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 #include "wx/msw/registry.h"
 412 // this should work if M$ Internet Exploiter is installed
 413 static long CharsetToCodepage(const wxChar *name)
 414 {
 415     if (!name)
 416         return GetACP();
 417
 418     long CP=-1;
 419
 420     wxString cn(name);
 421     do {
 422         wxString path(wxT("MIME\\Database\\Charset\\"));
 423         path += cn;
 424         wxRegKey key(wxRegKey::HKCR, path);
 425
 426         if (!key.Exists()) break;
 427
 428         // two cases: either there's an AliasForCharset string,
 429         // or there are Codepage and InternetEncoding dwords.
 430         // The InternetEncoding gives us the actual encoding,
 431         // the Codepage just says which Windows character set to
 432         // use when displaying the data.
 433         if (key.HasValue(wxT("InternetEncoding")) &&
 434             key.QueryValue(wxT("InternetEncoding"), &CP)) break;
 435
 436         // no encoding, see if it's an alias
 437         if (!key.HasValue(wxT("AliasForCharset")) ||
 438             !key.QueryValue(wxT("AliasForCharset"), cn)) break;
 439     } while (1);
 440
 441     return CP;
 442 }
 443 #endif
 444
 445 class wxCharacterSet
 446 {
 447 public:
 448     wxCharacterSet(const wxChar*name)
 449         : cname(name) {}
 450     virtual ~wxCharacterSet()
 451         {}
 452     virtual size_t MB2WC(wchar_t *buf, const char *psz, size_t n)
 453         { return (size_t)-1; }
 454     virtual size_t WC2MB(char *buf, const wchar_t *psz, size_t n)
 455         { return (size_t)-1; }
 456     virtual bool usable()
 457         { return FALSE; }
 458 public:
 459     const wxChar*cname;
 460 };
 461
 462 class ID_CharSet : public wxCharacterSet
 463 {
 464 public:
 465     ID_CharSet(const wxChar *name,wxMBConv *cnv)
 466         : wxCharacterSet(name), work(cnv) {}
 467
 468     size_t MB2WC(wchar_t *buf, const char *psz, size_t n)
 469         { return work ? work->MB2WC(buf,psz,n) : (size_t)-1; }
 470
 471     size_t WC2MB(char *buf, const wchar_t *psz, size_t n)
 472         { return work ? work->WC2MB(buf,psz,n) : (size_t)-1; }
 473
 474     bool usable()
 475         { return work!=NULL; }
 476 public:
 477     wxMBConv*work;
 478 };
 479
 480
 481 #ifdef HAVE_ICONV_H
 482
 483 bool g_wcNeedsSwap = FALSE;
 484 static const char *g_wcCharset = NULL;
 485
 486 // VS: glibc 2.1.3 is broken in that iconv() conversion to/from UCS4 fails with E2BIG
 487 //     if output buffer is _exactly_ as big as needed. Such case is (unless there's
 488 //     yet another bug in glibc) the only case when iconv() returns with (size_t)-1
 489 //     (which means error) and says there are 0 bytes left in the input buffer --
 490 //     when _real_ error occurs, bytes-left-in-input buffer is non-zero. Hence,
 491 //     this alternative test for iconv() failure.
 492 //     [This bug does not appear in glibc 2.2.]
