]> git.saurik.com Git - wxWidgets.git/blob - src/common/strconv.cpp
projects / wxWidgets.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
clarified From/To8BitData's purpose
[wxWidgets.git] / src / common / strconv.cpp
1 /////////////////////////////////////////////////////////////////////////////
2 // Name: src/common/strconv.cpp
3 // Purpose: Unicode conversion classes
4 // Author: Ove Kaaven, Robert Roebling, Vadim Zeitlin, Vaclav Slavik,
5 // Ryan Norton, Fredrik Roubert (UTF7)
6 // Modified by:
7 // Created: 29/01/98
8 // RCS-ID: $Id$
9 // Copyright: (c) 1999 Ove Kaaven, Robert Roebling, Vaclav Slavik
10 // (c) 2000-2003 Vadim Zeitlin
11 // (c) 2004 Ryan Norton, Fredrik Roubert
12 // Licence: wxWindows licence
13 /////////////////////////////////////////////////////////////////////////////
14
15 // For compilers that support precompilation, includes "wx.h".
16 #include "wx/wxprec.h"
17
18 #ifdef __BORLANDC__
19 #pragma hdrstop
20 #endif //__BORLANDC__
21
22 #ifndef WX_PRECOMP
23 #include "wx/intl.h"
24 #include "wx/log.h"
25 #include "wx/utils.h"
26 #include "wx/hashmap.h"
27 #endif
28
29 #include "wx/strconv.h"
30
31 #if wxUSE_WCHAR_T
32
33 #ifndef __WXWINCE__
34 #include <errno.h>
35 #endif
36
37 #include <ctype.h>
38 #include <string.h>
39 #include <stdlib.h>
40
41 #if defined(__WIN32__) && !defined(__WXMICROWIN__)
42 #include "wx/msw/private.h"
43 #include "wx/msw/missing.h"
44 #define wxHAVE_WIN32_MB2WC
45 #endif
46
47 #ifdef HAVE_ICONV
48 #include <iconv.h>
49 #include "wx/thread.h"
50 #endif
51
52 #include "wx/encconv.h"
53 #include "wx/fontmap.h"
54
55 #ifdef __DARWIN__
56 #include "wx/mac/corefoundation/private/strconv_cf.h"
57 #endif //def __DARWIN__
58
59
60 #define TRACE_STRCONV _T("strconv")
61
62 // WC_UTF16 is defined only if sizeof(wchar_t) == 2, otherwise it's supposed to
63 // be 4 bytes
64 #if SIZEOF_WCHAR_T == 2
65 #define WC_UTF16
66 #endif
67
68
69 // ============================================================================
70 // implementation
71 // ============================================================================
72
73 // helper function of cMB2WC(): check if n bytes at this location are all NUL
74 static bool NotAllNULs(const char *p, size_t n)
75 {
76 while ( n && *p++ == '\0' )
77 n--;
78
79 return n != 0;
80 }
81
82 // ----------------------------------------------------------------------------
83 // UTF-16 en/decoding to/from UCS-4 with surrogates handling
84 // ----------------------------------------------------------------------------
85
86 static size_t encode_utf16(wxUint32 input, wxUint16 *output)
87 {
88 if (input <= 0xffff)
89 {
90 if (output)
91 *output = (wxUint16) input;
92
93 return 1;
94 }
95 else if (input >= 0x110000)
96 {
97 return wxCONV_FAILED;
98 }
99 else
100 {
101 if (output)
102 {
103 *output++ = (wxUint16) ((input >> 10) + 0xd7c0);
104 *output = (wxUint16) ((input & 0x3ff) + 0xdc00);
105 }
106
107 return 2;
108 }
109 }
110
111 static size_t decode_utf16(const wxUint16* input, wxUint32& output)
112 {
113 if ((*input < 0xd800) || (*input > 0xdfff))
114 {
115 output = *input;
116 return 1;
117 }
118 else if ((input[1] < 0xdc00) || (input[1] > 0xdfff))
119 {
120 output = *input;
121 return wxCONV_FAILED;
122 }
123 else
124 {
125 output = ((input[0] - 0xd7c0) << 10) + (input[1] - 0xdc00);
126 return 2;
127 }
128 }
129
130 #ifdef WC_UTF16
131 typedef wchar_t wxDecodeSurrogate_t;
132 #else // !WC_UTF16
133 typedef wxUint16 wxDecodeSurrogate_t;
134 #endif // WC_UTF16/!WC_UTF16
135
136 // returns the next UTF-32 character from the wchar_t buffer and advances the
137 // pointer to the character after this one
138 //
139 // if an invalid character is found, *pSrc is set to NULL, the caller must
140 // check for this
141 static wxUint32 wxDecodeSurrogate(const wxDecodeSurrogate_t **pSrc)
142 {
143 wxUint32 out;
144 const size_t
145 n = decode_utf16(wx_reinterpret_cast(const wxUint16 *, *pSrc), out);
146 if ( n == wxCONV_FAILED )
147 *pSrc = NULL;
148 else
149 *pSrc += n;
150
151 return out;
152 }
153
154 // ----------------------------------------------------------------------------
155 // wxMBConv
156 // ----------------------------------------------------------------------------
157
158 size_t
159 wxMBConv::ToWChar(wchar_t *dst, size_t dstLen,
160 const char *src, size_t srcLen) const
161 {
162 // although new conversion classes are supposed to implement this function
163 // directly, the existins ones only implement the old MB2WC() and so, to
164 // avoid to have to rewrite all conversion classes at once, we provide a
165 // default (but not efficient) implementation of this one in terms of the
166 // old function by copying the input to ensure that it's NUL-terminated and
167 // then using MB2WC() to convert it
168
169 // the number of chars [which would be] written to dst [if it were not NULL]
170 size_t dstWritten = 0;
171
172 // the number of NULs terminating this string
173 size_t nulLen = 0; // not really needed, but just to avoid warnings
174
175 // if we were not given the input size we just have to assume that the
176 // string is properly terminated as we have no way of knowing how long it
177 // is anyhow, but if we do have the size check whether there are enough
178 // NULs at the end
179 wxCharBuffer bufTmp;
180 const char *srcEnd;
181 if ( srcLen != wxNO_LEN )
182 {
183 // we need to know how to find the end of this string
184 nulLen = GetMBNulLen();
185 if ( nulLen == wxCONV_FAILED )
186 return wxCONV_FAILED;
187
188 // if there are enough NULs we can avoid the copy
189 if ( srcLen < nulLen || NotAllNULs(src + srcLen - nulLen, nulLen) )
190 {
191 // make a copy in order to properly NUL-terminate the string
192 bufTmp = wxCharBuffer(srcLen + nulLen - 1 /* 1 will be added */);
193 char * const p = bufTmp.data();
194 memcpy(p, src, srcLen);
195 for ( char *s = p + srcLen; s < p + srcLen + nulLen; s++ )
196 *s = '\0';
197
198 src = bufTmp;
199 }
200
201 srcEnd = src + srcLen;
202 }
203 else // quit after the first loop iteration
204 {
205 srcEnd = NULL;
206 }
207
208 for ( ;; )
209 {
210 // try to convert the current chunk
211 size_t lenChunk = MB2WC(NULL, src, 0);
212 if ( lenChunk == wxCONV_FAILED )
213 return wxCONV_FAILED;
214
215 lenChunk++; // for the L'\0' at the end of this chunk
216
217 dstWritten += lenChunk;
218
219 if ( lenChunk == 1 )
220 {
221 // nothing left in the input string, conversion succeeded
222 break;
223 }
224
225 if ( dst )
226 {
227 if ( dstWritten > dstLen )
228 return wxCONV_FAILED;
229
230 if ( MB2WC(dst, src, lenChunk) == wxCONV_FAILED )
231 return wxCONV_FAILED;
232
233 dst += lenChunk;
234 }
235
236 if ( !srcEnd )
237 {
238 // we convert just one chunk in this case as this is the entire
239 // string anyhow
240 break;
241 }
242
243 // advance the input pointer past the end of this chunk
244 while ( NotAllNULs(src, nulLen) )
245 {
246 // notice that we must skip over multiple bytes here as we suppose
247 // that if NUL takes 2 or 4 bytes, then all the other characters do
248 // too and so if advanced by a single byte we might erroneously
249 // detect sequences of NUL bytes in the middle of the input
250 src += nulLen;
251 }
252
253 src += nulLen; // skipping over its terminator as well
254
255 // note that ">=" (and not just "==") is needed here as the terminator
256 // we skipped just above could be inside or just after the buffer
257 // delimited by inEnd
258 if ( src >= srcEnd )
259 break;
260 }
261
262 return dstWritten;
263 }
264
265 size_t
266 wxMBConv::FromWChar(char *dst, size_t dstLen,
267 const wchar_t *src, size_t srcLen) const
268 {
269 // the number of chars [which would be] written to dst [if it were not NULL]
270 size_t dstWritten = 0;
271
272 // make a copy of the input string unless it is already properly
273 // NUL-terminated
274 //
275 // if we don't know its length we have no choice but to assume that it is,
276 // indeed, properly terminated
277 wxWCharBuffer bufTmp;
278 if ( srcLen == wxNO_LEN )
279 {
280 srcLen = wxWcslen(src) + 1;
281 }
282 else if ( srcLen != 0 && src[srcLen - 1] != L'\0' )
283 {
284 // make a copy in order to properly NUL-terminate the string
285 bufTmp = wxWCharBuffer(srcLen);
286 memcpy(bufTmp.data(), src, srcLen * sizeof(wchar_t));
287 src = bufTmp;
288 }
289
290 const size_t lenNul = GetMBNulLen();
291 for ( const wchar_t * const srcEnd = src + srcLen;
292 src < srcEnd;
293 src += wxWcslen(src) + 1 /* skip L'\0' too */ )
294 {
295 // try to convert the current chunk
296 size_t lenChunk = WC2MB(NULL, src, 0);
297
298 if ( lenChunk == wxCONV_FAILED )
299 return wxCONV_FAILED;
300
301 lenChunk += lenNul;
302 dstWritten += lenChunk;
303
304 if ( dst )
305 {
306 if ( dstWritten > dstLen )
307 return wxCONV_FAILED;
308
309 if ( WC2MB(dst, src, lenChunk) == wxCONV_FAILED )
310 return wxCONV_FAILED;
311
312 dst += lenChunk;
313 }
314 }
315
316 return dstWritten;
317 }
318
319 size_t wxMBConv::MB2WC(wchar_t *outBuff, const char *inBuff, size_t outLen) const
320 {
321 size_t rc = ToWChar(outBuff, outLen, inBuff);
322 if ( rc != wxCONV_FAILED )
323 {
324 // ToWChar() returns the buffer length, i.e. including the trailing
325 // NUL, while this method doesn't take it into account
326 rc--;
327 }
328
329 return rc;
330 }
331
332 size_t wxMBConv::WC2MB(char *outBuff, const wchar_t *inBuff, size_t outLen) const
333 {
334 size_t rc = FromWChar(outBuff, outLen, inBuff);
335 if ( rc != wxCONV_FAILED )
336 {
337 rc -= GetMBNulLen();
338 }
339
340 return rc;
341 }
342
343 wxMBConv::~wxMBConv()
344 {
345 // nothing to do here (necessary for Darwin linking probably)
346 }
347
348 const wxWCharBuffer wxMBConv::cMB2WC(const char *psz) const
349 {
350 if ( psz )
351 {
352 // calculate the length of the buffer needed first
353 const size_t nLen = ToWChar(NULL, 0, psz);
354 if ( nLen != wxCONV_FAILED )
355 {
356 // now do the actual conversion
357 wxWCharBuffer buf(nLen - 1 /* +1 added implicitly */);
358
359 // +1 for the trailing NULL
360 if ( ToWChar(buf.data(), nLen, psz) != wxCONV_FAILED )
361 return buf;
362 }
363 }
364
365 return wxWCharBuffer();
366 }
367
368 const wxCharBuffer wxMBConv::cWC2MB(const wchar_t *pwz) const
369 {
370 if ( pwz )
371 {
372 const size_t nLen = FromWChar(NULL, 0, pwz);
373 if ( nLen != wxCONV_FAILED )
374 {
375 wxCharBuffer buf(nLen - 1);
376 if ( FromWChar(buf.data(), nLen, pwz) != wxCONV_FAILED )
377 return buf;
378 }
379 }
380
381 return wxCharBuffer();
382 }
383
384 const wxWCharBuffer
385 wxMBConv::cMB2WC(const char *inBuff, size_t inLen, size_t *outLen) const
386 {
387 const size_t dstLen = ToWChar(NULL, 0, inBuff, inLen);
388 if ( dstLen != wxCONV_FAILED )
389 {
390 // notice that we allocate space for dstLen+1 wide characters here
391 // because we want the buffer to always be NUL-terminated, even if the
392 // input isn't (as otherwise the caller has no way to know its length)
393 wxWCharBuffer wbuf(dstLen);
394 wbuf.data()[dstLen - 1] = L'\0';
395 if ( ToWChar(wbuf.data(), dstLen, inBuff, inLen) != wxCONV_FAILED )
396 {
397 if ( outLen )
398 {
399 *outLen = dstLen;
400 if ( wbuf[dstLen - 1] == L'\0' )
401 (*outLen)--;
402 }
403
404 return wbuf;
405 }
406 }
407
408 if ( outLen )
409 *outLen = 0;
410
411 return wxWCharBuffer();
412 }
413
414 const wxCharBuffer
415 wxMBConv::cWC2MB(const wchar_t *inBuff, size_t inLen, size_t *outLen) const
416 {
417 size_t dstLen = FromWChar(NULL, 0, inBuff, inLen);
418 if ( dstLen != wxCONV_FAILED )
419 {
420 const size_t nulLen = GetMBNulLen();
421
422 // as above, ensure that the buffer is always NUL-terminated, even if
423 // the input is not
424 wxCharBuffer buf(dstLen + nulLen - 1);
425 memset(buf.data() + dstLen, 0, nulLen);
426 if ( FromWChar(buf.data(), dstLen, inBuff, inLen) != wxCONV_FAILED )
427 {
428 if ( outLen )
429 {
430 *outLen = dstLen;
431
432 if ( dstLen >= nulLen &&
433 !NotAllNULs(buf.data() + dstLen - nulLen, nulLen) )
434 {
435 // in this case the output is NUL-terminated and we're not
436 // supposed to count NUL
437 *outLen -= nulLen;
438 }
439 }
440
441 return buf;
442 }
443 }
444
445 if ( outLen )
446 *outLen = 0;
447
448 return wxCharBuffer();
449 }
450
451 // ----------------------------------------------------------------------------
452 // wxMBConvLibc
453 // ----------------------------------------------------------------------------
454
455 size_t wxMBConvLibc::MB2WC(wchar_t *buf, const char *psz, size_t n) const
456 {
457 return wxMB2WC(buf, psz, n);
458 }
459
460 size_t wxMBConvLibc::WC2MB(char *buf, const wchar_t *psz, size_t n) const
461 {
462 return wxWC2MB(buf, psz, n);
463 }
464
465 // ----------------------------------------------------------------------------
466 // wxConvBrokenFileNames
467 // ----------------------------------------------------------------------------
468
469 #ifdef __UNIX__
470
471 wxConvBrokenFileNames::wxConvBrokenFileNames(const wxString& charset)
472 {
473 if ( wxStricmp(charset, _T("UTF-8")) == 0 ||
474 wxStricmp(charset, _T("UTF8")) == 0 )
475 m_conv = new wxMBConvUTF8(wxMBConvUTF8::MAP_INVALID_UTF8_TO_PUA);
476 else
477 m_conv = new wxCSConv(charset);
478 }
479
480 #endif // __UNIX__
481
482 // ----------------------------------------------------------------------------
483 // UTF-7
484 // ----------------------------------------------------------------------------
485
486 // Implementation (C) 2004 Fredrik Roubert
487
488 //
489 // BASE64 decoding table
490 //
491 static const unsigned char utf7unb64[] =
492 {
493 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
494 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
495 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
496 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
497 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
498 0xff, 0xff, 0xff, 0x3e, 0xff, 0xff, 0xff, 0x3f,
499 0x34, 0x35, 0x36, 0x37, 0x38, 0x39, 0x3a, 0x3b,
500 0x3c, 0x3d, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
501 0xff, 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06,
502 0x07, 0x08, 0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e,
503 0x0f, 0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16,
504 0x17, 0x18, 0x19, 0xff, 0xff, 0xff, 0xff, 0xff,
505 0xff, 0x1a, 0x1b, 0x1c, 0x1d, 0x1e, 0x1f, 0x20,
506 0x21, 0x22, 0x23, 0x24, 0x25, 0x26, 0x27, 0x28,
507 0x29, 0x2a, 0x2b, 0x2c, 0x2d, 0x2e, 0x2f, 0x30,
508 0x31, 0x32, 0x33, 0xff, 0xff, 0xff, 0xff, 0xff,
509 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
510 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
511 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
512 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
513 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
514 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
515 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
516 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
517 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
518 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
519 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
520 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
521 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
522 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
523 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
524 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff
525 };
526
527 size_t wxMBConvUTF7::MB2WC(wchar_t *buf, const char *psz, size_t n) const
528 {
529 size_t len = 0;
530
531 while ( *psz && (!buf || (len < n)) )
532 {
533 unsigned char cc = *psz++;
534 if (cc != '+')
535 {
536 // plain ASCII char
537 if (buf)
538 *buf++ = cc;
539 len++;
540 }
541 else if (*psz == '-')
542 {
543 // encoded plus sign
544 if (buf)
545 *buf++ = cc;
546 len++;
547 psz++;
548 }
549 else // start of BASE64 encoded string
550 {
551 bool lsb, ok;
552 unsigned int d, l;
553 for ( ok = lsb = false, d = 0, l = 0;
554 (cc = utf7unb64[(unsigned char)*psz]) != 0xff;
555 psz++ )
556 {
557 d <<= 6;
558 d += cc;
559 for (l += 6; l >= 8; lsb = !lsb)
560 {
561 unsigned char c = (unsigned char)((d >> (l -= 8)) % 256);
562 if (lsb)
563 {
564 if (buf)
565 *buf++ |= c;
566 len ++;
567 }
568 else
569 {
570 if (buf)
571 *buf = (wchar_t)(c << 8);
572 }
573
574 ok = true;
575 }
576 }
577
578 if ( !ok )
579 {
580 // in valid UTF7 we should have valid characters after '+'
581 return wxCONV_FAILED;
582 }
583
584 if (*psz == '-')
585 psz++;
586 }
587 }
588
589 if ( buf && (len < n) )
590 *buf = '\0';
591
592 return len;
593 }
594
595 //
596 // BASE64 encoding table
597 //
598 static const unsigned char utf7enb64[] =
599 {
600 'A', 'B', 'C', 'D', 'E', 'F', 'G', 'H',
601 'I', 'J', 'K', 'L', 'M', 'N', 'O', 'P',
602 'Q', 'R', 'S', 'T', 'U', 'V', 'W', 'X',
603 'Y', 'Z', 'a', 'b', 'c', 'd', 'e', 'f',
604 'g', 'h', 'i', 'j', 'k', 'l', 'm', 'n',
605 'o', 'p', 'q', 'r', 's', 't', 'u', 'v',
606 'w', 'x', 'y', 'z', '0', '1', '2', '3',
607 '4', '5', '6', '7', '8', '9', '+', '/'
608 };
609
610 //
611 // UTF-7 encoding table
612 //
613 // 0 - Set D (directly encoded characters)
614 // 1 - Set O (optional direct characters)
615 // 2 - whitespace characters (optional)
616 // 3 - special characters
617 //
618 static const unsigned char utf7encode[128] =
619 {
620 3, 3, 3, 3, 3, 3, 3, 3, 3, 2, 2, 3, 3, 2, 3, 3,
621 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,
622 2, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 3, 0, 0, 0, 3,
623 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 0,
624 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
625 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 3, 1, 1, 1,
626 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
627 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 3, 3
628 };
629
630 size_t wxMBConvUTF7::WC2MB(char *buf, const wchar_t *psz, size_t n) const
631 {
632 size_t len = 0;
633
634 while (*psz && ((!buf) || (len < n)))
635 {
636 wchar_t cc = *psz++;
637 if (cc < 0x80 && utf7encode[cc] < 1)
638 {
639 // plain ASCII char
640 if (buf)
641 *buf++ = (char)cc;
642
643 len++;
644 }
645 #ifndef WC_UTF16
646 else if (((wxUint32)cc) > 0xffff)
647 {
648 // no surrogate pair generation (yet?)
649 return wxCONV_FAILED;
650 }
651 #endif
652 else
653 {
654 if (buf)
655 *buf++ = '+';
656
657 len++;
658 if (cc != '+')
659 {
660 // BASE64 encode string
661 unsigned int lsb, d, l;
662 for (d = 0, l = 0; /*nothing*/; psz++)
663 {
664 for (lsb = 0; lsb < 2; lsb ++)
665 {
666 d <<= 8;
667 d += lsb ? cc & 0xff : (cc & 0xff00) >> 8;
668
669 for (l += 8; l >= 6; )
670 {
671 l -= 6;
672 if (buf)
673 *buf++ = utf7enb64[(d >> l) % 64];
674 len++;
675 }
676 }
677
678 cc = *psz;
679 if (!(cc) || (cc < 0x80 && utf7encode[cc] < 1))
680 break;
681 }
682
683 if (l != 0)
684 {
685 if (buf)
686 *buf++ = utf7enb64[((d % 16) << (6 - l)) % 64];
687
688 len++;
689 }
690 }
691
692 if (buf)
693 *buf++ = '-';
694 len++;
695 }
696 }
697
698 if (buf && (len < n))
699 *buf = 0;
700
701 return len;
702 }
703
704 // ----------------------------------------------------------------------------
705 // UTF-8
706 // ----------------------------------------------------------------------------
707
708 static const wxUint32 utf8_max[]=
709 { 0x7f, 0x7ff, 0xffff, 0x1fffff, 0x3ffffff, 0x7fffffff, 0xffffffff };
710
711 // boundaries of the private use area we use to (temporarily) remap invalid
712 // characters invalid in a UTF-8 encoded string
713 const wxUint32 wxUnicodePUA = 0x100000;
714 const wxUint32 wxUnicodePUAEnd = wxUnicodePUA + 256;
715
716 // this table gives the length of the UTF-8 encoding from its first character:
717 const unsigned char tableUtf8Lengths[256] = {
718 // single-byte sequences (ASCII):
719 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, // 00..0F
720 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, // 10..1F
721 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, // 20..2F
722 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, // 30..3F
723 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, // 40..4F
724 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, // 50..5F
725 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, // 60..6F
726 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, // 70..7F
727
728 // these are invalid:
729 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, // 80..8F
730 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, // 90..9F
731 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, // A0..AF
732 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, // B0..BF
733 0, 0, // C0,C1
734
735 // two-byte sequences:
736 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, // C2..CF
737 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, // D0..DF
738
739 // three-byte sequences:
740 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, // E0..EF
741
742 // four-byte sequences:
743 4, 4, 4, 4, 4, // F0..F4
744
745 // these are invalid again (5- or 6-byte
746 // sequences and sequences for code points
747 // above U+10FFFF, as restricted by RFC 3629):
748 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 // F5..FF
749 };
750
751 size_t
752 wxMBConvStrictUTF8::ToWChar(wchar_t *dst, size_t dstLen,
753 const char *src, size_t srcLen) const
754 {
755 wchar_t *out = dstLen ? dst : NULL;
756 size_t written = 0;
757
758 if ( srcLen == wxNO_LEN )
759 srcLen = strlen(src) + 1;
760
761 for ( const char *p = src; ; p++ )
762 {
763 if ( !(srcLen == wxNO_LEN ? *p : srcLen) )
764 {
765 // all done successfully, just add the trailing NULL if we are not
766 // using explicit length
767 if ( srcLen == wxNO_LEN )
768 {
769 if ( out )
770 {
771 if ( !dstLen )
772 break;
773
774 *out = L'\0';
775 }
776
777 written++;
778 }
779
780 return written;
781 }
782
783 if ( out && !dstLen-- )
784 break;
785
786 wxUint32 code;
787 unsigned char c = *p;
788
789 if ( c < 0x80 )
790 {
791 if ( srcLen == 0 ) // the test works for wxNO_LEN too
792 break;
793
794 if ( srcLen != wxNO_LEN )
795 srcLen--;
796
797 code = c;
798 }
799 else
800 {
801 unsigned len = tableUtf8Lengths[c];
802 if ( !len )
