从每日表格中构建每月汇总

Question

我编写了以下代码来检查过去 30 天内的唯一客户。我如何重新利用此代码来检查每月开始日期的唯一客户。我正在尝试使用具有每日粒度的 table_billing 构建每月聚合。你能指导一下吗？

select 'context.processingDate' as rptg_dt, COALESCE(item_type,'ALL_ITEMS') as item_type, unique_customers
from (select 
        (case when item_type_code in ('A') then 'Books'
         when item_type_code in ('B','C') then 'Toys'
         else 'Fruits' end
         ) as item_type,
        count(distinct person_id) as unique_customers
      from table_billing
      where rptg_dt between cast('context.processingDate' as date format 'YYYY-MM-DD')-30 
        AND cast('context.processingDate' as date format 'YYYY-MM-DD')
        and item_type_code in ('A','B','C','D','E')
      group by CUBE(1)
    ) a;

所需输出：

Monthly Start Date    | Item Type     | Unique Customers

5/1/14                | Books            | 100

5/1/14                | Toys             | 80

5/1/14                | Fruits           | 25

5/1/14                | ALL_ITEMS        | 175

6/1/14                | Books            | 80

6/1/14                | Toys             | 60

6/1/14                | Fruits           | 40

6/1/14                | ALL_ITEMS        | 95

我希望按如下方式重写此查询：

select 'context.processingDate' as month_start_dt, 
COALESCE(item_type,'ALL_ITEMS') as item_type, unique_customers
from (select 
(case when item_type_code in ('A') then 'Books'
when item_type_code in ('B','C') then 'Toys'
else 'Fruits' end
) as item_type,
count(distinct person_id) as unique_customers
from table_billing
where month_start_dt = cast('context.processingDate' as date format'YYYY-MM-DD') and item_type_code in ('A','B','C','D','E') group by CUBE(1)) a;

我将如何调整查询以使其成为可能？谢谢你！

Answer 1

我假设

context.processingDate

是您将参数传递给脚本的方式。对于每月聚合，您希望处理所有可用数据，因此不再需要此参数。

那么答案是：

select monthly_rptg_dt, COALESCE(item_type,'ALL_ITEMS') as item_type, unique_customers
from (select
        date_trunc('MONTH', rptg_dt) AS monthly_rptg_dt,
        (case when item_type_code in ('A') then 'Books'
         when item_type_code in ('B','C') then 'Toys'
         else 'Fruits' end
         ) as item_type,
        count(distinct person_id) as unique_customers
      from table_billing
      where item_type_code in ('A','B','C','D','E')
      group by 1, CUBE(2)
    ) a;

我还假设不存在间隙（没有数据的月份），或者不需要它们。否则，您必须将此数据外部连接到月份列表，或使用

TIMESERIES

。