Introduction

The vast majority of Perl's XML modules are built on top of XML::Parser, Larry Wall and Clark Cooper's Perl interface to James Clark's expat parser. The expat-XML::Parser combination is not the only full-featured XML parser available in the Perl World. This month we'll look at XML::LibXML, Matt Sergeant and Christian Glahn's Perl interface to Daniel Velliard's libxml2.

Why Would You Want Yet Another XML Parser?

Expat and XML::Parser have proven themselves to be quite capable, but they are not without limitations. Expat was among the first XML parsers available and, as a result, its interfaces reflect the expectations of users at the time it was written. Expat and XML::Parser do not implement the Document Object Model, SAX, or XPath language interfaces (things that most modern XML users take for granted) because either the given interface did not exist or was still being heavily evaluated and not considered "standard" at the time it was written.

The somewhat unfortunate result of this is that most of the available Perl XML modules are built upon one of XML::Parser's non- or not-quite-standard interfaces with the presumption that the input will be some sort of textual representation of an XML document (file, filehandle, string, socket stream) that must be parsed before proceeding. While this works for many simple cases, most advanced XML applications need to do more than one thing with a given document and that means that for each stage in the process, the document must be serialized to a string and then re-parsed by the next module.

By contrast libxml2 was written after the DOM, XPath, and SAX interfaces became common, and so it implements all three. In-memory trees can be built by parsing documents stored in files, strings, and so on, or generated from a series of SAX events. Those trees can then be operated on using the W3C DOM and XPath interfaces or used to generate SAX events that are handed off to external event handlers. This added flexibility, which reflects current XML processing expectations, makes XML::LibXML a strong contender for XML::Parser's throne.

Using XML::LibXML

This month's column may be seen as a addendum to the Perl/XML Quickstart Guide published earlier this year, when XML::LibXML was in its infancy, and we'll use the same tests from the Quickstart to put XML::LibXML though its paces. For a detailed overview of the test cases see the first installment in the Quickstart; but, to summarize, the two tests illustrate how to extract and print data from an XML document, and how to build and print, programmatically, an XML document from data stored in a Perl HASH using the facilities offered by a given XML module.

Reading

For accessing the data stored in XML documents, XML::LibXML provides a standard W3C DOM interface. Documents are treated as a tree of nodes and the data those nodes contain are accessed by calling methods on the node objects themselves.

use strict;
use XML::LibXML;

my $file = 'files/camelids.xml';
my $parser = XML::LibXML->new();
my $tree = $parser->parse_file($file);
my $root = $tree->getDocumentElement;
my @species = $root->getElementsByTagName('species');

foreach my $camelid (@species) {
    my $latin_name = $camelid->getAttribute('name');
    my @name_node  = $camelid->getElementsByTagName('common-name');
    my $common_name = $name_node[0]->getFirstChild->getData;
    my @c_node  = $camelid->getElementsByTagName('conservation');
    my $status =  $c_node[0]->getAttribute('status');
    print "$common_name ($latin_name) $status \n";
}

One of the more exciting features of XML::LibXML is that, in addition to the DOM interface, it allows you to select nodes using the XPath language. The following illustrates how to achieve the same effect as the previous example using XPath to select the desired nodes:

use strict;
use XML::LibXML;

my $file = 'files/camelids.xml';
my $parser = XML::LibXML->new();
my $tree = $parser->parse_file($file);
my $root = $tree->getDocumentElement;

foreach my $camelid ($root->findnodes('species')) {
    my $latin_name = $camelid->findvalue('@name');
    my $common_name = $camelid->findvalue('common-name');
    my $status =  $camelid->findvalue('conservation/@status');
    print "$common_name ($latin_name) $status \n";
}

What makes this exciting is that you can you can mix and match methods from the DOM and XPath interfaces to best suit the needs of your application, while operating on the same tree of nodes.

