We want to make contributing to this project as easy and transparent as possible.
We actively welcome your pull requests.
- Fork the repo and create your branch from
main
. - If you've added code that should be tested, add tests.
- If you've changed APIs, update the documentation.
- Ensure the test suite passes.
- Make sure your code lints.
- If you haven't already, complete the Contributor License Agreement ("CLA").
In order to accept your pull request, we need you to submit a CLA. You only need to do this once to work on any of Meta's open source projects.
Complete your CLA here: https://code.facebook.com/cla
We use GitHub issues to track public bugs. Please ensure your description is clear and has sufficient instructions to be able to reproduce the issue.
Meta has a bounty program for the safe disclosure of security bugs. In those cases, please go through the process outlined on that page and do not file a public issue.
- 4 spaces for indentation rather than tabs
- 80 character line length
Run the linter via flake8
(pip install flake8
) from the root of the Ax repository. Note that we have a custom flake8 configuration.
We use Pyre for static type checking and require code to be fully type annotated.
We strongly recommend adding unit testing when introducing new code. To run all unit tests, we recommend installing pytest using pip install pytest
and running pytest -ra
from the root of the balance repo.
- We require docstrings on all public functions and classes (those not prepended with
_
). - We use the Google docstring style & use Sphinx to compile API reference documentation.
- Our website leverages Docusaurus 2.0 + Sphinx + Jupyter notebook for generating our documentation content.
- To rule out parsing errors, we suggesting installing sphinx and running
make html
from the balance/sphinx folder.
balance's website is also open source and part of this repository. balance leverages several open source frameworks for website development.
- Docusaurus 2: The main site is generated using Docusaurus, with the code living under the website folder). This includes the website template (navbar, footer, sidebars), landing page, and main sections (Blog, Docs, Tutorials, API Reference).
- Markdown is used for the content of several sections, particularly the "Docs" section. Files are under the docs/ folder
- Jupyter notebook is used to generate the notebook tutorials under the "Tutorials" section, based on our ipynb tutorials in our tutorials folder.
- Sphinx is used for Python documentation generation, populated under the "API Reference" section. Files are under the sphinx folder.
To install the necessary dependencies for website development, run the following from the repo root:
python -m pip install git+https://github.com/bbalasub1/[email protected]
python -m pip install .[dev]
All our notebook tutorials are housed under the tutorials folder at the root of the repo. We use these notebooks as the source of truth for the "Tutorials" section of the website, executing & generating HTML pages for each notebook.
To add a new tutorial:
- Check in your notebook (.ipynb) to our tutorials folder. We strongly suggest clearing notebook output cells.
- Extend the "Building tutorial HTML" section of
scripts/make_docs.sh
to execute & generate HTML for the new tutorial e.g.jupyter nbconvert tutorials/my_tutorial.ipynb --execute --to html --output-dir website/static/html/tutorials
. - Introduce a new .mdx page under the website/docs/tutorials folder for the new tutorial. Use HTMLLoader to load the generated HTML e.g.
<HTMLLoader docFile={useBaseUrl('html/tutorials/my_tutorial.html')}/>
. quickstart.mdx is a good reference for the setup
To test the setup, see the Building & Testing Website Changes
section below.
Note: The generated HTML should not be checked into the main repo.
We've developed a helper script for running the full website build process:
./scripts/make_docs.sh
# To start up the local webserver
cd website
yarn serve
Once the local webserver is up, you'll get a link you can follow to visit the newly-built site. See Docusaurus docs for more info.
We rely on Github Actions to run our CI/CD. The workflow files can be found here.
In summary
- On every pull request, we run our "Build & Test" workflow, which includes PyTest tests, Wheels package builds, flake8 linting, and website build.
- We also run the same "Build & Test" suite nightly.
- On every push, we deploy a new version of the website. The
make_docs.sh
script is run from the main branch and the build artifacts are published to thegh-pages
branch, which is linked to our repo's Github Page's deployment.
To create a new release, simply navigate to the "Release" page of the repo, draft a new release, and publish. The Github Action workflow should be triggered on publish and you should see a new version of the package live on PyPi in ~10 mins. You can check the status of the job via the GH Actions tab.
Guidelines when drafting a new release:
- Follow semantic versioning conventions when chosing the next version.
- The release's tag should only be the version itself (e.g. "0.1.0"). Do not add any prefixes like "v" or "version". The build process relies on proper formatting of this tag.
The Github Actions job is configured at release.yml.
By contributing to balance, you agree that your contributions will be licensed under the LICENSE file in the root directory of this source tree.