Stephan Hoyer

This has gone stale.

Belatedly -- I like this idea!

Can we delete the "Flexible indexes" meeting? It doesn't happen anymore.

I think the bug here is the use of `self._attrs` in `__dask_tokenize__`. It should use `self.attrs` instead, which is basically equivalent to `self._attrs or {}`.

> As an alternative, we could instead perhaps use make_template() internally inside ChunksToZarr. I implemented this in https://github.com/google/xarray-beam/pull/62

You're welcome to act on this (or not), but I'm not planning to submit a PR. On Sun, Mar 24, 2024 at 8:46 PM Varun Godbole ***@***.***> wrote: > Assigned...

Here was my original reasoning: compiling a new gufunc helps because unlike reshaping you can be sure you'll never need to copy the data with the reshape. Then as long...

My guess is that NumPy is better at using vectorized CPU instructions for some reason. No idea why Numba can't do this, though...

I'm curious about the reasoning for putting this into map vs vmap? I do think the later is a more accessible/widely used API, but maybe that also makes it higher...

> @cgarciae FYI @inailuig noticed a while ago that the approach you are using here (which we also are using) breaks sharding propagation if you chunk an axis that is...