xmlparser/File.cpp

// Copyright 2017 Patrick Brosi
// info@patrickbrosi.de

#include <fcntl.h>
#include <sys/stat.h>
#include <sys/types.h>
#include <unistd.h>
#include <cassert>
#include <cstring>
#include <fstream>
#include <iostream>
#include <map>
#include "xml/File.h"
#include "xml/NamedEnts.h"

using namespace xml;

// _____________________________________________________________________________
File::File(const std::string& path)
    : _file(0), _c(0), _lastBytes(0), _which(0), _path(path), _totReadBef(0) {
  _buffer = new char*[2];
  _buffer[0] = new char[BUFFER_S + 1];
  _buffer[1] = new char[BUFFER_S + 1];

  reset();
}

// _____________________________________________________________________________
File::~File() {
  delete[] _buffer[0];
  delete[] _buffer[1];
  delete[] _buffer;
  close(_file);
}

// _____________________________________________________________________________
void File::reset() {
  _which = 0;
  _s.s = NONE;
  _s.hanging = 0;
  _totReadBef = 0;

  if (_file) close(_file);
  _file = open(_path.c_str(), O_RDONLY);
  if (_file < 0)
    throw XmlFileException(std::string("could not open file"), _path, 0, 0, 0);
#ifdef __unix__
  posix_fadvise(_file, 0, 0, POSIX_FADV_SEQUENTIAL);
#endif

  _lastBytes = read(_file, _buffer[_which], BUFFER_S);
  _lastNewData = _lastBytes;
  _c = _buffer[_which];
  while (!_s.tagStack.empty()) _s.tagStack.pop();
  _s.tagStack.push("[root]");
  _prevs = _s;
}

// _____________________________________________________________________________
size_t File::level() const { return _s.tagStack.size() - _s.hanging; }

// _____________________________________________________________________________
ParserState File::state() { return _prevs; }

// _____________________________________________________________________________
void File::setState(const ParserState& s) {
  _s = s;
  _prevs = s;

  lseek(_file, _s.off, SEEK_SET);
  _totReadBef = _s.off;
  _lastBytes = read(_file, _buffer[_which], BUFFER_S);
  _lastNewData = _lastBytes;
  _c = _buffer[_which];

  next();
}

// _____________________________________________________________________________
const Tag& File::get() const { return _ret; }

// _____________________________________________________________________________
bool File::next() {
  if (!_s.tagStack.size()) return false;
  // avoid too much stack copying
  if (_prevs.tagStack.size() != _s.tagStack.size() ||
      _prevs.tagStack.top() != _s.tagStack.top()) {
    _prevs.tagStack = _s.tagStack;
  }
  _prevs.s = _s.s;
  _prevs.hanging = _s.hanging;
  _prevs.off =
      _totReadBef + (_c - _buffer[_which]) - (_lastBytes - _lastNewData);

  if (_s.hanging) _s.hanging--;
  _ret.name = 0;
  _ret.text = emptyStr;
  _ret.attrs.clear();
  void* i;
  while (_lastBytes) {
    for (; _c - _buffer[_which] < _lastBytes; ++_c) {
      char c = *_c;
      switch (_s.s) {
        case NONE:
          if (std::isspace(c))
            continue;
          else if (c == '<') {
            _s.s = IN_TAG_TENTATIVE;
            continue;
          } else {
            _s.s = IN_TEXT;
            _ret.name = emptyStr;
            _tmp = _c;
            continue;
          }
        case IN_TEXT:
          i = memchr(_c, '<', _lastBytes - (_c - _buffer[_which]));
          if (!i) {
            _c = _buffer[_which] + _lastBytes;
            continue;
          } else {
            _c = (char*)i;
            *_c = 0;
            _ret.text = _tmp;
            _s.s = IN_TAG_TENTATIVE;
            _c++;
            return true;
          }
        case IN_COMMENT_TENTATIVE:
          if (c == '-') {
            _s.s = IN_COMMENT_TENTATIVE2;
            continue;
          }
          throw XmlFileException("Expected comment", _path, _c, _buffer[_which],
                                 _prevs.off);

        case IN_COMMENT_TENTATIVE2:
          if (c == '-') {
            _s.s = IN_COMMENT;
            continue;
          }
          throw XmlFileException("Expected comment", _path, _c, _buffer[_which],
                                 _prevs.off);

        case IN_COMMENT_CL_TENTATIVE:
          if (c == '-') {
            _s.s = IN_COMMENT_CL_TENTATIVE2;
            continue;
          }
          _s.s = IN_COMMENT;
          continue;

        case IN_COMMENT_CL_TENTATIVE2:
          if (c == '>') {
            _s.s = NONE;
            continue;
          }
          _s.s = IN_COMMENT;
        // fall through, we are still in comment

