Hi all,

Assume:
- SubClassA extends BaseClass.
- SubClassA implements Lifecycle but BaseClass does not.

If BaseClass is mapped as lazy=true (default) (and is not proxied, not final etc.) then you get back a CGLIB enhanced class, and the Lifecycle onLoad event is *not* called. If you set lazy=false then you get back the actual subclass and the events are called.

FYI.

This is a problem of using proxy when you have inheritance.

In some situations, the cglib proxy, proxies the base class and then is not trully a subclass of your subclass.

This leads to problems with usage of 'instanceof'. So to avoid this problem you have to put lazy=false on your classes.

Also note that Lifecycle is deprecated in hibernate3. Prefer using interceptor in hibernate2 or events in hibernate3 to achieve what you want to do

Hi Seb.

Yeah, I ran into the "instanceof" problem when I realised the CGLIB class implements *all possible subclasses* :)

Problem with lazy=false is all single-point-associations are eagerly loaded :( In my domain that would be disastrous.

Haven't looked at interceptors.....

Are you sure this would be disastrous. Did you test it ? Single point associations are often fetch with join then you don't have more selects. Although, you just have to do that on classes that have inheritance. you can also use second level cache to improve performance on bottlenecks.
Check the difference and this if this is truly disastrous. (in my case, this is not)

It is in my case ;)

I have ObjectA one2one objectB one2one objectC.

I also have many (10s-100s) of instances of ObjectA that never need access to objectB.

Just curious, what version of Hibernate are you using ?

3.0.5

If you use session.get() instead of session.load(), you don't get a proxy for your top-level object and your onLoad method will be called.

I'm not sure what all the differences are otherwise. I believe that load() will throw an Exception if the ID isn't found and get() will just return null, at least that was the behavior last time I looked into it.