如果事件每 X 分钟发生一次以上，则仅计算一次事件

Question

每次用户在我的应用程序中启动会话时，我都会填写一个表格。但如果他们在 10 分钟内完成，我不想多次计算他们的会话。我该怎么做？

这是一个从表中返回的示例

   select
    *
    from table
    limit 100


+----------+--------+---------+----------------+
| event_ID | userid | city_id |   created_at   |
+----------+--------+---------+----------------+
|        1 | a      |       1 | 15/08/19 10:10 |
|        2 | b      |       1 | 15/08/19 10:11 |
|        3 | a      |       1 | 15/08/19 10:14 |
|        4 | a      |       1 | 15/08/19 10:25 |
|        5 | b      |       1 | 15/08/19 10:27 |
|        6 | c      |       1 | 15/08/19 10:30 |
|        7 | c      |       1 | 15/08/19 10:35 |
|        8 | d      |       1 | 15/08/19 10:40 |
|        9 | d      |       1 | 15/08/19 10:49 |
|       10 | c      |       1 | 15/08/19 10:55 |
+----------+--------+---------+----------------+

最后，我想计算每个用户的 unique event_id，前提是唯一的 event_id 由每 10 分钟发生的次数定义

所以最后应该是这样的：

+--------+------------------+
| userid | unique_event_ids |
+--------+------------------+
| a      |                2 |
| b      |                2 |
| c      |                2 |
| d      |                1 |
+--------+------------------+
+--------+------------------+
| Total  |                7 |
+--------+------------------+

关于如何开始的任何建议？

Answer 1

使用lag() 确定为用户创建上一个事件的时间。然后是一些日期过滤和聚合：

select userid, count(*)
from (select t.*,
             lag(created_at) over (partition by userid order by created_at) as prev_created_at
      from t
     ) t
where prev_created_at is null or prev_created_at < created_at - interval '10 minute'
group by userid

Answer 2

我愿意：

select
  userid,
  sum(case when created_at - interval '10 minute' < prev then 0 else 1 end)
    as unique_events_ids
from (
  select
    *,
    lag(created_at) over(partition by userid order by created_at) as prev
  from t
) x
group by userid