From eec47cc6c45a2d9a474cae9da897ebfdb0b7be21 Mon Sep 17 00:00:00 2001 From: Vadim Zeitlin Date: Sat, 1 Apr 2006 12:43:03 +0000 Subject: [PATCH] Described in the comments and documented the semantics of the parameters and return values of wxMBConv methods and tried to make them more consistent. The only (intentional) backwards incompatible change is that cMB2WC/cWC2MB now return the length of the converted string in outLen parameter and not length+1 Added wxMBConv::GetMBNul() and use it instead of supposing that all multibyte strings are always terminated with a single NUL which is wrong for UTF-16/32. Using GetMBNul(), completely rewrote cMB2WC/cWC2MB() to accept a string of the specified length, whether it is NUL-terminated or not. This means that they don't overwrite the provided buffer any more and convert the entire string in all cases. Fixed bug in wxMBConvUTF16::WC2MB() which didn't NUL-terminate the string properlyv even if there was enough space. git-svn-id: https://svn.wxwidgets.org/svn/wx/wxWidgets/trunk@38498 c3d73ce0-8a6f-49c7-b76d-6d57e0e08775 --- docs/changes.txt | 3 + docs/latex/wx/mbconv.tex | 113 +++++++++---- include/wx/strconv.h | 111 +++++++++--- src/common/strconv.cpp | 353 ++++++++++++++++++++++++++------------- src/common/string.cpp | 114 ++++--------- 5 files changed, 446 insertions(+), 248 deletions(-) diff --git a/docs/changes.txt b/docs/changes.txt index 7a22272a8c..cc5d1829fb 100644 --- a/docs/changes.txt +++ b/docs/changes.txt @@ -11,6 +11,9 @@ INCOMPATIBLE CHANGES SINCE 2.6.x - For all wxInputStreams, Eof() becomes true after an attempt has been made to read _past_ the end of file. - wxCHECK family of macros now must be followed by a semicolon +- wxMBConv::cMB2WC() and cWC2MB() take size of the input buffer and return + length of the converted string in all cases now. + Deprecated methods since 2.6.x and their replacements ----------------------------------------------------- diff --git a/docs/latex/wx/mbconv.tex b/docs/latex/wx/mbconv.tex index 3d5e4a07b4..102186019a 100644 --- a/docs/latex/wx/mbconv.tex +++ b/docs/latex/wx/mbconv.tex @@ -1,14 +1,38 @@ -% -% automatically generated by HelpGen from -% ../include/wx/strconv.h at 25/Mar/00 10:20:56 -% +%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% +%% Name: mbconv.tex +%% Purpose: wxMBConv documentation +%% Author: Ove Kaaven, Vadim Zeitlin +%% Created: 2000-03-25 +%% RCS-ID: $Id$ +%% Copyright: (c) 2000 Ove Kaaven +%% (c) 2003-2006 Vadim Zeitlin +%% License: wxWindows license +%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% + \section{\class{wxMBConv}}\label{wxmbconv} This class is the base class of a hierarchy of classes capable of converting -text strings between multibyte (SBCS or DBCS) encodings and Unicode. It is itself -a wrapper around the standard libc mbstowcs() and wcstombs() routines, and has -one predefined instance, {\bf wxConvLibc}. +text strings between multibyte (SBCS or DBCS) encodings and Unicode. + +In the documentation for this and related classes please notice that +\emph{length} of the string refers to the number of characters in the string +not counting the terminating \NUL, if any. While the \emph{size} of the string +is the total number of bytes in the string, including any trailing {\NUL}s. +Thus, length of wide character string \texttt{L"foo"} is $3$ while its size can +be either $8$ or $16$ depending on whether \texttt{wchar\_t} is $2$ bytes (as +under Windows) or $4$ (Unix). + +\wxheading{Global variables} + +There are several predefined instances of this class: +\begin{twocollist} +\twocolitem{\textbf{wxConvLibc}}{Uses the standard ANSI C \texttt{mbstowcs()} and +\texttt{wcstombs()} functions to perform the conversions; thus depends on the +current locale.} +\twocolitem{\textbf{wxConvFile}}{The appropriate conversion for the file names, +depends on the system.