I do the follwing in my app:

- create once (static) a SessionFactory

- for all Get /Create (Save) / Update / Delete I create a new session, do my command and close it

Lets take the classic customer : order = 1 : n example: if I get all customers and all orders once from the DB, then I have for each record one object in memory. If I e.g. make a lazy loading for one Customer, it gets me only the Orders from this one. The problem is, because I already loaded them before (with the get all orders statement), I now have the orders from this customer twice in the memory.

My solution to that problem currently is, that I lock all orders from which I know that they are orders from that customer in the new session, and do then the lazy loading for the customer. So I still have only one object representation of each record in memory.

In the forum I was told, I should use the 2nd lvl cache, which sounds to be a perfect (and more professional :) ) solution. So I did everything like it says in the Wiki:

App.Config: <add key="hibernate.cache.provider_class" value="NHibernate.Caches.SysCache.SysCacheProvider, NHibernate.Caches.SysCache" />

All classes and bags: <cache usage="read-write" />

It compiles and runs without error, but it looks like it doesn't work. Any clue what I do wrong?

Another question about cache expiration: I want them as long as my session factory lives in the cache. Is this the default setting?

Thanks alot, Rainer.

Rainer,

The SysCacheProvider will only work when your assembly is running within the ASP.Net engine. If you want to use a second level cache in a WinForm app you will need to use the Prevalence library. Unfortunately there isn't much documentation on how to use it. Sorry.

Well, not much can be understood from such problem description :) But from the rest of the message, it looks like you are trying to ensure that there's only one instance of some object across your entire application. This is rather difficult to do with NHibernate and second-level cache won't help you with this. Second-level cache stores disassembled copies of objects, that is, just arrays of property values. If an object is found in second-level cache, a new object is constructed and initialized (assembled) from the disassembled values, so what you get is a fully independent copy of the original object.

The reason why second-level cache stores disassembled objects and gives out copies of objects is to avoid concurrency issues.

The lack of documentation is entirely my fault. I have been extremely busy lately. Please look at the test cases in CVS and see if that helps.

Have you tried it? What error does it give you? The tests don't run under ASP.NET and they pass, so I thought it would work.

The only property that is configurable is "prevalenceBase", which is the directory used to cache the objects. It will default to the current directory.

Again, the lack of documentation is my fault and I apologize for that.

k-dub, I assumed that Prevalence wasn't documented because no one uses it.

When I first started using SysCache the logs never showed that an object was placed or retrieved from second-level cache when my DAL was used in a non-ASP.Net app. No error was thrown, it just didn't look to be used. I'll have to try it again.

Sorry :oops: .



Hmmm. That means that my locking (i described it in more detail here: http://nhibernate.sourceforge.net/forum ... 2&start=15) is the best practice for my problem?

Just to confirm: SysCache works in Winforms all right. You should see it in debug log - something like: