source: trunk/gsdl/lib/gsdlunicode.cpp@ 100

/**********************************************************************
 *
 * gsdlunicode.cpp -- 
 * Copyright (C) 1999  The New Zealand Digital Library Project
 *
 * PUT COPYRIGHT NOTICE HERE
 *
 * $Id: gsdlunicode.cpp 100 1999-01-08 02:33:16Z rjmcnab $
 *
 *********************************************************************/

/*
   $Log$
   Revision 1.4  1999/01/08 02:33:15  rjmcnab

   Added standard header to source files.

 */

static char *RCSID = "$Id: gsdlunicode.cpp 100 1999-01-08 02:33:16Z rjmcnab $";


#include "gsdlunicode.h"


// unitool is currently in mg, if mg is not being used it should
// be moved into GSDLHOME/lib 
#include "unitool.h"

#include "fileutil.h"

#include <stdio.h>








// converts a unicode encode text_t string to a utf-8
// encoded text_t string
text_t to_utf8 (const text_t &in) {
  text_t::const_iterator here = in.begin();
  text_t::const_iterator end = in.end();
  text_t out;

  unsigned char thischar[MAXUTF8CHARLEN];
  int i, charlen;

  while (here != end) {
    charlen = output_utf8_char (*here, thischar, &thischar[MAXUTF8CHARLEN-1]);
    for (i=0; i<charlen; i++) out.push_back(thischar[i]);
    here++;
  }

  return out;
}

// converts a utf-8 encoded text_t string to a unicode
// encoded text_t string
text_t to_uni (const text_t &in) {
  text_t out;
  unsigned char *in_cstr = (unsigned char *)in.getcstr();
  unsigned char *here = in_cstr;
  unsigned char *end = in_cstr;

  unsigned short unichar;
  int charlen = 0;

  // get the last valid character in the string
  while (*end != '\0') end++;
  end--;

  while ((charlen = parse_utf8_char (here, end, &unichar)) > 0) {
    out.push_back(unichar);
    here += charlen;
  }

  delete in_cstr;

  return out;
}



utf8inconvertclass::utf8inconvertclass () {
  utf8buflen = 0;
}

void utf8inconvertclass::reset () {
  start = NULL;
  len = 0;
  utf8buflen=0;
}

void utf8inconvertclass::convert (text_t &output, status_t &status) {
  output.clear();

  if (start == NULL || len == 0) {
    if (utf8buflen == 0) status = finished;
    else status = stopped;
    return;
  }

  // don't want any funny sign conversions happening
  unsigned char *here = (unsigned char *)start;

  size_t charlen = getutf8charlen ();
  unsigned short c;
  size_t realcharlen;
  while (len > 0) {
    if (charlen == 0) {
      // start parsing a new character
      utf8buflen = 0;
      utf8buf[utf8buflen++] = *here;
      ++here;
      --len;
      charlen = getutf8charlen ();

    } else if (utf8buflen < charlen) {
      // assumes charlen is always less than MAXUTF8CHARLEN
      utf8buf[utf8buflen++] = *here;
      ++here;
      --len;
    }

    if (utf8buflen == charlen) {
      // got a complete character
      realcharlen = parse_utf8_char (utf8buf, &utf8buf[utf8buflen-1], &c);
      output.push_back (c);
      
      // move any unparsed characters. If an error occurred some of
      // the characters might be unused.
      int i;
      int diff = utf8buflen - realcharlen;
      for (i=0; i < diff; i++) utf8buf[i] = utf8buf[i+diff];
      utf8buflen = diff;
      charlen = getutf8charlen ();
    }
  }

  start = (char *)here; // save current position

  if (utf8buflen == 0) status = finished;
  else status = stopped;
}


// returns the length that the current contents of the 
// utf8buf should be
size_t utf8inconvertclass::getutf8charlen () {
  if (utf8buflen == 0) return 0;

  // one byte character
  if (utf8buf[0] < 0x80) return 1;

  // error, is not the start of a utf-8 character
  if (utf8buf[0] < 0xc0) return 1;

  // two bute character
  if (utf8buf[0] < 0xe0) return 2;

  // three byte character
  if (utf8buf[0] < 0xf0) return 3;

  // error, character too long for unicode
  return 1;
}


void utf8outconvertclass::reset () {
  input = NULL;
  outs = NULL;
  utf8buflen = 0;
  utf8bufhere = 0;
}

// note that convert does not null-terminate the
// output array of characters
void utf8outconvertclass::convert (char *output, size_t maxlen, 
                   size_t &len, status_t &status) {
  if (input == NULL || output == NULL) {
    if (utf8buflen == 0) status = finished;
    else status = unfinished;
    return;
  }

  // don't want any funny sign conversions happening
  unsigned char *uoutput = (unsigned char *)output;
  text_t::iterator textend = input->end();
  len = 0;
  while (len < maxlen) {
    // empty the contents of the internal buffer
    if (utf8buflen > 0) {
      while (len < maxlen && utf8bufhere < utf8buflen) {
    *uoutput = utf8buf[utf8bufhere];
    uoutput++;
    len++;
    utf8bufhere++;
      }

      if (utf8bufhere == utf8buflen) {
    utf8bufhere = 0;
    utf8buflen = 0;
      }
    }

    // fill up the buffer with the next character
    if (utf8buflen == 0) {
      if (texthere == textend) break; // finished!
      if (!rzws || (*texthere != 0x200b))
    utf8buflen = output_utf8_char (*texthere, utf8buf, 
                       &utf8buf[MAXUTF8CHARLEN-1]);
      texthere++;
      utf8bufhere = 0;
    }
  }
  
  if (texthere == textend && utf8buflen == 0) status = finished;
  else status = unfinished;
}






mapdata_t::mapdata_t () {
  int i;

  // reset all the map ptrs to be NULL
  for (i=0; i<256; i++) {
    ptrs[i] = (unsigned short *)NULL;
  }

  // say nothing has been loaded
  loaded = false;
}


mapconvert::mapconvert () {
  absentc = 0;
}


// loadmapfile should be called before any conversion is done
bool mapconvert::loadmapfile (const text_t &thegsdlhome, 
                  const text_t &theencoding, 
                  unsigned short theabsentc) {
  FILE *mapfilein = (FILE *)NULL;

  // check to see if the mapfile has been already loaded
  if (mapdata.loaded && gsdlhome == thegsdlhome &&
      encoding == theencoding && absentc == theabsentc) 
    return true;

  unloadmapfile ();
  gsdlhome = thegsdlhome;
  encoding = theencoding;
  absentc = theabsentc;

  // open the map file
  text_t filename = filename_cat (gsdlhome, "unicode");
  filename = filename_cat (filename, encoding);
  filename += ".ump";
  char *cfilename = filename.getcstr();
  if (cfilename == (char *)NULL) return false;
  mapfilein = fopen(cfilename, "rb");
  delete cfilename;

  if (mapfilein == (FILE *)NULL) return false;

  unsigned char c, n1, n2;
  unsigned short *arrptr;
  int i;
  c = fgetc (mapfilein);
  while (!feof (mapfilein)) {
    if (mapdata.ptrs[c] == (unsigned short *)NULL) {
      // allocate a new array
      arrptr = new unsigned short[256];
      mapdata.ptrs[c] = arrptr;
    } else arrptr = mapdata.ptrs[c];

    // clear the array
    for (i=0; i<256; i++) arrptr[i] = 0;
    
    // read in this block
    n1 = fgetc (mapfilein);
    n2 = fgetc (mapfilein);
    i=0;
    while (!feof (mapfilein)) {
      arrptr[i] = ((unsigned short)n1 << 8) | (unsigned short)n2;

      i++;
      if (i >= 256) break;
      n1 = fgetc (mapfilein);
      n2 = fgetc (mapfilein);
    }

    c = fgetc (mapfilein);
  }

  mapdata.loaded = true;

  return true;
}

void mapconvert::unloadmapfile () {
  if (!mapdata.loaded) return;

  int i;
  for (i=0; i<256; i++) {
    if (mapdata.ptrs[i] != (unsigned short *)NULL) {
      delete [] mapdata.ptrs[i];
      mapdata.ptrs[i] = (unsigned short *)NULL;
    }
  }

  mapdata.loaded = false;
}


unsigned short mapconvert::convert (unsigned short c) {
  if (!mapdata.loaded) return absentc;

  if (c == 0) return 0; // 0 always maps to 0...

