DateOffset¶

pd.DateOffset¶

day resets the actual day, days adds to the original day

print(pd.Timestamp("2015-01-01") - pd.DateOffset(day=3))  # 2015-01-03
print(pd.Timestamp("2015-01-01") + pd.DateOffset(day=3))  # 2015-01-03
print(pd.Timestamp("2015-01-01") + pd.DateOffset(days=3)) # 2015-01-04

period¶

Only honor the integer periods

t = pd.Timestamp('2022-08-01 12:25:56')
p = t.to_period('30T')
print(f'from {p.start_time} to {p.end_time}')
from 2022-08-01 12:25:00 to 2022-08-01 12:54:59.999999999

normalize¶

>>> pd.date_range('2022-08-01 23:59:59', '2022-09-16', freq=pd.tseries.frequencies.to_offset('MS'))
DatetimeIndex(['2022-08-01 23:59:59'], dtype='datetime64[ns]', freq='MS')

>>> pd.date_range('2022-08-01 23:59:59', '2022-09-16', freq=pd.tseries.frequencies.to_offset('MS'), normalize=True)
DatetimeIndex(['2022-08-01', '2022-09-01'], dtype='datetime64[ns]', freq='MS')

>>> pd.date_range(pd.Timestamp('2022-08-01 23:59:59') + pd.DateOffset(days=0, normalize=True), '2022-09-16', freq=pd.tseries.frequencies.to_offset('MS'))
DatetimeIndex(['2022-08-01', '2022-09-01'], dtype='datetime64[ns]', freq='MS')

normalize vs floor('D')¶

Really???

pd.Timestamp.normalize() sets the time component to midnight (00:00:00) in the local time zone, but does not change the time zone itself.
pd.Timestamp.floor('D') sets the time component to midnight (00:00:00) in the UTC time zone, and then converts the timestamp to the local time zone if necessary.

pd.DateOffset vs pd.Timedleta vs pd.tseries.offsets.Day¶

DateOffset respect calendar arithmetic while Timedelta respects absolute time arithmetic, might be different for datetime with timezone info
pd.DateOffset(days=1) increments a date by 1 calendar day (23, 24, or 25 hours depending on day light savings time)
pd.Timedelta(days=1) increments a date by 24 hours.

import pandas as pd
date = pd.Timestamp('2016-10-30 00:00:00+0300', tz='Europe/Helsinki')
date_plus_dateoffset = date + pd.DateOffset(days=1)
date_plus_timedelta = date + pd.Timedelta(days=1)
date_plus_offsetsday = date + pd.tseries.offsets.Day()

print(date_plus_dateoffset) #2016-10-31 00:00:00+02:00
print(date_plus_timedelta)  #2016-10-30 23:00:00+02:00
print(date_plus_offsetsday) #2016-10-30 23:00:00+02:00