//
// the original source is outline.c by Clark Cooper.
//

#include <stdio.h>
#include <expat.h>
#include <iostream>
#include <string>
#include <map>
#include <vector>

using namespace std;

#define BUFFSIZE        8192

char Buff[BUFFSIZE];

struct Tag {
  string name;
  string str;
  map<string, string> attributes;
  Tag(const string &n) : name(n) {}
};

struct Parseinfo
{
  enum TagType { OPEN, CLOSE };

  int skip;
  int depth;
  string cdata;
  vector<Tag> st;
  TagType last_tagtype;

  Parseinfo() {
    skip = 0;
    depth = 1;
    last_tagtype = CLOSE;
  }
};

vector<string> sentence_segmentation(const string & abs)
{
  vector<string> vs;
  int begin = 0;
  for (int i = 0; ; i++) {
    if (i == abs.size() ||
        (i > 1 && abs[i-2] == '.' && abs[i-1] == ' ' &&
         (isupper(abs[i]) || abs[i] == ' ') ) ) {
      vs.push_back(abs.substr(begin, i - begin));
      begin = i;
    }
    if (i == abs.size()) break;
  }
  return vs;
}

static void
print_cdata(const string& s)
{
  for (int i = 0; i < s.size(); i++) {
    if (s[i] == '<') {
      cout << "&lt;";
      continue;
    }
    if (s[i] == '>') {
      cout << "&gt;";
      continue;
    }
    if (s[i] == '&') {
      cout << "&amp;";
      continue;
    }
    cout << s[i];
  }

}

static void
process_cdata(Parseinfo* inf, bool tag_start)
{
  string s(inf->cdata);

  vector<Tag>& st = inf->st;

  for (int i = 1; i < st.size(); i++) {
    Tag &t = st[i];
    t.str += s;
  }

  return;

}

static void
start(void *data, const char *el, const char **attr)
{
  Parseinfo *inf = (Parseinfo *) data;

  Tag t(el);
  for (int i = 0; attr[i]; i += 2) {
    t.attributes[attr[i]] = attr[i+1];
  }

  process_cdata(inf, true);

  inf->st.push_back(t);
  inf->depth++;

  inf->cdata = "";
  inf->last_tagtype = Parseinfo::OPEN;
}

static void
end(void *data, const char *el)
{
  Parseinfo *inf = (Parseinfo *) data;

  process_cdata(inf, false);

  Tag& t = inf->st.back();

  static string pmid;
  if (t.name == "PMID") {
    pmid = t.str;
  }
  if (t.name == "AbstractText") {
    vector<string> vs(sentence_segmentation(t.str));
    for (vector<string>::const_iterator i = vs.begin(); i != vs.end(); i++) {
      cout << pmid << "\tA\t" << *i << endl;
    }
  }
  if (t.name == "ArticleTitle") {
    vector<string> vs(sentence_segmentation(t.str));
    for (vector<string>::const_iterator i = vs.begin(); i != vs.end(); i++) {
      cout << pmid << "\tT\t" << *i << endl;
    }
  }
  /*
  if (t.attributes.find("sem") != t.attributes.end()) {
    cout << t.str << "\t" << t.attributes["sem"] << endl;
    if (t.attributes.find("lex") != t.attributes.end()) {
      if ((t.attributes["lex"])[0] == '*')
        cerr << "warning: \""
             << t.str << " " << t.attributes["sem"] << "\" "
             << "looks like a fragment while it has a sem-value." << endl;
    }
  }
  */

  inf->st.pop_back();
  inf->depth--;

  inf->cdata = "";
  inf->last_tagtype = Parseinfo::CLOSE;
}

static void
chardata(void *data, const char *str, int len)
{
  Parseinfo *inf = (Parseinfo *) data;
  string s(str, len);

  //  Tag& t = inf->st.top();
  inf->cdata += s;
  //  cout << t.attributes["sem"] << " " << len << " " << s << endl;
}

int
main(int argc, char *argv[])
{
  //  cout << "<?xml-stylesheet href=\"a.css\" type=\"text/css\"?>";
  
  XML_Parser p = XML_ParserCreate(NULL);
  if (! p) {
    fprintf(stderr, "Couldn't allocate memory for parser\n");
    exit(-1);
  }

  Parseinfo inf;
  XML_SetUserData(p, &inf);
    
  XML_SetElementHandler(p, start, end);
  XML_SetCharacterDataHandler(p, chardata);
  
  for (;;) {
    int done;
    int len;

    len = fread(Buff, 1, BUFFSIZE, stdin);
    if (ferror(stdin)) {
      fprintf(stderr, "Read error\n");
      exit(-1);
    }
    done = feof(stdin);

    if (XML_Parse(p, Buff, len, done) == 0) {
      fprintf(stderr, "Parse error at line %d:\n%s\n",
              XML_GetCurrentLineNumber(p),
              XML_ErrorString(XML_GetErrorCode(p)));
      exit(-1);
    }

    if (done)
      break;
  }
  return 0;
}