Question

I have that problem:
I have a model with those 3 tables:
Linha, Itinerario and Rua
Into Itinerario i have the reference id to Linha table and one reference with id for Rua.
In my code i recieve two arguments about idRua and i must return all Linhas where i have Itinerario the references for Rua with both idRua… In this examples where i have idRua = 1 and idRua = 2.

SELECT l.codigo, l.linha, l.idEmpresa, l.idLinha 
FROM Linha l 
INNER JOIN Itinerario i1 ON i1.idLinha = l.idLinha 
INNER JOIN Itinerario i2 ON i2.idLinha = l.idLinha 
WHERE i1.ida = i2.ida and i1.idRua = 1 and i2.idRua = 2 
ORDER BY l.linha

The problem is that i get 2 inner join in that table Itinerario and the query gets to slow…
Is there somehow to optimize it?
Is there some “IN” operator with “AND” condition or something like this?
Im using SQLite.

Answer 1

There may be a way to eliminate the self-join.

If I’m reading this correctly, you want idLinha’s from Itinerario that contain both idRua = 1 and idRua = 2 for the same ida. I’m noticing that this is just a filter condition, since everything in the select comes from linha.

The following gets this condition:

SELECT idlinha
From itinerario
GROUP BY idlinha, ida
having max(case when idRua = 1 then 1 else 0 end) = 1 and
       max(case when idRua = 2 then 1 else 0 end) = 1

Now, we can use this in an “in” or “join” clause, as in:

SELECT l.codigo, l.linha, l.idEmpresa, l.idLinha 
FROM Linha l
where l.idlinha in (SELECT idlinha
                    From itinerario
                    GROUP BY idlinha
                    having max(case when idRua = 1 then 1 else 0 end) = 1 and
                           max(case when idRua = 2 then 1 else 0 end) = 1
                   )
order by l.linha

It is possible that the group by will be faster than a self-join.