[python] How to round the minute of a datetime object

I have a datetime object produced using strptime().

>>> tm
datetime.datetime(2010, 6, 10, 3, 56, 23)

What I need to do is round the minute to the closest 10th minute. What I have been doing up to this point was taking the minute value and using round() on it.

min = round(tm.minute, -1)

However, as with the above example, it gives an invalid time when the minute value is greater than 56. i.e.: 3:60

What is a better way to do this? Does datetime support this?

Those seem overly complex

def round_down_to():
    num = int(datetime.utcnow().replace(second=0, microsecond=0).minute)
    return num - (num%10)

From the best answer I modified to an adapted version using only datetime objects, this avoids having to do the conversion to seconds and makes the calling code more readable:

def roundTime(dt=None, dateDelta=datetime.timedelta(minutes=1)):
    """Round a datetime object to a multiple of a timedelta
    dt : datetime.datetime object, default now.
    dateDelta : timedelta object, we round to a multiple of this, default 1 minute.
    Author: Thierry Husson 2012 - Use it as you want but don't blame me.
            Stijn Nevens 2014 - Changed to use only datetime objects as variables
    """
    roundTo = dateDelta.total_seconds()

    if dt == None : dt = datetime.datetime.now()
    seconds = (dt - dt.min).seconds
    # // is a floor division, not a comment on following line:
    rounding = (seconds+roundTo/2) // roundTo * roundTo
    return dt + datetime.timedelta(0,rounding-seconds,-dt.microsecond)

Samples with 1 hour rounding & 15 minutes rounding:

print roundTime(datetime.datetime(2012,12,31,23,44,59),datetime.timedelta(hour=1))
2013-01-01 00:00:00

print roundTime(datetime.datetime(2012,12,31,23,44,49),datetime.timedelta(minutes=15))
2012-12-31 23:30:00

The shortest way I know

min = tm.minute // 10 * 10

yes, if your data belongs to a DateTime column in a pandas series, you can round it up using the built-in pandas.Series.dt.round function. See documentation here on pandas.Series.dt.round. In your case of rounding to 10min it will be Series.dt.round('10min') or Series.dt.round('600s') like so:

pandas.Series(tm).dt.round('10min')

Edit to add Example code:

import datetime
import pandas

tm = datetime.datetime(2010, 6, 10, 3, 56, 23)
tm_rounded = pandas.Series(tm).dt.round('10min')
print(tm_rounded)

>>> 0   2010-06-10 04:00:00
dtype: datetime64[ns]

A two line intuitive solution to round to a given time unit, here seconds, for a datetime object t:

format_str = '%Y-%m-%d %H:%M:%S'
t_rounded = datetime.strptime(datetime.strftime(t, format_str), format_str)

If you wish to round to a different unit simply alter format_str.

This approach does not round to arbitrary time amounts as above methods, but is a nicely Pythonic way to round to a given hour, minute or second.

Other solution:

def round_time(timestamp=None, lapse=0):
    """
    Round a timestamp to a lapse according to specified minutes

    Usage:

    >>> import datetime, math
    >>> round_time(datetime.datetime(2010, 6, 10, 3, 56, 23), 0)
    datetime.datetime(2010, 6, 10, 3, 56)
    >>> round_time(datetime.datetime(2010, 6, 10, 3, 56, 23), 1)
    datetime.datetime(2010, 6, 10, 3, 57)
    >>> round_time(datetime.datetime(2010, 6, 10, 3, 56, 23), -1)
    datetime.datetime(2010, 6, 10, 3, 55)
    >>> round_time(datetime.datetime(2019, 3, 11, 9, 22, 11), 3)
    datetime.datetime(2019, 3, 11, 9, 24)
    >>> round_time(datetime.datetime(2019, 3, 11, 9, 22, 11), 3*60)
    datetime.datetime(2019, 3, 11, 12, 0)
    >>> round_time(datetime.datetime(2019, 3, 11, 10, 0, 0), 3)
    datetime.datetime(2019, 3, 11, 10, 0)

