[FIXED] How to properly add hours to a pandas.tseries.index.DatetimeIndex?

Issue

I have a normal df.index that I would like to add some hours to it.

In [1]: test[1].index
Out[2]: 
<class 'pandas.tseries.index.DatetimeIndex'>
[2010-03-11, ..., 2014-08-14]
Length: 52, Freq: None, Timezone: None

This is how the first element looks like:

In [1]: test[1].index[0]
Out[2]: Timestamp('2010-03-11 00:00:00')

So I try this to add the hours:

In [1]: test[1].index[0] + pd.tseries.timedeltas.to_timedelta(16, unit='h')

However I get this:

Out[2]: Timestamp('2010-03-11 00:00:00.000000016')

But I would like to get this:

Out[2]: Timestamp('2010-03-11 16:00:00')

What I am missing?. The enviroment is Anaconda (latest) Python 2.7.7, iPython 2.2

Thanks a lot

Solution

You can use pd.DateOffset:

test[1].index + pd.DateOffset(hours=16)

pd.DateOffset accepts the same keyword arguments as dateutil.relativedelta.

---

The problem you encountered was due to this bug which has been fixed in Pandas version 0.14.1:

In [242]: pd.to_timedelta(16, unit='h')
Out[242]: numpy.timedelta64(16,'ns')

If you upgrade, your original code should work.

Answered By - unutbu