# Static weirdness..

- Thread: https://externals.io/message/7946
- Original archive: https://news-web.php.net/php.internals/7946
- Started by John Coggeshall on 2004-02-18T07:53:50+00:00
- Messages: 41
- Page 1 of 2: messages 1 to 25 (25 per page, add `?page=N` to the URL)
- Next page: https://externals.io/message/7946.md?page=2
- Reply by email to internals@lists.php.net
- More for agents: search with https://externals.io/search.md?q={query} · JSON API at https://externals.io/api · documentation at https://externals.io/llms.txt

---

## #7946 — John Coggeshall — 2004-02-18T07:53:50+00:00

URL: https://externals.io/message/7946#7946

http://bugs.php.net/bug.php?id=27304

Marcus says he's brought this up before, and i think it really needs to
be addressed before PHP 5 so I'm bringing it up again. I am told that
currently we are allowing static methods to be called from an object
context because of a performance hit if we check every call, but
currently because $this is undefined regardless of context there is no
way even for the developer to check if the method was called properly. 

John

-- 
-=~=--=~=--=~=--=~=--=~=--=~=--=~=--=~=--=~=--=~=--=~=--=~=-
John Coggeshall                   http://www.coggeshall.org/
The PHP Developer's Handbook    http://www.php-handbook.com/
-=~=--=~=--=~=--=~=--=~=--=~=--=~=--=~=--=~=--=~=--=~=--=~=

---

## #7949 — Pierre-Alain Joye — 2004-02-18T10:54:21+00:00

URL: https://externals.io/message/7946#7949 · In reply to: #7946

On Wed, 18 Feb 2004 02:53:50 -0500
john@coggeshall.org (John Coggeshall) wrote:

> http://bugs.php.net/bug.php?id=27304
> 
> Marcus says he's brought this up before, and i think it really needs
> to be addressed before PHP 5 so I'm bringing it up again. I am told
> that currently we are allowing static methods to be called from an
> object context because of a performance hit if we check every call,
> but currently because $this is undefined regardless of context there
> is no way even for the developer to check if the method was called
> properly. 

imho, this is the expected behavior. At least a notice should be raised.
I do not see a reason to define $this if a method is explicitly defined
as static (means it should not be called dynamically). Declare the
same funciton as public|private|protected and everything works fine.

pierre

---

## #7959 — John Coggeshall — 2004-02-18T17:14:04+00:00

URL: https://externals.io/message/7946#7959 · In reply to: #7949

> imho, this is the expected behavior. At least a notice should be raised.
> I do not see a reason to define $this if a method is explicitly defined
> as static (means it should not be called dynamically). Declare the
> same funciton as public|private|protected and everything works fine.

There is no reason I can possibly imagine why methods declared as static
should be allowed to be called from a non-static context. Sure, in my
own classes I know not to call a method I declared as static from an
instance. But IMO the whole point of static is to create a restriction
in then engine keeping from anyone calling that function in anything
other than a static context. 

class foo {
	static function bar() {
		echo "Hello.\n";
	}
	
	function foobar();
}

$a = new foo;
$b->bar();

This is a completely acceptable thing, and it is IMO completely wrong to
even allow in the engine. foo::bar() should be the only accepted syntax
for static functions. If I download a new class from some site, or
otherwise use someone else's code, how do I know other than digging
through the source on a method's context? If $this isn't set I can't do
the check, and if then engine doesn't do it there isn't a way to tell
the difference.


-- 
-=~=--=~=--=~=--=~=--=~=--=~=--=~=--=~=--=~=--=~=--=~=--=~=-
John Coggeshall                   http://www.coggeshall.org/
The PHP Developer's Handbook    http://www.php-handbook.com/
-=~=--=~=--=~=--=~=--=~=--=~=--=~=--=~=--=~=--=~=--=~=--=~=

---

## #7960 — George Schlossnagle — 2004-02-18T17:21:12+00:00

URL: https://externals.io/message/7946#7960 · In reply to: #7959

> This is a completely acceptable thing, and it is IMO completely wrong 
> to
> even allow in the engine. foo::bar() should be the only accepted syntax
> for static functions.

Maybe I'm having a slow-English day, but how is the first sentence here 
is not intrinsically self-contradictory?

George

---

## #7961 — John Coggeshall — 2004-02-18T17:23:04+00:00

URL: https://externals.io/message/7946#7961 · In reply to: #7960

> Maybe I'm having a slow-English day, but how is the first sentence here 
> is not intrinsically self-contradictory?

