Of course it is possible. I like using handlers because then you get rid of some of the tree before having to load it entirely in memory, plus I am used to it, and I find it often simpler than using XPath or navigation to get to the elements I want to process.

If size is not a problem you can certainly load the tree, and then write custom code to output the element and the tree "above it":

#!/usr/local/bin/perl -w

use strict;
use XML::Twig;

my $xml_twig = XML::Twig->new(
    pretty_print => 'indented',
    NoLWP => 1,
    discard_spaces => 1,
    EltClass => 'my_elt',
);

$xml_twig->parse ( \*DATA ) || die "\nError parsing data $@\n";

# here I split on the 'split' attribute, but any other
# way of getting the list of elements to split on is OK
foreach my $elt ($xml_twig->findnodes( '//*[@split]'))
  { print "Element: ", $elt->name, "\n\n", $elt->sub_doc, "\n\n";
    $elt->delete;
  }

package my_elt;

use base 'XML::Twig::Elt';

sub sub_doc    { my $elt= shift; return $elt->start_tags . $elt->sprin
+t . $elt->end_tags; }
sub start_tags { my $elt= shift; return join '', map { $_->start_tag }
+ reverse $elt->ancestors; }
sub end_tags   { my $elt= shift; return join '', map { $_->end_tag   }
+ reverse $elt->ancestors; }

package main;
__DATA__
<myconfig_data foo="bar">
        <level1 split="1">
                <tag>hello1</tag>
        </level1>
        <level2>
          <sub split="1">
                <tag>hello2</tag>
          </sub>
          <sub split="1">
                <tag>hello3</tag>
          </sub>
        </level2>
        <level3 split="1">
                <tag>hello4</tag>
        </level3>
</myconfig_data>
[download]