带有中间记录的 SQL Server 最小值最大值

Question

如果这个问题之前已经回答过，我很抱歉，但我找不到它。也许是我的错误关键字。

我有这张桌子：

CREATE TABLE test1 
(
    Employee VARCHAR(10),
    Band VARCHAR(10),
    StartDate DATE,
    EndDate DATE
)

INSERT INTO test1 
VALUES ('Emp1', 'Band1', '2009-01-01', '2010-12-31'),
       ('Emp1', 'Band1', '2011-01-01', '2012-12-31'), 
       ('Emp1', 'Band1', '2013-01-01', '2013-08-31'),
       ('Emp1', 'Band2', '2013-09-01', '2013-12-31'),
       ('Emp1', 'Band2', '2014-01-01', '2014-06-30'),
       ('Emp1', 'Band1', '2014-07-01', '2014-12-31'),
       ('Emp1', 'Band1', '2015-01-01', '2018-08-31'),
       ('Emp2', 'Band1', '2012-01-01', '2014-12-31'),
       ('Emp2', 'Band1', '2015-01-01', '2018-03-31')

此表中的结果：

Employee   Band   StartDate   EndDate
----------------------------------------
Emp1       Band1  2009-01-01  2010-12-31
Emp1       Band1  2011-01-01  2012-12-31
Emp1       Band1  2013-01-01  2013-08-31
Emp1       Band2  2013-09-01  2013-12-31
Emp1       Band2  2014-01-01  2014-06-30
Emp1       Band1  2014-07-01  2014-12-31
Emp1       Band1  2015-01-01  2018-08-31
Emp2       Band1  2012-01-01  2014-12-31
Emp2       Band1  2015-01-01  2018-03-31

我要创建的是一个结果表，将每个employee 与band 和最小start date 和最大end date 分组，但是当一个中间记录（band） 存在于类似band 之间的中间，end date 应该被封顶并且类似band 中下一组的start date 应该再次重置。

Employee   Band   StartDate   EndDate
----------------------------------------
Emp1       Band1  2009-01-01  2013-08-31
Emp1       Band2  2013-09-01  2014-06-30
Emp1       Band1  2014-07-01  2018-08-31
Emp2       Band1  2012-01-01  2018-03-31

我已经尝试 CTE 来获取每个波段的最大值和最小值，并将其与原始表进行比较，但我仍然失败了。我也尝试过使用领先和滞后，但仍然失败。

很高兴

假设第二条记录的EndDate是2012-02-01，我更喜欢结果仍然是第一个Band1组的一条记录。

Employee   Band   StartDate   EndDate
----------------------------------------
Emp1       Band1  2009-01-01  2013-08-31

结果的每一行仅在 Band 中有所不同，我可以计算出员工在移动到其他波段（不同波段）之前在某个波段停留了多长时间。

但拥有它是件好事。

Answer 1

显然，您希望在员工更改乐队时对行进行分组。这对于窗口函数来说是直截了当的。以下解决方案在带区更改时向行添加“更改”标志。请注意，它忽略了差距。将 DATEDIFF 检查添加到案例语句中，以查找此人与乐队关联的实际时间：

DECLARE @test1 TABLE(
    Employee VARCHAR(10),
    Band VARCHAR(10),
    StartDate DATE,
    EndDate DATE
);

INSERT INTO @test1 VALUES
('Emp1', 'Band1', '2009-01-01', '2010-12-31'),
('Emp1', 'Band1', '2011-01-01', '2012-12-31'),
('Emp1', 'Band1', '2013-01-01', '2013-08-31'),
('Emp1', 'Band2', '2013-09-01', '2013-12-31'),
('Emp1', 'Band2', '2014-01-01', '2014-06-30'),
('Emp1', 'Band1', '2014-07-01', '2014-12-31'),
('Emp1', 'Band1', '2015-01-01', '2018-08-31'),
('Emp2', 'Band1', '2012-01-01', '2014-12-31'),
('Emp2', 'Band1', '2015-01-01', '2018-03-31');

WITH cte1 AS (
    SELECT *,
        CASE WHEN LAG(Band) OVER (PARTITION BY Employee ORDER BY StartDate) = Band /* AND DATEDIFF(...) */ THEN 0 ELSE 1 END AS Chg
    FROM @test1
), cte2 AS (
    SELECT *,
        SUM(Chg) OVER (PARTITION BY Employee ORDER BY StartDate) AS Grp
    FROM cte1
)
SELECT Employee, Band, MIN(StartDate), Max(EndDate)
FROM cte2
GROUP BY Employee, Band, Grp

DB Fiddle

如果你想看看它是如何工作的，这是中间结果：

| Employee | Band  | StartDate           | EndDate             | Chg | Grp |
|----------|-------|---------------------|---------------------|-----|-----|
| Emp1     | Band1 | 01/01/2009 00:00:00 | 31/12/2010 00:00:00 | 1   | 1   |
| Emp1     | Band1 | 01/01/2011 00:00:00 | 31/12/2012 00:00:00 | 0   | 1   |
| Emp1     | Band1 | 01/01/2013 00:00:00 | 31/08/2013 00:00:00 | 0   | 1   |
| Emp1     | Band2 | 01/09/2013 00:00:00 | 31/12/2013 00:00:00 | 1   | 2   |
| Emp1     | Band2 | 01/01/2014 00:00:00 | 30/06/2014 00:00:00 | 0   | 2   |
| Emp1     | Band1 | 01/07/2014 00:00:00 | 31/12/2014 00:00:00 | 1   | 3   |
| Emp1     | Band1 | 01/01/2015 00:00:00 | 31/08/2018 00:00:00 | 0   | 3   |
| Emp2     | Band1 | 01/01/2012 00:00:00 | 31/12/2014 00:00:00 | 1   | 1   |
| Emp2     | Band1 | 01/01/2015 00:00:00 | 31/03/2018 00:00:00 | 0   | 1   |

Answer 2

这通常称为差距和岛屿。

一种方法

示例

Declare @YourTable Table ([Employee] varchar(50),[Band] varchar(50),[Start] date,[End] date)
Insert Into @YourTable Values 
 ('Emp1','Band1','2009-01-01','2010-12-31')
,('Emp1','Band1','2011-01-01','2012-12-31')
,('Emp1','Band1','2013-01-01','2013-08-31')
,('Emp1','Band2','2013-09-01','2013-12-31')
,('Emp1','Band2','2014-01-01','2014-06-30')
,('Emp1','Band1','2014-07-01','2014-12-31')
,('Emp1','Band1','2015-01-01','2018-08-31')
,('Emp2','Band3','2012-01-01','2014-12-31')
,('Emp2','Band3','2015-01-01','2018-03-31')

;with cte as (
Select *,Grp = sum(Flg) over (Partition By Employee Order by [End])
 From (
        Select *,Flg = IsNull(datediff(DAY,Lag([End],1) over (Partition By Employee,Band Order by [End]) ,[Start]) - 1,1)
         From  @YourTable
      ) A
)
Select Employee
      ,Band
      ,[Start] = min([Start])
      ,[End]   = max([End])
 From cte
 Group By Employee,Band,Grp
 Order by Employee,max([End])

退货

Employee    Band    Start       End
Emp1        Band1   2009-01-01  2013-08-31
Emp1        Band2   2013-09-01  2014-06-30
Emp1        Band1   2014-07-01  2018-08-31
Emp2        Band3   2012-01-01  2018-03-31

如果它有助于可视化，CTE 会生成以下内容

注意标志和组列