如何在 PostgreSQL 中通过具有开始和结束日期的日期来构建流失漏斗

Question

我在 PostgreSQL 中有一张表，上面有我们客户取消订阅我们服务的结束日期的开始日期

user_id	date	disactivated_at
1	October 28, 2021
2	October 28, 2021
3	October 28, 2021	August 26, 2022
4	October 28, 2021
5	October 28, 2021
6	October 28, 2021	March 29, 2022
7	October 28, 2021
8	October 28, 2021
9	October 28, 2021	August 26, 2022
10	October 28, 2021
11	October 28, 2021	March 30, 2022
12	October 28, 2021
13	October 28, 2021
14	October 28, 2021	February 4, 2022
15	October 28, 2021
16	October 28, 2021
17	October 28, 2021
18	October 28, 2021	January 19, 2022
19	October 28, 2021
20	October 28, 2021

我怎样才能有一个表格，显示我们每个月（或一周或一天）有多少活跃用户？

理想的结果是：

Month	Active_users
November	20
December	20
January	19
February	18
March	16
April	16
May	16
June	16
July	16
August	14
September	14

不幸的是，我没有自己的代码草稿，因为我不知道从哪里开始。基本分组方法在这里不起作用（如果我现在需要知道实际的活跃用户数，它可以工作，但我也需要之前的每个时间段）。

Answer 1

有点复杂，但一步一步很简单—— 建立一个月（或周，或天）列表 - t_months_list CTE，将其与"date"-s 修剪为月（或周或日）的用户列表一起添加为"month" - t_list CTE - 使用"month"，然后使用aggregate with filtering。

with
t_months_list as 
(
 select generate_series('2021-10-01', '2022-09-01', interval '1 month')::date "month"
),
t_list as 
(
 select date_trunc('month', "date")::date "month", "date", desactivated_at 
 from the_table
)
select tml."month",
  count(*) filter 
    (
      where tl."date" is not null 
      and (desactivated_at is null or desactivated_at > tml."month")
    ) as active_users
from t_months_list tml left join t_list tl on tml."month" >= tl."month"
group by tml."month"
order by tml."month";

db-fiddle 上的演示
请注意，使用保留字（即月份、日期）作为列名不是一个好主意。