The code as a source for models

I am taking a rest after some hours of work. I used the javassist library to extract information from the code of a layer in an enterprise application and use it to build a model to generate the successive layer. I wish I could extract some of the information using generics but the infamous “Java 1.4 only” constraint plagued me. Anyways, to use generics I would need to refactor a little bit the layer I am using as source of my model (aka the “source layer”) and I can’t.

There were moments where I wanted to bang my head so strongly on the wall but after some efforts I managed to extract the information in the 75% of the cases I needed to. Anyway it is information that comes for free because it is not expressed purposely to build my model, it is just normal code. The problem is that I can not obtain all the information automatically, I still need a human refinement of the model before I start the generation of the target layer. So I am considering if the information I obtained is worthy because it saved the 75% of the human effort needed to build the complete model or it useless because it is not enough to support automatization.

Mumble, mumble… opinions?