如何按时间顺序计算 SQL 中的事件？

Question

我想聚合 Talbe_A 并使它像 Table_B。我很难按照时间顺序来统计事件。我应该如何编写查询？

Table_A

client_id   event   time
   A        view        12:00:00
   A        view        12:02:00
   A        view        12:05:00
   A        purchase    14:02:00
   B        view        12:04:00
   B        view        12:07:00
   B        view        13:20:00
   C        view        12:00:00
   C        view        12:07:00
   C        add_to_cart 14:02:00
   C        view        14:19:00
   C        purchase    14:32:00
   C        view        15:32:00

Table_B

client_id   event   count
    A       view        3
    A       purchase    1
    B       view        3
    C       view        3
    C       add_to_cart 1
    C       view        1 
    C       purchase    1
    C       view        1

Answer 1

这是一个 gaps-and-islands 问题，您希望将“相邻”行组合在一起。

这里我认为最简单的做法是使用行号的不同来定义分组：

select client_id, event, count(*) cnt, min(time) start_time, max(time) end_time
from (
    select 
        t.*,
        row_number() over(partition by client_id order by time) rn1,
        row_number() over(partition by client_id, event order by time) rn2
    from mytable t
) t
group by client_id, event, rn1 - rn2
order by client_id, min(time)

Demo on DB Fiddle:

client_id | event       | cnt | start_time | end_time
:-------- | :---------- | --: | :--------- | :-------
A         | view        |   3 | 12:00:00   | 12:05:00
A         | purchase    |   1 | 14:02:00   | 14:02:00
B         | view        |   3 | 12:04:00   | 13:20:00
C         | view        |   2 | 12:00:00   | 12:07:00
C         | add_to_cart |   1 | 14:02:00   | 14:02:00
C         | view        |   1 | 14:19:00   | 14:19:00
C         | purchase    |   1 | 14:32:00   | 14:32:00
C         | view        |   1 | 15:32:00   | 15:32:00

如何按时间顺序计算 SQL 中的事件？

How do I count events in chronological order in SQL?

sql

time

window-functions

gaps-and-islands

amazon-redshift