Question

我在PostGre数据库上有此表：

CREATE TABLE testAAA(Id integer PRIMARY KEY, datum date, COLA text, COLB text, COLC text, COLD int);

/* Create few records in this table */
INSERT INTO testAAA VALUES(1,to_date('01/01/2018','dd/mm/yyyy'),'PLANTA','VENDORA','OPA','1');
INSERT INTO testAAA VALUES(2,to_date('01/02/2018','dd/mm/yyyy'),'PLANTA','VENDORA','OPA','1');
INSERT INTO testAAA VALUES(3,to_date('01/03/2018','dd/mm/yyyy'),'PLANTA','VENDORA','OPA','1');
INSERT INTO testAAA VALUES(4,null,'PLANTA','VENDORA','OPA','1');
INSERT INTO testAAA VALUES(5,to_date('01/10/2018','dd/mm/yyyy'),'PLANTA','VENDORA','OPA','5');
INSERT INTO testAAA VALUES(6,to_date('01/10/2018','dd/mm/yyyy'),'PLANTA','VENDORA','OPA','10');
INSERT INTO testAAA VALUES(7,to_date('01/11/2018','dd/mm/yyyy'),'PLANTA','VENDORB','OPA','50');
INSERT INTO testAAA VALUES(8,to_date('01/10/2018','dd/mm/yyyy'),'PLANTA','VENDORA','OPA','10');
INSERT INTO testAAA VALUES(9,to_date('01/11/2018','dd/mm/yyyy'),'PLANTA','VENDORA','OPB','5');
COMMIT;

我有这个查询：

SELECT COLA,COLB,COLC,to_char(datum,'YYYY-MM') AS datum_c,
    SUM(COLD) FILTER (WHERE (datum >= now() - interval '6 month') AND (datum <= now())) OVER (PARTITION BY COLA, COLB, COLC) / 6.0 AS AVG_6_month,
    /* TO Aggregate at month level we need to PARTITION IT */
    SUM(COLD) FILTER (WHERE datum is not null) OVER (PARTITION BY COLA,to_char(datum,'YYYY-MM')) AS TOT_ORDERED,
    SUM(COLD) FILTER (WHERE datum is not null) OVER (PARTITION BY COLA,COLB,to_char(datum,'YYYY-MM')) AS TOT_ORDERED_PLANT,
    (SUM(COLD) FILTER (WHERE datum >= now() - interval '6 month') OVER (PARTITION BY COLA, COLB, COLC) / 6.0) / (SUM(COLD) FILTER (WHERE datum is not null) OVER (PARTITION BY COLA))*100 AS Variation,
    CASE
      WHEN (SUM(COLD) FILTER (WHERE datum >= now() - interval '6 month') OVER (PARTITION BY COLA, COLB, COLC) / 6.0) / (SUM(COLD) FILTER (WHERE datum is not null) OVER (PARTITION BY COLA))*100 >= 10 THEN 'A'
     WHEN (SUM(COLD) FILTER (WHERE datum >= now() - interval '6 month') OVER (PARTITION BY COLA, COLB, COLC) / 6.0) / (SUM(COLD) FILTER (WHERE datum is not null) OVER (PARTITION BY COLA))*100 < 10  THEN 'B'
    END AS var_type
FROM testAAA;

我希望变异<10的PARTAL COLA / COLB的TOT COUNT。我试图在case语句中添加计数：

WHEN (SUM(COLD) FILTER (WHERE datum >= now() - interval '6 month') OVER (PARTITION BY COLA, COLB, COLC) / 6.0) / (SUM(COLD) FILTER (WHERE datum is not null) OVER (PARTITION BY COLA))*100 >= 10 THEN count((SUM(COLD) FILTER (WHERE datum >= now() - interval '6 month') OVER (PARTITION BY COLA, COLB, COLC) / 6.0) / (SUM(COLD) FILTER (WHERE datum is not null) OVER (PARTITION BY COLA))*100)

但是我又得到了这个错误：

42803：集合函数调用不能包含窗口函数调用。

在“ The Impaler”提示后进行###### EDITED

with
x as ( -- first, compute the window functions
  select COLA,COLB,COLC,to_char(datum,'YYYY-MM') AS datum_c,
    SUM(COLD) FILTER (WHERE (datum >= now() - interval '6 month') AND (datum <= now())) OVER (PARTITION BY COLA, COLB, COLC) / 6.0 AS AVG_6_month,
    /* TO Aggregate at month level we need to PARTITION IT */
    SUM(COLD) FILTER (WHERE datum is not null) OVER (PARTITION BY COLA,to_char(datum,'YYYY-MM')) AS TOT_ORDERED,
    SUM(COLD) FILTER (WHERE datum is not null) OVER (PARTITION BY COLA,COLB,to_char(datum,'YYYY-MM')) AS TOT_ORDERED_PLANT,
    (SUM(COLD) FILTER (WHERE datum >= now() - interval '6 month') OVER (PARTITION BY COLA, COLB, COLC) / 6.0) / (SUM(COLD) FILTER (WHERE datum is not null) OVER (PARTITION BY COLA))*100 AS Variation
  from testAAA
)
select*,
  COUNT(Variation) FILTER (WHERE Variation < 10) OVER (PARTITION BY COLA, COLB) AS below10,
  COUNT(Variation) FILTER (WHERE Variation > 10) OVER (PARTITION BY COLA, COLB) AS above10
from x

Answer 1

这不是答案，而是注释部分中不适合的注释。当您有多个窗口函数时，并且想要操作它们的结果（在这种情况下），我倾向于在CTE（公用表表达式）中计算它们；然后，我将它们用作主查询中的简单标量值。这使我的生活变得更加轻松。

例如，这样的查询：

select 
  sum(a) over(partition by region_id order by created_at) /
  sum(b) over(partition by region_id order by deleted_at) as score
from my_table

我将查询改为：a）预计算“复杂”值，然后2）在简单查询中使用它们，如：

with
x as ( -- first, compute the window functions
  select 
    sum(a) over(partition by region_id order by created_at) as sum_a
    sum(b) over(partition by region_id order by deleted_at) as sum_b
  from my_table
)
select
  sum_a,
  sum_b,
  sum_a / sum_b, -- then use them for any purposes
  sum_a * sum_b
from x

也许您可以使用此策略使查询看起来更简单，因此调试起来会更容易。

按条件对分区进行计数

1 个答案: