src/common/strconv.cpp

   1 /////////////////////////////////////////////////////////////////////////////
   2 // Name:        strconv.cpp
   3 // Purpose:     Unicode conversion classes
   4 // Author:      Ove Kaaven, Robert Roebling, Vadim Zeitlin, Vaclav Slavik
   5 // Modified by:
   6 // Created:     29/01/98
   7 // RCS-ID:      $Id$
   8 // Copyright:   (c) 1999 Ove Kaaven, Robert Roebling, Vadim Zeitlin, Vaclav Slavik
   9 // Licence:     wxWindows license
  10 /////////////////////////////////////////////////////////////////////////////
  11
  12 // ============================================================================
  13 // declarations
  14 // ============================================================================
  15
  16 // ----------------------------------------------------------------------------
  17 // headers
  18 // ----------------------------------------------------------------------------
  19
  20 #ifdef __GNUG__
  21   #pragma implementation "strconv.h"
  22 #endif
  23
  24 // For compilers that support precompilation, includes "wx.h".
  25 #include "wx/wxprec.h"
  26
  27 #ifdef __BORLANDC__
  28   #pragma hdrstop
  29 #endif
  30
  31 #ifndef WX_PRECOMP
  32     #include "wx/intl.h"
  33     #include "wx/log.h"
  34 #endif // WX_PRECOMP
  35
  36 #ifdef __WXMSW__
  37     #include "wx/msw/private.h"
  38 #endif
  39
  40 #include <errno.h>
  41 #include <ctype.h>
  42 #include <string.h>
  43 #include <stdlib.h>
  44
  45 #include "wx/module.h"
  46 #include "wx/strconv.h"
  47
  48 // ----------------------------------------------------------------------------
  49 // globals
  50 // ----------------------------------------------------------------------------
  51
  52 #if wxUSE_WCHAR_T
  53     WXDLLEXPORT_DATA(wxMBConv) wxConvLibc;
  54     WXDLLEXPORT_DATA(wxCSConv) wxConvLocal((const wxChar *)NULL);
  55 #else
  56     // stand-ins in absence of wchar_t
  57     WXDLLEXPORT_DATA(wxMBConv) wxConvLibc, wxConvFile, wxConvLocal;
  58 #endif // wxUSE_WCHAR_T
  59
  60 WXDLLEXPORT_DATA(wxMBConv *) wxConvCurrent = &wxConvLibc;
  61
  62 class wxStrConvModule: public wxModule
  63 {
  64 public:
  65     wxStrConvModule() : wxModule() { }
  66     virtual bool OnInit() { return TRUE; }
  67     virtual void OnExit()
  68     {
  69 #if wxUSE_WCHAR_T
  70          wxConvLocal.Clear();
  71 #endif
  72     }
  73
  74     DECLARE_DYNAMIC_CLASS(wxStrConvModule)
  75 };
  76
  77 IMPLEMENT_DYNAMIC_CLASS(wxStrConvModule, wxModule)
  78
  79
  80 // ----------------------------------------------------------------------------
  81 // headers
  82 // ----------------------------------------------------------------------------
  83
  84 #if wxUSE_WCHAR_T
  85
  86 #ifdef __SALFORDC__
  87     #include <clib.h>
  88 #endif
  89
  90 #ifdef HAVE_ICONV
  91     #include <iconv.h>
  92 #endif
  93
  94 #include "wx/encconv.h"
  95 #include "wx/fontmap.h"
  96
  97 // ----------------------------------------------------------------------------
  98 // macros
  99 // ----------------------------------------------------------------------------
 100
 101 #define BSWAP_UCS4(str, len) { unsigned _c; for (_c=0; _c<len; _c++) str[_c]=wxUINT32_SWAP_ALWAYS(str[_c]); }
 102 #define BSWAP_UTF16(str, len) { unsigned _c; for (_c=0; _c<len; _c++) str[_c]=wxUINT16_SWAP_ALWAYS(str[_c]); }
 103
 104 // under Unix SIZEOF_WCHAR_T is defined by configure, but under other platforms
 105 // it might be not defined - assume the most common value
 106 #ifndef SIZEOF_WCHAR_T
 107     #define SIZEOF_WCHAR_T 2
 108 #endif // !defined(SIZEOF_WCHAR_T)
 109
 110 #if SIZEOF_WCHAR_T == 4
 111     #define WC_NAME         "UCS4"
 112     #define WC_BSWAP         BSWAP_UCS4
 113     #ifdef WORDS_BIGENDIAN
 114       #define WC_NAME_BEST  "UCS-4BE"
 115     #else
 116       #define WC_NAME_BEST  "UCS-4LE"
 117     #endif
 118 #elif SIZEOF_WCHAR_T == 2
 119     #define WC_NAME         "UTF16"
 120     #define WC_BSWAP         BSWAP_UTF16
 121     #define WC_UTF16
