Tom White

> I'll update the examples to do this. Done in https://github.com/cubed-dev/cubed/commit/c6b30c8e959377683234432a952aa3a20931366a

I think `map_overlap` could be implemented using Cubed's `map_direct`, which allows you to read arbitrary parts of Zarr arrays (it's used for indexing and concatenation already for example).

I've started an implementation of `map_overlap` here: https://github.com/tomwhite/cubed/commit/cd1a15c2acadac22ef8854d7dfe5ac9e50d8cdf0, and it seems to be fairly straightforward. It only supports constant boundary values, but it should be possible to implement some of...

> > a combination of xp.diff and xp.pad since Cubed would use several intermediate Zarr arrays > > This statement is confusing to me. Wouldn't `diff` just use `map_overlap`? That's...

There's more detail in the [blog post](https://xarray.dev/blog/cubed-xarray#pros-and-cons-of-cubeds-model) and [notebook](https://gist.github.com/TomNicholas/c6a28f7c22c6981f75bce280d3e28283).

> This is very helpful for me to understand what optimization options we have, but for a user this would seem like an unmotivated exposure of pretty deep internals. Is...

> To be clear - as a temporary record of the state of possible optimizations, this is great. My criticisms are more looking towards what would be most useful for...

I've re-worked this page to reflect the new defaults (multiple-input fusion is enabled by default now), and also to move the advanced settings to the end since they should be...

@saulshanabrook that would be awesome - thanks for the offer!

This should be possible, but I'm not sure how much this occurs in practice. The only calls to `rechunk` in Cubed are in `reshape` and `from_array` (for some cases).