Bug related to multilevel index series ?

[halleygithub]

I am upgrading Pandas from 0.8.1 to 0.10.1.dev-f7f7e13 . My environment is Window XP with below: Python: 2.7.3 Numpy: 1.6.2 MPL: 1.1.1 Pandas: 0.10.1.dev-f7f7e13.

Then OK application on 0.8.1 now meets errors. I trace the root cause to filtering the duplicated index of Series. Detail in : http://stackoverflow.com/questions/14395678/how-to-drop-extra-copy-of-duplicate-index-of-pandas-series

simply put: below snippet has two issues :

import pandas as pd
idx_tp = [('600809', '20061231'), ('600809', '20070331'), ('600809', '20070630'), ('600809', '20070331')]
dt = ['demo','demo','demo','demo']
idx = pd.MultiIndex.from_tuples(idx_tp,names = ['STK_ID','RPT_Date'])
s = pd.Series(dt,index=idx)

Issue 1: s[s.index.unique()] works well on 0.8.1 but not 0.10.1

Issue 2: s.groupby(s.index).first() will crash on my machine

[wesm]

thanks will have a look

[wesm]

Fixed the second issue. I'm surprised the first ever worked, going to have a look at 0.8.1

[wesm]

The first is not a supported API and only worked by accident before. Please do something like:

In [16]: s[-Series(s.index.values, s.index).duplicated()]
Out[16]: 
STK_ID  RPT_Date
600809  20061231    demo
        20070331    demo
        20070630    demo

I need to add a top level function duplicated.

[halleygithub]

Thanks. but "s[s.index.unique()]" looks elegant than "s[-Series(s.index.values, s.index).duplicated()]" .