在Python中从日期中减去一个月的最简单方法是什么？

If only timedelta had a month argument in it’s constructor. So what’s the simplest way to do this?

EDIT: I wasn’t thinking too hard about this as was pointed out below. Really what I wanted was any day in the last month because eventually I’m going to grab the year and month only. So given a datetime object, what’s the simplest way to return any datetime object that falls in the previous month?

Try this:

def monthdelta(date, delta):
    m, y = (date.month+delta) % 12, date.year + ((date.month)+delta-1) // 12
    if not m: m = 12
    d = min(date.day, [31,
        29 if y%4==0 and (not y%100==0 or y%400 == 0) else 28,
        31,30,31,30,31,31,30,31,30,31][m-1])
    return date.replace(day=d,month=m, year=y)

>>> for m in range(-12, 12):
    print(monthdelta(datetime.now(), m))

    
2009-08-06 16:12:27.823000
2009-09-06 16:12:27.855000
2009-10-06 16:12:27.870000
2009-11-06 16:12:27.870000
2009-12-06 16:12:27.870000
2010-01-06 16:12:27.870000
2010-02-06 16:12:27.870000
2010-03-06 16:12:27.886000
2010-04-06 16:12:27.886000
2010-05-06 16:12:27.886000
2010-06-06 16:12:27.886000
2010-07-06 16:12:27.886000
2010-08-06 16:12:27.901000
2010-09-06 16:12:27.901000
2010-10-06 16:12:27.901000
2010-11-06 16:12:27.901000
2010-12-06 16:12:27.901000
2011-01-06 16:12:27.917000
2011-02-06 16:12:27.917000
2011-03-06 16:12:27.917000
2011-04-06 16:12:27.917000
2011-05-06 16:12:27.917000
2011-06-06 16:12:27.933000
2011-07-06 16:12:27.933000
>>> monthdelta(datetime(2010,3,30), -1)
datetime.datetime(2010, 2, 28, 0, 0)
>>> monthdelta(datetime(2008,3,30), -1)
datetime.datetime(2008, 2, 29, 0, 0)

Edit Corrected to handle the day as well.

Edit See also the answer from puzzlement which points out a simpler calculation for d:

d = min(date.day, calendar.monthrange(y, m)[1])

You can use the third party dateutil module (PyPI entry here).

import datetime
import dateutil.relativedelta

d = datetime.datetime.strptime("2013-03-31", "%Y-%m-%d")
d2 = d - dateutil.relativedelta.relativedelta(months=1)
print d2

output:

2013-02-28 00:00:00

After the original question’s edit to “any datetime object in the previous month”, you can do it pretty easily by subtracting 1 day from the first of the month.

from datetime import datetime, timedelta

def a_day_in_previous_month(dt):
   return dt.replace(day=1) - timedelta(days=1)

A variation on Duncan’s answer (I don’t have sufficient reputation to comment), which uses calendar.monthrange to dramatically simplify the computation of the last day of the month:

import calendar
def monthdelta(date, delta):
    m, y = (date.month+delta) % 12, date.year + ((date.month)+delta-1) // 12
    if not m: m = 12
    d = min(date.day, calendar.monthrange(y, m)[1])
    return date.replace(day=d,month=m, year=y)

Info on monthrange from Get Last Day of the Month in Python

A vectorized, pandas solution is very simple:

df['date'] - pd.DateOffset(months=1)

If only timedelta had a month argument in it’s constructor. So what’s the simplest way to do this?

What do you want the result to be when you subtract a month from, say, a date that is March 30? That is the problem with adding or subtracting months: months have different lengths! In some application an exception is appropriate in such cases, in others “the last day of the previous month” is OK to use (but that’s truly crazy arithmetic, when subtracting a month then adding a month is not overall a no-operation!), in others yet you’ll want to keep in addition to the date some indication about the fact, e.g., “I’m saying Feb 28 but I really would want Feb 30 if it existed”, so that adding or subtracting another month to that can set things right again (and the latter obviously requires a custom class holding a data plus s/thing else).

There can be no real solution that is tolerable for all applications, and you have not told us what your specific app’s needs are for the semantics of this wretched operation, so there’s not much more help that we can provide here.

I think the simple way is to use DateOffset from Pandas like so:

import pandas as pd
date_1 = pd.to_datetime("2013-03-31", format="%Y-%m-%d") - pd.DateOffset(months=1)

The result will be a datetime object

If all you want is any day in the last month, the simplest thing you can do is subtract the number of days from the current date, which will give you the last day of the previous month.

For instance, starting with any date:

>>> import datetime                                                                                                                                                                 
>>> today = datetime.date.today()                                                                                                                                                   
>>> today
datetime.date(2016, 5, 24)

Subtracting the days of the current date we get:

>>> last_day_previous_month = today - datetime.timedelta(days=today.day)
>>> last_day_previous_month
datetime.date(2016, 4, 30)

This is enough for your simplified need of any day in the last month.

