SYNOPSIS

use XML::TreePuller;

$pull = XML::TreePuller->new(location => '/what/ever/filename.xml');
$pull = XML::TreePuller->new(location => 'http://urls.too/data.xml');
$pull = XML::TreePuller->new(IO => \*FH);
$pull = XML::TreePuller->new(string => '<xml/>');
#parse the document and return the root element
#takes same arguments as new()
$element = XML::TreePuller->parse(%ARGS);
$pull->reader; #return the XML::LibXML::Reader object
$pull->iterate_at('/xml', 'short'); #read the first part of an element
$pull->iterate_at('/xml', 'subtree'); #read the element and subtree

while($element = $pull->next) { }

$element->name;
$element->text; #fetch text for the element and all children
$element->attribute('attribute_name'); #get attribute value
$element->attribute; #returns hashref of attributes
$element->get_elements; #return all child elements
$element->get_elements('element/path'); #elements from path
$element->xpath('/xml'); #search using a XPath

ABOUT

This module implements a tree oriented XML pull processor providing fast and convenient unmarshalling of extremely large XML documents serially. Unmarshalling means the module is intended to turn the XML document into datastructures, not transform it. Tree oriented means the data is returned from the engine as a tree of data replicating the structure of the original XML document. Pull processor means you sequentially ask the engine for more data (the opposite of SAX). This engine also supports breaking the document into fragments so the trees are small enough to fit into RAM.

Features

High speed

This framework has been benchmarked to process XML between 1 meg/sec and 70 meg/sec in real world scenarios using the high level interface.

Work with documents too big to fit into RAM

The interface is nearly identical for large documents and small documents.

High level

The document is mapped to a high level XML element class that is easy to use.

Low level

If you need lower level access to the XML document you can treat the element class as a set of arrays representing the structure of your document or you can work with the XML::LibXML::Reader instance directly.

Justification

``Another XML processing scheme? Why don't you create a new template parsing framework to go with it!?'' --- If I had a trillion dollars for every time I've heard this I could bail out the US Government (as of Apr 26, 2010 that is). When I set out to create the replacement for Parse::MediaWikiDump I started by benchmarking the performance of existing XML processing frameworks (XML::SAX (all of them), XML::Parser, and higher level frameworks such as XML::Twig). The results of my research was that there exists no very fast pull oriented high level framework for processing XML.

I set about building MediaWiki::DumpFile using a base of XML::LibXML::Reader and XML::CompactTree; I wound up with a reconfigurable XML processing engine that I rather liked so I decided to publish it on CPAN.

STATUS

This software is currently ALPHA quality - the only known use is MediaWiki::DumpFile which is itself becoming tested in production. The API is not stable and there may be bugs: please report success and failure to the author below.

XML::TreePuller

METHODS

new

The constructor for this class returns an instance of itself; all arguments are passed straight on to XML::LibXML::Reader when it is constructed. See the documentation for a full specification of what you can use but for quick reference:

new(location => '/what/ever/filename.xml');

new(location => 'http://urls.work.too/data.xml');

new(string => $xml_data);

new(IO => \*FH);

parse

This method takes the same arguments as new() but parses the entire document into an element and returns it; you can use this if you don't need to break the document into chunks.

iterate_at

This method allows you to control the configuration of the processing engine; you specify two arguments: a path to an XML element and an instruction. The engine will move along node by node through the document and keep track of the full path to the current element. The combination of the current path of the XML document in the reader and the instruction to use will cause instances of XML::TreePuller::Element to be available from the ``next'' method.

If iterate_at() is never called then the entire document will be read into a single element at the first invocation of next().

iterate_at('/path/to/element' => 'short');

When the path of the current XML element matches the path specified the ``next'' method will return an instance of XML::TreePuller::Element that holds any attributes and will contain textual data up to the start of another element; there will be no child elements in this element.

iterate_at('/ditto' => 'subtree');

When the path of the current XML element matches the path specified the ``next'' method will return an instance of XML::TreePuller::Element that holds the attributes for the element and all of the element textual data and child elements.

next

This method is the iterator for the processing system. Each time an instruction is matched it will return an instance of XML::TreePuller::Element. When called in scalar context returns a reference to the next available element or undef when no more data is available. When called in list context it returns a two item list with the first item being the path to the node that was matched and the second item being the next available element; returns an empty list when there is no more data to be processed.