    :param timestamp: Timestamp to round (default: now)
    :param lapse: Lapse to round in minutes (default: 0)
    """
    t = timestamp or datetime.datetime.now()  # type: Union[datetime, Any]
    surplus = datetime.timedelta(seconds=t.second, microseconds=t.microsecond)
    t -= surplus
    try:
        mod = t.minute % lapse
    except ZeroDivisionError:
        return t
    if mod:  # minutes % lapse != 0
        t += datetime.timedelta(minutes=math.ceil(t.minute / lapse) * lapse - t.minute)
    elif surplus != datetime.timedelta() or lapse < 0:
        t += datetime.timedelta(minutes=(t.minute / lapse + 1) * lapse - t.minute)
    return t

Hope this helps!

Pandas has a datetime round feature, but as with most things in Pandas it needs to be in Series format.

>>> ts = pd.Series(pd.date_range(Dt(2019,1,1,1,1),Dt(2019,1,1,1,4),periods=8))
>>> print(ts)
0   2019-01-01 01:01:00.000000000
1   2019-01-01 01:01:25.714285714
2   2019-01-01 01:01:51.428571428
3   2019-01-01 01:02:17.142857142
4   2019-01-01 01:02:42.857142857
5   2019-01-01 01:03:08.571428571
6   2019-01-01 01:03:34.285714285
7   2019-01-01 01:04:00.000000000
dtype: datetime64[ns]

>>> ts.dt.round('1min')
0   2019-01-01 01:01:00
1   2019-01-01 01:01:00
2   2019-01-01 01:02:00
3   2019-01-01 01:02:00
4   2019-01-01 01:03:00
5   2019-01-01 01:03:00
6   2019-01-01 01:04:00
7   2019-01-01 01:04:00
dtype: datetime64[ns]

Docs - Change the frequency string as needed.

Based on Stijn Nevens and modified for Django use to round current time to the nearest 15 minute.

from datetime import date, timedelta, datetime, time

    def roundTime(dt=None, dateDelta=timedelta(minutes=1)):

        roundTo = dateDelta.total_seconds()

        if dt == None : dt = datetime.now()
        seconds = (dt - dt.min).seconds
        # // is a floor division, not a comment on following line:
        rounding = (seconds+roundTo/2) // roundTo * roundTo
        return dt + timedelta(0,rounding-seconds,-dt.microsecond)

    dt = roundTime(datetime.now(),timedelta(minutes=15)).strftime('%H:%M:%S')

 dt = 11:45:00

if you need full date and time just remove the .strftime('%H:%M:%S')

Here is a simpler generalized solution without floating point precision issues and external library dependencies:

import datetime

def time_mod(time, delta, epoch=None):
    if epoch is None:
        epoch = datetime.datetime(1970, 1, 1, tzinfo=time.tzinfo)
    return (time - epoch) % delta

def time_round(time, delta, epoch=None):
    mod = time_mod(time, delta, epoch)
    if mod < (delta / 2):
       return time - mod
    return time + (delta - mod)

In your case:

>>> tm = datetime.datetime(2010, 6, 10, 3, 56, 23)
>>> time_round(tm, datetime.timedelta(minutes=10))
datetime.datetime(2010, 6, 10, 4, 0)

def get_rounded_datetime(self, dt, freq, nearest_type='inf'):

    if freq.lower() == '1h':
        round_to = 3600
    elif freq.lower() == '3h':
        round_to = 3 * 3600
    elif freq.lower() == '6h':
        round_to = 6 * 3600
    else:
        raise NotImplementedError("Freq %s is not handled yet" % freq)

    # // is a floor division, not a comment on following line:
    seconds_from_midnight = dt.hour * 3600 + dt.minute * 60 + dt.second
    if nearest_type == 'inf':
        rounded_sec = int(seconds_from_midnight / round_to) * round_to
    elif nearest_type == 'sup':
        rounded_sec = (int(seconds_from_midnight / round_to) + 1) * round_to
    else:
        raise IllegalArgumentException("nearest_type should be  'inf' or 'sup'")

    dt_midnight = datetime.datetime(dt.year, dt.month, dt.day)

    return dt_midnight + datetime.timedelta(0, rounded_sec)

