Joining multiple counts on the same table, from different columns? - sql

Question

I feel like this should be easy but it's late and I'm struggling.
Say (in an oracle 12 db) I have a table which represents which staff filled what roles in a bar, during different events, like this:
+----------+----------+-------+------------+----------+
| event_id | bar | doors | cloak_room | keg_room |
+----------+----------+-------+------------+----------+
| 2 | bob | bill | john | mary |
+----------+----------+-------+------------+----------+
| 3 | bob | bill | mary | kev |
+----------+----------+-------+------------+----------+
| 4 | bob | john | louise | mary |
+----------+----------+-------+------------+----------+
| 5 | kyle | kev | sarah | louise |
+----------+----------+-------+------------+----------+
| 6 | jennifer | bob | jay | john |
+----------+----------+-------+------------+----------+
| 7 | john | bill | mary | steve |
+----------+----------+-------+------------+----------+
and I want to get a count of, overall, how many events each staff member worked, like this:
+-------+--------+
| count | person |
+-------+--------+
| 4 | bob |
+-------+--------+
| 4 | john |
+-------+--------+
| 3 | bill |
+-------+--------+
| 3 | mary |
+-------+--------+
| 2 | kev |
+-------+--------+
| 2 | louise |
+-------+--------+
| 1 | jay |
+-------+--------+
| 1 | steve |
+-------+--------+
We see here that bob has a count of 4 - because he is associated with 4 distinct event_id: 3 as a barman, and 1 as a doorman.
(assuming no two staff members have the same name, and no one can work two jobs at once)
How do I do this?
for one 'role' it's clear:
select count(event_id), bar group by bar
but is there an elegant way to do this for all columns - without full joins and string concat?
thanks!

Answer 1

You should change the structure of your data, so you have one row per event/person/role. Then you could just use aggregation.
You can do that in a query as well:
select who, count(*)
from (select event_id, 'bar' as job, bar as who from t union all
select event_id, 'doors' as job, doors as who from t union all
select event_id, 'cloak_room' as job, cloak_room as who from t union all
select event_id, 'keg_room' as job, keg_room as who from t
) jw
group by who;
If someone could have multiple jobs in one event, then use count(distinct event_id).
EDIT:
I see you are using Oracle 12c. Then use a lateral join/cross apply:
select who, count(*)
from t cross apply
(select t.event_id, 'bar' as job, t.bar as who from dual union all
select t.event_id, 'doors' as job, t.doors as who from dual from dual union all
select event_id, 'cloak_room' as job, cloak_room as who from dual union all
select t.event_id, 'keg_room' as job, t.keg_room as who from dual
) jw
group by who;

Answer 2

You may count by string columns in the nested inner query and then sum them up outside with your desired order :
SELECT sum(count) count, person
FROM
(
SELECT count(event_id) count, bar person FROM mytable GROUP BY bar UNION ALL
--> P.S. Only aliasing as "person" is enough in this upper "select" for all
--> four "select" statements inside the parentheses.
SELECT count(event_id) , doors FROM mytable GROUP BY doors UNION ALL
SELECT count(event_id) , cloak_room FROM mytable GROUP BY cloak_room UNION ALL
SELECT count(event_id) , keg_room FROM mytable GROUP BY keg_room
)
GROUP BY person
ORDER BY 1 desc, 2;
COUNT PERSON
4 bob
4 john
3 bill
3 mary
2 kev
2 louise
1 jay
1 jennifer
1 kyle
1 mary2
1 sarah
1 steve
SQL Fiddle Demo