BUG: _flex_binary_moment() doesn't preserve column order or handle multiple columns with the same label

[seth-p]

_flex_binary_moment() sorts the columns alphabetically. I think it should preserve the order.

In [23]: df = DataFrame([[2,4],[1,2],[5,2],[8,1]], columns=['B','A'])

In [24]: expanding_corr(df, df, pairwise=True)[3]
Out[24]:
          A         B
A  1.000000 -0.670166
B -0.670166  1.000000

(I originally mentioned this in #7514, but then broke it out as a separate issue.)

Also, _flex_binary_moment() doesn't work if multiple columns have the same name. While it's not obvious to me what the outcome should be for two distinct DataFrames with pairwise=False, there's no reason things shouldn't work for all other cases (i.e. a DataFrame with itself with pairwise=True or False; two distinct DataFrames with pairwise=True; and a DataFrame with a Series).

In [13]: df = DataFrame([[2,4],[1,2],[5,2],[8,1]], columns=['C','C'])

In [14]: expanding_corr(df, df, pairwise=True)
---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-14-f20d94c2d8dc> in <module>()
----> 1 expanding_corr(df, df, pairwise=True)

C:\Python34\lib\site-packages\pandas\stats\moments.py in expanding_corr(arg1, arg2, min_periods, freq, center, pairwise)
    940     return rolling_corr(arg1, arg2, window,
    941                         min_periods=min_periods,
--> 942                         freq=freq, center=center, pairwise=pairwise)
    943
    944

C:\Python34\lib\site-packages\pandas\stats\moments.py in rolling_corr(arg1, arg2, window, min_periods, freq, center, pairwise, how)
    246         return num / den
    247
--> 248     return _flex_binary_moment(arg1, arg2, _get_corr, pairwise=bool(pairwise))
    249
    250

C:\Python34\lib\site-packages\pandas\stats\moments.py in _flex_binary_moment(arg1, arg2, f, pairwise)
    278                             results[k1][k2] = results[k2][k1]
    279                         else:
--> 280                             results[k1][k2] = f(*_prep_binary(arg1[k1], arg2[k2]))
    281                 return Panel.from_dict(results).swapaxes('items', 'major')
    282             else:

C:\Python34\lib\site-packages\pandas\stats\moments.py in _get_corr(a, b)
    239         adj_window = min(window, len(a), len(b))
    240         num = rolling_cov(a, b, adj_window, min_periods, freq=freq,
--> 241                           center=center)
    242         den = (rolling_std(a, adj_window, min_periods, freq=freq,
    243                            center=center) *

C:\Python34\lib\site-packages\pandas\stats\moments.py in rolling_cov(arg1, arg2, window, min_periods, freq, center, pairwise, how)
    217         bias_adj = count / (count - 1)
    218         return (mean(X * Y) - mean(X) * mean(Y)) * bias_adj
--> 219     rs = _flex_binary_moment(arg1, arg2, _get_cov, pairwise=bool(pairwise))
    220     return rs
    221

C:\Python34\lib\site-packages\pandas\stats\moments.py in _flex_binary_moment(arg1, arg2, f, pairwise)
    290                 results[col] = f(X[col], Y)
    291
--> 292         return DataFrame(results, index=X.index, columns=res_columns)
    293     else:
    294         return _flex_binary_moment(arg2, arg1, f)

C:\Python34\lib\site-packages\pandas\core\frame.py in __init__(self, data, index, columns, dtype, copy)
    201                                  dtype=dtype, copy=copy)
    202         elif isinstance(data, dict):
--> 203             mgr = self._init_dict(data, index, columns, dtype=dtype)
    204         elif isinstance(data, ma.MaskedArray):
    205             import numpy.ma.mrecords as mrecords

C:\Python34\lib\site-packages\pandas\core\frame.py in _init_dict(self, data, index, columns, dtype)
    325
    326         return _arrays_to_mgr(arrays, data_names, index, columns,
--> 327                               dtype=dtype)
    328
    329     def _init_ndarray(self, values, index, columns, dtype=None,

C:\Python34\lib\site-packages\pandas\core\frame.py in _arrays_to_mgr(arrays, arr_names, index, columns, dtype)
   4623
   4624     # don't force copy because getting jammed in an ndarray anyway
-> 4625     arrays = _homogenize(arrays, index, dtype)
   4626
   4627     # from BlockManager perspective

C:\Python34\lib\site-packages\pandas\core\frame.py in _homogenize(data, index, dtype)
   4932
   4933             v = _sanitize_array(v, index, dtype=dtype, copy=False,
-> 4934                                 raise_cast_failure=False)
   4935
   4936         homogenized.append(v)

C:\Python34\lib\site-packages\pandas\core\series.py in _sanitize_array(data, index, dtype, copy, raise_cast_failure)
   2507             raise Exception('Data must be 1-dimensional')
   2508         else:
-> 2509             subarr = _asarray_tuplesafe(data, dtype=dtype)
   2510
   2511     # This is to prevent mixed-type Series getting all casted to

C:\Python34\lib\site-packages\pandas\core\common.py in _asarray_tuplesafe(values, dtype)
   2191             except ValueError:
   2192                 # we have a list-of-list
-> 2193                 result[:] = [tuple(x) for x in values]
   2194
   2195     return result

ValueError: cannot copy sequence with size 2 to array axis with dimension 4