        case IN_COMMENT:
          i = memchr(_c, '-', _lastBytes - (_c - _buffer[_which]));
          if (!i) {
            _c = _buffer[_which] + _lastBytes;
            continue;
          } else {
            _c = (char*)i;
            _s.s = IN_COMMENT_CL_TENTATIVE;
            continue;
          }
        case IN_TAG_TENTATIVE:
          if (c == '/') {
            _s.s = IN_TAG_NAME_CLOSE;
            _tmp = _c + 1;
            continue;
          } else if (c == '?') {
            _s.s = IN_TAG_NAME_META;
            continue;
          } else if (c == '!') {
            _s.s = IN_COMMENT_TENTATIVE;
            continue;
          } else if (std::isalnum(c) || c == '-' || c == '_' || c == '.') {
            _s.s = IN_TAG_NAME;
            _ret.name = _c;
            continue;
          }

        case IN_TAG:
          if (std::isspace(c))
            continue;
          else if (std::isalnum(c) || c == '-' || c == '_' || c == '.') {
            _s.s = IN_ATTRKEY;
            _tmp = _c;
            continue;
          } else if (c == '/') {
            _s.s = AW_CLOSING;
            continue;
          } else if (c == '>') {
            _s.hanging++;
            _s.tagStack.push(_ret.name);
            _s.s = WS_SKIP;
            continue;
          } else {
            throw XmlFileException("Expected valid tag", _path, _c,
                                   _buffer[_which], _prevs.off);
          }

        case IN_ATTRVAL_SQ:
          i = memchr(_c, '\'', _lastBytes - (_c - _buffer[_which]));
          if (!i) {
            _c = _buffer[_which] + _lastBytes;
            continue;
          } else {
            _c = (char*)i;
            _s.s = IN_TAG;
            *_c = 0;
            _ret.attrs[_tmp] = _tmp2;
            continue;
          }

        case IN_ATTRVAL_DQ:
          i = memchr(_c, '"', _lastBytes - (_c - _buffer[_which]));
          if (!i) {
            _c = _buffer[_which] + _lastBytes;
            continue;
          } else {
            _c = (char*)i;
            _s.s = IN_TAG;
            *_c = 0;
            _ret.attrs[_tmp] = _tmp2;
            continue;
          }

        case AW_IN_ATTRVAL:
          if (std::isspace(c))
            continue;
          else if (c == '\'') {
            _s.s = IN_ATTRVAL_SQ;
            _tmp2 = _c + 1;
            continue;
          } else if (c == '"') {
            _s.s = IN_ATTRVAL_DQ;
            _tmp2 = _c + 1;
            continue;
          } else {
            throw XmlFileException("Expected attribute value", _path, _c,
                                   _buffer[_which], _prevs.off);
          }

        case IN_ATTRKEY:
          if (std::isspace(c)) {
            *_c = 0;
            _s.s = AFTER_ATTRKEY;
            continue;
          } else if (std::isalnum(c) || c == '-' || c == '_' || c == '.') {
            continue;
          } else if (c == '=') {
            *_c = 0;
            _s.s = AW_IN_ATTRVAL;
            continue;
          }

          throw XmlFileException("Expected attribute key char or =", _path, _c,
                                 _buffer[_which], _prevs.off);

        case AFTER_ATTRKEY:
          if (std::isspace(c))
            continue;
          else if (c == '=') {
            _s.s = AW_IN_ATTRVAL;
            continue;
          } else {
            throw XmlFileException(
                std::string("Expected attribute value for '") + _tmp + "'.",
                _path, _c, _buffer[_which], _prevs.off);
          }

        case IN_TAG_NAME:
          if (std::isspace(c)) {
            *_c = 0;
            _s.s = IN_TAG;
            continue;
          } else if (c == '>') {
            *_c = 0;
            _s.hanging++;
            _s.tagStack.push(_ret.name);
            _s.s = WS_SKIP;
            continue;
          } else if (c == '/') {
            *_c = 0;
            _s.s = AW_CLOSING;
            continue;
          } else if (std::isalnum(c) || c == '-' || c == '_' || c == '.') {
            continue;
          }

        case IN_TAG_NAME_META:
          // TODO: read meta tags!
          if (c == '>') {
            _s.s = NONE;
            continue;
          }

          continue;

        case IN_TAG_NAME_CLOSE:
          if (std::isspace(c)) {
            *_c = 0;
            _s.s = IN_TAG_CLOSE;
            continue;
          } else if (std::isalnum(c) || c == '-' || c == '_' || c == '.') {
            continue;
          } else if (c == '>') {
            *_c = 0;
            if (_tmp != _s.tagStack.top()) {
              throw XmlFileException(std::string("Closing wrong tag '<") + _tmp +
                                         ">', expected close of '<" +
                                         _s.tagStack.top() + ">'.",
                                     _path, _c, _buffer[_which], _prevs.off);
            }
            _s.tagStack.pop();
            _s.s = NONE;
            continue;
          }

        case IN_TAG_CLOSE:
          if (std::isspace(c))
            continue;
          else if (c == '>') {
            if (_tmp != _s.tagStack.top()) {
              throw XmlFileException(std::string("Closing wrong tag '<") + _tmp +
                                         ">', expected close of '<" +
                                         _s.tagStack.top() + ">'.",
                                     _path, _c, _buffer[_which], _prevs.off);
            }
            _s.tagStack.pop();
            _s.s = NONE;
            continue;
          } else {
            throw XmlFileException("Expected '>'", _path, _c, _buffer[_which],
                                   _prevs.off);
          }

        case AW_CLOSING:
          if (c == '>') {
            _s.s = WS_SKIP;
            continue;
          }

        case WS_SKIP:
          if (std::isspace(c))
            continue;
          else {
            _s.s = NONE;
            return true;
          }
      }
    }

