Contributing code

Local setup

To contribute code, basic knowledge of git is required, if you are unfamiliar with the git workflow, checkout GitHub’s tutorial. The following steps allow you to contribute code to DoWhy.

  1. Fork the dowhy main repository

  2. Clone this repository to your local machine using

    git clone https://github.com/<YOUR_GITHUB_USERNAME>/dowhy
    
  3. Install DoWhy and its requirements. Poetry will create a virtual environment automatically, but if preferred, it can be created in a different way. By default, Poetry will install DoWhy in interactive mode. This way, you can immediately test your changes to the codebase.

    cd dowhy
    pip install --upgrade pip
    poetry install -E "plotting"
    

    Note

    Installing pygraphviz can cause problems on some platforms. One way that works for most Linux distributions is to first install graphviz and then pygraphviz as shown below. Otherwise, please consult the documentation of pygraphviz.

    sudo apt install graphviz libgraphviz-dev graphviz-dev pkg-config
    pip install pygraphviz --install-option="--include-path=/usr/include/graphviz" \
    --install-option="--library-path=/usr/lib/graphviz/"
    
  4. (optional) add dowhy as an upstream remote to keep your fork up-to-date with DoWhy’s main branch.

    git remote add upstream http://www.github.com/py-why/dowhy
    

    You are now ready to make changes to the code base locally.

Pull request checklist

  1. Execute the flake8 linter for breaking warnings and fix all reported problems.

    poetry run poe lint
    
  2. Make sure the newly added code complies with the format requirements of black and isort.

    poetry run poe format_check
    

    You can use following commands to fix formatting automatically

    poetry run poe format
    
  3. Add tests for your new code and execute the unittests to make sure you did not introduce any breaking changes or bugs.

    poetry run poe test
    

    Note that you can also execute those tasks together

    poetry run poe verify
    

    A full list of available tasks can be obtained using

    poetry run poe -h
    

    The full test suite of DoWhy takes quite long. To speed up development cycles, you can restrict the tests executed as in the following example.

    poetry run pytest -v tests/causal_refuters
    
  4. Once your code is finished and it passes all checks successfully, commit your changes. Make sure to add an informative commit message and to sign off your commits:

    git commit --signoff -m "informative commit message"
    

    By including this sign-off step, a commit is enriched with a Developer Certificate of Origin (DCO), containing the author’s name and email address. The DCO is a lightweight alternative to a CLA and affirms that the author is the source of the committed code and has the right to contribute it to the project. For the full text, see DCO.

    Note

    Note the “–signoff” or shorthand “-s” is obligatory and commits without cannot be merged. By default, most IDEs won’t include this step within their git integration, so an additional setup may be required.

    In case you made a single commit without adding the required DCO, you can do

    git commit --amend --no-edit --signoff
    git push -f origin <BRANCH_NAME>
    

    In case of more commits, one way is to squash them together (example for 3 commits)

    git reset --soft HEAD~3
    git commit -s -m "new informative commit message of squashed commit"
    

    or use a rebase with as many “^” as commits to be changed.

    git rebase --signoff HEAD^^^
    
  5. (advanced) Poetry fixes its dependecies and their version with a poetry.lock file. Poetry’s dependencies should be updated regularly by maintainers via

    poetry update
    

    For most PRs, this is unnecessary. If a PR necessitates a lockfile change, we request that you provide a justification as to why a dependency update was necessary.