All project work (code or not), documentation, etc. should be done or tracked in a repository within the UBC Launch Pad GitHub organization. This page outlines general advice for how to maintain repositories.
Make sure your entire team is “Watching” your project repository - you can do this from the top right of the GitHub project page. This allows you to receive notifications in the web interface.
If you do not like the GitHub notification emails, you can disable the emails
and instead configure Pull Reminders for yourself
or for your team’s channel. This can be done via the web interface or the Slack
bot in the Launch Pad workspace - just shoot
@Pull Reminders a message.
The primary source of truth for any repository should be the README.
At the top of the README, you should provide a summary of your software, and in particular briefly and succinctly outline what role your project plays or what problem it solves.
Installation instructions should come immediately after this summary. Your installation instructions should be written in as simple and clear a manner as possible and any differences in installation between different operating systems or environments should be made clear. It is also usually helpful to include separate development setup instructions, especially if this involves any extra dependencies or tools.
Optionally, you may want to include sections below this that explain the high-level operation of the tool, list any major dependencies, and/or credit your hard-working developers and designers.
Many continuous integration, code coverage, and documentation sites provide “badges” of project status that you can place at the top of your README. These are highly recommended - they make your README look much more professional.
For a fantastic example of a README page, check out Launch Pad’s Inertia project.
You should enforce a clearly defined set of standards for your code; in particular, this should both be listed in your documentation and enforced during the build process.
Most modern programming languages have linters or other static analysis tools to catch “obvious” mistakes and enforce code style and documentation rules, and running these alongside your tests is highly recommended. Additionally, many languages, including Python and golang, have standardized code styles and tools to enforce them, such as pycodestyle and gofmt.
Adding a pull request template will increase the likelihood that pull request descriptions contain the information you and other maintainers need during the review process.
.gitignore file to reduce the chances that people push unwanted
content; in particular, it is highly recommended to include config files
produced by common IDEs, like IntelliJ, Xcode, Android Studio, and so on. It is
also best to ignore compiled assets, such as Python’s
It is highly recommended that you provide Docker and docker-compose files as a means of ensuring simple setup and standardized development/build environments.
You should have a continuous integration system in place to ensure that new code is tested and meets team standards. Launch Pad recommends Travis CI, which is free for open-source repositories; our GitHub organization is already connected to Travis and a number of project use it currently.
Your CI process should run automated tests, linters, code style enforcers, and anything else that can be automated and help maintain high code quality. Tests in particular help make sure that even when refactors or major changes are made, you can trust that most of your code base still behaves as expected, which means you spend less time chasing down obscure bugs, and allows you more freedom when improving your project’s codebase.
Code coverage of tests in particular is very valuable to collect during your CI builds. It’s particularly recommended to enforce a minimum threshold for code coverage (though this does not guarantee that you have good tests!). Sites that provide code coverage dashboards are a convenient way to summarize this data; we currently recommend Codecov, or alternatively Coveralls.
Ideally documentation about the inner workings of the tool should lie as close to the code it describes as possible.