 122     #ifdef WORDS_BIGENDIAN
 123       #define WC_NAME_BEST  "UTF-16BE"
 124     #else
 125       #define WC_NAME_BEST  "UTF-16LE"
 126     #endif
 127 #else // sizeof(wchar_t) != 2 nor 4
 128     // I don't know what to do about this
 129     #error "Weird sizeof(wchar_t): please report your platform details to wx-users mailing list"
 130 #endif
 131
 132 // ============================================================================
 133 // implementation
 134 // ============================================================================
 135
 136 // ----------------------------------------------------------------------------
 137 // UTF-16 en/decoding
 138 // ----------------------------------------------------------------------------
 139
 140 #ifdef WC_UTF16
 141
 142 static size_t encode_utf16(wxUint32 input, wchar_t *output)
 143 {
 144     if (input<=0xffff)
 145     {
 146         if (output) *output++ = (wchar_t) input;
 147         return 1;
 148     }
 149     else if (input>=0x110000)
 150     {
 151         return (size_t)-1;
 152     }
 153     else
 154     {
 155         if (output)
 156         {
 157             *output++ = (wchar_t) ((input >> 10)+0xd7c0);
 158             *output++ = (wchar_t) ((input&0x3ff)+0xdc00);
 159         }
 160         return 2;
 161     }
 162 }
 163
 164 static size_t decode_utf16(const wchar_t* input, wxUint32& output)
 165 {
 166     if ((*input<0xd800) || (*input>0xdfff))
 167     {
 168         output = *input;
 169         return 1;
 170     }
 171     else if ((input[1]<0xdc00) || (input[1]>=0xdfff))
 172     {
 173         output = *input;
 174         return (size_t)-1;
 175     }
 176     else
 177     {
 178         output = ((input[0] - 0xd7c0) << 10) + (input[1] - 0xdc00);
 179         return 2;
 180     }
 181 }
 182
 183 #endif // WC_UTF16
 184
 185 // ----------------------------------------------------------------------------
 186 // wxMBConv
 187 // ----------------------------------------------------------------------------
 188
 189 #define IGNORE_LIBC 0
 190
 191 size_t wxMBConv::MB2WC(wchar_t *buf, const char *psz, size_t n) const
 192 {
 193 #if IGNORE_LIBC
 194     if (buf)
 195     {
 196         for (size_t i = 0; i < strlen( psz )+1; i++)
 197             buf[i] = (wchar_t) psz[i];
 198         return strlen( psz );
 199     }
 200     else
 201     {
 202         return strlen( psz );
 203     }
 204 #else
 205     return wxMB2WC(buf, psz, n);
 206 #endif
 207 }
 208
 209 size_t wxMBConv::WC2MB(char *buf, const wchar_t *psz, size_t n) const
 210 {
 211 #if IGNORE_LIBC
 212     if (buf)
 213     {
 214         for (size_t i = 0; i < wxStrlen( psz )+1; i++)
 215             buf[i] = (char) psz[i];
 216         return wxStrlen( psz );
 217     }
 218     else
 219     {
 220         return wxStrlen( psz );
 221     }
 222 #else
 223     return wxWC2MB(buf, psz, n);
 224 #endif
 225 }
 226
 227 const wxWCharBuffer wxMBConv::cMB2WC(const char *psz) const
 228 {
 229     if (psz)
 230     {
 231         size_t nLen = MB2WC((wchar_t *) NULL, psz, 0);  // return value excludes /0
 232         if (nLen == (size_t)-1)
 233             return wxWCharBuffer((wchar_t *) NULL);
 234         wxWCharBuffer buf(nLen);                        // this allocates nLen1+
 235         MB2WC((wchar_t *)(const wchar_t *) buf, psz, nLen+1);
 236         return buf;
 237     }
 238     else
 239         return wxWCharBuffer((wchar_t *) NULL);
 240 }
 241
 242 const wxCharBuffer wxMBConv::cWC2MB(const wchar_t *pwz) const
 243 {
 244     // return value excludes NUL
 245     size_t nLen = pwz ? WC2MB((char *) NULL, pwz, 0) : (size_t)-1;
 246     if (nLen == (size_t)-1)
 247         return wxCharBuffer((const char *)NULL);
 248
 249     wxCharBuffer buf(nLen);                      // this allocates nLen+1
 250     WC2MB((char *)(const char *) buf, pwz, nLen+1);
 251     return buf;
 252 }
 253
 254 // ----------------------------------------------------------------------------
 255 // standard gdk conversion
