Working with file path strings on Linux, what encoding to use?

Question

What encoding does Linux use for its file APIs? How should I work with path strings in C++, what class to use? I mean paths with non-ASCII characters. On Windows I use UTF-16 and std::wstring, on Mac - UTF-8 and my own UTF-8 string class. But unfortunately my class is not available on Linux, so what should I use?

Answer 1

Internally, Linux permits to use any byte sequence for file name, except for null byte 0 and forward slash '/' (which is used as directory separator).

Common convention to permit Unicode file names on Linux is to use UTF-8 encoding for file names. Easiest way to achieve that is to use good old std::string (not std::wstring which is suggested on Windows), however, you may need to write your own class which will validate that it is indeed valid UTF-8.

There are few examples of ready-to-use libraries that provide handling of UTF-8 strings:

• ICU (robust but very heavy).
• Glib::ustring (has implicit casts to std::string, GPL).
• UTF8-CPP (very lightweight, header-only).

Answer 2

Linux does not enforce an encoding on file names. Using UTF-8 is common though.