align with broadcast_axis specified always uses inner join when aligning dataframe and series on other axis.

[kdebrab]

Pandas version 0.18.1:

In[1]: import pandas as pd
In[2]: df = pd.DataFrame(np.array([[1., 2.], [3., 4.]]), columns=list('AB'))
In[3]: ts = pd.Series([5., 6., 7.])
In[4]: df.align(ts, join='outer', axis=0, broadcast_axis=1)
Out[4]: 
(     A    B
 0  1.0  2.0
 1  3.0  4.0,      A    B
 0  5.0  5.0
 1  6.0  6.0)

However, as 'join' is specified to be 'outer', I would have expected the following result:

Out[4]: 
(     A    B
 0  1.0  2.0
 1  3.0  4.0
 2  NaN  NaN,      A    B
 0  5.0  5.0
 1  6.0  6.0
 2  7.0  7.0
 dtype: float64)

Same problem occurs for:

df.align(ts, join='right', axis=0, broadcast_axis=1)
ts.align(df, join='outer', axis=0, broadcast_axis=1)
ts.align(df, join='left', axis=0, broadcast_axis=1)

[jreback]

yeah this has very little testing. I'll mark it as a bug; seems like an easy fix if you'd like to take a crack at it.

[kdebrab]

The easy fix seems to replace in generic.py:
**other._construct_axes_dict() by **self._construct_axes_dict(**other._construct_axes_dict(axes=['columns']))
and
**self._construct_axes_dict() by **other._construct_axes_dict(**self._construct_axes_dict(axes=['columns']))

However, looking at the code I see another issue: though the documentation mentions as possible values for broadcast_axis: {0, 1, ‘index’, ‘columns’}, in reality only broadcast_axis == 1 has been implemented!

Unfortunately, I don't have the time now to implement that myself. I hope someone else will take this up...

[pfrcks]

Hi @jreback @kdebrab I would like to work on this as my first fix if that's not a problem?