Question

我不明白为什么在这种情况下列correct和incorrect具有不同的值。老实说-这似乎很奇怪，因为它是有效的SQL，因为cross apply部分不是由于group by而计算出的聚合函数。

with Foo as (
    select 1 as id, 'foo_1' as txt from Dual union all
    select 2, 'foo_2' from Dual),
Bar as (
    select 1 as id, 'bar_1' as txt, 1 as foo_id from Dual union all
    select 2, 'bar_2', 1 from Dual union all
    select 3, 'bar_3', 1 from Dual union all
    select 4, 'bar_4', 2 from Dual)
select 
    f.id,
    listagg(b.txt, ', ') within group (order by b.id) as correct,
    incorrect
from Foo f
join Bar b on f.id = b.foo_id
cross apply (
    select 
        listagg(txt, ', ') within group (order by id) as incorrect
    from Bar 
    where foo_id = f.id)
group by f.id
order by f.id;

结果是：

ID  CORRECT                 INCORRECT
1   bar_1, bar_2, bar_3     bar_4
2   bar_4                   bar_4

但是当我确实删除group by部分并添加distinct时，如下所示：

with Foo as (
    select 1 as id, 'foo_1' as txt from Dual union all
    select 2, 'foo_2' from Dual),
Bar as (
    select 1 as id, 'bar_1' as txt, 1 as foo_id from Dual union all
    select 2, 'bar_2', 1 from Dual union all
    select 3, 'bar_3', 1 from Dual union all
    select 4, 'bar_4', 2 from Dual)
select distinct
    f.id,
    incorrect_now_correct
from Foo f
join Bar b on f.id = b.foo_id
cross apply (
    select 
        listagg(txt, ', ') within group (order by id) as incorrect_now_correct
    from Bar 
    where foo_id = f.id)
order by f.id;

我确实得到以下结果：

ID  INCORRECT_NOW_CORRECT
1   bar_1, bar_2, bar_3
2   bar_4

这是我在第一个查询的incorrect列中期望的结果。

使用group by子句时对group by子句的处理是否有所不同？直到现在，我还只是认为这更像是与子查询的智能联接。将incorrect添加到group by部分使它按应有的方式工作，没有它，其行为类似于第一个查询的结果中所述。它会编译，运行并返回一个有点随机的值。

编辑：数据库版本为 Oracle Database 12c企业版12.1.0.2.0

与group by一起使用时，交叉应用的行为很奇怪

0 个答案: