OK, let's clear some misconceptions about XML::Twig with some explanations and a bit of code (I'll add this to the doc of the module for next version):

• if you are using twig_roots then the tree will be composed only of the document root, the twig_roots elements and all of their sub-elements,
• twig_handlers on the other hands do not impact the way the tree is built, the handler is called as soon as the element is completely parsed, and the entire tree that has been built previously, including all of the element sub-elements is accessible...
• except when when you use either flush or purge: in this case the module will delete as much of the tree as can possibly be, which means that only the direct ancestors of the current element will remain, each with one single child.

Now for something completely similar:

#!/bin/perl -w
use strict;
use XML::Twig;
my $doc=
'<doc>
  <l1>
    <l2>
      <e1>e1</e1>
      <e2>e2</e2>
      <e3>e3</e3>
    </l2>
  </l1>
</doc>';

my $twig= new XML::Twig( twig_roots    => { l2 => 1 }
                         twig_handlers => { e2 => \&e2}
                       );
$twig->parse( $doc);

sub e2
  { my( $t, $e2)= @_;

    print "ancestors: ";
    print map { $_->gi . " " } $e2->ancestors;   # print l2 doc
    print "\n";

    print "siblings: ";
    print $e2->prev_sibling->gi;                 # print e1
    print " no next sibling\n" unless $e2->next_sibling; # e3 is not t
+here yet

    print "purge!\n";
    $t->purge;

    print "ancestors: ";
    print map { $_->gi . " " } $e2->ancestors;   # print l2 doc (still
+ there)
    print "\n";


    print "siblings: ";
    print " no prev sibling\n" unless $e2->prev_sibling; # e1 is gone
    print " no next sibling\n" unless $e2->next_sibling; # e3 is still
+ not there
  }
[download]