Abstract Syntax Tree Merging in Git
Context
In classical version control systems (e.g., Git), software development can be separated using different branches. However, at some point, the source code developed inside a given branch needs to be reconciled with the one pushed by the other software developers. In a perfect world, this integration is easy, as the system will automatically merge together the source code seamlessly.
Unfortunately, the world is far from being perfect, and software developers fear the so-called merge conflict. The fact that Git works at the text level is one of the root causes of the total mess such a conflict can create in a project: the tool has no idea of the structure of the pieces of software to be merged! For example, imagine two developers, Bob and Alice. If Alice changes the contents if a method m in her branch, and Bob moves the method m to another part of the program, this is considered as a conflict in Git. But it should not be, as considering the semantics of their respective intention, we can automatically reconcile their editions, and yield an automated merged version of the source code!
In this project, we propose to leverage the Git merge algorithm to provide a tool able to merge source code at the abstract tree level. The key point is to identify how to reconcile classical conflicts (using real-world examples extracted from large scale Java projects) in a better way than the text-based algorithm.
Join the team, make git-merge great again!
Skills and Background
- loving to craft software;
- knowing how to use Git, and the meaning of the MVP acronym;
- fluency in object-oriented programming (e.g., Java);
- good knowledge in object-oriented software design;
- knowing software development good practices (i.e, Clean Code by Uncle Bob is on your night table).
Required role of the student
- contribution to a research project;
- contribution to the development of prototypes;
- discovering state-of-the-art literature;
- communicating about the work done in the lab (technical report, article).
References
- Fine-grained and accurate source code differencing. Jean-Rémy Falleri, Floréal Morandat, Xavier Blanc, Matias Martinez and Martin Monperrus. ASE 2014.
- On the Nature of Merge Conflicts: a Study of 2,731 Open Source Java Projects Hosted by GitHub. Gleiph Ghiotto, Leonardo Murta, Marcio Barros and Andre van der Hoek. TSE 2018, ICSE 2019 (Journal-First)
Supervision
Sébastien Mosser
Professor, UQAM