Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:51431
Mailing-List: contact internals-help@lists.php.net; run by ezmlm
Received-SPF: error (pb1.pair.com: domain fmethod.com from 209.85.215.170 cause and error)
Message-ID: <129B11E6393E43AA846265F1C2D2EAC9@pc>
To: "Rasmus Lerdorf" <rasmus@lerdorf.com>,
	"Chris Stockton" <chrisstocktonaz@gmail.com>
Cc: "Ben Schmidt" <mail_ben_schmidt@yahoo.com.au>,
	"PHP Developers Mailing List" <internals@lists.php.net>
References: <AANLkTik3HXHKCYEFgZk2pwUkxfStwFrrOLoX9hEgQ0H4@mail.gmail.com>	<4D4B82DD.9040209@yahoo.com.au> <AANLkTi=v-R2fib6VmwXv94bJYXpqFA-fkNC=HUPFCvhq@mail.gmail.com> <4D4B8637.2000807@lerdorf.com>
Date: Mon, 7 Feb 2011 17:39:26 +0200
MIME-Version: 1.0
Content-Type: text/plain;
	format=flowed;
	charset="iso-8859-1";
	reply-type=original
Content-Transfer-Encoding: 7bit
Subject: Re: [PHP-DEV] A quick consensus on magic method idea
From: sv_forums@fmethod.com ("Stan Vass")


> This creates problems for opcode caches since the classes are all
> precompiled and usually optimized to the point where the opcodes that
> loads them are NOP'ed away and you just have the full class table cached
> in memory for each op_array.  Walking through that class table on every
> request and looking for these magic methods to call would be quite
> annoying, especially since most of them wouldn't have such a method to
> begin with.
>
> -Rasmus
>

As a class is added to the table (on first execute of the opcodes) it can 
check at that point if the magic method exists, and put it in a separate 
compact list to call for every request. Walking all classes repeatedly is 
thus avoided.

However, a remaining problem is the user no longer controls when exactly are 
his classes initialized, and in what concrete order. We have the potential 
for race conditions, where classes B, and C rely on initialized class A, or 
on some other initialization routine called in userland, which has not been 
executed yet, producing "sometimes broken" code that's hard to debug.

Stan Vass