How developers copy.pdf

How Developers Copy In Proceedings of International Conference on Program Comprehension (ICPC 2006), pp. 56–65 Mihai Balint1, Tudor G??rba2 and Radu Marinescu1 1LOOSE Research Group, Politehnica University of Timis?oara, Romania 2Software Composition Group, University of Berne, Switzerland Abstract Copy-paste programming is dangerous as it may lead to hidden dependencies between different parts of the system. Modifying clones is not always straight forward, because we might not know all the places that need modification. This is even more of a problem when several developers need to know about how to change the clones. In this paper, we correlate the code clones with the time of the modifica- tion and with the developer that performed the modification to detect patterns of how developers copy from one another. We develop a visualization, named Clone Evolution View1, to represent the evolution of the duplicated code. We show the relevance of our approach on several large case studies and we distill our experience in forms of interesting copy patterns. Keywords: code duplication, reverse engineering, soft- ware evolution, visualization 1 Introduction Code duplication appears because of several reasons: time pressure, shortcomings of the programming languages, inexperienced developers etc. No matter the reason, du- plicating the code introduces implicit dependencies be- tween the duplicated parts (named clones hereafter) because changing one part requires changing the other parts as well. Detecting code clones has long been in the center of the reverse engineering community [1, 6, 10, 12]. Most of the work in this area has been focused on the detection of prob- lems in the code and in relating clones with their impact in terms of quality assurance. Kim et al. made an interesting observation by exper- imentally showing that “refactorings may not always im- prove the software with respect to clones” [13]. Indeed, their empirical study showed that the short-lived clones a