Mining Peer Code Review System for Computing Effort and Contribution Metrics for Patch Reviewers

Abstract

Peer code review is a software quality assurance activity followed in several open-source and closed-source software projects. Rietveld and Gerrit are the most popular peer code review systems used by open-source software projects. Despite the popularity and usefulness of these systems, they do not record or maintain the cost and effort information for a submitted patch review activity. Currently there are no formal or standard metrics available to measure effort and contribution of a patch review activity. We hypothesize that the size and complexity of modified files and patches are significant indicators of effort and contribution of patch reviewers in a patch review process. We propose a metric for computing the effort and contribution of a patch reviewer based on modified file size, patch size and program complexity variables. We conduct a survey of developers involved in peer code review activity to test our hypothesis of causal relationship between proposed indicators and effort. We employ the proposed model and conduct an empirical analysis using the proposed metrics on open-source Google Android project.