Series with MultiIndex with PeriodIndex can't do unstack()

[ruoyu0088]

Here is the code:

import pandas as pd

idx1 = pd.PeriodIndex(["2013-01", "2013-01", "2013-02", "2013-02"], freq="M")
idx2 = ["a","b","a","b"]
value = [1,2,3,4]

idx = pd.MultiIndex.from_arrays([idx1, idx2])
s = pd.Series(value, index=idx)

s.unstack()

It will raise following exception:

anaconda/lib/python2.7/site-packages/pandas-0.12.0.dev_424e187-py2.7-linux-x86_64.egg/pandas/tseries/period.pyc in _from_arraylike(cls, data, freq, tz)
    649 
    650                 if freq is None:
--> 651                     raise ValueError(('freq not specified and cannot be '
    652                                       'inferred from first element'))
    653 

ValueError: freq not specified and cannot be inferred from first element

[jreback]

marking as a bug

[kevinastone]

Any way to work around this issue?

[McRip2]

What about converting the Period-index back to a timestamp-index before unstacking? Afterwards we can go back to a Period-index.

s.index = pd.MultiIndex.from_tuples([(x[0].to_timestamp(),x[1]) for x in s.index])
s = s.unstack()
s.index = pd.PeriodIndex(list(s.index),freq='M')

[hayd]

This came up again in http://stackoverflow.com/questions/21352520/pandas-unstack-doesnt-work