:) 

class foo {
	static function bar() {
	}
}
$a = new foo();
$a->bar(); /* Unacceptable and contradictory to the concept of static */
foo::bar(); /* Acceptable */

John
-- 
-=~=--=~=--=~=--=~=--=~=--=~=--=~=--=~=--=~=--=~=--=~=--=~=-
John Coggeshall                   http://www.coggeshall.org/
The PHP Developer's Handbook    http://www.php-handbook.com/
-=~=--=~=--=~=--=~=--=~=--=~=--=~=--=~=--=~=--=~=--=~=--=~=

---

## #7962 — Pierre-Alain Joye — 2004-02-18T17:30:12+00:00

URL: https://externals.io/message/7946#7962 · In reply to: #7961

On Wed, 18 Feb 2004 12:23:04 -0500
John Coggeshall <john@coggeshall.org> wrote:

> class foo {
> 	static function bar() {
> 	}
> }
> $a = new foo();
> $a->bar(); /* Unacceptable and contradictory to the concept of static
> */ foo::bar(); /* Acceptable */

I have the same problem as George :) I did not understand well your
original post.

We are talking about the same thing. A notice (error sounds too
drastic here) should be raised if a static method is called from the
instanciated object.

pierre

---

## #7963 — John Coggeshall — 2004-02-18T17:33:35+00:00

URL: https://externals.io/message/7946#7963 · In reply to: #7962

> We are talking about the same thing. A notice (error sounds too
> drastic here) should be raised if a static method is called from the
> instanciated object.

No. There is *no* reason why static methods should be called from an
object context. Doing so is more than a notice -- its flat out wrong and
defeats the entire purpose of having static in the first place... 

John

-- 
-=~=--=~=--=~=--=~=--=~=--=~=--=~=--=~=--=~=--=~=--=~=--=~=-
John Coggeshall                   http://www.coggeshall.org/
The PHP Developer's Handbook    http://www.php-handbook.com/
-=~=--=~=--=~=--=~=--=~=--=~=--=~=--=~=--=~=--=~=--=~=--=~=

---

## #7964 — Pierre-Alain Joye — 2004-02-18T17:34:32+00:00

URL: https://externals.io/message/7946#7964 · In reply to: #7959

On Wed, 18 Feb 2004 12:14:04 -0500
John Coggeshall <john@coggeshall.org> wrote:

> if I download a new class from some site, or otherwise use someone
> else's code, how do I know other than digging through the source on
> a method's context? If $this isn't set I can't do the check, and if
> then engine doesn't do it there isn't a way to tell the difference.

To help John to do not read the sources ;-) (and many of us as well) it
should be nice if the reflection api returns this infos in getMethods()
result. For now only the methods name and the class name are returned.
If it's possible.

pierre

---

## #7966 — Pierre-Alain Joye — 2004-02-18T17:38:28+00:00

URL: https://externals.io/message/7946#7966 · In reply to: #7963

On Wed, 18 Feb 2004 12:33:35 -0500
John Coggeshall <john@coggeshall.org> wrote:

> No. There is *no* reason why static methods should be called from an
> object context. Doing so is more than a notice -- its flat out wrong
> and defeats the entire purpose of having static in the first place... 

I only say that a error (fatal or no) may be not the good choice. But I
do not care that much as far as there is at least a notice :)

pierre

---

## #7967 — Hans Lellelid — 2004-02-18T17:40:03+00:00

URL: https://externals.io/message/7946#7967 · In reply to: #7962

Hi -

Pierre-Alain Joye wrote:
> On Wed, 18 Feb 2004 12:23:04 -0500
> John Coggeshall <john@coggeshall.org> wrote:
> 
> > […]
> 
> 
> I have the same problem as George :) I did not understand well your
> original post.
> 
> We are talking about the same thing. A notice (error sounds too
> drastic here) should be raised if a static method is called from the
> instanciated object.
> 

I disagree w/ any sort of error being raised by calling a static method 
from object context.  I think it should be allowed because

1) there is already the stuff in static to disallow any reference to 
$this, so there is no danger of a static method messing w/ instance 
variables. (i.e. if you're calling it by accident, you'll probably find 
out pretty quickly)

2) In some cases it is useful to instantiate an otherwise static class 
for the sake of passing it to some code that doesn't know the name of 
the static class.

