Evgeni Burovski

> while browsing the source code, this was found: `overwrite_b` is a typo indeed, would be nice to fix (same or separate quick PR). That `overwrite_?` arguments are ignored: yes,...

If push comes to shove, yes it's an option. I would argue to not go that route though. A minor technical reason is that numpy does extra work to return...

As discussed in the Oct 16 consortium meeting, the preference is to require eigenvalues to be always of a complex dtype. The last commit updates the text accordingly, and fixes...

As a data point, `scipy.interpolate` typically adds batch dimensions to the end. This did cause some mild inconveniences, mostly because it's not what usual broadcasting does.

@lucascolley you mean this: https://github.com/scipy/scipy/actions/runs/20090652780/job/57637273985?pr=24119 ? Then yes, the error message comes straigh from pytorch, no idea if torch changed it recently. Anyhow, +1 for adding the current message to...

> changing is_lazy_array to actually check if a JAX array is a tracer or a concrete array, I guess we should discuss whether it's OK to branch based on whether...

The behavior looks good to me, +1. ``` In [16]: a = np.array(42) In [17]: xp_assert_close(a, a.copy()) # great! In [18]: xp_assert_close(a, 2*(a//2)) # uhm, NumPy is not reasonable, OK...

That was me emulating your PR on a smaller scale. IIUC, the wrappers just mimic your asserions, and only need allow_0d->check_0d change. They simply opt out of the same np...

I was not going to comment here (see below for reasons) but since I've been pinged: - There was clearly no ill will, it's an unfortunate interaction with the ndimage...

> 20959 (comment) perhaps? Yes, thanks. Again, why was that one closed as "not planned", I've no idea.