[Python-3000] [PythonInfo Wiki] Update of "GoogleSprintPy3k" by

Guido van Rossum guido at python.org
Tue Aug 22 05:55:49 CEST 2006

On 8/21/06, Collin Winter <collinw at gmail.com> wrote:
> On 8/21/06, Michael Urman <murman at gmail.com> wrote:
> > On 8/21/06, Guido van Rossum <guido at python.org> wrote:
> > > I'd like map(f, a, b) to be the same as to (f(*x) for x in zip(a, b))
> > > so we have to explain less. (And I think even map(f, *args) === (f(*x)
> > > for x in zip(*args)).)
> >
> > Should map(None, a, b) == zip(a, b), leaving python with multiple ways
> > to do one thing? Or should the surprising but useful map(None, ...)
> > behavior disappear or become even more surprising by padding? Is there
> > any reason at all for map to take multiple sequences now that we have
> > starmap and (i)zip?
> FWIW, I'm ambivalent as to whether map() accepts multiple sequences,
> but I'm strongly in favor of map(None, ....) disappearing. Similarly,
> I'd want to see filter(None, ...) go away, too; fastpathing the case
> of filter(bool, ....) will achieve the same performance benefit.

I think map(f, a, b, ...) and filter(p, a, b, ...) should stay, but
the None cases should be gotten rid of. I don't want to move starmap()
out of itertools into builtins.

I expect that filter(bool, a) is fast enough without greasing the
tracks, but if you don't, feel free to benchmark it.

--Guido van Rossum (home page: http://www.python.org/~guido/)

More information about the Python-3000 mailing list