Question

I’m using SQLite.
I need an help for an easy issue.
Here’s my three tables:

--------------
problem
--------------
id (primary key)
question_id (foreign key)

--------------
question
--------------
id (primary key)
answer_id (foreign key)

--------------
answer
--------------
id (primary key)

I would like to get ALL problems that have at least N answers in every question of a problem. I’ll give you an example:

-------
problem
id 
1
2


-------
question 
id   problem_id
1    1
2    1
3    1
4    2

-------
answer
id   question_id
1    1
2    1
3    1
4    2
5    2
6    3
7    4
8    4

If n=2, my result should be problem_id=2.

I’ve tried this:

   select distinct question.problem_id 
   from answer, question
   where answer.question_id = question.id
   group by answer.question_id
   having count(*) >= 2

but it doesn’t work because it gets problems with at least one question with at least 2 answers. All questions must satisfy that condition.

Any problems?

Answer 1

Here’s my go at the issue in T-SQL:

declare @problem table(id bigint not null primary key clustered)
declare @question table(id bigint not null primary key clustered, problem_id bigint)
declare @answer table(id bigint not null primary key clustered, question_id bigint)

declare @n int = 2

insert @problem
      select 1 
union select 2

insert @question
      select 1, 1 
union select 2, 1
union select 3, 1
union select 4, 2

insert @answer 
      select 1, 1 
union select 2, 1
union select 3, 1
union select 4, 2
union select 5, 2
union select 6, 3
union select 7, 4
union select 8, 4

select p.id --, p.name, p.description, p.etc
from @problem p
where @n >= ALL --http://msdn.microsoft.com/en-us/library/ms178543.aspx
(
    select COUNT(a.id) 
    from @question q
    left outer join @answer a
        on q.id = a.question_id
    where p.id = q.problem_id
    group by q.id
)

NB: the table schema differs slightly from the question as the schema in the question doesn’t match the example data.

ALTERNATIVE

(based on @RichardTheKiwi’s answer with the inner SQL moved to a temp table)

declare @tempTable table (pid bigint, qid bigint, aidCount bigint)

insert @tempTable
select q.problem_id, q.id, count(a.id) answercount
from @question q
left join @answer a on a.question_id = q.id
group by q.problem_id, q.id

select pid
from @tempTable
group by pid 
having min(aidCount) >= @n