问题描述
我有一个sql Server表,如下所示。我想根据上述分组,按名称和参加考试的地点分组,按日期升序作为分区。
现在提供了例如4天的可配置窗口。在下表中,如果第一次考试的日期是 02/01/2019(2月1日)-已取得其分数,并且在接下来的4天之内已重新获得的任何其他测试分数将不予考虑。如果记录也位于已经排除的项目示例行ID-4的4天窗口内,则该记录也应排除在外。
对此逻辑的任何sql语句都表示赞赏。
CREATE TABLE test(
[recordid] int IDENTITY(1,1) PRIMARY KEY,[name] [nvarchar](25) NULL,[testcentre] [nvarchar](25) NULL,[testdate] [smalldatetime] NOT NULL,[testscore] [int],[Preferred_Output] [int],[Result] [nvarchar](75) NULL
)
GO
INSERT INTO test
(
[name],[testcentre],[testdate],[testscore],[Preferred_Output],[Result] )
VALUES
('George','bangalore',' 02/01/2019',1,'Selected as first item -grouped by name and location'),('George',' 02/02/2019','ignore as within 4 days'),' 02/04/2019',' 02/06/2019',3,'ignore as within 4 days from already ignored item -04-02-2019'),' 02/15/2019',2,'Selected as second item -grouped by name and location'),' 02/18/2019',5,'ignore as within 4 days of prevIoUs'),'Pune',4,'Selected as third item'),6,' 02/19/2019',7,' 02/20/2019',8,'ignore as within 4 days of prevIoUs')
GO
select * from test
GO
+----------+--------+------------+------------+-----------+------------------+
| recordid | name | testcentre | testdate | testscore | Preferred_Output |
+----------+--------+------------+------------+-----------+------------------+
| 1 | George | bangalore | 02/01/2019 | 1 | 1 |
| 2 | George | bangalore | 02/02/2019 | 0 | 0 |
| 3 | George | bangalore | 02/04/2019 | 1 | 0 |
| 4 | George | bangalore | 02/06/2019 | 3 | 0 |
| 5 | George | bangalore | 02/15/2019 | 2 | 2 |
| 6 | George | bangalore | 02/18/2019 | 5 | 0 |
| 7 | George | Pune | 02/15/2019 | 4 | 3 |
| 8 | George | Pune | 02/18/2019 | 6 | 0 |
| 9 | George | Pune | 02/19/2019 | 7 | 0 |
| 10 | George | Pune | 02/20/2019 | 8 | 0 |
+----------+--------+------------+------------+-----------+------------------+
解决方法
我认为不需要为此进行递归查询。您想比较连续记录中的日期,所以这是一种空白问题,需要确定每个岛的起点。
窗口函数可以做到这一点:
select t.*,case when lag_testdate is null or testdate > dateadd(day,4,lag_testdate)
then testscore
else 0
end new_core
from (
select t.*,lag(testdate) over(partition by name,testcentre order by testdate) lag_testdate
from test t
) t