Hi,

By design, NHibernate doesn't (second level-)cache collections that have filters enabled and I think I understand the difficulties this would cause.

But I noticed a comment in the java version that suggests this might be a possibility by including filter info in the cache key. Do you think this is feasible? Would it be a difficult change? I'd love to contribute but would need a lot of help.

log.debug( "Refusing to add to cache due to enabled filters" );
// todo : add the notion of enabled filters to the CacheKey to differentiate filtered collections from non-filtered;
// but CacheKey is currently used for both collections and entities; would ideally need to define two seperate ones;
// currently this works in conjuction with the check on
// DefaultInitializeCollectionEventHandler.initializeCollectionFromCache() (which makes sure to not read from
// cache with enabled filters).

If this change isn't feasible or isn't on the radar, do you have any suggestions for caching a subset of data in a one-to-many relationship, specifically with temporal data. I really need the flexibility to shift date perspective, but still get the performance benefits of caching.

Thanks a bunch.

Matt
matt_cameron@invesco.com

To get started on the NHibernate source code, you may have a look here:
http://www.hibernate.org/428.html

To port a Java feature, I would suggest checking out the Hibernate trunk and port over the related tests first. Assuming that you are not familiar with the neither the Hibernate or NHibernate code base, you may have to debug Hibernate and NHibernate side-by-side to identify the differences. I personally use Eclipse for Hibernate, but I am sure any Java IDE would do.

There is a learning curve involved for sure, but it is not as steep as one might think. If you run into a brick wall, you can always ask questions on the NHibernate-Development mailing list. Provided that the questions are related to the development of NHibernate itself, one usually gets a response pretty quickly.

Hope this helps.