BUG: pd.to_timedelta raises instead of returning NaT

[danielballan]

In [1]: Series(['2011-01-01', np.nan]).apply(pd.to_datetime)
Out[1]: 
0   2011-01-01 00:00:00
1                   NaT
dtype: datetime64[ns]

In [2]: Series(['00:00:01', np.nan]).apply(pd.to_timedelta)
AssertionError: Invalid type for timedelta scalar: <type 'float'>

May not be a trivial fix. So far I have seen that inserting np.nan and pd.tslib.iNaT into the result turns the dtype to object, which is no good.

[danielballan]

That said, if this can be squeezed into 0.13, I'm willing to rush to get it done.

[jreback]

hmm though has done this oh well

it's easy just change this routine

coerce_scalar_to_timedelta_type(

in tseries/time delta

just need a isnull check and return tslib.iNaT
(it will work now if u pass NaT)