Hello everybody
i've been busy upgrading our n* stack to a more recent version. We'd been using FluentNhibernate for configuration and Nhibernate.search coupled with Lucene.Net for full-text search. Everything worked fine until i changed the various versions of the libraries to the following:
- FluentNHibernate.dll: 1.0.0.593
- NHibernate.dll: 2.1.0.4000
- NHibernate.Search.dll: 2.0.0.1001
- Lucene.Net.dll: 2.3.1.3
The stack has been working as previously, but i noticed something strange; whereas before the Lucene index directory would contain one subdirectory for each indexed class, it's now using two subdirectories, suffixed by a integer.
We went from
LuceneDirectory
.Class1
.Class2
to
LuceneDirectory
.Class1.0
.Class1.1
.Class2.0
.Class2.1
I poked a bit around the configuration building of FluentNhibernate and found out that for each mapping, i have two imports created, one with the fullname of the class mapped, one without (if i map the "User" class, i'll have one mapping, and two imports "POCOAssembly.User", and "User"). The directories are created at the Initialize call of the FullTextIndexEventListener.
Has anyone stumbled upon the same problem? i read through the nhibernate.search release notes, but didn't find any info regarding a change in mappings; is there something i'm missing? Has there been a breaking change in the libraries?
Edit:
I realized it may be important to point out that my mapped classes and my mappings derive from a common base object used for auditing purposes (creation date/update time).