Tips from the Basic debugging checklist: print and Data::Dumper. This will show you that $x and $y attempt to access elements beyond the end of your @rt and @pd arrays.

What should x and y be?

I ran your code thru perltidy, then fixed some split's and cluttered it up with ugly debug code to show you how I came to those conclusions:

use Graph::Undirected;
use warnings;
use strict;

package Graph::Base;

# Instantiate
my $G = new Graph::Undirected;

open( IN, "Input.txt" ) || die $!;

while (<IN>) {
    chomp;
    my @rxn = split /\#/, $_;

    # "$rxn[1]\n"; #debug
    my @cmp = split /\s*\<=>\s*/, $rxn[1];
#use Data::Dumper; $Data::Dumper::Sortkeys=1; print Dumper(\@cmp);

    #print"$cmp[0]\n"; # debug first element for reactants.
#    my @rt = split / \+ /, $cmp[0];    # split by space and '+'. if i
+ don't put space b/n \+ i get bad formated output.
    my @rt = split /[+\s]+/, $cmp[0];    # split by space and '+'. if 
+i don't put space b/n \+ i get bad formated output.

    my $x;                             #declaring $x b/c the use of st
+rict wouldn't allow me to $x or my ($x)!
    for ( $x = 0 ; $x <= $#rt ; $x++ ) {
        #print "$rt[$x]\n"; # debug prints the reactant compounds.
    }
#    my @pd = split /\+/, $cmp[1];
    my @pd = split /[+\s]+/, $cmp[1];
#use Data::Dumper; $Data::Dumper::Sortkeys=1; print Dumper(\@pd);
    my $y;
    for ( $y = 0 ; $y <= $#pd ; $y++ ) {
        #print "$pd[$y]\n";
    }

#use Data::Dumper; $Data::Dumper::Sortkeys=1; print Dumper(\@rt);
#use Data::Dumper; $Data::Dumper::Sortkeys=1; print Dumper(\@pd);
#print "$x $y\n";

$x=1; $y = 1;  # <----------------------------- ????????????
    $G = $G->add_edge( $rt[$x], $pd[$y] );
    print "$G\n";
}


__END__

My output:

C02=C04
C02=C04,C04=C08
[download]

Things I fixed:

• chomp
• split

toolic has been faster with posting. But as I also looked at the problem, I post my part of solution:

use Data::Dumper;
open (my $infile, "<", "Input.txt") || die $!;

while (defined(my $line = <$infile>)) {
    chomp $line;
    my @rxn = split /#/, $line;
    next if length $rxn[1] == 0;
    my ($left_side, $right_side) = split /\s*<=>\s*/, $rxn[1], 2;
    my @left_elements = split /\s*\+\s*/, $left_side;
    my @right_elements = split /\s*\+\s*/, $right_side;

    my @combinations;
    for my $left_element (@left_elements) {
        for my $right_element (@right_elements) {
            push @combinations,  [$left_element, $right_element];
        }
    }
    print join(", ", map { "$_->[0]=$_->[1]" } @combinations), "\n";
    # print Dumper(\@combinations);
}
[download]

In @combinations you now have a list of combinations, you can create edges from.

McA

G'day filipo,

I see ++toolic has pointed out issues in you code. Here's my take on a solution (pm_graph_undirected.pl) which uses glob to generate the combinations.

#!/usr/bin/env perl -l

use strict;
use warnings;
use autodie;

use Graph::Undirected;

my $graph = Graph::Undirected::->new();

open my $input_fh, '<', './pm_graph_undirected.txt';

while (<$input_fh>) {
    my ($x, $y) = /^\w+[#]([^<]+)\s+<=>\s+(.*)$/;
    
    $graph = $graph->add_edge(split /,/) for glob join ',' =>
        map { '{' . join(',' => split /[+ ]+/ => $_) . '}' } ($x, $y);
}   

print $graph;
[download]

Input:

$ cat pm_graph_undirected.txt
R1#C01+ C02 <=> C03 + C04
R2#C01 + C04 + C05 <=> C07 + C08 + C09
[download]

Output:

$ pm_graph_undirected.pl
C01=C03,C01=C04,C01=C07,C01=C08,C01=C09,C02=C03,C02=C04,C04=C07,C04=C0
+8,C04=C09,C05=C07,C05=C08,C05=C09
[download]

Update: ++choroba has pointed out a limitation with glob (below). I've tracked down details: see GLOB_LIMIT in File::Glob - POSIX FLAGS.

[Update2: very minor code change to remove an unnecessary comma s['}',]['}'] in the map. Retested and got the same output.]

Note that glob only generates strings of a limited length (4096 on my machine). Probably not an issue with a dozen of nodes with two letter names.
لսႽ† ᥲᥒ⚪⟊Ⴙᘓᖇ Ꮅᘓᖇ⎱ Ⴙᥲ𝇋ƙᘓᖇ

That great, thanks for the help.