static int
html_processor(const char *filename)
{
struct stat statbuf = {0};
htmlDocPtr htmldoc;
htmlctx_t hctx = {0};
hctx.state = YS_FIRST_PASS;
htmldoc = htmlSAXParseFile(filename, NULL,
html_handler, &hctx);
if (htmldoc != NULL) {
xmlFreeDoc(htmldoc);
}
if (hctx.title[0] == 0)
strncpy(hctx.title, filebasename(filename), sizeof hctx.title);
printf("<?xml version=\"1.0\"?>\n");
printf("<!DOCTYPE YASEFILE SYSTEM \"yase.dtd\">\n");
printf("<YASEFILE title=\"%s\" type=\"HTML\" author=\"%s\" keywords=\"%s\">\n", hctx.title, hctx.author, hctx.keywords);
printf("<YASEDOC title=\"%s\">\n", hctx.title);
hctx.state = YS_IN_DOC;
htmldoc = htmlSAXParseFile(filename, NULL,
html_handler, &hctx);
if (htmldoc != NULL) {
xmlFreeDoc(htmldoc);
}
printf("</YASEDOC>\n");
printf("</YASEFILE>\n");
xmlCleanupParser();
return 0;
}
static void f_parse_html(INT32 args)
{
xmlDocPtr doc = NULL;
char * encoding = "utf-8";
struct pike_string *encode_data = NULL;
if ( args == 1 ) {
if ( ARG(1).type != T_STRING )
Pike_error("Incorrect type for argument 0: expected string (encoding)\n");
encode_data = ARG(1).u.string;
encoding = encode_data->str;
}
// do nothing
if ( THIS->input_data->len == 0 )
push_int(0);
switch (THIS->parsing_method) {
case PARSE_PUSH_PARSER:
Pike_error("Push parser not implemented yet. Please bug [email protected] to implement it.");
case PARSE_MEMORY_PARSER:
htmlHandleOmittedElem(1);
doc=htmlSAXParseDoc(THIS->input_data->str, encoding, THIS->sax, NULL);
break;
case PARSE_FILE_PARSER:
htmlHandleOmittedElem(1);
doc=htmlSAXParseFile(THIS->input_data->str, "utf-8", THIS->sax, NULL);
break;
}
if ( doc != NULL )
xmlFreeDoc(doc);
push_int(0);
}
static VALUE native_parse_file(VALUE self, VALUE data, VALUE encoding)
{
xmlSAXHandlerPtr handler;
Data_Get_Struct(self, xmlSAXHandler, handler);
htmlSAXParseFile( StringValuePtr(data),
(const char *)StringValuePtr(encoding),
(htmlSAXHandlerPtr)handler,
(void *)self );
return data;
}
/*
* call-seq:
* native_parse_file(data, encoding)
*
* Parse +data+ with +encoding+
*/
static VALUE native_parse_file(VALUE self, VALUE data, VALUE encoding)
{
xmlSAXHandlerPtr handler;
htmlDocPtr hdoc ;
Data_Get_Struct(self, xmlSAXHandler, handler);
hdoc = htmlSAXParseFile( StringValuePtr(data),
(const char *)StringValuePtr(encoding),
(htmlSAXHandlerPtr)handler,
NOKOGIRI_SAX_TUPLE_NEW(NULL, self) );
xmlFreeDoc(hdoc);
return data;
}
static GSList *
html_to_ascii(GFile *file, gboolean is_recipebook)
{
Ctxt *ctxt = g_slice_new0(Ctxt);
ctxt->chars = g_string_new("");
DocText *doctext = g_slice_new0(DocText);
doctext->is_recipebook = is_recipebook;
doctext->file = g_object_ref(file);
ctxt->doctext = doctext;
char *path = g_file_get_path(file);
htmlSAXParseFile(path, NULL, &i7_html_sax, ctxt);
g_free(path);
doctext->body = g_string_free(ctxt->chars, FALSE);
GSList *retval = g_slist_prepend(ctxt->completed_doctexts, ctxt->doctext);
g_slice_free(Ctxt, ctxt);
return retval;
}
G_MODULE_EXPORT gboolean
tracker_extract_get_metadata (TrackerExtractInfo *info)
{
TrackerSparqlBuilder *metadata;
GFile *file;
TrackerConfig *config;
htmlDocPtr doc;
parser_data pd;
gchar *filename;
xmlSAXHandler handler = {
NULL, /* internalSubset */
NULL, /* isStandalone */
NULL, /* hasInternalSubset */
NULL, /* hasExternalSubset */
NULL, /* resolveEntity */
NULL, /* getEntity */
NULL, /* entityDecl */
NULL, /* notationDecl */
NULL, /* attributeDecl */
NULL, /* elementDecl */
NULL, /* unparsedEntityDecl */
NULL, /* setDocumentLocator */
NULL, /* startDocument */
NULL, /* endDocument */
parser_start_element, /* startElement */
parser_end_element, /* endElement */
NULL, /* reference */
parser_characters, /* characters */
NULL, /* ignorableWhitespace */
NULL, /* processingInstruction */
NULL, /* comment */
NULL, /* xmlParserWarning */
NULL, /* xmlParserError */
NULL, /* xmlParserError */
NULL, /* getParameterEntity */
NULL, /* cdataBlock */
NULL, /* externalSubset */
1, /* initialized */
NULL, /* private */
NULL, /* startElementNsSAX2Func */
NULL, /* endElementNsSAX2Func */
NULL /* xmlStructuredErrorFunc */
};
metadata = tracker_extract_info_get_metadata_builder (info);
file = tracker_extract_info_get_file (info);
tracker_sparql_builder_predicate (metadata, "a");
tracker_sparql_builder_object (metadata, "nfo:HtmlDocument");
pd.metadata = metadata;
pd.current = -1;
pd.in_body = FALSE;
pd.plain_text = g_string_new (NULL);
pd.title = g_string_new (NULL);
config = tracker_main_get_config ();
pd.n_bytes_remaining = tracker_config_get_max_bytes (config);
filename = g_file_get_path (file);
doc = htmlSAXParseFile (filename, NULL, &handler, &pd);
g_free (filename);
if (doc) {
xmlFreeDoc (doc);
}
g_strstrip (pd.plain_text->str);
g_strstrip (pd.title->str);
if (pd.title->str &&
*pd.title->str != '\0') {
tracker_sparql_builder_predicate (metadata, "nie:title");
tracker_sparql_builder_object_unvalidated (metadata, pd.title->str);
}
if (pd.plain_text->str &&
*pd.plain_text->str != '\0') {
tracker_sparql_builder_predicate (metadata, "nie:plainTextContent");
tracker_sparql_builder_object_unvalidated (metadata, pd.plain_text->str);
}
g_string_free (pd.plain_text, TRUE);
g_string_free (pd.title, TRUE);
return TRUE;
}
