Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:32906
Mailing-List: contact internals-help@lists.php.net; run by ezmlm
Received-SPF: pass (pb1.pair.com: domain horde.org designates 66.92.78.250 as permitted sender)
Message-ID: <20071021203755.11105vn0ycds6bs4@technest.org>
Date: Sun, 21 Oct 2007 20:37:55 -0400
To: Stanislav Malyshev <stas@zend.com>
Cc: internals@lists.php.net
References: <20071019224202.20245u5zry52h5c8@technest.org>
	<471A6B32.3080401@zend.com> <20071020172658.1680416b83wxgv6s@technest.org>
	<471A7C88.6090904@zend.com> <20071020184554.69044keujpgfelus@technest.org>
	<471A8A12.3050407@zend.com> <20071020191948.19935z9hoqcn9c84@technest.org>
	<471A9049.1020408@zend.com> <20071020194217.98794nbyt3bwrlnk@technest.org>
	<471AEF26.5000601@zend.com> <20071021105124.85283gxiqqx9zse8@technest.org>
	<471BA74C.1040103@zend.com>
In-Reply-To: <471BA74C.1040103@zend.com>
MIME-Version: 1.0
Content-Type: text/plain;
	charset=ISO-8859-1;
	DelSp="Yes";
	format="flowed"
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable
User-Agent: Internet Messaging Program (IMP) H3 (4.2-cvs)
Subject: Re: [PHP-DEV] Order of
	class	resolution	with	namespaces	and	autoload
From: chuck@horde.org (Chuck Hagenbuch)

Quoting Stanislav Malyshev <stas@zend.com>:

> I'm not sure I follow you. If you intend to always use internal one, =20
> why you define the local one? If you sometimes intend to use one and =20
> sometimes another - I propose you a way to ensure the right one is =20
> always used, this without using potentially long full name. I must =20
> be missing something here - can you specify what exactly you want it =20
> to be? You do understand that if you insist on always using =20
> unqualified name, it must work some consistent way, and this way can =20
> not satisfy all possible scenarios because they are contradictory?

Yes. I think that if you use an unqualified name it should always be =20
relative to the namespace (and importing internal classes into your =20
namespace lets you use short names for them, avoiding ::Exception).

>> This unpredictability is I think the worst part of the current =20
>> implementation.
>
> There's no "unpredictability" - the behavior is entirely =20
> predictable, you just don't like it :) That's like saying that =20
> function call is unpredictable, because if you don't define the =20
> function it produces an error.

Here is what I mean:

1.php
-----

<?php

namespace Test;
import Test::Exception;

throw new Exception();


2.php
-----

<?php

namespace Test;

throw new Exception();


parent.php
----------

<?php

function __autoload($class) {
     include './test_exception.php';
}

try {
     include '2.php';
} catch (Exception $e) {
     echo get_class($e) . "\n";
}

try {
     include '1.php';
} catch (Exception $e) {
     echo get_class($e) . "\n";
}



test_exception.php defines Test::Exception.

If you run parent.php as written above, the first exception will be of =20
type Exception and the 2nd will be of type Test::Exception. However if =20
you edit parent.php so that 1.php is included before 2.php, then both =20
exceptions will be of type Test::Exception.

To me, this is unpredictable because include order changes how =20
otherwise identical code behaves.

I can solve this in my case by always writing import Test::Exception;. =20
However, if I really want to use the builtin Exception class, it seems =20
like I _must_ type it as ::Exception every time - which is something =20
you wanted to avoid as well as I.

Other options I tried:

import Exception;
-> name conflict, which seems correct

import Exception as Error;
-> Fatal error: Import name 'Error' conflicts with defined class in =20
/Users/chuck/Desktop/php namespaces/2.php on line 4
(this I don't understand)

import ::Exception as Error;
-> parse error (can't import a top-level class name that way)


Therefore, in my mind, in order to write durable code that will not =20
break no matter what other classes other developers define or import, =20
I should always prefix builtin classes with ::.

Given this, the only difference between how I think things should work =20
is that I think that, when I have declared a namespace, I should be =20
able to rely on any short name referring to that namespace or one of =20
my explicit imports. The key being that _I_ declared that namespace =20
and the imports and they will affect only that file, so I am =20
completely in control of what name means what.

The way things currently are, using a shortname means look inside the =20
namespace, UNLESS there's a builtin class with the same short name and =20
the namespaced version hasn't been autoloaded yet, or there's an =20
explicit import...

I really think it's unintuitive that when you declare a namespace =20
names don't necessarily refer to that namespace. It doesn't match =20
other languages, either - you always need to import classes/names if =20
you want to use short names for functionality outside your =20
namespace/package/etc.

Sincerely,
-chuck