Hello,

This is just a general question for getting more familiar with Hibernate System.

How Hibernate manage for dirty checking?

As in JDO, those people maintains a bitsets with every object and used to set the bit from every mutator method. And before saving, they ask from statemanager for dirty status of any bean, based on this bitset.

Is there any same algorithm in hibernate, or it is handled by some other way?

No, Hibernate has a more sophisticated approach that works not only for immutable objects, but also for mutable objects like arrays. (JDO dirty checking does not work for arrays.)

Hibernate uses snapshot comparison instead.

The other advantage of our approach is that it does not require buildtime bytecode manipulation.

... and the disadvantage being that it doesn't work for detached objects?

OTOH JDO doesn't support detached objects at all, so I guess this is not really a point in favour of JDO. ;-)

Not that I'm complaining, I can't come up with any sane way to do automatic dirty checking for detached objects.

Anyways, I opened a thread about how to do dirty checking for detached objects here: http://forum.hibernate.org/viewtopic.php?t=929910&highlight=

Actually JDO requires that the enhanced classes be serialization-compatible with unenhanced classes, so they can't use this approach for dirty-checking detached objects either.

Though I agree, it is theoretically possible.

And, indeed, if you wrote your own buildtime bytecode enhancer, you could use Hibernate as the persistence engine with this dirty checking strategy! Just implement Interceptor.findDirty() ... easy as pie...

That would be an interesting project for someone ;)

Hi

Sorry to bug about more JDO stuff, but I'm a student doing a schoolpaper on JDO vs Hibernate, and I get a bit confused about this runtime/buildtime bytecode processing.
I've also posted a question on the JDOCentral forum,


http://www.jdocentral.com/forums/index. ... #entry6269

and David Tinker seems to be of another opinion regarding wich is better of runtime buildtime processing.

Do anyone have any comments on what David and also Patrick Linskey has to say regarding bytecode processing.

regards dag