Question

我有一组数据有2个字段“ActivationTime”，它是DATETIME＆amp;的类型。 “得分”是INT的类型。

我想根据这条规则对所有数据进行分组：

（1）每组中的数据必须相互间隔最多5天（对应于“ActivationTime”值。请注意，第一个“ActivationTime”是数据集中的最新“ActivationTime”。例如，在下面样品首先“ActivationTime”将是'2017-11-24 09：45：00.000'，其得分= 1）

（2）分组数据的每个部分中的项目必须按“分数”值从“数据的其他分组部分”“单独”排序

如果有可能，有人可以帮助我吗？

编辑：例如想象@ExampleTabel如下：

Score                  ActivationTime
-------------------------------------------
  9               '2017-11-20 09:37:00.000'
  5               '2017-11-21 09:39:00.000'
  4               '2017-11-22 09:40:00.000'
  3               '2017-11-23 09:43:00.000'
  1               '2017-11-24 09:45:00.000'
  12              '2017-11-22 09:49:00.000'
  7               '2017-11-17 09:38:00.000'
  2               '2017-11-15 09:41:00.000'
  0               '2017-11-14 09:44:00.000'
  15              '2017-11-13 09:47:00.000'
  7               '2017-11-12 09:46:00.000'
  20              '2017-11-11 09:42:00.000'
  4               '2017-11-10 09:78:00.000'

除了生成之外的结果：

Score                 ActivationTime             Category
---------------------------------------------------------------
  12             '2017-11-20 09:49:00.000'          1
  9              '2017-11-20 09:37:00.000'          1
  5              '2017-11-21 09:39:00.000'          1
  4              '2017-11-22 09:40:00.000'          1
  3              '2017-11-23 09:43:00.000'          1
  1              '2017-11-24 09:45:00.000'          1

  15             '2017-11-13 09:47:00.000'          2
  7              '2017-11-17 09:38:00.000'          2
  2              '2017-11-15 09:41:00.000'          2
  0              '2017-11-14 09:44:00.000'          2

  20             '2017-11-11 09:42:00.000'          3
  7              '2017-11-12 09:46:00.000'          3
  3              '2017-11-20 09:37:57.570'          3
  4              '2017-11-10 09:78:00.000'          3

Answer 1

从你所说的话来看，你似乎想要这样的东西。

select
Score, ActivationTime, (rnk - 1) / 5 grp
from (
    select
        Score, ActivationTime, dense_rank()over(order by cast(ActivationTime as date) desc) rnk
from
    mytable
) t
order by grp, Score desc

您应该提供一些示例数据以及您希望获得更好帮助的结果

Answer 2

根据您的评论，这是您在N Days返回时的方式。

首先，我创建一个虚拟表，其中包含我想要返回多少天的数字。这个表然后在我的光标中循环以创建我的CTE语句。最后我联合我所有的CTE，然后通过激活时间和分数来订购

/*SCRIPT */

DECLARE @NDaysAgo nvarchar(2) = 20 --Change this based on how many days you want to go back
DECLARE @Number nvarchar(2)
DECLARE @sqlCreateNumbers nvarchar(max)
DECLARE @AliasName nvarchar(50)
DECLARE @SQL nvarchar(max)
DECLARE @SQLUnion nvarchar(max)
 DECLARE @crlf nvarchar(max)
 SET @crlf = CHAR(13)
DECLARE @SQLTotal nvarchar(MAX)

Set @sqlCreateNumbers = '

IF EXISTS (SELECT * FROM sys.objects 
WHERE object_id = OBJECT_ID(N''[dbo].[numbers]'') AND type in (N''U''))
DROP TABLE [dbo].[numbers];

SELECT TOP '+@NDaysAgo+' IDENTITY(int,0,1) AS n 
into dbo.numbers
FROM MASTER..spt_values a, MASTER..spt_values b;

CREATE UNIQUE CLUSTERED INDEX idx_1 ON numbers(n);


'
PRINT @sqlCreateNumbers

EXEC (@sqlCreateNumbers)


DECLARE SqlCur CURSOR FOR
Select DaysAgo = N,Alias = 'Today'+cast(N as nvarchar(50))
from dbo.Numbers
OPEN SqlCur
FETCH NEXT FROM SqlCur
INTO @Number,@AliasName
IF (@@FETCH_STATUS>=0)
BEGIN
SET @SQL = 'with '+@AliasName+' as (
Select '''+@AliasName+''' as SourceName,Score,ActivationTime
from dbo.testactivation
Where cast(ActivationTime as date) = cast(dateadd(dd,'+@Number+',GETDATE()) as date)

)'

SET @SQLUnion = 'Select * from '+@AliasName+''

--PRINT @SQL
FETCH NEXT FROM SqlCur INTO @Number,@AliasName
 END
 WHILE (@@FETCH_STATUS<>-1)

 BEGIN
 IF (@@FETCH_STATUS<>-2)

 SET @SQL = @SQL + ','+@AliasName+' as (
Select '''+@AliasName+''' as SourceName,Score,ActivationTime
from dbo.testactivation
Where cast(ActivationTime as date) = cast(dateadd(dd,-'+@Number+',GETDATE()) as date)

)'
SET @SQLUnion =  @SQLUnion + @crlf+ ' union all Select * from '+@AliasName+''
--PRINT @SQL

FETCH NEXT FROM SqlCur INTO @Number,@AliasName
END

 CLOSE SqlCur
 DEALLOCATE SqlCur


SET @SQLTotal = @SQL + 'Select * from ( '+@SQLUnion+ ' ) x ORDER BY ACTIVATIONTIME DESC, SCORE DESC'

PRINT @SQLTotal --This will not show you all records since there is a limit to how much text it can show
Select @SQLTotal -- If you select it you can see the query

EXEC(@SQLTotal) --This will execute your statement, so you can see your descired result

修改如果您无法使用变量而无法获得NDays，那么您必须对其进行硬编码。

使用一个查询

Select 'Data' as SourceName,Score,ActivationTime from dbo.testactivation Where cast(ActivationTime as date) between cast(dateadd(dd,0,GETDATE()) as date) and cast(dateadd(dd,-20,GETDATE()) as date) order by ActivationTime desc,score desc

<强>结果

将每个部分中的一组数据和订单项分组，与T SQL中的其他部分分开

2 个答案: