preserve chunked data when creating DataArray from itself

[FabianHofmann]

What happened:

When creating a new DataArray from a DataArray with chunked data, the underlying dask array is converted to a numpy array.

What you expected to happen:

I expected the underlying dask array to be preseved when creating a new DataArray instance.

Minimal Complete Verifiable Example:

import xarray as xr
import numpy as np
from dask import array

d = np.ones((10, 10))
x = array.from_array(d, chunks=5)

da = xr.DataArray(x) # this is chunked
xr.DataArray(da) # this is not chunked anymore

Anything else we need to know?:

Environment:

Output of xr.show_versions()

INSTALLED VERSIONS ------------------ commit: None python: 3.9.7 (default, Sep 16 2021, 13:09:58) [GCC 7.5.0] python-bits: 64 OS: Linux OS-release: 5.11.0-40-generic machine: x86_64 processor: x86_64 byteorder: little LC_ALL: None LANG: en_US.UTF-8 LOCALE: ('en_US', 'UTF-8') libhdf5: 1.10.6 libnetcdf: 4.7.4

xarray: 0.19.0
pandas: 1.3.3
numpy: 1.20.3
scipy: 1.7.1
netCDF4: 1.5.6
pydap: None
h5netcdf: 0.11.0
h5py: 3.2.1
Nio: None
zarr: 2.10.1
cftime: 1.5.0
nc_time_axis: None
PseudoNetCDF: None
rasterio: 1.2.6
cfgrib: None
iris: None
bottleneck: 1.3.2
dask: 2021.09.1
distributed: 2021.09.1
matplotlib: 3.4.3
cartopy: 0.19.0.post1
seaborn: 0.11.2
numbagg: None
pint: None
setuptools: 58.0.4
pip: 21.2.4
conda: 4.10.3
pytest: 6.2.5
IPython: 7.27.0
sphinx: 4.2.0

[dcherian]

Can you give us a little more context about why this might be useful? IIRC we disallowed creating dataarrays from dataarrays in some other place because it leads to ambiguous situations like the following

xr.DataArray(da, attrs={"a": 1})  # does the result have da.attrs or the provided attrs?

[FabianHofmann]

Ah yes, this is indeed ambiguous. On the other hand, as long it is still supported to create DataArray's from DataArray's they should at least preserve the data format. I need this as I am creating a subclass from the xarray.DataArray (see https://github.com/PyPSA/linopy/blob/8ac34d9fdbddc1fec0c7b4781f3d49e9c5ae064e/linopy/constraints.py#L18). In case I want to convert a lazy DataArray to my custom class the chunked data is directly computed, which seems a bit weird...

[dcherian]

IMO we should raise an error asking the user to pass da.data instead

[FabianHofmann]

Not sure, but I'd argue to keep the DataArray-from-self-construction as I imagine many convenience cases where arrays maybe DataArray, numpy arrays or dask arrays, and one wants to ensure a DataArray type. Many other packages like pandas/numpy have that. Also the xarray.Dataset supports from-self-construction.

Perhaps it is better to raise an error when ambiguities occur? Meaning don't allowing to pass attrs, coords when data is an DataArray...