Excluding raw arrays of basic character types from ranges.

[OptimumCpp]

Raw arrays of basic characters have a dual role in C. They can represent strings as well as ordinary arrays. This ambiguity can result in surprises - dealing with ranges library:

constexpr auto JupiterMoons =
    "Callisto, Europa, Ganymede, Io"sv; 
constexpr auto TrySepJupMoons =
    JupiterMoons | ", "; //oops! '\0' 
constexpr auto SepJupMoons =
    JupiterMoons | ", "sv; //ok! sv

In the above example, TrySepJupMoons is a single element range containing the entire input. While the intended behavior was SepJupMoons. This is a hard to find mistake caused by the ambiguity of raw character arrays. IMO, the proper solution would be to exclude raw character arrays for all basic character types(char, wchar_t ...) from the std::ranges::range concept:

template <typename C>
concept basic_char_type = /*...*/;

template <typename R>
concept range = /*...*/ 
    and not
      ( is_array_v<T> 
    and requires (R & arr)
        { { auto{front(arr)} }
            ->basic_char_type; } );

This exclusion prevents accidental use of null terminate raw arrays as ranges. L-value raw character arrays can be used as span(arr) preserving the null charater or basic_string_view(arr) dropping it. R-value character arrays can use the "str view"sv UDL or std::array{"char arr"} CTAD.
It is also possible to define UDLs for std::array to initialize from string literals:

//null terminated UDL
template<array value>
consteval auto& operator""_0()
{  return value; };

//Drop the last null charater 
template<array charlit>
consteval auto operator""_a(){
   return span(charlit)
        . subspan<0, size(charlit)-1>();
};

Now we can unambiguously handle every case. Normally ""sv UDL creates a string view, but we can have std::array with ""_0 UDL as a null terminated sequence or std::span of limited extent for none-null terminated case with ""_a UDL.

The risk of ambiguous use is eliminated via exclusion in the primary std::ranges::range concept.

PS: the split example is modified sample snippet at cppreference. The modification and shortening is intended to show a problem, that the original carefully and smartly trying to avoid.

[JeffGarland]

Hi @OptimumCpp - Beman project uses discourse https://discourse.bemanproject.org/ for discussions. If you'd like restart this on discourse, although I'll admit I don't see how this is on topic. I'm going to lock this thread since we don't use github for this purpose. Thanks.