A comparative study of test code clones and production code clones

Abstract Clones are one of the most widespread code smells, known to negatively affect the evolution of software systems. While there is a large body of research on detecting, managing, and refactoring clones in production code, clones in test code are often neglected in today’s literature. In this paper we provide empirical evidence that further research on clones in test code is warranted. By analysing the clones in five representative open-source systems and comparing production code clones to test code clones, we observe that test code contains twice as many clones as production code. A detailed analysis reveals that most test clones are of Type II and Type III, and that many tests are duplicated multiple times with slight modifications. Moreover, current clone detection tools suffer from false negatives, and that this occurs more frequently in test code than in production code (NiCad = 76%, CPD = 90%, iClones = 12%). So even from a tools perspective, specific fine-tuning for test code is needed.