Writing

To create an XML document programmatically with XML::LibXML you simply use the provided DOM interface:

use strict;
use XML::LibXML;

my $doc = XML::LibXML::Document->new();
my $root = $doc->createElement('html');
$doc->setDocumentElement($root);
my $body = $doc->createElement('body');
$root->appendChild($body);

foreach my $item (keys (%camelid_links)) {
   my $link = $doc->createElement('a');
   $link->setAttribute('href', $camelid_links{$item}->{url});
   my $text = XML::LibXML::Text->new($camelid_links{$item}->{description});
   $link->appendChild($text);
   $body->appendChild($link);
}
print $doc->toString;

An important difference between XML::LibXML and XML::DOM is that libxml2's object model conforms to the W3C DOM Level 2 interface, which is better able to cope with documents containing XML Namespaces. So, where XML::DOM is limited to:

@nodeset = getElementsByTagName($element_name);

and

$node = $doc->createElement($element_name);

XML::LibXML also provides:

@nodeset = getElementsByTagNameNS($namespace_uri, $element_name);

and

$node = $doc->createElementNS($namespace_uri, $element_name);

The Joy of SAX

We've seen the DOM and XPath goodness that XML::LibXML provides, but the story does not end there. The libxml2 library also offers a SAX interface that can be used to create DOM trees from SAX events or generate SAX events from DOM trees.

The following creates a DOM tree programmatically from a SAX driver built on XML::SAX::Base. In this example, the initial SAX events are generated from a custom driver implemented in the CamelDriver class that calls the handler events in the XML::LibXML::SAX::Builder class to build the DOM tree.

use XML::LibXML;
use XML::LibXML::SAX::Builder;

my $builder = XML::LibXML::SAX::Builder->new();
my $driver = CamelDriver->new(Handler => $builder);
my $doc = $driver->parse(%camelid_links);

# doc is an XML::LibXML::Document object
print $doc->toString;

package CamelDriver;
use base qw(XML::SAX::Base);

sub parse {
  my $self = shift;
  my %links = @_;
  $self->SUPER::start_document;
  $self->SUPER::start_element({Name => 'html'});
  $self->SUPER::start_element({Name => 'body'});

  foreach my $item (keys (%camelid_links)) {
    $self->SUPER::start_element({Name => 'a',
                                   Attributes => {
                                     'href' => $links{$item}->{url}
                                               }
                                });
    $self->SUPER::characters({Data => $links{$item}->{description}});
    $self->SUPER::end_element({Name => 'a'});
  }

  $self->SUPER::end_element({Name => 'body'});
  $self->SUPER::end_element({Name => 'html'});
  $self->SUPER::end_document;

}
1;

You can also generate SAX events from an existing DOM tree using XML::LibXML::SAX::Generator. In the following snippet, the DOM tree created by parsing the file camelids.xml is handed to XML::LibXML::SAX::Generator's generate() method which in turn calls the event handlers in XML::Handler::XMLWriter to print the document to STDOUT.

use strict;
use XML::LibXML;
use XML::LibXML::SAX::Generator;
use XML::Handler::XMLWriter;

my $file = 'files/camelids.xml';
my $parser = XML::LibXML->new();
my $doc = $parser->parse_file($file);
my $handler = XML::Handler::XMLWriter->new();
my $driver = XML::LibXML::SAX::Generator->new(Handler => $handler);

# generate SAX events that are captured
# by a SAX Handler or Filter.
$driver->generate($doc);

Resources

Download the sample code. Perl XML Quickstart: The Standard XML Interfaces Writing SAX Drivers for Non-XML Data Transforming Data With SAX Filters

This ability to accept and emit SAX events is especially useful in light of the recent discussion in this column of generating SAX events from non-XML data and writing SAX filter chains. You could, for example, use a SAX driver written in Perl to emit events based on data returned from a database query that creates a DOM object, which is then transformed in C-space for display using XSLT and the mind-numbingly fast libxslt library (which expects libxml2 DOM objects), and then emit SAX events from that transformed DOM tree for further processing using custom SAX filters to provide the finishing touches -- all without once having had to serialize the document to a string for re-parsing. Wow.

Conclusions

As we have seen, XML::LibXML offers a fast, updated approach to XML processing that may be superior to the first-generation XML::Parser for many cases. Do not misunderstand, XML::Parser and its dependents are still quite useful, well-supported, and are not likely to go away any time soon. But it is not the only game in town, and given the added flexibility that XML::LibXML provides, I would strongly encourage you to give XML::LibXML a closer look before beginning your next Perl/XML project.

XML.com Copyright © 1998-2006 O'Reilly Media, Inc.