After reading and experimenting with the 2nd level cache quite extensively, I've gotten to a situation where I want to simply count the number of children of an object, but not initialize those children. In fact, I have the join cached in a collection cache and it would be totally fine with me if I could simply get a size() call from the collection in the cache (i.e. no actual data from the child objects). However, I can't seem to find how to do this. Even when the getChildren().size() results in a collection cache hit the full list of child objects is initialized to complete the size() call.

Other things I have tried:

1. SQL only solution - didn't work due to performance and SQL version limitations
2. Making the child object a proxy - didn't work since the full objects still seem to get initialized by the size() call
3. Just getting by with caching everything, including the child objects - doesn't work in my case because there is too much data, and even loading from cache those child objects is too slow in this use

Does anyone have any ideas or alternate ways to get a size of that collection without initializing it?

Thanks for any advice.

Matt

Thanks for the quick reply Michael, I should have been more specific. That method came under #1 of the methods I tried since it generates SQL to count the size of the collection. Unfortunately the table that containts the elements of that collection is so large that running the necessary series of "select count(*)" statements is just too slow.

So really, what I seem to need specifically is a way to get a size() of the collection that's in the collection cache without initializing the underlying collection and it's objects. That seems to be the only way that'll be fast enough in this case.

Thanks for any other input.

Matt

P.s. Just as a quick background on the problem, on application startup I cache all the objects related to this problem so when they're accessed later they achieve 100% cache hit rates. However, it's too slow even to initialize the child collection from the cache when really all I need is the size of the collection. So that's how I got to this point.

I have done this and some other related things a few times using a formula property. In terms of performance it works out much better.

http://www.hibernate.org/hib_docs/reference/html/or-mapping.html#or-mapping-s1-7; check out the formula attribute

Thanks for the info Steve, I hadn't seen the formula attribute before. One thing I'm concerned about is caching that object with the formula property, and whether that property will become incorrect as the tables the property is based on change. Also, the 2.1.1 release notes say:

* don't cache objects with formula properties immediately after insert/update

So does that mean that no object with a formula will be cached? I guess I'm just not clear on cache interaction with the forumla property in general.

Thanks for the tip, let me know if you have any experience with the above.

Matt

No, it means the object will be evicted and recached next time it is used.

Gavin,

Does it make sense to you that perhaps a size() or empty() call to a persistent collection should return the size of the cached array if found rather than initializing the whole collection? I'm figuring that the size() call initializing the colleciton comes from the days of no second level collection cache and is probably still around since it's mentioned in the docs as a way to initialize collections. Is that correct?

Is there another reason for it to still happen that I'm not seeing from the code? Thanks for the help.

Matt