Evgeni Burovski

As Nikolay said above and linked SO answers demonstrate, this is doable in specific cases with not too much difficulty using numpy broadcasting. What are the memory/speed tradeoffs needs to...

Edited in the interp2d deprecation plan from gh-17180 and the related mailing list discussion. Please let me know and/or adjust the landing page if what I added is not right.

One other thing I'd do here is keep the original python implementations as private functions somewhere in test_rgi, and actually add a test that a cython implementation gives consistent results.

Here's some teaching/demo materials for cython, in case it helps: https://github.com/ev-br/np_cy (`planet` stuff is heavily based on Pauli's lecture notes from 2011, adapted to jupyter notebooks) Re meson/cpp : it's...

Can you push the attempt which shows the error? IME with cython it's always taking several attempts, and it's way easier to figure out with a (failing) code snippet.

Here's a small step: https://github.com/scipy/scipy/commit/ca3febff43d0f1df9826b564ef7ce3313c2a9f5c Basically, it gets way easier if we convert everything to a double. At least for now. At least it now compiles even if fails all...

Oh indeed. Let's *not* ask Kai to join Cython Anonymous just yet though :-). For this PR specifically it's all premature, let's first see if there is a promise at...

Several `Buffer type mismatch` errors are due to the `xi` array being integer, so fixed by ```diff $ git diff diff --git a/scipy/interpolate/_rgi.py b/scipy/interpolate/_rgi.py index 6e1fc6d04e..354ed23d76 100644 --- a/scipy/interpolate/_rgi.py +++...

> Comparing the different indices arrays they do all seem to be off by 1. Maybe the way find_interval_ascending works is slightly different to that of np.searchsorted? Looking at https://github.com/scipy/scipy/blob/main/scipy/interpolate/tests/test_interpolate.py#L2268...

@Kai-Striega will do.