我参加了之前的一项练习并增加了一些复杂性。 您可以在select case with "over partition by"
下找到我原来的问题场景:(使用SQL Server 2012)
create table #testing (b varchar (20), a date, c int, e int)
insert into #testing (b,a,c,e)
values
('xf_1m','2015-03-02','1','3'),
('xf_3m','2015-03-02','2','5'),
('xf_5y','2015-03-02','4','2'),
('xf_10y','2015-03-02','3','6'),
('ay_10y','2015-03-02','7','2'),
('adfe_1m','2015-03-02','2','5'),
('xm_1m','2013-02-01','7','2'),
('xf_15y','2013-02-01','1','8'),
('xf_20y','2013-02-01','10','1')
使用此查询后:
select
b, a, c, e,
substring (b, 1, CHARINDEX ('_', b) - 1) rnc,
substring(b, CHARINDEX('_', b) + 1, LEN (b)) rnb,
case
when b like 'xf%' then --
(sum(c * e) over (partition by a )) end as sumProduct
into #testing2
from #testing
select
*,
case
when b like 'xf%' then --
(sum(c * e) over (partition by a )) end as sumProduct
into #testing3
from #testing2
select *
from #testing3
我得到了这个:
现在我只想计算由rnc和date(列a)分区的sumProduct。 这该怎么做?我尝试使用group by,但是我遇到了选择中的不等号和我正在分组的项目数量的麻烦。
所以,我喜欢以某种方式重写这个:
(sum(c * e) over (partition by a and partition by rnc )) as sumProduct
答案 0 :(得分:1)
我不确定您使用临时表的原因,但只需将它们包含在partition by
列表中即可按多列进行分区:
select *,
(case when b like 'xf%'
then sum(c * e) over (partition by a, rnd )
end) as sumProduct
into #testing3
from #testing2;