Req #55266 [Com]: Traits vs Type hints

From: Date: Sat, 23 Jul 2011 11:47:49 +0000
Subject: Req #55266 [Com]: Traits vs Type hints
References: 1  Groups: php.bugs 
Request: Send a blank email to php-bugs+get-162125@lists.php.net to get a copy of this message
Edit report at https://bugs.php.net/bug.php?id=55266&edit=1

 ID:                 55266
 Comment by:         gron@php.net
 Reported by:        mchlpl at gmail dot com
 Summary:            Traits vs Type hints
 Status:             Open
 Type:               Feature/Change Request
 Package:            Class/Object related
 PHP Version:        5.4.0alpha2
 Block user comment: N
 Private report:     N

 New Comment:

My view on this issue is that traits do not constitute types and do not provide interfaces. (I
proposed the later one in the beginning in the sense of traits are interfaces with implementation,
but this was disregarded to make clear that traits are neither classes nor interfaces.)

Thus, is_a should also not say anything about a trait.
Traits are not units of encapsulation, they do not guarantee to provide/protect any invariants.

However, if you need to know what traits are used by a class, please refer to:
ReflectionClass::getTraits()

I just noticed that we have the following function in the SPL:
http://php.net/manual/en/function.class-implements.php
That should probably be mirrored to provide the same functionality as ReflectionClass::getTraits().

Not sure what the design policies are here. From a symmetry perspective there should be a
class_uses() function, but from my personal perspective, class_implements should get nuked and uses
should transition to the reflection extension if they need such meta programming facilities. Well,
the later is not practical, so we will probably need to have class_uses().


Previous Comments:
------------------------------------------------------------------------
[2011-07-22 14:51:09] alex dot howansky at gmail dot com

> $a is not an instance of the trait but rather of
> the class that utilizes the trait.

You can say the same of interfaces and abstracts, but is_a returns true for 
them.

Test script:
---------------
trait someTrait {}

interface someInterface {}

abstract class someAbstract {}

class someClass extends someAbstract implements someInterface {
    use someTrait;
}

$a = new someClass();
var_dump(is_a($a, 'someClass'));
var_dump(is_a($a, 'someAbstract'));
var_dump(is_a($a, 'someInterface'));
var_dump(is_a($a, 'someTrait'));

Expected result:
----------------
bool(true)
bool(true)
bool(true)
bool(?)

Actual result:
----------------
bool(true)
bool(true)
bool(true)
bool(false)

------------------------------------------------------------------------
[2011-07-22 14:50:36] mchlpl at gmail dot com

Mike: That's one way of looking at it. My point of view is that a trait adds methods to
class' interface (where interface is a set of public members of a class - both methods and
fields) and this should be reflected in class type.

Currently (unless I missed something that was not mentioned in RFC - I admit to not have gone
through SVN version of docs) there seem to be no way to check if class uses a trait or no. Only new
functions mentioned are trait_exists() and get_declared_traits(). Checking if the object has a
method will not work, if trait is meant to override method in host class. Having traits reflected in
object's type, would solve this problem nicely. Adding another function to glabal namespace
(uses_trait() ?) is not something I would like to see.

------------------------------------------------------------------------
[2011-07-22 13:36:59] me at mikestowe dot com

I believe this is the correct result as $a is not an instance of the trait but rather of the class
that utilizes the trait.

------------------------------------------------------------------------
[2011-07-22 07:44:34] mchlpl at gmail dot com

Description:
------------
Traits, when used in a class, are not added to class' type, and therefore can not be use in
type hints. The RFCs on traits https://wiki.php.net/rfc/traits  and on horizontal reuse
https://wiki.php.net/rfc/horizontalreuse do
not explain if this is by design.

On the other hand Traits seem to share the same namespace as Classes and Interfaces, since it is
impossible to have an Interface and a Trait that share the name ('Fatal error: Cannot redeclare
class' is raised when this is attempted).

Test script:
---------------
<?php
trait SomeTrait {}

class SomeClass {
  use SomeTrait;
}

$a = new SomeClass();

var_dump(is_a($a,'SomeTrait'));

Expected result:
----------------
bool(true)

Actual result:
--------------
bool(false)


------------------------------------------------------------------------



-- 
Edit this bug report at https://bugs.php.net/bug.php?id=55266&edit=1


Thread (9 messages)

« previous php.bugs (#162125) next »