What's the problem exactly? You don't know how to download the XML documents whose urls you extract, or you don't know how to parse the downloaded documents with the code I posted?

use strict;
use warnings;
use feature qw( say );

use LWP::UserAgent            qw( );
use URI                       qw( );
use URI::file                 qw( );
use XML::LibXML               qw( );
use XML::LibXML::XPathContext qw( );

my $parser = XML::LibXML->new();
my $xpc = XML::LibXML::XPathContext->new();
$xpc->registerNs('xi', 'http://www.w3.org/2001/XInclude');
my $ua = LWP::UserAgent->new();

my $root_url = URI->new_abs($ARGV[0], URI::file->cwd());

my @todo = $root_url;
my %found;
while (@todo) {
   my $url = pop(@todo);

   my $response = $ua->get($url);
   if (!$response->is_success()) {
      warn("Can't get $url: " . $response->status_line() . "\n");
      next;
   }

   my $xml = $response->decoded_content( charset => 'none' );
   my $doc = $parser->parse_string($xml);
   for ($xpc->findnodes('//xi:include/@href', $doc)) {
      my $child_url = URI->new_abs($_->getValue(), $url);
      push @todo, $child_url
         if !$found{$child_url}++;
   }
}

say for sort keys %found;
[download]

Update: Fixed constructor. Made url absolute as required.

Sorry, you misunderstood me - I'm not trying to download any files from remote sites, all my files are local

My point is that in order to make the parser to expand XInclude's I have to call it with option "xinclude => 1" - but in this case the resulting tree doesn't contain any original filenames, from which XML pieces have been included

If I call the parser without this option (which is default), then it won't expand these XInclude's, so only the first-level filenames will be here (because no expansion will be happening)

So, in both cases I can't find ALL filenames I need

Alex

I'm not trying to download any files from remote sites, all my files are local

It makes no difference if the urls point to remote or local resources. LWP::UserAgent will properly handle file: urls.

So, in both cases I can't find ALL filenames I need

Wee! I've done the impossible!

I'm not trying to download any files from remote sites, all my files are local

It makes no difference if the urls point to remote or local resources. LWP::UserAgent will properly handle file: urls.

So, in both cases I can't find ALL filenames I need

Wee! I've done the impossible!