BUG: reindex on level of multi-index does not respect order of new index ?

[jorisvandenbossche]

When having a MultiIndex, and reindexing on a level of it, reindex does not respect the order of the new index. Is this a bug? Or a limitation of the multi-index reindexing (as in one level you don't have unique elements?) which should be documented?

In [69]: s = pd.Series([0,1,2,3,4,5,6,7], index=pd.MultiIndex.from_product([[1,2,3,4],[7,8]]))

In [70]: s
Out[70]:
1  7    0
   8    1
2  7    2
   8    3
3  7    4
   8    5
4  7    6
   8    7
dtype: int64

In [71]: s.reindex([3,2], level=0)
Out[71]:
2  7    2
   8    3
3  7    4
   8    5
dtype: int64

Als with a dataframe multi-indexed columns:

In [72]: df = pd.DataFrame(np.random.randn(5,6), columns=pd.MultiIndex.from_prod
uct([list('abc'), list('ef')]))

In [73]: df
Out[73]:
          a                   b                   c
          e         f         e         f         e         f
0 -0.968421  0.343670  0.353754  0.211625  1.964620  0.696301
1 -1.082284  0.532646  0.099657 -1.355934 -0.748171 -0.766453
2  0.674002 -0.440444 -0.296365  0.225170  0.592352 -0.691878
3  0.880808  0.470677  0.330382  0.937828 -0.225192 -1.059934
4 -0.616588  0.501665 -0.232722  1.300403  0.235369  2.318810

In [74]: df.reindex(columns=['c', 'b'], level=0)
Out[74]:
          b                   c
          e         f         e         f
0  0.353754  0.211625  1.964620  0.696301
1  0.099657 -1.355934 -0.748171 -0.766453
2 -0.296365  0.225170  0.592352 -0.691878
3  0.330382  0.937828 -0.225192 -1.059934
4 -0.232722  1.300403  0.235369  2.318810

[jreback]

this is a dupe of #4088 (currently an open issue)