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 #ifdef __WXMSW__
  32   #include "wx/msw/private.h"
  33 #endif
  34
  35 #include <errno.h>
  36 #include <ctype.h>
  37 #include <string.h>
  38 #include <stdlib.h>
  39
  40
  41 #include "wx/debug.h"
  42 #include "wx/strconv.h"
  43 #include "wx/intl.h"
  44 #include "wx/log.h"
  45
  46 // ----------------------------------------------------------------------------
  47 // globals
  48 // ----------------------------------------------------------------------------
  49
  50 WXDLLEXPORT_DATA(wxMBConv *) wxConvCurrent = &wxConvLibc;
  51
  52
  53 // ============================================================================
  54 // implementation
  55 // ============================================================================
  56
  57 #if wxUSE_WCHAR_T
  58
  59 #ifdef __SALFORDC__
  60   #include <clib.h>
  61 #endif
  62
  63 #ifdef HAVE_ICONV_H
  64   #include <iconv.h>
  65 #endif
  66
  67 #ifdef __WXMSW__
  68   #include <windows.h>
  69 #endif
  70
  71 #define BSWAP_UCS4(str, len) { unsigned _c; for (_c=0; _c<len; _c++) str[_c]=wxUINT32_SWAP_ALWAYS(str[_c]); }
  72 #define BSWAP_UTF16(str, len) { unsigned _c; for (_c=0; _c<len; _c++) str[_c]=wxUINT16_SWAP_ALWAYS(str[_c]); }
  73
  74 // under Unix SIZEOF_WCHAR_T is defined by configure, but under other platforms
  75 // it might be not defined - assume the most common value
  76 #ifndef SIZEOF_WCHAR_T
  77     #define SIZEOF_WCHAR_T 2
  78 #endif // !defined(SIZEOF_WCHAR_T)
  79
  80 #if SIZEOF_WCHAR_T == 4
  81     #define WC_NAME         "UCS4"
  82     #define WC_BSWAP         BSWAP_UCS4
  83     #ifdef WORDS_BIGENDIAN
  84       #define WC_NAME_BEST  "UCS-4BE"
  85     #else
  86       #define WC_NAME_BEST  "UCS-4LE"
  87     #endif
  88 #elif SIZEOF_WCHAR_T == 2
  89     #define WC_NAME         "UTF16"
  90     #define WC_BSWAP         BSWAP_UTF16
  91     #define WC_UTF16
  92     #ifdef WORDS_BIGENDIAN
  93       #define WC_NAME_BEST  "UTF-16BE"
  94     #else
  95       #define WC_NAME_BEST  "UTF-16LE"
  96     #endif
  97 #else // sizeof(wchar_t) != 2 nor 4
  98     // I don't know what to do about this
  99     #error "Weird sizeof(wchar_t): please report your platform details to wx-users mailing list"
 100 #endif
 101
 102
 103 #ifdef WC_UTF16
 104
 105 static size_t encode_utf16(wxUint32 input, wchar_t *output)
 106 {
 107     if (input<=0xffff)
 108     {
 109         if (output) *output++ = input;
 110         return 1;
 111     }
 112     else if (input>=0x110000)
 113     {
 114         return (size_t)-1;
 115     }
 116     else
 117     {
 118         if (output)
 119         {
 120             *output++ = (input >> 10)+0xd7c0;
 121             *output++ = (input&0x3ff)+0xdc00;
 122         }
 123         return 2;
 124     }
 125 }
 126
 127 static size_t decode_utf16(const wchar_t* input, wxUint32& output)
 128 {
 129     if ((*input<0xd800) || (*input>0xdfff))
 130     {
 131         output = *input;
 132         return 1;
 133     }
 134     else if ((input[1]<0xdc00) || (input[1]>=0xdfff))
 135     {
 136         output = *input;
 137         return (size_t)-1;
 138     }
 139     else
 140     {
 141         output = ((input[0] - 0xd7c0) << 10) + (input[1] - 0xdc00);
 142         return 2;
 143     }
 144 }
 145
 146 #endif // WC_UTF16
 147
 148 // ----------------------------------------------------------------------------
 149 // wxMBConv
 150 // ----------------------------------------------------------------------------
 151
 152 WXDLLEXPORT_DATA(wxMBConv) wxConvLibc;
 153
 154 size_t wxMBConv::MB2WC(wchar_t *buf, const char *psz, size_t n) const
 155 {
 156     return wxMB2WC(buf, psz, n);
 157 }
 158
 159 size_t wxMBConv::WC2MB(char *buf, const wchar_t *psz, size_t n) const
 160 {
 161     return wxWC2MB(buf, psz, n);
 162 }
 163
 164 const wxWCharBuffer wxMBConv::cMB2WC(const char *psz) const
 165 {
 166     if (psz)
 167     {
 168         size_t nLen = MB2WC((wchar_t *) NULL, psz, 0);
