Project names on pypi and Python package names are a rather terrible mess: in many cases, a useful package name is already taken by some abandoned project with version 0.0.1.
So in order to publish a package with a rather common name, let’s for argument’s sake say “maths” (which has a version 0.0.0 project on PyPi already), the safest approach seems to be to instead use a personal/organizational namespace, e.g. xyz.maths.
A bit later on, maybe we want to publish yet another package xyz.graphs ( of course there is already a package graphs 0.1.3 on PyPi).
If I understand correctly there is already a recommendation for using namespace packages to indicate ownership in PEP 0423
My question is: how to do this properly on PyPi? All the information I found so far is confusing and partly contradictory (e.g. PEP0423 seems to say that a PyPi project name should include the dot while Packaging namespace packages — Python Packaging User Guide seems to recommend using names whith underscores?
So is there a best practice for:
- choosing a pypi project name for
- the namespace packe xyz (is this even necessary?)
- all the subpackages like xyz.maths, xyz.graphs
- creating the proper source code structures and setup.py for those projects
- creating corresponding projects on RTD (readthedocs)
- maybe also naming/organizing the corresponding source code on GitHub/GitLab?