803 break;
804
805 if ( srcLen < len ) // the test works for wxNO_LEN too
806 break;
807
808 if ( srcLen != wxNO_LEN )
809 srcLen -= len;
810
811 // Char. number range | UTF-8 octet sequence
812 // (hexadecimal) | (binary)
813 // ----------------------+----------------------------------------
814 // 0000 0000 - 0000 007F | 0xxxxxxx
815 // 0000 0080 - 0000 07FF | 110xxxxx 10xxxxxx
816 // 0000 0800 - 0000 FFFF | 1110xxxx 10xxxxxx 10xxxxxx
817 // 0001 0000 - 0010 FFFF | 11110xxx 10xxxxxx 10xxxxxx 10xxxxxx
818 //
819 // Code point value is stored in bits marked with 'x',
820 // lowest-order bit of the value on the right side in the diagram
821 // above. (from RFC 3629)
822
823 // mask to extract lead byte's value ('x' bits above), by sequence
824 // length:
825 static const unsigned char leadValueMask[] = { 0x7F, 0x1F, 0x0F, 0x07 };
826
827 // mask and value of lead byte's most significant bits, by length:
828 static const unsigned char leadMarkerMask[] = { 0x80, 0xE0, 0xF0, 0xF8 };
829 static const unsigned char leadMarkerVal[] = { 0x00, 0xC0, 0xE0, 0xF0 };
830
831 len--; // it's more convenient to work with 0-based length here
832
833 // extract the lead byte's value bits:
834 if ( (c & leadMarkerMask[len]) != leadMarkerVal[len] )
835 break;
836
837 code = c & leadValueMask[len];
838
839 // all remaining bytes, if any, are handled in the same way
840 // regardless of sequence's length:
841 for ( ; len; --len )
842 {
843 c = *++p;
844 if ( (c & 0xC0) != 0x80 )
845 return wxCONV_FAILED;
846
847 code <<= 6;
848 code |= c & 0x3F;
849 }
850 }
851
852 #ifdef WC_UTF16
853 // cast is ok because wchar_t == wxUint16 if WC_UTF16
854 if ( encode_utf16(code, (wxUint16 *)out) == 2 )
855 {
856 if ( out )
857 out++;
858 written++;
859 }
860 #else // !WC_UTF16
861 if ( out )
862 *out = code;
863 #endif // WC_UTF16/!WC_UTF16
864
865 if ( out )
866 out++;
867
868 written++;
869 }
870
871 return wxCONV_FAILED;
872 }
873
874 size_t
875 wxMBConvStrictUTF8::FromWChar(char *dst, size_t dstLen,
876 const wchar_t *src, size_t srcLen) const
877 {
878 char *out = dstLen ? dst : NULL;
879 size_t written = 0;
880
881 for ( const wchar_t *wp = src; ; wp++ )
882 {
883 if ( !(srcLen == wxNO_LEN ? *wp : srcLen--) )
884 {
885 // all done successfully, just add the trailing NULL if we are not
886 // using explicit length
887 if ( srcLen == wxNO_LEN )
888 {
889 if ( out )
890 {
891 if ( !dstLen )
892 break;
893
894 *out = '\0';
895 }
896
897 written++;
898 }
899
900 return written;
901 }
902
903
904 wxUint32 code;
905 #ifdef WC_UTF16
906 // cast is ok for WC_UTF16
907 if ( decode_utf16((const wxUint16 *)wp, code) == 2 )
908 {
909 // skip the next char too as we decoded a surrogate
910 wp++;
911 }
912 #else // wchar_t is UTF-32
913 code = *wp & 0x7fffffff;
914 #endif
915
916 unsigned len;
917 if ( code <= 0x7F )
918 {
919 len = 1;
920 if ( out )
921 {
922 if ( dstLen < len )
923 break;
924
925 out[0] = (char)code;
926 }
927 }
928 else if ( code <= 0x07FF )
929 {
930 len = 2;
931 if ( out )
932 {
933 if ( dstLen < len )
934 break;
935
936 // NB: this line takes 6 least significant bits, encodes them as
937 // 10xxxxxx and discards them so that the next byte can be encoded:
938 out[1] = 0x80 | (code & 0x3F); code >>= 6;
939 out[0] = 0xC0 | code;
940 }
941 }
942 else if ( code < 0xFFFF )
943 {
944 len = 3;
945 if ( out )
946 {
947 if ( dstLen < len )
948 break;
949
950 out[2] = 0x80 | (code & 0x3F); code >>= 6;
951 out[1] = 0x80 | (code & 0x3F); code >>= 6;
952 out[0] = 0xE0 | code;
953 }
954 }
955 else if ( code <= 0x10FFFF )
956 {
957 len = 4;
958 if ( out )
959 {
960 if ( dstLen < len )
961 break;
962
963 out[3] = 0x80 | (code & 0x3F); code >>= 6;
964 out[2] = 0x80 | (code & 0x3F); code >>= 6;
965 out[1] = 0x80 | (code & 0x3F); code >>= 6;
966 out[0] = 0xF0 | code;
967 }
968 }
969 else
970 {
971 wxFAIL_MSG( _T("trying to encode undefined Unicode character") );
972 break;
973 }
974
975 if ( out )
976 {
977 out += len;
978 dstLen -= len;
979 }
980
981 written += len;
982 }
983
984 // we only get here if an error occurs during decoding
985 return wxCONV_FAILED;
986 }
987
988 size_t wxMBConvUTF8::ToWChar(wchar_t *buf, size_t n,
989 const char *psz, size_t srcLen) const
990 {
991 if ( m_options == MAP_INVALID_UTF8_NOT )
992 return wxMBConvStrictUTF8::ToWChar(buf, n, psz, srcLen);
993
994 size_t len = 0;
995
996 while ((srcLen == wxNO_LEN ? *psz : srcLen--) && ((!buf) || (len < n)))
997 {
998 const char *opsz = psz;
999 bool invalid = false;
1000 unsigned char cc = *psz++, fc = cc;
1001 unsigned cnt;
1002 for (cnt = 0; fc & 0x80; cnt++)
1003 fc <<= 1;
1004
1005 if (!cnt)
1006 {
1007 // plain ASCII char
1008 if (buf)
1009 *buf++ = cc;
1010 len++;
1011
1012 // escape the escape character for octal escapes
1013 if ((m_options & MAP_INVALID_UTF8_TO_OCTAL)
1014 && cc == '\\' && (!buf || len < n))
1015 {
1016 if (buf)
1017 *buf++ = cc;
1018 len++;
1019 }
1020 }
1021 else
1022 {
1023 cnt--;
1024 if (!cnt)
1025 {
1026 // invalid UTF-8 sequence
1027 invalid = true;
1028 }
1029 else
1030 {
1031 unsigned ocnt = cnt - 1;
1032 wxUint32 res = cc & (0x3f >> cnt);
1033 while (cnt--)
1034 {
1035 cc = *psz;
1036 if ((cc & 0xC0) != 0x80)
1037 {
1038 // invalid UTF-8 sequence
1039 invalid = true;
1040 break;
1041 }
1042
1043 psz++;
1044 res = (res << 6) | (cc & 0x3f);
1045 }
1046
1047 if (invalid || res <= utf8_max[ocnt])
1048 {
1049 // illegal UTF-8 encoding
1050 invalid = true;
1051 }
1052 else if ((m_options & MAP_INVALID_UTF8_TO_PUA) &&
1053 res >= wxUnicodePUA && res < wxUnicodePUAEnd)
1054 {
1055 // if one of our PUA characters turns up externally
1056 // it must also be treated as an illegal sequence
1057 // (a bit like you have to escape an escape character)
1058 invalid = true;
1059 }
1060 else
1061 {
1062 #ifdef WC_UTF16
1063 // cast is ok because wchar_t == wxUint16 if WC_UTF16
1064 size_t pa = encode_utf16(res, (wxUint16 *)buf);
1065 if (pa == wxCONV_FAILED)
1066 {
1067 invalid = true;
1068 }
1069 else
1070 {
1071 if (buf)
1072 buf += pa;
1073 len += pa;
1074 }
1075 #else // !WC_UTF16
1076 if (buf)
1077 *buf++ = (wchar_t)res;
1078 len++;
1079 #endif // WC_UTF16/!WC_UTF16
1080 }
1081 }
1082
1083 if (invalid)
1084 {
1085 if (m_options & MAP_INVALID_UTF8_TO_PUA)
1086 {
1087 while (opsz < psz && (!buf || len < n))
1088 {
1089 #ifdef WC_UTF16
1090 // cast is ok because wchar_t == wxUuint16 if WC_UTF16
1091 size_t pa = encode_utf16((unsigned char)*opsz + wxUnicodePUA, (wxUint16 *)buf);
1092 wxASSERT(pa != wxCONV_FAILED);
1093 if (buf)
1094 buf += pa;
1095 opsz++;
1096 len += pa;
1097 #else
1098 if (buf)
1099 *buf++ = (wchar_t)(wxUnicodePUA + (unsigned char)*opsz);
1100 opsz++;
1101 len++;
1102 #endif
1103 }
1104 }
1105 else if (m_options & MAP_INVALID_UTF8_TO_OCTAL)
1106 {
1107 while (opsz < psz && (!buf || len < n))
1108 {
1109 if ( buf && len + 3 < n )
1110 {
1111 unsigned char on = *opsz;
1112 *buf++ = L'\\';
1113 *buf++ = (wchar_t)( L'0' + on / 0100 );
1114 *buf++ = (wchar_t)( L'0' + (on % 0100) / 010 );
1115 *buf++ = (wchar_t)( L'0' + on % 010 );
1116 }
1117
1118 opsz++;
1119 len += 4;
1120 }
1121 }
1122 else // MAP_INVALID_UTF8_NOT
1123 {
1124 return wxCONV_FAILED;
1125 }
1126 }
1127 }
1128 }
1129
1130 if (srcLen == wxNO_LEN && buf && (len < n))
1131 *buf = 0;
1132
1133 return len + 1;
1134 }
1135
1136 static inline bool isoctal(wchar_t wch)
1137 {
1138 return L'0' <= wch && wch <= L'7';
1139 }
1140
1141 size_t wxMBConvUTF8::FromWChar(char *buf, size_t n,
1142 const wchar_t *psz, size_t srcLen) const
1143 {
1144 if ( m_options == MAP_INVALID_UTF8_NOT )
1145 return wxMBConvStrictUTF8::FromWChar(buf, n, psz, srcLen);
1146
1147 size_t len = 0;
1148
1149 while ((srcLen == wxNO_LEN ? *psz : srcLen--) && ((!buf) || (len < n)))
1150 {
1151 wxUint32 cc;
1152
1153 #ifdef WC_UTF16
1154 // cast is ok for WC_UTF16
1155 size_t pa = decode_utf16((const wxUint16 *)psz, cc);
1156 psz += (pa == wxCONV_FAILED) ? 1 : pa;
1157 #else
1158 cc = (*psz++) & 0x7fffffff;
1159 #endif
1160
1161 if ( (m_options & MAP_INVALID_UTF8_TO_PUA)
1162 && cc >= wxUnicodePUA && cc < wxUnicodePUAEnd )
1163 {
1164 if (buf)
1165 *buf++ = (char)(cc - wxUnicodePUA);
1166 len++;
1167 }
1168 else if ( (m_options & MAP_INVALID_UTF8_TO_OCTAL)
1169 && cc == L'\\' && psz[0] == L'\\' )
1170 {
1171 if (buf)
1172 *buf++ = (char)cc;
1173 psz++;
1174 len++;
1175 }
1176 else if ( (m_options & MAP_INVALID_UTF8_TO_OCTAL) &&
1177 cc == L'\\' &&
1178 isoctal(psz[0]) && isoctal(psz[1]) && isoctal(psz[2]) )
1179 {
1180 if (buf)
1181 {
1182 *buf++ = (char) ((psz[0] - L'0') * 0100 +
1183 (psz[1] - L'0') * 010 +
1184 (psz[2] - L'0'));
1185 }
1186
1187 psz += 3;
1188 len++;
1189 }
1190 else
1191 {
1192 unsigned cnt;
1193 for (cnt = 0; cc > utf8_max[cnt]; cnt++)
1194 {
1195 }
1196
1197 if (!cnt)
1198 {
1199 // plain ASCII char
1200 if (buf)
1201 *buf++ = (char) cc;
1202 len++;
1203 }
1204 else
1205 {
1206 len += cnt + 1;
1207 if (buf)
1208 {
1209 *buf++ = (char) ((-128 >> cnt) | ((cc >> (cnt * 6)) & (0x3f >> cnt)));
1210 while (cnt--)
1211 *buf++ = (char) (0x80 | ((cc >> (cnt * 6)) & 0x3f));
1212 }
1213 }
1214 }
1215 }
1216
1217 if (srcLen == wxNO_LEN && buf && (len < n))
1218 *buf = 0;
1219
1220 return len + 1;
1221 }
1222
1223 // ============================================================================
1224 // UTF-16
1225 // ============================================================================
1226
1227 #ifdef WORDS_BIGENDIAN
1228 #define wxMBConvUTF16straight wxMBConvUTF16BE
1229 #define wxMBConvUTF16swap wxMBConvUTF16LE
1230 #else
1231 #define wxMBConvUTF16swap wxMBConvUTF16BE
1232 #define wxMBConvUTF16straight wxMBConvUTF16LE
1233 #endif
1234
1235 /* static */
1236 size_t wxMBConvUTF16Base::GetLength(const char *src, size_t srcLen)
1237 {
1238 if ( srcLen == wxNO_LEN )
1239 {
1240 // count the number of bytes in input, including the trailing NULs
1241 const wxUint16 *inBuff = wx_reinterpret_cast(const wxUint16 *, src);
1242 for ( srcLen = 1; *inBuff++; srcLen++ )
1243 ;
1244
1245 srcLen *= BYTES_PER_CHAR;
1246 }
1247 else // we already have the length
1248 {
1249 // we can only convert an entire number of UTF-16 characters
1250 if ( srcLen % BYTES_PER_CHAR )
1251 return wxCONV_FAILED;
1252 }
1253
1254 return srcLen;
1255 }
1256
1257 // case when in-memory representation is UTF-16 too
1258 #ifdef WC_UTF16
1259
1260 // ----------------------------------------------------------------------------
1261 // conversions without endianness change
1262 // ----------------------------------------------------------------------------
1263
1264 size_t
1265 wxMBConvUTF16straight::ToWChar(wchar_t *dst, size_t dstLen,
1266 const char *src, size_t srcLen) const
1267 {
1268 // set up the scene for using memcpy() (which is presumably more efficient
1269 // than copying the bytes one by one)
1270 srcLen = GetLength(src, srcLen);
1271 if ( srcLen == wxNO_LEN )
1272 return wxCONV_FAILED;
1273
1274 const size_t inLen = srcLen / BYTES_PER_CHAR;
1275 if ( dst )
1276 {
1277 if ( dstLen < inLen )
1278 return wxCONV_FAILED;
1279
1280 memcpy(dst, src, srcLen);
1281 }
1282
1283 return inLen;
1284 }
1285
1286 size_t
1287 wxMBConvUTF16straight::FromWChar(char *dst, size_t dstLen,
1288 const wchar_t *src, size_t srcLen) const
1289 {
1290 if ( srcLen == wxNO_LEN )
1291 srcLen = wxWcslen(src) + 1;
1292
1293 srcLen *= BYTES_PER_CHAR;
1294
1295 if ( dst )
1296 {
1297 if ( dstLen < srcLen )
1298 return wxCONV_FAILED;
1299
1300 memcpy(dst, src, srcLen);
1301 }
1302
1303 return srcLen;
1304 }
1305
1306 // ----------------------------------------------------------------------------
1307 // endian-reversing conversions
1308 // ----------------------------------------------------------------------------
1309
1310 size_t
1311 wxMBConvUTF16swap::ToWChar(wchar_t *dst, size_t dstLen,
1312 const char *src, size_t srcLen) const
1313 {
1314 srcLen = GetLength(src, srcLen);
1315 if ( srcLen == wxNO_LEN )
1316 return wxCONV_FAILED;
1317
1318 srcLen /= BYTES_PER_CHAR;
1319
1320 if ( dst )
1321 {
1322 if ( dstLen < srcLen )
1323 return wxCONV_FAILED;
1324
1325 const wxUint16 *inBuff = wx_reinterpret_cast(const wxUint16 *, src);
1326 for ( size_t n = 0; n < srcLen; n++, inBuff++ )
1327 {
1328 *dst++ = wxUINT16_SWAP_ALWAYS(*inBuff);
1329 }
1330 }
1331
1332 return srcLen;
1333 }
1334
1335 size_t
1336 wxMBConvUTF16swap::FromWChar(char *dst, size_t dstLen,
1337 const wchar_t *src, size_t srcLen) const
1338 {
1339 if ( srcLen == wxNO_LEN )
1340 srcLen = wxWcslen(src) + 1;
1341
1342 srcLen *= BYTES_PER_CHAR;
1343
1344 if ( dst )
1345 {
1346 if ( dstLen < srcLen )
1347 return wxCONV_FAILED;
1348
1349 wxUint16 *outBuff = wx_reinterpret_cast(wxUint16 *, dst);
1350 for ( size_t n = 0; n < srcLen; n += BYTES_PER_CHAR, src++ )
1351 {
1352 *outBuff++ = wxUINT16_SWAP_ALWAYS(*src);
1353 }
1354 }
1355
1356 return srcLen;
1357 }
1358
1359 #else // !WC_UTF16: wchar_t is UTF-32
1360
1361 // ----------------------------------------------------------------------------
1362 // conversions without endianness change
1363 // ----------------------------------------------------------------------------
1364
1365 size_t
1366 wxMBConvUTF16straight::ToWChar(wchar_t *dst, size_t dstLen,
1367 const char *src, size_t srcLen) const
1368 {
1369 srcLen = GetLength(src, srcLen);
1370 if ( srcLen == wxNO_LEN )
1371 return wxCONV_FAILED;
1372
1373 const size_t inLen = srcLen / BYTES_PER_CHAR;
1374 if ( !dst )
1375 {
1376 // optimization: return maximal space which could be needed for this
1377 // string even if the real size could be smaller if the buffer contains
1378 // any surrogates
1379 return inLen;
1380 }
1381
1382 size_t outLen = 0;
1383 const wxUint16 *inBuff = wx_reinterpret_cast(const wxUint16 *, src);
1384 for ( const wxUint16 * const inEnd = inBuff + inLen; inBuff < inEnd; )
1385 {
1386 const wxUint32 ch = wxDecodeSurrogate(&inBuff);
1387 if ( !inBuff )
1388 return wxCONV_FAILED;
1389
1390 if ( ++outLen > dstLen )
1391 return wxCONV_FAILED;
1392
1393 *dst++ = ch;
1394 }
1395
1396
1397 return outLen;
1398 }
1399
1400 size_t
1401 wxMBConvUTF16straight::FromWChar(char *dst, size_t dstLen,
1402 const wchar_t *src, size_t srcLen) const
1403 {
1404 if ( srcLen == wxNO_LEN )
1405 srcLen = wxWcslen(src) + 1;
1406
1407 size_t outLen = 0;
1408 wxUint16 *outBuff = wx_reinterpret_cast(wxUint16 *, dst);
1409 for ( size_t n = 0; n < srcLen; n++ )
1410 {
1411 wxUint16 cc[2];
1412 const size_t numChars = encode_utf16(*src++, cc);
1413 if ( numChars == wxCONV_FAILED )
1414 return wxCONV_FAILED;
1415
1416 outLen += numChars * BYTES_PER_CHAR;
1417 if ( outBuff )
1418 {
1419 if ( outLen > dstLen )
1420 return wxCONV_FAILED;
1421
1422 *outBuff++ = cc[0];
1423 if ( numChars == 2 )
1424 {
1425 // second character of a surrogate
1426 *outBuff++ = cc[1];
1427 }
1428 }
1429 }
1430
1431 return outLen;
1432 }
1433
1434 // ----------------------------------------------------------------------------
1435 // endian-reversing conversions
1436 // ----------------------------------------------------------------------------
1437
1438 size_t
1439 wxMBConvUTF16swap::ToWChar(wchar_t *dst, size_t dstLen,
1440 const char *src, size_t srcLen) const
1441 {
1442 srcLen = GetLength(src, srcLen);
1443 if ( srcLen == wxNO_LEN )
1444 return wxCONV_FAILED;
1445
1446 const size_t inLen = srcLen / BYTES_PER_CHAR;
1447 if ( !dst )
1448 {
1449 // optimization: return maximal space which could be needed for this
1450 // string even if the real size could be smaller if the buffer contains
1451 // any surrogates
1452 return inLen;
1453 }
1454
1455 size_t outLen = 0;
1456 const wxUint16 *inBuff = wx_reinterpret_cast(const wxUint16 *, src);
1457 for ( const wxUint16 * const inEnd = inBuff + inLen; inBuff < inEnd; )
1458 {
1459 wxUint32 ch;
1460 wxUint16 tmp[2];
1461
1462 tmp[0] = wxUINT16_SWAP_ALWAYS(*inBuff);
1463 inBuff++;
1464 tmp[1] = wxUINT16_SWAP_ALWAYS(*inBuff);
1465
1466 const size_t numChars = decode_utf16(tmp, ch);
1467 if ( numChars == wxCONV_FAILED )
1468 return wxCONV_FAILED;
1469
1470 if ( numChars == 2 )
1471 inBuff++;
1472
1473 if ( ++outLen > dstLen )
1474 return wxCONV_FAILED;
1475
1476 *dst++ = ch;
1477 }
1478
1479
1480 return outLen;
1481 }
1482
1483 size_t
1484 wxMBConvUTF16swap::FromWChar(char *dst, size_t dstLen,
1485 const wchar_t *src, size_t srcLen) const
1486 {
1487 if ( srcLen == wxNO_LEN )
1488 srcLen = wxWcslen(src) + 1;
1489
1490 size_t outLen = 0;
1491 wxUint16 *outBuff = wx_reinterpret_cast(wxUint16 *, dst);
1492 for ( const wchar_t *srcEnd = src + srcLen; src < srcEnd; src++ )
1493 {
1494 wxUint16 cc[2];
1495 const size_t numChars = encode_utf16(*src, cc);
1496 if ( numChars == wxCONV_FAILED )
1497 return wxCONV_FAILED;
1498
1499 outLen += numChars * BYTES_PER_CHAR;
1500 if ( outBuff )
1501 {
1502 if ( outLen > dstLen )
1503 return wxCONV_FAILED;
1504
1505 *outBuff++ = wxUINT16_SWAP_ALWAYS(cc[0]);
1506 if ( numChars == 2 )
1507 {
1508 // second character of a surrogate
1509 *outBuff++ = wxUINT16_SWAP_ALWAYS(cc[1]);
1510 }
1511 }
1512 }
1513
1514 return outLen;
1515 }
1516
1517 #endif // WC_UTF16/!WC_UTF16
1518
1519
1520 // ============================================================================
1521 // UTF-32
1522 // ============================================================================
1523
1524 #ifdef WORDS_BIGENDIAN
1525 #define wxMBConvUTF32straight wxMBConvUTF32BE
1526 #define wxMBConvUTF32swap wxMBConvUTF32LE
1527 #else
1528 #define wxMBConvUTF32swap wxMBConvUTF32BE
1529 #define wxMBConvUTF32straight wxMBConvUTF32LE
1530 #endif
1531
1532
1533 WXDLLIMPEXP_DATA_BASE(wxMBConvUTF32LE) wxConvUTF32LE;
1534 WXDLLIMPEXP_DATA_BASE(wxMBConvUTF32BE) wxConvUTF32BE;
1535
1536 /* static */
1537 size_t wxMBConvUTF32Base::GetLength(const char *src, size_t srcLen)
1538 {
1539 if ( srcLen == wxNO_LEN )
1540 {
1541 // count the number of bytes in input, including the trailing NULs
1542 const wxUint32 *inBuff = wx_reinterpret_cast(const wxUint32 *, src);
1543 for ( srcLen = 1; *inBuff++; srcLen++ )
1544 ;
1545
1546 srcLen *= BYTES_PER_CHAR;
1547 }
1548 else // we already have the length
1549 {
1550 // we can only convert an entire number of UTF-32 characters
1551 if ( srcLen % BYTES_PER_CHAR )
1552 return wxCONV_FAILED;
1553 }
1554
1555 return srcLen;
1556 }
1557
1558 // case when in-memory representation is UTF-16
1559 #ifdef WC_UTF16
1560
1561 // ----------------------------------------------------------------------------
1562 // conversions without endianness change
1563 // ----------------------------------------------------------------------------
1564
1565 size_t
1566 wxMBConvUTF32straight::ToWChar(wchar_t *dst, size_t dstLen,
1567 const char *src, size_t srcLen) const
1568 {
1569 srcLen = GetLength(src, srcLen);
1570 if ( srcLen == wxNO_LEN )
1571 return wxCONV_FAILED;
1572
1573 const wxUint32 *inBuff = wx_reinterpret_cast(const wxUint32 *, src);
1574 const size_t inLen = srcLen / BYTES_PER_CHAR;
1575 size_t outLen = 0;
1576 for ( size_t n = 0; n < inLen; n++ )
1577 {
1578 wxUint16 cc[2];
1579 const size_t numChars = encode_utf16(*inBuff++, cc);
1580 if ( numChars == wxCONV_FAILED )
1581 return wxCONV_FAILED;
1582
1583 outLen += numChars;
1584 if ( dst )
1585 {
1586 if ( outLen > dstLen )
1587 return wxCONV_FAILED;
1588
1589 *dst++ = cc[0];
1590 if ( numChars == 2 )
1591 {
1592 // second character of a surrogate
1593 *dst++ = cc[1];
1594 }
1595 }
1596 }
1597
1598 return outLen;
1599 }
1600
1601 size_t
1602 wxMBConvUTF32straight::FromWChar(char *dst, size_t dstLen,
1603 const wchar_t *src, size_t srcLen) const
1604 {
1605 if ( srcLen == wxNO_LEN )
1606 srcLen = wxWcslen(src) + 1;
1607
1608 if ( !dst )
1609 {
1610 // optimization: return maximal space which could be needed for this
1611 // string instead of the exact amount which could be less if there are
1612 // any surrogates in the input
1613 //
1614 // we consider that surrogates are rare enough to make it worthwhile to
1615 // avoid running the loop below at the cost of slightly extra memory
1616 // consumption
1617 return srcLen * BYTES_PER_CHAR;
1618 }
1619
1620 wxUint32 *outBuff = wx_reinterpret_cast(wxUint32 *, dst);
1621 size_t outLen = 0;
1622 for ( const wchar_t * const srcEnd = src + srcLen; src < srcEnd; )
1623 {
1624 const wxUint32 ch = wxDecodeSurrogate(&src);
1625 if ( !src )
1626 return wxCONV_FAILED;
1627
1628 outLen += BYTES_PER_CHAR;
1629
1630 if ( outLen > dstLen )
1631 return wxCONV_FAILED;
1632
1633 *outBuff++ = ch;
1634 }
1635
1636 return outLen;
1637 }
1638
1639 // ----------------------------------------------------------------------------
1640 // endian-reversing conversions
1641 // ----------------------------------------------------------------------------
1642
1643 size_t
1644 wxMBConvUTF32swap::ToWChar(wchar_t *dst, size_t dstLen,
1645 const char *src, size_t srcLen) const
1646 {
1647 srcLen = GetLength(src, srcLen);
1648 if ( srcLen == wxNO_LEN )
1649 return wxCONV_FAILED;
1650
1651 const wxUint32 *inBuff = wx_reinterpret_cast(const wxUint32 *, src);
1652 const size_t inLen = srcLen / BYTES_PER_CHAR;
1653 size_t outLen = 0;
1654 for ( size_t n = 0; n < inLen; n++, inBuff++ )
1655 {
1656 wxUint16 cc[2];
1657 const size_t numChars = encode_utf16(wxUINT32_SWAP_ALWAYS(*inBuff), cc);
1658 if ( numChars == wxCONV_FAILED )
1659 return wxCONV_FAILED;
1660
1661 outLen += numChars;
1662 if ( dst )
1663 {
1664 if ( outLen > dstLen )
1665 return wxCONV_FAILED;
1666
1667 *dst++ = cc[0];
1668 if ( numChars == 2 )
1669 {
1670 // second character of a surrogate
1671 *dst++ = cc[1];
1672 }
1673 }
1674 }
1675
1676 return outLen;
1677 }
1678
1679 size_t
1680 wxMBConvUTF32swap::FromWChar(char *dst, size_t dstLen,
1681 const wchar_t *src, size_t srcLen) const
1682 {
1683 if ( srcLen == wxNO_LEN )
1684 srcLen = wxWcslen(src) + 1;
1685
1686 if ( !dst )
1687 {
1688 // optimization: return maximal space which could be needed for this
1689 // string instead of the exact amount which could be less if there are
1690 // any surrogates in the input
1691 //
1692 // we consider that surrogates are rare enough to make it worthwhile to