 493 #if defined(__GLIBC__) && __GLIBC__ == 2 && __GLIBC_MINOR__ <= 1
 494 #define ICONV_FAILED(cres, bufLeft) ((cres == (size_t)-1) && \
 495                                      (errno != E2BIG || bufLeft != 0))
 496 #else
 497 #define ICONV_FAILED(cres, bufLeft)  (cres == (size_t)-1)
 498 #endif
 499
 500 class IC_CharSet : public wxCharacterSet
 501 {
 502 public:
 503     IC_CharSet(const wxChar *name)
 504         : wxCharacterSet(name)
 505     {
 506         // check for charset that represents wchar_t:
 507         if (g_wcCharset == NULL)
 508         {
 509             g_wcNeedsSwap = FALSE;
 510
 511             // try charset with explicit bytesex info (e.g. "UCS-4LE"):
 512             g_wcCharset = WC_NAME_BEST;
 513             m2w = iconv_open(g_wcCharset, wxConvLibc.cWX2MB(name));
 514
 515             if (m2w == (iconv_t)-1)
 516             {
 517                 // try charset w/o bytesex info (e.g. "UCS4")
 518                 // and check for bytesex ourselves:
 519                 g_wcCharset = WC_NAME;
 520                 m2w = iconv_open(g_wcCharset, wxConvLibc.cWX2MB(name));
 521
 522                 // last bet, try if it knows WCHAR_T pseudo-charset
 523                 if (m2w == (iconv_t)-1)
 524                 {
 525                     g_wcCharset = "WCHAR_T";
 526                     m2w = iconv_open(g_wcCharset, wxConvLibc.cWX2MB(name));
 527                 }
 528
 529                 if (m2w != (iconv_t)-1)
 530                 {
 531                     char    buf[2], *bufPtr;
 532                     wchar_t wbuf[2], *wbufPtr;
 533                     size_t  insz, outsz;
 534                     size_t  res;
 535
 536                     buf[0] = 'A';
 537                     buf[1] = 0;
 538                     wbuf[0] = 0;
 539                     insz = 2;
 540                     outsz = SIZEOF_WCHAR_T * 2;
 541                     wbufPtr = wbuf;
 542                     bufPtr = buf;
 543
 544                     #ifdef WX_ICONV_TAKES_CHAR
 545                     res = iconv(m2w, (char**)&bufPtr, &insz, (char**)&wbufPtr, &outsz);
 546                     #else
 547                     res = iconv(m2w, (const char**)&bufPtr, &insz, (char**)&wbufPtr, &outsz);
 548                     #endif
 549                     if (ICONV_FAILED(res, insz))
 550                     {
 551                         g_wcCharset = NULL;
 552                         wxLogLastError(wxT("iconv"));
 553                         wxLogError(_("Convertion to charset '%s' doesn't work."), name);
 554                     }
 555                     else
 556                     {
 557                         g_wcNeedsSwap = (wbuf[0] != (wchar_t)buf[0]);
 558                     }
 559                 }
 560                 else
 561                 {
 562                     g_wcCharset = NULL;
 563                     wxLogError(_("Don't know how to convert to/from charset '%s'."), name);
 564                 }
 565             }
 566             wxLogTrace(wxT("strconv"), wxT("wchar_t charset is '%s', needs swap: %i"), g_wcCharset, g_wcNeedsSwap);
 567         }
 568         else
 569             m2w = iconv_open(g_wcCharset, wxConvLibc.cWX2MB(name));
 570
 571         w2m = iconv_open(wxConvLibc.cWX2MB(name), g_wcCharset);
 572     }
 573
 574     ~IC_CharSet()
 575     {
 576         if ( m2w != (iconv_t)-1 )
 577             iconv_close(m2w);
 578         if ( w2m != (iconv_t)-1 )
 579             iconv_close(w2m);
 580     }
 581
 582     size_t MB2WC(wchar_t *buf, const char *psz, size_t n)
 583     {
 584         size_t inbuf = strlen(psz);
 585         size_t outbuf = n * SIZEOF_WCHAR_T;
 586         size_t res, cres;
 587         // VS: Use these instead of psz, buf because iconv() modifies its arguments:
 588         wchar_t *bufPtr = buf;
 589         const char *pszPtr = psz;
