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
  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 // ============================================================================
 501 // wxCharacterSet and derived classes
 502 // ============================================================================
 503
 504 // ----------------------------------------------------------------------------
 505 // wxCharacterSet is the ABC for the classes below
 506 // ----------------------------------------------------------------------------
 507
 508 class wxCharacterSet
 509 {
 510 public:
 511     wxCharacterSet(const wxChar*name) : cname(name) {}
 512     virtual ~wxCharacterSet() {}
 513     virtual size_t MB2WC(wchar_t *buf, const char *psz, size_t n) = 0;
 514     virtual size_t WC2MB(char *buf, const wchar_t *psz, size_t n) = 0;
 515     virtual bool usable() const = 0;
 516 public:
 517     const wxChar*cname;
 518 };
 519
 520 // ----------------------------------------------------------------------------
 521 // ID_CharSet: implementation of wxCharacterSet using an existing wxMBConv
 522 // ----------------------------------------------------------------------------
 523
 524 class ID_CharSet : public wxCharacterSet
 525 {
 526 public:
 527     ID_CharSet(const wxChar *name, wxMBConv *cnv)
 528         : wxCharacterSet(name), work(cnv) {}
 529
 530     size_t MB2WC(wchar_t *buf, const char *psz, size_t n)
 531         { return work ? work->MB2WC(buf,psz,n) : (size_t)-1; }
 532
 533     size_t WC2MB(char *buf, const wchar_t *psz, size_t n)
 534         { return work ? work->WC2MB(buf,psz,n) : (size_t)-1; }
 535
 536     bool usable() const
 537         { return work!=NULL; }
 538 public:
 539     wxMBConv*work;
 540 };
 541
 542
 543 // ============================================================================
 544 // The classes doing conversion using the iconv_xxx() functions
 545 // ============================================================================
 546
 547 #ifdef HAVE_ICONV
 548
 549 // VS: glibc 2.1.3 is broken in that iconv() conversion to/from UCS4 fails with E2BIG
 550 //     if output buffer is _exactly_ as big as needed. Such case is (unless there's
 551 //     yet another bug in glibc) the only case when iconv() returns with (size_t)-1
 552 //     (which means error) and says there are 0 bytes left in the input buffer --
 553 //     when _real_ error occurs, bytes-left-in-input buffer is non-zero. Hence,
 554 //     this alternative test for iconv() failure.
 555 //     [This bug does not appear in glibc 2.2.]
 556 #if defined(__GLIBC__) && __GLIBC__ == 2 && __GLIBC_MINOR__ <= 1
 557 #define ICONV_FAILED(cres, bufLeft) ((cres == (size_t)-1) && \
 558                                      (errno != E2BIG || bufLeft != 0))
 559 #else
 560 #define ICONV_FAILED(cres, bufLeft)  (cres == (size_t)-1)
 561 #endif
 562
 563 #define ICONV_CHAR_CAST(x)  ((ICONV_CONST char **)(x))
 564
 565 // ----------------------------------------------------------------------------
 566 // IC_CharSet: encapsulates an iconv character set
 567 // ----------------------------------------------------------------------------
 568
 569 class IC_CharSet : public wxCharacterSet
 570 {
 571 public:
 572     IC_CharSet(const wxChar *name);
 573     virtual ~IC_CharSet();
 574
 575     virtual size_t MB2WC(wchar_t *buf, const char *psz, size_t n);
 576     virtual size_t WC2MB(char *buf, const wchar_t *psz, size_t n);
 577
 578     bool usable() const
 579         { return (m2w != (iconv_t)-1) && (w2m != (iconv_t)-1); }
 580
