Mining and Assessing Issue Resolution Processes in Open Source Software Repositories

Resolving issues is a key activity in open-source software (OSS) development, but the ad hoc nature of issue handling on platforms like GitHub makes it challenging to understand the underlying processes of issue resolution, how well such processes are organized, and how to effectively improve issue resolution efficiency. In this work, we quantitatively examine the relationship between issues' organization levels, modeled as process uncertainties, and resolution efficiency measured by issue lifetime and event transition time. We propose an information-theoretical approach to assess the uncertainty lies in issue processing by calculating the entropy of Direct Follow Graphs (DFGs) that represent issue processes. Instead of performing analysis on a project's all issues as a whole, which may lead to over complex models, we find concise yet representative DFG-based models for a project's issues with an entropy-guided KMeans++ clustering algorithm. Findings from real-world OSS projects' issues suggest that higher process uncertainty of issues is associated with longer issue lifetime and extended event transition time. The result also indicates that enhancing issue process organization can potentially improve issue resolution efficiency.