I made a module that.. essentially is the base for an 'indexer'. It's beta but quite functional. Has tests etc.. I call it Metadata::ByInode , you inherit it and then you can overrite the method index_extra() - it gets passed a hash with the current file's path info (absolute path, filename )- you can use that to _set() whatever you want- like inspect the file's mime and feed it back to the indexer. I'll put it on cpan and post a link. It has examples and is fully documented.

Metadata::ByInode