39

Does python offer a way to easily get the current week of the month (1:4) ?

martineau
  • 112,593
  • 23
  • 157
  • 280
Joao Figueiredo
  • 2,962
  • 3
  • 30
  • 39
  • 1
    When you reach the 29th, you are actually in the fifth week, right? – erikbstack Sep 27 '10 at 17:59
  • 1
    Does a week start on the first day of the month or always on Monday? Or always on Sunday? Or... ? – Mark Byers Sep 27 '10 at 18:02
  • "week of the month" is not exactly a commonly used concept. You need to tell us what is your definition, with examples, of "week of the month" ... in particular, week of WHICH month e.g. week starts on Monday, suppose today is both Sunday and the first day of the month: do you want (previous month, week 5) or (current month, week -1 or 0) or something else? – John Machin Sep 28 '10 at 00:47
  • Here's one definition: http://download-llnw.oracle.com/javase/1.3/docs/api/java/util/GregorianCalendar.html – John Machin Sep 28 '10 at 01:04
  • Mark Byers answer went straight to the point. I'm accessing an Oracle DB with cx_Oracle and trying to optimize the computation time passing queries already with the month and week of the month(the table is partitioned by month and subpartitioned by week of the month, which John Machin link has a definition as: "WEEK_OF_MONTH field range from .. 1 to .. 5") – Joao Figueiredo Sep 28 '10 at 11:07
  • Just to clarify, this was to avoid using PARTITION_KEY = ((TO_CHAR(SYSDATE, 'MM'))-1) and SUBPARTITION_WEEK = TO_NUMBER(TO_CHAR(SYSDATE, 'W')) as it would invalidate the index use. Passing the query with the correct numbers allows using the indexes. – Joao Figueiredo Sep 28 '10 at 11:19

17 Answers17

47

In order to use straight division, the day of month for the date you're looking at needs to be adjusted according to the position (within the week) of the first day of the month. So, if your month happens to start on a Monday (the first day of the week), you can just do division as suggested above. However, if the month starts on a Wednesday, you'll want to add 2 and then do the division. This is all encapsulated in the function below.

from math import ceil

def week_of_month(dt):
    """ Returns the week of the month for the specified date.
    """

    first_day = dt.replace(day=1)

    dom = dt.day
    adjusted_dom = dom + first_day.weekday()

    return int(ceil(adjusted_dom/7.0))
Josh
  • 11,888
  • 4
  • 46
  • 48
24

I know this is years old, but I spent a lot of time trying to find this answer. I made my own method and thought I should share.

The calendar module has a monthcalendar method that returns a 2D array where each row represents a week. For example:

import calendar
calendar.monthcalendar(2015,9)

result:

[[0,0,1,2,3,4,5],
 [6,7,8,9,10,11,12],
 [13,14,15,16,17,18,19],
 [20,21,22,23,24,25,26],
 [27,28,29,30,0,0,0]]

So numpy's where is your friend here. And I'm in USA so I want the week to start on Sunday and the first week to be labelled 1:

import calendar
import numpy as np
calendar.setfirstweekday(6)

def get_week_of_month(year, month, day):
    x = np.array(calendar.monthcalendar(year, month))
    week_of_month = np.where(x==day)[0][0] + 1
    return(week_of_month)

get_week_of_month(2015,9,14)

returns

3
16

If your first week starts on the first day of the month you can use integer division:

import datetime
day_of_month = datetime.datetime.now().day
week_number = (day_of_month - 1) // 7 + 1
Mark Byers
  • 767,688
  • 176
  • 1,542
  • 1,434
  • maybe datetime.datetime.now().day would give the today's day in the month – aeter Sep 27 '10 at 18:11
  • Thanks. It turns out it wasn't about Python after all, but about arithmetics :) – Joao Figueiredo Sep 28 '10 at 11:00
  • 2
    It's not clear to me that this works. Assume your week starts on a Monday and the 1st for the month of interest is a Sunday, you'd want day of month 1 to map to week of month 1. The algorithm above will return the correct answer in this case. However, if you move to the second day, a Monday the 2nd, you'd want the algorithm to return a 2. In this case, the algorithm returns the wrong answer. Please see my answer below which returns the correct result in both cases by accounting for the offset. – Josh May 29 '13 at 02:23
  • It fails for the date Feb 10th 2020 case and most other cases too .It is not a generic one – saravanan saminathan Mar 11 '20 at 10:44
14

