improve iota_view's iterator's binary operator+ (LWG-3580) - #2417
Conversation
Casey Carter (CaseyCarter)
left a comment
There was a problem hiding this comment.
I've already pushed a change to correct the pre-existing issues in my comments.
| } | ||
| _NODISCARD friend constexpr _Ioterator operator+(const difference_type _Off, _Ioterator _It) noexcept( | ||
| noexcept(static_cast<_Wi>(_It._Current + _Off))) /* strengthened */ requires _Advanceable<_Wi> { | ||
| is_nothrow_move_constructible_v<_Wi>&& noexcept( |
There was a problem hiding this comment.
Casey Carter (@CaseyCarter) I observe that this one says is_nothrow_move_constructible_v<_Wi> but the other two said is_nothrow_move_constructible_v<_Ioterator>. You mentioned that they're equivalent, and there seems to be a rationale (as this one constructs an _Ioterator prvalue), but I wanted to point this out in case it was unintentional and should be cleaned up later. No change requested here.
There was a problem hiding this comment.
The other two functions return an lvalue denoting a variable with automatic storage duration, so they are implicitly moved, and that variable is a function argument, so they cannot benefit from NRVO or RVO. Consequently they move construct an _Ioterator.
This function returns a prvalue, so no _Ioterators are moved or copied - C++17 delayed temporary materialization kicks in. It does need to account for potential throws from the _Ioterator constructor from _Wi, however. That constructor is annotated with noexcept(is_nothrow_move_constructible_v<_Wi>. I'm not sure why I decided to "inline" that noexcept-specifier instead of using the more direct is_nothrow_constructible_v<_Ioterator, _Wi>. I really need to work at making my conditional noexcept style more consistent.
|
I'm mirroring this to the MSVC-internal repo - please notify me if any further changes are pushed. |
|
Thanks for implementing a seventh LWG issue resolution in today's batch! 🚀 🚀 😹 |
Fixes #2398