Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:63056
Mailing-List: contact internals-help@lists.php.net; run by ezmlm
Received-SPF: error (pb1.pair.com: domain mindplay.dk from 209.85.220.170 cause and error)
MIME-Version: 1.0
Date: Tue, 18 Sep 2012 08:19:19 -0400
Message-ID: <CADqTB_hWnXfPq+pTTHPVNB==G-rnVu9DQ=dJVQgeJKdmDqRoDQ@mail.gmail.com>
To: internals@lists.php.net
Content-Type: multipart/alternative; boundary=20cf307c9afe4405b804c9f8e6e8
Subject: constructor hook
From: rasmus@mindplay.dk (Rasmus Schultz)

--20cf307c9afe4405b804c9f8e6e8
Content-Type: text/plain; charset=ISO-8859-1

Hey,

I'm going to make this brief, because I suspect a lot of people are going
to jump at the opportunity to cry bloody murder when I suggest this.

I wonder if it would make sense to have a way to globally hook into
__construct() - sort of like how you can hook into the autoloaders with
spl_autoload_register() ...

This could be useful for things like dependency injection and debugging.

Lithium (for one) uses a unified constructor to accomplish something like
this:

http://lithify.me/docs/manual/lithium-basics/architecture.wiki

That's bad for various reasons I won't delve into.

Various DI containers use a replacement for the "new" keyword, e.g.
DI::create('MyClass') which is bad (worse) for other reasons...

So basically, what if you could solve these design patterns by hooking into
constructors directly instead - something like:

<?php
> constructor_hook_register(function($object) {
>   echo 'an object of type '.get_class($object).' was constructed!';
> });
> $test = new Foo();


would print:

an object of type Foo was constructed!


One other area where this could help, is in event-driven systems, where
this sort of thing could be used for "co-construction".

Of course, you could accomplish the same thing in a more controlled
fashion, by simply establishing conventions - for example, DI::inject(new
My Class()) would work just as well, it's just inconvenient to type; on the
other hand, it's explicit about what's happening.

So I'm not at all demanding this feature - I'm merely posting this for
debate.

Thanks,
  Rasmus Schultz

--20cf307c9afe4405b804c9f8e6e8--