# [RFC] Anonymous functions

- Thread: https://externals.io/message/34631
- Original archive: https://news-web.php.net/php.internals/34631
- Started by Ryusuke SEKIYAMA on 2008-01-10T10:09:26+00:00
- Messages: 21
- 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

---

## #34631 — Ryusuke SEKIYAMA — 2008-01-10T10:09:26+00:00

URL: https://externals.io/message/34631#34631

Hello, lists,

We have discussed about implementing anonymous functions and
closures in PHP.
However, I consider that implementing anonymous functions and
implementing lexical scopes should be discussed separately.
# Even though I like closure. ;-)
So I wrote anonymous function patch for PHP 5.3 and 6.0.


Patches (inlcude tests):
for PHP 5.3: http://www.opendogs.org/pub/php-5.3dev-080109-anon.patch
for PHP 6.0: http://www.opendogs.org/pub/php-6.0dev-080109-anon.patch


Featues:
- Unlike create_function(), there is no need to take care of
  quotes, backslashes and dollars .
- Can be used in loop, but be compiled only once.
- Supports recursive call using __FUNCTION__.
- Supports inline execution. It works like a block scope.
- Works with opcode caches.
  Since I couldn't build APC against PHP 5.3, I have tested
  PHP 5.2.5 + anonymous function patch and APC 3.0.16.


Example:
<?php
// callback
$arr = array(5, 3, 6, 0);
usort($arr, function($a, $b){ return $a - $b; });
print_r($arr); // 0, 3, 5, 6

// assign to a variable
$lambda = function(){
  echo "Hello, World!\n";
};
$lambda(); // "Hello, World!"

// inline execution
$result = function($a, $b){
  return $a + $b;
}(1, 2);
var_dump($result); // "int(3)"
?>


The anonymous function name is composed of '\0', "ZEND_ANON<#serial>",
filename and character position.
- Leading '\0' blocks to be displayed by `get_defined_functions()`.
- "ZEND_ANON<" helps determine wheter the function is anonymous or not
  because '<' cannot be used for user-defined function name.
- Serial number and file informations make the name unique.
- char anon_key_buf[32] is long enough because
  `snprintf(anon_key_buf, 32, "ZEND_ANON<%llu>", (unsigned long
long)UINT64_MAX)'
  returns 31.


Regards,


2008/1/6, Marcus Boerger <helly@php.net>:
> Hello Stanislav,
>
>   tha makesw three then already, how about we ask around again?
> Ryusuke, can you please start a new '[RFC] Square brackets shortcut' thread
> to collect opinions and pass along the patch for that?
>
> I like the anonymous function patch too. It is clean and simple. Maybe you
> want to start a second '[RFC] Anonymous functions' thread with that patch.
>
> Can you also please add tests for both?
>
> marcus
>
> Wednesday, January 2, 2008, 7:51:06 PM, you wrote:
>
> > […]
>
> > […]
>
>
>
> Best regards,
>  Marcus
>
>


-- 
/**
 * Ryusuke SEKIYAMA
 * rsky0711@gmail.com
 */

---

## #34633 — Marcus Boerger — 2008-01-10T10:20:53+00:00

URL: https://externals.io/message/34631#34633 · In reply to: #34631

Hello Ryusuke,

  +1, looks pretty nice, makes my life easier in some cases and is closer to
  other languages like pythin so i can transfer concepts easily which is a
  big plus for me.

marcus

Thursday, January 10, 2008, 11:09:26 AM, you wrote:

> Hello, lists,

> We have discussed about implementing anonymous functions and
> closures in PHP.
> However, I consider that implementing anonymous functions and
> implementing lexical scopes should be discussed separately.
> # Even though I like closure.  
> So I wrote anonymous function patch for PHP 5.3 and 6.0.


> Patches (inlcude tests):
> for PHP 5.3: http://www.opendogs.org/pub/php-5.3dev-080109-anon.patch
> for PHP 6.0: http://www.opendogs.org/pub/php-6.0dev-080109-anon.patch


> Featues:
> - Unlike create_function(), there is no need to take care of
>   quotes, backslashes and dollars .
> - Can be used in loop, but be compiled only once.
> - Supports recursive call using __FUNCTION__.
> - Supports inline execution. It works like a block scope.
> - Works with opcode caches.
>   Since I couldn't build APC against PHP 5.3, I have tested
>   PHP 5.2.5 + anonymous function patch and APC 3.0.16.


> Example:
> <?php
> // callback
> $arr = array(5, 3, 6, 0);
> usort($arr, function($a, $b){ return $a - $b; });
> print_r($arr); // 0, 3, 5, 6

> // assign to a variable
> $lambda = function(){
>   echo "Hello, World!\n";
> };
> $lambda(); // "Hello, World!"

> // inline execution
> $result = function($a, $b){
>   return $a + $b;
> }(1, 2);
> var_dump($result); // "int(3)"
?>>


> The anonymous function name is composed of '\0', "ZEND_ANON<#serial>",
> filename and character position.
> - Leading '\0' blocks to be displayed by `get_defined_functions()`.
> - "ZEND_ANON<" helps determine wheter the function is anonymous or not
>   because '<' cannot be used for user-defined function name.
> - Serial number and file informations make the name unique.
> - char anon_key_buf[32] is long enough because
>   `snprintf(anon_key_buf, 32, "ZEND_ANON<%llu>", (unsigned long
> long)UINT64_MAX)'
>   returns 31.


> Regards,


> 2008/1/6, Marcus Boerger <helly@php.net>:
> > […]


> -- 
> /**
>  * Ryusuke SEKIYAMA
>  * rsky0711@gmail.com
>  */




Best regards,
 Marcus

---

## #34637 — Richard Quadling — 2008-01-10T12:10:08+00:00

URL: https://externals.io/message/34631#34637 · In reply to: #34633

> Hello Ryusuke,
>
>   +1, looks pretty nice, makes my life easier in some cases and is closer to
>   other languages like pythin so i can transfer concepts easily which is a
>   big plus for me.
>
> marcus
>
> Thursday, January 10, 2008, 11:09:26 AM, you wrote:
>
> > […]
>
> > […]
>
>
> > […]
>
>
> > […]
>
>
> > […]
>
> > […]
>
> > […]
> ?>>
>
>
> > […]
>
>
> > […]
>
>
> > […]
>
>
> > […]
>
>
>
>
> Best regards,
>  Marcus
>
> --
> 
> 
>
>

A nice feature. Especially for array callbacks.





-- 
-----
Richard Quadling
Zend Certified Engineer : http://zend.com/zce.php?c=ZEND002498&r=213474731
"Standing on the shoulders of some very clever giants!"

---

## #35006 — KOYAMA Tetsuji — 2008-01-29T18:26:21+00:00

URL: https://externals.io/message/34631#35006 · In reply to: #34631

Hi lists,

Is this discussion stopping?

I want this feature in PHP. Please start to talk.


