Microsoft SQL Server 2016，T-SQL：根据各个日期获取数据集的日期范围

Question

我在 SQL Server 2016 中有一个有趣的情况。我正在使用 T-SQL 语言。

我有一个名为 (#dataset) 的数据集：

名为 ContinuousDates 的最后一列将始终具有连续的日期值，例如 2021 年 1 月 1 日至 2021 年 12 月 31 日。它永远不会有相同 ID 或姓名的重复日期，即给定的一个人一天只能有一行数据。（在这个例子中，我只显示了一个人，ID = 1 和 Name = X。在我的实际数据中，我有多个人）。

请注意，纽约市出现在数据集中较早的位置，并在最后 4 行重复出现。

我需要根据日期范围获取以下数据集：

我尝试在数据集上使用简单的 MINIMUM 和 MAXIMUM，但我意识到有时我会得到 错误的 输出，如下所示：

我使用 RANK() 和 DENSE_RANK() 函数尝试了一些选项，但无法找到解决方案。有人可以为我提供帮助吗？

我在这里附上代码：

CREATE TABLE #dataset

(

ID int,
Name varchar(20),
City varchar(20),
ContinuousDates date

)


INSERT INTO #dataset
VALUES(1,'X','NYC','1/1/2021')

INSERT INTO #dataset
VALUES(1,'X','NYC','1/2/2021')

INSERT INTO #dataset
VALUES(1,'X','NYC','1/3/2021')

INSERT INTO #dataset
VALUES(1,'X','SFO','1/4/2021')

INSERT INTO #dataset
VALUES(1,'X','SFO','1/5/2021')

INSERT INTO #dataset
VALUES(1,'X','PHY','1/6/2021')

INSERT INTO #dataset
VALUES(1,'X','PHY','1/7/2021')

INSERT INTO #dataset
VALUES(1,'X','PHY','1/8/2021')

INSERT INTO #dataset
VALUES(1,'X','NYC','1/9/2021')

INSERT INTO #dataset
VALUES(1,'X','NYC','1/10/2021')

INSERT INTO #dataset
VALUES(1,'X','NYC','1/11/2021')

INSERT INTO #dataset
VALUES(1,'X','NYC','1/12/2021')




SELECT *
FROM #dataset
ORDER BY ContinuousDates

我有一套新的代码，为了更好的演示：

CREATE TABLE #dataset

(

ID int,
Name varchar(20),
City varchar(20),
ContinuousDates date

)


INSERT INTO #dataset
VALUES(1,'X','NYC','1/1/2021')

INSERT INTO #dataset
VALUES(1,'X','NYC','1/2/2021')

INSERT INTO #dataset
VALUES(1,'X','NYC','1/3/2021')

INSERT INTO #dataset
VALUES(1,'X','SFO','1/4/2021')

INSERT INTO #dataset
VALUES(1,'X','SFO','1/5/2021')

INSERT INTO #dataset
VALUES(1,'X','PHY','1/6/2021')

INSERT INTO #dataset
VALUES(1,'X','PHY','1/7/2021')

INSERT INTO #dataset
VALUES(1,'X','PHY','1/8/2021')

INSERT INTO #dataset
VALUES(1,'X','NYC','1/9/2021')

INSERT INTO #dataset
VALUES(1,'X','NYC','1/10/2021')

INSERT INTO #dataset
VALUES(1,'X','NYC','1/11/2021')

INSERT INTO #dataset
VALUES(1,'X','NYC','1/12/2021')

INSERT INTO #dataset
VALUES(2,'Y','MEL','1/13/2021')

INSERT INTO #dataset
VALUES(3,'Z','SYD','1/14/2021')

INSERT INTO #dataset
VALUES(3,'Z','SYD','1/15/2021')

INSERT INTO #dataset
VALUES(3,'Z','PER','1/16/2021')

INSERT INTO #dataset
VALUES(4,'A',NULL,'1/16/2021')

INSERT INTO #dataset
VALUES(4,'A', NULL,'1/17/2021')



