XML::TokeParser - Simplified interface to XML::Parser


    use XML::TokeParser;
    #parse from file
    my $p = XML::TokeParser->new('file.xml')
    #parse from open handle
    open IN, 'file.xml' or die $!;
    my $p = XML::TokeParser->new( \*IN, Noempty => 1 );
    #parse literal text
    my $text = '<tag xmlns="">text</tag>';
    my $p    = XML::TokeParser->new( \$text, Namespaces => 1 );
    #read next token
    my $token = $p->get_token();
    #skip to <title> and read text
    #read text of next <para>, ignoring any internal markup
    #process <para> if interesting text
    $t = $p->get_tag('para');
    if ( $p->get_trimmed_text('/para') =~ /interesting stuff/ ) {


XML::TokeParser provides a procedural ("pull mode") interface to XML::Parser in much the same way that Gisle Aas' HTML::TokeParser provides a procedural interface to HTML::Parser. XML::TokeParser splits its XML input up into "tokens," each corresponding to an XML::Parser event.

A token is a bless'd reference to an array whose first element is an event-type string and whose last element is the literal text of the XML input that generated the event, with intermediate elements varying according to the event type.

Each token is an object of type XML::TokeParser::Token. Read "XML::TokeParser::Token" to learn what methods are available for inspecting the token, and retrieving data from it.


$p = XML::TokeParser->new($input, [options])

Creates a new parser, specifying the input source and any options. If $input is a string, it is the name of the file to parse. If $input is a reference to a string, that string is the actual text to parse. If $input is a reference to a typeglob or an IO::Handle object corresponding to an open file or socket, the text read from the handle will be parsed.

Options are name=>value pairs and can be any of the following:


If set to a true value, namespace processing is enabled.


This option is passed on to the underlying XML::Parser object; see that module's documentation for details.


If set to a true value, text tokens consisting of only whitespace (such as those created by indentation and line breaks in between tags) will be ignored.


If set to a true value, all text other than the literal text elements of tokens will be translated into the ISO 8859-1 (Latin-1) character encoding rather than the normal UTF-8 encoding.


The value is the URI of a catalog file used to resolve PUBLIC and SYSTEM identifiers. See XML::Catalog for details.

$token = $p->get_token()

Returns the next token, as an array reference, from the input. Returns undef if there are no remaining tokens.


Pushes tokens back so they will be re-read. Useful if you've read one or more tokens too far. Correctly handles "partial" tokens returned by get_tag().

$token = $p->get_tag( [$token] )

If no argument given, skips tokens until the next start tag or end tag token. If an argument is given, skips tokens until the start tag or end tag (if the argument begins with '/') for the named element. The returned token does not include an event type code; its first element is the element name, prefixed by a '/' if the token is for an end tag.

$text = $p->get_text( [$token] )

If no argument given, returns the text at the current position, or an empty string if the next token is not a 'T' token. If an argument is given, gathers up all text between the current position and the specified start or end tag, stripping out any intervening tags (much like the way a typical Web browser deals with unknown tags).

$text = $p->get_trimmed_text( [$token] )

Like get_text(), but deletes any leading or trailing whitespaces and collapses multiple whitespace (including newlines) into single spaces.

$p->begin_saving( [$token] )

Causes subsequent calls to get_token(), get_tag(), get_text(), and get_trimmed_text() to save the returned tokens. In conjunction with restore_saved(), allows you to "back up" within a token stream. If an argument is supplied, it is placed at the beginning of the list of saved tokens (useful because you often won't know you want to begin saving until you've already read the first token you want saved).


Pushes all the tokens saved by begin_saving() back onto the token stream. Stops saving tokens. To cancel saving without backing up, call begin_saving() and restore_saved() in succession.


A token is a blessed array reference, that you acquire using $p->get_token or $p->get_tag, and that might look like:

    ["S",  $tag, $attr, $attrseq, $raw]
    ["E",  $tag, $raw]
    ["T",  $text, $raw]
    ["C",  $text, $raw]
    ["PI", $target, $data, $raw]

If you don't like remembering array indices (you're a real programmer), you may access the attributes of a token like:

$t->tag, $t->attr, $t->attrseq, $t->raw, $t->text, $t->target, $t->data.

