如果我写这样的查询:
with foo as (
select
'foo' as name,
1 as value
union all select
'foo', 2
union all select
'foo', 3
union all select
'foo', 4
), bar as (
select
'bar' as name,
3 as value
union all select
'bar', 4
union all select
'bar', 5
union all select
'bar', 6
)
select
f.name,
b.value,
b.name
from
foo as f
full outer join
bar as b
on b.value = f.value;
我明白了:
name value name
foo (null) (null)
foo (null) (null)
foo 3 bar
foo 4 bar
(null) 5 bar
(null) 6 bar
这就是我所期望的。但是,如果我尝试对表值函数执行类似的操作,例如:
with allSizes as (
select cast('120X60' as character varying) as size
union all select '160X600'
union all select '1X1'
union all select '300X250'
union all select '728X90'
union all select '88X32'
union all select 'PEEL'
)
select
msd.region,
msd.market,
s.size,
msd.target,
msd.actual
from
marketdata('2013-01-05') as msd
full outer join
allSizes as s
on s.size = msd.size
where
msd.market = 'Foo, USA';
其中
marketdata()
是一个表值函数,我没有得到与size
函数中不存在的marketdata()
列对应的空行。为什么不呢? 最佳答案
这是因为:
where
msd.market = 'Foo, USA';
它生成
full outer join
,然后过滤掉marketdata()
中不存在的行(因为它们具有msd.market
=NULL
)您可以将查询更改为:
with allSizes as (
select cast('120X60' as character varying) as size
union all select '160X600'
union all select '1X1'
union all select '300X250'
union all select '728X90'
union all select '88X32'
union all select 'PEEL'
)
select
msd.region,
msd.market,
s.size,
msd.target,
msd.actual
from
(select marketdata('2013-01-05')
where market = 'Foo, USA') as msd
full outer join
allSizes as s
on s.size = msd.size;
或者干脆重写
where
msd.market = 'Foo, USA' OR msd.market IS NULL;
如果
marketdata()
不返回带有msd.market IS NULL
的行。关于sql - Postgres Full Outer Join Headscratcher,我们在Stack Overflow上找到一个类似的问题:https://stackoverflow.com/questions/14219841/