swh:1:snp:7d9f1bc35e120776766db9334626062f837c20ad
Raw File
Tip revision: 965b76582d41f5a7e67241b113bbadfbd00393e6 authored by Duncan Temple Lang on 01 May 2003, 00:00:00 UTC
version 0.93-4
Tip revision: 965b765
HTMLParse.c
/*
  This file uses the HTML parser in libxml to provide an HTML
  parser in R that is basically identical to the XML parsing interface.
  It can handle files, URLs, compressed files, and raw HTML text.
  It drops the DTD and validation options since these are not very relevant
  for HTML. (We can add put them back if anyone wants!)
 */

#include "DocParse.h"
#include "Utils.h"

#include "libxml/HTMLparser.h"

#include <sys/stat.h>
#include <unistd.h>  

USER_OBJECT_
RS_XML(HtmlParseTree)(USER_OBJECT_ fileName, USER_OBJECT_ converterFunctions, 
                       USER_OBJECT_ skipBlankLines, USER_OBJECT_ replaceEntities,
                       USER_OBJECT_ asText, USER_OBJECT_ trim, USER_OBJECT_ isURL)
{
  char *name;
  xmlDocPtr doc;
  USER_OBJECT_ rdoc;
  USER_OBJECT_ className;
  R_XMLSettings parserSettings;

  int asTextBuffer = LOGICAL_DATA(asText)[0];
  int isURLDoc = LOGICAL_DATA(isURL)[0];

  parserSettings.skipBlankLines = LOGICAL_DATA(skipBlankLines)[0];
  parserSettings.converters = converterFunctions;
  parserSettings.trim = LOGICAL_DATA(trim)[0];

  if(asTextBuffer == 0) {
    struct stat tmp_stat;  
#ifdef USE_R
    name = R_ExpandFileName(CHAR(STRING_ELT(fileName, 0)));
#else
    name = CHARACTER_DATA(fileName)[0];
#endif
    if(!isURLDoc && (name == NULL || stat(name, &tmp_stat) < 0)) {
      PROBLEM "Can't find file %s", CHAR_DEREF(STRING_ELT(fileName, 0))
      ERROR;
    }
  } else {
     name = strdup(CHAR_DEREF(STRING_ELT(fileName, 0)));
  }

    /* If one wants entities expanded directly and to appear as text.  */
  if(LOGICAL_DATA(replaceEntities)[0])
    xmlSubstituteEntitiesDefault(1);   

  if(asTextBuffer) {
   doc = htmlParseDoc(name, NULL);
   if(doc != NULL) {
      doc->name = xmlStrdup("<buffer>");
   }
  } else {
      doc = htmlParseFile(name, NULL);
  }

  if(doc == NULL) {
    if(asTextBuffer && name)
       free(name);
    PROBLEM "error in creating parser for %s", name
    ERROR;
  }

  PROTECT(rdoc = RS_XML(convertXMLDoc)(name, doc, converterFunctions, &parserSettings));

  if(asTextBuffer && name)
    free(name);


  xmlFreeDoc(doc);

     /* Set the class for the document. */
  className = NEW_CHARACTER(1);
  PROTECT(className);
    SET_STRING_ELT(className, 0, COPY_TO_USER_STRING("HTMLDocument"));   
    SET_CLASS(rdoc, className);
  UNPROTECT(1);


 UNPROTECT(1); 
 return(rdoc);
}
back to top