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 #if wxUSE_WCHAR_T
  70          wxConvLocal.Clear();
  71 #endif
  72     }
  73
  74     DECLARE_DYNAMIC_CLASS(wxStrConvModule)
  75 };
  76
  77 IMPLEMENT_DYNAMIC_CLASS(wxStrConvModule, wxModule)
  78
  79
  80 // ----------------------------------------------------------------------------
  81 // headers
  82 // ----------------------------------------------------------------------------
  83
  84 #if wxUSE_WCHAR_T
  85
  86 #ifdef __SALFORDC__
  87     #include <clib.h>
  88 #endif
  89
  90 #ifdef HAVE_ICONV
  91     #include <iconv.h>
  92 #endif
  93
  94 #include "wx/encconv.h"
  95 #include "wx/fontmap.h"
  96
  97 // ----------------------------------------------------------------------------
  98 // macros
  99 // ----------------------------------------------------------------------------
 100
 101 #define BSWAP_UCS4(str, len) { unsigned _c; for (_c=0; _c<len; _c++) str[_c]=wxUINT32_SWAP_ALWAYS(str[_c]); }
 102 #define BSWAP_UTF16(str, len) { unsigned _c; for (_c=0; _c<len; _c++) str[_c]=wxUINT16_SWAP_ALWAYS(str[_c]); }
 103
 104 // under Unix SIZEOF_WCHAR_T is defined by configure, but under other platforms
 105 // it might be not defined - assume the most common value
 106 #ifndef SIZEOF_WCHAR_T
 107     #define SIZEOF_WCHAR_T 2
 108 #endif // !defined(SIZEOF_WCHAR_T)
 109
 110 #if SIZEOF_WCHAR_T == 4
 111     #define WC_NAME         "UCS4"
 112     #define WC_BSWAP         BSWAP_UCS4
 113     #ifdef WORDS_BIGENDIAN
 114       #define WC_NAME_BEST  "UCS-4BE"
 115     #else
 116       #define WC_NAME_BEST  "UCS-4LE"
 117     #endif
 118 #elif SIZEOF_WCHAR_T == 2
 119     #define WC_NAME         "UTF16"
 120     #define WC_BSWAP         BSWAP_UTF16
 121     #define WC_UTF16
 122     #ifdef WORDS_BIGENDIAN
 123       #define WC_NAME_BEST  "UTF-16BE"
 124     #else
 125       #define WC_NAME_BEST  "UTF-16LE"
 126     #endif
 127 #else // sizeof(wchar_t) != 2 nor 4
 128     // I don't know what to do about this
 129     #error "Weird sizeof(wchar_t): please report your platform details to wx-users mailing list"
 130 #endif
 131
 132 // ============================================================================
 133 // implementation
 134 // ============================================================================
 135
 136 // ----------------------------------------------------------------------------
 137 // UTF-16 en/decoding
 138 // ----------------------------------------------------------------------------
 139
 140 #ifdef WC_UTF16
 141
 142 static size_t encode_utf16(wxUint32 input, wchar_t *output)
 143 {
 144     if (input<=0xffff)
 145     {
 146         if (output) *output++ = (wchar_t) input;
 147         return 1;
 148     }
 149     else if (input>=0x110000)
 150     {
 151         return (size_t)-1;
 152     }
 153     else
 154     {
 155         if (output)
 156         {
 157             *output++ = (wchar_t) ((input >> 10)+0xd7c0);
 158             *output++ = (wchar_t) ((input&0x3ff)+0xdc00);
 159         }
 160         return 2;
 161     }
 162 }
 163
 164 static size_t decode_utf16(const wchar_t* input, wxUint32& output)
 165 {
 166     if ((*input<0xd800) || (*input>0xdfff))
 167     {
 168         output = *input;
 169         return 1;
 170     }
 171     else if ((input[1]<0xdc00) || (input[1]>=0xdfff))
 172     {
 173         output = *input;
 174         return (size_t)-1;
 175     }
 176     else
 177     {
 178         output = ((input[0] - 0xd7c0) << 10) + (input[1] - 0xdc00);
 179         return 2;
 180     }
 181 }
 182
 183 #endif // WC_UTF16
 184
 185 // ----------------------------------------------------------------------------
 186 // wxMBConv
 187 // ----------------------------------------------------------------------------
 188
 189 #define IGNORE_LIBC 0
 190
 191 size_t wxMBConv::MB2WC(wchar_t *buf, const char *psz, size_t n) const
 192 {
 193 #if IGNORE_LIBC
 194     if (buf)
 195     {
 196         for (size_t i = 0; i < strlen( psz )+1; i++)
