XML::LibXML - Interface to the gnome libxml2 library


  use XML::LibXML;
  my $parser = XML::LibXML->new();

  my $doc = $parser->parse_string(<<'EOT');


This module is an interface to the gnome libxml2 DOM parser (no SAX parser support yet), and the DOM tree. It also provides an XML::XPath-like findnodes() interface, providing access to the XPath API in libxml2.


LibXML options are global (unfortunately this is a limitation of the underlying implementation, not this interface). They can either be set using $parser->option(...), or XML::LibXML->option(...), both are treated in the same manner. Note that even two forked processes will share some of the same options, so be careful out there!

Every option returns the previous value, and can be called without parameters to get the current value.



Turn validation on (or off). Defaults to off.



Turn the parsers recover mode on (or off). Defaults to off.

This allows to parse broken XML data into memory. This switch will only work with XML data rather than HTML data. Also the validation will be switched off automaticly.

The recover mode helps to recover documents that are almost wellformed very efficiently. That is for example a document that forgets to close the document tag (or any other tag inside the document). The recover mode of XML::LibXML has problems though to restore documents that are more like well ballanced chunks. In that case XML::LibXML will only parse the first tag of the chunk.



Turn entity expansion on or off, enabled by default. If entity expansion is off, any external parsed entities in the document are left as entities. Probably not very useful for most purposes.



Allows you to turn off XML::LibXML's default behaviour of maintaining whitespace in the document.



You can make XML::LibXML more pedantic if you want to.



Load external DTD subsets while parsing.



Complete the elements attributes lists with the ones defaulted from the DTDs. By default, this option is enabled.



Expands XIinclude tags imidiatly while parsing the document. This flag ashures that the parser callbacks are used while parsing the included Document.


  $parser->load_catalog( $catalog_file );

Will use $catalog_file as a catalog during all parsing processes. Using a catalog will significantly speed up parsing processes if many external ressources are loaded into the parsed documents (such as DTDs or XIncludes)

Note that catalogs will not be available if an external entity handler was specified. At the current state it is not possible to make use of both types of resolving systems at the same time.


  $parser->base_uri( $your_base_uri );

In case of parsing strings or file handles, XML::LibXML doesn't know about the base uri of the document. To make relative references such as XIncludes work, one has to set a separate base URI, that is then used for the parsed documents.



Although quite powerful XML:LibXML's DOM implementation is limited if one needs or wants full DOM level 2 or level 3 support. XML::GDOME is based on libxml2 as well but provides a rather complete DOM implementation by wrapping libgdome. This allows you to make use of XML::LibXML's full parser options and XML::GDOME's DOM implementation at the same time.

All XML::LibXML parser functions recognize this switch.



Sets a "match" callback. See "Input Callbacks" below.



Sets an open callback. See "Input Callbacks" below.



Sets a read callback. See "Input Callbacks" below.



Sets a close callback. See "Input Callbacks" below.


The XML::LibXML constructor, new(), takes the following parameters:


  my $parser = XML::LibXML->new(ext_ent_handler => sub { ... });

The ext_ent_handler sub is called whenever libxml needs to load an external parsed entity. The handler sub will be passed two parameters: a URL (SYSTEM identifier) and an ID (PUBLIC identifier). It should return a string containing the resource at the given URI.

Note that you do not need to enable this - if not supplied libxml will get the resource either directly from the filesystem, or using an internal http client library.


  my $parser = XML::LibXML->new( catalog => $private_catalog );

Alternatively to ext_ent_handler the catalog parameter allows to use libxml2's catalog interface directly. The parameter takes a filename to a catalog file. This catalog is loaded by libxml2 and will be used during parsing processes.

Note that catalogs will not be available if an external entity handler was specified. At the current state it is not possible to make use of both types of resolving systems at the same time.


The following table gives an overview about the default values of the parser attributes.

validation == off (0)
recover == off (0)
expand_entities == on (1)
keep_blanks == on (1)
pedantic_parser == off (0)
load_ext_dtd == on (1)
complete_attributes == on (1)
expand_xinclude == off (0)
base_uri == ""
gdome_dom == off (0)

By default no callback handler is set.


