Single Save with JPA CascadeType.PERSIST resulting to wrong order of insetion

Question

I have the following scenario:

class A {
 @OneToMany(mappedBy = "a", cascade = {CascadeType.PERSIST})
 List<B> bsOfA;
 @OneToMany(mappedBy = "a", cascade = {CascadeType.PERSIST}) 
 List<C> csOfA;
}

class B {
 @OneToMany(mappedBy = "b", cascade = {CascadeType.PERSIST})
 List<D> dsOfB;
 @ManyToOne
 A a;
}

class C {
 @OneToMany(mappedBy = "c", cascade = {CascadeType.PERSIST})
 List<D> dsOfC;
 @ManyToOne
 A a;
}

class D {
 @ManyToOne
 C c;
 @ManyToOne
 B b;
 @ManyToOne
 A a;
}

Then, I am attempting to save A along with all its associated entities as follows:

a = new A(aId);
b = new B(bId);
c = new C(cId);
d = new D(dId);
a.setBsOfA(List.of(b));
a.setCsOfA(List.of(c));
b.setDsOfB(List.of(d));
c.setDsOfC(List.of(d));
repo.save(a)

Now my problem is, I am getting a foreign key violation because the order of insertion is incorrect. Checking my logs, I am seeing the following order of insertion: a -> b -> d (error c not yet existing)

How can I enforce Hibernate to insert in the correct order: a -> b -> c -> d OR a -> c -> b -> d

My first inclination would be to suggest refactoring your persistence schema to avoid the kind of cycles that you show. I have no idea whether that's actually plausible for you, though, the entities presented being so abstract. — John Bollinger
– John Bollinger, Commented Sep 22 at 16:12
There shouldn't be a cycle. This might not be the best analogy but... A has 2 sets of children birthed from different partners: set B and set C. Now for some reasons, a member of b and a member of c births an incestuous offspring D... — Guppy_00
– Guppy_00, Commented Sep 22 at 16:20
You aren't showing exactly how you set D's reference to A, B, and Cs - are they set? If they aren't, no JPA provider can tell at the object level that D is dependent on A B or C being inserted first. What happens if you Save D and cascade from that side instead? You could persist/save A, B, C, Ds independently before setting the relationships if you need more control - but I know this works in other JPA providers like EclipseLink, so I'm sure there is something overlooked here too. — Chris
– Chris, Commented Sep 22 at 16:43
Actually, I am not setting them. I am only setting the OneToMany relationships since I assumed that those are where the cascade persists relationships depend on. — Guppy_00
– Guppy_00, Commented Sep 22 at 16:48
There is definitely a cycle, by which I mean in the graph theory sense: A->B->D->C->A. This is why you observe the problem you describe. I find such a schema pretty questionable, and I'm inclined to think that most real-world schemas having such a feature could be refactored to remove it. If yours is not such a schema, or if refactoring is impractical, then you're probably going to need to live with less cascading. — John Bollinger
– John Bollinger, Commented Sep 22 at 17:00

Guppy_00 · Accepted Answer · 2025-09-23 15:21:12Z

0

For anyone who might be encountering the same problem as me, the answer is to make sure all references are assigned correctly. Both forward and back. Also, I was initially wondering why a particular foreign key is causing problems while others are not. It turns out this particular foreign key is NOT NULLABLE while the others are so it didn't cause a problem. Initially, I was not setting the back references, I only set the references sufficient to perform cascade persist. Example, in the case above, I linked D to B so that when B is saved, D will also be saved. BUT, I did not link D to C so JPA had no idea that D is dependent on both B and C. Also, I removed the reference from C to D because in my case I didn't really need the relationship to be bidirectional.

answered Sep 23 at 15:21

Guppy_00

3071 gold badge2 silver badges9 bronze badges

Sign up to request clarification or add additional context in comments.

2 Comments

Chris Sep 23 at 16:10

I don't believe you needed to modify the object model - this wasn't a cycle. A cycle if when each table has a foreign key to another such as if A->B->C->A. OneToMany and M:M are not included in cycle dependencies as they are not holding FKs themselves. In your model D->C->A, D->B->A, and D->A. There is no cycle back from A to B, C or D.

Guppy_00 Sep 25 at 13:16

You're right. I did not make the modification because of there is no cycle in the directed graph. I just realized that I didn't really need it to be bidirectional so I removed it.

Collectives™ on Stack Overflow

Single Save with JPA CascadeType.PERSIST resulting to wrong order of insetion

1 Answer 1

2 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

2 Comments

Your Answer

Sign up or log in

Post as a guest

Related