Unexpected result from rolling_median when center = True

[jonblunt]

s = pd.Series( range(10), index = pd.PeriodIndex(start = '2013-10-1', end = '2013-10-10', freq = 'D'))
pd.rolling_median(s,5,  min_periods = 1)

2013-10-01    0.0
2013-10-02    0.5
2013-10-03    1.0
2013-10-04    1.5
2013-10-05    2.0
2013-10-06    3.0
2013-10-07    4.0
2013-10-08    5.0
2013-10-09    6.0
2013-10-10    7.0
Freq: D, dtype: float64

When center is specified we get Nans in the last window//2 positions

s = pd.Series( range(10), index = pd.PeriodIndex(start = '2013-10-1', end = '2013-10-10', freq = 'D'))
pd.rolling_median(s,5,  min_periods = 1, center = True)

2013-10-01    1.0
2013-10-02    1.5
2013-10-03    2.0
2013-10-04    3.0
2013-10-05    4.0
2013-10-06    5.0
2013-10-07    6.0
2013-10-08    7.0
2013-10-09    NaN
2013-10-10    NaN
Freq: D, dtype: float64

I was expecting the last two values to be 7.5 and 8.

I got the same results, nans at the end with rolling_mean as well

This is with pandas 12.0 and windows 64.

[jonblunt]

found issue #3136 .Closing