Tracking the decisions to select repositories for Mining Software Repositories experiments

Abstract

Mining Software Repositories analyzes and cross-links the data available in software repositories. This enables MRS to recognize patterns in software repositories. For example, to study how developers resolve conflicting merges. However, two main problems exist in the selection process of repositories: the limitations presented in traditional approaches used when selecting repositories and the lack of a systematic process for choosing repositories, turning off the experiments' reproducibility. This approach is proposed to resolve identified limitations and assist users in software repositories' selection. Initial results show that this approach returns at least 1.8 times more repositories, overcoming, for instance, the main language restriction in searches.