Automatic extraction of product line architecture and feature models from UML class diagram variants

Abstract Context Software Product Lines (SPLs) are families of related products developed for specific domains. SPLs commonly emerge from existing variants when their individual maintenance and/or evolution become complex. Even though there exists a vast research literature on SPL extraction, the majority of the approaches have only focused on source code, are partially automated, or do not reflect domain constraints. Such limitations can make more difficult the extraction, management, documentation and generation of some important SPL artifacts such as the product line architecture, a fact that can impact negatively the evolution and maintenance of SPLs. Objective To tackle these limitations, this work presents ModelVars2SPL (Model Variants to SPL Core Assets), an automated approach to aid the development of SPLs from existing system variants. Method The input for ModelVars2SPL is a set of Unified Modeling Language (UML) class diagrams and the list of features they implement. The approach extracts two main assets: (i) Feature Model (FM), which represents the combinations of features, and (ii) a Product Line Architecture (PLA), which represents a global structure of the variants. ModelVars2SPL is composed of four automated steps. We conducted a thorough evaluation of ModelVars2SPL to analyze the artefacts it generates and its performance. Results The results show that the FMs well-represent the features organization, providing useful information to define and manage commonalities and variabilities. The PLAs show a global structure of current variants, facilitating the understanding of existing implementations of all variants. Conclusions An advantage of ModelVars2SPL is to exploit the use of UML design models, that is, it is independent of the programming language, and supports the re-engineering process in the design level, allowing practitioners to have a broader view of the SPL.