Recovering Architectural Variability from Source Code

Abstract

Context: Systematic variability management helps efficiently manage commonalities and differences in software systems (e.g., in software product lines and families). This enables the reuse of development artifacts in organizations and increases the quality of product variants. In software product lines, the product line architecture (PLA) is the core architecture for all product line variants. In practice, software architectures are often not documented in detail. Architecture recovery techniques can recover a system's architecture from development artifacts (e.g., source code). To recover the architecture of product lines, we need recovery techniques that are able to identify variability from different sources. Goal: We present SAVaR, an approach to recover architectural variability from the source code of product variants of a product line. SAVaR aims to help developers to (a) create architectural documentation for a product line, and (b) understand and improve the implementation of variability. SAVaR identifies the smallest subset of architectural information that is common across products of a product line. To limit the explosion of variability (and hence the complexity of architecture documentation) in the product line architecture, SAVaR allows architects to exclude architecture elements that appear in only a few product variants. Method: We performed an exploratory study with SAVaR to recover the architectures in ten academic product line projects. We verified how the elimination of exclusive optional modules improves the results of SAVaR. Results: The results showed that SAVaR is able to present improvements for the recovered PLAs and it helped to identify that some projects maintained the variability under control.