But now that you have it, you can also get any day in the month, including the same day you started with (i.e. more or less the same as subtracting a month):

>>> same_day_last_month = last_day_previous_month.replace(day=today.day)
>>> same_day_last_month
datetime.date(2016, 4, 24)

Of course, you need to be careful with 31st on a 30 day month or the days missing from February (and take care of leap years), but that’s also easy to do:

>>> a_date = datetime.date(2016, 3, 31)                                                                                                                                             
>>> last_day_previous_month = a_date - datetime.timedelta(days=a_date.day)
>>> a_date_minus_month = (
...     last_day_previous_month.replace(day=a_date.day)
...     if a_date.day < last_day_previous_month.day
...     else last_day_previous_month
... )
>>> a_date_minus_month
datetime.date(2016, 2, 29)

I use this for government fiscal years where Q4 starts October 1st. Note I convert the date into quarters and undo it as well.

import pandas as pd

df['Date'] = '1/1/2020'
df['Date'] = pd.to_datetime(df['Date'])              #returns 2020-01-01
df['NewDate'] = df.Date - pd.DateOffset(months=3)    #returns 2019-10-01 <---- answer

# For fun, change it to FY Quarter '2019Q4'
df['NewDate'] = df['NewDate'].dt.year.astype(str) + 'Q' + df['NewDate'].dt.quarter.astype(str)

# Convert '2019Q4' back to 2019-10-01
df['NewDate'] = pd.to_datetime(df.NewDate)

Here is some code to do just that. Haven’t tried it out myself…

def add_one_month(t):
    """Return a `datetime.date` or `datetime.datetime` (as given) that is
    one month earlier.

    Note that the resultant day of the month might change if the following
    month has fewer days:

        >>> add_one_month(datetime.date(2010, 1, 31))
        datetime.date(2010, 2, 28)
    """
    import datetime
    one_day = datetime.timedelta(days=1)
    one_month_later = t + one_day
    while one_month_later.month == t.month:  # advance to start of next month
        one_month_later += one_day
    target_month = one_month_later.month
    while one_month_later.day < t.day:  # advance to appropriate day
        one_month_later += one_day
        if one_month_later.month != target_month:  # gone too far
            one_month_later -= one_day
            break
    return one_month_later

def subtract_one_month(t):
    """Return a `datetime.date` or `datetime.datetime` (as given) that is
    one month later.

    Note that the resultant day of the month might change if the following
    month has fewer days:

        >>> subtract_one_month(datetime.date(2010, 3, 31))
        datetime.date(2010, 2, 28)
    """
    import datetime
    one_day = datetime.timedelta(days=1)
    one_month_earlier = t - one_day
    while one_month_earlier.month == t.month or one_month_earlier.day > t.day:
        one_month_earlier -= one_day
    return one_month_earlier

Given a (year,month) tuple, where month goes from 1-12, try this:

>>> from datetime import datetime
>>> today = datetime.today()
>>> today
datetime.datetime(2010, 8, 6, 10, 15, 21, 310000)
>>> thismonth = today.year, today.month
>>> thismonth
(2010, 8)
>>> lastmonth = lambda (yr,mo): [(y,m+1) for y,m in (divmod((yr*12+mo-2), 12),)][0]
>>> lastmonth(thismonth)
(2010, 7)
>>> lastmonth( (2010,1) )
(2009, 12)

Assumes there are 12 months in every year.

def month_sub(year, month, sub_month):
    result_month = 0
    result_year = 0
    if month > (sub_month % 12):
        result_month = month - (sub_month % 12)
        result_year = year - (sub_month / 12)
    else:
        result_month = 12 - (sub_month % 12) + month
        result_year = year - (sub_month / 12 + 1)
    return (result_year, result_month)

>>> month_sub(2015, 7, 1)    
(2015, 6)
>>> month_sub(2015, 7, -1)
(2015, 8)
>>> month_sub(2015, 7, 13)
(2014, 6)
>>> month_sub(2015, 7, -14)
(2016, 9)

I Used the following code to go back n Months from a specific Date:

your_date =  datetime.strptime(input_date, "%Y-%m-%d")  #to convert date(2016-01-01) to timestamp
start_date=your_date    #start from current date

#Calculate Month
for i in range(0,n):    #n = number of months you need to go back
    start_date=start_date.replace(day=1)    #1st day of current month
    start_date=start_date-timedelta(days=1) #last day of previous month

#Calculate Day
if(start_date.day>your_date.day):   
    start_date=start_date.replace(day=your_date.day)            

print start_date

For eg: input date = 28/12/2015 Calculate 6 months previous date.

I) CALCULATE MONTH: This step will give you the start_date as 30/06/2015.
Note that after the calculate month step you will get the last day of the required month.