1693 // avoid running the loop below at the cost of slightly extra memory
1694 // consumption
1695 return srcLen*BYTES_PER_CHAR;
1696 }
1697
1698 wxUint32 *outBuff = wx_reinterpret_cast(wxUint32 *, dst);
1699 size_t outLen = 0;
1700 for ( const wchar_t * const srcEnd = src + srcLen; src < srcEnd; )
1701 {
1702 const wxUint32 ch = wxDecodeSurrogate(&src);
1703 if ( !src )
1704 return wxCONV_FAILED;
1705
1706 outLen += BYTES_PER_CHAR;
1707
1708 if ( outLen > dstLen )
1709 return wxCONV_FAILED;
1710
1711 *outBuff++ = wxUINT32_SWAP_ALWAYS(ch);
1712 }
1713
1714 return outLen;
1715 }
1716
1717 #else // !WC_UTF16: wchar_t is UTF-32
1718
1719 // ----------------------------------------------------------------------------
1720 // conversions without endianness change
1721 // ----------------------------------------------------------------------------
1722
1723 size_t
1724 wxMBConvUTF32straight::ToWChar(wchar_t *dst, size_t dstLen,
1725 const char *src, size_t srcLen) const
1726 {
1727 // use memcpy() as it should be much faster than hand-written loop
1728 srcLen = GetLength(src, srcLen);
1729 if ( srcLen == wxNO_LEN )
1730 return wxCONV_FAILED;
1731
1732 const size_t inLen = srcLen/BYTES_PER_CHAR;
1733 if ( dst )
1734 {
1735 if ( dstLen < inLen )
1736 return wxCONV_FAILED;
1737
1738 memcpy(dst, src, srcLen);
1739 }
1740
1741 return inLen;
1742 }
1743
1744 size_t
1745 wxMBConvUTF32straight::FromWChar(char *dst, size_t dstLen,
1746 const wchar_t *src, size_t srcLen) const
1747 {
1748 if ( srcLen == wxNO_LEN )
1749 srcLen = wxWcslen(src) + 1;
1750
1751 srcLen *= BYTES_PER_CHAR;
1752
1753 if ( dst )
1754 {
1755 if ( dstLen < srcLen )
1756 return wxCONV_FAILED;
1757
1758 memcpy(dst, src, srcLen);
1759 }
1760
1761 return srcLen;
1762 }
1763
1764 // ----------------------------------------------------------------------------
1765 // endian-reversing conversions
1766 // ----------------------------------------------------------------------------
1767
1768 size_t
1769 wxMBConvUTF32swap::ToWChar(wchar_t *dst, size_t dstLen,
1770 const char *src, size_t srcLen) const
1771 {
1772 srcLen = GetLength(src, srcLen);
1773 if ( srcLen == wxNO_LEN )
1774 return wxCONV_FAILED;
1775
1776 srcLen /= BYTES_PER_CHAR;
1777
1778 if ( dst )
1779 {
1780 if ( dstLen < srcLen )
1781 return wxCONV_FAILED;
1782
1783 const wxUint32 *inBuff = wx_reinterpret_cast(const wxUint32 *, src);
1784 for ( size_t n = 0; n < srcLen; n++, inBuff++ )
1785 {
1786 *dst++ = wxUINT32_SWAP_ALWAYS(*inBuff);
1787 }
1788 }
1789
1790 return srcLen;
1791 }
1792
1793 size_t
1794 wxMBConvUTF32swap::FromWChar(char *dst, size_t dstLen,
1795 const wchar_t *src, size_t srcLen) const
1796 {
1797 if ( srcLen == wxNO_LEN )
1798 srcLen = wxWcslen(src) + 1;
1799
1800 srcLen *= BYTES_PER_CHAR;
1801
1802 if ( dst )
1803 {
1804 if ( dstLen < srcLen )
1805 return wxCONV_FAILED;
1806
1807 wxUint32 *outBuff = wx_reinterpret_cast(wxUint32 *, dst);
1808 for ( size_t n = 0; n < srcLen; n += BYTES_PER_CHAR, src++ )
1809 {
1810 *outBuff++ = wxUINT32_SWAP_ALWAYS(*src);
1811 }
1812 }
1813
1814 return srcLen;
1815 }
1816
1817 #endif // WC_UTF16/!WC_UTF16
1818
1819
1820 // ============================================================================
1821 // The classes doing conversion using the iconv_xxx() functions
1822 // ============================================================================
1823
1824 #ifdef HAVE_ICONV
1825
1826 // VS: glibc 2.1.3 is broken in that iconv() conversion to/from UCS4 fails with
1827 // E2BIG if output buffer is _exactly_ as big as needed. Such case is
1828 // (unless there's yet another bug in glibc) the only case when iconv()
1829 // returns with (size_t)-1 (which means error) and says there are 0 bytes
1830 // left in the input buffer -- when _real_ error occurs,
1831 // bytes-left-in-input buffer is non-zero. Hence, this alternative test for
1832 // iconv() failure.
1833 // [This bug does not appear in glibc 2.2.]
1834 #if defined(__GLIBC__) && __GLIBC__ == 2 && __GLIBC_MINOR__ <= 1
1835 #define ICONV_FAILED(cres, bufLeft) ((cres == (size_t)-1) && \
1836 (errno != E2BIG || bufLeft != 0))
1837 #else
1838 #define ICONV_FAILED(cres, bufLeft) (cres == (size_t)-1)
1839 #endif
1840
1841 #define ICONV_CHAR_CAST(x) ((ICONV_CONST char **)(x))
1842
1843 #define ICONV_T_INVALID ((iconv_t)-1)
1844
1845 #if SIZEOF_WCHAR_T == 4
1846 #define WC_BSWAP wxUINT32_SWAP_ALWAYS
1847 #define WC_ENC wxFONTENCODING_UTF32
1848 #elif SIZEOF_WCHAR_T == 2
1849 #define WC_BSWAP wxUINT16_SWAP_ALWAYS
1850 #define WC_ENC wxFONTENCODING_UTF16
1851 #else // sizeof(wchar_t) != 2 nor 4
1852 // does this ever happen?
1853 #error "Unknown sizeof(wchar_t): please report this to wx-dev@lists.wxwindows.org"
1854 #endif
1855
1856 // ----------------------------------------------------------------------------
1857 // wxMBConv_iconv: encapsulates an iconv character set
1858 // ----------------------------------------------------------------------------
1859
1860 class wxMBConv_iconv : public wxMBConv
1861 {
1862 public:
1863 wxMBConv_iconv(const char *name);
1864 virtual ~wxMBConv_iconv();
1865
1866 virtual size_t MB2WC(wchar_t *buf, const char *psz, size_t n) const;
1867 virtual size_t WC2MB(char *buf, const wchar_t *psz, size_t n) const;
1868
1869 // classify this encoding as explained in wxMBConv::GetMBNulLen() comment
1870 virtual size_t GetMBNulLen() const;
1871
1872 #if wxUSE_UNICODE_UTF8
1873 virtual bool IsUTF8() const;
1874 #endif
1875
1876 virtual wxMBConv *Clone() const
1877 {
1878 wxMBConv_iconv *p = new wxMBConv_iconv(m_name.ToAscii());
1879 p->m_minMBCharWidth = m_minMBCharWidth;
1880 return p;
1881 }
1882
1883 bool IsOk() const
1884 { return (m2w != ICONV_T_INVALID) && (w2m != ICONV_T_INVALID); }
1885
1886 protected:
1887 // the iconv handlers used to translate from multibyte
1888 // to wide char and in the other direction
1889 iconv_t m2w,
1890 w2m;
1891
1892 #if wxUSE_THREADS
1893 // guards access to m2w and w2m objects
1894 wxMutex m_iconvMutex;
1895 #endif
1896
1897 private:
1898 // the name (for iconv_open()) of a wide char charset -- if none is
1899 // available on this machine, it will remain NULL
1900 static wxString ms_wcCharsetName;
1901
1902 // true if the wide char encoding we use (i.e. ms_wcCharsetName) has
1903 // different endian-ness than the native one
1904 static bool ms_wcNeedsSwap;
1905
1906
1907 // name of the encoding handled by this conversion
1908 wxString m_name;
1909
1910 // cached result of GetMBNulLen(); set to 0 meaning "unknown"
1911 // initially
1912 size_t m_minMBCharWidth;
1913 };
1914
1915 // make the constructor available for unit testing
1916 WXDLLIMPEXP_BASE wxMBConv* new_wxMBConv_iconv( const char* name )
1917 {
1918 wxMBConv_iconv* result = new wxMBConv_iconv( name );
1919 if ( !result->IsOk() )
1920 {
1921 delete result;
1922 return 0;
1923 }
1924
1925 return result;
1926 }
1927
1928 wxString wxMBConv_iconv::ms_wcCharsetName;
1929 bool wxMBConv_iconv::ms_wcNeedsSwap = false;
1930
1931 wxMBConv_iconv::wxMBConv_iconv(const char *name)
1932 : m_name(name)
1933 {
1934 m_minMBCharWidth = 0;
1935
1936 // check for charset that represents wchar_t:
1937 if ( ms_wcCharsetName.empty() )
1938 {
1939 wxLogTrace(TRACE_STRCONV, _T("Looking for wide char codeset:"));
1940
1941 #if wxUSE_FONTMAP
1942 const wxChar **names = wxFontMapperBase::GetAllEncodingNames(WC_ENC);
1943 #else // !wxUSE_FONTMAP
1944 static const wxChar *names_static[] =
1945 {
1946 #if SIZEOF_WCHAR_T == 4
1947 _T("UCS-4"),
1948 #elif SIZEOF_WCHAR_T = 2
1949 _T("UCS-2"),
1950 #endif
1951 NULL
1952 };
1953 const wxChar **names = names_static;
1954 #endif // wxUSE_FONTMAP/!wxUSE_FONTMAP
1955
1956 for ( ; *names && ms_wcCharsetName.empty(); ++names )
1957 {
1958 const wxString nameCS(*names);
1959
1960 // first try charset with explicit bytesex info (e.g. "UCS-4LE"):
1961 wxString nameXE(nameCS);
1962
1963 #ifdef WORDS_BIGENDIAN
1964 nameXE += _T("BE");
1965 #else // little endian
1966 nameXE += _T("LE");
1967 #endif
1968
1969 wxLogTrace(TRACE_STRCONV, _T(" trying charset \"%s\""),
1970 nameXE.c_str());
1971
1972 m2w = iconv_open(nameXE.ToAscii(), name);
1973 if ( m2w == ICONV_T_INVALID )
1974 {
1975 // try charset w/o bytesex info (e.g. "UCS4")
1976 wxLogTrace(TRACE_STRCONV, _T(" trying charset \"%s\""),
1977 nameCS.c_str());
1978 m2w = iconv_open(nameCS.ToAscii(), name);
1979
1980 // and check for bytesex ourselves:
1981 if ( m2w != ICONV_T_INVALID )
1982 {
1983 char buf[2], *bufPtr;
1984 wchar_t wbuf[2];
1985 size_t insz, outsz;
1986 size_t res;
1987
1988 buf[0] = 'A';
1989 buf[1] = 0;
1990 wbuf[0] = 0;
1991 insz = 2;
1992 outsz = SIZEOF_WCHAR_T * 2;
1993 char* wbufPtr = (char*)wbuf;
1994 bufPtr = buf;
1995
1996 res = iconv(
1997 m2w, ICONV_CHAR_CAST(&bufPtr), &insz,
1998 &wbufPtr, &outsz);
1999
2000 if (ICONV_FAILED(res, insz))
2001 {
2002 wxLogLastError(wxT("iconv"));
2003 wxLogError(_("Conversion to charset '%s' doesn't work."),
2004 nameCS.c_str());
2005 }
2006 else // ok, can convert to this encoding, remember it
2007 {
2008 ms_wcCharsetName = nameCS;
2009 ms_wcNeedsSwap = wbuf[0] != (wchar_t)buf[0];
2010 }
2011 }
2012 }
2013 else // use charset not requiring byte swapping
2014 {
2015 ms_wcCharsetName = nameXE;
2016 }
2017 }
2018
2019 wxLogTrace(TRACE_STRCONV,
2020 wxT("iconv wchar_t charset is \"%s\"%s"),
2021 ms_wcCharsetName.empty() ? wxString("<none>")
2022 : ms_wcCharsetName,
2023 ms_wcNeedsSwap ? _T(" (needs swap)")
2024 : _T(""));
2025 }
2026 else // we already have ms_wcCharsetName
2027 {
2028 m2w = iconv_open(ms_wcCharsetName.ToAscii(), name);
2029 }
2030
2031 if ( ms_wcCharsetName.empty() )
2032 {
2033 w2m = ICONV_T_INVALID;
2034 }
2035 else
2036 {
2037 w2m = iconv_open(name, ms_wcCharsetName.ToAscii());
2038 if ( w2m == ICONV_T_INVALID )
2039 {
2040 wxLogTrace(TRACE_STRCONV,
2041 wxT("\"%s\" -> \"%s\" works but not the converse!?"),
2042 ms_wcCharsetName.c_str(), name);
2043 }
2044 }
2045 }
2046
2047 wxMBConv_iconv::~wxMBConv_iconv()
2048 {
2049 if ( m2w != ICONV_T_INVALID )
2050 iconv_close(m2w);
2051 if ( w2m != ICONV_T_INVALID )
2052 iconv_close(w2m);
2053 }
2054
2055 size_t wxMBConv_iconv::MB2WC(wchar_t *buf, const char *psz, size_t n) const
2056 {
2057 // find the string length: notice that must be done differently for
2058 // NUL-terminated strings and UTF-16/32 which are terminated with 2/4 NULs
2059 size_t inbuf;
2060 const size_t nulLen = GetMBNulLen();
2061 switch ( nulLen )
2062 {
2063 default:
2064 return wxCONV_FAILED;
2065
2066 case 1:
2067 inbuf = strlen(psz); // arguably more optimized than our version
2068 break;
2069
2070 case 2:
2071 case 4:
2072 // for UTF-16/32 not only we need to have 2/4 consecutive NULs but
2073 // they also have to start at character boundary and not span two
2074 // adjacent characters
2075 const char *p;
2076 for ( p = psz; NotAllNULs(p, nulLen); p += nulLen )
2077 ;
2078 inbuf = p - psz;
2079 break;
2080 }
2081
2082 #if wxUSE_THREADS
2083 // NB: iconv() is MT-safe, but each thread must use its own iconv_t handle.
2084 // Unfortunately there are a couple of global wxCSConv objects such as
2085 // wxConvLocal that are used all over wx code, so we have to make sure
2086 // the handle is used by at most one thread at the time. Otherwise
2087 // only a few wx classes would be safe to use from non-main threads
2088 // as MB<->WC conversion would fail "randomly".
2089 wxMutexLocker lock(wxConstCast(this, wxMBConv_iconv)->m_iconvMutex);
2090 #endif // wxUSE_THREADS
2091
2092 size_t outbuf = n * SIZEOF_WCHAR_T;
2093 size_t res, cres;
2094 const char *pszPtr = psz;
2095
2096 if (buf)
2097 {
2098 char* bufPtr = (char*)buf;
2099
2100 // have destination buffer, convert there
2101 cres = iconv(m2w,
2102 ICONV_CHAR_CAST(&pszPtr), &inbuf,
2103 &bufPtr, &outbuf);
2104 res = n - (outbuf / SIZEOF_WCHAR_T);
2105
2106 if (ms_wcNeedsSwap)
2107 {
2108 // convert to native endianness
2109 for ( unsigned i = 0; i < res; i++ )
2110 buf[n] = WC_BSWAP(buf[i]);
2111 }
2112
2113 // NUL-terminate the string if there is any space left
2114 if (res < n)
2115 buf[res] = 0;
2116 }
2117 else
2118 {
2119 // no destination buffer... convert using temp buffer
2120 // to calculate destination buffer requirement
2121 wchar_t tbuf[8];
2122 res = 0;
2123
2124 do
2125 {
2126 char* bufPtr = (char*)tbuf;
2127 outbuf = 8 * SIZEOF_WCHAR_T;
2128
2129 cres = iconv(m2w,
2130 ICONV_CHAR_CAST(&pszPtr), &inbuf,
2131 &bufPtr, &outbuf );
2132
2133 res += 8 - (outbuf / SIZEOF_WCHAR_T);
2134 }
2135 while ((cres == (size_t)-1) && (errno == E2BIG));
2136 }
2137
2138 if (ICONV_FAILED(cres, inbuf))
2139 {
2140 //VS: it is ok if iconv fails, hence trace only
2141 wxLogTrace(TRACE_STRCONV, wxT("iconv failed: %s"), wxSysErrorMsg(wxSysErrorCode()));
2142 return wxCONV_FAILED;
2143 }
2144
2145 return res;
2146 }
2147
2148 size_t wxMBConv_iconv::WC2MB(char *buf, const wchar_t *psz, size_t n) const
2149 {
2150 #if wxUSE_THREADS
2151 // NB: explained in MB2WC
2152 wxMutexLocker lock(wxConstCast(this, wxMBConv_iconv)->m_iconvMutex);
2153 #endif
2154
2155 size_t inlen = wxWcslen(psz);
2156 size_t inbuflen = inlen * SIZEOF_WCHAR_T;
2157 size_t outbuflen = n;
2158 size_t res, cres;
2159
2160 wchar_t *tmpbuf = 0;
2161
2162 if (ms_wcNeedsSwap)
2163 {
2164 // need to copy to temp buffer to switch endianness
2165 // (doing WC_BSWAP twice on the original buffer won't help, as it
2166 // could be in read-only memory, or be accessed in some other thread)
2167 tmpbuf = (wchar_t *)malloc(inbuflen + SIZEOF_WCHAR_T);
2168 for ( size_t i = 0; i < inlen; i++ )
2169 tmpbuf[n] = WC_BSWAP(psz[i]);
2170
2171 tmpbuf[inlen] = L'\0';
2172 psz = tmpbuf;
2173 }
2174
2175 char* inbuf = (char*)psz;
2176 if (buf)
2177 {
2178 // have destination buffer, convert there
2179 cres = iconv(w2m, ICONV_CHAR_CAST(&inbuf), &inbuflen, &buf, &outbuflen);
2180
2181 res = n - outbuflen;
2182
2183 // NB: iconv was given only wcslen(psz) characters on input, and so
2184 // it couldn't convert the trailing zero. Let's do it ourselves
2185 // if there's some room left for it in the output buffer.
2186 if (res < n)
2187 buf[0] = 0;
2188 }
2189 else
2190 {
2191 // no destination buffer: convert using temp buffer
2192 // to calculate destination buffer requirement
2193 char tbuf[16];
2194 res = 0;
2195 do
2196 {
2197 buf = tbuf;
2198 outbuflen = 16;
2199
2200 cres = iconv(w2m, ICONV_CHAR_CAST(&inbuf), &inbuflen, &buf, &outbuflen);
2201
2202 res += 16 - outbuflen;
2203 }
2204 while ((cres == (size_t)-1) && (errno == E2BIG));
2205 }
2206
2207 if (ms_wcNeedsSwap)
2208 {
2209 free(tmpbuf);
2210 }
2211
2212 if (ICONV_FAILED(cres, inbuflen))
2213 {
2214 wxLogTrace(TRACE_STRCONV, wxT("iconv failed: %s"), wxSysErrorMsg(wxSysErrorCode()));
2215 return wxCONV_FAILED;
2216 }
2217
2218 return res;
2219 }
2220
2221 size_t wxMBConv_iconv::GetMBNulLen() const
2222 {
2223 if ( m_minMBCharWidth == 0 )
2224 {
2225 wxMBConv_iconv * const self = wxConstCast(this, wxMBConv_iconv);
2226
2227 #if wxUSE_THREADS
2228 // NB: explained in MB2WC
2229 wxMutexLocker lock(self->m_iconvMutex);
2230 #endif
2231
2232 const wchar_t *wnul = L"";
2233 char buf[8]; // should be enough for NUL in any encoding
2234 size_t inLen = sizeof(wchar_t),
2235 outLen = WXSIZEOF(buf);
2236 char *inBuff = (char *)wnul;
2237 char *outBuff = buf;
2238 if ( iconv(w2m, ICONV_CHAR_CAST(&inBuff), &inLen, &outBuff, &outLen) == (size_t)-1 )
2239 {
2240 self->m_minMBCharWidth = (size_t)-1;
2241 }
2242 else // ok
2243 {
2244 self->m_minMBCharWidth = outBuff - buf;
2245 }
2246 }
2247
2248 return m_minMBCharWidth;
2249 }
2250
2251 #if wxUSE_UNICODE_UTF8
2252 bool wxMBConv_iconv::IsUTF8() const
2253 {
2254 return wxStricmp(m_name, "UTF-8") == 0 ||
2255 wxStricmp(m_name, "UTF8") == 0;
2256 }
2257 #endif
2258
2259 #endif // HAVE_ICONV
2260
2261
2262 // ============================================================================
2263 // Win32 conversion classes
2264 // ============================================================================
2265
2266 #ifdef wxHAVE_WIN32_MB2WC
2267
2268 // from utils.cpp
2269 #if wxUSE_FONTMAP
2270 extern WXDLLIMPEXP_BASE long wxCharsetToCodepage(const char *charset);
2271 extern WXDLLIMPEXP_BASE long wxEncodingToCodepage(wxFontEncoding encoding);
2272 #endif
2273
2274 class wxMBConv_win32 : public wxMBConv
2275 {
2276 public:
2277 wxMBConv_win32()
2278 {
2279 m_CodePage = CP_ACP;
2280 m_minMBCharWidth = 0;
2281 }
2282
2283 wxMBConv_win32(const wxMBConv_win32& conv)
2284 : wxMBConv()
2285 {
2286 m_CodePage = conv.m_CodePage;
2287 m_minMBCharWidth = conv.m_minMBCharWidth;
2288 }
2289
2290 #if wxUSE_FONTMAP
2291 wxMBConv_win32(const char* name)
2292 {
2293 m_CodePage = wxCharsetToCodepage(name);
2294 m_minMBCharWidth = 0;
2295 }
2296
2297 wxMBConv_win32(wxFontEncoding encoding)
2298 {
2299 m_CodePage = wxEncodingToCodepage(encoding);
2300 m_minMBCharWidth = 0;
2301 }
2302 #endif // wxUSE_FONTMAP
2303
2304 virtual size_t MB2WC(wchar_t *buf, const char *psz, size_t n) const
2305 {
2306 // note that we have to use MB_ERR_INVALID_CHARS flag as it without it
2307 // the behaviour is not compatible with the Unix version (using iconv)
2308 // and break the library itself, e.g. wxTextInputStream::NextChar()
2309 // wouldn't work if reading an incomplete MB char didn't result in an
2310 // error
2311 //
2312 // Moreover, MB_ERR_INVALID_CHARS is only supported on Win 2K SP4 or
2313 // Win XP or newer and it is not supported for UTF-[78] so we always
2314 // use our own conversions in this case. See
2315 // http://blogs.msdn.com/michkap/archive/2005/04/19/409566.aspx
2316 // http://msdn.microsoft.com/library/en-us/intl/unicode_17si.asp
2317 if ( m_CodePage == CP_UTF8 )
2318 {
2319 return wxMBConvUTF8().MB2WC(buf, psz, n);
2320 }
2321
2322 if ( m_CodePage == CP_UTF7 )
2323 {
2324 return wxMBConvUTF7().MB2WC(buf, psz, n);
2325 }
2326
2327 int flags = 0;
2328 if ( (m_CodePage < 50000 && m_CodePage != CP_SYMBOL) &&
2329 IsAtLeastWin2kSP4() )
2330 {
2331 flags = MB_ERR_INVALID_CHARS;
2332 }
2333
2334 const size_t len = ::MultiByteToWideChar
2335 (
2336 m_CodePage, // code page
2337 flags, // flags: fall on error
2338 psz, // input string
2339 -1, // its length (NUL-terminated)
2340 buf, // output string
2341 buf ? n : 0 // size of output buffer
2342 );
2343 if ( !len )
2344 {
2345 // function totally failed
2346 return wxCONV_FAILED;
2347 }
2348
2349 // if we were really converting and didn't use MB_ERR_INVALID_CHARS,
2350 // check if we succeeded, by doing a double trip:
2351 if ( !flags && buf )
2352 {
2353 const size_t mbLen = strlen(psz);
2354 wxCharBuffer mbBuf(mbLen);
2355 if ( ::WideCharToMultiByte
2356 (
2357 m_CodePage,
2358 0,
2359 buf,
2360 -1,
2361 mbBuf.data(),
2362 mbLen + 1, // size in bytes, not length
2363 NULL,
2364 NULL
2365 ) == 0 ||
2366 strcmp(mbBuf, psz) != 0 )
2367 {
2368 // we didn't obtain the same thing we started from, hence
2369 // the conversion was lossy and we consider that it failed
2370 return wxCONV_FAILED;
2371 }
2372 }
2373
2374 // note that it returns count of written chars for buf != NULL and size
2375 // of the needed buffer for buf == NULL so in either case the length of
2376 // the string (which never includes the terminating NUL) is one less
2377 return len - 1;
2378 }
2379
2380 virtual size_t WC2MB(char *buf, const wchar_t *pwz, size_t n) const
2381 {
2382 /*
2383 we have a problem here: by default, WideCharToMultiByte() may
2384 replace characters unrepresentable in the target code page with bad
2385 quality approximations such as turning "1/2" symbol (U+00BD) into
2386 "1" for the code pages which don't have it and we, obviously, want
2387 to avoid this at any price
2388
2389 the trouble is that this function does it _silently_, i.e. it won't
2390 even tell us whether it did or not... Win98/2000 and higher provide
2391 WC_NO_BEST_FIT_CHARS but it doesn't work for the older systems and
2392 we have to resort to a round trip, i.e. check that converting back
2393 results in the same string -- this is, of course, expensive but
2394 otherwise we simply can't be sure to not garble the data.
2395 */
2396
2397 // determine if we can rely on WC_NO_BEST_FIT_CHARS: according to MSDN
2398 // it doesn't work with CJK encodings (which we test for rather roughly
2399 // here...) nor with UTF-7/8 nor, of course, with Windows versions not
2400 // supporting it
2401 BOOL usedDef wxDUMMY_INITIALIZE(false);
2402 BOOL *pUsedDef;
2403 int flags;
2404 if ( CanUseNoBestFit() && m_CodePage < 50000 )
2405 {
2406 // it's our lucky day
2407 flags = WC_NO_BEST_FIT_CHARS;
2408 pUsedDef = &usedDef;
2409 }
2410 else // old system or unsupported encoding
2411 {
2412 flags = 0;
2413 pUsedDef = NULL;
2414 }
2415
2416 const size_t len = ::WideCharToMultiByte
2417 (
2418 m_CodePage, // code page
2419 flags, // either none or no best fit
2420 pwz, // input string
2421 -1, // it is (wide) NUL-terminated
2422 buf, // output buffer
2423 buf ? n : 0, // and its size
2424 NULL, // default "replacement" char
2425 pUsedDef // [out] was it used?
2426 );
2427
2428 if ( !len )
2429 {
2430 // function totally failed
2431 return wxCONV_FAILED;
2432 }
2433
2434 // we did something, check if we really succeeded
2435 if ( flags )
2436 {
2437 // check if the conversion failed, i.e. if any replacements
2438 // were done
2439 if ( usedDef )
2440 return wxCONV_FAILED;
2441 }
2442 else // we must resort to double tripping...