SELECT *
FROM #dataset
ORDER BY ID, ContinuousDates

Answer 1

这是一种空岛问题。

有许多不同的解决方案。这是一个简单的

使用 LAG 来识别每个岛的起始行
运行条件计数为我们提供了每个岛屿的 ID
然后简单地按该 ID 分组（连同任何其他分区列）

WITH StartPoints AS (
    SELECT *,
      IsStart = CASE WHEN LAG(City, 1, '') OVER (PARTITION BY ID ORDER BY ContinuousDates)
                         <> City THEN 1 END
    FROM #dataset ds
),
Groups AS (
    SELECT *,
      GroupId = COUNT(IsStart) OVER (PARTITION BY ID ORDER BY ContinuousDates ROWS UNBOUNDED PRECEDING)
    FROM StartPoints
)
SELECT
  ID,
  Name,
  City = MIN(City),
  DateStart = MIN(ContinuousDates),
  DateEnd = MAX(ContinuousDates)
FROM Groups
GROUP BY
  ID,
  Name,
  GroupId;

db<>fiddle

Answer 2

解决步骤：

ID 和名称按日期排序的编号部分 (row_id)
包含 ID、姓名和城市的数字部分按日期排序 (p_row_id)
计算row_id - p_row_id

现在您在一组唯一的值中拥有每个时期的组号。

您只需按此号码、ID、姓名和城市分组

ID	Name	City	ContinuousDates	p_row_id	row_id	row_id - p_row_id
1	X	NYC	2021-01-01	1	1	0
1	X	NYC	2021-01-02	2	2	0
1	X	NYC	2021-01-03	3	3	0
1	X	SFO	2021-01-04	1	4	3
1	X	SFO	2021-01-05	2	5	3
1	X	PHY	2021-01-06	1	6	5
1	X	PHY	2021-01-07	2	7	5
1	X	PHY	2021-01-08	3	8	5
1	X	NYC	2021-01-09	4	9	5
1	X	NYC	2021-01-10	5	10	5
1	X	NYC	2021-01-11	6	11	5
1	X	NYC	2021-01-12	7	12	5

select
     CD.ID
    ,CD.[Name]
    ,CD.City
    ,min(CD.ContinuousDates) as DateStart
    ,max(CD.ContinuousDates) as DateEnd
from
    (
        select *
            ,row_number() over(partition by CD.ID, CD.[Name], CD.City order by CD.ContinuousDates) as p_row_id
            ,row_number() over(partition by CD.ID, CD.[Name] order by CD.ContinuousDates) as row_id
        from #dataset CD
    ) CD
group by CD.row_id - CD.p_row_id
        ,CD.ID
        ,CD.[Name]
        ,CD.City
order by DateStart

多列模板：

select
     CD.GroupColumn1
    ,CD.GroupColumn2
    ..
    ,CD.Column1
    ,CD.Column2
    ,CD.Column3
    ,CD.Column4
    ..
    ,min(CD.ContinuousDates) as DateStart
    ,max(CD.ContinuousDates) as DateEnd
from
    (
        select *
            ,row_number() over(partition by
                 CD.GroupColumn1
                ,CD.GroupColumn2
                ..
                ,CD.Column1
                ,CD.Column2
                ,CD.Column3
                ,CD.Column4
                ..
                order by CD.ContinuousDates) as p_row_id
            ,row_number() over(partition by
                 CD.GroupColumn1
                ,CD.GroupColumn2
                ..
                order by CD.ContinuousDates) as row_id
        from #dataset CD
    ) CD
group by CD.row_id - CD.p_row_id
        ,CD.GroupColumn1
        ,CD.GroupColumn2
        ..
         CD.Column1
        ,CD.Column2
        ,CD.Column3
        ,CD.Column4
        ..
order by DateStart

Microsoft SQL Server 2016，T-SQL：根据各个日期获取数据集的日期范围

Microsoft SQL Server 2016, T-SQL : obtain date range for a dataset based on individual dates

tsql

sql-server

date

gaps-and-islands

sql-server-2016