src/common/strconv.cpp

   1 /////////////////////////////////////////////////////////////////////////////
   2 // Name:        strconv.cpp
   3 // Purpose:     Unicode conversion classes
   4 // Author:      Ove Kaaven, Robert Roebling, Vadim Zeitlin, Vaclav Slavik
   5 // Modified by:
   6 // Created:     29/01/98
   7 // RCS-ID:      $Id$
   8 // Copyright:   (c) 1999 Ove Kaaven, Robert Roebling, Vadim Zeitlin, Vaclav Slavik
   9 // Licence:     wxWindows license
  10 /////////////////////////////////////////////////////////////////////////////
  11
  12 // ============================================================================
  13 // declarations
  14 // ============================================================================
  15
  16 // ----------------------------------------------------------------------------
  17 // headers
  18 // ----------------------------------------------------------------------------
  19
  20 #ifdef __GNUG__
  21   #pragma implementation "strconv.h"
  22 #endif
  23
  24 // For compilers that support precompilation, includes "wx.h".
  25 #include "wx/wxprec.h"
  26
  27 #ifdef __BORLANDC__
  28   #pragma hdrstop
  29 #endif
  30
  31 #ifndef WX_PRECOMP
  32     #include "wx/intl.h"
  33     #include "wx/log.h"
  34 #endif // WX_PRECOMP
  35
  36 #ifdef __WXMSW__
  37     #include "wx/msw/private.h"
  38 #endif
  39
  40 #include <errno.h>
  41 #include <ctype.h>
  42 #include <string.h>
  43 #include <stdlib.h>
  44
  45 #include "wx/module.h"
  46 #include "wx/strconv.h"
  47
  48 // ----------------------------------------------------------------------------
  49 // globals
  50 // ----------------------------------------------------------------------------
  51
  52 #if wxUSE_WCHAR_T
  53     WXDLLEXPORT_DATA(wxMBConv) wxConvLibc;
  54     WXDLLEXPORT_DATA(wxCSConv) wxConvLocal((const wxChar *)NULL);
  55 #else
  56     // stand-ins in absence of wchar_t
  57     WXDLLEXPORT_DATA(wxMBConv) wxConvLibc, wxConvFile;
  58 #endif // wxUSE_WCHAR_T
  59
  60 WXDLLEXPORT_DATA(wxMBConv *) wxConvCurrent = &wxConvLibc;
  61
  62 class wxStrConvModule: public wxModule
  63 {
  64 public:
  65     wxStrConvModule() : wxModule() { }
  66     virtual bool OnInit() { return TRUE; }
  67     virtual void OnExit()
  68     {
  69         wxConvLocal.Clear();
  70     }
  71
  72     DECLARE_DYNAMIC_CLASS(wxStrConvModule)
  73 };
  74
  75 IMPLEMENT_DYNAMIC_CLASS(wxStrConvModule, wxModule)
  76
  77
  78 // ----------------------------------------------------------------------------
  79 // headers
  80 // ----------------------------------------------------------------------------
  81
  82 #if wxUSE_WCHAR_T
  83
  84 #ifdef __SALFORDC__
  85     #include <clib.h>
  86 #endif
  87
  88 #ifdef HAVE_ICONV
  89     #include <iconv.h>
  90 #endif
  91
  92 #include "wx/encconv.h"
  93 #include "wx/fontmap.h"
  94
  95 // ----------------------------------------------------------------------------
  96 // macros
  97 // ----------------------------------------------------------------------------
  98
  99 #define BSWAP_UCS4(str, len) { unsigned _c; for (_c=0; _c<len; _c++) str[_c]=wxUINT32_SWAP_ALWAYS(str[_c]); }
 100 #define BSWAP_UTF16(str, len) { unsigned _c; for (_c=0; _c<len; _c++) str[_c]=wxUINT16_SWAP_ALWAYS(str[_c]); }
 101
 102 // under Unix SIZEOF_WCHAR_T is defined by configure, but under other platforms
 103 // it might be not defined - assume the most common value
 104 #ifndef SIZEOF_WCHAR_T
 105     #define SIZEOF_WCHAR_T 2
 106 #endif // !defined(SIZEOF_WCHAR_T)
 107
 108 #if SIZEOF_WCHAR_T == 4
 109     #define WC_NAME         "UCS4"
 110     #define WC_BSWAP         BSWAP_UCS4
 111     #ifdef WORDS_BIGENDIAN
 112       #define WC_NAME_BEST  "UCS-4BE"
 113     #else
 114       #define WC_NAME_BEST  "UCS-4LE"
 115     #endif
 116 #elif SIZEOF_WCHAR_T == 2
 117     #define WC_NAME         "UTF16"
 118     #define WC_BSWAP         BSWAP_UTF16
 119     #define WC_UTF16
