X-Git-Url: https://git.saurik.com/wxWidgets.git/blobdiff_plain/fd5e476dbb9245a9ec755eeaa5b0e4b4556d2f8c..eea4d01c65f9b29baa1193db762b4c6b8144af24:/src/common/strconv.cpp diff --git a/src/common/strconv.cpp b/src/common/strconv.cpp index dfd5295d29..170ad719bd 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) @@ -12,41 +12,21 @@ // 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 + #pragma hdrstop +#endif //__BORLANDC__ #ifndef WX_PRECOMP #include "wx/intl.h" #include "wx/log.h" -#endif // WX_PRECOMP - -#include "wx/strconv.h" - -#if wxUSE_WCHAR_T - -#ifdef __WXMSW__ - #include "wx/msw/private.h" + #include "wx/utils.h" + #include "wx/hashmap.h" #endif -#ifdef __WINDOWS__ - #include "wx/msw/missing.h" -#endif +#include "wx/strconv.h" #ifndef __WXWINCE__ #include @@ -57,104 +37,86 @@ #include #if defined(__WIN32__) && !defined(__WXMICROWIN__) + #include "wx/msw/private.h" + #include "wx/msw/missing.h" #define wxHAVE_WIN32_MB2WC -#endif // __WIN32__ but !__WXMICROWIN__ - -// ---------------------------------------------------------------------------- -// headers -// ---------------------------------------------------------------------------- - -#ifdef __SALFORDC__ - #include #endif #ifdef HAVE_ICONV #include + #include "wx/thread.h" #endif #include "wx/encconv.h" #include "wx/fontmap.h" -#include "wx/utils.h" -#ifdef __WXMAC__ -#include -#include -#include +#ifdef __DARWIN__ +#include "wx/osx/core/private/strconv_cf.h" +#endif //def __DARWIN__ -#include "wx/mac/private.h" // includes mac headers -#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 +125,290 @@ 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(reinterpret_cast(*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 existing 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 + // + // moreover, some conversion classes simply can't implement ToWChar() + // directly, the primary example is wxConvLibc: mbstowcs() only handles + // NUL-terminated strings + + // 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; + } + + // the idea of this code is straightforward: it converts a NUL-terminated + // chunk of the string during each iteration and updates the output buffer + // with the result + // + // all the complication come from the fact that this function, for + // historical reasons, must behave in 2 subtly different ways when it's + // called with a fixed number of characters and when it's called for the + // entire NUL-terminated string: in the former case (srcEnd != NULL) we + // must count all characters we convert, NUL or not; but in the latter we + // do not count the trailing NUL -- but still count all the NULs inside the + // string + // + // so for the (simple) former case we just always count the trailing NUL, + // but for the latter we need to wait until we see if there is going to be + // another loop iteration and only count it then + for ( ;; ) + { + // try to convert the current chunk + size_t lenChunk = MB2WC(NULL, src, 0); + if ( lenChunk == wxCONV_FAILED ) + return wxCONV_FAILED; + + dstWritten += lenChunk; + if ( !srcEnd ) + dstWritten++; + + if ( !lenChunk ) + { + // nothing left in the input string, conversion succeeded + break; + } + + if ( dst ) + { + if ( dstWritten > dstLen ) + return wxCONV_FAILED; + + // +1 is for trailing NUL + if ( MB2WC(dst, src, lenChunk + 1) == wxCONV_FAILED ) + return wxCONV_FAILED; + + dst += lenChunk; + if ( !srcEnd ) + dst++; + } + + if ( !srcEnd ) + { + // we convert just one chunk in this case as this is the entire + // string anyhow (and we don't count the trailing NUL in this case) + break; + } + + // advance the input pointer past the end of this chunk: notice that we + // will always stop before srcEnd because we know that the chunk is + // always properly NUL-terminated + 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; + } + + // if the buffer ends before this NUL, we shouldn't count it in our + // output so skip the code below + if ( src == srcEnd ) + break; + + // do count this terminator as it's inside the buffer we convert + dstWritten++; + if ( dst ) + dst++; + + src += nulLen; // skip the terminator itself + + 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; + + // if we don't know its length we have no choice but to assume that it is + // NUL-terminated (notice that it can still be NUL-terminated even if + // explicit length is given but it doesn't change our return value) + const bool isNulTerminated = srcLen == wxNO_LEN; + + // make a copy of the input string unless it is already properly + // NUL-terminated + wxWCharBuffer bufTmp; + if ( isNulTerminated ) + { + 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++ /* skip L'\0' too */ ) + { + // try to convert the current chunk + size_t lenChunk = WC2MB(NULL, src, 0); + if ( lenChunk == wxCONV_FAILED ) + return wxCONV_FAILED; + + dstWritten += lenChunk; + + const wchar_t * const + chunkEnd = isNulTerminated ? srcEnd - 1 : src + wxWcslen(src); + + // our return value accounts for the trailing NUL(s), unlike that of + // WC2MB(), however don't do it for the last NUL we artificially added + // ourselves above + if ( chunkEnd < srcEnd ) + dstWritten += lenNul; + + if ( dst ) + { + if ( dstWritten > dstLen ) + return wxCONV_FAILED; + + // if we know that there is enough space in the destination buffer + // (because we accounted for lenNul in dstWritten above), we can + // convert directly in place -- but otherwise we need another + // temporary buffer to ensure that we don't overwrite the output + wxCharBuffer dstBuf; + char *dstTmp; + if ( chunkEnd == srcEnd ) + { + dstBuf = wxCharBuffer(lenChunk + lenNul - 1); + dstTmp = dstBuf.data(); + } + else + { + dstTmp = dst; + } + + if ( WC2MB(dstTmp, src, lenChunk + lenNul) == wxCONV_FAILED ) + return wxCONV_FAILED; + + if ( dstTmp != dst ) + { + // copy everything up to but excluding the terminating NUL(s) + // into the real output buffer + memcpy(dst, dstTmp, lenChunk); + + // micro-optimization: if dstTmp != dst it means that chunkEnd + // == srcEnd and so we're done, no need to update anything below + break; + } + + dst += lenChunk; + if ( chunkEnd < srcEnd ) + dst += lenNul; + } + + src = chunkEnd; + } + + 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 +419,143 @@ 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 = ToWChar(NULL, 0, psz); + 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 /* +1 added implicitly */); + + // +1 for the trailing NULL + if ( ToWChar(buf.data(), nLen, psz) != 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 = FromWChar(NULL, 0, pwz); + 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 ) - { + wxCharBuffer buf(nLen - 1); + if ( FromWChar(buf.data(), nLen, pwz) != wxCONV_FAILED ) return buf; + } + } + + return wxCharBuffer(); +} + +const wxWCharBuffer +wxMBConv::cMB2WC(const char *inBuff, size_t inLen, size_t *outLen) const +{ + const size_t dstLen = ToWChar(NULL, 0, inBuff, inLen); + if ( dstLen != wxCONV_FAILED ) + { + // notice that we allocate space for dstLen+1 wide characters here + // because we want the buffer to always be NUL-terminated, even if the + // input isn't (as otherwise the caller has no way to know its length) + wxWCharBuffer wbuf(dstLen); + wbuf.data()[dstLen] = L'\0'; + if ( ToWChar(wbuf.data(), dstLen, inBuff, inLen) != wxCONV_FAILED ) + { + if ( outLen ) + { + *outLen = dstLen; + + // we also need to handle NUL-terminated input strings + // specially: for them the output is the length of the string + // excluding the trailing NUL, however if we're asked to + // convert a specific number of characters we return the length + // of the resulting output even if it's NUL-terminated + if ( inLen == wxNO_LEN ) + (*outLen)--; } + + return wbuf; } } - wxCharBuffer buf((char *)NULL); + if ( outLen ) + *outLen = 0; - return buf; + return wxWCharBuffer(); } -size_t wxMBConv::MB2WC(wchar_t* szBuffer, const char* szString, - size_t outsize, size_t nStringLen) const +const wxCharBuffer +wxMBConv::cWC2MB(const wchar_t *inBuff, size_t inLen, size_t *outLen) const { - const char* szEnd = szString + nStringLen + 1; - const char* szPos = szString; - const char* szStart = szPos; + size_t dstLen = FromWChar(NULL, 0, inBuff, inLen); + if ( dstLen != wxCONV_FAILED ) + { + const size_t nulLen = GetMBNulLen(); - size_t nActualLength = 0; + // as above, ensure that the buffer is always NUL-terminated, even if + // the input is not + wxCharBuffer buf(dstLen + nulLen - 1); + memset(buf.data() + dstLen, 0, nulLen); + if ( FromWChar(buf.data(), dstLen, inBuff, inLen) != wxCONV_FAILED ) + { + if ( outLen ) + { + *outLen = dstLen; - //Convert the string until the length() is reached, continuing the - //loop every time a null character is reached - while(szPos != szEnd) - { - wxASSERT(szPos < szEnd); //something is _really_ screwed up if this rings true + if ( inLen == wxNO_LEN ) + { + // in this case both input and output are NUL-terminated + // and we're not supposed to count NUL + *outLen -= nulLen; + } + } - //Get the length of the current (sub)string - size_t nLen = MB2WC(NULL, szPos, 0); + return buf; + } + } - //Invalid conversion? - if( nLen == (size_t)-1 ) - return nLen; + if ( outLen ) + *outLen = 0; - //Increase the actual length (+1 for current null character) - nActualLength += nLen + 1; + return wxCharBuffer(); +} - //Only copy data in if buffer size is big enough - if (szBuffer != NULL && - nActualLength <= outsize) +const wxWCharBuffer wxMBConv::cMB2WC(const wxScopedCharBuffer& buf) const +{ + const size_t srcLen = buf.length(); + if ( srcLen ) + { + const size_t dstLen = ToWChar(NULL, 0, buf, srcLen); + if ( dstLen != wxCONV_FAILED ) { - //Convert the current (sub)string - if ( MB2WC(&szBuffer[szPos - szStart], szPos, nLen + 1) == (size_t)-1 ) - return (size_t)-1; + wxWCharBuffer wbuf(dstLen); + wbuf.data()[dstLen] = L'\0'; + if ( ToWChar(wbuf.data(), dstLen, buf, srcLen) != wxCONV_FAILED ) + return wbuf; } - - //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 nActualLength - 1; //success - return actual length + return wxScopedWCharBuffer::CreateNonOwned(L"", 0); } -size_t wxMBConv::WC2MB(char* szBuffer, const wchar_t* szString, - size_t outsize, size_t nStringLen) const +const wxCharBuffer wxMBConv::cWC2MB(const wxScopedWCharBuffer& wbuf) 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) + const size_t srcLen = wbuf.length(); + if ( srcLen ) { - 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); - - //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) + const size_t dstLen = FromWChar(NULL, 0, wbuf, srcLen); + if ( dstLen != wxCONV_FAILED ) { - //Convert the current (sub)string - if(WC2MB(&szBuffer[szPos - szStart], szPos, nLen + 1) == (size_t)-1 ) - return (size_t)-1; + wxCharBuffer