PRCollector: Facilitating On-Demand Collection of Pull Request Data From GitHub

As pull-based software development has become popular, many researchers have analyzed pull request (PR) data in empirical studies. Consequently, they are keen to easily acquire and curate many PRs from software repositories. Although GitHub hosts numerous software projects, researchers have found online acquisition through the GitHub API troublesome. To overcome this problem, several offline datasets mirroring GitHub’s persistent data have been proposed. Unfortunately, these datasets incur high costs to maintain the latest data. Our proposed <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">PRCollector</i>, a PR data collection platform, enables users to write a simple Java program that automatically collects the desired PR data on demand and stores them in local storage. The collected data are organized according to our proposed metamodel, which illustrates the relationships between elements in PR activities. It is designed to bypass further analysis of Markdown descriptions in PR conversations and cumbersome tracking of source code elements corresponding to PR comments. The PR models stored locally reduce the time required to conduct investigations of PR data. Experimental results with 64,361 PRs across 23 repositories demonstrate that <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">PRCollector</i> is a superior choice for researchers seeking more efficient analyses of PRs.