Solved: Multiplying the result of a SUM() on one table by the result of a COUNT() from another table - sql

Question

I'm trying to multiply the sum computed in one table by the count of values found in another table linked to it.
I have a table with entries like this:
Table_A
Id
SubId
Start
Stop
For each Id there can be several SubId. The Id is linked through a whole series of other tables to a second table which basically looks like this:
Table_B
Id
Val
Id in Table_B can occur from one to multiple times.
The idea is that I want to mulitply the result of the sum of time spans found in Table_A for each SubId by the count of distinct Val found in Table_B by linking it through the Id.
I already have the sum of time spans from Table_A like this (I'm no developper so it might be very shoddy code, I apologize in advance):
SELECT SUM(duration) FROM (
SELECT (mod_stop - mod_start) AS duration FROM (
SELECT
(CASE
WHEN date_start < '#begin' THEN '#begin'
ELSE date_start
END) AS mod_start,
(CASE
WHEN date_stop > '#end' THEN '#end'
WHEN date_stop = '0' THEN '#end'
ELSE date_stop
END) AS mod_stop
FROM table_a
WHERE (
state = 'Launching' OR
state = 'Running' OR
state = 'Finishing'
)
AND table_a_id IN (
SELECT
DISTINCT(table_a_id)
FROM table_a
WHERE
(to_timestamp(date_start), to_timestamp(date_stop)) OVERLAPS (to_timestamp('#begin'), to_timestamp('#end'))
AND
(state = 'Running' OR state = 'Launching' OR state = 'Finishing')
AND
NOT date_stop = '0'
)
) AS t
) AS d
;
This works but I now need to multiply each duration by the number of Val associated to it's Id and I can't work out how to do this.
I thought having another AND ... IN clause in the WHERE with all the table linking mechanisms returning the individual Val count would do it but the query does not return anything despite running for over an hour whereas without it it returns in approximately ten minutes (there is no index on date_* and Table_A is several tens of million records long which explains why it is so slow) so I fear I might have got something wrong.
Is there a way to do this? Thanks!

Answer 1

Thanks to the comments posted and what I managed to find I've cobbled a solution together. It might be crude but it works. I build a temp table whose values I use to build a CTE. Feel free to comment if you think this can be done better.
BEGIN;
CREATE TEMP TABLE my_jsl ON COMMIT DROP AS
SELECT job_id, job_state, mod_start, mod_stop, duration FROM (
SELECT job_id, job_state, mod_start, mod_stop, mod_stop - mod_start duration FROM (
SELECT job_id, job_state,
(CASE
WHEN date_start < :my_begin THEN :my_begin
ELSE date_start
END) AS mod_start,
(CASE
WHEN date_stop > :my_end THEN :my_end
WHEN date_stop = '0' THEN :my_end
ELSE date_stop
END) AS mod_stop
FROM job_state_logs
WHERE (
job_state = 'Launching' OR
job_state = 'Running' OR
job_state = 'Finishing'
)
AND job_state_log_id IN (
SELECT
DISTINCT(job_state_log_id)
FROM job_state_logs
WHERE
(to_timestamp(date_start), to_timestamp(date_stop)) OVERLAPS (to_timestamp(:my_begin), to_timestamp(:my_end))
AND
(job_state = 'Running' OR job_state = 'Launching' OR job_state = 'Finishing')
AND
NOT date_stop = '0'
)
) AS job_temp_log
) AS job_log
;
WITH my_jr AS (
SELECT COUNT(a.resource_id) jrc, j.job_id
FROM assigned_resources a
INNER JOIN moldable_job_descriptions m ON m.moldable_id = a.moldable_job_id
INNER JOIN jobs j ON j.assigned_moldable_job = m.moldable_id
WHERE j.job_id IN (
SELECT DISTINCT(job_id) FROM my_jsl
)
GROUP BY j.job_id
),
my_surf AS (
SELECT my_jr.job_id, my_jsl.duration, my_jr.jrc, (my_jsl.duration * my_jr.jrc) AS surface
FROM my_jr, my_jsl
WHERE my_jr.job_id = my_jsl.job_id
)
SELECT SUM(surface) job_surface FROM my_surf
;
COMMIT;
;