Contributions, whether big or small, are appreciated! You can get involved by submitting an issue, making a suggestion, or adding code to the project. PyKEEN is young and wants to address the problems the KGE community is currently facing, and has a lot of excited people working on it!
- Check that you have the latest version of
pykeen
- Check that StackOverflow hasn't already solved your problem
- Go here: https://github.com/pykeen/pykeen/issues
- Check that this issue hasn't been solved
- Click "new issue"
- Choose the appropriate issue template then follow its instructions. Issues not following the template may be discarded without review.
Same drill! Submit an issue and we'll have a nice conversation in the thread.
This project uses the GitHub Flow model for code contributions. Follow these steps:
- Create a fork of the upstream
repository
at
pykeen/pykeen
on your GitHub account (or in one of your organizations) - Clone your fork
with
git clone https://github.com/<your namespace here>/pykeen.git
- Make and commit changes to your fork with
git commit
- Push changes to your fork with
git push
- Repeat steps 3 and 4 as needed
- Submit a pull request back to the upstream repository
This project uses squash merges to group all related commits in a given pull request into a single commit upon acceptance and merge into the main branch. This has several benefits:
- Keeps the commit history on the main branch focused on high-level narrative
- Enables people to make lots of small commits without worrying about muddying up the commit history
- Commits correspond 1-to-1 with pull requests
This project encourages the use of optional static typing. It
uses mypy
as a type checker
and sphinx_autodoc_typehints
to automatically generate documentation based on type hints. You can check if
your code passes mypy
with tox -e mypy
.
This project uses black
to automatically
enforce a consistent code style. You can apply black
and other pre-configured
linters with tox -e lint
.
This project uses flake8
and several plugins for
additional checks of documentation style, security issues, good variable
nomenclature, and more (
see tox.ini
for a list of flake8 plugins). You can check if your
code passes flake8
with tox -e flake8
.
Each of these checks are run on each commit using GitHub Actions as a continuous integration service. Passing all of them is required for accepting a contribution. If you're unsure how to address the feedback from one of these tools, please say so either in the description of your pull request or in a comment, and we will help you.
Python's builtin print()
should not be used (except when writing to files),
it's checked by the
flake8-print
plugin to flake8
. If
you're in a command line setting or main()
function for a module, you can use
click.echo()
. Otherwise, you can use the builtin logging
module by adding
logger = logging.getLogger(__name__)
below the imports at the top of your
file.
All public functions (i.e., not starting with an underscore _
) should be
documented using
the sphinx documentation format.
The darglint
plugin to flake8
reports on functions that are not fully documented.
This project uses sphinx
to automatically build
documentation into a narrative structure. You can check that the documentation
properly builds with tox -e docs-test
and build the docs locally with
tox -e docs
.
Functions in this repository should be unit tested. These can either be written
using the unittest
framework in the tests/
directory or as embedded
doctests. You can check that the unit tests pass with tox -e py
and that the
doctests pass with tox -e doctests
. These tests are required to pass for
accepting a contribution.
If other code is updated before your contribution gets merged, you might need to resolve conflicts against the main branch. After cloning, you should add the upstream repository with
$ git remote add pykeen https://github.com/pykeen/pykeen.git
Then, you can merge upstream code into your branch. You can also use the GitHub UI to do this by following this tutorial.
This project aims to support all versions of Python that have not passed their
end-of-life dates. After end-of-life, the version will be removed from the Trove
qualifiers in the setup.cfg
and from the GitHub Actions testing
configuration.
See https://endoflife.date/python for a timeline of Python release and end-of-life dates.
PyKEEN uses single source versioning. This means that there's a variable
pykeen.version.VERSION
which is the canonical value used as the version.
Management of this value is done by bumpversion
via tox
. When you're
ready to make a release, do the following:
- Make sure there are no uncommitted changes.
- Run
tox -e bumpversion release
- Push to GitHub
- Draft a new release at https://github.com/pykeen/pykeen/releases/new.
Name the release based on the version that was just bumped to with the form
vX.Y.Z where X is major release, Y is minor release, and Z is patch. By default,
there's a box that says
Target: master
. If you're not 100% sure the last commit made before making a tag/release was the bump commit, click it, click "Recent Commits" then click the commit with the textBump version: X.Y.Z-dev -> X.Y.Z
.
Directly after making a release, you can easily upload to PyPI using another tox
command:
tox -e release
prepares the code and uploads it to PyPI.tox -e bumpversion patch
to bump the version. DO NOT do this before uploading to PyPI, otherwise the version on PyPI will have-dev
as a suffix.- Push to GitHub
The process of bumping the version (release), pushing to GitHub, making a release to PyPI,
bumping the version (patch), and pushing to GitHub one more time has been automated with
tox -e finish
. If you use this, make sure you go to GitHub and manually find the right
commit for making a tag/release, since it will not be the most recent one.