Python Poetry Rules for Bazel

These rules are designed to allow you to easily use the Poetry Package Manager with Bazel. It does this whilst still allowing you to use Poetry as usual with poetry add and poetry run.

Getting started

To illustrate how to use this package, there's an example project included which shows all the relevant wiring.

Poetry Setup

In order to smoothen out the interactions between Bazel and Poetry, we use the common Python location of .venv for the Virtual Environment. This makes it easier for both tools to find it, this is configured using the virtualenvs.in-project configuration with Poetry:

poetry config virtualenvs.in-project true --local

Which results in the poetry.toml file found in our example project. You can then use the normal Poetry commands.

Bazel Setup

To enable Bazel to manage the Poetry Virtual Environment, we use the managed_directories property in our example WORKSPACE; this lets Bazel recreate the environment within our workspace, and symlinks it into the Bazel environment:

workspace(
    name = "basic_project",
    managed_directories = {
        "@poetry_environment": [".venv"]
    }
)

Afterwards, we use the http_archive package to download the rules:

http_archive(
    name = "rules_python_poetry",
    url = "https://github.com/DaMouse404/rules_python_poetry/releases/download/0.0.1/rules_python_poetry-0.0.1.tar.gz",
    sha256 = "3ac54f1e9b070d2ed727c58f30798f4cea1123242279e7d6e1a48e1f06ca16d6",
)

Then we run the poetry_enviromment rule from environment.bzl, which generates the Virtual Environment if necessary, alongside symlinking the various Poetry configuration files into the Bazel environment:

load("@rules_python_poetry//:environment.bzl", "poetry_environment")
poetry_environment(
    name="poetry_environment",
    project="//:pyproject.toml",
    lock="//:poetry.lock",
    config="//:poetry.toml"
)

Then we can register the Poetry Python intrepreter as a toolchain for PY3 in Bazel:

register_toolchains("@poetry_environment//:poetry_toolchain")

And finally we use poetry export to export requirements to requirements.txt format:

load("@poetry_environment//:runtime.bzl", "interpreter_path")
load("@poetry_environment//:export.bzl", "poetry_export")
poetry_export(
    name="poetry_requirements"
)

Which can then be used to interact with the standard rules_python (take note of the python_interpreter passed to pip_import here to use the virtual env one):

http_archive(
    name = "rules_python",
    url = "https://github.com/bazelbuild/rules_python/releases/download/0.0.2/rules_python-0.0.2.tar.gz",
    strip_prefix = "rules_python-0.0.2",
    sha256 = "b5668cde8bb6e3515057ef465a35ad712214962f0b3a314e551204266c7be90c",
)
load("@rules_python//python:repositories.bzl", "py_repositories")
py_repositories()
load("@rules_python//python:pip.bzl", "pip_repositories", "pip_import")
pip_repositories()

pip_import(
    name = "basic_project_pip",
    requirements = "@poetry_requirements//:requirements.txt",
    python_interpreter = interpreter_path
)
load("@basic_project_pip//:requirements.bzl", "pip_install")
pip_install()