General function to round a datetime at any time lapse in seconds:

def roundTime(dt=None, roundTo=60):
   """Round a datetime object to any time lapse in seconds
   dt : datetime.datetime object, default now.
   roundTo : Closest number of seconds to round to, default 1 minute.
   Author: Thierry Husson 2012 - Use it as you want but don't blame me.
   """
   if dt == None : dt = datetime.datetime.now()
   seconds = (dt.replace(tzinfo=None) - dt.min).seconds
   rounding = (seconds+roundTo/2) // roundTo * roundTo
   return dt + datetime.timedelta(0,rounding-seconds,-dt.microsecond)

Samples with 1 hour rounding & 30 minutes rounding:

print roundTime(datetime.datetime(2012,12,31,23,44,59,1234),roundTo=60*60)
2013-01-01 00:00:00

print roundTime(datetime.datetime(2012,12,31,23,44,59,1234),roundTo=30*60)
2012-12-31 23:30:00

This will get the 'floor' of a datetime object stored in tm rounded to the 10 minute mark before tm.

tm = tm - datetime.timedelta(minutes=tm.minute % 10,
                             seconds=tm.second,
                             microseconds=tm.microsecond)

If you want classic rounding to the nearest 10 minute mark, do this:

discard = datetime.timedelta(minutes=tm.minute % 10,
                             seconds=tm.second,
                             microseconds=tm.microsecond)
tm -= discard
if discard >= datetime.timedelta(minutes=5):
    tm += datetime.timedelta(minutes=10)

or this:

tm += datetime.timedelta(minutes=5)
tm -= datetime.timedelta(minutes=tm.minute % 10,
                         seconds=tm.second,
                         microseconds=tm.microsecond)

Not the best for speed when the exception is caught, however this would work.

def _minute10(dt=datetime.utcnow()):
    try:
        return dt.replace(minute=round(dt.minute, -1))
    except ValueError:
        return dt.replace(minute=0) + timedelta(hours=1)

Timings

%timeit _minute10(datetime(2016, 12, 31, 23, 55))
100000 loops, best of 3: 5.12 µs per loop

%timeit _minute10(datetime(2016, 12, 31, 23, 31))
100000 loops, best of 3: 2.21 µs per loop

I used Stijn Nevens code (thank you Stijn) and have a little add-on to share. Rounding up, down and rounding to nearest.

update 2019-03-09 = comment Spinxz incorporated; thank you.

update 2019-12-27 = comment Bart incorporated; thank you.

Tested for date_delta of "X hours" or "X minutes" or "X seconds".

import datetime

def round_time(dt=None, date_delta=datetime.timedelta(minutes=1), to='average'):
    """
    Round a datetime object to a multiple of a timedelta
    dt : datetime.datetime object, default now.
    dateDelta : timedelta object, we round to a multiple of this, default 1 minute.
    from:  http://stackoverflow.com/questions/3463930/how-to-round-the-minute-of-a-datetime-object-python
    """
    round_to = date_delta.total_seconds()
    if dt is None:
        dt = datetime.now()
    seconds = (dt - dt.min).seconds

    if seconds % round_to == 0 and dt.microsecond == 0:
        rounding = (seconds + round_to / 2) // round_to * round_to
    else:
        if to == 'up':
            # // is a floor division, not a comment on following line (like in javascript):
            rounding = (seconds + dt.microsecond/1000000 + round_to) // round_to * round_to
        elif to == 'down':
            rounding = seconds // round_to * round_to
        else:
            rounding = (seconds + round_to / 2) // round_to * round_to

    return dt + datetime.timedelta(0, rounding - seconds, - dt.microsecond)

