陷入 t-SQL 中单个表的选择性 DELETE 语句中

Question

说，我有一张这样的表（SQL Server 2008）：

CREATE TABLE tbl (ID INT, dtIn DATETIME2, dtOut DATETIME2, Type INT)

INSERT tbl VALUES
(1, '05:00', '6:00', 1),  -- will be removed
(2, '05:00', '7:00', 1),  -- will be removed
(3, '05:01', '8:00', 1),
(4, '05:00', '8:00', 1),
(5, '05:00', '6:00', 2),  -- will be removed
(6, '05:00', '7:00', 2),
(7, '05:00', '7:00', 3),
(8, '04:00', '7:00', 3)

我需要删除所有具有相同“类型”的记录（如果找到 2 条或更多条），它们的“类型”具有相同的“dtIn”，但具有最大“dtOut”的记录除外。换句话说，上表应该是这样的：

(3, '05:01', '8:00', 1),   -- no matching 'dtIn' for 'type' = 1
(4, '05:00', '8:00', 1),   -- largest 'dtOut' for 'type' = 1
(6, '05:00', '7:00', 2),   -- largest 'dtOut' for 'type' = 2
(7, '05:00', '7:00', 3),   -- no matching 'dtIn' for 'type' = 3
(8, '04:00', '7:00', 3)    -- no matching 'dtIn' for 'type' = 4

您如何甚至从同一个表中选择多个具有相同类型的行。你不能 select * from tbl where type=type.... 无论如何，我会很感激这方面的帮助...

Answer 1

这是选择要删除的行的一种方法：

SELECT *
FROM tbl T1
WHERE EXISTS
(
    SELECT *
    FROM tbl T2
    WHERE T1.Type = T2.Type
    AND T1.dtIn = T2.dtIn
    AND (
            T1.dtOut < T2.dtOut
            OR (T1.dtOut = T2.dtOut AND T1.id < T2.id)
        )
)

此查询也可以轻松更改为实际删除行。只需将SELECT * 更改为DELETE T1。但请在实际运行 delete 语句之前测试它是否符合您的要求。

在线查看：sqlfiddle

---

更新

这是一种使用 ROW_NUMBER 的方法：

;WITH T1 AS (
    SELECT id, ROW_NUMBER() OVER (PARTITION BY Type, dtIn
                                  ORDER BY dtOut DESC, ID DESC) AS rn
    FROM tbl
)
SELECT * FROM tbl
WHERE id IN
(
    SELECT id
    FROM T1
    WHERE rn > 1
)

在线查看：sqlfiddle

Answer 2

不是很优雅（+ 条件与马克的答案相同），但它是一种替代方案。

Fiddle is here

;with my_cte(ID)
as
(
  SELECT T1.ID
  FROM 
      tbl T1
      JOIN tbl T2 on
        T1.Type = T2.Type
        AND T1.dtIn = T2.dtIn
        AND T1.dtOut < T2.dtOut
  )
select *
from tbl t
where not exists 
              (
              select 1 
              from my_cte c 
              where t.ID = c.ID
              ) 

Answer 3

重复行的查询

我刚刚看到您可以有重复的行。所以：

DELETE X
FROM
   (SELECT Row_Number() OVER (PARTITION BY Type, dtIn ORDER BY dtOut DESC) Item, *
   FROM #tbl) X
WHERE Item > 1

请注意，以上版本不需要连接，因此应该比页面上的所有其他查询执行得更好。

或

DELETE T
FROM #tbl T
   CROSS APPLY (
      SELECT TOP 1 *
      FROM #tbl T2
      WHERE
         T.dtIn = T2.dtIn
         AND T.Type = T2.Type
      ORDER BY T2.dtOut DESC, T2.ID DESC
   ) X
WHERE T.ID < X.ID

或

DELETE T
FROM #tbl T
   INNER JOIN #tbl T2
      ON T.dtIn = T2.dtIn
      AND T.Type = T2.Type
      AND (
         T.dtOut < T2.dtOut
         OR (T.tdOut = T2.dtOut AND T.ID > T2.ID)
      )

查询无重复行

如果您不能有重复的行（ID 除外），那么它会更简单：

DELETE T
FROM #tbl T
WHERE EXISTS (
   SELECT * FROM #tbl T2
   WHERE
      T.dtIn = T2.dtIn
      AND T.Type = T2.Type
      AND T.dtOut < T2.dtOut
)

甚至

DELETE T
FROM #tbl T
   INNER JOIN #tbl T2
      ON T.dtIn = T2.dtIn
      AND T.Type = T2.Type
      AND T.dtOut < T2.dtOut

Answer 4

我的和@Mark 的有点不同，但我相信它应该会得到相同的结果（只是因为我认为id 是独一无二的）：

DELETE a
FROM dbo.tbl a
WHERE id <> (
    SELECT MAX(id)
    from dbo.tbl b
    WHERE a.type = b.type
      AND a.dtIn = b.dtIn
)

Answer 5

DECLARE @Table TABLE     
(  
ID INT,  
newno int  
)   

insert @table  
SELECT T1.ID,   
ROW_NUMBER() OVER (PARTITION BY Type, dtIn
                              ORDER BY dtOut DESC, ID DESC) AS newno  
 FROM tbl T1      

 select *  
from tbl t
where t.ID IN
             (
              select ID from @Table where newno > 1
              )