Hi, I'm using the <Any> mapping in my project and I just realized that the information that get inserted in the DB is the fully qualified name of the class.

This is a big problem with version updates. If I upgrade my software version I will have to update every row with the new version number. Is there a way NOT to use the FullyQualifiedName?? Should I rise a jira issue? This is kind of urgent for me since we are scheduled to deploy the application in january.

Thanks!

Hey roniburd,

See the posts I made on this subject:
http://nhibernate.sourceforge.net/forum/viewtopic.php?t=1093
http://nhibernate.sourceforge.net/forum/viewtopic.php?t=1101&highlight=manytoany

I would have preferred to find a way where I didn't have to explicitly specifiy the descriminators, and got an answer I couldn't understand on the JIRA issue I raised about it after the second thread. At the end of the day I was forced to use the meta-value tag to specify a discriminator.

Hope this helps,

Symon.

What Peter meant in that comment is that you should specify a fixed version for your domain assembly.

Here is another solution you can try. The piece of code that writes assembly-qualified type names to the database is TypeType.NullSafeSet. You can try deriving your own type from TypeType, overriding its NullSafeSet and NullSafeGet operations and specifying your new type as the meta-type for the <any> tag.

Thanks you both...

Sergey, forcing a version freeze on my domain isn't a bit extreme? I understand I could build my own type, but that seems a bit overkill just for droping the fully qualified name.

After all .NET can (AFAIK) instantiate by reflection without using the fully qualified name given an assembly. Could this feature be added?

What are other people doing??

I think what Peter was getting at is not actually doing a version freeze, but setting something in the assembly to force it to report a specific version all the time rather than the version that would normally be added to a build.

He gave me a couple of URLs to get information from but I didn't have the time to dig into it in any detail, so I walked away. If you want to read about it you can have a look at the JIRA issue I reported and Peter's response here: http://jira.nhibernate.org/browse/NH-422

If you figure it out let me know!

Cheers,

Symon.

Why is it overkill? As I said, it only means overriding two methods (maybe only one). Writing your own types is one of the means of customizing NHibernate. Alternatively you can create a JIRA issue and have a vote there for it to be changed in the next major/minor version (probably 1.1). I personally don't want to change this in the 1.0.x line.

It's been a while since I looked at this and I'm wondering whether anyone out there can explain how Peter intended that a fixed (non-running) version be specified for a domain assembly. Is there a way for the version to be set specifically so that it will only affect the type names as they are persisted?

I've re-read the links Peter supplied, but they seem to have more to do with remoting and I can't figure out how they apply to this problem.

Can anyone help clarify this?

Cheers,

Symon.