Solution

Use a lexical analyzer.

Break the code down into tokens:
Determine which (non-reserved) word tokens (/[a-zA-Z_]\w*/) represent identifiers, as opposed to fields.

Use a standard string comparison algorithm, such as Boyer-Moore to compare sections. To compare identifiers, keep symbol tables for each segment:

 1 sub tokeq {
 2   my ($t1, $t2, $symtab1, $symtab2, $vars) = @_;
 3   #...
 4   my $name = $t1->{name};
 5   if (!$symtab1->{$name}) {
 6     $symtab1->{$name} = $vars->[0]++;
 7   }
 8 
 9   $name = $t2->{name};
10   if (!$symtab2->{$name}) {
11     $symtab2->{$name} = $vars->[1]++;
12   }
13   #...
14   return $symtab1->{$name} == $symtab2->{$name};
15 }