BEGIN
{
$HTML::Data::Parser::AUTHORITY = 'cpan:TOBYINK';
$HTML::Data::Parser::VERSION = '0.006';
$RDF::Trine::Parser::parser_names{ 'htmldata' } = __PACKAGE__;
}
sub new
{
my ($class, %options) = @_;
my $self = bless {
dom_parser => 'html5',
parse_grddl => 0,
parse_rdfa => 1,
parse_microdata => undef,
parse_microformats => undef,
parse_n3 => undef,
parse_outline => 0,
on_error => 'warn',
%options,
}, $class;
return $self;
}
sub parse
{
my ($self, $base, $string, $handler) = @_;
$self->{'_document_context'} = {};
return unless reftype($handler) eq 'CODE';
my $modules = {
grddl => ['XML::GRDDL'],
rdfa => ['RDF::RDFa::Parser'],
microdata => ['HTML::HTML5::Microdata::Parser'],
outline => ['HTML::HTML5::Outline'],
microformats => ['HTML::Microformats'],
n3 => ['HTML::Embedded::Turtle'],
};
my $dom;
if (blessed($string) and $string->isa('XML::LibXML::Document'))
{
$dom = $string;
$self->{'options_rdfa_default'} = RDF::RDFa::Parser::Config->new('xhtml');
}
elsif ($self->{dom_parser} =~ /^(xml|xhtml|x)$/i)
{
$dom = XML::LibXML->new->parse_string("$string");
$self->{'options_rdfa_default'} = RDF::RDFa::Parser::Config->new('xhtml');
}
else
{
$dom = HTML::HTML5::Parser->new->parse_string("$string");
$self->{'options_rdfa_default'} = RDF::RDFa::Parser::Config->new('html');
}
foreach my $type (qw[rdfa grddl microdata microformats n3 outline])
{
my $should_parse = $self->{"parse_${type}"};
unless (defined $should_parse and !$should_parse)
{
local $@ = undef;
my @module = @{$modules->{$type}};
eval join(' ', 'require', @module).';'
unless $module[0] eq 'RDF::RDFa::Parser';
if (! $@)
{
my $sub = "_parse_${type}";
$self->$sub($base, $dom, $handler);
}
elsif (defined $should_parse)
{
$self->_handle_error("Could not require @module: ".$@);
}
}
}
}
sub _handle_triple
{
my ($self, $handler, $g, $st) = @_;
my @nodes = $st->nodes;
if ($self->{named_graphs} and $g)
{
$st = RDF::Trine::Statement->new(
@nodes[0..2],
RDF::Trine::Node::Resource->new($g),
);
}
elsif (defined $nodes[3])
{
$st = RDF::Trine::Statement->new(@nodes[0..2]);
}
$handler->($st);
}
sub _handle_error
{
my ($self, $err) = @_;
if (reftype($self->{on_error}) eq 'CODE')
{
$self->{on_error}->($err);
}
elsif ($self->{on_error} =~ /^warn$/i)
{
warn $err;
}
elsif ($self->{on_error} =~ /^die$/i)
{
die $err;
}
}
sub _parse_rdfa
{
my ($self, $base, $dom, $handler) = @_;
my $parser = RDF::RDFa::Parser->new($dom, $base, $self->{'options_rdfa'}||$self->{'options_rdfa_default'});
$parser->set_callbacks({
ontriple => sub {
my ($p, $e, $st) = @_;
$self->_handle_triple($handler, "${base}#graph/rdfa", $st);
},
onerror => sub {
my ($p, $l, $c, $e) = @_;
$self->_handle_error("RDF::RDFa::Parser Error: ".$e);
},
});
$parser->consume;
$self->{'_document_context'}->{'RDFA'} = $parser;
}
sub _parse_microdata
{
my ($self, $base, $dom, $handler) = @_;
my $parser = HTML::HTML5::Microdata::Parser->new($dom, $base, $self->{'options_microdata'});
$parser->set_callbacks({
ontriple => sub {
my ($p, $e, $st) = @_;
$self->_handle_triple($handler, "${base}#graph/microdata", $st);
},
});
$parser->consume;
}
sub _parse_n3
{
my ($self, $base, $dom, $handler) = @_;
my $het;
if ($self->{'_document_context'}->{'RDFA'})
{
$het = bless {
rdfa_parser => $self->{'_document_context'}->{'RDFA'},
dom => $self->{'_document_context'}->{'RDFA'}->dom,
base_uri => $self->{'_document_context'}->{'RDFA'}->uri,
}, 'HTML::Embedded::Turtle';
$het->_find_endorsed;
$het->_extract_graphs;
}
else
{
$het = HTML::Embedded::Turtle->new($dom->toString, $base, {markup=>'xhtml'});
}
my $model = $het->union_graph;
$model->as_stream->each(sub {
my ($st) = @_;
$self->_handle_triple($handler, "${base}#graph/n3", $st);
});
$self->{'_document_context'}->{'RDFA'} ||= $het->{rdfa_parser};
}
sub _parse_microformats
{
my ($self, $base, $dom, $handler) = @_;
my $doc = HTML::Microformats->new_document($dom, $base)->assume_all_profiles;
my $model = $doc->model;
$model->as_stream->each(sub {
my ($st) = @_;
$self->_handle_triple($handler, "${base}#graph/microformats", $st);
});
}
sub _parse_grddl
{
my ($self, $base, $dom, $handler) = @_;
$self->{'_instance_context'}->{'GRDDL'} = XML::GRDDL->new;
my $model = $self->{'_instance_context'}->{'GRDDL'}->data($dom, $base);
$model->as_stream->each(sub {
my ($st) = @_;
$self->_handle_triple($handler, "${base}#graph/grddl", $st);
});
}
sub _parse_outline
{
my ($self, $base, $dom, $handler) = @_;
my $es = {};
if ($self->{'_document_context'}->{'RDFA'})
{
$es = $self->{'_document_context'}->{'RDFA'}->element_subjects;
}
HTML::HTML5::Outline->new(
$dom,
uri => $base,
element_subjects => $es,
%{ $self->{'options_outline'} }
)
->to_rdf
->as_stream
->each(sub {
my ($st) = @_;
$self->_handle_triple($handler, "${base}#graph/outline", $st);
});
}
1;