选择总和达到700的记录

Question

I have a score table where every record comes with a variable number of points, a timestamp and a mobile user ID.

I have the task of figuring out the average time it takes a user to obtain 700 points.

How can I use SQL find out the record id where the SUM of total_points reaches 700 in order to compare timestamps and do this for every user.

Is a python script the best approach to solve this problem? Say I get the 2 timestamps (first ever and the one where total_points reached 700) for users in my database who made >= 700 points in order to come up with an average time for all my users.

Or is this possible to do without having to write a script?

Any help or guidance is appreciated.

Answer 1

Here you go. If the data is:

create table score (
  id int,
  mobile_user_id int,
  report_date datetime,
  total_points int
);

insert into score (id, mobile_user_id, report_date, total_points)
  values
  (1, 123, '2018-07-23', 100),
  (1, 123, '2018-07-24', 200),
  (1, 123, '2018-07-25', 500),
  (1, 123, '2018-07-26', 200),
  (2, 124, '2018-06-03', 800),
  (3, 125, '2018-06-17', 150);

The query is:

with a as (
  select
      id, mobile_user_id, report_date,
      sum(total_points) over(partition by id order by report_date)
        as points_so_far
    from score
), 
b as (
  select id, min(report_date) as obtain_date
    from a where points_so_far >= 700
    group by id
)
select s.id, s.initial_date, b.obtain_date
  from b join (
    select id, min(report_date) as initial_date 
      from score group by id  
  ) s on s.id = b.id;

Result:

id           initial_date         obtain_date          
-----------  -------------------  ---------------------
1            2018-07-22 20:00:00  2018-07-24 20:00:00  
2            2018-06-02 20:00:00  2018-06-02 20:00:00  

Answer 2

This will list all users with a total score equal to or over 700 points. It also list the start date and end date for the user, and the total number of days in that period.

SELECT
    `mobile_user_id`,
    MIN(`report_date`) as `points_start`,
    MAX(`report_date`) as `points_end`,
    DATEDIFF(MIN(`report_date`),MIN(`report_date`)) as `duration_days`
    count(`total_points`) as `total`
FROM `points_table`
GROUP BY `mobile_user_id`
HAVING `total` >= 700

Answer 3

I assume your on version 5.7 or below, that is window functions are out of the window (pun, ha, ha). So you'd need to calculate the rolling sum using a subquery, that sums all points from earlier reports.

Use that to filter for only reports at or after the sum of points reached 700.

Group that result by user and get the minimum time stamp for each user, when they reached the 700 points or topped them.

Inner join the overall minimum time stamp per user, which you again get by grouping. The inner join will also eliminate rows from the second grouping subquery, that are from users with less than 700 points.

Get the difference of the two time stamps and calculate the average of that difference.

SELECT avg(timestampdiff(SECOND, x2.report_date, x1.report_date)) duration
       FROM (SELECT t1.mobile_user_id,
                    min(t1.report_date) report_date
                    FROM elbat t1
                    WHERE (SELECT sum(t2.total_points)
                                  FROM elbat t2
                                  WHERE t2.mobile_user_id = t1.mobile_user_id
                                        AND (t2.report_date < t1.report_date
                                              OR t2.report_date = t1.report_date
                                                 AND t2.id < t1.id)) >= 700
                    GROUP BY t1.mobile_user_id) x1
            INNER JOIN (SELECT t1.mobile_user_id,
                               min(t1.report_date) report_date
                               FROM elbat t1
                               GROUP BY t1.mobile_user_id) x2
                                        ON x2.mobile_user_id = x1.mobile_user_id;

Answer 4

What you need is a table with this sort of schema -

user_id, time_when_700_points_achieved (timestamp) , user_start_time (timestamp)

Here is the query for it -

select user_id, max(report_date) as time_when_700_points_achieved , min(report_date) as user_start_time  
from tablename
where points <= 700
group by user

After you achieve this, you can easily take the minute difference between two timestamps.

Assumption - There will be a hit when a user reaches 700