#!/usr/bin/perl 

use strict;
use warnings;

use WWW::Mechanize;

my $url = shift || die "Please pass in base url as argument to $0\n";
my %visited;
my @links;
my $max_depth = 3;
my $depth = 0;

my $mech = WWW::Mechanize->new();

# This helps prevent following off-site links.
# Note, assumes that url's passed in will represent the
# highest level in a website hierarchy that will be visited.
# i.e. http://www.example.com/dir/ will record a link to
# http://www.example.com/, but will not follow it and report
# subsequent links.
my( $base_uri ) = $url =~ m|^(.*/)|;

get_links( $url );

sub get_links {
  my @urls = @_;
  my @found_links;
  for( @urls ){
    # This prevents following off-site or off-parent links.
    next unless m/^$base_uri/;
    $mech->get( $_ );
    # Filters out links we've already visited, plus mailto's and
    # javascript:etc hrefs.  Adjust to suit.
    @found_links = grep { ++$visited{$_} == 1 && ! /^(mailto|javascript)/i  }
                   map  { $_->url_abs() }
                   $mech->links();
    push @links, @found_links;
  }
  # Keep going, as long as we should.
  get_links( @found_links ) if $depth++ < $max_depth;
}

# Instead of printing them, you could insert them into the database.
print $_ . "\n" for @links;