Reverse Engineering Feature Models from Programs' Feature Sets

Successful software is more and more rarely developed as a one-of-a-kind system. Instead, different system variants are built from a common set of assets and customized for catering to the different functionality or technology needs of the distinct clients and users. The Software Product Line Engineering (SPLE) paradigm has proven effective to cope with the variability described for this scenario. However, evolving a Software Product Line (SPL) from a family of systems is not a simple endeavor. A crucial requirement is accurately capturing the variability present in the family of systems and representing it with Feature Models (FMs), the de facto standard for variability modeling. Current research has focused on extracting FMs from configuration scripts, propositional logic expressions or natural language. In contrast, in this short paper we present an algorithm that reverse engineers a basic feature model from the feature sets which describe the features each system provides. We perform an evaluation of our approach using several case studies and outline the issues that still need to be addressed.