 120     #ifdef WORDS_BIGENDIAN
 121       #define WC_NAME_BEST  "UTF-16BE"
 122     #else
 123       #define WC_NAME_BEST  "UTF-16LE"
 124     #endif
 125 #else // sizeof(wchar_t) != 2 nor 4
 126     // I don't know what to do about this
 127     #error "Weird sizeof(wchar_t): please report your platform details to wx-users mailing list"
 128 #endif
 129
 130 // ============================================================================
 131 // implementation
 132 // ============================================================================
 133
 134 // ----------------------------------------------------------------------------
 135 // UTF-16 en/decoding
 136 // ----------------------------------------------------------------------------
 137
 138 #ifdef WC_UTF16
 139
 140 static size_t encode_utf16(wxUint32 input, wchar_t *output)
 141 {
 142     if (input<=0xffff)
 143     {
 144         if (output) *output++ = (wchar_t) input;
 145         return 1;
 146     }
 147     else if (input>=0x110000)
 148     {
 149         return (size_t)-1;
 150     }
 151     else
 152     {
 153         if (output)
 154         {
 155             *output++ = (wchar_t) ((input >> 10)+0xd7c0);
 156             *output++ = (wchar_t) ((input&0x3ff)+0xdc00);
 157         }
 158         return 2;
 159     }
 160 }
 161
 162 static size_t decode_utf16(const wchar_t* input, wxUint32& output)
 163 {
 164     if ((*input<0xd800) || (*input>0xdfff))
 165     {
 166         output = *input;
 167         return 1;
 168     }
 169     else if ((input[1]<0xdc00) || (input[1]>=0xdfff))
 170     {
 171         output = *input;
 172         return (size_t)-1;
 173     }
 174     else
 175     {
 176         output = ((input[0] - 0xd7c0) << 10) + (input[1] - 0xdc00);
 177         return 2;
 178     }
 179 }
 180
 181 #endif // WC_UTF16
 182
 183 // ----------------------------------------------------------------------------
 184 // wxMBConv
 185 // ----------------------------------------------------------------------------
 186
 187 size_t wxMBConv::MB2WC(wchar_t *buf, const char *psz, size_t n) const
 188 {
 189     return wxMB2WC(buf, psz, n);
 190 }
 191
 192 size_t wxMBConv::WC2MB(char *buf, const wchar_t *psz, size_t n) const
 193 {
 194     return wxWC2MB(buf, psz, n);
 195 }
 196
 197 const wxWCharBuffer wxMBConv::cMB2WC(const char *psz) const
 198 {
 199     if (psz)
 200     {
 201         size_t nLen = MB2WC((wchar_t *) NULL, psz, 0);
 202         if (nLen == (size_t)-1)
 203             return wxWCharBuffer((wchar_t *) NULL);
 204         wxWCharBuffer buf(nLen);
 205         MB2WC((wchar_t *)(const wchar_t *) buf, psz, nLen);
 206         return buf;
 207     }
 208     else
 209         return wxWCharBuffer((wchar_t *) NULL);
 210 }
 211
 212 const wxCharBuffer wxMBConv::cWC2MB(const wchar_t *psz) const
 213 {
 214     if (psz)
 215     {
 216         size_t nLen = WC2MB((char *) NULL, psz, 0);
 217         if (nLen == (size_t)-1)
 218             return wxCharBuffer((char *) NULL);
 219         wxCharBuffer buf(nLen);
 220         WC2MB((char *)(const char *) buf, psz, nLen);
 221         return buf;
 222     }
 223     else
 224         return wxCharBuffer((char *) NULL);
 225 }
 226
 227 // ----------------------------------------------------------------------------
 228 // standard file conversion
 229 // ----------------------------------------------------------------------------
 230
 231 WXDLLEXPORT_DATA(wxMBConvFile) wxConvFile;
 232
 233 // just use the libc conversion for now
 234 size_t wxMBConvFile::MB2WC(wchar_t *buf, const char *psz, size_t n) const
 235 {
 236     return wxMB2WC(buf, psz, n);
 237 }
 238
 239 size_t wxMBConvFile::WC2MB(char *buf, const wchar_t *psz, size_t n) const
 240 {
 241     return wxWC2MB(buf, psz, n);
 242 }
 243
 244 // ----------------------------------------------------------------------------
 245 // standard gdk conversion
 246 // ----------------------------------------------------------------------------