buf(dstLen); + buf.data()[dstLen] = '\0'; + if ( FromWChar(buf.data(), dstLen, wbuf, srcLen) != wxCONV_FAILED ) + 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 + return wxScopedCharBuffer::CreateNonOwned("", 0); } // ---------------------------------------------------------------------------- @@ -318,11 +571,31 @@ size_t wxMBConvLibc::WC2MB(char *buf, const wchar_t *psz, size_t n) const { return wxWC2MB(buf, psz, n); } + +// ---------------------------------------------------------------------------- +// wxConvBrokenFileNames +// ---------------------------------------------------------------------------- + +#ifdef __UNIX__ + +wxConvBrokenFileNames::wxConvBrokenFileNames(const wxString& charset) +{ + if ( wxStricmp(charset, wxT("UTF-8")) == 0 || + wxStricmp(charset, wxT("UTF8")) == 0 ) + m_conv = new wxMBConvUTF8(wxMBConvUTF8::MAP_INVALID_UTF8_TO_PUA); + else + m_conv = new wxCSConv(charset); +} + +#endif // __UNIX__ + // ---------------------------------------------------------------------------- -// UTF-7 +// UTF-7 // ---------------------------------------------------------------------------- // Implementation (C) 2004 Fredrik Roubert +// +// Changes to work in streaming mode (C) 2008 Vadim Zeitlin // // BASE64 decoding table @@ -363,60 +636,150 @@ static const unsigned char utf7unb64[] = 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; -size_t wxMBConvUTF7::MB2WC(wchar_t *buf, const char *psz, size_t n) const +size_t wxMBConvUTF7::ToWChar(wchar_t *dst, size_t dstLen, + const char *src, size_t srcLen) const { + DecoderState stateOrig, + *statePtr; + if ( srcLen == wxNO_LEN ) + { + // convert the entire string, up to and including the trailing NUL + srcLen = strlen(src) + 1; + + // when working on the entire strings we don't update nor use the shift + // state from the previous call + statePtr = &stateOrig; + } + else // when working with partial strings we do use the shift state + { + statePtr = const_cast(&m_stateDecoder); + + // also save the old state to be able to rollback to it on error + stateOrig = m_stateDecoder; + } + // but to simplify the code below we use this variable in both cases + DecoderState& state = *statePtr; + + + // number of characters [which would have been] written to dst [if it were + // not NULL] size_t len = 0; - while (*psz && ((!buf) || (len < n))) + const char * const srcEnd = src + srcLen; + + while ( (src < srcEnd) && (!dst || (len < dstLen)) ) { - unsigned char cc = *psz++; - if (cc != '+') - { - // plain ASCII char - if (buf) - *buf++ = cc; - len++; - } - else if (*psz == '-') - { - // encoded plus sign - if (buf) - *buf++ = cc; - len++; - psz++; - } - else + const unsigned char cc = *src++; + + if ( state.IsShifted() ) { - // BASE64 encoded string - bool lsb; - unsigned char c; - unsigned int d, l; - for (lsb = false, d = 0, l = 0; - (cc = utf7unb64[(unsigned char)*psz]) != 0xff; psz++) + const unsigned char dc = utf7unb64[cc]; + if ( dc == 0xff ) + { + // end of encoded part, check that nothing was left: there can + // be up to 4 bits of 0 padding but nothing else (we also need + // to check isLSB as we count bits modulo 8 while a valid UTF-7 + // encoded sequence must contain an integral number of UTF-16 + // characters) + if ( state.isLSB || state.bit > 4 || + (state.accum & ((1 << state.bit) - 1)) ) + { + if ( !len ) + state = stateOrig; + + return wxCONV_FAILED; + } + + state.ToDirect(); + + // re-parse this character normally below unless it's '-' which + // is consumed by the decoder + if ( cc == '-' ) + continue; + } + else // valid encoded character { - d <<= 6; - d += cc; - for (l += 6; l >= 8; lsb = !lsb) + // mini base64 decoder: each character is 6 bits + state.bit += 6; + state.accum <<= 6; + state.accum += dc; + + if ( state.bit >= 8 ) { - c = (d >> (l -= 8)) % 256; - if (lsb) + // got the full byte, consume it + state.bit -= 8; + unsigned char b = (state.accum >> state.bit) & 0x00ff; + + if ( state.isLSB ) { - if (buf) - *buf++ |= c; - len ++; + // we've got the full word, output it + if ( dst ) + *dst++ = (state.msb << 8) | b; + len++; + state.isLSB = false; } - else - if (buf) - *buf = c << 8; + else // MSB + { + // just store it while we wait for LSB + state.msb = b; + state.isLSB = true; + } + } + } + } + + if ( state.IsDirect() ) + { + // start of an encoded segment? + if ( cc == '+' ) + { + if ( *src == '-' ) + { + // just the encoded plus sign, don't switch to shifted mode + if ( dst ) + *dst++ = '+'; + len++; + src++; + } + else if ( utf7unb64[(unsigned)*src] == 0xff ) + { + // empty encoded chunks are not allowed + if ( !len ) + state = stateOrig; + + return wxCONV_FAILED; + } + else // base-64 encoded chunk follows + { + state.ToShifted(); } } - if (*psz == '-') - psz++; + else // not '+' + { + // only printable 7 bit ASCII characters (with the exception of + // NUL, TAB, CR and LF) can be used directly + if ( cc >= 0x7f || (cc < ' ' && + !(cc == '\0' || cc == '\t' || cc == '\r' || cc == '\n')) ) + return wxCONV_FAILED; + + if ( dst ) + *dst++ = cc; + len++; + } } } - if (buf && (len < n)) - *buf = 0; + + if ( !len ) + { + // as we didn't read any characters we should be called with the same + // data (followed by some more new data) again later so don't save our + // state + state = stateOrig; + + return wxCONV_FAILED; + } + return len; } @@ -445,7 +808,7 @@ static const unsigned char utf7enb64[] = // static const unsigned char utf7encode[128] = { - 3, 3, 3, 3, 3, 3, 3, 3, 3, 2, 2, 3, 3, 2, 3, 3, + 0, 3, 3, 3, 3, 3, 3, 3, 3, 2, 2, 3, 3, 2, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 2, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 3, 0, 0, 0, 3, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 0, @@ -455,72 +818,121 @@ 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 +static inline bool wxIsUTF7Direct(wchar_t wc) +{ + return wc < 0x80 && utf7encode[wc] < 1; +} + +size_t wxMBConvUTF7::FromWChar(char *dst, size_t dstLen, + const wchar_t *src, size_t srcLen) const { + EncoderState stateOrig, + *statePtr; + if ( srcLen == wxNO_LEN ) + { + // we don't apply the stored state when operating on entire strings at + // once + statePtr = &stateOrig; + + srcLen = wxWcslen(src) + 1; + } + else // do use the mode we left the output in previously + { + stateOrig = m_stateEncoder; + statePtr = const_cast(&m_stateEncoder); + } + + EncoderState& state = *statePtr; size_t len = 0; - while (*psz && ((!buf) || (len < n))) + const wchar_t * const srcEnd = src + srcLen; + while ( src < srcEnd && (!dst || len < dstLen) ) { - wchar_t cc = *psz++; - if (cc < 0x80 && utf7encode[cc] < 1) - { - // plain ASCII char - if (buf) - *buf++ = (char)cc; - len++; - } -#ifndef WC_UTF16 - else if (cc > ((const wchar_t)0xffff)) + wchar_t cc = *src++; + if ( wxIsUTF7Direct(cc) ) { - // no surrogate pair generation (yet?) - return (size_t)-1; - } -#endif + if ( state.IsShifted() ) + { + // pad with zeros the last encoded block if necessary + if ( state.bit ) + { + if ( dst ) + *dst++ = utf7enb64[((state.accum % 16) << (6 - state.bit)) % 64]; + len++; + } + + state.ToDirect(); + + if ( dst ) + *dst++ = '-'; + len++; + } + + if ( dst ) + *dst++ = (char)cc; + len++; + } + else if ( cc == '+' && state.IsDirect() ) + { + if ( dst ) + { + *dst++ = '+'; + *dst++ = '-'; + } + + len += 2; + } +#ifndef WC_UTF16 + else if (((wxUint32)cc) > 0xffff) + { + // no surrogate pair generation (yet?) + return wxCONV_FAILED; + } +#endif else { - if (buf) - *buf++ = '+'; - len++; - if (cc != '+') + if ( state.IsDirect() ) { - // BASE64 encode string - unsigned int lsb, d, l; - for (d = 0, l = 0;; psz++) + state.ToShifted(); + + if ( dst ) + *dst++ = '+'; + len++; + } + + // BASE64 encode string + for ( ;; ) + { + for ( unsigned lsb = 0; lsb < 2; lsb++ ) { - for (lsb = 0; lsb < 2; lsb ++) - { - d <<= 8; - d += lsb ? cc & 0xff : (cc & 0xff00) >> 8; + state.accum <<= 8; + state.accum += lsb ? cc & 0xff : (cc & 0xff00) >> 8; - for (l += 8; l >= 6; ) - { - l -= 6; - if (buf) - *buf++ = utf7enb64[(d >> l) % 64]; - len++; - } + for (state.bit += 8; state.bit >= 6; ) + { + state.bit -= 6; + if ( dst ) + *dst++ = utf7enb64[(state.accum >> state.bit) % 64]; + 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 ( src == srcEnd || wxIsUTF7Direct(cc = *src) ) + break; + + src++; } - if (buf) - *buf++ = '-'; - len++; } } - if (buf && (len < n)) - *buf = 0; + + // we need to restore the original encoder state if we were called just to + // calculate the amount of space needed as we will presumably be called + // again to really convert the data now + if ( !dst ) + state = stateOrig; + return len; } @@ -528,25 +940,320 @@ size_t wxMBConvUTF7::WC2MB(char *buf, const wchar_t // UTF-8 // ---------------------------------------------------------------------------- -static wxUint32 utf8_max[]= +static const wxUint32 utf8_max[]= { 0x7f, 0x7ff, 0xffff, 0x1fffff, 0x3ffffff, 0x7fffffff, 0xffffffff }; -size_t wxMBConvUTF8::MB2WC(wchar_t *buf, const char *psz, size_t n) const +// 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; + +// this table gives the length of the UTF-8 encoding from its first character: +const unsigned char tableUtf8Lengths[256] = { + // single-byte sequences (ASCII): + 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, // 00..0F + 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, // 10..1F + 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, // 20..2F + 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, // 30..3F + 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, // 40..4F + 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, // 50..5F + 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, // 60..6F + 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, // 70..7F + + // these are invalid: + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, // 80..8F + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, // 90..9F + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, // A0..AF + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, // B0..BF + 0, 0, // C0,C1 + + // two-byte sequences: + 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, // C2..CF + 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, // D0..DF + + // three-byte sequences: + 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, // E0..EF + + // four-byte sequences: + 4, 4, 4, 4, 4, // F0..F4 + + // these are invalid again (5- or 6-byte + // sequences and sequences for code points + // above U+10FFFF, as restricted by RFC 3629): + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 // F5..FF +}; + +size_t +wxMBConvStrictUTF8::ToWChar(wchar_t *dst, size_t dstLen, + const char *src, size_t srcLen) const +{ + wchar_t *out = dstLen ? dst : NULL; + size_t written = 0; + + if ( srcLen == wxNO_LEN ) + srcLen = strlen(src) + 1; + + for ( const char *p = src; ; p++ ) + { + if ( (srcLen == wxNO_LEN ? !