Studying Co-evolution of Production & Test Code Using Association Rule Mining

Unit tests are generally acknowledged as an important aid to produce high quality code, as they provide quick feedback to developers on the correctness of their code. In order to achieve high quality, well-maintained tests are needed. Ideally, tests co-evolve with the production code to test changes as soon as possible. In this paper, we explore an approach to determine whether production and test code co-evolve synchronously. Our approach is based on applying association rule mining to the change history of product and test code classes. Based on these co-evolution rules, we introduce a number of measures to assess the co-evolution of product and test code classes. Through two case studies, one with an open source and another one with an industrial software system, we show that association rule mining and our set of measures allows one to assess the co-evolution of product and test code in a software project and, moreover, to uncover the distribution of programmer effort over pure coding, pure testing, or a more test-driven-like practice.