 256 // ----------------------------------------------------------------------------
 257
 258 #ifdef __WXGTK12__
 259
 260 WXDLLEXPORT_DATA(wxMBConvGdk) wxConvGdk;
 261
 262 #include <gdk/gdk.h>
 263
 264 size_t wxMBConvGdk::MB2WC(wchar_t *buf, const char *psz, size_t n) const
 265 {
 266     if (buf)
 267     {
 268         return gdk_mbstowcs((GdkWChar *)buf, psz, n);
 269     }
 270     else
 271     {
 272         GdkWChar *nbuf = new GdkWChar[n=strlen(psz)];
 273         size_t len = gdk_mbstowcs(nbuf, psz, n);
 274         delete[] nbuf;
 275         return len;
 276     }
 277 }
 278
 279 size_t wxMBConvGdk::WC2MB(char *buf, const wchar_t *psz, size_t n) const
 280 {
 281     char *mbstr = gdk_wcstombs((GdkWChar *)psz);
 282     size_t len = mbstr ? strlen(mbstr) : 0;
 283     if (buf)
 284     {
 285         if (len > n)
 286             len = n;
 287         memcpy(buf, psz, len);
 288         if (len < n)
 289             buf[len] = 0;
 290     }
 291     return len;
 292 }
 293
 294 #endif // GTK > 1.0
 295
 296 // ----------------------------------------------------------------------------
 297 // UTF-7
 298 // ----------------------------------------------------------------------------
 299
 300 WXDLLEXPORT_DATA(wxMBConvUTF7) wxConvUTF7;
 301
 302 #if 0
 303 static char utf7_setD[]="ABCDEFGHIJKLMNOPQRSTUVWXYZ"
 304                         "abcdefghijklmnopqrstuvwxyz"
 305                         "0123456789'(),-./:?";
 306 static char utf7_setO[]="!\"#$%&*;<=>@[]^_`{|}";
 307 static char utf7_setB[]="ABCDEFGHIJKLMNOPQRSTUVWXYZ"
 308                         "abcdefghijklmnopqrstuvwxyz"
 309                         "0123456789+/";
 310 #endif
 311
 312 // TODO: write actual implementations of UTF-7 here
 313 size_t wxMBConvUTF7::MB2WC(wchar_t * WXUNUSED(buf),
 314                            const char * WXUNUSED(psz),
 315                            size_t WXUNUSED(n)) const
 316 {
 317   return 0;
 318 }
 319
 320 size_t wxMBConvUTF7::WC2MB(char * WXUNUSED(buf),
 321                            const wchar_t * WXUNUSED(psz),
 322                            size_t WXUNUSED(n)) const
 323 {
 324   return 0;
 325 }
 326
 327 // ----------------------------------------------------------------------------
 328 // UTF-8
 329 // ----------------------------------------------------------------------------
 330
 331 WXDLLEXPORT_DATA(wxMBConvUTF8) wxConvUTF8;
 332
 333 static wxUint32 utf8_max[]=
 334     { 0x7f, 0x7ff, 0xffff, 0x1fffff, 0x3ffffff, 0x7fffffff, 0xffffffff };
 335
 336 size_t wxMBConvUTF8::MB2WC(wchar_t *buf, const char *psz, size_t n) const
 337 {
 338     size_t len = 0;
 339
 340     while (*psz && ((!buf) || (len < n)))
 341     {
 342         unsigned char cc = *psz++, fc = cc;
 343         unsigned cnt;
 344         for (cnt = 0; fc & 0x80; cnt++)
 345             fc <<= 1;
 346         if (!cnt)
 347         {
 348             // plain ASCII char
 349             if (buf)
 350                 *buf++ = cc;
 351             len++;
 352         }
 353         else
 354         {
 355             cnt--;
 356             if (!cnt)
 357             {
 358                 // invalid UTF-8 sequence
 359                 return (size_t)-1;
 360             }
 361             else
 362             {
 363                 unsigned ocnt = cnt - 1;
 364                 wxUint32 res = cc & (0x3f >> cnt);
 365                 while (cnt--)
 366                 {
 367                     cc = *psz++;
 368                     if ((cc & 0xC0) != 0x80)
 369                     {
 370                         // invalid UTF-8 sequence
 371                         return (size_t)-1;
 372                     }
 373                     res = (res << 6) | (cc & 0x3f);
 374                 }
 375                 if (res <= utf8_max[ocnt])
 376                 {
 377                     // illegal UTF-8 encoding
 378                     return (size_t)-1;
 379                 }
 380 #ifdef WC_UTF16
 381                 size_t pa = encode_utf16(res, buf);
