如何在 bigquery sql 中使用 windows function 在一段时间内获取聚合值的差异？

Question

日期键	cust_id	销售量
2022-01-01	1	30
2022-01-02	1	35
2022-01-05	1	38
2022-01-10	1	20
2022-01-11	1	35
2022-01-01	2	20
2022-01-02	2	25
2022-01-04	2	38
2022-01-09	2	20
2022-01-15	1	35
2022-01-11	3	35

我想获取当前期间的所有 customer_ids，然后加入 2022-01-01 -2022-01-05 期间的总和（销售额）与 2022-01-06 至 2022-01 期间的总和（销售额）之间的差异-11。

您将如何在 windows function 中实现这一点？ 目前我正在使用 ctes

with 
users as(
 select 
  distinct cust_id 
 from 
  tableSales 
  where date_key between date('2022-01-06) and date('2022-01-11)),
currentPeriod as(
 select
  distinct cust_id
  ,sum(sales) sales
 from users
  left join tableSales using (customer_id)
  where date_key between date('2022-01-06) and date('2022-01-11)
),
previousPeriod as(
 select
  distinct cust_id
  ,sum(sales) sales
 from users
 left join tableSales using (customer_id)
 where date_key between date('2022-01-05) and date('2022-01-01)
)
#-----------------------
Select 
 distinct cust_id 
 ,cp.sales - pp.sales deltaSales
 from users
left join currentperiod cp using(customer_id)
left join previousperiod pp using(customer_id)

使用 windows function 必须有更短的方法来实现这一点？ 请帮忙。

Answer 1

在您的查询中缺少引号'字段customer_id和cust_id应该相同，对吗？

日期切换： between date('2022-01-05) and date('2022-01-01)

给定的时间间隔很奇怪，因为不清楚用户为什么需要它们。

使用window function：

with tableSales as 
(Select date_sub(date("2022-01-11"), interval cast(rand()*10 as int64) day ) date_key, cust_id,
cast(rand()*100 as int64) as sales
from unnest([1,2,3]) cust_id, unnest(generate_array(1,10,1)) a
)
,tmp as 
(Select *,
sum(if(date_key between date('2022-01-06') and date('2022-01-11'), sales ,0 ) ) over (partition by cust_id) as currentperiod ,
sum(if(date_key between date('2022-01-01') and date('2022-01-05'), sales ,0 ) ) over (partition by cust_id) as previousperiod 
 from tableSales
 )
 Select distinct cust_id, currentperiod, previousperiod from tmp

好吧，做一个“分组”要好得多：

with tableSales as 
(Select date_sub(date("2022-01-11"), interval cast(rand()*10 as int64) day ) date_key, cust_id,
cast(rand()*100 as int64) as sales
from unnest([1,2,3]) cust_id, unnest(generate_array(1,10,1)) a
)
,tmp as 
(Select cust_id,
sum(if(date_key between date('2022-01-06') and date('2022-01-11'), sales ,0 ) )   currentperiod ,
sum(if(date_key between date('2022-01-01') and date('2022-01-05'), sales ,0 ) )   previousperiod 
 from tableSales
 group by 1
 )
 Select distinct cust_id, currentperiod, previousperiod from tmp