> Hello, lists,
>
> We have discussed about implementing anonymous functions and
> closures in PHP.
> However, I consider that implementing anonymous functions and
> implementing lexical scopes should be discussed separately.
> # Even though I like closure. ;-)
> So I wrote anonymous function patch for PHP 5.3 and 6.0.
>
>
> Patches (inlcude tests):
> for PHP 5.3: http://www.opendogs.org/pub/php-5.3dev-080109-anon.patch
> for PHP 6.0: http://www.opendogs.org/pub/php-6.0dev-080109-anon.patch
>
>
> Featues:
> - Unlike create_function(), there is no need to take care of
>  quotes, backslashes and dollars .
> - Can be used in loop, but be compiled only once.
> - Supports recursive call using __FUNCTION__.
> - Supports inline execution. It works like a block scope.
> - Works with opcode caches.
>  Since I couldn't build APC against PHP 5.3, I have tested
>  PHP 5.2.5 + anonymous function patch and APC 3.0.16.
>
>
> Example:
> <?php
> // callback
> $arr = array(5, 3, 6, 0);
> usort($arr, function($a, $b){ return $a - $b; });
> print_r($arr); // 0, 3, 5, 6
>
> // assign to a variable
> $lambda = function(){
>  echo "Hello, World!\n";
> };
> $lambda(); // "Hello, World!"
>
> // inline execution
> $result = function($a, $b){
>  return $a + $b;
> }(1, 2);
> var_dump($result); // "int(3)"
> ?>
>
>
> The anonymous function name is composed of '\0', "ZEND_ANON<#serial>",
> filename and character position.
> - Leading '\0' blocks to be displayed by `get_defined_functions()`.
> - "ZEND_ANON<" helps determine wheter the function is anonymous or not
>  because '<' cannot be used for user-defined function name.
> - Serial number and file informations make the name unique.
> - char anon_key_buf[32] is long enough because
>  `snprintf(anon_key_buf, 32, "ZEND_ANON<%llu>", (unsigned long
> long)UINT64_MAX)'
>  returns 31.
>
>
> Regards,
>
>
> 2008/1/6, Marcus Boerger <helly@php.net>:
> > […]
>
>
> --
> /**
>  * Ryusuke SEKIYAMA
>  * rsky0711@gmail.com
>  */
>
> --
> 
> 
>
>

-
KOYAMA, Tetsuji
koyama@hoge.org

---

## #35559 — Matvey Arye — 2008-02-18T16:26:33+00:00

URL: https://externals.io/message/34631#35559 · In reply to: #35006

Hi All,

I second this. Can we please re-open the discussion on anonymous 
functions as well as closures. That would be an awesome feature. 
create_function is a bit ugly semantically to be sufficient for all 
anonymous function needs.

Thanks,
Mat

KOYAMA Tetsuji wrote:
> Hi lists,
>
> Is this discussion stopping?
>
> I want this feature in PHP. Please start to talk.
>
> 
>   
> > […]
>
> -
> KOYAMA, Tetsuji
> koyama@hoge.org

---

## #35743 — Andrei Zmievski — 2008-02-22T18:00:11+00:00

URL: https://externals.io/message/34631#35743 · In reply to: #35559

Yes, let's re-open, I'd like to get a closure on that issue (pun  
intended).

-Andrei



> Hi All,
>
> I second this. Can we please re-open the discussion on anonymous  
> functions as well as closures. That would be an awesome feature.  
> create_function is a bit ugly semantically to be sufficient for all  
> anonymous function needs.
>
> Thanks,
> Mat
>
> KOYAMA Tetsuji wrote:
> > […]

---

## #35747 — Marcus Boerger — 2008-02-23T11:43:46+00:00

URL: https://externals.io/message/34631#35747 · In reply to: #35743

Hello Andrei,

  +1

Friday, February 22, 2008, 7:00:11 PM, you wrote:

> Yes, let's re-open, I'd like to get a closure on that issue (pun  
> intended).

> -Andrei

> 

> > […]
[...]
> > […]
[...]
> > […]


Best regards,
 Marcus

---

## #35748 — dz@bitxtender.com — 2008-02-23T11:50:02+00:00

URL: https://externals.io/message/34631#35748 · In reply to: #35747

And +1 from me :)


David



Am 23.02.2008 um 12:43 schrieb Marcus Boerger:

> Hello Andrei,
>
>  +1
>
> Friday, February 22, 2008, 7:00:11 PM, you wrote:
>
> > […]
>
> > […]
>
> > […]
>
> > […]
> [...]
> > […]
> [...]
> > […]
>
>
> Best regards,
> Marcus

---

## #35749 — Lars Strojny — 2008-02-23T12:50:18+00:00

URL: https://externals.io/message/34631#35749 · In reply to: #35743

Hi Andrei,

Am Freitag, den 22.02.2008, 10:00 -0800 schrieb Andrei Zmievski:
> Yes, let's re-open, I'd like to get a closure on that issue (pun  
> intended).