 382                 if (pa == (size_t)-1)
 383                   return (size_t)-1;
 384                 if (buf)
 385                     buf += pa;
 386                 len += pa;
 387 #else // !WC_UTF16
 388                 if (buf)
 389                     *buf++ = res;
 390                 len++;
 391 #endif // WC_UTF16/!WC_UTF16
 392             }
 393         }
 394     }
 395     if (buf && (len < n))
 396         *buf = 0;
 397     return len;
 398 }
 399
 400 size_t wxMBConvUTF8::WC2MB(char *buf, const wchar_t *psz, size_t n) const
 401 {
 402     size_t len = 0;
 403
 404     while (*psz && ((!buf) || (len < n)))
 405     {
 406         wxUint32 cc;
 407 #ifdef WC_UTF16
 408         size_t pa = decode_utf16(psz, cc);
 409         psz += (pa == (size_t)-1) ? 1 : pa;
 410 #else
 411         cc=(*psz++) & 0x7fffffff;
 412 #endif
 413         unsigned cnt;
 414         for (cnt = 0; cc > utf8_max[cnt]; cnt++) {}
 415         if (!cnt)
 416         {
 417             // plain ASCII char
 418             if (buf)
 419                 *buf++ = (char) cc;
 420             len++;
 421         }
 422
 423         else
 424         {
 425             len += cnt + 1;
 426             if (buf)
 427             {
 428                 *buf++ = (char) ((-128 >> cnt) | ((cc >> (cnt * 6)) & (0x3f >> cnt)));
 429                 while (cnt--)
 430                     *buf++ = (char) (0x80 | ((cc >> (cnt * 6)) & 0x3f));
 431             }
 432         }
 433     }
 434
 435     if (buf && (len<n)) *buf = 0;
 436
 437     return len;
 438 }
 439
 440 // ============================================================================
 441 // wxCharacterSet and derived classes
 442 // ============================================================================
 443
 444 // ----------------------------------------------------------------------------
 445 // wxCharacterSet is the ABC for the classes below
 446 // ----------------------------------------------------------------------------
 447
 448 class wxCharacterSet
 449 {
 450 public:
 451     wxCharacterSet(const wxChar*name) : cname(name) {}
 452     virtual ~wxCharacterSet() {}
 453     virtual size_t MB2WC(wchar_t *buf, const char *psz, size_t n) = 0;
 454     virtual size_t WC2MB(char *buf, const wchar_t *psz, size_t n) = 0;
 455     virtual bool usable() const = 0;
 456 public:
 457     const wxChar*cname;
 458 };
 459
 460 // ----------------------------------------------------------------------------
 461 // ID_CharSet: implementation of wxCharacterSet using an existing wxMBConv
 462 // ----------------------------------------------------------------------------
 463
 464 class ID_CharSet : public wxCharacterSet
 465 {
 466 public:
 467     ID_CharSet(const wxChar *name, wxMBConv *cnv)
 468         : wxCharacterSet(name), work(cnv) {}
 469
 470     size_t MB2WC(wchar_t *buf, const char *psz, size_t n)
 471         { return work ? work->MB2WC(buf,psz,n) : (size_t)-1; }
 472
 473     size_t WC2MB(char *buf, const wchar_t *psz, size_t n)
 474         { return work ? work->WC2MB(buf,psz,n) : (size_t)-1; }
 475
 476     bool usable() const
 477         { return work!=NULL; }
 478 public:
 479     wxMBConv*work;
 480 };
 481
 482
 483 // ============================================================================
 484 // The classes doing conversion using the iconv_xxx() functions
 485 // ============================================================================
 486
 487 #ifdef HAVE_ICONV
 488
 489 // VS: glibc 2.1.3 is broken in that iconv() conversion to/from UCS4 fails with E2BIG
 490 //     if output buffer is _exactly_ as big as needed. Such case is (unless there's
 491 //     yet another bug in glibc) the only case when iconv() returns with (size_t)-1
 492 //     (which means error) and says there are 0 bytes left in the input buffer --
 493 //     when _real_ error occurs, bytes-left-in-input buffer is non-zero. Hence,
 494 //     this alternative test for iconv() failure.
 495 //     [This bug does not appear in glibc 2.2.]