} +\end{twocollist} \wxheading{Derived from} @@ -35,30 +59,31 @@ Constructor. \membersection{wxMBConv::MB2WC}\label{wxmbconvmb2wc} -\constfunc{virtual size\_t}{MB2WC}{\param{wchar\_t *}{outputBuf}, \param{const char *}{psz}, \param{size\_t }{outputSize}} +\constfunc{virtual size\_t}{MB2WC}{\param{wchar\_t *}{out}, \param{const char *}{in}, \param{size\_t }{outLen}} + +Converts from a string \arg{in} in multibyte encoding to Unicode putting up to +\arg{outLen} characters into the buffer \arg{out}. -Converts from a string {\it psz} in multibyte encoding to Unicode putting the -output into the buffer {\it outputBuf} of the maximum size {\it outputSize} (in wide -characters, not bytes). If {\it outputBuf} is {\tt NULL}, only the length of the -string which would result from the conversion is calculated and returned. -Note that this is the length and not size, i.e. the returned value does -{\bf not} include the trailing NUL. But when the function is called with a -non-{\tt NULL} {\it outputBuf}, the {\it outputSize} parameter should be the size of the buffer -and so it {\bf should} take into account the trailing NUL. +If \arg{out} is \NULL, only the length of the string which would result from +the conversion is calculated and returned. Note that this is the length and not +size, i.e. the returned value does \emph{not} include the trailing \NUL. But +when the function is called with a non-\NULL \arg{out} buffer, the \arg{outLen} +parameter should be one more to allow to properly \NUL-terminate the string. \wxheading{Parameters} -\docparam{outputBuf}{the output buffer, may be {\tt NULL} if the caller is only +\docparam{out}{The output buffer, may be \NULL if the caller is only interested in the length of the resulting string} -\docparam{psz}{the {\tt NUL}-terminated input string, cannot be {\tt NULL}} +\docparam{in}{The \NUL-terminated input string, cannot be \NULL} -\docparam{outputSize}{the size of the output buffer (in wide characters, {\bf including} the -NUL) , ignored if {\it outputBuf} is {\tt NULL}} +\docparam{outLen}{The length of the output buffer but \emph{including} +\NUL, ignored if \arg{out} is \NULL} \wxheading{Return value} -The length of the converted string (in wide characters, {\bf excluding} the NUL) +The length of the converted string \emph{excluding} the trailing {\NUL}. + \membersection{wxMBConv::WC2MB}\label{wxmbconvwc2mb} @@ -68,25 +93,54 @@ Converts from Unicode to multibyte encoding. The semantics of this function (including the return value meaning) is the same as for \helpref{MB2WC}{wxmbconvmb2wc}. -Notice that when the function is called with a non-{\tt NULL} buffer, the -{\it n} parameter should be the size of the buffer and so it {\bf should} take +Notice that when the function is called with a non-\NULL buffer, the +{\it n} parameter should be the size of the buffer and so it \emph{should} take into account the trailing NUL, which might take two or four bytes for some -encodings (UTF-16 and UTF-32). +encodings (UTF-16 and UTF-32) and not one. + \membersection{wxMBConv::cMB2WC}\label{wxmbconvcmb2wc} -\constfunc{const wxWCharBuffer}{cMB2WC}{\param{const char* }{psz}} +\constfunc{const wxWCharBuffer}{cMB2WC}{\param{const char *}{in}} + +\constfunc{const wxWCharBuffer}{cMB2WC}{\param{const char *}{in}, \param{size\_t }{inLen}, \param{size\_t }{*outLen}} + +Converts from multibyte encoding to Unicode by calling +\helpref{MB2WC}{wxmbconvmb2wc}, allocating a temporary wxWCharBuffer to hold +the result. + +The first overload takes a \NUL-terminated input string. The second one takes a +string of exactly the specified length and the string may include or not the +trailing {\NUL}s. If the string is not \NUL-terminated, a temporary +\NUL-terminated copy of it suitable for passing to \helpref{MB2WC}{wxmbconvmb2wc} +is made, so it is more efficient to ensure that the string is does have the +appropriate number of \NUL bytes (which is usually $1$ but may be $2$ or $4$ +for UTF-16 or UTF-32), especially for long strings. + +If \arg{outLen} is not-\NULL, it receives the length of the converted +string. -Converts from multibyte encoding to Unicode by calling MB2WC, -allocating a temporary wxWCharBuffer to hold the result. \membersection{wxMBConv::cWC2MB}\label{wxmbconvcwc2mb} -\constfunc{const wxCharBuffer}{cWC2MB}{\param{const wchar\_t* }{psz}} +\constfunc{const wxCharBuffer}{cWC2MB}{\param{const wchar\_t* }{in}} + +\constfunc{const wxCharBuffer}{cWC2MB}{\param{const wchar\_t* }{in}, \param{size\_t }{inLen}, \param{size\_t }{*outLen}} Converts from Unicode to multibyte encoding by calling WC2MB, allocating a temporary wxCharBuffer to hold the result. +The second overload of this function