按 SQL 中两列值的计数过滤行

Question

在 SQL 服务器中，我有下表：

CREATE TABLE x
(
    ColumnA CHAR NOT NULL
  , ColumnB CHAR NOT NULL
  , ColumnC INT NOT NULL
  , CONSTRAINT CompositeKey PRIMARY KEY (ColumnA, ColumnB)
);

使用以下数据：

INSERT INTO x
VALUES
 ('A', 'X', 2)
,('A', 'Y', 1)
,('B', 'X', 4)
,('B', 'Z', 2)
,('C', 'X', 9)
,('C', 'Y', 3)
,('C', 'P', 2)
,('D', 'X', 6)
,('D', 'Y', 4)
,('E', 'P', 2)

我正在使用以下 2 个谓词编写查询：

1. Count(ColumnA) > 1
2. Count(ColumnB) > 2

我的查询

SELECT *
FROM
(
    SELECT *
          ,COUNT(*) OVER (PARTITION BY [ColumnA]) AS columnA_cnt
          ,COUNT(*) OVER (PARTITION BY [ColumnB]) AS columnB_cnt
    FROM @DataSource
) DS
WHERE columnA_cnt >= 2
  AND columnB_cnt >= 3

预期结果

一个	乙	C
一个	X	2
一个	是的	1
C	X	9
C	是的	3
D	X	6
D	是的	4

实际结果

一个	乙	C
一个	X	2
一个	是的	1
乙	X	4
C	X	9
C	是的	3
D	X	6
D	是的	4

Answer 1

(ColumnA, ColumnB)是主键。 因此

COUNT([ColumnA]) OVER (PARTITION BY [ColumnB]) AS columnA_cnt

是相同的

COUNT(*) OVER (PARTITION BY [ColumnB]) AS columnA_cnt

您查看一行的ColumnB并计算它在表中出现的次数。 因此，比columnA_cnt更好的名称可能是columnB_cnt ;-)

SELECT *
FROM
(
    SELECT *
          ,COUNT(*) OVER (PARTITION BY [ColumnA]) AS columnA_cnt
          ,COUNT(*) OVER (PARTITION BY [ColumnB]) AS columnB_cnt
    FROM @DataSource
) DS
WHERE columnA_cnt >= 2
  AND columnB_cnt >= 3

Answer 2

此查询符合您预期结果的要求

CREATE TABLE x
(
    ColumnA CHAR NOT NULL
  , ColumnB CHAR NOT NULL
  , ColumnC INT NOT NULL
  , CONSTRAINT CompositeKey PRIMARY KEY (ColumnA, ColumnB)
);

INSERT INTO x
VALUES
 ('A', 'X', 2)
,('A', 'Y', 1)
,('B', 'X', 4)
,('B', 'Z', 2)
,('C', 'X', 9)
,('C', 'Y', 3)
,('C', 'P', 2)
,('D', 'X', 6)
,('D', 'Y', 4)
,('E', 'P', 2)

with cte as (
select * from x
where ColumnA in (select ColumnA from x group by ColumnA having count(ColumnA) >1)
and ColumnB in (select ColumnB from x group by ColumnB having count(ColumnA) >2)
)
select * from cte
where ColumnA in (select ColumnA from cte group by ColumnA having count(ColumnA) >1)
and ColumnB in (select ColumnB from cte group by ColumnB having count(ColumnA) >2)

Answer 3

这对你有用吗（第二个条件columnA > 1 ， columnB > 2 ）？：

select 
  ColumnA
  ,ColumnB
  ,ColumnC
from (
  select 
    ColumnA
    ,ColumnB
    ,ColumnC
    ,COUNT(ColumnA) OVER (PARTITION BY ColumnA) AS columnA_cnt
    ,COUNT(ColumnB) OVER (PARTITION BY ColumnB) AS columnB_cnt
  from (
      SELECT *
            ,COUNT(ColumnA) OVER (PARTITION BY ColumnA) AS columnA_cnt
            ,COUNT(ColumnB) OVER (PARTITION BY ColumnB) AS columnB_cnt
      FROM x
  ) ds where columnA_cnt >= 2 and columnB_cnt >= 3
) ds2 where columnA_cnt > 1 and columnB_cnt > 2
order by columnA, columnB;