int
html2text(char *text, const char *content)
{
int ret;
xmlNodePtr root;
xmlErrorPtr err;
htmlParserCtxtPtr parser;
parser = htmlCreatePushParserCtxt(NULL, NULL, NULL, 0, NULL, 0);
/* htmlCtxtUseOptions(parser, HTML_PARSE_NOBLANKS | HTML_PARSE_NOERROR | HTML_PARSE_NOWARNING | HTML_PARSE_NONET); */
htmlCtxtUseOptions(parser, HTML_PARSE_RECOVER | HTML_PARSE_NOBLANKS | HTML_PARSE_NOERROR | HTML_PARSE_NOWARNING | HTML_PARSE_NONET);
ret = htmlParseChunk(parser, content, xmlStrlen(content), 0);
if (ret != 0) {
err = xmlCtxtGetLastError(parser);
fprintf(stderr, "htmlParseChunk failure: %d: %s\n", \
ret, err->message);
}
ret = htmlParseChunk(parser, NULL, 0, 1);
if (ret != 0) {
err = xmlCtxtGetLastError(parser);
fprintf(stderr, "htmlParseChunk failure 2: %d: %s\n", \
ret, err->message);
}
root = xmlDocGetRootElement(parser->myDoc);
walkTree(parser->myDoc, root, text);
return 0;
}
void se_parser::parse_output(char *output,
std::vector<search_snippet*> *snippets,
const int &count_offset) throw (sp_exception)
{
_count = count_offset;
htmlParserCtxtPtr ctxt = NULL;
parser_context pc;
pc._parser = this;
pc._snippets = snippets;
pc._current_snippet = NULL;
htmlSAXHandler saxHandler =
{
NULL,
NULL,
NULL,
NULL,
NULL,
NULL,
NULL,
NULL,
NULL,
NULL,
NULL,
NULL,
NULL,
NULL,
start_element_wrapper,
end_element_wrapper,
NULL,
characters_wrapper,
NULL,
NULL,
NULL,
NULL,
NULL,
NULL,
NULL,
cdata_wrapper,
NULL,
NULL,
NULL,
NULL,
NULL,
NULL
};
//mutex_lock(&se_parser::_se_parser_mutex);
int status = 0;
try
{
ctxt = htmlCreatePushParserCtxt(&saxHandler, &pc, "", 0, "",
XML_CHAR_ENCODING_UTF8); // encoding here.
htmlCtxtUseOptions(ctxt,HTML_PARSE_NOERROR);
status = htmlParseChunk(ctxt,output,strlen(output),0);
}
catch (std::exception e)
{
errlog::log_error(LOG_LEVEL_PARSER,"Error %s in xml/html parsing of search results.",
e.what());
throw sp_exception(WB_ERR_PARSE,e.what());
}
catch (...) // catch everything else to avoid crashes.
{
std::string msg = "Unknown error in xml/html parsing of search results";
errlog::log_error(LOG_LEVEL_PARSER,msg.c_str());
throw sp_exception(WB_ERR_PARSE,msg);
}
if (status == 0)
{
if (ctxt)
xmlFreeParserCtxt(ctxt);
//mutex_unlock(&se_parser::_se_parser_mutex);
}
else // an error occurred.
{
xmlErrorPtr xep = xmlCtxtGetLastError(ctxt);
if (xep)
{
std::string err_msg = std::string(xep->message);
miscutil::replace_in_string(err_msg,"\n","");
errlog::log_error(LOG_LEVEL_PARSER, "html level parsing error (libxml2): %s",
err_msg.c_str());
// check on error level.
if (xep->level == 3) // fatal or recoverable error.
{
std::string msg = "libxml2 fatal error";
errlog::log_error(LOG_LEVEL_PARSER,msg.c_str());
if (ctxt)
xmlFreeParserCtxt(ctxt);
//mutex_unlock(&se_parser::_se_parser_mutex);
throw sp_exception(WB_ERR_PARSE,msg);
}
// XXX: too verbose, and confusing to users.
else if (xep->level == 2)
{
std::string msg = "libxml2 recoverable error";
errlog::log_error(LOG_LEVEL_DEBUG,msg.c_str());
if (ctxt)
xmlFreeParserCtxt(ctxt);
//mutex_unlock(&se_parser::_se_parser_mutex);
//throw sp_exception(WB_ERR_PARSE,msg);
}
}
}
}