# PHP5: hook cast_object handler in userspace - Thread: https://externals.io/message/3697 - Original archive: https://news-web.php.net/php.internals/3697 - Started by Cristiano Duarte on 2003-08-02T22:50:01+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 --- ## #3697 — Cristiano Duarte — 2003-08-02T22:50:01+00:00 URL: https://externals.io/message/3697#3697 Hi internals, Is there a way to hook the cast_object handler in userspace ? What I want is a default "to_string" method called when I do something like: class my_object extends class_with_default_to_string_method { private $value = "Hello"; public function to_string() { //overhiding default to_string wich prints "Object id #n" return $this->value; } } $o = new my_object(); echo $o; or: $o = new xxx(); echo "My stringfied value is=$o"; Thanx, Cristiano Duarte --- ## #3698 — Cristiano Duarte — 2003-08-02T23:00:10+00:00 URL: https://externals.io/message/3697#3698 · In reply to: #3697 Sorry guys, > $o = new xxx(); should be $o = new my_object(); And if there is a way to hook into zend_standard_class (stdClass) what would make all objects inherit the to_string method, it would be great ! Cristiano Duarte "Cristiano Duarte" escreveu na mensagem news:20030802224959.58552.qmail@pb1.pair.com... > Hi internals, > > Is there a way to hook the cast_object handler in userspace ? What I want is > a default "to_string" method called when I do something like: > > class my_object extends class_with_default_to_string_method { > private $value = "Hello"; > public function to_string() { //overhiding default to_string wich prints > "Object id #n" > return $this->value; > } > } > > $o = new my_object(); > echo $o; > > or: > > $o = new xxx(); > echo "My stringfied value is=$o"; > > Thanx, > > Cristiano Duarte --- ## #3699 — Cristiano Duarte — 2003-08-03T03:33:25+00:00 URL: https://externals.io/message/3697#3699 · In reply to: #3698 I made a patch to the latest PHP5-CVS wich implements this hook. I think it's useful to make a meaningful object stringfied value. Can anyone verify it and maybe commit it to CVS ? --- zend.c 2003-07-30 14:07:36.000000000 -0300 +++ zend.c.new 2003-08-03 00:27:43.000000000 -0300 @@ -227,10 +227,24 @@ if (expr->value.obj.handlers->cast_object) { TSRMLS_FETCH(); expr->value.obj.handlers->cast_object(expr, expr_copy, IS_STRING, 0 TSRMLS_CC); } else { - expr_copy->value.str.val = (char *) emalloc(sizeof("Object id #")-1 + MAX_LENGTH_OF_LONG); - expr_copy->value.str.len = sprintf(expr_copy->value.str.val, "Object id #%ld", (long)expr->value.obj.handle); + //call to_string method + zval *fname, *retval; + MAKE_STD_ZVAL(fname); + ZVAL_STRING(fname, "to_string", 1); + TSRMLS_FETCH(); + if (call_user_function_ex(NULL, &expr, fname, &retval, 0, NULL, 0, `NULL` TSRMLS_CC) == SUCCESS) { + if (Z_TYPE_P(retval) != IS_STRING) { + convert_to_string(retval); + } + ZVAL_STRINGL(expr_copy, Z_STRVAL_P(retval), Z_STRLEN_P(retval), 1); + zval_ptr_dtor(&retval); + } else { + Z_STRVAL_P(expr_copy) = (char *) emalloc(sizeof("Object id #")-1 + MAX_LENGTH_OF_LONG); + Z_STRLEN_P(expr_copy) = sprintf(Z_STRVAL_P(expr_copy), "Object id #%ld", (long)Z_OBJ_HANDLE_P(expr)); + } + zval_ptr_dtor(&fname); } #if 0 /* FIXME: This might break BC for some people */ expr_copy->value.str.len = sizeof("Object")-1; Thanx, Cristiano Duarte. "Cristiano Duarte" escreveu na mensagem news:20030802230007.65232.qmail@pb1.pair.com... > Sorry guys, > > > […] > should be $o = new my_object(); > > And if there is a way to hook into zend_standard_class (stdClass) what would > make all objects inherit the to_string method, it would be great ! > > Cristiano Duarte > > "Cristiano Duarte" escreveu na mensagem > news:20030802224959.58552.qmail@pb1.pair.com... > > […] want > is > > […] > prints > > […] --- ## #3703 — Sterling Hughes — 2003-08-03T06:57:40+00:00 URL: https://externals.io/message/3697#3703 So, being the author of the C-level cast_object(), I thought about exposing this at one time. I decided against it mainly because I needed it for simplexml, and I didn't want an internals decision to be clouded by userspace discussions (how's my whitespace btw?) I really like the idea of PHP having an object interface that is completely transparent. It would be really nice if the language could (preferably via interfaces), provide a way for userspace people to design classes and objects that behave like base types. Is this a bit magical? Yes. But its also the only reason I'd ever use objects. If I need something to map to an external datasource, or an abstract data type, objects are the most powerful container. They allow for more efficiency (fewer function calls), and a more