src/common/strconv.cpp

   1 /////////////////////////////////////////////////////////////////////////////
   2 // Name:        strconv.cpp
   3 // Purpose:     Unicode conversion classes
   4 // Author:      Ove Kaaven, Robert Roebling, Vadim Zeitlin
   5 // Modified by:
   6 // Created:     29/01/98
   7 // RCS-ID:      $Id$
   8 // Copyright:   (c) 1999 Ove Kaaven, Robert Roebling, Vadim Zeitlin
   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 #ifdef __SALFORDC__
  41   #include <clib.h>
  42 #endif
  43
  44 #ifdef HAVE_ICONV_H
  45   #include <iconv.h>
  46 #endif
  47 #ifdef HAVE_LANGINFO_H
  48   #include <langinfo.h>
  49 #endif
  50
  51 #ifdef __WXMSW__
  52   #include <windows.h>
  53 #endif
  54
  55 #include "wx/debug.h"
  56 #include "wx/strconv.h"
  57 #include "wx/intl.h"
  58 #include "wx/log.h"
  59
  60 #if defined(WORDS_BIGENDIAN) || defined(__STDC_ISO_10646__)
  61 #define BSWAP_UCS4(str, len)
  62 #define BSWAP_UCS2(str, len)
  63 #else
  64 #define BSWAP_UCS4(str, len) { unsigned _c; for (_c=0; _c<len; _c++) str[_c]=wxUINT32_SWAP_ALWAYS(str[_c]); }
  65 #define BSWAP_UCS2(str, len) { unsigned _c; for (_c=0; _c<len; _c++) str[_c]=wxUINT16_SWAP_ALWAYS(str[_c]); }
  66 #define WC_NEED_BSWAP
  67 #endif
  68 #define BSWAP_UTF32(str, len) BSWAP_UCS4(str, len)
  69 #define BSWAP_UTF16(str, len) BSWAP_UCS2(str, len)
  70
  71 #if SIZEOF_WCHAR_T == 4
  72 #define WC_NAME "UCS4"
  73 #define WC_BSWAP BSWAP_UCS4
  74 #elif SIZEOF_WCHAR_T == 2
  75 #define WC_NAME "UTF16"
  76 #define WC_BSWAP BSWAP_UTF16
  77 #define WC_UTF16
  78 #endif
  79
  80 // ----------------------------------------------------------------------------
  81 // globals
  82 // ----------------------------------------------------------------------------
  83
  84 WXDLLEXPORT_DATA(wxMBConv *) wxConvCurrent = &wxConvLibc;
  85
  86 // ============================================================================
  87 // implementation
  88 // ============================================================================
  89
  90 #if wxUSE_WCHAR_T
  91
  92 #ifdef WC_UTF16
  93
  94 static size_t encode_utf16(wxUint32 input,wxUint16*output)
  95 {
  96   if (input<=0xffff) {
  97     if (output) *output++ = input;
  98     return 1;
  99   } else
 100   if (input>=0x110000) {
 101     return (size_t)-1;
 102   } else {
 103     if (output) {
 104       *output++ = (input >> 10)+0xd7c0;
 105       *output++ = (input&0x3ff)+0xdc00;
 106     }
 107     return 2;
 108   }
 109 }
 110
 111 static size_t decode_utf16(wxUint16*input,wxUint32&output)
 112 {
 113   if ((*input<0xd800) || (*input>0xdfff)) {
 114     output = *input;
 115     return 1;
 116   } else
 117   if ((input[1]<0xdc00) || (input[1]>=0xdfff)) {
 118     output = *input;
 119     return (size_t)-1;
 120   } else {
 121     output = ((input[0] - 0xd7c0) << 10) + (input[1] - 0xdc00);
 122     return 2;
 123   }
 124 }
 125
 126 #endif // WC_UTF16
 127
 128 // ----------------------------------------------------------------------------
 129 // wxMBConv
 130 // ----------------------------------------------------------------------------
 131
 132 WXDLLEXPORT_DATA(wxMBConv) wxConvLibc;
 133
 134 size_t wxMBConv::MB2WC(wchar_t *buf, const char *psz, size_t n) const
 135 {
 136     return wxMB2WC(buf, psz, n);
 137 }
 138
 139 size_t wxMBConv::WC2MB(char *buf, const wchar_t *psz, size_t n) const
 140 {
 141     return wxWC2MB(buf, psz, n);
 142 }
 143
 144 const wxWCharBuffer wxMBConv::cMB2WC(const char *psz) const
 145 {
 146     if (psz)
 147     {
 148         size_t nLen = MB2WC((wchar_t *) NULL, psz, 0);
 149         if (nLen == (size_t)-1)
 150             return wxWCharBuffer((wchar_t *) NULL);
 151         wxWCharBuffer buf(nLen);
 152         MB2WC((wchar_t *)(const wchar_t *) buf, psz, nLen);
 153         return buf;
 154     }
 155     else
 156         return wxWCharBuffer((wchar_t *) NULL);
 157 }
 158
 159 const wxCharBuffer wxMBConv::cWC2MB(const wchar_t *psz) const
 160 {
 161     if (psz)
 162     {
 163         size_t nLen = WC2MB((char *) NULL, psz, 0);
 164         if (nLen == (size_t)-1)
 165             return wxCharBuffer((char *) NULL);
 166         wxCharBuffer buf(nLen);
 167         WC2MB((char *)(const char *) buf, psz, nLen);
 168         return buf;
 169     }
 170     else
 171         return wxCharBuffer((char *) NULL);
 172 }
 173
 174 // ----------------------------------------------------------------------------
 175 // standard file conversion
 176 // ----------------------------------------------------------------------------
 177
 178 WXDLLEXPORT_DATA(wxMBConvFile) wxConvFile;
 179
 180 // just use the libc conversion for now
 181 size_t wxMBConvFile::MB2WC(wchar_t *buf, const char *psz, size_t n) const
 182 {
 183     return wxMB2WC(buf, psz, n);
 184 }
 185
 186 size_t wxMBConvFile::WC2MB(char *buf, const wchar_t *psz, size_t n) const
 187 {
 188     return wxWC2MB(buf, psz, n);
 189 }
 190
 191 // ----------------------------------------------------------------------------
 192 // standard gdk conversion
 193 // ----------------------------------------------------------------------------
 194
 195 #ifdef __WXGTK12__
 196
 197 WXDLLEXPORT_DATA(wxMBConvGdk) wxConvGdk;
 198
 199 #include <gdk/gdk.h>
 200
 201 size_t wxMBConvGdk::MB2WC(wchar_t *buf, const char *psz, size_t n) const
 202 {
 203   if (buf) {
 204     return gdk_mbstowcs((GdkWChar *)buf, psz, n);
 205   } else {
 206     GdkWChar *nbuf = new GdkWChar[n=strlen(psz)];
 207     size_t len = gdk_mbstowcs(nbuf, psz, n);
 208     delete [] nbuf;
 209     return len;
 210   }
 211 }
 212
 213 size_t wxMBConvGdk::WC2MB(char *buf, const wchar_t *psz, size_t n) const
 214 {
 215   char *mbstr = gdk_wcstombs((GdkWChar *)psz);
 216   size_t len = mbstr ? strlen(mbstr) : 0;
 217   if (buf) {
 218     if (len > n) len = n;
 219     memcpy(buf, psz, len);
 220     if (len < n) buf[len] = 0;
 221   }
 222   return len;
 223 }
 224
 225 #endif // GTK > 1.0
 226
 227 // ----------------------------------------------------------------------------
 228 // UTF-7
 229 // ----------------------------------------------------------------------------
 230
 231 WXDLLEXPORT_DATA(wxMBConvUTF7) wxConvUTF7;
 232
 233 #if 0
 234 static char utf7_setD[]="ABCDEFGHIJKLMNOPQRSTUVWXYZ"
 235                         "abcdefghijklmnopqrstuvwxyz"
 236                         "0123456789'(),-./:?";
 237 static char utf7_setO[]="!\"#$%&*;<=>@[]^_`{|}";
 238 static char utf7_setB[]="ABCDEFGHIJKLMNOPQRSTUVWXYZ"
 239                         "abcdefghijklmnopqrstuvwxyz"
 240                         "0123456789+/";
 241 #endif
 242
 243 // TODO: write actual implementations of UTF-7 here
 244 size_t wxMBConvUTF7::MB2WC(wchar_t * WXUNUSED(buf),
 245                            const char * WXUNUSED(psz),
 246                            size_t WXUNUSED(n)) const
 247 {
 248   return 0;
 249 }
 250
 251 size_t wxMBConvUTF7::WC2MB(char * WXUNUSED(buf),
 252                            const wchar_t * WXUNUSED(psz),
 253                            size_t WXUNUSED(n)) const
 254 {
 255   return 0;
 256 }
 257
 258 // ----------------------------------------------------------------------------
