Python 2.4.1 (#1, May 16 2005, 15:19:29)
[GCC 4.0.0 20050512 (Red Hat 4.0.0-5)] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> from time import strptime
>>> strptime('2006-53-0', '%Y-%U-%w')
(2006, 12, 31, 0, 0, 0, 6, 365, -1)
>>> strptime('2007-00-0', '%Y-%U-%w')
(2006, 12, 24, 0, 0, 0, 6, -7, -1)
>>> strptime('2007-01-0', '%Y-%U-%w')
(2007, 1, 7, 0, 0, 0, 6, 7, -1)
>>> strptime('2007-02-0', '%Y-%U-%w')
(2007, 1, 7, 0, 0, 0, 6, 7, -1)
>>> strptime('2007-03-0', '%Y-%U-%w')
(2007, 1, 14, 0, 0, 0, 6, 14, -1)
>>>


Note that in the above test, Sunday of week 1 and week 2 for 2007 reported the date as 2007-01-07 and Sunday of week 0 was reported as 2006-12-24, not 2006-12-31.  I'm not sure exactly what is correct, but the inconsistencies are bothersome.

Same results on:
Python 2.4.4c1 (#70, Oct 11 2006, 10:59:14) [MSC v.1310 32 bit (Intel)] on win32

I will try to fix this when I can.  Just to warn you, Brian, I really doubt I will put the effort into backporting this to 2.4.

No worries.

Here's what I'm doing as a work-around.  I needed to convert the results of a mysql YEARWEEK field to the sunday at the start of that week:

import datetime
def mysqlWeekToSundayDate(yearweek):
    year = int(yearweek[0:4])
    week = int(yearweek[4:6])
    day = datetime.date(year, 1, 1)
    dayDelta = datetime.timedelta(1)
    weekDelta = datetime.timedelta(7)
    while day.strftime("%w") != "0":
        day = day + dayDelta
    
    day = day + ((week - 1) * weekDelta)
        
    return day

I'm relatively new to Python so it is probably not the most efficient method but it does the job.

Rev. 53564 (trunk) has the fix and 2.5 will as soon as a commit problem I am having is fixed.  I basically rewrote the algorithm to have a generic calculation for the Julian day and just shifted the length of week 0 and the day of the week based on whether %U or %W was specified.  Cut out all the other edge cases which were messy and confusing.