allows to convert a string of the given +length \arg{inLen}, whether it is \NUL-terminated or not (for wide character +strings, unlike for the multibyte ones, a single \NUL is always enough). +But notice that just as with \helpref{cMB2WC}{wxmbconvmb2wc}, it is more +efficient to pass an already terminated string to this function as otherwise a +copy is made internally. + +If \arg{outLen} is not-\NULL, it receives the length of the converted +string. + + \membersection{wxMBConv::cMB2WX}\label{wxmbconvcmb2wx} \constfunc{const char*}{cMB2WX}{\param{const char* }{psz}} @@ -99,6 +153,7 @@ it returns the parameter unaltered. If wxChar is wchar\_t, it returns the result in a wxWCharBuffer. The macro wxMB2WXbuf is defined as the correct return type (without const). + \membersection{wxMBConv::cWX2MB}\label{wxmbconvcwx2mb} \constfunc{const char*}{cWX2MB}{\param{const wxChar* }{psz}} @@ -110,6 +165,7 @@ it returns the parameter unaltered. If wxChar is wchar\_t, it returns the result in a wxCharBuffer. The macro wxWX2MBbuf is defined as the correct return type (without const). + \membersection{wxMBConv::cWC2WX}\label{wxmbconvcwc2wx} \constfunc{const wchar\_t*}{cWC2WX}{\param{const wchar\_t* }{psz}} @@ -121,6 +177,7 @@ it returns the parameter unaltered. If wxChar is char, it returns the result in a wxCharBuffer. The macro wxWC2WXbuf is defined as the correct return type (without const). + \membersection{wxMBConv::cWX2WC}\label{wxmbconvcwx2wc} \constfunc{const wchar\_t*}{cWX2WC}{\param{const wxChar* }{psz}} diff --git a/include/wx/strconv.h b/include/wx/strconv.h index d9e89d5e3c..ae701a15e5 100644 --- a/include/wx/strconv.h +++ b/include/wx/strconv.h @@ -35,23 +35,38 @@ class WXDLLIMPEXP_BASE wxMBConv { public: - // the actual conversion takes place here + // The functions doing actual conversion. On success, the return value is + // the length (i.e. the number of characters, not bytes, and not counting + // the trailing L'\0') of the converted string. On failure, (size_t)-1 is + // returned. In the special case when outputBuf is NULL the return value is + // the same one but nothing is written to the buffer. // - // note that outputSize is the size of the output buffer, not the length of input - // (the latter is always supposed to be NUL-terminated) - virtual size_t MB2WC(wchar_t *outputBuf, const char *psz, size_t outputSize) const = 0; - virtual size_t WC2MB(char *outputBuf, const wchar_t *psz, size_t outputSize) const = 0; + // Note that outLen is the length of the output buffer, not the length of + // the input (which is always supposed to be terminated by one or more + // NULs, as appropriate for the encoding)! + virtual size_t MB2WC(wchar_t *out, const char *in, size_t outLen) const = 0; + virtual size_t WC2MB(char *out, const wchar_t *in, size_t outLen) const = 0; // MB <-> WC - const wxWCharBuffer cMB2WC(const char *psz) const; - const wxCharBuffer cWC2MB(const wchar_t *psz) const; + const wxWCharBuffer cMB2WC(const char *in) const; + const wxCharBuffer cWC2MB(const wchar_t *in) const; - // MB <-> WC for strings with embedded null characters + // Functions converting strings which may contain embedded NULs and don't + // have to be NUL-terminated. // - // pszLen length of the input string - // pOutSize gets the final size of the converted string - const wxWCharBuffer cMB2WC(const char *psz, size_t pszLen, size_t* pOutSize) const; - const wxCharBuffer cWC2MB(const wchar_t *psz, size_t pszLen, size_t* pOutSize) const; + // inLen is the length of the buffer including trailing NUL if any: if the + // last 4 bytes of the buffer are all NULs, these functions are more + // efficient as they avoid copying the string, but otherwise a copy is made + // internally which could be quite bad for (very) long strings. + // + // outLen receives, if not NULL, the length of the converted string or 0 if + // the conversion failed (returning 0 and not -1 in this case makes it + // difficult to distinguish between failed conversion and empty input but + // this is done for backwards compatibility) + const wxWCharBuffer + cMB2WC(const char *in, size_t inLen, size_t *outLen) const; + const wxCharBuffer + cWC2MB(const wchar_t *in, size_t inLen, size_t *outLen) const; // convenience functions for converting MB or