ExtensionArray series.align(frame) not working

[jorisvandenbossche]

Aligning a Series containing extension array data with a dataframe does not work if there is actually something to be aligned:

In [27]: from pandas.tests.extension.decimal.array import DecimalArray, make_data

In [28]: dec_arr = DecimalArray(make_data()[:3])

In [29]: s = pd.Series(dec_arr)

In [30]: s
Out[30]: 
0    0.29242561210243966929311909552779980003833770...
1    0.34798224977276304148432473084540106356143951...
2    0.04963128775050906771326708621927537024021148...
dtype: decimal

In [31]: frame = pd.DataFrame({'col': np.arange(3)})

In [32]: s.align(frame)
Out[32]: 
(0    0.29242561210243966929311909552779980003833770...
 1    0.34798224977276304148432473084540106356143951...
 2    0.04963128775050906771326708621927537024021148...
 dtype: decimal,    col
 0    0
 1    1
 2    2)

In [33]: frame = pd.DataFrame({'col': np.arange(4)})

In [34]: s.align(frame)
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-34-94a43a65d337> in <module>()
----> 1 s.align(frame)

/home/joris/scipy/pandas/pandas/core/series.py in align(self, other, join, axis, level, copy, fill_value, method, limit, fill_axis, broadcast_axis)
   3231                                          fill_value=fill_value, method=method,
   3232                                          limit=limit, fill_axis=fill_axis,
-> 3233                                          broadcast_axis=broadcast_axis)
   3234 
   3235     def rename(self, index=None, **kwargs):

/home/joris/scipy/pandas/pandas/core/generic.py in align(self, other, join, axis, level, copy, fill_value, method, limit, fill_axis, broadcast_axis)
   7174                                      copy=copy, fill_value=fill_value,
   7175                                      method=method, limit=limit,
-> 7176                                      fill_axis=fill_axis)
   7177         elif isinstance(other, Series):
   7178             return self._align_series(other, join=join, axis=axis, level=level,

/home/joris/scipy/pandas/pandas/core/generic.py in _align_frame(self, other, join, axis, level, copy, fill_value, method, limit, fill_axis)
   7210         left = self._reindex_with_indexers(reindexers, copy=copy,
   7211                                            fill_value=fill_value,
-> 7212                                            allow_dups=True)
   7213         # other must be always DataFrame
   7214         right = other._reindex_with_indexers({0: [join_index, iridx],

/home/joris/scipy/pandas/pandas/core/generic.py in _reindex_with_indexers(self, reindexers, fill_value, copy, allow_dups)
   3812                                                 fill_value=fill_value,
   3813                                                 allow_dups=allow_dups,
-> 3814                                                 copy=copy)
   3815 
   3816         if copy and new_data is self._data:

/home/joris/scipy/pandas/pandas/core/internals.py in reindex_indexer(self, new_axis, indexer, axis, fill_value, allow_dups, copy)
   4377         if axis == 0:
   4378             new_blocks = self._slice_take_blocks_ax0(indexer,
-> 4379                                                      fill_tuple=(fill_value,))
   4380         else:
   4381             new_blocks = [blk.take_nd(indexer, axis=axis, fill_tuple=(

/home/joris/scipy/pandas/pandas/core/internals.py in _slice_take_blocks_ax0(self, slice_or_indexer, fill_tuple)
   4411             elif not allow_fill or self.ndim == 1:
   4412                 if allow_fill and fill_tuple[0] is None:
-> 4413                     _, fill_value = maybe_promote(blk.dtype)
   4414                     fill_tuple = (fill_value, )
   4415 

/home/joris/scipy/pandas/pandas/core/dtypes/cast.py in maybe_promote(dtype, fill_value)
    334     elif is_datetimetz(dtype):
    335         pass
--> 336     elif issubclass(np.dtype(dtype).type, string_types):
    337         dtype = np.object_
    338 

TypeError: data type not understood

Other combinations of aligning (frame with series, series with series) do work correctly:

In [36]: frame.align(s, axis=0)
Out[36]: 
(   col
 0    0
 1    1
 2    2
 3    3, 0    0.29242561210243966929311909552779980003833770...
 1    0.34798224977276304148432473084540106356143951...
 2    0.04963128775050906771326708621927537024021148...
 3                                                  NaN
 dtype: decimal)

In [37]: s.align(frame['col'])
Out[37]: 
(0    0.29242561210243966929311909552779980003833770...
 1    0.34798224977276304148432473084540106356143951...
 2    0.04963128775050906771326708621927537024021148...
 3                                                  NaN
 dtype: decimal, 0    0
 1    1
 2    2
 3    3
 Name: col, dtype: int64)