Edit report at https://bugs.php.net/bug.php?id=53727&edit=1
ID: 53727
Comment by: ralph at smashlabs dot com
Reported by: mattknight at xymail dot co dot uk
Summary: Inconsistent behavior of is_subclass_of with
interfaces
Status: Open
Type: Bug
Package: Class/Object related
Operating System: Gentoo x86_64
PHP Version: 5.3.5
Block user comment: N
Private report: N
New Comment:
Dmitry,
The patch called fix-is_subclass_of-PHP_5_3.diff does the following:
* Alters the existing Zend/tests/is_a.phpt to alter the bad
expectation,
* Adds a new test file for is_subclass_of() as
Zend/tests/is_subclass_of.phpt
* Finally alters is_a_impl() in the following ways:
* If an object was provided as the first parameter, and we are
inside of an is_subclass_of call (known by the flag
only_subclass), then assign the class entry of the object
* Next, determine if the instance_ce provided is the same class
entry as that we are testing against (parameter 2 of the
is_subclass_of() call ... if it is, return false
* Now that we know the instance_ce is not the same as the class
entry for the test class, allow is_a_impl to pass these class
entries to instanceof_function, and return that value
It is pretty straight forward, and any way I look at it, there are
few consequences to this patch.
Thanks,
Ralph
Previous Comments:
------------------------------------------------------------------------
[2011-01-13 09:13:50] mattknight at xymail dot co dot uk
Yes, it would also be fine to return false when comparing against an ancestral
interface, just so long as the behaviour is consistent.
Expected result:
----------------
false
true
false
true
------------------------------------------------------------------------
[2011-01-12 18:06:36] chris at cmbuckley dot co dot uk
Apologies; I actually meant:
false
true
false
true
------------------------------------------------------------------------
[2011-01-12 17:58:37] chris at cmbuckley dot co dot uk
One could argue that the expected result is actually:
true
true
false
false
since neither classes are technically "subclasses" of the interface. Either
way, the actual behaviour is undesirable :-)
------------------------------------------------------------------------
[2011-01-12 17:08:59] mattknight at xymail dot co dot uk
Description:
------------
is_subclass_of() can take a string as the first parameter, as well as a second
string parameter of the class name. It will check to see whether a class with
the name specified in the first parameter is a subclass of the class with the
name specified in the second parameter.
This logic, however, doesn't apply with interfaces.
The function behaves as expected for classes that don't directly implement an
interface, but won't recognise the interface on a class that does directly
implement it.
Test script:
---------------
<?php
interface MyInterface {
const TEST_CONSTANT = true;
}
class ParentClass implements MyInterface { }
class ChildClass extends ParentClass { }
echo (is_subclass_of('ChildClass', 'MyInterface') ? 'true' : 'false') . "\n";
echo (defined('ChildClass::TEST_CONSTANT') ? 'true' : 'false') . "\n";
echo (is_subclass_of('ParentClass', 'MyInterface') ? 'true' : 'false') . "\n";
echo (defined('ParentClass::TEST_CONSTANT') ? 'true' : 'false') . "\n";
Expected result:
----------------
true
true
true
true
Actual result:
--------------
true
true
false
true
------------------------------------------------------------------------
--
Edit this bug report at https://bugs.php.net/bug.php?id=53727&edit=1
