Hi there,

I am a bit lost. I don't know how to model this right.

Scenario:

* Objects A,B,C
* Object B is a child of A (so A.getBs() and B.getA(), one-to-many)
* Object B reference C (unidirectional assoziation).

Here comes the problem:
Since B reference C, the reference should be set to null if C gots deleted. (SQL: on delete set null).

But I was not able to get things right on first try. So I used a hack using an integer column (primary key of C) without a foreign key constraint. But now it's time to make things right.

So here are the options I can think of:

-Add a second bidirectional mapping between B and C (C is parent) and use onDelete methods on both side to remove the link before either B or C gets deleted (adds the burdon of utilizing the lifecycle, which I think is not such a good option).

-Life with the hack of integer and ensure that if any new C is generated using the same id, things gets corrected (logical link, but this smells that kind of badly that I would like to avoid this).

-Using an unidrectional mapping but again I ran into the problem, how I can create a behaviour like 'on delete set null'.

How is the 'best' way you know, this can be modeled using hibernate?


Thanks

I implemented it using bidirectional mapping and also utilizes the lifecycle method onDelete() on B and C. So C removes all the Bs from its set if it gots deleted and B removes itself from C on its deletion. Works well but I am still suffering from being forced to introduce a feature (C knows about the Bs) without the need of it.

An other idea is to drop one side of the mapping and alter the onDelete method of C. So all what C has to perform is to find every B referencing the current C and dereference it. But again C has to be aware of B. But the set disappears. Is it that kind of smart? I had to introduce hibernate specific queries to the POJO. I would like to avoid it since I try to decouple the hibernate aspect as much as possible.

Any ideas on that?