Easy Python Packaging (7 steps to publish on PyPI)

This document aims at giving a quick tutorial for structuring, packaging, and publishing your Python package on https://pypi.org/.

To illustrate this tutorial, let's work on distribution of a full Python package called otquickmodule example, based on the following tutorial https://github.com/judy2k/publishing_python_packages_talk.

Our illustrative package, otquickmodule, includes two classes:

• KernelHerding
• QuadratureWeighting

0. Create a new environment

Before strating, creating a new environment for this occasion is highly recommended to properly manage dependencies (we decided to name ours otqm_env). For more information, check-out the conda environments documentation https://docs.conda.io/projects/conda/en/latest/user-guide/tasks/manage-environments.html.

~$ conda create -n otqm_env python=3.9
~$ conda activate otqm_env

Do not forget to activate this environment for all the following steps.

1. Write __init__.py file

This file will allow the auto-completion your methods when using the package. Note that the version defined here will be the package version (including on pypi).

"""otquickmodule module"""

from .KernelHerding import KernelHerding
from .QuadratureWeighting import QuadratureWeighting

__all__ = [
"KernelHerding",
"QuadratureWeighting",
]
__version__ = "0.0.1"

2. Create the package structure

Before creating the package structure, we recommend to follow a few good practices in Python:

• Avoid unnecessary dependencies to simplify package maintenance
• Follow the PEP 8 coding standards and use the package black to format your code according to PEP 8.
• Write explicit docstrings on each method using the numpydoc syntax. This formalism will allow you to easily create a nice documentation in the future (e.g., https://efekhari27.github.io/otkerneldesign/master/)

Here is the recommended Python package structure, do not forget to place the __init__.py file along your classes.

PACKAGE STRUCTURE (step 2)
==========================
├── otquickmodule
|    ├── example
|    ├── test
|    ├── otquickmodule
|    |   ├── __init__.py
|    |   ├── QuadratureWeighting.py
|    |   └── KernelHerding.py

3. Add license, readme and gitignore files

• The README.md file should describe the package and will appear on your github repository and your pypi page.

• To help you choosing a license: https://choosealicense.com/. In our case, we chose a GPL license in the form of the file LICENSE.

• To create a .gitignore file (to make sure git doesn't follow auxillary files): copy ours or check out https://www.toptal.com/developers/gitignore/

PACKAGE STRUCTURE (step 3)
==========================
├── otquickmodule
|    ├── example
|    ├── test
|    ├── otquickmodule
|    |   ├── __init__.py
|    |   ├── QuadratureWeighting.py
|    |   └── KernelHerding.py
|    ├── LICENSE
|    ├── README.md
|    └── .gitignore

4. Write a setup.py file

The setup.py is the most important file for installing your package. Note that the name defined here will be the name published on pypi.

# coding: utf8
"""
Setup script for a Python package
"""
import os
import re
from setuptools import setup

# Get the version from __init__.py
path = os.path.join(os.path.dirname(__file__), 'otquickmodule', '__init__.py')
with open(path) as f:
  version_file = f.read()

version = re.search(r"^\s*__version__\s*=\s*['\"]([^'\"]+)['\"]",
version_file, re.M)
if version:
  version = version.group(1)
else:
  raise RuntimeError("Unable to find version string.")

# Long description
with open("README.md", "r") as fh:
  long_description = fh.read()

setup(
name="otquickmodule",
version=version,
author="Elias Fekhari",
author_email="elias.fekhari@edf.fr",
description="This repository is a tutorial for easy Python packaging",
license='GPLv3+',
keywords=['OpenTURNS', 'KernelHerding'],
url="https://github.com/efekhari27/otquickmodule",
packages=['otquickmodule', 'test'],
long_description=long_description,
long_description_content_type="text/markdown",
classifiers=[
"Programming Language :: Python :: 3",
"License :: OSI Approved :: GNU General Public License v3 or later (GPLv3+)",
"Intended Audience :: Science/Research",
"Topic :: Software Development",
"Topic :: Scientific/Engineering",
],
install_requires=[
"numpy",
"scipy", 
"openturns>=1.17"
],
)

At this step, you should be able to run the following commands without error. It will first create a build repertory with the sources of your package. Then, you can try to install the package locally (the option -e imposes a local install).

~/otquickmodule$ python setup.py bdist_wheel
~/otquickmodule$ pip install -e .

You can now import the module from anywhere with the Python interpreter of the otqm_env conda environment:

~$ conda activate otqm_env
~$ python 
>>> import otquickmodule as otqm
>>>

5. Write tests and examples

Each class should have a corresponding test file. Tests can either be writen using the unittest or pytest package. pytest is very popular and present various practical features but requires to be added to the package requirements. For our needs, we chose to use the native Python module unnitest.

Additionally, users appreciate to get started with a simple usage example which can be a ipython notebook file.

PACKAGE STRUCTURE (step 5)
==========================
├── otquickmodule
|    ├── build
|    |   ├── ...
|    |   └── ...
|    ├── example
|    |   └── kernel_herding_example.ipynb
|    ├── test
|    |   ├── TestQuadratureWeighting.py
|    |   └── TestKernelHerding.py
|    ├── otquickmodule
|    |   ├── __init__.py
|    |   ├── QuadratureWeighting.py
|    |   └── KernelHerding.py
|    ├── LICENSE
|    ├── README.md
|    ├── setup.py
|    └── .gitignore

At this step you can repeat the build and install commands from the previous step.

6. Build, distribute and publish

Create the source distribution of the package: (creates a tar gz archive and the hosting dist repertory)

~/otquickmodule$ python setup.py bdist_wheel sdist

Check if it worked: (you should see the content of the archive)

~/otquickmodule$ tar tzf dist/otquickmodule-0.0.1.tar.gz 

Create a PyPI account online and remember your username and password.

Install the package twine:

~/otquickmodule$ conda install twine

Publish the package on PyPI:

~/otquickmodule$ twine upload dist/*
Uploading distributions to https://upload.pypi.org/legacy/
Enter your username: YOUR_USER_NAME
Enter your password: YOUR_PASSWORD
Uploading otquickmodule-0.0.1.tar.gz
100%|██████████████████████████████████████████████████████████████| 3.49k/3.49k [00:01<00:00, 2.17kB/s]

View at:
https://pypi.org/project/otquickmodule/0.0.1/

Your package is available online!

7. Try your package

Create or switch to a different conda environment and install your package from pypi:

~$ conda activate other_env
~$ pip install otquickmodule