For example in Propel (PHP5 Torque port) the "Peer" classes are static 
classes that do work on entities -- like inserting them in the database, 
etc.  If I have an array of Entity objects, I might want to get access 
to the Peer class that handles those entities, so I have:

foreach($entities as $ent) {
	$peer = $ent->getPeer(); // e.g. might be BookPeer
	$peer->doSelect(new Criteria()); //static: BookPeer::doSelect()
}

This [taken from Java Torque impl] is particularly useful because the 
alternative is having to invoke the method using `call_user_func()` or 
eval() (yuk!).

Hans

---

## #7980 — Greg Beaver — 2004-02-19T00:38:17+00:00

URL: https://externals.io/message/7946#7980 · In reply to: #7946

Hi,

I agree that allowing $a->bar() with a static method is too confusing, 
and should not be allowed.  However, the ability to call a static method 
of an object (variable class name, in other words), is invaluable.  What 
if PHP simply allowed $object::staticMethod() syntax?

I did some playing around, and although I wouldn't know how to disable 
$a->staticMethod(), this patch enables $a::staticMethod().  It also has 
another side effect

<?php
class foo {
     static function bar()
     {
	echo 'hello';
     }
}

$bar = 'hello';
$a = new foo;
$a->bar(); // this could be disabled to error out easily
$a::bar(); // my patch now allows this to print "hello"
$a::$bar(); // this also works
$a::{'bar'}(); // even this works
?>

If someone could get the patch to work with a `T_STRING` instead of an 
object_property, then it would disable the $a::$bar() and 
$a::{expression}() syntax.

This is my first attempt at a patch of the core, please be gentle if it 
sucks. :)

Thanks,
Greg

John Coggeshall wrote:
> http://bugs.php.net/bug.php?id=27304
> 
> Marcus says he's brought this up before, and i think it really needs to
> be addressed before PHP 5 so I'm bringing it up again. I am told that
> currently we are allowing static methods to be called from an object
> context because of a performance hit if we check every call, but
> currently because $this is undefined regardless of context there is no
> way even for the developer to check if the method was called properly. 
> 
> John

---

## #7984 — Hartmut Holzgraefe — 2004-02-19T09:03:33+00:00

URL: https://externals.io/message/7946#7984 · In reply to: #7980

Greg Beaver wrote:
> Hi,
> 
> I agree that allowing $a->bar() with a static method is too confusing, 
> and should not be allowed.  However, the ability to call a static method 
> of an object (variable class name, in other words), is invaluable.  What 
> if PHP simply allowed $object::staticMethod() syntax?

Why do i (as a user of a class) have to know whether a member function
is static or not? Having two different calling conventions is what
appears to be confusing to me ...


-- 
Hartmut Holzgraefe  <hartmut@php.net

---

## #7985 — Lukas Smith — 2004-02-19T09:03:49+00:00

URL: https://externals.io/message/7946#7985 · In reply to: #7984

Hartmut Holzgraefe wrote:
> Greg Beaver wrote:
> 
> > […]
> 
> 
> Why do i (as a user of a class) have to know whether a member function
> is static or not? Having two different calling conventions is what
> appears to be confusing to me ...

I agree with Hartmut.

regards,
Lukas Smith
smith@backendmedia.com
_______________________________
   BackendMedia
   www.backendmedia.com
   berlin@backendmedia.com

   Linn Zwoch Smith GbR
   Pariser Str. 44
   D-10707 Berlin

   Tel +49 30 83 22 50 00
   Fax +49 30 83 22 50 07

---

## #7988 — Timm Friebe — 2004-02-19T13:31:55+00:00

URL: https://externals.io/message/7946#7988 · In reply to: #7967

> Hi -
[...]
> foreach($entities as $ent) {
> 	$peer = $ent->getPeer(); // e.g. might be BookPeer
> 	$peer->doSelect(new Criteria()); //static: BookPeer::doSelect()
> }

$m= new Reflection_Method($ent->getPeer(), 'doSelect');
$m->invoke(NULL, new Criteria());

/* untested */

- Timm

---

## #7991 — Marcus Boerger — 2004-02-19T14:49:42+00:00

URL: https://externals.io/message/7946#7991 · In reply to: #7984

Hello Hartmut,

Thursday, February 19, 2004, 10:03:33 AM, you wrote:

> Greg Beaver wrote:
> > […]

> Why do i (as a user of a class) have to know whether a member function
> is static or not? Having two different calling conventions is what
> appears to be confusing to me ...