 259 // UTF-8
 260 // ----------------------------------------------------------------------------
 261
 262 WXDLLEXPORT_DATA(wxMBConvUTF8) wxConvUTF8;
 263
 264 static wxUint32 utf8_max[]={0x7f,0x7ff,0xffff,0x1fffff,0x3ffffff,0x7fffffff,0xffffffff};
 265
 266 size_t wxMBConvUTF8::MB2WC(wchar_t *buf, const char *psz, size_t n) const
 267 {
 268   size_t len = 0;
 269
 270   while (*psz && ((!buf) || (len<n))) {
 271     unsigned char cc=*psz++, fc=cc;
 272     unsigned cnt;
 273     for (cnt=0; fc&0x80; cnt++) fc<<=1;
 274     if (!cnt) {
 275       // plain ASCII char
 276       if (buf) *buf++=cc;
 277       len++;
 278     } else {
 279       cnt--;
 280       if (!cnt) {
 281         // invalid UTF-8 sequence
 282         return (size_t)-1;
 283       } else {
 284         unsigned ocnt=cnt-1;
 285         wxUint32 res=cc&(0x3f>>cnt);
 286         while (cnt--) {
 287           cc = *psz++;
 288           if ((cc&0xC0)!=0x80) {
 289             // invalid UTF-8 sequence
 290             return (size_t)-1;
 291           }
 292           res=(res<<6)|(cc&0x3f);
 293         }
 294         if (res<=utf8_max[ocnt]) {
 295           // illegal UTF-8 encoding
 296           return (size_t)-1;
 297         }
 298 #ifdef WC_UTF16
 299         size_t pa = encode_utf16(res, buf);
 300         if (pa == (size_t)-1)
 301           return (size_t)-1;
 302         if (buf) buf+=pa;
 303         len+=pa;
 304 #else
 305         if (buf) *buf++=res;
 306         len++;
 307 #endif
 308       }
 309     }
 310   }
 311   if (buf && (len<n)) *buf = 0;
 312   return len;
 313 }
 314
 315 size_t wxMBConvUTF8::WC2MB(char *buf, const wchar_t *psz, size_t n) const
 316 {
 317   size_t len = 0;
 318
 319   while (*psz && ((!buf) || (len<n))) {
 320     wxUint32 cc;
 321 #ifdef WC_UTF16
 322     size_t pa = decode_utf16(psz,cc);
 323     psz += (pa == (size_t)-1) ? 1 : pa;
 324 #else
 325     cc=(*psz++)&0x7fffffff;
 326 #endif
 327     unsigned cnt;
 328     for (cnt=0; cc>utf8_max[cnt]; cnt++);
 329     if (!cnt) {
 330       // plain ASCII char
 331       if (buf) *buf++=cc;
 332       len++;
 333     } else {
 334       len+=cnt+1;
 335       if (buf) {
 336         *buf++=(-128>>cnt)|((cc>>(cnt*6))&(0x3f>>cnt));
 337         while (cnt--)
 338           *buf++=0x80|((cc>>(cnt*6))&0x3f);
 339       }
 340     }
 341   }
 342   if (buf && (len<n)) *buf = 0;
 343   return len;
 344 }
 345
 346 // ----------------------------------------------------------------------------
 347 // specified character set
 348 // ----------------------------------------------------------------------------
 349
 350 WXDLLEXPORT_DATA(wxCSConv) wxConvLocal((const wxChar *)NULL);
 351
 352 #include "wx/encconv.h"
 353 #include "wx/fontmap.h"
 354
 355 // TODO: add some tables here
 356 // - perhaps common encodings to common codepages (for Win32)
 357 // - perhaps common encodings to objects ("UTF8" -> wxConvUTF8)
 358 // - move wxEncodingConverter meat in here
 359
 360 #ifdef __WIN32__
 361 #include "wx/msw/registry.h"
 362 // this should work if M$ Internet Exploiter is installed
 363 static long CharsetToCodepage(const wxChar *name)
 364 {
 365     if (!name)
 366         return GetACP();
 367
 368     long CP=-1;
 369
 370     wxString cn(name);
 371     do {
 372         wxString path(wxT("MIME\\Database\\Charset\\"));
 373         path += cn;
 374         wxRegKey key(wxRegKey::HKCR, path);
 375
 376         if (!key.Exists()) continue;
 377
 378         // two cases: either there's an AliasForCharset string,
 379         // or there are Codepage and InternetEncoding dwords.
