Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:44769
Mailing-List: contact internals-help@lists.php.net; run by ezmlm
Received-SPF: error (pb1.pair.com: domain chiaraquartet.net from 209.85.198.226 cause and error)
Message-ID: <4A536183.407@chiaraquartet.net>
Date: Tue, 07 Jul 2009 09:53:55 -0500
User-Agent: Thunderbird 2.0.0.6 (Macintosh/20070807)
MIME-Version: 1.0
To: Christian Schneider <cschneid@cschneid.com>
CC: PHP internals <internals@lists.php.net>
References: <2D0F5226-EBCA-4B45-BF01-8ED1C643976C@prohost.org> <4A535468.2000606@chiaraquartet.net> <A4DD837EEB0441B1932380ABE8AE8146@pc> <4A535BEF.8060302@cschneid.com>
In-Reply-To: <4A535BEF.8060302@cschneid.com>
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit
Subject: Re: [PHP-DEV] Re: Type hinting/casting request for vote
From: greg@chiaraquartet.net (Greg Beaver)

Christian Schneider wrote:
> Stan Vassilev wrote:
>>> +1 if the object type hint is change to use T_CLASS so we don't break
>>> every external package using "Object" as a base class.
>>> http://www.google.com/codesearch?as_q=class\s%2Bobject&btnG=Search+Code&hl=en&as_lang=php&as_license_restrict=i&as_license=&as_package=&as_filename=&as_case=
>> Or this can wait until 6.0, when (as I hear) we'll have case-sensitive
>> class names, so Object/object, Int/int won't cause collisions.
> 
> This won't help much because of functions:

This changes my vote to -1 in any version without a technical fix in the
patch to avoid this problem.

There are 2 ways to avoid this problem

1) create a patch allowing reserved words as function/class names.  I've
already tried this for a previous reason, and it's difficult to do
without vastly over-complicating the lexer.

2) have 1 syntax, T_*_CAST as in function ((int) $a) {} and have that
mean strict type hinting.

Ilia: if you can include a lexer patch that allows reserved words in key
places, I'd be willing to modify my vote.  The problem is supporting
static class things requires lookahead tokens, thus one would need to
specify that this pseudo-regex is a T_STRING:

/(LABEL)(?:\s*->|::)/

not to mention having a new "looking for label" state that is entered
whenever we encounter class|interface\s+(extends|implements|{) or T_NEW

Although having this freedom would be nice, I prefer #2, it's a whole
lot simpler.

Greg