as said above, if you have any control over the format of this data you should get it changed to either an easier to parse proprietary format or get it into valid xml. As it stands a full xml parser wouldn't touch it with a barge poll (and rightly so)

having said that, you could approach your regex from the other end and work backwards. I think this does what you want.

while ( $content =~ s/(.*<level1 id="([^"]*).*?)<level2>/$1<level2 id=
+\"$2\">/gsi ) {  }
[download]

I feel dirty just posting that though, as it's so flaky a single white space would break it. Think of it as an example as to why you should fix your file format :)

UPDATE

I couldn't shake the feeling inflicted on myself by the above post. I seek to redeem myself with a full xml version :)

#!/usr/bin/perl
use XML::DOM;
use warnings;
use strict;
 

my $xml = q|<root>
<level1 id="L1_0001">
    <level2>
        <level3>
            <level4/>
        </level3>
        <level3>
            <level4/>
        </level3>
    </level2>
    <level2/>
</level1>   
<level1 id="L1_0002">
</level1>   
<level1 id="L1_0003">
    <level2>
        <level3/>
    </level2>
</level1>
</root>|;

my $parser = new XML::DOM::Parser;
my $doc = $parser->parse($xml);

foreach my $l1_node ($doc->getElementsByTagName ('level1') ) {
    my $current_id = $l1_node->getAttribute('id');
    foreach my $l2_node ($l1_node->getElementsByTagName ('level2') ) {
        $l2_node->setAttribute('id', $current_id);
    }    
}

print $doc->toString;
exit();
[download]