 247
 248 #ifdef __WXGTK12__
 249
 250 WXDLLEXPORT_DATA(wxMBConvGdk) wxConvGdk;
 251
 252 #include <gdk/gdk.h>
 253
 254 size_t wxMBConvGdk::MB2WC(wchar_t *buf, const char *psz, size_t n) const
 255 {
 256     if (buf)
 257     {
 258         return gdk_mbstowcs((GdkWChar *)buf, psz, n);
 259     }
 260     else
 261     {
 262         GdkWChar *nbuf = new GdkWChar[n=strlen(psz)];
 263         size_t len = gdk_mbstowcs(nbuf, psz, n);
 264         delete[] nbuf;
 265         return len;
 266     }
 267 }
 268
 269 size_t wxMBConvGdk::WC2MB(char *buf, const wchar_t *psz, size_t n) const
 270 {
 271     char *mbstr = gdk_wcstombs((GdkWChar *)psz);
 272     size_t len = mbstr ? strlen(mbstr) : 0;
 273     if (buf)
 274     {
 275         if (len > n)
 276             len = n;
 277         memcpy(buf, psz, len);
 278         if (len < n)
 279             buf[len] = 0;
 280     }
 281     return len;
 282 }
 283
 284 #endif // GTK > 1.0
 285
 286 // ----------------------------------------------------------------------------
 287 // UTF-7
 288 // ----------------------------------------------------------------------------
 289
 290 WXDLLEXPORT_DATA(wxMBConvUTF7) wxConvUTF7;
 291
 292 #if 0
 293 static char utf7_setD[]="ABCDEFGHIJKLMNOPQRSTUVWXYZ"
 294                         "abcdefghijklmnopqrstuvwxyz"
 295                         "0123456789'(),-./:?";
 296 static char utf7_setO[]="!\"#$%&*;<=>@[]^_`{|}";
 297 static char utf7_setB[]="ABCDEFGHIJKLMNOPQRSTUVWXYZ"
 298                         "abcdefghijklmnopqrstuvwxyz"
 299                         "0123456789+/";
 300 #endif
 301
 302 // TODO: write actual implementations of UTF-7 here
 303 size_t wxMBConvUTF7::MB2WC(wchar_t * WXUNUSED(buf),
 304                            const char * WXUNUSED(psz),
 305                            size_t WXUNUSED(n)) const
 306 {
 307   return 0;
 308 }
 309
 310 size_t wxMBConvUTF7::WC2MB(char * WXUNUSED(buf),
 311                            const wchar_t * WXUNUSED(psz),
 312                            size_t WXUNUSED(n)) const
 313 {
 314   return 0;
 315 }
 316
 317 // ----------------------------------------------------------------------------
 318 // UTF-8
 319 // ----------------------------------------------------------------------------
 320
 321 WXDLLEXPORT_DATA(wxMBConvUTF8) wxConvUTF8;
 322
 323 static wxUint32 utf8_max[]=
 324     { 0x7f, 0x7ff, 0xffff, 0x1fffff, 0x3ffffff, 0x7fffffff, 0xffffffff };
 325
 326 size_t wxMBConvUTF8::MB2WC(wchar_t *buf, const char *psz, size_t n) const
 327 {
 328     size_t len = 0;
 329
 330     while (*psz && ((!buf) || (len < n)))
 331     {
 332         unsigned char cc = *psz++, fc = cc;
 333         unsigned cnt;
 334         for (cnt = 0; fc & 0x80; cnt++)
 335             fc <<= 1;
 336         if (!cnt)
 337         {
 338             // plain ASCII char
 339             if (buf)
 340                 *buf++ = cc;
 341             len++;
 342         }
 343         else
 344         {
 345             cnt--;
 346             if (!cnt)
 347             {
 348                 // invalid UTF-8 sequence
 349                 return (size_t)-1;
 350             }
 351             else
 352             {
 353                 unsigned ocnt = cnt - 1;
 354                 wxUint32 res = cc & (0x3f >> cnt);
 355                 while (cnt--)
 356                 {
 357                     cc = *psz++;
 358                     if ((cc & 0xC0) != 0x80)
 359                     {
 360                         // invalid UTF-8 sequence
 361                         return (size_t)-1;
 362                     }
 363                     res = (res << 6) | (cc & 0x3f);
 364                 }
 365                 if (res <= utf8_max[ocnt])
 366                 {
 367                     // illegal UTF-8 encoding
 368                     return (size_t)-1;
 369                 }
 370 #ifdef WC_UTF16
 371                 size_t pa = encode_utf16(res, buf);
 372                 if (pa == (size_t)-1)