WC to/from wxWin default #if wxUSE_UNICODE @@ -68,6 +83,17 @@ public: // virtual dtor for any base class virtual ~wxMBConv(); + +private: + // this function must return the multibyte representation of L'\0' + // + // on error, nulLen should be set to -1 + virtual const char *GetMBNul(size_t *nulLen) const + { + *nulLen = 1; + + return ""; + } }; // ---------------------------------------------------------------------------- @@ -97,15 +123,28 @@ public: wxConvBrokenFileNames(const wxChar *charset); virtual ~wxConvBrokenFileNames() { delete m_conv; } - virtual size_t MB2WC(wchar_t *outputBuf, const char *psz, size_t outputSize) const; - virtual size_t WC2MB(char *outputBuf, const wchar_t *psz, size_t outputSize) const; + virtual size_t MB2WC(wchar_t *out, const char *in, size_t outLen) const + { + return m_conv->MB2WC(out, in, outLen); + } + + virtual size_t WC2MB(char *out, const wchar_t *in, size_t outLen) const + { + return m_conv->WC2MB(out, in, outLen); + } private: + virtual wxCharBuffer GetMBNul(size_t *nulLen) const + { + return m_conv->GetMBNul(nulLen); + } + + // the conversion object we forward to wxMBConv *m_conv; }; -#endif +#endif // __UNIX__ // ---------------------------------------------------------------------------- // wxMBConvUTF7 (for conversion using UTF7 encoding) @@ -125,7 +164,7 @@ public: class WXDLLIMPEXP_BASE wxMBConvUTF8 : public wxMBConv { public: - enum { + enum { MAP_INVALID_UTF8_NOT = 0, MAP_INVALID_UTF8_TO_PUA = 1, MAP_INVALID_UTF8_TO_OCTAL = 2 @@ -134,16 +173,30 @@ public: wxMBConvUTF8(int options = MAP_INVALID_UTF8_NOT) : m_options(options) { } virtual size_t MB2WC(wchar_t *outputBuf, const char *psz, size_t outputSize) const; virtual size_t WC2MB(char *outputBuf, const wchar_t *psz, size_t outputSize) const; - + private: int m_options; }; +// ---------------------------------------------------------------------------- +// wxMBConvUTF16Base: for both LE and BE variants +// ---------------------------------------------------------------------------- + +class WXDLLIMPEXP_BASE wxMBConvUTF16Base : public wxMBConv +{ +private: + virtual const char *GetMBNul(size_t *nulLen) const + { + *nulLen = 2; + return "\0"; + } +}; + // ---------------------------------------------------------------------------- // wxMBConvUTF16LE (for conversion using UTF16 Little Endian encoding) // ---------------------------------------------------------------------------- -class WXDLLIMPEXP_BASE wxMBConvUTF16LE : public wxMBConv +class WXDLLIMPEXP_BASE wxMBConvUTF16LE : public wxMBConvUTF16Base { public: virtual size_t MB2WC(wchar_t *outputBuf, const char *psz, size_t outputSize) const; @@ -154,18 +207,32 @@ public: // wxMBConvUTF16BE (for conversion using UTF16 Big Endian encoding) // ---------------------------------------------------------------------------- -class WXDLLIMPEXP_BASE wxMBConvUTF16BE : public wxMBConv +class WXDLLIMPEXP_BASE wxMBConvUTF16BE : public wxMBConvUTF16Base { public: virtual size_t MB2WC(wchar_t *outputBuf, const char *psz, size_t outputSize) const; virtual size_t WC2MB(char *outputBuf, const wchar_t *psz, size_t outputSize) const; }; +// ---------------------------------------------------------------------------- +// wxMBConvUTF32Base: base class for both LE and BE variants +// ---------------------------------------------------------------------------- + +class WXDLLIMPEXP_BASE wxMBConvUTF32Base : public wxMBConv +{ +private: + virtual const char *GetMBNul(size_t *nulLen) const + { + *nulLen = 4; + return "\0\0\0"; + } +}; + // ---------------------------------------------------------------------------- // wxMBConvUTF32LE (for conversion using UTF32 Little Endian encoding) // ---------------------------------------------------------------------------- -class WXDLLIMPEXP_BASE wxMBConvUTF32LE : public wxMBConv +class WXDLLIMPEXP_BASE wxMBConvUTF32LE : public wxMBConvUTF32Base { public: virtual size_t MB2WC(wchar_t *outputBuf, const char *psz, size_t outputSize) const; @@ -176,7 +243,7 @@ public: // wxMBConvUTF32BE (for conversion using UTF32 Big Endian encoding) // ---------------------------------------------------------------------------- -class WXDLLIMPEXP_BASE wxMBConvUTF32BE : public wxMBConv +class WXDLLIMPEXP_BASE wxMBConvUTF32BE : public wxMBConvUTF32Base { public: virtual size_t MB2WC(wchar_t *outputBuf, const char *psz, size_t outputSize) const; @@ -221,6 +288,8 @@ private: // charset string void SetName(const wxChar *charset); + virtual const char *GetMBNul(size_t *nulLen) const; + // note that we can't use wxString here because of compilation // dependencies: we're included from wx/string.h diff --git a/src/common/strconv.cpp b/src/common/strconv.cpp index 9d8481349f..dfc8a40dac 100644 --- a/src/common/strconv.cpp +++ b/src/common/strconv.cpp @@ -187,129 +187,163 @@ const wxCharBuffer wxMBConv::cWC2MB(const wchar_t *pwz) const return buf; } -const wxWCharBuffer wxMBConv::cMB2WC(const char *szString, size_t nStringLen, size_t* pOutSize) const +const wxWCharBuffer +wxMBConv::cMB2WC(const char *in, size_t inLen, size_t *outLen) const { - wxASSERT(pOutSize != NULL); - - const char* szEnd = szString + nStringLen + 1; - const char* szPos = szString; - const char* szStart = szPos; - - size_t nActualLength = 0; - size_t nCurrentSize = nStringLen; //try normal size first (should never resize?) - - wxWCharBuffer theBuffer(nCurrentSize); + // the currently accumulated wide characters + wxWCharBuffer wbuf; + + // the current length of wbuf + size_t lenBuf = 0; + + // we need to know the representation of L'\0' for this conversion + size_t nulLen; + const char * const nul = GetMBNul(&nulLen); + if ( nulLen == (size_t)-1 || nulLen == 0 ) + return wxWCharBuffer(); + + // make a copy of the input string unless it is already properly + // NUL-terminated + wxCharBuffer bufTmp; + + // now we can compute the input size if we were not given it: notice that + // in this case the string must be properly NUL-terminated, of course, as + // otherwise we have no way of knowing how long it is + if ( inLen == (size_t)-1 ) + { + // not the most efficient algorithm but it shouldn't matter as normally + // there are not many NULs in the string and so normally memcmp() + // should stop on the first character + for ( const char *p = in; ; p++ ) + { + if ( memcmp(p, nul, nulLen) == 0 ) + break; + } - //Convert the string until the length() is reached, continuing the - //loop every time a null character is reached - while(szPos != szEnd) + inLen = p - in + nulLen; + } + else // we already have the size { - wxASSERT(szPos < szEnd); //something is _really_ screwed up if this rings true + // check if it's not already NUL-terminated too to avoid the copy + if ( inLen < nulLen || memcmp(in + inLen - nulLen, nul, nulLen) != 0 ) + { + // make a copy in order to properly NUL-terminate the string + bufTmp = wxCharBuffer(inLen + nulLen - 1 /* 1 will be added */); + memcpy(bufTmp.data(), in, inLen); + memcpy(bufTmp.data() + inLen, nul, nulLen); + } + } - //Get the length of the current (sub)string - size_t nLen = MB2WC(NULL, szPos, 0); + if ( bufTmp ) + in = bufTmp; - //Invalid conversion? - if( nLen == (size_t)-1 ) + for ( const char * const inEnd = in + inLen;; ) + { + // try to convert the current chunk if anything left + size_t lenChunk = in < inEnd ? MB2WC(NULL, in, 0) : 0; + if ( lenChunk == 0 ) { - *pOutSize = 0; - theBuffer.data()[0u] = wxT('\0'); - return theBuffer; + // nothing left in the input string, conversion succeeded + if ( outLen ) + { + // we shouldn't include the last NUL in the result length + *outLen = lenBuf ? lenBuf - 1 : 0; + } + + return wbuf; } + if ( lenChunk == (size_t)-1 ) + break; - //Increase the actual length (+1 for current null character) - nActualLength += nLen + 1; + const size_t lenBufNew = lenBuf + lenChunk; + if ( !wbuf.extend(lenBufNew) ) + break; - //if buffer too big, realloc the buffer - if (nActualLength > (nCurrentSize+1)) - { - wxWCharBuffer theNewBuffer(nCurrentSize << 1); - memcpy(theNewBuffer.data(), theBuffer.data(), nCurrentSize * sizeof(wchar_t)); - theBuffer = theNewBuffer; - nCurrentSize <<= 1; - } + lenChunk = MB2WC(wbuf.data() + lenBuf, in, lenChunk + 1 /* for NUL */); + if ( lenChunk == (size_t)-1 ) + break; - //Convert the current (sub)string - if ( MB2WC(&theBuffer.data()[szPos - szStart], szPos, nLen + 1) == (size_t)-1 ) - { - *pOutSize = 0; - theBuffer.data()[0u] = wxT('\0'); - return theBuffer; - } + // +! for the embedded NUL (if something follows) + lenBuf = lenBufNew + 1; + + // advance the input pointer past the