An Empirical Study on GitHub Pull Requests’ Reactions

The pull request mechanism is commonly used to propose source code modifications and get feedback from the community before merging them into a software repository. On GitHub, practitioners can provide on by either commenting or simply reacting it using set of pre-defined GitHub reactions, i.e., “Thumbs-up”, “Laugh”, “Hooray”, “Heart”, “Rocket”, “Thumbs-down”, “Confused” “Eyes”. While large number prior studies investigated how improve different engineering activities (e.g., review integration) investigating requests, they focused only requests’ comments as feedback. However, according our preliminary study, contain that not manifested within requests. In fact, analysis six popular projects shows median 100% who reacted did leave any comment suggesting reactions be unique further integration process. To help future better leverage mechanism, we conduct an empirical study understand usage their promises limitations. We investigate in this paper are used, when use what types requests for purposes. Our considers quantitative 380k 63k open-source three qualitative analyses total 989 same projects. find most common positive ones (i.e., “Rocket” “Laugh”). observe reactors express attitude approval, appreciation, excitement) proposed changes A just 1.95% negative ones, which disagree with reasons, such feature might have more downsides than upsides wrong approach address certain problems. Most (a 78.40%) come closing corresponding Interestingly, non-contributors outsiders potentially “end-users” software) also active top that, core contributors, peripheral casual contributors behaviors For instance, react early stages request, while around time or, some cases, after merged. Contributors tend request’s code, concerned about impact end-user experience. findings shed light patterns taxonomies intention reactors, inspire reactions.