 590
 591         if (buf)
 592         {
 593             // have destination buffer, convert there
 594 #ifdef WX_ICONV_TAKES_CHAR
 595             cres = iconv(m2w, (char**)&pszPtr, &inbuf, (char**)&bufPtr, &outbuf);
 596 #else
 597             cres = iconv(m2w, &pszPtr, &inbuf, (char**)&bufPtr, &outbuf);
 598 #endif
 599             res = n - (outbuf / SIZEOF_WCHAR_T);
 600
 601             if (g_wcNeedsSwap)
 602             {
 603                 // convert to native endianness
 604                 WC_BSWAP(buf /* _not_ bufPtr */, res)
 605             }
 606         }
 607         else
 608         {
 609             // no destination buffer... convert using temp buffer
 610             // to calculate destination buffer requirement
 611             wchar_t tbuf[8];
 612             res = 0;
 613             do {
 614                 bufPtr = tbuf; outbuf = 8*SIZEOF_WCHAR_T;
 615 #ifdef WX_ICONV_TAKES_CHAR
 616                 cres = iconv( m2w, (char**)&pszPtr, &inbuf, (char**)&bufPtr, &outbuf );
 617 #else
 618                 cres = iconv( m2w, &pszPtr, &inbuf, (char**)&bufPtr, &outbuf );
 619 #endif
 620                 res += 8-(outbuf/SIZEOF_WCHAR_T);
 621             } while ((cres==(size_t)-1) && (errno==E2BIG));
 622         }
 623
 624         if (ICONV_FAILED(cres, inbuf))
 625         {
 626             //VS: it is ok if iconv fails, hence trace only
 627             wxLogTrace(wxT("strconv"), wxT("iconv failed: %s"), wxSysErrorMsg(wxSysErrorCode()));
 628             return (size_t)-1;
 629         }
 630
 631         return res;
 632     }
 633
 634     size_t WC2MB(char *buf, const wchar_t *psz, size_t n)
 635     {
 636 #if defined(__BORLANDC__) && (__BORLANDC__ > 0x530)
 637         size_t inbuf = std::wcslen(psz) * SIZEOF_WCHAR_T;
 638 #else
 639         size_t inbuf = ::wcslen(psz) * SIZEOF_WCHAR_T;
 640 #endif
 641         size_t outbuf = n;
 642         size_t res, cres;
 643
 644         wchar_t *tmpbuf;
 645
 646         if (g_wcNeedsSwap)
 647         {
 648             // need to copy to temp buffer to switch endianness
 649             // this absolutely doesn't rock!
 650             // (no, doing WC_BSWAP twice on the original buffer won't help, as it
 651             //  could be in read-only memory, or be accessed in some other thread)
 652             tmpbuf=(wchar_t*)malloc((inbuf+1)*SIZEOF_WCHAR_T);
 653             memcpy(tmpbuf,psz,(inbuf+1)*SIZEOF_WCHAR_T);
 654             WC_BSWAP(tmpbuf, inbuf)
 655             psz=tmpbuf;
 656         }
 657
 658         if (buf)
 659         {
 660             // have destination buffer, convert there
 661 #ifdef WX_ICONV_TAKES_CHAR
 662             cres = iconv( w2m, (char**)&psz, &inbuf, &buf, &outbuf );
 663 #else
 664             cres = iconv( w2m, (const char**)&psz, &inbuf, &buf, &outbuf );
 665 #endif
 666             res = n-outbuf;
 667         }
 668         else
 669         {
 670             // no destination buffer... convert using temp buffer
 671             // to calculate destination buffer requirement
 672             char tbuf[16];
 673             res = 0;
 674             do {
 675                 buf = tbuf; outbuf = 16;
 676 #ifdef WX_ICONV_TAKES_CHAR
 677                 cres = iconv( w2m, (char**)&psz, &inbuf, &buf, &outbuf );
 678 #else
 679                 cres = iconv( w2m, (const char**)&psz, &inbuf, &buf, &outbuf );
 680 #endif
 681                 res += 16 - outbuf;
 682             } while ((cres==(size_t)-1) && (errno==E2BIG));
 683         }
 684
 685         if (g_wcNeedsSwap)
 686         {
 687             free(tmpbuf);
 688         }
 689
 690         if (ICONV_FAILED(cres, inbuf))
 691         {
 692             //VS: it is ok if iconv fails, hence trace only
 693             wxLogTrace(wxT("strconv"), wxT("iconv failed: %s"), wxSysErrorMsg(wxSysErrorCode()));