+1. I would love it.

cu, Lars
-- 
Lars Strojny
Senior Software Developer Media Ventures GmbH (http://mediaventures.de)

---

## #35751 — Andrei Zmievski — 2008-02-23T16:09:18+00:00

URL: https://externals.io/message/34631#35751 · In reply to: #35747

Do we have a working patch somewhere then?

-Andrei



> Hello Andrei,
>
>   +1
>
> Friday, February 22, 2008, 7:00:11 PM, you wrote:
>
> > […]
>
> > […]

---

## #35752 — Pierre Joye — 2008-02-23T16:13:26+00:00

URL: https://externals.io/message/34631#35752 · In reply to: #35751

> Do we have a working patch somewhere then?

Did we not get one from Wez? and an alternative from <I don't remember> :)
-- 
Pierre
http://blog.thepimp.net | http://www.libgd.org

---

## #35753 — Tobias Tom — 2008-02-23T17:32:31+00:00

URL: https://externals.io/message/34631#35753 · In reply to: #35752

Am 23.02.2008 um 17:13 schrieb Pierre Joye:
> On Sat, Feb 23, 2008 at 5:09 PM, Andrei Zmievski <andrei@gravitonic.com 
> > […]
>
> Did we not get one from Wez? and an alternative from <I don't  
> remember> :)

http://news.php.net/php.internals/34216

I think the last version should be http://www.christian-seiler.de/temp/closures-php-5-3-v2.patch 
. AFAIK he has planned to improve the patch from user feedback, but  
I'm not exactly sure about that.

Tobias

---

## #35756 — Ryusuke SEKIYAMA — 2008-02-23T18:46:34+00:00

URL: https://externals.io/message/34631#35756 · In reply to: #35751

2008/2/24, Andrei Zmievski <andrei@gravitonic.com>:
> Do we have a working patch somewhere then?

Hi, I have added support for closures to my patches.

http://www.opendogs.org/pub/php-5.3-080223-anon.patch
http://www.opendogs.org/pub/php-6.0-080223-anon.patch
http://www.opendogs.org/pub/php-anon-tests-080223.tgz

Example:
<?php
function get_counter($i = 0)
{
    /**
     * 'static' modified anonymous function's
     *  uninitialized (or null) static variables
     *  are copied from the declared scope.
     */
    return static function(){
        static $i;
        return ++$i;
    };
}

$c1 = get_counter();
$c2 = get_counter(100);
var_dump($c1(), $c2());
var_dump($c1(), $c2());
var_dump($c1(), $c2());
?>

For recursive calling of anonymous functions and closures,
added function get_current_function_key().

<?php
$f = function($i = 0)
{
    echo "$i ";
    if (++$i > 10) {
        return;
    }
    $f = get_current_function_key();
    $f($i);
};

$f();
?>


And here is a Zend extension for PHP 5.2 which enables
anonymous functions, closures and square bracket arrays.

http://www.opendogs.org/pub/php_qiq-0.1.0pre.tgz

This has a few more features, but, at present, there is
no document...


Regards,

-- 
/**
 * Ryusuke SEKIYAMA
 * rsky0711@gmail.com
 */

---

## #35759 — Andi Gutmans — 2008-02-23T19:36:18+00:00

URL: https://externals.io/message/34631#35759 · In reply to: #35743

I think we have to be very clear whether we are going to provide just a
sexier notation for anonymous functions or closures (and if the latter
what the semantics are).
Just doing one and not figuring out the long term piece doesn't make
sense because we may end up having two completely distinct features
and/or feature creep over the next year which doesn't make much sense to
me.

I think in last discussion most ppl just preferred a sexier anonymous
function esp. as there weren't very clear/clean proposals for how
closures would work in PHP and what the benefits would be.