 197             buf[i] = (wchar_t) psz[i];
 198         return strlen( psz );
 199     }
 200     else
 201     {
 202         return strlen( psz );
 203     }
 204 #else
 205     return wxMB2WC(buf, psz, n);
 206 #endif
 207 }
 208
 209 size_t wxMBConv::WC2MB(char *buf, const wchar_t *psz, size_t n) const
 210 {
 211 #if IGNORE_LIBC
 212     if (buf)
 213     {
 214         for (size_t i = 0; i < wxStrlen( psz )+1; i++)
 215             buf[i] = (char) psz[i];
 216         return wxStrlen( psz );
 217     }
 218     else
 219     {
 220         return wxStrlen( psz );
 221     }
 222 #else
 223     return wxWC2MB(buf, psz, n);
 224 #endif
 225 }
 226
 227 const wxWCharBuffer wxMBConv::cMB2WC(const char *psz) const
 228 {
 229     if (psz)
 230     {
 231         size_t nLen = MB2WC((wchar_t *) NULL, psz, 0);  // return value excludes /0
 232         if (nLen == (size_t)-1)
 233             return wxWCharBuffer((wchar_t *) NULL);
 234         wxWCharBuffer buf(nLen);                        // this allocates nLen1+
 235         MB2WC((wchar_t *)(const wchar_t *) buf, psz, nLen+1);
 236         return buf;
 237     }
 238     else
 239         return wxWCharBuffer((wchar_t *) NULL);
 240 }
 241
 242 const wxCharBuffer wxMBConv::cWC2MB(const wchar_t *psz) const
 243 {
 244     if (psz)
 245     {
 246         size_t nLen = WC2MB((char *) NULL, psz, 0);  // return value excludes /0
 247         if (nLen == (size_t)-1)
 248             return wxCharBuffer((char *) NULL);
 249         wxCharBuffer buf(nLen);                      // this allocates nLen+1
 250         WC2MB((char *)(const char *) buf, psz, nLen+1);
 251         return buf;
 252     }
 253     else
 254         return wxCharBuffer((char *) NULL);
 255 }
 256
 257 // ----------------------------------------------------------------------------
 258 // standard gdk conversion
 259 // ----------------------------------------------------------------------------
 260
 261 #ifdef __WXGTK12__
 262
 263 WXDLLEXPORT_DATA(wxMBConvGdk) wxConvGdk;
 264
 265 #include <gdk/gdk.h>
 266
 267 size_t wxMBConvGdk::MB2WC(wchar_t *buf, const char *psz, size_t n) const
 268 {
 269     if (buf)
 270     {
 271         return gdk_mbstowcs((GdkWChar *)buf, psz, n);
 272     }
 273     else
 274     {
 275         GdkWChar *nbuf = new GdkWChar[n=strlen(psz)];
 276         size_t len = gdk_mbstowcs(nbuf, psz, n);
 277         delete[] nbuf;
 278         return len;
 279     }
 280 }
 281
 282 size_t wxMBConvGdk::WC2MB(char *buf, const wchar_t *psz, size_t n) const
 283 {
 284     char *mbstr = gdk_wcstombs((GdkWChar *)psz);
 285     size_t len = mbstr ? strlen(mbstr) : 0;
 286     if (buf)
 287     {
 288         if (len > n)
 289             len = n;
 290         memcpy(buf, psz, len);
 291         if (len < n)
 292             buf[len] = 0;
 293     }
 294     return len;
 295 }
 296
 297 #endif // GTK > 1.0
 298
 299 // ----------------------------------------------------------------------------
 300 // UTF-7
 301 // ----------------------------------------------------------------------------
 302
 303 WXDLLEXPORT_DATA(wxMBConvUTF7) wxConvUTF7;
 304
 305 #if 0
 306 static char utf7_setD[]="ABCDEFGHIJKLMNOPQRSTUVWXYZ"
 307                         "abcdefghijklmnopqrstuvwxyz"
 308                         "0123456789'(),-./:?";
 309 static char utf7_setO[]="!\"#$%&*;<=>@[]^_`{|}";
 310 static char utf7_setB[]="ABCDEFGHIJKLMNOPQRSTUVWXYZ"
 311                         "abcdefghijklmnopqrstuvwxyz"
 312                         "0123456789+/";
 313 #endif
 314
 315 // TODO: write actual implementations of UTF-7 here
 316 size_t wxMBConvUTF7::MB2WC(wchar_t * WXUNUSED(buf),
 317                            const char * WXUNUSED(psz),
 318                            size_t WXUNUSED(n)) const
 319 {
 320   return 0;
 321 }
 322
 323 size_t wxMBConvUTF7::WC2MB(char * WXUNUSED(buf),
 324                            const wchar_t * WXUNUSED(psz),
 325                            size_t WXUNUSED(n)) const