*p : !srcLen) ) + { + // all done successfully, just add the trailing NULL if we are not + // using explicit length + if ( srcLen == wxNO_LEN ) + { + if ( out ) + { + if ( !dstLen ) + break; + + *out = L'\0'; + } + + written++; + } + + return written; + } + + if ( out && !dstLen-- ) + break; + + wxUint32 code; + unsigned char c = *p; + + if ( c < 0x80 ) + { + if ( srcLen == 0 ) // the test works for wxNO_LEN too + break; + + if ( srcLen != wxNO_LEN ) + srcLen--; + + code = c; + } + else + { + unsigned len = tableUtf8Lengths[c]; + if ( !len ) + break; + + if ( srcLen < len ) // the test works for wxNO_LEN too + break; + + if ( srcLen != wxNO_LEN ) + srcLen -= len; + + // Char. number range | UTF-8 octet sequence + // (hexadecimal) | (binary) + // ----------------------+---------------------------------------- + // 0000 0000 - 0000 007F | 0xxxxxxx + // 0000 0080 - 0000 07FF | 110xxxxx 10xxxxxx + // 0000 0800 - 0000 FFFF | 1110xxxx 10xxxxxx 10xxxxxx + // 0001 0000 - 0010 FFFF | 11110xxx 10xxxxxx 10xxxxxx 10xxxxxx + // + // Code point value is stored in bits marked with 'x', + // lowest-order bit of the value on the right side in the diagram + // above. (from RFC 3629) + + // mask to extract lead byte's value ('x' bits above), by sequence + // length: + static const unsigned char leadValueMask[] = { 0x7F, 0x1F, 0x0F, 0x07 }; + + // mask and value of lead byte's most significant bits, by length: + static const unsigned char leadMarkerMask[] = { 0x80, 0xE0, 0xF0, 0xF8 }; + static const unsigned char leadMarkerVal[] = { 0x00, 0xC0, 0xE0, 0xF0 }; + + len--; // it's more convenient to work with 0-based length here + + // extract the lead byte's value bits: + if ( (c & leadMarkerMask[len]) != leadMarkerVal[len] ) + break; + + code = c & leadValueMask[len]; + + // all remaining bytes, if any, are handled in the same way + // regardless of sequence's length: + for ( ; len; --len ) + { + c = *++p; + if ( (c & 0xC0) != 0x80 ) + return wxCONV_FAILED; + + code <<= 6; + code |= c & 0x3F; + } + } + +#ifdef WC_UTF16 + // cast is ok because wchar_t == wxUint16 if WC_UTF16 + if ( encode_utf16(code, (wxUint16 *)out) == 2 ) + { + if ( out ) + out++; + written++; + } +#else // !WC_UTF16 + if ( out ) + *out = code; +#endif // WC_UTF16/!WC_UTF16 + + if ( out ) + out++; + + written++; + } + + return wxCONV_FAILED; +} + +size_t +wxMBConvStrictUTF8::FromWChar(char *dst, size_t dstLen, + const wchar_t *src, size_t srcLen) const { + char *out = dstLen ? dst : NULL; + size_t written = 0; + + for ( const wchar_t *wp = src; ; wp++ ) + { + if ( (srcLen == wxNO_LEN ? !*wp : !srcLen) ) + { + // all done successfully, just add the trailing NULL if we are not + // using explicit length + if ( srcLen == wxNO_LEN ) + { + if ( out ) + { + if ( !dstLen ) + break; + + *out = '\0'; + } + + written++; + } + + return written; + } + + if ( srcLen != wxNO_LEN ) + srcLen--; + + wxUint32 code; +#ifdef WC_UTF16 + // cast is ok for WC_UTF16 + if ( decode_utf16((const wxUint16 *)wp, code) == 2 ) + { + // skip the next char too as we decoded a surrogate + wp++; + } +#else // wchar_t is UTF-32 + code = *wp & 0x7fffffff; +#endif + + unsigned len; + if ( code <= 0x7F ) + { + len = 1; + if ( out ) + { + if ( dstLen < len ) + break; + + out[0] = (char)code; + } + } + else if ( code <= 0x07FF ) + { + len = 2; + if ( out ) + { + if ( dstLen < len ) + break; + + // NB: this line takes 6 least significant bits, encodes them as + // 10xxxxxx and discards them so that the next byte can be encoded: + out[1] = 0x80 | (code & 0x3F); code >>= 6; + out[0] = 0xC0 | code; + } + } + else if ( code < 0xFFFF ) + { + len = 3; + if ( out ) + { + if ( dstLen < len ) + break; + + out[2] = 0x80 | (code & 0x3F); code >>= 6; + out[1] = 0x80 | (code & 0x3F); code >>= 6; + out[0] = 0xE0 | code; + } + } + else if ( code <= 0x10FFFF ) + { + len = 4; + if ( out ) + { + if ( dstLen < len ) + break; + + out[3] = 0x80 | (code & 0x3F); code >>= 6; + out[2] = 0x80 | (code & 0x3F); code >>= 6; + out[1] = 0x80 | (code & 0x3F); code >>= 6; + out[0] = 0xF0 | code; + } + } + else + { + wxFAIL_MSG( wxT("trying to encode undefined Unicode character") ); + break; + } + + if ( out ) + { + out += len; + dstLen -= len; + } + + written += len; + } + + // we only get here if an error occurs during decoding + return wxCONV_FAILED; +} + +size_t wxMBConvUTF8::ToWChar(wchar_t *buf, size_t n, + const char *psz, size_t srcLen) const +{ + if ( m_options == MAP_INVALID_UTF8_NOT ) + return wxMBConvStrictUTF8::ToWChar(buf, n, psz, srcLen); + size_t len = 0; - while (*psz && ((!buf) || (len < n))) + while ((srcLen == wxNO_LEN ? *psz : srcLen--) && ((!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 { @@ -554,7 +1261,7 @@ 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 { @@ -562,87 +1269,197 @@ 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 == wxUint16 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)) + + if (srcLen == wxNO_LEN && buf && (len < n)) *buf = 0; - return len; + + return len + 1; +} + +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 wxMBConvUTF8::FromWChar(char *buf, size_t n, + const wchar_t *psz, size_t srcLen) const { + if ( m_options == MAP_INVALID_UTF8_NOT ) + return wxMBConvStrictUTF8::FromWChar(buf, n, psz, srcLen); + size_t len = 0; - while (*psz && ((!buf) || (len < n))) + while ((srcLen == wxNO_LEN ? *psz : srcLen--) && ((!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(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-16 characters + if ( srcLen % BYTES_PER_CHAR ) + return wxCONV_FAILED; + } + + return srcLen; +} + +// case when in-memory representation is UTF-16 too #ifdef WC_UTF16 -// copy 16bit MB to 16bit String -size_t wxMBConvUTF16straight::MB2WC(wchar_t *buf, const char *psz, size_t n) const +// ---------------------------------------------------------------------------- +// conversions without endianness change +// ---------------------------------------------------------------------------- + +size_t +wxMBConvUTF16straight::ToWChar(wchar_t *dst, size_t dstLen, + const char *src, size_t srcLen) const { - size_t len=0; + // set up the scene for using memcpy() (which is presumably more efficient + // than copying the bytes one by one) + 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 ) { - if (buf) - *buf++ = *(wxUint16*)psz; - len++; + if ( dstLen < inLen ) + return wxCONV_FAILED; - psz += sizeof(wxUint16); + memcpy(dst, src, srcLen); } - if (buf && len(src); + for ( size_t n = 0; n < srcLen; n++, inBuff++ ) { - ((char *)buf)[0] = psz[1]; - ((char *)buf)[1] = psz[0]; - buf++; + *dst++ = wxUINT16_SWAP_ALWAYS(*inBuff); } - len++; - psz += sizeof(wxUint16); } - if (buf && len(dst); + for ( size_t n = 0; n < srcLen; n += BYTES_PER_CHAR, src++ ) { - *buf++ = ((char*)psz)[1]; - *buf++ = ((char*)psz)[0]; + *outBuff++ = wxUINT16_SWAP_ALWAYS(*src); } - len += sizeof(wxUint16); - psz++; } - if (buf && len<=n-sizeof(wxUint16)) *(wxUint16*)buf=0; - return len; + return srcLen; } +#else // !WC_UTF16: wchar_t is UTF-32 -#else // WC_UTF16 - +// ---------------------------------------------------------------------------- +// conversions without endianness change +// ---------------------------------------------------------------------------- -// copy 16bit MB to 32bit String -size_t wxMBConvUTF16straight::MB2WC(wchar_t *buf, const char *psz, size_t n) const +size_t +wxMBConvUTF16straight::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; - size_t pa=decode_utf16((wxUint16*)psz, cc); - if (pa == (size_t)-1) - return pa; + // 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; + } - if (buf) - *buf++ = cc; - len++; - psz += pa * sizeof(wxUint16); + size_t outLen = 0; + const wxUint16 *inBuff = reinterpret_cast(src); + for ( const wxUint16 * const inEnd = inBuff + inLen; inBuff < inEnd; ) + { + const wxUint32 ch = wxDecodeSurrogate(&inBuff); + if ( !inBuff ) + return wxCONV_FAILED; + + if ( ++outLen > dstLen ) + return wxCONV_FAILED; + + *dst++ = ch; } - if (buf && len(dst); + for ( size_t n = 0; n < srcLen; n++ ) { - wxUint16 cc[2]; - size_t pa=encode_utf16(*psz, cc); + wxUint16 cc[2] = { 0 }; + const size_t numChars = encode_utf16(*src++, cc); + if ( numChars == wxCONV_FAILED ) + return wxCONV_FAILED; - if (pa == (size_t)-1) - return pa; - - if (buf) + outLen += numChars * BYTES_PER_CHAR; + if ( outBuff ) { - *(wxUint16*)buf = cc[0]; - buf += sizeof(wxUint16); - if (pa > 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 = reinterpret_cast(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(dst); + for ( const wchar_t *srcEnd = src + srcLen; src < srcEnd; src++ ) { - wxUint16 cc[2]; - size_t pa=encode_utf16(*psz, cc); - - if (pa == (size_t)-1) - return pa; + wxUint16 cc[2] = { 0 }; + const size_t numChars = encode_utf16(*src, cc); + if ( numChars == wxCONV_FAILED ) + return wxCONV_FAILED; - if (buf) + outLen += numChars * BYTES_PER_CHAR; + if ( outBuff ) { - *buf++ = ((char*)cc)[1]; - *buf++ = ((char*)cc)[0]; - if (pa > 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; - -#ifdef WC_UTF16 - -// copy 32bit MB to 16bit String -size_t wxMBConvUTF32straight::MB2WC(wchar_t *buf, const char *psz, size_t n) const +/* static */ +size_t wxMBConvUTF32Base::GetLength(const char *src, size_t srcLen) { - size_t len=0; - - while (*(wxUint32*)psz && (!buf || len < n)) + if ( srcLen == wxNO_LEN ) { - wxUint16 cc[2]; + // count the number of bytes in input, including the trailing NULs + const wxUint32 *inBuff = reinterpret_cast(src); + for ( srcLen = 1; *inBuff++; srcLen++ ) + ; - size_t pa=encode_utf16(*(wxUint32*)psz, cc); - if (pa == (size_t)-1) - return pa; - - if (buf) - { - *buf++ = cc[0]; - if (pa > 1) - *buf++ = cc[1]; - } - len += pa; - psz += sizeof(wxUint32); + 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; } - if (buf && len(src); + const size_t inLen = srcLen / BYTES_PER_CHAR; + size_t outLen = 0; + for ( size_t n = 0; n < inLen; n++ ) + { + wxUint16 cc[2] = { 0 }; + const size_t numChars = encode_utf16(*inBuff++, cc); + if ( numChars == wxCONV_FAILED ) + return wxCONV_FAILED; + + outLen += numChars; + if ( dst ) { - *(wxUint32*)buf = cc; - buf += sizeof(wxUint32); + if ( outLen > dstLen ) + return wxCONV_FAILED; + + *dst++ = cc[0]; + if ( numChars == 2 ) + { + // second character of a surrogate + *dst++ = cc[1]; + } } - len += sizeof(wxUint32); - psz += pa; } - if (buf && len<=n-sizeof(wxUint32)) - *(wxUint32*)buf=0; - - return len; + return outLen; } - - -// swap 32bit MB to 16bit String -size_t wxMBConvUTF32swap::MB2WC(wchar_t *buf, const char *psz, size_t n) const +size_t +wxMBConvUTF32straight::FromWChar(char *dst, size_t dstLen, + const wchar_t *src, size_t srcLen) const { - size_t