But it would be expected (at least by me) that it does.

Bug in conversion frompd.series

MCVE Code Sample

```python import pandas as pd import xarray as xr

data3 = pd.DataFrame([(1, 2, 1), (2, 2, 1)],columns=['x', 'B', 'A']) data3 = data3.set_index(['x']) data3.rename_axis('tag', axis=1, inplace=True) data3 = data3.stack() print(data3) print(xr.DataArray.from_series(data3).sel(tag='B')) ```

Expected Output

[2,2]

Problem Description

The same code gives the expected output in xarray 0.14.1 Also works correctly if "tag" dimension is properly sorted during DataFrame cretion.

Versions

MCVE Code Sample

```python import xarray as xr

da1 = xr.DataArray([1,2],dims=['x'],coords={'x':[0,1]})

print(da1.indexes) da1['x'].values = [1,2] print(da1.indexes) print('--------') print(da1['x'])

x: Int64Index([0, 1], dtype='int64', name='x') x: Int64Index([0, 1], dtype='int64', name='x') -------- <xarray.DataArray 'x' (x: 2)> array([1, 2]) Coordinates: * x (x) int32 1 2 ```

Expected Output

```python x: Int64Index([0, 1], dtype='int64', name='x') x: Int64Index([1, 2], dtype='int64', name='x')

<xarray.DataArray 'x' (x: 2)> array([1, 2]) Coordinates: * x (x) int32 1 2 ```

Problem Description

• .indexes gets out of sync with the values stored in the array
• this is unfortunate as alignment with other array is done through .indexes

The issue does not occur if I a, either do not call ".indexes" before setting the value, or b, call da1['x'] = [1,2] instead of da1['x'].values = [1,2]

Output of xr.show_versions()

Code Sample

```python import numpy as np import xarray as xr

def noop(da): return da

def expd(da): return da.expand_dims(['A'])

na=np.array([[1,2],[3,4]]) xa=xr.DataArray(na,dims=['A','B'])

This works:

r1 = xa.groupby('A').apply(noop) print("!!!!")

This doesn't:

r2 = xa.groupby('A').apply(expd) ```

Problem description

Is this intended behavior? The reason I would need this is to be able to increase the size of each group.

Expected Output

I expect r1 to be equivalent to r2

This issue may be related to #1386

I hope this part of the stacktrace is sufficient to demonstrate my point:

File "C:\src\validation\main\risk_reporting\data_access\db_model.py", line 246, in get_data ds = ds.groupby('pvshort').apply(select_latest_pv) File "C:\Anaconda2\lib\site-packages\xarray\core\common.py", line 345, in groupby return self.groupby_cls(self, group, squeeze=squeeze) File "C:\Anaconda2\lib\site-packages\xarray\core\groupby.py", line 226, in init unique_values, group_indices = unique_value_groups(group, sort=sort) File "C:\Anaconda2\lib\site-packages\xarray\core\groupby.py", line 34, in unique_value_groups inverse, values = pd.factorize(ar, sort=sort) File "C:\Anaconda2\lib\site-packages\pandas\core\algorithms.py", line 552, in factorize values = _ensure_arraylike(values) File "C:\Anaconda2\lib\site-packages\pandas\core\algorithms.py", line 166, in _ensure_arraylike values = lib.list_to_object_array(values) TypeError: Argument 'obj' has incorrect type (expected list, got DataArray)