 496 #if defined(__GLIBC__) && __GLIBC__ == 2 && __GLIBC_MINOR__ <= 1
 497 #define ICONV_FAILED(cres, bufLeft) ((cres == (size_t)-1) && \
 498                                      (errno != E2BIG || bufLeft != 0))
 499 #else
 500 #define ICONV_FAILED(cres, bufLeft)  (cres == (size_t)-1)
 501 #endif
 502
 503 #define ICONV_CHAR_CAST(x)  ((ICONV_CONST char **)(x))
 504
 505 // ----------------------------------------------------------------------------
 506 // IC_CharSet: encapsulates an iconv character set
 507 // ----------------------------------------------------------------------------
 508
 509 class IC_CharSet : public wxCharacterSet
 510 {
 511 public:
 512     IC_CharSet(const wxChar *name);
 513     virtual ~IC_CharSet();
 514
 515     virtual size_t MB2WC(wchar_t *buf, const char *psz, size_t n);
 516     virtual size_t WC2MB(char *buf, const wchar_t *psz, size_t n);
 517
 518     bool usable() const
 519         { return (m2w != (iconv_t)-1) && (w2m != (iconv_t)-1); }
 520
 521 protected:
 522     // the iconv handlers used to translate from multibyte to wide char and in
 523     // the other direction
 524     iconv_t m2w,
 525             w2m;
 526
 527 private:
 528     // the name (for iconv_open()) of a wide char charset - if none is
 529     // available on this machine, it will remain NULL
 530     static const char *ms_wcCharsetName;
 531
 532     // true if the wide char encoding we use (i.e. ms_wcCharsetName) has
 533     // different endian-ness than the native one
 534     static bool ms_wcNeedsSwap;
 535 };
 536
 537 const char *IC_CharSet::ms_wcCharsetName = NULL;
 538 bool IC_CharSet::ms_wcNeedsSwap = FALSE;
 539
 540 IC_CharSet::IC_CharSet(const wxChar *name)
 541           : wxCharacterSet(name)
 542 {
 543     // Do it the hard way
 544     char cname[100];
 545     for (size_t i = 0; i < wxStrlen(name)+1; i++)
 546         cname[i] = (char) name[i];
 547
 548     // check for charset that represents wchar_t:
 549     if (ms_wcCharsetName == NULL)
 550     {
 551         ms_wcNeedsSwap = FALSE;
 552
 553         // try charset with explicit bytesex info (e.g. "UCS-4LE"):
 554         ms_wcCharsetName = WC_NAME_BEST;
 555         m2w = iconv_open(ms_wcCharsetName, cname);
 556
 557         if (m2w == (iconv_t)-1)
 558         {
 559             // try charset w/o bytesex info (e.g. "UCS4")
 560             // and check for bytesex ourselves:
 561             ms_wcCharsetName = WC_NAME;
 562             m2w = iconv_open(ms_wcCharsetName, cname);
 563
 564             // last bet, try if it knows WCHAR_T pseudo-charset
 565             if (m2w == (iconv_t)-1)
 566             {
 567                 ms_wcCharsetName = "WCHAR_T";
 568                 m2w = iconv_open(ms_wcCharsetName, cname);
 569             }
 570
 571             if (m2w != (iconv_t)-1)
 572             {
 573                 char    buf[2], *bufPtr;
 574                 wchar_t wbuf[2], *wbufPtr;
 575                 size_t  insz, outsz;
 576                 size_t  res;
 577
 578                 buf[0] = 'A';
 579                 buf[1] = 0;
 580                 wbuf[0] = 0;
 581                 insz = 2;
 582                 outsz = SIZEOF_WCHAR_T * 2;
 583                 wbufPtr = wbuf;
 584                 bufPtr = buf;
 585
 586                 res = iconv(m2w, ICONV_CHAR_CAST(&bufPtr), &insz,
 587                             (char**)&wbufPtr, &outsz);
 588
 589                 if (ICONV_FAILED(res, insz))
 590                 {
 591                     ms_wcCharsetName = NULL;
 592                     wxLogLastError(wxT("iconv"));
 593                     wxLogError(_("Convertion to charset '%s' doesn't work."), name);
 594                 }
 595                 else
 596                 {
 597                     ms_wcNeedsSwap = wbuf[0] != (wchar_t)buf[0];
 598                 }
 599             }
 600             else
 601             {
 602                 ms_wcCharsetName = NULL;
 603
 604                 // VS: we must not output an error here, since wxWindows will safely
 605                 //     fall back to using wxEncodingConverter.
 606                 wxLogTrace(wxT("strconv"), wxT("Impossible to convert to/from charset '%s' with iconv, falling back to wxEncodingConverter."), name);
 607                 //wxLogError(
 608             }
 609         }
 610         wxLogTrace(wxT("strconv"), wxT("wchar_t charset is '%s', needs swap: %i"), ms_wcCharsetName, ms_wcNeedsSwap);
 611     }
 612     else // we already have ms_wcCharsetName
 613     {
 614         m2w = iconv_open(ms_wcCharsetName, cname);
 615     }
 616
 617     // NB: don't ever pass NULL to iconv_open(), it may crash!
 618     if ( ms_wcCharsetName )
 619     {
 620         w2m = iconv_open( cname, ms_wcCharsetName);
 621     }
 622     else
 623     {
 624         w2m = (iconv_t)-1;
 625     }
 626 }
 627
 628 IC_CharSet::~IC_CharSet()
 629 {
 630     if ( m2w != (iconv_t)-1 )
 631         iconv_close(m2w);
 632     if ( w2m != (iconv_t)-1 )
 633         iconv_close(w2m);
 634 }
 635
 636 size_t IC_CharSet::MB2WC(wchar_t *buf, const char *psz, size_t n)
 637 {
 638     size_t inbuf = strlen(psz);
 639     size_t outbuf = n * SIZEOF_WCHAR_T;
 640     size_t res, cres;
 641     // VS: Use these instead of psz, buf because iconv() modifies its arguments:
 642     wchar_t *bufPtr = buf;
 643     const char *pszPtr = psz;
 644
 645     if (buf)
 646     {
 647         // have destination buffer, convert there
 648         cres = iconv(m2w,
 649                      ICONV_CHAR_CAST(&pszPtr), &inbuf,
 650                      (char**)&bufPtr, &outbuf);
 651         res = n - (outbuf / SIZEOF_WCHAR_T);
 652
 653         if (ms_wcNeedsSwap)
 654         {
 655             // convert to native endianness
 656             WC_BSWAP(buf /* _not_ bufPtr */, res)
 657         }
 658
 659         // NB: iconv was given only strlen(psz) characters on input, and so
 660         //     it couldn't convert the trailing zero. Let's do it ourselves
 661         //     if there's some room left for it in the output buffer.
