Added get_debug_type as new function #5143

markrandall · 2020-02-02T21:50:44Z

Alternative to #5142 exposing a new function get_debug_type which will return the class name if the type is an object.

RFC: https://wiki.php.net/rfc/get_debug_type

markrandall · 2020-02-02T21:57:22Z

Will be changing the scalar types to represent their in-code versions.

KalleZ · 2020-02-02T22:41:08Z

This looks to be just as easily implemented in userland code, what is the rationale for implementing this in the core (the other PR didn't really give me any clue)?

The issue mentioned in regards to resources is notable, unless it returned something like: resource(xxx) (with xxx being the result of get_resource_type()).

markrandall · 2020-02-02T22:48:53Z

This looks to be just as easily implemented in userland code, what is the rationale for implementing this in the core (the other PR didn't really give me any clue)?

Just a request from Beberlei that cropped up in a discussion with Derick. I've had to implement this pattern a few times in the past week, and Beberlei mentioned it appears almost 200 times in Symfony and almost 400 times in Magento.

Perhaps more significantly, it's an opportunity to change the return values on this to match what is expected from type hints e.g. int not integer, float not double. It might make more sense for just hitting up zend_get_type_by_const for everything other than IS_OBJECT.

ext/standard/tests/general_functions/get_debug_type_basic.phpt

ext/standard/type.c

nikic · 2020-02-03T09:13:02Z

@KalleZ Basically this is the function you need to generate the expected XXX, YYY given style messages that PHP uses without adding a bunch of boilerplate for every exception you throw.

KalleZ · 2020-02-03T14:25:47Z

@nikic that makes sense to me, +1.

Side question, do we want to distinguish resource types or just flat out return them as "resource"? Given resource is a valid class name, then identifying the resource type could potentially solve that collision.

kocsismate · 2020-02-04T07:56:48Z

@KalleZ I was basically wondering on the same thing in the other PR.

So I'm curious if it would be worth to reserve the resource class name from use? Or is there any reason I am not aware of why it's the only type that is a valid class name in the same time (ok, along with ~~double~~ array - but at least the parser fails in that case)?

EDIT: I understood it in the meanwhile: as resource is not a real type (it's just a "confusable" type :) ), there is no reason not to support a class with this name as it is going to work as a type declaration.

KalleZ · 2020-02-04T13:39:35Z

@kocsismate thinking about it, I think it might be better to just leave the issue open but instead invest time in looking at changing resources into opaque objects (similar to how ext/gmp and ext/gd does) over the course of the 8.x series and then the issues fixes itself and we don't reserve a keyword while allowing for a more powerful typing with functionality implemented in the core.

nikic · 2020-02-05T09:26:42Z

Implementation looks fine. This is going to need a (small) RFC though. Expect some bikeshedding on names...

ext/standard/tests/general_functions/get_debug_type_basic.phpt

nicolas-grekas · 2020-02-05T12:50:35Z

Thank you for opening this PR, we really miss this helper IMHO!
What about improving the display name for anonymous classes?

the polyfilling code could be like that (inspired from what we do in Symfony):

function get_debug_type($value): string
{
    if (!is_object($value)) {
        return gettype($value);
    }

    $class = get_class($value);

    if (0 !== strpos($class, "class@anonymous\0")) {
        return $class;
    }

    return (get_parent_class($class) ?: key(class_implements($class))).'@anonymous';
}

nikic · 2020-02-05T13:37:15Z

Cutting class@anonmyous off at the \0 makes sense to me, as we do that for error messages as well.

Regarding use of the parent class / implemented interface, I'd only do that if we actually change the class name to look like that.

nicolas-grekas · 2020-02-05T13:42:07Z

Cutting class@anonmyous off at the \0 makes sense to me, as we do that for error messages as well.

ONLY if using the parent/interface as proposed! otherwise, we'll be unable to implement using them in userland with this function (and this would void the benefit of introducing it)

Regarding use of the parent class / implemented interface, I'd only do that if we actually change the class name to look like that.

