在Python中遍历一系列日期

I have the following code to do this, but how can I do it better? Right now I think it’s better than nested loops, but it starts to get Perl-one-linerish when you have a generator in a list comprehension.

day_count = (end_date - start_date).days + 1
for single_date in [d for d in (start_date + timedelta(n) for n in range(day_count)) if d <= end_date]:
    print strftime("%Y-%m-%d", single_date.timetuple())

Notes

• I’m not actually using this to print. That’s just for demo purposes.
• The start_date and end_date variables are datetime.date objects because I don’t need the timestamps. (They’re going to be used to generate a report).

Sample Output

For a start date of 2009-05-30 and an end date of 2009-06-09:

2009-05-30
2009-05-31
2009-06-01
2009-06-02
2009-06-03
2009-06-04
2009-06-05
2009-06-06
2009-06-07
2009-06-08
2009-06-09

Why are there two nested iterations? For me it produces the same list of data with only one iteration:

for single_date in (start_date + timedelta(n) for n in range(day_count)):
    print ...

And no list gets stored, only one generator is iterated over. Also the “if” in the generator seems to be unnecessary.

After all, a linear sequence should only require one iterator, not two.

Update after discussion with John Machin:

Maybe the most elegant solution is using a generator function to completely hide/abstract the iteration over the range of dates:

from datetime import timedelta, date

def daterange(start_date, end_date):
    for n in range(int ((end_date - start_date).days)):
        yield start_date + timedelta(n)

start_date = date(2013, 1, 1)
end_date = date(2015, 6, 2)
for single_date in daterange(start_date, end_date):
    print(single_date.strftime("%Y-%m-%d"))

NB: For consistency with the built-in range() function this iteration stops before reaching the end_date. So for inclusive iteration use the next day, as you would with range().

This might be more clear:

from datetime import date, timedelta

start_date = date(2019, 1, 1)
end_date = date(2020, 1, 1)
delta = timedelta(days=1)
while start_date <= end_date:
    print (start_date.strftime("%Y-%m-%d"))
    start_date += delta

Use the dateutil library:

from datetime import date
from dateutil.rrule import rrule, DAILY

a = date(2009, 5, 30)
b = date(2009, 6, 9)

for dt in rrule(DAILY, dtstart=a, until=b):
    print dt.strftime("%Y-%m-%d")

This python library has many more advanced features, some very useful, like relative deltas—and is implemented as a single file (module) that’s easily included into a project.

Pandas is great for time series in general, and has direct support for date ranges.

import pandas as pd
daterange = pd.date_range(start_date, end_date)

You can then loop over the daterange to print the date:

for single_date in daterange:
    print (single_date.strftime("%Y-%m-%d"))

It also has lots of options to make life easier. For example if you only wanted weekdays, you would just swap in bdate_range. See http://pandas.pydata.org/pandas-docs/stable/timeseries.html#generating-ranges-of-timestamps

The power of Pandas is really its dataframes, which support vectorized operations (much like numpy) that make operations across large quantities of data very fast and easy.

EDIT: You could also completely skip the for loop and just print it directly, which is easier and more efficient:

print(daterange)

import datetime

def daterange(start, stop, step=datetime.timedelta(days=1), inclusive=False):
  # inclusive=False to behave like range by default
  if step.days > 0:
    while start < stop:
      yield start
      start = start + step
      # not +=! don't modify object passed in if it's mutable
      # since this function is not restricted to
      # only types from datetime module
  elif step.days < 0:
    while start > stop:
      yield start
      start = start + step
  if inclusive and start == stop:
    yield start

# ...

for date in daterange(start_date, end_date, inclusive=True):
  print strftime("%Y-%m-%d", date.timetuple())

This function does more than you strictly require, by supporting negative step, etc. As long as you factor out your range logic, then you don’t need the separate day_count and most importantly the code becomes easier to read as you call the function from multiple places.

This is the most human-readable solution I can think of.

import datetime

def daterange(start, end, step=datetime.timedelta(1)):
    curr = start
    while curr < end:
        yield curr
        curr += step

Why not try:

import datetime as dt

start_date = dt.datetime(2012, 12,1)
end_date = dt.datetime(2012, 12,5)

total_days = (end_date - start_date).days + 1 #inclusive 5 days

for day_number in range(total_days):
    current_date = (start_date + dt.timedelta(days = day_number)).date()
    print current_date

Numpy’s arange function can be applied to dates:

import numpy as np
from datetime import datetime, timedelta
d0 = datetime(2009, 1,1)
d1 = datetime(2010, 1,1)
dt = timedelta(days = 1)
dates = np.arange(d0, d1, dt).astype(datetime)

The use of astype is to convert from numpy.datetime64 to an array of datetime.datetime objects.

