Contributing

Contributions are welcome, and they are greatly appreciated. Every little bit helps, and credit will always be given.

Types of Contributions

  • Report Bugs: Refer to the Issue Tracker. Please include details such as operating system, Python version, and ribs version, as well as detailed steps to reproduce the bug.

  • Fix Bugs: Look through the Issue Tracker for bugs. Anything tagged with “bug” and “help wanted” is open to whoever wants to implement it.

  • Propose features: To request new features in pyribs, submit a Feature Request on the Issue Tracker. In the request, please:

    • Explain in detail how the feature would work.

    • Keep the scope as narrow as possible, to make the features easier to implement.

  • Implement Features: Look through the Issue Tracker for features. Anything tagged with “enhancement” and “help wanted” is open to whoever wants to implement it.

  • Write Documentation: pyribs could always use more documentation, whether as part of the official pyribs docs, in docstrings, or even on the web in blog posts, articles, and such. For the website, refer to the website repo.

  • Submit Feedback: The best way to send feedback is to file an issue on the Issue Tracker.

Developing pyribs

Ready to contribute? Here’s how to set up pyribs for local development.

  1. Fork the pyribs repo on GitHub.

  2. Clone the fork locally:

    # With SSH:
    git clone git@github.com:USERNAME/pyribs.git
    
    # Without SSH:
    git clone https://github.com/USERNAME/pyribs.git
    
  3. Install the local copy and dev requirements into an environment. For instance, with Conda, the following creates an environment at ./env.

    cd pyribs
    conda create --prefix ./env python=3.7 # 3.7 is the minimum version pyribs supports.
    conda activate ./env
    pip install -e .[dev]
    
  4. Create a branch for local development:

    git checkout -b name-of-bugfix-or-feature
    

    Now make the appropriate changes locally.

    • Please follow the Google Style Guide (particularly when writing docstrings).

    • Make sure to auto-format the code using YAPF. We highly recommend installing an editor plugin that auto-formats on save, but YAPF also runs on the command line:

      yapf -i FILES
      
  5. After making changes, check that the changes pass the tests:

    pytest tests/
    make test # ^ same as above
    

    And to run the benchmarks:

    pytest -c pytest_benchmark.ini
    make benchmark # ^ same as above
    

    Finally, to lint the code:

    pylint ribs tests benchmarks examples
    make lint # ^ same as above
    

    To get pytest and pylint, pip install them into the environment. However, they should already install with pip install -e .[dev].

  6. Add your change to the changelog for the current version in HISTORY.md.

  7. Commit the changes and push the branch to GitHub:

    git add .
    git commit -m "Detailed description of changes."
    git push origin name-of-bugfix-or-feature
    
  8. Submit a pull request through the GitHub website.

Pull Request Guidelines

Before submitting a pull request, check that it meets these guidelines:

  1. Style: Code should follow the Google Style Guide and be auto-formatted with YAPF.

  2. The pull request should include tests.

  3. If the pull request adds functionality, corresponding docstrings and other documentation should be updated.

  4. The pull request should work for Python 3.7 and higher. GitHub Actions will display test results at the bottom of the pull request page. Check there for test results.

Instructions

Running a Subset of Tests

To run a subset of tests, use pytest with the directory name, such as:

pytest tests/core/archives

Documentation

Documentation is primarily written in Markdown, as we use the MyST Sphinx plugin.

To preview documentation, use:

make servedocs

This will open up a browser window and automatically reload as changes are made to the docs.

Adding a Tutorial

Tutorials are created in Jupyter notebooks that are stored under tutorials/ in the repo. To create a tutorial:

  1. Write the notebook and save it under tutorials/.

  2. Use cell magic (e.g. %pip install ribs) to install pyribs and other dependencies.

    • Installation cells tend to produce a lot of output. Make sure to clear this output in Jupyter lab so that it does not clutter the documentation.

  3. Before the main loop of the QD algorithm, include a line like total_itrs = 500 (any other integer will work). This line will be replaced during testing (see tests/tutorials.sh) in order to test that the notebook runs end-to-end. By default, the tests run the notebook with total_itrs = 5. If this tutorial needs more (or less), modify the switch-case statement in tests/tutorials.sh.

  4. Make sure that the only level 1 heading (e.g. # Awesome Tutorial) is the title at the top of the notebook. Subsequent titles should be level 2 (e.g. ## Level 2 Heading) or higher.

  5. If linking to the pyribs documentation, make sure to link to pages in the latest version on ReadTheDocs, i.e. your links should start with https://docs.pyribs.org/en/latest/

  6. Add an entry into the toctree in docs/tutorials.md and add it to one of the lists of tutorials.

  7. Check that the tutorial shows up on the Tutorials page when serving the docs.

  8. Create a PR into the website repo that adds the tutorial onto the home page, specifically this file. In the PR, include a square image that represents the tutorial.

Adding an Example

Examples are created in Python scripts stored under examples/ in the repo, and their source is shown in the docs. To create an example:

  1. Write the Python script and save it under examples/.

  2. Add any dependencies at the top of the script with a pip install command (see existing examples for a sample of how to do this).

  3. Add a shell command to tests/examples.sh that calls the script with parameters that will make it run as quickly as possible. This helps us ensure that the script has basic correctness. Also call the install_deps function on the script file before running the script.

  4. Add a Markdown file in the docs/examples directory with the same name as the Python file – if the example is examples/foobar.py, the Markdown file will be docs/examples/foobar.md.

  5. Add a title to the Markdown file, such as:

    # My Awesome Example
    
  6. In the markdown file, include the following so that the source code of the example is displayed.

    ```{eval-rst}
    .. literalinclude:: ../../examples/EXAMPLE.py
        :language: python
        :linenos:
    ```
    
  7. Add any other relevant info to the Markdown file.

  8. Add an entry into the toctree and list of examples in docs/examples.md.

  9. Check that the example shows up on the Examples page when serving the docs.

Referencing Papers

When referencing papers, refer to them as Lastname YEAR, e.g. Smith 2004. Also, prefer to link to the paper’s website, rather than just the PDF. This is particularly relevant when linking to arXiv papers.

Deploying

  1. Create a PR into master after doing the following:

    1. Switch tutorial links from latest to stable with:

      make tutorial_links
      

      See #300 for why we do this.

    2. Update the version with bump2version by running the following for minor versions:

      bump2version minor
      

      or the following for patch versions:

      bump2version patch
      
    3. Add all necessary info on the version to HISTORY.md.

  2. (Optional) Once the PR has passed CI/CD and been squashed-and-merged into master, check out the squash commit and locally run make release-test. This uploads the code to TestPyPI to check that the deployment works. If this fails, make fixes as appropriate.

  3. Once the PR in step 1 and any changes in step 2 have passed CI/CD and been squashed-and-merged into master, locally tag the master branch with a tag like v0.2.1, e.g.

    git tag v0.2.1 HEAD
    
  4. Now push the tag with

    git push --tags
    
  5. Check that the version was deployed to PyPI. If it failed, delete the tag, make appropriate fixes, and repeat steps 2 and 3.

  6. Write up the release on GitHub, and attach it to the tag.

  7. Submit another PR which reverts the changes to the tutorial links. Specifically, while on master, make sure your workspace is clean, then revert the changes with:

    git checkout HEAD~ tutorials/
    

    And commit the result.

Our deployment process may change in the future as pyribs becomes more complex.