 694             return (size_t)-1;
 695         }
 696
 697         return res;
 698     }
 699
 700     bool usable()
 701         { return (m2w != (iconv_t)-1) && (w2m != (iconv_t)-1); }
 702
 703 protected:
 704     iconv_t m2w, w2m;
 705 };
 706 #endif
 707
 708 #if defined(__WIN32__) && !defined(__WXMICROWIN__)
 709 class CP_CharSet : public wxCharacterSet
 710 {
 711 public:
 712     CP_CharSet(const wxChar*name)
 713         : wxCharacterSet(name), CodePage(CharsetToCodepage(name)) {}
 714
 715     size_t MB2WC(wchar_t *buf, const char *psz, size_t n)
 716     {
 717         size_t len =
 718             MultiByteToWideChar(CodePage, 0, psz, -1, buf, buf ? n : 0);
 719         //VS: returns # of written chars for buf!=NULL and *size*
 720         //    needed buffer for buf==NULL
 721         return len ? (buf ? len : len-1) : (size_t)-1;
 722     }
 723
 724     size_t WC2MB(char *buf, const wchar_t *psz, size_t n)
 725     {
 726         size_t len = WideCharToMultiByte(CodePage, 0, psz, -1, buf,
 727                                          buf ? n : 0, NULL, NULL);
 728         //VS: returns # of written chars for buf!=NULL and *size*
 729         //    needed buffer for buf==NULL
 730         return len ? (buf ? len : len-1) : (size_t)-1;
 731     }
 732
 733     bool usable()
 734         { return CodePage != -1; }
 735
 736 public:
 737     long CodePage;
 738 };
 739 #endif // __WIN32__
 740
 741 #if wxUSE_FONTMAP
 742
 743 class EC_CharSet : public wxCharacterSet
 744 {
 745 public:
 746     // temporarily just use wxEncodingConverter stuff,
 747     // so that it works while a better implementation is built
 748     EC_CharSet(const wxChar*name) : wxCharacterSet(name),
 749                                     enc(wxFONTENCODING_SYSTEM)
 750     {
 751         if (name)
 752             enc = wxTheFontMapper->CharsetToEncoding(name, FALSE);
 753         m2w.Init(enc, wxFONTENCODING_UNICODE);
 754         w2m.Init(wxFONTENCODING_UNICODE, enc);
 755     }
 756
 757     size_t MB2WC(wchar_t *buf, const char *psz, size_t n)
 758     {
 759         size_t inbuf = strlen(psz);
 760         if (buf)
 761             m2w.Convert(psz,buf);
 762         return inbuf;
 763     }
 764
 765     size_t WC2MB(char *buf, const wchar_t *psz, size_t n)
 766     {
 767 #if defined(__BORLANDC__) && (__BORLANDC__ > 0x530)
 768         size_t inbuf = std::wcslen(psz);
 769 #else
 770         size_t inbuf = ::wcslen(psz);
 771 #endif
 772         if (buf)
 773             w2m.Convert(psz,buf);
 774
 775         return inbuf;
 776     }
 777
 778     bool usable()
 779         { return (enc!=wxFONTENCODING_SYSTEM) && (enc!=wxFONTENCODING_DEFAULT); }
 780
 781 public:
 782     wxFontEncoding enc;
 783     wxEncodingConverter m2w, w2m;
 784 };
 785
 786 #endif // wxUSE_FONTMAP
 787
 788 static wxCharacterSet *wxGetCharacterSet(const wxChar *name)
 789 {
 790     wxCharacterSet *cset = NULL;
 791     if (name)
 792     {
 793         if (wxStricmp(name, wxT("UTF8")) == 0 || wxStricmp(name, wxT("UTF-8")) == 0)
 794         {
 795             cset = new ID_CharSet(name, &wxConvUTF8);
 796         }
 797         else
 798         {
 799 #ifdef HAVE_ICONV_H
 800             cset = new IC_CharSet(name); // may not take NULL
 801 #endif
 802         }
 803     }
 804
 805     if (cset && cset->usable())
 806         return cset;
 807
 808     if (cset)
 809     {
 810         delete cset;
 811         cset = NULL;
 812     }
 813
 814 #if defined(__WIN32__) && !defined(__WXMICROWIN__)
 815     cset = new CP_CharSet(name); // may take NULL
 816     if (cset->usable())
 817         return cset;
 818
 819     delete cset;
 820 #endif // __WIN32__
 821
 822 #if wxUSE_FONTMAP
 823     cset = new EC_CharSet(name);
 824     if (cset->usable())
 825         return cset;
 826 #endif // wxUSE_FONTMAP
 827
