Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:8406
Mailing-List: contact internals-help@lists.php.net; run by ezmlm
To: internals@lists.php.net
In-Reply-To: <5.1.0.14.2.20040304140729.02098198@127.0.0.1>
References: <1078074613.1440.9.camel@localhost>
	 <1078074613.1440.9.camel@localhost>
	 <5.1.0.14.2.20040304140729.02098198@127.0.0.1>
Content-Type: text/plain
Message-ID: <1078584828.266.48.camel@localhost>
Mime-Version: 1.0
Date: Sat, 06 Mar 2004 15:53:49 +0100
Content-Transfer-Encoding: 7bit
Subject: Re: [PHP-DEV] BC break: $o= new stdclass; empty($o)
From: thekid@thekid.de (Timm Friebe)

On Thu, 2004-03-04 at 13:10, Andi Gutmans wrote:
> Hey,
Hi,

> The reason for this behavior is that in PHP 3 and 4, objects were treated 
> very much like arrays (this is the main reason for redesigning the object 
> model for PHP 5).
> I don't think that the old behavior is correct and an empty should always 
> return true (i.e. empty($var) should return true if the variable is NULL 
> and false if $var is an object).

to clarify: Maybe I was misusing PHP by doing the following:

  class null {
    function null() { }

    function __call($name, $args) {
      throw(new NullPointerException('Method.invokation('.$name.')'));
    }

    function __set($name, $value) {
      throw(new NullPointerException('Property.write('.$name.')'));
    }

    function __get($name) {
      throw(new NullPointerException('Property.read('.$name.')'));
    }
  }

  class registry {
    public static $null;
  }

  registry::$null= new null();

  class ClassType {

    function forName($name) {
      // ... abbreviated ...
      return new ClassType($name);
    }

    function getConstructor() {
      if (!$this->hasConstructor()) return registry::$null;
      return $this->constructor;
    }
  }
 
  $i= ClassType::forName($class)->getConstructor()->newInstance();

What I am doing here is fatal error prevention. To come to my point: The
following statement:

  if (ClassType::forName($class)->getConstructor()) {
    // ...
  }

would now always evaluate to (bool)true. In PHP4 (where, of course, the
syntax need be a bit different) the above statement would evaluate to
(bool)false due to the "emptiness" of the null object.

I know this is kind of hacky but was a nice way of being able to have a
way of implementing a "Fatal error: Call to a member function of a
non-object"-safe "NULL" *and* being able to check for it using boolean
operators.

> I don't think we should by default support the PHP 4 behavior because it's 
> just not right. What I suggest is to either break BC completely or support 
> the old behavior in compatibility mode.
> What option do you prefer?

I saw you already committed a patch for the latter option (return 0 in
compat mode).

I'm not sure whether this affects a lot of users, and as noone has
commented on it, I guess PHP5 will get away with this break:)

I'll just have to find a workaround if you insist the PHP4 behaviour is
broken:)

- Timm