    // buffer ended, read new stuff, but copy remaining if needed
    size_t off = 0;
    if (_s.s == IN_TAG_NAME) {  //|| IN_TAG_NAME_META) {
      off = _lastBytes - (_ret.name - _buffer[_which]);
      memmove(_buffer[!_which], _ret.name, off);
      _ret.name = _buffer[!_which];
    } else if (_s.s == IN_TAG_NAME_CLOSE || _s.s == IN_ATTRKEY ||
               _s.s == IN_TEXT) {
      off = _lastBytes - (_tmp - _buffer[_which]);
      memmove(_buffer[!_which], _tmp, off);
      _tmp = _buffer[!_which];
    } else if (_s.s == IN_ATTRVAL_SQ || _s.s == IN_ATTRVAL_DQ) {
      off = _lastBytes - (_tmp2 - _buffer[_which]);
      memmove(_buffer[!_which], _tmp2, off);
      _tmp2 = _buffer[!_which];
    }

    assert(off <= BUFFER_S);

    size_t readb = read(_file, _buffer[!_which] + off, BUFFER_S - off);
    if (!readb) break;
    _totReadBef += _lastNewData;
    _which = !_which;
    _lastNewData = readb;
    _lastBytes = _lastNewData + off;
    _c = _buffer[_which] + off;
  }

  if (_s.tagStack.size()) {
    if (_s.tagStack.top() != "[root]") {
      throw XmlFileException("XML tree not complete", _path, _c,
                             _buffer[_which], _prevs.off);
    } else {
      _s.tagStack.pop();
    }
  }
  _s.s = NONE;
  _ret.name = "[root]";
  return false;
}

// _____________________________________________________________________________
std::string File::decode(const std::string& str) { return decode(str.c_str()); }

// _____________________________________________________________________________
std::string File::decode(const char* str) {
  const char* c = strchr(str, '&');
  if (!c) return str;

  char* decRet = new char[strlen(str) + 1];
  const char* last = str;
  char* dstPt = decRet;

  for (; c != 0; c = strchr(c + 1, '&')) {
    memcpy(dstPt, last, c - last);
    dstPt += c - last;
    last = c;

    if (*(c + 1) == '#') {
      uint64_t cp = -1;
      char* tail;
      errno = 0;
      if (*(c + 2) == 'x' || *(c + 2) == 'X')
        cp = strtoul(c + 3, &tail, 16);
      else
        cp = strtoul(c + 2, &tail, 10);

      if (*tail == ';' && cp <= 0x1FFFFF && !errno) {
        dstPt += utf8(cp, dstPt);
        last = tail + 1;
      }
    } else {
      const char* e = strchr(c, ';');
      if (e) {
        char* ent = new char[e - 1 - c + 1];
        memcpy(ent, c + 1, e - 1 - c);
        ent[e - 1 - c] = 0;
        const auto it = xml::ENTITIES.find(ent);
        if (it != xml::ENTITIES.end()) {
          const char* utf8 = it->second;
          memcpy(dstPt, utf8, strlen(utf8));
          dstPt += strlen(utf8);
          last += strlen(ent) + 2;
        }
        delete[] ent;
      }
    }
  }

  strcpy(dstPt, last);
  std::string ret(decRet);
  delete[] decRet;
  return ret;
}

// _____________________________________________________________________________
size_t File::utf8(size_t cp, char* out) {
  if (cp <= 0x7F) {
    out[0] = cp & 0x7F;
    return 1;
  } else if (cp <= 0x7FF) {
    out[0] = 0xC0 | (cp >> 6);
    out[1] = 0x80 | (cp & 0x3F);
    return 2;
  } else if (cp <= 0xFFFF) {
    out[0] = 0xE0 | (cp >> 12);
    out[1] = 0x80 | ((cp >> 6) & 0x3F);
    out[2] = 0x80 | (cp & 0x3F);
    return 3;
  } else if (cp <= 0x1FFFFF) {
    out[0] = 0xF0 | (cp >> 18);
    out[1] = 0x80 | ((cp >> 12) & 0x3F);
    out[2] = 0x80 | ((cp >> 6) & 0x3F);
    out[3] = 0x80 | (cp & 0x3F);
    return 4;
  }

  return 0;
}