在项目中，我们经常遇到或用到分页，那么在大数据量（百万级以上）下，哪种分页算法效率最优呢？我们不妨用事实说话。
测试环境
硬件：CPU 酷睿双核T5750 内存：2G
软件:Windows server 2003 + sql server 2005
OK,我们首先创建一数据库：data_Test,并在此数据库中创建一表：tb_TestTable
代码如下:
create database data_Test --创建数据库
data_Test
GO
use data_Test
GO
create table tb_TestTable --创建表
(id int identity(1,1) primary key,
userName nvarchar(20) not null,
userPWD nvarchar(20) not null,
userEmail nvarchar(40) null)
GO
然后我们在数据表中插入2000000条数据:
代码如下:
--插入数据
set identity_insert tb_TestTable on
declare @count int
set @count=1
while @countbegin
insert into tb_TestTable(id,userName,userPWD,userEmail) values(@count,'admin','admin888','lli0077@yahoo.com.cn')
set @count=@count+1
end
set identity_insert tb_TestTable off
我首先写了五个常用存储过程：
1，利用select top 和select not in进行分页，具体代码如下：
代码如下:create procedure proc_paged_with_notin --利用select top and select not in
(
@pageIndex int, --页索引
@pageSize int --每页记录数
)
as
begin
set nocount on;
declare @timediff datetime --耗时
declare @sql nvarchar(500)
select @timediff=Getdate()
set @sql='select top '+str(@pageSize)+' * from tb_TestTable where(ID not in(select top '+str(@pageSize*@pageIndex)+' id from tb_TestTable order by ID ASC)) order by ID'
execute(@sql) --因select top后不支技直接接参数，所以写成了字符串@sql
select datediff(ms,@timediff,GetDate()) as 耗时
set nocount off;
end
2,利用select top 和 select max(列键)
代码如下:create procedure proc_paged_with_selectMax --利用select top and select max(列)
(
@pageIndex int, --页索引
@pageSize int --页记录数
)
as
begin
set nocount on;
declare @timediff datetime
declare @sql nvarchar(500)
select @timediff=Getdate()
set @sql='select top '+str(@pageSize)+' * From tb_TestTable where(ID>(select max(id) From (select top '+str(@pageSize*@pageIndex)+' id From tb_TestTable order by ID) as TempTable)) order by ID'
execute(@sql)
select datediff(ms,@timediff,GetDate()) as 耗时
set nocount off;
end
3,利用select top和中间变量--此方法因网上有人说效果最佳，所以贴出来一同测试
代码如下:create procedure proc_paged_with_Midvar --利用ID>最大ID值和中间变量
(
@pageIndex int,
@pageSize int
)
as
declare @count int
declare @ID int
declare @timediff datetime
declare @sql nvarchar(500)
begin
set nocount on;
select @count=0,@ID=0,@timediff=getdate()
select @count=@count+1,@ID=case when @countset @sql='select top '+str(@pageSize)+' * from tb_testTable where ID>'+str(@ID)
execute(@sql)
select datediff(ms,@timediff,getdate()) as 耗时
set nocount off;
end
4,利用Row_number() 此方法为sql server 2005中新的方法,利用Row_number()给数据行加上索引
代码如下:create procedure proc_paged_with_Rownumber --利用SQL 2005中的Row_number()
(
@pageIndex int,
@pageSize int
)
as
declare @timediff datetime
begin
set nocount on;
select @timediff=getdate()
select * from (select *,Row_number() over(order by ID asc) as IDRank from tb_testTable) as IDWithRowNumber where IDRank>@pageSize*@pageIndex and IDRankselect datediff(ms,@timediff,getdate()) as 耗时