 326 {
 327   return 0;
 328 }
 329
 330 // ----------------------------------------------------------------------------
 331 // UTF-8
 332 // ----------------------------------------------------------------------------
 333
 334 WXDLLEXPORT_DATA(wxMBConvUTF8) wxConvUTF8;
 335
 336 static wxUint32 utf8_max[]=
 337     { 0x7f, 0x7ff, 0xffff, 0x1fffff, 0x3ffffff, 0x7fffffff, 0xffffffff };
 338
 339 size_t wxMBConvUTF8::MB2WC(wchar_t *buf, const char *psz, size_t n) const
 340 {
 341     size_t len = 0;
 342
 343     while (*psz && ((!buf) || (len < n)))
 344     {
 345         unsigned char cc = *psz++, fc = cc;
 346         unsigned cnt;
 347         for (cnt = 0; fc & 0x80; cnt++)
 348             fc <<= 1;
 349         if (!cnt)
 350         {
 351             // plain ASCII char
 352             if (buf)
 353                 *buf++ = cc;
 354             len++;
 355         }
 356         else
 357         {
 358             cnt--;
 359             if (!cnt)
 360             {
 361                 // invalid UTF-8 sequence
 362                 return (size_t)-1;
 363             }
 364             else
 365             {
 366                 unsigned ocnt = cnt - 1;
 367                 wxUint32 res = cc & (0x3f >> cnt);
 368                 while (cnt--)
 369                 {
 370                     cc = *psz++;
 371                     if ((cc & 0xC0) != 0x80)
 372                     {
 373                         // invalid UTF-8 sequence
 374                         return (size_t)-1;
 375                     }
 376                     res = (res << 6) | (cc & 0x3f);
 377                 }
 378                 if (res <= utf8_max[ocnt])
 379                 {
 380                     // illegal UTF-8 encoding
 381                     return (size_t)-1;
 382                 }
 383 #ifdef WC_UTF16
 384                 size_t pa = encode_utf16(res, buf);
 385                 if (pa == (size_t)-1)
 386                   return (size_t)-1;
 387                 if (buf)
 388                     buf += pa;
 389                 len += pa;
 390 #else // !WC_UTF16
 391                 if (buf)
 392                     *buf++ = res;
 393                 len++;
 394 #endif // WC_UTF16/!WC_UTF16
 395             }
 396         }
 397     }
 398     if (buf && (len < n))
 399         *buf = 0;
 400     return len;
 401 }
 402
 403 size_t wxMBConvUTF8::WC2MB(char *buf, const wchar_t *psz, size_t n) const
 404 {
 405     size_t len = 0;
 406
 407     while (*psz && ((!buf) || (len < n)))
 408     {
 409         wxUint32 cc;
 410 #ifdef WC_UTF16
 411         size_t pa = decode_utf16(psz, cc);
 412         psz += (pa == (size_t)-1) ? 1 : pa;
 413 #else
 414         cc=(*psz++) & 0x7fffffff;
 415 #endif
 416         unsigned cnt;
 417         for (cnt = 0; cc > utf8_max[cnt]; cnt++) {}
 418         if (!cnt)
 419         {
 420             // plain ASCII char
 421             if (buf)
 422                 *buf++ = (char) cc;
 423             len++;
 424         }
 425
 426         else
 427         {
 428             len += cnt + 1;
 429             if (buf)
 430             {
 431                 *buf++ = (char) ((-128 >> cnt) | ((cc >> (cnt * 6)) & (0x3f >> cnt)));
 432                 while (cnt--)
 433                     *buf++ = (char) (0x80 | ((cc >> (cnt * 6)) & 0x3f));
 434             }
 435         }
 436     }
 437
 438     if (buf && (len<n)) *buf = 0;
 439
 440     return len;
 441 }
 442
 443 // ============================================================================
 444 // wxCharacterSet and derived classes
 445 // ============================================================================
 446
 447 // ----------------------------------------------------------------------------
 448 // wxCharacterSet is the ABC for the classes below
 449 // ----------------------------------------------------------------------------
 450
 451 class wxCharacterSet
 452 {
 453 public:
 454     wxCharacterSet(const wxChar*name) : cname(name) {}
