Private member/method access inconsistency View as Markdown

2009/1/6 David Coallier davidc@php.net:

call to __call() on private method invocation. The former approach presents
more of a BC
problem IMHO, so I am advocating the latter. I've attached a simple patch
for consideration.

I'd say go ahead, this sounds like common sense to be consistent in
both methods and members.

Cool, looks like this has gone in already. A couple of comments:

1. I think the change has not been made for callbacks (so callbacks
   now behave differently from normal method invocations).
2. Perhaps on 5_3 and HEAD, the equivalent change should be made for
   __callStatic()?

Addressing those two points would make things even more consistent.
I'm adding some testcases to illustrate.

--TEST--
Trigger __call() in lieu of non visible methods when called via a callback.
--FILE--
<?php
class C {
protected function prot() { }
private function priv() { }
public function __call($name, $args) {
echo "In __call() for method $name()\n";
}
}

$c = new C;
call_user_func(array($c, 'none'));
call_user_func(array($c, 'prot'));
call_user_func(array($c, 'priv'));
?>
--EXPECTF--
In __call() for method none()
In __call() for method prot()
In __call() for method priv()

--TEST--
Trigger __callStatic() in lieu of non visible methods when called
directly or via a callback.
--FILE--
<?php
class C {
protected static function prot() { }
private static function priv() { }
public static function __callStatic($name, $args) {
echo "In __callStatic() for method $name()\n";
}
}

echo "Using direct method invocation:\n";
C::none();
C::prot();
C::priv();

echo "\nUsing callbacks:\n";
call_user_func(array('C', 'none'));
call_user_func(array('C', 'prot'));
call_user_func(array('C', 'priv'));
?>
--EXPECTF--
Using direct method invocation:
In __callStatic() for method none()
In __callStatic() for method prot()
In __callStatic() for method priv()

Using callbacks:
In __callStatic() for method none()
In __callStatic() for method prot()
In __callStatic() for method priv()