An Empirical Analysis of Code Clones in GitHub Actions Workflows

Collaborative software development relies on social coding platforms that integrate CI/CD tools to automate repetitive development tasks. This article focuses on GitHub and its CI/CD service GitHub Actions, that supports and promotes reusable Actions and reusable workflows to avoid duplication of workflow configuration code. Despite these reuse mechanisms, copy-paste practices persist, leading to a high amount of Type I and Type II clones in workflow files. We conduct the first large-scale quantitative clone analysis across 118K GitHub Actions workflows in 34K GitHub repositories, containing 364K instances of clones. We study the prevalence of such clones in workflows, identify the specific workflow entities that are more prone to cloning, the distribution of clones at different levels of granularity, and the extent to which clones could be mitigated by the reuse mechanisms offered by GitHub. Our findings reveal that two-thirds of workflows contain clones, the majority of which could be refactored using built-in reuse mechanisms.