 169         if (nLen == (size_t)-1)
 170             return wxWCharBuffer((wchar_t *) NULL);
 171         wxWCharBuffer buf(nLen);
 172         MB2WC((wchar_t *)(const wchar_t *) buf, psz, nLen);
 173         return buf;
 174     }
 175     else
 176         return wxWCharBuffer((wchar_t *) NULL);
 177 }
 178
 179 const wxCharBuffer wxMBConv::cWC2MB(const wchar_t *psz) const
 180 {
 181     if (psz)
 182     {
 183         size_t nLen = WC2MB((char *) NULL, psz, 0);
 184         if (nLen == (size_t)-1)
 185             return wxCharBuffer((char *) NULL);
 186         wxCharBuffer buf(nLen);
 187         WC2MB((char *)(const char *) buf, psz, nLen);
 188         return buf;
 189     }
 190     else
 191         return wxCharBuffer((char *) NULL);
 192 }
 193
 194 // ----------------------------------------------------------------------------
 195 // standard file conversion
 196 // ----------------------------------------------------------------------------
 197
 198 WXDLLEXPORT_DATA(wxMBConvFile) wxConvFile;
 199
 200 // just use the libc conversion for now
 201 size_t wxMBConvFile::MB2WC(wchar_t *buf, const char *psz, size_t n) const
 202 {
 203     return wxMB2WC(buf, psz, n);
 204 }
 205
 206 size_t wxMBConvFile::WC2MB(char *buf, const wchar_t *psz, size_t n) const
 207 {
 208     return wxWC2MB(buf, psz, n);
 209 }
 210
 211 // ----------------------------------------------------------------------------
 212 // standard gdk conversion
 213 // ----------------------------------------------------------------------------
 214
 215 #ifdef __WXGTK12__
 216
 217 WXDLLEXPORT_DATA(wxMBConvGdk) wxConvGdk;
 218
 219 #include <gdk/gdk.h>
 220
 221 size_t wxMBConvGdk::MB2WC(wchar_t *buf, const char *psz, size_t n) const
 222 {
 223     if (buf)
 224     {
 225         return gdk_mbstowcs((GdkWChar *)buf, psz, n);
 226     }
 227     else
 228     {
 229         GdkWChar *nbuf = new GdkWChar[n=strlen(psz)];
 230         size_t len = gdk_mbstowcs(nbuf, psz, n);
 231         delete[] nbuf;
 232         return len;
 233     }
 234 }
 235
 236 size_t wxMBConvGdk::WC2MB(char *buf, const wchar_t *psz, size_t n) const
 237 {
 238     char *mbstr = gdk_wcstombs((GdkWChar *)psz);
 239     size_t len = mbstr ? strlen(mbstr) : 0;
 240     if (buf)
 241     {
 242         if (len > n)
 243             len = n;
 244         memcpy(buf, psz, len);
 245         if (len < n)
 246             buf[len] = 0;
 247     }
 248     return len;
 249 }
 250
 251 #endif // GTK > 1.0
 252
 253 // ----------------------------------------------------------------------------
 254 // UTF-7
 255 // ----------------------------------------------------------------------------
 256
 257 WXDLLEXPORT_DATA(wxMBConvUTF7) wxConvUTF7;
 258
 259 #if 0
 260 static char utf7_setD[]="ABCDEFGHIJKLMNOPQRSTUVWXYZ"
 261                         "abcdefghijklmnopqrstuvwxyz"
 262                         "0123456789'(),-./:?";
 263 static char utf7_setO[]="!\"#$%&*;<=>@[]^_`{|}";
 264 static char utf7_setB[]="ABCDEFGHIJKLMNOPQRSTUVWXYZ"
 265                         "abcdefghijklmnopqrstuvwxyz"
 266                         "0123456789+/";
 267 #endif
 268
 269 // TODO: write actual implementations of UTF-7 here
 270 size_t wxMBConvUTF7::MB2WC(wchar_t * WXUNUSED(buf),
 271                            const char * WXUNUSED(psz),
 272                            size_t WXUNUSED(n)) const
 273 {
 274   return 0;
 275 }
 276
 277 size_t wxMBConvUTF7::WC2MB(char * WXUNUSED(buf),
 278                            const wchar_t * WXUNUSED(psz),
 279                            size_t WXUNUSED(n)) const
 280 {
 281   return 0;
 282 }
 283
 284 // ----------------------------------------------------------------------------
 285 // UTF-8
 286 // ----------------------------------------------------------------------------
 287
 288 WXDLLEXPORT_DATA(wxMBConvUTF8) wxConvUTF8;
 289
 290 static wxUint32 utf8_max[]=
 291     { 0x7f, 0x7ff, 0xffff, 0x1fffff, 0x3ffffff, 0x7fffffff, 0xffffffff };
 292
 293 size_t wxMBConvUTF8::MB2WC(wchar_t *buf, const char *psz, size_t n) const
