P2988R12 std::optional<T&> #7979
Conversation
|
Thanks, @steve-downey, I can take it from here! |
28c97ee to
c81e631
Compare
source/utilities.tex
Outdated
| \tcode{T} shall be a type | ||
| other than \cv{} \tcode{in_place_t} or \cv{} \tcode{nullopt_t} | ||
| that meets the \oldconcept{Destructible} requirements (\tref{cpp17.destructible}). | ||
| A type \tcode{X} is a \defnx{valid contained type}{valid contained type!\idxcode{optional}} for \tcode{optional} if \tcode{X} is an lvalue reference type |
There was a problem hiding this comment.
Let's see if ISO lets us get away with this definition. If not, this is a local-enough term that we can just not mark it up, like we do elsewhere (e.g. shared_ptr's "empty").
c81e631 to
b15b2ef
Compare
I will keep an eye so as to improve the next one. |
dee99a1 to
bd3fcd7
Compare
bd3fcd7 to
f9fe472
Compare
| @@ -4442,6 +4451,551 @@ | |||
| \tcode{*this} does not contain a value. | |||
| \end{itemdescr} | |||
|
|
|||
| \rSec2[optional.optional.ref]{Partial specialization of \tcode{optional} for reference types} | |||
|
|
|||
| \rSec3[optional.optional.ref.general]{General} | |||
There was a problem hiding this comment.
Should we call this [optional.ref.general] instead? It's a sibling of [optional.ref.ctor], [optional.ref.assign], etc.
There was a problem hiding this comment.
Yes, good call. What about [optional.optional.ref], though?
There was a problem hiding this comment.
Same.
[optional.optional.ref*] is just stupid; we spent plenty of effort to get rid of [stmt.stmt].
There was a problem hiding this comment.
We could change that too, but my main concern was the excessive length of optional.optional.ref.general
There was a problem hiding this comment.
Hm, good point that we shouldn't immediately create a new set of problematic names that will invite ridicule and disbelief.
Can we come up with an acceptable complete set of new labels for this paper?
There was a problem hiding this comment.
The challenge is the parallelism with the existing mess that has [optional.optional.general] and then the ctors in [optional.ctor]. I had optionalref instead of optional.ref at one point, but that is not better.
\rSec1[optional]{Optional objects}
\rSec2[optional.general]{General}
\rSec2[optional.syn]{Header \tcode{<optional>} synopsis}
\rSec2[optional.optional]{Class template \tcode{optional}}
\rSec3[optional.optional.general]{General}
\rSec3[optional.ctor]{Constructors}
\rSec3[optional.dtor]{Destructor}
\rSec3[optional.assign]{Assignment}
\rSec3[optional.swap]{Swap}
\rSec3[optional.iterators]{Iterator support}
\rSec3[optional.observe]{Observers}
\rSec3[optional.monadic]{Monadic operations}
\rSec3[optional.mod]{Modifiers}
+\rSec2[optional.optional.ref]{Partial specialization of \tcode{optional} for reference types}
+\rSec3[optional.optional.ref.general]{General}
+\rSec3[optional.ref.ctor]{Constructors}
+\rSec3[optional.ref.assign]{Assignment}
+\rSec3[optional.ref.swap]{Swap}
+\rSec3[optional.ref.iterators]{Iterator support}
+\rSec3[optional.ref.observe]{Observers}
+\rSec3[optional.ref.monadic]{Monadic operations}
+\rSec3[optional.ref.mod]{Modifiers}
+\rSec3[optional.ref.expos]{Exposition only helper functions}
\rSec2[optional.nullopt]{No-value state indicator}
\rSec2[optional.bad.access]{Class \tcode{bad_optional_access}}
\rSec2[optional.relops]{Relational operators}Possibly changing the [optional.optional.ref] ones to [optional.ref] would be least worst?
There was a problem hiding this comment.
Possibly changing the [optional.optional.ref] ones to [optional.ref] would be least worst?
Good enough for me
| if (has_value()) { | ||
| return invoke(std::forward<F>(f), *@\exposid{val}@); | ||
| } else { | ||
| return remove_cvref_t<U>(); | ||
| } |
There was a problem hiding this comment.
IIRC LWG didn't want these braces here, but it's consistent with the equivalent functions for optional<T>. We can remove them from both places later if necessary.
There was a problem hiding this comment.
Hm, surely this is editorial? Why is LWG opposed to braces? Have braces ever decreased clarity of presentation? goto fail; goto fail.
There was a problem hiding this comment.
Yes definitely editorial. It's just shorter without the braces (only by one line, admittedly).
Fixes #7939
Fixes cplusplus/papers#1661
Note -- fixed a overfull hbox issue by forcing a break for the constraints at around 5469
is_reference_v ||
(is_move_constructible_v && is_swappable_v)
is true.