cookiecutter-pypackage-minimal
An opinionated, minimal cookiecutter template for Python packages, and some guidelines for Python packaging.
Usage
pip install cookiecutter
cookiecutter https://github.com/joamatab/cookiecutter-pypackage-minimal
You should then change the classifiers in {{ package_name }}/setup.py
- it is assumed that the project will run on the latest versions of Python 2 and 3, so you should remove any classifiers that do not apply. The full list of PyPI classifiers can be found here.
Fill out the README, and - if necessary - choose a license for the project.
Explanation
The decisions cookiecutter-pypackage-minimal
makes should all be explained here.
README
- README should use reStructuredText format This is the format used by most Python tools, is expected by setuptools, and can be used by Sphinx.
- As few README files as possible Additional README files (AUTHORS, CHANGELOG, etc) should be left to the user to create when necessary.
LICENSE
- MIT license by default
This template provides you the classic MIT licence: it lets people do almost anything they want with your project, including to make and distribute closed source versions.
If you choose another license, you also need to update the
{{ package_name }}/setup.py
file: adjust theclassifiers
andlicense
fields accordingly. - A license is a requirement
Nowadays, people who want to use your library/application want to make sure they can do it legally.
If your library is a private library, you can use a private license. In the
{{ package_name }}/setup.py
file, setlicense="Proprietary"
, and choose'License :: Other/Proprietary License'
in the trove classifiers.
setup.py
- Use setuptools
It's the standard packaging library for Python.
distribute
has merged back intosetuptools
, anddistutils
is less capable. - setup.py should not import anything from the package
When installing from source, the user may not have the packages dependencies installed, and importing the package is likely to raise an
ImportError
. - setup.py should be the canonical source of package dependencies
There is no reason to duplicate dependency specifiers (i.e. also using a
requirements.txt
file). See the testing section below for testing dependencies.
Testing
- Use Tox to manage test environments Tox provides isolation, runs tests across multiple Python versions, and ensures the package can be installed.
- Uses pytest as the default test runner This can be changed easily, though pytest is a easier, more powerful test library and runner than the standard library's unittest.
- Define testing dependencies in
tox.ini
Avoid duplicating dependency definitions, and usetox.ini
as the canonical description of how the unittests should be run. tests
directory should not be a package Thetests
directory should not be a Python package unless you want to define some fixtures. But the best practices are to use PyTest fixtures which provides a better solution. Therefore, thetests
directory has no__init__.py
file.