net/lihonggen0/archive/2004/09/14/103511.aspx
只是做了更大数据量、不同位置页的对比。
建立表:
CREATE
TABLE
[
TestTable
]
(
[ ID ] [ int ] IDENTITY ( 1 , 1 ) NOT NULL ,
[ FirstName ] [ nvarchar ] ( 100 ) COLLATE Chinese_PRC_CI_AS NULL ,
[ LastName ] [ nvarchar ] ( 100 ) COLLATE Chinese_PRC_CI_AS NULL ,
[ Country ] [ nvarchar ] ( 50 ) COLLATE Chinese_PRC_CI_AS NULL ,
[ Note ] [ nvarchar ] ( 2000 ) COLLATE Chinese_PRC_CI_AS NULL
) ON [ PRIMARY ]
GO
[ ID ] [ int ] IDENTITY ( 1 , 1 ) NOT NULL ,
[ FirstName ] [ nvarchar ] ( 100 ) COLLATE Chinese_PRC_CI_AS NULL ,
[ LastName ] [ nvarchar ] ( 100 ) COLLATE Chinese_PRC_CI_AS NULL ,
[ Country ] [ nvarchar ] ( 50 ) COLLATE Chinese_PRC_CI_AS NULL ,
[ Note ] [ nvarchar ] ( 2000 ) COLLATE Chinese_PRC_CI_AS NULL
) ON [ PRIMARY ]
GO
插入数据:(100万条)
SET
IDENTITY_INSERT
TestTable
ON
declare @i int
set @i = 1
while @i <= 1000000
begin
insert into TestTable( [ id ] , FirstName, LastName, Country,Note) values (@i, ' FirstName_XXX ' , ' LastName_XXX ' , ' Country_XXX ' , ' Note_XXX ' )
set @i = @i + 1
end
SET IDENTITY_INSERT TestTable OFF
declare @i int
set @i = 1
while @i <= 1000000
begin
insert into TestTable( [ id ] , FirstName, LastName, Country,Note) values (@i, ' FirstName_XXX ' , ' LastName_XXX ' , ' Country_XXX ' , ' Note_XXX ' )
set @i = @i + 1
end
SET IDENTITY_INSERT TestTable OFF
------------------------------------
分页方案一:(利用Not In和SELECT TOP分页)
语句形式:
SELECT
TOP
页大小
*
FROM TestTable
WHERE (ID NOT IN
( SELECT TOP 页大小 * 页数 id
FROM 表
ORDER BY id))
ORDER BY ID
FROM TestTable
WHERE (ID NOT IN
( SELECT TOP 页大小 * 页数 id
FROM 表
ORDER BY id))
ORDER BY ID
-------------------------------------
分页方案二:(利用ID大于多少和SELECT TOP分页)
SELECT
TOP
页大小
*
FROM TestTable
WHERE (ID >
( SELECT MAX (id)
FROM ( SELECT TOP 页大小 * 页数 id
FROM 表
ORDER BY id) AS T))
ORDER BY ID
FROM TestTable
WHERE (ID >
( SELECT MAX (id)
FROM ( SELECT TOP 页大小 * 页数 id
FROM 表
ORDER BY id) AS T))
ORDER BY ID
分页方案三:(利用SQL的游标存储过程分页)
create
procedure
XiaoZhengGe
@sqlstr nvarchar ( 4000 ), -- 查询字符串
@currentpage int , -- 第N页
@pagesize int -- 每页行数
as
set nocount on
declare @P1 int , -- P1是游标的id
@rowcount int
exec sp_cursoropen @P1 output,@sqlstr,@scrollopt = 1 ,@ccopt = 1 ,@rowcount = @rowcount output
select ceiling ( 1.0 * @rowcount / @pagesize) as 总页数 -- ,@rowcount as 总行数,@currentpage as 当前页
set @currentpage = (@currentpage - 1 ) * @pagesize + 1
exec sp_cursorfetch @P1, 16 ,@currentpage,@pagesize
exec sp_cursorclose @P1
set nocount off
@sqlstr nvarchar ( 4000 ), -- 查询字符串
@currentpage int , -- 第N页
@pagesize int -- 每页行数
as
set nocount on
declare @P1 int , -- P1是游标的id
@rowcount int
exec sp_cursoropen @P1 output,@sqlstr,@scrollopt = 1 ,@ccopt = 1 ,@rowcount = @rowcount output
select ceiling ( 1.0 * @rowcount / @pagesize) as 总页数 -- ,@rowcount as 总行数,@currentpage as 当前页
set @currentpage = (@currentpage - 1 ) * @pagesize + 1
exec sp_cursorfetch @P1, 16 ,@currentpage,@pagesize
exec sp_cursorclose @P1
set nocount off
测试结果:
测试均为每页10条,三个数字依次为三种方案出结果需要的时间,单位为秒:
第2页:18,10,29
第500页:12,8,21
第50000页:16,18,22
第500000页:24,16,22
这次测试的主要目的是对大数据量不同部分的翻页效率的测试。本以为应该是一个线性的结果,结果发现变化很奇怪。多测试几次结果误差在1、2秒之内,估计sql server对于翻页也是根据不同位置有优化的。看了查询分析,主要的花销还是order by,这还是主键的,如果不是主键,或者是字符串,估计更慢。
因为还有别的事情要忙,也没有做进一步的测试,有兴趣的朋友可以继续做做10万条、无索引、字符串内容的各种测试,记得告诉我一下结果。