SQL帮助按组和计数查找唯一对

Question

Need some help with SQL, possibly using group and count, or whatever it needs. Just could not find a way. Thanks a lot.

A simple table:

ColA   ColB
  1      A
  1      A
  2      B
  3      B
  4      C
  4      C
  5      C

Return all unique pairs of ColA and ColB, where for the same ColB there are more than one distinct ColA values.

For given data above, it shall return

ColA  ColB
  2     B
  3     B
  4     C
  5     C

Answer 1

EDIT: Apologies about the first answer, I hadn't noticed your stipulation for ColA to have more than one distinct value across ColB . Here's my updated answer:

SELECT ColA, ColB
FROM test
WHERE EXISTS (
    SELECT ColB
    FROM test AS subtest
    WHERE test.ColB = subtest.ColB
    GROUP BY ColB
    HAVING COUNT(DISTINCT ColA) > 1
)
GROUP BY ColA, ColB

Answer 2

You can use APPLY to check if ColB has more than 1 disinct ColA values:

WITH Cte(ColA, ColB) AS(
    SELECT * FROM( VALUES
        (1, 'A'), (1, 'A'), (2, 'B'), (3, 'B'), (4, 'C'), (4, 'C'), (5, 'C')
    ) t(a,b)
)
SELECT DISTINCT c1.*
FROM Cte c1
CROSS APPLY(
    SELECT COUNT(*) AS cnt
    FROM Cte c2
    WHERE
        c2.ColB = c1.ColB
        AND c2.ColA <> c1.ColA
    GROUP BY c2.ColB
    HAVING COUNT(*) > 0
) x

ONLINE DEMO

Answer 3

First using Group By Clause and Row_Number() I will get a number sequenced result set. This way I can identify unique pairs of ColumnA and ColumnB where for the same ColumnB there is more than one distinct value of ColumnA.

Select
    ColumnA,
    ColumnB,
    Row_Number() Over(Partition By ColumnB Order By ColumnA) As RowNum
From SimpleTable
Group By ColumnB, ColumnA;

Output:

ColumnA ColumnB RowNum
1       A       1
2       B       1
3       B       2
4       C       1
5       C       2

Now you can put this result in a table expression - CTE or a derived table (I'm choosing CTE) and filter out only those ColumnB values for which RowNum is greater than or equal to 2. So the final query will be --

;With CTE
As
(
    Select
        ColumnA,
        ColumnB,
        Row_Number() Over(Partition By ColumnB Order By ColumnA) As RowNum
    From SimpleTable
    Group By ColumnB, ColumnA
)
Select ColumnA, ColumnB From CTE
Where ColumnB In (Select ColumnB From CTE Where RowNum >=2)
Order By ColumnA, ColumnB;

Final output:

ColumnA ColumnB
2       B
3       B
4       C
5       C

Hope this is helpful :)

Answer 4

I would suggest a simple approach using min() and max() as window functions:

select colA, colB
from (select t.*,
             min(colA) over (partition by colB) as mincolA,
             max(colA) over (partition by colB) as maxcolA
      from t
     ) t
where mincolA <> maxcolA;