 294 {
 295     size_t len = 0;
 296
 297     while (*psz && ((!buf) || (len < n)))
 298     {
 299         unsigned char cc = *psz++, fc = cc;
 300         unsigned cnt;
 301         for (cnt = 0; fc & 0x80; cnt++)
 302             fc <<= 1;
 303         if (!cnt)
 304         {
 305             // plain ASCII char
 306             if (buf)
 307                 *buf++ = cc;
 308             len++;
 309         }
 310         else
 311         {
 312             cnt--;
 313             if (!cnt)
 314             {
 315                 // invalid UTF-8 sequence
 316                 return (size_t)-1;
 317             }
 318             else
 319             {
 320                 unsigned ocnt = cnt - 1;
 321                 wxUint32 res = cc & (0x3f >> cnt);
 322                 while (cnt--)
 323                 {
 324                     cc = *psz++;
 325                     if ((cc & 0xC0) != 0x80)
 326                     {
 327                         // invalid UTF-8 sequence
 328                         return (size_t)-1;
 329                     }
 330                     res = (res << 6) | (cc & 0x3f);
 331                 }
 332                 if (res <= utf8_max[ocnt])
 333                 {
 334                     // illegal UTF-8 encoding
 335                     return (size_t)-1;
 336                 }
 337 #ifdef WC_UTF16
 338                 size_t pa = encode_utf16(res, buf);
 339                 if (pa == (size_t)-1)
 340                   return (size_t)-1;
 341                 if (buf)
 342                     buf += pa;
 343                 len += pa;
 344 #else
 345                 if (buf)
 346                     *buf++ = res;
 347                 len++;
 348 #endif
 349             }
 350         }
 351     }
 352     if (buf && (len < n))
 353         *buf = 0;
 354     return len;
 355 }
 356
 357 size_t wxMBConvUTF8::WC2MB(char *buf, const wchar_t *psz, size_t n) const
 358 {
 359     size_t len = 0;
 360
 361     while (*psz && ((!buf) || (len < n)))
 362     {
 363         wxUint32 cc;
 364 #ifdef WC_UTF16
 365         size_t pa = decode_utf16(psz, cc);
 366         psz += (pa == (size_t)-1) ? 1 : pa;
 367 #else
 368         cc=(*psz++) & 0x7fffffff;
 369 #endif
 370         unsigned cnt;
 371         for (cnt = 0; cc > utf8_max[cnt]; cnt++) {}
 372         if (!cnt)
 373         {
 374             // plain ASCII char
 375             if (buf)
 376                 *buf++ = cc;
 377             len++;
 378         }
 379
 380         else
 381         {
 382             len += cnt + 1;
 383             if (buf)
 384             {
 385                 *buf++ = (-128 >> cnt) | ((cc >> (cnt * 6)) & (0x3f >> cnt));
 386                 while (cnt--)
 387                     *buf++ = 0x80 | ((cc >> (cnt * 6)) & 0x3f);
 388             }
 389         }
 390     }
 391
 392     if (buf && (len<n)) *buf = 0;
 393     return len;
 394 }
 395
 396 // ----------------------------------------------------------------------------
 397 // specified character set
 398 // ----------------------------------------------------------------------------
 399
 400 WXDLLEXPORT_DATA(wxCSConv) wxConvLocal((const wxChar *)NULL);
 401
 402 #include "wx/encconv.h"
 403 #include "wx/fontmap.h"
 404
 405 // TODO: add some tables here
 406 // - perhaps common encodings to common codepages (for Win32)
 407 // - perhaps common encodings to objects ("UTF8" -> wxConvUTF8)
 408 // - move wxEncodingConverter meat in here
 409
 410 #if defined(__WIN32__) && !defined(__WXMICROWIN__)
 411
 412 #if wxUSE_GUI
 413
 414 // VZ: the new version of wxCharsetToCodepage() is more politically correct
 415 //     and should work on other Windows versions as well but the old version is
 416 //     still needed for !wxUSE_FONTMAP || !wxUSE_GUI case
 417
 418 extern long wxEncodingToCodepage(wxFontEncoding encoding)
 419 {
 420     // translate encoding into the Windows CHARSET
 421     wxNativeEncodingInfo natveEncInfo;