 373                   return (size_t)-1;
 374                 if (buf)
 375                     buf += pa;
 376                 len += pa;
 377 #else // !WC_UTF16
 378                 if (buf)
 379                     *buf++ = res;
 380                 len++;
 381 #endif // WC_UTF16/!WC_UTF16
 382             }
 383         }
 384     }
 385     if (buf && (len < n))
 386         *buf = 0;
 387     return len;
 388 }
 389
 390 size_t wxMBConvUTF8::WC2MB(char *buf, const wchar_t *psz, size_t n) const
 391 {
 392     size_t len = 0;
 393
 394     while (*psz && ((!buf) || (len < n)))
 395     {
 396         wxUint32 cc;
 397 #ifdef WC_UTF16
 398         size_t pa = decode_utf16(psz, cc);
 399         psz += (pa == (size_t)-1) ? 1 : pa;
 400 #else
 401         cc=(*psz++) & 0x7fffffff;
 402 #endif
 403         unsigned cnt;
 404         for (cnt = 0; cc > utf8_max[cnt]; cnt++) {}
 405         if (!cnt)
 406         {
 407             // plain ASCII char
 408             if (buf)
 409                 *buf++ = (char) cc;
 410             len++;
 411         }
 412
 413         else
 414         {
 415             len += cnt + 1;
 416             if (buf)
 417             {
 418                 *buf++ = (char) ((-128 >> cnt) | ((cc >> (cnt * 6)) & (0x3f >> cnt)));
 419                 while (cnt--)
 420                     *buf++ = (char) (0x80 | ((cc >> (cnt * 6)) & 0x3f));
 421             }
 422         }
 423     }
 424
 425     if (buf && (len<n)) *buf = 0;
 426     return len;
 427 }
 428
 429 // ============================================================================
 430 // wxCharacterSet and derived classes
 431 // ============================================================================
 432
 433 // ----------------------------------------------------------------------------
 434 // wxCharacterSet is the ABC for the classes below
 435 // ----------------------------------------------------------------------------
 436
 437 class wxCharacterSet
 438 {
 439 public:
 440     wxCharacterSet(const wxChar*name) : cname(name) {}
 441     virtual ~wxCharacterSet() {}
 442     virtual size_t MB2WC(wchar_t *buf, const char *psz, size_t n) = 0;
 443     virtual size_t WC2MB(char *buf, const wchar_t *psz, size_t n) = 0;
 444     virtual bool usable() const = 0;
 445 public:
 446     const wxChar*cname;
 447 };
 448
 449 // ----------------------------------------------------------------------------
 450 // ID_CharSet: implementation of wxCharacterSet using an existing wxMBConv
 451 // ----------------------------------------------------------------------------
 452
 453 class ID_CharSet : public wxCharacterSet
 454 {
 455 public:
 456     ID_CharSet(const wxChar *name, wxMBConv *cnv)
 457         : wxCharacterSet(name), work(cnv) {}
 458
 459     size_t MB2WC(wchar_t *buf, const char *psz, size_t n)
 460         { return work ? work->MB2WC(buf,psz,n) : (size_t)-1; }
 461
 462     size_t WC2MB(char *buf, const wchar_t *psz, size_t n)
 463         { return work ? work->WC2MB(buf,psz,n) : (size_t)-1; }
 464
 465     bool usable() const
 466         { return work!=NULL; }
 467 public:
 468     wxMBConv*work;
 469 };
 470
 471
 472 // ============================================================================
 473 // The classes doing conversion using the iconv_xxx() functions
 474 // ============================================================================
 475
 476 #ifdef HAVE_ICONV
 477
 478 // VS: glibc 2.1.3 is broken in that iconv() conversion to/from UCS4 fails with E2BIG
 479 //     if output buffer is _exactly_ as big as needed. Such case is (unless there's
 480 //     yet another bug in glibc) the only case when iconv() returns with (size_t)-1
 481 //     (which means error) and says there are 0 bytes left in the input buffer --
 482 //     when _real_ error occurs, bytes-left-in-input buffer is non-zero. Hence,
 483 //     this alternative test for iconv() failure.
 484 //     [This bug does not appear in glibc 2.2.]
