Increase year of datetime array without using for loop

Question

I have a numpy array which has a long list of datetimes. I was wondering is there a way to add a year to all values of the array at once without using a for loop? eg. Using some numpy or datetime module?

>>> import datetime
>>> import numpy as np

>>> dts.shape
(64580,)

>>> dts[:5]
array([datetime.date(2000, 1, 15), datetime.date(2000, 1, 15),
       datetime.date(2000, 1, 15), datetime.date(2000, 1, 15),
       datetime.date(2000, 1, 15)], dtype=object)

>>> new_dts = somemodule.somefunctionforaddingyearorsomething(dts, year=1)
>>> new_dts
array([datetime.date(2001, 1, 15), datetime.date(2001, 1, 15),
       datetime.date(2001, 1, 15), datetime.date(2001, 1, 15),
       datetime.date(2001, 1, 15)], dtype=object)

Note: Day of each date is always set to day 15 as the dates represent monthly mean data.

I have implemented it using a for loop however this can be computationally slow..

The code for that is here:

def add_year_to_Datelist(dl):

dts = dl.dates.copy()
for idx, date in enumerate(dts):
    dts[idx] = date.replace(year=date.year + 1)

dl.set_dates(dts)
return dl

Cheers

jpp · Accepted Answer · 2019-01-04 02:44:25Z

3

NumPy doesn't support "1 year" as a timedelta, since this can be 365 or 366 days. If you are willing to use Pandas, you can use pd.to_datetime with pd.DateOffset, which does support year-based offsets:

import datetime, numpy as np, pandas as pd

dts = np.array([datetime.date(2000, 1, 15), datetime.date(2000, 1, 15),
                datetime.date(2000, 1, 15), datetime.date(2000, 1, 15),
                datetime.date(2000, 1, 15)], dtype=object)

dts = (pd.to_datetime(dts) + pd.DateOffset(years=1)).values.astype('datetime64[D]')

# array(['2001-01-15', '2001-01-15', '2001-01-15', '2001-01-15',
#        '2001-01-15'], dtype='datetime64[D]')

With NumPy, you can add a fixed number of days after converting your object dtype array to np.datetime64:

dts = dts.astype(np.datetime64) + np.timedelta64(365, 'D')

# array(['2001-01-14', '2001-01-14', '2001-01-14', '2001-01-14',
#        '2001-01-14'], dtype='datetime64[D]')

Note the 1-day mismatch between the two results.

edited Jan 4, 2019 at 2:44

answered Jan 4, 2019 at 1:51

jpp

166k37 gold badges301 silver badges362 bronze badges

Sign up to request clarification or add additional context in comments.

2 Comments

Max Collier Over a year ago

Your pandas answer works perfectly! Although I tried the first answer and noticed that for any leap year the data will be incorrect. i.e months Jan/Feb have day value of 14 whereas the other months have day values of 15..

jpp Over a year ago

@MaxCollier, Yes, that's exactly the problem with using 365 days fixed. There's no native NumPy solution as far as I'm aware.

GeeTransit · Accepted Answer · 2019-01-04 01:42:18Z

2

Making things faster while using a for loop is kind of hard. Using map however brings the calculations into C and can speed things up considerably.

Try this:

def add_year(dl, add=1):
    """Return new list with year increased by specified amount."""
    dts = list(map(lambda date: date.replace(year = date.year + add), dl.dates.copy()))
    return dts

Not sure how to get a numpy list because I haven't worked with numpy yet (I know, pathetic...) but you know so yeah :D

edited Jan 4, 2019 at 1:42

answered Jan 4, 2019 at 1:36

GeeTransit

1,46812 silver badges25 bronze badges

Collectives™ on Stack Overflow

Increase year of datetime array without using for loop

2 Answers 2

2 Comments

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

2 Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Related