[Tickets #7031] Re: Colons within quoted parameter values are not handled correctly

[bugs at horde.org]

DO NOT REPLY TO THIS MESSAGE. THIS EMAIL ADDRESS IS NOT MONITORED.

Ticket URL: http://bugs.horde.org/ticket/7031
------------------------------------------------------------------------------
  Ticket             | 7031
  Updated By         | Jan Schneider <jan at horde.org>
  Summary            | Colons within quoted parameter values are not handled
                     | correctly
  Queue              | Horde Framework Packages
  Version            | FRAMEWORK_3
  Type               | Bug
-State              | Unconfirmed
+State              | Feedback
  Priority           | 2. Medium
  Milestone          |
-Patch              | 1
+Patch              |
  Owners             |
------------------------------------------------------------------------------


Jan Schneider <jan at horde.org> (2008-07-06 18:05) wrote:

It's far more complicated than that unfortunately, because the same  
regex must also parse vCalendar 1.0 properties which are escaped  
differently. Those only require semicolons in parameter values to be  
escaped by a preceding backslash.

I came up with the following so far:

         $name_re = '[-a-zA-Z0-9]+';
         $param_text_re = '[^";:,]*';
         $quoted_string_re = '"[^"]*"';
         $param_value_re = $param_text_re . '|' . $quoted_string_re;
         $param_re = $name_re . '=' . $param_value_re;
         $old_param_re = '(?:' . $name_re . '=)?(?:[^;]|(?<=\\\\);)*';

And then run the preg_match like so:

preg_match('/(' . $name_re . ')(;(?:' . $param_re . '|' .  
$old_param_re . '))*:([^\r\n]*)[\r\n]*/',
            $attribute, $parts);

This doesn't work unfortunately either, because the regex for the old  
format is too greedy. If using an ungreedy modifier, it becomes too  
ungreedy.