TSQL - 过滤组，如果它包含所有空

Posted 2023-03-27

【中文标题】TSQL - 过滤组，如果它包含所有空

【英文标题】：TSQL - Filter group if it contains all null

【发布时间】：2018-11-15 10:44:47

【问题描述】：

我想过滤掉包含 TASK_START 和 TASK_END 的所有空值的组，即 B 和 D。

示例表数据

+----+-------+-------------------------+-------------------------+-------------------------+-------------------------+
| ID | STATE |       ENTER_STATE       |       LEAVE_STATE       |       TASK_START        |        TASK_END         |
+----+-------+-------------------------+-------------------------+-------------------------+-------------------------+
| A  | UP    | 2018-11-11 08:00:00.000 | 2018-11-11 08:30:00.000 | 2018-11-11 08:00:00.000 | 2018-11-11 08:10:00.000 |
| A  | UP    | 2018-11-11 09:00:00.000 | 2018-11-11 09:30:00.000 | NULL                    | NULL                    |
| A  | UP    | 2018-11-11 10:00:00.000 | 2018-11-11 10:30:00.000 | 2018-11-11 08:20:00.000 | 2018-11-11 08:30:00.000 |
| B  | UP    | 2018-11-11 08:00:00.000 | 2018-11-11 09:00:00.000 | NULL                    | NULL                    |
| B  | UP    | 2018-11-11 09:00:00.000 | 2018-11-11 10:00:00.000 | NULL                    | NULL                    |
| B  | UP    | 2018-11-11 10:20:00.000 | 2018-11-11 11:00:00.000 | NULL                    | NULL                    |
| B  | UP    | 2018-11-11 11:00:00.000 | 2018-11-11 12:00:00.000 | NULL                    | NULL                    |
| C  | UP    | 2018-11-11 08:00:00.000 | 2018-11-11 08:20:00.000 | 2018-11-11 08:15:00.000 | 2018-11-11 08:30:00.000 |
| C  | UP    | 2018-11-11 08:20:00.000 | 2018-11-11 08:30:00.000 | 2018-11-11 08:20:00.000 | 2018-11-11 08:35:00.000 |
| D  | UP    | 2018-11-11 08:00:00.000 | 2018-11-11 08:10:00.000 | NULL                    | NULL                    |
| D  | UP    | 2018-11-11 08:10:00.000 | 2018-11-11 09:10:00.000 | NULL                    | NULL                    |
+----+-------+-------------------------+-------------------------+-------------------------+-------------------------+

【参考方案1】：

试试这个：

DECLARE @DataSource TABLE
(
    [ID] CHAR(1)
   ,[STATE] CHAR(2)
   ,[ENTER_STATE] DATETIME2(0)
   ,[LEAVE_STATE] DATETIME2(0)
   ,[TASK_START] DATETIME2(0)
   ,[TASK_END] DATETIME2(0)
);

INSERT INTO @DataSource
VALUES ('A', 'UP', '2018-11-11 08:00:00.000', '2018-11-11 08:30:00.000', '2018-11-11 08:00:00.000', '2018-11-11 08:10:00.000')
      ,('A', 'UP', '2018-11-11 09:00:00.000', '2018-11-11 09:30:00.000', NULL, NULL)
      ,('A', 'UP', '2018-11-11 10:00:00.000', '2018-11-11 10:30:00.000', '2018-11-11 08:20:00.000', '2018-11-11 08:30:00.000')
      ,('B', 'UP', '2018-11-11 08:00:00.000', '2018-11-11 09:00:00.000', NULL, NULL)
      ,('B', 'UP', '2018-11-11 09:00:00.000', '2018-11-11 10:00:00.000', NULL, NULL)
      ,('B', 'UP', '2018-11-11 10:20:00.000', '2018-11-11 11:00:00.000', NULL, NULL)
      ,('B', 'UP', '2018-11-11 11:00:00.000', '2018-11-11 12:00:00.000', NULL, NULL)
      ,('C', 'UP', '2018-11-11 08:00:00.000', '2018-11-11 08:20:00.000', '2018-11-11 08:15:00.000', '2018-11-11 08:30:00.000')
      ,('C', 'UP', '2018-11-11 08:20:00.000', '2018-11-11 08:30:00.000', '2018-11-11 08:20:00.000', '2018-11-11 08:35:00.000')
      ,('D', 'UP', '2018-11-11 08:00:00.000', '2018-11-11 08:10:00.000', NULL, NULL)
      ,('D', 'UP', '2018-11-11 08:10:00.000', '2018-11-11 09:10:00.000', NULL, NULL);

WITH DataSource AS
(
    SELECT *
         ,MAX([TASK_START]) OVER (PARTITION BY [ID]) AS [DateStart]
         ,MAX([TASK_END]) OVER (PARTITION BY [ID]) AS [DateEnd]
    FROM @DataSource
)
SELECT *
FROM DataSource
WHERE NOT ([DateStart] IS NULL AND [DateEnd] IS NULL);   

我们的想法是获取每个组的最大（或最小）日期，然后，如果存在此值为 NULL 的行，则排除它们。

【参考方案2】：

你可以使用not exists：

select t.*
from table t
where not exists (select 1 
                  from table t1 
                  where t1.id = t.id and t1.task_start is not null and t1.task_end is not null
                 );

其他选项将使用GROUP BY：

select id
from table t
group by id
having sum(case when task_start is not null then 1 else 0  end) = 0 and
       sum(case when task_end is not null then 1 else 0  end) = 0;

【讨论】：

不会也从 A 中删除该字段吗？

@Kyrive。 . .是的，它会删除A，因为TASK_START & TASK_END 不为空。

【参考方案3】：

我会简单地将count() 与having 一起使用：

select id
from table t
group by id
having count(task_start) = 0 and
       count(task_end) = 0;