 581 protected:
 582     // the iconv handlers used to translate from multibyte to wide char and in
 583     // the other direction
 584     iconv_t m2w,
 585             w2m;
 586
 587 private:
 588     // the name (for iconv_open()) of a wide char charset - if none is
 589     // available on this machine, it will remain NULL
 590     static const char *ms_wcCharsetName;
 591
 592     // true if the wide char encoding we use (i.e. ms_wcCharsetName) has
 593     // different endian-ness than the native one
 594     static bool ms_wcNeedsSwap;
 595 };
 596
 597 const char *IC_CharSet::ms_wcCharsetName = NULL;
 598 bool IC_CharSet::ms_wcNeedsSwap = FALSE;
 599
 600 IC_CharSet::IC_CharSet(const wxChar *name)
 601           : wxCharacterSet(name)
 602 {
 603     // check for charset that represents wchar_t:
 604     if (ms_wcCharsetName == NULL)
 605     {
 606         ms_wcNeedsSwap = FALSE;
 607
 608         // try charset with explicit bytesex info (e.g. "UCS-4LE"):
 609         ms_wcCharsetName = WC_NAME_BEST;
 610         m2w = iconv_open(ms_wcCharsetName, wxConvLibc.cWX2MB(name));
 611
 612         if (m2w == (iconv_t)-1)
 613         {
 614             // try charset w/o bytesex info (e.g. "UCS4")
 615             // and check for bytesex ourselves:
 616             ms_wcCharsetName = WC_NAME;
 617             m2w = iconv_open(ms_wcCharsetName, wxConvLibc.cWX2MB(name));
 618
 619             // last bet, try if it knows WCHAR_T pseudo-charset
 620             if (m2w == (iconv_t)-1)
 621             {
 622                 ms_wcCharsetName = "WCHAR_T";
 623                 m2w = iconv_open(ms_wcCharsetName, wxConvLibc.cWX2MB(name));
 624             }
 625
 626             if (m2w != (iconv_t)-1)
 627             {
 628                 char    buf[2], *bufPtr;
 629                 wchar_t wbuf[2], *wbufPtr;
 630                 size_t  insz, outsz;
 631                 size_t  res;
 632
 633                 buf[0] = 'A';
 634                 buf[1] = 0;
 635                 wbuf[0] = 0;
 636                 insz = 2;
 637                 outsz = SIZEOF_WCHAR_T * 2;
 638                 wbufPtr = wbuf;
 639                 bufPtr = buf;
 640
 641                 res = iconv(m2w, ICONV_CHAR_CAST(&bufPtr), &insz,
 642                             (char**)&wbufPtr, &outsz);
 643
 644                 if (ICONV_FAILED(res, insz))
 645                 {
 646                     ms_wcCharsetName = NULL;
 647                     wxLogLastError(wxT("iconv"));
 648                     wxLogError(_("Convertion to charset '%s' doesn't work."), name);
 649                 }
 650                 else
 651                 {
 652                     ms_wcNeedsSwap = wbuf[0] != (wchar_t)buf[0];
 653                 }
 654             }
 655             else
 656             {
 657                 ms_wcCharsetName = NULL;
 658
 659                 // VS: we must not output an error here, since wxWindows will safely
 660                 //     fall back to using wxEncodingConverter.
 661                 wxLogTrace(wxT("strconv"), wxT("Impossible to convert to/from charset '%s' with iconv, falling back to wxEncodingConverter."), name);
 662                 //wxLogError(
 663             }
 664         }
 665         wxLogTrace(wxT("strconv"), wxT("wchar_t charset is '%s', needs swap: %i"), ms_wcCharsetName, ms_wcNeedsSwap);
 666     }
 667     else // we already have ms_wcCharsetName
 668     {
 669         m2w = iconv_open(ms_wcCharsetName, wxConvLibc.cWX2MB(name));
 670     }
 671
 672     // NB: don't ever pass NULL to iconv_open(), it may crash!