So I definitely suggest to have a discussion, maybe RFC based like
Stefan did but really decide once and for all on where we'd like to go
with this. We all know that no features map 1:1 to PHP and there are
always unique behaviors of our language and environment which require a
fresh view at things.

Andi

> -----Original Message-----
> From: Andrei Zmievski [mailto:andrei@gravitonic.com]
> Sent: Friday, February 22, 2008 10:00 AM
> To: Matvey Arye
> Cc: KOYAMA Tetsuji; internals@lists.php.net
> Subject: Re: [PHP-DEV] [RFC] Anonymous functions
> 
> Yes, let's re-open, I'd like to get a closure on that issue (pun
> intended).
> 
> -Andrei
> 
> 
> 
> > […]
> wrote:
> > […]

---

## #35765 — Andrei Zmievski — 2008-02-24T10:23:22+00:00

URL: https://externals.io/message/34631#35765 · In reply to: #35759

To be clear, my statement was about the closures and not sexier  
anonymous functions. I agree that we need to do an RFC on this matter.

-Andrei



> I think we have to be very clear whether we are going to provide  
> just a
> sexier notation for anonymous functions or closures (and if the latter
> what the semantics are).
> Just doing one and not figuring out the long term piece doesn't make
> sense because we may end up having two completely distinct features
> and/or feature creep over the next year which doesn't make much  
> sense to
> me.
>
> I think in last discussion most ppl just preferred a sexier anonymous
> function esp. as there weren't very clear/clean proposals for how
> closures would work in PHP and what the benefits would be.
>
> So I definitely suggest to have a discussion, maybe RFC based like
> Stefan did but really decide once and for all on where we'd like to go
> with this. We all know that no features map 1:1 to PHP and there are
> always unique behaviors of our language and environment which  
> require a
> fresh view at things.

---

## #35776 — Stanislav Malyshev — 2008-02-25T00:03:19+00:00

URL: https://externals.io/message/34631#35776 · In reply to: #35765

Hi!

> To be clear, my statement was about the closures and not sexier 
> anonymous functions. I agree that we need to do an RFC on this matter.

I'm curious - for which PHP version all these things are targeted?
-- 
Stanislav Malyshev, Zend Software Architect
stas@zend.com   http://www.zend.com/
(408)253-8829   MSN: stas@zend.com

---

## #35779 — Andrei Zmievski — 2008-02-25T08:50:20+00:00

URL: https://externals.io/message/34631#35779 · In reply to: #35776

Whichever one the newly developed features are going into these days.  
Seems to be 5.3 or 5.4.

-Andrei



> Hi!
>
> > […]
>
> I'm curious - for which PHP version all these things are targeted?
> -- 
> Stanislav Malyshev, Zend Software Architect
> stas@zend.com   http://www.zend.com/
> (408)253-8829   MSN: stas@zend.com

---

## #35784 — Marcus Boerger — 2008-02-25T09:29:23+00:00

URL: https://externals.io/message/34631#35784 · In reply to: #35759

Hello Andi,

  my understanding was that we did not like the anon function syntax sugar
and could not agree on a feature set for closures upfront and thus decided
to disagree and not do anything. Now that the topic came up again it seems
that we actually should work on an agreeable featureset in a form of a RFC.

My current opinion is that we should have a local scope created on anon
function creation that would even contain $this. Now this is even more so
premature so I lea ve it for a RFC.

What we need is someone to start that RFC, any volunteers?

marcus

Saturday, February 23, 2008, 8:36:18 PM, you wrote:

> I think we have to be very clear whether we are going to provide just a
> sexier notation for anonymous functions or closures (and if the latter
> what the semantics are).
> Just doing one and not figuring out the long term piece doesn't make
> sense because we may end up having two completely distinct features
> and/or feature creep over the next year which doesn't make much sense to
> me.

> I think in last discussion most ppl just preferred a sexier anonymous
> function esp. as there weren't very clear/clean proposals for how
> closures would work in PHP and what the benefits would be.