 662         if (res < n)
 663             buf[res] = 0;
 664     }
 665     else
 666     {
 667         // no destination buffer... convert using temp buffer
 668         // to calculate destination buffer requirement
 669         wchar_t tbuf[8];
 670         res = 0;
 671         do {
 672             bufPtr = tbuf;
 673             outbuf = 8*SIZEOF_WCHAR_T;
 674
 675             cres = iconv(m2w,
 676                          ICONV_CHAR_CAST(&pszPtr), &inbuf,
 677                          (char**)&bufPtr, &outbuf );
 678
 679             res += 8-(outbuf/SIZEOF_WCHAR_T);
 680         } while ((cres==(size_t)-1) && (errno==E2BIG));
 681     }
 682
 683     if (ICONV_FAILED(cres, inbuf))
 684     {
 685         //VS: it is ok if iconv fails, hence trace only
 686         wxLogTrace(wxT("strconv"), wxT("iconv failed: %s"), wxSysErrorMsg(wxSysErrorCode()));
 687         return (size_t)-1;
 688     }
 689
 690     return res;
 691 }
 692
 693 size_t IC_CharSet::WC2MB(char *buf, const wchar_t *psz, size_t n)
 694 {
 695 #if defined(__BORLANDC__) && (__BORLANDC__ > 0x530)
 696     size_t inbuf = std::wcslen(psz) * SIZEOF_WCHAR_T;
 697 #else
 698     size_t inbuf = ::wcslen(psz) * SIZEOF_WCHAR_T;
 699 #endif
 700     size_t outbuf = n;
 701     size_t res, cres;
 702
 703     wchar_t *tmpbuf = 0;
 704
 705     if (ms_wcNeedsSwap)
 706     {
 707         // need to copy to temp buffer to switch endianness
 708         // this absolutely doesn't rock!
 709         // (no, doing WC_BSWAP twice on the original buffer won't help, as it
 710         //  could be in read-only memory, or be accessed in some other thread)
 711         tmpbuf=(wchar_t*)malloc((inbuf+1)*SIZEOF_WCHAR_T);
 712         memcpy(tmpbuf,psz,(inbuf+1)*SIZEOF_WCHAR_T);
 713         WC_BSWAP(tmpbuf, inbuf)
 714         psz=tmpbuf;
 715     }
 716
 717     if (buf)
 718     {
 719         // have destination buffer, convert there
 720         cres = iconv( w2m, ICONV_CHAR_CAST(&psz), &inbuf, &buf, &outbuf );
 721
 722         res = n-outbuf;
 723
 724         // NB: iconv was given only wcslen(psz) characters on input, and so
 725         //     it couldn't convert the trailing zero. Let's do it ourselves
 726         //     if there's some room left for it in the output buffer.
