Justus Magin

I think I found it. `dask-expr>=1.0` sets `pd.set_option("mode.copy_on_write", True)` on import, which causes the change in behavior. This reproduces as well (note how the writable flag changes from `True` to...

thanks for the report, and we'd definitely like to be able to use reductions while ignoring `nan` values. However, as far as I can tell the array API does [not...

> would this be considered a bug? indeed, that's a bug. See #8566 and #8573 Edit: and #7243

I agree that debugging the somewhat complex piece of code you linked to is a bit much to ask for an issue (especially given that we're already somewhat behind in...

thanks for the minimal example. That reduces this issue to just "Should we allow backends to return chunked datasets?", which is much easier to discuss. Back when we introduced the...

can you explain why you think we require it? Is it that the tests fail for the released version (meaning we need a `importorskip` for the relevant tests)? Or does...

no worries, we apparently forgot to skip the tests that need it, so this is a good hint.

same as above: does this cause anything to fail? If it does, we can try to fix these. (For reference, one of the examples above is a helper function for...

right. I suppose we could remove the typing import until `DataTree` actually exists, and release a version immediately after merging the PR.

> If people import the private `core._datatree` package it's at their own risk. `xarray.datatree_` and `xarray.core.datatree` are both internal API, and anyone importing these two directly is on their own...