 485 #if defined(__GLIBC__) && __GLIBC__ == 2 && __GLIBC_MINOR__ <= 1
 486 #define ICONV_FAILED(cres, bufLeft) ((cres == (size_t)-1) && \
 487                                      (errno != E2BIG || bufLeft != 0))
 488 #else
 489 #define ICONV_FAILED(cres, bufLeft)  (cres == (size_t)-1)
 490 #endif
 491
 492 #define ICONV_CHAR_CAST(x)  ((ICONV_CONST char **)(x))
 493
 494 // ----------------------------------------------------------------------------
 495 // IC_CharSet: encapsulates an iconv character set
 496 // ----------------------------------------------------------------------------
 497
 498 class IC_CharSet : public wxCharacterSet
 499 {
 500 public:
 501     IC_CharSet(const wxChar *name);
 502     virtual ~IC_CharSet();
 503
 504     virtual size_t MB2WC(wchar_t *buf, const char *psz, size_t n);
 505     virtual size_t WC2MB(char *buf, const wchar_t *psz, size_t n);
 506
 507     bool usable() const
 508         { return (m2w != (iconv_t)-1) && (w2m != (iconv_t)-1); }
 509
 510 protected:
 511     // the iconv handlers used to translate from multibyte to wide char and in
 512     // the other direction
 513     iconv_t m2w,
 514             w2m;
 515
 516 private:
 517     // the name (for iconv_open()) of a wide char charset - if none is
 518     // available on this machine, it will remain NULL
 519     static const char *ms_wcCharsetName;
 520
 521     // true if the wide char encoding we use (i.e. ms_wcCharsetName) has
 522     // different endian-ness than the native one
 523     static bool ms_wcNeedsSwap;
 524 };
 525
 526 const char *IC_CharSet::ms_wcCharsetName = NULL;
 527 bool IC_CharSet::ms_wcNeedsSwap = FALSE;
 528
 529 IC_CharSet::IC_CharSet(const wxChar *name)
 530           : wxCharacterSet(name)
 531 {
 532     // check for charset that represents wchar_t:
 533     if (ms_wcCharsetName == NULL)
 534     {
 535         ms_wcNeedsSwap = FALSE;
 536
 537         // try charset with explicit bytesex info (e.g. "UCS-4LE"):
 538         ms_wcCharsetName = WC_NAME_BEST;
 539         m2w = iconv_open(ms_wcCharsetName, wxConvLibc.cWX2MB(name));
 540
 541         if (m2w == (iconv_t)-1)
 542         {
 543             // try charset w/o bytesex info (e.g. "UCS4")
 544             // and check for bytesex ourselves:
 545             ms_wcCharsetName = WC_NAME;
 546             m2w = iconv_open(ms_wcCharsetName, wxConvLibc.cWX2MB(name));
 547
 548             // last bet, try if it knows WCHAR_T pseudo-charset
 549             if (m2w == (iconv_t)-1)
 550             {
 551                 ms_wcCharsetName = "WCHAR_T";
 552                 m2w = iconv_open(ms_wcCharsetName, wxConvLibc.cWX2MB(name));
 553             }
 554
 555             if (m2w != (iconv_t)-1)
 556             {
 557                 char    buf[2], *bufPtr;
 558                 wchar_t wbuf[2], *wbufPtr;
 559                 size_t  insz, outsz;
 560                 size_t  res;
 561
 562                 buf[0] = 'A';
 563                 buf[1] = 0;
 564                 wbuf[0] = 0;
 565                 insz = 2;
 566                 outsz = SIZEOF_WCHAR_T * 2;
 567                 wbufPtr = wbuf;
 568                 bufPtr = buf;
 569
 570                 res = iconv(m2w, ICONV_CHAR_CAST(&bufPtr), &insz,
 571                             (char**)&wbufPtr, &outsz);
 572
 573                 if (ICONV_FAILED(res, insz))
 574                 {
 575                     ms_wcCharsetName = NULL;
 576                     wxLogLastError(wxT("iconv"));
 577                     wxLogError(_("Convertion to charset '%s' doesn't work."), name);
 578                 }
 579                 else
 580                 {
 581                     ms_wcNeedsSwap = wbuf[0] != (wchar_t)buf[0];
 582                 }
 583             }
 584             else
 585             {
 586                 ms_wcCharsetName = NULL;
 587
 588                 // VS: we must not output an error here, since wxWindows will safely
 589                 //     fall back to using wxEncodingConverter.
 590                 wxLogTrace(wxT("strconv"), wxT("Impossible to convert to/from charset '%s' with iconv, falling back to wxEncodingConverter."), name);
 591                 //wxLogError(
 592             }
 593         }
 594         wxLogTrace(wxT("strconv"), wxT("wchar_t charset is '%s', needs swap: %i"), ms_wcCharsetName, ms_wcNeedsSwap);
 595     }
 596     else // we already have ms_wcCharsetName
 597     {
 598         m2w = iconv_open(ms_wcCharsetName, wxConvLibc.cWX2MB(name));
 599     }
 600
 601     // NB: don't ever pass NULL to iconv_open(), it may crash!
