Rules for `char8_t` et al.

[Eisenwave]

Bit by bit, the C++ standard is gaining more support for Unicode, and the charN_t family of types is gaining relevance. We should have some rules that guide the usage of the types.

A good starting point would be something like

Don't mix different character types in expressions

Mixing character types is often wrong, even when no narrowing occurs. Consider the following example:

bool contains_oe(std::u8string_view str) {
    for (char8_t c : str)
        if (c == U'ö') // comparison always fails
            return true;
    return false;
}

The comparison always fails because ö is UTF-8-encoded as 0xC3 0xB6, so even if str contains a u8"ö" somewhere, you wouldn't be able to find it this way.

There are certain instances where mixing character types is safe; for instance u8'x' == U'x' is true. However, safe use of this property requires the developer to memorize the set of ASCII characters.

Mixing char, wchar_t, and other character types in expressions is generally bug-prone because it's encoding-dependent. Treating char as char8_t may be safe if char is UTF-8 anyway, but that's far from universally true.

[hsutter]

Editors call: We agree there could be room for such a rule, but as you point out C++ is still in the process of gaining Unicode support, and the Guidelines are for established best practices with known features. We should pursue this again once the standard finishes adding enough Unicode support including basics like printing and formatting of char8_t.