Add chunks argument to {zeros/ones/empty}_like.

[nbren12]

Describe the solution you'd like

We have started using xarray objects as "schema" for initializing zarrs that will be written to using the region argument of to_zarr. For example,

output_schema.to_zarr(path, compute=False)
for region in regions:
    output = func(input_data.isel(region))
    output.to_zarr(path, region=region)

Currently, xarray's tools for computing the output_schema Dataset are a lacking since rechunking existing datasets can be slow. dask.array.zeros_like takes a chunks argument, can we add one here too?

Describe alternatives you've considered

.chunk

[max-sixty]

If #8251 were solved, we could do:

xr.DataArray().chunk(42).broadcast_like(ds)

In the place of that, we can do a hack:

xr.DataArray([1]).chunk(2).broadcast_like(ds).squeeze('dim_0')

(where broadcast_like can also be adding on dimensions with .expand_dims — not sure exactly how you prefer to initialize the arrays)

I would have a preference for using orthogonal methods like .chunk, rather than adding more kwargs to methods.

Do you think that works for your case @nbren12 ?

[dcherian]

rather than adding more kwargs to methods.

I think these array creation functions are almost always going to be messy if we want to support all the array types in them. It just happens that there is a dask-specific method. Like we might want to pass format for pydata/sparse (example)

We could accept a array_like argument, and pass **kwargs to the appropriate library though it turns out numpy.zeros does accept like but numpy.zeros_like does not accept like. I think the inconsistency is OK. I've never missed xarray.zeros

[max-sixty]

We could accept a array_like argument, and pass **kwargs to the appropriate library

+1

[dcherian]

Wait this was added at some point: https://docs.xarray.dev/en/stable/generated/xarray.zeros_like.html

xarray.zeros_like(other, dtype=None, *, chunks=None, chunked_array_type=None, from_array_kwargs=None)

[max-sixty]

Ha! OK, well, ask and ye shall receive...

(I certainly don't feel strongly enough to suggest removing it...)