 602     if ( ms_wcCharsetName )
 603     {
 604         w2m = iconv_open(wxConvLibc.cWX2MB(name), ms_wcCharsetName);
 605     }
 606     else
 607     {
 608         w2m = (iconv_t)-1;
 609     }
 610 }
 611
 612 IC_CharSet::~IC_CharSet()
 613 {
 614     if ( m2w != (iconv_t)-1 )
 615         iconv_close(m2w);
 616     if ( w2m != (iconv_t)-1 )
 617         iconv_close(w2m);
 618 }
 619
 620 size_t IC_CharSet::MB2WC(wchar_t *buf, const char *psz, size_t n)
 621 {
 622     size_t inbuf = strlen(psz);
 623     size_t outbuf = n * SIZEOF_WCHAR_T;
 624     size_t res, cres;
 625     // VS: Use these instead of psz, buf because iconv() modifies its arguments:
 626     wchar_t *bufPtr = buf;
 627     const char *pszPtr = psz;
 628
 629     if (buf)
 630     {
 631         // have destination buffer, convert there
 632         cres = iconv(m2w,
 633                      ICONV_CHAR_CAST(&pszPtr), &inbuf,
 634                      (char**)&bufPtr, &outbuf);
 635         res = n - (outbuf / SIZEOF_WCHAR_T);
 636
 637         if (ms_wcNeedsSwap)
 638         {
 639             // convert to native endianness
 640             WC_BSWAP(buf /* _not_ bufPtr */, res)
 641         }
 642     }
 643     else
 644     {
 645         // no destination buffer... convert using temp buffer
 646         // to calculate destination buffer requirement
 647         wchar_t tbuf[8];
 648         res = 0;
 649         do {
 650             bufPtr = tbuf;
 651             outbuf = 8*SIZEOF_WCHAR_T;
 652
 653             cres = iconv(m2w,
 654                          ICONV_CHAR_CAST(&pszPtr), &inbuf,
 655                          (char**)&bufPtr, &outbuf );
 656
 657             res += 8-(outbuf/SIZEOF_WCHAR_T);
 658         } while ((cres==(size_t)-1) && (errno==E2BIG));
 659     }
 660
 661     if (ICONV_FAILED(cres, inbuf))
 662     {
 663         //VS: it is ok if iconv fails, hence trace only
 664         wxLogTrace(wxT("strconv"), wxT("iconv failed: %s"), wxSysErrorMsg(wxSysErrorCode()));
 665         return (size_t)-1;
 666     }
 667
 668     return res;
 669 }
 670
 671 size_t IC_CharSet::WC2MB(char *buf, const wchar_t *psz, size_t n)
 672 {
 673 #if defined(__BORLANDC__) && (__BORLANDC__ > 0x530)
 674     size_t inbuf = std::wcslen(psz) * SIZEOF_WCHAR_T;
 675 #else
 676     size_t inbuf = ::wcslen(psz) * SIZEOF_WCHAR_T;
 677 #endif
 678     size_t outbuf = n;
 679     size_t res, cres;
 680
 681     wchar_t *tmpbuf = 0;
 682
 683     if (ms_wcNeedsSwap)
 684     {
 685         // need to copy to temp buffer to switch endianness
 686         // this absolutely doesn't rock!
 687         // (no, doing WC_BSWAP twice on the original buffer won't help, as it
 688         //  could be in read-only memory, or be accessed in some other thread)
 689         tmpbuf=(wchar_t*)malloc((inbuf+1)*SIZEOF_WCHAR_T);
 690         memcpy(tmpbuf,psz,(inbuf+1)*SIZEOF_WCHAR_T);
 691         WC_BSWAP(tmpbuf, inbuf)
 692         psz=tmpbuf;
 693     }
 694
 695     if (buf)
 696     {
 697         // have destination buffer, convert there
 698         cres = iconv( w2m, ICONV_CHAR_CAST(&psz), &inbuf, &buf, &outbuf );
 699
 700         res = n-outbuf;
 701     }
 702     else
 703     {
 704         // no destination buffer... convert using temp buffer
 705         // to calculate destination buffer requirement
 706         char tbuf[16];
 707         res = 0;
 708         do {
 709             buf = tbuf; outbuf = 16;
 710
 711             cres = iconv( w2m, ICONV_CHAR_CAST(&psz), &inbuf, &buf, &outbuf );
 712
 713             res += 16 - outbuf;
 714         } while ((cres==(size_t)-1) && (errno==E2BIG));