len=0; + if ( srcLen == wxNO_LEN ) + srcLen = wxWcslen(src) + 1; - while (*(wxUint32*)psz && (!buf || len < n)) + if ( !dst ) { - char tmp[4]; - tmp[0] = psz[3]; tmp[1] = psz[2]; - tmp[2] = psz[1]; tmp[3] = psz[0]; + // optimization: return maximal space which could be needed for this + // string instead of the exact amount which could be less if there are + // any surrogates in the input + // + // we consider that surrogates are rare enough to make it worthwhile to + // avoid running the loop below at the cost of slightly extra memory + // consumption + return srcLen * BYTES_PER_CHAR; + } + wxUint32 *outBuff = reinterpret_cast(dst); + size_t outLen = 0; + for ( const wchar_t * const srcEnd = src + srcLen; src < srcEnd; ) + { + const wxUint32 ch = wxDecodeSurrogate(&src); + if ( !src ) + return wxCONV_FAILED; - wxUint16 cc[2]; + outLen += BYTES_PER_CHAR; - size_t pa=encode_utf16(*(wxUint32*)tmp, cc); - if (pa == (size_t)-1) - return pa; + if ( outLen > dstLen ) + return wxCONV_FAILED; - if (buf) + *outBuff++ = ch; + } + + return outLen; +} + +// ---------------------------------------------------------------------------- +// endian-reversing conversions +// ---------------------------------------------------------------------------- + +size_t +wxMBConvUTF32swap::ToWChar(wchar_t *dst, size_t dstLen, + const char *src, size_t srcLen) const +{ + srcLen = GetLength(src, srcLen); + if ( srcLen == wxNO_LEN ) + return wxCONV_FAILED; + + const wxUint32 *inBuff = reinterpret_cast(src); + const size_t inLen = srcLen / BYTES_PER_CHAR; + size_t outLen = 0; + for ( size_t n = 0; n < inLen; n++, inBuff++ ) + { + wxUint16 cc[2] = { 0 }; + const size_t numChars = encode_utf16(wxUINT32_SWAP_ALWAYS(*inBuff), cc); + if ( numChars == wxCONV_FAILED ) + return wxCONV_FAILED; + + 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(dst); + size_t outLen = 0; + for ( const wchar_t * const srcEnd = src + srcLen; src < srcEnd; ) + { + const wxUint32 ch = wxDecodeSurrogate(&src); + if ( !src ) + return wxCONV_FAILED; - if (buf) - { - *buf++ = cc[3]; - *buf++ = cc[2]; - *buf++ = cc[1]; - *buf++ = cc[0]; - } - len += sizeof(wxUint32); - psz += pa; - } + outLen += BYTES_PER_CHAR; - if (buf && len<=n-sizeof(wxUint32)) - *(wxUint32*)buf=0; + if ( outLen > 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 && len(src); + for ( size_t n = 0; n < srcLen; n++, inBuff++ ) { - ((char *)buf)[0] = psz[3]; - ((char *)buf)[1] = psz[2]; - ((char *)buf)[2] = psz[1]; - ((char *)buf)[3] = psz[0]; - buf++; + *dst++ = wxUINT32_SWAP_ALWAYS(*inBuff); } - len++; - psz += sizeof(wxUint32); } - if (buf && len(dst); + for ( size_t n = 0; n < srcLen; n += BYTES_PER_CHAR, src++ ) { - *buf++ = ((char *)psz)[3]; - *buf++ = ((char *)psz)[2]; - *buf++ = ((char *)psz)[1]; - *buf++ = ((char *)psz)[0]; + *outBuff++ = wxUINT32_SWAP_ALWAYS(*src); } - len += sizeof(wxUint32); - psz++; } - if (buf && len<=n-sizeof(wxUint32)) - *(wxUint32*)buf=0; - - return len; + return srcLen; } - -#endif // WC_UTF16 +#endif // WC_UTF16/!WC_UTF16 // ============================================================================ @@ -1111,12 +2060,13 @@ size_t wxMBConvUTF32swap::WC2MB(char *buf, const wchar_t *psz, size_t n) const #ifdef HAVE_ICONV -// VS: glibc 2.1.3 is broken in that iconv() conversion to/from UCS4 fails with E2BIG -// if output buffer is _exactly_ as big as needed. Such case is (unless there's -// yet another bug in glibc) the only case when iconv() returns with (size_t)-1 -// (which means error) and says there are 0 bytes left in the input buffer -- -// when _real_ error occurs, bytes-left-in-input buffer is non-zero. Hence, -// this alternative test for iconv() failure. +// VS: glibc 2.1.3 is broken in that iconv() conversion to/from UCS4 fails with +// E2BIG if output buffer is _exactly_ as big as needed. Such case is +// (unless there's yet another bug in glibc) the only case when iconv() +// returns with (size_t)-1 (which means error) and says there are 0 bytes +// left in the input buffer -- when _real_ error occurs, +// bytes-left-in-input buffer is non-zero. Hence, this alternative test for +// iconv() failure. // [This bug does not appear in glibc 2.2.] #if defined(__GLIBC__) && __GLIBC__ == 2 && __GLIBC_MINOR__ <= 1 #define ICONV_FAILED(cres, bufLeft) ((cres == (size_t)-1) && \ @@ -1127,6 +2077,19 @@ size_t wxMBConvUTF32swap::WC2MB(char *buf, const wchar_t *psz, size_t n) const #define ICONV_CHAR_CAST(x) ((ICONV_CONST char **)(x)) +#define ICONV_T_INVALID ((iconv_t)-1) + +#if SIZEOF_WCHAR_T == 4 + #define WC_BSWAP wxUINT32_SWAP_ALWAYS + #define WC_ENC wxFONTENCODING_UTF32 +#elif SIZEOF_WCHAR_T == 2 + #define WC_BSWAP wxUINT16_SWAP_ALWAYS + #define WC_ENC wxFONTENCODING_UTF16 +#else // sizeof(wchar_t) != 2 nor 4 + // does this ever happen? + #error "Unknown sizeof(wchar_t): please report this to wx-dev@lists.wxwindows.org" +#endif + // ---------------------------------------------------------------------------- // wxMBConv_iconv: encapsulates an iconv character set // ---------------------------------------------------------------------------- @@ -1134,190 +2097,319 @@ size_t wxMBConvUTF32swap::WC2MB(char *buf, const wchar_t *psz, size_t n) const class wxMBConv_iconv : public wxMBConv { public: - wxMBConv_iconv(const wxChar *name); + wxMBConv_iconv(const char *name); virtual ~wxMBConv_iconv(); - virtual size_t MB2WC(wchar_t *buf, const char *psz, size_t n) const; - virtual size_t WC2MB(char *buf, const wchar_t *psz, size_t n) const; + // implement base class virtual methods + virtual size_t ToWChar(wchar_t *dst, size_t dstLen, + const char *src, size_t srcLen = wxNO_LEN) const; + virtual size_t FromWChar(char *dst, size_t dstLen, + const wchar_t *src, size_t srcLen = wxNO_LEN) const; + virtual size_t GetMBNulLen() const; + +#if wxUSE_UNICODE_UTF8 + virtual bool IsUTF8() const; +#endif + + virtual wxMBConv *Clone() const + { + wxMBConv_iconv *p = new wxMBConv_iconv(m_name); + p->m_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 + const char *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 char* 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) +wxMBConv_iconv::wxMBConv_iconv(const char *name) + : m_name(wxStrdup(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; // check for charset that represents wchar_t: - if (ms_wcCharsetName == NULL) + if ( ms_wcCharsetName.empty() ) { - ms_wcNeedsSwap = false; + wxLogTrace(TRACE_STRCONV, wxT("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 *const *names = wxFontMapperBase::GetAllEncodingNames(WC_ENC); +#else // !wxUSE_FONTMAP + static const wxChar *const names_static[] = + { +#if SIZEOF_WCHAR_T == 4 + wxT("UCS-4"), +#elif SIZEOF_WCHAR_T == 2 + wxT("UCS-2"), +#endif + NULL + }; + const wxChar *const *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 += wxT("BE"); +#else // little endian + nameXE += wxT("LE"); +#endif - if (m2w != (iconv_t)-1) + wxLogTrace(TRACE_STRCONV, wxT(" trying charset \"%s\""), + nameXE.c_str()); + + m2w = iconv_open(nameXE.ToAscii(), name); + 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, wxT(" trying charset \"%s\""), + nameCS.c_str()); + m2w = iconv_open(nameCS.ToAscii(), name); + + // 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]; + size_t insz, outsz; + size_t res; + + buf[0] = 'A'; + buf[1] = 0; + wbuf[0] = 0; + insz = 2; + outsz = SIZEOF_WCHAR_T * 2; + char* wbufPtr = (char*)wbuf; + bufPtr = buf; + + res = iconv( + m2w, ICONV_CHAR_CAST(&bufPtr), &insz, + &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() ? wxString("") + : ms_wcCharsetName, + ms_wcNeedsSwap ? wxT(" (needs swap)") + : wxT("")); } else // we already have ms_wcCharsetName { - m2w = iconv_open(ms_wcCharsetName, cname); + m2w = iconv_open(ms_wcCharsetName.ToAscii(), name); } - // 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(name, ms_wcCharsetName.ToAscii()); + if ( w2m == ICONV_T_INVALID ) + { + wxLogTrace(TRACE_STRCONV, + wxT("\"%s\" -> \"%s\" works but not the converse!?"), + ms_wcCharsetName.c_str(), name); + } } } wxMBConv_iconv::~wxMBConv_iconv() { - if ( m2w != (iconv_t)-1 ) + free(const_cast(m_name)); + + 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 +wxMBConv_iconv::ToWChar(wchar_t *dst, size_t dstLen, + const char *src, size_t srcLen) const { - size_t inbuf = strlen(psz); - size_t outbuf = n * SIZEOF_WCHAR_T; + if ( srcLen == wxNO_LEN ) + { + // find the string length: notice that must be done differently for + // NUL-terminated strings and UTF-16/32 which are terminated with 2/4 + // consecutive NULs + const size_t nulLen = GetMBNulLen(); + switch ( nulLen ) + { + default: + return wxCONV_FAILED; + + case 1: + srcLen = strlen(src); // 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 = src; NotAllNULs(p, nulLen); p += nulLen ) + ; + srcLen = p - src; + break; + } + + // when we're determining the length of the string ourselves we count + // the terminating NUL(s) as part of it and always NUL-terminate the + // output + srcLen += nulLen; + } + + // we express length in the number of (wide) characters but iconv always + // counts buffer sizes it in bytes + dstLen *= SIZEOF_WCHAR_T; + +#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 res, cres; - // VS: Use these instead of psz, buf because iconv() modifies its arguments: - wchar_t *bufPtr = buf; - const char *pszPtr = psz; + const char *pszPtr = src; - if (buf) + if ( dst ) { + char* bufPtr = (char*)dst; + // have destination buffer, convert there + size_t dstLenOrig = dstLen; cres = iconv(m2w, - ICONV_CHAR_CAST(&pszPtr), &inbuf, - (char**)&bufPtr, &outbuf); - res = n - (outbuf / SIZEOF_WCHAR_T); + ICONV_CHAR_CAST(&pszPtr), &srcLen, + &bufPtr, &dstLen); + + // convert the number of bytes converted as returned by iconv to the + // number of (wide) characters converted that we need + res = (dstLenOrig - dstLen) / SIZEOF_WCHAR_T; if (ms_wcNeedsSwap) { // convert to native endianness - WC_BSWAP(buf /* _not_ bufPtr */, res) + for ( unsigned i = 0; i < res; i++ ) + dst[i] = WC_BSWAP(dst[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. - if (res < n) - buf[res] = 0; } - else + else // no destination buffer { - // no destination buffer... convert using temp buffer - // to calculate destination buffer requirement - wchar_t tbuf[8]; + // convert using temp buffer to calculate the size of the buffer needed + wchar_t tbuf[256]; res = 0; - do { - bufPtr = tbuf; - outbuf = 8*SIZEOF_WCHAR_T; + + do + { + char* bufPtr = (char*)tbuf; + dstLen = 8 * SIZEOF_WCHAR_T; cres = iconv(m2w, - ICONV_CHAR_CAST(&pszPtr), &inbuf, - (char**)&bufPtr, &outbuf ); + ICONV_CHAR_CAST(&pszPtr), &srcLen, + &bufPtr, &dstLen ); - res += 8-(outbuf/SIZEOF_WCHAR_T); - } while ((cres==(size_t)-1) && (errno==E2BIG)); + res += 8 - (dstLen / SIZEOF_WCHAR_T); + } + while ((cres == (size_t)-1) && (errno == E2BIG)); } - if (ICONV_FAILED(cres, inbuf)) + if (ICONV_FAILED(cres, srcLen)) { //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::WC2MB(char *buf, const wchar_t *psz, size_t n) const +size_t wxMBConv_iconv::FromWChar(char *dst, size_t dstLen, + const wchar_t *src, size_t srcLen) const { - size_t inbuf = wxWcslen(psz) * SIZEOF_WCHAR_T; - size_t outbuf = n; +#if wxUSE_THREADS + // NB: explained in MB2WC + wxMutexLocker lock(wxConstCast(this, wxMBConv_iconv)->m_iconvMutex); +#endif + + if ( srcLen == wxNO_LEN ) + srcLen = wxWcslen(src) + 1; + + size_t inbuflen = srcLen * SIZEOF_WCHAR_T; + size_t outbuflen = dstLen; size_t res, cres; wchar_t *tmpbuf = 0; @@ -1325,41 +2417,38 @@ 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 work, 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(inbuflen); + for ( size_t i = 0; i < srcLen; i++ ) + tmpbuf[i] = WC_BSWAP(src[i]); + + src = tmpbuf; } - if (buf) + char* inbuf = (char*)src; + if ( dst ) { // have destination buffer, convert there - cres = iconv( w2m, ICONV_CHAR_CAST(&psz), &inbuf, &buf, &outbuf ); + cres = iconv(w2m, ICONV_CHAR_CAST(&inbuf), &inbuflen, &dst, &outbuflen); - 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 - // if there's some room left for it in the output buffer. - if (res < n) - buf[0] = 0; + res = dstLen - outbuflen; } - else + else // no destination buffer { - // no destination buffer... convert using temp buffer - // to calculate destination buffer requirement - char tbuf[16]; + // convert using temp buffer to calculate the size of the buffer needed + char tbuf[256]; res = 0; - do { - buf = tbuf; outbuf = 16; + do + { + dst = tbuf; + outbuflen = WXSIZEOF(tbuf); - cres = iconv( w2m, ICONV_CHAR_CAST(&psz), &inbuf, &buf, &outbuf ); + cres = iconv(w2m, ICONV_CHAR_CAST(&inbuf), &inbuflen, &dst, &outbuflen); - res += 16 - outbuf; - } while ((cres==(size_t)-1) && (errno==E2BIG)); + res += WXSIZEOF(tbuf) - outbuflen; + } + while ((cres == (size_t)-1) && (errno == E2BIG)); } if (ms_wcNeedsSwap) @@ -1367,16 +2456,53 @@ size_t wxMBConv_iconv::WC2MB(char *buf, const wchar_t *psz, size_t n) const free(tmpbuf); } - if (ICONV_FAILED(cres, inbuf)) + if (ICONV_FAILED(cres, inbuflen)) { - //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 + + const 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; +} + +#if wxUSE_UNICODE_UTF8 +bool wxMBConv_iconv::IsUTF8() const +{ + return wxStricmp(m_name, "UTF-8") == 0 || + wxStricmp(m_name, "UTF8") == 0; +} +#endif + #endif // HAVE_ICONV @@ -1388,7 +2514,7 @@ size_t wxMBConv_iconv::WC2MB(char *buf, const wchar_t *psz, size_t n) const // from utils.cpp #if wxUSE_FONTMAP -extern WXDLLIMPEXP_BASE long wxCharsetToCodepage(const wxChar *charset); +extern WXDLLIMPEXP_BASE long wxCharsetToCodepage(const char *charset); extern WXDLLIMPEXP_BASE long wxEncodingToCodepage(wxFontEncoding encoding); #endif @@ -1398,44 +2524,107 @@ 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) + wxMBConv_win32(const char* 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 wxMBConvUTF8().MB2WC(buf, psz, n); + } + + if ( m_CodePage == CP_UTF7 ) + { + return wxMBConvUTF7().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 @@ -1461,939 +2650,184 @@ public: int flags; if ( CanUseNoBestFit() && m_CodePage < 50000 ) { - // it's our lucky day - flags = WC_NO_BEST_FIT_CHARS; - pUsedDef = &usedDef; - } - else // old system or unsupported encoding - { - flags = 0; - pUsedDef = NULL; - } - - const size_t len = ::WideCharToMultiByte - ( - m_CodePage, // code page - flags, // either none or no best fit - pwz, // input string - -1, // it is (wide) NUL-terminated - buf, // output buffer - buf ? n : 0, // and its size - NULL, // default "replacement" char - pUsedDef // [out] was it used? - ); - - if ( !len ) - { - // function totally failed - return (size_t)-1; - } - - // if we were really converting, check if we succeeded - if ( buf ) - { - if ( flags ) - { - // check if the conversion failed, i.e. if any replacements - // were done - if ( usedDef ) - return (size_t)-1; - } - else // we must resort to double tripping... - { - wxWCharBuffer wcBuf(n); - if ( MB2WC(wcBuf.data(), buf, n) == (size_t)-1 || - 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; - } - } - } - - // see the comment above for the reason of "len - 1" - return len - 1; - } - - bool IsOk() const { return m_CodePage != -1; } - -private: - static bool CanUseNoBestFit() - { - static int s_isWin98Or2k = -1; - - if ( s_isWin98Or2k == -1 ) - { - int verMaj, verMin; - switch ( wxGetOsVersion(&verMaj, &verMin) ) - { - case wxWIN95: - s_isWin98Or2k = verMaj >= 4 && verMin >= 10; - break; - - case wxWINDOWS_NT: - s_isWin98Or2k = verMaj >= 5; - break; - - default: - // unknown, be conseravtive by default - s_isWin98Or2k = 0; - } - - wxASSERT_MSG( s_isWin98Or2k != -1, _T("should be set above") ); - } - - return s_isWin98Or2k == 1; - } - - long m_CodePage; -}; - -#endif // wxHAVE_WIN32_MB2WC - -// ============================================================================ -// Cocoa conversion classes -// ============================================================================ - -#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). - -#include -#include - -CFStringEncoding wxCFStringEncFromFontEnc(wxFontEncoding encoding) -{ - CFStringEncoding enc = 0 ; - if ( encoding == wxFONTENCODING_DEFAULT ) - { -#if wxUSE_GUI - encoding = wxFont::GetDefaultEncoding() ; -#else - encoding = wxLocale::GetSystemEncoding() ; -#endif - } - else switch( encoding) - { - case wxFONTENCODING_ISO8859_1 : - enc = kCFStringEncodingISOLatin1 ; - break ; - case wxFONTENCODING_ISO8859_2 : - enc = kCFStringEncodingISOLatin2; - break ; - case wxFONTENCODING_ISO8859_3 : - enc = kCFStringEncodingISOLatin3 ; - break ; - case wxFONTENCODING_ISO8859_4 : - enc = kCFStringEncodingISOLatin4; - break ; - case wxFONTENCODING_ISO8859_5 : - enc = kCFStringEncodingISOLatinCyrillic; - break ; - case wxFONTENCODING_ISO8859_6 : - enc = kCFStringEncodingISOLatinArabic; - break ; - case wxFONTENCODING_ISO8859_7 : - enc = kCFStringEncodingISOLatinGreek; - break ; - case wxFONTENCODING_ISO8859_8 : - enc = kCFStringEncodingISOLatinHebrew; - break ; - case wxFONTENCODING_ISO8859_9 : - enc = kCFStringEncodingISOLatin5; - break ; - case wxFONTENCODING_ISO8859_10 : - enc = kCFStringEncodingISOLatin6; - break ; - case wxFONTENCODING_ISO8859_11 : - enc = kCFStringEncodingISOLatinThai; - break ; - case wxFONTENCODING_ISO8859_13 : - enc = kCFStringEncodingISOLatin7; - break ; - case wxFONTENCODING_ISO8859_14 : - enc = kCFStringEncodingISOLatin8; - break ; - case wxFONTENCODING_ISO8859_15 : - enc = kCFStringEncodingISOLatin9; - break ; - - case wxFONTENCODING_KOI8 : - enc = kCFStringEncodingKOI8_R; - break ; - case wxFONTENCODING_ALTERNATIVE : // MS-DOS CP866 - enc = kCFStringEncodingDOSRussian; - break ; - -// case wxFONTENCODING_BULGARIAN : -// enc = ; -// break ; - - case wxFONTENCODING_CP437 : - enc =kCFStringEncodingDOSLatinUS ; - break ; - case wxFONTENCODING_CP850 : - enc = kCFStringEncodingDOSLatin1; - break ; - case wxFONTENCODING_CP852 : - enc = kCFStringEncodingDOSLatin2; - break ; - case wxFONTENCODING_CP855 : - enc = kCFStringEncodingDOSCyrillic; - break ; - case wxFONTENCODING_CP866 : - enc =kCFStringEncodingDOSRussian ; - break ; - case wxFONTENCODING_CP874 : - enc = kCFStringEncodingDOSThai; - break ; - case wxFONTENCODING_CP932 : - enc = kCFStringEncodingDOSJapanese; - break ; - case wxFONTENCODING_CP936 : - enc =kCFStringEncodingDOSChineseSimplif ; - break ; - case wxFONTENCODING_CP949 : - enc = kCFStringEncodingDOSKorean; - break ; - case wxFONTENCODING_CP950 : - enc = kCFStringEncodingDOSChineseTrad; - break ; - - case wxFONTENCODING_CP1250 : - enc = kCFStringEncodingWindowsLatin2; - break ; - case wxFONTENCODING_CP1251 : - enc =kCFStringEncodingWindowsCyrillic ; - break ; - case wxFONTENCODING_CP1252 : - enc =kCFStringEncodingWindowsLatin1 ; - break ; - case wxFONTENCODING_CP1253 : - enc = kCFStringEncodingWindowsGreek; - break ; - case wxFONTENCODING_CP1254 : - enc = kCFStringEncodingWindowsLatin5; - break ; - case wxFONTENCODING_CP1255 : - enc =kCFStringEncodingWindowsHebrew ; - break ; - case wxFONTENCODING_CP1256 : - enc =kCFStringEncodingWindowsArabic ; - break ; - case wxFONTENCODING_CP1257 : - enc = kCFStringEncodingWindowsBalticRim; - break ; - case wxFONTENCODING_UTF7 : - enc = kCFStringEncodingNonLossyASCII ; - break ; - case wxFONTENCODING_UTF8 : - enc = kCFStringEncodingUTF8 ; - break ; - case wxFONTENCODING_EUC_JP : - enc = kCFStringEncodingEUC_JP; - break ; - case wxFONTENCODING_UTF16 : - enc = kCFStringEncodingUnicode ; - break ; - case wxFONTENCODING_MACROMAN : - enc = kCFStringEncodingMacRoman ; - break ; - case wxFONTENCODING_MACJAPANESE : - enc = kCFStringEncodingMacJapanese ; - break ; - case wxFONTENCODING_MACCHINESETRAD : - enc = kCFStringEncodingMacChineseTrad ; - break ; - case wxFONTENCODING_MACKOREAN : - enc = kCFStringEncodingMacKorean ; - break ; - case wxFONTENCODING_MACARABIC : - enc = kCFStringEncodingMacArabic ; - break ; - case wxFONTENCODING_MACHEBREW : - enc = kCFStringEncodingMacHebrew ; - break ; - case wxFONTENCODING_MACGREEK : - enc = kCFStringEncodingMacGreek ; - break ; - case wxFONTENCODING_MACCYRILLIC : - enc = kCFStringEncodingMacCyrillic ; - break ; - case wxFONTENCODING_MACDEVANAGARI : - enc = kCFStringEncodingMacDevanagari ; - break ; - case wxFONTENCODING_MACGURMUKHI : - enc = kCFStringEncodingMacGurmukhi ; - break ; - case wxFONTENCODING_MACGUJARATI : - enc = kCFStringEncodingMacGujarati ; - break ; - case wxFONTENCODING_MACORIYA : - enc = kCFStringEncodingMacOriya ; - break ; - case wxFONTENCODING_MACBENGALI : - enc = kCFStringEncodingMacBengali ; - break ; - case wxFONTENCODING_MACTAMIL : - enc = kCFStringEncodingMacTamil ; - break ; - case wxFONTENCODING_MACTELUGU : - enc = kCFStringEncodingMacTelugu ; - break ; - case wxFONTENCODING_MACKANNADA : - enc = kCFStringEncodingMacKannada ; - break ; - case wxFONTENCODING_MACMALAJALAM : - enc = kCFStringEncodingMacMalayalam ; - break ; - case wxFONTENCODING_MACSINHALESE : - enc = kCFStringEncodingMacSinhalese ; - break ; - case wxFONTENCODING_MACBURMESE : - enc = kCFStringEncodingMacBurmese ; - break ; - case wxFONTENCODING_MACKHMER : - enc = kCFStringEncodingMacKhmer ; - break ; - case wxFONTENCODING_MACTHAI : - enc = kCFStringEncodingMacThai ; - break ; - case wxFONTENCODING_MACLAOTIAN : - enc = kCFStringEncodingMacLaotian ; - break ; - case wxFONTENCODING_MACGEORGIAN : - enc = kCFStringEncodingMacGeorgian ; - break ; - case wxFONTENCODING_MACARMENIAN : - enc = kCFStringEncodingMacArmenian ; - break ; - case wxFONTENCODING_MACCHINESESIMP : - enc = kCFStringEncodingMacChineseSimp ; - break ; - case