set nocount off;
end
5，利用临时表及Row_number
代码如下:create procedure proc_CTE --利用临时表及Row_number
(
@pageIndex int, --页索引
@pageSize int --页记录数
)
as
set nocount on;
declare @ctestr nvarchar(400)
declare @strSql nvarchar(400)
declare @datediff datetime
begin
select @datediff=GetDate()
set @ctestr='with Table_CTE as
(select ceiling((Row_number() over(order by ID ASC))/'+str(@pageSize)+') as page_num,* from tb_TestTable)';
set @strSql=@ctestr+' select * From Table_CTE where page_num='+str(@pageIndex)
end
begin
execute sp_executesql @strSql
select datediff(ms,@datediff,GetDate())
set nocount off;
end
OK,至此，存储过程创建完毕，我们分别在每页10条数据的情况下在第2页，第1000页，第10000页，第100000页，第199999页进行测试，耗时单位：ms 每页测试5次取其平均值
存过 第2页耗时 第1000页耗时 第10000页耗时 第100000页耗时 第199999页耗时 效率排行
1用not in 0ms 16ms 47ms 475ms 953ms 3
2用select max 5ms 16ms 35ms 325ms 623ms 1
3中间变量 966ms 970ms 960ms 945ms 933ms 5
4row_number 0ms 0ms 34ms 365ms 710ms 2
4临时表 780ms 796ms 798ms 780ms 805ms 4
测试结果显示：select max >row_number>not in>临时表>中间变量
于是我对效率最高的select max方法用2分法进行了扩展，代码取自互联网，我修改了ASC排序时取不到值的BUG，测试结果:
2分法 156ms 156ms 180ms 470ms 156ms 1*
从测试结果来看，使用2分法确实可以提高效率并使效率更为稳定，我又增加了第159999页的测试，用时仅296ms，效果相当的不错！
下面是2分法使用select max的代码，已相当完善。
代码如下:
--/*-----存储过程 分页处理 孙伟 2005-03-28创建 -------*/
--/*-----存储过程 分页处理 浪尘 2008-9-1修改----------*/
--/*----- 对数据进行了2分处理使查询前半部分数据与查询后半部分数据性能相同 -------*/
alter PROCEDURE proc_paged_2part_selectMax
(
@tblName nvarchar(200), ----要显示的表或多个表的连接
@fldName nvarchar(500) = '*', ----要显示的字段列表
@pageSize int = 10, ----每页显示的记录个数
@page int = 1, ----要显示那一页的记录
@fldSort nvarchar(200) = null, ----排序字段列表或条件
@Sort bit = 0, ----排序方法，0为升序，1为降序(如果是多字段排列Sort指代最后一个排序字段的排列顺序(最后一个排序字段不加排序标记)--程序传参如：' SortA Asc,SortB Desc,SortC ')
@strCondition nvarchar(1000) = null, ----查询条件,不需where
@ID nvarchar(150), ----主表的主键
@Dist bit = 0, ----是否添加查询字段的 DISTINCT 默认0不添加/1添加
@pageCount int = 1 output, ----查询结果分页后的总页数
@Counts int = 1 output ----查询到的记录数
)
AS
SET NOCOUNT ON
Declare @sqlTmp nvarchar(1000) ----存放动态生成的SQL语句
Declare @strTmp nvarchar(1000) ----存放取得查询结果总数的查询语句
Declare @strID nvarchar(1000) ----存放取得查询开头或结尾ID的查询语句
Declare @strSortType nvarchar(10) ----数据排序规则A
Declare @strFSortType nvarchar(10) ----数据排序规则B
Declare @SqlSelect nvarchar(50) ----对含有DISTINCT的查询进行SQL构造
Declare @SqlCounts nvarchar(50) ----对含有DISTINCT的总数查询进行SQL构造
declare @timediff datetime --耗时测试时间差
select @timediff=getdate()
if @Dist = 0
begin
set @SqlSelect = 'select '
set @SqlCounts = 'Count(*)'
end
else
begin
set @SqlSelect = 'select distinct '
set @SqlCounts = 'Count(DISTINCT '+@ID+')'
end
if @Sort=0
begin
set @strFSortType=' ASC '
set @strSortType=' DESC '
end
else
begin
set @strFSortType=' DESC '
set @strSortType=' ASC '
end
--------生成查询语句--------
