X-Git-Url: https://git.saurik.com/wxWidgets.git/blobdiff_plain/55d99c7a77789ff4904bf96eddca3715eb5af9b9..403e664e5b3749060df8f526172a01e488cde6f3:/src/common/textfile.cpp diff --git a/src/common/textfile.cpp b/src/common/textfile.cpp index 219b1b26f7..f61220dc06 100644 --- a/src/common/textfile.cpp +++ b/src/common/textfile.cpp @@ -13,10 +13,6 @@ // headers // ============================================================================ -#ifdef __GNUG__ - #pragma implementation "textfile.h" -#endif - #include "wx/wxprec.h" #ifdef __BORLANDC__ @@ -39,6 +35,7 @@ #include "wx/textfile.h" #include "wx/filename.h" +#include "wx/buffer.h" // ============================================================================ // wxTextFile class implementation @@ -89,98 +86,138 @@ bool wxTextFile::OnClose() } -bool wxTextFile::OnRead(wxMBConv& conv) +bool wxTextFile::OnRead(const wxMBConv& conv) { // file should be opened and we must be in it's beginning wxASSERT( m_file.IsOpened() && m_file.Tell() == 0 ); - char *strBuf, *strPtr, *strEnd; - char ch, chLast = '\0'; - char buf[1024]; - int n, nRead; - - strPtr = strBuf = new char[1024]; - strEnd = strBuf + 1024; - - do + // read the entire file in memory: this is not the most efficient thing to + // do but there is no good way to avoid it in Unicode build because if we + // read the file block by block we can't convert each block to Unicode + // separately (the last multibyte char in the block might be only partially + // read and so the conversion would fail) and, as the file contents is kept + // in memory by wxTextFile anyhow, it shouldn't be a big problem to read + // the file entirely + const size_t bufSize = (size_t)(m_file.Length() + 4 /* for trailing NULs */ ); + size_t bufPos = 0; + wxCharBuffer buf(bufSize - 1 /* it adds 1 internally */); + + char block[1024]; + for ( bool eof = false; !eof; ) { - nRead = m_file.Read(buf, WXSIZEOF(buf)); - if ( nRead == wxInvalidOffset ) + // try to read up to the size of the entire block + ssize_t nRead = m_file.Read(block, WXSIZEOF(block)); + + if ( nRead == wxInvalidOffset ) { // read error (error message already given in wxFile::Read) - delete[] strBuf; - return FALSE; + return false; } - for (n = 0; n < nRead; n++) + if ( nRead == 0 ) + break; + + // this shouldn't happen but don't overwrite the buffer if it does + wxCHECK_MSG( bufPos + nRead <= bufSize, false, + _T("read more than file length?") ); + + // append to the buffer + memcpy(buf.data() + bufPos, block, nRead); + bufPos += nRead; + } + + const wxString str(buf, conv, bufPos); + + // this doesn't risk to happen in ANSI build +#if wxUSE_UNICODE + if ( bufSize > 4 && str.empty() ) + { + wxLogError(_("Failed to convert file contents to Unicode.")); + return false; + } +#endif // wxUSE_UNICODE + + free(buf.release()); // we don't need this memory any more + + + // now break the buffer in lines + + // last processed character, we need to know if it was a CR or not + wxChar chLast = '\0'; + + // the beginning of the current line, changes inside the loop + wxString::const_iterator lineStart = str.begin(); + const wxString::const_iterator end = str.end(); + for ( wxString::const_iterator p = lineStart; p != end; p++ ) + { + const wxChar ch = *p; + switch ( ch ) { - ch = buf[n]; - switch ( ch ) - { - case '\n': - // Dos/Unix line termination - *strPtr = '\0'; - AddLine(wxString(strBuf, conv), - chLast == '\r' ? wxTextFileType_Dos - : wxTextFileType_Unix); - strPtr = strBuf; - chLast = '\n'; - break; - - case '\r': - if ( chLast == '\r' ) + case '\n': + // could be a DOS or Unix EOL + if ( chLast == '\r' ) + { + if ( p - 1 >= lineStart ) { - // Mac empty line - AddLine(wxEmptyString, wxTextFileType_Mac); + AddLine(wxString(lineStart, p - 1), wxTextFileType_Dos); } else - chLast = '\r'; - break; - - default: - if ( chLast == '\r' ) { - // Mac line termination - *strPtr = '\0'; - AddLine(wxString(strBuf, conv), wxTextFileType_Mac); - chLast = ch; - strPtr = strBuf; - *(strPtr++) = ch; + // there were two line endings, so add an empty line: + AddLine(wxEmptyString, wxTextFileType_Dos); } - else + } + else // bare '\n', Unix style + { + AddLine(wxString(lineStart, p), wxTextFileType_Unix); + } + + lineStart = p + 1; + break; + + case '\r': + if ( chLast == '\r' ) + { + // Mac empty line + AddLine(wxEmptyString, wxTextFileType_Mac); + lineStart = p + 1; + } + //else: we don't know what this is yet -- could be a Mac EOL or + // start of DOS EOL so wait for next char + break; + + default: + if ( chLast == '\r' ) + { + // Mac line termination + if ( p - 1 >= lineStart ) { - // add to the current line - *(strPtr++) = ch; - if ( strPtr == strEnd ) - { - // we must allocate more memory - size_t size = strEnd - strBuf; - char *newBuf = new char[size + 1024]; - memcpy(newBuf, strBuf, size); - delete[] strBuf; - strBuf = newBuf; - strEnd = strBuf + size + 1024; - strPtr = strBuf + size; - } + AddLine(wxString(lineStart, p - 1), wxTextFileType_Mac); } - } + else + { + // there were two line endings, so add an empty line: + AddLine(wxEmptyString, wxTextFileType_Mac); + } + lineStart = p; + } } - } while ( nRead == WXSIZEOF(buf) ); + + chLast = ch; + } // anything in the last line? - if ( strPtr != strBuf ) + if ( lineStart != end ) { - *strPtr = '\0'; - AddLine(wxString(strBuf, conv), - wxTextFileType_None); // no line terminator + // add unterminated last line + AddLine(wxString(lineStart, end), wxTextFileType_None); } - delete[] strBuf; - return TRUE; + return true; } -bool wxTextFile::OnWrite(wxTextFileType typeNew, wxMBConv& conv) +bool wxTextFile::OnWrite(wxTextFileType typeNew, const wxMBConv& conv) { wxFileName fn = m_strBufferName; @@ -194,7 +231,7 @@ bool wxTextFile::OnWrite(wxTextFileType typeNew, wxMBConv& conv) if ( !fileTmp.IsOpened() ) { wxLogError(_("can't write buffer '%s' to disk."), m_strBufferName.c_str()); - return FALSE; + return false; } size_t nCount = GetLineCount(); @@ -210,4 +247,3 @@ bool wxTextFile::OnWrite(wxTextFileType typeNew, wxMBConv& conv) } #endif // wxUSE_TEXTFILE -