@http

There are a lot of #tools and #libraries - in #Python and other languages - that are basically #wrappers around #compiled libraries written in C, C++, or other compiled languages. In general, installing the Python package from a repository declares the binary library package as a #dependency.

A name collision between the Python package and the underlying C library would be problematic. You could argue that a Python library that exposes the functionality of `libfrobnicate`, which is part of the `frobnicate` package, shouldn't itself be called `frobnicate` but something totally different - but people go searching for "Python for Frobnicate" so it's a natural enough name. And therefore the repository maintainers have to make it `python-frobnicate` etc.

@ianhopkinson

In general, no, but in data science your chances of problems are a little higher. If you're relying on any binary wheels in your work, you may find it more difficult because many projects don't produce wheels for non-dominant architectures on any given OS. It was really common on MacOS when the Mx chips were newer, to find only x86 binary wheels.

So you may have to build wheels for compiled extensions, which can be fiddly.

#compiled #extension #wheel #binary