Question

我想按客户分组日期并为2个单独的值生成计数列（Flag = Y和Flag = N）。输入表如下所示：

Customer  Date   Flag
------- ------- -----
001      201201  Y
001      201202  Y
001      201203  Y
001      201204  N
001      201205  N
001      201206  Y
001      201207  Y
001      201208  Y
001      201209  N
002         201201  N
002         201202  Y
002         201203  Y
002         201205  N

输出应如下所示：

Customer MinDate  MaxDate Count_Y
------- ------ ------- -------
001     201201  201203     3  
001     201206  201208     3     
002     201202  201203     2

如何编写SQL查询？任何形式的帮助表示赞赏！谢谢！

Answer 1

您想要找到＆＃34; Y＆＃34;的连续值。这是一个＆＃34;缺口和岛屿＆＃34;问题，有两种基本方法：

确定第一个＆＃34; Y＆＃34;在每个组中并使用此信息来定义一组连续的＆＃34; Y＆＃34;值。
使用row_number()值的差异进行计算。

第一个依赖于SQL Server 2012+并且您还没有指定版本。所以，第二个看起来像这样：

select customer, min(date) as mindate, max(date) as maxdate,
       count(*) as numYs
from (select t.*,
             row_number() over (partition by customer order by date) as seqnum_cd,
             row_number() over (partition by customer, flag order by date) as seqnum_cfd
      from t
     ) t
where flag = 'Y'
group by customer, (seqnum_cd - seqnum_cfd), flag;

解释这是如何工作有点棘手。根据我的经验，想一想，如果你运行子查询，你会看到如何计算seqnum列并且＆＃34;得到它＆＃34;通过观察结果。

注意：这假设每天最多只有一条记录。如果还有更多内容，您可以使用dense_rank()代替row_number()来获得相同的效果。

Answer 2

尝试使用以下查询，它将为您提供您想要的内容。

DROP TABLE [GroupCustomer]
GO

CREATE TABLE [dbo].[GroupCustomer](
     Customer VARCHAR(50),
     [Date] [datetime] NULL,
     Flag VARCHAR(1)
       )

INSERT INTO [dbo].[GroupCustomer]  (Customer ,[Date],Flag)
VALUES   ('001','201201','Y'),('001','201202','Y'),
         ('001','201203','Y'),('001','201204','N'),
         ('001','201205','N'),('001','201206','Y'),
         ('001','201207','Y'),('001','201208','Y'),
         ('001','201209','N'),('002','201201','N'),
         ('002','201202','Y'),('002','201203','Y'),
         ('002','201205','N')
GO


;WITH cte_cnt
AS
(
 SELECT Customer,Format(MIN([Date]),'yyMMdd') AS MinDate
   ,Format(MAX([Date]),'yyMMdd') AS MaxDate
   , COUNT('A') AS Count_Y
 FROM (
     SELECT Customer,Flag,[Date],
        ROW_NUMBER() OVER(Partition by customer ORDER BY [Date]) AS ROW_NUMBER,
        DATEDIFF(D, ROW_NUMBER() OVER(Partition by customer ORDER BY [Date])
        , [Date]) AS Diff
    FROM [GroupCustomer]
    WHERE Flag='Y') AS dt
   GROUP BY Customer,Flag, Diff )
SELECT *
FROM  cte_cnt  c
ORDER BY Customer

GO

SQL组由数据与行分开

2 个答案: