X-Git-Url: https://git.saurik.com/wxWidgets.git/blobdiff_plain/55d99c7a77789ff4904bf96eddca3715eb5af9b9..e5ef1cae0f424be0bffbd651cb10052dfb208396:/src/common/strconv.cpp diff --git a/src/common/strconv.cpp b/src/common/strconv.cpp index bbd8ce8369..507bf1307b 100644 --- a/src/common/strconv.cpp +++ b/src/common/strconv.cpp @@ -1,183 +1,124 @@ ///////////////////////////////////////////////////////////////////////////// -// Name: strconv.cpp +// Name: src/common/strconv.cpp // Purpose: Unicode conversion classes -// Author: Ove Kaaven, Robert Roebling, Vadim Zeitlin, Vaclav Slavik +// Author: Ove Kaaven, Robert Roebling, Vadim Zeitlin, Vaclav Slavik, +// Ryan Norton, Fredrik Roubert (UTF7) // Modified by: // Created: 29/01/98 // RCS-ID: $Id$ -// Copyright: (c) 1999 Ove Kaaven, Robert Roebling, Vadim Zeitlin, Vaclav Slavik +// Copyright: (c) 1999 Ove Kaaven, Robert Roebling, Vaclav Slavik +// (c) 2000-2003 Vadim Zeitlin +// (c) 2004 Ryan Norton, Fredrik Roubert // Licence: wxWindows licence ///////////////////////////////////////////////////////////////////////////// -// ============================================================================ -// declarations -// ============================================================================ - -// ---------------------------------------------------------------------------- -// headers -// ---------------------------------------------------------------------------- - -#ifdef __GNUG__ - #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 - -#ifdef __WXMSW__ - #include "wx/msw/private.h" + #include "wx/utils.h" + #include "wx/hashmap.h" #endif -#include -#include -#include -#include - -#include "wx/module.h" #include "wx/strconv.h" -// ---------------------------------------------------------------------------- -// globals -// ---------------------------------------------------------------------------- - #if wxUSE_WCHAR_T - WXDLLEXPORT_DATA(wxMBConv) wxConvLibc; - WXDLLEXPORT_DATA(wxCSConv) wxConvLocal((const wxChar *)NULL); - WXDLLEXPORT_DATA(wxCSConv) wxConvISO8859_1(_T("iso-8859-1")); -#else - // stand-ins in absence of wchar_t - WXDLLEXPORT_DATA(wxMBConv) wxConvLibc, - wxConvFile, - wxConvISO8859_1, - wxConvLocal, - wxConvUTF8; -#endif // wxUSE_WCHAR_T - -WXDLLEXPORT_DATA(wxMBConv *) wxConvCurrent = &wxConvLibc; -class wxStrConvModule: public wxModule -{ -public: - wxStrConvModule() : wxModule() { } - virtual bool OnInit() { return TRUE; } - virtual void OnExit() - { -#if wxUSE_WCHAR_T - wxConvLocal.Clear(); - wxConvISO8859_1.Clear(); +#ifndef __WXWINCE__ +#include #endif - } - - DECLARE_DYNAMIC_CLASS(wxStrConvModule) -}; - -IMPLEMENT_DYNAMIC_CLASS(wxStrConvModule, wxModule) - -// ---------------------------------------------------------------------------- -// headers -// ---------------------------------------------------------------------------- - -#if wxUSE_WCHAR_T +#include +#include +#include -#ifdef __SALFORDC__ - #include +#if defined(__WIN32__) && !defined(__WXMICROWIN__) + #include "wx/msw/private.h" + #include "wx/msw/missing.h" + #define wxHAVE_WIN32_MB2WC #endif #ifdef HAVE_ICONV #include + #include "wx/thread.h" #endif #include "wx/encconv.h" #include "wx/fontmap.h" -// ---------------------------------------------------------------------------- -// macros -// ---------------------------------------------------------------------------- +#ifdef __DARWIN__ +#include "wx/osx/core/private/strconv_cf.h" +#endif //def __DARWIN__ -#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++ = (wchar_t) ((input >> 10)+0xd7c0); - *output++ = (wchar_t) ((input&0x3ff)+0xdc00); + *output++ = (wxUint16) ((input >> 10) + 0xd7c0); + *output = (wxUint16) ((input & 0x3ff) + 0xdc00); } + return 2; } } -static size_t decode_utf16(const wchar_t* input, wxUint32& output) +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 { @@ -186,53 +127,222 @@ static size_t decode_utf16(const wchar_t* input, wxUint32& output) } } -#endif // WC_UTF16 +#ifdef WC_UTF16 + typedef wchar_t wxDecodeSurrogate_t; +#else // !WC_UTF16 + typedef wxUint16 wxDecodeSurrogate_t; +#endif // WC_UTF16/!WC_UTF16 + +// returns the next UTF-32 character from the wchar_t buffer and advances the +// pointer to the character after this one +// +// if an invalid character is found, *pSrc is set to NULL, the caller must +// check for this +static wxUint32 wxDecodeSurrogate(const wxDecodeSurrogate_t **pSrc) +{ + wxUint32 out; + const size_t + n = decode_utf16(wx_reinterpret_cast(const wxUint16 *, *pSrc), out); + if ( n == wxCONV_FAILED ) + *pSrc = NULL; + else + *pSrc += n; + + return out; +} // ---------------------------------------------------------------------------- // wxMBConv // ---------------------------------------------------------------------------- -#define IGNORE_LIBC 0 - -wxMBConv::~wxMBConv() +size_t +wxMBConv::ToWChar(wchar_t *dst, size_t dstLen, + const char *src, size_t srcLen) const { - // nothing to do here + // although new conversion classes are supposed to implement this function + // directly, the existins ones only implement the old MB2WC() and so, to + // avoid to have to rewrite all conversion classes at once, we provide a + // default (but not efficient) implementation of this one in terms of the + // old function by copying the input to ensure that it's NUL-terminated and + // then using MB2WC() to convert it + + // the number of chars [which would be] written to dst [if it were not NULL] + size_t dstWritten = 0; + + // the number of NULs terminating this string + size_t nulLen = 0; // not really needed, but just to avoid warnings + + // if we were not given the input size we just have to assume that the + // string is properly terminated as we have no way of knowing how long it + // is anyhow, but if we do have the size check whether there are enough + // NULs at the end + wxCharBuffer bufTmp; + const char *srcEnd; + if ( srcLen != wxNO_LEN ) + { + // we need to know how to find the end of this string + nulLen = GetMBNulLen(); + if ( nulLen == wxCONV_FAILED ) + return wxCONV_FAILED; + + // if there are enough NULs we can avoid the copy + if ( srcLen < nulLen || NotAllNULs(src + srcLen - nulLen, nulLen) ) + { + // make a copy in order to properly NUL-terminate the string + bufTmp = wxCharBuffer(srcLen + nulLen - 1 /* 1 will be added */); + char * const p = bufTmp.data(); + memcpy(p, src, srcLen); + for ( char *s = p + srcLen; s < p + srcLen + nulLen; s++ ) + *s = '\0'; + + src = bufTmp; + } + + srcEnd = src + srcLen; + } + else // quit after the first loop iteration + { + srcEnd = NULL; + } + + for ( ;; ) + { + // try to convert the current chunk + size_t lenChunk = MB2WC(NULL, src, 0); + if ( lenChunk == wxCONV_FAILED ) + return wxCONV_FAILED; + + lenChunk++; // for the L'\0' at the end of this chunk + + dstWritten += lenChunk; + + if ( lenChunk == 1 ) + { + // nothing left in the input string, conversion succeeded + break; + } + + if ( dst ) + { + if ( dstWritten > dstLen ) + return wxCONV_FAILED; + + if ( MB2WC(dst, src, lenChunk) == wxCONV_FAILED ) + return wxCONV_FAILED; + + dst += lenChunk; + } + + if ( !srcEnd ) + { + // we convert just one chunk in this case as this is the entire + // string anyhow + break; + } + + // advance the input pointer past the end of this chunk + while ( NotAllNULs(src, nulLen) ) + { + // notice that we must skip over multiple bytes here as we suppose + // that if NUL takes 2 or 4 bytes, then all the other characters do + // too and so if advanced by a single byte we might erroneously + // detect sequences of NUL bytes in the middle of the input + src += nulLen; + } + + src += nulLen; // skipping over its terminator as well + + // note that ">=" (and not just "==") is needed here as the terminator + // we skipped just above could be inside or just after the buffer + // delimited by inEnd + if ( src >= srcEnd ) + break; + } + + return dstWritten; } -size_t wxMBConv::MB2WC(wchar_t *buf, const char *psz, size_t n) const +size_t +wxMBConv::FromWChar(char *dst, size_t dstLen, + const wchar_t *src, size_t srcLen) const { -#if IGNORE_LIBC - if (buf) + // the number of chars [which would be] written to dst [if it were not NULL] + size_t dstWritten = 0; + + // make a copy of the input string unless it is already properly + // NUL-terminated + // + // if we don't know its length we have no choice but to assume that it is, + // indeed, properly terminated + wxWCharBuffer bufTmp; + if ( srcLen == wxNO_LEN ) { - for (size_t i = 0; i < strlen( psz )+1; i++) - buf[i] = (wchar_t) psz[i]; - return strlen( psz ); + srcLen = wxWcslen(src) + 1; } - else + else if ( srcLen != 0 && src[srcLen - 1] != L'\0' ) { - return strlen( psz ); + // make a copy in order to properly NUL-terminate the string + bufTmp = wxWCharBuffer(srcLen); + memcpy(bufTmp.data(), src, srcLen * sizeof(wchar_t)); + src = bufTmp; } -#else - return wxMB2WC(buf, psz, n); -#endif + + const size_t lenNul = GetMBNulLen(); + for ( const wchar_t * const srcEnd = src + srcLen; + src < srcEnd; + src += wxWcslen(src) + 1 /* skip L'\0' too */ ) + { + // try to convert the current chunk + size_t lenChunk = WC2MB(NULL, src, 0); + + if ( lenChunk == wxCONV_FAILED ) + return wxCONV_FAILED; + + lenChunk += lenNul; + dstWritten += lenChunk; + + if ( dst ) + { + if ( dstWritten > dstLen ) + return wxCONV_FAILED; + + if ( WC2MB(dst, src, lenChunk) == wxCONV_FAILED ) + return wxCONV_FAILED; + + dst += lenChunk; + } + } + + return dstWritten; } -size_t wxMBConv::WC2MB(char *buf, const wchar_t *psz, size_t n) const +size_t wxMBConv::MB2WC(wchar_t *outBuff, const char *inBuff, size_t outLen) const { -#if IGNORE_LIBC - if (buf) + size_t rc = ToWChar(outBuff, outLen, inBuff); + if ( rc != wxCONV_FAILED ) { - for (size_t i = 0; i < wxStrlen( psz )+1; i++) - buf[i] = (char) psz[i]; - return wxStrlen( psz ); + // ToWChar() returns the buffer length, i.e. including the trailing + // NUL, while this method doesn't take it into account + rc--; } - else + + 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 ) { - return wxStrlen( psz ); + rc -= GetMBNulLen(); } -#else - return wxWC2MB(buf, psz, n); -#endif + + return rc; +} + +wxMBConv::~wxMBConv() +{ + // nothing to do here (necessary for Darwin linking probably) } const wxWCharBuffer wxMBConv::cMB2WC(const char *psz) const @@ -240,268 +350,1580 @@ 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); - MB2WC(buf.data(), psz, nLen + 1); // with the trailing NUL + wxWCharBuffer buf(nLen - 1 /* +1 added implicitly */); - return buf; + // +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); - WC2MB(buf.data(), pwz, nLen + 1); - - return buf; + wxCharBuffer buf(nLen - 1); + if ( FromWChar(buf.data(), nLen, pwz) != wxCONV_FAILED ) + return buf; } } - wxCharBuffer buf((char *)NULL); - - return buf; + return wxCharBuffer(); } -// ---------------------------------------------------------------------------- -// standard gdk conversion -// ---------------------------------------------------------------------------- +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 - 1] = L'\0'; + if ( ToWChar(wbuf.data(), dstLen, inBuff, inLen) != wxCONV_FAILED ) + { + if ( outLen ) + { + *outLen = dstLen; + if ( wbuf[dstLen - 1] == L'\0' ) + (*outLen)--; + } -#ifdef __WXGTK12__ + return wbuf; + } + } -WXDLLEXPORT_DATA(wxMBConvGdk) wxConvGdk; + if ( outLen ) + *outLen = 0; -#include + return wxWCharBuffer(); +} -size_t wxMBConvGdk::MB2WC(wchar_t *buf, const char *psz, size_t n) const +const wxCharBuffer +wxMBConv::cWC2MB(const wchar_t *inBuff, size_t inLen, size_t *outLen) const { - if (buf) - { - return gdk_mbstowcs((GdkWChar *)buf, psz, n); - } - else + size_t dstLen = FromWChar(NULL, 0, inBuff, inLen); + if ( dstLen != wxCONV_FAILED ) { - GdkWChar *nbuf = new GdkWChar[n=strlen(psz)]; - size_t len = gdk_mbstowcs(nbuf, psz, n); - delete[] nbuf; - return len; + const size_t nulLen = GetMBNulLen(); + + // 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; + + if ( dstLen >= nulLen && + !NotAllNULs(buf.data() + dstLen - nulLen, nulLen) ) + { + // in this case the output is NUL-terminated and we're not + // supposed to count NUL + *outLen -= nulLen; + } + } + + return buf; + } } + + if ( outLen ) + *outLen = 0; + + return wxCharBuffer(); } -size_t wxMBConvGdk::WC2MB(char *buf, const wchar_t *psz, size_t n) const +// ---------------------------------------------------------------------------- +// wxMBConvLibc +// ---------------------------------------------------------------------------- + +size_t wxMBConvLibc::MB2WC(wchar_t *buf, const char *psz, size_t n) const { - char *mbstr = gdk_wcstombs((GdkWChar *)psz); - size_t len = mbstr ? strlen(mbstr) : 0; - if (buf) - { - if (len > n) - len = n; - memcpy(buf, psz, len); - if (len < n) - buf[len] = 0; - } - return len; + return wxMB2WC(buf, psz, n); } -#endif // GTK > 1.0 +size_t wxMBConvLibc::WC2MB(char *buf, const wchar_t *psz, size_t n) const +{ + return wxWC2MB(buf, psz, n); +} // ---------------------------------------------------------------------------- -// UTF-7 +// wxConvBrokenFileNames // ---------------------------------------------------------------------------- -WXDLLEXPORT_DATA(wxMBConvUTF7) wxConvUTF7; - -#if 0 -static char utf7_setD[]="ABCDEFGHIJKLMNOPQRSTUVWXYZ" - "abcdefghijklmnopqrstuvwxyz" - "0123456789'(),-./:?"; -static char utf7_setO[]="!\"#$%&*;<=>@[]^_`{|}"; -static char utf7_setB[]="ABCDEFGHIJKLMNOPQRSTUVWXYZ" - "abcdefghijklmnopqrstuvwxyz" - "0123456789+/"; -#endif +#ifdef __UNIX__ -// TODO: write actual implementations of UTF-7 here -size_t wxMBConvUTF7::MB2WC(wchar_t * WXUNUSED(buf), - const char * WXUNUSED(psz), - size_t WXUNUSED(n)) const +wxConvBrokenFileNames::wxConvBrokenFileNames(const wxString& charset) { - return 0; + if ( wxStricmp(charset, _T("UTF-8")) == 0 || + wxStricmp(charset, _T("UTF8")) == 0 ) + m_conv = new wxMBConvUTF8(wxMBConvUTF8::MAP_INVALID_UTF8_TO_PUA); + else + m_conv = new wxCSConv(charset); } -size_t wxMBConvUTF7::WC2MB(char * WXUNUSED(buf), - const wchar_t * WXUNUSED(psz), - size_t WXUNUSED(n)) const -{ - return 0; -} +#endif // __UNIX__ // ---------------------------------------------------------------------------- -// UTF-8 +// UTF-7 // ---------------------------------------------------------------------------- -WXDLLEXPORT_DATA(wxMBConvUTF8) wxConvUTF8; +// Implementation (C) 2004 Fredrik Roubert +// +// Changes to work in streaming mode (C) 2008 Vadim Zeitlin -static wxUint32 utf8_max[]= - { 0x7f, 0x7ff, 0xffff, 0x1fffff, 0x3ffffff, 0x7fffffff, 0xffffffff }; +// +// BASE64 decoding table +// +static const unsigned char utf7unb64[] = +{ + 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, + 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, + 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, + 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, + 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, + 0xff, 0xff, 0xff, 0x3e, 0xff, 0xff, 0xff, 0x3f, + 0x34, 0x35, 0x36, 0x37, 0x38, 0x39, 0x3a, 0x3b, + 0x3c, 0x3d, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, + 0xff, 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, + 0x07, 0x08, 0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, + 0x0f, 0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, + 0x17, 0x18, 0x19, 0xff, 0xff, 0xff, 0xff, 0xff, + 0xff, 0x1a, 0x1b, 0x1c, 0x1d, 0x1e, 0x1f, 0x20, + 0x21, 0x22, 0x23, 0x24, 0x25, 0x26, 0x27, 0x28, + 0x29, 0x2a, 0x2b, 0x2c, 0x2d, 0x2e, 0x2f, 0x30, + 0x31, 0x32, 0x33, 0xff, 0xff, 0xff, 0xff, 0xff, + 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, + 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, + 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, + 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, + 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, + 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, + 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, + 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, + 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, + 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, + 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, + 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, + 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, + 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, + 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, + 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff +}; -size_t wxMBConvUTF8::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 = wx_const_cast(DecoderState *, &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++, fc = cc; - unsigned cnt; - for (cnt = 0; fc & 0x80; cnt++) - fc <<= 1; - if (!cnt) - { - // plain ASCII char - if (buf) - *buf++ = cc; - len++; - } - else + const unsigned char cc = *src++; + + if ( state.IsShifted() ) { - cnt--; - if (!cnt) + const unsigned char dc = utf7unb64[cc]; + if ( dc == 0xff ) { - // invalid UTF-8 sequence - return (size_t)-1; + // end of encoded part + state.ToDirect(); + + // re-parse this character normally below unless it's '-' which + // is consumed by the decoder + if ( cc == '-' ) + continue; } - else + else // valid encoded character { - unsigned ocnt = cnt - 1; - wxUint32 res = cc & (0x3f >> cnt); - while (cnt--) + // mini base64 decoder: each character is 6 bits + state.bit += 6; + state.accum <<= 6; + state.accum += dc; + + if ( state.bit >= 8 ) { - cc = *psz++; - if ((cc & 0xC0) != 0x80) + // got the full byte, consume it + state.bit -= 8; + unsigned char b = (state.accum >> state.bit) & 0x00ff; + + if ( state.isLSB ) { - // invalid UTF-8 sequence - return (size_t)-1; + // we've got the full word, output it + if ( dst ) + *dst++ = (state.msb << 8) | b; + len++; + state.isLSB = false; } - res = (res << 6) | (cc & 0x3f); + 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 == srcEnd ) + return wxCONV_FAILED; // can't have '+' at the end + + if ( *src == '-' ) + { + // just the encoded plus sign, don't switch to shifted mode + if ( dst ) + *dst++ = '+'; + len++; + src++; } - if (res <= utf8_max[ocnt]) + else { - // illegal UTF-8 encoding - return (size_t)-1; + state.ToShifted(); } -#ifdef WC_UTF16 - size_t pa = encode_utf16(res, buf); - if (pa == (size_t)-1) - return (size_t)-1; - if (buf) - buf += pa; - len += pa; -#else // !WC_UTF16 - if (buf) - *buf++ = res; + } + 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++; -#endif // WC_UTF16/!WC_UTF16 } } } - 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; } -size_t wxMBConvUTF8::WC2MB(char *buf, const wchar_t *psz, size_t n) const +// +// BASE64 encoding table +// +static const unsigned char utf7enb64[] = { - size_t len = 0; + 'A', 'B', 'C', 'D', 'E', 'F', 'G', 'H', + 'I', 'J', 'K', 'L', 'M', 'N', 'O', 'P', + 'Q', 'R', 'S', 'T', 'U', 'V', 'W', 'X', + 'Y', 'Z', 'a', 'b', 'c', 'd', 'e', 'f', + 'g', 'h', 'i', 'j', 'k', 'l', 'm', 'n', + 'o', 'p', 'q', 'r', 's', 't', 'u', 'v', + 'w', 'x', 'y', 'z', '0', '1', '2', '3', + '4', '5', '6', '7', '8', '9', '+', '/' +}; + +// +// UTF-7 encoding table +// +// 0 - Set D (directly encoded characters) +// 1 - Set O (optional direct characters) +// 2 - whitespace characters (optional) +// 3 - special characters +// +static const unsigned char utf7encode[128] = +{ + 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, + 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 3, 1, 1, 1, + 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 3, 3 +}; + +static inline bool wxIsUTF7Direct(wchar_t wc) +{ + return wc < 0x80 && utf7encode[wc] < 1; +} - while (*psz && ((!buf) || (len < n))) +size_t wxMBConvUTF7::FromWChar(char *dst, size_t dstLen, + const wchar_t *src, size_t srcLen) const +{ + EncoderState stateOrig, + *statePtr; + if ( srcLen == wxNO_LEN ) { - wxUint32 cc; -#ifdef WC_UTF16 - size_t pa = decode_utf16(psz, cc); - psz += (pa == (size_t)-1) ? 1 : pa; -#else - cc=(*psz++) & 0x7fffffff; -#endif - unsigned cnt; - for (cnt = 0; cc > utf8_max[cnt]; cnt++) {} - if (!cnt) + // 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 = wx_const_cast(EncoderState *, &m_stateEncoder); + } + + EncoderState& state = *statePtr; + + + size_t len = 0; + + const wchar_t * const srcEnd = src + srcLen; + while ( src < srcEnd && (!dst || len < dstLen) ) + { + wchar_t cc = *src++; + if ( wxIsUTF7Direct(cc) ) + { + 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 ( state.IsDirect() ) + { + state.ToShifted(); + + if ( dst ) + *dst++ = '+'; + len++; + } + + // BASE64 encode string + for ( ;; ) + { + for ( unsigned lsb = 0; lsb < 2; lsb++ ) + { + state.accum <<= 8; + state.accum += lsb ? cc & 0xff : (cc & 0xff00) >> 8; + + for (state.bit += 8; state.bit >= 6; ) + { + state.bit -= 6; + if ( dst ) + *dst++ = utf7enb64[(state.accum >> state.bit) % 64]; + len++; + } + } + + if ( src == srcEnd || wxIsUTF7Direct(cc = *src) ) + break; + + src++; + } + } + } + + // 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; +} + +// ---------------------------------------------------------------------------- +// UTF-8 +// ---------------------------------------------------------------------------- + +static const wxUint32 utf8_max[]= + { 0x7f, 0x7ff, 0xffff, 0x1fffff, 0x3ffffff, 0x7fffffff, 0xffffffff }; + +// boundaries of the private use area we use to (temporarily) remap invalid +// characters invalid in a UTF-8 encoded string +const wxUint32 wxUnicodePUA = 0x100000; +const wxUint32 wxUnicodePUAEnd = wxUnicodePUA + 256; + +// 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( _T("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 ((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 + { + cnt--; + if (!cnt) + { + // invalid UTF-8 sequence + invalid = true; + } + else + { + unsigned ocnt = cnt - 1; + wxUint32 res = cc & (0x3f >> cnt); + while (cnt--) + { + cc = *psz; + if ((cc & 0xC0) != 0x80) + { + // invalid UTF-8 sequence + invalid = true; + break; + } + + psz++; + res = (res << 6) | (cc & 0x3f); + } + + if (invalid || res <= utf8_max[ocnt]) + { + // illegal UTF-8 encoding + 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 == 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++ = (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 (srcLen == wxNO_LEN && buf && (len < n)) + *buf = 0; + + return len + 1; +} + +static inline bool isoctal(wchar_t wch) +{ + return L'0' <= wch && wch <= L'7'; +} + +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 ((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 == wxCONV_FAILED) ? 1 : pa; +#else + cc = (*psz++) & 0x7fffffff; +#endif + + if ( (m_options & MAP_INVALID_UTF8_TO_PUA) + && cc >= wxUnicodePUA && cc < wxUnicodePUAEnd ) + { + if (buf) + *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 + { + unsigned cnt; + for (cnt = 0; cc > utf8_max[cnt]; cnt++) + { + } + + 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 (srcLen == wxNO_LEN && buf && (len < n)) + *buf = 0; + + return len + 1; +} + +// ============================================================================ +// UTF-16 +// ============================================================================ + +#ifdef WORDS_BIGENDIAN + #define wxMBConvUTF16straight wxMBConvUTF16BE + #define wxMBConvUTF16swap wxMBConvUTF16LE +#else + #define wxMBConvUTF16swap wxMBConvUTF16BE + #define wxMBConvUTF16straight wxMBConvUTF16LE +#endif + +/* static */ +size_t wxMBConvUTF16Base::GetLength(const char *src, size_t srcLen) +{ + if ( srcLen == wxNO_LEN ) + { + // count the number of bytes in input, including the trailing NULs + const wxUint16 *inBuff = wx_reinterpret_cast(const wxUint16 *, 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 + +// ---------------------------------------------------------------------------- +// conversions without endianness change +// ---------------------------------------------------------------------------- + +size_t +wxMBConvUTF16straight::ToWChar(wchar_t *dst, size_t dstLen, + const char *src, size_t srcLen) const +{ + // 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; + + const size_t inLen = srcLen / BYTES_PER_CHAR; + if ( dst ) + { + if ( dstLen < inLen ) + return wxCONV_FAILED; + + memcpy(dst, src, srcLen); + } + + return inLen; +} + +size_t +wxMBConvUTF16straight::FromWChar(char *dst, size_t dstLen, + const wchar_t *src, size_t srcLen) const +{ + if ( srcLen == wxNO_LEN ) + srcLen = wxWcslen(src) + 1; + + srcLen *= BYTES_PER_CHAR; + + if ( dst ) + { + if ( dstLen < srcLen ) + return wxCONV_FAILED; + + memcpy(dst, src, srcLen); + } + + return srcLen; +} + +// ---------------------------------------------------------------------------- +// endian-reversing conversions +// ---------------------------------------------------------------------------- + +size_t +wxMBConvUTF16swap::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; + + srcLen /= BYTES_PER_CHAR; + + if ( dst ) + { + if ( dstLen < srcLen ) + return wxCONV_FAILED; + + const wxUint16 *inBuff = wx_reinterpret_cast(const wxUint16 *, src); + for ( size_t n = 0; n < srcLen; n++, inBuff++ ) + { + *dst++ = wxUINT16_SWAP_ALWAYS(*inBuff); + } + } + + return srcLen; +} + +size_t +wxMBConvUTF16swap::FromWChar(char *dst, size_t dstLen, + const wchar_t *src, size_t srcLen) const +{ + if ( srcLen == wxNO_LEN ) + srcLen = wxWcslen(src) + 1; + + srcLen *= BYTES_PER_CHAR; + + if ( dst ) + { + if ( dstLen < srcLen ) + return wxCONV_FAILED; + + wxUint16 *outBuff = wx_reinterpret_cast(wxUint16 *, dst); + for ( size_t n = 0; n < srcLen; n += BYTES_PER_CHAR, src++ ) + { + *outBuff++ = wxUINT16_SWAP_ALWAYS(*src); + } + } + + return srcLen; +} + +#else // !WC_UTF16: wchar_t is UTF-32 + +// ---------------------------------------------------------------------------- +// conversions without endianness change +// ---------------------------------------------------------------------------- + +size_t +wxMBConvUTF16straight::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 size_t inLen = srcLen / BYTES_PER_CHAR; + if ( !dst ) + { + // 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 outLen = 0; + const wxUint16 *inBuff = wx_reinterpret_cast(const wxUint16 *, 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; + } + + + return outLen; +} + +size_t +wxMBConvUTF16straight::FromWChar(char *dst, size_t dstLen, + const wchar_t *src, size_t srcLen) const +{ + if ( srcLen == wxNO_LEN ) + srcLen = wxWcslen(src) + 1; + + size_t outLen = 0; + wxUint16 *outBuff = wx_reinterpret_cast(wxUint16 *, dst); + for ( size_t n = 0; n < srcLen; n++ ) + { + wxUint16 cc[2]; + const size_t numChars = encode_utf16(*src++, cc); + if ( numChars == wxCONV_FAILED ) + return wxCONV_FAILED; + + outLen += numChars * BYTES_PER_CHAR; + if ( outBuff ) + { + if ( outLen > dstLen ) + return wxCONV_FAILED; + + *outBuff++ = cc[0]; + if ( numChars == 2 ) + { + // second character of a surrogate + *outBuff++ = cc[1]; + } + } + } + + return outLen; +} + +// ---------------------------------------------------------------------------- +// endian-reversing conversions +// ---------------------------------------------------------------------------- + +size_t +wxMBConvUTF16swap::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 size_t inLen = srcLen / BYTES_PER_CHAR; + if ( !dst ) + { + // 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 outLen = 0; + const wxUint16 *inBuff = wx_reinterpret_cast(const wxUint16 *, src); + for ( const wxUint16 * const inEnd = inBuff + inLen; inBuff < inEnd; ) + { + wxUint32 ch; + wxUint16 tmp[2]; + + 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++; + + if ( ++outLen > dstLen ) + return wxCONV_FAILED; + + *dst++ = ch; + } + + + return outLen; +} + +size_t +wxMBConvUTF16swap::FromWChar(char *dst, size_t dstLen, + const wchar_t *src, size_t srcLen) const +{ + if ( srcLen == wxNO_LEN ) + srcLen = wxWcslen(src) + 1; + + size_t outLen = 0; + wxUint16 *outBuff = wx_reinterpret_cast(wxUint16 *, dst); + for ( const wchar_t *srcEnd = src + srcLen; src < srcEnd; src++ ) + { + wxUint16 cc[2]; + const size_t numChars = encode_utf16(*src, cc); + if ( numChars == wxCONV_FAILED ) + return wxCONV_FAILED; + + outLen += numChars * BYTES_PER_CHAR; + if ( outBuff ) + { + if ( outLen > dstLen ) + return wxCONV_FAILED; + + *outBuff++ = wxUINT16_SWAP_ALWAYS(cc[0]); + if ( numChars == 2 ) + { + // second character of a surrogate + *outBuff++ = wxUINT16_SWAP_ALWAYS(cc[1]); + } + } + } + + return outLen; +} + +#endif // WC_UTF16/!WC_UTF16 + + +// ============================================================================ +// UTF-32 +// ============================================================================ + +#ifdef WORDS_BIGENDIAN + #define wxMBConvUTF32straight wxMBConvUTF32BE + #define wxMBConvUTF32swap wxMBConvUTF32LE +#else + #define wxMBConvUTF32swap wxMBConvUTF32BE + #define wxMBConvUTF32straight wxMBConvUTF32LE +#endif + + +WXDLLIMPEXP_DATA_BASE(wxMBConvUTF32LE) wxConvUTF32LE; +WXDLLIMPEXP_DATA_BASE(wxMBConvUTF32BE) wxConvUTF32BE; + +/* static */ +size_t wxMBConvUTF32Base::GetLength(const char *src, size_t srcLen) +{ + if ( srcLen == wxNO_LEN ) + { + // count the number of bytes in input, including the trailing NULs + const wxUint32 *inBuff = wx_reinterpret_cast(const wxUint32 *, src); + for ( srcLen = 1; *inBuff++; srcLen++ ) + ; + + srcLen *= BYTES_PER_CHAR; + } + else // we already have the length + { + // we can only convert an entire number of UTF-32 characters + if ( srcLen % BYTES_PER_CHAR ) + return wxCONV_FAILED; + } + + return srcLen; +} + +// case when in-memory representation is UTF-16 +#ifdef WC_UTF16 + +// ---------------------------------------------------------------------------- +// conversions without endianness change +// ---------------------------------------------------------------------------- + +size_t +wxMBConvUTF32straight::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 = wx_reinterpret_cast(const wxUint32 *, src); + const size_t inLen = srcLen / BYTES_PER_CHAR; + size_t outLen = 0; + for ( size_t n = 0; n < inLen; n++ ) + { + wxUint16 cc[2]; + const size_t numChars = encode_utf16(*inBuff++, cc); + if ( numChars == wxCONV_FAILED ) + return wxCONV_FAILED; + + outLen += numChars; + if ( dst ) { - // plain ASCII char - if (buf) - *buf++ = (char) cc; - len++; + if ( outLen > dstLen ) + return wxCONV_FAILED; + + *dst++ = cc[0]; + if ( numChars == 2 ) + { + // second character of a surrogate + *dst++ = cc[1]; + } } + } - else + return outLen; +} + +size_t +wxMBConvUTF32straight::FromWChar(char *dst, size_t dstLen, + const wchar_t *src, size_t srcLen) const +{ + if ( srcLen == wxNO_LEN ) + srcLen = wxWcslen(src) + 1; + + if ( !dst ) + { + // 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 = wx_reinterpret_cast(wxUint32 *, 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; + + outLen += BYTES_PER_CHAR; + + if ( outLen > dstLen ) + return wxCONV_FAILED; + + *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 = wx_reinterpret_cast(const wxUint32 *, src); + const size_t inLen = srcLen / BYTES_PER_CHAR; + size_t outLen = 0; + for ( size_t n = 0; n < inLen; n++, inBuff++ ) + { + wxUint16 cc[2]; + const size_t numChars = encode_utf16(wxUINT32_SWAP_ALWAYS(*inBuff), cc); + if ( numChars == wxCONV_FAILED ) + return wxCONV_FAILED; + + outLen += numChars; + if ( dst ) { - len += cnt + 1; - if (buf) + if ( outLen > dstLen ) + return wxCONV_FAILED; + + *dst++ = cc[0]; + if ( numChars == 2 ) { - *buf++ = (char) ((-128 >> cnt) | ((cc >> (cnt * 6)) & (0x3f >> cnt))); - while (cnt--) - *buf++ = (char) (0x80 | ((cc >> (cnt * 6)) & 0x3f)); + // second character of a surrogate + *dst++ = cc[1]; } } } - if (buf && (len dstLen ) + return wxCONV_FAILED; + + *outBuff++ = wxUINT32_SWAP_ALWAYS(ch); + } + + return outLen; } -// ============================================================================ -// wxCharacterSet and derived classes -// ============================================================================ +#else // !WC_UTF16: wchar_t is UTF-32 // ---------------------------------------------------------------------------- -// wxCharacterSet is the ABC for the classes below +// conversions without endianness change // ---------------------------------------------------------------------------- -class wxCharacterSet +size_t +wxMBConvUTF32straight::ToWChar(wchar_t *dst, size_t dstLen, + const char *src, size_t srcLen) const { -public: - wxCharacterSet(const wxChar*name) : cname(name) {} - virtual ~wxCharacterSet() {} - virtual size_t MB2WC(wchar_t *buf, const char *psz, size_t n) = 0; - virtual size_t WC2MB(char *buf, const wchar_t *psz, size_t n) = 0; - virtual bool usable() const = 0; -public: - const wxChar*cname; -}; + // use memcpy() as it should be much faster than hand-written loop + srcLen = GetLength(src, srcLen); + if ( srcLen == wxNO_LEN ) + return wxCONV_FAILED; + + const size_t inLen = srcLen/BYTES_PER_CHAR; + if ( dst ) + { + if ( dstLen < inLen ) + return wxCONV_FAILED; + + memcpy(dst, src, srcLen); + } + + return inLen; +} + +size_t +wxMBConvUTF32straight::FromWChar(char *dst, size_t dstLen, + const wchar_t *src, size_t srcLen) const +{ + if ( srcLen == wxNO_LEN ) + srcLen = wxWcslen(src) + 1; + + srcLen *= BYTES_PER_CHAR; + + if ( dst ) + { + if ( dstLen < srcLen ) + return wxCONV_FAILED; + + memcpy(dst, src, srcLen); + } + + return srcLen; +} // ---------------------------------------------------------------------------- -// ID_CharSet: implementation of wxCharacterSet using an existing wxMBConv +// endian-reversing conversions // ---------------------------------------------------------------------------- -class ID_CharSet : public wxCharacterSet +size_t +wxMBConvUTF32swap::ToWChar(wchar_t *dst, size_t dstLen, + const char *src, size_t srcLen) const { -public: - ID_CharSet(const wxChar *name, wxMBConv *cnv) - : wxCharacterSet(name), work(cnv) {} + srcLen = GetLength(src, srcLen); + if ( srcLen == wxNO_LEN ) + return wxCONV_FAILED; - size_t MB2WC(wchar_t *buf, const char *psz, size_t n) - { return work ? work->MB2WC(buf,psz,n) : (size_t)-1; } + srcLen /= BYTES_PER_CHAR; - size_t WC2MB(char *buf, const wchar_t *psz, size_t n) - { return work ? work->WC2MB(buf,psz,n) : (size_t)-1; } + if ( dst ) + { + if ( dstLen < srcLen ) + return wxCONV_FAILED; - bool usable() const - { return work!=NULL; } -public: - wxMBConv*work; -}; + const wxUint32 *inBuff = wx_reinterpret_cast(const wxUint32 *, src); + for ( size_t n = 0; n < srcLen; n++, inBuff++ ) + { + *dst++ = wxUINT32_SWAP_ALWAYS(*inBuff); + } + } + + return srcLen; +} + +size_t +wxMBConvUTF32swap::FromWChar(char *dst, size_t dstLen, + const wchar_t *src, size_t srcLen) const +{ + if ( srcLen == wxNO_LEN ) + srcLen = wxWcslen(src) + 1; + + srcLen *= BYTES_PER_CHAR; + + if ( dst ) + { + if ( dstLen < srcLen ) + return wxCONV_FAILED; + + wxUint32 *outBuff = wx_reinterpret_cast(wxUint32 *, dst); + for ( size_t n = 0; n < srcLen; n += BYTES_PER_CHAR, src++ ) + { + *outBuff++ = wxUINT32_SWAP_ALWAYS(*src); + } + } + + return srcLen; +} + +#endif // WC_UTF16/!WC_UTF16 // ============================================================================ @@ -510,12 +1932,13 @@ public: #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) && \ @@ -526,198 +1949,337 @@ public: #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 + // ---------------------------------------------------------------------------- -// IC_CharSet: encapsulates an iconv character set +// wxMBConv_iconv: encapsulates an iconv character set // ---------------------------------------------------------------------------- -class IC_CharSet : public wxCharacterSet +class wxMBConv_iconv : public wxMBConv { public: - IC_CharSet(const wxChar *name); - virtual ~IC_CharSet(); + wxMBConv_iconv(const char *name); + virtual ~wxMBConv_iconv(); + + // 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 size_t MB2WC(wchar_t *buf, const char *psz, size_t n); - virtual size_t WC2MB(char *buf, const wchar_t *psz, size_t n); + virtual wxMBConv *Clone() const + { + wxMBConv_iconv *p = new wxMBConv_iconv(m_name.ToAscii()); + p->m_minMBCharWidth = m_minMBCharWidth; + return p; + } - bool usable() const - { return (m2w != (iconv_t)-1) && (w2m != (iconv_t)-1); } + bool IsOk() const + { 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 + // the name (for iconv_open()) of a wide char charset -- if none is // available on this machine, it will remain NULL - static const char *ms_wcCharsetName; + static wxString ms_wcCharsetName; // true if the wide char encoding we use (i.e. ms_wcCharsetName) has // different endian-ness than the native one static bool ms_wcNeedsSwap; + + + // name of the encoding handled by this conversion + wxString m_name; + + // cached result of GetMBNulLen(); set to 0 meaning "unknown" + // initially + size_t m_minMBCharWidth; }; -const char *IC_CharSet::ms_wcCharsetName = NULL; -bool IC_CharSet::ms_wcNeedsSwap = FALSE; +// 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; -IC_CharSet::IC_CharSet(const wxChar *name) - : wxCharacterSet(name) +wxMBConv_iconv::wxMBConv_iconv(const char *name) + : m_name(name) { - // Do it the hard way - char cname[100]; - for (size_t i = 0; i < wxStrlen(name)+1; i++) - cname[i] = (char) name[i]; + m_minMBCharWidth = 0; // check for charset that represents wchar_t: - if (ms_wcCharsetName == NULL) + if ( ms_wcCharsetName.empty() ) { - ms_wcNeedsSwap = FALSE; + wxLogTrace(TRACE_STRCONV, _T("Looking for wide char codeset:")); - // try charset with explicit bytesex info (e.g. "UCS-4LE"): - ms_wcCharsetName = WC_NAME_BEST; - m2w = iconv_open(ms_wcCharsetName, cname); +#if wxUSE_FONTMAP + const wxChar **names = wxFontMapperBase::GetAllEncodingNames(WC_ENC); +#else // !wxUSE_FONTMAP + static const wxChar *names_static[] = + { +#if SIZEOF_WCHAR_T == 4 + _T("UCS-4"), +#elif SIZEOF_WCHAR_T = 2 + _T("UCS-2"), +#endif + NULL + }; + const wxChar **names = names_static; +#endif // wxUSE_FONTMAP/!wxUSE_FONTMAP - if (m2w == (iconv_t)-1) + for ( ; *names && ms_wcCharsetName.empty(); ++names ) { - // try charset w/o bytesex info (e.g. "UCS4") - // and check for bytesex ourselves: - ms_wcCharsetName = WC_NAME; - m2w = iconv_open(ms_wcCharsetName, cname); + const wxString nameCS(*names); - // last bet, try if it knows WCHAR_T pseudo-charset - if (m2w == (iconv_t)-1) - { - ms_wcCharsetName = "WCHAR_T"; - m2w = iconv_open(ms_wcCharsetName, cname); - } + // first try charset with explicit bytesex info (e.g. "UCS-4LE"): + wxString nameXE(nameCS); + +#ifdef WORDS_BIGENDIAN + nameXE += _T("BE"); +#else // little endian + nameXE += _T("LE"); +#endif - if (m2w != (iconv_t)-1) + wxLogTrace(TRACE_STRCONV, _T(" trying charset \"%s\""), + nameXE.c_str()); + + m2w = iconv_open(nameXE.ToAscii(), 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, _T(" 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 wxWindows 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 ? _T(" (needs swap)") + : _T("")); } 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); + } } } -IC_CharSet::~IC_CharSet() +wxMBConv_iconv::~wxMBConv_iconv() { - if ( m2w != (iconv_t)-1 ) + if ( m2w != ICONV_T_INVALID ) iconv_close(m2w); - if ( w2m != (iconv_t)-1 ) + if ( w2m != ICONV_T_INVALID ) iconv_close(w2m); } -size_t IC_CharSet::MB2WC(wchar_t *buf, const char *psz, size_t n) +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 + // convert using temp buffer to calculate the size of the buffer needed wchar_t tbuf[8]; 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 IC_CharSet::WC2MB(char *buf, const wchar_t *psz, size_t n) +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; @@ -725,41 +2287,39 @@ size_t IC_CharSet::WC2MB(char *buf, const wchar_t *psz, size_t n) if (ms_wcNeedsSwap) { // need to copy to temp buffer to switch endianness - // this absolutely doesn't rock! - // (no, doing WC_BSWAP twice on the original buffer won't help, as it + // (doing WC_BSWAP twice on the original buffer won't help, as it // could be in read-only memory, or be accessed in some other thread) - tmpbuf=(wchar_t*)malloc((inbuf+1)*SIZEOF_WCHAR_T); - memcpy(tmpbuf,psz,(inbuf+1)*SIZEOF_WCHAR_T); - WC_BSWAP(tmpbuf, inbuf) - psz=tmpbuf; + tmpbuf = (wchar_t *)malloc(inbuflen + SIZEOF_WCHAR_T); + for ( size_t i = 0; i < srcLen; i++ ) + tmpbuf[i] = WC_BSWAP(src[i]); + + tmpbuf[srcLen] = L'\0'; + 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 ); - - res = n-outbuf; + cres = iconv(w2m, ICONV_CHAR_CAST(&inbuf), &inbuflen, &dst, &outbuflen); - // 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 + // convert using temp buffer to calculate the size of the buffer needed char tbuf[16]; res = 0; - do { - buf = tbuf; outbuf = 16; + do + { + dst = tbuf; + outbuflen = 16; - 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 += 16 - outbuflen; + } + while ((cres == (size_t)-1) && (errno == E2BIG)); } if (ms_wcNeedsSwap) @@ -767,77 +2327,378 @@ size_t IC_CharSet::WC2MB(char *buf, const wchar_t *psz, size_t n) 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 + // ============================================================================ // Win32 conversion classes // ============================================================================ -#if defined(__WIN32__) && !defined(__WXMICROWIN__) && !defined(__WXUNIVERSAL__) +#ifdef wxHAVE_WIN32_MB2WC -extern long wxCharsetToCodepage(const wxChar *charset); // from utils.cpp +// from utils.cpp +#if wxUSE_FONTMAP +extern WXDLLIMPEXP_BASE long wxCharsetToCodepage(const char *charset); +extern WXDLLIMPEXP_BASE long wxEncodingToCodepage(wxFontEncoding encoding); +#endif -class CP_CharSet : public wxCharacterSet +class wxMBConv_win32 : public wxMBConv { public: - CP_CharSet(const wxChar* name) - : wxCharacterSet(name) + 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 char* name) + { + m_CodePage = wxCharsetToCodepage(name); + m_minMBCharWidth = 0; + } + + wxMBConv_win32(wxFontEncoding encoding) + { + m_CodePage = wxEncodingToCodepage(encoding); + m_minMBCharWidth = 0; + } +#endif // wxUSE_FONTMAP + + 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 ) { - m_CodePage = wxCharsetToCodepage(name); + 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; } - size_t MB2WC(wchar_t *buf, const char *psz, size_t n) - { const size_t len = ::MultiByteToWideChar ( m_CodePage, // code page - 0, // flags (none) + 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 - 1; + } + + virtual size_t WC2MB(char *buf, const wchar_t *pwz, size_t n) const + { + /* + we have a problem here: by default, WideCharToMultiByte() may + replace characters unrepresentable in the target code page with bad + quality approximations such as turning "1/2" symbol (U+00BD) into + "1" for the code pages which don't have it and we, obviously, want + to avoid this at any price + + the trouble is that this function does it _silently_, i.e. it won't + even tell us whether it did or not... Win98/2000 and higher provide + WC_NO_BEST_FIT_CHARS but it doesn't work for the older systems and + we have to resort to a round trip, i.e. check that converting back + results in the same string -- this is, of course, expensive but + otherwise we simply can't be sure to not garble the data. + */ + + // determine if we can rely on WC_NO_BEST_FIT_CHARS: according to MSDN + // it doesn't work with CJK encodings (which we test for rather roughly + // here...) nor with UTF-7/8 nor, of course, with Windows versions not + // supporting it + BOOL usedDef wxDUMMY_INITIALIZE(false); + BOOL *pUsedDef; + 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 wxCONV_FAILED; + } + + // 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; + } + + 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; + } + } + + // see the comment above for the reason of "len - 1" + return len - 1; + } + + virtual size_t GetMBNulLen() const + { + if ( m_minMBCharWidth == 0 ) + { + int len = ::WideCharToMultiByte + ( + m_CodePage, // code page + 0, // no flags + L"", // input string + 1, // translate just the NUL + NULL, // output buffer + 0, // and its size + NULL, // no replacement char + NULL // [out] don't care if it was used + ); + + wxMBConv_win32 * const self = wxConstCast(this, wxMBConv_win32); + switch ( len ) + { + default: + wxLogDebug(_T("Unexpected NUL length %d"), len); + self->m_minMBCharWidth = (size_t)-1; + break; + + case 0: + self->m_minMBCharWidth = (size_t)-1; + break; + + case 1: + case 2: + case 4: + self->m_minMBCharWidth = len; + break; + } + } + + return m_minMBCharWidth; + } + + virtual wxMBConv *Clone() const { return new wxMBConv_win32(*this); } + + bool IsOk() const { return m_CodePage != -1; } + +private: + static bool CanUseNoBestFit() + { + static int s_isWin98Or2k = -1; + + if ( s_isWin98Or2k == -1 ) + { + int verMaj, verMin; + switch ( wxGetOsVersion(&verMaj, &verMin) ) + { + case wxOS_WINDOWS_9X: + s_isWin98Or2k = verMaj >= 4 && verMin >= 10; + break; + + case wxOS_WINDOWS_NT: + s_isWin98Or2k = verMaj >= 5; + break; + + default: + // unknown: be conservative by default + s_isWin98Or2k = 0; + break; + } + + wxASSERT_MSG( s_isWin98Or2k != -1, _T("should be set above") ); + } - // note that it returns # of written chars for buf != NULL and *size* - // of the needed buffer for buf == NULL - return len ? (buf ? len : len - 1) : (size_t)-1; + return s_isWin98Or2k == 1; } - size_t WC2MB(char *buf, const wchar_t *psz, size_t n) + static bool IsAtLeastWin2kSP4() { - const size_t len = ::WideCharToMultiByte - ( - m_CodePage, // code page - 0, // flags (none) - psz, // input string - -1, // it is (wide) NUL-terminated - buf, // output buffer - buf ? n : 0, // and its size - NULL, // default "replacement" char - NULL // [out] was it used? - ); +#ifdef __WXWINCE__ + return false; +#else + static int s_isAtLeastWin2kSP4 = -1; + + if ( s_isAtLeastWin2kSP4 == -1 ) + { + OSVERSIONINFOEX ver; + + memset(&ver, 0, sizeof(ver)); + ver.dwOSVersionInfoSize = sizeof(ver); + GetVersionEx((OSVERSIONINFO*)&ver); + + s_isAtLeastWin2kSP4 = + ((ver.dwMajorVersion > 5) || // Vista+ + (ver.dwMajorVersion == 5 && ver.dwMinorVersion > 0) || // XP/2003 + (ver.dwMajorVersion == 5 && ver.dwMinorVersion == 0 && + ver.wServicePackMajor >= 4)) // 2000 SP4+ + ? 1 : 0; + } - // see the comment above! - return len ? (buf ? len : len - 1) : (size_t)-1; + return s_isAtLeastWin2kSP4 == 1; +#endif } - bool usable() const - { return m_CodePage != -1; } -public: + // the code page we're working with long m_CodePage; + + // cached result of GetMBNulLen(), set to 0 initially meaning + // "unknown" + size_t m_minMBCharWidth; }; -#endif // defined(__WIN32__) && !defined(__WXMICROWIN__) && !defined(__WXUNIVERSAL__) + +#endif // wxHAVE_WIN32_MB2WC + // ============================================================================ // wxEncodingConverter based conversion classes @@ -845,132 +2706,147 @@ public: #if wxUSE_FONTMAP -class EC_CharSet : public wxCharacterSet +class wxMBConv_wxwin : public wxMBConv { +private: + void Init() + { + // 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 - EC_CharSet(const wxChar* name) : wxCharacterSet(name), - enc(wxFONTENCODING_SYSTEM) + wxMBConv_wxwin(const char* name) { if (name) - enc = wxFontMapper::Get()->CharsetToEncoding(name, FALSE); + m_enc = wxFontMapperBase::Get()->CharsetToEncoding(name, false); + else + m_enc = wxFONTENCODING_SYSTEM; + + Init(); + } + + wxMBConv_wxwin(wxFontEncoding enc) + { + m_enc = enc; - m_ok = m2w.Init(enc, wxFONTENCODING_UNICODE) && - w2m.Init(wxFONTENCODING_UNICODE, enc); + Init(); } - size_t MB2WC(wchar_t *buf, const char *psz, size_t WXUNUSED(n)) + size_t MB2WC(wchar_t *buf, const char *psz, size_t WXUNUSED(n)) const { size_t inbuf = strlen(psz); if (buf) - m2w.Convert(psz,buf); + { + if (!m2w.Convert(psz, buf)) + return wxCONV_FAILED; + } return inbuf; } - size_t WC2MB(char *buf, const wchar_t *psz, size_t WXUNUSED(n)) + size_t WC2MB(char *buf, const wchar_t *psz, size_t WXUNUSED(n)) const { const size_t inbuf = wxWcslen(psz); if (buf) - w2m.Convert(psz,buf); + { + if (!w2m.Convert(psz, buf)) + return wxCONV_FAILED; + } return inbuf; } - bool usable() const { return m_ok; } + 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 enc; + wxFontEncoding m_enc; wxEncodingConverter m2w, w2m; +private: // were we initialized successfully? bool m_ok; -}; - -#endif // wxUSE_FONTMAP -// ---------------------------------------------------------------------------- -// the function creating the wxCharacterSet for the specified charset on the -// current system, trying all possibilities -// ---------------------------------------------------------------------------- + DECLARE_NO_COPY_CLASS(wxMBConv_wxwin) +}; -static wxCharacterSet *wxGetCharacterSet(const wxChar *name) +// make the constructors available for unit testing +WXDLLIMPEXP_BASE wxMBConv* new_wxMBConv_wxwin( const char* name ) { - // check for the special case of ASCII charset -#if wxUSE_FONTMAP - if ( wxFontMapper::Get()->CharsetToEncoding(name) == wxFONTENCODING_DEFAULT ) -#else // wxUSE_FONTMAP - if ( !name ) -#endif // wxUSE_FONTMAP/!wxUSE_FONTMAP + wxMBConv_wxwin* result = new wxMBConv_wxwin( name ); + if ( !result->IsOk() ) { - // don't convert at all - return NULL; + delete result; + return 0; } - // the test above must have taken care of this case - wxCHECK_MSG( name, NULL, _T("NULL name must be wxFONTENCODING_DEFAULT") ); - - wxCharacterSet *cset; + return result; +} - if ( wxStricmp(name, wxT("UTF8")) == 0 || wxStricmp(name, wxT("UTF-8")) == 0) - { - cset = new ID_CharSet(name, &wxConvUTF8); - } - else - { -#ifdef HAVE_ICONV - cset = new IC_CharSet(name); -#else // !HAVE_ICONV - cset = NULL; -#endif // HAVE_ICONV/!HAVE_ICONV - } +#endif // wxUSE_FONTMAP - // it can only be NULL in this case -#ifndef HAVE_ICONV - if ( cset ) -#endif // !HAVE_ICONV - { - if ( cset->usable() ) - return cset; +// ============================================================================ +// wxCSConv implementation +// ============================================================================ - delete cset; - cset = NULL; - } +void wxCSConv::Init() +{ + m_name = NULL; + m_convReal = NULL; + m_deferred = true; +} -#if defined(__WIN32__) && !defined(__WXMICROWIN__) && !defined(__WXUNIVERSAL__) - cset = new CP_CharSet(name); - if ( cset->usable() ) - return cset; +wxCSConv::wxCSConv(const wxString& charset) +{ + Init(); - delete cset; - cset = NULL; -#endif // defined(__WIN32__) && !defined(__WXMICROWIN__) && !defined(__WXUNIVERSAL__) + if ( !charset.empty() ) + { + SetName(charset.ToAscii()); + } #if wxUSE_FONTMAP - cset = new EC_CharSet(name); - if ( cset->usable() ) - return cset; - - delete cset; - cset = NULL; -#endif // wxUSE_FONTMAP - - wxLogError(_("Cannot convert from encoding '%s'!"), name); - - return NULL; + m_encoding = wxFontMapperBase::GetEncodingFromName(charset); +#else + m_encoding = wxFONTENCODING_SYSTEM; +#endif } -// ============================================================================ -// wxCSConv implementation -// ============================================================================ - -wxCSConv::wxCSConv(const wxChar *charset) +wxCSConv::wxCSConv(wxFontEncoding encoding) { - m_name = (wxChar *)NULL; - m_cset = (wxCharacterSet *) NULL; - m_deferred = TRUE; + if ( encoding == wxFONTENCODING_MAX || encoding == wxFONTENCODING_DEFAULT ) + { + wxFAIL_MSG( _T("invalid encoding value in wxCSConv ctor") ); + + encoding = wxFONTENCODING_SYSTEM; + } - SetName(charset); + Init(); + + m_encoding = encoding; } wxCSConv::~wxCSConv() @@ -979,67 +2855,323 @@ wxCSConv::~wxCSConv() } wxCSConv::wxCSConv(const wxCSConv& conv) - : wxMBConv() + : wxMBConv() { - Clear(); + Init(); + SetName(conv.m_name); + m_encoding = conv.m_encoding; } wxCSConv& wxCSConv::operator=(const wxCSConv& conv) { Clear(); + SetName(conv.m_name); + m_encoding = conv.m_encoding; + return *this; } void wxCSConv::Clear() { - if (m_name) - free(m_name); - if (m_cset) - delete m_cset; + free(m_name); + delete m_convReal; + m_name = NULL; - m_cset = NULL; + m_convReal = NULL; } -void wxCSConv::SetName(const wxChar *charset) +void wxCSConv::SetName(const char *charset) { if (charset) { m_name = wxStrdup(charset); - m_deferred = TRUE; + m_deferred = true; } } -void wxCSConv::LoadNow() +#if wxUSE_FONTMAP + +WX_DECLARE_HASH_MAP( wxFontEncoding, wxString, wxIntegerHash, wxIntegerEqual, + wxEncodingNameCache ); + +static wxEncodingNameCache gs_nameCache; +#endif + +wxMBConv *wxCSConv::DoCreate() const { - if ( m_deferred ) +#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 + if ( m_encoding == wxFONTENCODING_ISO8859_1 || + m_encoding == wxFONTENCODING_DEFAULT ) { - // it would probably be better to make GetSystemEncodingName() always - // available (i.e. even when wxUSE_INTL == 0)? -#if wxUSE_INTL - if ( !m_name ) + // don't convert at all + return NULL; + } + + // we trust OS to do conversion better than we can so try external + // conversion methods first + // + // the full order is: + // 1. OS conversion (iconv() under Unix or Win32 API) + // 2. hard coded conversions for UTF + // 3. wxEncodingConverter as fall back + + // step (1) +#ifdef HAVE_ICONV +#if !wxUSE_FONTMAP + if ( m_name ) +#endif // !wxUSE_FONTMAP + { +#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 + encoding = + wxFontMapperBase::Get()->CharsetToEncoding(m_name, false); +#endif // wxUSE_FONTMAP + } +#if wxUSE_FONTMAP { - wxString name = wxLocale::GetSystemEncodingName(); - if ( !name.empty() ) + 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(it->second.ToAscii()); + if ( conv->IsOk() ) + return conv; + + delete conv; + } + + const wxChar** names = wxFontMapperBase::GetAllEncodingNames(encoding); + // CS : in case this does not return valid names (eg for MacRoman) + // encoding got a 'failure' entry in the cache all the same, + // although it just has to be created using a different method, so + // only store failed iconv creation attempts (or perhaps we + // shoulnd't do this at all ?) + if ( names[0] != NULL ) { - SetName(name); + 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] = _T(""); // cache the failure } } +#endif // wxUSE_FONTMAP + } +#endif // HAVE_ICONV + +#ifdef wxHAVE_WIN32_MB2WC + { +#if wxUSE_FONTMAP + wxMBConv_win32 *conv = m_name ? new wxMBConv_win32(m_name) + : new wxMBConv_win32(m_encoding); + if ( conv->IsOk() ) + return conv; + + delete conv; +#else + return NULL; +#endif + } +#endif // wxHAVE_WIN32_MB2WC + +#ifdef __DARWIN__ + { + // 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 + + if ( conv->IsOk() ) + return conv; + + delete conv; + } + } +#endif // __DARWIN__ + + // step (2) + wxFontEncoding enc = m_encoding; +#if wxUSE_FONTMAP + if ( enc == wxFONTENCODING_SYSTEM && m_name ) + { + // 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 = wxFontMapperBase::Get()->CharsetToEncoding(m_name, false); + } +#endif // wxUSE_FONTMAP + + switch ( enc ) + { + case wxFONTENCODING_UTF7: + return new wxMBConvUTF7; + + case wxFONTENCODING_UTF8: + return new wxMBConvUTF8; + + case wxFONTENCODING_UTF16BE: + return new wxMBConvUTF16BE; + + case wxFONTENCODING_UTF16LE: + return new wxMBConvUTF16LE; + + case wxFONTENCODING_UTF32BE: + return new wxMBConvUTF32BE; + + case wxFONTENCODING_UTF32LE: + return new wxMBConvUTF32LE; + + default: + // nothing to do but put here to suppress gcc warnings + break; + } + + // step (3) +#if wxUSE_FONTMAP + { + wxMBConv_wxwin *conv = m_name ? new wxMBConv_wxwin(m_name) + : new wxMBConv_wxwin(m_encoding); + if ( conv->IsOk() ) + return conv; + + 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 a timestamp, for which it will need wxConvLocal (to convert + // time to char* and then wchar_t*), but that fails, tries to log the + // error, but wxLog has an (already locked) critical section that + // guards the static buffer. + static bool alreadyLoggingError = false; + if (!alreadyLoggingError) + { + alreadyLoggingError = true; + wxLogError(_("Cannot convert from the charset '%s'!"), + m_name ? m_name + : +#if wxUSE_FONTMAP + (const char*)wxFontMapperBase::GetEncodingDescription(m_encoding).ToAscii() +#else // !wxUSE_FONTMAP + (const char*)wxString::Format(_("encoding %i"), m_encoding).ToAscii() +#endif // wxUSE_FONTMAP/!wxUSE_FONTMAP + ); + + alreadyLoggingError = false; + } + + return NULL; +} + +void wxCSConv::CreateConvIfNeeded() const +{ + if ( m_deferred ) + { + wxCSConv *self = (wxCSConv *)this; // const_cast + + // 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 ) + { +#if wxUSE_INTL + self->m_encoding = wxLocale::GetSystemEncoding(); +#else + // fallback to some reasonable default: + self->m_encoding = wxFONTENCODING_ISO8859_1; #endif // wxUSE_INTL + } - // wxGetCharacterSet() complains about NULL name - m_cset = m_name ? wxGetCharacterSet(m_name) : NULL; - m_deferred = FALSE; + self->m_convReal = DoCreate(); + self->m_deferred = false; } } +bool wxCSConv::IsOk() const +{ + CreateConvIfNeeded(); + + // special case: no convReal created for wxFONTENCODING_ISO8859_1 + if ( m_encoding == wxFONTENCODING_ISO8859_1 ) + return true; // always ok as we do it ourselves + + // 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::ToWChar(wchar_t *dst, size_t dstLen, + const char *src, size_t srcLen) const +{ + CreateConvIfNeeded(); + + if (m_convReal) + return m_convReal->ToWChar(dst, dstLen, src, srcLen); + + // latin-1 (direct) + return wxMBConv::ToWChar(dst, dstLen, src, srcLen); +} + +size_t wxCSConv::FromWChar(char *dst, size_t dstLen, + const wchar_t *src, size_t srcLen) const +{ + CreateConvIfNeeded(); + + if (m_convReal) + return m_convReal->FromWChar(dst, dstLen, src, srcLen); + + // latin-1 (direct) + return wxMBConv::FromWChar(dst, dstLen, src, srcLen); +} + size_t wxCSConv::MB2WC(wchar_t *buf, const char *psz, size_t n) const { - ((wxCSConv *)this)->LoadNow(); // discard constness + CreateConvIfNeeded(); - if (m_cset) - return m_cset->MB2WC(buf, psz, n); + if (m_convReal) + return m_convReal->MB2WC(buf, psz, n); // latin-1 (direct) size_t len = strlen(psz); @@ -1055,22 +3187,167 @@ size_t wxCSConv::MB2WC(wchar_t *buf, const char *psz, size_t n) const size_t wxCSConv::WC2MB(char *buf, const wchar_t *psz, size_t n) const { - ((wxCSConv *)this)->LoadNow(); // discard constness + CreateConvIfNeeded(); - if (m_cset) - return m_cset->WC2MB(buf, psz, n); + if (m_convReal) + return m_convReal->WC2MB(buf, psz, n); // latin-1 (direct) const size_t len = wxWcslen(psz); if (buf) { for (size_t c = 0; c <= len; c++) - buf[c] = (psz[c] > 0xff) ? '?' : psz[c]; + { + if (psz[c] > 0xFF) + return wxCONV_FAILED; + + buf[c] = (char)psz[c]; + } + } + else + { + for (size_t c = 0; c <= len; c++) + { + if (psz[c] > 0xFF) + return wxCONV_FAILED; + } } return len; } -#endif // wxUSE_WCHAR_T +size_t wxCSConv::GetMBNulLen() const +{ + CreateConvIfNeeded(); + + if ( m_convReal ) + { + return m_convReal->GetMBNulLen(); + } + + // otherwise, we are ISO-8859-1 + return 1; +} + +#if wxUSE_UNICODE_UTF8 +bool wxCSConv::IsUTF8() const +{ + CreateConvIfNeeded(); + + if ( m_convReal ) + { + return m_convReal->IsUTF8(); + } + + // otherwise, we are ISO-8859-1 + return false; +} +#endif + + +#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; +} + +#endif // wxUSE_UNICODE + +// ---------------------------------------------------------------------------- +// globals +// ---------------------------------------------------------------------------- + +// 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 __WINDOWS__ + WX_DEFINE_GLOBAL_CONV2(wxMBConv, wxMBConv_win32, wxConvLibc, wxEMPTY_PARAMETER_VALUE); +#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__ +// The xnu kernel always communicates file paths in decomposed UTF-8. +// WARNING: Are we sure that CFString's conversion will cause decomposition? +static wxMBConv_cf wxConvMacUTF8DObj(wxFONTENCODING_UTF8); +#endif + +WXDLLIMPEXP_DATA_BASE(wxMBConv *) wxConvFileName = +#ifdef __DARWIN__ + &wxConvMacUTF8DObj; +#else // !__DARWIN__ + wxGet_wxConvLibcPtr(); +#endif // __DARWIN__/!__DARWIN__ + +#else // !wxUSE_WCHAR_T +// FIXME-UTF8: remove this, wxUSE_WCHAR_T is required now +// stand-ins in absence of wchar_t +WXDLLIMPEXP_DATA_BASE(wxMBConv) wxConvLibc, + wxConvISO8859_1, + wxConvLocal, + wxConvUTF8; +#endif // wxUSE_WCHAR_T/!wxUSE_WCHAR_T