SQL查詢前行優化

Question

這是我的表結構

MyTable
-----------

ObjectID int (Identity),           -- Primary Key
FileName varchar(10),
CreatedDate datetime
...........
...........
...........

我需要花時間在文件中創建記錄...即...同一文件中的上一條記錄與同一文件的當前記錄之間經過的時間

即...如果記錄是

ObjectID    FileName    CreatedDate (just showing the time part here)
--------    --------    -----------
1           ABC         10:23
2           ABC         10:25
3           DEF         10:26
4           ABC         10:30
5           DEF         10:31
6           DEF         10:35

所需的輸出是...

ObjectID    FileName    CreatedDate     PrevRowCreatedDate
--------    --------    ----------- ---------------
1           ABC         10:23           NULL
2           ABC         10:25           10:23
3           DEF         10:26           NULL
4           ABC         10:30           10:25
5           DEF         10:31           10:26
6           DEF         10:35           10:31

到目前為止，我得到了這個查詢，但是它花費的時間比預期的要長。有更好的方法嗎？

    Select  A.ObjectID, 
        A.FileName
        A.CreatedDate as CreatedDate, 
        B.PrevRowCreatedDate,
        datediff("SS", '1900-01-01 00:00:00', Coalesce((A.CreatedDate - B.PrevRowCreatedDate),0)) as secondsTaken
    from MyTable as A 
        Cross Apply (       
        (Select PrevRowCreatedDate = Max(CreatedDate) from MyTable as BB 
                        where   BB.FileName = A.FileName and 
                                BB.CreatedDate < A.CreatedDate
        )
        ) as B  

如果您需要更多信息，請告訴我

謝謝

Answer 1

SELECT t1.FileName, t1.CreatedDate, t2.CreatedDate as PrevCreatedDate
FROM 
   (SELECT FileName, CreateDate,
          ROW_NUMBER() OVER(PARTITION BY FileName ORDER BY CreatedDate) AS OrderNo
   FROM MyTable) t1
LEFT JOIN
   (SELECT FileName, CreateDate,
     ROW_NUMBER() OVER(PARTITION BY FileName ORDER BY CreatedDate) AS OrderNo
     FROM MyTable) t2
ON (t1.FileName = t2.FileName AND t1.OrderNo = t2.OrderNo - 1)

或者最好使用“ WITH”，因為查詢是相同的：

WITH t(ObjectID, FileName, CreatedDate, OrderNo) AS
   (SELECT ObjectID, FileName, CreatedDate,
          ROW_NUMBER() OVER(PARTITION BY FileName ORDER BY CreatedDate) AS OrderNo
   FROM MyTable) 
SELECT t1.ObjectID, t1.FileName, t1.CreatedDate, t2.CreatedDate AS PrevCreatedDate,
        DATEDIFF("SS", '1900-01-01 00:00:00', 
           COALESCE((t1.CreatedDate - t2.CreatedDate),0)) AS secondsTaken
FROM t t1 LEFT JOIN t t2 
ON (t1.FileName = t2.FileName AND t1.OrderNo = t2.OrderNo + 1)

Answer 2

我認為Michael的回答確實應該更有效。 盡管在評估效率時，我只想提請注意Management Studio中顯示的查詢成本（相對於批次）的問題。

我建立了一個包含23,174行的測試表，並在問題和邁克爾的問題中運行了查詢。 查看實際執行計划中的“查詢成本（相對於批次）”，原始查詢為1％，邁克爾的成本為99％，因此效率極低。

但是實際數據卻完全不同

交叉應用方法

SQL Server執行時間：CPU時間= 0毫秒，經過的時間= 0毫秒。
表“ MyTable”。 掃描計數23175 ，邏輯讀取49335 ，物理讀取0，預讀讀取0，lob邏輯讀取0，lob物理讀取0，lob提前讀取0。

ROW_NUMBER方法

SQL Server執行時間：CPU時間= 391毫秒，經過的時間= 417毫秒。
表“工作表”。 掃描計數0，邏輯讀0，物理讀0，預讀0，lob邏輯讀0，lob物理讀0，lob預讀0。
表“ MyTable”。 掃描計數2，邏輯讀148，物理讀0，預讀0，lob邏輯讀0，lob物理讀0，lob預讀0。

在ROW_NUMBER計划中， rownumber=rownumber+1上的Merge Join有23,174行進入雙方。 此值是唯一的，實際的行數也為23,174。 但是，SQL Server估計從該聯接產生的行將為34,812,000，因此其在計划中稍后插入的估計成本非常不准確。

測試腳本

BEGIN TRAN

CREATE TABLE MyTable
  (
     [ObjectID]    [INT] IDENTITY(1, 1) NOT NULL PRIMARY KEY CLUSTERED,
     [FileName]    [VARCHAR](50) NULL,
     [CreatedDate] [DATETIME] NULL
  )

GO

INSERT INTO MyTable
SELECT ISNULL(type, NEWID()),
       DATEADD(DAY, CAST(RAND(CAST(NEWID() AS VARBINARY)) * 10000 AS INT), GETDATE())
FROM   master.dbo.spt_values,
       (SELECT TOP 10 1 AS X FROM  master.dbo.spt_values) V


DELETE FROM MyTable
WHERE  EXISTS(SELECT *
              FROM   MyTable m2
              WHERE  MyTable.CreatedDate = m2.CreatedDate
                     AND MyTable.FileName = m2.FileName
                     AND MyTable.ObjectID < m2.ObjectID)

CREATE UNIQUE NONCLUSTERED INDEX [IX_MyTable]
  ON MyTable ([FileName] ASC, [CreatedDate] ASC)

SET STATISTICS IO ON
SET STATISTICS TIME ON

SELECT A.ObjectID,
       A.FileName,
       A.CreatedDate                                                                                AS CreatedDate,
       B.PrevRowCreatedDate,
       DATEDIFF("SS", '1900-01-01 00:00:00', COALESCE(( A.CreatedDate - B.PrevRowCreatedDate ), 0)) AS secondsTaken
INTO   #A
FROM   MyTable AS A
       CROSS APPLY ((SELECT PrevRowCreatedDate = MAX(CreatedDate)
                     FROM   MyTable AS BB
                     WHERE  BB.FileName = A.FileName
                            AND BB.CreatedDate < A.CreatedDate)) AS B;

WITH t(ObjectID, FileName, CreatedDate, OrderNo)
     AS (SELECT ObjectID,
                FileName,
                CreatedDate,
                RANK() OVER(PARTITION BY FileName ORDER BY CreatedDate) AS OrderNo
         FROM   MyTable)
SELECT t1.ObjectID,
       t1.FileName,
       t1.CreatedDate,
       t2.CreatedDate                                                                          AS PrevCreatedDate,
       DATEDIFF("SS", '1900-01-01 00:00:00', COALESCE(( t1.CreatedDate - t2.CreatedDate ), 0)) AS secondsTaken
INTO   #B
FROM   t t1
       LEFT JOIN t t2
         ON ( t1.FileName = t2.FileName
              AND t1.OrderNo = t2.OrderNo + 1 )

/*Test the 2 queries give the same result*/
SELECT *
FROM   #A
EXCEPT
SELECT *
FROM   #B

SELECT *
FROM   #B
EXCEPT
SELECT *
FROM   #A

ROLLBACK