How to convert numpy.timedelta64 to minutes

Question

I have a date time column in a Pandas DataFrame and I'd like to convert it to minutes or seconds.

For example: I want to convert 00:27:00 to 27 mins.

example = data['duration'][0]
example

result: numpy.timedelta64(1620000000000,'ns')

What's the best way to achieve this?

In the general case, do you want the fractional part, too? E.g. if the input timedelta is `00:27:45`, do you want 27 or 27.75? — Warren Weckesser, Oct 25 '14 at 19:22
I'm more interested in minutes as my data doesn't include seconds, so no I don't need the fraction. — Behzad Shahbaz, Oct 25 '14 at 19:29
I figured you can get to the minutes by executing this line: mins = np.array([data['duration']], dtype = "timedelta64[m]")[0]. How can I append the values I receive in this array to my original data frame? — Behzad Shahbaz, Oct 25 '14 at 19:30
see docs here: very easy in 0.15 http://pandas.pydata.org/pandas-docs/stable/timedeltas.html#frequency-conversion — Jeff, Oct 25 '14 at 20:10

score 46 · Accepted Answer · answered Oct 25 '14 at 19:36

46

Use array.astype() to convert the type of an array safely:

>>> import numpy as np
>>> a = np.timedelta64(1620000000000,'ns')
>>> a.astype('timedelta64[m]')
numpy.timedelta64(27,'m')

answered Oct 25 '14 at 19:36

sebix

1

Note that this will loose accuracy (based on the comments, this is ok for the questor, but should be highlighted for the general question): try `np.timedelta64(1621111111110,'ns').astype('timedelta64[m]')` – ntg Aug 26 '22 at 04:28

1 Answers1