Question

From $PYTHONHOME/lib/python2.7/calendar.py, timegm is defined as

EPOCH = 1970
_EPOCH_ORD = datetime.date(EPOCH, 1, 1).toordinal()


def timegm(tuple):
    """Unrelated but handy function to calculate Unix timestamp from GMT."""
    year, month, day, hour, minute, second = tuple[:6]
    days = datetime.date(year, month, 1).toordinal() - _EPOCH_ORD + day - 1
    hours = days*24 + hour
    minutes = hours*60 + minute
    seconds = minutes*60 + second
    return seconds

Is there any reason why days is not calculated as :

days = datetime.date(year, month, day).toordinal() - _EPOCH_ORD

Can anyone think of cases where the previous expression would break?

Answer 1

datetime.date(year, month, day).toordinal()

“range checks” that the day (combined with year and month) is part of a valid datetime.date.

datetime.date(year, month, 1) does not — or rather, only requires that year and month are valid.

Apparently some user was exploiting this “feature” (of not checking the day) and for the sake of backwards compatibility (with Python2.2!) the developers kept this behavior. See Raymond Hettinger’s comment in commit #27881.

PS. You can find the commit that changed this line in calendar.py by running

hg blame cpython/Lib/calendar.py 

(provided you’ve hg cloned the CPython repository).

10940: def timegm(tuple):
11691:     """Unrelated but handy function to calculate Unix timestamp from GMT."""
11691:     year, month, day, hour, minute, second = tuple[:6]
27881:     days = datetime.date(year, month, 1).toordinal() - _EPOCH_ORD + day - 1

Then, to find more about commit #27881, you can search http://hg.python.org/