 455     virtual ~wxCharacterSet() {}
 456     virtual size_t MB2WC(wchar_t *buf, const char *psz, size_t n) = 0;
 457     virtual size_t WC2MB(char *buf, const wchar_t *psz, size_t n) = 0;
 458     virtual bool usable() const = 0;
 459 public:
 460     const wxChar*cname;
 461 };
 462
 463 // ----------------------------------------------------------------------------
 464 // ID_CharSet: implementation of wxCharacterSet using an existing wxMBConv
 465 // ----------------------------------------------------------------------------
 466
 467 class ID_CharSet : public wxCharacterSet
 468 {
 469 public:
 470     ID_CharSet(const wxChar *name, wxMBConv *cnv)
 471         : wxCharacterSet(name), work(cnv) {}
 472
 473     size_t MB2WC(wchar_t *buf, const char *psz, size_t n)
 474         { return work ? work->MB2WC(buf,psz,n) : (size_t)-1; }
 475
 476     size_t WC2MB(char *buf, const wchar_t *psz, size_t n)
 477         { return work ? work->WC2MB(buf,psz,n) : (size_t)-1; }
 478
 479     bool usable() const
 480         { return work!=NULL; }
 481 public:
 482     wxMBConv*work;
 483 };
 484
 485
 486 // ============================================================================
 487 // The classes doing conversion using the iconv_xxx() functions
 488 // ============================================================================
 489
 490 #ifdef HAVE_ICONV
 491
 492 // VS: glibc 2.1.3 is broken in that iconv() conversion to/from UCS4 fails with E2BIG
 493 //     if output buffer is _exactly_ as big as needed. Such case is (unless there's
 494 //     yet another bug in glibc) the only case when iconv() returns with (size_t)-1
 495 //     (which means error) and says there are 0 bytes left in the input buffer --
 496 //     when _real_ error occurs, bytes-left-in-input buffer is non-zero. Hence,
 497 //     this alternative test for iconv() failure.
 498 //     [This bug does not appear in glibc 2.2.]
 499 #if defined(__GLIBC__) && __GLIBC__ == 2 && __GLIBC_MINOR__ <= 1
 500 #define ICONV_FAILED(cres, bufLeft) ((cres == (size_t)-1) && \
 501                                      (errno != E2BIG || bufLeft != 0))
 502 #else
 503 #define ICONV_FAILED(cres, bufLeft)  (cres == (size_t)-1)
 504 #endif
 505
 506 #define ICONV_CHAR_CAST(x)  ((ICONV_CONST char **)(x))
 507
 508 // ----------------------------------------------------------------------------
 509 // IC_CharSet: encapsulates an iconv character set
 510 // ----------------------------------------------------------------------------
 511
 512 class IC_CharSet : public wxCharacterSet
 513 {
 514 public:
 515     IC_CharSet(const wxChar *name);
 516     virtual ~IC_CharSet();
 517
 518     virtual size_t MB2WC(wchar_t *buf, const char *psz, size_t n);
 519     virtual size_t WC2MB(char *buf, const wchar_t *psz, size_t n);
 520
 521     bool usable() const
 522         { return (m2w != (iconv_t)-1) && (w2m != (iconv_t)-1); }
 523
 524 protected:
 525     // the iconv handlers used to translate from multibyte to wide char and in
 526     // the other direction
 527     iconv_t m2w,
 528             w2m;
 529
 530 private:
 531     // the name (for iconv_open()) of a wide char charset - if none is
 532     // available on this machine, it will remain NULL
 533     static const char *ms_wcCharsetName;
 534
 535     // true if the wide char encoding we use (i.e. ms_wcCharsetName) has
 536     // different endian-ness than the native one
 537     static bool ms_wcNeedsSwap;
 538 };
 539
 540 const char *IC_CharSet::ms_wcCharsetName = NULL;
 541 bool IC_CharSet::ms_wcNeedsSwap = FALSE;
 542
 543 IC_CharSet::IC_CharSet(const wxChar *name)
 544           : wxCharacterSet(name)
 545 {
 546     // Do it the hard way
 547     char cname[100];
 548     for (size_t i = 0; i < wxStrlen(name)+1; i++)
 549         cname[i] = (char) name[i];
 550
 551     // check for charset that represents wchar_t:
 552     if (ms_wcCharsetName == NULL)
 553     {
 554         ms_wcNeedsSwap = FALSE;
 555