end of this chunk + while ( memcmp(in, nul, nulLen) != 0 ) + in++; - //Increment to next (sub)string - //Note that we have to use strlen instead of nLen here - //because XX2XX gives us the size of the output buffer, - //which is not necessarily the length of the string - szPos += strlen(szPos) + 1; + in += nulLen; // skipping over its terminator as well } - //success - return actual length and the buffer - *pOutSize = nActualLength; - return theBuffer; + // conversion failed + if ( outLen ) + *outLen = 0; + + return wxWCharBuffer(); } -const wxCharBuffer wxMBConv::cWC2MB(const wchar_t *szString, size_t nStringLen, size_t* pOutSize) const +const wxCharBuffer +wxMBConv::cWC2MB(const wchar_t *in, size_t inLen, size_t *outLen) const { - wxASSERT(pOutSize != NULL); - - const wchar_t* szEnd = szString + nStringLen + 1; - const wchar_t* szPos = szString; - const wchar_t* szStart = szPos; - - size_t nActualLength = 0; - size_t nCurrentSize = nStringLen << 2; //try * 4 first + // the currently accumulated multibyte characters + wxCharBuffer buf; - wxCharBuffer theBuffer(nCurrentSize); + // the current length of buf + size_t lenBuf = 0; - //Convert the string until the length() is reached, continuing the - //loop every time a null character is reached - while(szPos != szEnd) + // 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 ( inLen == (size_t)-1 ) { - wxASSERT(szPos < szEnd); //something is _really_ screwed up if this rings true + inLen = wxWcslen(in) + 1; + } + else if ( inLen != 0 && in[inLen - 1] != L'\0' ) + { + // make a copy in order to properly NUL-terminate the string + bufTmp = wxWCharBuffer(inLen); + memcpy(bufTmp.data(), in, inLen*sizeof(wchar_t)); + } - //Get the length of the current (sub)string - size_t nLen = WC2MB(NULL, szPos, 0); + if ( bufTmp ) + in = bufTmp; - //Invalid conversion? - if( nLen == (size_t)-1 ) + for ( const wchar_t * const inEnd = in + inLen;; ) + { + // try to convert the current chunk, if anything left + size_t lenChunk = in < inEnd ? WC2MB(NULL, in, 0) : 0; + if ( lenChunk == 0 ) { - *pOutSize = 0; - theBuffer.data()[0u] = wxT('\0'); - return theBuffer; + // nothing left in the input string, conversion succeeded + if ( outLen ) + *outLen = lenBuf ? lenBuf - 1 : lenBuf; + + return buf; } - //Increase the actual length (+1 for current null character) - nActualLength += nLen + 1; + if ( lenChunk == (size_t)-1 ) + break; - //if buffer too big, realloc the buffer - if (nActualLength > (nCurrentSize+1)) - { - wxCharBuffer theNewBuffer(nCurrentSize << 1); - memcpy(theNewBuffer.data(), theBuffer.data(), nCurrentSize); - theBuffer = theNewBuffer; - nCurrentSize <<= 1; - } + const size_t lenBufNew = lenBuf + lenChunk; + if ( !buf.extend(lenBufNew) ) + break; - //Convert the current (sub)string - if(WC2MB(&theBuffer.data()[szPos - szStart], szPos, nLen + 1) == (size_t)-1 ) - { - *pOutSize = 0; - theBuffer.data()[0u] = wxT('\0'); - return theBuffer; - } + lenChunk = WC2MB(buf.data() + lenBuf, in, lenChunk + 1 /* for NUL */); + if ( lenChunk == (size_t)-1 ) + break; - //Increment to next (sub)string - //Note that we have to use wxWcslen instead of nLen here - //because XX2XX gives us the size of the output buffer, - //which is not necessarily the length of the string - szPos += wxWcslen(szPos) + 1; + // chunk successfully converted, go to the next one + in += wxWcslen(in) + 1 /* skip NUL too */; + lenBuf = lenBufNew + 1; } - //success - return actual length and the buffer - *pOutSize = nActualLength; - return theBuffer; + // conversion failed + if ( outLen ) + *outLen = 0; + + return wxCharBuffer(); } // ---------------------------------------------------------------------------- @@ -326,12 +360,12 @@ size_t wxMBConvLibc::WC2MB(char *buf, const wchar_t *psz, size_t n) const return wxWC2MB(buf, psz, n); } -#ifdef __UNIX__ - // ---------------------------------------------------------------------------- // wxConvBrokenFileNames // ---------------------------------------------------------------------------- +#ifdef __UNIX__ + wxConvBrokenFileNames::wxConvBrokenFileNames(const wxChar *charset) { if ( !charset || wxStricmp(charset, _T("UTF-8")) == 0 @@ -341,23 +375,7 @@ wxConvBrokenFileNames::wxConvBrokenFileNames(const wxChar *charset) m_conv = new wxCSConv(charset); } -size_t -wxConvBrokenFileNames::MB2WC(wchar_t *outputBuf, - const char *psz, - size_t outputSize) const -{ - return m_conv->MB2WC( outputBuf, psz, outputSize ); -} - -size_t -wxConvBrokenFileNames::WC2MB(char *outputBuf, - const wchar_t *psz, - size_t outputSize) const -{ - return m_conv->WC2MB( outputBuf, psz, outputSize ); -} - -#endif +#endif // __UNIX__ // ---------------------------------------------------------------------------- // UTF-7 @@ -509,8 +527,6 @@ static const unsigned char utf7encode[128] = size_t wxMBConvUTF7::WC2MB(char *buf, const wchar_t *psz, size_t n) const { - - size_t len = 0; while (*psz && ((!buf) || (len < n))) @@ -888,19 +904,21 @@ size_t wxMBConvUTF16swap::MB2WC(wchar_t *buf, const char *psz, size_t n) const // swap 16bit MB to 16bit String size_t wxMBConvUTF16swap::WC2MB(char *buf, const wchar_t *psz, size_t n) const { - size_t len=0; + size_t len = 0; - while (*psz && (!buf || len < n)) + while ( *psz && (!buf || len < n) ) { - if (buf) + if ( buf ) { *buf++ = ((char*)psz)[1]; *buf++ = ((char*)psz)[0]; } - len += sizeof(wxUint16); + len += 2; psz++; } - if (buf && len<=n-sizeof(wxUint16)) *(wxUint16*)buf=0; + + if ( buf && len < n ) + *buf = '\0'; return len; } @@ -1336,6 +1354,8 @@ protected: #endif private: + virtual const char *GetMBNul(size_t *nulLen) const; + // the name (for iconv_open()) of a wide char charset -- if none is // available on this machine, it will remain NULL static wxString ms_wcCharsetName; @@ -1343,6 +1363,10 @@ private: // true if the wide char encoding we use (i.e. ms_wcCharsetName) has // different endian-ness than the native one static bool ms_wcNeedsSwap; + + // NUL representation + size_t m_nulLen; + char m_nulBuf[8]; }; // make the constructor available for unit testing @@ -1362,6 +1386,8 @@ bool wxMBConv_iconv::ms_wcNeedsSwap = false; wxMBConv_iconv::wxMBConv_iconv(const wxChar *name) { + m_nulLen = (size_t)-2; + // iconv operates with chars, not wxChars, but luckily it uses only ASCII // names for the charsets const wxCharBuffer cname(wxString(name).ToAscii()); @@ -1618,6 +1644,27 @@ size_t wxMBConv_iconv::WC2MB(char *buf, const wchar_t *psz, size_t n) const return res; } +const char *wxMBConv_iconv::GetMBNul(size_t *nulLen) const +{ + if ( m_nulLen == (size_t)-2 ) + { + wxMBConv_iconv * const self = wxConstCast(this, wxMBConv_iconv); + +#if wxUSE_THREADS + // NB: explained in MB2WC + wxMutexLocker lock(self->m_iconvMutex); +#endif + + size_t inLen = 1, + outLen = WXSIZEOF(m_nulBuf); + self->m_nulLen = iconv(w2m, ICONV_CHAR_CAST(L""), &inLen, + &self->m_nulBuf, &outLen); + } + + *nulLen = m_nulLen; + return m_nulBuf; +} + #endif // HAVE_ICONV @@ -1639,19 +1686,22 @@ public: wxMBConv_win32() { m_CodePage = CP_ACP; + m_nulLen = (size_t)-2; } #if wxUSE_FONTMAP wxMBConv_win32(const wxChar* name) { m_CodePage = wxCharsetToCodepage(name); + m_nulLen = (size_t)-2; } wxMBConv_win32(wxFontEncoding encoding) { m_CodePage = wxEncodingToCodepage(encoding); + m_nulLen = (size_t)-2; } -#endif +#endif // wxUSE_FONTMAP size_t MB2WC(wchar_t *buf, const char *psz, size_t n) const { @@ -1876,7 +1926,35 @@ private: #endif } + virtual const char *GetMBNul(size_t *nulLen) const + { + if ( m_nulLen == (size_t)-2 ) + { + wxMBConv_win32 * const self = wxConstCast(this, wxMBConv_win32); + + self->m_nulLen = ::WideCharToMultiByte + ( + m_CodePage, // code page + 0, // no flags + L"", // input string + 1, // translate just NUL + self->m_nulBuf, // output buffer + WXSIZEOF(m_nulBuf), // and its size + NULL, // "replacement" char + NULL // [out] was it used? + ); + + if ( m_nulLen == 0 ) + self->m_nulLen = (size_t)-1; + } + + *nulLen = m_nulLen; + return m_nulBuf; + } + long m_CodePage; + size_t m_nulLen; + char m_nulBuf[8]; }; #endif // wxHAVE_WIN32_MB2WC @@ -2516,6 +2594,27 @@ public: wxFontEncoding m_enc; wxEncodingConverter m2w, w2m; +private: + virtual const char *GetMBNul(size_t *nulLen) const + { + switch ( m_enc ) + { + case wxFONTENCODING_UTF16BE: + case wxFONTENCODING_UTF16LE: + *nulLen = 2; + return "\0"; + + case wxFONTENCODING_UTF32BE: + case wxFONTENCODING_UTF32LE: + *nulLen = 4; + return "\0\0\0"; + + default: + *nulLen = 1; + return ""; + } + } + // were we initialized successfully? bool m_ok; @@ -2908,6 +3007,20 @@ size_t wxCSConv::WC2MB(char *buf, const wchar_t *psz, size_t n) const return len; } +const char *wxCSConv::GetMBNul(size_t *nulLen) const +{ + CreateConvIfNeeded(); + + if ( m_convReal ) + { + // cast needed just to call private function of m_convReal + return ((wxCSConv *)m_convReal)->GetMBNul(nulLen); + } + + *nulLen = 1; + return ""; +} + // ---------------------------------------------------------------------------- // globals // ---------------------------------------------------------------------------- diff --git a/src/common/string.cpp b/src/common/string.cpp index 5bb461cb17..4e128a124f 100644 --- a/src/common/string.cpp +++ b/src/common/string.cpp @@ -1008,103 +1008,60 @@ int STRINGCLASS::compare(size_t nStart, size_t nLen, // from multibyte string wxString::wxString(const char *psz, wxMBConv& conv, size_t nLength) { - // if nLength != npos, then we have to make a NULL-terminated copy - // of first nLength bytes of psz first because the input buffer to MB2WC - // must always be NULL-terminated: - wxCharBuffer inBuf((const char *)NULL); - if (nLength != npos) - { - wxASSERT( psz != NULL ); - wxCharBuffer tmp(nLength); - memcpy(tmp.data(), psz, nLength); - tmp.data()[nLength] = '\0'; - inBuf = tmp; - psz = inBuf.data(); - } - - // first get the size of the buffer we need - size_t nLen; - if ( psz ) - { - // calculate the needed size ourselves or use the provided one - if (nLength == npos) - nLen = strlen(psz); - else - nLen = nLength; - } - else - { - // nothing to convert - nLen = 0; - } - - // anything to do? - if ( (nLen != 0) && (nLen != (size_t)-1) ) + if ( psz && nLength != 0 ) { - //Convert string - size_t nRealSize; - wxWCharBuffer theBuffer = conv.cMB2WC(psz, nLen, &nRealSize); + if ( nLength == npos ) + { + nLength = (size_t)-1; + } + else if ( nLength == length() ) + { + // this is important to avoid copying the string in cMB2WC: we're + // already NUL-terminated so we can pass this NUL with the data + nLength++; + } + + size_t nLenWide; + wxWCharBuffer wbuf = conv.cMB2WC(psz, nLength, &nLenWide); - //Copy - if (nRealSize) - assign( theBuffer.data() , nRealSize - 1 ); + if ( nLenWide ) + assign(wbuf, nLenWide); } } //Convert wxString in Unicode mode to a multi-byte string const wxCharBuffer wxString::mb_str(wxMBConv& conv) const { - size_t dwOutSize; - return conv.cWC2MB(c_str(), length(), &dwOutSize); + return conv.cWC2MB(c_str(), length() + 1 /* size, not length */, NULL); } #else // ANSI #if wxUSE_WCHAR_T + // from wide string wxString::wxString(const wchar_t *pwz, wxMBConv& conv, size_t nLength) { - // if nLength != npos, then we have to make a NULL-terminated copy - // of first nLength chars of psz first because the input buffer to WC2MB - // must always be NULL-terminated: - wxWCharBuffer inBuf((const wchar_t *)NULL); - if (nLength != npos) - { - wxASSERT( pwz != NULL ); - wxWCharBuffer tmp(nLength); - memcpy(tmp.data(), pwz, nLength * sizeof(wchar_t)); - tmp.data()[nLength] = '\0'; - inBuf = tmp; - pwz = inBuf.data(); - } - - // first get the size of the buffer we need - size_t nLen; - if ( pwz ) - { - // calculate the needed size ourselves or use the provided one - if (nLength == npos) - nLen = wxWcslen(pwz); - else - nLen = nLength; - } - else - { - // nothing to convert - nLen = 0; - } - // anything to do? - if ( (nLen != 0) && (nLen != (size_t)-1) ) + if ( pwz && nLength != 0 ) { - //Convert string - size_t nRealSize; - wxCharBuffer theBuffer = conv.cWC2MB(pwz, nLen, &nRealSize); + if ( nLength == npos ) + { + nLength = (size_t)-1; + } + else if ( nLength == length() ) + { + // this is important to avoid copying the string in cMB2WC: we're + // already NUL-terminated so we can pass this NUL with the data + nLength++; + } + + size_t nLenMB; + wxCharBuffer buf = conv.cWC2MB(pwz, nLength, &nLenMB); - //Copy - if (nRealSize) - assign( theBuffer.data() , nRealSize - 1 ); + if ( nLenMB ) + assign(buf, nLenMB); } } @@ -1112,8 +1069,7 @@ wxString::wxString(const wchar_t *pwz, wxMBConv& conv, size_t nLength) //mode is not enabled and wxUSE_WCHAR_T is enabled const wxWCharBuffer wxString::wc_str(wxMBConv& conv) const { - size_t dwOutSize; - return conv.cMB2WC(c_str(), length(), &dwOutSize); + return conv.cMB2WC(c_str(), length() + 1 /* size, not length */, NULL); } #endif // wxUSE_WCHAR_T -- 2.45.2