# Constant and expression ? - Thread: https://externals.io/message/48339 - Original archive: https://news-web.php.net/php.internals/48339 - Started by Frederic Hardy on 2010-05-19T08:54:03+00:00 - Messages: 12 - 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 --- ## #48339 — Frederic Hardy — 2010-05-19T08:54:03+00:00 URL: https://externals.io/message/48339#48339 Hello ! Is there any feature request about constant and expression ? I would like to write something like that : Best regards, Fred. --- ## #48340 — Tjerk Anne Meesters — 2010-05-19T09:03:49+00:00 URL: https://externals.io/message/48339#48340 · In reply to: #48339 AFAIK class constants are optimized for speed and therefore don't support expressions; don't expect this to change any time soon ;-) On Wed, May 19, 2010 at 4:54 PM, Frederic Hardy wrote: > Hello ! > > Is there any feature request about constant and expression ? > I would like to write something like that : > > > namespace foo\bar\directories { >    const tmp = __DIR__ . '/tmp'; > } > > ?> > > Best regards, > Fred. > > -- > > > > -- -- Tjerk --- ## #48341 — Pierre Joye — 2010-05-19T09:10:50+00:00 URL: https://externals.io/message/48339#48341 · In reply to: #48340 hi, > AFAIK class constants are optimized for speed and therefore don't > support expressions; don't expect this to change any time soon ;-) It would require to do a two passes compilation (or something along this line) as you can define constants later in your script. They will also be defined at runtime, not at buildtime. pecl's hidef may help :). Cheers, -- Pierre @pierrejoye | http://blog.thepimp.net | http://www.libgd.org --- ## #48343 — Tjerk Anne Meesters — 2010-05-19T10:21:27+00:00 URL: https://externals.io/message/48339#48343 On Wed, May 19, 2010 at 5:52 PM, Frederic Hardy wrote: > Hello ! > > […] > > So the best solution is something like define('foo\bar\directories\tmp', __DIR__ . '/tmp'); ?>... > It 's strange that "define()" function can do something that the "const" >  keyword can't do, even if "const" keyword evaluation is done at compilation > time and "define()" evaluation is done at runtime. That's why there's a cost penalty to using `define()`. I wrote a small article that gives an idea of the speed differences: http://shwup.blogspot.com/2010/04/about-constants.html > > Best regards, > Fred. > > -- > ======================================================================== > Frédéric Hardy : Architecte d'application/Admin. système/Ergonome >        Status : En recherche d'emploi > ======================================================================== > > -- -- Tjerk --- ## #48344 — fqqdk — 2010-05-19T11:03:59+00:00 URL: https://externals.io/message/48339#48344 · In reply to: #48343 2010/5/19 Tjerk Anne Meesters > On Wed, May 19, 2010 at 5:52 PM, Frederic Hardy > wrote: > > […] > compilation > > […] > > That's why there's a cost penalty to using `define()`. > > I wrote a small article that gives an idea of the speed differences: > http://shwup.blogspot.com/2010/04/about-constants.html > > > > […] > > > > -- > -- > Tjerk > > -- > > > How about extending the usage of the 'final' keyword to support a java-like syntax? class Foo { private static final $bar = 'ex' . 'pression'; // and also private final $baz = 'another'.'expression'; // and maybe private final $inTheCtor; public function __construct($aValue) { $this->inTheCtor = $value; } } --- ## #48345 — Richard Quadling — 2010-05-19T11:04:20+00:00 URL: https://externals.io/message/48339#48345 · In reply to: #48343 > On Wed, May 19, 2010 at 5:52 PM, Frederic Hardy > wrote: > > […] > > That's why there's a cost penalty to using `define()`. > > I wrote a small article that gives an idea of the speed differences: > http://shwup.blogspot.com/2010/04/about-constants.html > > > > […] > > > > -- > -- > Tjerk With regard to HiDef and the hidef.ini_path setting, and the comment on [1] regarding "However, all this goodness comes with a drawback: configuration settings are instance wide, which impacts all setups of PHP that runs beyond a single web request (e.g. Apache module or FastCGI).", could the ini entry be placed in a ... [HOST=] section in the ini file? [2] If so, could this be the best solution? Regards, Richard. [1] http://shwup.blogspot.com/2010/04/about-constants.html [2] http://docs.php.net/manual/en/ini.sections.php -- ----- Richard Quadling "Standing on the shoulders of some very clever giants!" EE : http://www.experts-exchange.com/M_248814.html EE4Free : http://www.experts-exchange.com/becomeAnExpert.jsp Zend Certified Engineer : http://zend.com/zce.php?c=ZEND002498&r=213474731 ZOPA : http://uk.zopa.com/member/RQuadling --- ## #48347 — Christian Schneider — 2010-05-19T11:49:54+00:00 URL: https://externals.io/message/48339#48347 · In reply to: #48344 fqqdk wrote: > How about extending the usage of the 'final' keyword to support a java-like > syntax? > > class Foo { > private static final $bar = 'ex' . 