Question

On Server Fault, How to list symbolic link chains? (not my question) talks about listing all the symbolic links and following them. To make this doable, let’s consider a single directory at first.

I want to write a short utility that does this. It looks easy to put pairs from symbolic links into a hash and then process the hash.

But then I might have something like:

ls -l
total 0
lrwxrwxrwx 1 pjb pjb 1 2010-02-23 08:48 a -> b
lrwxrwxrwx 1 pjb pjb 1 2010-02-23 08:48 b -> c
lrwxrwxrwx 1 pjb pjb 1 2010-02-23 09:03 c -> a
lrwxrwxrwx 1 pjb pjb 1 2010-02-23 09:17 trap -> b
lrwxrwxrwx 1 pjb pjb 1 2010-02-23 09:17 x -> y
lrwxrwxrwx 1 pjb pjb 1 2010-02-23 09:17 y -> b

where it is obvious that a->b->c is a loop, and that trap points into a loop, but to know x points into a loop I need to follow a bit.

One hash representation is:

a => b
b => c
c => a
trap => b
x => y
y => b

But the reverse representation is better for marking loops to bad starting points, once I know what the loops are.

So here’s some questions:

• Is a hash the best structure to represent symbolic links?
• what’s the best way to separate the graph of the file system to tell the loopy components from the tree components to the twig with a loop type pieces?
• Is there a better algorithm than manually searching for all the loops from all the starting points?
• From a graph-theory perspective — is this sort of thing in the CPAN already? If not, what are some good helper modules?

Answer 1

There’s a Graph module on CPAN that you might use as in the following:

#! /usr/bin/perl

use warnings;
use strict;

use Graph;

my $g = Graph->new;
my $dir = @ARGV ? shift : ".";

opendir my $dh, $dir or die "$0: opendir $dir: $!";
while (defined(my $name = readdir $dh)) {
  my $path = $dir . "/" . $name;

  if (-l $path) {
    my $dest = readlink $path;
    die "$0: readlink $path: $!" unless defined $dest;

    $g->add_edge($name => $dest);
  }
  else {
    $g->add_vertex($name);
  }
}

my @cycle = $g->find_a_cycle;
if (@cycle) {
  $" = ' -> '; #" # highlighting error
  print "$0: $dir: at least one cycle: @cycle\n";
}
else {
  print "$0: $dir: no cycles\n";
}

For example, in a directory similar in structure to the one in your question, the output is

$ ../has-cycle 
../has-cycle: .: at least one cycle: c -> a -> b