How to create a DataArray from a dask.array?

[ianliu]

Is it possible to create a DataArray from a dask.array? The API documentation for xarray.DataArray says the following for the data argument:

data (array_like) – Values for this array. Must be an numpy.ndarray, ndarray like, or castable to an ndarray. If a self-described xarray or pandas object, attempts are made to use this array’s metadata to fill in other unspecified arguments. A view of the array’s data is used instead of a copy if possible.

It doesn't mention anything about passing a dask.array, but the following snippet of code seems to work as expected:

import numpy as np
import dask
import dask.array as da
import xarray as xr

arr = da.stack([
    da.from_delayed(dask.delayed(np.ones)(5),
                    shape=(5,),
                    dtype=float)
    for _ in range(3)])

xr.DataArray(arr, coords={"y": list("abc"), "x": list(range(5))})

Can I assume that the aforementioned snippet of code doesn't compute the dask array to create the DataArray?

[keewis]

yes, you can rely on dask not to compute if you pass dask arrays to the constructor of DataArray.

I guess this part of the docstring has not been updated since quite a while? It is supposed to refer to duck array instead of array_like, since that's what we actually support.