There are three ways to parse documents - as a string, as a Perl filehandle, or as a filename. The return value from each is a XML::LibXML::Document object, which is a DOM object (although not all DOM methods are implemented yet). See "XML::LibXML::Document" below for more details on the methods available on documents.

Each of the below methods will throw an exception if the document is invalid. To prevent this causing your program exiting, wrap the call in an eval{} block.


  my $doc = $parser->parse_string($string);

or, passing in a directory to use as the "base":

  my $doc = $parser->parse_string($string, $dir);


  my $doc = $parser->parse_fh($fh);

Here, $fh can be an IOREF, or a subclass of IO::Handle.

And again, you can pass in a directory as the "base":

  my $doc = $parser->parse_fh($fh, $dir);

Note in the above two cases, $dir must end in a trailing slash, otherwise the parent of that directory is used. This can actually be useful, in that it will accept the filename of what you're parsing.


  my $doc = $parser->parse_file($filename);

This function reads an absolute filename into the memory. It causes XML::LibXML to use libxml2's file parser instead of letting perl reading the file such as with parse_fh(). If you need to parse files directly, this function would be the faster choice, since this function is about 6-8 times faster then parse_fh().

Parsing Html

As of version 0.96, XML::LibXML is capable of parsing HTML into a regular XML DOM. This gives you the full power of XML::LibXML on HTML documents.

The methods work in exactly the same way as the methods above, and return exactly the same type of object. If you wish to dump the resulting document as HTML again, you can use $doc-toStringHTML()> to do that.


  my $doc = $parser->parse_html_string($string);


  my $doc = $parser->parse_html_fh($fh);


  my $doc = $parser->parse_html_file($filename);

Push Parser

XML::LibXML supports also a push parser interface. This allows one to parse large documents without actually loading the entire document into memory.

The interface is devided into two parts:

  • pushing the data into the parser

  • finish the parse

The user has no chance to access the document while still pushing the data to the parser. The resulting document will be returned when the parser is told to finish the parsing process.

$parser->push( @data )

This function pushs the data stored inside the array to libxml2's parse. Each entry in @data must be a normal scalar!

$parser->finish_push( $restore );

This function returns the result of the parsing process. If this function is called without a parameter it will complain about non wellformed documents. If $restore is 1, the push parser can be used to restore broken or non well formed (XML) documents as the following example shows:

  $parser->push( "<foo>", "bar" );
  eval { $doc = $parser->finish_push(); };      # will complain
  if ( $@ ) {
     # ...

This can be anoing if the closing tag misses by accident. The following code will restore the document:

  $parser->push( "<foo>", "bar" );
  eval { $doc = $parser->finish_push(1); };      # will not complain

  warn $doc->toString(); # returns "<foo>bar</foo>"

of course finish_push() will return nothing if there was no data pushed to the parser before.

Extra parsing methods


  $parser->processXIncludes( $doc );

While the document class implements a separate XInclude processing, this method, is stricly related to the parser. The use of this method is only required, if the parser implements special callbacks that should to be used for the XInclude as well.

If expand_xincludes is set to 1, the method is only required to process XIncludes appended to the DOM after its original parsing.

Error Handling

XML::LibXML throws exceptions during parseing, validation or XPath processing. These errors can be catched by useing eval blocks. The error then will be stored in $@. Alternatively one can use the get_last_error() function of XML::LibXML. It will return the same string that is stored in $@. Using get_last_error() makes it still nessecary to eval the statement, since these function groups will die() on errors.

get_last_error() can be called either by the class itself or by a parser instance:

   $errstring = XML::LibXML->get_last_error();
   $errstring = $parser->get_last_error();

Note that XML::LibXML exceptions are global. That means if get_last_error is called on an parser instance, the last global error will be returned. This is not nessecarily the error caused by the parser instance itself.


