需要为相同的产品分组选择具有重叠日期范围的记录答案

【问题标题】：Need to select records that have overlapping date ranges for same product groupings需要为相同的产品分组选择具有重叠日期范围的记录
【发布时间】：2017-06-15 16:17:28
【问题描述】：

create table #productcov  
(  
[group] varchar(20), -- as business  
account varchar(10), -- as business location  
member  varchar(10), -- as member  
product varchar(10), -- as health product  
[plan]  varchar(10), -- as product type  
StartTime datetime,  -- as beginning of coverage period  
EndTime datetime     -- as end of coverage period  
) 
go  

insert into #productcov values 

('NN00001','10001','1234567890','ADD','A1000','2010-10-01 00:00:00' , '2012-09-30 00:00:00'),  
('NN00001','10001','1234567890','ADD','A1000','2012-10-01 00:00:00' , '2013-09-30 00:00:00'),  
('NN00001','10001','1234567890','ADD','A1000','2013-01-01 00:00:00' , '2013-12-31 00:00:00'),   
('NN00001','10001','1234567890','ADD','B1000','2010-01-01 00:00:00' , '2014-12-31 00:00:00'),    
('NN00001','10001','1234567890','ADD','B1000','2015-01-01 00:00:00' , '2016-12-31 00:00:00'),  
('NN00001','10001','1234567890','LIF','A1000','2010-10-01 00:00:00' , '2012-09-30 00:00:00'),  
('NN00001','10001','1234567890','LIF','A1000','2012-10-01 00:00:00' , '2013-09-30 00:00:00'),  
('NN00001','10001','1234567890','LIF','A1000','2013-01-01 00:00:00' , '2013-12-31 00:00:00'),  
('NN00001','10001','1234567890','LIF','B1000','2010-01-01 00:00:00' , '2014-12-31 00:00:00'),  
('NN00001','10001','1234567890','LIF','B1000','2015-01-01 00:00:00' , '2016-12-31 00:00:00')

select * from   #productcov

上面的示例表适用于单个组中的单个成员。我试图从#productcov 表中的十条记录中仅选择第二条和第三条以及第七条和第八条记录，其中医疗产品的承保日期与成员的产品重叠。这是一个审计过程，用于清理覆盖日期中的重叠数据。表中重要的重叠日期需要按组、帐户、成员、产品和计划进行分组。

我几天来一直在尝试寻找解决方案，但结果一无所获！任何帮助表示赞赏。

所需的输出将是下面列出的 4 条记录：

组、帐户、成员、产品、计划、开始日期、结束日期

'NN00001','10001','1234567890','ADD','A1000','2012-10-01 00:00:00','2013-09-30 00:00:00',

'NN00001','10001','1234567890','ADD','A1000','2013-01-01 00:00:00','2013-12-31 00:00:00'

'NN00001','10001','1234567890','LIF','A1000','2012-10-01 00:00:00','2013-09-30 00:00:00'

'NN00001','10001','1234567890','LIF','A1000','2013-01-01 00:00:00','2013-12-31 00:00:00'

【问题讨论】：

标签： sql date datetime select overlap

【解决方案1】：

您可以使用exists：

select pc.*
from #productcov pc
where exists (select 1
              from #productcov pc2
              where pc2.[group] = pc.[group] and  -- don't know if this is needed
                    pc2.member = pc.member and
                    pc2.product = pc.product and
                    pc.endtime > pc2.starttime and pc.starttime < pc2.endtime
             ) ;

注意：您可能需要将> 调整为>= 和< 调整为<=，具体取决于您的重叠规则。

【讨论】：

我尝试使用存在并使用了与您提供的非常相似的 select 语句，但使用您的 select 语句得到的结果与我自己尝试时得到的结果相同。您提供的 select 语句的结果产生了 20 行，即重复两次的 10 条记录。我试图只获得重叠的 4 条记录。感谢您的帮助。

【解决方案2】：

我终于能够为我们的生产数据库自己解决这个审计问题了！我在一个大型查询中使用了多个 Common Table Expressions、Dense_Ranks、Row_Numbers 来解决！稍后将发布测试数据查询的结果。

with cte as ( 
   select distinct 
   DENSE_RANK() OVER (ORDER BY CONCAT(rtrim([group]),rtrim    (account),rtrim(member), rtrim(product), rtrim([plan]))) AS [Key Count],
   ROW_NUMBER() OVER (PARTITION BY CONCAT(rtrim([group]),rtrim(account), rtrim(member), rtrim(product), rtrim([plan])) order by CONCAT(rtrim([group]),rtrim(account), rtrim(member), rtrim(product), rtrim([plan])), StartTime) as [Overlap],
   CONCAT(DENSE_RANK() OVER (ORDER BY CONCAT(rtrim([group]),rtrim(account), rtrim(member), rtrim(product), rtrim([plan]))), ROW_NUMBER() OVER (PARTITION BY CONCAT(rtrim([group]),rtrim(account), rtrim(member), rtrim(product), rtrim([plan])) order by CONCAT(rtrim([group]),rtrim(account), rtrim(member), rtrim(product), rtrim([plan])), StartTime)) as Identifier,
   CONCAT(rtrim([group]),rtrim(account),rtrim(member), rtrim(product), rtrim([plan])) as [Key],
   StartTime as [Eff Date],
   case when EndTime = 0 then '29991231' else EndTime end as [Term Date],
   [group],
   account,
   member,
   product,
   [plan]
from   #productcov
where  [group] like 'N%'), 

cte2 as (
   select distinct 
   DENSE_RANK() OVER (ORDER BY CONCAT(rtrim([group]),rtrim(account),rtrim(member), rtrim(product), rtrim([plan]))) AS [Key Count],
   ROW_NUMBER() OVER (PARTITION BY CONCAT(rtrim([group]),rtrim(account), rtrim(member), rtrim(product), rtrim([plan])) order by CONCAT(rtrim([group]),rtrim(account), rtrim(member), rtrim(product), rtrim([plan])), StartTime) as [Overlap],
   CONCAT(DENSE_RANK() OVER (ORDER BY CONCAT(rtrim([group]),rtrim(account), rtrim(member), rtrim(product), rtrim([plan]))), ROW_NUMBER() OVER (PARTITION BY CONCAT(rtrim([group]),rtrim(account), rtrim(member), rtrim(product), rtrim([plan])) order by CONCAT(rtrim([group]),rtrim(account), rtrim(member), rtrim(product), rtrim([plan])), StartTime)) as Identifier,
   CONCAT(rtrim([group]),rtrim(account),rtrim(member), rtrim(product), rtrim([plan])) as [Key],
   StartTime as [Eff Date],
   case when EndTime = 0 then '29991231' else EndTime end as [Term Date],
   [group],
   account,
   member,
   product,
   [plan]
from   #productcov
where  [group] like 'N%')

select distinct 
            cte2.[Group], 
            cte2.[Account], 
            cte2.[Member],
            cte2.[Product],
            cte2.[Plan]
from   cte
left outer join cte2 on cte.[Key] = cte2.[Key]
where  (cte2.[Eff Date] between cte.[Eff Date] and cte.[Term Date]
or      cte2.[Term Date] between cte.[Eff Date] and cte.[Term Date])
and    cte.[Key Count] = cte2.[Key Count]
and    cte.[Identifier] <> cte2.[Identifier]
group by cte2.[Group], cte2.[Account], cte2.[Member], cte2.[Product], cte2.[Plan]
order by cte2.[Group], cte2.[Account], cte2.[Member], cte2.[Product], cte2.[Plan];

【讨论】：