2443 {
2444 // first we need to ensure that we really have the MB data: this is
2445 // not the case if we're called with NULL buffer, in which case we
2446 // need to do the conversion yet again
2447 wxCharBuffer bufDef;
2448 if ( !buf )
2449 {
2450 bufDef = wxCharBuffer(len);
2451 buf = bufDef.data();
2452 if ( !::WideCharToMultiByte(m_CodePage, flags, pwz, -1,
2453 buf, len, NULL, NULL) )
2454 return wxCONV_FAILED;
2455 }
2456
2457 if ( !n )
2458 n = wcslen(pwz);
2459 wxWCharBuffer wcBuf(n);
2460 if ( MB2WC(wcBuf.data(), buf, n + 1) == wxCONV_FAILED ||
2461 wcscmp(wcBuf, pwz) != 0 )
2462 {
2463 // we didn't obtain the same thing we started from, hence
2464 // the conversion was lossy and we consider that it failed
2465 return wxCONV_FAILED;
2466 }
2467 }
2468
2469 // see the comment above for the reason of "len - 1"
2470 return len - 1;
2471 }
2472
2473 virtual size_t GetMBNulLen() const
2474 {
2475 if ( m_minMBCharWidth == 0 )
2476 {
2477 int len = ::WideCharToMultiByte
2478 (
2479 m_CodePage, // code page
2480 0, // no flags
2481 L"", // input string
2482 1, // translate just the NUL
2483 NULL, // output buffer
2484 0, // and its size
2485 NULL, // no replacement char
2486 NULL // [out] don't care if it was used
2487 );
2488
2489 wxMBConv_win32 * const self = wxConstCast(this, wxMBConv_win32);
2490 switch ( len )
2491 {
2492 default:
2493 wxLogDebug(_T("Unexpected NUL length %d"), len);
2494 self->m_minMBCharWidth = (size_t)-1;
2495 break;
2496
2497 case 0:
2498 self->m_minMBCharWidth = (size_t)-1;
2499 break;
2500
2501 case 1:
2502 case 2:
2503 case 4:
2504 self->m_minMBCharWidth = len;
2505 break;
2506 }
2507 }
2508
2509 return m_minMBCharWidth;
2510 }
2511
2512 virtual wxMBConv *Clone() const { return new wxMBConv_win32(*this); }
2513
2514 bool IsOk() const { return m_CodePage != -1; }
2515
2516 private:
2517 static bool CanUseNoBestFit()
2518 {
2519 static int s_isWin98Or2k = -1;
2520
2521 if ( s_isWin98Or2k == -1 )
2522 {
2523 int verMaj, verMin;
2524 switch ( wxGetOsVersion(&verMaj, &verMin) )
2525 {
2526 case wxOS_WINDOWS_9X:
2527 s_isWin98Or2k = verMaj >= 4 && verMin >= 10;
2528 break;
2529
2530 case wxOS_WINDOWS_NT:
2531 s_isWin98Or2k = verMaj >= 5;
2532 break;
2533
2534 default:
2535 // unknown: be conservative by default
2536 s_isWin98Or2k = 0;
2537 break;
2538 }
2539
2540 wxASSERT_MSG( s_isWin98Or2k != -1, _T("should be set above") );
2541 }
2542
2543 return s_isWin98Or2k == 1;
2544 }
2545
2546 static bool IsAtLeastWin2kSP4()
2547 {
2548 #ifdef __WXWINCE__
2549 return false;
2550 #else
2551 static int s_isAtLeastWin2kSP4 = -1;
2552
2553 if ( s_isAtLeastWin2kSP4 == -1 )
2554 {
2555 OSVERSIONINFOEX ver;
2556
2557 memset(&ver, 0, sizeof(ver));
2558 ver.dwOSVersionInfoSize = sizeof(ver);
2559 GetVersionEx((OSVERSIONINFO*)&ver);
2560
2561 s_isAtLeastWin2kSP4 =
2562 ((ver.dwMajorVersion > 5) || // Vista+
2563 (ver.dwMajorVersion == 5 && ver.dwMinorVersion > 0) || // XP/2003
2564 (ver.dwMajorVersion == 5 && ver.dwMinorVersion == 0 &&
2565 ver.wServicePackMajor >= 4)) // 2000 SP4+
2566 ? 1 : 0;
2567 }
2568
2569 return s_isAtLeastWin2kSP4 == 1;
2570 #endif
2571 }
2572
2573
2574 // the code page we're working with
2575 long m_CodePage;
2576
2577 // cached result of GetMBNulLen(), set to 0 initially meaning
2578 // "unknown"
2579 size_t m_minMBCharWidth;
2580 };
2581
2582 #endif // wxHAVE_WIN32_MB2WC
2583
2584
2585 // ============================================================================
2586 // wxEncodingConverter based conversion classes
2587 // ============================================================================
2588
2589 #if wxUSE_FONTMAP
2590
2591 class wxMBConv_wxwin : public wxMBConv
2592 {
2593 private:
2594 void Init()
2595 {
2596 // Refuse to use broken wxEncodingConverter code for Mac-specific encodings.
2597 // The wxMBConv_cf class does a better job.
2598 m_ok = (m_enc < wxFONTENCODING_MACMIN || m_enc > wxFONTENCODING_MACMAX) &&
2599 m2w.Init(m_enc, wxFONTENCODING_UNICODE) &&
2600 w2m.Init(wxFONTENCODING_UNICODE, m_enc);
2601 }
2602
2603 public:
2604 // temporarily just use wxEncodingConverter stuff,
2605 // so that it works while a better implementation is built
2606 wxMBConv_wxwin(const char* name)
2607 {
2608 if (name)
2609 m_enc = wxFontMapperBase::Get()->CharsetToEncoding(name, false);
2610 else
2611 m_enc = wxFONTENCODING_SYSTEM;
2612
2613 Init();
2614 }
2615
2616 wxMBConv_wxwin(wxFontEncoding enc)
2617 {
2618 m_enc = enc;
2619
2620 Init();
2621 }
2622
2623 size_t MB2WC(wchar_t *buf, const char *psz, size_t WXUNUSED(n)) const
2624 {
2625 size_t inbuf = strlen(psz);
2626 if (buf)
2627 {
2628 if (!m2w.Convert(psz, buf))
2629 return wxCONV_FAILED;
2630 }
2631 return inbuf;
2632 }
2633
2634 size_t WC2MB(char *buf, const wchar_t *psz, size_t WXUNUSED(n)) const
2635 {
2636 const size_t inbuf = wxWcslen(psz);
2637 if (buf)
2638 {
2639 if (!w2m.Convert(psz, buf))
2640 return wxCONV_FAILED;
2641 }
2642
2643 return inbuf;
2644 }
2645
2646 virtual size_t GetMBNulLen() const
2647 {
2648 switch ( m_enc )
2649 {
2650 case wxFONTENCODING_UTF16BE:
2651 case wxFONTENCODING_UTF16LE:
2652 return 2;
2653
2654 case wxFONTENCODING_UTF32BE:
2655 case wxFONTENCODING_UTF32LE:
2656 return 4;
2657
2658 default:
2659 return 1;
2660 }
2661 }
2662
2663 virtual wxMBConv *Clone() const { return new wxMBConv_wxwin(m_enc); }
2664
2665 bool IsOk() const { return m_ok; }
2666
2667 public:
2668 wxFontEncoding m_enc;
2669 wxEncodingConverter m2w, w2m;
2670
2671 private:
2672 // were we initialized successfully?
2673 bool m_ok;
2674
2675 DECLARE_NO_COPY_CLASS(wxMBConv_wxwin)
2676 };
2677
2678 // make the constructors available for unit testing
2679 WXDLLIMPEXP_BASE wxMBConv* new_wxMBConv_wxwin( const char* name )
2680 {
2681 wxMBConv_wxwin* result = new wxMBConv_wxwin( name );
2682 if ( !result->IsOk() )
2683 {
2684 delete result;
2685 return 0;
2686 }
2687
2688 return result;
2689 }
2690
2691 #endif // wxUSE_FONTMAP
2692
2693 // ============================================================================
2694 // wxCSConv implementation
2695 // ============================================================================
2696
2697 void wxCSConv::Init()
2698 {
2699 m_name = NULL;
2700 m_convReal = NULL;
2701 m_deferred = true;
2702 }
2703
2704 wxCSConv::wxCSConv(const wxString& charset)
2705 {
2706 Init();
2707
2708 if ( !charset.empty() )
2709 {
2710 SetName(charset.ToAscii());
2711 }
2712
2713 #if wxUSE_FONTMAP
2714 m_encoding = wxFontMapperBase::GetEncodingFromName(charset);
2715 #else
2716 m_encoding = wxFONTENCODING_SYSTEM;
2717 #endif
2718 }
2719
2720 wxCSConv::wxCSConv(wxFontEncoding encoding)
2721 {
2722 if ( encoding == wxFONTENCODING_MAX || encoding == wxFONTENCODING_DEFAULT )
2723 {
2724 wxFAIL_MSG( _T("invalid encoding value in wxCSConv ctor") );
2725
2726 encoding = wxFONTENCODING_SYSTEM;
2727 }
2728
2729 Init();
2730
2731 m_encoding = encoding;
2732 }
2733
2734 wxCSConv::~wxCSConv()
2735 {
2736 Clear();
2737 }
2738
2739 wxCSConv::wxCSConv(const wxCSConv& conv)
2740 : wxMBConv()
2741 {
2742 Init();
2743
2744 SetName(conv.m_name);
2745 m_encoding = conv.m_encoding;
2746 }
2747
2748 wxCSConv& wxCSConv::operator=(const wxCSConv& conv)
2749 {
2750 Clear();
2751
2752 SetName(conv.m_name);
2753 m_encoding = conv.m_encoding;
2754
2755 return *this;
2756 }
2757
2758 void wxCSConv::Clear()
2759 {
2760 free(m_name);
2761 delete m_convReal;
2762
2763 m_name = NULL;
2764 m_convReal = NULL;
2765 }
2766
2767 void wxCSConv::SetName(const char *charset)
2768 {
2769 if (charset)
2770 {
2771 m_name = wxStrdup(charset);
2772 m_deferred = true;
2773 }
2774 }
2775
2776 #if wxUSE_FONTMAP
2777
2778 WX_DECLARE_HASH_MAP( wxFontEncoding, wxString, wxIntegerHash, wxIntegerEqual,
2779 wxEncodingNameCache );
2780
2781 static wxEncodingNameCache gs_nameCache;
2782 #endif
2783
2784 wxMBConv *wxCSConv::DoCreate() const
2785 {
2786 #if wxUSE_FONTMAP
2787 wxLogTrace(TRACE_STRCONV,
2788 wxT("creating conversion for %s"),
2789 (m_name ? m_name
2790 : (const char*)wxFontMapperBase::GetEncodingName(m_encoding).mb_str()));
2791 #endif // wxUSE_FONTMAP
2792
2793 // check for the special case of ASCII or ISO8859-1 charset: as we have
2794 // special knowledge of it anyhow, we don't need to create a special
2795 // conversion object
2796 if ( m_encoding == wxFONTENCODING_ISO8859_1 ||
2797 m_encoding == wxFONTENCODING_DEFAULT )
2798 {
2799 // don't convert at all
2800 return NULL;
2801 }
2802
2803 // we trust OS to do conversion better than we can so try external
2804 // conversion methods first
2805 //
2806 // the full order is:
2807 // 1. OS conversion (iconv() under Unix or Win32 API)
2808 // 2. hard coded conversions for UTF
2809 // 3. wxEncodingConverter as fall back
2810
2811 // step (1)
2812 #ifdef HAVE_ICONV
2813 #if !wxUSE_FONTMAP
2814 if ( m_name )
2815 #endif // !wxUSE_FONTMAP
2816 {
2817 #if wxUSE_FONTMAP
2818 wxFontEncoding encoding(m_encoding);
2819 #endif
2820
2821 if ( m_name )
2822 {
2823 wxMBConv_iconv *conv = new wxMBConv_iconv(m_name);
2824 if ( conv->IsOk() )
2825 return conv;
2826
2827 delete conv;
2828
2829 #if wxUSE_FONTMAP
2830 encoding =
2831 wxFontMapperBase::Get()->CharsetToEncoding(m_name, false);
2832 #endif // wxUSE_FONTMAP
2833 }
2834 #if wxUSE_FONTMAP
2835 {
2836 const wxEncodingNameCache::iterator it = gs_nameCache.find(encoding);
2837 if ( it != gs_nameCache.end() )
2838 {
2839 if ( it->second.empty() )
2840 return NULL;
2841
2842 wxMBConv_iconv *conv = new wxMBConv_iconv(it->second.ToAscii());
2843 if ( conv->IsOk() )
2844 return conv;
2845
2846 delete conv;
2847 }
2848
2849 const wxChar** names = wxFontMapperBase::GetAllEncodingNames(encoding);
2850 // CS : in case this does not return valid names (eg for MacRoman)
2851 // encoding got a 'failure' entry in the cache all the same,
2852 // although it just has to be created using a different method, so
2853 // only store failed iconv creation attempts (or perhaps we
2854 // shoulnd't do this at all ?)