Show the last n days from today:

import datetime
for i in range(0, 100):
    print((datetime.date.today() + datetime.timedelta(i)).isoformat())

Output:

2016-06-29
2016-06-30
2016-07-01
2016-07-02
2016-07-03
2016-07-04

import datetime

def daterange(start, stop, step_days=1):
    current = start
    step = datetime.timedelta(step_days)
    if step_days > 0:
        while current < stop:
            yield current
            current += step
    elif step_days < 0:
        while current > stop:
            yield current
            current += step
    else:
        raise ValueError("daterange() step_days argument must not be zero")

if __name__ == "__main__":
    from pprint import pprint as pp
    lo = datetime.date(2008, 12, 27)
    hi = datetime.date(2009, 1, 5)
    pp(list(daterange(lo, hi)))
    pp(list(daterange(hi, lo, -1)))
    pp(list(daterange(lo, hi, 7)))
    pp(list(daterange(hi, lo, -7))) 
    assert not list(daterange(lo, hi, -1))
    assert not list(daterange(hi, lo))
    assert not list(daterange(lo, hi, -7))
    assert not list(daterange(hi, lo, 7)) 

for i in range(16):
    print datetime.date.today() + datetime.timedelta(days=i)

For completeness, Pandas also has a period_range function for timestamps that are out of bounds:

import pandas as pd

pd.period_range(start='1/1/1626', end='1/08/1627', freq='D')

I have a similar problem, but I need to iterate monthly instead of daily.

This is my solution

import calendar
from datetime import datetime, timedelta

def days_in_month(dt):
    return calendar.monthrange(dt.year, dt.month)[1]

def monthly_range(dt_start, dt_end):
    forward = dt_end >= dt_start
    finish = False
    dt = dt_start

    while not finish:
        yield dt.date()
        if forward:
            days = days_in_month(dt)
            dt = dt + timedelta(days=days)            
            finish = dt > dt_end
        else:
            _tmp_dt = dt.replace(day=1) - timedelta(days=1)
            dt = (_tmp_dt.replace(day=dt.day))
            finish = dt < dt_end

Example #1

date_start = datetime(2016, 6, 1)
date_end = datetime(2017, 1, 1)

for p in monthly_range(date_start, date_end):
    print(p)

Output

2016-06-01
2016-07-01
2016-08-01
2016-09-01
2016-10-01
2016-11-01
2016-12-01
2017-01-01

Example #2

date_start = datetime(2017, 1, 1)
date_end = datetime(2016, 6, 1)

for p in monthly_range(date_start, date_end):
    print(p)

Output

2017-01-01
2016-12-01
2016-11-01
2016-10-01
2016-09-01
2016-08-01
2016-07-01
2016-06-01

Can‘t* believe this question has existed for 9 years without anyone suggesting a simple recursive function:

from datetime import datetime, timedelta

def walk_days(start_date, end_date):
    if start_date <= end_date:
        print(start_date.strftime("%Y-%m-%d"))
        next_date = start_date + timedelta(days=1)
        walk_days(next_date, end_date)

#demo
start_date = datetime(2009, 5, 30)
end_date   = datetime(2009, 6, 9)

walk_days(start_date, end_date)

Output:

2009-05-30
2009-05-31
2009-06-01
2009-06-02
2009-06-03
2009-06-04
2009-06-05
2009-06-06
2009-06-07
2009-06-08
2009-06-09

Edit: *Now I can believe it — see Does Python optimize tail recursion? . Thank you Tim.

You can generate a series of date between two dates using the pandas library simply and trustfully

import pandas as pd

print pd.date_range(start='1/1/2010', end='1/08/2018', freq='M')

You can change the frequency of generating dates by setting freq as D, M, Q, Y (daily, monthly, quarterly, yearly )

> pip install DateTimeRange

from datetimerange import DateTimeRange

def dateRange(start, end, step):
        rangeList = []
        time_range = DateTimeRange(start, end)
        for value in time_range.range(datetime.timedelta(days=step)):
            rangeList.append(value.strftime('%m/%d/%Y'))
        return rangeList

    dateRange("2018-09-07", "2018-12-25", 7)  

    Out[92]: 
    ['09/07/2018',
     '09/14/2018',
     '09/21/2018',
     '09/28/2018',
     '10/05/2018',
     '10/12/2018',
     '10/19/2018',
     '10/26/2018',
     '11/02/2018',
     '11/09/2018',
     '11/16/2018',
     '11/23/2018',
     '11/30/2018',
     '12/07/2018',
     '12/14/2018',
     '12/21/2018']

