Matt Haberland

Thanks @chrisb83! @AtsushiSakai took a look at gh-19923 and approved, but was unable to test locally, so a second opinion was needed. In case you're interested, it provides a huge...

This is the source of the problem. ```python3 from scipy.special import ive ive(1, 1e10) # nan ``` Would you like to see a temporary workaround in stats, or should this...

I thought that these benchmarks are relevant to mention here. https://github.com/scipy/scipy/blob/f0de2e78e6f45fa5b24b86229ed1e2d29738a7b1/benchmarks/benchmarks/stats.py#L192 https://github.com/scipy/scipy/blob/f0de2e78e6f45fa5b24b86229ed1e2d29738a7b1/benchmarks/benchmarks/stats.py#L279

The first failure suggests an edge case bug in `jax.scipy.special.ndtri`: ```python3 from scipy import special from jax import numpy as jnp from jax.scipy.special import ndtri, gammainc, gammaincc special.ndtri(-1.) # nan...

I'll fix this shortly. In the meantime, please use: ``` stats.wilcoxon(A, method='approx') ``` The problem is just with `method='auto'` (and `method='exact'` if there are zeros).

What is the shape of the input, and does it have NaNs? Try `stats.wilcoxon(A, _no_deco=True)` if you want a speed boost, but it will behave as though it omits NaNs....

Great to hear : ) Hopefully it will get array API support before too long, too.

This is getting a little out of hand with lint complaining about `UP031` in CI but not locally. I fix one thing and it starts complaining about another : (

Probably not. I thought we would have configured a rule set, though. I guess as ruff has added these, and we don't exclude them.

Interesting. No, I don't think it is picking up in CI. I am running with CuPy locally, and I wasn't getting those. Upgrading ruff seems to have broken my in-place...