I modified my code today to use NHibernate.Generics. So far I’m very pleased with the simplicity it brings to my code. However, there are some things that I discovered:

• Always initialize the EntitySet & EntityRef in the default constructor, and make sure that all your constructors call to the default constructor.
• Always use EntityRef when you are connecting persistent objects, even if currently the connection is one sided. Sometime in the future you may need to make the connection bidirectional and it might save you some pain. Actually, this advice is wrong. The problem is not switching from normal property access to EntityRef. That is quite simple; the problem is that it may be confusing to have some EntityRef and some domain objects. If everything is an EntityRef, there are no problems, it’s clear that there is no connection since the EntityRef’s change & add delegates are empty.
• Do not use the EntityRef’s constructor overload which pass an initial object, this overload is going to go away, and it’s bad idea to use it since then the delegates are not called on it.

I added the ability to specify a custom access strategy to Active Record and now you can do things like this:

[HasMany(typeof (Post), "PostId", "Posts", Inverse=true,

CustomAccess = Generics.Access,

RelationType = RelationType.Set)]

public ICollection<Post> Post

{

      get { return  post; }

}

As you can see, it’s one big ass attribute, but you specify it once, and then just forget about it.

Some notes about it:

• Generics.Access – This is a static class with a const field that contains “NHibernate.Generics.GenericAccessor, NHibernate.Generics”. It makes the code much more readable, IMO.
• RelationType – Active Record can no longer guess at the type of the relation, so you need to tell it explicitly.