--此处@strTmp为取得查询结果数量的语句
if @strCondition is null or @strCondition='' --没有设置显示条件
begin
set @sqlTmp = @fldName + ' From ' + @tblName
set @strTmp = @SqlSelect+' @Counts='+@SqlCounts+' FROM '+@tblName
set @strID = ' From ' + @tblName
end
else
begin
set @sqlTmp = + @fldName + 'From ' + @tblName + ' where (1>0) ' + @strCondition
set @strTmp = @SqlSelect+' @Counts='+@SqlCounts+' FROM '+@tblName + ' where (1>0) ' + @strCondition
set @strID = ' From ' + @tblName + ' where (1>0) ' + @strCondition
end
----取得查询结果总数量-----
exec sp_executesql @strTmp,N'@Counts int out ',@Counts out
declare @tmpCounts int
if @Counts = 0
set @tmpCounts = 1
else
set @tmpCounts = @Counts
--取得分页总数
set @pageCount=(@tmpCounts+@pageSize-1)/@pageSize
/**//**//**//**当前页大于总页数 取最后一页**/
if @page>@pageCount
set @page=@pageCount
--/*-----数据分页2分处理-------*/
declare @pageIndex int --总数/页大小
declare @lastcount int --总数%页大小
set @pageIndex = @tmpCounts/@pageSize
set @lastcount = @tmpCounts%@pageSize
if @lastcount > 0
set @pageIndex = @pageIndex + 1
else
set @lastcount = @pagesize
--//***显示分页
if @strCondition is null or @strCondition='' --没有设置显示条件
begin
if @pageIndexbegin
if @page=1
set @strTmp=@SqlSelect+' top '+ CAST(@pageSize as VARCHAR(4))+' '+ @fldName+' from '+@tblName
+' order by '+ @fldSort +' '+ @strFSortType
else
begin
if @Sort=1
begin
set @strTmp=@SqlSelect+' top '+ CAST(@pageSize as VARCHAR(4))+' '+ @fldName+' from '+@tblName
+' where '+@ID+' +' order by '+ @fldSort +' '+ @strFSortType+') AS TBMinID)'
+' order by '+ @fldSort +' '+ @strFSortType
end
else
begin
set @strTmp=@SqlSelect+' top '+ CAST(@pageSize as VARCHAR(4))+' '+ @fldName+' from '+@tblName
+' where '+@ID+' >(select max('+ @ID +') from ('+ @SqlSelect+' top '+ CAST(@pageSize*(@page-1) as Varchar(20)) +' '+ @ID +' from '+@tblName
+' order by '+ @fldSort +' '+ @strFSortType+') AS TBMinID)'
+' order by '+ @fldSort +' '+ @strFSortType
end
end
end
else
begin
set @page = @pageIndex-@page+1 --后半部分数据处理
if @page set @strTmp=@SqlSelect+' * from ('+@SqlSelect+' top '+ CAST(@lastcount as VARCHAR(4))+' '+ @fldName+' from '+@tblName
+' order by '+ @fldSort +' '+ @strSortType+') AS TempTB'+' order by '+ @fldSort +' '+ @strFSortType
else
if @Sort=1
begin
set @strTmp=@SqlSelect+' * from ('+@SqlSelect+' top '+ CAST(@pageSize as VARCHAR(4))+' '+ @fldName+' from '+@tblName
+' where '+@ID+' >(select max('+ @ID +') from('+ @SqlSelect+' top '+ CAST(@pageSize*(@page-2)+@lastcount as Varchar(20)) +' '+ @ID +' from '+@tblName
+' order by '+ @fldSort +' '+ @strSortType+') AS TBMaxID)'
+' order by '+ @fldSort +' '+ @strSortType+') AS TempTB'+' order by '+ @fldSort +' '+ @strFSortType
end
else
begin
set @strTmp=@SqlSelect+' * from ('+@SqlSelect+' top '+ CAST(@pageSize as VARCHAR(4))+' '+ @fldName+' from '+@tblName
