Splitting out yearly sums across columns - sql

Question

I have 2 tables where I'm trying to grab counts of student interactions by year, then sum them in their respective year. The code I've attached works, but I'm wondering if I've neglected a much easier way of doing this calculation. For example, if I wanted to do this for more than 2 years I'd be cursing myself doing it this way.
select s.id
, coalesce(cl2016.cl2016, 0) + coalesce(wf2016.wf2016, 0) as s2016
, coalesce(cl2017.cl2017, 0) + coalesce(wf2017.wf2017, 0) as s2017
from students s
left join (
select dm.student_id
, count(dm.meeting_id) as cl2016
from dim_meetings dm
where dm.start_time between '2016-01-01' and '2016-12-31'
group by dm.student_id
) cl2016 on cl2016.student_id = s.id
left join (
select dm.student_id
, count(dm.meeting_id) as cl2017
from dim_meetings dm
where dm.start_time between '2017-01-01' and '2017-12-31'
group by dm.student_id
) cl2017 on cl2017.student_id = s.id
left join (
select sub.student_id
, count(sub.id) as wf2016
from submissions sub
where sub.submitted_at between '2016-01-01' and '2016-12-31'
group by sub.student_id
) wf2016 on wf2016.student_id = s.id
left join (
select sub.student_id
, count(sub.id) as wf2017
from submissions sub
where sub.submitted_at between '2017-01-01' and '2017-12-31'
group by sub.student_id
) wf2017 on wf2017.student_id = s.id

Answer 1

Use conditional aggregation:
select s.id, dm.cl2016, cl2017, su.wf2016, su.wf2017,
coalesce(dm.cl2016, 0) + coalesce(su.wf2016, 0) as s2016,
coalesce(dm.cl2017, 0) + coalesce(su.wf2017, 0) as s2017
from students s left join
(select dm.student_id,
sum( (dm.start_time between '2016-01-01' and '2016-12-31')::int) as cl2016,
sum( (dm.start_time between '2017-01-01' and '2017-12-31')::int) as cl2017
from dim_meetings dm
where dm.start_time between '2016-01-01' and '2017-12-31'
group by dm.student_id
) dm
on dm.student_id = s.id left join
(select su.student_id,
sum( (su.start_time between '2016-01-01' and '2016-12-31')::int) as wf2016,
sum( (su.start_time between '2017-01-01' and '2017-12-31')::int) as wf2017
from submissions su
where su.start_time between '2016-01-01' and '2017-12-31'
group by su.student_id
) su
on su.student_id = s.id ;

Answer 2

It is useful to place a case expression inside an aggregate function as shown in the example below:
select
dm.student_id
, count(dm.start_time '2016-01-01'
and < dm.start_time '2017-01-01' then dm.meeting_id end) as cl2016
, count(dm.start_time '2017-01-01'
and < dm.start_time '2018-01-01' then dm.meeting_id end) as cl2017
from dim_meetings dm
where dm.start_time between '2016-01-01' and dm.start_time < '2018-01-01'
group by dm.student_id
;
A case expression returns a value, then that value is used within the aggregate function just as it would any other value.
Please also note that "between" is evil for date ranges, instead use explicit ranges via >= and < as shown above, which involves moving the upper boundary date up to the the next day. See:
Bad habits to kick : mis-handling date / range queries
What do BETWEEN and the devil have in common?