X-Git-Url: https://git.saurik.com/wxWidgets.git/blobdiff_plain/c1981a2fa408700243400606f93679b2befe1ac2..fa3b08caf17a6310540d623b6ce508c201efb9a5:/src/common/textfile.cpp diff --git a/src/common/textfile.cpp b/src/common/textfile.cpp index f9cd34077d..ef27836e89 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 @@ -85,27 +86,48 @@ 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) ); + // 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 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, + bufPos = 0; + char block[1024]; + 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(); + } - static const size_t BUF_SIZE = 1024; -#if wxUSE_UNICODE - static const size_t NUL_SIZE = 4; -#else - static const size_t NUL_SIZE = 1; -#endif + // 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 ) + { + // we know the required length, so set the buffer size in advance + bufSize = fileLength; + if ( !buf.extend(bufSize - 1 /* it adds 1 internally */) ) + return false; - char buf[BUF_SIZE + NUL_SIZE]; - wxChar chLast = '\0'; - wxString str; + // 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") ); + } for ( ;; ) { - // leave space for trailing NUL - ssize_t nRead = m_file.Read(buf, BUF_SIZE); + ssize_t nRead = m_file.Read(block, WXSIZEOF(block)); if ( nRead == wxInvalidOffset ) { @@ -114,97 +136,130 @@ bool wxTextFile::OnRead(wxMBConv& conv) } if ( nRead == 0 ) - break; + { + // if no bytes have been read, presumably this is a valid-but-empty file + if ( bufPos == 0 ) + return true; - // save the number characters which we already processed during the - // last loop iteration - const size_t lenOld = str.length(); + // otherwise we've finished reading the file + break; + } -#if wxUSE_UNICODE - // we have to properly NUL-terminate the string for any encoding it may - // use -- 4 NULs should be enough for everyone (this is why we add 4 - // extra bytes to the buffer) - buf[nRead] = - buf[nRead + 1] = - buf[nRead + 2] = - buf[nRead + 3] = '\0'; - - // append to the remains of the last block, don't overwrite - wxString strbuf(buf, conv); - if ( strbuf.empty() ) + if ( seekable ) { - // conversion failed - return false; + // 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?") ); + } + else // !seekable + { + // for non-seekable files we have to allocate more memory on the go + if ( !buf.extend(bufPos + nRead - 1 /* it adds 1 internally */) ) + return false; } - str += strbuf; -#else // ANSI - wxUnusedVar(conv); - buf[nRead] = '\0'; - str += buf; -#endif // wxUSE_UNICODE/!wxUSE_UNICODE + // append to the buffer + memcpy(buf.data() + bufPos, block, nRead); + bufPos += nRead; + } + + if ( !seekable ) + { + bufSize = bufPos; + } + + const wxString str(buf, conv, bufPos); + + // 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 + + 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 + lenOld; p != end; p++ ) + // 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 ) { - const wxChar ch = *p; - switch ( ch ) - { - case '\n': - // could be a DOS or Unix EOL - if ( chLast == '\r' ) + 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; @@ -234,4 +289,3 @@ bool wxTextFile::OnWrite(wxTextFileType typeNew, wxMBConv& conv) } #endif // wxUSE_TEXTFILE -