 556         // try charset with explicit bytesex info (e.g. "UCS-4LE"):
 557         ms_wcCharsetName = WC_NAME_BEST;
 558         m2w = iconv_open(ms_wcCharsetName, cname);
 559
 560         if (m2w == (iconv_t)-1)
 561         {
 562             // try charset w/o bytesex info (e.g. "UCS4")
 563             // and check for bytesex ourselves:
 564             ms_wcCharsetName = WC_NAME;
 565             m2w = iconv_open(ms_wcCharsetName, cname);
 566
 567             // last bet, try if it knows WCHAR_T pseudo-charset
 568             if (m2w == (iconv_t)-1)
 569             {
 570                 ms_wcCharsetName = "WCHAR_T";
 571                 m2w = iconv_open(ms_wcCharsetName, cname);
 572             }
 573
 574             if (m2w != (iconv_t)-1)
 575             {
 576                 char    buf[2], *bufPtr;
 577                 wchar_t wbuf[2], *wbufPtr;
 578                 size_t  insz, outsz;
 579                 size_t  res;
 580
 581                 buf[0] = 'A';
 582                 buf[1] = 0;
 583                 wbuf[0] = 0;
 584                 insz = 2;
 585                 outsz = SIZEOF_WCHAR_T * 2;
 586                 wbufPtr = wbuf;
 587                 bufPtr = buf;
 588
 589                 res = iconv(m2w, ICONV_CHAR_CAST(&bufPtr), &insz,
 590                             (char**)&wbufPtr, &outsz);
 591
 592                 if (ICONV_FAILED(res, insz))
 593                 {
 594                     ms_wcCharsetName = NULL;
 595                     wxLogLastError(wxT("iconv"));
 596                     wxLogError(_("Convertion to charset '%s' doesn't work."), name);
 597                 }
 598                 else
 599                 {
 600                     ms_wcNeedsSwap = wbuf[0] != (wchar_t)buf[0];
 601                 }
 602             }
 603             else
 604             {
 605                 ms_wcCharsetName = NULL;
 606
 607                 // VS: we must not output an error here, since wxWindows will safely
 608                 //     fall back to using wxEncodingConverter.
 609                 wxLogTrace(wxT("strconv"), wxT("Impossible to convert to/from charset '%s' with iconv, falling back to wxEncodingConverter."), name);
 610                 //wxLogError(
 611             }
 612         }
 613         wxLogTrace(wxT("strconv"), wxT("wchar_t charset is '%s', needs swap: %i"), ms_wcCharsetName, ms_wcNeedsSwap);
 614     }
 615     else // we already have ms_wcCharsetName
 616     {
 617         m2w = iconv_open(ms_wcCharsetName, cname);
 618     }
 619
 620     // NB: don't ever pass NULL to iconv_open(), it may crash!
 621     if ( ms_wcCharsetName )
 622     {
 623         w2m = iconv_open( cname, ms_wcCharsetName);
 624     }
 625     else
 626     {
 627         w2m = (iconv_t)-1;
 628     }
 629 }
 630
 631 IC_CharSet::~IC_CharSet()
 632 {
 633     if ( m2w != (iconv_t)-1 )
 634         iconv_close(m2w);
 635     if ( w2m != (iconv_t)-1 )
 636         iconv_close(w2m);
 637 }
 638
 639 size_t IC_CharSet::MB2WC(wchar_t *buf, const char *psz, size_t n)
 640 {
 641     size_t inbuf = strlen(psz);
 642     size_t outbuf = n * SIZEOF_WCHAR_T;
 643     size_t res, cres;
 644     // VS: Use these instead of psz, buf because iconv() modifies its arguments:
 645     wchar_t *bufPtr = buf;
 646     const char *pszPtr = psz;
 647
 648     if (buf)
 649     {
 650         // have destination buffer, convert there
 651         cres = iconv(m2w,
 652                      ICONV_CHAR_CAST(&pszPtr), &inbuf,
 653                      (char**)&bufPtr, &outbuf);
 654         res = n - (outbuf / SIZEOF_WCHAR_T);
 655
 656         if (ms_wcNeedsSwap)
 657         {
 658             // convert to native endianness
 659             WC_BSWAP(buf /* _not_ bufPtr */, res)
 660         }
 661
 662         // NB: iconv was given only strlen(psz) characters on input, and so
 663         //     it couldn't convert the trailing zero. Let's do it ourselves
 664         //     if there's some room left for it in the output buffer.
