Is code co-committal an indicator of evolutionary coupling in software repositories?

Software repositories such as Git are significant sources of metadata about software projects, containing information such as the files changed, change authors, and often commentary on the change. An emerging approach to augment software change impact analysis is to exploit this metadata to determine which files are linked by co-committal, i.e. when updates are frequently made to both files in the same Git commit. This information as useful as, for example, an indicator of change impact sets for future changes. The aims of this study are to determine whether co-committal is an indicator of links between software artifacts stored in Git, and if so, to determine whether these links continue as said software artifacts evolve and thus could provide a valuable dimension for use in change impact analysis. To empirically investigate the objective of the study as discussed above, we mined the metadata of five large Git code repositories, with over 14k commits and extracted co-change sets from the data. The results have show that co-committal links between artifacts vary in the strength and frequency of the link, in a way which was highly dependent on the development style and level of activity of the involved software developer(s). Results also indicate that while this approach may be applicable in some scenarios, it cannot be applied generally, and is only effective on projects with specific patterns of developer behavior.