 715     }
 716
 717     if (ms_wcNeedsSwap)
 718     {
 719         free(tmpbuf);
 720     }
 721
 722     if (ICONV_FAILED(cres, inbuf))
 723     {
 724         //VS: it is ok if iconv fails, hence trace only
 725         wxLogTrace(wxT("strconv"), wxT("iconv failed: %s"), wxSysErrorMsg(wxSysErrorCode()));
 726         return (size_t)-1;
 727     }
 728
 729     return res;
 730 }
 731
 732 #endif // HAVE_ICONV
 733
 734 // ============================================================================
 735 // Win32 conversion classes
 736 // ============================================================================
 737
 738 #if defined(__WIN32__) && !defined(__WXMICROWIN__)
 739
 740 extern long wxCharsetToCodepage(const wxChar *charset); // from utils.cpp
 741
 742 class CP_CharSet : public wxCharacterSet
 743 {
 744 public:
 745     CP_CharSet(const wxChar* name)
 746         : wxCharacterSet(name)
 747         {
 748             m_CodePage = wxCharsetToCodepage(name);
 749         }
 750
 751     size_t MB2WC(wchar_t *buf, const char *psz, size_t n)
 752     {
 753         size_t len =
 754             MultiByteToWideChar(m_CodePage, 0, psz, -1, buf, buf ? n : 0);
 755         //VS: returns # of written chars for buf!=NULL and *size*
 756         //    needed buffer for buf==NULL
 757         return len ? (buf ? len : len-1) : (size_t)-1;
 758     }
 759
 760     size_t WC2MB(char *buf, const wchar_t *psz, size_t n)
 761     {
 762         size_t len = WideCharToMultiByte(m_CodePage, 0, psz, -1, buf,
 763                                          buf ? n : 0, NULL, NULL);
 764         //VS: returns # of written chars for buf!=NULL and *size*
 765         //    needed buffer for buf==NULL
 766         return len ? (buf ? len : len-1) : (size_t)-1;
 767     }
 768
 769     bool usable() const
 770         { return m_CodePage != -1; }
 771
 772 public:
 773     long m_CodePage;
 774 };
 775 #endif // __WIN32__
 776
 777 // ============================================================================
 778 // wxEncodingConverter based conversion classes
 779 // ============================================================================
 780
 781 #if wxUSE_FONTMAP
 782
 783 class EC_CharSet : public wxCharacterSet
 784 {
 785 public:
 786     // temporarily just use wxEncodingConverter stuff,
 787     // so that it works while a better implementation is built
 788     EC_CharSet(const wxChar* name) : wxCharacterSet(name),
 789                                      enc(wxFONTENCODING_SYSTEM)
 790     {
 791         if (name)
 792             enc = wxTheFontMapper->CharsetToEncoding(name, FALSE);
 793
 794         m_ok = m2w.Init(enc, wxFONTENCODING_UNICODE) &&
 795                w2m.Init(wxFONTENCODING_UNICODE, enc);
 796     }
 797
 798     size_t MB2WC(wchar_t *buf, const char *psz, size_t WXUNUSED(n))
 799     {
 800         size_t inbuf = strlen(psz);
 801         if (buf)
 802             m2w.Convert(psz,buf);
 803         return inbuf;
 804     }
 805
 806     size_t WC2MB(char *buf, const wchar_t *psz, size_t WXUNUSED(n))
 807     {
 808 #if ( defined(__BORLANDC__) && (__BORLANDC__ > 0x530) ) \
 809     || ( defined(__MWERKS__) && defined(__WXMSW__) )
 810         size_t inbuf = std::wcslen(psz);
 811 #else
 812         size_t inbuf = ::wcslen(psz);
 813 #endif
 814         if (buf)
 815             w2m.Convert(psz,buf);
 816
 817         return inbuf;
 818     }
 819
 820     bool usable() const { return m_ok; }
 821
 822 public:
 823     wxFontEncoding enc;
 824     wxEncodingConverter m2w, w2m;
 825
 826     // were we initialized successfully?
