[Python-Dev] Disallow float arguments where an integer is expected in Python 2.7.

[Gregory P. Smith]

On Mon, Dec 21, 2009 at 7:02 AM, Mark Dickinson <dickinsm at gmail.com> wrote:
> In Python 2.7, PyArg_ParseTuple and friends currently accept a float
> argument where an integer is expected, but produce a
> DeprecationWarning in this case.  This can be seen in various places
> in Python proper:
>
>>>> itertools.combinations(range(5), 2.0)
> __main__:1: DeprecationWarning: integer argument expected, got float
> <itertools.combinations object at 0x10059f638>
>
> Are there any objections to turning this DeprecationWarning into a
> TypeError for Python 2.7?  The behaviour has been deprecated since
> Python 2.3, it's gone in 3.x, and having it produce an error in 2.7
> might slightly reduce the number of surprises involved in porting from
> 2.x to 3.x.  There's a patch at http://bugs.python.org/issue5080
>
> There's one fly in the ointment:  the deprecation warning is produced
> for all integer codes except for one---the 'L' code.  The patch adds a
> deprecation warning for this code.
>
> Mark

+0.5 from me assuming that warning has been present by default in
python 2.5 and 2.6.  (the fly prevents me from giving it a full +1 as
it effectively means some APIs have never issued the warning at all).
No hard feelings if the 'L' forces us to back off and not do this.  At
least its fixed in 3.x.