df.groupby(...).transform(func) breaks when func renames input df

[DiegoAlbertoTorres]

Calling df.groupby(...).transform(func) breaks when func is capable of taking in a pd.DataFrame but renames its columns.

Code sample

def demean_rename(x):
    result = x - x.mean()

    if isinstance(x, pd.Series):
        return result

    result = result.rename(columns=
        {c: '{}_demeaned'.format(c) for c in result.columns})

    return result

df = pd.DataFrame({'group': list('ababa'),
                   'value': [1, 1, 1, 2, 2]})
expected = pd.DataFrame({'value': [-1./3., -0.5, -1./3., 0.5, 2./3.]})

result = df.groupby('group').transform(demean_rename)
tm.assert_frame_equal(result, expected)
# Instead, this prints:
# E   DataFrame.iloc[:, 0] values are different (40.0 %)
# E   [left]:  [-0.33333333333333326, nan, -0.33333333333333326, nan, 0.6666666666666667]

Problem description

The current behavior gives wrong results (with nan) for everything except the first group. This happens even when func can be called successfully with each column and not return any nans.

This problem is present in master. I already have the fix: there is a bug in how results from the slow-path and the fast-path are joined in df.groupby(...).transform(func). I will be putting up a PR very soon with the fix.