 727         if (res < n)
 728             buf[0] = 0;
 729     }
 730     else
 731     {
 732         // no destination buffer... convert using temp buffer
 733         // to calculate destination buffer requirement
 734         char tbuf[16];
 735         res = 0;
 736         do {
 737             buf = tbuf; outbuf = 16;
 738
 739             cres = iconv( w2m, ICONV_CHAR_CAST(&psz), &inbuf, &buf, &outbuf );
 740
 741             res += 16 - outbuf;
 742         } while ((cres==(size_t)-1) && (errno==E2BIG));
 743     }
 744
 745     if (ms_wcNeedsSwap)
 746     {
 747         free(tmpbuf);
 748     }
 749
 750     if (ICONV_FAILED(cres, inbuf))
 751     {
 752         //VS: it is ok if iconv fails, hence trace only
 753         wxLogTrace(wxT("strconv"), wxT("iconv failed: %s"), wxSysErrorMsg(wxSysErrorCode()));
 754         return (size_t)-1;
 755     }
 756
 757     return res;
 758 }
 759
 760 #endif // HAVE_ICONV
 761
 762 // ============================================================================
 763 // Win32 conversion classes
 764 // ============================================================================
 765
 766 #if defined(__WIN32__) && !defined(__WXMICROWIN__)
 767
 768 extern long wxCharsetToCodepage(const wxChar *charset); // from utils.cpp
 769
 770 class CP_CharSet : public wxCharacterSet
 771 {
 772 public:
 773     CP_CharSet(const wxChar* name)
 774         : wxCharacterSet(name)
 775         {
 776             m_CodePage = wxCharsetToCodepage(name);
 777         }
 778
 779     size_t MB2WC(wchar_t *buf, const char *psz, size_t n)
 780     {
 781         size_t len =
 782 #ifdef __WXWINE__
 783             MultiByteToWideChar(m_CodePage, 0, psz, -1, (WCHAR*) buf, buf ? n : 0);
 784 #else
 785             MultiByteToWideChar(m_CodePage, 0, psz, -1, buf, buf ? n : 0);
 786 #endif
 787         //VS: returns # of written chars for buf!=NULL and *size*
 788         //    needed buffer for buf==NULL
 789         return len ? (buf ? len : len-1) : (size_t)-1;
 790     }
 791
 792     size_t WC2MB(char *buf, const wchar_t *psz, size_t n)
 793     {
 794 #ifdef __WXWINE__
 795         size_t len = WideCharToMultiByte(m_CodePage, 0, (const WCHAR*) psz, -1, buf,
 796                                          buf ? n : 0, NULL, NULL);
 797 #else
 798         size_t len = WideCharToMultiByte(m_CodePage, 0, psz, -1, buf,
 799                                          buf ? n : 0, NULL, NULL);
 800 #endif
 801         //VS: returns # of written chars for buf!=NULL and *size*
 802         //    needed buffer for buf==NULL
 803         return len ? (buf ? len : len-1) : (size_t)-1;
 804     }
 805
 806     bool usable() const
 807         { return m_CodePage != -1; }
 808
 809 public:
 810     long m_CodePage;
 811 };
 812 #endif // __WIN32__
 813
 814 // ============================================================================
 815 // wxEncodingConverter based conversion classes
 816 // ============================================================================
 817
 818 #if wxUSE_FONTMAP
 819
 820 class EC_CharSet : public wxCharacterSet
 821 {
 822 public:
 823     // temporarily just use wxEncodingConverter stuff,
 824     // so that it works while a better implementation is built
 825     EC_CharSet(const wxChar* name) : wxCharacterSet(name),
 826                                      enc(wxFONTENCODING_SYSTEM)
 827     {
 828         if (name)
 829             enc = wxFontMapper::Get()->CharsetToEncoding(name, FALSE);
 830
 831         m_ok = m2w.Init(enc, wxFONTENCODING_UNICODE) &&
 832                w2m.Init(wxFONTENCODING_UNICODE, enc);
 833     }
 834
 835     size_t MB2WC(wchar_t *buf, const char *psz, size_t WXUNUSED(n))
 836     {
 837         size_t inbuf = strlen(psz);
 838         if (buf)
 839             m2w.Convert(psz,buf);
 840         return inbuf;
 841     }
 842
 843     size_t WC2MB(char *buf, const wchar_t *psz, size_t WXUNUSED(n))
 844     {
 845 #if ( defined(__BORLANDC__) && (__BORLANDC__ > 0x530) ) \
 846     || ( defined(__MWERKS__) && defined(__WXMSW__) )
 847         size_t inbuf = std::wcslen(psz);
 848 #else
 849         size_t inbuf = ::wcslen(psz);
 850 #endif
 851         if (buf)
 852             w2m.Convert(psz,buf);
 853
 854         return inbuf;
 855     }
 856
 857     bool usable() const { return m_ok; }
 858
 859 public:
 860     wxFontEncoding enc;
 861     wxEncodingConverter m2w, w2m;
 862
 863     // were we initialized successfully?
