<valarray>: magic static

[AlexGuteniev]

Describe the bug
valarray potentially uses magic static

Command-line test case

#include <string>
#include <iostream>
#include <valarray>

int main()
{
	std::valarray<std::string> a{"strings", "are", "numeric", "types"};
	std::valarray<std::string> b = a.shift(2);
	for (std::string s : b) {
		std::cout << s << " ";
	}
}

Expected behavior
It works as expected.
But in C++17 where std::string constructor is not constexpr it should rely on magic static in debug mode:

STL/stl/inc/valarray

Line 353 in e000d3f

__PURE_APPDOMAIN_GLOBAL static _Ty _Dflt;

STL version
e000d3f

Additional context
Probably not a real problem, only from lawyering point of view.

[StephanTLavavej]

_Ty{} might or might not be ideal here (because it would theoretically construct, move-assign, and destroy), but I would expect to see an ordinary local variable otherwise.

[AlexGuteniev]

I'd expect _Construct_in_place(_Myptr[_Idx]);

[AlexGuteniev]

Actually figured it out. Yes, _Construct_in_place, but no, don't need to call it.
It was already called, since a new valarray for destination is created.

[AlexGuteniev]

Yeah,

    _NODISCARD valarray shift(int _Count) const {
        const size_t _Size = size();
        valarray<_Ty> _Ans(_Size);
        const size_t _Min = (_Count < 0) ? static_cast<size_t>(-_Count) : 0;
        const size_t _Max = (_Count > 0) ? _Size - static_cast<size_t>(_Count) : _Size;
        for (size_t _Idx = _Min; _Idx < _Max; ++_Idx) {
            _Ans[_Idx] = _Myptr[_Idx + _Count];
        }
        return _Ans;
    }

Was it ever worth to encapsulate a cryptic ternary into a mighty macro?