+' where '+@ID+' +' order by '+ @fldSort +' '+ @strSortType+') AS TBMaxID)'
+' order by '+ @fldSort +' '+ @strSortType+') AS TempTB'+' order by '+ @fldSort +' '+ @strFSortType
end
end
end
else --有查询条件
begin
if @pageIndexbegin
if @page=1
set @strTmp=@SqlSelect+' top '+ CAST(@pageSize as VARCHAR(4))+' '+ @fldName+' from '+@tblName
+' where 1=1 ' + @strCondition + ' order by '+ @fldSort +' '+ @strFSortType
else if(@Sort=1)
begin
set @strTmp=@SqlSelect+' top '+ CAST(@pageSize as VARCHAR(4))+' '+ @fldName+' from '+@tblName
+' where '+@ID+' +' where (1=1) ' + @strCondition +' order by '+ @fldSort +' '+ @strFSortType+') AS TBMinID)'
+' '+ @strCondition +' order by '+ @fldSort +' '+ @strFSortType
end
else
begin
set @strTmp=@SqlSelect+' top '+ CAST(@pageSize as VARCHAR(4))+' '+ @fldName+' from '+@tblName
+' where '+@ID+' >(select max('+ @ID +') from ('+ @SqlSelect+' top '+ CAST(@pageSize*(@page-1) as Varchar(20)) +' '+ @ID +' from '+@tblName
+' where (1=1) ' + @strCondition +' order by '+ @fldSort +' '+ @strFSortType+') AS TBMinID)'
+' '+ @strCondition +' order by '+ @fldSort +' '+ @strFSortType
end
end
else
begin
set @page = @pageIndex-@page+1 --后半部分数据处理
if @page set @strTmp=@SqlSelect+' * from ('+@SqlSelect+' top '+ CAST(@lastcount as VARCHAR(4))+' '+ @fldName+' from '+@tblName
+' where (1=1) '+ @strCondition +' order by '+ @fldSort +' '+ @strSortType+') AS TempTB'+' order by '+ @fldSort +' '+ @strFSortType
else if(@Sort=1)
set @strTmp=@SqlSelect+' * from ('+@SqlSelect+' top '+ CAST(@pageSize as VARCHAR(4))+' '+ @fldName+' from '+@tblName
+' where '+@ID+' >(select max('+ @ID +') from('+ @SqlSelect+' top '+ CAST(@pageSize*(@page-2)+@lastcount as Varchar(20)) +' '+ @ID +' from '+@tblName
+' where (1=1) '+ @strCondition +' order by '+ @fldSort +' '+ @strSortType+') AS TBMaxID)'
+' '+ @strCondition+' order by '+ @fldSort +' '+ @strSortType+') AS TempTB'+' order by '+ @fldSort +' '+ @strFSortType
else
set @strTmp=@SqlSelect+' * from ('+@SqlSelect+' top '+ CAST(@pageSize as VARCHAR(4))+' '+ @fldName+' from '+@tblName
+' where '+@ID+' +' where (1=1) '+ @strCondition +' order by '+ @fldSort +' '+ @strSortType+') AS TBMaxID)'
+' '+ @strCondition+' order by '+ @fldSort +' '+ @strSortType+') AS TempTB'+' order by '+ @fldSort +' '+ @strFSortType
end
end
------返回查询结果-----
exec sp_executesql @strTmp
select datediff(ms,@timediff,getdate()) as 耗时
--print @strTmp
SET NOCOUNT OFF
GO
执行示例：exec proc_paged_2part_selectMax 'tb_testTable','ID,userName,userPWD,userEmail',10,100000,'ID',0,null,'ID',0
这种测试只在单机进行，并且没有在实际开发WEB项目中分页测试，测试项也比较单一，所以不够全面系统，但从其效率相比上，我们可以在数据库分页算法上进行有效的控制。
![[Web front-end] Node.js quick start](https://img.php.cn/upload/course/000/000/067/662b5d34ba7c0227.png)
