On 26 April 2025 00:58:21 BST, Arthur O'Dwyer via Std-Proposals <std-proposals_at_[hidden]> wrote:
>A quick Google search for "is_structural wg21" turns up:
>https://github.com/cplusplus/papers/issues/139 — P1154 "Type traits for
>structural equality"
><https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2019/p1154r2.html>,
>coauthored by Jeff Snyder and myself. I forget if I ever knew the details;
>but it looks like we got this into C++20 and then it was removed again as a
>result of
>https://cplusplus.github.io/LWG/issue3354
>and neither of us were paying attention by then, so that revert never got
>un-reverted.
>Meanwhile, `std::is_structural` is mentioned (and I *think* proposed? it's
>hard to tell at a quick glance) in P3380R1 "Extending support for class
>types as non-type template parameters"
><https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2024/p3380r1.html> (Barry
>Revzin, December 2024) — FWIW, since I don't follow Reflection, I don't
>understand the exact connection between that paper (which seems to be
>deeply involved with Reflection) and the idea of adding a
>std::is_structural{,_type} trait to the "normal" <type_traits> header. But
>it's definitely mentioned in that paper for *some* reason.
>
>Anyway, yeah, there's tons of history here; but I'm afraid I don't remember
>(or never knew) what all the history *is*.
>
>One possible problem with `is_structural` is that it's like the old
>now-removed `is_literal_type` — it tells you something about the type, but
>not a useful thing. Suppose I know that `T` is structural. Does that mean
>that I can write `MyTemplate<T{}>`? No, of course not: `T` might not have a
>constexpr default constructor. Or a constexpr destructor. Or its constexpr
>default constructor might set its non-static data members to something that
>can't be mangled (such as the address of a string literal). Or something
>else I'm not thinking of. So, if you can't use `is_structural` to tell you
>anything usable about the type, what exactly *is* the point of it?

You can write []<T>(){} exactly when T is structural. This is not SFINAE-friendly. I do agree that the trait probably has limited usability.

>
>Contrariwise, if all you want to know is whether `MyTemplate<T{}>` will
>compile, I *think in practice* you can (generally) use SFINAE or a
>requires-expression to tell you that, even without a dedicated trait.
>
>–Arthur
>
>On Fri, Apr 25, 2025 at 7:28 PM Braden Ganetsky via Std-Proposals <
>std-proposals_at_[hidden]> wrote:
>
>> Hi all,
>>
>> I haven't posted to this list before. I searched the archive to see if
>> this topic has been discussed here before, and I couldn't find anything.
>>
>> In C++20 we can now take class types as non-type template parameters, as
>> long as it is a "structural type". In my own code, I've wanted to directly
>> check this property at compile-time. Preferably, I'd like to static_assert
>> this after the definition of my type, instead of it simply failing to work
>> in the NTTP.
>>
>> Since "structural type" is defined in the standard (
>> https://eel.is/c++draft/temp.param#11) and not just a colloquial term, I
>> would like to see a trait that corresponds to this definition,
>> "std::is_structural". Would anyone else see value in this? Has this been
>> discussed before somewhere?
>>
>> Thanks,
>> Braden
>> --
>> Std-Proposals mailing list
>> Std-Proposals_at_[hidden]
>> https://lists.isocpp.org/mailman/listinfo.cgi/std-proposals
>>