  unsigned short n1 = c >> 8;
  unsigned short n2 = c & 0xff;

  unsigned short *arrptr = mapdata.ptrs[n1];
  if (arrptr == (unsigned short *)NULL) return absentc;

  if (arrptr[n2] == 0) return absentc;
  return arrptr[n2];
}

text_t mapconvert::convert (const text_t &instr) {
  if (!mapdata.loaded) return absentc;

  text_t outstr;
  text_t::const_iterator here = instr.begin();
  text_t::const_iterator end = instr.end();

  while (here != end) {
    outstr.push_back(this->convert(*here));
    here++;
  }
  
  return outstr;
}




mapinconvertclass::mapinconvertclass () {
  mapbuflen = 0;
}

void mapinconvertclass::reset () {
  start = NULL;
  len = 0;
  mapbuflen=0;
}

void mapinconvertclass::convert (text_t &output, status_t &status) {
  output.clear();

  if (start == NULL || len == 0) {
    if (mapbuflen == 0) status = finished;
    else status = stopped;
    return;
  }

  // don't want any funny sign conversions happening
  unsigned char *here = (unsigned char *)start;

  size_t charlen = getmapcharlen ();
  unsigned short c;
  size_t realcharlen;
  while (len > 0) {
    if (charlen == 0) {
      // start parsing a new character
      mapbuflen = 0;
      mapbuf[mapbuflen++] = *here;
      ++here;
      --len;
      charlen = getmapcharlen ();

    } else if (mapbuflen < charlen) {
      // assumes charlen is always less than MAXMAPCHARLEN
      mapbuf[mapbuflen++] = *here;
      ++here;
      --len;
    }

    if (mapbuflen == charlen) {
      // got a complete character
      if (charlen == 1) {
    // ascii character
    output.push_back (mapbuf[0]);

      } else {
    // two byte character
    output.push_back (converter.convert(((unsigned short)mapbuf[0] << 8) |
                        (unsigned short)mapbuf[1]));
      }

      mapbuflen = 0;
      charlen = 0;
    }
  }

  start = (char *)here; // save current position

  if (mapbuflen == 0) status = finished;
  else status = stopped;
}



mapoutconvertclass::mapoutconvertclass () {
  mapbuflen=0; 
  mapbufhere=0;
}

void mapoutconvertclass::reset () {
  input = NULL;
  outs = NULL;
  mapbuflen = 0;
  mapbufhere = 0;
}

// note that convert does not null-terminate the
// output array of characters
void mapoutconvertclass::convert (char *output, size_t maxlen, 
                 size_t &len, status_t &status) {
  unsigned short outc;

  if (input == NULL || output == NULL) {
    if (mapbuflen == 0) status = finished;
    else status = unfinished;
    return;
  }

  // don't want any funny sign conversions happening
  unsigned char *uoutput = (unsigned char *)output;
  text_t::iterator textend = input->end();
  len = 0;
  while (len < maxlen) {
    // empty the contents of the internal buffer
    if (mapbuflen > 0) {
      while (len < maxlen && mapbufhere < mapbuflen) {
    *uoutput = mapbuf[mapbufhere];
    uoutput++;
    len++;
    mapbufhere++;
      }

      if (mapbufhere == mapbuflen) {
    mapbufhere = 0;
    mapbuflen = 0;
      }
    }

    // fill up the buffer with the next character
    if (mapbuflen == 0) {
      if (texthere == textend) break; // finished!
      if (!rzws || (*texthere != 0x200b)) {
    if (*texthere < 0x80) {
      mapbuf[0] = (unsigned char)*texthere;
      mapbuflen = 1;
    } else {
      outc = converter.convert (*texthere);
      mapbuf[0] = (unsigned char)(outc >> 8);
      mapbuf[1] = (unsigned char)(outc & 0xff);
      mapbuflen = 2;
    }
      }

      texthere++;
      mapbufhere = 0;
    }
  }
  
  if (texthere == textend && mapbuflen == 0) status = finished;
  else status = unfinished;
}