 673     if ( ms_wcCharsetName )
 674     {
 675         w2m = iconv_open(wxConvLibc.cWX2MB(name), ms_wcCharsetName);
 676     }
 677     else
 678     {
 679         w2m = (iconv_t)-1;
 680     }
 681 }
 682
 683 IC_CharSet::~IC_CharSet()
 684 {
 685     if ( m2w != (iconv_t)-1 )
 686         iconv_close(m2w);
 687     if ( w2m != (iconv_t)-1 )
 688         iconv_close(w2m);
 689 }
 690
 691 size_t IC_CharSet::MB2WC(wchar_t *buf, const char *psz, size_t n)
 692 {
 693     size_t inbuf = strlen(psz);
 694     size_t outbuf = n * SIZEOF_WCHAR_T;
 695     size_t res, cres;
 696     // VS: Use these instead of psz, buf because iconv() modifies its arguments:
 697     wchar_t *bufPtr = buf;
 698     const char *pszPtr = psz;
 699
 700     if (buf)
 701     {
 702         // have destination buffer, convert there
 703         cres = iconv(m2w,
 704                      ICONV_CHAR_CAST(&pszPtr), &inbuf,
 705                      (char**)&bufPtr, &outbuf);
 706         res = n - (outbuf / SIZEOF_WCHAR_T);
 707
 708         if (ms_wcNeedsSwap)
 709         {
 710             // convert to native endianness
 711             WC_BSWAP(buf /* _not_ bufPtr */, res)
 712         }
 713     }
 714     else
 715     {
 716         // no destination buffer... convert using temp buffer
 717         // to calculate destination buffer requirement
 718         wchar_t tbuf[8];
 719         res = 0;
 720         do {
 721             bufPtr = tbuf;
 722             outbuf = 8*SIZEOF_WCHAR_T;
 723
 724             cres = iconv(m2w,
 725                          ICONV_CHAR_CAST(&pszPtr), &inbuf,
 726                          (char**)&bufPtr, &outbuf );
 727
 728             res += 8-(outbuf/SIZEOF_WCHAR_T);
 729         } while ((cres==(size_t)-1) && (errno==E2BIG));
 730     }
 731
 732     if (ICONV_FAILED(cres, inbuf))
 733     {
 734         //VS: it is ok if iconv fails, hence trace only
 735         wxLogTrace(wxT("strconv"), wxT("iconv failed: %s"), wxSysErrorMsg(wxSysErrorCode()));
 736         return (size_t)-1;
 737     }
 738
 739     return res;
 740 }
 741
 742 size_t IC_CharSet::WC2MB(char *buf, const wchar_t *psz, size_t n)
 743 {
 744 #if defined(__BORLANDC__) && (__BORLANDC__ > 0x530)
 745     size_t inbuf = std::wcslen(psz) * SIZEOF_WCHAR_T;
 746 #else
 747     size_t inbuf = ::wcslen(psz) * SIZEOF_WCHAR_T;
 748 #endif
 749     size_t outbuf = n;
 750     size_t res, cres;
 751
 752     wchar_t *tmpbuf = 0;
 753
 754     if (ms_wcNeedsSwap)
 755     {
 756         // need to copy to temp buffer to switch endianness
 757         // this absolutely doesn't rock!
 758         // (no, doing WC_BSWAP twice on the original buffer won't help, as it
 759         //  could be in read-only memory, or be accessed in some other thread)
 760         tmpbuf=(wchar_t*)malloc((inbuf+1)*SIZEOF_WCHAR_T);
 761         memcpy(tmpbuf,psz,(inbuf+1)*SIZEOF_WCHAR_T);
 762         WC_BSWAP(tmpbuf, inbuf)
 763         psz=tmpbuf;
 764     }
 765
 766     if (buf)
 767     {
 768         // have destination buffer, convert there
 769         cres = iconv( w2m, ICONV_CHAR_CAST(&psz), &inbuf, &buf, &outbuf );
 770
 771         res = n-outbuf;
 772     }
 773     else
 774     {
 775         // no destination buffer... convert using temp buffer
 776         // to calculate destination buffer requirement
 777         char tbuf[16];
 778         res = 0;
 779         do {
 780             buf = tbuf; outbuf = 16;
 781
 782             cres = iconv( w2m, ICONV_CHAR_CAST(&psz), &inbuf, &buf, &outbuf );
 783
 784             res += 16 - outbuf;
 785         } while ((cres==(size_t)-1) && (errno==E2BIG));
 786     }
 787
 788     if (ms_wcNeedsSwap)
 789     {
 790         free(tmpbuf);
 791     }
 792
 793     if (ICONV_FAILED(cres, inbuf))
 794     {
 795         //VS: it is ok if iconv fails, hence trace only
 796         wxLogTrace(wxT("strconv"), wxT("iconv failed: %s"), wxSysErrorMsg(wxSysErrorCode()));
 797         return (size_t)-1;
