An opinionated template repo for Python projects using Bazel, built around the principle that wherever possible projects should be hermetic and reproducible. It includes:
- Automated linting for Bazel & Python sources with Buildifier, Flake8, Black, and Isort
- Python test helpers to simplify unit test targets
- Hermetic Python toolchain
- Hermetic pip dependencies sourced from requirements.txt
- Python Docker image support, including a matching runtime version as the hermetic toolchain.
It does not seek to provide a continual submodule that can pick up upstream changes from the
template, but is more like a create-react-app template where you can eject from the template
and proceed with your project. That being said, if you don't change any paths you can likely
git cherry-pick or otherwise patch in changes from upstream and then run eject.sh again
to update paths for your project.
- Install Bazel. Instructions for doing so can be found here: https://docs.bazel.build/versions/master/install.html
- Run the template eject script, providing your project's name.
./eject.sh <project_name>
Note to Linux users:
- Python pip requires libssl-dev to be installed (sudo apt-get install libssl-dev).
py_binary()uses the host's Python as a bootstrap for execing the hermetic Python under Bazel. If you only have python3 installed you will need to symlink python to python3 (i.e. ln -s /usr/bin/python3 /usr/bin/python)
Pull requests are welcome! Open one against this repo and it will be reviewed & merged.
In general, lint by running the lint.sh script.
The first time you run lint, you'll see a warning like:
UserWarning: `known_${project}` setting is defined, but ${project} is not included in `sections`
config option: ('FUTURE', 'STDLIB', 'THIRDPARTY', '${project}', 'FIRSTPARTY', 'LOCALFOLDER').
To resolve this, open pyproject.toml in the repo root, and update the <project> reference in
SECTIONS on line 11 to be capitalized. This is a known issue with the basic templating script
used in the repo, and PRs are welcome if you have a resolution :)
You can manually run ${Buildifier/Flake8/Black/Isort} with the following if you want to, though you'll have to handle paths manually to prevent them from working on the Bazel sandbox.
bazel run //tools/${tool}
Unit tests are run using the bazel test command. A test wrapper is provided that invokes pytest
with reasonable defaults and configures the entrypoint. You just write pytest test cases in normal
Python files, and pass these as srcs to a ${project}_py_test rule. The test runner handles the rest.
Currently this Bazel workspace supports building/running on macOS and Linux. Windows support is explicitly unscoped. The currently-supported languages are:
- Python
- Golang
- Golang is limited at the moment and mostly exists just for hermetic Buildifier. Additional work is likely required if you want to use arbitrary Go dependencies with Gazelle.
You can run an interactive python session via either an ipython shell or a Jupyter notebook. By
default, both include the //${project} bazel target so you can make use of ${project} code
interactively.
- iPython -
bazel run //tools/ipython. You'll be dropped into a python shell in the Bazel sandbox - Jupyter -
bazel run //tools/jupyter. You'll be dropped into a Jupyter web interface. You can create a new Notebook to experiment. In general, consider these environments to be ephemeral, but if needed the notebook is created insidebazel-bin/tools/jupyter/jupyter.runfiles/__main__. Alternatively, consider downloading the notebook from the web UI if you will continue to need it.
By default this includes the //${project} target, so you can make use of ${project} code interactively