To emphasise and make a little clearer what previous replies have said: It is almost never appropriate to use a regex to manipulate XML. The two biggest problems are whitespace and nested elements. It is very hard, and in many cases not possible in any sensible way, to write regexen to perform edits on XML.

As mentioned by marto, XML::Twig is the way to do this sort of thing:

use strict;
use warnings;
use XML::Twig;

my $t= XML::Twig->new
  (
  twig_roots => {'figure' => \&doSwap},
  twig_print_outside_roots => 1,
  );
  
my $source = do {local $/ = ''; <DATA>};
$t->set_pretty_print ('record');
$t->parse ($source);


sub doSwap
{
my ($t, $figure)= @_;
my @title = $figure->cut_children ('title');
$title[0]->paste ('last_child', $figure);
$figure->print;
}
[download]

use strict;
use warnings;
use XML::Twig;

my $infile = "in-test.xml";
my $outfile = "out-test.xml";

open(OUT, ">$outfile") or die "can\'t open output file $outfile:$!";
open(IN,  "$infile")   or die "can\'t open input file $infile:$!";

my $t = XML::Twig->new
    (
    twig_roots => {'figure' => \&doSwap},
    twig_print_outside_roots => \*OUT,
    );
$t->parse (\*IN);

close OUT or die "Could not close $outfile";
close IN or  die "Could not close $infile";

sub doSwap {
    my ($t, $figure) = @_;                          # required argumen
+ts for twig handlers
    my @title = $figure->cut_children ('title');    # extract (the onl
+y) <title> node from the <figure> node
    $title[0]->paste (last_child => $figure);       # make it the last
+ child of that <figure> node
    $figure->print(\*OUT);                          # print the rearra
+nged <figure> node
}
[download]

Below is an XSLT solution to my swapping problem. It was quite interesting that it took the XSLT engine I was using (MSXML4.0) over 21 seconds to generate an output file on my 3MB input (45K+ XML nodes). The Perl program I posted last week generated an output file in a couple of seconds.

<xsl:stylesheet version="1.0"  xmlns:xsl="http://www.w3.org/1999/XSL/T
+ransform">
<!-- 12/01/2005 jkulas@lsijax.com -->
    <!-- This transforms figure by moving title to last_child position
+ -->
    <!-- Functionally similar to twig_roots -->
    <xsl:template match ="figure">
        <xsl:element name="figure">
            <xsl:copy-of select="@*"/>
            <xsl:apply-templates/>
            <!-- Add the title node as last child-->
            <xsl:if test="title">
                <xsl:element name ="title">
                    <xsl:copy-of select="title/@*"/>
                    <xsl:copy-of select="title/text()"/>
                </xsl:element>
            </xsl:if>
        </xsl:element>
    </xsl:template>

    <!-- Ignore any figure/title's since they are handled in the figur
+e template-->
    <xsl:template match ="figure/title">
    </xsl:template>

    <!-- Copy all nodes, except figure nodes (of course)-->
    <!-- Adapted from Michael Kay, _XSLT 2.0_, 3/e, p. 243-->
    <!-- Functionally similar to twig_print_outside_roots -->
    <xsl:template match ="@* | node()">
        <xsl:copy> 
            <xsl:copy-of select="@*"/>
            <xsl:apply-templates/>
        </xsl:copy>
    </xsl:template>

</xsl:stylesheet>
[download]

Peace,
--Jack

Here's what I observe: There may or may not be a line break between the elements. However, your regex assumes that there is not. Somehow, you need to get the entire document into one variable, and run the substitutions on that variable, making sure you account for possible newlines in the variable.

    -Bryan