Um estudo sobre a correlação entre acoplamentos de software entre ramos e esforços de merge

Abstract

Collaborative software development demands parallel work, and the concurrent changes need to be merged afterwards. When a merge fails, either due to direct or indirect conflicts, the developer must intervene manually. Thus far, existing work in the literature provides some initial evidence that structural coupling is one of the reasons for indirect conflicts. However, the existing work does not formally assess the extent in which structural coupling correlates with merge effort and does not consider other types of couplings, such as logical and conceptual. In this work, we propose metrics to quantify the structural, logical, and conceptual couplings among changes that occur across branches. Then, we investigate the correlation between the software coupling metrics and the merge effort metrics (i.e., rework, wasted work, and extra work) over 12 open-source projects, summing up 6,376 merges. We could observe a weak correlation between all coupling metrics and the extra work effort. However, aligned to the literature, we note that the correlation between the structural coupling and the extra work effort is greater than the other two. In addition, we observed a moderate correlation between structural and conceptual couplings, and rework and wasted work efforts, but a low correlation between logical coupling and rework and wasted work efforts. Finally, we could observe a moderate correlation was observed between logical and conceptual couplings and a weak correlation between structural and logical couplings, as well as between structural and conceptual couplings.