 422     if ( !wxGetNativeFontEncoding(encoding, &natveEncInfo) )
 423         return -1;
 424
 425     // translate CHARSET to code page
 426     CHARSETINFO csetInfo;
 427     if ( !::TranslateCharsetInfo((DWORD *)(DWORD)natveEncInfo.charset,
 428                                  &csetInfo,
 429                                  TCI_SRCCHARSET) )
 430     {
 431         wxLogLastError(_T("TranslateCharsetInfo(TCI_SRCCHARSET)"));
 432
 433         return -1;
 434     }
 435
 436     return csetInfo.ciACP;
 437 }
 438
 439 #if wxUSE_FONTMAP
 440
 441 extern long wxCharsetToCodepage(const wxChar *name)
 442 {
 443     // first get the font encoding for this charset
 444     if ( !name )
 445         return -1;
 446
 447     wxFontEncoding enc = wxTheFontMapper->CharsetToEncoding(name, FALSE);
 448     if ( enc == wxFONTENCODING_SYSTEM )
 449         return -1;
 450
 451     // the use the helper function
 452     return wxEncodingToCodepage(enc);
 453 }
 454
 455 #endif // wxUSE_FONTMAP
 456
 457 #endif // wxUSE_GUI
 458
 459 // include old wxCharsetToCodepage() by OK if needed
 460 #if !wxUSE_GUI || !wxUSE_FONTMAP
 461
 462 #include "wx/msw/registry.h"
 463
 464 // this should work if Internet Exploiter is installed
 465 extern long wxCharsetToCodepage(const wxChar *name)
 466 {
 467     if (!name)
 468         return GetACP();
 469
 470     long CP=-1;
 471
 472     wxString cn(name);
 473     do {
 474         wxString path(wxT("MIME\\Database\\Charset\\"));
 475         path += cn;
 476         wxRegKey key(wxRegKey::HKCR, path);
 477
 478         if (!key.Exists()) break;
 479
 480         // two cases: either there's an AliasForCharset string,
 481         // or there are Codepage and InternetEncoding dwords.
 482         // The InternetEncoding gives us the actual encoding,
 483         // the Codepage just says which Windows character set to
 484         // use when displaying the data.
 485         if (key.HasValue(wxT("InternetEncoding")) &&
 486             key.QueryValue(wxT("InternetEncoding"), &CP)) break;
 487
 488         // no encoding, see if it's an alias
 489         if (!key.HasValue(wxT("AliasForCharset")) ||
 490             !key.QueryValue(wxT("AliasForCharset"), cn)) break;
 491     } while (1);
 492
 493     return CP;
 494 }
 495
 496 #endif // !wxUSE_GUI || !wxUSE_FONTMAP
 497
 498 #endif // Win32
 499
 500 class wxCharacterSet
 501 {
 502 public:
 503     wxCharacterSet(const wxChar*name)
 504         : cname(name) {}
 505     virtual ~wxCharacterSet()
 506         {}
 507     virtual size_t MB2WC(wchar_t *buf, const char *psz, size_t n)
 508         { return (size_t)-1; }
 509     virtual size_t WC2MB(char *buf, const wchar_t *psz, size_t n)
 510         { return (size_t)-1; }
 511     virtual bool usable()
 512         { return FALSE; }
 513 public:
 514     const wxChar*cname;
 515 };
 516
 517 class ID_CharSet : public wxCharacterSet
 518 {
 519 public:
 520     ID_CharSet(const wxChar *name,wxMBConv *cnv)
 521         : wxCharacterSet(name), work(cnv) {}
 522
 523     size_t MB2WC(wchar_t *buf, const char *psz, size_t n)
 524         { return work ? work->MB2WC(buf,psz,n) : (size_t)-1; }
 525
 526     size_t WC2MB(char *buf, const wchar_t *psz, size_t n)
 527         { return work ? work->WC2MB(buf,psz,n) : (size_t)-1; }
 528
 529     bool usable()
 530         { return work!=NULL; }
 531 public:
 532     wxMBConv*work;
 533 };
 534
 535
 536 #ifdef HAVE_ICONV_H
 537
 538 bool g_wcNeedsSwap = FALSE;
 539 static const char *g_wcCharset = NULL;
 540
 541 // VS: glibc 2.1.3 is broken in that iconv() conversion to/from UCS4 fails with E2BIG
 542 //     if output buffer is _exactly_ as big as needed. Such case is (unless there's
 543 //     yet another bug in glibc) the only case when iconv() returns with (size_t)-1
 544 //     (which means error) and says there are 0 bytes left in the input buffer --
 545 //     when _real_ error occurs, bytes-left-in-input buffer is non-zero. Hence,
 546 //     this alternative test for iconv() failure.
