Often data is stored in XML files must be massaged into other formats (e.g. DocBook to roff). There are well developed procedures for defining such transformations (XSLT) and a number of tools to apply them (e.g. xsltproc).

Besides the W3 tutorial, there is also a nice introduction by Paul Grosso and Norman Walsh. I've copied a simple example from this intro and also included a slightly more complicated setup for generating online help for a list of macros.

XSLT is also useful for standardizing XML content. For example, I was recently trying to compare to Gramps XML files, to see what had changed between two database backups. Unfortunately, the backup XML was not sorted by id, so there were many diff chunks due to node shuffling that didn't represent any useful information. With the following XSLT:

<?xml version="1.0"?>
<xsl:stylesheet version="1.0"
    <xsl:output method="xml" indent="yes"/>
    <xsl:strip-space elements="*"/>
    <!-- identity transform -->
    <xsl:template match="@*|node()">
            <xsl:apply-templates select="@*|node()"/>
    <!-- sort node children by their `id` attributes -->
    <xsl:template match="node()">
            <xsl:apply-templates select="@*"/>
            <xsl:for-each select="node()">
                <xsl:sort select="@id" order="ascending"/>
                <xsl:apply-templates select="."/>

With the above saved as sort-by-id.xsl, you can sort some.xml using

$ xsltproc --nonet --novalid sort-by-id.xsl some.xml

You can compare two Gramps XML files with

$ diff -u <(zcat a.gramps | xsltproc --nonet --novalid sort-by-id.xsl -)
          <(zcat b.gramps | xsltproc --nonet --novalid sort-by-id.xsl -) | less

Jesper Tverskov has a nice page about the identity template and related tricks if you want more examples of quasi-copy transforms.