Metric Evolution Maps: Multidimensional Attribute-driven Exploration of Software Repositories

Abstract

Understanding how software entities in a repository evolve over time is challenging, as an entity has many aspects that undergo such changes. We cast this problem in a multidimensional visualization context: First, we capture change by extracting quality metrics from all software entities in all revisions in a software repository, yielding a multidimensional time-dependent dataset. Next, we propose Metric Evolution Maps (MEMs), a new visual approach to create dynamic maps that show the similarity of entities in a revision and changes across revisions. We enrich MEMs with visual cues to show which metrics and metric values are key to formation of similar-entity patterns. Additionally, we show how entities change between revisions, and due to which metrics. We illustrate our approach by exploring changes in two real-world software repositories.