Committing to Writing Good Commit Messages: Supporting the Creation of Human- and Machine-Readable Commit Messages with Python
Have you ever read the commit messages that you previously wrote for a Python
project that you maintain on GitHub? If you are like me, then you might find
that some of these commit messages are confusing, incorrect, or downright
hilarious! Since many software developers find it difficult to write a succinct,
yes descriptive commit message, there are a number of standards that constrain
the format a Git commit message should take. Many of these standards, such as
the one called Conventional Commits or the one adopted by the Angular project,
require a programmer to attach a label, like fix or feat, to designate that
a specific commit, for instance, fixes a defect or adds a new feature.
Since it is often difficult for programmers to confidently pick the right label for a commit, this presentation introduces a tool that we built, called CommitCanvas, that uses machine learning to automatically predict the most suitable label for a commit message. After overviewing the techniques that CommitCanvas uses to predict the best label for a commit message, this talk will show people how to integrate it into their workflow as either a pre-commit hook or as a command-line application. The presentation will also share insights into the effectiveness of CommitCanvas, thereby helping developers to understand the situations in which it works best. People who watch this talk will gain the knowledge that they need to start using CommitCanvas to automatically label their own commit messages.
- Introduction to the challenges associated with writing good Git commit messages
- How Conventional Commits and the Angular standard partially address the challenge of writing commit messages
- Explanation of how machine learning can automatically predict the label for a Git commit message
- How to integrate the CommitCanvas tool into real-world workflows for developing Python projects
- Insights into when the machine learning methods in CommitCanvas will and will not work well
Gregory M. Kapfhammer is an Associate Professor of Computer Science at Allegheny College. Along with conducting research in the field of software engineering, he teaches classes in a variety of technical and scientific areas and enjoys programming in Python. In the Python community, Gregory has previously given talks at PyGotham and PyOhio, presented posters at PyCon, and presented at PyCon's Education Summit. You can learn more about Gregory by visiting his web site at https://www.gregorykapfhammer.com/, checking out several of his Python projects at https://github.com/gkapfham, and viewing the slides of his previous presentations at https://speakerdeck.com/gkapfham.