****Please note that this may change in the future, where as there will be 4 token types, XML::TokeParser::Token::StartTag ....

What kind of token is it?

To find out, inspect your token using any of these is_* methods (1 == true, 0 == false, d'oh):

is_pi which is short for is_process_instruction

What's that token made of? To retrieve data from your token, use any of the following methods, depending on the kind of token you have:


only for process instructions


only for process instructions


for all tokens


only for start tags, returns a hashref ( print "#link ", $t->attr->{href} ).

my $attrseq = $t->attrseq

only for start tags, returns an array ref of the keys found in $t->attr in the order they originally appeared in.

my $tagname = $t->tag

only for tags ( print "opening ", $t->tag if $t->is_start_tag ).

my $text = $token->text

only for tokens of type text and comment

Here's more detailed info about the tokens.

Start tag

The token has five elements: 'S', the element's name, a reference to a hash of attribute values keyed by attribute names, a reference to an array of attribute names in the order in which they appeared in the tag, and the literal text.

End tag

The token has three elements: 'E', the element's name, and the literal text.

Character data (text)

The token has three elements: 'T', the parsed text, and the literal text. All contiguous runs of text are gathered into single tokens; there will never be two 'T' tokens in a row.


The token has three elements: 'C', the parsed text of the comment, and the literal text.

Processing instruction

The token has four elements: 'PI', the target, the data, and the literal text.

The literal text includes any markup delimiters (pointy brackets, <![CDATA[, etc.), entity references, and numeric character references and is in the XML document's original character encoding. All other text is in UTF-8 (unless the Latin option is set, in which case it's in ISO-8859-1) regardless of the original encoding, and all entity and character references are expanded.

If the Namespaces option is set, element and attribute names are prefixed by their (possibly empty) namespace URIs enclosed in curly brackets and xmlns:* attributes do not appear in 'S' tokens.


Uses a true XML parser rather than a modified HTML parser.

Text and comment tokens include extracted text as well as literal text.

PI tokens include target and data as well as literal text.

No tokens for declarations.

No "textify" hash.

unget_token correctly handles partial tokens returned by get_tag().

begin_saving() and restore_saved()



    use XML::TokeParser;
    use strict;
    my $text = '<tag foo="bar" foy="floy"> some text <!--comment--></tag>';
    my $p    = XML::TokeParser->new( \$text );
    print $/;
    while( defined( my $t = $p->get_token() ) ){
        local $\="\n";
        print '         raw = ', $t->raw;
        if( $t->tag ){
            print '         tag = ', $t->tag;
            if( $t->is_start_tag ) {
                print '        attr = ', join ',', %{$t->attr};
                print '     attrseq = ', join ',', @{$t->attrseq};
            print 'is_tag       ', $t->is_tag;
            print 'is_start_tag ', $t->is_start_tag;
            print 'is_end_tag   ', $t->is_end_tag;
        elsif( $t->is_pi ){
            print '      target = ', $t->target;
            print '        data = ', $t->data;
            print 'is_pi        ', $t->is_pi;
        else {
            print '        text = ', $t->text;
            print 'is_text      ', $t->is_text;
            print 'is_comment   ', $t->is_comment;
        print $/;


             raw = <tag foo="bar" foy="floy">
             tag = tag
            attr = foo,bar,foy,floy
         attrseq = foo,foy
    is_tag       1
    is_start_tag 1
    is_end_tag   0

             raw =  some text 
            text =  some text 
    is_text      1
    is_comment   0

             raw = <!--comment-->
            text = comment
    is_text      0
    is_comment   1

             raw = </tag>
             tag = tag
    is_tag       1
    is_start_tag 0
    is_end_tag   1


To report bugs, go to <> or send mail to <>


Copyright (c) 2003 D.H. aka PodMaster (current maintainer). Copyright (c) 2001 Eric Bohlman (original author).

All rights reserved. This program is free software; you can redistribute it and/or modify it under the same terms as Perl itself. If you don't know what this means, visit <> or <>.


HTML::TokeParser, XML::Parser, XML::Catalog, XML::Smart, XML::Twig.