Python, Pandas, Numpy: Date_range: passing a np.timedelta as freq. argument

Question

I have a time delta that is in format numpy.timedelta(64): value = numpy.timedelta64(30,'m')
(here 30min).

Is there a way to pass that value to the construction of a pandas date_range: pd.date_range(start_time, end_time, freq=value) ?

If not how is it possible to convert that value in order to pass it in the date_range? I tried value.astype('str') but it returns '0:30:00' that is also not suited as an argument for the pd.date_range.

Answer 1

You can convert it to a Timedelta and pass it into freq.

>>> pd.date_range('20160101', '20170101', freq=pd.Timedelta(value)) 
DatetimeIndex(['2016-01-01 00:00:00', '2016-01-01 00:30:00',
               '2016-01-01 01:00:00', '2016-01-01 01:30:00',
               '2016-01-01 02:00:00', '2016-01-01 02:30:00',
               '2016-01-01 03:00:00', '2016-01-01 03:30:00',
               '2016-01-01 04:00:00', '2016-01-01 04:30:00',
               ...
               '2016-12-31 19:30:00', '2016-12-31 20:00:00',
               '2016-12-31 20:30:00', '2016-12-31 21:00:00',
               '2016-12-31 21:30:00', '2016-12-31 22:00:00',
               '2016-12-31 22:30:00', '2016-12-31 23:00:00',
               '2016-12-31 23:30:00', '2017-01-01 00:00:00'],
              dtype='datetime64[ns]', length=17569, freq='30T')

I think older versions of pandas didn't support this, but could still use a frequency string:

 pd.date_range('20160101', '20170101', freq='30min')

Answer 2

Since you are starting with a numpy.timedelta64, I wondered how np.arange handled this.

With datetime64 start and stop values arange handles this nicely:

In [847]: x=np.arange(np.datetime64('2016-01-01'), np.datetime64('2017-01-01'),value)
In [848]: x
Out[848]: 
array(['2016-01-01T00:00', '2016-01-01T00:30', '2016-01-01T01:00', ...,
       '2016-12-31T22:30', '2016-12-31T23:00', '2016-12-31T23:30'], dtype='datetime64[m]')

and DatetimeIndex accepts such an array:

In [849]: pd.DatetimeIndex(x)
Out[849]: 
DatetimeIndex(['2016-01-01 00:00:00', '2016-01-01 00:30:00',
               '2016-01-01 01:00:00', '2016-01-01 01:30:00',
               '2016-01-01 02:00:00', '2016-01-01 02:30:00',
               '2016-01-01 03:00:00', '2016-01-01 03:30:00',
               '2016-01-01 04:00:00', '2016-01-01 04:30:00',
               ...
               '2016-12-31 19:00:00', '2016-12-31 19:30:00',
               '2016-12-31 20:00:00', '2016-12-31 20:30:00',
               '2016-12-31 21:00:00', '2016-12-31 21:30:00',
               '2016-12-31 22:00:00', '2016-12-31 22:30:00',
               '2016-12-31 23:00:00', '2016-12-31 23:30:00'],
              dtype='datetime64[ns]', length=17568, freq=None)