// TODO: process CDATA, scripts

static List<String> htmlcoarsetok(String s) {
  List<String> tok = new ArrayList<String>();
  int l = s == null ? 0 : s.length();
  
  int i = 0;
  while (i < l) {
    int j = i;
    char c;
    
    // scan for non-tags
    while (j < l) {
      if (s.charAt(j) != '<')
        // regular character
        ++j;
      else if (s.substring(j, Math.min(j+4, l)).equals("<!--")) {
        // HTML comment
        j = j+4;
        do ++j; while (j < l && !s.substring(j, Math.min(j+3, l)).equals("-->"));
        j = Math.min(j+3, l);
      } else
        // it's a tag
        break;
    }
    tok.add(s.substring(i, j));
    i = j;
    if (i >= l) break;
    c = s.charAt(i);

    // scan for tags
    if (c == '<') {
      ++j;
      
      while (j < l && s.charAt(j) != '>') ++j; // TODO: strings?
      if (j < l) ++j;
    }

    tok.add(s.substring(i, j));
    i = j;
  }
  
  if ((tok.size() % 2) == 0) tok.add("");
  return tok;
}

Began life as a copy of #1000670