Let's Supercharge the Workflows: An Empirical Study of GitHub Actions
Tingting Chen, Yang Zhang, Shu Chen, Tao Wang, Yiwen Wu
Abstract
Automation has become a norm in software development practices, especially in CI/CD practices. Recently, GitHub introduced GitHub Actions (GA) to provide automated workflows for software maintainers. However, few researches have been proposed to evaluate its capability and impact, even though several GA have been built by practitioners. In this paper, we conduct a large-scale empirical study of GitHub projects, to help practitioners gain deep insights into the GA. We quantitatively investigate the basic adoption of GA and its potential correlation with project properties. We also analyze the usage details of GA, including its component scale and action sequences. Finally, using regression modeling, we investigate the impact of GA on the commit frequency, pull request and issue's resolution efficiency. Our findings suggest a nuanced picture of how practitioners are adapting to, and benefiting from GA.