libxml2_parse_html.c

#include <libxml/tree.h>
#include <libxml/xpath.h>
#include <libxml/HTMLparser.h>
#include <stdio.h>

#define FILENAME "index.html"

inline static
void initialize()
{
	xmlInitParser() ;
}

inline static
void shutdown()
{
	xmlCleanupParser() ;
}


static
htmlDocPtr parseHtmlDocument(const char *filename)
{
	htmlParserCtxtPtr parser_context = htmlNewParserCtxt() ;

	htmlDocPtr document = htmlCtxtReadFile(parser_context, FILENAME, NULL, HTML_PARSE_NOWARNING | HTML_PARSE_NOERROR | HTML_PARSE_RECOVER) ;
	htmlFreeParserCtxt(parser_context) ;

	return document ;
}

static
xmlXPathObjectPtr findNodes(htmlDocPtr document, const char *xpath_query)
{
	xmlXPathContextPtr xpath_ctx = xmlXPathNewContext(document) ;
	xmlXPathObjectPtr  nodes = xmlXPathEvalExpression(xpath_query, xpath_ctx) ;
	xmlXPathFreeContext(xpath_ctx) ;
	return nodes ;

}

typedef void (*node_function_t)(xmlNodePtr node, void *data) ;

void printLinkNode(xmlNodePtr node, void *data)
{
	if (node->type == XML_ELEMENT_NODE) {
		xmlAttrPtr href = xmlHasProp(node, "href") ;
		if (href) {
			printf("-> Link to '%s'\n", xmlGetProp(node, "href")) ;
		}
	}
}

static
void eachNode(xmlXPathObjectPtr nodes, node_function_t f, void *data)
{
	xmlNodeSetPtr      nodeset = nodes->nodesetval ;
	int i, size = nodeset->nodeNr ;
	for (i = 0 ; i < size ; i++) {
		xmlNodePtr cur ;
		cur = (xmlNodePtr)nodeset->nodeTab[i] ;
		f(cur, data) ;
	}
}

int main(int argc, char *argv[])
{
	initialize() ;
	htmlDocPtr document = parseHtmlDocument(FILENAME) ;

	xmlXPathObjectPtr  nodes = findNodes(document, "/html/body//a") ;
	eachNode(nodes, printLinkNode, NULL) ;
	
	xmlXPathFreeObject(nodes) ;
	xmlFreeDoc(document) ;

	shutdown() ;
	return 0 ;
}