-- Randal L. Schwartz, Perl hacker

Hi all again,
Thanks for the great input. Based on your comments, I've given the code another go. This time I've "compiled" the code into the form (op_name, arglist...), which is then run using an "eval". I've also establised a symbol table hash (which currently holds the lvalue type and value). I'd like to have your opinion on this design. (Oh yeah, the grammar syntax is still very ugly and basic. Needs to be redone.)
Thanks for your time.


use Parse::RecDescent;
use vars qw(@lines $curline %stash);
use strict;

$RD_HINT++;
$RD_WARN++;
#$RD_TRACE++;


sub do_print
{
    my $a=shift;
    print "$a\n";
}

sub do_goto
{
    $curline = int(shift)-1;
}

my $grammar = q{
  Start: Expression(s) /\Z/
  Expression: LineNum Print
  { 
      $::lines[int($item{LineNum})]= ["print", [$item{Print}]];
  }
  | LineNum Goto
  {
     $::lines[int($item{LineNum})]= ["goto", [$item[2]]];
  }
  | LineNum Ass
  | <error>
  Ass: StringAss
  | NumAss
  StringAss: Identifier /\$/ /=/ /\"/ String /\"/
  {
      $::stash{$item{Identifier}}{type}="String";
      $::stash{$item{Identifier}}{value}=$item{String};
  }
  NumAss: Identifier /=/ Number
  {
      $::stash{$item{Identifier}}{type}="Number";
      $::stash{$item{Identifier}}{value}=$item{Number};
  }
  Goto: /goto/i LineNum
  Print: /print/i /\"/ String /\"/
  { 
      $item{String};
  }
    | /print/i Identifier /\$/
    {
    if(defined($::stash{$item{Identifier}}{value}))
    {
        $::stash{$item{Identifier}}{value};
    }
    else {die "A nasty and horrible death... Bad string identifier.\n"
+;}
    }
    | /print/i Identifier
    {
    if(defined($::stash{$item{Identifier}}{value}))
    {
        $::stash{$item{Identifier}}{value};
    }
    else {die "A nasty and horrible death... Bad number identifier.\n"
+;}
    }
  String: /[\w\s]+/
  LineNum: /\d+/
  Identifier: /[a-zA-Z][a-zA-Z0-9]*/
  Number: /[+-]?\d+(\.\d+)?/
  };

my $parser = new Parse::RecDescent($grammar);
undef $/;
my $text = <>;
$parser->Start($text);
$curline=0;

while(1)
{
    if($lines[$curline])
    {
    my $todo = "do_".${$lines[$curline]}[0]."(\"";
        my $args = join(', ', @{${$lines[$curline]}[1]});
        $todo = $todo.$args."\")";
        eval $todo;
    }
  last if $curline>$#lines;
  $curline++;
}
[download]


Here are some test files:
test1.bas

5 print "This "
10 print "is "
15 print "a "
20 print "test"
25 goto 5
[download]


test2.bas

5 a$ = "This is a string"
10 print a$
15 b = 35
20 print b
[download]

Very important next step: compile once, run many. :)

-- Randal L. Schwartz, Perl hacker

If you want to learn how to write interpreters, read Structure and Interpretation of Computer Programs, and, if you want still more, Essentials of Programming Languages.

Christian Lemburg
Brainbench MVP for Perl
http://www.brainbench.com

Oh lord, there must be an easier way to do this in Parse::RecDescent, so why can't I see it? I think Merlyn's right - it's better to compile the data into another form rather then try to screw around with it on-the-fly.

What this does is save every line seen, and, when a GOTO is encountered, it's all pasted it all back into to top of $text (which holds the unparsed text.) There is no error checking to make sure the line to go to actually exists.

use Parse::RecDescent;

$RD_HINT++;
$RD_WARN++;

my $grammar = q{
  Start: Expression(s) /\Z/

  Expression: 
    LineNum Print    
    {                
      $thisparser->{_lines}->{$item[1]} = $item[1].' PRINT "'.$item[2]
+.'"';
    }
    | LineNum Goto  
    {               
      $text = $item[1] . " GOTO " . $item[2] . $text;
      foreach ( sort { $b <=> $a } grep { $_ >= $item[2] } keys %{$thi
+sparser->{_lines}} ) {
        $text = $thisparser->{_lines}->{$_} . $text;
      }                                             
    }
    | <error>

  Print: /print/i /\"/ /[\s\w]+/ /\"/  
  {
      print "$item[3]\n";
      $item[3];          
  }
  Goto: /goto/i LineNum
  {
      $item[2];
               
      # print "Seen a goto line $item[2]\n";
      # How do I seek back to line $item[2] of
      # the input?                            
  }
  LineNum: /\d+/
};

my $parser = new Parse::RecDescent($grammar);

undef $/;
my $text = <DATA>;

$parser->Start($text);

__END__
10 print "hello"
20 print "this"
30 print "is"
40 print "a"
50 print "test"
60 print "goodbye"
70 goto 30
[download]

There's a general method or something that I'm missing or have skipped over or have forgotten ...