That could be nice, but it could have BC consequences (breaking the code above)...

nikic · 2020-02-05T13:43:35Z

ONLY if using the parent/interface as proposed! otherwise, we'll be unable to implement using them in userland with this function (at this will void the benefit of introducing it)

Why? Your code should still work just fine as long as you drop the match on the \0 at the end.

nicolas-grekas · 2020-02-05T13:46:36Z

get_debug_type($anonymousClass) should NOT return class@anonymous. If it were, we wouldn't be able to provide the improvement we do with parent/interface from the output of the function. But I think I didn't get what you meant :)

nikic · 2020-02-05T13:50:31Z

function get_extended_debug_type($value) {
    $type = get_debug_type($value);
    if ($type === 'class@anonymous') {
        $class = get_class($value);
        return (get_parent_class($class) ?: key(class_implements($class))).'@anonymous';
    }
    return $type;
}

I don't think there's a lot of value in ^-- that though, it doesn't seems like important information.

Basically get_debug_type() should match the way PHP would display the value in an error message -- if that is going to be ParentClass@anonymous, that's fine. But if it's class@anonymous (as it is right now), then that's what the function should return.

nicolas-grekas · 2020-02-05T13:58:55Z

Yes, it's true that we can implement get_extended_debug_type(), but we could also totally implement get_debug_type() right now. We didn't because that'd mean yet another package/dependency, with all the associated maintenance overhead. We'd better keep the current code, which is fine...

I don't think there's a lot of value in ^-- that though, it doesn't seems like important information.

Actually it's critical information: when using anonymous classes as exception classes, the default error message is totally useless as far as the class name is concerned. Same for anonymous objects found in stack traces. This is the reason why I tend to avoid using them in portable classes, and why I implemented this parent/interface display in the tools I'm using (Symfony Debug/ErrorHandler). Not everyone uses Symfony, so it would be awesome if PHP could borrow this.

nicolas-grekas · 2020-02-06T11:07:27Z

Could it be worth adding the resource type?
get_debug_type($resource) => 'resource (stream)' ?

ext/standard/tests/general_functions/get_debug_type_basic.phpt

ext/standard/type.c

nicolas-grekas

Really nice!

derickr · 2020-03-01T14:34:56Z

ext/standard/type.c

+				RETURN_NEW_STR(zend_strpprintf(0, "resource (%s)", name));
+			} else {
+				RETURN_INTERNED_STR(ZSTR_KNOWN(ZEND_STR_CLOSED_RESOURCE));
+			}


Doesn't it miss "IS_UNINITIALIZED" used with typed properties?

They throw upon trying to read them, so could not be used in an expression.

ext/standard/php_type.h

This PR was merged into the 1.15-dev branch. Discussion ---------- [Php80] Add get_debug_type() Waiting for - [ ] php/php-src#5143 - [ ] https://wiki.php.net/rfc/get_debug_type - [x] php/php-src#5153 Commits ------- 61168ea [Php80] Add get_debug_type()

This PR was merged into the 1.15-dev branch. Discussion ---------- [Php80] Add get_debug_type() Waiting for - [ ] php/php-src#5143 - [ ] https://wiki.php.net/rfc/get_debug_type - [x] php/php-src#5153 Commits ------- 61168ea [Php80] Add get_debug_type()

ext/standard/tests/general_functions/get_debug_type_basic.phpt

nikic

Please add an UPGRADING note and squash when merging.

nikic · 2020-04-23T08:46:34Z

Rebased and merged as ef0e447.

Added get_debug_type as new function

9c346eb

Your Name added 2 commits February 2, 2020 22:03

Resolve type names to in-code version

5756a6d

Team tabs supremacy

bd32f07

markrandall mentioned this pull request Feb 2, 2020

Allow passing a second argument to gettype to resolve class names #5142

Closed

nikic reviewed Feb 3, 2020

View reviewed changes

ext/standard/tests/general_functions/get_debug_type_basic.phpt Show resolved Hide resolved

ext/standard/type.c Outdated Show resolved Hide resolved

Return known interned strings

9d5d333

nikic added the RFC label Feb 5, 2020

nicolas-grekas reviewed Feb 5, 2020

View reviewed changes

ext/standard/tests/general_functions/get_debug_type_basic.phpt Show resolved Hide resolved