 828     delete cset;
 829     wxLogError(_("Unknown encoding '%s'!"), name);
 830     return NULL;
 831 }
 832
 833 wxCSConv::wxCSConv(const wxChar *charset)
 834 {
 835     m_name = (wxChar *)NULL;
 836     m_cset = (wxCharacterSet *) NULL;
 837     m_deferred = TRUE;
 838
 839     SetName(charset);
 840 }
 841
 842 wxCSConv::~wxCSConv()
 843 {
 844     free(m_name);
 845     delete m_cset;
 846 }
 847
 848 void wxCSConv::SetName(const wxChar *charset)
 849 {
 850     if (charset)
 851     {
 852         m_name = wxStrdup(charset);
 853         m_deferred = TRUE;
 854     }
 855 }
 856
 857 void wxCSConv::LoadNow()
 858 {
 859     if (m_deferred)
 860     {
 861         if ( !m_name )
 862         {
 863             wxString name = wxLocale::GetSystemEncodingName();
 864             if ( !name.empty() )
 865                 SetName(name);
 866         }
 867
 868         // wxGetCharacterSet() complains about NULL name
 869         m_cset = m_name ? wxGetCharacterSet(m_name) : NULL;
 870         m_deferred = FALSE;
 871     }
 872 }
 873
 874 size_t wxCSConv::MB2WC(wchar_t *buf, const char *psz, size_t n) const
 875 {
 876     ((wxCSConv *)this)->LoadNow(); // discard constness
 877
 878     if (m_cset)
 879         return m_cset->MB2WC(buf, psz, n);
 880
 881     // latin-1 (direct)
 882     size_t len = strlen(psz);
 883
 884     if (buf)
 885     {
 886         for (size_t c = 0; c <= len; c++)
 887             buf[c] = (unsigned char)(psz[c]);
 888     }
 889
 890     return len;
 891 }
 892
 893 size_t wxCSConv::WC2MB(char *buf, const wchar_t *psz, size_t n) const
 894 {
 895     ((wxCSConv *)this)->LoadNow(); // discard constness
 896
 897     if (m_cset)
 898         return m_cset->WC2MB(buf, psz, n);
 899
 900     // latin-1 (direct)
 901 #if defined(__BORLANDC__) && (__BORLANDC__ > 0x530)
 902     size_t len=std::wcslen(psz);
 903 #else
 904     size_t len=::wcslen(psz);
 905 #endif
 906     if (buf)
 907     {
 908         for (size_t c = 0; c <= len; c++)
 909             buf[c] = (psz[c] > 0xff) ? '?' : psz[c];
 910     }
 911
 912     return len;
 913 }
 914
 915 #ifdef HAVE_ICONV_H
 916
 917 class IC_CharSetConverter
 918 {
 919 public:
 920     IC_CharSetConverter(IC_CharSet *from, IC_CharSet *to)
 921     {
 922         cnv = iconv_open(wxConvLibc.cWX2MB(to->cname),
 923                          wxConvLibc.cWX2MB(from->cname));
 924     }
 925
 926     ~IC_CharSetConverter()
 927     {
 928         if (cnv != (iconv_t)-1)
 929             iconv_close(cnv);
 930     }
 931
 932     size_t Convert(char *buf, const char *psz, size_t n)
 933     {
 934         size_t inbuf = strlen(psz);
 935         size_t outbuf = n;
 936 #ifdef WX_ICONV_TAKES_CHAR
 937         size_t res = iconv( cnv, (char**)&psz, &inbuf, &buf, &outbuf );
 938 #else
 939         size_t res = iconv( cnv, &psz, &inbuf, &buf, &outbuf );
 940 #endif
 941         if (res == (size_t)-1)
 942             return (size_t)-1;
 943         return (n - outbuf);
 944     }
 945
 946 public:
 947     iconv_t cnv;
 948 };
 949
 950 #endif // HAVE_ICONV_H
 951
 952 class EC_CharSetConverter
 953 {
 954 public:
 955     EC_CharSetConverter(EC_CharSet* from,EC_CharSet* to)
 956         { cnv.Init(from->enc,to->enc); }
 957
 958     size_t Convert(char* buf, const char* psz, size_t n)
 959     {
 960         size_t inbuf = strlen(psz);
 961         if (buf) cnv.Convert(psz,buf);
 962         return inbuf;
 963     }
 964
 965 public:
 966     wxEncodingConverter cnv;
 967 };
 968
 969 #else // !wxUSE_WCHAR_T
 970
 971 // ----------------------------------------------------------------------------
 972 // stand-ins in absence of wchar_t
 973 // ----------------------------------------------------------------------------
 974
 975 WXDLLEXPORT_DATA(wxMBConv) wxConvLibc, wxConvFile;
 976
 977 #endif // wxUSE_WCHAR_T
 978
 979