wxFONTENCODING_MACTIBETAN : - enc = kCFStringEncodingMacTibetan ; - break ; - case wxFONTENCODING_MACMONGOLIAN : - enc = kCFStringEncodingMacMongolian ; - break ; - case wxFONTENCODING_MACETHIOPIC : - enc = kCFStringEncodingMacEthiopic ; - break ; - case wxFONTENCODING_MACCENTRALEUR : - enc = kCFStringEncodingMacCentralEurRoman ; - break ; - case wxFONTENCODING_MACVIATNAMESE : - enc = kCFStringEncodingMacVietnamese ; - break ; - case wxFONTENCODING_MACARABICEXT : - enc = kCFStringEncodingMacExtArabic ; - break ; - case wxFONTENCODING_MACSYMBOL : - enc = kCFStringEncodingMacSymbol ; - break ; - case wxFONTENCODING_MACDINGBATS : - enc = kCFStringEncodingMacDingbats ; - break ; - case wxFONTENCODING_MACTURKISH : - enc = kCFStringEncodingMacTurkish ; - break ; - case wxFONTENCODING_MACCROATIAN : - enc = kCFStringEncodingMacCroatian ; - break ; - case wxFONTENCODING_MACICELANDIC : - enc = kCFStringEncodingMacIcelandic ; - break ; - case wxFONTENCODING_MACROMANIAN : - enc = kCFStringEncodingMacRomanian ; - break ; - case wxFONTENCODING_MACCELTIC : - enc = kCFStringEncodingMacCeltic ; - break ; - case wxFONTENCODING_MACGAELIC : - enc = kCFStringEncodingMacGaelic ; - break ; -// case wxFONTENCODING_MACKEYBOARD : -// enc = kCFStringEncodingMacKeyboardGlyphs ; -// break ; - default : - // because gcc is picky - break ; - } ; - return enc ; -} - -wxFontEncoding wxFontEncFromCFStringEnc(CFStringEncoding encoding) -{ - wxFontEncoding enc = wxFONTENCODING_DEFAULT ; - - switch( encoding) - { - case kCFStringEncodingISOLatin1 : - enc = wxFONTENCODING_ISO8859_1 ; - break ; - case kCFStringEncodingISOLatin2 : - enc = wxFONTENCODING_ISO8859_2; - break ; - case kCFStringEncodingISOLatin3 : - enc = wxFONTENCODING_ISO8859_3 ; - break ; - case kCFStringEncodingISOLatin4 : - enc = wxFONTENCODING_ISO8859_4; - break ; - case kCFStringEncodingISOLatinCyrillic : - enc = wxFONTENCODING_ISO8859_5; - break ; - case kCFStringEncodingISOLatinArabic : - enc = wxFONTENCODING_ISO8859_6; - break ; - case kCFStringEncodingISOLatinGreek : - enc = wxFONTENCODING_ISO8859_7; - break ; - case kCFStringEncodingISOLatinHebrew : - enc = wxFONTENCODING_ISO8859_8; - break ; - case kCFStringEncodingISOLatin5 : - enc = wxFONTENCODING_ISO8859_9; - break ; - case kCFStringEncodingISOLatin6 : - enc = wxFONTENCODING_ISO8859_10; - break ; - case kCFStringEncodingISOLatin7 : - enc = wxFONTENCODING_ISO8859_13; - break ; - case kCFStringEncodingISOLatin8 : - enc = wxFONTENCODING_ISO8859_14; - break ; - case kCFStringEncodingISOLatin9 : - enc =wxFONTENCODING_ISO8859_15 ; - break ; - - case kCFStringEncodingKOI8_R : - enc = wxFONTENCODING_KOI8; - break ; - -// case : -// enc = wxFONTENCODING_BULGARIAN; -// break ; - - case kCFStringEncodingDOSLatinUS : - enc = wxFONTENCODING_CP437; - break ; - case kCFStringEncodingDOSLatin1 : - enc = wxFONTENCODING_CP850; - break ; - case kCFStringEncodingDOSLatin2 : - enc =wxFONTENCODING_CP852 ; - break ; - case kCFStringEncodingDOSCyrillic : - enc = wxFONTENCODING_CP855; - break ; - case kCFStringEncodingDOSRussian : - enc = wxFONTENCODING_CP866; - break ; - case kCFStringEncodingDOSThai : - enc =wxFONTENCODING_CP874 ; - break ; - case kCFStringEncodingDOSJapanese : - enc = wxFONTENCODING_CP932; - break ; - case kCFStringEncodingDOSChineseSimplif : - enc = wxFONTENCODING_CP936; - break ; - case kCFStringEncodingDOSKorean : - enc = wxFONTENCODING_CP949; - break ; - case kCFStringEncodingDOSChineseTrad : - enc = wxFONTENCODING_CP950; - break ; - - case kCFStringEncodingWindowsLatin2 : - enc = wxFONTENCODING_CP1250; - break ; - case kCFStringEncodingWindowsCyrillic : - enc = wxFONTENCODING_CP1251; - break ; - case kCFStringEncodingWindowsLatin1 : - enc = wxFONTENCODING_CP1252; - break ; - case kCFStringEncodingWindowsGreek : - enc = wxFONTENCODING_CP1253; - break ; - case kCFStringEncodingWindowsLatin5 : - enc = wxFONTENCODING_CP1254; - break ; - case kCFStringEncodingWindowsHebrew : - enc = wxFONTENCODING_CP1255; - break ; - case kCFStringEncodingWindowsArabic : - enc = wxFONTENCODING_CP1256; - break ; - case kCFStringEncodingWindowsBalticRim : - enc =wxFONTENCODING_CP1257 ; - break ; - case kCFStringEncodingEUC_JP : - enc = wxFONTENCODING_EUC_JP; - break ; - case kCFStringEncodingUnicode : - enc = wxFONTENCODING_UTF16; - break; - case kCFStringEncodingMacRoman : - enc = wxFONTENCODING_MACROMAN ; - break ; - case kCFStringEncodingMacJapanese : - enc = wxFONTENCODING_MACJAPANESE ; - break ; - case kCFStringEncodingMacChineseTrad : - enc = wxFONTENCODING_MACCHINESETRAD ; - break ; - case kCFStringEncodingMacKorean : - enc = wxFONTENCODING_MACKOREAN ; - break ; - case kCFStringEncodingMacArabic : - enc =wxFONTENCODING_MACARABIC ; - break ; - case kCFStringEncodingMacHebrew : - enc = wxFONTENCODING_MACHEBREW ; - break ; - case kCFStringEncodingMacGreek : - enc = wxFONTENCODING_MACGREEK ; - break ; - case kCFStringEncodingMacCyrillic : - enc = wxFONTENCODING_MACCYRILLIC ; - break ; - case kCFStringEncodingMacDevanagari : - enc = wxFONTENCODING_MACDEVANAGARI ; - break ; - case kCFStringEncodingMacGurmukhi : - enc = wxFONTENCODING_MACGURMUKHI ; - break ; - case kCFStringEncodingMacGujarati : - enc = wxFONTENCODING_MACGUJARATI ; - break ; - case kCFStringEncodingMacOriya : - enc =wxFONTENCODING_MACORIYA ; - break ; - case kCFStringEncodingMacBengali : - enc =wxFONTENCODING_MACBENGALI ; - break ; - case kCFStringEncodingMacTamil : - enc = wxFONTENCODING_MACTAMIL ; - break ; - case kCFStringEncodingMacTelugu : - enc = wxFONTENCODING_MACTELUGU ; - break ; - case kCFStringEncodingMacKannada : - enc = wxFONTENCODING_MACKANNADA ; - break ; - case kCFStringEncodingMacMalayalam : - enc = wxFONTENCODING_MACMALAJALAM ; - break ; - case kCFStringEncodingMacSinhalese : - enc = wxFONTENCODING_MACSINHALESE ; - break ; - case kCFStringEncodingMacBurmese : - enc = wxFONTENCODING_MACBURMESE ; - break ; - case kCFStringEncodingMacKhmer : - enc = wxFONTENCODING_MACKHMER ; - break ; - case kCFStringEncodingMacThai : - enc = wxFONTENCODING_MACTHAI ; - break ; - case kCFStringEncodingMacLaotian : - enc = wxFONTENCODING_MACLAOTIAN ; - break ; - case kCFStringEncodingMacGeorgian : - enc = wxFONTENCODING_MACGEORGIAN ; - break ; - case kCFStringEncodingMacArmenian : - enc = wxFONTENCODING_MACARMENIAN ; - break ; - case kCFStringEncodingMacChineseSimp : - enc = wxFONTENCODING_MACCHINESESIMP ; - break ; - case kCFStringEncodingMacTibetan : - enc = wxFONTENCODING_MACTIBETAN ; - break ; - case kCFStringEncodingMacMongolian : - enc = wxFONTENCODING_MACMONGOLIAN ; - break ; - case kCFStringEncodingMacEthiopic : - enc = wxFONTENCODING_MACETHIOPIC ; - break ; - case kCFStringEncodingMacCentralEurRoman: - enc = wxFONTENCODING_MACCENTRALEUR ; - break ; - case kCFStringEncodingMacVietnamese: - enc = wxFONTENCODING_MACVIATNAMESE ; - break ; - case kCFStringEncodingMacExtArabic : - enc = wxFONTENCODING_MACARABICEXT ; - break ; - case kCFStringEncodingMacSymbol : - enc = wxFONTENCODING_MACSYMBOL ; - break ; - case kCFStringEncodingMacDingbats : - enc = wxFONTENCODING_MACDINGBATS ; - break ; - case kCFStringEncodingMacTurkish : - enc = wxFONTENCODING_MACTURKISH ; - break ; - case kCFStringEncodingMacCroatian : - enc = wxFONTENCODING_MACCROATIAN ; - break ; - case kCFStringEncodingMacIcelandic : - enc = wxFONTENCODING_MACICELANDIC ; - break ; - case kCFStringEncodingMacRomanian : - enc = wxFONTENCODING_MACROMANIAN ; - break ; - case kCFStringEncodingMacCeltic : - enc = wxFONTENCODING_MACCELTIC ; - break ; - case kCFStringEncodingMacGaelic : - enc = wxFONTENCODING_MACGAELIC ; - break ; -// case kCFStringEncodingMacKeyboardGlyphs : -// enc = wxFONTENCODING_MACKEYBOARD ; -// break ; - } ; - return enc ; -} - -class wxMBConv_cocoa : public wxMBConv -{ -public: - wxMBConv_cocoa() - { - Init(CFStringGetSystemEncoding()) ; - } - - wxMBConv_cocoa(const wxChar* name) - { - Init( wxCFStringEncFromFontEnc(wxFontMapper::Get()->CharsetToEncoding(name, false) ) ) ; - } - - wxMBConv_cocoa(wxFontEncoding encoding) - { - Init( wxCFStringEncFromFontEnc(encoding) ); - } - - ~wxMBConv_cocoa() - { - } - - void Init( CFStringEncoding encoding) - { - m_char_encoding = encoding ; - m_unicode_encoding = kCFStringEncodingUnicode; - } - - size_t MB2WC(wchar_t * szOut, const char * szUnConv, size_t nOutSize) const - { - wxASSERT(szUnConv); - - size_t nBufSize = strlen(szUnConv) + 1; - size_t nRealOutSize; - - UniChar* szUniCharBuffer = (UniChar*) szOut; - wchar_t* szConvBuffer = szOut; - - if (szConvBuffer == NULL && nOutSize != 0) - { - szConvBuffer = new wchar_t[nOutSize] ; - } - -#if SIZEOF_WCHAR_T == 4 - szUniCharBuffer = new UniChar[nOutSize]; -#endif - - CFDataRef theData = CFDataCreateWithBytesNoCopy ( - NULL, //allocator - (const UInt8*)szUnConv, - nBufSize - 1, - NULL //deallocator - ); - - wxASSERT(theData); - - CFStringRef theString = CFStringCreateFromExternalRepresentation ( - NULL, - theData, - m_char_encoding - ); - - wxASSERT(theString); - - if (nOutSize == 0) - { - nRealOutSize = CFStringGetLength(theString) + 1; - CFRelease(theString); - return nRealOutSize - 1; - } - - CFRange theRange = { 0, CFStringGetLength(theString) }; - - CFStringGetCharacters(theString, theRange, szUniCharBuffer); - - - nRealOutSize = (CFStringGetLength(theString) + 1); - - CFRelease(theString); - - szUniCharBuffer[nRealOutSize-1] = '\0' ; - -#if SIZEOF_WCHAR_T == 4 - wxMBConvUTF16 converter ; - converter.MB2WC(szConvBuffer , (const char*)szUniCharBuffer , nRealOutSize ) ; - delete[] szUniCharBuffer; -#endif - if ( szOut == NULL ) - delete [] szConvBuffer; - - return nRealOutSize ; - } - - size_t WC2MB(char *szOut, const wchar_t *szUnConv, size_t nOutSize) const - { - size_t nBufSize = wxWcslen(szUnConv) + 1; - size_t nRealOutSize; - char* szBuffer = szOut; - UniChar* szUniBuffer = (UniChar*) szUnConv; - - if (szOut == NULL) + // it's our lucky day + flags = WC_NO_BEST_FIT_CHARS; + pUsedDef = &usedDef; + } + else // old system or unsupported encoding { - // worst case - nRealOutSize = ((nBufSize - 1) * 8) +1 ; - szBuffer = new char[ nRealOutSize ] ; + flags = 0; + pUsedDef = NULL; } - else - nRealOutSize = nOutSize; - -#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)) ; - nBufSize /= sizeof(UniChar); - ++nBufSize; -#endif - - CFStringRef theString = CFStringCreateWithCharactersNoCopy( - NULL, //allocator - szUniBuffer, - nBufSize, - NULL //deallocator - ); - - wxASSERT(theString); - - //Note that CER puts a BOM when converting to unicode - //so we may want to check and use getchars instead in that case - CFDataRef theData = CFStringCreateExternalRepresentation( - NULL, //allocator - theString, - m_char_encoding, - 0 //what to put in characters that can't be converted - - //0 tells CFString to return NULL if it meets such a character - ); - - if(!theData) - return (size_t)-1; - - CFRelease(theString); - - nRealOutSize = CFDataGetLength(theData); - if ( szOut == NULL ) - delete[] szBuffer; + const size_t len = ::WideCharToMultiByte + ( + m_CodePage, // code page + flags, // either none or no best fit + pwz, // input string + -1, // it is (wide) NUL-terminated + buf, // output buffer + buf ? n : 0, // and its size + NULL, // default "replacement" char + pUsedDef // [out] was it used? + ); - if(nOutSize == 0) + if ( !len ) { -//TODO: This gets flagged as a non-malloced address by the debugger... -//#if SIZEOF_WCHAR_T == 4 -// delete[] szUniBuffer; -//#endif - CFRelease(theData); - return nRealOutSize - 1; + // function totally failed + return wxCONV_FAILED; } - CFRange theRange = {0, CFDataGetLength(theData) }; - CFDataGetBytes(theData, theRange, (UInt8*) szBuffer); + // we did something, check if we really succeeded + if ( flags ) + { + // check if the conversion failed, i.e. if any replacements + // were done + if ( usedDef ) + return wxCONV_FAILED; + } + else // we must resort to double tripping... + { + // first we need to ensure that we really have the MB data: this is + // not the case if we're called with NULL buffer, in which case we + // need to do the conversion yet again + wxCharBuffer bufDef; + if ( !buf ) + { + bufDef = wxCharBuffer(len); + buf = bufDef.data(); + if ( !::WideCharToMultiByte(m_CodePage, flags, pwz, -1, + buf, len, NULL, NULL) ) + return wxCONV_FAILED; + } - CFRelease(theData); + if ( !n ) + n = wcslen(pwz); + wxWCharBuffer wcBuf(n); + if ( MB2WC(wcBuf.data(), buf, n + 1) == 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 wxCONV_FAILED; + } + } -//TODO: This gets flagged as a non-malloced address by the debugger... -//#if SIZEOF_WCHAR_T == 4 -// delete[] szUniBuffer; -//#endif - return nRealOutSize - 1; + // see the comment above for the reason of "len - 1" + return len - 1; } - bool IsOk() const + virtual size_t GetMBNulLen() const { - //TODO: check for invalid en/de/coding - return true; - } - -private: - CFStringEncoding m_char_encoding ; - CFStringEncoding m_unicode_encoding ; -}; + 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 + ); -#endif // defined(__WXCOCOA__) + wxMBConv_win32 * const self = wxConstCast(this, wxMBConv_win32); + switch ( len ) + { + default: + wxLogDebug(wxT("Unexpected NUL length %d"), len); + self->m_minMBCharWidth = (size_t)-1; + break; -// ============================================================================ -// Mac conversion classes -// ============================================================================ + case 0: + self->m_minMBCharWidth = (size_t)-1; + break; -#if defined(__WXMAC__) && defined(TARGET_CARBON) + case 1: + case 2: + case 4: + self->m_minMBCharWidth = len; + break; + } + } -class wxMBConv_mac : public wxMBConv -{ -public: - wxMBConv_mac() - { - Init(CFStringGetSystemEncoding()) ; + return m_minMBCharWidth; } - wxMBConv_mac(const wxChar* name) - { - Init( wxMacGetSystemEncFromFontEnc(wxFontMapper::Get()->CharsetToEncoding(name, false) ) ) ; - } + virtual wxMBConv *Clone() const { return new wxMBConv_win32(*this); } - wxMBConv_mac(wxFontEncoding encoding) - { - Init( wxMacGetSystemEncFromFontEnc(encoding) ); - } + bool IsOk() const { return m_CodePage != -1; } - ~wxMBConv_mac() +private: + static bool CanUseNoBestFit() { - OSStatus status = noErr ; - status = TECDisposeConverter(m_MB2WC_converter); - status = TECDisposeConverter(m_WC2MB_converter); - } - + static int s_isWin98Or2k = -1; - void Init( TextEncodingBase encoding) - { - OSStatus status = noErr ; - m_char_encoding = encoding ; - m_unicode_encoding = CreateTextEncoding(kTextEncodingUnicodeDefault,0,kUnicode16BitFormat) ; + if ( s_isWin98Or2k == -1 ) + { + int verMaj, verMin; + switch ( wxGetOsVersion(&verMaj, &verMin) ) + { + case wxOS_WINDOWS_9X: + s_isWin98Or2k = verMaj >= 4 && verMin >= 10; + break; - status = TECCreateConverter(&m_MB2WC_converter, - m_char_encoding, - m_unicode_encoding); - status = TECCreateConverter(&m_WC2MB_converter, - m_unicode_encoding, - m_char_encoding); - } + case wxOS_WINDOWS_NT: + s_isWin98Or2k = verMaj >= 5; + break; - size_t MB2WC(wchar_t *buf, const char *psz, size_t n) const - { - OSStatus status = noErr ; - ByteCount byteOutLen ; - ByteCount byteInLen = strlen(psz) ; - wchar_t *tbuf = NULL ; - UniChar* ubuf = NULL ; - size_t res = 0 ; + default: + // unknown: be conservative by default + s_isWin98Or2k = 0; + break; + } - if (buf == NULL) - { - n = byteInLen ; - tbuf = (wchar_t*) malloc( n * SIZEOF_WCHAR_T) ; + wxASSERT_MSG( s_isWin98Or2k != -1, wxT("should be set above") ); } - 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); -#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 ) ; - free( ubuf ) ; -#else - res = byteOutLen / sizeof( UniChar ) ; -#endif - if ( buf == NULL ) - free(tbuf) ; - - if ( buf && res < n) - buf[res] = 0; - return res ; + return s_isWin98Or2k == 1; } - //NB: This is _broken_ - in invalid conversions, instead of returning -1 - //like it should, it (sometimes?) converts invalid characters of the encoding to a question - //mark character '?'. - // - //We need to do the msw double-pass check for the question marks as Vadim - //lines out above (RN: I don't recall this happening in the core foundation version, - //but it might do it there also, ick) - size_t WC2MB(char *buf, const wchar_t *psz, size_t n) const + static bool IsAtLeastWin2kSP4() { - OSStatus status = noErr ; - ByteCount byteOutLen ; - ByteCount byteInLen = wxWcslen(psz) * SIZEOF_WCHAR_T ; - - char *tbuf = NULL ; +#ifdef __WXWINCE__ + return false; +#else + static int s_isAtLeastWin2kSP4 = -1; - if (buf == NULL) + if ( s_isAtLeastWin2kSP4 == -1 ) { - // worst case - n = ((byteInLen / SIZEOF_WCHAR_T) * 8) + SIZEOF_WCHAR_T; - tbuf = (char*) malloc( n ) ; + 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; } - ByteCount byteBufferLen = n ; - UniChar* ubuf = NULL ; -#if SIZEOF_WCHAR_T == 4 - wxMBConvUTF16BE 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 - status = TECConvertText(m_WC2MB_converter, (ConstTextPtr) ubuf , byteInLen, &byteInLen, - (TextPtr) (buf ? buf : tbuf) , byteBufferLen, &byteOutLen); -#if SIZEOF_WCHAR_T == 4 - free( ubuf ) ; + return s_isAtLeastWin2kSP4 == 1; #endif - if ( buf == NULL ) - free(tbuf) ; - - size_t res = byteOutLen ; - if ( buf && res < n) - buf[res] = 0; - - return res ; } - bool IsOk() const - { return m_MB2WC_converter != NULL && m_WC2MB_converter != NULL ; } -private: - TECObjectRef m_MB2WC_converter ; - TECObjectRef m_WC2MB_converter ; + // the code page we're working with + long m_CodePage; - TextEncodingBase m_char_encoding ; - TextEncodingBase m_unicode_encoding ; + // cached result of GetMBNulLen(), set to 0 initially meaning + // "unknown" + size_t m_minMBCharWidth; }; -#endif // defined(__WXMAC__) && defined(TARGET_CARBON) +#endif // wxHAVE_WIN32_MB2WC + // ============================================================================ // wxEncodingConverter based conversion classes @@ -2406,17 +2840,20 @@ class wxMBConv_wxwin : public wxMBConv private: void Init() { - m_ok = m2w.Init(m_enc, wxFONTENCODING_UNICODE) && + // Refuse to use broken wxEncodingConverter code for Mac-specific encodings. + // The wxMBConv_cf class does a better job. + m_ok = (m_enc < wxFONTENCODING_MACMIN || m_enc > wxFONTENCODING_MACMAX) && + m2w.Init(m_enc, wxFONTENCODING_UNICODE) && w2m.Init(wxFONTENCODING_UNICODE, m_enc); } public: // temporarily just use wxEncodingConverter stuff, // so that it works while a better implementation is built - wxMBConv_wxwin(const wxChar* name) + wxMBConv_wxwin(const char* name) { if (name) - m_enc = wxFontMapper::Get()->CharsetToEncoding(name, false); + m_enc = wxFontMapperBase::Get()->CharsetToEncoding(name, false); else m_enc = wxFONTENCODING_SYSTEM; @@ -2434,7 +2871,10 @@ public: { size_t inbuf = strlen(psz); if (buf) - m2w.Convert(psz,buf); + { + if (!m2w.Convert(psz, buf)) + return wxCONV_FAILED; + } return inbuf; } @@ -2442,23 +2882,59 @@ 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) + wxDECLARE_NO_COPY_CLASS(wxMBConv_wxwin); }; +// make the constructors available for unit testing +WXDLLIMPEXP_BASE wxMBConv* new_wxMBConv_wxwin( const char* name ) +{ + wxMBConv_wxwin* result = new wxMBConv_wxwin( name ); + if ( !result->IsOk() ) + { + delete result; + return 0; + } + + return result; +} + #endif // wxUSE_FONTMAP // ============================================================================ @@ -2469,33 +2945,75 @@ void wxCSConv::Init() { m_name = NULL; m_convReal = NULL; - m_deferred = true; } -wxCSConv::wxCSConv(const wxChar *charset) +void wxCSConv::SetEncoding(wxFontEncoding encoding) +{ + switch ( encoding ) + { + case wxFONTENCODING_MAX: + case wxFONTENCODING_SYSTEM: + if ( m_name ) + { + // It's ok to not have encoding value if we have a name for it. + m_encoding = wxFONTENCODING_SYSTEM; + } + else // No name neither. + { + // Fall back to the system default encoding in this case (not + // sure how much sense does this make but this is how the old + // code used to behave). +#if wxUSE_INTL + m_encoding = wxLocale::GetSystemEncoding(); + if ( m_encoding == wxFONTENCODING_SYSTEM ) +#endif // wxUSE_INTL + m_encoding = wxFONTENCODING_ISO8859_1; + } + break; + + case wxFONTENCODING_DEFAULT: + // wxFONTENCODING_DEFAULT is same as US-ASCII in this context + m_encoding = wxFONTENCODING_ISO8859_1; + break; + + default: + // Just use the provided encoding. + m_encoding = encoding; + } +} + +wxCSConv::wxCSConv(const wxString& charset) { Init(); - if ( charset ) + if ( !charset.empty() ) { - SetName(charset); + SetName(charset.ToAscii()); } - m_encoding = wxFONTENCODING_SYSTEM; +#if wxUSE_FONTMAP + SetEncoding(wxFontMapperBase::GetEncodingFromName(charset)); +#else + SetEncoding(wxFONTENCODING_SYSTEM); +#endif + + m_convReal = DoCreate(); } wxCSConv::wxCSConv(wxFontEncoding encoding) { if ( encoding == wxFONTENCODING_MAX || encoding == wxFONTENCODING_DEFAULT ) { - wxFAIL_MSG( _T("invalid encoding value in wxCSConv ctor") ); + wxFAIL_MSG( wxT("invalid encoding value in wxCSConv ctor") ); encoding = wxFONTENCODING_SYSTEM; } Init(); - m_encoding = encoding; + SetEncoding(encoding); + + m_convReal = DoCreate(); } wxCSConv::~wxCSConv() @@ -2509,7 +3027,9 @@ wxCSConv::wxCSConv(const wxCSConv& conv) Init(); SetName(conv.m_name); - m_encoding = conv.m_encoding; + SetEncoding(conv.m_encoding); + + m_convReal = DoCreate(); } wxCSConv& wxCSConv::operator=(const wxCSConv& conv) @@ -2517,7 +3037,9 @@ wxCSConv& wxCSConv::operator=(const wxCSConv& conv) Clear(); SetName(conv.m_name); - m_encoding = conv.m_encoding; + SetEncoding(conv.m_encoding); + + m_convReal = DoCreate(); return *this; } @@ -2525,23 +3047,34 @@ wxCSConv& wxCSConv::operator=(const