The oposite of parsing is serialization. In XML::LibXML this can be done by using the functions toString(), toFile() and toFH(). All serialization functions understand the flag setTagCompression. if this Flag is set to 1 empty tags are displayed as <foo></foo> rather than <foo/>.

toString() additionally checks two other flags:

skipDTD and skipXMLDeclaration

If skipDTD is specified and any DTD node is found in the document this will not be serialized.

If skipXMLDeclaration is set to 1 the documents xml declaration is not serialized. This flag will cause the document to be serialized as UTF8 even if the document has an other encoding specified.

XML::LibXML does not define these flags itself, therefore they have to specify them manually by the caller:

 local $XML::LibXML::skipXMLDeclaration = 1;
 local $XML::LibXML::skipDTD = 1;
 local $XML::LibXML::setTagCompression = 1;

will cause the serializer to avoid the XML declaration for a document, skip the DTD if found, and expand empty tags.

*NOTE* $XML::LibXML::skipXMLDeclaration and $XML::LibXML::skipDTD are only recognized by the Documents toString() function.

Additionally it is possible to serialize single nodes by using toString() for the node. Since a node has no DTD and no XML Declaration the related flags will take no effect. Nevertheless setTagCompression is supported.

All basic serialization function recognize an additional formating flag. This flag is an easy way to format complex xml documents without adding ignoreable whitespaces.

Input Callbacks

The input callbacks are used whenever LibXML has to get something other than external parsed entities from somewhere. The input callbacks in LibXML are stacked on top of the original input callbacks within the libxml library. This means that if you decide not to use your own callbacks (see match()), then you can revert to the default way of handling input. This allows, for example, to only handle certain URI schemes.

Callbacks are only used on files, but not on strings or filehandles. This is because LibXML requires the match event to find out about which callback set is shall be used for the current input stream. LibXML can decide this only before the stream is open. For LibXML strings and filehandles are already opened streams.

The following callbacks are defined:


If you want to handle the URI, simply return a true value from this callback.


Open something and return it to handle that resource.

read(handle, bytes)

Read a certain number of bytes from the resource. This callback is called even if the entire Document has already read.


Close the handle associated with the resource.


This is a purely fictitious example that uses a MyScheme::Handler object that responds to methods similar to an IO::Handle.

  sub match_uri {
    my $uri = shift;
    return $uri =~ /^myscheme:/;
  sub open_uri {
    my $uri = shift;
    return MyScheme::Handler->new($uri);
  sub read_uri {
    my $handler = shift;
    my $length = shift;
    my $buffer;
    read($handler, $buffer, $length);
    return $buffer;
  sub close_uri {
    my $handler = shift;

A more realistic example can be found in the "example" directory

Since the parser requires all callbacks defined it is also possible to set all callbacks with a single call of callbacks(). This would simplify the example code to:

  $parser->callbacks( \&match_uri, \&open_uri, \&read_uri, \&close_uri);

All functions that are used to set the callbacks, can also be used to retrieve the callbacks from the parser.

Global Callbacks

Optionaly it is possible to apply global callback on the XML::LibXML class level. This allows multiple parses to share the same callbacks. To set these global callbacks one can use the callback access functions directly on the class.

  XML::LibXML->callbacks( \&match_uri, \&open_uri, \&read_uri, \&close_uri);

The previous code snippet will set the callbacks from the first example as global callbacks.


All data will be stored UTF-8 encoded. Nevertheless the input and output functions are aware about the encoding of the owner document. By default all functions will assume, UTF-8 encoding of the passed strings unless the owner document has a different encoding. In such a case the functions will assume the encoding of the document to be valid.

At the current state of implementation query functions like findnodes(), getElementsByTagName() or getAttribute() accept only UTF-8 encoded strings, even if the underlaying document has a different encoding. At first this seems to be a limitation, but on application level there is no way to make save asumptations about the encoding of the strings.

Future releases will offer the opportunity to force an application wide encoding, so make shure that you installed the latest version of XML::LibXML.

