Path: csiph.com!usenet.pasdenom.info!weretis.net!feeder4.news.weretis.net!rt.uk.eu.org!newsfeed.xs4all.nl!newsfeed3.news.xs4all.nl!xs4all!post.news.xs4all.nl!not-for-mail
MIME-Version: 1.0
In-Reply-To: <51FED74E.6080003@markusrother.de>
References: <51FED74E.6080003@markusrother.de>
Date: Mon, 5 Aug 2013 00:06:16 +0100
Subject: Re: Bug? ( () == [] ) != ( ().__eq__([]) )
From: Chris Angelico <rosuav@gmail.com>
To: python-list@python.org
Content-Type: text/plain; charset=ISO-8859-1
Precedence: list
Newsgroups: comp.lang.python
Message-ID: <mailman.195.1375658082.1251.python-list@python.org>
Lines: 28
NNTP-Posting-Host: 2001:888:2000:d::a6
Xref: csiph.com comp.lang.python:51943

On Sun, Aug 4, 2013 at 11:35 PM, Markus Rother <python@markusrother.de> wrote:
> Hello,
>
> The following behaviour seen in 3.2 seems very strange to me:
>
> As expected:
>>>> () == []
> False
>
> However:
>>>> ().__eq__([])
> NotImplemented
>>>> [].__eq__(())
> NotImplemented

You don't normally want to be calling dunder methods directly. The
reasoning behind this behaviour goes back to a few things, including a
way to handle "1 == Foo()" where Foo is a custom type that implements
__eq__; obviously the integer 1 won't know whether it's equal to a Foo
instance or not, so it has to defer to the second operand to get a
result. This deferral is done by returning NotImplemented, which is an
object, and so is true by default. I don't see any particular reason
for it to be false, as you shouldn't normally be using it; it's more
like a "null" state, it means "I don't know if we're equal or not". If
neither side knows whether they're equal, then they're presumed to be
unequal, but you can't determine that from a single call to __eq__.

ChrisA