C++ Logo


Advanced search

Re: Allow virtual base classes in constexpr

From: Brian Bi <bbi5291_at_[hidden]>
Date: Tue, 4 Jun 2019 15:55:37 -0500
I think this is a reasonable thing to ask for. Since P1064 was accepted, in
C++20 we will be able to call virtual methods at compile time. The idea is
to unify compile-time and run-time programming styles as much as
possible---the fact that someone *could* achieve polymorphism at
compile-time using CRTP was not a compelling reason to not accept P1064.
Someone might want their multiple-inheritance-based design to work at
compile time, and that might require virtual inheritance. The main
objection I can see is that changing the language to allow it could be a
bit complicated (what are all the consequences of changing the definition
of "literal type" to include classes with virtual base classes?)

On Fri, May 31, 2019 at 10:59 PM Jack Adrian Zappa via Std-Proposals <
std-proposals_at_[hidden]> wrote:

> Perhaps you might be able to implement this in a CRTP manner, just like
> you could implement many constexpr functions like recursive ternary
> operation. However, I've been seeing many of the wonderful cppcon videos
> and am seeing that just because one can dig with spoons, doesn't mean that
> we should when a shovel is a much better tool (
> https://youtu.be/4AfRAVcThyA). Though, as CRTP is a static "inheritance",
> you'd have to use tuples instead of other homogenous containers.
> I don't pretend to understand the inner workings of the compiler, but as a
> developer, I would think that anything is possible, it's just the level of
> difficulty and if the feature is deemed worth the trouble.
> A
> On Fri, May 31, 2019, 22:33 Jake Arkinstall via Std-Proposals, <
> std-proposals_at_[hidden]> wrote:
>> Immediate thought: this doesn't make sense to me. Can you provide a more
>> detailed description of your particular use case? If it is just the
>> inheritance of the I/O child classes then CRTP is the canonical solution.
>> At least that way the type of the child class is known by the compiler,
>> whereas the entire point of virtual inheritance is to allow flexibility at
>> runtime.
>> As such, I don't think this is remotely possible apart from in the
>> circumstances that only one type is possible (E.g. in a branch where the
>> type is determined). If those circumstances are commonplace, virtual
>> inheritance is likely avoidable.
>> --
>> Std-Proposals mailing list
>> Std-Proposals_at_[hidden]
>> http://lists.isocpp.org/mailman/listinfo.cgi/std-proposals
> --
> Std-Proposals mailing list
> Std-Proposals_at_[hidden]
> http://lists.isocpp.org/mailman/listinfo.cgi/std-proposals

*Brian Bi*

Received on 2019-06-04 15:57:36