wxCSConv& conv) void wxCSConv::Clear() { free(m_name); - delete m_convReal; - m_name = NULL; - m_convReal = NULL; + + wxDELETE(m_convReal); } -void wxCSConv::SetName(const wxChar *charset) +void wxCSConv::SetName(const char *charset) { - if (charset) - { + if ( charset ) m_name = wxStrdup(charset); - m_deferred = true; - } } +#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 + : (const char*)wxFontMapperBase::GetEncodingName(m_encoding).mb_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 @@ -2565,18 +3098,65 @@ wxMBConv *wxCSConv::DoCreate() const if ( m_name ) #endif // !wxUSE_FONTMAP { - wxString name(m_name); +#if wxUSE_FONTMAP + wxFontEncoding encoding(m_encoding); +#endif + + if ( m_name ) + { + wxMBConv_iconv *conv = new wxMBConv_iconv(m_name); + if ( conv->IsOk() ) + return conv; + + delete conv; #if wxUSE_FONTMAP - if ( name.empty() ) - name = wxFontMapper::Get()->GetEncodingName(m_encoding); + encoding = + wxFontMapperBase::Get()->CharsetToEncoding(m_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.ToAscii()); + if ( conv->IsOk() ) + return conv; - delete conv; + delete conv; + } + + const wxChar* const* 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 ) + { + // FIXME-UTF8: wxFontMapperBase::GetAllEncodingNames() + // will need changes that will obsolete this + wxString name(*names); + wxMBConv_iconv *conv = new wxMBConv_iconv(name.ToAscii()); + if ( conv->IsOk() ) + { + gs_nameCache[encoding] = *names; + return conv; + } + + delete conv; + } + + gs_nameCache[encoding] = wxT(""); // cache the failure + } + } +#endif // wxUSE_FONTMAP } #endif // HAVE_ICONV @@ -2594,34 +3174,28 @@ wxMBConv *wxCSConv::DoCreate() const #endif } #endif // wxHAVE_WIN32_MB2WC -#if defined(__WXMAC__) - { - if ( m_name || ( m_encoding < wxFONTENCODING_UTF16BE ) ) - { - - wxMBConv_mac *conv = m_name ? new wxMBConv_mac(m_name) - : new wxMBConv_mac(m_encoding); - if ( conv->IsOk() ) - return conv; - delete conv; - } - } -#endif -#if defined(__WXCOCOA__) +#ifdef __DARWIN__ { - if ( m_name || ( m_encoding <= wxFONTENCODING_UTF16 ) ) + // 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_cf *conv = m_name ? new wxMBConv_cf(m_name) + : new wxMBConv_cf(m_encoding); +#else + wxMBConv_cf *conv = new wxMBConv_cf(m_encoding); +#endif - wxMBConv_cocoa *conv = m_name ? new wxMBConv_cocoa(m_name) - : new wxMBConv_cocoa(m_encoding); if ( conv->IsOk() ) return conv; delete conv; } } -#endif +#endif // __DARWIN__ + // step (2) wxFontEncoding enc = m_encoding; #if wxUSE_FONTMAP @@ -2630,7 +3204,7 @@ 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 @@ -2656,7 +3230,7 @@ wxMBConv *wxCSConv::DoCreate() const default: // nothing to do but put here to suppress gcc warnings - ; + break; } // step (3) @@ -2669,136 +3243,206 @@ wxMBConv *wxCSConv::DoCreate() const delete conv; } -#endif // wxUSE_FONTMAP - // 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. - static bool alreadyLoggingError = false; - if (!alreadyLoggingError) - { - alreadyLoggingError = true; - wxLogError(_("Cannot convert from the charset '%s'!"), - m_name ? m_name - : -#if wxUSE_FONTMAP - wxFontMapper::GetEncodingDescription(m_encoding).c_str() -#else // !wxUSE_FONTMAP - wxString::Format(_("encoding %s"), m_encoding).c_str() -#endif // wxUSE_FONTMAP/!wxUSE_FONTMAP - ); - alreadyLoggingError = false; - } + wxLogTrace(TRACE_STRCONV, + wxT("encoding \"%s\" is not supported by this system"), + (m_name ? wxString(m_name) + : wxFontMapperBase::GetEncodingName(m_encoding))); +#endif // wxUSE_FONTMAP return NULL; } -void wxCSConv::CreateConvIfNeeded() const +bool wxCSConv::IsOk() const { - if ( m_deferred ) - { - wxCSConv *self = (wxCSConv *)this; // const_cast - -#if wxUSE_INTL - // if we don't have neither the name nor the encoding, use the default - // encoding for this system - if ( !m_name && m_encoding == wxFONTENCODING_SYSTEM ) - { - self->m_name = wxStrdup(wxLocale::GetSystemEncodingName()); - } -#endif // wxUSE_INTL + // special case: no convReal created for wxFONTENCODING_ISO8859_1 + if ( m_encoding == wxFONTENCODING_ISO8859_1 ) + return true; // always ok as we do it ourselves - self->m_convReal = DoCreate(); - self->m_deferred = false; - } + // m_convReal->IsOk() is called at its own creation, so we know it must + // be ok if m_convReal is non-NULL + return m_convReal != NULL; } -size_t wxCSConv::MB2WC(wchar_t *buf, const char *psz, size_t n) const +size_t wxCSConv::ToWChar(wchar_t *dst, size_t dstLen, + const char *src, size_t srcLen) const { - CreateConvIfNeeded(); - if (m_convReal) - return m_convReal->MB2WC(buf, psz, n); + return m_convReal->ToWChar(dst, dstLen, src, srcLen); // latin-1 (direct) - size_t len = strlen(psz); + if ( srcLen == wxNO_LEN ) + srcLen = strlen(src) + 1; // take trailing NUL too - if (buf) + if ( dst ) { - for (size_t c = 0; c <= len; c++) - buf[c] = (unsigned char)(psz[c]); + if ( dstLen < srcLen ) + return wxCONV_FAILED; + + for ( size_t n = 0; n < srcLen; n++ ) + dst[n] = (unsigned char)(src[n]); } - return len; + return srcLen; } -size_t wxCSConv::WC2MB(char *buf, const wchar_t *psz, size_t n) const +size_t wxCSConv::FromWChar(char *dst, size_t dstLen, + const wchar_t *src, size_t srcLen) const { - CreateConvIfNeeded(); - if (m_convReal) - return m_convReal->WC2MB(buf, psz, n); + return m_convReal->FromWChar(dst, dstLen, src, srcLen); // latin-1 (direct) - const size_t len = wxWcslen(psz); - if (buf) + if ( srcLen == wxNO_LEN ) + srcLen = wxWcslen(src) + 1; + + if ( dst ) { - for (size_t c = 0; c <= len; c++) + if ( dstLen < srcLen ) + return wxCONV_FAILED; + + for ( size_t n = 0; n < srcLen; n++ ) { - if (psz[c] > 0xFF) - return (size_t)-1; - buf[c] = (char)psz[c]; + if ( src[n] > 0xFF ) + return wxCONV_FAILED; + + dst[n] = (char)src[n]; } + } - else + else // still need to check the input validity { - for (size_t c = 0; c <= len; c++) + for ( size_t n = 0; n < srcLen; n++ ) { - if (psz[c] > 0xFF) - return (size_t)-1; + if ( src[n] > 0xFF ) + return wxCONV_FAILED; } } - return len; + return srcLen; } -// ---------------------------------------------------------------------------- -// globals -// ---------------------------------------------------------------------------- +size_t wxCSConv::GetMBNulLen() const +{ + if ( m_convReal ) + return m_convReal->GetMBNulLen(); -#ifdef __WINDOWS__ - static wxMBConv_win32 wxConvLibcObj; -#elif defined(__WXMAC__) && !defined(__MACH__) - static wxMBConv_mac wxConvLibcObj ; -#else - static wxMBConvLibc wxConvLibcObj; + // otherwise, we are ISO-8859-1 + return 1; +} + +#if wxUSE_UNICODE_UTF8 +bool wxCSConv::IsUTF8() const +{ + if ( m_convReal ) + return m_convReal->IsUTF8(); + + // otherwise, we are ISO-8859-1 + return false; +} #endif -static wxCSConv wxConvLocalObj(wxFONTENCODING_SYSTEM); -static wxCSConv wxConvISO8859_1Obj(wxFONTENCODING_ISO8859_1); -static wxMBConvUTF7 wxConvUTF7Obj; -static wxMBConvUTF8 wxConvUTF8Obj; +#if wxUSE_UNICODE + +wxWCharBuffer wxSafeConvertMB2WX(const char *s) +{ + if ( !s ) + return wxWCharBuffer(); + + wxWCharBuffer wbuf(wxConvLibc.cMB2WX(s)); + if ( !wbuf ) + wbuf = wxMBConvUTF8().cMB2WX(s); + if ( !wbuf ) + wbuf = wxConvISO8859_1.cMB2WX(s); + + return wbuf; +} + +wxCharBuffer wxSafeConvertWX2MB(const wchar_t *ws) +{ + if ( !ws ) + return wxCharBuffer(); + + wxCharBuffer buf(wxConvLibc.cWX2MB(ws)); + if ( !buf ) + buf = wxMBConvUTF8(wxMBConvUTF8::MAP_INVALID_UTF8_TO_OCTAL).cWX2MB(ws); + + return buf; +} -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; +#endif // wxUSE_UNICODE -#else // !wxUSE_WCHAR_T +// ---------------------------------------------------------------------------- +// globals +// ---------------------------------------------------------------------------- -// stand-ins in absence of wchar_t -WXDLLIMPEXP_DATA_BASE(wxMBConv) wxConvLibc, - wxConvISO8859_1, - wxConvLocal, - wxConvUTF8; +// NB: The reason why we create converted objects in this convoluted way, +// using a factory function instead of global variable, is that they +// may be used at static initialization time (some of them are used by +// wxString ctors and there may be a global wxString object). In other +// words, possibly _before_ the converter global object would be +// initialized. + +#undef wxConvLibc +#undef wxConvUTF8 +#undef wxConvUTF7 +#undef wxConvLocal +#undef wxConvISO8859_1 + +#define WX_DEFINE_GLOBAL_CONV2(klass, impl_klass, name, ctor_args) \ + WXDLLIMPEXP_DATA_BASE(klass*) name##Ptr = NULL; \ + WXDLLIMPEXP_BASE klass* wxGet_##name##Ptr() \ + { \ + static impl_klass name##Obj ctor_args; \ + return &name##Obj; \ + } \ + /* this ensures that all global converter objects are created */ \ + /* by the time static initialization is done, i.e. before any */ \ + /* thread is launched: */ \ + static klass* gs_##name##instance = wxGet_##name##Ptr() + +#define WX_DEFINE_GLOBAL_CONV(klass, name, ctor_args) \ + WX_DEFINE_GLOBAL_CONV2(klass, klass, name, ctor_args) + +#ifdef __INTELC__ + // disable warning "variable 'xxx' was declared but never referenced" + #pragma warning(disable: 177) +#endif // Intel C++ -#endif // wxUSE_WCHAR_T/!wxUSE_WCHAR_T +#ifdef __WINDOWS__ + WX_DEFINE_GLOBAL_CONV2(wxMBConv, wxMBConv_win32, wxConvLibc, wxEMPTY_PARAMETER_VALUE); +#elif 0 // defined(__WXOSX__) + WX_DEFINE_GLOBAL_CONV2(wxMBConv, wxMBConv_cf, wxConvLibc, (wxFONTENCODING_UTF8)); +#else + WX_DEFINE_GLOBAL_CONV2(wxMBConv, wxMBConvLibc, wxConvLibc, wxEMPTY_PARAMETER_VALUE); +#endif +// NB: we can't use wxEMPTY_PARAMETER_VALUE as final argument here because it's +// passed to WX_DEFINE_GLOBAL_CONV2 after a macro expansion and so still +// provokes an error message about "not enough macro parameters"; and we +// can't use "()" here as the name##Obj declaration would be parsed as a +// function declaration then, so use a semicolon and live with an extra +// empty statement (and hope that no compilers warns about this) +WX_DEFINE_GLOBAL_CONV(wxMBConvStrictUTF8, wxConvUTF8, ;); +WX_DEFINE_GLOBAL_CONV(wxMBConvUTF7, wxConvUTF7, ;); + +WX_DEFINE_GLOBAL_CONV(wxCSConv, wxConvLocal, (wxFONTENCODING_SYSTEM)); +WX_DEFINE_GLOBAL_CONV(wxCSConv, wxConvISO8859_1, (wxFONTENCODING_ISO8859_1)); + +WXDLLIMPEXP_DATA_BASE(wxMBConv *) wxConvCurrent = wxGet_wxConvLibcPtr(); +WXDLLIMPEXP_DATA_BASE(wxMBConv *) wxConvUI = wxGet_wxConvLocalPtr(); + +#ifdef __DARWIN__ +// It is important to use this conversion object under Darwin as it ensures +// that Unicode strings are (re)composed correctly even though xnu kernel uses +// decomposed form internally (at least for the file names). +static wxMBConv_cf wxConvMacUTF8DObj(wxFONTENCODING_UTF8); +#endif +WXDLLIMPEXP_DATA_BASE(wxMBConv *) wxConvFileName = +#ifdef __DARWIN__ + &wxConvMacUTF8DObj; +#else // !__DARWIN__ + wxGet_wxConvLibcPtr(); +#endif // __DARWIN__/!__DARWIN__