 665         if (res < n)
 666             buf[res] = 0;
 667     }
 668     else
 669     {
 670         // no destination buffer... convert using temp buffer
 671         // to calculate destination buffer requirement
 672         wchar_t tbuf[8];
 673         res = 0;
 674         do {
 675             bufPtr = tbuf;
 676             outbuf = 8*SIZEOF_WCHAR_T;
 677
 678             cres = iconv(m2w,
 679                          ICONV_CHAR_CAST(&pszPtr), &inbuf,
 680                          (char**)&bufPtr, &outbuf );
 681
 682             res += 8-(outbuf/SIZEOF_WCHAR_T);
 683         } while ((cres==(size_t)-1) && (errno==E2BIG));
 684     }
 685
 686     if (ICONV_FAILED(cres, inbuf))
 687     {
 688         //VS: it is ok if iconv fails, hence trace only
 689         wxLogTrace(wxT("strconv"), wxT("iconv failed: %s"), wxSysErrorMsg(wxSysErrorCode()));
 690         return (size_t)-1;
 691     }
 692
 693     return res;
 694 }
 695
 696 size_t IC_CharSet::WC2MB(char *buf, const wchar_t *psz, size_t n)
 697 {
 698 #if defined(__BORLANDC__) && (__BORLANDC__ > 0x530)
 699     size_t inbuf = std::wcslen(psz) * SIZEOF_WCHAR_T;
 700 #else
 701     size_t inbuf = ::wcslen(psz) * SIZEOF_WCHAR_T;
 702 #endif
 703     size_t outbuf = n;
 704     size_t res, cres;
 705
 706     wchar_t *tmpbuf = 0;
 707
 708     if (ms_wcNeedsSwap)
 709     {
 710         // need to copy to temp buffer to switch endianness
 711         // this absolutely doesn't rock!
 712         // (no, doing WC_BSWAP twice on the original buffer won't help, as it
 713         //  could be in read-only memory, or be accessed in some other thread)
 714         tmpbuf=(wchar_t*)malloc((inbuf+1)*SIZEOF_WCHAR_T);
 715         memcpy(tmpbuf,psz,(inbuf+1)*SIZEOF_WCHAR_T);
 716         WC_BSWAP(tmpbuf, inbuf)
 717         psz=tmpbuf;
 718     }
 719
 720     if (buf)
 721     {
 722         // have destination buffer, convert there
 723         cres = iconv( w2m, ICONV_CHAR_CAST(&psz), &inbuf, &buf, &outbuf );
 724
 725         res = n-outbuf;
 726
 727         // NB: iconv was given only wcslen(psz) characters on input, and so
 728         //     it couldn't convert the trailing zero. Let's do it ourselves
 729         //     if there's some room left for it in the output buffer.
 730         if (res < n)
 731             buf[0] = 0;
 732     }
 733     else
 734     {
 735         // no destination buffer... convert using temp buffer
 736         // to calculate destination buffer requirement
 737         char tbuf[16];
 738         res = 0;
 739         do {
 740             buf = tbuf; outbuf = 16;
 741
 742             cres = iconv( w2m, ICONV_CHAR_CAST(&psz), &inbuf, &buf, &outbuf );
 743
 744             res += 16 - outbuf;
 745         } while ((cres==(size_t)-1) && (errno==E2BIG));
 746     }
 747
 748     if (ms_wcNeedsSwap)
 749     {
 750         free(tmpbuf);
 751     }
 752
 753     if (ICONV_FAILED(cres, inbuf))
 754     {
 755         //VS: it is ok if iconv fails, hence trace only
 756         wxLogTrace(wxT("strconv"), wxT("iconv failed: %s"), wxSysErrorMsg(wxSysErrorCode()));
 757         return (size_t)-1;
 758     }
 759
 760     return res;
 761 }
 762
 763 #endif // HAVE_ICONV
 764
 765 // ============================================================================
 766 // Win32 conversion classes
 767 // ============================================================================
 768
 769 #if defined(__WIN32__) && !defined(__WXMICROWIN__)
 770
 771 extern long wxCharsetToCodepage(const wxChar *charset); // from utils.cpp
 772
 773 class CP_CharSet : public wxCharacterSet
 774 {
 775 public:
 776     CP_CharSet(const wxChar* name)
 777         : wxCharacterSet(name)
 778         {
 779             m_CodePage = wxCharsetToCodepage(name);