To encode or decode a string to or from UTF-8 XML::LibXML exports two functions, which use the encoding mechanism of the underlaying implementation. These functions should be used, if external encoding is required (e.g. for queryfunctions).


    $encodedstring = encodeToUTF8( $name_of_encoding, $sting_to_encode );

The function will encode a string from the specified encoding to UTF-8.


    $decodedstring = decodeFromUTF8($name_of_encoding, $string_to_decode );

This Function transforms an UTF-8 encoded string the specified encoding. While transforms to ISO encodings may cause errors if the given stirng contains unsupported characters, this function can transform to UTF-16 encodings as well.



Although both modules make use of libxml2's XML capabilities, the DOM implementation of both modules are not compatible. But still it is possible to exchange nodes from one DOM to the other. The concept of this exchange is pretty similar to the function cloneNode(): The particular node is copied on the lowlevel to the opposite DOM implementation.

Since the DOM implementations cannot coexist with in one document, one is forced to copy each node that should be used. Because of keeping allways two nodes this may cause quite an impact on a machines memory useage.

XML::LibXML provides two functions to export or import GDOME nodes: import_GDOME() and export_GDOME(). Both function have two parameters: the node and a flag for recursive import. The flag works as in cloneNode().


  XML::LibXML->import_GDOME( $node, $deep );

This converts an XML::GDOME node to XML::LibXML explicitly.


  XML::LibXML->export_GDOME( $node, $deep );

Allows to export an XML::LibXML node to XML::GDOME explicitly.

Although these two explicit functions exist, XML::LibXML allows also the transparent import of XML::GDOME nodes in functions such as appendChild(), insertAfter() and so on. While native nodes are automaticly adopted in most functions XML::GDOME nodes are allways cloned in advance. Thus if the original node is modified after the operation, the node in the XML::LibXML document will not have this information.


This module allows you to parse and return a DTD object. It has one method right now, new().


  my $dtd = XML::LibXML::Dtd->new($public, $system);

Creates a new DTD object from the public and system identifiers. It will automatically load the objects from the filesystem, or use the input callbacks (see "Input Callbacks" below) to load the DTD.

Processing Instructions - XML::LibXML::PI

Processing instructions are implemented with XML::LibXML with read and write access ;) The PI data is the PI without the PI target (as specified in XML 1.0 [17]) as a string. This string can be accessed with getData as implemented in XML::LibXML::Node.

The write access is aware about the fact, that many processing instructions have attribute like data. Therefor setData provides besides the DOM spec conform Interface to pass a set of named parameter. So the code segment

    my $pi = $dom->createProcessingInstruction("abc");
    $pi->setData(foo=>'bar', foobar=>'foobar');
    $dom->appendChild( $pi );

will result the following PI in the DOM:

    <?abc foo="bar" foobar="foobar"?>

The same can be done with

   $pi->setData( 'foo="bar" foobar="foobar"' );

Which is how it is specified in the "DOM specification". This three step interface creates temporary a node in perl space. This can be avoided while using the insertProcessingInstruction method. Instead of the three calls described above, the call $dom-insertProcessingInstruction("abc",'foo="bar" foobar="foobar"');> will have the same result as above.

Currently only the setData() function accepts named parameters, while only strings are accepted by the other methods.



   $pinode = $dom->createProcessingInstruction( $target );


   $pinode = $dom->createProcessingInstruction( $target, $data );

This function creates a new PI and returns this node. The PI is bound to the DOM, but is not appended to the DOM itself. To add the PI to the DOM, one needs to use appendChild() directly on the dom itself.



  $dom->insertProcessingInstruction( $target, $data );

Creates a processing instruction and inserts it directly to the DOM. The function does not return a node.


alias for createProcessingInstruction


alias for insertProcessingInstruction



   $pinode->setData( $data_string );


   $pinode->setData( name=>string_value [...] );

This method allows to change the content data of a PI. Additionaly to the interface specified for DOM Level2, the method provides a named parameter interface to set the data. This parameterlist is converted into a string before it is appended to the PI.


Matt Sergeant,

Copyright 2001, Ltd. All rights reserved.