isnull loads dask array

[fujiisoup]

From gitter
cc. @davidh-ssec

In [1]: import numpy as np
   ...: import xarray as xr
   ...: 
   ...: da = xr.DataArray(np.arange(100), dims='x').chunk({'x': 10})
   ...: da.isnull()
   ...: 
Out[1]: 
<xarray.DataArray (x: 100)>
array([False, False, False, False, False, False, False, False, False, False,
       False, False, False, False, False, False, False, False, False, False,
       False, False, False, False, False, False, False, False, False, False,
       False, False, False, False, False, False, False, False, False, False,
       False, False, False, False, False, False, False, False, False, False,
       False, False, False, False, False, False, False, False, False, False,
       False, False, False, False, False, False, False, False, False, False,
       False, False, False, False, False, False, False, False, False, False,
       False, False, False, False, False, False, False, False, False, False,
       False, False, False, False, False, False, False, False, False, False], dtype=bool)
Dimensions without coordinates: x

Problem description

DataArray.isnull() automatically computes dask arrays.

Expected Output

Out[2]: 
<xarray.DataArray (x: 100)>
dask.array<shape=(100,), dtype=bool, chunksize=(10,)>
Dimensions without coordinates: x

Cause

xarray/xarray/core/ops.py

Lines 322 to 324 in 697cc74

	for name in PANDAS_UNARY_FUNCTIONS:
	f = _func_slash_method_wrapper(getattr(pd, name), name=name)
	setattr(cls, name, cls._unary_op(f))

Here, getattr(pd, name) should be getattr(duck_array_ops, name).