Check out the package Pendulum

>>> dt = pendulum.parse('2018-09-30')
>>> dt.week_of_month
5
Prateek Dorwal
  • 193
  • 2
  • 10
  • I've been using that solution until 2021-01-10, 2021-01-17, 2021-01-24, 2021-01-31 dates return minus values as that, respectively: -51, -50, -49, -48.. So it's kind of buggy. – Gunner1905 Jul 06 '21 at 08:20
5

This version could be improved but as a first look in python modules (datetime and calendar), I make this solution, I hope could be useful:

from datetime import datetime
n = datetime.now()
#from django.utils.timezone import now
#n = now() #if you use django with timezone

from calendar import Calendar
cal = Calendar() # week starts Monday
#cal = Calendar(6) # week stars Sunday

weeks = cal.monthdayscalendar(n.year, n.month)
for x in range(len(weeks)):
    if n.day in weeks[x]:
        print x+1
manus
  • 467
  • 10
  • 11
3

Josh's answer has to be tweaked slightly to accomodate the first day falling on a Sunday.

def get_week_of_month(date):
   first_day = date.replace(day=1)

   day_of_month = date.day

   if(first_day.weekday() == 6):
       adjusted_dom = (1 + first_day.weekday()) / 7
   else:
       adjusted_dom = day_of_month + first_day.weekday()

   return int(ceil(adjusted_dom/7.0))
crazyDiamond
  • 932
  • 1
  • 13
  • 20
3

Check out the python calendar module

vito huang
  • 3,956
  • 6
  • 24
  • 22
2
def week_of_month(date_value):
    week = date_value.isocalendar()[1] - date_value.replace(day=1).isocalendar()[1] + 1
    return date_value.isocalendar()[1] if week < 0 else week

date_value should in timestamp format This will give the perfect answer in all the cases. It is purely based on ISO calendar

1

I found a quite simple way:

import datetime
def week(year, month, day):
    first_week_month = datetime.datetime(year, month, 1).isocalendar()[1]
    if month == 1 and first_week_month > 10:
        first_week_month = 0
    user_date = datetime.datetime(year, month, day).isocalendar()[1]
    if month == 1 and user_date > 10:
        user_date = 0
    return user_date - first_week_month

returns 0 if first week

1

Josh' answer seems the best but I think that we should take into account the fact that a week belongs to a month only if its Thursday falls into that month. At least that's what the iso says.

According to that standard, a month can have up to 5 weeks. A day could belong to a month, but the week it belongs to may not.

I have taken into account that just by adding a simple

if (first_day.weekday()>3) :
        return ret_val-1
    else:
        return ret_val

where ret_val is exactly Josh's calculated value. Tested on June 2017 (has 5 weeks) and on September 2017. Passing '2017-09-01' returns 0 because that day belongs to a week that does not belong to September.

The most correct way would be to have the method return both the week number and the month name the input day belongs to.

benelgiac
  • 781
  • 5
  • 9
1

A variation on @Manuel Solorzano's answer:

from calendar import monthcalendar
def get_week_of_month(year, month, day):
    return next(
        (
            week_number
            for week_number, days_of_week in enumerate(monthcalendar(year, month), start=1)
            if day in days_of_week
        ),
        None,
    )

E.g.:

>>> get_week_of_month(2020, 9, 1)
1
>>> get_week_of_month(2020, 9, 30)
5
>>> get_week_of_month(2020, 5, 35)
None
vekerdyb
  • 1,085
  • 10
  • 25
1

Say we have some month's calender as follows:

Mon Tue Wed Thur Fri Sat Sun
                 1   2   3 
4   5   6   7    8   9   10

We say day 1 ~ 3 belongs to week 1 and day 4 ~ 10 belongs to week 2 etc.

