Question

First, sorry if the title is confusing.
SQL is not my strong suit and I’ve been working on this for a while, my thoughts at the mmoment is something with a join, and group maybe.

Soto Example:

record | type | key1  | key2    | data1
---------------------------------------
1      | 1    | joe   | smoe    | 10
2      | 2    | homer | simpson | 20
3      | 1    | null  | null    | 30
4      | 3    | bart  | simpson | 40 

Where primary key is made up of id, key1, key2.

I only want rows of ‘type’ WHERE key1 is not null AND key2 is not null.

So since in record 3, type 1 has null keys, I therefore want all records of type 1 to not be included in the derived table.

Answer 1

Here’s a correlated, “not exists” approach:

select *
from T as t1
where not exists (
    select *
    from T as t2
    where t2.type = t1.type and (t2.key1 is null or t2.key2 is null)
)

And here’s one that uses a non-correlated query along with grouping. Perhaps it’s what you had in mind:

select *
from T as t1
where t1.type in (
    select t2.type
    from T as t2
    group by t2.type
    having count(*) = count(t2.key1) and count(*) = count(t2.key2)
)

Since I understand mysql query plans can be sensitive to these things. Here’s the equivalent with a join:

select t1.*
from T as t1
    inner join
    (
        select t2.type
        from T as t2
        group by t2.type
        having count(*) = count(t2.key1) and count(*) = count(t2.key2)
    ) as goodtypes
        on goodtypes.type = t1.type