> So I definitely suggest to have a discussion, maybe RFC based like
> Stefan did but really decide once and for all on where we'd like to go
> with this. We all know that no features map 1:1 to PHP and there are
> always unique behaviors of our language and environment which require a
> fresh view at things.

> Andi

> > […]




Best regards,
 Marcus

---

## #36385 — Marcus Boerger — 2008-03-23T14:46:34+00:00

URL: https://externals.io/message/34631#36385 · In reply to: #35756

Hello Ryusuke,

 I have put your proposal as a link to a PHP GSoC 2008 idea here:
http://wiki.php.net/gsoc/2008

Feel invited to add to this idea in whatever way you want :-)

marcus

Saturday, February 23, 2008, 7:46:34 PM, you wrote:

> 2008/2/24, Andrei Zmievski <andrei@gravitonic.com>:
> > […]

> Hi, I have added support for closures to my patches.

> http://www.opendogs.org/pub/php-5.3-080223-anon.patch
> http://www.opendogs.org/pub/php-6.0-080223-anon.patch
> http://www.opendogs.org/pub/php-anon-tests-080223.tgz

> Example:
> <?php
> function get_counter($i = 0)
> {
>     /**
>      * 'static' modified anonymous function's
>      *  uninitialized (or null) static variables
>      *  are copied from the declared scope.
>      */
>     return static function(){
>         static $i;
>         return ++$i;
>     };
> }

> $c1 = get_counter();
> $c2 = get_counter(100);
> var_dump($c1(), $c2());
> var_dump($c1(), $c2());
> var_dump($c1(), $c2());
?>>

> For recursive calling of anonymous functions and closures,
> added function get_current_function_key().

> <?php
> $f = function($i = 0)
> {
>     echo "$i ";
>     if (++$i > 10) {
>         return;
>     }
>     $f = get_current_function_key();
>     $f($i);
> };

> $f();
?>>


> And here is a Zend extension for PHP 5.2 which enables
> anonymous functions, closures and square bracket arrays.

> http://www.opendogs.org/pub/php_qiq-0.1.0pre.tgz

> This has a few more features, but, at present, there is
> no document...


> Regards,




Best regards,
 Marcus

---

## #36415 — Alexey Zakhlestin — 2008-03-24T09:21:57+00:00

URL: https://externals.io/message/34631#36415 · In reply to: #36385

> Hello Ryusuke,
>
>   I have put your proposal as a link to a PHP GSoC 2008 idea here:
>  http://wiki.php.net/gsoc/2008
>
>  Feel invited to add to this idea in whatever way you want :-)

how is it related to gsoc?
patches are already there — all we need is to decide which one is
better, and what needs to be fixed. no?

>
>  marcus
>
>
>  Saturday, February 23, 2008, 7:46:34 PM, you wrote:
>
> > […]
>
> > […]
>
> > […]
>
> > […]
>
> > […]
>  ?>>
>
> > […]
>
> > […]
>
> > […]
>  ?>>
>
>
> > […]
>
> > […]
>
> > […]
>
>
> > […]
>
>
>
>
>
> Best regards,
>
>  Marcus
>
>
>
>  --
>  
>  
>
>


-- 
Alexey Zakhlestin
http://blog.milkfarmsoft.com/

---

## #36422 — Marcus Boerger — 2008-03-24T13:03:01+00:00

URL: https://externals.io/message/34631#36422 · In reply to: #36415

Hello Alexey,

  my last update on this was that people were unhappy with closures. If
that is not the case than maybe we need to write this as an RFC on the
wiki: http://wiki.php.net/rfc

So well maybe you're right and it only requires selecting between the two
and checking whether there are tiny parts missing.

marcus

Monday, March 24, 2008, 10:21:57 AM, you wrote:

> 
> > […]

> how is it related to gsoc?
> patches are already there — all we need is to decide which one is
> better, and what needs to be fixed. no?

> > […]





Best regards,
 Marcus