I'm having a situation here where I have a web application backend library, based on NHibernate, which is shared among multiple web applications. Each of these web apps has each own, but very similar databases. The difference is: some tables might not exist in all databases.

The applications with the missing tables does of course not try to access these tables, but they all share the same .hbm.xml file (which is embedded in the backend library).

Now, the problem is that when NHibernate starts up and loads the persisters I get an error about missing object in database, even before the first real request for persistent objects. Is this correct, does NHibernate check the tables during initialization?

Is there an easy way to avoid this, like a configuration option or something?

NHibernate shouldn't check the tables on startup, can you provide a debug or info log to help understand what's happening?

I think I see the problem now. The class/table in question is mapped as a joined-subclass. The error might occur during the first query, not the startup.

The log tells me that NHibernate.Loader.Loaders runs a query that left outer joins the base class table with all the subclasses mapped in the .hbm.xml to figure out which subclass is requested. If I remove an unused table, the query will always fail.

I guess there's no option to avoid this, so I guess the solution will be to only keep the base class mapping in the shared library and then add required subclass mappings in the client apps as needed, although this is not as transparent as I would like to.

NHibernate has worked great for me so far, and I know the this problem might sound a bit weird, but there is a different application that relies on
checking which tables are present in order to load the respective modules.

Any other suggestions? Or just to re-design it?[/i]

That was me, sorry, my session timed out.