Broken support for `write_empty_chunks` with Zarr v3

[sotosoul]

What happened?

The to_zarr() method accepts a keyword argument write_empty_chunks but the Zarr 3 backend no longer supports it. Instead, it can accept config={'write_empty_chunks': False} but to_zarr does not.

xarray 2025.1.2, zarr 3.0.3

What did you expect to happen?

It would write the data in Zarr skipping the empty chunks..

Minimal Complete Verifiable Example

ds = xr.Dataset(...)
ds.to_zarr(write_empty_chunks=False, ...)  # Raises: TypeError: Group.create_array() got an unexpected keyword argument 'write_empty_chunks'

ds.to_zarr(config={'write_empty_chunks': False}, ...)  # Unexpected kwarg 'config'

MVCE confirmation

• Minimal example — the example is as focused as reasonably possible to demonstrate the underlying issue in xarray.
• Complete example — the example is self-contained, including all data and the text of any traceback.
• Verifiable example — the example copy & pastes into an IPython prompt or Binder notebook, returning the result.
• New issue — a search of GitHub Issues suggests this is not a duplicate.
• Recent environment — the issue occurs with the latest version of xarray and its dependencies.

Relevant log output

Anything else we need to know?

May I also suggest that you add a zarr_backend_kwargs argument that gets passed to the Zarr backend?

Environment

xarray=2025.1.2, zarr=3.0.3

[welcome]

Thanks for opening your first issue here at xarray! Be sure to follow the issue template!
If you have an idea for a solution, we would really welcome a Pull Request with proposed changes.
See the Contributing Guide for more.
It may take us a while to respond here, but we really value your contribution. Contributors like you help make xarray better.
Thank you!

[ShaneMill1]

Having this problem as well, so following this issue

[sotosoul]

I'm working on a fix in my fork.

The fix entails passing a config argument (dict) to the Zarr backend, which expects it as documented here (create) and here (create_array) (and the respective sync methods).

It is my first time contributing to xarray, so I'll appreciate any help or hints that come this way :)

[sotosoul]

My approach is to add a zarr_store_config optional kwarg as follows:

def to_zarr(
    dataset: Dataset,
    ...
    zarr_store_config: dict[str, Any] | None = None,

I got that part working and backwards compatible with Zarr 2. This change alone, however, won't solve the issue because of the fill_value issue, i.e., that Zarr's perception of "fill value" is different from Xarray's. In the "eyes" of Zarr, empty chunks contain only fill value cells. To my understanding, a decision was made to not use the Dataset's _FillValue as Zarr's fill_value in v3.

I propose that we add another optional parameteruse_zarr_fill_value_as_mask to the ds.to_zarr method that will default to True for Zarr v2 and False for v3 so that it can be overriden if storing GeoTIFFs, for example. The parameter could be a kwarg or an expected key in the zarr_store_config dictionary.

Thoughs?