Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:16197
Mailing-List: contact internals-help@lists.php.net; run by ezmlm
Reply-To: lpedretti@suserver.com
Organization: Axon Sistemas
To: Rory Browne <rory.browne@gmail.com>
Date: Tue, 10 May 2005 10:46:19 -0300
User-Agent: KMail/1.7.2
References: <200505091130.56300.lpedretti@suserver.com> <427FA7A5.6050803@iamjochem.com> <fa07888a05050916286059d049@mail.gmail.com>
In-Reply-To: <fa07888a05050916286059d049@mail.gmail.com>
Cc: internals@lists.php.net
MIME-Version: 1.0
Content-Type: text/plain;
	charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable
Content-Disposition: inline
Message-ID: <200505101046.19099.lpedretti@suserver.com>
Subject: Re: [PHP] Re: [PHP-DEV] make new return a reference to existent object
From: lpedretti@suserver.com (Leonardo Pedretti)

It would not be such a problem to write something that does the same task=20
through a static or global function, but unfortunately it is impossible, du=
e=20
to the following: you can build a cache in an array, then in a __new()=20
function you could check if the requested object is already loaded and retu=
rn=20
a reference. However, if this is the case, objects that enter the cache nev=
er=20
leave it because they are still referenced in the array when all other=20
references have gone, so they never destruct, except when the array is=20
destroyed at program's end.
So, the 'cache' would grow to include all possible objects, that's not good.
Now, the question is not if there should be an 'existing' keyword for=20
replacing singletons... singletons are ok, but there may be some=20
circumstances when one would like to return some already existing instance =
of=20
some class (between other existing instances, determined by the $id member=
=20
for example) and get the destructor executed when all references to that=20
instance are gone. However, it is not possible to do it because there is no=
t=20
any way to detect the removal of a reference to an object instance, except=
=20
implementing specific (and really ugly) methods for assigning and deassigni=
ng=20
the variables, such as __new(&$var) and __release(&$var) functions that do=
=20
the task of incrementing and decrementing a reference count for the object=
=20
and remove it from the cache when the refcount reaches 0. The main=20
disadvantage of this is that any request for an object of the mentioned cla=
ss=20
should be matched with a __release() function much like in a C++ program=20
using new and delete.
Is there any way of checking the existing instances of a class and returnin=
g=20
the same instance in a function without implementing that cache system, so=
=20
when all references are gone it naturally gets destroyed by the normal and=
=20
natural php code flux?

Leo


On Monday 09 May 2005 20:28, Rory Browne wrote:
> I think Leonardo might have a point here. Built-in support for
> Singletons could be a nice feature, without the use of static
> functions, would be nice. I wouldn't use the 'new' keyword however. I
> think something along the lines of existing, or ref
>
> $ref =3D existing TEST; // which would return an existing instance of
> TEST, if one exists, and a new one if it doesn't. ( pretty much like
> the singleton code, except no need for singleton code)
>
> I don't think it's anything important, and there are more pressing
> matters, to be worked on, I'm sure, but if someone created the code,
> and it was secure / bug-free / compliant-with-coding-standards / etc,
> I can't see why it couldn't(or shouldn't) be included. When all's said
> and done though, it's still just syntactic sugar. It all depends on
> how sweet the devs consider it.
>
> On 5/9/05, Jochem Maas <jochem@iamjochem.com> wrote:
> > Leonardo Pedretti wrote:
> > > I would like (for code cleanliness purposes) to make 'new' return a
> > > reference to an already created object under certain circumstances
> > > without using a factory, is it possible?
> >
> > not unless you hack the php engine (in which case
> > your code will only work on your custom php build),
> > I'll bet money that none of the php/zend devs feel anything for
> > introducing this kind of 'magic' into engine.
> >
> > IMHO a factory would be a clean method of handling this behaviour and
> > has the benefit that no programmer looking at your code will
> > misinterpret occurances of the 'new' keyword. e.g.:
> >
> > class Test
> > {
> >         /* ... */
> >         function __construct() {}
> >         function get() {}
> > }
> >
> > // and do:
> >
> > $var =3D Test::get( /* pass ctor args */ );
> >
> > // instead of:
> >
> > $var =3D new Test( /* pass ctor args */ );
> >
> > ... in this example its only 1 extra char to type when
> > using the factory method and you could turn that around by
> > doing something evil(tm) like:
> >
> > $var =3D Test::_( /* pass ctor args */ ); // '_' is a function name!!! =
(ala
> > wordpress-CS)
> >
> > maybe reply to generals describing what you are (trying to) do[ing],
> > I for one am always interested in other people's realworld php5/oo
> > problems/ideas/etc :-)
> >
> > rgds,
> > Jochem
> >
> > ps - I redirected this reply to generals because I didn't really think
> > that it was an internals question (and those guys are busy enough :-)
> >
> > > Thanx
> >
> > --
> > PHP General Mailing List (http://www.php.net/)
> > To unsubscribe, visit: http://www.php.net/unsub.php

=2D-=20

Leonardo Pedretti
Axon Sistemas
L=EDder de Equipo Proyecto Basalto