 780         }
 781
 782     size_t MB2WC(wchar_t *buf, const char *psz, size_t n)
 783     {
 784         size_t len =
 785             MultiByteToWideChar(m_CodePage, 0, psz, -1, buf, buf ? n : 0);
 786         //VS: returns # of written chars for buf!=NULL and *size*
 787         //    needed buffer for buf==NULL
 788         return len ? (buf ? len : len-1) : (size_t)-1;
 789     }
 790
 791     size_t WC2MB(char *buf, const wchar_t *psz, size_t n)
 792     {
 793         size_t len = WideCharToMultiByte(m_CodePage, 0, psz, -1, buf,
 794                                          buf ? n : 0, NULL, NULL);
 795         //VS: returns # of written chars for buf!=NULL and *size*
 796         //    needed buffer for buf==NULL
 797         return len ? (buf ? len : len-1) : (size_t)-1;
 798     }
 799
 800     bool usable() const
 801         { return m_CodePage != -1; }
 802
 803 public:
 804     long m_CodePage;
 805 };
 806 #endif // __WIN32__
 807
 808 // ============================================================================
 809 // wxEncodingConverter based conversion classes
 810 // ============================================================================
 811
 812 #if wxUSE_FONTMAP
 813
 814 class EC_CharSet : public wxCharacterSet
 815 {
 816 public:
 817     // temporarily just use wxEncodingConverter stuff,
 818     // so that it works while a better implementation is built
 819     EC_CharSet(const wxChar* name) : wxCharacterSet(name),
 820                                      enc(wxFONTENCODING_SYSTEM)
 821     {
 822         if (name)
 823             enc = wxFontMapper::Get()->CharsetToEncoding(name, FALSE);
 824
 825         m_ok = m2w.Init(enc, wxFONTENCODING_UNICODE) &&
 826                w2m.Init(wxFONTENCODING_UNICODE, enc);
 827     }
 828
 829     size_t MB2WC(wchar_t *buf, const char *psz, size_t WXUNUSED(n))
 830     {
 831         size_t inbuf = strlen(psz);
 832         if (buf)
 833             m2w.Convert(psz,buf);
 834         return inbuf;
 835     }
 836
 837     size_t WC2MB(char *buf, const wchar_t *psz, size_t WXUNUSED(n))
 838     {
 839 #if ( defined(__BORLANDC__) && (__BORLANDC__ > 0x530) ) \
 840     || ( defined(__MWERKS__) && defined(__WXMSW__) )
 841         size_t inbuf = std::wcslen(psz);
 842 #else
 843         size_t inbuf = ::wcslen(psz);
 844 #endif
 845         if (buf)
 846             w2m.Convert(psz,buf);
 847
 848         return inbuf;
 849     }
 850
 851     bool usable() const { return m_ok; }
 852
 853 public:
 854     wxFontEncoding enc;
 855     wxEncodingConverter m2w, w2m;
 856
 857     // were we initialized successfully?
 858     bool m_ok;
 859 };
 860
 861 #endif // wxUSE_FONTMAP
 862
 863 // ----------------------------------------------------------------------------
 864 // the function creating the wxCharacterSet for the specified charset on the
 865 // current system, trying all possibilities
 866 // ----------------------------------------------------------------------------
 867
 868 static wxCharacterSet *wxGetCharacterSet(const wxChar *name)
 869 {
 870     // check for the special case of ASCII charset
 871 #if wxUSE_FONTMAP
 872     if ( wxFontMapper::Get()->CharsetToEncoding(name) == wxFONTENCODING_DEFAULT )
 873 #else // wxUSE_FONTMAP
 874     if ( !name )
 875 #endif // wxUSE_FONTMAP/!wxUSE_FONTMAP
 876     {
 877         // don't convert at all
 878         return NULL;
 879     }
 880
 881     // the test above must have taken care of this case
 882     wxCHECK_MSG( name, NULL, _T("NULL name must be wxFONTENCODING_DEFAULT") );
 883
 884     wxCharacterSet *cset;
 885
 886     if ( wxStricmp(name, wxT("UTF8")) == 0 || wxStricmp(name, wxT("UTF-8")) == 0)
 887     {
 888         cset = new ID_CharSet(name, &wxConvUTF8);
 889     }
