Empirical Analysis of Pull Requests for Google Summer of Code

TL;DR

This paper empirically analyzes 17,232 pull requests from Google Summer of Code interns, revealing common activities, review feedback types, and educational implications for open-source contribution training.

Contribution

It provides a comprehensive empirical analysis of GSoC pull requests, highlighting the nature of tasks, review feedback, and educational insights, which were previously underexplored.

Findings

Most tasks involve feature addition and bug fixing.

Reviewers focus on code functionality, testing, and readability.

The study offers insights into open-source contribution practices during GSoC.

Abstract

Internship and industry-affiliated capstone projects are popular ways to expose students to real world experiences and bridge the gap between academic training and industry requirements. However, these two approaches often require active industry collaboration, and many students struggle to find industry placements. Open-source contributions are a crucial alternative to gain real world experience, earn publicly verifiable contribution with real-world impact, and learn from experienced open-source contributors. The Google Summer of Code (GSoC) is a global initiative that matches students or new contributors with experienced mentors to work on open-source projects. The program aims to introduce the students to open-source development, help them gain valuable skills under the guidance of mentors, and hopefully encourage them to continue contributing to open-source projects. The realization of the program objectives will provide a continuous pool of talented new contributors necessary for maintaining open-source projects. This study presents an empirical analysis of pull requests created by interns during the GSoC program. We extracted and analyzed 17,232 pull requests from 2,456 interns across 1,937 open-source projects. The results show most tasks involve both code-intensive activities like adding new features and fixing bugs, as well as non-code tasks like updating documentation and restructuring the codebase. Feedback from reviewers covers code functionality and programming logic, testing coverage, error handling, code readability, and adherence to best practices. Finally, we discuss the implications of these results for software engineering education.