nicolas-grekas mentioned this pull request Feb 5, 2020

[Php80] Add get_debug_type() symfony/polyfill#226

Merged

3 tasks

nikic mentioned this pull request Feb 6, 2020

Improve generated names for anonymous classes #5153

Closed

nicolas-grekas reviewed Feb 6, 2020

View reviewed changes

ext/standard/tests/general_functions/get_debug_type_basic.phpt Outdated Show resolved Hide resolved

nikic reviewed Feb 15, 2020

View reviewed changes

ext/standard/type.c Outdated Show resolved Hide resolved

Mark Randall added 2 commits February 15, 2020 16:48

Cutting off anonymous class names, resolving resource names

b2d43dd

Blessed tests

ee672b1

nikic reviewed Feb 15, 2020

View reviewed changes

ext/standard/type.c Outdated Show resolved Hide resolved

ext/standard/type.c Outdated Show resolved Hide resolved

Inlined logic, return existing string if not anon class

d05ac57

nicolas-grekas approved these changes Feb 27, 2020

View reviewed changes

derickr reviewed Mar 1, 2020

View reviewed changes

ext/standard/php_type.h Show resolved Hide resolved

nikic approved these changes Mar 16, 2020

View reviewed changes

ext/standard/tests/general_functions/get_debug_type_basic.phpt Outdated Show resolved Hide resolved

Mark Randall added 2 commits March 22, 2020 12:58

Merge remote-tracking branch 'origin/master' into get_debug_type

f59ad7c

Added tests for extending and implementing

0fe1344

nikic approved these changes Mar 23, 2020

View reviewed changes

nikic closed this Apr 23, 2020

carusogabriel added this to the PHP 8.0 milestone May 29, 2020

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Added get_debug_type as new function #5143

Added get_debug_type as new function #5143

markrandall commented Feb 2, 2020 •

edited by nikic

Loading

markrandall commented Feb 2, 2020

KalleZ commented Feb 2, 2020

markrandall commented Feb 2, 2020

nikic commented Feb 3, 2020

KalleZ commented Feb 3, 2020

kocsismate commented Feb 4, 2020 •

edited

Loading

KalleZ commented Feb 4, 2020

nikic commented Feb 5, 2020

nicolas-grekas commented Feb 5, 2020

nikic commented Feb 5, 2020

nicolas-grekas commented Feb 5, 2020 •

edited

Loading

nikic commented Feb 5, 2020

nicolas-grekas commented Feb 5, 2020

nikic commented Feb 5, 2020

nicolas-grekas commented Feb 5, 2020 •

edited

Loading

nicolas-grekas commented Feb 6, 2020

nicolas-grekas left a comment

derickr Mar 1, 2020

markrandall Mar 1, 2020

nikic left a comment

nikic commented Apr 23, 2020

Added get_debug_type as new function #5143

Added get_debug_type as new function #5143

Conversation

markrandall commented Feb 2, 2020 • edited by nikic Loading

markrandall commented Feb 2, 2020

KalleZ commented Feb 2, 2020

markrandall commented Feb 2, 2020

nikic commented Feb 3, 2020

KalleZ commented Feb 3, 2020

kocsismate commented Feb 4, 2020 • edited Loading

KalleZ commented Feb 4, 2020

nikic commented Feb 5, 2020

nicolas-grekas commented Feb 5, 2020

nikic commented Feb 5, 2020

nicolas-grekas commented Feb 5, 2020 • edited Loading

nikic commented Feb 5, 2020

nicolas-grekas commented Feb 5, 2020

nikic commented Feb 5, 2020

nicolas-grekas commented Feb 5, 2020 • edited Loading

nicolas-grekas commented Feb 6, 2020

nicolas-grekas left a comment

Choose a reason for hiding this comment

derickr Mar 1, 2020

Choose a reason for hiding this comment

markrandall Mar 1, 2020

Choose a reason for hiding this comment

nikic left a comment

Choose a reason for hiding this comment

nikic commented Apr 23, 2020

markrandall commented Feb 2, 2020 •

edited by nikic

Loading

kocsismate commented Feb 4, 2020 •

edited

Loading

nicolas-grekas commented Feb 5, 2020 •

edited

Loading

nicolas-grekas commented Feb 5, 2020 •

edited

Loading