Awesome

bumphub version 1.2.1

The project contains scripts to automate distribution and packaging of python modules on PyPI and conda.

pybump

The first utility, pybump, contains a script to automatically publish a package on PyPI. More specifically the following steps are carried out:

• Bump the version number of a python module (in __init__.py, in setup.py and README.md).
• Regenerate the documentation with pdoc3.
• Commit these changes with git.
• Add a git tag with the new version number to the latest commit.
• Push these changes upstream.
• Build the package.
• Upload the package to PyPI.
• Bump the version number again.
• Commit these changes with git.

These other python packages are dependencies:

• plumbing
• autopaths
• sh

These executables are required on the $PATH:

• gsed
• twine
• pdoc

You can use it from the shell like this:

$ pybump myproject

Command line options are:

• --test to upload to testpypi instead of the real PyPI.
• --infinite to keep incrementing the minor version number and never increment the major version number.

See https://packaging.python.org/guides/using-testpypi/ for the test option.

To get the dependencies, you can use:

$ conda env create -f environment.yml

anacondabump

The second utility, anacondabump, contains a script to automatically retrieve a package that was previously uploaded to PyPI, and upload an exact copy on a personal anaconda channel.

You can use it from the shell like this:

$ anacondabump myproject

To use this script you first need to install conda and some utilities (see next section).

These executables are required on the $PATH:

• conda
• conda-build
• grayskull
• anaconda

To get the dependencies, you can use:

$ conda env create -f environment.yml

Setup conda on macOS

To set up conda on macOS do the following:

$ brew install miniconda
$ conda init "$(basename "${SHELL}")"
$ mv ~/.bash_profile ~/.conda/conda_init.sh
$ source ~/.conda/conda_init.sh
$ conda install conda-build
$ conda install anaconda-client

You also need your anaconda credentials to be recorded by running the command:

$ anaconda login

Interesting blog article containing possible improvements to this script:

https://blog.gishub.org/how-to-publish-a-python-package-on-conda-forge

Uploading a package to PyPI manually

If you want to do the steps manually instead, follow this guide. Let's say your project is named myproject.

cd ~/repos/myproject
git init
git add *
git add .gitignore
git commit -m "First commit"

Now create a new empty repository on GitHub, let's say your username is bob. Then push to it:

git remote add origin git@github.com:bob/myproject.git
git push -u origin master

Upgrade the setuptools, and the wheel packages:

pip install --user --upgrade setuptools wheel

Build a wheel:

python setup.py sdist bdist_wheel

Test installation locally:

virtualenv test_env
source ./test_env/bin/activate
pip install dist/myproject-1.0.0-py2-none-any.whl
python -c "import myproject"

Clean up:

deactivate
rm -rf test_env

Upload the universal tar:

twine upload dist/*.tar.gz

Or if you prefer the version specific wheel:

twine upload dist/*.whl

Clean up:

rm -rf *.egg-info
rm -rf build
rm -rf dist

Dependencies

$ mamba install grayskull
$ mamba install conda-build
$ mamba install anaconda-client

Additional resources can be found here.

Uploading a package to anaconda manually

Resources on this topic can be found here.