 547 //     [This bug does not appear in glibc 2.2.]
 548 #if defined(__GLIBC__) && __GLIBC__ == 2 && __GLIBC_MINOR__ <= 1
 549 #define ICONV_FAILED(cres, bufLeft) ((cres == (size_t)-1) && \
 550                                      (errno != E2BIG || bufLeft != 0))
 551 #else
 552 #define ICONV_FAILED(cres, bufLeft)  (cres == (size_t)-1)
 553 #endif
 554
 555 class IC_CharSet : public wxCharacterSet
 556 {
 557 public:
 558     IC_CharSet(const wxChar *name)
 559         : wxCharacterSet(name)
 560     {
 561         // check for charset that represents wchar_t:
 562         if (g_wcCharset == NULL)
 563         {
 564             g_wcNeedsSwap = FALSE;
 565
 566             // try charset with explicit bytesex info (e.g. "UCS-4LE"):
 567             g_wcCharset = WC_NAME_BEST;
 568             m2w = iconv_open(g_wcCharset, wxConvLibc.cWX2MB(name));
 569
 570             if (m2w == (iconv_t)-1)
 571             {
 572                 // try charset w/o bytesex info (e.g. "UCS4")
 573                 // and check for bytesex ourselves:
 574                 g_wcCharset = WC_NAME;
 575                 m2w = iconv_open(g_wcCharset, wxConvLibc.cWX2MB(name));
 576
 577                 // last bet, try if it knows WCHAR_T pseudo-charset
 578                 if (m2w == (iconv_t)-1)
 579                 {
 580                     g_wcCharset = "WCHAR_T";
 581                     m2w = iconv_open(g_wcCharset, wxConvLibc.cWX2MB(name));
 582                 }
 583
 584                 if (m2w != (iconv_t)-1)
 585                 {
 586                     char    buf[2], *bufPtr;
 587                     wchar_t wbuf[2], *wbufPtr;
 588                     size_t  insz, outsz;
 589                     size_t  res;
 590
 591                     buf[0] = 'A';
 592                     buf[1] = 0;
 593                     wbuf[0] = 0;
 594                     insz = 2;
 595                     outsz = SIZEOF_WCHAR_T * 2;
 596                     wbufPtr = wbuf;
 597                     bufPtr = buf;
 598
 599                     #ifdef WX_ICONV_TAKES_CHAR
 600                     res = iconv(m2w, (char**)&bufPtr, &insz, (char**)&wbufPtr, &outsz);
 601                     #else
 602                     res = iconv(m2w, (const char**)&bufPtr, &insz, (char**)&wbufPtr, &outsz);
 603                     #endif
 604                     if (ICONV_FAILED(res, insz))
 605                     {
 606                         g_wcCharset = NULL;
 607                         wxLogLastError(wxT("iconv"));
 608                         wxLogError(_("Convertion to charset '%s' doesn't work."), name);
 609                     }
 610                     else
 611                     {
 612                         g_wcNeedsSwap = (wbuf[0] != (wchar_t)buf[0]);
 613                     }
 614                 }
 615                 else
 616                 {
 617                     g_wcCharset = NULL;
 618                     wxLogError(_("Don't know how to convert to/from charset '%s'."), name);
 619                 }
 620             }
 621             wxLogTrace(wxT("strconv"), wxT("wchar_t charset is '%s', needs swap: %i"), g_wcCharset, g_wcNeedsSwap);
 622         }
 623         else
 624             m2w = iconv_open(g_wcCharset, wxConvLibc.cWX2MB(name));
 625
 626         w2m = iconv_open(wxConvLibc.cWX2MB(name), g_wcCharset);
 627     }
 628
 629     ~IC_CharSet()
 630     {
 631         if ( m2w != (iconv_t)-1 )
 632             iconv_close(m2w);
 633         if ( w2m != (iconv_t)-1 )
 634             iconv_close(w2m);
 635     }
 636
 637     size_t MB2WC(wchar_t *buf, const char *psz, size_t n)
 638     {
 639         size_t inbuf = strlen(psz);
 640         size_t outbuf = n * SIZEOF_WCHAR_T;
 641         size_t res, cres;
 642         // VS: Use these instead of psz, buf because iconv() modifies its arguments:
 643         wchar_t *bufPtr = buf;
