Re[4]: mixing of asign() and assign_by_ref() overwrites external vars

B> 1) I don't agree that it is unsuspected; it follows the PHP4 paradigm.
Do, $smarty->assign("a", "b") has EQUAL paradigm as PHP's $a="b".
Agree?  Do  you  plan  exactly  this  while  developing  assign()
function? Is it documented?

B> 2) I don't agree that it is transparent: messju already pointed out the
B> problem with deprecated pass-by-reference.
It is not a problem. You may not use deprecated syntax at all.

B> 3) I agree with Monte's sentiment: don't fix what isn't broken. In my
B> last post, I show the standard work-around for getting the behaviour
B> that was desired in the case originally given. It uses an existing API
B> call which merely wraps what you would have to do with straight PHP:
B> use unset().
Oh.

B> 4) You are overloading the meaning of assign() to allow
B> pass-by-reference whereas the Smarty API already decided long ago to
B> break that into a different API call, namely assign_by_ref(). I think
B> it is more confusing when you have two methods to do ostensibly the
B> same thing.
But you surely can not modify API nor use deprecated syntax.

    function assign($tpl_var, $val = null)
    {
        // First copy - always!
        $value = $val;
        // Then - assign by ref, to avoid unexpected (?) behavour.
        if (is_array($tpl_var)){
            foreach ($tpl_var as $key => $val) {
                if ($key != '') {
                    $this->_tpl_vars[$key] =& $val;
                }
            }
        } else {
            if ($tpl_var != '')
                $this->_tpl_vars[$tpl_var] =& $value;
        }
    }

So, now API is not touched, deprecated & call works as without &,
but  assign()  after  assign_by_ref()  does  not  modify  source
variable.

P.S.
It is now question of principle, of course, not practical.

-- 
Best regards,
  Dmitry Koteroff.

-- 
Smarty Development Mailing List (http://smarty.php.net/)
To unsubscribe, visit: http://www.php.net/unsub.php