Using GitHub Actions to run tests for Python packages


GitHub recently launched GitHub Actions, a way to automate software workflows and to run continuous integration or continuous delivery with a deep integration with the GitHub platform. It’s currently in beta and the general availability is planned for November 2019. Like CircleCI, jobs are free for public repositories, a chance for open source projects. Workflows are expressed in YAML.

GitHub develops some actions you can reuse and you can build yours. GitHub provides suggestions for common workflow needs: running tests on a Node package, pushing a Docker image to Docker hub when creating a tag etc. The Actions Marketplace has an interesting list of actions to help you get started in various tasks: linting, security, publishing, building, notifications, code reviews etc.

I decided to give it a spin with a Python package. My goal was to run unit tests on various Python versions. GitHub Actions has the concept of build matrix, something coming from Travis CI, which allows you to run a job in different environments (OS, Python version, architecture etc.). It makes it a breeze to test your code in various environments, something you could not easily do locally. You can find the YAML code I wrote to install dependencies and run tests on various Python versions.

You can head to the GitHub Actions documentation for a complete tour of the available features.

Sounds great? Give me a follow on Twitter or learn more about me.