fancy indexing with dupe columns yields unexpected ordering

[y-p]

In [16]: from pandas.util.testing import makeCustomDataframe as mkdf
    ...: df= mkdf(10, 3)
    ...: df.columns = ['a','a','b']
    ...: cols = ['b','a']
    ...: df[['b','a']].columns
Out[16]: Index([a, a, b], dtype=object)

expected to see

['b','a','a']

Once fixed, can relax format.py:CSVFormatter raising NotImplementedError()
on cols selection with dupe columns, for the new engine only. see #3457 .

edit: #3509 removed that limitation from CSVFormatter

[jreback]

I see where this is broken....

[jreback]

looks like it always returns in the index in the original order......

[jreback]

closed by #3459