Question

我有一张桌子

CREATE TABLE #tblA (mem_id int, type varchar(20), address1 varchar(20),group_id int)

insert into #tblA (mem_id, type, address1,group_id)
values (1,'self','abc St',1),
       (2,'Child','abc St',1),
       (3,'Child','xyz st',1),
       (4,'spouse','pqr st',1),
       (5,'Child','abc St',1),
       (6,'Child','xyz st',1),

       (7,'self','mno st',2),
       (8,'Child','def St',2),
       (9,'Child','def st',2),

       (10,'self','loi st',3),
       (11,'Child','loi St',3),
       (12,'Child','ewr st',3)

       (13,'self','ooo st',NULL),

所以，我想选择：

对于每个组，请选择“self”。
如果没有组，请选择自己。
如果其他人住在与自己不同的地址，请选择这些成员。
如果多个成员住在不同的地址，请选择1名成员在不同的地址

预期结果：

   (1,'self','abc St',1),
   (3,'Child','xyz st',1),
   (4,'spouse','pqr st',1),
   (7,'self','mno st',2),
   (8,'Child','def St',2),
   (10,'self','loi st',3),
   (12,'Child','ewr st',3)
   (13,'self','ooo st',NULL),

由于

我的代码：不工作：

select  
   mem_id
from 
   (select 
        a.*,
        RANK() over (partition by group_id order by AddressCnt DESC) as AddressRank
    from 
        (select 
             a.*,
             (case when max(address1) over (partition by group_id) = min(address1) over (partition by group_id)
              then 1 else 0
              end) as AddressSame,
             count(*) over (partition by group_id, address1) as AddressCnt
         from #tblA a) a
   ) a
where 
   (AddressSame = 1 or type in ('Self') or
    (AddressRank > 1 OR  type in ('Self') or group_id is null

Answer 1

这是一个解决方案。

创建一个识别“自我”记录的CTE。
使用从self到tableA的连接创建另一个标识“其他”的cte
使用row_number只查找其他
联合两个cte的

WITH self 
     AS (SELECT * 
         FROM   tbla 
         WHERE  type = 'self'), 
     others 
     AS (SELECT a.mem_id, 
                a.type, 
                a.address1, 
                a.group_id, 
                Row_number() 
                  OVER ( 
                    partition BY a.group_id, a.address1 
                    ORDER BY a.mem_id) rn 
         FROM   self s 
                INNER JOIN tbla a 
                        ON s.group_id = a.group_id 
                           AND s.mem_id <> a.mem_id --Exclude self records
                           AND s.address1 <> a.address1 --Exclude the same address as self) 
SELECT mem_id, 
       type, 
       address1, 
       group_id 
FROM   self 
UNION 
SELECT mem_id, 
       type, 
       address1, 
       group_id 
FROM   others 
WHERE  rn = 1 
ORDER  BY mem_id

DEMO

Answer 2

此MySQL查询提供了所需的输出：

从tblA中选择*，其中type ='self'UNION select * from tblA group by address1，address1 not in in（select address1 from tblA where type ='self'）ORDER by mem_id;

SQL Server 2008：根据条件选择成员组

2 个答案: