PyPi package

Preparation

1. Adapt the version number in _meta

2. Check if docs/contributors and readme is up to date

Note

The Cython module for peakTree_fastreader.pyx is left out from the pypi package to simplify the build process.

Build the .tar and .whl (new)

python3.11 -m build
python3.11 -m twine upload -r testpypi --verbose dist/pyLARDA-3.3.7-*
python3.11 -m twine upload -r testpypi --verbose dist/pylarda-3.3.7*

Legacy build process

Build the .tar package

python3 setup.py sdist
# check build files
python3 -m twine check dist/*
tar tzf dist/pyLARDA-3.3.tar.gz
# upload to test server
python3 -m twine upload --repository testpypi dist/*
#
python3 -m twine upload dist/*

Build the windows binaries

Using the respective anaconda version:

python3 setup.py bdist_wheel

python3 setup.py bdist_wheel

set CONDA_FORCE_32BIT=1
conda create -n py38_32bit python=3.8

Build the manylinux binaries

pypi only accepts the manylinux wheels. Most conveniently they are packaged with the manylinux docker images (I only got the manylinux2014 working).

docker run -it -v $(pwd):/io 90ac8ec # or whatever id your manylinux image has

# might be necessary to update some dependencies
python3.9 -m pip install Cython
python3.9 -m pip install numba -U

python3.8 setup.py bdist_wheel
# and/or
python3.9 setup.py bdist_wheel

auditwheel repair <name-of-the-produced-wheel>.whl