Feature and variability extraction from Agile specifications and their related source code for software product line migration

Abstract

Migrating a set of similar software products into a Software Product Line is a time-consuming and costly process which, ultimately, provides an important gain in time and customization. Conducting this migration within an agile development process is a complex process which requires discipline and adaptation. We think it can be beneficial to drive the migration by leveraging agile software specifications and the source code versioning platform. Currently, we are working on a method, whose design is explained in this paper, which exploits: (1) Epics and User stories to identify features and variability and (2) the source code associated to code merges related to User stories and Epics to locate them. We plan to extract features and variability inside Epics and User stories using Natural Language Processing (NLP) techniques. Then we plan to investigate how formal concept analysis (FCA) and relational concept analysis (RCA) can assist feature model synthesis and establish mappings between features and source code. These knowledge discovery methods have been chosen for their ability to highlight and hierarchically organize groups of similar artefacts. FCA only considers artefact description to establish groups of similar artefacts. RCA groups similarly described artefacts that, in addition, share similar relationships to other artefact groups. We also plan to evaluate the method within the context of a company (ITK) with which we collaborate, using its code base and the associated project management artifacts. We also will assess how the method can be generalized to public projects in source code versioning platforms.