If statement in mysql query with inner join - mysql

Question

I'm currently showing users that got unfinished jobs and based on the results I run a while loop and a switch case statement to come with the final results. I'm wondering if it is possible to move that statement in the mysql query.
select
sum(cnt_jobs_unfinished = 0) cnt_users_no_unfinished_jobs,
sum(cnt_jobs_unfinished_30d > 0) cnt_users_unfinished_30d,
sum(cnt_jobs_unfinished_31_60d > 0) cnt_users_unfinished_31_60d,
sum(cnt_jobs_unfinished_61_90d > 0) cnt_users_unfinished_61_90d,
sum(cnt_jobs_unfinished_90d_more > 0) cnt_users_unfinished_90d_more
from (
select
u.user_id,
sum(l.job_id is null) cnt_jobs_unfinished,
sum(l.job_id is null and j.date >= curdate() - interval 30 day) cnt_jobs_unfinished_30d,
sum(
l.job_id is null
and j.date < curdate() - interval 30 day
and j.date >= curdate() - interval 60 day
) cnt_jobs_unfinished_31_60d,
sum(
l.job_id is null
and j.date < curdate() - interval 60 day
and j.date >= curdate() - interval 90 day
) cnt_jobs_unfinished_61_90d,
sum(
l.job_id is null
and j.date < curdate() - interval 90 day
) cnt_jobs_unfinished_90d_more
from users u
inner join scheduled_jobs j
on j.date <= curdate()
and j.user_id = u.user_id
left join last_update l
on l.job_id = j.job_id
group by u.user_id
) t
Here is the dbfiddle: https://dbfiddle.uk/?rdbms=mysql_8.0&fiddle=d2f217e074a391d8b5f769e08b1d2c87
As you can see because a user got both an unfinished job between 61-90 days and 90+ days the final table shows both results which is a mistake. The correct one would be 61-90 days: 0 users and 90+: 1 user.

Answer 1

This is what I came up with (fiddle):
SELECT
sum(IF(t.nb_days IS NULL OR t.nb_days <= 0, 1, 0)) cnt_users_no_unfinished_jobs,
sum(IF(t.nb_days > 0 AND t.nb_days <= 30, 1, 0)) cnt_users_unfinished_30d,
sum(IF(t.nb_days > 30 AND t.nb_days <= 60, 1, 0)) cnt_users_unfinished_31_60d,
sum(IF(t.nb_days > 60 AND t.nb_days <= 90, 1, 0)) cnt_users_unfinished_61_90d,
sum(IF(t.nb_days > 90, 1, 0)) cnt_users_unfinished_90d_more
FROM users u
LEFT JOIN (
SELECT j.user_id, DATEDIFF(curdate(), MIN(j.date)) AS nb_days
FROM scheduled_jobs j
LEFT JOIN last_update l
ON l.job_id = j.job_id
WHERE l.job_id IS NULL
GROUP BY j.user_id
) AS t
ON u.user_id = t.user_id
Instead of counting how many jobs are in each range for each user, I'm only looking at their oldest unfinished job, and extracting the number of days since the deadline: DATEDIFF(curdate(), MIN(j.date)) AS nb_days).
Then this result is LEFT JOIN-ed to users (that way, in case a user doesn't have any unfinished jobs, he'll still show up in the cnt_users_no_unfinished_jobs column by checking for NULL values)
Finally, it's just a matter of SELECT-ing how many nb_days are in each range.