X-Git-Url: https://git.saurik.com/wxWidgets.git/blobdiff_plain/c543817b3280c560d53307c0fa98e7a99aa00bf9..5ac1d44aa499901b1a2e194292ba7a93112196ae:/src/common/strconv.cpp diff --git a/src/common/strconv.cpp b/src/common/strconv.cpp index c75888b960..742feb387f 100644 --- a/src/common/strconv.cpp +++ b/src/common/strconv.cpp @@ -1,5 +1,5 @@ ///////////////////////////////////////////////////////////////////////////// -// Name: strconv.cpp +// Name: src/common/strconv.cpp // Purpose: Unicode conversion classes // Author: Ove Kaaven, Robert Roebling, Vadim Zeitlin, Vaclav Slavik, // Ryan Norton, Fredrik Roubert (UTF7) @@ -12,39 +12,22 @@ // Licence: wxWindows licence ///////////////////////////////////////////////////////////////////////////// -// ============================================================================ -// declarations -// ============================================================================ - -// ---------------------------------------------------------------------------- -// headers -// ---------------------------------------------------------------------------- - -#if defined(__GNUG__) && !defined(NO_GCC_PRAGMA) - #pragma implementation "strconv.h" -#endif - // For compilers that support precompilation, includes "wx.h". #include "wx/wxprec.h" -#ifdef __BORLANDC__ - #pragma hdrstop -#endif - #ifndef WX_PRECOMP #include "wx/intl.h" #include "wx/log.h" -#endif // WX_PRECOMP + #include "wx/utils.h" + #include "wx/hashmap.h" +#endif #include "wx/strconv.h" #if wxUSE_WCHAR_T -#ifdef __WXMSW__ - #include "wx/msw/private.h" -#endif - #ifdef __WINDOWS__ + #include "wx/msw/private.h" #include "wx/msw/missing.h" #endif @@ -58,11 +41,7 @@ #if defined(__WIN32__) && !defined(__WXMICROWIN__) #define wxHAVE_WIN32_MB2WC -#endif // __WIN32__ but !__WXMICROWIN__ - -// ---------------------------------------------------------------------------- -// headers -// ---------------------------------------------------------------------------- +#endif #ifdef __SALFORDC__ #include @@ -70,91 +49,86 @@ #ifdef HAVE_ICONV #include + #include "wx/thread.h" #endif #include "wx/encconv.h" #include "wx/fontmap.h" -#include "wx/utils.h" #ifdef __WXMAC__ +#ifndef __DARWIN__ #include #include #include +#endif -#include "wx/mac/private.h" // includes mac headers +// includes Mac headers +#include "wx/mac/private.h" #endif -// ---------------------------------------------------------------------------- -// macros -// ---------------------------------------------------------------------------- -#define BSWAP_UCS4(str, len) { unsigned _c; for (_c=0; _c=0x110000) + else if (input >= 0x110000) { - return (size_t)-1; + return wxCONV_FAILED; } else { if (output) { - *output++ = (wxUint16) ((input >> 10)+0xd7c0); - *output = (wxUint16) ((input&0x3ff)+0xdc00); + *output++ = (wxUint16) ((input >> 10) + 0xd7c0); + *output = (wxUint16) ((input & 0x3ff) + 0xdc00); } + return 2; } } static size_t decode_utf16(const wxUint16* input, wxUint32& output) { - if ((*input<0xd800) || (*input>0xdfff)) + if ((*input < 0xd800) || (*input > 0xdfff)) { output = *input; return 1; } - else if ((input[1]<0xdc00) || (input[1]>=0xdfff)) + else if ((input[1] < 0xdc00) || (input[1] > 0xdfff)) { output = *input; - return (size_t)-1; + return wxCONV_FAILED; } else { @@ -163,11 +137,219 @@ static size_t decode_utf16(const wxUint16* input, wxUint32& output) } } +#ifdef WC_UTF16 + typedef wchar_t wxDecodeSurrogate_t; +#else // !WC_UTF16 + typedef wxUint16 wxDecodeSurrogate_t; +#endif // WC_UTF16/!WC_UTF16 + +// returns the next UTF-32 character from the wchar_t buffer and advances the +// pointer to the character after this one +// +// if an invalid character is found, *pSrc is set to NULL, the caller must +// check for this +static wxUint32 wxDecodeSurrogate(const wxDecodeSurrogate_t **pSrc) +{ + wxUint32 out; + const size_t + n = decode_utf16(wx_reinterpret_cast(const wxUint16 *, *pSrc), out); + if ( n == wxCONV_FAILED ) + *pSrc = NULL; + else + *pSrc += n; + + return out; +} // ---------------------------------------------------------------------------- // wxMBConv // ---------------------------------------------------------------------------- +size_t +wxMBConv::ToWChar(wchar_t *dst, size_t dstLen, + const char *src, size_t srcLen) const +{ + // although new conversion classes are supposed to implement this function + // directly, the existins ones only implement the old MB2WC() and so, to + // avoid to have to rewrite all conversion classes at once, we provide a + // default (but not efficient) implementation of this one in terms of the + // old function by copying the input to ensure that it's NUL-terminated and + // then using MB2WC() to convert it + + // the number of chars [which would be] written to dst [if it were not NULL] + size_t dstWritten = 0; + + // the number of NULs terminating this string + size_t nulLen = 0; // not really needed, but just to avoid warnings + + // if we were not given the input size we just have to assume that the + // string is properly terminated as we have no way of knowing how long it + // is anyhow, but if we do have the size check whether there are enough + // NULs at the end + wxCharBuffer bufTmp; + const char *srcEnd; + if ( srcLen != wxNO_LEN ) + { + // we need to know how to find the end of this string + nulLen = GetMBNulLen(); + if ( nulLen == wxCONV_FAILED ) + return wxCONV_FAILED; + + // if there are enough NULs we can avoid the copy + if ( srcLen < nulLen || NotAllNULs(src + srcLen - nulLen, nulLen) ) + { + // make a copy in order to properly NUL-terminate the string + bufTmp = wxCharBuffer(srcLen + nulLen - 1 /* 1 will be added */); + char * const p = bufTmp.data(); + memcpy(p, src, srcLen); + for ( char *s = p + srcLen; s < p + srcLen + nulLen; s++ ) + *s = '\0'; + + src = bufTmp; + } + + srcEnd = src + srcLen; + } + else // quit after the first loop iteration + { + srcEnd = NULL; + } + + for ( ;; ) + { + // try to convert the current chunk + size_t lenChunk = MB2WC(NULL, src, 0); + if ( lenChunk == wxCONV_FAILED ) + return wxCONV_FAILED; + + lenChunk++; // for the L'\0' at the end of this chunk + + dstWritten += lenChunk; + + if ( lenChunk == 1 ) + { + // nothing left in the input string, conversion succeeded + break; + } + + if ( dst ) + { + if ( dstWritten > dstLen ) + return wxCONV_FAILED; + + if ( MB2WC(dst, src, lenChunk) == wxCONV_FAILED ) + return wxCONV_FAILED; + + dst += lenChunk; + } + + if ( !srcEnd ) + { + // we convert just one chunk in this case as this is the entire + // string anyhow + break; + } + + // advance the input pointer past the end of this chunk + while ( NotAllNULs(src, nulLen) ) + { + // notice that we must skip over multiple bytes here as we suppose + // that if NUL takes 2 or 4 bytes, then all the other characters do + // too and so if advanced by a single byte we might erroneously + // detect sequences of NUL bytes in the middle of the input + src += nulLen; + } + + src += nulLen; // skipping over its terminator as well + + // note that ">=" (and not just "==") is needed here as the terminator + // we skipped just above could be inside or just after the buffer + // delimited by inEnd + if ( src >= srcEnd ) + break; + } + + return dstWritten; +} + +size_t +wxMBConv::FromWChar(char *dst, size_t dstLen, + const wchar_t *src, size_t srcLen) const +{ + // the number of chars [which would be] written to dst [if it were not NULL] + size_t dstWritten = 0; + + // make a copy of the input string unless it is already properly + // NUL-terminated + // + // if we don't know its length we have no choice but to assume that it is, + // indeed, properly terminated + wxWCharBuffer bufTmp; + if ( srcLen == wxNO_LEN ) + { + srcLen = wxWcslen(src) + 1; + } + else if ( srcLen != 0 && src[srcLen - 1] != L'\0' ) + { + // make a copy in order to properly NUL-terminate the string + bufTmp = wxWCharBuffer(srcLen); + memcpy(bufTmp.data(), src, srcLen * sizeof(wchar_t)); + src = bufTmp; + } + + const size_t lenNul = GetMBNulLen(); + for ( const wchar_t * const srcEnd = src + srcLen; + src < srcEnd; + src += wxWcslen(src) + 1 /* skip L'\0' too */ ) + { + // try to convert the current chunk + size_t lenChunk = WC2MB(NULL, src, 0); + + if ( lenChunk == wxCONV_FAILED ) + return wxCONV_FAILED; + + lenChunk += lenNul; + dstWritten += lenChunk; + + if ( dst ) + { + if ( dstWritten > dstLen ) + return wxCONV_FAILED; + + if ( WC2MB(dst, src, lenChunk) == wxCONV_FAILED ) + return wxCONV_FAILED; + + dst += lenChunk; + } + } + + return dstWritten; +} + +size_t wxMBConv::MB2WC(wchar_t *outBuff, const char *inBuff, size_t outLen) const +{ + size_t rc = ToWChar(outBuff, outLen, inBuff); + if ( rc != wxCONV_FAILED ) + { + // ToWChar() returns the buffer length, i.e. including the trailing + // NUL, while this method doesn't take it into account + rc--; + } + + return rc; +} + +size_t wxMBConv::WC2MB(char *outBuff, const wchar_t *inBuff, size_t outLen) const +{ + size_t rc = FromWChar(outBuff, outLen, inBuff); + if ( rc != wxCONV_FAILED ) + { + rc -= GetMBNulLen(); + } + + return rc; +} + wxMBConv::~wxMBConv() { // nothing to do here (necessary for Darwin linking probably) @@ -178,131 +360,99 @@ const wxWCharBuffer wxMBConv::cMB2WC(const char *psz) const if ( psz ) { // calculate the length of the buffer needed first - size_t nLen = MB2WC(NULL, psz, 0); - if ( nLen != (size_t)-1 ) + const size_t nLen = MB2WC(NULL, psz, 0); + if ( nLen != wxCONV_FAILED ) { // now do the actual conversion - wxWCharBuffer buf(nLen); - nLen = MB2WC(buf.data(), psz, nLen + 1); // with the trailing NULL - if ( nLen != (size_t)-1 ) - { + wxWCharBuffer buf(nLen /* +1 added implicitly */); + + // +1 for the trailing NULL + if ( MB2WC(buf.data(), psz, nLen + 1) != wxCONV_FAILED ) return buf; - } } } - wxWCharBuffer buf((wchar_t *)NULL); - - return buf; + return wxWCharBuffer(); } const wxCharBuffer wxMBConv::cWC2MB(const wchar_t *pwz) const { if ( pwz ) { - size_t nLen = WC2MB(NULL, pwz, 0); - if ( nLen != (size_t)-1 ) + const size_t nLen = WC2MB(NULL, pwz, 0); + if ( nLen != wxCONV_FAILED ) { - wxCharBuffer buf(nLen+3); // space for a wxUint32 trailing zero - nLen = WC2MB(buf.data(), pwz, nLen + 4); - if ( nLen != (size_t)-1 ) - { + // extra space for trailing NUL(s) + static const size_t extraLen = GetMaxMBNulLen(); + + wxCharBuffer buf(nLen + extraLen - 1); + if ( WC2MB(buf.data(), pwz, nLen + extraLen) != wxCONV_FAILED ) return buf; - } } } - wxCharBuffer buf((char *)NULL); - - return buf; + return wxCharBuffer(); } -size_t wxMBConv::MB2WC(wchar_t* szBuffer, const char* szString, - size_t outsize, size_t nStringLen) const +const wxWCharBuffer +wxMBConv::cMB2WC(const char *inBuff, size_t inLen, size_t *outLen) const { - const char* szEnd = szString + nStringLen + 1; - const char* szPos = szString; - const char* szStart = szPos; - - size_t nActualLength = 0; - - //Convert the string until the length() is reached, continuing the - //loop every time a null character is reached - while(szPos != szEnd) + const size_t dstLen = ToWChar(NULL, 0, inBuff, inLen); + if ( dstLen != wxCONV_FAILED ) { - wxASSERT(szPos < szEnd); //something is _really_ screwed up if this rings true - - //Get the length of the current (sub)string - size_t nLen = MB2WC(NULL, szPos, 0); - - //Invalid conversion? - if( nLen == (size_t)-1 ) - return nLen; - - //Increase the actual length (+1 for current null character) - nActualLength += nLen + 1; - - //Only copy data in if buffer size is big enough - if (szBuffer != NULL && - nActualLength <= outsize) + wxWCharBuffer wbuf(dstLen - 1); + if ( ToWChar(wbuf.data(), dstLen, inBuff, inLen) != wxCONV_FAILED ) { - //Convert the current (sub)string - if ( MB2WC(&szBuffer[szPos - szStart], szPos, nLen + 1) == (size_t)-1 ) - return (size_t)-1; - } + if ( outLen ) + { + *outLen = dstLen; + if ( wbuf[dstLen - 1] == L'\0' ) + (*outLen)--; + } - //Increment to next (sub)string - //Note that we have to use strlen here instead of nLen - //here because XX2XX gives us the size of the output buffer, - //not neccessarly the length of the string - szPos += strlen(szPos) + 1; + return wbuf; + } } - return nActualLength - 1; //success - return actual length + if ( outLen ) + *outLen = 0; + + return wxWCharBuffer(); } -size_t wxMBConv::WC2MB(char* szBuffer, const wchar_t* szString, - size_t outsize, size_t nStringLen) const +const wxCharBuffer +wxMBConv::cWC2MB(const wchar_t *inBuff, size_t inLen, size_t *outLen) const { - const wchar_t* szEnd = szString + nStringLen + 1; - const wchar_t* szPos = szString; - const wchar_t* szStart = szPos; - - size_t nActualLength = 0; - - //Convert the string until the length() is reached, continuing the - //loop every time a null character is reached - while(szPos != szEnd) + size_t dstLen = FromWChar(NULL, 0, inBuff, inLen); + if ( dstLen != wxCONV_FAILED ) { - wxASSERT(szPos < szEnd); //something is _really_ screwed up if this rings true - - //Get the length of the current (sub)string - size_t nLen = WC2MB(NULL, szPos, 0); + // special case of empty input: can't allocate 0 size buffer below as + // wxCharBuffer insists on NUL-terminating it + wxCharBuffer buf(dstLen ? dstLen - 1 : 1); + if ( FromWChar(buf.data(), dstLen, inBuff, inLen) != wxCONV_FAILED ) + { + if ( outLen ) + { + *outLen = dstLen; - //Invalid conversion? - if( nLen == (size_t)-1 ) - return nLen; + const size_t nulLen = GetMBNulLen(); + if ( dstLen >= nulLen && + !NotAllNULs(buf.data() + dstLen - nulLen, nulLen) ) + { + // in this case the output is NUL-terminated and we're not + // supposed to count NUL + *outLen -= nulLen; + } + } - //Increase the actual length (+1 for current null character) - nActualLength += nLen + 1; - - //Only copy data in if buffer size is big enough - if (szBuffer != NULL && - nActualLength <= outsize) - { - //Convert the current (sub)string - if(WC2MB(&szBuffer[szPos - szStart], szPos, nLen + 1) == (size_t)-1 ) - return (size_t)-1; + return buf; } - - //Increment to next (sub)string - //Note that we have to use wxWcslen here instead of nLen - //here because XX2XX gives us the size of the output buffer, - //not neccessarly the length of the string - szPos += wxWcslen(szPos) + 1; } - return nActualLength - 1; //success - return actual length + if ( outLen ) + *outLen = 0; + + return wxCharBuffer(); } // ---------------------------------------------------------------------------- @@ -318,8 +468,26 @@ size_t wxMBConvLibc::WC2MB(char *buf, const wchar_t *psz, size_t n) const { return wxWC2MB(buf, psz, n); } + // ---------------------------------------------------------------------------- -// UTF-7 +// wxConvBrokenFileNames +// ---------------------------------------------------------------------------- + +#ifdef __UNIX__ + +wxConvBrokenFileNames::wxConvBrokenFileNames(const wxChar *charset) +{ + if ( !charset || wxStricmp(charset, _T("UTF-8")) == 0 + || wxStricmp(charset, _T("UTF8")) == 0 ) + m_conv = new wxMBConvUTF8(wxMBConvUTF8::MAP_INVALID_UTF8_TO_OCTAL); + else + m_conv = new wxCSConv(charset); +} + +#endif // __UNIX__ + +// ---------------------------------------------------------------------------- +// UTF-7 // ---------------------------------------------------------------------------- // Implementation (C) 2004 Fredrik Roubert @@ -365,10 +533,9 @@ static const unsigned char utf7unb64[] = size_t wxMBConvUTF7::MB2WC(wchar_t *buf, const char *psz, size_t n) const { - size_t len = 0; - while (*psz && ((!buf) || (len < n))) + while ( *psz && (!buf || (len < n)) ) { unsigned char cc = *psz++; if (cc != '+') @@ -386,20 +553,19 @@ size_t wxMBConvUTF7::MB2WC(wchar_t *buf, const char *psz, size_t n) const len++; psz++; } - else + else // start of BASE64 encoded string { - // BASE64 encoded string - bool lsb; - unsigned char c; + bool lsb, ok; unsigned int d, l; - for (lsb = false, d = 0, l = 0; - (cc = utf7unb64[(unsigned char)*psz]) != 0xff; psz++) + for ( ok = lsb = false, d = 0, l = 0; + (cc = utf7unb64[(unsigned char)*psz]) != 0xff; + psz++ ) { d <<= 6; d += cc; for (l += 6; l >= 8; lsb = !lsb) { - c = (d >> (l -= 8)) % 256; + unsigned char c = (unsigned char)((d >> (l -= 8)) % 256); if (lsb) { if (buf) @@ -407,16 +573,29 @@ size_t wxMBConvUTF7::MB2WC(wchar_t *buf, const char *psz, size_t n) const len ++; } else + { if (buf) - *buf = c << 8; + *buf = (wchar_t)(c << 8); + } + + ok = true; } } + + if ( !ok ) + { + // in valid UTF7 we should have valid characters after '+' + return wxCONV_FAILED; + } + if (*psz == '-') psz++; } } - if (buf && (len < n)) - *buf = 0; + + if ( buf && (len < n) ) + *buf = '\0'; + return len; } @@ -455,11 +634,8 @@ static const unsigned char utf7encode[128] = 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 3, 3 }; -size_t wxMBConvUTF7::WC2MB(char *buf, const wchar_t -*psz, size_t n) const +size_t wxMBConvUTF7::WC2MB(char *buf, const wchar_t *psz, size_t n) const { - - size_t len = 0; while (*psz && ((!buf) || (len < n))) @@ -470,29 +646,27 @@ size_t wxMBConvUTF7::WC2MB(char *buf, const wchar_t // plain ASCII char if (buf) *buf++ = (char)cc; + len++; } #ifndef WC_UTF16 -#ifdef __VMS - else if (cc > 0xffff) -#else - else if (cc > ((const wchar_t)0xffff)) -#endif - { + else if (((wxUint32)cc) > 0xffff) + { // no surrogate pair generation (yet?) - return (size_t)-1; + return wxCONV_FAILED; } #endif else { if (buf) *buf++ = '+'; + len++; if (cc != '+') { // BASE64 encode string unsigned int lsb, d, l; - for (d = 0, l = 0;; psz++) + for (d = 0, l = 0; /*nothing*/; psz++) { for (lsb = 0; lsb < 2; lsb ++) { @@ -507,24 +681,30 @@ size_t wxMBConvUTF7::WC2MB(char *buf, const wchar_t len++; } } + cc = *psz; if (!(cc) || (cc < 0x80 && utf7encode[cc] < 1)) break; } + if (l != 0) { if (buf) *buf++ = utf7enb64[((d % 16) << (6 - l)) % 64]; + len++; } } + if (buf) *buf++ = '-'; len++; } } + if (buf && (len < n)) *buf = 0; + return len; } @@ -535,22 +715,39 @@ size_t wxMBConvUTF7::WC2MB(char *buf, const wchar_t static wxUint32 utf8_max[]= { 0x7f, 0x7ff, 0xffff, 0x1fffff, 0x3ffffff, 0x7fffffff, 0xffffffff }; +// boundaries of the private use area we use to (temporarily) remap invalid +// characters invalid in a UTF-8 encoded string +const wxUint32 wxUnicodePUA = 0x100000; +const wxUint32 wxUnicodePUAEnd = wxUnicodePUA + 256; + size_t wxMBConvUTF8::MB2WC(wchar_t *buf, const char *psz, size_t n) const { size_t len = 0; while (*psz && ((!buf) || (len < n))) { + const char *opsz = psz; + bool invalid = false; unsigned char cc = *psz++, fc = cc; unsigned cnt; for (cnt = 0; fc & 0x80; cnt++) fc <<= 1; + if (!cnt) { // plain ASCII char if (buf) *buf++ = cc; len++; + + // escape the escape character for octal escapes + if ((m_options & MAP_INVALID_UTF8_TO_OCTAL) + && cc == '\\' && (!buf || len < n)) + { + if (buf) + *buf++ = cc; + len++; + } } else { @@ -558,7 +755,7 @@ size_t wxMBConvUTF8::MB2WC(wchar_t *buf, const char *psz, size_t n) const if (!cnt) { // invalid UTF-8 sequence - return (size_t)-1; + invalid = true; } else { @@ -566,40 +763,112 @@ size_t wxMBConvUTF8::MB2WC(wchar_t *buf, const char *psz, size_t n) const wxUint32 res = cc & (0x3f >> cnt); while (cnt--) { - cc = *psz++; + cc = *psz; if ((cc & 0xC0) != 0x80) { // invalid UTF-8 sequence - return (size_t)-1; + invalid = true; + break; } + + psz++; res = (res << 6) | (cc & 0x3f); } - if (res <= utf8_max[ocnt]) + + if (invalid || res <= utf8_max[ocnt]) { // illegal UTF-8 encoding - return (size_t)-1; + invalid = true; + } + else if ((m_options & MAP_INVALID_UTF8_TO_PUA) && + res >= wxUnicodePUA && res < wxUnicodePUAEnd) + { + // if one of our PUA characters turns up externally + // it must also be treated as an illegal sequence + // (a bit like you have to escape an escape character) + invalid = true; } + else + { #ifdef WC_UTF16 - // cast is ok because wchar_t == wxUuint16 if WC_UTF16 - size_t pa = encode_utf16(res, (wxUint16 *)buf); - if (pa == (size_t)-1) - return (size_t)-1; - if (buf) - buf += pa; - len += pa; + // cast is ok because wchar_t == wxUuint16 if WC_UTF16 + size_t pa = encode_utf16(res, (wxUint16 *)buf); + if (pa == wxCONV_FAILED) + { + invalid = true; + } + else + { + if (buf) + buf += pa; + len += pa; + } #else // !WC_UTF16 - if (buf) - *buf++ = res; - len++; + if (buf) + *buf++ = (wchar_t)res; + len++; #endif // WC_UTF16/!WC_UTF16 + } + } + + if (invalid) + { + if (m_options & MAP_INVALID_UTF8_TO_PUA) + { + while (opsz < psz && (!buf || len < n)) + { +#ifdef WC_UTF16 + // cast is ok because wchar_t == wxUuint16 if WC_UTF16 + size_t pa = encode_utf16((unsigned char)*opsz + wxUnicodePUA, (wxUint16 *)buf); + wxASSERT(pa != wxCONV_FAILED); + if (buf) + buf += pa; + opsz++; + len += pa; +#else + if (buf) + *buf++ = (wchar_t)(wxUnicodePUA + (unsigned char)*opsz); + opsz++; + len++; +#endif + } + } + else if (m_options & MAP_INVALID_UTF8_TO_OCTAL) + { + while (opsz < psz && (!buf || len < n)) + { + if ( buf && len + 3 < n ) + { + unsigned char on = *opsz; + *buf++ = L'\\'; + *buf++ = (wchar_t)( L'0' + on / 0100 ); + *buf++ = (wchar_t)( L'0' + (on % 0100) / 010 ); + *buf++ = (wchar_t)( L'0' + on % 010 ); + } + + opsz++; + len += 4; + } + } + else // MAP_INVALID_UTF8_NOT + { + return wxCONV_FAILED; + } } } } + if (buf && (len < n)) *buf = 0; + return len; } +static inline bool isoctal(wchar_t wch) +{ + return L'0' <= wch && wch <= L'7'; +} + size_t wxMBConvUTF8::WC2MB(char *buf, const wchar_t *psz, size_t n) const { size_t len = 0; @@ -607,46 +876,80 @@ size_t wxMBConvUTF8::WC2MB(char *buf, const wchar_t *psz, size_t n) const while (*psz && ((!buf) || (len < n))) { wxUint32 cc; + #ifdef WC_UTF16 // cast is ok for WC_UTF16 size_t pa = decode_utf16((const wxUint16 *)psz, cc); - psz += (pa == (size_t)-1) ? 1 : pa; + psz += (pa == wxCONV_FAILED) ? 1 : pa; #else - cc=(*psz++) & 0x7fffffff; + cc = (*psz++) & 0x7fffffff; #endif - unsigned cnt; - for (cnt = 0; cc > utf8_max[cnt]; cnt++) {} - if (!cnt) + + if ( (m_options & MAP_INVALID_UTF8_TO_PUA) + && cc >= wxUnicodePUA && cc < wxUnicodePUAEnd ) { - // plain ASCII char if (buf) - *buf++ = (char) cc; + *buf++ = (char)(cc - wxUnicodePUA); len++; } + else if ( (m_options & MAP_INVALID_UTF8_TO_OCTAL) + && cc == L'\\' && psz[0] == L'\\' ) + { + if (buf) + *buf++ = (char)cc; + psz++; + len++; + } + else if ( (m_options & MAP_INVALID_UTF8_TO_OCTAL) && + cc == L'\\' && + isoctal(psz[0]) && isoctal(psz[1]) && isoctal(psz[2]) ) + { + if (buf) + { + *buf++ = (char) ((psz[0] - L'0') * 0100 + + (psz[1] - L'0') * 010 + + (psz[2] - L'0')); + } + psz += 3; + len++; + } else { - len += cnt + 1; - if (buf) + unsigned cnt; + for (cnt = 0; cc > utf8_max[cnt]; cnt++) { - *buf++ = (char) ((-128 >> cnt) | ((cc >> (cnt * 6)) & (0x3f >> cnt))); - while (cnt--) - *buf++ = (char) (0x80 | ((cc >> (cnt * 6)) & 0x3f)); + } + + if (!cnt) + { + // plain ASCII char + if (buf) + *buf++ = (char) cc; + len++; + } + else + { + len += cnt + 1; + if (buf) + { + *buf++ = (char) ((-128 >> cnt) | ((cc >> (cnt * 6)) & (0x3f >> cnt))); + while (cnt--) + *buf++ = (char) (0x80 | ((cc >> (cnt * 6)) & 0x3f)); + } } } } - if (buf && (len dstLen ) + return wxCONV_FAILED; + + *dst++ = ch; } - if (buf && len 1) + if ( outLen > dstLen ) + return wxCONV_FAILED; + + *outBuff++ = cc[0]; + if ( numChars == 2 ) { - *(wxUint16*)buf = cc[1]; - buf += sizeof(wxUint16); + // second character of a surrogate + *outBuff++ = cc[1]; } } - - len += pa*sizeof(wxUint16); - psz++; } - if (buf && len<=n-sizeof(wxUint16)) *(wxUint16*)buf=0; - return len; + return outLen; } +// ---------------------------------------------------------------------------- +// endian-reversing conversions +// ---------------------------------------------------------------------------- -// swap 16bit MB to 32bit String -size_t wxMBConvUTF16swap::MB2WC(wchar_t *buf, const char *psz, size_t n) const +size_t +wxMBConvUTF16swap::ToWChar(wchar_t *dst, size_t dstLen, + const char *src, size_t srcLen) const { - size_t len=0; + srcLen = GetLength(src, srcLen); + if ( srcLen == wxNO_LEN ) + return wxCONV_FAILED; - while (*(wxUint16*)psz && (!buf || len < n)) + const size_t inLen = srcLen / BYTES_PER_CHAR; + if ( !dst ) { - wxUint32 cc; - char tmp[4]; - tmp[0]=psz[1]; tmp[1]=psz[0]; - tmp[2]=psz[3]; tmp[3]=psz[2]; + // optimization: return maximal space which could be needed for this + // string even if the real size could be smaller if the buffer contains + // any surrogates + return inLen; + } - size_t pa=decode_utf16((wxUint16*)tmp, cc); - if (pa == (size_t)-1) - return pa; + size_t outLen = 0; + const wxUint16 *inBuff = wx_reinterpret_cast(const wxUint16 *, src); + for ( const wxUint16 * const inEnd = inBuff + inLen; inBuff < inEnd; ) + { + wxUint32 ch; + wxUint16 tmp[2]; - if (buf) - *buf++ = cc; + tmp[0] = wxUINT16_SWAP_ALWAYS(*inBuff); + inBuff++; + tmp[1] = wxUINT16_SWAP_ALWAYS(*inBuff); + + const size_t numChars = decode_utf16(tmp, ch); + if ( numChars == wxCONV_FAILED ) + return wxCONV_FAILED; + + if ( numChars == 2 ) + inBuff++; - len++; - psz += pa * sizeof(wxUint16); + if ( ++outLen > dstLen ) + return wxCONV_FAILED; + + *dst++ = ch; } - if (buf && len 1) + if ( outLen > dstLen ) + return wxCONV_FAILED; + + *outBuff++ = wxUINT16_SWAP_ALWAYS(cc[0]); + if ( numChars == 2 ) { - *buf++ = ((char*)cc)[3]; - *buf++ = ((char*)cc)[2]; + // second character of a surrogate + *outBuff++ = wxUINT16_SWAP_ALWAYS(cc[1]); } } - - len += pa*sizeof(wxUint16); - psz++; } - if (buf && len<=n-sizeof(wxUint16)) *(wxUint16*)buf=0; - return len; + return outLen; } -#endif // WC_UTF16 +#endif // WC_UTF16/!WC_UTF16 -// ---------------------------------------------------------------------------- +// ============================================================================ // UTF-32 -// ---------------------------------------------------------------------------- +// ============================================================================ #ifdef WORDS_BIGENDIAN -#define wxMBConvUTF32straight wxMBConvUTF32BE -#define wxMBConvUTF32swap wxMBConvUTF32LE + #define wxMBConvUTF32straight wxMBConvUTF32BE + #define wxMBConvUTF32swap wxMBConvUTF32LE #else -#define wxMBConvUTF32swap wxMBConvUTF32BE -#define wxMBConvUTF32straight wxMBConvUTF32LE + #define wxMBConvUTF32swap wxMBConvUTF32BE + #define wxMBConvUTF32straight wxMBConvUTF32LE #endif WXDLLIMPEXP_DATA_BASE(wxMBConvUTF32LE) wxConvUTF32LE; WXDLLIMPEXP_DATA_BASE(wxMBConvUTF32BE) wxConvUTF32BE; +/* static */ +size_t wxMBConvUTF32Base::GetLength(const char *src, size_t srcLen) +{ + if ( srcLen == wxNO_LEN ) + { + // count the number of bytes in input, including the trailing NULs + const wxUint32 *inBuff = wx_reinterpret_cast(const wxUint32 *, src); + for ( srcLen = 1; *inBuff++; srcLen++ ) + ; + + srcLen *= BYTES_PER_CHAR; + } + else // we already have the length + { + // we can only convert an entire number of UTF-32 characters + if ( srcLen % BYTES_PER_CHAR ) + return wxCONV_FAILED; + } + + return srcLen; +} +// case when in-memory representation is UTF-16 #ifdef WC_UTF16 -// copy 32bit MB to 16bit String -size_t wxMBConvUTF32straight::MB2WC(wchar_t *buf, const char *psz, size_t n) const +// ---------------------------------------------------------------------------- +// conversions without endianness change +// ---------------------------------------------------------------------------- + +size_t +wxMBConvUTF32straight::ToWChar(wchar_t *dst, size_t dstLen, + const char *src, size_t srcLen) const { - size_t len=0; + srcLen = GetLength(src, srcLen); + if ( srcLen == wxNO_LEN ) + return wxCONV_FAILED; - while (*(wxUint32*)psz && (!buf || len < n)) + const wxUint32 *inBuff = wx_reinterpret_cast(const wxUint32 *, src); + const size_t inLen = srcLen / BYTES_PER_CHAR; + size_t outLen = 0; + for ( size_t n = 0; n < inLen; n++ ) { wxUint16 cc[2]; + const size_t numChars = encode_utf16(*inBuff++, cc); + if ( numChars == wxCONV_FAILED ) + return wxCONV_FAILED; - size_t pa=encode_utf16(*(wxUint32*)psz, cc); - if (pa == (size_t)-1) - return pa; - - if (buf) + outLen += numChars; + if ( dst ) { - *buf++ = cc[0]; - if (pa > 1) - *buf++ = cc[1]; + if ( outLen > dstLen ) + return wxCONV_FAILED; + + *dst++ = cc[0]; + if ( numChars == 2 ) + { + // second character of a surrogate + *dst++ = cc[1]; + } } - len += pa; - psz += sizeof(wxUint32); } - if (buf && len dstLen ) + return wxCONV_FAILED; - return len; -} + *outBuff++ = ch; + } + return outLen; +} +// ---------------------------------------------------------------------------- +// endian-reversing conversions +// ---------------------------------------------------------------------------- -// swap 32bit MB to 16bit String -size_t wxMBConvUTF32swap::MB2WC(wchar_t *buf, const char *psz, size_t n) const +size_t +wxMBConvUTF32swap::ToWChar(wchar_t *dst, size_t dstLen, + const char *src, size_t srcLen) const { - size_t len=0; + srcLen = GetLength(src, srcLen); + if ( srcLen == wxNO_LEN ) + return wxCONV_FAILED; - while (*(wxUint32*)psz && (!buf || len < n)) + const wxUint32 *inBuff = wx_reinterpret_cast(const wxUint32 *, src); + const size_t inLen = srcLen / BYTES_PER_CHAR; + size_t outLen = 0; + for ( size_t n = 0; n < inLen; n++, inBuff++ ) { - char tmp[4]; - tmp[0] = psz[3]; tmp[1] = psz[2]; - tmp[2] = psz[1]; tmp[3] = psz[0]; - - wxUint16 cc[2]; + const size_t numChars = encode_utf16(wxUINT32_SWAP_ALWAYS(*inBuff), cc); + if ( numChars == wxCONV_FAILED ) + return wxCONV_FAILED; - size_t pa=encode_utf16(*(wxUint32*)tmp, cc); - if (pa == (size_t)-1) - return pa; - - if (buf) + outLen += numChars; + if ( dst ) { - *buf++ = cc[0]; - if (pa > 1) - *buf++ = cc[1]; + if ( outLen > dstLen ) + return wxCONV_FAILED; + + *dst++ = cc[0]; + if ( numChars == 2 ) + { + // second character of a surrogate + *dst++ = cc[1]; + } } - len += pa; - psz += sizeof(wxUint32); } - if (buf && len dstLen ) + return wxCONV_FAILED; - return len; + *outBuff++ = wxUINT32_SWAP_ALWAYS(ch); + } + + return outLen; } -#else // WC_UTF16 +#else // !WC_UTF16: wchar_t is UTF-32 +// ---------------------------------------------------------------------------- +// conversions without endianness change +// ---------------------------------------------------------------------------- -// copy 32bit MB to 32bit String -size_t wxMBConvUTF32straight::MB2WC(wchar_t *buf, const char *psz, size_t n) const +size_t +wxMBConvUTF32straight::ToWChar(wchar_t *dst, size_t dstLen, + const char *src, size_t srcLen) const { - size_t len=0; + // use memcpy() as it should be much faster than hand-written loop + srcLen = GetLength(src, srcLen); + if ( srcLen == wxNO_LEN ) + return wxCONV_FAILED; - while (*(wxUint32*)psz && (!buf || len < n)) + const size_t inLen = srcLen/BYTES_PER_CHAR; + if ( dst ) { - if (buf) - *buf++ = *(wxUint32*)psz; - len++; - psz += sizeof(wxUint32); - } + if ( dstLen < inLen ) + return wxCONV_FAILED; - if (buf && lenm_minMBCharWidth = m_minMBCharWidth; + return p; + } + bool IsOk() const - { return (m2w != (iconv_t)-1) && (w2m != (iconv_t)-1); } + { return (m2w != ICONV_T_INVALID) && (w2m != ICONV_T_INVALID); } protected: - // the iconv handlers used to translate from multibyte to wide char and in - // the other direction + // the iconv handlers used to translate from multibyte + // to wide char and in the other direction iconv_t m2w, w2m; +#if wxUSE_THREADS + // guards access to m2w and w2m objects + wxMutex m_iconvMutex; +#endif + private: // the name (for iconv_open()) of a wide char charset -- if none is // available on this machine, it will remain NULL - static const char *ms_wcCharsetName; + static wxString ms_wcCharsetName; // true if the wide char encoding we use (i.e. ms_wcCharsetName) has // different endian-ness than the native one static bool ms_wcNeedsSwap; + + + // name of the encoding handled by this conversion + wxString m_name; + + // cached result of GetMBNulLen(); set to 0 meaning "unknown" + // initially + size_t m_minMBCharWidth; }; -const char *wxMBConv_iconv::ms_wcCharsetName = NULL; +// make the constructor available for unit testing +WXDLLIMPEXP_BASE wxMBConv* new_wxMBConv_iconv( const wxChar* name ) +{ + wxMBConv_iconv* result = new wxMBConv_iconv( name ); + if ( !result->IsOk() ) + { + delete result; + return 0; + } + + return result; +} + +wxString wxMBConv_iconv::ms_wcCharsetName; bool wxMBConv_iconv::ms_wcNeedsSwap = false; wxMBConv_iconv::wxMBConv_iconv(const wxChar *name) + : m_name(name) { - // Do it the hard way - char cname[100]; - for (size_t i = 0; i < wxStrlen(name)+1; i++) - cname[i] = (char) name[i]; + m_minMBCharWidth = 0; + + // iconv operates with chars, not wxChars, but luckily it uses only ASCII + // names for the charsets + const wxCharBuffer cname(wxString(name).ToAscii()); // check for charset that represents wchar_t: - if (ms_wcCharsetName == NULL) + if ( ms_wcCharsetName.empty() ) { - ms_wcNeedsSwap = false; + wxLogTrace(TRACE_STRCONV, _T("Looking for wide char codeset:")); - // try charset with explicit bytesex info (e.g. "UCS-4LE"): - ms_wcCharsetName = WC_NAME_BEST; - m2w = iconv_open(ms_wcCharsetName, cname); +#if wxUSE_FONTMAP + const wxChar **names = wxFontMapperBase::GetAllEncodingNames(WC_ENC); +#else // !wxUSE_FONTMAP + static const wxChar *names_static[] = + { +#if SIZEOF_WCHAR_T == 4 + _T("UCS-4"), +#elif SIZEOF_WCHAR_T = 2 + _T("UCS-2"), +#endif + NULL + }; + const wxChar **names = names_static; +#endif // wxUSE_FONTMAP/!wxUSE_FONTMAP - if (m2w == (iconv_t)-1) + for ( ; *names && ms_wcCharsetName.empty(); ++names ) { - // try charset w/o bytesex info (e.g. "UCS4") - // and check for bytesex ourselves: - ms_wcCharsetName = WC_NAME; - m2w = iconv_open(ms_wcCharsetName, cname); + const wxString nameCS(*names); - // last bet, try if it knows WCHAR_T pseudo-charset - if (m2w == (iconv_t)-1) - { - ms_wcCharsetName = "WCHAR_T"; - m2w = iconv_open(ms_wcCharsetName, cname); - } + // first try charset with explicit bytesex info (e.g. "UCS-4LE"): + wxString nameXE(nameCS); + +#ifdef WORDS_BIGENDIAN + nameXE += _T("BE"); +#else // little endian + nameXE += _T("LE"); +#endif - if (m2w != (iconv_t)-1) + wxLogTrace(TRACE_STRCONV, _T(" trying charset \"%s\""), + nameXE.c_str()); + + m2w = iconv_open(nameXE.ToAscii(), cname); + if ( m2w == ICONV_T_INVALID ) { - char buf[2], *bufPtr; - wchar_t wbuf[2], *wbufPtr; - size_t insz, outsz; - size_t res; - - buf[0] = 'A'; - buf[1] = 0; - wbuf[0] = 0; - insz = 2; - outsz = SIZEOF_WCHAR_T * 2; - wbufPtr = wbuf; - bufPtr = buf; - - res = iconv(m2w, ICONV_CHAR_CAST(&bufPtr), &insz, - (char**)&wbufPtr, &outsz); - - if (ICONV_FAILED(res, insz)) - { - ms_wcCharsetName = NULL; - wxLogLastError(wxT("iconv")); - wxLogError(_("Conversion to charset '%s' doesn't work."), name); - } - else + // try charset w/o bytesex info (e.g. "UCS4") + wxLogTrace(TRACE_STRCONV, _T(" trying charset \"%s\""), + nameCS.c_str()); + m2w = iconv_open(nameCS.ToAscii(), cname); + + // and check for bytesex ourselves: + if ( m2w != ICONV_T_INVALID ) { - ms_wcNeedsSwap = wbuf[0] != (wchar_t)buf[0]; + char buf[2], *bufPtr; + wchar_t wbuf[2], *wbufPtr; + size_t insz, outsz; + size_t res; + + buf[0] = 'A'; + buf[1] = 0; + wbuf[0] = 0; + insz = 2; + outsz = SIZEOF_WCHAR_T * 2; + wbufPtr = wbuf; + bufPtr = buf; + + res = iconv( + m2w, ICONV_CHAR_CAST(&bufPtr), &insz, + (char**)&wbufPtr, &outsz); + + if (ICONV_FAILED(res, insz)) + { + wxLogLastError(wxT("iconv")); + wxLogError(_("Conversion to charset '%s' doesn't work."), + nameCS.c_str()); + } + else // ok, can convert to this encoding, remember it + { + ms_wcCharsetName = nameCS; + ms_wcNeedsSwap = wbuf[0] != (wchar_t)buf[0]; + } } } - else + else // use charset not requiring byte swapping { - ms_wcCharsetName = NULL; - - // VS: we must not output an error here, since wxWidgets will safely - // fall back to using wxEncodingConverter. - wxLogTrace(wxT("strconv"), wxT("Impossible to convert to/from charset '%s' with iconv, falling back to wxEncodingConverter."), name); - //wxLogError( + ms_wcCharsetName = nameXE; } } - wxLogTrace(wxT("strconv"), wxT("wchar_t charset is '%s', needs swap: %i"), ms_wcCharsetName, ms_wcNeedsSwap); + + wxLogTrace(TRACE_STRCONV, + wxT("iconv wchar_t charset is \"%s\"%s"), + ms_wcCharsetName.empty() ? _T("") + : ms_wcCharsetName.c_str(), + ms_wcNeedsSwap ? _T(" (needs swap)") + : _T("")); } else // we already have ms_wcCharsetName { - m2w = iconv_open(ms_wcCharsetName, cname); + m2w = iconv_open(ms_wcCharsetName.ToAscii(), cname); } - // NB: don't ever pass NULL to iconv_open(), it may crash! - if ( ms_wcCharsetName ) + if ( ms_wcCharsetName.empty() ) { - w2m = iconv_open( cname, ms_wcCharsetName); + w2m = ICONV_T_INVALID; } else { - w2m = (iconv_t)-1; + w2m = iconv_open(cname, ms_wcCharsetName.ToAscii()); + if ( w2m == ICONV_T_INVALID ) + { + wxLogTrace(TRACE_STRCONV, + wxT("\"%s\" -> \"%s\" works but not the converse!?"), + ms_wcCharsetName.c_str(), cname.data()); + } } } wxMBConv_iconv::~wxMBConv_iconv() { - if ( m2w != (iconv_t)-1 ) + if ( m2w != ICONV_T_INVALID ) iconv_close(m2w); - if ( w2m != (iconv_t)-1 ) + if ( w2m != ICONV_T_INVALID ) iconv_close(w2m); } size_t