II)CALCULATE DAY: Condition if(start_date.day>your_date.day) checks whether the day from input_date is present in the required month. This handles condition where input date is 31(or 30) and the required month has less than 31(or 30 in case of feb) days. It handles leap year case as well(For Feb). After this step you will get result as 28/06/2015

If this condition is not satisfied, the start_date remains the last date of the previous month. So if you give 31/12/2015 as input date and want 6 months previous date, it will give you 30/06/2015

You can use below given function to get date before/after X month.

from datetime import date

def next_month(given_date, month):
    yyyy = int(((given_date.year * 12 + given_date.month) + month)/12)
    mm = int(((given_date.year * 12 + given_date.month) + month)%12)

    if mm == 0:
        yyyy -= 1
        mm = 12

    return given_date.replace(year=yyyy, month=mm)


if __name__ == "__main__":
    today = date.today()
    print(today)

    for mm in [-12, -1, 0, 1, 2, 12, 20 ]:
        next_date = next_month(today, mm)
        print(next_date)

I think this answer is quite readable:

def month_delta(dt, delta):
    year_delta, month = divmod(dt.month + delta, 12)

    if month == 0:
        # convert a 0 to december
        month = 12
        if delta < 0:
            # if moving backwards, then it's december of last year
            year_delta -= 1

    year = dt.year + year_delta

    return dt.replace(month=month, year=year)

for delta in range(-20, 21):
    print(delta, "->", month_delta(datetime(2011, 1, 1), delta))

-20 -> 2009-05-01 00:00:00
-19 -> 2009-06-01 00:00:00
-18 -> 2009-07-01 00:00:00
-17 -> 2009-08-01 00:00:00
-16 -> 2009-09-01 00:00:00
-15 -> 2009-10-01 00:00:00
-14 -> 2009-11-01 00:00:00
-13 -> 2009-12-01 00:00:00
-12 -> 2010-01-01 00:00:00
-11 -> 2010-02-01 00:00:00
-10 -> 2010-03-01 00:00:00
-9 -> 2010-04-01 00:00:00
-8 -> 2010-05-01 00:00:00
-7 -> 2010-06-01 00:00:00
-6 -> 2010-07-01 00:00:00
-5 -> 2010-08-01 00:00:00
-4 -> 2010-09-01 00:00:00
-3 -> 2010-10-01 00:00:00
-2 -> 2010-11-01 00:00:00
-1 -> 2010-12-01 00:00:00
0 -> 2011-01-01 00:00:00
1 -> 2011-02-01 00:00:00
2 -> 2011-03-01 00:00:00
3 -> 2011-04-01 00:00:00
4 -> 2011-05-01 00:00:00
5 -> 2011-06-01 00:00:00
6 -> 2011-07-01 00:00:00
7 -> 2011-08-01 00:00:00
8 -> 2011-09-01 00:00:00
9 -> 2011-10-01 00:00:00
10 -> 2011-11-01 00:00:00
11 -> 2012-12-01 00:00:00
12 -> 2012-01-01 00:00:00
13 -> 2012-02-01 00:00:00
14 -> 2012-03-01 00:00:00
15 -> 2012-04-01 00:00:00
16 -> 2012-05-01 00:00:00
17 -> 2012-06-01 00:00:00
18 -> 2012-07-01 00:00:00
19 -> 2012-08-01 00:00:00
20 -> 2012-09-01 00:00:00

One liner ?

previous_month_date = (current_date - datetime.timedelta(days=current_date.day+1)).replace(day=current_date.day)

Simplest Way that i have tried Just now

from datetime import datetime
from django.utils import timezone





current = timezone.now()
if current.month == 1:
     month = 12
else:
     month = current.month - 1
current = datetime(current.year, month, current.day)

Some time ago I came across the following algorithm which works very well for incrementing and decrementing months on either a date or datetime.

CAVEAT: This will fail if day is not available in the new month. I use this on date objects where day == 1 always.

Python 3.x:

def increment_month(d, add=1):
    return date(d.year+(d.month+add-1)//12, (d.month+add-1) % 12+1, 1)

For Python 2.7 change the //12 to just /12 since integer division is implied.

I recently used this in a defaults file when a script started to get these useful globals:

MONTH_THIS = datetime.date.today()
MONTH_THIS = datetime.date(MONTH_THIS.year, MONTH_THIS.month, 1)

MONTH_1AGO = datetime.date(MONTH_THIS.year+(MONTH_THIS.month-2)//12,
                           (MONTH_THIS.month-2) % 12+1, 1)

MONTH_2AGO = datetime.date(MONTH_THIS.year+(MONTH_THIS.month-3)//12,
                           (MONTH_THIS.month-3) % 12+1, 1)

import datetime
date_str = '08/01/2018'
format_str = '%d/%m/%Y'
datetime_obj = datetime.datetime.strptime(date_str, format_str)   
datetime_obj.replace(month=datetime_obj.month-1)

Simple solution, no need for special libraries.

You could do it in two lines like this:

now = datetime.now()
last_month = datetime(now.year, now.month - 1, now.day)

remember the imports

from datetime import datetime