Skip to content

Latest commit

 

History

History
125 lines (91 loc) · 5.31 KB

CONTRIBUTING.md

File metadata and controls

125 lines (91 loc) · 5.31 KB
Raw

Contributing to CausalML

The CausalML project welcome community contributors. To contribute to it, please follow guidelines here.

The codebase is hosted on Github at https://github.com/uber/causalml.

We use black as a formatter to keep the coding style and format across all Python files consistent and compliant with PEP8. We recommend that you add black to your IDE as a formatter (see the instruction) or run black on the command line before submitting a PR as follows:

# move to the top directory of the causalml repository
$ cd causalml 
$ pip install -U black
$ black .

As a start, please check out outstanding issues. If you'd like to contribute to something else, open a new issue for discussion first.

Development Workflow 💻

  1. Fork the causalml repo. This will create your own copy of the causalml repo. For more details about forks, please check this guide at GitHub.
  2. Clone the forked repo locally
  3. Create a branch for the change:
$ git checkout -b branch_name
  1. Make a change
  2. Test your change as described below in the Test section
  3. Commit the change to your local branch
$ git add file1_changed file2_changed
$ git commit -m "Issue number: message to describe the change."
  1. Push your local branch to remote
$ git push origin branch_name
  1. Go to GitHub and create PR from your branch in your forked repo to the original causalml repo. An instruction to create a PR from a fork is available here

Documentation 📚

CausalML documentation is generated with Sphinx and hosted on Read the Docs.

Docstrings

All public classes and functions should have docstrings to specify their inputs, outputs, behaviors and/or examples. For docstring conventions in Python, please refer to PEP257.

CausalML supports the NumPy and Google style docstrings in addition to Python's original docstring with sphinx.ext.napoleon. Google style docstrings are recommended for simplicity. You can find examples of Google style docstrings here

Generating Documentation Locally

You can generate documentation in HTML locally as follows:

$ cd docs/
$ pip install -r requirements.txt
$ make html

Documentation will be available in docs/_build/html/index.html.

Test 🔧

If you added a new inference method, add test code to the tests/ folder.

Prerequisites

CausalML uses pytest for tests. Install pytest and pytest-cov, and the package dependencies:

$ pip install .[test]

See details for test dependencies in pyproject.toml

Building Cython

In order to run tests, you need to build the Cython modules

$ python setup.py build_ext --inplace

This is important because during testing causalml modules are imported from the source code.

Testing

Before submitting a PR, make sure the change to pass all tests and test coverage to be at least 70%.

$ pytest -vs tests/ --cov causalml/

To run tests that require tensorflow (i.e. DragonNet), make sure tensorflow is installed and include the --runtf option with the pytest command. For example:

$ pytest --runtf -vs tests/test_dragonnet.py

You can also run tests via make:

$ make test

Submission 🎉

In your PR, please include:

  • Changes made
  • Links to related issues/PRs
  • Tests
  • Dependencies
  • References

Please add the core Causal ML contributors as reviewers.

Maintain in conda-forge 🐍

We are supporting to install the package through conda, in order to maintain the packages in conda we need to keep the package's version in conda's recipe repository here in sync with CausalML. You can follow the instruction from conda or below steps:

  1. After a new release of the package, fork the repo.
  2. Create a new branch from the master branch.
  3. Edit the recipe:
    • Update the version number here in meta.yaml
    • Generate the new sha256 hash and update it here: the sha256 hash can get from PyPi; look for the SHA256 link next to the download link on PyPi package’s files page, e.g. https://pypi.org/project/causalml/#files
    • Reset the build number to 0
    • Update the dependencies if needed
  4. Submit the PR and the recipe will automatically be built;

Once the recipe is ready it will be merged. The recipe will then automatically be built and uploaded to the conda-forge channel.