从numpy timedelta64获得秒数

Question

I have a datetime index in pandas

index = np.array(['2013-11-11T12:36:00.078757888-0800',
                  '2013-11-11T12:36:03.692692992-0800',
                  '2013-11-11T12:36:07.085489920-0800',
                  '2013-11-11T12:36:08.957488128-0800'], dtype='datetime64[ns]')

I want to calculate the time difference in seconds. The way I came up with is:

diff(index).astype('float64')/1e9

is there a better/cleaner way?

Answer 1

Your own answer is correct and good. Slightly different way is to specify scale constants with timedelta expression.

For example, to scale to seconds:

>>> np.diff(index)/np.timedelta64(1, 's')
array([ 3.6139351 ,  3.39279693,  1.87199821])

To minutes:

>>> np.diff(index)/np.timedelta64(1, 'm')
array([ 0.06023225,  0.05654662,  0.03119997])