SQL Server：基于一个列进行分组，并根据另一列获取列

Question

My data:

ColumnA                 Column B            ColumnC           
A                          Hi                  Yes
A                          Hello               NULL
B                          Hola                Yes
B                          Hola                Yes
B                          Hi                  NULL
C                          ABCD                Yes

My SQL should be able to do following:

1. Group by Column A
2. Select only those groups of A that contain Yes and Null both in the result of a

Expected Result is:

ColumnA                 Column B            ColumnC           
A                          Hi                  Yes
A                          Hello               NULL
B                          Hola                Yes
B                          Hola                Yes
B                          Hi                  NULL

How do I select all values that satisfy above rule?

Answer 1

One method uses GROUP BY and HAVING :

select columnA
from t
group by columnA
having sum(case when columnC is null then 1 else 0 end) > 0 and
       sum(case when columnC = 'yes' then 1 else 0 end) > 0;

Answer 2

    select * from t 
    where 
columnA in 
    (select columnA from T where columnC is NULL) and 
columnA in 
    (select columnA from T where columnC = 'Yes') 

Answer 3

I think the filtering would actually be better handled by the 'WHERE' clause instead of 'HAVING'.
True the original request stated "Select only those groups of A that contain Yes and Null both in the result of a" BUT no aggr is needed to limit the rows. I don't know if there is an additional cost to using the SUM function but why do the additional work.
I don't know if it is a far comparison but why would you go to the store, pick up a bunch of apples and when you get to the register, start pulling some of the out that are going bad?

Try this:

SELECT ColumnA, ColumnB, ColumnC
FROM myTable
WHERE ColumnC IS NULL || ColumnC = 'yes'
--GROUP BY ColumnA, ColumnB, ColumnC --commented per additional comment below.

One more note. Please double check your sample data, if you are grouping I cant tell as all you did was exclude the row with 'c' as the value for columnA. Also, I would assume that the two rows:

ColumnA                 Column B            ColumnC
B                          Hola                Yes

would have been only showing as one as they would have been grouped together, but in you sample output you have it there twice...

Answer 4

create table #t(ColumnA varchar(10),ColumnB varchar(10),ColumnC varchar(10))
insert into #t
    select 'A','Hi','Yes' union all
    select 'A','Hello',NULL union all
    select 'B','Hola','Yes' union all
    select 'B','Hola','Yes' union all
    select 'B','Hi',NULL union all
    select 'C','ABCD','Yes'
 select * from ( 
     select *,sum(case when ColumnC='YES' THEN 1 else 0 end)over(partition by ColumnA) as YesCount
             ,sum(case when ColumnC is null then 1 else 0 end)over(partition by ColumnA) as NULLCount 
     from #t
) as t where t.YesCount>0 and t.NULLCount>0

ColumnA ColumnB ColumnC YesCount    NULLCount
1   A   Hi  Yes 1   1
2   A   Hello   NULL    1   1
3   B   Hola    Yes 2   1
4   B   Hola    Yes 2   1
5   B   Hi  NULL    2   1

Answer 5

try this without group by and having cluase,

create table #t(ColumnA varchar(10),ColumnB varchar(10),ColumnC varchar(10))
insert into #t
    select 'A','Hi','Yes' union all
    select 'A','Hello',NULL union all
    select 'B','Hola','Yes' union all
    select 'B','Hola','Yes' union all
    select 'B','Hi',NULL union all
    select 'C','ABCD','Yes'

    ;With CTE as
    (
    select * 
    ,ROW_NUMBER()over(partition by ColumnA order by ColumnA)rn
    from #t
    where ColumnC is null or ColumnC='Yes'
    )
    select * from cte c
    where exists(select columnA from cte c1 
    where c.ColumnA=c1.ColumnA and rn>1)

    drop table #t

Answer 6

select t1.* 
from t t1 
inner join 
    (select sum(case when t3.colc is null then 1  when t3.colc = 'Yes' then 1 else 0 end) as cnt, 
     t3.cola  
     from t t3 
     group by cola) t2 on t1.cola = t2.cola 
     where t2.cnt >= 2;