 380         // The InternetEncoding gives us the actual encoding,
 381         // the Codepage just says which Windows character set to
 382         // use when displaying the data.
 383         if (key.HasValue(wxT("InternetEncoding")) &&
 384             key.QueryValue(wxT("InternetEncoding"), &CP)) break;
 385
 386         // no encoding, see if it's an alias
 387         if (!key.HasValue(wxT("AliasForCharset")) ||
 388             !key.QueryValue(wxT("AliasForCharset"), cn)) break;
 389     } while (1);
 390
 391     return CP;
 392 }
 393 #endif
 394
 395 class wxCharacterSet
 396 {
 397 public:
 398     wxCharacterSet(const wxChar*name)
 399         : cname(name) {}
 400     virtual ~wxCharacterSet()
 401         {}
 402     virtual size_t MB2WC(wchar_t*buf, const char*psz, size_t n)
 403         { return (size_t)-1; }
 404     virtual size_t WC2MB(char*buf, const wchar_t*psz, size_t n)
 405         { return (size_t)-1; }
 406     virtual bool usable()
 407         { return FALSE; }
 408 public:
 409     const wxChar*cname;
 410 };
 411
 412 class ID_CharSet : public wxCharacterSet
 413 {
 414 public:
 415     ID_CharSet(const wxChar*name,wxMBConv*cnv)
 416         : wxCharacterSet(name), work(cnv) {}
 417
 418     size_t MB2WC(wchar_t*buf, const char*psz, size_t n)
 419         { return work ? work->MB2WC(buf,psz,n) : (size_t)-1; }
 420
 421     size_t WC2MB(char*buf, const wchar_t*psz, size_t n)
 422         { return work ? work->WC2MB(buf,psz,n) : (size_t)-1; }
 423
 424     bool usable()
 425         { return work!=NULL; }
 426 public:
 427     wxMBConv*work;
 428 };
 429
 430
 431 #ifdef HAVE_ICONV_H
 432
 433 // VS: glibc 2.1.3 is broken in that iconv() conversion to/from UCS4 fails with E2BIG
 434 //     if output buffer is _exactly_ as big as needed. Such case is (unless there's
 435 //     yet another bug in glibc) the only case when iconv() returns with (size_t)-1
 436 //     (which means error) and says there are 0 bytes left in the input buffer --
 437 //     when _real_ error occurs, bytes-left-in-input buffer is non-zero. Hence,
 438 //     this alternative test for iconv() failure.
 439 //     [This bug does not appear in glibc 2.2.]
 440 #if defined(__GLIBC__) && __GLIBC__ == 2 && __GLIBC_MINOR__ <= 1
 441 #define ICONV_FAILED(cres, bufLeft) ((cres == (size_t)-1) && \
 442                                      (errno != E2BIG || bufLeft != 0))
 443 #else
 444 #define ICONV_FAILED(cres, bufLeft)  (cres == (size_t)-1)
 445 #endif
 446
 447 class IC_CharSet : public wxCharacterSet
 448 {
 449 public:
 450     IC_CharSet(const wxChar*name)
 451         : wxCharacterSet(name)
 452     {
 453         m2w = iconv_open(WC_NAME, wxConvLibc.cWX2MB(cname));
 454         w2m = iconv_open(wxConvLibc.cWX2MB(cname), WC_NAME);
 455     }
 456
 457     ~IC_CharSet()
 458     {
 459         if ( m2w != (iconv_t)-1 )
 460             iconv_close(m2w);
 461         if ( w2m != (iconv_t)-1 )
 462             iconv_close(w2m);
 463     }
 464
 465     size_t MB2WC(wchar_t *buf, const char *psz, size_t n)
 466     {
 467         size_t inbuf = strlen(psz);
 468         size_t outbuf = n * SIZEOF_WCHAR_T;
 469         size_t res, cres;
 470         // VS: Use these instead of psz, buf because iconv() modifies its arguments:
 471         wchar_t *bufPtr = buf;
 472         const char *pszPtr = psz;
 473
 474         if (buf)
 475         {
 476             // have destination buffer, convert there
 477 #ifdef WX_ICONV_TAKES_CHAR
 478             cres = iconv(m2w, (char**)&pszPtr, &inbuf, (char**)&bufPtr, &outbuf);
