Pamphile Roy

Interesting take on Alpine. In the end, there are other issues than package support. https://pythonspeed.com/articles/alpine-docker-python/ In any case, we would only consider it after NumPy does build wheels for this...

Yep, that the premise from the article I linked. Unless you put in a lot of effort to clean the image, you will not gain much with all that's needed...

> Dependencies required to build `numpy` or any other package should not end up in the final image, only the ones required to run it. And also if `musl` `numpy`...

Hi @oldestprogrammer, thank you for reporting. From your log, you are using macOS Sierra (macOS 10.12), but we only distribute wheels for macOS Monterey (macOS 12). So you just cannot...

Right, some misreading on my part. Did you try to install in a conda environment instead? Or building from source? https://scipy.github.io/devdocs/dev/dev_quickstart.html. You could also try with some nightly builds https://github.com/scipy/scipy/actions/runs/2908455453#artifacts

I am +100 on this proposal. There are too many issues with the current infrastructure and it's very hard to maintain. Starting off with a clean slate will definitely be...

@josef-pkt to > Allowing both in a generic way looks to me very difficult. One of the main point of this proposal is to remove the concept of frozen distributions....

> @josef-pkt I think simply initializing a new class for different parameters to achieve the stateless class behavior isn't too bad. We can also add static methods in the new...

Following your example. You could write a small wrapper that would re-use a given class instance and just vary attributes (shape parameters). It might even be faster than instantiating a...

I don't believe that having to modify the shape parameters of a distribution is the common use of distributions. I am not saying this is not useful, but that it...