Understanding transaction session with lazy loading in Spring JPA Hibernate

Question

I would like to get some clarification regarding lazy loading and session boundaries etc.

My code structure is as follows

@Entity

class A {

....

  @OneToOne(fetch=LAZY)
  private B b;

  ..
}

@Entity
class B {

 private id;

 private name;    

}

@Transactional(SUPPORTS)
ADao {
  A findById(int id);  
}

@Transactional(SUPPORTS)
LayerDB {

    A getAForId(int i) {
      return adao.findById(i);
    }

}

//Note that there is no transactional attribute here
LayerB {

   public boolean doSomethingWithAandB(int aId) {
    A a = LayerDB.getAForId(aId);
    if(a.getB().getName().equals("HIGH"))
     return true;
    return false;
   }

}

//start transaction here
@Transaction(REQUIRED)
LayerC {

    LayerB layerb;

    private handleRequest(int id) {

       layerb.doSomethingWithAandB(id);

    }
}

Now when we try to access B in entity A within the method

doSomethingWithAandB

Am getting a lazy initialization exception when trying to access B.

Even though the method is within the transaction created in LayerC, still i get the following exception

Exception : org.hibernate.LazyInitializationException: could not initialize proxy - no Session

But on changing the following two methods as :

@Transactional(SUPPORTS)
LayerDB {

   A getAForId(int i) {
      A a = adao.findById(i);
      a.getB().getName();
      return a;
    }

}

//Note that there is no transactional attribute here
LayerB {

   public boolean doSomethingWithAandB(int aId) {
     A a = LayerDB.getAForId(aId);
     if(a.getB().getName().equals("HIGH"))
     return true;
    return false;
  }

}

Why is it not using the transaction / session created in LayerC ?

Even though we have SUPPORTS on the DBLayer, is it creating a separate 'session'.

Any pointers for proper understand would help me a great deal.

Thank you.

Answer 1

With lazy loading, when you request an object a of type A, you get an object a of type A. a.getB() however, will not be of type B, instead a.getB() is a proxy for B that can be resolved later on (that's the lazy loading part), but only in the persistence context in which a lives in.

Your second implementation does just that: it resolves B by calling a.getB().getName() while you are still in the @Transaction. Hibernate can now make a second request to the database to fetch B, and now a.getB() is really of type B and stays that way, so you can use it outside the persistence context.

Your first implementation skips that. A is fetched from the database, the @Transactional block ends, then you call a.getB().getName(), but now the persistence context is gone, a.getB() can not be fetched from the database, and an exception is thrown.