elegant userspace interface. I think it would be really great if the PHP object model contained all the necessary hooks so that something like simplexml could be implemented in userspace. -Sterling Am Son, 2003-08-03 um 04.59 schrieb Marcus Börger: > Hello Cristiano, > > Sunday, August 3, 2003, 5:33:25 AM, you wrote: > > CD> I made a patch to the latest PHP5-CVS wich implements this hook. I think > CD> it's useful to make a meaningful object stringfied value. > > CD> Can anyone verify it and maybe commit it to CVS ? > > CD> --- zend.c 2003-07-30 14:07:36.000000000 -0300 > CD> +++ zend.c.new 2003-08-03 00:27:43.000000000 -0300 > CD> @@ -227,10 +227,24 @@ > CD> if (expr->value.obj.handlers->cast_object) { > CD> TSRMLS_FETCH(); > CD> expr->value.obj.handlers->cast_object(expr, expr_copy, IS_STRING, 0 > CD> TSRMLS_CC); > CD> } else { > - expr_copy->>value.str.val = (char *) emalloc(sizeof("Object id #")-1 + > CD> MAX_LENGTH_OF_LONG); > - expr_copy->>value.str.len = sprintf(expr_copy->value.str.val, "Object id > #%ld", (long)expr->>value.obj.handle); > CD> + //call to_string method > CD> + zval *fname, *retval; > CD> + MAKE_STD_ZVAL(fname); > CD> + ZVAL_STRING(fname, "to_string", 1); > CD> + TSRMLS_FETCH(); > CD> + if (call_user_function_ex(NULL, &expr, fname, &retval, 0, NULL, 0, > CD> `NULL` TSRMLS_CC) == SUCCESS) { > CD> + if (Z_TYPE_P(retval) != IS_STRING) { > CD> + convert_to_string(retval); > CD> + } > CD> + ZVAL_STRINGL(expr_copy, Z_STRVAL_P(retval), Z_STRLEN_P(retval), > CD> 1); > CD> + zval_ptr_dtor(&retval); > CD> + } else { > CD> + Z_STRVAL_P(expr_copy) = (char *) > CD> emalloc(sizeof("Object id #")-1 + MAX_LENGTH_OF_LONG); > CD> + Z_STRLEN_P(expr_copy) = > CD> sprintf(Z_STRVAL_P(expr_copy), "Object id #%ld", > CD> (long)Z_OBJ_HANDLE_P(expr)); > CD> + } > CD> + zval_ptr_dtor(&fname); > CD> } > CD> #if 0 > CD> /* FIXME: This might break BC for some people */ > CD> expr_copy->value.str.len = sizeof("Object")-1; > > > CD> Thanx, > > CD> Cristiano Duarte. > > CD> "Cristiano Duarte" escreveu na mensagem > CD> news:20030802230007.65232.qmail@pb1.pair.com... > > […] > CD> would > > […] > CD> want > > […] > > > > > The patch doesn't compile (always use confugure --enable-maintainer-zts). > Then you assume that every class has a function to_string(), this is not what > we want. > > -- > Best regards, > Marcus mailto:helly@php.net -- UNIX was not designed to stop you from doing stupid things, because that would also stop you from doing clever things. - Doug Gwyn --- ## #3715 — Cristiano Duarte — 2003-08-04T02:45:20+00:00 URL: https://externals.io/message/3697#3715 Hello Marcus, > CD> switch/case removed and patch attached. Is it ok? > CD> Should the method be named "to_string()", "tostring()" or "__tostring()" ? > CD> Do you commit it to CVS ? > > No. Last Problem i see is that even when to_string() is available the default > behavior can get called that seems wrong. If that function doesn't return a > string it's an error and it shouldn't fall back to default in that case. Do you think it's an error even if the value returned got cast to string ? I don't know how to throw an error/exception from inside the core of Zend2. I have some code to throw an exception/error from extensions... > However with that last change the thing would be ok from my point of view. > Perhaps you could also add an interface for that function as sterling once > suggested. But still you need to discuss this with the rest of the ppl here. > Because eventually we might want to do it with overloadable object handlers. I modified the code to just call to_string if the object implements the "printable" interface. But I don't know how and where to register the interface in ZE2. Are there any interfaces registered inside ZE2? I cound't find any... If you could point me some code to dig... Cristiano Duarte --- ## #3729 — Alan Knowles — 2003-08-05T06:43:53+00:00 URL: https://externals.io/message/3697#3729 > switch/case removed and patch attached. Is it ok? > Should the method be named "to_string()", "tostring()" or "__tostring()" ? According to the current Coding Standard document.. it should be toString() : even though php is not Case-sensitive:).. Would be nice to see this work.. Regards Alan > Do you commit it to CVS ? > > Cristiano Duarte > > > -- Can you help out? Need Consulting Services or Know of a Job? http://www.akbkhome.com --- ## #3742 — Cristiano Duarte — 2003-08-05T23:23:15+00:00 URL: https://externals.io/message/3697#3742 · In reply to: #3729 Hi all, SK> Maybe it will be better to have this method in all classes and define SK> default behavior such as existing ( string )$object cast? I guess that using an interface is better. But if the guys here think different I would revert to the patch with no interfaces. AK> According to the current Coding Standard document.. it should be AK> toString() : even though php is not Case-sensitive:).. What about __construct, __destruct and __clone ? Shouldn't it be named __to_string() or __tostring() ? AK> Would be nice to see this work.. In fact, this does work. The method is called to_string(). Two questions: 1) I can't figure out a better name for the interface. Any ideas ? 