Can't figure out how to full join two queries with different values in both tables - sql

Question

I have two queries that look like this:
SELECT
sem.Sem_Jahr,
sem.Sem_KW,
COUNT(*) AS Seminars,
bearb.MA_ID
FROM acc_seminar.t_Seminar sem
JOIN acc_seminar.t_Seminar_Thema semth ON sem.Sem_SemTh_ID = semth.SemTh_ID
JOIN acc_ma.t_Mitarbeiter bearb ON sem.Sem_Berb_MA_ID = bearb.MA_ID
WHERE sem.Sem_Sto != 1 AND semth.SemTh_Typ = 7 AND sem.Sem_Jahr = #Jahr and MA_ID = 372
GROUP BY bearb.MA_ID, sem.Sem_KW, sem.Sem_Jahr
the second query is exactly the same, except the condition is WHERE sem.Sem_Sto != 1 AND semth.SemTh_Typ = 7 AND sem.Sem_Jahr = #Jahr and MA_ID = 372
KW refers to week
I want to show results from both queries, combined. The problem is that the first query may have seminar count value for KW 2, but the second one would have NULL. The problem is, I can't figure out how to join them to get the following desired result:
KW | Seminars from query 1 | Seminars from query 2
----------------------------------------------------
2 | NULL | 5
3 | 8 | NULL
4 | 1 | 4
What I tried:
I tried just putting UNION between these two, but then I only get results from first query.
I also tried to write first query normally and then doing a FULL OUTER JOIN with second query as subquery in JOIN, but then I get results for the first query and results from second query only where the week matches with row from first query.
This whole request seems so banal to me, but I just can't figure it out, it doesn't click in my head on how to join them. Any suggestions?

Answer 1

Alway aim for a minimal, reproducable example. My sample data has way less joins, but should still show your issue and possible solutions.
Sample data
create table data
(
year int,
week int,
flag bit
);
insert into data (year, week, flag) values
(2021, 1, 0),
(2021, 1, 1),
(2021, 1, 1),
(2021, 2, 0),
(2021, 2, 0),
(2021, 2, 0),
(2021, 2, 0),
(2021, 3, 1);
Issue reproduction
Second query as subquery:
select coalesce(f.year, t.year) as year,
coalesce(f.week, t.week) as week,
count(1) as countFalse,
t.countTrue
from data f
full join ( select d.year,
d.week,
count(1) as countTrue
from data d
where d.flag = 1
group by d.year,
d.week ) t
on t.year = f.year
and t.week = f.week
where f.flag = 0 --> issue: week 3 not available for flag = 0, results limited...
group by f.year,
t.year,
f.week,
t.week,
t.countTrue
order by f.year,
f.week;
Result missing week = 3:
year week countFalse countTrue
---- ---- ---------- ---------
2021 1 1 2
2021 2 4 null
Solution 1
Isolate both queries in common table expressions (cte_false, cte_true) and join them without where clause in final select.
with cte_false as
(
select d.year,
d.week,
count(1) as countFalse
from data d
where d.flag = 0
group by d.year,
d.week
),
cte_true as
(
select d.year,
d.week,
count(1) as countTrue
from data d
where d.flag = 1
group by d.year,
d.week
)
select coalesce(f.year, t.year) as year,
coalesce(f.week, t.week) as week,
f.countFalse,
t.countTrue
from cte_false f
full join cte_true t
on t.year = f.year
and t.week = f.week;
Solution 2
Perform all calculations first (cte_count), then use pivot to transform the data.
with cte_count as
(
select d.year,
d.week,
d.flag,
count(1) as countFlag
from data d
group by d.year,
d.week,
d.flag
)
select piv.year,
piv.week,
piv.[0] as countFalse,
piv.[1] as countTrue
from cte_count cc
pivot (max(cc.countFlag) for cc.flag in ([0], [1])) piv;
Result
year week countFalse countTrue
---- ---- ---------- ---------
2021 1 1 2
2021 2 4 null
2021 3 null 1
Fiddle to see things in action.

Answer 2

You can do this using conditional aggregation:
SELECT sem.Sem_Jahr,
sem.Sem_KW,
SUM(CASE WHEN sem.Sem_Sto <> 1 AND semth.SemTh_Typ = 7 AND sem.Sem_Jahr = #Jahr and MA_ID = 372 THEN 1 ELSE 0 END) AS Seminars,
SUM( <whatever the second condition is> THEN 1 ELSE 0 END),
bearb.MA_ID
FROM acc_seminar.t_Seminar sem JOIN
acc_seminar.t_Seminar_Thema semth
ON sem.Sem_SemTh_ID = semth.SemTh_ID JOIN
acc_ma.t_Mitarbeiter bearb
ON sem.Sem_Berb_MA_ID = bearb.MA_ID
GROUP BY bearb.MA_ID, sem.Sem_KW, sem.Sem_Jahr;
If I speculate that the difference is one of the columns, such as semth.SemTh_Typ = 8, then this can be simplified by moving common conditions to the WHERE clause:
SELECT sem.Sem_Jahr,
sem.Sem_KW,
SUM(CASE WHEN semth.SemTh_Typ = 7 THEN 1 ELSE 0 END) AS Seminars,
SUM(CASE WHEN semth.SemTh_Typ = 8 THEN 1 ELSE 0 END),
bearb.MA_ID
FROM acc_seminar.t_Seminar sem JOIN
acc_seminar.t_Seminar_Thema semth
ON sem.Sem_SemTh_ID = semth.SemTh_ID JOIN
acc_ma.t_Mitarbeiter bearb
ON sem.Sem_Berb_MA_ID = bearb.MA_ID
WHERE sem.Sem_Sto <> 1 AND
semth.SemTh_Typ IN (7, 8) AND
sem.Sem_Jahr = #Jahr AND
MA_ID = 372
GROUP BY bearb.MA_ID, sem.Sem_KW, sem.Sem_Jahr;