 644         const char *pszPtr = psz;
 645
 646         if (buf)
 647         {
 648             // have destination buffer, convert there
 649 #ifdef WX_ICONV_TAKES_CHAR
 650             cres = iconv(m2w, (char**)&pszPtr, &inbuf, (char**)&bufPtr, &outbuf);
 651 #else
 652             cres = iconv(m2w, &pszPtr, &inbuf, (char**)&bufPtr, &outbuf);
 653 #endif
 654             res = n - (outbuf / SIZEOF_WCHAR_T);
 655
 656             if (g_wcNeedsSwap)
 657             {
 658                 // convert to native endianness
 659                 WC_BSWAP(buf /* _not_ bufPtr */, res)
 660             }
 661         }
 662         else
 663         {
 664             // no destination buffer... convert using temp buffer
 665             // to calculate destination buffer requirement
 666             wchar_t tbuf[8];
 667             res = 0;
 668             do {
 669                 bufPtr = tbuf; outbuf = 8*SIZEOF_WCHAR_T;
 670 #ifdef WX_ICONV_TAKES_CHAR
 671                 cres = iconv( m2w, (char**)&pszPtr, &inbuf, (char**)&bufPtr, &outbuf );
 672 #else
 673                 cres = iconv( m2w, &pszPtr, &inbuf, (char**)&bufPtr, &outbuf );
 674 #endif
 675                 res += 8-(outbuf/SIZEOF_WCHAR_T);
 676             } while ((cres==(size_t)-1) && (errno==E2BIG));
 677         }
 678
 679         if (ICONV_FAILED(cres, inbuf))
 680         {
 681             //VS: it is ok if iconv fails, hence trace only
 682             wxLogTrace(wxT("strconv"), wxT("iconv failed: %s"), wxSysErrorMsg(wxSysErrorCode()));
 683             return (size_t)-1;
 684         }
 685
 686         return res;
 687     }
 688
 689     size_t WC2MB(char *buf, const wchar_t *psz, size_t n)
 690     {
 691 #if defined(__BORLANDC__) && (__BORLANDC__ > 0x530)
 692         size_t inbuf = std::wcslen(psz) * SIZEOF_WCHAR_T;
 693 #else
 694         size_t inbuf = ::wcslen(psz) * SIZEOF_WCHAR_T;
 695 #endif
 696         size_t outbuf = n;
 697         size_t res, cres;
 698
 699         wchar_t *tmpbuf = 0;
 700
 701         if (g_wcNeedsSwap)
 702         {
 703             // need to copy to temp buffer to switch endianness
 704             // this absolutely doesn't rock!
 705             // (no, doing WC_BSWAP twice on the original buffer won't help, as it
 706             //  could be in read-only memory, or be accessed in some other thread)
 707             tmpbuf=(wchar_t*)malloc((inbuf+1)*SIZEOF_WCHAR_T);
 708             memcpy(tmpbuf,psz,(inbuf+1)*SIZEOF_WCHAR_T);
 709             WC_BSWAP(tmpbuf, inbuf)
 710             psz=tmpbuf;
 711         }
 712
 713         if (buf)
 714         {
 715             // have destination buffer, convert there
 716 #ifdef WX_ICONV_TAKES_CHAR
 717             cres = iconv( w2m, (char**)&psz, &inbuf, &buf, &outbuf );
 718 #else
 719             cres = iconv( w2m, (const char**)&psz, &inbuf, &buf, &outbuf );
 720 #endif
 721             res = n-outbuf;
 722         }
 723         else
 724         {
 725             // no destination buffer... convert using temp buffer
 726             // to calculate destination buffer requirement
 727             char tbuf[16];
 728             res = 0;
 729             do {
 730                 buf = tbuf; outbuf = 16;
 731 #ifdef WX_ICONV_TAKES_CHAR
 732                 cres = iconv( w2m, (char**)&psz, &inbuf, &buf, &outbuf );
 733 #else
 734                 cres = iconv( w2m, (const char**)&psz, &inbuf, &buf, &outbuf );
 735 #endif
 736                 res += 16 - outbuf;
 737             } while ((cres==(size_t)-1) && (errno==E2BIG));
 738         }
 739
 740         if (g_wcNeedsSwap)
 741         {
 742             free(tmpbuf);
 743         }
 744
 745         if (ICONV_FAILED(cres, inbuf))
 746         {
 747             //VS: it is ok if iconv fails, hence trace only
 748             wxLogTrace(wxT("strconv"), wxT("iconv failed: %s"), wxSysErrorMsg(wxSysErrorCode()));
