Hibernate version: 1.0.2

Hi all,

In my design, the following relationships exists:
EntityA has a one to many relationship with EntityB, which in turn has a one to many relationship with EntityC.

The number of objects involved runs into hundreds at best, ten thousands at worst. EntityA has to be passed through a webservice, so the entire objectgraph needs to be loaded.

Since this is rather costly, I was thinking of reversing the relations into a many to one relations: EntityB has a many to one to EntityA, EntityC has a many to one to EntityB, and EntityA knows nothing (in a mapping sense) about either EntityB and EntityC.

Functionally, this is perfectly acceptable, since users don't work with the entire objectgraph at once anyway, but I was wondering if this is considered an acceptable or (despised) practise from a design and NHibernate point of view.

Any thoughts?

This is perfectly fine from NHibernate's point of view, see http://blog.hibernate.org/cgi-bin/blosxom.cgi/2006/01/25#collectionmapping. As from the design point of view, I would say it depends. In your case I think the compromise is justified.

Thanks for corroborating my thoughts on the subject (and the link).

I'll take the many-to-one route. The design incurs a slight hit by dropping the link, but I too felt it to be a justifiable choice.