Automatic Decomposition of Java Open Source Pull Requests: A Replication Study

The presence of large changesets containing several independent modifications (e.g. bug fixes, refactorings, features) can negatively affect the efficacy of code review. To cope with this problem, Barnett et al. developed ClusterChanges — a lightweight static analysis technique for decomposing changesets in different partitions that can be reviewed independently. They have found that ClusterChanges can indeed decompose such changesets and that developers agree with their decomposition. However, the authors’ restricted their analysis to software that is: (i) closed source, (ii) written in C# and (iii) developed by a single organization. To address this threat to validity, we implemented JClusterChanges, a free and open source (FOSS) implementation of ClusterChanges for Java software, and replicated the original Barnett et al. study using changesets from Java open source projects hosted on GitHub. We found that open source changesets are similar to the changesets of the original study. Thus, our research confirms that the problem is relevant to other environments and provides a FOSS implementation of ClusterChanges that not only shows the feasibility of the technique in other contexts but can also be used to help future research.