 798     }
 799
 800     return res;
 801 }
 802
 803 #endif // HAVE_ICONV
 804
 805 // ============================================================================
 806 // Win32 conversion classes
 807 // ============================================================================
 808
 809 #if defined(__WIN32__) && !defined(__WXMICROWIN__)
 810 class CP_CharSet : public wxCharacterSet
 811 {
 812 public:
 813     CP_CharSet(const wxChar* name)
 814         : wxCharacterSet(name)
 815         {
 816             m_CodePage = wxCharsetToCodepage(name);
 817         }
 818
 819     size_t MB2WC(wchar_t *buf, const char *psz, size_t n)
 820     {
 821         size_t len =
 822             MultiByteToWideChar(m_CodePage, 0, psz, -1, buf, buf ? n : 0);
 823         //VS: returns # of written chars for buf!=NULL and *size*
 824         //    needed buffer for buf==NULL
 825         return len ? (buf ? len : len-1) : (size_t)-1;
 826     }
 827
 828     size_t WC2MB(char *buf, const wchar_t *psz, size_t n)
 829     {
 830         size_t len = WideCharToMultiByte(m_CodePage, 0, psz, -1, buf,
 831                                          buf ? n : 0, NULL, NULL);
 832         //VS: returns # of written chars for buf!=NULL and *size*
 833         //    needed buffer for buf==NULL
 834         return len ? (buf ? len : len-1) : (size_t)-1;
 835     }
 836
 837     bool usable() const
 838         { return m_CodePage != -1; }
 839
 840 public:
 841     long m_CodePage;
 842 };
 843 #endif // __WIN32__
 844
 845 // ============================================================================
 846 // wxEncodingConverter based conversion classes
 847 // ============================================================================
 848
 849 #if wxUSE_FONTMAP
 850
 851 class EC_CharSet : public wxCharacterSet
 852 {
 853 public:
 854     // temporarily just use wxEncodingConverter stuff,
 855     // so that it works while a better implementation is built
 856     EC_CharSet(const wxChar* name) : wxCharacterSet(name),
 857                                      enc(wxFONTENCODING_SYSTEM)
 858     {
 859         if (name)
 860             enc = wxTheFontMapper->CharsetToEncoding(name, FALSE);
 861
 862         m_ok = m2w.Init(enc, wxFONTENCODING_UNICODE) &&
 863                w2m.Init(wxFONTENCODING_UNICODE, enc);
 864     }
 865
 866     size_t MB2WC(wchar_t *buf, const char *psz, size_t n)
 867     {
 868         size_t inbuf = strlen(psz);
 869         if (buf)
 870             m2w.Convert(psz,buf);
 871         return inbuf;
 872     }
 873
 874     size_t WC2MB(char *buf, const wchar_t *psz, size_t n)
 875     {
 876 #if ( defined(__BORLANDC__) && (__BORLANDC__ > 0x530) ) \
 877     || ( defined(__MWERKS__) && defined(__WXMSW__) )
 878         size_t inbuf = std::wcslen(psz);
 879 #else
 880         size_t inbuf = ::wcslen(psz);
 881 #endif
 882         if (buf)
 883             w2m.Convert(psz,buf);
 884
 885         return inbuf;
 886     }
 887
 888     bool usable() const { return m_ok; }
 889
 890 public:
 891     wxFontEncoding enc;
 892     wxEncodingConverter m2w, w2m;
 893
 894     // were we initialized successfully?
 895     bool m_ok;
 896 };
 897
 898 #endif // wxUSE_FONTMAP
 899
 900 // ----------------------------------------------------------------------------
 901 // the function creating the wxCharacterSet for the specified charset on the
 902 // current system, trying all possibilities
 903 // ----------------------------------------------------------------------------
 904
 905 static wxCharacterSet *wxGetCharacterSet(const wxChar *name)
 906 {
 907     // check for the special case of ASCII charset
 908 #if wxUSE_FONTMAP
 909     if ( wxTheFontMapper->CharsetToEncoding(name) == wxFONTENCODING_DEFAULT )
 910 #else // wxUSE_FONTMAP
 911     if ( !name )
 912 #endif // wxUSE_FONTMAP/!wxUSE_FONTMAP
 913     {
 914         // don't convert at all
 915         return NULL;
 916     }
 917
