How can I make dynamic strings to work with UTF-8 in console?

Question

Most of answers and questions here on SO use to put L before any UTF-8 string. I found no explantion of what it is, in the source code, the constant is, according to my IDE, defined in winnt.h.

This is how I use it, without knowing what it is:

std::wcout<<L"\"Přetečení zásobníku\" is Stack overflow in Czech.";

Obviously, constant concatenation cannot be applied on variables:

void printUTF8(const char* str) {
  //Does not make the slightest bit of sense
  std::wcout<<L str; 
}

So what is it and how to add it to dynamic strings?

Answer 1

Re your actual question

” what is [the L prefix] and how to add it to dynamic strings?

This is very different from the title of the question at the time I’m writing this, namely “How can I make dynamic strings to work with UTF-8 in console?”

In short, UTF-8 is an encoding of Unicode where the basic encoding unit is 8 bits, commonly called a byte (more precisely it's an octet), while the L prefix forms a wide character or string literal, where the encoding unit typically is 16 or 32 bits – in Windows it’s 16 bits, as in original Unicode.

A wide character or string literal is based on the wchar_t type instead of char.

In Windows a wide string is encoded as UTF-16. The most common sixty thousand or so Unicode characters are represented with single wchar_t values, but some seldom used Chinese ideograms etc. require two successive wchar_t values, called a surrogate pair.

The use of 16 bit encoding unit in Windows was established around 1992. I am not sure when UTF-16 was adopted (as an extension of then UCS-2 encoding), it was just a bit later. So this was established long before C99 required that all characters of the wide character set should be representable with single wchar_t values. That requirement appears to have been a pure political maneuver, ensuring that no Windows C compiler could be formally conforming, a general ISO programming language standard that applied only to Unix-land. Unfortunately, since C++11 was based on C99 we now have that also in C++11, ensuring that no Windows C++ compiler can be fully conforming. Pure idiocy. If you ask me.

Errata, re deleted text above: according to Wikipedia’s article about it the wording about a single wchar_t being sufficient for any character in the “extended character set” was there already in C90. Which makes the incompatibility between Windows and the C and C++ standards the fault of Microsoft, not the fault of the C committee. It still appears to be political and fairly idiotic, but (enlightened) with others to blame than I maintained at first…

---

One way to work with wide dynamic strings is to use std::wstring, from the <string> header.

With Visual C++ you can use a wmain function instead of standard main, as an easy way to get wide command line arguments.

wmain is also supported by MinGW64 (IIRC) g++, although not yet by ordinary MinGW g++, as of g++ 4.8.something. It is however easy to implement in terms of the Windows API. Unless you require strict standard-conforming code that provides the special main function features such as ability to declare it with or without arguments, but hey, let's be practical about things.

---

Example that compiles fine with both Visual C++ 12.0 and g++ 4.8.2:

// Source encoding: UTF-8 with BOM.

#include <io.h>         // _setmode
#include <fcntl.h>      // _O_WTEXT

#include <iostream>     // std::wcout, std::endl
#include <string>       // std::wstring
using namespace std;

auto main()
    -> int
{
    _setmode( _fileno( stdin ), _O_WTEXT );
    _setmode( _fileno( stdout ), _O_WTEXT );

    wcout << L"Hi, what’s your name? ";
    wstring username;
    getline( wcin, username );
    wcout << L"Welcome to Windows C++, " << username << "!" << endl;
}

Note that with Windows ANSI source this won’t compile with g++ unless you specify the source encoding with the appropriate compiler option.

Answer 2

L is an indication to the C compiler that the string is composed of "wide characters". In Windows, these would be UTF-16 - each character that you put in the string is 16 bits, or two bytes, wide:

L"This is a wide string"

In contrast, a UTF-8 string is always a string composed of bytes. ASCII characters (A-Z 0-9 etc) are encoded the way they have always been - in the range 0x00 to 0x7F (or 0 to 127). International characters (like ř) are encoded using multiple bytes in the range 0x80 to 0xFF - there is a very good explanation on wikipedia. The advantage is that it can be represented using ordinary C strings.

"This is an ordinary string, but also a UTF-8 string"

"This is a C cedilla in UTF-8: \xc3\x87"

However, if you are typing these international characters in to actual code, your editor needs to know that you are typing in UTF-8 so it can encode the characters correctly - like the C cedilla above. Then the string will be passed correctly to your function.

In your case, your comment indicates that you are using UTF-16. In which case there are two other issues:

• The console will, by default, not output Unicode characters correctly. You need to change the font to a truetype font like Lucida Console

• You also need to change the output mode to a Unicode UTF-16 one. You can do this with:

  _setmode(_fileno(stdout), _O_U16TEXT);

Code example:

#include <iostream>
#include <io.h>
#include <fcntl.h>

int wmain(int argc, wchar_t* argv[])
{
    _setmode(_fileno(stdout), _O_U16TEXT);
    std::wcout << L"Přetečení zásobníku is Stack overflow in Czech." << std::endl;
}

Answer 3

L"" is a WIDE string. That is to say, it's a a wchar_t[1]. UTF-8 strings can't be wide, since they are multi-byte (variable length). VC++ is slightly wrong and made wide strings variable length, UTF-16 to be precise. But usually they're UTF-32.

The problem with multi-byte strings is that there are many different encodings, and UTF-8 is only one of them. Windows does not in fact natively support UTF-8 encodings. MessageBoxA() for instance can use any encoding but UTF-8. There's just one exception to that, which is MultiByteToWideChar(CP_UTF8, ...) which is what you'd need here.