Skip to content

Latest commit

 

History

History
59 lines (44 loc) · 3.52 KB

README.md

File metadata and controls

59 lines (44 loc) · 3.52 KB

cookiecutter-pypackage-minimal

An opinionated, minimal cookiecutter template for Python packages, and some guidelines for Python packaging.

Usage

pip install cookiecutter
git clone https://github.com/kragniz/cookiecutter-pypackage-minimal.git
cookiecutter 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 the classifiers and license 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, set license="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 into setuptools, and distutils 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 use tox.ini as the canonical description of how the unittests should be run.
  • tests directory should not be a package The tests 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, the tests directory has no __init__.py file.