 918     // the test above must have taken care of this case
 919     wxCHECK_MSG( name, NULL, _T("NULL name must be wxFONTENCODING_DEFAULT") );
 920
 921     wxCharacterSet *cset;
 922
 923     if ( wxStricmp(name, wxT("UTF8")) == 0 || wxStricmp(name, wxT("UTF-8")) == 0)
 924     {
 925         cset = new ID_CharSet(name, &wxConvUTF8);
 926     }
 927     else
 928     {
 929 #ifdef HAVE_ICONV
 930         cset = new IC_CharSet(name);
 931 #else // !HAVE_ICONV
 932         cset = NULL;
 933 #endif // HAVE_ICONV/!HAVE_ICONV
 934     }
 935
 936     // it can only be NULL in this case
 937 #ifndef HAVE_ICONV
 938     if ( cset )
 939 #endif // !HAVE_ICONV
 940     {
 941         if ( cset->usable() )
 942             return cset;
 943
 944         delete cset;
 945         cset = NULL;
 946     }
 947
 948 #if defined(__WIN32__) && !defined(__WXMICROWIN__)
 949     cset = new CP_CharSet(name);
 950     if ( cset->usable() )
 951         return cset;
 952
 953     delete cset;
 954     cset = NULL;
 955 #endif // __WIN32__
 956
 957 #if wxUSE_FONTMAP
 958     cset = new EC_CharSet(name);
 959     if ( cset->usable() )
 960         return cset;
 961
 962     delete cset;
 963     cset = NULL;
 964 #endif // wxUSE_FONTMAP
 965
 966     wxLogError(_("Cannot convert from encoding '%s'!"), name);
 967
 968     return NULL;
 969 }
 970
 971 // ============================================================================
 972 // wxCSConv implementation
 973 // ============================================================================
 974
 975 wxCSConv::wxCSConv(const wxChar *charset)
 976 {
 977     m_name = (wxChar *)NULL;
 978     m_cset = (wxCharacterSet *) NULL;
 979     m_deferred = TRUE;
 980
 981     SetName(charset);
 982 }
 983
 984 wxCSConv::~wxCSConv()
 985 {
 986     free(m_name);
 987     delete m_cset;
 988 }
 989
 990 void wxCSConv::SetName(const wxChar *charset)
 991 {
 992     if (charset)
 993     {
 994         m_name = wxStrdup(charset);
 995         m_deferred = TRUE;
 996     }
 997 }
 998
 999 void wxCSConv::LoadNow()
1000 {
1001     if (m_deferred)
1002     {
1003         if ( !m_name )
1004         {
1005             wxString name = wxLocale::GetSystemEncodingName();
1006             if ( !name.empty() )
1007                 SetName(name);
1008         }
1009
1010         // wxGetCharacterSet() complains about NULL name
1011         m_cset = m_name ? wxGetCharacterSet(m_name) : NULL;
1012         m_deferred = FALSE;
1013     }
1014 }
1015
1016 size_t wxCSConv::MB2WC(wchar_t *buf, const char *psz, size_t n) const
1017 {
1018     ((wxCSConv *)this)->LoadNow(); // discard constness
1019
1020     if (m_cset)
1021         return m_cset->MB2WC(buf, psz, n);
1022
1023     // latin-1 (direct)
1024     size_t len = strlen(psz);
1025
1026     if (buf)
1027     {
1028         for (size_t c = 0; c <= len; c++)
1029             buf[c] = (unsigned char)(psz[c]);
1030     }
1031
1032     return len;
1033 }
1034
1035 size_t wxCSConv::WC2MB(char *buf, const wchar_t *psz, size_t n) const
1036 {
1037     ((wxCSConv *)this)->LoadNow(); // discard constness
1038
1039     if (m_cset)
1040         return m_cset->WC2MB(buf, psz, n);
1041
1042     // latin-1 (direct)
1043 #if ( defined(__BORLANDC__) && (__BORLANDC__ > 0x530) ) \
1044     || ( defined(__MWERKS__) && defined(__WXMSW__) )
1045     size_t len=std::wcslen(psz);
1046 #else
1047     size_t len=::wcslen(psz);
1048 #endif
1049     if (buf)
1050     {
1051         for (size_t c = 0; c <= len; c++)
1052             buf[c] = (psz[c] > 0xff) ? '?' : psz[c];
1053     }
1054
1055     return len;
1056 }
1057
1058 #else // !wxUSE_WCHAR_T
1059
1060 // ----------------------------------------------------------------------------
1061 // stand-ins in absence of wchar_t
1062 // ----------------------------------------------------------------------------
1063
1064 WXDLLEXPORT_DATA(wxMBConv) wxConvLibc, wxConvFile;
1065
1066 #endif // wxUSE_WCHAR_T
1067
1068