检查查询是否包含 oracle 中子查询的所有值

Question

假设您有一个查询，并且您只想显示具有子查询所有值的那个。例如，我们有下表：

CREATE TABLE test
(
code VARCHAR2(4),
year VARCHAR2(4),
action VARCHAR2(50),
CONSTRAINT pk PRIMARY KEY (code, year)
);

还有以下寄存器：

INSERT INTO test
VALUES ('1','2020','Departure');
INSERT INTO test
VALUES ('1','2021','Arrival');

INSERT INTO test
VALUES ('2','2020','Departure');

想象一个子查询返回以下值：

('Departure','Arrival')

所以我想做一个查询，它只返回与子查询中返回的两个值相匹配的代码和年份。查看寄存器，它应该只返回 return ('1','2020') 和 ('1','2021') 因为它们是唯一的动作是 'Arrival' 和 'Departure' 的。我该怎么办？

Answer 1

将您的查询包含在这样的 CTE 中：

with cte as (
  <your query here>
)
select t.*
from test t 
where
  t.action in (select action from cte)
  and 
  code in (
    select code
    from test
    where action in (select action from cte)
    group by code
    having count(distinct action) = (select count(*) from cte)
  ) 

IN 的子查询返回包含查询返回的所有操作的所有代码。
请参阅demo。
结果：

> CODE | YEAR | ACTION   
> :--- | :--- | :--------
> 1    | 2021 | Arrival  
> 1    | 2020 | Departure

Answer 2

作为一般解决方案，我会寻找使用HAVING count(*) = # 的机会。大概，

SELECT code FROM table WHERE action in (SUBSELECT) GROUP BY code HAVING count(*) = (SELECT count(*) from SUBSELECT)

当然，如果单个code 可以有多个Arrival，则还必须包含DISTINCT。大概，

SELECT code FROM (SELECT code, distinct(action) FROM table) WHERE action in...

我担心这样的查询的性能，但您必须在现场检查它，因为大多数数据库引擎都能够转换这样的复杂查询以提高效率。

===

如果您确定每个代码的每种类型只有一个操作，我认为 JOIN 也可以工作。例如。

SELECT t.code FROM SUBQUERY as s RIGHT JOIN table as t on t.action = s.action GROUP by t.code HAVING count(*) = (SELECT COUNT(*) FROM SUBQUERY)

（在任何一种情况下都可以，如果要内联子查询以避免重复执行它，您应该使用 @forpas 建议的 CTE 功能）。

Answer 3

稍加扩展的样本数据，其中CODEs 1 和 3 都有 Arrival 和 Departure：

SQL> with test (code, year, action) as
  2    (select 1, 2020, 'Departure' from dual union all
  3     select 1, 2021, 'Arrival'   from dual union all
  4     select 2, 2020, 'Departure' from dual union all
  5     --
  6     select 3, 2018, 'Arrival'   from dual union all
  7     select 3, 2019, 'Departure' from dual
  8    ),
  9  subq as
 10    (select distinct action,
 11            count(distinct action) over () cnt_da
 12     from test
 13    )
 14  select a.code, a.year
 15  from test a join subq s on a.action = s.action
 16  where s.cnt_da = (select count(distinct action)
 17                    from test b
 18                    where b.code = a.code
 19                   );

      CODE       YEAR
---------- ----------
         1       2021
         1       2020
         3       2019
         3       2018

SQL>

---

另一个选项，使用MINUS 集合运算符：

SQL> with test (code, year, action) as
  2    (select 1, 2020, 'Departure' from dual union all
  3     select 1, 2021, 'Arrival'   from dual union all
  4     select 2, 2020, 'Departure' from dual union all
  5     --
  6     select 3, 2018, 'Arrival'   from dual union all
  7     select 3, 2019, 'Departure' from dual
  8    ),
  9  subq as
 10    (select distinct action from test)       --> this is your "subquery"
 11  select code, year
 12  from test a
 13  where (select s.action from subq s
 14         minus
 15         select b.action from test b where b.code = a.code
 16        ) is null;

      CODE       YEAR
---------- ----------
         1       2020
         1       2021
         3       2018
         3       2019

SQL>