The returned path will always be a full path in the document starting at the root element and ending in the element that ultimately matched.

reader

Returns the instance of XML::LibXML::Reader that we are using to parse the XML document. You can move the cursor of the reader if you want but keep this in mind: if you move the cursor of the reader to an element in the document that is at a higher level than the reader was sitting at when you moved it then the reader must move the cursor to an element that was at the same depth in the document as it was at the start; this may cause some parts of the document to be thrown out that you are not expecting.

XML::TreePuller::Element

This class is how you access the data from XML::TreePuller. XML::TreePuller::Element is implemented as a set of methods that operate on arrays as returned by XML::CompactTree; you are free to work with XML::TreePuller::Element objects just as you would work with data returned from XML::CompactTree::readSubtreeToPerl() and such.

METHODS

name

Returns the name of the element as a string

text

Returns the text stored in the element and all subelements as a string; returns an empty string if there is no text

attribute

If called with out any arguments returns a hash reference containing the attribute names as keys and the attribute values as the data. If called with an argument returns the value for the attribute by that name or undef if there is no attribute by that name.

get_elements

Searches this element for any child elements as matched by the path supplied as an argument; the path is relative to the current element. The path is of the format 'element1/element2/element3' where each element name is separated by a forward slash and there are no trailing or leading forward slashes. If no path is specified it returns all of the child elements for the current element.

If called in scalar context returns the first element that matches the path; if called in array context returns a list of all elements that matched.

xpath

Perform an XPath query on the element and return the results; if called in list context you'll get all of the elements that matched. If called in scalar context you'll get the first element that matched. XPath support is currently EXPERIMENTAL.

The XPath query is rooted at the element so you must include the current element name as part of the path if you are specifying an absolute path to a subelement.

IMPROVING PERFORMANCE

First of all if you want to improve the throughput of this XML processing system be sure to install XML::CompactTree::XS - once installed this module is used automatically and drastically improves overall performance of unmarshalling the XML from the document (this does not involve XML::TreePuller::Element).

Secondly there are a number of ways to solve problems with this module, see XML::TreePuller::CookBook::Performance for information.

FURTHER READING

XML::TreePuller::CookBook::Intro

Gentle introduction to parsing using Atom as an example.

XML::TreePuller::CookBook::Performance

High performance processing of Wikipedia dump files.

XML::TreePuller::CookBook::Patterns

XPath Tutorial

• http://www.zvon.org/xxl/XPathTutorial/Output/example1.html
• http://www.w3schools.com/xpath/

MediaWiki::DumpFile::Pages

Object oriented recursive descent parser that maps Mediawiki XML dump files into high level Perl objects for working with the data.

LIMITATIONS

• This module is not XML compliant though it is built from XML compliant components. There may be unexpected behavior compared to proper XML behavior and if this is encountered please open a bug report.
• XPath support is EXPERIMENTAL (even more so than the rest of this module)
• There is only support for elements, text in elements, and CDATA blocks - other features of XML are not part of the API and are not tested but may bleed through from the underlying modules used to build this system. If you have an idea on how to add support for these extra features the author is soliciting feedback and patches.
• Things are pretty arbitrary right now as this module started life as the heart of MediaWiki::DumpFile; it would be nice to bring in more formal XML processing concepts.

ATTRIBUTION

With out the following people this module would not be possible:

Andrew Rodland

My Perl mentor and friend, his influence has helped me everywhere.

Petr Pajas

As the maintainer of XML::LibXML and creator of XML::CompactTree this module would not be possible with out building on his great work.

Michel Rodriguez

For creating Tree::XPathEngine which made adding XPath support a one day exercise.

AUTHOR

Tyler Riddle, "<triddle at cpan.org>"