MySQL find entries matching multiple pivot rows - mysql

Question

I have a schema with 3 regular tables and 2 link tables like so:
tutorial
id
exam
id
user
id
exam_user
user
exam
tutorial_user
user
tutorial
Where the link tables determine which users have attended which exam/tutorial.
Starting with an exam ID, I'm trying to find tutorials whose members have ALL attended the exam.
For example:
exam_user
+----+-------+----------+
| id | exam | user |
+----+-------+----------+
| 23 | 8 | 1 |
| 24 | 8 | 5 |
| 25 | 8 | 8 |
| 26 | 8 | 11 |
| 27 | 8 | 12 |
+----+-------+----------+
tutorial_user
+----+----------+----------+
| id | tutorial | user |
+----+----------+-----------
| 56 | 2 | 1 |
| 57 | 2 | 5 |
| 58 | 2 | 8 |
| 59 | 2 | 11 |
+----+----------+----------+
+----+----------+----------+
| id | tutorial | user |
+----+----------+-----------
| 60 | 3 | 1 |
| 61 | 3 | 5 |
| 62 | 3 | 8 |
| 63 | 3 | 15 |
+----+----------+----------+
Tutorial 2 would match because all entries are in exam 8, but tutorial 3 would not because user 11 is not present.
Is there any (relatively) efficient way to achieve the above?

Answer 1

How about you try something like this:
select * from tutorial_user where tutorial not in
(
select tutorial from
(
(
select ex.exam, ex.user as ex_user, tut.tutorial as tutorial,
tut.user as tut_user from exam_user ex
inner join tutorial_user tut on
ex.User = tut.User
)
union all
(
select null, null, tutorial, user as tut_user from tutorial_user
)
) src
group by tutorial, tut_user
having count(tut_user) = 1
)
Now the first query should give you a list that looks like this (essentially a combination of exams and tutorials for every user):
+----+-------+----------+-------------+---------+
| id | exam | user | Tutorial | User |
+----+-------+----------+-------------+---------+
| 23 | 8 | 1 | 2 | 1 |
| 23 | 8 | 1 | 3 | 1 |
| 24 | 8 | 5 | 2 | 5 |
| 24 | 8 | 5 | 3 | 5 |
...
Next, by doing a union all with the next query, you'll end up with something like this:
+-----+-------+----------+-------------+---------+
| id | exam | user | Tutorial | User |
+-----+-------+----------+-------------+---------+
| 23 | 8 | 1 | 2 | 1 |
| 23 | 8 | 1 | 3 | 1 |
| 24 | 8 | 5 | 2 | 5 |
| 24 | 8 | 5 | 3 | 5 |
| null| null | null | 2 | 1 |
| null| null | null | 2 | 5 |
| null| null | null | 2 | 8 |
...
Basically, what this allows us to do is a group by statement that's going to highlight users in the tutorial table and not in the exam table. The said highlight is going to be with the having() clause. If the count is 1, then you have someone in the tutorial table that's not in the exam table. All you have to do now is select the tutorials from the tutorial table that are NOT part of what you found and you should be good.