PostgreSQL 9.5：计算总和并存储到HSTORE

下表有三列：

表:

给定记录：

注意：我想计算前3个月每个colid的colvalue之和并存储它 转换成HSTORE格式

预期结果：

我的尝试：

WITH a
as
(
    select colid,coldate
    from cal_3month
    group by colid,coldate
),
b 
as
(
    select colid,SUM(colvalue) as sumvalue
    from cal_3month td
    WHERE extract('month' from td.coldate) between extract('month' from now() - interval '3 month')  and extract('month' from now() - interval '1 month')  
    and extract('year' from td.coldate) = extract('year' from now()) --and td.st_exchng = 'NSE'
    group by td.coldate,td.colid
    order by td.coldate desc
)
,
c
as
(
    SELECT a.colid,string_agg((extract('year' from a.coldate)::varchar||extract('month' from a.coldate)::varchar)::varchar,',') as dt,
    string_agg(b.sumvalue::varchar,',') as vl
    from a
    inner join b 
    on a.colid = b.colid
    group by a.colid
)
select c.colid,hstore(c.dt,c.vl)
from c;

结果:

---

我没有得到预期结果中包含月份的零件，可能是因为id=6的位置以及为什么id=1月1日没有数据？@VaoTsun，id=1月1日没有数据？->因为我只需要最后三个月的数据。啊-从现在起三个月了？。@VaoTsun，是的！从现在起过去3个月。

colid   coldate         colvalue
-----------------------------------------------
1   2017-01-31 00:00:00 100
1   2017-03-27 00:00:00 238
1   2017-02-12 00:00:00 500
1   2017-02-23 00:00:00 50
2   2017-03-22 00:00:00 78
2   2017-04-22 00:00:00 400
2   2017-04-22 00:00:00 123
3   2017-04-02 00:00:00 990
4   2017-03-22 00:00:00 890
5   2017-02-11 00:00:00 1000
6   2017-01-02 00:00:00 2300
7   2017-02-25 00:00:00 780
8   2017-02-26 00:00:00 90

colid   date and sum value(hstore)
-----------------------------------------------
1   "201702"=>"550","201703"=>"238"
2   "201703"=>"78","201704"=>"523"
3   "201704"=>"990"
4   "201703"=>"890"
5   "201702"=>"1000"
6
7   "201702"=>"780"
8   "201702"=>"90"

WITH a
as
(
    select colid,coldate
    from cal_3month
    group by colid,coldate
),
b 
as
(
    select colid,SUM(colvalue) as sumvalue
    from cal_3month td
    WHERE extract('month' from td.coldate) between extract('month' from now() - interval '3 month')  and extract('month' from now() - interval '1 month')  
    and extract('year' from td.coldate) = extract('year' from now()) --and td.st_exchng = 'NSE'
    group by td.coldate,td.colid
    order by td.coldate desc
)
,
c
as
(
    SELECT a.colid,string_agg((extract('year' from a.coldate)::varchar||extract('month' from a.coldate)::varchar)::varchar,',') as dt,
    string_agg(b.sumvalue::varchar,',') as vl
    from a
    inner join b 
    on a.colid = b.colid
    group by a.colid
)
select c.colid,hstore(c.dt,c.vl)
from c;

colid   hstore
-----------------------------------------------
4   "20173"=>"890"
5   "20172"=>"1000"
8   "20172"=>"90"
1   "20171,20171,20171,20173,20173,20173,20172,20172,20172,20172,20172,20172"=>"500,50,238,500,50,238,500,50,238,500,50,238"
2   "20173,20173,20174,20174"=>"78,523,78,523"
3   "20174"=>"990"
7   "20172"=>"780"

t=# with e as (
with c as (
  SELECT distinct colid
   , case when coldate < now() - '3 month'::interval then null else sum(colvalue) over (partition by date_trunc('month',coldate),colid) end summ
  , case when coldate < now() - '3 month'::interval then null else replace(substr(date_trunc('month',coldate)::text,1,7),'-','') end m
  FROM Cal_3Month
order by 1
)
select
  colid, hstore(m,summ::text)
from c
)
select colid,string_agg(hstore::text,',') from e group by colid order by 1 ;
 colid |           string_agg
-------+---------------------------------
     1 | "201702"=>"550","201703"=>"238"
     2 | "201703"=>"78","201704"=>"523"
     3 | "201704"=>"990"
     4 | "201703"=>"890"
     5 | "201702"=>"1000"
     6 |
     7 | "201702"=>"780"
     8 | "201702"=>"90"
(8 rows)