 749             return (size_t)-1;
 750         }
 751
 752         return res;
 753     }
 754
 755     bool usable()
 756         { return (m2w != (iconv_t)-1) && (w2m != (iconv_t)-1); }
 757
 758 protected:
 759     iconv_t m2w, w2m;
 760 };
 761 #endif
 762
 763 #if defined(__WIN32__) && !defined(__WXMICROWIN__)
 764 class CP_CharSet : public wxCharacterSet
 765 {
 766 public:
 767     CP_CharSet(const wxChar* name)
 768         : wxCharacterSet(name)
 769         {
 770             m_CodePage = wxCharsetToCodepage(name);
 771         }
 772
 773     size_t MB2WC(wchar_t *buf, const char *psz, size_t n)
 774     {
 775         size_t len =
 776             MultiByteToWideChar(m_CodePage, 0, psz, -1, buf, buf ? n : 0);
 777         //VS: returns # of written chars for buf!=NULL and *size*
 778         //    needed buffer for buf==NULL
 779         return len ? (buf ? len : len-1) : (size_t)-1;
 780     }
 781
 782     size_t WC2MB(char *buf, const wchar_t *psz, size_t n)
 783     {
 784         size_t len = WideCharToMultiByte(m_CodePage, 0, psz, -1, buf,
 785                                          buf ? n : 0, NULL, NULL);
 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     bool usable()
 792         { return m_CodePage != -1; }
 793
 794 public:
 795     long m_CodePage;
 796 };
 797 #endif // __WIN32__
 798
 799 #if wxUSE_FONTMAP
 800
 801 class EC_CharSet : public wxCharacterSet
 802 {
 803 public:
 804     // temporarily just use wxEncodingConverter stuff,
 805     // so that it works while a better implementation is built
 806     EC_CharSet(const wxChar* name) : wxCharacterSet(name),
 807                                      enc(wxFONTENCODING_SYSTEM)
 808     {
 809         if (name)
 810             enc = wxTheFontMapper->CharsetToEncoding(name, FALSE);
 811         m2w.Init(enc, wxFONTENCODING_UNICODE);
 812         w2m.Init(wxFONTENCODING_UNICODE, enc);
 813     }
 814
 815     size_t MB2WC(wchar_t *buf, const char *psz, size_t n)
 816     {
 817         size_t inbuf = strlen(psz);
 818         if (buf)
 819             m2w.Convert(psz,buf);
 820         return inbuf;
 821     }
 822
 823     size_t WC2MB(char *buf, const wchar_t *psz, size_t n)
 824     {
 825 #if ( defined(__BORLANDC__) && (__BORLANDC__ > 0x530) ) \
 826     || ( defined(__MWERKS__) && defined(__WXMSW__) )
 827         size_t inbuf = std::wcslen(psz);
 828 #else
 829         size_t inbuf = ::wcslen(psz);
 830 #endif
 831         if (buf)
 832             w2m.Convert(psz,buf);
 833
 834         return inbuf;
 835     }
 836
 837     bool usable()
 838         { return (enc!=wxFONTENCODING_SYSTEM) && (enc!=wxFONTENCODING_DEFAULT); }
 839
 840 public:
 841     wxFontEncoding enc;
 842     wxEncodingConverter m2w, w2m;
 843 };
 844
 845 #endif // wxUSE_FONTMAP
 846
 847 static wxCharacterSet *wxGetCharacterSet(const wxChar *name)
 848 {
 849     wxCharacterSet *cset = NULL;
 850     if (name)
 851     {
 852         if (wxStricmp(name, wxT("UTF8")) == 0 || wxStricmp(name, wxT("UTF-8")) == 0)
 853         {
 854             cset = new ID_CharSet(name, &wxConvUTF8);
 855         }
 856         else
 857         {
 858 #ifdef HAVE_ICONV_H
 859             cset = new IC_CharSet(name); // may not take NULL
 860 #endif
 861         }
 862     }
 863
 864     if (cset && cset->usable())
 865         return cset;
 866
 867     if (cset)
 868     {
 869         delete cset;
 870         cset = NULL;
 871     }
 872
 873 #if defined(__WIN32__) && !defined(__WXMICROWIN__)
 874     cset = new CP_CharSet(name); // may take NULL
 875     if (cset->usable())
 876         return cset;
 877
 878     delete cset;
 879 #endif // __WIN32__
 880
 881 #if wxUSE_FONTMAP
 882     cset = new EC_CharSet(name);
 883     if (cset->usable())
 884         return cset;
 885 #endif // wxUSE_FONTMAP
 886
 887     delete cset;