A static and a non static member function are two absolute completley
different things. Anyway calling a static method from an instance as
$instance->staticmethod() or $instance::staticmethod() should be allowed.
I guess the latter is what Lukas and Hartmut meant.


-- 
Best regards,
 Marcus                            mailto:helly@php.net

---

## #7992 — Pierre-Alain Joye — 2004-02-19T14:55:41+00:00

URL: https://externals.io/message/7946#7992 · In reply to: #7991

On Thu, 19 Feb 2004 15:49:42 +0100
Marcus Boerger <helly@php.net> wrote:

> A static and a non static member function are two absolute completley
> different things. Anyway calling a static method from an instance as
> $instance->staticmethod() or $instance::staticmethod() should be
> allowed. I guess the latter is what Lukas and Hartmut meant.

I discussed with Lukas what he meant by "I agree". A lot of
misunderstanding in this thread :)

One sure thing, $this should not be set in static methods.

I do not care that much but I like to see a notice in `E_STRICT` mode if a
method explicitly declared as static is called from an instance of the
object.

About this topic, see my other post. It's a patch that add fn_flags
(method flags like public, private) to the getMethod() result.

pierre

---

## #7993 — Derick Rethans — 2004-02-19T15:02:10+00:00

URL: https://externals.io/message/7946#7993 · In reply to: #7992

> On Thu, 19 Feb 2004 15:49:42 +0100
> Marcus Boerger <helly@php.net> wrote:
>
> > […]
>
> I discussed with Lukas what he meant by "I agree". A lot of
> misunderstanding in this thread :)
>
> One sure thing, $this should not be set in static methods.

But it should get set in non-static methods...which doesn't happen now
:)

Derick

---

## #7994 — Hartmut Holzgraefe — 2004-02-19T15:04:43+00:00

URL: https://externals.io/message/7946#7994 · In reply to: #7991

Marcus Boerger wrote:
> A static and a non static member function are two absolute completley
> different things. 

sure they are (in certain ways), but when *calling* them (from a class
users point of view, not an implementors point of view) the only
difference is that the static member is guaranteed to not change
the state of an instance while a non-static member is not
besides that there is no difference for a caller/user of an instance
and forcing him to know which member functions are static and which
aren't just adds a level of inconvenience without gain

so IMHO

$instance->static() has to work
classname::static() obviously has to work, too
$instance::static() is optional

-- 
Hartmut Holzgraefe  <hartmut@php.net

---

## #7995 — Hartmut Holzgraefe — 2004-02-19T15:05:50+00:00

URL: https://externals.io/message/7946#7995 · In reply to: #7992

Pierre-Alain Joye wrote:
> One sure thing, $this should not be set in static methods.

ACK

-- 
Hartmut Holzgraefe  <hartmut@php.net

---

## #7996 — Pierre-Alain Joye — 2004-02-19T15:10:09+00:00

URL: https://externals.io/message/7946#7996 · In reply to: #7993

On Thu, 19 Feb 2004 16:02:10 +0100 (CET)
Derick Rethans <derick@php.net> wrote:

> But it should get set in non-static methods...which doesn't happen now
> :)

$this is set in non static methods if the method is called from an
instance of the object and not set if called statically. It's exactly
the same behavior as php4. That's why I like the notice on a non static
call of a static method, cleaner and no BC as the static methods did not
exist in php4.

See the tests script in the beginning of this thread.

hth

pierre

---

## #7997 — Greg Beaver — 2004-02-19T15:11:58+00:00

URL: https://externals.io/message/7946#7997 · In reply to: #7992

Pierre-Alain Joye wrote:

>One sure thing, $this should not be set in static methods.
>
I completely agree - the headache of fixing PEAR's "kind of static" 
methods is really annoying, and not really fixable.  A method should 
either be static, non-static, or split into two methods, one that is 
static, and one that is not.  Language-level enforcement of not allowing 
$this in static methods is very nice.

Grge

---

## #7998 — Hans Lellelid — 2004-02-19T15:14:53+00:00

URL: https://externals.io/message/7946#7998 · In reply to: #7988

Timm Friebe wrote:
> 
> 
> > […]
> 
> 
> $m= new Reflection_Method($ent->getPeer(), 'doSelect');
> $m->invoke(NULL, new Criteria());
> 
> /* untested */
> 

Yeah, I'm sure there's a way to get something like that to work also. 
I'm not sure that's prefereable to `call_user_func()`, honestly -- and it 
doesn't get around the basic problem which is that calling code needs to 
know whether methods are static.