'pression'; In my book "final" in Java gets an award for sickest abuse of a keyword. So no, I don't consider this an option. Simply use a class variable (or `define()`) if you need to store some runtime calculated value. "But can do this" alone is not sufficient as a reason to bloat PHP IMHO. - Chris --- ## #48349 — Ferenc Kovacs — 2010-05-19T12:04:13+00:00 URL: https://externals.io/message/48339#48349 · In reply to: #48347 On Wed, May 19, 2010 at 1:49 PM, Christian Schneider wrote: > fqqdk wrote: > > […] > java-like > > […] > > In my book "final" in Java gets an award for sickest abuse of a keyword. > So no, I don't consider this an option. Simply use a class variable (or > `define()`) if you need to store some runtime calculated value. > > "But can do this" alone is not sufficient as a > reason to bloat PHP IMHO. > > - Chris > I think that is pretty subjective, but I did suprised when I found out, that the global scope constants behaves differently than the class consts. Allowing the class consts to be initialized in runtime(if neccessary) wouldn't breake any php application, only slow them. Can't we do something, to allow broader usage of consts without sacrificing performance? So I propose to add a new syntax: literal. literal should behave how the const behaved, and const should be initialized in runtime. maybe we could check from runtime, that the const was a dynamic value, or literal, and generate a notice for the developer to change the const to literal for gaining performance. What do you think? Tyrael --- ## #48352 — johannes@schlueters.de — 2010-05-19T12:43:14+00:00 URL: https://externals.io/message/48339#48352 · In reply to: #48344 Hi, > 2010/5/19 Tjerk Anne Meesters > > […] Unfortunately this doesn't tell, what you actually measured. And note that as of 5.3 the "class constant" syntax can be used in the global scope, too. > How about extending the usage of the 'final' keyword to support a java-like > syntax? No. We have a const keyword. Rather somebody interested in this topic might try to come up with a patch to at least allow "constant expressions" this should serve most needs. constant expressions might be stuff like __DIR__.'/foobar.php' or CONST_A | CONST_B. We already have two steps in the constant resolution for constructs like these: php > class A { const C = FOOBAR; } php > define('FOOBAR', 42); php > echo A::C; 42 See zval_update_constant_ex in zend_API.c as a starting point. While such an approach would also require some parser work and some clever idea where/how to store the expression to be evaluated etc. johannes --- ## #48355 — Tjerk Anne Meesters — 2010-05-19T14:57:27+00:00 URL: https://externals.io/message/48339#48355 · In reply to: #48345 On Wed, May 19, 2010 at 7:04 PM, Richard Quadling wrote: > > > […] > > With regard to HiDef and the hidef.ini_path setting, and the comment > on [1] regarding "However, all this goodness comes with a drawback: > configuration settings are instance wide, which impacts all setups of > PHP that runs beyond a single web request (e.g. Apache module or > FastCGI).", could the ini entry be placed in a ... > > [HOST=] > > section in the ini file? [2] > > If so, could this be the best solution? The speed of hidef is accomplished by having only MINIT and MSHUTDOWN handlers; it would be interesting to see how speed is affected by moving some logic to the request init/shutdown; it should still have the benefit of constants being defined at compile time. > > Regards, > > Richard. > > [1] http://shwup.blogspot.com/2010/04/about-constants.html > [2] http://docs.php.net/manual/en/ini.sections.php > > -- > ----- > Richard Quadling > "Standing on the shoulders of some very clever giants!" > EE : http://www.experts-exchange.com/M_248814.html > EE4Free : http://www.experts-exchange.com/becomeAnExpert.jsp > Zend Certified Engineer : http://zend.com/zce.php?c=ZEND002498&r=213474731 > ZOPA : http://uk.zopa.com/member/RQuadling --- ## #48696 — Frederic Hardy — 2010-06-09T08:51:36+00:00 URL: https://externals.io/message/48339#48696 · In reply to: #48352 > Hi, > > > > > […] > Unfortunately this doesn't tell, what you actually measured. And note > that as of 5.3 the "class constant" syntax can be used in the global > scope, too. > > > > […] > No. We have a const keyword. Rather somebody interested in this topic > might try to come up with a patch to at least allow "constant > expressions" this should serve most needs. > > constant expressions might be stuff like __DIR__.'/foobar.php' or > CONST_A | CONST_B. > > We already have two steps in the constant resolution for constructs like > these: > > php> class A { const C = FOOBAR; } > php> define('FOOBAR', 42); > php> echo A::C; > 42 > > See zval_update_constant_ex in zend_API.c as a starting point. While > such an approach would also require some parser work and some clever > idea where/how to store the expression to be evaluated etc. > > johannes > So, it's possible ? Very good news. Best regards, Fred. -- ======================================================================== Frédéric Hardy : Architecte d'application/Admin. système/Ergonome ======================================================================== --- ## #48697 — johannes@schlueters.de — 2010-06-09T09:36:17+00:00 URL: https://externals.io/message/48339#48697 · In reply to: #48696 > > […] > So, it's possible ? > Very good news. Everything is "possible". The question is how maintainable the solution would be, how much of an performance impact it has etc. That I don't know. It might turn out not to be acceptable. I just gave a starting point in case somebody has the time needed for some experiments. As always: Ideas, there are many. They need developers who have time and motivation to do it. johannes