RE: evaluating parsed expressions

> { $item[1] || $item[3] } but that would not work as the parser would

These two variables hold the TEXT that was matched by the appropriate rule.

What you need to do is maintain a symbol table. (A hash, I believe you
already have that).  Then when you have a statement like 

(b|c)

you would return 

{$return = $symbols{$item[1]} || $symbols{$item[2]}; 1;}

The 1 is there only because I'm not really sure what your doing (I admit I
didn't read your code that closely) and it ensure that the rule succeeds.
You could avoid it if you knew for sure that $symbols{$item[1]} and [2] as
well would return defined values.

Actually on that note you might do something like this:
(assuming that there is a declaration phase in your language, oh and this
UNTESTED code in particular I cant remember if im using <error> correctly,
check the docs.)  

my $grammar=<<'GRAMMER';
{
# Init code for grammar... Handles a symbol table
my %symbols;

sub symbol_add{
   my $symbol=shift;
   my $value =shift;
   if ($symbols{$symbol}) {
         warn "Attempt to declare $symbol more than once\n";
	   return undef;
   } 
   $symbols{$symbol}=$value;
   return $value
}

sub symbol_fetch{
   my $symbol=shift;
   unless (exists $symbols{$symbol}) {
         warn "Attempt to fetch from undeclared $symbol\n";
	   return undef;
   } 
   return $symbols{$symbol}
}  

sub symbol_store{
   my $symbol=shift;
   my $value =shift;
   unless (exists $symbols{$symbol}) {
         warn "Attempt to store to an undeclared $symbol\n";
	   return undef;
   } 
   $symbols{$symbol}=$value;
   return $value
}  

}

conjunction :l_identifier '|' r_identifier 
 
{$return=symbol_fetch($item{l_identifier})||symbol_fetch($item{r_identifier}
)}
            |<error>
declaration :'set' identifier '=' value
{$return=symbol_add($item{identifier},$item{value});}
            | <error>
l_identifier:identifier {$return=$item[1]}
r_identifier:identifier {$return=$item[1]}
identifier  :/[A-Za-z]+/
value       :/\w+/

 
GRAMMER


BTW:  This is basically how the red-dragon suggest one should handle
symbols, (with a perlish flavour)

HTH (even though it only addresses 1 line of your question)

Yves