 479 #else
 480             cres = iconv(m2w, &pszPtr, &inbuf, (char**)&bufPtr, &outbuf);
 481 #endif
 482             res = n - (outbuf / SIZEOF_WCHAR_T);
 483             // convert to native endianness
 484 #ifdef WC_NEED_BSWAP
 485             WC_BSWAP(buf /* _not_ bufPtr */, res)
 486 #endif
 487         }
 488         else
 489         {
 490             // no destination buffer... convert using temp buffer
 491             // to calculate destination buffer requirement
 492             wchar_t tbuf[8];
 493             res = 0;
 494             do {
 495                 bufPtr = tbuf; outbuf = 8*SIZEOF_WCHAR_T;
 496 #ifdef WX_ICONV_TAKES_CHAR
 497                 cres = iconv( m2w, (char**)&pszPtr, &inbuf, (char**)&bufPtr, &outbuf );
 498 #else
 499                 cres = iconv( m2w, &pszPtr, &inbuf, (char**)&bufPtr, &outbuf );
 500 #endif
 501                 res += 8-(outbuf/SIZEOF_WCHAR_T);
 502             } while ((cres==(size_t)-1) && (errno==E2BIG));
 503         }
 504
 505         if (ICONV_FAILED(cres, inbuf))
 506             return (size_t)-1;
 507
 508         return res;
 509     }
 510
 511     size_t WC2MB(char*buf, const wchar_t*psz, size_t n)
 512     {
 513 #if defined(__BORLANDC__) && (__BORLANDC__ > 0x530)
 514         size_t inbuf = std::wcslen(psz) * SIZEOF_WCHAR_T;
 515 #else
 516         size_t inbuf = ::wcslen(psz) * SIZEOF_WCHAR_T;
 517 #endif
 518         size_t outbuf = n;
 519         size_t res, cres;
 520
 521 #ifdef WC_NEED_BSWAP
 522         // need to copy to temp buffer to switch endianness
 523         // this absolutely doesn't rock!
 524         // (no, doing WC_BSWAP twice on the original buffer won't help, as it
 525         //  could be in read-only memory, or be accessed in some other thread)
 526         wchar_t*tmpbuf=(wchar_t*)malloc((inbuf+1)*SIZEOF_WCHAR_T);
 527         memcpy(tmpbuf,psz,(inbuf+1)*SIZEOF_WCHAR_T);
 528         WC_BSWAP(tmpbuf, inbuf)
 529         psz=tmpbuf;
 530 #endif
 531         if (buf)
 532         {
 533             // have destination buffer, convert there
 534 #ifdef WX_ICONV_TAKES_CHAR
 535             cres = iconv( w2m, (char**)&psz, &inbuf, &buf, &outbuf );
 536 #else
 537             cres = iconv( w2m, (const char**)&psz, &inbuf, &buf, &outbuf );
 538 #endif
 539             res = n-outbuf;
 540         }
 541         else
 542         {
 543             // no destination buffer... convert using temp buffer
 544             // to calculate destination buffer requirement
 545             char tbuf[16];
 546             res = 0;
 547             do {
 548                 buf = tbuf; outbuf = 16;
 549 #ifdef WX_ICONV_TAKES_CHAR
 550                 cres = iconv( w2m, (char**)&psz, &inbuf, &buf, &outbuf );
 551 #else
 552                 cres = iconv( w2m, (const char**)&psz, &inbuf, &buf, &outbuf );
 553 #endif
 554                 res += 16 - outbuf;
 555             } while ((cres==(size_t)-1) && (errno==E2BIG));
 556         }
 557 #ifdef WC_NEED_BSWAP
 558         free(tmpbuf);
 559 #endif
 560         if (ICONV_FAILED(cres, inbuf))
 561             return (size_t)-1;
 562
 563         return res;
 564     }
 565
 566     bool usable()
 567         { return (m2w != (iconv_t)-1) && (w2m != (iconv_t)-1); }
 568
 569 public:
 570     iconv_t m2w, w2m;
 571 };
 572 #endif
 573
 574 #ifdef __WIN32__
 575 class CP_CharSet : public wxCharacterSet
 576 {
 577 public:
 578     CP_CharSet(const wxChar*name)
 579         : wxCharacterSet(name), CodePage(CharsetToCodepage(name)) {}