# test data
print(round_time(datetime.datetime(2019,11,1,14,39,00), date_delta=datetime.timedelta(seconds=30), to='up'))
print(round_time(datetime.datetime(2019,11,2,14,39,00,1), date_delta=datetime.timedelta(seconds=30), to='up'))
print(round_time(datetime.datetime(2019,11,3,14,39,00,776980), date_delta=datetime.timedelta(seconds=30), to='up'))
print(round_time(datetime.datetime(2019,11,4,14,39,29,776980), date_delta=datetime.timedelta(seconds=30), to='up'))
print(round_time(datetime.datetime(2018,11,5,14,39,00,776980), date_delta=datetime.timedelta(seconds=30), to='down'))
print(round_time(datetime.datetime(2018,11,6,14,38,59,776980), date_delta=datetime.timedelta(seconds=30), to='down'))
print(round_time(datetime.datetime(2017,11,7,14,39,15), date_delta=datetime.timedelta(seconds=30), to='average'))
print(round_time(datetime.datetime(2017,11,8,14,39,14,999999), date_delta=datetime.timedelta(seconds=30), to='average'))
print(round_time(datetime.datetime(2019,11,9,14,39,14,999999), date_delta=datetime.timedelta(seconds=30), to='up'))
print(round_time(datetime.datetime(2012,12,10,23,44,59,7769),to='average'))
print(round_time(datetime.datetime(2012,12,11,23,44,59,7769),to='up'))
print(round_time(datetime.datetime(2010,12,12,23,44,59,7769),to='down',date_delta=datetime.timedelta(seconds=1)))
print(round_time(datetime.datetime(2011,12,13,23,44,59,7769),to='up',date_delta=datetime.timedelta(seconds=1)))
print(round_time(datetime.datetime(2012,12,14,23,44,59),date_delta=datetime.timedelta(hours=1),to='down'))
print(round_time(datetime.datetime(2012,12,15,23,44,59),date_delta=datetime.timedelta(hours=1),to='up'))
print(round_time(datetime.datetime(2012,12,16,23,44,59),date_delta=datetime.timedelta(hours=1)))
print(round_time(datetime.datetime(2012,12,17,23,00,00),date_delta=datetime.timedelta(hours=1),to='down'))
print(round_time(datetime.datetime(2012,12,18,23,00,00),date_delta=datetime.timedelta(hours=1),to='up'))
print(round_time(datetime.datetime(2012,12,19,23,00,00),date_delta=datetime.timedelta(hours=1)))

i'm using this. it has the advantage of working with tz aware datetimes.

def round_minutes(some_datetime: datetime, step: int):
    """ round up to nearest step-minutes """
    if step > 60:
        raise AttrbuteError("step must be less than 60")

    change = timedelta(
        minutes= some_datetime.minute % step,
        seconds=some_datetime.second,
        microseconds=some_datetime.microsecond
    )

    if change > timedelta():
        change -= timedelta(minutes=step)

    return some_datetime - change

it has the disadvantage of only working for timeslices less than an hour.

A straightforward approach:

def round_time(dt, round_to_seconds=60):
    """Round a datetime object to any number of seconds
    dt: datetime.datetime object
    round_to_seconds: closest number of seconds for rounding, Default 1 minute.
    """
    rounded_epoch = round(dt.timestamp() / round_to_seconds) * round_to_seconds
    rounded_dt = datetime.datetime.fromtimestamp(rounded_epoch).astimezone(dt.tzinfo)
    return rounded_dt

if you don't want to use condition, you can use modulo operator:

minutes = int(round(tm.minute, -1)) % 60

UPDATE

did you want something like this?

def timeround10(dt):
    a, b = divmod(round(dt.minute, -1), 60)
    return '%i:%02i' % ((dt.hour + a) % 24, b)

timeround10(datetime.datetime(2010, 1, 1, 0, 56, 0)) # 0:56
# -> 1:00

timeround10(datetime.datetime(2010, 1, 1, 23, 56, 0)) # 23:56
# -> 0:00

.. if you want result as string. for obtaining datetime result, it's better to use timedelta - see other responses ;)