SQL聚合和累积求和

Question

I have a user table which looks something like this

There are many users with unique id's, the table is first ordered by users and then by datetime. Each instance is either a training or a test session, test sessions have additional arguments, most importantly score.

I would like to aggregate the rows of this table, so that the time of all the training sessions before a test session are summed and only one instance is returned with the cumulative training time and the test score. Every additional test before a new training session is just a repeat instance with the new test score. This cumulative time summation continues until a new user (user id), when it is reset.

The resulting table would look something like this

Is there an SQL command that can achieve this? (self-join)?
The query doesn't have to be overly optimized, simple solutions are welcome.

Answer 1

Assign a group using a cumulative sum of the test s. Then aggregate to get the summaries you want within each group:

select grp, user_id, sum(time), max(case when task = 'test' then time end) as test_time, max(score)
from (select t.*,
             sum(case when task = 'test' then 1 else 0 end) over (partition by user_id order by datetime) as grp
      from t
     ) t
group by grp, user_id;