Change Impact Analysis of Code Clones
Copying a code fragment and reusing it with or without modifications is known to be a frequent activity in software development. This results in exact or closely similar copies of code fragments, known as code clones, to exist in the software systems. Developers leverage the code reuse opportunity by code cloning for increased productivity. However, different studies on code clones report important concerns regarding the impacts of clones on software maintenance. One of the key concerns is to maintain consistent evolution of the clone fragments as inconsistent changes to clones may introduce bugs. Challenges to the consistent evolution of clones involve the identification of all related clone fragments for change propagation when a cloned fragment is changed. The task of identifying the ripple effects (i.e., all the related components to change) is known as Change Impact Analysis (CIA). In this thesis, we evaluate the impacts of clones on software systems from new perspectives and then we propose an evolutionary coupling based technique for change impact analysis of clones. First, we empirically evaluate the comparative stability of cloned and non-cloned code using fine-grained syntactic change types. Second, we assess the impacts of clones from the perspective of coupling at the domain level. Third, we carry out a comprehensive analysis of the comparative stability of cloned and non-cloned code within a uniform framework. We compare stability metrics with the results from the original experimental settings with respect to the clone detection tools and the subject systems. Fourth, we investigate the relationships between stability and bug-proneness of clones to assess whether and how stability contribute to the bug-proneness of different types of clones. Next, in the fifth study, we analyzed the impacts of co-change coupling on the bug-proneness of different types of clones. After a comprehensive evaluation of the impacts of clones on software systems, we propose an evolutionary coupling based CIA approach to support the consistent evolution of clones. In the sixth study, we propose a solution to minimize the effects of atypical commits (extra large commits) on the accuracy of the detection of evolutionary coupling. We propose a clustering-based technique to split atypical commits into pseudo-commits of related entities. This considerably reduces the number of incorrect couplings introduced by the atypical commits. Finally, in the seventh study, we propose an evolutionary coupling based change impact analysis approach for clones. In addition to handling the atypical commits, we use the history of fine-grained syntactic changes extracted from the software repositories to detect typed evolutionary coupling of clones. Conventional approaches consider only the frequency of co-change of the entities to detect evolutionary coupling. We consider both change frequencies and the fine-grained change types in the detection of evolutionary coupling. Findings from our studies give important insights regarding the impacts of clones and our proposed typed evolutionary coupling based CIA approach has the potential to support the consistent evolution of clones for better clone management.
Code Clones, Change Impact Analysis, Evolutionary Coupling
Doctor of Philosophy (Ph.D.)