What's the difference between printf("%s"), printf("%ls"), wprintf("%s"), and wprintf("%ls")?

Question

Consider this sample program:

#include <cstdio>
#include <cwchar>
#include <string>

int main()
{
    std::string narrowstr = "narrow";
    std::wstring widestr = L"wide";
    printf("1 %s \n", narrowstr.c_str());
    printf("2 %ls \n", widestr.c_str());
    wprintf(L"3 %s \n", narrowstr.c_str());
    wprintf(L"4 %ls \n", widestr.c_str());
   
   return 0;
}

The output of this is:

1 narrow 
2 wide

I'm wondering:

1. Why didn't 3 & 4 print?
2. What's the differences between 1 & 3 and 2 & 4?
3. Does it make any difference if narrowstr is in UTF8 and widestr is in UTF16?

Solution 1

You need to do:

wprintf(L"3 %hs \n", narrowstr.c_str());
wprintf(L"4 %s \n", widestr.c_str());

Why? Because for printf, %s says narrow-char-string. For wprintf, %ls says wide.

But, for wprintf, %s implies wide, %ls would mean wide itself. %hs would mean narrow (for both). For printf, %s, in this manner would simply mean %hs

On VC++/Windows, %S (capital S), would reverse the effect. Therfore for printf("%S") it would mean wide, and wprintf("%S") would mean narrow. This is useful for _tprintf.

Solution 2

Note that you're using C streams. C streams have a very special quality called "orientation". A stream is either unoriented, wide, or narrow. Orientation is decided by the first output made to any particular stream (see http://en.cppreference.com/w/cpp/io/c for a summary of C I/O streams)

In your case, stdout starts out unoriented, and by executing the first printf, you're setting it narrow. Once narrow, it's stuck narrow, and wprintf fails (check its return code!). The only way to change a C stream is to freopen it, which doesn't quite work with stdout. That's why 3 and 4 didn't print.

The differences between 1 and 3 is that 1 is a narrow output function which is using narrow string conversion specifier %s: it reads bytes from the char array and sends bytes into a byte stream. 3 is a wide output function with a narrow string conversion specifier %s: it first reads bytes from the char array and mbtowcs them into wchar_ts, then sends wchar_ts into a wide stream, which then wctombs them into bytes or multibyte sequences that are then pushed into the standard out with a write

Finally, if widestr is in utf16, you must be using Windows, and all bets are off; there is very little support for anything beyond ASCII on that platform. You may as well give in and use WinAPI (you can get by with standard C++11 for some Unicode things, and even do this C output, with magic words _setmode(_fileno(stdout), _O_U16TEXT);, that's been discussed enough times)