CStdioFile-derived class for multibyte and Unicode reading and writing

• Download demo project - 21.7 KB
• Download source files - 18.4 KB

Introduction

This is a class derived from CStdioFile which transparently handles the reading and writing of Unicode text files as well as ordinary multibyte text files.

The code compiles as both multibyte and Unicode. In Unicode, multibyte files will be read and their content converted to Unicode using the current code page. In multibyte compilations, Unicode files will be read and converted to multibyte text.

The identification of a Unicode text file depends entirely on the presence of the Unicode byte order mark (0xFEFF). Its absence is not an absolute guarantee that a file is not Unicode, but it's the only method I use here. Feel free to suggest improvements.

By default, the class writes multibyte files, but can optionally write Unicode.

Background

The ability to transparently handle both multibyte and Unicode seems to be such a fundamental requirement, that I was sure that there would already be something similar on offer, and yet nothing turned up. Did I miss something?

I needed it for a translation tool I wrote, and knocked together an implementation that was good enough for my needs. This is little more than a cleaned up version of that, so expect bugs and all manner of deficiencies. I've tested the demo app though with the basic combinations -- Unicode files in a multibyte compilation, Unicode-Unicode, Multibyte-Unicode, and Multibyte-Multibyte, and they all seem to work.

Using the code

The use of the class is pretty simple. It overrides three functions of CStdioFile: Open(), ReadString() and WriteString(). To write a Unicode file, add the flag CStdioFileEx::modeWriteUnicode to the flags when calling the Open() function.

In other respects, usage is identical to CStdioFile.

To find out if a file you have opened is Unicode, you can call IsFileUnicodeText().

To get the number of characters in the file, you can call GetCharCount(). This is unreliable for multibyte/UTF-8, however.

An example of writing in Unicode:

// Test writing
CStdioFileEx fileWriteUnicode;

if (fileWriteUnicode.Open(_T("c:\\testwrite_unicode.txt"), 
    CFile::modeCreate | CFile::modeWrite | CStdioFileEx::modeWriteUnicode))
{
    fileWriteUnicode.WriteString(_T("Unicode test file\n"));
    fileWriteUnicode.WriteString(_T("Writing data\n"));
    fileWriteUnicode.Close();
}

You can now also specify the code page for multibyte file reading or writing. Simply call SetCodePage() before a read to tell CStdioFileEx which code page the file is coded in, or before a write, to tell it which code page you want it written in. Specifying CP_UTF8 as the code page allows you to read or write UTF-8 files.

The demo app is a dialog which opens a file, tells you whether it's Unicode or not and how many characters it contains, and shows the first fifteen lines from it. In the last couple of iterations I've added the option to convert a Unicode file to multibyte, and a multibyte file to Unicode, and a combo to specify the code page when reading.

As of v1.6, there is no limitation on the length of the line that can be read in any mode (Multibyte/Unicode, Unicode/Multibyte, etc.).

I'd love to hear of people's experiences with it, as well as reports of bugs, problems, improvements, etc.

Oh, and if I've accidentally included something offensive in the demo dialog, let me know. My Arabic and Chinese are not all that good.

History

• v1.0 - Posted 14 May 2003
• v1.1 - 23 August 2003. Incorporated fixes from Dennis Jeryd
• v1.2 - 06 January 2005. Fixed garbage at end of file bug (Howard J Oh)
• v1.3 - 19 February 2005. Howard J Oh's fix mysteriously failed to make it into the last release. Improved the test program. Fixed miscellaneous bugs
  Very important: In this release, ANSI files written in ANSI are no longer written using WriteString. This means \n will no longer be "interpreted" as \r\n. What you write is what you get
• v1.4 - 26 February 2005. Fixed submission screw-up
• v1.5 - 18 November 2005. Code page can be specified for reading and writing (inc. UTF-8). Multibyte buffers properly calculated. Fix from Andy Goodwin
• v1.6 - 19 July 2007. Major rewrite: Maximum line length restriction removed; Use of strlen/lstrlen eliminated. Conversion functions always used to calculate required buffers; \r or \n characters no longer lost; BOM writing now optional; UTF-8 reading and writing works properly; systematic tests are now included with the demo project