arabica_document xml_parser::_parse(sax_source &is) {
Arabica::SAX2DOM::Parser<string_type> parser;
Arabica::SAX::CatchErrorHandler<string_type> eh;
parser.setErrorHandler(eh);
parser.parse(is);
if (eh.errorsReported()) {
throw exception( eh.errors() );
}
return parser.getDocument();
}
NistXmlTestset::NistXmlTestset(const std::string &file)
: logger_(logkw::channel = "NistXmlTestset") {
Arabica::SAX2DOM::Parser<std::string> domParser;
Arabica::SAX::InputSource<std::string> is(file);
Arabica::SAX::CatchErrorHandler<std::string> errh;
domParser.setErrorHandler(errh);
domParser.parse(is);
if(errh.errorsReported())
BOOST_LOG_SEV(logger_, error) << errh.errors();
Arabica::DOM::Document<std::string> doc = domParser.getDocument();
if(doc == 0) {
BOOST_LOG_SEV(logger_, error) << "Error parsing input file: " << file;
exit(1);
}
doc.getDocumentElement().normalize();
Arabica::XPath::XPath<std::string> xp;
Arabica::XPath::NodeSet<std::string> docnodes =
xp.compile("/mteval/srcset/doc").evaluateAsNodeSet(doc.getDocumentElement());
docnodes.to_document_order();
BOOST_FOREACH(Arabica::DOM::Node<std::string> n, docnodes)
documents_.push_back(boost::make_shared<NistXmlDocument>(n));
outdoc_ = static_cast<Arabica::DOM::Document<std::string> >(doc.cloneNode(true));
Arabica::DOM::Element<std::string> srcset =
static_cast<Arabica::DOM::Element<std::string> >(
xp.compile("/mteval/srcset").evaluateAsNodeSet(outdoc_.getDocumentElement())[0]);
Arabica::DOM::Element<std::string> tstset = outdoc_.createElement("tstset");
int docno = 0;
while(srcset.hasChildNodes()) {
Arabica::DOM::Node<std::string> n = srcset.removeChild(srcset.getFirstChild());
tstset.appendChild(n);
if(n.getNodeType() == Arabica::DOM::Node<std::string>::ELEMENT_NODE &&
n.getNodeName() == "doc")
documents_[docno++]->setOutputNode(n);
}
tstset.setAttribute("setid", srcset.getAttribute("setid"));
tstset.setAttribute("srclang", srcset.getAttribute("srclang"));
tstset.setAttribute("trglang", "TRGLANG");
tstset.setAttribute("sysid", "SYSID");
srcset.getParentNode().replaceChild(tstset, srcset);
}
arabica_document html_parser::_parse(sax_source &is) {
typedef Arabica::SAX::Taggle<string_type> Taggle;
Arabica::SAX2DOM::Parser<string_type, Taggle> parser;
Arabica::SAX::CatchErrorHandler<string_type> eh;
parser.setErrorHandler(eh);
// Dont explicitly set default attributes.
// Without this:
// <a href="foo">bar</a>
// gets turned into
// <a href="foo" shape="rect">bar</a>
parser.setFeature(Taggle::defaultAttributesFeature, false);
// This one disabled the xmlns attr on the html element
parser.setFeature(Taggle::useSchemaNSFeature, false);
// This one lets 'inline' contain blocks: <a><div>foo</div></a>
parser.setFeature(Taggle::flexibleElementHierarchyFeautre, true);
parser.parse(is);
if (eh.errorsReported()) {
throw exception( eh.errors() );
}
return parser.getDocument();
}