 888     wxLogError(_("Unknown encoding '%s'!"), name);
 889     return NULL;
 890 }
 891
 892 wxCSConv::wxCSConv(const wxChar *charset)
 893 {
 894     m_name = (wxChar *)NULL;
 895     m_cset = (wxCharacterSet *) NULL;
 896     m_deferred = TRUE;
 897
 898     SetName(charset);
 899 }
 900
 901 wxCSConv::~wxCSConv()
 902 {
 903     free(m_name);
 904     delete m_cset;
 905 }
 906
 907 void wxCSConv::SetName(const wxChar *charset)
 908 {
 909     if (charset)
 910     {
 911         m_name = wxStrdup(charset);
 912         m_deferred = TRUE;
 913     }
 914 }
 915
 916 void wxCSConv::LoadNow()
 917 {
 918     if (m_deferred)
 919     {
 920         if ( !m_name )
 921         {
 922             wxString name = wxLocale::GetSystemEncodingName();
 923             if ( !name.empty() )
 924                 SetName(name);
 925         }
 926
 927         // wxGetCharacterSet() complains about NULL name
 928         m_cset = m_name ? wxGetCharacterSet(m_name) : NULL;
 929         m_deferred = FALSE;
 930     }
 931 }
 932
 933 size_t wxCSConv::MB2WC(wchar_t *buf, const char *psz, size_t n) const
 934 {
 935     ((wxCSConv *)this)->LoadNow(); // discard constness
 936
 937     if (m_cset)
 938         return m_cset->MB2WC(buf, psz, n);
 939
 940     // latin-1 (direct)
 941     size_t len = strlen(psz);
 942
 943     if (buf)
 944     {
 945         for (size_t c = 0; c <= len; c++)
 946             buf[c] = (unsigned char)(psz[c]);
 947     }
 948
 949     return len;
 950 }
 951
 952 size_t wxCSConv::WC2MB(char *buf, const wchar_t *psz, size_t n) const
 953 {
 954     ((wxCSConv *)this)->LoadNow(); // discard constness
 955
 956     if (m_cset)
 957         return m_cset->WC2MB(buf, psz, n);
 958
 959     // latin-1 (direct)
 960 #if ( defined(__BORLANDC__) && (__BORLANDC__ > 0x530) ) \
 961     || ( defined(__MWERKS__) && defined(__WXMSW__) )
 962     size_t len=std::wcslen(psz);
 963 #else
 964     size_t len=::wcslen(psz);
 965 #endif
 966     if (buf)
 967     {
 968         for (size_t c = 0; c <= len; c++)
 969             buf[c] = (psz[c] > 0xff) ? '?' : psz[c];
 970     }
 971
 972     return len;
 973 }
 974
 975 #ifdef HAVE_ICONV_H
 976
 977 class IC_CharSetConverter
 978 {
 979 public:
 980     IC_CharSetConverter(IC_CharSet *from, IC_CharSet *to)
 981     {
 982         cnv = iconv_open(wxConvLibc.cWX2MB(to->cname),
 983                          wxConvLibc.cWX2MB(from->cname));
 984     }
 985
 986     ~IC_CharSetConverter()
 987     {
 988         if (cnv != (iconv_t)-1)
 989             iconv_close(cnv);
 990     }
 991
 992     size_t Convert(char *buf, const char *psz, size_t n)
 993     {
 994         size_t inbuf = strlen(psz);
 995         size_t outbuf = n;
 996 #ifdef WX_ICONV_TAKES_CHAR
 997         size_t res = iconv( cnv, (char**)&psz, &inbuf, &buf, &outbuf );
 998 #else
 999         size_t res = iconv( cnv, &psz, &inbuf, &buf, &outbuf );
1000 #endif
1001         if (res == (size_t)-1)
1002             return (size_t)-1;
1003         return (n - outbuf);
1004     }
1005
1006 public:
1007     iconv_t cnv;
1008 };
1009
1010 #endif // HAVE_ICONV_H
1011
1012 class EC_CharSetConverter
1013 {
1014 public:
1015     EC_CharSetConverter(EC_CharSet* from,EC_CharSet* to)
1016         { cnv.Init(from->enc,to->enc); }
1017
1018     size_t Convert(char* buf, const char* psz, size_t n)
1019     {
1020         size_t inbuf = strlen(psz);
1021         if (buf) cnv.Convert(psz,buf);
1022         return inbuf;
1023     }
1024
1025 public:
1026     wxEncodingConverter cnv;
1027 };
1028
1029 #else // !wxUSE_WCHAR_T
1030
1031 // ----------------------------------------------------------------------------
1032 // stand-ins in absence of wchar_t
1033 // ----------------------------------------------------------------------------
1034
1035 WXDLLEXPORT_DATA(wxMBConv) wxConvLibc, wxConvFile;
1036
1037 #endif // wxUSE_WCHAR_T
1038
1039