In case you haven’t followed (I would not blame you if you haven’t) what has transpired since I created this subject: that is not because Zig (in the form of ziglang) did cause problems, but because of how some tools/users use PyPI. As a result I decided to change the automatic dependeny of the package with the Zig compiled code from the main package (ruamel.yaml), that caused these downloads in the first place, to an optional one. I could do this because all functionality provided by the Zig compiled code is also available as pure Python in the main package (and the most important features, only as such).
I have started outlining a tool that can be used to remedy this (and thus also some of the problems Loris brought up in talk at Pycon Italia 2024), and that involves an even bigger role for Zig.
Sorry, my comment probably came across a bit more nasty than I meant it.
I’d followed the github thread, and (to turn it back to zig) I think it highlights the poor mindset a lot of people (not you) have with respect to publicly available packages. I’m hopeful the mindset in the zig community develops into a healthy scepticism of dependency bloat.
Hmm well anyone not pinning dependencies needs to be prepared to fix upgrade issues (myself included in a few places )
Regarding the missing python-dev packages, maybe there is a usecase to ship the headers for all python versions using pip build dependencies as well….Not so much the windows linkable library since it usually comes with the interpreter’s distribution.
With apologies to @Anthon, I think it’s best for me to lock this thread.
Normally a Showcase is a good place for the original author to come back to, from time to time, when said author has further updates to share. In that event I invite you specifically to open a second thread for that purpose.
)