wxMBConv_iconv::MB2WC(wchar_t *buf, const char *psz, size_t n) const { - size_t inbuf = strlen(psz); + // find the string length: notice that must be done differently for + // NUL-terminated strings and UTF-16/32 which are terminated with 2/4 NULs + size_t inbuf; + const size_t nulLen = GetMBNulLen(); + switch ( nulLen ) + { + default: + return wxCONV_FAILED; + + case 1: + inbuf = strlen(psz); // arguably more optimized than our version + break; + + case 2: + case 4: + // for UTF-16/32 not only we need to have 2/4 consecutive NULs but + // they also have to start at character boundary and not span two + // adjacent characters + const char *p; + for ( p = psz; NotAllNULs(p, nulLen); p += nulLen ) + ; + inbuf = p - psz; + break; + } + +#if wxUSE_THREADS + // NB: iconv() is MT-safe, but each thread must use its own iconv_t handle. + // Unfortunately there are a couple of global wxCSConv objects such as + // wxConvLocal that are used all over wx code, so we have to make sure + // the handle is used by at most one thread at the time. Otherwise + // only a few wx classes would be safe to use from non-main threads + // as MB<->WC conversion would fail "randomly". + wxMutexLocker lock(wxConstCast(this, wxMBConv_iconv)->m_iconvMutex); +#endif // wxUSE_THREADS + size_t outbuf = n * SIZEOF_WCHAR_T; size_t res, cres; // VS: Use these instead of psz, buf because iconv() modifies its arguments: @@ -1281,12 +1833,11 @@ size_t wxMBConv_iconv::MB2WC(wchar_t *buf, const char *psz, size_t n) const if (ms_wcNeedsSwap) { // convert to native endianness - WC_BSWAP(buf /* _not_ bufPtr */, res) + for ( unsigned i = 0; i < res; i++ ) + buf[n] = WC_BSWAP(buf[i]); } - // NB: iconv was given only strlen(psz) characters on input, and so - // it couldn't convert the trailing zero. Let's do it ourselves - // if there's some room left for it in the output buffer. + // NUL-terminate the string if there is any space left if (res < n) buf[res] = 0; } @@ -1296,23 +1847,26 @@ size_t wxMBConv_iconv::MB2WC(wchar_t *buf, const char *psz, size_t n) const // to calculate destination buffer requirement wchar_t tbuf[8]; res = 0; - do { + + do + { bufPtr = tbuf; - outbuf = 8*SIZEOF_WCHAR_T; + outbuf = 8 * SIZEOF_WCHAR_T; cres = iconv(m2w, ICONV_CHAR_CAST(&pszPtr), &inbuf, (char**)&bufPtr, &outbuf ); - res += 8-(outbuf/SIZEOF_WCHAR_T); - } while ((cres==(size_t)-1) && (errno==E2BIG)); + res += 8 - (outbuf / SIZEOF_WCHAR_T); + } + while ((cres == (size_t)-1) && (errno == E2BIG)); } if (ICONV_FAILED(cres, inbuf)) { //VS: it is ok if iconv fails, hence trace only - wxLogTrace(wxT("strconv"), wxT("iconv failed: %s"), wxSysErrorMsg(wxSysErrorCode())); - return (size_t)-1; + wxLogTrace(TRACE_STRCONV, wxT("iconv failed: %s"), wxSysErrorMsg(wxSysErrorCode())); + return wxCONV_FAILED; } return res; @@ -1320,7 +1874,13 @@ size_t wxMBConv_iconv::MB2WC(wchar_t *buf, const char *psz, size_t n) const size_t wxMBConv_iconv::WC2MB(char *buf, const wchar_t *psz, size_t n) const { - size_t inbuf = wxWcslen(psz) * SIZEOF_WCHAR_T; +#if wxUSE_THREADS + // NB: explained in MB2WC + wxMutexLocker lock(wxConstCast(this, wxMBConv_iconv)->m_iconvMutex); +#endif + + size_t inlen = wxWcslen(psz); + size_t inbuf = inlen * SIZEOF_WCHAR_T; size_t outbuf = n; size_t res, cres; @@ -1329,13 +1889,14 @@ size_t wxMBConv_iconv::WC2MB(char *buf, const wchar_t *psz, size_t n) const if (ms_wcNeedsSwap) { // need to copy to temp buffer to switch endianness - // this absolutely doesn't rock! - // (no, doing WC_BSWAP twice on the original buffer won't help, as it + // (doing WC_BSWAP twice on the original buffer won't help, as it // could be in read-only memory, or be accessed in some other thread) - tmpbuf=(wchar_t*)malloc((inbuf+1)*SIZEOF_WCHAR_T); - memcpy(tmpbuf,psz,(inbuf+1)*SIZEOF_WCHAR_T); - WC_BSWAP(tmpbuf, inbuf) - psz=tmpbuf; + tmpbuf = (wchar_t *)malloc(inbuf + SIZEOF_WCHAR_T); + for ( size_t i = 0; i < inlen; i++ ) + tmpbuf[n] = WC_BSWAP(psz[i]); + + tmpbuf[inlen] = L'\0'; + psz = tmpbuf; } if (buf) @@ -1343,7 +1904,7 @@ size_t wxMBConv_iconv::WC2MB(char *buf, const wchar_t *psz, size_t n) const // have destination buffer, convert there cres = iconv( w2m, ICONV_CHAR_CAST(&psz), &inbuf, &buf, &outbuf ); - res = n-outbuf; + res = n - outbuf; // NB: iconv was given only wcslen(psz) characters on input, and so // it couldn't convert the trailing zero. Let's do it ourselves @@ -1353,17 +1914,20 @@ size_t wxMBConv_iconv::WC2MB(char *buf, const wchar_t *psz, size_t n) const } else { - // no destination buffer... convert using temp buffer + // no destination buffer: convert using temp buffer // to calculate destination buffer requirement char tbuf[16]; res = 0; - do { - buf = tbuf; outbuf = 16; + do + { + buf = tbuf; + outbuf = 16; cres = iconv( w2m, ICONV_CHAR_CAST(&psz), &inbuf, &buf, &outbuf ); res += 16 - outbuf; - } while ((cres==(size_t)-1) && (errno==E2BIG)); + } + while ((cres == (size_t)-1) && (errno == E2BIG)); } if (ms_wcNeedsSwap) @@ -1373,14 +1937,43 @@ size_t wxMBConv_iconv::WC2MB(char *buf, const wchar_t *psz, size_t n) const if (ICONV_FAILED(cres, inbuf)) { - //VS: it is ok if iconv fails, hence trace only - wxLogTrace(wxT("strconv"), wxT("iconv failed: %s"), wxSysErrorMsg(wxSysErrorCode())); - return (size_t)-1; + wxLogTrace(TRACE_STRCONV, wxT("iconv failed: %s"), wxSysErrorMsg(wxSysErrorCode())); + return wxCONV_FAILED; } return res; } +size_t wxMBConv_iconv::GetMBNulLen() const +{ + if ( m_minMBCharWidth == 0 ) + { + wxMBConv_iconv * const self = wxConstCast(this, wxMBConv_iconv); + +#if wxUSE_THREADS + // NB: explained in MB2WC + wxMutexLocker lock(self->m_iconvMutex); +#endif + + wchar_t *wnul = L""; + char buf[8]; // should be enough for NUL in any encoding + size_t inLen = sizeof(wchar_t), + outLen = WXSIZEOF(buf); + char *inBuff = (char *)wnul; + char *outBuff = buf; + if ( iconv(w2m, ICONV_CHAR_CAST(&inBuff), &inLen, &outBuff, &outLen) == (size_t)-1 ) + { + self->m_minMBCharWidth = (size_t)-1; + } + else // ok + { + self->m_minMBCharWidth = outBuff - buf; + } + } + + return m_minMBCharWidth; +} + #endif // HAVE_ICONV @@ -1402,44 +1995,107 @@ public: wxMBConv_win32() { m_CodePage = CP_ACP; + m_minMBCharWidth = 0; + } + + wxMBConv_win32(const wxMBConv_win32& conv) + : wxMBConv() + { + m_CodePage = conv.m_CodePage; + m_minMBCharWidth = conv.m_minMBCharWidth; } #if wxUSE_FONTMAP wxMBConv_win32(const wxChar* name) { m_CodePage = wxCharsetToCodepage(name); + m_minMBCharWidth = 0; } wxMBConv_win32(wxFontEncoding encoding) { m_CodePage = wxEncodingToCodepage(encoding); + m_minMBCharWidth = 0; } -#endif +#endif // wxUSE_FONTMAP - size_t MB2WC(wchar_t *buf, const char *psz, size_t n) const + virtual size_t MB2WC(wchar_t *buf, const char *psz, size_t n) const { // note that we have to use MB_ERR_INVALID_CHARS flag as it without it // the behaviour is not compatible with the Unix version (using iconv) // and break the library itself, e.g. wxTextInputStream::NextChar() // wouldn't work if reading an incomplete MB char didn't result in an // error + // + // Moreover, MB_ERR_INVALID_CHARS is only supported on Win 2K SP4 or + // Win XP or newer and it is not supported for UTF-[78] so we always + // use our own conversions in this case. See + // http://blogs.msdn.com/michkap/archive/2005/04/19/409566.aspx + // http://msdn.microsoft.com/library/en-us/intl/unicode_17si.asp + if ( m_CodePage == CP_UTF8 ) + { + return wxConvUTF8.MB2WC(buf, psz, n); + } + + if ( m_CodePage == CP_UTF7 ) + { + return wxConvUTF7.MB2WC(buf, psz, n); + } + + int flags = 0; + if ( (m_CodePage < 50000 && m_CodePage != CP_SYMBOL) && + IsAtLeastWin2kSP4() ) + { + flags = MB_ERR_INVALID_CHARS; + } + const size_t len = ::MultiByteToWideChar ( m_CodePage, // code page - MB_ERR_INVALID_CHARS, // flags: fall on error + flags, // flags: fall on error psz, // input string -1, // its length (NUL-terminated) buf, // output string buf ? n : 0 // size of output buffer ); + if ( !len ) + { + // function totally failed + return wxCONV_FAILED; + } + + // if we were really converting and didn't use MB_ERR_INVALID_CHARS, + // check if we succeeded, by doing a double trip: + if ( !flags && buf ) + { + const size_t mbLen = strlen(psz); + wxCharBuffer mbBuf(mbLen); + if ( ::WideCharToMultiByte + ( + m_CodePage, + 0, + buf, + -1, + mbBuf.data(), + mbLen + 1, // size in bytes, not length + NULL, + NULL + ) == 0 || + strcmp(mbBuf, psz) != 0 ) + { + // we didn't obtain the same thing we started from, hence + // the conversion was lossy and we consider that it failed + return wxCONV_FAILED; + } + } // note that it returns count of written chars for buf != NULL and size // of the needed buffer for buf == NULL so in either case the length of // the string (which never includes the terminating NUL) is one less - return len ? len - 1 : (size_t)-1; + return len - 1; } - size_t WC2MB(char *buf, const wchar_t *pwz, size_t n) const + virtual size_t WC2MB(char *buf, const wchar_t *pwz, size_t n) const { /* we have a problem here: by default, WideCharToMultiByte() may @@ -1490,7 +2146,7 @@ public: if ( !len ) { // function totally failed - return (size_t)-1; + return wxCONV_FAILED; } // if we were really converting, check if we succeeded @@ -1501,17 +2157,17 @@ public: // check if the conversion failed, i.e. if any replacements // were done if ( usedDef ) - return (size_t)-1; + return wxCONV_FAILED; } else // we must resort to double tripping... { wxWCharBuffer wcBuf(n); - if ( MB2WC(wcBuf.data(), buf, n) == (size_t)-1 || + if ( MB2WC(wcBuf.data(), buf, n) == wxCONV_FAILED || wcscmp(wcBuf, pwz) != 0 ) { // we didn't obtain the same thing we started from, hence // the conversion was lossy and we consider that it failed - return (size_t)-1; + return wxCONV_FAILED; } } } @@ -1520,6 +2176,47 @@ public: return len - 1; } + virtual size_t GetMBNulLen() const + { + if ( m_minMBCharWidth == 0 ) + { + int len = ::WideCharToMultiByte + ( + m_CodePage, // code page + 0, // no flags + L"", // input string + 1, // translate just the NUL + NULL, // output buffer + 0, // and its size + NULL, // no replacement char + NULL // [out] don't care if it was used + ); + + wxMBConv_win32 * const self = wxConstCast(this, wxMBConv_win32); + switch ( len ) + { + default: + wxLogDebug(_T("Unexpected NUL length %d"), len); + self->m_minMBCharWidth = (size_t)-1; + break; + + case 0: + self->m_minMBCharWidth = (size_t)-1; + break; + + case 1: + case 2: + case 4: + self->m_minMBCharWidth = len; + break; + } + } + + return m_minMBCharWidth; + } + + virtual wxMBConv *Clone() const { return new wxMBConv_win32(*this); } + bool IsOk() const { return m_CodePage != -1; } private: @@ -1541,8 +2238,9 @@ private: break; default: - // unknown, be conseravtive by default + // unknown: be conservative by default s_isWin98Or2k = 0; + break; } wxASSERT_MSG( s_isWin98Or2k != -1, _T("should be set above") ); @@ -1551,7 +2249,40 @@ private: return s_isWin98Or2k == 1; } + static bool IsAtLeastWin2kSP4() + { +#ifdef __WXWINCE__ + return false; +#else + static int s_isAtLeastWin2kSP4 = -1; + + if ( s_isAtLeastWin2kSP4 == -1 ) + { + OSVERSIONINFOEX ver; + + memset(&ver, 0, sizeof(ver)); + ver.dwOSVersionInfoSize = sizeof(ver); + GetVersionEx((OSVERSIONINFO*)&ver); + + s_isAtLeastWin2kSP4 = + ((ver.dwMajorVersion > 5) || // Vista+ + (ver.dwMajorVersion == 5 && ver.dwMinorVersion > 0) || // XP/2003 + (ver.dwMajorVersion == 5 && ver.dwMinorVersion == 0 && + ver.wServicePackMajor >= 4)) // 2000 SP4+ + ? 1 : 0; + } + + return s_isAtLeastWin2kSP4 == 1; +#endif + } + + + // the code page we're working with long m_CodePage; + + // cached result of GetMBNulLen(), set to 0 initially meaning + // "unknown" + size_t m_minMBCharWidth; }; #endif // wxHAVE_WIN32_MB2WC @@ -1562,9 +2293,9 @@ private: #if defined(__WXCOCOA__) -// RN: There is no UTF-32 support in either Core Foundation or -// Cocoa. Strangely enough, internally Core Foundation uses -// UTF 32 internally quite a bit - its just not public (yet). +// RN: There is no UTF-32 support in either Core Foundation or Cocoa. +// Strangely enough, internally Core Foundation uses +// UTF-32 internally quite a bit - its just not public (yet). #include #include @@ -1572,12 +2303,13 @@ private: CFStringEncoding wxCFStringEncFromFontEnc(wxFontEncoding encoding) { CFStringEncoding enc = kCFStringEncodingInvalidId ; - if ( encoding == wxFONTENCODING_DEFAULT ) - { - enc = CFStringGetSystemEncoding(); - } - else switch( encoding) + + switch (encoding) { + case wxFONTENCODING_DEFAULT : + enc = CFStringGetSystemEncoding(); + break ; + case wxFONTENCODING_ISO8859_1 : enc = kCFStringEncodingISOLatin1 ; break ; @@ -1633,7 +2365,7 @@ CFStringEncoding wxCFStringEncFromFontEnc(wxFontEncoding encoding) // break ; case wxFONTENCODING_CP437 : - enc =kCFStringEncodingDOSLatinUS ; + enc = kCFStringEncodingDOSLatinUS ; break ; case wxFONTENCODING_CP850 : enc = kCFStringEncodingDOSLatin1; @@ -1645,7 +2377,7 @@ CFStringEncoding wxCFStringEncFromFontEnc(wxFontEncoding encoding) enc = kCFStringEncodingDOSCyrillic; break ; case wxFONTENCODING_CP866 : - enc =kCFStringEncodingDOSRussian ; + enc = kCFStringEncodingDOSRussian ; break ; case wxFONTENCODING_CP874 : enc = kCFStringEncodingDOSThai; @@ -1654,7 +2386,7 @@ CFStringEncoding wxCFStringEncFromFontEnc(wxFontEncoding encoding) enc = kCFStringEncodingDOSJapanese; break ; case wxFONTENCODING_CP936 : - enc =kCFStringEncodingDOSChineseSimplif ; + enc = kCFStringEncodingDOSChineseSimplif ; break ; case wxFONTENCODING_CP949 : enc = kCFStringEncodingDOSKorean; @@ -1666,10 +2398,10 @@ CFStringEncoding wxCFStringEncFromFontEnc(wxFontEncoding encoding) enc = kCFStringEncodingWindowsLatin2; break ; case wxFONTENCODING_CP1251 : - enc =kCFStringEncodingWindowsCyrillic ; + enc = kCFStringEncodingWindowsCyrillic ; break ; case wxFONTENCODING_CP1252 : - enc =kCFStringEncodingWindowsLatin1 ; + enc = kCFStringEncodingWindowsLatin1 ; break ; case wxFONTENCODING_CP1253 : enc = kCFStringEncodingWindowsGreek; @@ -1678,10 +2410,10 @@ CFStringEncoding wxCFStringEncFromFontEnc(wxFontEncoding encoding) enc = kCFStringEncodingWindowsLatin5; break ; case wxFONTENCODING_CP1255 : - enc =kCFStringEncodingWindowsHebrew ; + enc = kCFStringEncodingWindowsHebrew ; break ; case wxFONTENCODING_CP1256 : - enc =kCFStringEncodingWindowsArabic ; + enc = kCFStringEncodingWindowsArabic ; break ; case wxFONTENCODING_CP1257 : enc = kCFStringEncodingWindowsBalticRim; @@ -1819,10 +2551,12 @@ CFStringEncoding wxCFStringEncFromFontEnc(wxFontEncoding encoding) // case wxFONTENCODING_MACKEYBOARD : // enc = kCFStringEncodingMacKeyboardGlyphs ; // break ; + default : // because gcc is picky break ; - } ; + } + return enc ; } @@ -1834,10 +2568,17 @@ public: Init(CFStringGetSystemEncoding()) ; } + wxMBConv_cocoa(const wxMBConv_cocoa& conv) + { + m_encoding = conv.m_encoding; + } + +#if wxUSE_FONTMAP wxMBConv_cocoa(const wxChar* name) { - Init( wxCFStringEncFromFontEnc(wxFontMapper::Get()->CharsetToEncoding(name, false) ) ) ; + Init( wxCFStringEncFromFontEnc(wxFontMapperBase::Get()->CharsetToEncoding(name, false) ) ) ; } +#endif wxMBConv_cocoa(wxFontEncoding encoding) { @@ -1880,35 +2621,35 @@ public: #if SIZEOF_WCHAR_T == 4 UniChar* szUniCharBuffer = new UniChar[nOutSize]; #endif - + CFStringGetCharacters(theString, theRange, szUniCharBuffer); - + CFRelease(theString); - szUniCharBuffer[nOutLength] = '\0' ; + szUniCharBuffer[nOutLength] = '\0'; #if SIZEOF_WCHAR_T == 4 - wxMBConvUTF16 converter ; - converter.MB2WC(szOut, (const char*)szUniCharBuffer , nOutSize ) ; - delete[] szUniCharBuffer; + wxMBConvUTF16 converter; + converter.MB2WC( szOut, (const char*)szUniCharBuffer, nOutSize ); + delete [] szUniCharBuffer; #endif - + return nOutLength; } size_t WC2MB(char *szOut, const wchar_t *szUnConv, size_t nOutSize) const { wxASSERT(szUnConv); - + size_t nRealOutSize; size_t nBufSize = wxWcslen(szUnConv); UniChar* szUniBuffer = (UniChar*) szUnConv; #if SIZEOF_WCHAR_T == 4 - wxMBConvUTF16BE converter ; - nBufSize = converter.WC2MB( NULL , szUnConv , 0 ); - szUniBuffer = new UniChar[ (nBufSize / sizeof(UniChar)) + 1] ; - converter.WC2MB( (char*) szUniBuffer , szUnConv, nBufSize + sizeof(UniChar)) ; + wxMBConvUTF16 converter ; + nBufSize = converter.WC2MB( NULL, szUnConv, 0 ); + szUniBuffer = new UniChar[ (nBufSize / sizeof(UniChar)) + 1]; + converter.WC2MB( (char*) szUniBuffer, szUnConv, nBufSize + sizeof(UniChar)); nBufSize /= sizeof(UniChar); #endif @@ -1927,7 +2668,7 @@ public: { if (szOut != NULL) CFStringGetCharacters(theString, CFRangeMake(0, nOutSize - 1), (UniChar*) szOut); - + nRealOutSize = CFStringGetLength(theString) + 1; } else @@ -1940,7 +2681,7 @@ public: //0 tells CFString to return NULL if it meets such a character false, //not an external representation (UInt8*) szOut, - nOutSize, + nOutSize, (CFIndex*) &nRealOutSize ); } @@ -1954,9 +2695,11 @@ public: return nRealOutSize - 1; } + virtual wxMBConv *Clone() const { return new wxMBConv_cocoa(*this); } + bool IsOk() const { - return m_encoding != kCFStringEncodingInvalidId && + return m_encoding != kCFStringEncodingInvalidId && CFStringIsEncodingAvailable(m_encoding); } @@ -1980,10 +2723,17 @@ public: Init(CFStringGetSystemEncoding()) ; } + wxMBConv_mac(const wxMBConv_mac& conv) + { + Init(conv.m_char_encoding); + } + +#if wxUSE_FONTMAP wxMBConv_mac(const wxChar* name) { - Init( wxMacGetSystemEncFromFontEnc(wxFontMapper::Get()->CharsetToEncoding(name, false) ) ) ; + Init( wxMacGetSystemEncFromFontEnc( wxFontMapperBase::Get()->CharsetToEncoding(name, false) ) ); } +#endif wxMBConv_mac(wxFontEncoding encoding) { @@ -1993,58 +2743,77 @@ public: ~wxMBConv_mac() { OSStatus status = noErr ; - status = TECDisposeConverter(m_MB2WC_converter); - status = TECDisposeConverter(m_WC2MB_converter); + if (m_MB2WC_converter) + status = TECDisposeConverter(m_MB2WC_converter); + if (m_WC2MB_converter) + status = TECDisposeConverter(m_WC2MB_converter); } - - void Init( TextEncodingBase encoding) + void Init( TextEncodingBase encoding,TextEncodingVariant encodingVariant = kTextEncodingDefaultVariant , + TextEncodingFormat encodingFormat = kTextEncodingDefaultFormat) { - OSStatus status = noErr ; - m_char_encoding = encoding ; - m_unicode_encoding = CreateTextEncoding(kTextEncodingUnicodeDefault,0,kUnicode16BitFormat) ; + m_MB2WC_converter = NULL ; + m_WC2MB_converter = NULL ; + m_char_encoding = CreateTextEncoding(encoding, encodingVariant, encodingFormat) ; + m_unicode_encoding = CreateTextEncoding(kTextEncodingUnicodeDefault, 0, kUnicode16BitFormat) ; + } - status = TECCreateConverter(&m_MB2WC_converter, + virtual void CreateIfNeeded() const + { + if ( m_MB2WC_converter == NULL && m_WC2MB_converter == NULL ) + { + OSStatus status = noErr ; + status = TECCreateConverter(&m_MB2WC_converter, m_char_encoding, m_unicode_encoding); - status = TECCreateConverter(&m_WC2MB_converter, + wxASSERT_MSG( status == noErr , _("Unable to create TextEncodingConverter")) ; + status = TECCreateConverter(&m_WC2MB_converter, m_unicode_encoding, m_char_encoding); + wxASSERT_MSG( status == noErr , _("Unable to create TextEncodingConverter")) ; + } } - + size_t MB2WC(wchar_t *buf, const char *psz, size_t n) const { + CreateIfNeeded() ; OSStatus status = noErr ; ByteCount byteOutLen ; - ByteCount byteInLen = strlen(psz) ; + ByteCount byteInLen = strlen(psz) + 1; wchar_t *tbuf = NULL ; UniChar* ubuf = NULL ; size_t res = 0 ; if (buf == NULL) { - //apple specs say at least 32 - n = wxMax( 32 , byteInLen ) ; - tbuf = (wchar_t*) malloc( n * SIZEOF_WCHAR_T) ; + // Apple specs say at least 32 + n = wxMax( 32, byteInLen ) ; + tbuf = (wchar_t*) malloc( n * SIZEOF_WCHAR_T ) ; } + ByteCount byteBufferLen = n * sizeof( UniChar ) ; + #if SIZEOF_WCHAR_T == 4 ubuf = (UniChar*) malloc( byteBufferLen + 2 ) ; #else ubuf = (UniChar*) (buf ? buf : tbuf) ; #endif - status = TECConvertText(m_MB2WC_converter, (ConstTextPtr) psz , byteInLen, &byteInLen, - (TextPtr) ubuf , byteBufferLen, &byteOutLen); + + status = TECConvertText( + m_MB2WC_converter, (ConstTextPtr) psz, byteInLen, &byteInLen, + (TextPtr) ubuf, byteBufferLen, &byteOutLen); + #if SIZEOF_WCHAR_T == 4 // we have to terminate here, because n might be larger for the trailing zero, and if UniChar // is not properly terminated we get random characters at the end ubuf[byteOutLen / sizeof( UniChar ) ] = 0 ; - wxMBConvUTF16BE converter ; - res = converter.MB2WC( (buf ? buf : tbuf) , (const char*)ubuf , n ) ; + wxMBConvUTF16 converter ; + res = converter.MB2WC( (buf ? buf : tbuf), (const char*)ubuf, n ) ; free( ubuf ) ; #else res = byteOutLen / sizeof( UniChar ) ; #endif + if ( buf == NULL ) free(tbuf) ; @@ -2056,6 +2825,7 @@ public: size_t WC2MB(char *buf, const wchar_t *psz, size_t n) const { + CreateIfNeeded() ; OSStatus status = noErr ; ByteCount byteOutLen ; ByteCount byteInLen = wxWcslen(psz) * SIZEOF_WCHAR_T ; @@ -2064,27 +2834,32 @@ public: if (buf == NULL) { - //apple specs say at least 32 - n = wxMax( 32 , ((byteInLen / SIZEOF_WCHAR_T) * 8) + SIZEOF_WCHAR_T ); + // Apple specs say at least 32 + n = wxMax( 32, ((byteInLen / SIZEOF_WCHAR_T) * 8) + SIZEOF_WCHAR_T ); tbuf = (char*) malloc( n ) ; } ByteCount byteBufferLen = n ; UniChar* ubuf = NULL ; + #if SIZEOF_WCHAR_T == 4 - wxMBConvUTF16BE converter ; - size_t unicharlen = converter.WC2MB( NULL , psz , 0 ) ; + wxMBConvUTF16 converter ; + size_t unicharlen = converter.WC2MB( NULL, psz, 0 ) ; byteInLen = unicharlen ; ubuf = (UniChar*) malloc( byteInLen + 2 ) ; - converter.WC2MB( (char*) ubuf , psz, unicharlen + 2 ) ; + converter.WC2MB( (char*) ubuf, psz, unicharlen + 2 ) ; #else ubuf = (UniChar*) psz ; #endif - status = TECConvertText(m_WC2MB_converter, (ConstTextPtr) ubuf , byteInLen, &byteInLen, - (TextPtr) (buf ? buf : tbuf) , byteBufferLen, &byteOutLen); + + status = TECConvertText( + m_WC2MB_converter, (ConstTextPtr) ubuf, byteInLen, &byteInLen, + (TextPtr) (buf ? buf : tbuf), byteBufferLen, &byteOutLen); + #if SIZEOF_WCHAR_T == 4 free( ubuf ) ; #endif + if ( buf == NULL ) free(tbuf) ; @@ -2092,35 +2867,213 @@ public: if ( buf && res < n) { buf[res] = 0; - + //we need to double-trip to verify it didn't insert any ? in place //of bogus characters wxWCharBuffer wcBuf(n); size_t pszlen = wxWcslen(psz); - if ( MB2WC(wcBuf.data(), buf, n) == (size_t)-1 || + if ( MB2WC(wcBuf.data(), buf, n) == wxCONV_FAILED || wxWcslen(wcBuf) != pszlen || memcmp(wcBuf, psz, pszlen * sizeof(wchar_t)) != 0 ) { // we didn't obtain the same thing we started from, hence // the conversion was lossy and we consider that it failed - return (size_t)-1; + return wxCONV_FAILED; } } return res ; } + virtual wxMBConv *Clone() const { return new wxMBConv_mac(*this); } + bool IsOk() const - { return m_MB2WC_converter != NULL && m_WC2MB_converter != NULL ; } + { + CreateIfNeeded() ; + return m_MB2WC_converter != NULL && m_WC2MB_converter != NULL; + } -private: - TECObjectRef m_MB2WC_converter ; - TECObjectRef m_WC2MB_converter ; +protected : + mutable TECObjectRef m_MB2WC_converter; + mutable TECObjectRef m_WC2MB_converter; - TextEncodingBase m_char_encoding ; - TextEncodingBase m_unicode_encoding ; + TextEncodingBase m_char_encoding; + TextEncodingBase m_unicode_encoding; }; +// MB is decomposed (D) normalized UTF8 + +class wxMBConv_macUTF8D : public wxMBConv_mac +{ +public : + wxMBConv_macUTF8D() + { + Init( kTextEncodingUnicodeDefault , kUnicodeNoSubset , kUnicodeUTF8Format ) ; + m_uni = NULL; + m_uniBack = NULL ; + } + + ~wxMBConv_macUTF8D() + { + if (m_uni!=NULL) + DisposeUnicodeToTextInfo(&m_uni); + if (m_uniBack!=NULL) + DisposeUnicodeToTextInfo(&m_uniBack); + } + + size_t WC2MB(char *buf, const wchar_t *psz, size_t n) const + { + CreateIfNeeded() ; + OSStatus status = noErr ; + ByteCount byteOutLen ; + ByteCount byteInLen = wxWcslen(psz) * SIZEOF_WCHAR_T ; + + char *tbuf = NULL ; + + if (buf == NULL) + { + // Apple specs say at least 32 + n = wxMax( 32, ((byteInLen / SIZEOF_WCHAR_T) * 8) + SIZEOF_WCHAR_T ); + tbuf = (char*) malloc( n ) ; + } + + ByteCount byteBufferLen = n ; + UniChar* ubuf = NULL ; + +#if SIZEOF_WCHAR_T == 4 + wxMBConvUTF16 converter ; + size_t unicharlen = converter.WC2MB( NULL, psz, 0 ) ; + byteInLen = unicharlen ; + ubuf = (UniChar*) malloc( byteInLen + 2 ) ; + converter.WC2MB( (char*) ubuf, psz, unicharlen + 2 ) ; +#else + ubuf = (UniChar*) psz ; +#endif + + // ubuf is a non-decomposed UniChar buffer + + ByteCount dcubuflen = byteInLen * 2 + 2 ; + ByteCount dcubufread , dcubufwritten ; + UniChar *dcubuf = (UniChar*) malloc( dcubuflen ) ; + + ConvertFromUnicodeToText( m_uni , byteInLen , ubuf , + kUnicodeDefaultDirectionMask, 0, NULL, NULL, NULL, dcubuflen , &dcubufread , &dcubufwritten , dcubuf ) ; + + // we now convert that decomposed buffer into UTF8 + + status = TECConvertText( + m_WC2MB_converter, (ConstTextPtr) dcubuf, dcubufwritten, &dcubufread, + (TextPtr) (buf ? buf : tbuf), byteBufferLen, &byteOutLen); + + free( dcubuf ); + +#if SIZEOF_WCHAR_T == 4 + free( ubuf ) ; +#endif + + if ( buf == NULL ) + free(tbuf) ; + + size_t res = byteOutLen ; + if ( buf && res < n) + { + buf[res] = 0; + // don't test for round-trip fidelity yet, we cannot guarantee it yet + } + + return res ; + } + + size_t MB2WC(wchar_t *buf, const char *psz, size_t n) const + { + CreateIfNeeded() ; + OSStatus status = noErr ; + ByteCount byteOutLen ; + ByteCount byteInLen = strlen(psz) + 1; + wchar_t *tbuf = NULL ; + UniChar* ubuf = NULL ; + size_t res = 0 ; + + if (buf == NULL) + { + // Apple specs say at least 32 + n = wxMax( 32, byteInLen ) ; + tbuf = (wchar_t*) malloc( n * SIZEOF_WCHAR_T ) ; + } + + ByteCount byteBufferLen = n * sizeof( UniChar ) ; + +#if SIZEOF_WCHAR_T == 4 + ubuf = (UniChar*) malloc( byteBufferLen + 2 ) ; +#else + ubuf = (UniChar*) (buf ? buf : tbuf) ; +#endif + + ByteCount dcubuflen = byteBufferLen * 2 + 2 ; + ByteCount dcubufread , dcubufwritten ; + UniChar *dcubuf = (UniChar*) malloc( dcubuflen ) ; + + status = TECConvertText( + m_MB2WC_converter, (ConstTextPtr) psz, byteInLen, &byteInLen, + (TextPtr) dcubuf, dcubuflen, &byteOutLen); + // we have to terminate here, because n might be larger for the trailing zero, and if UniChar + // is not properly terminated we get random characters at the end + dcubuf[byteOutLen / sizeof( UniChar ) ] = 0 ; + + // now from the decomposed UniChar to properly composed uniChar + ConvertFromUnicodeToText( m_uniBack , byteOutLen , dcubuf , + kUnicodeDefaultDirectionMask, 0, NULL, NULL, NULL, dcubuflen , &dcubufread , &dcubufwritten , ubuf ) ; + + free( dcubuf ); + byteOutLen = dcubufwritten ; + ubuf[byteOutLen / sizeof( UniChar ) ] = 0 ; + + +#if SIZEOF_WCHAR_T == 4 + wxMBConvUTF16 converter ; + res = converter.MB2WC( (buf ? buf : tbuf), (const char*)ubuf, n ) ; + free( ubuf ) ; +#else + res = byteOutLen / sizeof( UniChar ) ; +#endif + + if ( buf == NULL ) + free(tbuf) ; + + if ( buf && res < n) + buf[res] = 0; + + return res ; + } + + virtual void CreateIfNeeded() const + { + wxMBConv_mac::CreateIfNeeded() ; + if ( m_uni == NULL ) + { + m_map.unicodeEncoding = CreateTextEncoding(kTextEncodingUnicodeDefault, + kUnicodeNoSubset, kTextEncodingDefaultFormat); + m_map.otherEncoding = CreateTextEncoding(kTextEncodingUnicodeDefault, + kUnicodeCanonicalDecompVariant, kTextEncodingDefaultFormat); + m_map.mappingVersion = kUnicodeUseLatestMapping; + + OSStatus err = CreateUnicodeToTextInfo(&m_map, &m_uni); + wxASSERT_MSG( err == noErr , _(" Couldn't create the UnicodeConverter")) ; + + m_map.unicodeEncoding = CreateTextEncoding(kTextEncodingUnicodeDefault, + kUnicodeNoSubset, kTextEncodingDefaultFormat); + m_map.otherEncoding = CreateTextEncoding(kTextEncodingUnicodeDefault, + kUnicodeCanonicalCompVariant, kTextEncodingDefaultFormat); + m_map.mappingVersion = kUnicodeUseLatestMapping; + err = CreateUnicodeToTextInfo(&m_map, &m_uniBack); + wxASSERT_MSG( err == noErr , _(" Couldn't create the UnicodeConverter")) ; + } + } +protected : + mutable UnicodeToTextInfo m_uni; + mutable UnicodeToTextInfo m_uniBack; + mutable UnicodeMapping m_map; +}; #endif // defined(__WXMAC__) && defined(TARGET_CARBON) // ============================================================================ @@ -2144,7 +3097,7 @@ public: wxMBConv_wxwin(const wxChar* name) { if (name) - m_enc = wxFontMapper::Get()->CharsetToEncoding(name, false); + m_enc = wxFontMapperBase::Get()->CharsetToEncoding(name, false); else m_enc = wxFONTENCODING_SYSTEM; @@ -2162,7 +3115,10 @@ public: { size_t inbuf = strlen(psz); if (buf) - m2w.Convert(psz,buf); + { + if (!m2w.Convert(psz, buf)) + return wxCONV_FAILED; + } return inbuf; } @@ -2170,23 +3126,59 @@ public: { const size_t inbuf = wxWcslen(psz); if (buf) - w2m.Convert(psz,buf); + { + if (!w2m.Convert(psz, buf)) + return wxCONV_FAILED; + } return inbuf; } + virtual size_t GetMBNulLen() const + { + switch ( m_enc ) + { + case wxFONTENCODING_UTF16BE: + case wxFONTENCODING_UTF16LE: + return 2; + + case wxFONTENCODING_UTF32BE: + case wxFONTENCODING_UTF32LE: + return 4; + + default: + return 1; + } + } + + virtual wxMBConv *Clone() const { return new wxMBConv_wxwin(m_enc); } + bool IsOk() const { return m_ok; } public: wxFontEncoding m_enc; wxEncodingConverter m2w, w2m; +private: // were we initialized successfully? bool m_ok; DECLARE_NO_COPY_CLASS(wxMBConv_wxwin) }; +// make the constructors available for unit testing +WXDLLIMPEXP_BASE wxMBConv* new_wxMBConv_wxwin( const wxChar* name ) +{ + wxMBConv_wxwin* result = new wxMBConv_wxwin( name ); + if ( !result->IsOk() ) + { + delete result; + return 0; + } + + return result; +} + #endif // wxUSE_FONTMAP // ============================================================================ @@ -2209,7 +3201,11 @@ wxCSConv::wxCSConv(const wxChar *charset) SetName(charset); } +#if wxUSE_FONTMAP + m_encoding = wxFontMapperBase::GetEncodingFromName(charset); +#else m_encoding = wxFONTENCODING_SYSTEM; +#endif } wxCSConv::wxCSConv(wxFontEncoding encoding) @@ -2268,12 +3264,28 @@ void wxCSConv::SetName(const wxChar *charset) } } +#if wxUSE_FONTMAP + +WX_DECLARE_HASH_MAP( wxFontEncoding, wxString, wxIntegerHash, wxIntegerEqual, + wxEncodingNameCache ); + +static wxEncodingNameCache gs_nameCache; +#endif + wxMBConv *wxCSConv::DoCreate() const { +#if wxUSE_FONTMAP + wxLogTrace(TRACE_STRCONV, + wxT("creating conversion for %s"), + (m_name ? m_name + : wxFontMapperBase::GetEncodingName(m_encoding).c_str())); +#endif // wxUSE_FONTMAP + // check for the special case of ASCII or ISO8859-1 charset: as we have // special knowledge of it anyhow, we don't need to create a special // conversion object - if ( m_encoding == wxFONTENCODING_ISO8859_1 ) + if ( m_encoding == wxFONTENCODING_ISO8859_1 || + m_encoding == wxFONTENCODING_DEFAULT ) { // don't convert at all return NULL; @@ -2294,17 +3306,61 @@ wxMBConv *wxCSConv::DoCreate() const #endif // !wxUSE_FONTMAP { wxString name(m_name); +#if wxUSE_FONTMAP + wxFontEncoding encoding(m_encoding); +#endif + + if ( !name.empty() ) + { + wxMBConv_iconv *conv = new wxMBConv_iconv(name); + if ( conv->IsOk() ) + return conv; + + delete conv; #if wxUSE_FONTMAP - if ( name.empty() ) - name = wxFontMapper::Get()->GetEncodingName(m_encoding); + encoding = + wxFontMapperBase::Get()->CharsetToEncoding(name, false); #endif // wxUSE_FONTMAP + } +#if wxUSE_FONTMAP + { + const wxEncodingNameCache::iterator it = gs_nameCache.find(encoding); + if ( it != gs_nameCache.end() ) + { + if ( it->second.empty() ) + return NULL; - wxMBConv_iconv *conv = new wxMBConv_iconv(name); - if ( conv->IsOk() ) - return conv; + wxMBConv_iconv *conv = new wxMBConv_iconv(it->second); + if ( conv->IsOk() ) + return conv; - delete conv; + delete conv; + } + + const wxChar** names = wxFontMapperBase::GetAllEncodingNames(encoding); + // CS : in case this does not return valid names (eg for MacRoman) encoding + // got a 'failure' entry in the cache all the same, although it just has to + // be created using a different method, so only store failed iconv creation + // attempts (or perhaps we shoulnd't do this at all ?) + if ( names[0] != NULL ) + { + for ( ; *names; ++names ) + { + wxMBConv_iconv *conv = new wxMBConv_iconv(*names); + if ( conv->IsOk() ) + { + gs_nameCache[encoding] = *names; + return conv; + } + + delete conv; + } + + gs_nameCache[encoding] = _T(""); // cache the failure + } + } +#endif // wxUSE_FONTMAP } #endif // HAVE_ICONV @@ -2322,13 +3378,19 @@ wxMBConv *wxCSConv::DoCreate() const #endif } #endif // wxHAVE_WIN32_MB2WC + #if defined(__WXMAC__) { - if ( m_name || ( m_encoding < wxFONTENCODING_UTF16BE ) ) + // leave UTF16 and UTF32 to the built-ins of wx + if ( m_name || ( m_encoding < wxFONTENCODING_UTF16BE || + ( m_encoding >= wxFONTENCODING_MACMIN && m_encoding <= wxFONTENCODING_MACMAX ) ) ) { - +#if wxUSE_FONTMAP wxMBConv_mac *conv = m_name ? new wxMBConv_mac(m_name) : new wxMBConv_mac(m_encoding); +#else + wxMBConv_mac *conv = new wxMBConv_mac(m_encoding); +#endif if ( conv->IsOk() ) return conv; @@ -2336,13 +3398,18 @@ wxMBConv *wxCSConv::DoCreate() const } } #endif + #if defined(__WXCOCOA__) { if ( m_name || ( m_encoding <= wxFONTENCODING_UTF16 ) ) { - +#if wxUSE_FONTMAP wxMBConv_cocoa *conv = m_name ? new wxMBConv_cocoa(m_name) : new wxMBConv_cocoa(m_encoding); +#else + wxMBConv_cocoa *conv = new wxMBConv_cocoa(m_encoding); +#endif + if ( conv->IsOk() ) return conv; @@ -2358,7 +3425,7 @@ wxMBConv *wxCSConv::DoCreate() const // use "false" to suppress interactive dialogs -- we can be called from // anywhere and popping up a dialog from here is the last thing we want to // do - enc = wxFontMapper::Get()->CharsetToEncoding(m_name, false); + enc = wxFontMapperBase::Get()->CharsetToEncoding(m_name, false); } #endif // wxUSE_FONTMAP @@ -2384,7 +3451,7 @@ wxMBConv *wxCSConv::DoCreate() const default: // nothing to do but put here to suppress gcc warnings - ; + break; } // step (3) @@ -2402,10 +3469,10 @@ wxMBConv *wxCSConv::DoCreate() const // NB: This is a hack to prevent deadlock. What could otherwise happen // in Unicode build: wxConvLocal creation ends up being here // because of some failure and logs the error. But wxLog will try to - // attach timestamp, for which it will need wxConvLocal (to convert - // time to char* and then wchar_t*), but that fails, tries to log - // error, but wxLog has a (already locked) critical section that - // guards static buffer. + // attach a timestamp, for which it will need wxConvLocal (to convert + // time to char* and then wchar_t*), but that fails, tries to log the + // error, but wxLog has an (already locked) critical section that + // guards the static buffer. static bool alreadyLoggingError = false; if (!alreadyLoggingError) { @@ -2414,11 +3481,12 @@ wxMBConv *wxCSConv::DoCreate() const m_name ? m_name : #if wxUSE_FONTMAP - wxFontMapper::GetEncodingDescription(m_encoding).c_str() + wxFontMapperBase::GetEncodingDescription(m_encoding).c_str() #else // !wxUSE_FONTMAP - wxString::Format(_("encoding %s"), m_encoding).c_str() + wxString::Format(_("encoding %i"), m_encoding).c_str() #endif // wxUSE_FONTMAP/!wxUSE_FONTMAP ); + alreadyLoggingError = false; } @@ -2478,7 +3546,8 @@ size_t wxCSConv::WC2MB(char *buf, const wchar_t *psz, size_t n) const for (size_t c = 0; c <= len; c++) { if (psz[c] > 0xFF) - return (size_t)-1; + return wxCONV_FAILED; + buf[c] = (char)psz[c]; } } @@ -2487,13 +3556,25 @@ size_t wxCSConv::WC2MB(char *buf, const wchar_t *psz, size_t n) const for (size_t c = 0; c <= len; c++) { if (psz[c] > 0xFF) - return (size_t)-1; + return wxCONV_FAILED; } } return len; } +size_t wxCSConv::GetMBNulLen() const +{ + CreateConvIfNeeded(); + + if ( m_convReal ) + { + return m_convReal->GetMBNulLen(); + } + + return 1; +} + // ---------------------------------------------------------------------------- // globals // ---------------------------------------------------------------------------- @@ -2510,14 +3591,26 @@ static wxCSConv wxConvLocalObj(wxFONTENCODING_SYSTEM); static wxCSConv wxConvISO8859_1Obj(wxFONTENCODING_ISO8859_1); static wxMBConvUTF7 wxConvUTF7Obj; static wxMBConvUTF8 wxConvUTF8Obj; - - +#if defined(__WXMAC__) && defined(TARGET_CARBON) +static wxMBConv_macUTF8D wxConvMacUTF8DObj; +#endif WXDLLIMPEXP_DATA_BASE(wxMBConv&) wxConvLibc = wxConvLibcObj; WXDLLIMPEXP_DATA_BASE(wxCSConv&) wxConvLocal = wxConvLocalObj; WXDLLIMPEXP_DATA_BASE(wxCSConv&) wxConvISO8859_1 = wxConvISO8859_1Obj; WXDLLIMPEXP_DATA_BASE(wxMBConvUTF7&) wxConvUTF7 = wxConvUTF7Obj; WXDLLIMPEXP_DATA_BASE(wxMBConvUTF8&) wxConvUTF8 = wxConvUTF8Obj; WXDLLIMPEXP_DATA_BASE(wxMBConv *) wxConvCurrent = &wxConvLibcObj; +WXDLLIMPEXP_DATA_BASE(wxMBConv *) wxConvUI = &wxConvLocal; +WXDLLIMPEXP_DATA_BASE(wxMBConv *) wxConvFileName = & +#ifdef __WXOSX__ +#if defined(__WXMAC__) && defined(TARGET_CARBON) + wxConvMacUTF8DObj; +#else + wxConvUTF8Obj; +#endif +#else + wxConvLibcObj; +#endif #else // !wxUSE_WCHAR_T @@ -2528,5 +3621,3 @@ WXDLLIMPEXP_DATA_BASE(wxMBConv) wxConvLibc, wxConvUTF8; #endif // wxUSE_WCHAR_T/!wxUSE_WCHAR_T - -