Automatic extraction of features and generation of feature models from java programs

Abstract

Feature modelling is a key technique for identifying common and variable features in software (software component families). The result of feature modelling is a feature model: a concise specification of product features and their relationships. Feature models have been proven to be useful for software variability modelling and management. However, there is a wide gap between feature models and program source code. Here we focus on reverse engineering of source code to feature models. We present a framework for the automated derivation of feature models from the existing software artefacts (components, libraries, etc.), which includes a formal description of a feature model, a program-feature relation meta-model, and a method for feature model generation based on feature dependency extraction and clustering. Feature models are generated in Feature Description Language (FDL) and as Prolog rules.