RevMate: Towards an optimized review process in AAA game development

In software development, code reviewing is an essential code validation step during integration [1, 2]. When developers submit their code for review, human reviewers need to validate the changes made to source code
and other artifacts (e.g., data files, artifacts created by artists). This review process can go through several iterations in which the reviewers comment on the submission and the submission’s developer revise the
submission by submitting a new version (“revision”) addressing the suggested changes. Eventually, the (revised) submission either is accepted and forwarded to integration in the main development branch, or
rejected.
In this research project, we study different automation approaches to improve the velocity of the code review process at Ubisoft, our industrial partner. More specifically, we intend to build techniques for code chunk
ordering, hot-spot detection and review comment prediction. We will also establish a pre-review protocol where the developer validates and/or modifies our approaches’ results, as to have actual developers’ input and insights
in the review process. We aim to conduct this study on several AAA game productions, both ongoing and past games, to empirically assess the generalization of our results.

Faculty Supervisor:

Bram Adams

Student:

Partner:

Ubisoft Toronto

Discipline:

Computer science

Sector:

Information and cultural industries

University:

Queen's University

Program: