Question

I use JPA with Hibernate implementation. My question is probably basic for experienced hibernate users: what is the most efficient way to write queries based on Hibernate first level cache?

For example, I have entity A and entity B:

@Entity
class A{
      private int ida;
      private int x;
      private String s;
    @OneToMany(mappedBy = "ida", cascade = CascadeType.ALL, fetch = FetchType.LAZY)
      private Set<B> Bset;
    }

@Entity
class B{
      private int ida;
      private String s2;
    }

Suppose I have several flows that may happen in the same session:

1. get A.x
2. get the whole A entity
3. check if A contains B with s2="...";

For every one of these queries I can

• write a specific query that gets A.x / B by ida and s2
  or
• suppose that Hibernate saves cache and always get A object, or A.Bset() and then java loops to get the needed B inside Bset()

What is the most efficient way?

thank you

Answer 1

For use cases 1 and 2 I would just load the whole entity. Second level cache will work equally well. Unless your A object is really big (has a lot of properties) you won’t see see any difference compared to:

SELECT a.x FROM A a WHERE a.id = :id

Even worse, using the query above will not take advantage of L2 cache.

Third use case is more interesting. It heavily depends on your requirements, but the reasonable balance would be to use a query like this:

SELECT B b
WHERE b.s2 == :s2
  AND b.a = :a

If the query returns something it means a contains b with given s2. This should be much faster than lazily loading Bset and iterating over it. Consider enabling query cache.

However if Bset is typically small and you use eager fetching, simple filtering in Java might be better. It really depends on your architecture.