 890     else
 891     {
 892 #ifdef HAVE_ICONV
 893         cset = new IC_CharSet(name);
 894 #else // !HAVE_ICONV
 895         cset = NULL;
 896 #endif // HAVE_ICONV/!HAVE_ICONV
 897     }
 898
 899     // it can only be NULL in this case
 900 #ifndef HAVE_ICONV
 901     if ( cset )
 902 #endif // !HAVE_ICONV
 903     {
 904         if ( cset->usable() )
 905             return cset;
 906
 907         delete cset;
 908         cset = NULL;
 909     }
 910
 911 #if defined(__WIN32__) && !defined(__WXMICROWIN__)
 912     cset = new CP_CharSet(name);
 913     if ( cset->usable() )
 914         return cset;
 915
 916     delete cset;
 917     cset = NULL;
 918 #endif // __WIN32__
 919
 920 #if wxUSE_FONTMAP
 921     cset = new EC_CharSet(name);
 922     if ( cset->usable() )
 923         return cset;
 924
 925     delete cset;
 926     cset = NULL;
 927 #endif // wxUSE_FONTMAP
 928
 929     wxLogError(_("Cannot convert from encoding '%s'!"), name);
 930
 931     return NULL;
 932 }
 933
 934 // ============================================================================
 935 // wxCSConv implementation
 936 // ============================================================================
 937
 938 wxCSConv::wxCSConv(const wxChar *charset)
 939 {
 940     m_name = (wxChar *)NULL;
 941     m_cset = (wxCharacterSet *) NULL;
 942     m_deferred = TRUE;
 943
 944     SetName(charset);
 945 }
 946
 947 wxCSConv::~wxCSConv()
 948 {
 949     Clear();
 950 }
 951
 952 wxCSConv::wxCSConv(const wxCSConv& conv)
 953     : wxMBConv()
 954 {
 955     Clear();
 956     SetName(conv.m_name);
 957 }
 958
 959 wxCSConv& wxCSConv::operator=(const wxCSConv& conv)
 960 {
 961     Clear();
 962     SetName(conv.m_name);
 963     return *this;
 964 }
 965
 966 void wxCSConv::Clear()
 967 {
 968     if (m_name)
 969         free(m_name);
 970     if (m_cset)
 971         delete m_cset;
 972     m_name = NULL;
 973     m_cset = NULL;
 974 }
 975
 976 void wxCSConv::SetName(const wxChar *charset)
 977 {
 978     if (charset)
 979     {
 980         m_name = wxStrdup(charset);
 981         m_deferred = TRUE;
 982     }
 983 }
 984
 985 void wxCSConv::LoadNow()
 986 {
 987     if (m_deferred)
 988     {
 989         if ( !m_name )
 990         {
 991             wxString name = wxLocale::GetSystemEncodingName();
 992             if ( !name.empty() )
 993             {
 994                 SetName(name);
 995             }
 996         }
 997
 998         // wxGetCharacterSet() complains about NULL name
 999         m_cset = m_name ? wxGetCharacterSet(m_name) : NULL;
1000         m_deferred = FALSE;
1001     }
1002 }
1003
1004 size_t wxCSConv::MB2WC(wchar_t *buf, const char *psz, size_t n) const
1005 {
1006     ((wxCSConv *)this)->LoadNow(); // discard constness
1007
1008     if (m_cset)
1009         return m_cset->MB2WC(buf, psz, n);
1010
1011     // latin-1 (direct)
1012     size_t len = strlen(psz);
1013
1014     if (buf)
1015     {
1016         for (size_t c = 0; c <= len; c++)
1017             buf[c] = (unsigned char)(psz[c]);
1018     }
1019
1020     return len;
1021 }
1022
1023 size_t wxCSConv::WC2MB(char *buf, const wchar_t *psz, size_t n) const
1024 {
1025     ((wxCSConv *)this)->LoadNow(); // discard constness
1026
1027     if (m_cset)
1028         return m_cset->WC2MB(buf, psz, n);
1029
1030     // latin-1 (direct)
1031 #if ( defined(__BORLANDC__) && (__BORLANDC__ > 0x530) ) \
1032     || ( defined(__MWERKS__) && defined(__WXMSW__) )
1033     size_t len=std::wcslen(psz);
1034 #else
1035     size_t len=::wcslen(psz);
1036 #endif
1037     if (buf)
1038     {
1039         for (size_t c = 0; c <= len; c++)
1040             buf[c] = (psz[c] > 0xff) ? '?' : psz[c];
1041     }
1042
1043     return len;
1044 }
1045
1046 #endif // wxUSE_WCHAR_T
1047
1048