Instead of using "global" variables, you can use sub-local pseudo-global variables. These are "local" to your function by definition, but are available to sub-functions, such as anonymous subs that you define. This avoids problems when you use the sub more than once, such as if you are using it recursively.

Here's an example of the pseudo-global variables using HTML::Parser, and hopefully it will help you with your particular application.

The theory is that you build a stack of tags that you are inside. The only catch is that some tags don't have pairs, so they don't get shut off properly. An exception table might be handy for this, so in this program there is a rudimentary list of simple tags that do not get closed. This prevents them from accumulating endlessly, as BR would tend to do.

sub parseur
{
      my $stuff  = shift;

      # Build a hash out of a list of "atomic" tags,
      # i.e. tags which are not normally closed.
      my %atomic = map {($_=>1)} qw [ hr br p img li option ];
      my @stack;

      my $start_h = sub
      {
            my ($tagname,$attr) = @_;
            push (@stack, $tagname) unless $atomic{$tagname};
      };

      my $end_h = sub
      {
           my ($tagname) = @_;

           # Find the last entry in the stack which corresponds...
           foreach my $i ($#stack .. 0)
           {
                 if ($stack[$i] eq $tagname)
                 {
                       # ...and remove it.
                       splice (@stack, $i, 1);
                 }
           }
      };

      my $parser = new HTML::Parser (
                       start_h => [ $start_h, 'tagname,attr' ],
                       end_h   => [ $end_h,   'tagname' ],
                    );

      $parser->parse ($stuff);
}
[download]