Question

我的数据集如下 - 在此数据集中，缺少start_location，因为activity_type为No。所以基本上我想创建一个列，其中每个间隙的最小日期为activity_date（表示没有start_location的记录）。如果5/07/2015,6 / 07 / 2015,7 / 07/2015将start_location设为null，那么新列即new_activity_date必须是05/07/2015的第一个差距日期。

基本上这是找到差距期的第一个日期。数据如下 -

id  activity_type   activity_date   start_location  end_location
27151   Yes 4/07/2015   18  18
27151   No  5/07/2015       
27151   No  6/07/2015       
27151   Yes 7/07/2015   18  17
27151   Yes 8/07/2015   18  17
27151   Yes 9/07/2015   18  17
27151   Others  19/07/2015  17  17
27151   Others  20/07/2015  17  17
27151   No  21/07/2015      
27151   No  22/07/2015      
27151   No  23/07/2015      
27151   Yes 24/07/2015  17  17
27151   Yes 25/07/2015  17  17
27151   Yes 26/07/2015  17  17
27151   Yes 27/07/2015  17  17

我的数据应该是 -

id  activity_type activity_date start_location  end_location new_activity_date
27151   Yes 4/07/2015   18  18  4/07/2015
27151   No  5/07/2015           4/07/2015
27151   No  6/07/2015           4/07/2015
27151   Yes 7/07/2015   18  17  7/07/2015
27151   Yes 8/07/2015   18  17  8/07/2015
27151   Yes 9/07/2015   18  17  9/07/2015
27151   Others  19/07/2015  17  17  19/07/2015
27151   Others  20/07/2015  17  17  20/07/2015
27151   No  21/07/2015          20/07/2015
27151   No  22/07/2015          20/07/2015
27151   No  23/07/2015          20/07/2015
27151   Yes 24/07/2015  17  17  24/07/2015
27151   Yes 25/07/2015  17  17  25/07/2015
27151   Yes 26/07/2015  17  17  26/07/2015
27151   Yes 27/07/2015  17  17  27/07/2015

先谢谢，不知道我在哪里误解了这个伎俩。

Answer 1

我想我知道你想要什么。我认为函数最有效，然后你可以在插入或更新中使用它：

create function dbo.get_New_Activity_Date(@Activity_Date datetime)
returns datetime
As
Begin 
  declare @result datetime
  set @result=
   (select min(activity_date)
    from myTable
    where activity_date<=@activity_Date
    and activity_date>
        (select max(activity_date)
           from myTable t2 
           where t2.start_location is not null -- if using null else test for blank
           and t2.activity_date <=@activity_date
        )
    )
  return isnull(@result,@activity_date)  
end

如果您只是想要一个select语句在不使用函数的情况下将此表提取为ETL作业的一部分：

select 
  id
  ,activity_type
  ,activity_date
  ,start_location
  ,end_location
  ,isnull((select min(t1.activity_date)
    from myTable t1
        where t1.id=myTable.id
        and t1.activity_date<=myTable.activity_Date
        and t1.activity_date>
    (select max(t2.activity_date)
       from myTable t2 
       where t2.id=myTable.id
       and t2.start_location is not null -- if using null else test for blank
       and t2.activity_date <=myTable.activity_date
    )
),activity_date) new_activity_date
from
  myTable

我的结果：

Answer 2

嗯。这对我来说似乎是条件累积最大值：

select t.*,
       max(case when activity_type <> 'No' then activity_date end) over (order by id, date) as new_activity_date
from t;

查找差距记录范围的第一个日期

2 个答案: