/pyfoobar

Python project template and best practices

Primary LanguagePython

pyfoobar

CircleCI travis codecov Code style: black awesome PyPi Version GitHub stars PyPi downloads

A Python project template that highlights some best practices in Python packaging. Can be used as a GitHub template for your new Python project.

Best practices

  • The name of the git repo should be the PyPi name of the package should be what you you type as import mypackagename. That means no hyphens in package names!

  • Your package should be a one-trick pony. Nobody wants to install a huge toolbox if all they need is the image converter in it.

  • After import yourpackagename, people should be able to call yourpackagename.__version__ plus some other meta data. This helps with debugging.

  • Choose a license for your code and provide a LICENSE[.txt] in the root level of your package as well as a statement in your main README. choosealicense.com can help you make a decision.

  • Use linting and formatting, include those in your integration tests.

    • black is a formatter that I like because you cannot configure it -- black is black.
    • Good linters are flake8 and pylint.
    • isort sorts your imports alphabetically.

  • Once you have tests in order, make sure they are executed with every git push. Two popular CI services that run your tests are travis and circleci. This repository contains the config files for each of the two.

  • Make sure that nobody can push to master. On GitHub, go to Settings -> Branches -> Add rule and select Require status checks to pass before merging and Include administrators. Development happens in pull requests, this makes sure that nobody -- including yourself -- ever accidentally pushes something broken to master.

  • Use a tool for measuring test coverage. codecov is one, and the circleci config submits the data to it.

  • Bitches love badges. If you have CI set up, want to show test coverage, or advertise the availability on PyPi, do so using a badge at the top of your README. Check out shields.io for what's available.

What you can do with this template

First run

find . -type f -name "*.py" -o -name Makefile -o -name "*.yml" -print0 | xargs -0 sed -i 's/pyfoobar/your-project-name/g'

and rename the folder pyfoobar to customize the name.

There is a simple Makefile that can help you with certain tasks:

  • Run make black to apply black formatting.

  • Run make lint to run flake8 linting

  • Run make publish to

    • tag your project on git (make tag)
    • upload your package to PyPi (make upload)

    After publishing, people can install your package with

    pip3 install --user pyfoobar

Testing

To run the pyfoobar unit tests, check out this repository and type

pytest

License

pyfoobar is published under the MIT license.