BUG: loc/ix indexers with empty list key lose names of respective axes

[immerrr]

This issue was encountered when fixing #6536. It seems that .loc indexer with empty list key seems to lose names of respective axes:

In [2]: pd.DataFrame(np.arange(9).reshape(3,3), columns=pd.Series(list('abc'), name='foobar'))
Out[2]: 
foobar  a  b  c
0       0  1  2
1       3  4  5
2       6  7  8

[3 rows x 3 columns]

In [3]: _2.iloc[:,[]]
Out[3]: 
Empty DataFrame
Columns: []
Index: [0, 1, 2]

[3 rows x 0 columns]

In [4]: _2.iloc[:,[]].columns.names
Out[4]: FrozenList([u'foobar'])

In [5]: _2.loc[:,[]].columns.names
Out[5]: FrozenList([None])

It's not important, but currently #6551 has two tests skipped because of this bug.

[immerrr]

This happens because Index.reindex discards names:

In [24]: idx = pd.Index(list('abc'), name='foobar')

In [25]: idx.names
Out[25]: FrozenList([u'foobar'])

In [27]: idx.reindex([])[0].names
Out[27]: FrozenList([None])

In [28]: idx.reindex(list('abc'))[0].names
Out[28]: FrozenList([None])

In [29]: idx.reindex(idx)[0].names
Out[29]: FrozenList([u'foobar'])

In [30]: idx.reindex(pd.Index([], name='xyzzy'))[0].names
Out[30]: FrozenList([u'xyzzy'])

I think it's more appropriate to preserve names unless the target iterable is an Index itself and in that case its names should be used.

On an unrelated topic, the same logic may be applied to other metadata, too, like timezone.

[jreback]

@immerrr doable for 0.15.0?

[immerrr]

Yes, it is.