Test-driven semantical similarity analysis for software product line extraction

Abstract

Software product line engineering rests upon the assumption that a set of products share a common base of similar functionality. The correct identification of similarities between different products can be a time-intensive task. Hence, this paper proposes an automated semantical similarity analysis supporting software product line extraction and maintenance. Under the assumption of an already identified compatible interface, the degree of semantical similarity is identified based on provided test cases. Therefore, the analysis can also be applied in a test-driven development. This is done by translating available test sequences for both components into two I/O extended finite automata and performing an abstraction of the defined behavior until a simulation relation is established. The test-based approach avoids complexity issues regarding the state space explosion problem, a common issue in model checking. The proposed approach is applied on different variants and versions of industrially used software components provided by an automotive supplier to demonstrate the method's applicability.