Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:57567
Mailing-List: contact internals-help@lists.php.net; run by ezmlm
Received-SPF: error (pb1.pair.com: domain schlueters.de from 217.114.211.66 cause and error)
To: Ryan McCue <lists@rotorised.com>
Cc: internals@lists.php.net
In-Reply-To: <4F25F6E5.3050404@rotorised.com>
References:
	 <CADqTB_hEP4yQHa8jfm4s7YCvbBwUdH4eHiXZ68PDc_S-n4Zm4Q@mail.gmail.com>
	 <1327883576.6780.722.camel@guybrush>  <4F25F6E5.3050404@rotorised.com>
Content-Type: text/plain; charset="UTF-8"
Date: Mon, 30 Jan 2012 12:15:36 +0100
Message-ID: <1327922136.1876.181.camel@guybrush>
Mime-Version: 1.0
Content-Transfer-Encoding: 8bit
Subject: Re: [PHP-DEV] ReflectionFile missing
From: johannes@schlueters.de (Johannes =?ISO-8859-1?Q?Schl=FCter?=)

On Mon, 2012-01-30 at 11:48 +1000, Ryan McCue wrote:
> Johannes Schlüter wrote:
> > PS. Mind that the example you've given even works on files not included
> > by parsing files, whereas internal reflection provides information what
> > actually is available from engine point of view ...
> 
> The original reason I wrote that class was for a documentation tool, so
> that we could discover classes before loading the files.

Right, I can see the need. But I think it is a different project from
the simple "ReflectionFile" thing, as long as that is supposed to work
with the current reflection system.

Doing a "static reflection" from inside PHP is not easy as currently the
parser is bound quite tight to the engine and our compiler, directly
affecting function and class tables. We'd have to split that out and
even if we do that there's one benefit(*) when doing it in PHP userland:
It will always be bound to valid syntax for the given PHP interpreter.
It won't work cross-version when we change the language. (so running 5.3
you couldn't analyze 5.4 code which might use traits etc.) a userland
implementation can be made version independent and updated with support
for new versions without breaking support for the old version.

Of course doing this in userland means cost of performance, but I expect
that IO on the one side and client code using the analyzed data on the
other side have a bigger impact and the difference can almost be
neglected, but that's a guess.

As a remark, there's another static reflection project:
https://github.com/manuelpichler/staticReflection maybe you can work
together? To create the best :-)

johannes

*) of course there are all those other benefits of writing userland code
like being simpler for the user to understand and debug, being simpler
to maintain, which means more potential contributors, ...