Path: csiph.com!x330-a1.tempe.blueboxinc.net!usenet.pasdenom.info!weretis.net!feeder1.news.weretis.net!feeder.erje.net!xlned.com!feeder3.xlned.com!amsnews11.chello.com!newsgate.cistron.nl!newsgate.news.xs4all.nl!post.news.xs4all.nl!not-for-mail
MIME-Version: 1.0
In-Reply-To: <Xns9FB880CEA6F2Fduncanbooth@127.0.0.1>
References: <4EE48AB3.7070807@gmail.com> <mailman.3510.1323602145.27778.python-list@python.org> <Xns9FB880CEA6F2Fduncanbooth@127.0.0.1>
Date: Mon, 12 Dec 2011 00:09:56 +1100
Subject: Re: Verbose and flexible args and kwargs syntax
From: Chris Angelico <rosuav@gmail.com>
To: python-list@python.org
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: quoted-printable
Precedence: list
Newsgroups: comp.lang.python
Message-ID: <mailman.3511.1323608999.27778.python-list@python.org>
Lines: 63
NNTP-Posting-Host: 2001:888:2000:d::a6
Xref: x330-a1.tempe.blueboxinc.net comp.lang.python:16983

On Sun, Dec 11, 2011 at 11:39 PM, Duncan Booth
<duncan.booth@invalid.invalid> wrote:
> Chris Angelico <rosuav@gmail.com> wrote:
> If it used keywords then you could keep symmetry quite easily:
>
> =A0 =A0def anyargs(arglist args, argdict kwargs):
> =A0 =A0 =A0 =A0return wrappedfunc(arglist args, argdict kwargs)
>
> and you would have the advantage of two new keywords that people could
> actually search on.

Yes, that's just a strict keywordification of the * and ** symbols.
The same argument could be made for eliminating the standard algebraic
+ operator and replacing it with a keyword "__add__". I don't think
that's worthwhile.

The OP suggested using 'dict' and 'list' themselves as the keywords,
thus allowing the use of subclasses. This would make it unsearchable,
or else rather verbose:

def anyargs(pack list args, pack dict kwargs):
    return wrappedfunc(pack list args, pack dict kwargs)

With this syntax, what happens if you muck up list/dict? Or
alternatively, the briefer syntax:

def anyargs(pack list args, pack dict kwargs):
    return wrappedfunc(pack args, pack kwargs)

 which breaks the symmetry, and doesn't say which one you're doing -
for instance, if you only use one out of list and dict args, it would
make sense to have a variable "options" or "args" that gets unpacked
to the function's arguments - and there's no way to see whether it's
going to be keyword or positional.

The verbose syntax has a lot going for it, but it's rather verbose.
Why say "pack list args" when you can just say "*args"? Compare
initializer syntax between Python and PHP:

foo =3D ["asdf", "qwer", {1:2, 3:4}]

$foo =3D array("asdf", "qwer", array(1=3D>2, 3=3D>4))

Is it more Pythonic to use explicitly-named types, or to have simple
notation that's clear and easy to read? Or is this a matter for
personal preference?

>> Another issue: You suggest being able to use "attrdict" or some other
>> dict subclass. This means that, rather than being a language
>> construct, this will involve a name lookup. And what happens if you
>> have a class that subclasses both list and dict? Will it get the
>> positional args, the keyword args, or both?
>
> Irrelevant, you can't subclass both list and dict:
> TypeError: multiple bases have instance lay-out conflict

Ah. Curious. I've not done much with multiple inheritance in Python
(come to think of it, I don't recall when I last used MI in _any_
language). In any case, there's still the potential unclarity as to
_which_ of dict and list is the one that's been inherited, which
requires a run-time lookup to solve.

ChrisA