 580
 581     size_t MB2WC(wchar_t*buf, const char*psz, size_t n)
 582     {
 583         size_t len = MultiByteToWideChar(CodePage,0,psz,-1,buf,buf?n:0);
 584         return len ? len : (size_t)-1;
 585     }
 586
 587     size_t WC2MB(char*buf, const wchar_t*psz, size_t n)
 588     {
 589         size_t len = WideCharToMultiByte(CodePage,0,psz,-1,buf,buf?n:0,NULL,NULL);
 590         return len ? len : (size_t)-1;
 591     }
 592
 593     bool usable()
 594         { return CodePage!=-1; }
 595
 596 public:
 597     long CodePage;
 598 };
 599 #endif
 600
 601 class EC_CharSet : public wxCharacterSet
 602 {
 603 public:
 604     // temporarily just use wxEncodingConverter stuff,
 605     // so that it works while a better implementation is built
 606     EC_CharSet(const wxChar*name) : wxCharacterSet(name), enc(wxFONTENCODING_SYSTEM)
 607     {
 608         if (name)
 609             enc = wxTheFontMapper->CharsetToEncoding(name, FALSE);
 610         m2w.Init(enc, wxFONTENCODING_UNICODE);
 611         w2m.Init(wxFONTENCODING_UNICODE, enc);
 612     }
 613
 614     size_t MB2WC(wchar_t*buf, const char*psz, size_t n)
 615     {
 616         size_t inbuf = strlen(psz);
 617         if (buf) m2w.Convert(psz,buf);
 618         return inbuf;
 619     }
 620
 621     size_t WC2MB(char*buf, const wchar_t*psz, size_t n)
 622     {
 623 #if defined(__BORLANDC__) && (__BORLANDC__ > 0x530)
 624         size_t inbuf = std::wcslen(psz);
 625 #else
 626         size_t inbuf = ::wcslen(psz);
 627 #endif
 628         if (buf)
 629             w2m.Convert(psz,buf);
 630
 631         return inbuf;
 632     }
 633
 634     bool usable()
 635         { return (enc!=wxFONTENCODING_SYSTEM) && (enc!=wxFONTENCODING_DEFAULT); }
 636
 637 public:
 638     wxFontEncoding enc;
 639     wxEncodingConverter m2w, w2m;
 640 };
 641
 642 static wxCharacterSet *wxGetCharacterSet(const wxChar *name)
 643 {
 644     wxCharacterSet *cset = NULL;
 645     if (name)
 646     {
 647         if (!wxStricmp(name, wxT("UTF8")) || !wxStricmp(name, wxT("UTF-8")))
 648         {
 649             cset = new ID_CharSet(name, &wxConvUTF8);
 650         }
 651         else
 652         {
 653 #ifdef HAVE_ICONV_H
 654             cset = new IC_CharSet(name); // may not take NULL
 655 #endif
 656         }
 657     }
 658
 659     if (cset && cset->usable()) return cset;
 660     if (cset) delete cset;
 661     cset = NULL;
 662 #ifdef __WIN32__
 663     cset = new CP_CharSet(name); // may take NULL
 664     if (cset->usable()) return cset;
 665 #endif
 666     if (cset) delete cset;
 667     cset = new EC_CharSet(name);
 668     if (cset->usable()) return cset;
 669     delete cset;
 670     wxLogError(_("Unknown encoding '%s'!"), name);
 671     return NULL;
 672 }
 673
 674 wxCSConv::wxCSConv(const wxChar *charset)
 675 {
 676     m_name = (wxChar *) NULL;
 677     m_cset = (wxCharacterSet *) NULL;
 678     m_deferred = TRUE;
 679     SetName(charset);
 680 }
 681
 682 wxCSConv::~wxCSConv()
 683 {
 684     if (m_name) free(m_name);
 685     if (m_cset) delete m_cset;
 686 }
 687
 688 void wxCSConv::SetName(const wxChar *charset)
 689 {
 690     if (charset)
 691     {
 692         m_name = wxStrdup(charset);
 693         m_deferred = TRUE;
 694     }
 695 }
 696
 697 void wxCSConv::LoadNow()
 698 {
 699     if (m_deferred)
 700     {
 701         if (!m_name)
 702         {
 703 #ifdef __UNIX__
 704 #if defined(HAVE_LANGINFO_H) && defined(CODESET)
 705             // GNU libc provides current character set this way
