在SQL Server 2000、2005、2008、2012中分页结果的最佳方法(性能方面)是什么?如果你还想获得结果的总数(在分页之前)?


当前回答

用例方面,以下内容似乎易于使用和快速。只需设置页码。

use AdventureWorks
DECLARE @RowsPerPage INT = 10, @PageNumber INT = 6;
with result as(
SELECT SalesOrderDetailID, SalesOrderID, ProductID,
ROW_NUMBER() OVER (ORDER BY SalesOrderDetailID) AS RowNum
FROM Sales.SalesOrderDetail
where 1=1
)
select SalesOrderDetailID, SalesOrderID, ProductID from result
WHERE result.RowNum BETWEEN ((@PageNumber-1)*@RowsPerPage)+1
AND @RowsPerPage*(@PageNumber)

同样没有CTE

use AdventureWorks
DECLARE @RowsPerPage INT = 10, @PageNumber INT = 6
SELECT SalesOrderDetailID, SalesOrderID, ProductID
FROM (
SELECT SalesOrderDetailID, SalesOrderID, ProductID,
ROW_NUMBER() OVER (ORDER BY SalesOrderDetailID) AS RowNum
FROM Sales.SalesOrderDetail
where 1=1
 ) AS SOD
WHERE SOD.RowNum BETWEEN ((@PageNumber-1)*@RowsPerPage)+1
AND @RowsPerPage*(@PageNumber)

其他回答

获取结果总数和分页是两个不同的操作。对于这个示例,让我们假设您正在处理的查询是

SELECT * FROM Orders WHERE OrderDate >= '1980-01-01' ORDER BY OrderDate

在这种情况下,您将使用以下命令确定结果的总数:

SELECT COUNT(*) FROM Orders WHERE OrderDate >= '1980-01-01'

...这可能看起来效率很低,但实际上性能很好,假设所有索引等都正确设置。

接下来,要以分页方式返回实际结果,下面的查询将是最有效的:

SELECT  *
FROM    ( SELECT    ROW_NUMBER() OVER ( ORDER BY OrderDate ) AS RowNum, *
          FROM      Orders
          WHERE     OrderDate >= '1980-01-01'
        ) AS RowConstrainedResult
WHERE   RowNum >= 1
    AND RowNum < 20
ORDER BY RowNum

这将返回原始查询的第1-19行。这里很酷的事情是,尤其是对于web应用程序,你不需要保留任何状态,除了要返回的行号。

对于SQL Server 2000,您可以使用一个带有IDENTITY列的表变量来模拟ROW_NUMBER():

DECLARE @pageNo int -- 1 based
DECLARE @pageSize int
SET @pageNo = 51
SET @pageSize = 20

DECLARE @firstRecord int
DECLARE @lastRecord int
SET @firstRecord = (@pageNo - 1) * @pageSize + 1 -- 1001
SET @lastRecord = @firstRecord + @pageSize - 1   -- 1020

DECLARE @orderedKeys TABLE (
  rownum int IDENTITY NOT NULL PRIMARY KEY CLUSTERED,
  TableKey int NOT NULL
)

SET ROWCOUNT @lastRecord
INSERT INTO @orderedKeys (TableKey) SELECT ID FROM Orders WHERE OrderDate >= '1980-01-01' ORDER BY OrderDate

SET ROWCOUNT 0

SELECT t.*
FROM Orders t
  INNER JOIN @orderedKeys o ON o.TableKey = t.ID
WHERE o.rownum >= @firstRecord
ORDER BY o.rownum

这种方法可以扩展到具有多列键的表,并且不会引起使用OR(跳过索引使用)的性能开销。缺点是,如果数据集非常大,并且其中一个接近最后一页,则会占用大量临时空间。在这种情况下,我没有测试游标的性能,但它可能会更好。

注意,这种方法可以针对数据的第一页进行优化。另外,使用ROWCOUNT是因为TOP在SQL Server 2000中不接受变量。

这个位使您能够使用SQL Server和新版本的MySQL进行分页,并携带每行中的总行数。 使用您的类推键来计数唯一行数。

WITH T AS
(  
  SELECT TABLE_ID, ROW_NUMBER() OVER (ORDER BY TABLE_ID) AS RN
  , (SELECT COUNT(TABLE_ID) FROM TABLE) AS TOTAL 
  FROM TABLE (NOLOCK)
)

SELECT T2.FIELD1, T2.FIELD2, T2.FIELD3, T.TOTAL 
FROM TABLE T2 (NOLOCK)
INNER JOIN T ON T2.TABLE_ID=T.TABLE_ID
WHERE T.RN >= 100
AND T.RN < 200

用例方面,以下内容似乎易于使用和快速。只需设置页码。

use AdventureWorks
DECLARE @RowsPerPage INT = 10, @PageNumber INT = 6;
with result as(
SELECT SalesOrderDetailID, SalesOrderID, ProductID,
ROW_NUMBER() OVER (ORDER BY SalesOrderDetailID) AS RowNum
FROM Sales.SalesOrderDetail
where 1=1
)
select SalesOrderDetailID, SalesOrderID, ProductID from result
WHERE result.RowNum BETWEEN ((@PageNumber-1)*@RowsPerPage)+1
AND @RowsPerPage*(@PageNumber)

同样没有CTE

use AdventureWorks
DECLARE @RowsPerPage INT = 10, @PageNumber INT = 6
SELECT SalesOrderDetailID, SalesOrderID, ProductID
FROM (
SELECT SalesOrderDetailID, SalesOrderID, ProductID,
ROW_NUMBER() OVER (ORDER BY SalesOrderDetailID) AS RowNum
FROM Sales.SalesOrderDetail
where 1=1
 ) AS SOD
WHERE SOD.RowNum BETWEEN ((@PageNumber-1)*@RowsPerPage)+1
AND @RowsPerPage*(@PageNumber)
   CREATE view vw_sppb_part_listsource as 
    select row_number() over (partition by sppb_part.init_id order by sppb_part.sppb_part_id asc ) as idx, * from (
      select 
          part.SPPB_PART_ID
          , 0 as is_rev
          , part.part_number 
          , part.init_id 
      from t_sppb_init_part part 
      left join t_sppb_init_partrev prev on ( part.SPPB_PART_ID = prev.SPPB_PART_ID )
      where prev.SPPB_PART_ID is null 
      union 
      select 
          part.SPPB_PART_ID
          , 1 as is_rev
          , prev.part_number 
          , part.init_id 
      from t_sppb_init_part part 
      inner join t_sppb_init_partrev prev on ( part.SPPB_PART_ID = prev.SPPB_PART_ID )
    ) sppb_part

当涉及到不同的init_id时会重新启动idx吗