2) Should all objects implent this method like __construct, __destruct and __clone or should it be left to the user to decide whether the object implements the printable interface or not.(IMHO a prefer the second) Cristiano Duarte. "Alan Knowles" escreveu na mensagem news:3F2F5229.5070004@akbkhome.com... > > > […] ? > According to the current Coding Standard document.. it should be > toString() : even though php is not Case-sensitive:).. > > Would be nice to see this work.. > > Regards > Alan > > > […] > > > -- > Can you help out? > Need Consulting Services or Know of a Job? > http://www.akbkhome.com --- ## #3749 — Alan Knowles — 2003-08-06T00:15:50+00:00 URL: https://externals.io/message/3697#3749 · In reply to: #3742 >AK> According to the current Coding Standard document.. it should be >AK> toString() : even though php is not Case-sensitive:).. >What about __construct, __destruct and __clone ? > these are built in. - and done to avoid conflicts. >Shouldn't it be named >__to_string() or __tostring() ? > since you implicitly make this available by adding an interface Morphable?? or soemthing, - toString() is ok.. I think it's alot better to make it available via an interface.. - It at least give you some chance of working out why it works as it does.. It would be nice to hear some noise from Zeev and Andi on this :) Regards Alan --- ## #3750 — Cristiano Duarte — 2003-08-06T02:08:42+00:00 URL: https://externals.io/message/3697#3750 · In reply to: #3749 "Alan Knowles" escreveu na mensagem news:3F3048B6.2060903@akbkhome.com... > > > […] AK> these are built in. - and done to avoid conflicts. All right. > > […] AK> since you implicitly make this available by adding an interface AK> Morphable?? or soemthing, - toString() is ok.. AK> I think it's alot better to make it available via an interface.. - It at AK> least give you some chance of working out why it works as it does.. Ok. New patch included. AK> It would be nice to hear some noise from Zeev and Andi on this :) I agree. Regards, Cristiano Duarte --- ## #3753 — Simeon Koptelov — 2003-08-06T08:07:36+00:00 URL: https://externals.io/message/3697#3753 · In reply to: #3742 Hello Cristiano, Here's why i think that interface is not good solution. We need the toString method like java's one to safely assume that _every_ object has the string representation. This means we can say $obj1->toString() and $obj2->toString() and don't care if $obj1 and $obj2 do support this method. In fact we _already_ have this functionality in ( string )$obj cast -- it's applicable to any object, doesn't it?. All we need is to formalize this cast in OOP manner. Interface is only the way to describe what class can do. If every class can provide it's string representation why should you implement some interface? It's only my $.20 though. -- Best regards, Simeon mailto:xi@newmail.ru --- ## #3761 — Cristiano Duarte — 2003-08-06T13:19:34+00:00 URL: https://externals.io/message/3697#3761 · In reply to: #3753 Simeon Koptelov wrote: > Hello Cristiano, > > Here's why i think that interface is not good solution. > > We need the toString method like java's one to > safely assume that _every_ object has the string representation. This > means we can say $obj1->toString() and $obj2->toString() and don't > care if $obj1 and $obj2 do support this method. In fact we _already_ > have this functionality in ( string )$obj cast -- it's > applicable to any object, doesn't it?. All we need is to formalize > this cast in OOP manner. Interface is only the way to describe what class > can do. If every class can provide it's string representation why should you > implement some interface? > > It's only my $.20 though. > Maybe you're right. I can move the code away from the "cast" function and create a toString method on the "Object" class wich returns "Object id #n". Let's see what the others think about it... Cristiano Duarte --- ## #3839 — Cristiano Duarte — 2003-08-09T20:48:22+00:00 URL: https://externals.io/message/3697#3839 Hello Marcus, Did you think of a better interface name instead of "Printable" ? Do you think Zeev or Andi will accept this patch ? Cristiano.