Trying to create an array from a startdate to an enddate with a slot for every second. Pretty much like numpy.linspace(startdate,enddate,number_of_seconds_inbetween). Whats the easiest/fastest way to do this?
Add a comment
|
2 Answers
np.arange does it more or less out of the box:
np.arange("2000-01-01","2000-01-02",dtype="M8[s]")
# array(['2000-01-01T00:00:00', '2000-01-01T00:00:01',
# '2000-01-01T00:00:02', ..., '2000-01-01T23:59:57',
# '2000-01-01T23:59:58', '2000-01-01T23:59:59'],
# dtype='datetime64[s]')
5 Comments
yatu
Ah nice... didn't think this was possible directly using
np.arange
Mateen Ulhaq
Now the question is why there isn't a
np.linspace implementation for this. XD
hpaulj
linspace produces evenly spaced floats. The datetime steps are discrete - days, minutes etc
Paul Panzer
@hpaulj
linspace can output "exact" dtypes like int, so in principle it should be possible to make datetime64 work in the same way.hpaulj
linspace is Python code. The current linspace (1.16) works by creating a arange(N) and scaling with the end points (which may be arrays). Math with datetime64 is limited to subtraction. But np.array("2000-01-01",'M8[D]') + np.linspace(0, 10, 11).astype("timedelta64[D]") does work.def linspace_datetime64(start_date, end_date, n):
return np.linspace(0, 1, n) * (end_date - start_date) + start_date
Example usage:
>>> start_date = np.datetime64('2008-01-01')
>>> end_date = np.datetime64('2009-01-01')
>>> linspace_datetime64(start_date, end_date, 12)
array(['2008-01-01', '2008-02-03', '2008-03-07', '2008-04-09',
'2008-05-13', '2008-06-15', '2008-07-18', '2008-08-20',
'2008-09-23', '2008-10-26', '2008-11-28', '2009-01-01'],
dtype='datetime64[D]')
1 Comment
ba_ul
The answer is not complete. It doesn't show how to get every second between any two dates.