True Variability Shining Through Taxonomy Mining

Software variants of a Software Product Line (SPL) consist of a set of artifacts specified by features. Variability models document the valid relationships between features and their mapping to artifacts. However, research has shown inconsistencies between the variability of variants in features and artifacts, with negative effects on system safety and development effort. To analyze this mismatch in variability, the causal relationships between features, artifacts, and variants must be uncovered, which has only been addressed to a limited extent. In this paper, we propose taxonomy graphs as novel variability models that reflect the composition of variants from artifacts and features, making mismatches in variability explicit. Our evaluation with two SPL case studies demonstrates the usefulness of our variability model and shows that mismatches in variability can vary significantly in detail and severity.