Thu, 09 Apr 2026 18:03:12 +0000 https://bugs.horde.org/ticket/3957 call_user_func always passes copy of Turba object to _turba_hook_encode_{attribute} The hook functions _turba_hook_encode_{attribute} and _turba_hook_decode_{attribute} always get a copy of the Turba object since the hook functions are called with call_user_func. It's not possible to get a reference to the object and thus it isn't possible to modify the object in the hook function. For example, I'd like to concatenate the first name and last name and store that value in the attributes "displayname" and "cn". I'd also like to construct the "postalAddress" from "street", "postalCode", "location" etc. This is only possible if I can get a reference to the Turba object instead of a copy. Here's an example hook function: function _turba_hook_encode_firstname($new_value, $old_value, &$turba_object) { $firstname = $new_value; $lastname = $turba_object->getValue('firstname'); $name = $firstname . ' ' . $lastname; $turba_object->setValue('displayname', $name); $turba_object->setValue('cn', $name); return $new_value; } I've attached a small patch that replaces call_user_func in turba/lib/Object.php with direct function calls. It's highly unlikely that this change will break existing hook functions. I don't think that anyone has requested a reference to the Turba object but then expects to get a copy of the object. Fri, 19 May 2006 09:44:28 +0000 https://bugs.horde.org/ticket/3957#t20397 On my way back home after work it crossed my mind that I could subclass Turba's LDAP driver instead of "abusing" the encode hooks to set additional LDAP attributes. I think that subclassing the LDAP driver is a much better solution since it doesn't require any changes to Horde. I'm sorry for bothering you! Fri, 19 May 2006 17:36:25 +0000 https://bugs.horde.org/ticket/3957#t20410 True, but it's also useful to be able to modify the object, and part of the reason the $this parameter was added. Can you test with using &$this instead of $this with the call_user_func syntax? Sun, 21 May 2006 04:36:25 +0000 https://bugs.horde.org/ticket/3957#t20443 > True, but it's also useful to be able to modify the object, and part of the reason the $this > parameter was added. Can you test with using &$this instead of $this with the call_user_func > syntax? That works, but you'll always get a reference. Here's an example: function foo($x) { $x = 42; } function bar(&$x) { $x = 43; } $y = 0; call_user_func('foo', &$y); print "$y\n"; # -> 42 call_user_func('bar', &$y); print "$y\n"; # -> 43 If the function is called directly you can decide in the hook function whether you'd like to get a reference or a copy: function foo($x) { $x = 42; } function bar(&$x) { $x = 43; } $y = 0; foo($y); print "$y\n"; # -> 0 bar($y); print "$y\n"; # -> 43 I don't know what's more sensible in this case. Mon, 22 May 2006 07:07:38 +0000 https://bugs.horde.org/ticket/3957#t20467 > That works, but you'll always get a reference. Sure, but I'm not sure why that's a problem. Any other opinions here? Jan? Mon, 22 May 2006 17:54:08 +0000 https://bugs.horde.org/ticket/3957#t20494 Doesn't this produce call-time pass-by-reference notices? Mon, 22 May 2006 21:06:38 +0000 https://bugs.horde.org/ticket/3957#t20500 No, that's how you need to wrap objects if you want a reference in an array. The array is what's being passed to the function. Mon, 22 May 2006 21:16:40 +0000 https://bugs.horde.org/ticket/3957#t20502 Fine for me then. Mon, 22 May 2006 21:19:30 +0000 https://bugs.horde.org/ticket/3957#t20503 Okay, I was obviously on something; of course you'd get a call-time-blah-blah error with our current syntax. But using &$this with call_user_func_array() works fine. Mon, 22 May 2006 23:34:06 +0000 https://bugs.horde.org/ticket/3957#t20507