 706             char*alang = nl_langinfo(CODESET);
 707             if (alang)
 708             {
 709                 SetName(wxConvLibc.cMB2WX(alang));
 710             }
 711             else
 712 #endif
 713             {
 714                 // if we can't get at the character set directly,
 715                 // try to see if it's in the environment variables
 716                 // (in most cases this won't work, but I was out of ideas)
 717                 wxChar *lang = wxGetenv(wxT("LC_ALL"));
 718                 if (!lang) lang = wxGetenv(wxT("LC_CTYPE"));
 719                 if (!lang) lang = wxGetenv(wxT("LANG"));
 720                 wxChar *dot = lang ? wxStrchr(lang, wxT('.')) : (wxChar *)NULL;
 721                 if (dot) SetName(dot+1);
 722             }
 723 #endif
 724         }
 725         m_cset = wxGetCharacterSet(m_name);
 726         m_deferred = FALSE;
 727     }
 728 }
 729
 730 size_t wxCSConv::MB2WC(wchar_t *buf, const char *psz, size_t n) const
 731 {
 732     ((wxCSConv *)this)->LoadNow(); // discard constness
 733
 734     if (m_cset)
 735         return m_cset->MB2WC(buf, psz, n);
 736
 737     // latin-1 (direct)
 738     size_t len=strlen(psz);
 739
 740     if (buf)
 741     {
 742         for (size_t c=0; c<=len; c++)
 743             buf[c] = (unsigned char)(psz[c]);
 744     }
 745
 746     return len;
 747 }
 748
 749 size_t wxCSConv::WC2MB(char *buf, const wchar_t *psz, size_t n) const
 750 {
 751     ((wxCSConv *)this)->LoadNow(); // discard constness
 752
 753     if (m_cset)
 754         return m_cset->WC2MB(buf, psz, n);
 755
 756     // latin-1 (direct)
 757 #if defined(__BORLANDC__) && (__BORLANDC__ > 0x530)
 758     size_t len=std::wcslen(psz);
 759 #else
 760     size_t len=::wcslen(psz);
 761 #endif
 762     if (buf)
 763     {
 764         for (size_t c=0; c<=len; c++)
 765             buf[c] = (psz[c]>0xff) ? '?' : psz[c];
 766     }
 767
 768     return len;
 769 }
 770
 771 #ifdef HAVE_ICONV_H
 772 class IC_CharSetConverter
 773 {
 774 public:
 775     IC_CharSetConverter(IC_CharSet*from,IC_CharSet*to)
 776         { cnv = iconv_open(wxConvLibc.cWX2MB(to->cname),wxConvLibc.cWX2MB(from->cname)); }
 777
 778     ~IC_CharSetConverter()
 779         { if (cnv!=(iconv_t)-1) iconv_close(cnv); }
 780
 781     size_t Convert(char*buf, const char*psz, size_t n)
 782     {
 783         size_t inbuf = strlen(psz);
 784         size_t outbuf = n;
 785 #ifdef WX_ICONV_TAKES_CHAR
 786         size_t res = iconv( cnv, (char**)&psz, &inbuf, &buf, &outbuf );
 787 #else
 788         size_t res = iconv( cnv, &psz, &inbuf, &buf, &outbuf );
 789 #endif
 790         if (res==(size_t)-1) return (size_t)-1;
 791         return n-outbuf;
 792     }
 793
 794 public:
 795     iconv_t cnv;
 796 };
 797 #endif
 798
 799 class EC_CharSetConverter
 800 {
 801 public:
 802     EC_CharSetConverter(EC_CharSet*from,EC_CharSet*to)
 803         { cnv.Init(from->enc,to->enc); }
 804
 805     size_t Convert(char*buf, const char*psz, size_t n)
 806     {
 807         size_t inbuf = strlen(psz);
 808         if (buf) cnv.Convert(psz,buf);
 809         return inbuf;
 810     }
 811
 812 public:
 813     wxEncodingConverter cnv;
 814 };
 815
 816 #else // !wxUSE_WCHAR_T
 817
 818 // ----------------------------------------------------------------------------
 819 // stand-ins in absence of wchar_t
 820 // ----------------------------------------------------------------------------
 821
 822 WXDLLEXPORT_DATA(wxMBConv) wxConvLibc, wxConvFile;
 823
 824 #endif // wxUSE_WCHAR_T
 825
 826