NAME

HTML::Gumbo - HTML5 parser based on gumbo C library

SYNOPSIS

    use HTML::Gumbo;
    say HTML::Gumbo->new->parse('<div></div>');

    say HTML::Gumbo->new->parse('<h1>Hello</h1>', format => 'tree')->as_HTML;

DESCRIPTION

Gumbo is an implementation of the HTML5 parsing algorithm implemented as a pure C99 library with no outside dependencies.

Goals and features of the C library:

  • Fully conformant with the HTML5 spec.

  • Robust and resilient to bad input.

  • Simple API that can be easily wrapped by other languages. (This is one of such wrappers.)

  • Support for source locations and pointers back to the original text. (Not exposed by this implementation at the moment.)

  • Relatively lightweight, with no outside dependencies.

  • Passes all html5lib-0.95 tests.

  • Tested on over 2.5 billion pages from Google's index.

METHODS

new

    my $parser = HTML::Gumbo->new;

No options at the moment.

parse

    my $res = $parser->parse(
        "<h1>hello world!</h1>",
        format => 'tree',
        input_is => 'string',
    );

Takes html string and pairs of named arguments:

format

Output format, default is string. See "SUPPORTED OUTPUT FORMATS".

fragment_namespace

Enables fragments parsing algorithm. Pass either 'HTML', 'SVG' or 'MATHML' to enable and set namespace. Without this input is parsed as html document, so html, head, title and body tags are added if absent.

Note that fragment_enclosing_tag is set to '<body>' and can not be changed at the moment. Feel free to send patches implementing this part.

See "SUPPORTED OUTPUT FORMATS" for additional details.

Note that SVG and MATHML parsing is not tested, feel free to file bug reports with tests in case it doesn't work.

input_is

Whether html is perl 'string', 'octets' or 'utf8' (octets known to be utf8). See "CHARACTER ENCODING OF THE INPUT".

encoding, encoding_content_type, encoding_tentative

See "CHARACTER ENCODING OF THE INPUT".

...

Some formatters may have additional arguments, see "SUPPORTED OUTPUT FORMATS"

Return value depends on the picked format.

SUPPORTED OUTPUT FORMATS

string

HTML is parsed and re-built from the tree, so tags are balanced (except void elements).

No additional arguments specific for this format.

    $html = HTML::Gumbo->new->parse( $html );

callback

HTML::Parser like interface. Pass a sub as callback argument to "parse" method and it will be called for every node in the document:

    HTML::Gumbo->new->parse( $html, format => 'callback', callback => sub {
        my ($event) = shift;
        if ( $event eq 'document start' ) {
            my ($doctype) = @_;
        }
        elsif ( $event eq 'document end' ) {
        }
        elsif ( $event eq 'start' ) {
            my ($tag, $attrs) = @_;
        }
        elsif ( $event eq 'end' ) {
            my ($tag) = @_;
        }
        elsif ( $event eq /^(text|space|cdata|comment)$/ ) {
            my ($text) = @_;
        }
        else {
            die "Unknown event";
        }
    } );

Note that 'end' events are not generated for void elements, for example hr, br and img.

No additional arguments except mentioned callback.

Fragment parsing still generates 'document start' and 'document end' events what can be handy to initialize your parsing callback.

tree

Alpha stage.

Produces tree based on HTML::Elements, like HTML::TreeBuilder.

There is major difference from HTML::TreeBuilder, this method produces top level element with tag name 'document' which may have doctype, comments and html tags as children.

Fragments parsing still produces top level 'document' element as fragment can be a list of tags, for example: '<p>hello</p><p>world</p'.

Yes, it's not ready to use as drop in replacement of tree builder. Patches are wellcome as I don't use this formatter at the moment. Note that it's hard to get rid of top level element because of situations described above. So not bad idea is to write HTML::Gumbo::Document class that is either subclass of HTML::Element or implements a small subset of methods of HTML::Element.

CHARACTER ENCODING OF THE INPUT

The C parser works only with UTF-8, so you have several options to make sure input is UTF-8. First of all define input_is argument:

string

Input is Perl string, for example obtained from "decoded_content" in HTTP::Response. Default value.

    $gumbo->parse( decode_utf8($octets) );
octets

Input are octets. Partial implementation of encoding sniffing algorithm is used. First thing wins:

encoding argument

Use it to hardcode a specific encoding.

    $gumbo->parse( $octets, input_is => 'octets', encoding => 'latin-1' );
BOM

UTF-8/UTF-16 BOMs are checked.

encoding_content_type argument

Encdoning from rransport layer, charset in content-type header.

    $gumbo->parse( $octets, input_is => 'octets', encoding_content_type => 'latin-1' );
Prescan

Not implemented, follow issue 58.

HTML5 defines prescan algorithm that extracts encoding from meta tags in the head.

It would be cool to get it in the C library, but I will accept a patch that impements it in pure perl.

encoding_tentative argument

The likely encoding for this page, e.g. based on the encoding of the page when it was last visited.

    $gumbo->parse( $octets, input_is => 'octets', encoding_tentative => 'latin-1' );
nested browsing context

Not implemented. Fragment parsing with or without context is not implemented. Parser also has no origin information, so it wouldn't be implemented.

autodetection

Not implemented.

Can be implemented using Encode::Detect::Detector. Patches are welcome.

otherwise

It dies.

utf8

Use utf8 as input_is when you're sure input is UTF-8, but octets. No pre-processing at all. Should only be used on trusted input or when it's preprocessed already.

AUTHOR

Ruslan Zakirov <ruz@bestpractical.com>

LICENSE

Under the same terms as perl itself.