I guess I'm not sure why this is necessary ...

As Hartmut & Lukas mention, I don't know why the class consumer *needs* 
to know that it's static.  Clearly internally there should be no 
possibility of referencing $this.

Also in my example the entire class is static -- i.e. all methods are 
static -- but what about cases where one or two methods in a class are 
static.  Now $a->methodName() will work for most but all of a sudden you 
have to do $a::staticMethodName() for the static members.

What I see as the workaround is creating static methods that aren't 
marked as static -- e.g. that would address the issue in my example 
above.  I think that's a lot worse than allowing invocation of static 
methods using instances, however, because I do want developers who read 
the API docs to know these are static functions.

Hans

---

## #7999 — Marcus Boerger — 2004-02-19T15:21:50+00:00

URL: https://externals.io/message/7946#7999 · In reply to: #7996

Hello Pierre-Alain,

Thursday, February 19, 2004, 4:10:09 PM, you wrote:

> On Thu, 19 Feb 2004 16:02:10 +0100 (CET)
> Derick Rethans <derick@php.net> wrote:

> > […]

> $this is set in non static methods if the method is called from an
> instance of the object and not set if called statically. It's exactly
> the same behavior as php4. That's why I like the notice on a non static
> call of a static method, cleaner and no BC as the static methods did not
> exist in php4.

But why the hell do you want a message, calling a static method from
anywhere is absolutley fine. This of course includes calling a static
method of an instance's class, however that is syntactically accomplished.

I heared pear already mentioned in this thread. That cannot be of any
reason with the ongoing effort to make statics work. First as i heared
most of pear doesn't run on php5 and second pear of today uses php4 which
didn't support statics and lots of other oo goodies. So it clear that pear
has some bc problems when moving to version where all the workarounds are
now solved by syntax.


Best regards,
 Marcus                            mailto:helly@php.net

---

## #8000 — Marcus Boerger — 2004-02-19T15:24:55+00:00

URL: https://externals.io/message/7946#8000 · In reply to: #7998

Hello Hans,

Thursday, February 19, 2004, 4:14:53 PM, you wrote:


> Timm Friebe wrote:
> > […]

> Yeah, I'm sure there's a way to get something like that to work also. 
> I'm not sure that's prefereable to `call_user_func()`, honestly -- and it 
> doesn't get around the basic problem which is that calling code needs to 
> know whether methods are static.

> I guess I'm not sure why this is necessary ...

> As Hartmut & Lukas mention, I don't know why the class consumer *needs* 
> to know that it's static.  Clearly internally there should be no 
> possibility of referencing $this.

> Also in my example the entire class is static -- i.e. all methods are 
> static -- but what about cases where one or two methods in a class are 
> static.  Now $a->methodName() will work for most but all of a sudden you 
> have to do $a::staticMethodName() for the static members.

> What I see as the workaround is creating static methods that aren't 
> marked as static -- e.g. that would address the issue in my example 
> above.  I think that's a lot worse than allowing invocation of static 
> methods using instances, however, because I do want developers who read 
> the API docs to know these are static functions.

I hope nobody wants such strange things. Calling a static method from
an instance is just one of the ways to get to the class of that method.


-- 
Best regards,
 Marcus                            mailto:helly@php.net

---

## #8001 — Pierre-Alain Joye — 2004-02-19T15:28:50+00:00

URL: https://externals.io/message/7946#8001 · In reply to: #7999

On Thu, 19 Feb 2004 16:21:50 +0100
Marcus Boerger <helly@php.net> wrote:

> But why the hell do you want a message, calling a static method from
> anywhere is absolutley fine. This of course includes calling a static
> method of an instance's class, however that is syntactically
> accomplished.

I think it's cleaner, but as said, I do not care. There is no chance to
get a compromise on that as it is the case in most OO topic here, too
much OO pov and fanatics ;-))

However I like to get the flags from getMethods :)


> I heared pear already mentioned in this thread. That cannot be of any
> reason with the ongoing effort to make statics work. First as i heared
> most of pear doesn't run on php5 and second pear of today uses php4
> which didn't support statics and lots of other oo goodies. So it clear
> that pear has some bc problems when moving to version where all the
> workarounds are now solved by syntax.

There is no problem. PEAR runs with PHP5 (I use only php5 these days)
And PEAR does not lead PHP devs but the counter, imho.

pierre

---

Next page: https://externals.io/message/7946.md?page=2