根据条件过滤掉某些行

Question

我有此输入表：

+-------------+--------------+-------+
| ProjectName | ProjectOwner | Col2  |
+-------------+--------------+-------+
| A           | Sara         | Sara  |
| B           | Jack         | Maria |
| B           | Jack         | Jack  |
| C           | Laura        | May   |
| C           | Laura        | Laura |
| D           | Paul         | Jack  |
| E           | May          | May   |
| E           | May          | May   |
+-------------+--------------+-------+

我想分别检查每个项目然后排除

1. 如果项目所有者的名字在col2中，并且其他人与他同在，则删除重复的行Jack Jack，

   a。例如，在Project B中，Jack在Col2中有他的名字，并且他正在与Maria一起工作，因此请删除Jack Jack

   b。还要在项目C中删除Laura Laura

2. 如果项目所有者仅自己在项目内部工作，则保留该项目，例如Project A和Project E

3. 如果项目所有者在col2中没有他的名字，则排除项目D，例如项目D

4. 删除重复项，例如Project E

输出表：

+-------------+--------------+-------+
| ProjectName | ProjectOwner | Col2  |
+-------------+--------------+-------+
| A           | Sara         | Sara  |
| B           | Jack         | Maria |
| C           | Laura        | May   |
| E           | May          | May   | 
+-------------+--------------+-------+

我尝试过此here，但结果没有向我显示Project A和E。

Answer 1

请尝试此代码：

select distinct  *
from mytable m
where (ProjectOwner!=Col2 or not exists(select 1 from mytable m1 where m1.ProjectName=m.ProjectName and  m1.ProjectOwner=m.ProjectOwner and m.Col2!=m1.Col2) )
and ProjectName in (select ProjectName from mytable where ProjectOwner=Col2 )

Demo

Answer 2

您可以通过使用GROUPING和COUNT聚合来通过此嵌套SQL获得所需的结果：

select ProjectName, ProjectOwner, Col2
  from
(  
   select ProjectName, ProjectOwner, Col2,
          count(1) over ( partition by ProjectName ) cnt
     from mytable
    group by ProjectName, ProjectOwner, Col2
) q
 where ( q.cnt = 1 or ProjectOwner != col2 )
   and ProjectOwner in ( select col2 from mytable );

SQL Fiddle Demo

Answer 3

使用EXISTS和ROW_NUMBER()功能：

select top (1) with ties *
from table t
where exists (select 1 
              from table t1 
              where t1.ProjectName = t.ProjectName and 
                    t1.Col2 = t.ProjectOwner
             )
order by dense_rank() over (partition by ProjectName 
                            order by (case when t.ProjectOwner = Col2 
                                           then 1 else 0 
                                       end)
                           );