SQL SUM() Calculates only one row in sub query - mysql

Question

I'm making statistic for 5 tables. I have made the example with one client data.
loan
id | status
------------
1454 | payed
payment schedule
id | loan_id | user_client_id
-----------------------------
1456 | 1454 | 3113
payment_schedule_row
id | payment_schedule_id | payment | payment_date
---------------------------------------------------
5013 | 1456 | 32 | 2013-11-06
5014 | 1456 | 32 | 2013-12-06
5015 | 1456 | 32 | 2013-01-05
5016 | 1456 | 32 | 2013-02-04
5017 | 1456 | 32 | 2013-03-06
5018 | 1456 | 32 | 2013-04-05
5019 | 1456 | 32 | 2013-05-05
5020 | 1456 | 32 | 2013-06-04
5021 | 1456 | 32 | 2013-07-04
5022 | 1456 | 32 | 2013-08-03
5023 | 1456 | 32 | 2013-09-02
5014 | 1456 | 32 | 2013-10-02
payment_schedule_cover
id | payment_schedule_id | date | sum
----------------------------------------------
2282 | 1456 | 2013-11-08 | 34
3054 | 1456 | 2013-12-07 | 40
3776 | 1456 | 2013-01-04 | 38
4871 | 1456 | 2013-02-06 | 49
5954 | 1456 | 2013-03-06 | 40
7070 | 1456 | 2013-04-25 | 49
9029 | 1456 | 2013-05-21 | 52
10377 | 1456 | 2013-06-20 | 30
10391 | 1456 | 2013-06-21 | 30
10927 | 1456 | 2013-07-07 | 60
payment_schedule_delay
id | payment_schedule_row_id | start_date | end_date | delay
----------------------------------------------------------------
1135 | 5013 | 2013-11-07 | 2013-11-08 | 0.07
1548 | 5014 | 2013-12-07 | 2013-12-07 | 0.03
2628 | 5016 | 2014-02-05 | 2014-02-06 | 0.01
And the query is :
SELECT period, loan_sum, covers, delay
FROM
(SELECT MAX(EXTRACT(YEAR_MONTH FROM psc.date)) AS period,
(SELECT SUM(psr2.payment) FROM payment_schedule_row AS psr2 WHERE psr.payment_schedule_id = psr2.payment_schedule_id) AS loan_sum,
(SELECT SUM(psc2.sum) FROM payment_schedule_cover AS psc2 WHERE psc.payment_schedule_id = psc2.payment_schedule_id) AS covers,
(SELECT SUM(psd2.delay) FROM payment_schedule_delay AS psd2 WHERE psr.id = psd2.payment_schedule_row_id) AS delay
FROM loan
INNER JOIN payment_schedule AS ps ON ps.loan_id = loan.id
INNER JOIN payment_schedule_row AS psr ON psr.payment_schedule_id = ps.id
INNER JOIN payment_schedule_cover AS psc ON psc.payment_schedule_id = ps.id
WHERE loan.status = 'payed'
GROUP BY ps.id) AS sum_by_id
GROUP BY period
Result for the query:
period | loan_sum | covers | delay
-----------------------------------
201407 | 384 | 422 | 0.07
Everything is right except the delay. It should be 0.11 (0.07 + 0.03 + 0.01)
So I have been trying to find the error from the query for days now. Maybe someone can tell me what I'm doing wrong.
Sqlfiddle link: http://sqlfiddle.com/#!2/21585/2

Answer 1

SELECT period, loan_sum, covers, delay
FROM
(SELECT MAX(EXTRACT(YEAR_MONTH FROM psc.date)) AS period,
(SELECT SUM(psr2.payment) FROM payment_schedule_row AS psr2 WHERE psr.payment_schedule_id = psr2.payment_schedule_id) AS loan_sum,
(SELECT SUM(psc2.sum) FROM payment_schedule_cover AS psc2 WHERE psc.payment_schedule_id = psc2.payment_schedule_id) AS covers,
(SELECT SUM(psd2.delay) FROM payment_schedule_delay AS psd2 WHERE psr.id = psd2.payment_schedule_row_id) AS delay
FROM loan
INNER JOIN payment_schedule AS ps ON ps.loan_id = loan.id
INNER JOIN payment_schedule_row AS psr ON psr.payment_schedule_id = ps.id
INNER JOIN payment_schedule_cover AS psc ON psc.payment_schedule_id = ps.id
INNER JOIN payment_schedule_delay AS psd ON psr.id = psd.payment_schedule_row_id
WHERE loan.status = 'payed'
GROUP BY ps.id) AS sum_by_id
GROUP BY period

Answer 2

SELECT period, loan_sum, covers, delay
FROM
(SELECT MAX(EXTRACT(YEAR_MONTH FROM psc.date)) AS period,
(SELECT SUM(psr2.payment) FROM payment_schedule_row AS psr2 WHERE psr.payment_schedule_id = psr2.payment_schedule_id) AS loan_sum,
(SELECT SUM(psc2.sum) FROM payment_schedule_cover AS psc2 WHERE psc.payment_schedule_id = psc2.payment_schedule_id) AS covers,
(SELECT SUM(psd2.delay) FROM payment_schedule_delay AS psd2 WHERE psr.id IN /* IN operator will allow for multiple values like psr.id IN (5013,5014,5016) */ psd2.payment_schedule_row_id) AS delay
FROM loan
INNER JOIN payment_schedule AS ps ON ps.loan_id = loan.id
INNER JOIN payment_schedule_row AS psr ON psr.payment_schedule_id = ps.id
INNER JOIN payment_schedule_cover AS psc ON psc.payment_schedule_id = ps.id
WHERE loan.status = 'paid'
GROUP BY ps.id) AS sum_by_id
GROUP BY period
Change = to IN in the line where you are summarizing the delay values.

Answer 3

I finally got an answer from a MySQL forum. The answer what fixed my problem was:
... there are problems ...
The Group By operator in the subquery does not see aggregation inside the correlated sub-subquery sums. Those sums need to be moved out a level.
There's no aggregation for the outer query's Group By to group; it just functions as an Order By
A query like select a,b,c sum(d) ... group by a can return arbitrary results for b and c unless a strictly 1:1 relationship holds between a and each of b and c, which looks unlikely to be the case in your subquery.
Correlated subqueries are inefficient, as yours illustrate with their two-stage joins
The delay correlated subquery doesn't join to anything
So move correlated subquery logic to the FROM clause, join the delay query, touch up the Group By clause, and we have ...
select psc.period, psc.sum, psr.payments, sum(psd.delay) as delay
from loan
join payment_schedule as ps on ps.loan_id = loan.id
join(
select payment_schedule_id, sum(payment) as payments
from payment_schedule_row
group by payment_schedule_id
) as psr on psr.payment_schedule_id = ps.id
join (
select payment_schedule_id, sum(sum) as sum, max( extract(year_month from date) ) as period
from payment_schedule_cover
group by payment_schedule_id
) psc on ps.id = psc.payment_schedule_id
join payment_schedule_row psr2 on ps.id = psr2.payment_schedule_id
join (
select payment_schedule_row_id, sum(delay) as delay
from payment_schedule_delay
group by payment_schedule_row_id
) as psd on psr2.id = psd.payment_schedule_row_id
where loan.status = 'payed'
group by psc.period, psc.sum, psr.payments;