[ConceptGCC]|
[ConceptGCC] |
ConceptGCC :Re: variadic templates and concepts |
From: Sylvain Pion (Sylvain.Pion_at_[hidden])
Date: 2007-11-20 11:39:06
Douglas Gregor a écrit :
> Hello Sylvain,
>
> Sylvain Pion wrote:
>> I am working on specifying an extension of std::min() to a variable
>> number of arguments using variadic templates. For the concept-enabled
>> version, I need to define a concept that constrains all arguments
>> to be of the same type. That is, an extension of std::SameType to the
>> variadic case.
>>
>> I am stuck with a problem using conceptgcc, which I fear is a bug,
>> or maybe a misunderstanding on my side. A reduced test-case is below,
>> which generates the following error message:
>>
>> bug_report.cpp: In function 'const T& f(const T&, const T&, const
>> Args& ...)':
>> bug_report.cpp:23: error: no matching function for call to 'f(const
>> T&, #'type_pack_expansion' not supported by
>> dump_type_prefix#<typeprefixerror>&#'type_pack_expansion' not
>> supported by dump_type_suffix#)'
>>
>> ------------------------------
>> concept SameType <typename T, typename... Args> {}
>>
>> template <typename T>
>> concept_map SameType<T> {}
>>
>> template <typename T>
>> concept_map SameType<T, T> {}
>>
>> template <typename T, typename... Args>
>> requires SameType<T, Args...>
>> concept_map SameType<T, T, Args...> {}
> Hmmm. SameType is one of the built-in concepts, so its form is fixed by
> the language to the two-type variant:
>
> concept SameType<typename T, typename U> { }
>
> Another version of SameType would either have no built-in meaning for
> the compiler (if it's not in namespace std) or be ill-formed (if it is
> in namespace std).
>
> That said, we actually have two interesting options here. First of all,
> we could write your SameType concept in terms of std::SameType:
>
> auto concept SameType<typename T, typename... Rest> {
> requires std::SameType<T, Rest>...;
> }
>
> That expands to requirements std::SameType<T, Rest1> && std::SameType<T,
> Rest2> & ... & std::SameType<T, RestN>.
>
> Or, of course, we could change the specification to make std::SameType
> variadic.
>> template <typename T, typename... Args>
>> requires SameType<T, Args...>
>> const T&
>> f(const T&a, const T&b, const Args&... args)
>> { return f(a, args...); }
> Using std::SameType, one could write this as:
>
> template <typename T, typename... Args>
> requires std::SameType<T, Args>...
Thank you Doug, I had missed this option. This one at least does not
require the specification of a new concept, nor the extension of
std::SameType to be variadic, so it has some minimal needs.
> const T&
> f(const T&a, const T&b, const Args&... args)
> { return f(a, args...); }
>
> Unfortunately, ConceptGCC's support for variadic templates with concepts
> is essentially non-existent, so you aren't going to be able to test this
> with an actual implementation. Sorry!
As long as I can still test using the
<conceptgcc_at_[hidden]> compiler... :)
-- Sylvain Pion INRIA Sophia-Antipolis Geometrica Project-Team CGAL, http://cgal.org/