Resampling with NaT in PeriodIndex raises ValueError · Issue #13224 · pandas-dev/pandas (original) (raw)

PeriodIndex can't handle NaT:

In [2]: pi = pd.PeriodIndex(['1970-01-01 00:00:00', pd.NaT, '1970-01-01 00:00:02'], freq='S')

In [3]: pd.DataFrame([2, 3, 5], pi).resample('1s').mean()
...
ValueError: Start and end must not be NaT

But DatetimeIndex obtained from PeriodIndex can:

In [4]: pd.DataFrame([2, 3, 5], pi.to_datetime()).resample('1s').mean()
Out[4]: 
                       0
1970-01-01 00:00:00  2.0
1970-01-01 00:00:01  NaN
1970-01-01 00:00:02  5.0

Shouldn't these behave the same?

Versions:

python: 2.7.11.final.0
OS: Linux
machine: x86_64
pandas: 0.18.1+38.gc7e6f5f.dirty
numpy: 1.10.2