Decide on packaging approach for python (sdist, vs binary distribution)

[jurra]

What are the options for packaging, what pros and cons do you see?

What option should we go for, sdist, vs binary distribution.

Requirement:
The main requirement is that the solution has the lowest friction for the users and can work for different os.

[jurra]

One tool that can help with this is cibuildwheel. It runs on your CI server and builds and tests your wheels across all of your platforms. It currently supports GitHub Actions, Azure Pipelines, Travis CI, AppVeyor, CircleCI, and GitLab CI3.

Here is a step-by-step approach to publishing a Python package with a C++ dependency:

• Write your C++ code and create a setup.py file that specifies how to build the C++ code into a Python extension module.-
• Use cibuildwheel to build wheels for your package on the desired platforms.
• Test the built wheels to ensure they work correctly.
• Use twine to upload the built wheels to PyPI.