 827     bool m_ok;
 828 };
 829
 830 #endif // wxUSE_FONTMAP
 831
 832 // ----------------------------------------------------------------------------
 833 // the function creating the wxCharacterSet for the specified charset on the
 834 // current system, trying all possibilities
 835 // ----------------------------------------------------------------------------
 836
 837 static wxCharacterSet *wxGetCharacterSet(const wxChar *name)
 838 {
 839     // check for the special case of ASCII charset
 840 #if wxUSE_FONTMAP
 841     if ( wxTheFontMapper->CharsetToEncoding(name) == wxFONTENCODING_DEFAULT )
 842 #else // wxUSE_FONTMAP
 843     if ( !name )
 844 #endif // wxUSE_FONTMAP/!wxUSE_FONTMAP
 845     {
 846         // don't convert at all
 847         return NULL;
 848     }
 849
 850     // the test above must have taken care of this case
 851     wxCHECK_MSG( name, NULL, _T("NULL name must be wxFONTENCODING_DEFAULT") );
 852
 853     wxCharacterSet *cset;
 854
 855     if ( wxStricmp(name, wxT("UTF8")) == 0 || wxStricmp(name, wxT("UTF-8")) == 0)
 856     {
 857         cset = new ID_CharSet(name, &wxConvUTF8);
 858     }
 859     else
 860     {
 861 #ifdef HAVE_ICONV
 862         cset = new IC_CharSet(name);
 863 #else // !HAVE_ICONV
 864         cset = NULL;
 865 #endif // HAVE_ICONV/!HAVE_ICONV
 866     }
 867
 868     // it can only be NULL in this case
 869 #ifndef HAVE_ICONV
 870     if ( cset )
 871 #endif // !HAVE_ICONV
 872     {
 873         if ( cset->usable() )
 874             return cset;
 875
 876         delete cset;
 877         cset = NULL;
 878     }
 879
 880 #if defined(__WIN32__) && !defined(__WXMICROWIN__)
 881     cset = new CP_CharSet(name);
 882     if ( cset->usable() )
 883         return cset;
 884
 885     delete cset;
 886     cset = NULL;
 887 #endif // __WIN32__
 888
 889 #if wxUSE_FONTMAP
 890     cset = new EC_CharSet(name);
 891     if ( cset->usable() )
 892         return cset;
 893
 894     delete cset;
 895     cset = NULL;
 896 #endif // wxUSE_FONTMAP
 897
 898     wxLogError(_("Cannot convert from encoding '%s'!"), name);
 899
 900     return NULL;
 901 }
 902
 903 // ============================================================================
 904 // wxCSConv implementation
 905 // ============================================================================
 906
 907 wxCSConv::wxCSConv(const wxChar *charset)
 908 {
 909     m_name = (wxChar *)NULL;
 910     m_cset = (wxCharacterSet *) NULL;
 911     m_deferred = TRUE;
 912
 913     SetName(charset);
 914 }
 915
 916 wxCSConv::~wxCSConv()
 917 {
 918     Clear();
 919 }
 920
 921 void wxCSConv::Clear()
 922 {
 923     if (m_name)
 924         free(m_name);
 925     if (m_cset)
 926         delete m_cset;
 927     m_name = NULL;
 928     m_cset = NULL;
 929 }
 930
 931 void wxCSConv::SetName(const wxChar *charset)
 932 {
 933     if (charset)
 934     {
 935         m_name = wxStrdup(charset);
 936         m_deferred = TRUE;
 937     }
 938 }
 939
 940 void wxCSConv::LoadNow()
 941 {
 942     if (m_deferred)
 943     {
 944         if ( !m_name )
 945         {
 946             wxString name = wxLocale::GetSystemEncodingName();
 947             if ( !name.empty() )
 948                 SetName(name);
 949         }
 950
 951         // wxGetCharacterSet() complains about NULL name
 952         m_cset = m_name ? wxGetCharacterSet(m_name) : NULL;
 953         m_deferred = FALSE;
 954     }
 955 }
 956
 957 size_t wxCSConv::MB2WC(wchar_t *buf, const char *psz, size_t n) const
 958 {
 959     ((wxCSConv *)this)->LoadNow(); // discard constness
 960
 961     if (m_cset)
 962         return m_cset->MB2WC(buf, psz, n);
 963
 964     // latin-1 (direct)
 965     size_t len = strlen(psz);
 966
 967     if (buf)
 968     {
 969         for (size_t c = 0; c <= len; c++)
 970             buf[c] = (unsigned char)(psz[c]);
 971     }
 972
 973     return len;
 974 }
 975
 976 size_t wxCSConv::WC2MB(char *buf, const wchar_t *psz, size_t n) const
 977 {
 978     ((wxCSConv *)this)->LoadNow(); // discard constness
 979
 980     if (m_cset)
 981         return m_cset->WC2MB(buf, psz, n);
 982
 983     // latin-1 (direct)
 984 #if ( defined(__BORLANDC__) && (__BORLANDC__ > 0x530) ) \
 985     || ( defined(__MWERKS__) && defined(__WXMSW__) )
 986     size_t len=std::wcslen(psz);
 987 #else
 988     size_t len=::wcslen(psz);
 989 #endif
 990     if (buf)
 991     {
 992         for (size_t c = 0; c <= len; c++)
 993             buf[c] = (psz[c] > 0xff) ? '?' : psz[c];
 994     }
 995
 996     return len;
 997 }
 998
 999 #endif // wxUSE_WCHAR_T
1000
1001