On 22 May 2023 11:29:33 BST, Oleksandr Koval via Std-Proposals <std-proposals_at_[hidden]> wrote:
>>
>> The element type is pair<K const, V> and the data member layout of pair is
>> fully prescribed.
>
>I think we were talking about std::set here, not about std::map.
>Anyway, are you sure about the `pair<const T1, T2>` case? const_cast says:
>
>> const_cast makes it possible to form a reference or pointer to non-const
>> type that is actually referring to a const object
>> <https://en.cppreference.com/w/cpp/language/cv>
>>
>Const object says:
>
>> *const object* - an object whose type is const-qualified, or a
>> non-mutable subobject of a const object.
>>
>To me it doesn't look like `pair<const T1, T2>` satisfies any of the above.

That pair holds the "first" data member with type const T1. That's very much a const object.

It's an other question that the standard library has to do some magic to make "extract" work.

>
>On Mon, May 22, 2023 at 1:07 PM Edward Catmur <ecatmur_at_[hidden]>
>wrote:
>
>>
>>
>> On Mon, May 22, 2023, 04:44 Oleksandr Koval via Std-Proposals <
>> std-proposals_at_[hidden]> wrote:
>>
>>> I don't think it's UB since the underlying memory is not read-only.
>>>
>>
>> The underlying memory very much is read only. The element type is pair<K
>> const, V> and the data member layout of pair is fully prescribed.
>>
>> Arthur's solution can be improved using hints but I think that would still
>>> require checking node(s) around the hint to verify that position is correct.
>>> Btw, boost::container::set works in the way you need.
>>>
>>> On Mon, May 22, 2023 at 9:32 AM Thiago Macieira via Std-Proposals <
>>> std-proposals_at_[hidden]> wrote:
>>>
>>>> On Sunday, 21 May 2023 14:09:30 PDT Frederick Virchanza Gotham via Std-
>>>> Proposals wrote:
>>>> > I wrote the following code today; I have a struct for storing an IP
>>>> > address as well as two booleans (one for a FTP server and one for a
>>>> > Samba file server):
>>>> >
>>>> > struct IPFileShare {
>>>> > std::uint32_t ip;
>>>>
>>>> <rant>IP addresses aren't 32-bit and haven't been since the late 1990s.
>>>> Is
>>>> your code stuck in the 80s?</rant>
>>>>
>>>> Let's not ever write or even suggest code that fails to work with IPv6.
>>>>
>>>> > I think strictly speaking, the Standard says that this is undefined
>>>> > behaviour. How about we change the Standard to say that this
>>>> > const_cast is allowed so long as it has no effect on the sorting?
>>>>
>>>> In addition to what Arthur said, std::set is not just about sorting,
>>>> it's
>>>> about identity. The fact that your type seems to carry extra information
>>>> that
>>>> is not part of the identity does not mean it's the case for all other
>>>> types.
>>>>
>>>> Let's not break the requirements without compelling reason. I didn't see
>>>> a
>>>> compelling reason why it has to be this way. What other alternatives
>>>> have you
>>>> explored and what were their drawbacks?
>>>>
>>>> --
>>>> Thiago Macieira - thiago (AT) macieira.info - thiago (AT) kde.org
>>>> Software Architect - Intel DCAI Cloud Engineering
>>>>
>>>>
>>>>
>>>> --
>>>> Std-Proposals mailing list
>>>> Std-Proposals_at_[hidden]
>>>> https://lists.isocpp.org/mailman/listinfo.cgi/std-proposals
>>>>
>>>
>>>
>>> --
>>> Regards,
>>> Oleksandr Koval.
>>> --
>>> Std-Proposals mailing list
>>> Std-Proposals_at_[hidden]
>>> https://lists.isocpp.org/mailman/listinfo.cgi/std-proposals
>>>
>>
>