tubular / rules_pygen

Rules for generating native Bazel Python libraries from requirements.txt

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

rules_pygen

Rules for generating native Bazel Python libraries from a requirements.txt file.

Currently, using pip and a requirements.txt file is the standard for installing dependencies in a Python project. Unfortunately, Bazel doesn't understand requirements.txt files and only understands the concept of a py_library; which is one or more Python source files. This library aims to bridge that gap. The rules_pygen generator takes a requirements.txt file as an input and generates a dependency graph of py_library rules based on the dependencies and subdependencies that pip would install given that requirements.txt file. In cases where a dependency is platform-specific, this tool generates two variants, one for MacOS and another for Linux.

The most common way to use this script is in a monorepo where one requirements.txt defines all the Python dependencies in your project, though you could potentially have more than one requirements file and run the script multiple times. Each time you change your requirements.txt (to add/modify a dependency) you re-run this script manually to generate the requirements.bzl variant which Bazel understands.

Limitations

  • The script itself runs on Python3.5+
  • Only works with wheels right now
  • Does not work on Windows
  • Config settings for linux / macos are assumed to live at //tool_bazel

Usage & Set up

  1. Add a git_repository rule to your WORKSPACE file to import the generator and rules_python for py_* rules:
git_repository(
    name = "rules_python",
    remote = "https://github.com/bazelbuild/rules_python.git",
    commit = "4b84ad270387a7c439ebdccfd530e2339601ef27",  # (2019-08-02 or later)
)

git_repository(
    name = "rules_pygen",
    remote = "https://github.com/tubular/rules_pygen.git",
    commit = "28835b7d278744916890f1ab3d974e7f5d75836c",
)
  1. Use the generator with local inputs:
bazel run @rules_pygen//:generator -- $(pwd)/path/to/python/requirements.txt $(pwd)/path/to/python/requirements.bzl //3rdparty/python --python=37
  1. Add to your WORKSPACE:
load("@//path/to:requirements.bzl", pypi_deps = "pypi_archives")
pypi_deps()
  1. Add two config_setting rules to differentiate linux and macos wheels (hardcoded at //tool_bazel for now)
config_setting(
    name = "linux",
    constraint_values = ["@bazel_tools//platforms:linux"],
    visibility=["//visibility:public"],
)

config_setting(
    name = "macos",
    constraint_values = ["@bazel_tools//platforms:osx"],
    visibility=["//visibility:public"],
)
  1. Use in a BUILD file:

alternative 1

py_library(
    name = "foo",
    srcs = glob([
        "src/foo/**/*.py",
    ]),
    deps = [
        "//3rdparty/python:pytest",
    ],
)

alternative 2

load("@//3rdparty/python:requirements.bzl", requirement = "requirement")
py_library(
    name = "baz",
    srcs = glob([
        "src/baz/**/*.py",
    ]),
    deps = [
        requirement("pytest"),
    ],
)

Development

Design choices

  • Generated build files should follow Skylark style guide (https://docs.bazel.build/versions/master/skylark/bzl-style.html) as much as possible
  • Code should be Python3.5+ compatible
  • No external dependencies; this is a library for generating imports -- we don't to make it hard to use by having it have imports of its own

Running tests

Unit

bazel run :generator_tests

Integration (run example project)

bazel run :generator -- $(pwd)/_examples/monorepo_foo/3rdparty/python/requirements.txt $(pwd)/_examples/monorepo_foo/3rdparty/python/requirements.bzl //3rdparty/python --python=37
cd _examples/monorepo_foo/
bazel run //component_a:greet

About

Rules for generating native Bazel Python libraries from requirements.txt

License:Apache License 2.0


Languages

Language:Python 97.0%Language:Starlark 3.0%