Granularity-based Comparison of the Change-proneness of Micro-Clones

Regular clones (i.e., identical or substantially equivalent code fragments greater than 4 LOC) inside a software system’s code-base have been thoroughly examined by several studies. In our research, we investigate the change-proneness of micro-clones (i.e., code clones of at most 4 LOC). In particular, we compare the change-proneness of function and block level micro-clones. An empirical study shows that micro-clones are a growing concern since the number of micro-clones is three times as much as the number of regular clones in a software system. Analyzing the impacts of function and block level micro-clones is very useful for developers because they can realize which granularity (function or block) is more important for software maintenance and evolution. When it comes to managing clones, developers should pay more attention to code clones that are subject to frequent modifications, sometimes called unstable clones, from an evolutionary perspective. They can prioritize the more change-prone micro-clones for refactoring or tracking. We investigated the micro-clone category of function level and block level for 5 diverse subject systems from the SVN repository and analyzed their change-proneness by using UNIX diff. We found that micro-clones of function granularity are generally more change-prone than micro-clones of block granularity. Thus, function level micro-clones should be prioritized more for management, such as refactoring and tracking.