2855 if ( names[0] != NULL )
2856 {
2857 for ( ; *names; ++names )
2858 {
2859 // FIXME-UTF8: wxFontMapperBase::GetAllEncodingNames()
2860 // will need changes that will obsolete this
2861 wxString name(*names);
2862 wxMBConv_iconv *conv = new wxMBConv_iconv(name.ToAscii());
2863 if ( conv->IsOk() )
2864 {
2865 gs_nameCache[encoding] = *names;
2866 return conv;
2867 }
2868
2869 delete conv;
2870 }
2871
2872 gs_nameCache[encoding] = _T(""); // cache the failure
2873 }
2874 }
2875 #endif // wxUSE_FONTMAP
2876 }
2877 #endif // HAVE_ICONV
2878
2879 #ifdef wxHAVE_WIN32_MB2WC
2880 {
2881 #if wxUSE_FONTMAP
2882 wxMBConv_win32 *conv = m_name ? new wxMBConv_win32(m_name)
2883 : new wxMBConv_win32(m_encoding);
2884 if ( conv->IsOk() )
2885 return conv;
2886
2887 delete conv;
2888 #else
2889 return NULL;
2890 #endif
2891 }
2892 #endif // wxHAVE_WIN32_MB2WC
2893
2894 #ifdef __DARWIN__
2895 {
2896 // leave UTF16 and UTF32 to the built-ins of wx
2897 if ( m_name || ( m_encoding < wxFONTENCODING_UTF16BE ||
2898 ( m_encoding >= wxFONTENCODING_MACMIN && m_encoding <= wxFONTENCODING_MACMAX ) ) )
2899 {
2900 #if wxUSE_FONTMAP
2901 wxMBConv_cf *conv = m_name ? new wxMBConv_cf(m_name)
2902 : new wxMBConv_cf(m_encoding);
2903 #else
2904 wxMBConv_cf *conv = new wxMBConv_cf(m_encoding);
2905 #endif
2906
2907 if ( conv->IsOk() )
2908 return conv;
2909
2910 delete conv;
2911 }
2912 }
2913 #endif // __DARWIN__
2914
2915 // step (2)
2916 wxFontEncoding enc = m_encoding;
2917 #if wxUSE_FONTMAP
2918 if ( enc == wxFONTENCODING_SYSTEM && m_name )
2919 {
2920 // use "false" to suppress interactive dialogs -- we can be called from
2921 // anywhere and popping up a dialog from here is the last thing we want to
2922 // do
2923 enc = wxFontMapperBase::Get()->CharsetToEncoding(m_name, false);
2924 }
2925 #endif // wxUSE_FONTMAP
2926
2927 switch ( enc )
2928 {
2929 case wxFONTENCODING_UTF7:
2930 return new wxMBConvUTF7;
2931
2932 case wxFONTENCODING_UTF8:
2933 return new wxMBConvUTF8;
2934
2935 case wxFONTENCODING_UTF16BE:
2936 return new wxMBConvUTF16BE;
2937
2938 case wxFONTENCODING_UTF16LE:
2939 return new wxMBConvUTF16LE;
2940
2941 case wxFONTENCODING_UTF32BE:
2942 return new wxMBConvUTF32BE;
2943
2944 case wxFONTENCODING_UTF32LE:
2945 return new wxMBConvUTF32LE;
2946
2947 default:
2948 // nothing to do but put here to suppress gcc warnings
2949 break;
2950 }
2951
2952 // step (3)
2953 #if wxUSE_FONTMAP
2954 {
2955 wxMBConv_wxwin *conv = m_name ? new wxMBConv_wxwin(m_name)
2956 : new wxMBConv_wxwin(m_encoding);
2957 if ( conv->IsOk() )
2958 return conv;
2959
2960 delete conv;
2961 }
2962 #endif // wxUSE_FONTMAP
2963
2964 // NB: This is a hack to prevent deadlock. What could otherwise happen
2965 // in Unicode build: wxConvLocal creation ends up being here
2966 // because of some failure and logs the error. But wxLog will try to
2967 // attach a timestamp, for which it will need wxConvLocal (to convert
2968 // time to char* and then wchar_t*), but that fails, tries to log the
2969 // error, but wxLog has an (already locked) critical section that
2970 // guards the static buffer.
2971 static bool alreadyLoggingError = false;
2972 if (!alreadyLoggingError)
2973 {
2974 alreadyLoggingError = true;
2975 wxLogError(_("Cannot convert from the charset '%s'!"),
2976 m_name ? m_name
2977 :
2978 #if wxUSE_FONTMAP
2979 (const char*)wxFontMapperBase::GetEncodingDescription(m_encoding).ToAscii()
2980 #else // !wxUSE_FONTMAP
2981 (const char*)wxString::Format(_("encoding %i"), m_encoding).ToAscii()
2982 #endif // wxUSE_FONTMAP/!wxUSE_FONTMAP
2983 );
2984
2985 alreadyLoggingError = false;
2986 }
2987
2988 return NULL;
2989 }
2990
2991 void wxCSConv::CreateConvIfNeeded() const
2992 {
2993 if ( m_deferred )
2994 {
2995 wxCSConv *self = (wxCSConv *)this; // const_cast
2996
2997 // if we don't have neither the name nor the encoding, use the default
2998 // encoding for this system
2999 if ( !m_name && m_encoding == wxFONTENCODING_SYSTEM )
3000 {
3001 #if wxUSE_INTL
3002 self->m_encoding = wxLocale::GetSystemEncoding();
3003 #else
3004 // fallback to some reasonable default:
3005 self->m_encoding = wxFONTENCODING_ISO8859_1;
3006 #endif // wxUSE_INTL
3007 }
3008
3009 self->m_convReal = DoCreate();
3010 self->m_deferred = false;
3011 }
3012 }
3013
3014 bool wxCSConv::IsOk() const
3015 {
3016 CreateConvIfNeeded();
3017
3018 // special case: no convReal created for wxFONTENCODING_ISO8859_1
3019 if ( m_encoding == wxFONTENCODING_ISO8859_1 )
3020 return true; // always ok as we do it ourselves
3021
3022 // m_convReal->IsOk() is called at its own creation, so we know it must
3023 // be ok if m_convReal is non-NULL
3024 return m_convReal != NULL;
3025 }
3026
3027 size_t wxCSConv::ToWChar(wchar_t *dst, size_t dstLen,
3028 const char *src, size_t srcLen) const
3029 {
3030 CreateConvIfNeeded();
3031
3032 if (m_convReal)
3033 return m_convReal->ToWChar(dst, dstLen, src, srcLen);
3034
3035 // latin-1 (direct)
3036 return wxMBConv::ToWChar(dst, dstLen, src, srcLen);
3037 }
3038
3039 size_t wxCSConv::FromWChar(char *dst, size_t dstLen,
3040 const wchar_t *src, size_t srcLen) const
3041 {
3042 CreateConvIfNeeded();
3043
3044 if (m_convReal)
3045 return m_convReal->FromWChar(dst, dstLen, src, srcLen);
3046
3047 // latin-1 (direct)
3048 return wxMBConv::FromWChar(dst, dstLen, src, srcLen);
3049 }
3050
3051 size_t wxCSConv::MB2WC(wchar_t *buf, const char *psz, size_t n) const
3052 {
3053 CreateConvIfNeeded();
3054
3055 if (m_convReal)
3056 return m_convReal->MB2WC(buf, psz, n);
3057
3058 // latin-1 (direct)
3059 size_t len = strlen(psz);
3060
3061 if (buf)
3062 {
3063 for (size_t c = 0; c <= len; c++)
3064 buf[c] = (unsigned char)(psz[c]);
3065 }
3066
3067 return len;
3068 }
3069
3070 size_t wxCSConv::WC2MB(char *buf, const wchar_t *psz, size_t n) const
3071 {
3072 CreateConvIfNeeded();
3073
3074 if (m_convReal)
3075 return m_convReal->WC2MB(buf, psz, n);
3076
3077 // latin-1 (direct)
3078 const size_t len = wxWcslen(psz);
3079 if (buf)
3080 {
3081 for (size_t c = 0; c <= len; c++)
3082 {
3083 if (psz[c] > 0xFF)
3084 return wxCONV_FAILED;
3085
3086 buf[c] = (char)psz[c];
3087 }
3088 }
3089 else
3090 {
3091 for (size_t c = 0; c <= len; c++)
3092 {
3093 if (psz[c] > 0xFF)
3094 return wxCONV_FAILED;
3095 }
3096 }
3097
3098 return len;
3099 }
3100
3101 size_t wxCSConv::GetMBNulLen() const
3102 {
3103 CreateConvIfNeeded();
3104
3105 if ( m_convReal )
3106 {
3107 return m_convReal->GetMBNulLen();
3108 }
3109
3110 // otherwise, we are ISO-8859-1
3111 return 1;
3112 }
3113
3114 #if wxUSE_UNICODE_UTF8
3115 bool wxCSConv::IsUTF8() const
3116 {
3117 CreateConvIfNeeded();
3118
3119 if ( m_convReal )
3120 {
3121 return m_convReal->IsUTF8();
3122 }
3123
3124 // otherwise, we are ISO-8859-1
3125 return false;
3126 }
3127 #endif
3128
3129
3130 #if wxUSE_UNICODE
3131
3132 wxWCharBuffer wxSafeConvertMB2WX(const char *s)
3133 {
3134 if ( !s )
3135 return wxWCharBuffer();
3136
3137 wxWCharBuffer wbuf(wxConvLibc.cMB2WX(s));
3138 if ( !wbuf )
3139 wbuf = wxMBConvUTF8().cMB2WX(s);
3140 if ( !wbuf )
3141 wbuf = wxConvISO8859_1.cMB2WX(s);
3142
3143 return wbuf;
3144 }
3145
3146 wxCharBuffer wxSafeConvertWX2MB(const wchar_t *ws)
3147 {
3148 if ( !ws )
3149 return wxCharBuffer();
3150
3151 wxCharBuffer buf(wxConvLibc.cWX2MB(ws));
3152 if ( !buf )
3153 buf = wxMBConvUTF8(wxMBConvUTF8::MAP_INVALID_UTF8_TO_OCTAL).cWX2MB(ws);
3154
3155 return buf;
3156 }
3157
3158 #endif // wxUSE_UNICODE
3159
3160 // ----------------------------------------------------------------------------
3161 // globals
3162 // ----------------------------------------------------------------------------
3163
3164 // NB: The reason why we create converted objects in this convoluted way,
3165 // using a factory function instead of global variable, is that they
3166 // may be used at static initialization time (some of them are used by
3167 // wxString ctors and there may be a global wxString object). In other
3168 // words, possibly _before_ the converter global object would be
3169 // initialized.
3170
3171 #undef wxConvLibc
3172 #undef wxConvUTF8
3173 #undef wxConvUTF7
3174 #undef wxConvLocal
3175 #undef wxConvISO8859_1
3176
3177 #define WX_DEFINE_GLOBAL_CONV2(klass, impl_klass, name, ctor_args) \
3178 WXDLLIMPEXP_DATA_BASE(klass*) name##Ptr = NULL; \
3179 WXDLLIMPEXP_BASE klass* wxGet_##name##Ptr() \
3180 { \
3181 static impl_klass name##Obj ctor_args; \
3182 return &name##Obj; \
3183 } \
3184 /* this ensures that all global converter objects are created */ \
3185 /* by the time static initialization is done, i.e. before any */ \
3186 /* thread is launched: */ \
3187 static klass* gs_##name##instance = wxGet_##name##Ptr()
3188
3189 #define WX_DEFINE_GLOBAL_CONV(klass, name, ctor_args) \
3190 WX_DEFINE_GLOBAL_CONV2(klass, klass, name, ctor_args)
3191
3192 #ifdef __WINDOWS__
3193 WX_DEFINE_GLOBAL_CONV2(wxMBConv, wxMBConv_win32, wxConvLibc, wxEMPTY_PARAMETER_VALUE);
3194 #else
3195 WX_DEFINE_GLOBAL_CONV2(wxMBConv, wxMBConvLibc, wxConvLibc, wxEMPTY_PARAMETER_VALUE);
3196 #endif
3197
3198 // NB: we can't use wxEMPTY_PARAMETER_VALUE as final argument here because it's
3199 // passed to WX_DEFINE_GLOBAL_CONV2 after a macro expansion and so still
3200 // provokes an error message about "not enough macro parameters"; and we
3201 // can't use "()" here as the name##Obj declaration would be parsed as a
3202 // function declaration then, so use a semicolon and live with an extra
3203 // empty statement (and hope that no compilers warns about this)
3204 WX_DEFINE_GLOBAL_CONV(wxMBConvStrictUTF8, wxConvUTF8, ;);
3205 WX_DEFINE_GLOBAL_CONV(wxMBConvUTF7, wxConvUTF7, ;);
3206
3207 WX_DEFINE_GLOBAL_CONV(wxCSConv, wxConvLocal, (wxFONTENCODING_SYSTEM));
3208 WX_DEFINE_GLOBAL_CONV(wxCSConv, wxConvISO8859_1, (wxFONTENCODING_ISO8859_1));
3209
3210 WXDLLIMPEXP_DATA_BASE(wxMBConv *) wxConvCurrent = wxGet_wxConvLibcPtr();
3211 WXDLLIMPEXP_DATA_BASE(wxMBConv *) wxConvUI = wxGet_wxConvLocalPtr();
3212
3213 #ifdef __DARWIN__
3214 // The xnu kernel always communicates file paths in decomposed UTF-8.
3215 // WARNING: Are we sure that CFString's conversion will cause decomposition?
3216 static wxMBConv_cf wxConvMacUTF8DObj(wxFONTENCODING_UTF8);
3217 #endif
3218
3219 WXDLLIMPEXP_DATA_BASE(wxMBConv *) wxConvFileName =
3220 #ifdef __DARWIN__
3221 &wxConvMacUTF8DObj;
3222 #else // !__DARWIN__
3223 wxGet_wxConvLibcPtr();
3224 #endif // __DARWIN__/!__DARWIN__
3225
3226 #else // !wxUSE_WCHAR_T
3227
3228 // FIXME-UTF8: remove this, wxUSE_WCHAR_T is required now
3229 // stand-ins in absence of wchar_t
3230 WXDLLIMPEXP_DATA_BASE(wxMBConv) wxConvLibc,
3231 wxConvISO8859_1,
3232 wxConvLocal,
3233 wxConvUTF8;
3234
3235 #endif // wxUSE_WCHAR_T/!wxUSE_WCHAR_T
wxWidgets (Impactor)