I persisting a graph with Hibernate 2.1. I have few objects that need to belong to only one of many objects at a time. In other words:

A (1) ---- (1) C OR B (1) ---- (1) C

If casacading is needed this will behave like:

A (1) ---- (1) C (1) ---- (1) B

meaning that C will belong to BOTH A and B at the same time.

One way to design these associations is using many-to-many:

A (*) ---- (*) C (*) ---- (*) B

Is there a better way to model these associations?

mota

I'm not sure I understand your question correctly, but why many-to-one mapping from A to C (and many-to-one from B to C) doesn't work?

In my design C needs to belong to either A or B. This means that C holds the foreign keys from A and B. If I need Hibernate to handle cascading (save or update) both these froeign keys should be not null. Therefore C must belong to BOTH A AND B at the same time. I need C to belong to EITHER A or B and have cascading.

This corresponds to the following case:
Customer has an Address
Company has an Address
Order has an Address

the same persistent entity Address can belong to a Customer, a Company, OR an Order. I also need cascading save/update from a parent to Address. One way to handle this is to use many-to-many instead of one-to-one. Is there a better suggestion?

mota

Look, some time I had a similar problem - Customer, Employee and alot of other entities had a Location. Not sure if it was a correct way, but I didn't specify foreign keys from Location to all other tables. All other entities had foreign key to Location.

Yes, in this scenario Location did not know which object it belongs to. But it wasn't necessary in my case.

Btw, if you need reverse mapping (from Address to Order or Company), you may consider using <any> mapping.

But again, this isn't really necessary if all you need is cascading updates from Company/Order to Address.

In my case Employee mapping looks like:



And Location class does not know anything about Employee/Company