X-Git-Url: https://git.saurik.com/wxWidgets.git/blobdiff_plain/86948c99a6f0fd177b09b3db0ef702739ec62a27..b53aea81d2e102224b452ef5bf7aee1132f37c6f:/src/common/textfile.cpp diff --git a/src/common/textfile.cpp b/src/common/textfile.cpp index b7ab9e06ca..79d837f631 100644 --- a/src/common/textfile.cpp +++ b/src/common/textfile.cpp @@ -35,6 +35,7 @@ #include "wx/textfile.h" #include "wx/filename.h" +#include "wx/buffer.h" // ============================================================================ // wxTextFile class implementation @@ -85,95 +86,205 @@ 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.GetKind() != wxFILE_KIND_DISK || m_file.Tell() == 0) ); - - char buf[1025]; - wxChar chLast = '\0'; - wxString str; + // file should be opened + wxASSERT_MSG( m_file.IsOpened(), _T("can't read closed file") ); + + // read the entire file in memory: this is not the most efficient thing to + // do it 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 + size_t bufSize = 0; + + // number of bytes to (try to) read from disk at once + static const size_t BLOCK_SIZE = 4096; + + wxCharBuffer buf; + + // first determine if the file is seekable or not and so whether we can + // determine its length in advance + wxFileOffset fileLength; + { + wxLogNull logNull; + fileLength = m_file.Length(); + } - for ( ;; ) + // some non-seekable files under /proc under Linux pretend that they're + // seekable but always return 0; others do return an error + const bool seekable = fileLength != wxInvalidOffset && fileLength != 0; + if ( seekable ) { - // leave space for trailing NUL - ssize_t nRead = m_file.Read(buf, WXSIZEOF(buf) - 1); + // we know the required length, so set the buffer size in advance + bufSize = fileLength; + if ( !buf.extend(bufSize) ) + return false; + + // if the file is seekable, also check that we're at its beginning + wxASSERT_MSG( m_file.Tell() == 0, _T("should be at start of file") ); - if ( nRead == wxInvalidOffset ) + char *dst = buf.data(); + for ( size_t nRemaining = bufSize; nRemaining > 0; ) { - // read error (error message already given in wxFile::Read) - return false; - } + size_t nToRead = BLOCK_SIZE; - if ( nRead == 0 ) - break; + // the file size could have changed, avoid overflowing the buffer + // even if it did + if ( nToRead > nRemaining ) + nToRead = nRemaining; + + ssize_t nRead = m_file.Read(dst, nToRead); + + if ( nRead == wxInvalidOffset ) + { + // read error (error message already given in wxFile::Read) + return false; + } - buf[nRead] = '\0'; + if ( nRead == 0 ) + { + // this file can't be empty because we checked for this above + // so this must be the end of file + break; + } - // append to the remains of the last block, don't overwrite - str += wxString(buf, conv); + dst += nRead; + nRemaining -= nRead; + } - // the beginning of the current line, changes inside the loop - const wxChar *lineStart = str.begin(); - const wxChar * const end = str.end(); - for ( const wxChar *p = lineStart; p != end; p++ ) + wxASSERT_MSG( dst - buf.data() == (wxFileOffset)bufSize, + _T("logic error") ); + } + else // file is not seekable + { + char block[BLOCK_SIZE]; + for ( ;; ) { - const wxChar ch = *p; - switch ( ch ) + ssize_t nRead = m_file.Read(block, WXSIZEOF(block)); + + if ( nRead == wxInvalidOffset ) { - case '\n': - // could be a DOS or Unix EOL - if ( chLast == '\r' ) + // read error (error message already given in wxFile::Read) + return false; + } + + if ( nRead == 0 ) + { + // if no bytes have been read, presumably this is a + // valid-but-empty file + if ( bufSize == 0 ) + return true; + + // otherwise we've finished reading the file + break; + } + + // extend the buffer for new data + if ( !buf.extend(bufSize + nRead) ) + return false; + + // and append it to the buffer + memcpy(buf.data() + bufSize, block, nRead); + bufSize += nRead; + } + } + + const wxString str(buf, conv, bufSize); + + // there's no risk of this happening in ANSI build +#if wxUSE_UNICODE + if ( bufSize > 4 && str.empty() ) + { + wxLogError(_("Failed to convert file \"%s\" to Unicode."), GetName()); + return false; + } +#endif // wxUSE_UNICODE + + // we don't need this memory any more + buf.reset(); + + + // 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 ) + { + case '\n': + // could be a DOS or Unix EOL + if ( chLast == '\r' ) + { + if ( p - 1 >= lineStart ) { AddLine(wxString(lineStart, p - 1), wxTextFileType_Dos); } - else // bare '\n', Unix style + else { - AddLine(wxString(lineStart, p), wxTextFileType_Unix); + // there were two line endings, so add an empty line: + AddLine(wxEmptyString, wxTextFileType_Dos); } - + } + 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; - break; - - case '\r': - if ( chLast == '\r' ) + } + //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 ) { - // Mac empty line - AddLine(wxEmptyString, wxTextFileType_Mac); - lineStart = p + 1; + AddLine(wxString(lineStart, p - 1), wxTextFileType_Mac); } - //else: we don't what this is yet -- could be a Mac EOL or - // start of DOS EOL so wait for next char - break; - - default: - if ( chLast == '\r' ) + else { - // Mac line termination - AddLine(wxString(lineStart, p - 1), wxTextFileType_Mac); - lineStart = p; + // there were two line endings, so add an empty line: + AddLine(wxEmptyString, wxTextFileType_Mac); } - } - - chLast = ch; + lineStart = p; + } } - // remove the part we already processed - str.erase(0, lineStart - str.begin()); + chLast = ch; } // anything in the last line? - if ( !str.empty() ) + if ( lineStart != end ) { - AddLine(str, wxTextFileType_None); // no line terminator + // add unterminated last line + AddLine(wxString(lineStart, end), wxTextFileType_None); } return true; } -bool wxTextFile::OnWrite(wxTextFileType typeNew, wxMBConv& conv) +bool wxTextFile::OnWrite(wxTextFileType typeNew, const wxMBConv& conv) { wxFileName fn = m_strBufferName; @@ -203,4 +314,3 @@ bool wxTextFile::OnWrite(wxTextFileType typeNew, wxMBConv& conv) } #endif // wxUSE_TEXTFILE -