In this case, I believe the week_of_month for a specific day should be calculated as follows:

import datetime
def week_of_month(year, month, day):
    weekday_of_day_one = datetime.date(year, month, 1).weekday()
    weekday_of_day = datetime.date(year, month, day)
    return (day - 1)//7 + 1 + (weekday_of_day < weekday_of_day_one)

However, if instead we want to get the nth of the weekday that date is, such as day 1 is the 1st Friday, day 8 is the 2nd Friday, and day 6 is the 1st Wednesday, then we can simply return (day - 1)//7 + 1

1

The answer you are looking for is (dm-dw+(dw-dm)%7)/7+1 where dm is the day of the month, dw is the day of the week, and % is the positive remainder.

This comes from relating the month offset (mo) and the week of the month (wm), where the month offset is how far into the week the first day starts. If we consider all of these variables to start at 0 we have

wm*7+dw = dm+mo

You can solve this modulo 7 for mo as that causes the wm variable drops out as it only appears as a multiple of 7

dw = dm+mo   (%7)
mo = dw-dm   (%7)
mo = (dw-dm)%7  (since the month offset is 0-6)

Then you just substitute the month offset into the original equation and solve for wm

wm*7+dw = dm+mo
wm*7 = dm-dw + mo
wm*7 = dm-dw + (dw-dm)%7
wm = (dm-dw + (dw-dm)%7) / 7

As dm and dw are always paired, these can be offset by any amount, so, switching everything to start a 1 only changes the the equation to (dm-dw + (dw-dm)%7)/7 + 1.

Of course the python datetime library starts dm at 1 and dw at 0. So, assuming date is a datatime.date object, you can go with

(date.day-1-date.dayofweek() + (date.dayofweek()-date.day+1)%7) / 7 + 1

As the inner bit is always a multiple of 7 (it is literally dw*7), you can see that the first -date.dayofweek() simply adjusts the value backwards to closest multiple of 7. Integer division does this too, so it can be further simplified to

(date.day-1 + (date.dayofweek()-date.day+1)%7) // 7 + 1

Be aware that dayofweek() puts Sunday at the end of the week.

  • Thank you for you answer. It is 100% correct. I've tested this one by using the calendar.monthcalendar as a source of truth and iterating from datetime.date.min up to datetime.date.max and the results were the same. I just needed to change daysofweek() -> weekday() and remove the +1 at the end. Please check out this Gist for more details. https://gist.github.com/klement97/0912e56a8023741e87201fac0da261cd – klement omeri Aug 07 '21 at 20:47
0

This should do it.

#! /usr/bin/env python2

import calendar, datetime

#FUNCTIONS
def week_of_month(date):
    """Determines the week (number) of the month"""

    #Calendar object. 6 = Start on Sunday, 0 = Start on Monday
    cal_object = calendar.Calendar(6)
    month_calendar_dates = cal_object.itermonthdates(date.year,date.month)

    day_of_week = 1
    week_number = 1

    for day in month_calendar_dates:
        #add a week and reset day of week
        if day_of_week > 7:
            week_number += 1
            day_of_week = 1

        if date == day:
            break
        else:
            day_of_week += 1

    return week_number


#MAIN
example_date = datetime.date(2015,9,21)

print "Week",str(week_of_month(example_date))
#Returns 'Week 4'
Dan
  • 1
  • 1
0

Move to last day of week in month and divide to 7

from math import ceil

def week_of_month(dt):
    """ Returns the week of the month for the specified date.
    """
    # weekday from monday == 0 ---> sunday == 6
    last_day_of_week_of_month = dt.day + (7 - (1 + dt.weekday()))
    return int(ceil(last_day_of_week_of_month/7.0))
Tan Nguyen
  • 567
  • 4
  • 8
0

You can simply do as follow:

  1. First extract the month and the week of year number
df['month'] = df['Date'].dt.month
df['week'] = df['Date'].dt.week 
  1. Then group by month and rank the week numbers
df['weekOfMonth'] = df.groupby('month')["week"].rank("dense", ascending=False)
-1
  import datetime
    
  def week_number_of_month(date_value):
            week_number = (date_value.isocalendar()[1] - date_value.replace(day=1).isocalendar()[1] + 1)
            if week_number == -46:
                week_number = 6
           return week_number
                
 date_given = datetime.datetime(year=2018, month=12, day=31).date()
                
 week_number_of_month(date_given)
kamran kausar
  • 3,467
  • 21
  • 17