SQL查詢來排除基於類型的記錄和日期

Question

ORIGINAL

match_date actual_date  colA  type type_desc 
09/16/11 10/1/2011  ABC12345 A  TTT222 
09/16/11 10/8/2011  ABC12345 S  BADTYPE_123 
09/16/11 11/8/2011  ABC12345 A  YYY222 
09/16/11 11/8/2011  ABC12345 A  WWW333 
09/16/11 11/8/2011  ABC12345 B  YYY222 
09/16/11 11/8/2011  ABC12345 B  WWW333 
05/11/12 9/17/2012  ABC12345 B  ZZZ222 
05/11/12 9/17/2012  ABC12345 A  ZZZ222 
05/11/12 9/17/2012  MNO12345 B  CCC222 
05/11/12 9/17/2012  MNO12345 A  CCC222 
08/16/12 10/8/2011  MNO12345 S  BADTYPE_789 
08/16/12 10/9/2011  MNO12345 A  CCC111 
11/11/12 11/17/2012  MNO12345 S  BADTYPE_790 
12/01/12 9/17/2012  MNO12345 A  DDD222 
11/20/12 1/06/2013  XYZ98765 B  TST111 
11/20/12 1/06/2013  XYZ98765 A  TST111 
01/15/13 3/17/2013  XYZ98765 A  TST222 
05/11/13 6/15/2013  XYZ98765 B  TST111 
05/11/13 9/15/2013  XYZ98765 A  TST111 

對於每個給定match_date「組」（可樂和比賽日期，ABC12345和11年9月16日和ABC12345 05/11/12在第一「group」示例）中，如果記錄類型爲「S」且包含「BADTYPE *」作爲type_desc，我想排除/刪除給定「match_date組」的「BADTYPE」actual_date和下一個match_date之間的所有記錄。

如果給定的「match_date組」沒有「BADTYPE」記錄，則應該忽略它。如果給定match_date組的唯一記錄是BADTYPE記錄，則它將保持並繼續。

記錄到要刪除的紀錄

match_date actual_date  colA  type type_desc 
09/16/11 11/8/2011  ABC12345 A  YYY222 
09/16/11 11/8/2011  ABC12345 A  WWW333 
09/16/11 11/8/2011  ABC12345 B  YYY222 
09/16/11 11/8/2011  ABC12345 B  WWW333 
08/16/12 10/9/2011  MNO12345 A  CCC111 

最終結果中移除

match_date actual_date  colA  type type_desc 
09/16/11 10/1/2011  ABC12345 A  TTT222 
09/16/11 10/8/2011  ABC12345 S  BADTYPE_123 
05/11/12 9/17/2012  ABC12345 B  ZZZ222 
05/11/12 9/17/2012  ABC12345 A  ZZZ222 
05/11/12 9/17/2012  MNO12345 B  CCC222 
05/11/12 9/17/2012  MNO12345 A  CCC222 
08/16/12 10/8/2011  MNO12345 S  BADTYPE_789 
11/11/12 11/17/2012  MNO12345 S  BADTYPE_790 
12/01/12 9/17/2012  MNO12345 A  DDD222 
11/20/12 1/06/2013  XYZ98765 B  TST111 
11/20/12 1/06/2013  XYZ98765 A  TST111 
01/15/13 3/17/2013  XYZ98765 A  TST222 
05/11/13 6/15/2013  XYZ98765 B  TST111 
05/11/13 9/15/2013  XYZ98765 A  TST111 

希望我已經充分說明了正是我試圖做的。

任何幫助將不勝感激。

Answer 1

我認爲最簡單快速的方法是使用窗口函數LEAD and LAG，但就您使用SQL Server 2008而言，這是不可能的。

在SQL Server 2008中，你可以使用outer apply - 通常，你需要是獲得由actual_date下降下令給「match_group」以前的記錄，刪除記錄，如果以前type_desc like 'BADTYPE%' and type = 'S'

當你決定你需要什麼，你可以很輕鬆錄製該算法到SQL，和感覺聲明語言的力量：

delete Table1 
from Table1 as T1 
    cross apply 
    (
     select top 1 T2.[type_desc], T2.[type] 
     from Table1 as T2 
     where T2.ColA = T1.ColA and T2.match_date = T1.match_date and T2.[actual_date] < T1.[actual_date] 
     order by T2.[actual_date] desc 
    ) as PR 
where PR.[type_desc] like 'BADTYPE%' and PR.[type] = 'S' 

你也可以查看SQL FIDDLE EXAMPLE，看看它是如何工作的

Answer 2

我建議你爲此使用遞歸查詢。
事情是這樣的：

SELECT *, 
     Row_number() 
     OVER ( 
      PARTITION BY COLA, MATCH_DATE 
      ORDER BY ACTUAL_DATE) RN 
INTO #TEMP1 
FROM TABLE1; 

WITH CTE 
    AS (SELECT *, 
       Cast(1 AS INT) AS flag 
     FROM #TEMP1 
     WHERE RN = 1 
     UNION ALL 
     SELECT t1.*, 
       CASE 
        WHEN T2.FLAG = 2 
         OR (t2.TYPE_DESC LIKE 'BADTYPE%' 
          AND t2.TYPE = 'S') THEN 2 
        ELSE 1 
       END flag 
     FROM #TEMP1 T1 
       INNER JOIN CTE T2 
         ON T1.COLA = T2.COLA 
          AND t1.MATCH_DATE = t2.MATCH_DATE 
          AND t1.RN = T2.RN + 1) 
SELECT [MATCH_DATE], 
     [ACTUAL_DATE], 
     [COLA], 
     [TYPE], 
     [TYPE_DESC] 
FROM CTE 
WHERE FLAG = 1 
ORDER BY COLA, 
      MATCH_DATE 

您可以在SQL Fiddle玩弄工作的例子。

祝你好運！