 864     bool m_ok;
 865 };
 866
 867 #endif // wxUSE_FONTMAP
 868
 869 // ----------------------------------------------------------------------------
 870 // the function creating the wxCharacterSet for the specified charset on the
 871 // current system, trying all possibilities
 872 // ----------------------------------------------------------------------------
 873
 874 static wxCharacterSet *wxGetCharacterSet(const wxChar *name)
 875 {
 876     // check for the special case of ASCII charset
 877 #if wxUSE_FONTMAP
 878     if ( wxFontMapper::Get()->CharsetToEncoding(name) == wxFONTENCODING_DEFAULT )
 879 #else // wxUSE_FONTMAP
 880     if ( !name )
 881 #endif // wxUSE_FONTMAP/!wxUSE_FONTMAP
 882     {
 883         // don't convert at all
 884         return NULL;
 885     }
 886
 887     // the test above must have taken care of this case
 888     wxCHECK_MSG( name, NULL, _T("NULL name must be wxFONTENCODING_DEFAULT") );
 889
 890     wxCharacterSet *cset;
 891
 892     if ( wxStricmp(name, wxT("UTF8")) == 0 || wxStricmp(name, wxT("UTF-8")) == 0)
 893     {
 894         cset = new ID_CharSet(name, &wxConvUTF8);
 895     }
 896     else
 897     {
 898 #ifdef HAVE_ICONV
 899         cset = new IC_CharSet(name);
 900 #else // !HAVE_ICONV
 901         cset = NULL;
 902 #endif // HAVE_ICONV/!HAVE_ICONV
 903     }
 904
 905     // it can only be NULL in this case
 906 #ifndef HAVE_ICONV
 907     if ( cset )
 908 #endif // !HAVE_ICONV
 909     {
 910         if ( cset->usable() )
 911             return cset;
 912
 913         delete cset;
 914         cset = NULL;
 915     }
 916
 917 #if defined(__WIN32__) && !defined(__WXMICROWIN__)
 918     cset = new CP_CharSet(name);
 919     if ( cset->usable() )
 920         return cset;
 921
 922     delete cset;
 923     cset = NULL;
 924 #endif // __WIN32__
 925
 926 #if wxUSE_FONTMAP
 927     cset = new EC_CharSet(name);
 928     if ( cset->usable() )
 929         return cset;
 930
 931     delete cset;
 932     cset = NULL;
 933 #endif // wxUSE_FONTMAP
 934
 935     wxLogError(_("Cannot convert from encoding '%s'!"), name);
 936
 937     return NULL;
 938 }
 939
 940 // ============================================================================
 941 // wxCSConv implementation
 942 // ============================================================================
 943
 944 wxCSConv::wxCSConv(const wxChar *charset)
 945 {
 946     m_name = (wxChar *)NULL;
 947     m_cset = (wxCharacterSet *) NULL;
 948     m_deferred = TRUE;
 949
 950     SetName(charset);
 951 }
 952
 953 wxCSConv::~wxCSConv()
 954 {
 955     Clear();
 956 }
 957
 958 wxCSConv::wxCSConv(const wxCSConv& conv)
 959     : wxMBConv()
 960 {
 961     Clear();
 962     SetName(conv.m_name);
 963 }
 964
 965 wxCSConv& wxCSConv::operator=(const wxCSConv& conv)
 966 {
 967     Clear();
 968     SetName(conv.m_name);
 969     return *this;
 970 }
 971
 972 void wxCSConv::Clear()
 973 {
 974     if (m_name)
 975         free(m_name);
 976     if (m_cset)
 977         delete m_cset;
 978     m_name = NULL;
 979     m_cset = NULL;
 980 }
 981
 982 void wxCSConv::SetName(const wxChar *charset)
 983 {
 984     if (charset)
 985     {
 986         m_name = wxStrdup(charset);
 987         m_deferred = TRUE;
 988     }
 989 }
 990
 991 void wxCSConv::LoadNow()
 992 {
 993     if (m_deferred)
 994     {
 995         if ( !m_name )
 996         {
 997             wxString name = wxLocale::GetSystemEncodingName();
 998             if ( !name.empty() )
 999             {
1000                 SetName(name);
1001             }
1002         }
1003
1004         // wxGetCharacterSet() complains about NULL name
1005         m_cset = m_name ? wxGetCharacterSet(m_name) : NULL;
1006         m_deferred = FALSE;
1007     }
1008 }
1009
1010 size_t wxCSConv::MB2WC(wchar_t *buf, const char *psz, size_t n) const
1011 {
1012     ((wxCSConv *)this)->LoadNow(); // discard constness
1013
1014     if (m_cset)
1015         return m_cset->MB2WC(buf, psz, n);
1016
1017     // latin-1 (direct)
1018     size_t len = strlen(psz);
1019
1020     if (buf)
1021     {
1022         for (size_t c = 0; c <= len; c++)
1023             buf[c] = (unsigned char)(psz[c]);
1024     }
1025
1026     return len;
1027 }
1028
1029 size_t wxCSConv::WC2MB(char *buf, const wchar_t *psz, size_t n) const
1030 {
1031     ((wxCSConv *)this)->LoadNow(); // discard constness
1032
1033     if (m_cset)
1034         return m_cset->WC2MB(buf, psz, n);
1035
1036     // latin-1 (direct)
1037 #if ( defined(__BORLANDC__) && (__BORLANDC__ > 0x530) ) \
1038     || ( defined(__MWERKS__) && defined(__WXMSW__) )
1039     size_t len=std::wcslen(psz);
1040 #else
1041     size_t len=::wcslen(psz);
1042 #endif
1043     if (buf)
1044     {
1045         for (size_t c = 0; c <= len; c++)
1046             buf[c] = (psz[c] > 0xff) ? '?' : psz[c];
1047     }
1048
1049     return len;
1050 }
1051
1052 #endif // wxUSE_WCHAR_T
1053
1054