This function has some extra features:

• can pass a string matching the DATE_FORMAT for start or end and it is converted to a date object
• can pass a date object for start or end

• error checking in case the end is older than the start

  import datetime
  from datetime import timedelta
  
  
  DATE_FORMAT = '%Y/%m/%d'
  
  def daterange(start, end):
        def convert(date):
              try:
                    date = datetime.datetime.strptime(date, DATE_FORMAT)
                    return date.date()
              except TypeError:
                    return date
  
        def get_date(n):
              return datetime.datetime.strftime(convert(start) + timedelta(days=n), DATE_FORMAT)
  
        days = (convert(end) - convert(start)).days
        if days <= 0:
              raise ValueError('The start date must be before the end date.')
        for n in range(0, days):
              yield get_date(n)
  
  
  start = '2014/12/1'
  end = '2014/12/31'
  print list(daterange(start, end))
  
  start_ = datetime.date.today()
  end = '2015/12/1'
  print list(daterange(start, end))
  

Here’s code for a general date range function, similar to Ber’s answer, but more flexible:

def count_timedelta(delta, step, seconds_in_interval):
    """Helper function for iterate.  Finds the number of intervals in the timedelta."""
    return int(delta.total_seconds() / (seconds_in_interval * step))


def range_dt(start, end, step=1, interval='day'):
    """Iterate over datetimes or dates, similar to builtin range."""
    intervals = functools.partial(count_timedelta, (end - start), step)

    if interval == 'week':
        for i in range(intervals(3600 * 24 * 7)):
            yield start + datetime.timedelta(weeks=i) * step

    elif interval == 'day':
        for i in range(intervals(3600 * 24)):
            yield start + datetime.timedelta(days=i) * step

    elif interval == 'hour':
        for i in range(intervals(3600)):
            yield start + datetime.timedelta(hours=i) * step

    elif interval == 'minute':
        for i in range(intervals(60)):
            yield start + datetime.timedelta(minutes=i) * step

    elif interval == 'second':
        for i in range(intervals(1)):
            yield start + datetime.timedelta(seconds=i) * step

    elif interval == 'millisecond':
        for i in range(intervals(1 / 1000)):
            yield start + datetime.timedelta(milliseconds=i) * step

    elif interval == 'microsecond':
        for i in range(intervals(1e-6)):
            yield start + datetime.timedelta(microseconds=i) * step

    else:
        raise AttributeError("Interval must be 'week', 'day', 'hour' 'second', \
            'microsecond' or 'millisecond'.")

What about the following for doing a range incremented by days:

for d in map( lambda x: startDate+datetime.timedelta(days=x), xrange( (stopDate-startDate).days ) ):
  # Do stuff here

• startDate and stopDate are datetime.date objects

For a generic version:

for d in map( lambda x: startTime+x*stepTime, xrange( (stopTime-startTime).total_seconds() / stepTime.total_seconds() ) ):
  # Do stuff here

• startTime and stopTime are datetime.date or datetime.datetime object (both should be the same type)
• stepTime is a timedelta object

Note that .total_seconds() is only supported after python 2.7 If you are stuck with an earlier version you can write your own function:

def total_seconds( td ):
  return float(td.microseconds + (td.seconds + td.days * 24 * 3600) * 10**6) / 10**6

Slightly different approach to reversible steps by storing range args in a tuple.

def date_range(start, stop, step=1, inclusive=False):
    day_count = (stop - start).days
    if inclusive:
        day_count += 1

    if step > 0:
        range_args = (0, day_count, step)
    elif step < 0:
        range_args = (day_count - 1, -1, step)
    else:
        raise ValueError("date_range(): step arg must be non-zero")

    for i in range(*range_args):
        yield start + timedelta(days=i)

import datetime
from dateutil.rrule import DAILY,rrule

date=datetime.datetime(2019,1,10)

date1=datetime.datetime(2019,2,2)

for i in rrule(DAILY , dtstart=date,until=date1):
     print(i.strftime('%Y%b%d'),sep='\n')

OUTPUT:

2019Jan10
2019Jan11
2019Jan12
2019Jan13
2019Jan14
2019Jan15
2019Jan16
2019Jan17
2019Jan18
2019Jan19
2019Jan20
2019Jan21
2019Jan22
2019Jan23
2019Jan24
2019Jan25
2019Jan26
2019Jan27
2019Jan28
2019Jan29
2019Jan30
2019Jan31
2019Feb01
2019Feb02