不想在过滤聚合中重复计算

Question

示例数据：

shopper_id	last_purchase_timestamp	active_p30	active_p60	active_over_p90
1	2022-03-02 1:20:00	TRUE	TRUE	TRUE
2	2022-03-01 1:30:00	TRUE	TRUE	TRUE
3	2022-02-28 1:24:03	TRUE	TRUE	TRUE
4	2022-02-02 21:22:26	FALSE	TRUE	TRUE

我想统计购物者在过去 30 天（从 3 月 5 日开始）、过去 60 天等期间是否活跃（如上次购买）

我的目标是找出有多少购物者在过去 30 天内购买了他们最后一件商品，有多少购物者在过去 60 天内购买了他们最后一件商品等等。但是我不想重复计算。

我尝试过的：

AS total_active_p30,

count(*) FILTER (where last_purchase_timestamp >= DATE '2022-03-05' - INTERVAL '60' day) 
AS total_active_p60,

count(*) FILTER (where last_purchase_timestamp >= DATE '2022-03-05' - INTERVAL '90' day) AS 
total_active_p90

结果：

total_active_p30	total_active_p60	total_active_p90
3	4	4

然而，这导致它重复计算。我怎样才能防止它重复计算？计数总数应为 4。

我理想的输出是：

total_active_p30	total_active_p60	total_active_p90
3	1	0

提前谢谢大家！我正在使用 Trino！

Answer 1

将上限和下限都添加到过滤器中，使它们不相交。沿着这条线：

-- sample data
WITH dataset (last_purchase_timestamp) AS (
    VALUES (timestamp '2022-03-02 1:20:00'),
        (timestamp '2022-03-01 1:30:00'),
        (timestamp '2022-02-28 1:24:03'),
        (timestamp '2022-02-02 21:22:26')
)

-- query
select count_if(last_purchase_timestamp >= DATE '2022-03-05' - INTERVAL '30' day) total_active_p30,
    count_if(last_purchase_timestamp >= DATE '2022-03-05' - INTERVAL '60' day and last_purchase_timestamp < DATE '2022-03-05' - INTERVAL '30' day) total_active_p60,
    count_if(last_purchase_timestamp >= DATE '2022-03-05' - INTERVAL '90' day and last_purchase_timestamp < DATE '2022-03-05' - INTERVAL '60' day) total_active_p90 
from dataset

输出：

total_active_p30	total_active_p60	total_active_p90
3	1	0

Answer 2

您的查询逻辑条件不正确。因为产生这个 >= DATE 2022-03-05 - INTERVAL 90 day 条件的数据总是有产生这个 >= DATE 2022-03-05 - INTERVAL 60 day 条件的数据。为此，我们必须编写查询：

count(*) filter (where last_purchase_timestamp >= ('2022-03-05'::date - INTERVAL '30' day)) 
as total_active_p30,

count(*) filter (where last_purchase_timestamp >= ('2022-03-05'::date - INTERVAL '60' day)
                            and last_purchase_timestamp < ('2022-03-05'::date - INTERVAL '30' day)) 
as total_active_p60,

count(*) filter (where last_purchase_timestamp >= ('2022-03-05'::date - INTERVAL '90' day)
                        and last_purchase_timestamp < ('2022-03-05'::date - INTERVAL '60' day)) 
as total_active_p90

不想在过滤聚合中重复计算

Don't want to double count in Filtered Aggregation

sql

presto

trino