Re: [boost] [type_erasure]bug in computed_bases?

Subject: Re: [boost] [type_erasure]bug in computed_bases?
From: Larry Evans (cppljevans_at_[hidden])
Date: 2012-08-28 08:22:47

On 08/27/12 22:28, Steven Watanabe wrote:

Thanks for the prompt reply Steven.

> AMDG
>
> On 08/27/2012 05:39 PM, Larry Evans wrote:
>>
>>   1. typename Concept
>>
>>      The concept that we're specializing concept_interface for.
>>
>> which uses the singular Concept instead of the plural Concepts,
>> suggesting the Concept is not an mpl::vector<> of any sort.
>>
>
> I think I stated quite clearly in several
> places that an MPL sequence of Concepts
> *is a* Concept.
>

Ah yes.  I remember now.  Here:

http://steven_watanabe.users.sourceforge.net/type_erasure/libs/type_erasure/doc/html/boost_typeerasure/conceptdef.html

it says:

  Any MPL Forward Sequence whose elements are concepts is
  also a concept.

and here:

http://steven_watanabe.users.sourceforge.net/type_erasure/libs/type_erasure/doc/html/boost_typeerasure/concept.html#boost_typeerasure.concept.compose

it says:

  the library allows us to combine several concepts into a single
  concept using an MPL sequence.

Sorry for not remembering.  Maybe the concept_interface reference:

http://steven_watanabe.users.sourceforge.net/type_erasure/libs/type_erasure/doc/html/boost/type_erasure/concept_interface.html

should provide a link to conceptdef.html. After all, when reading such
a reference, a person generally expects to see all that's needed to
understand that reference or a link to additional information needed
to understand the reference.  Having to read all the docs and remember
what was said somewhere in those docs is, IMO, too hard, and liable to
lead to someone, such as myself, misunderstanding.

Now, the following questions concern the test code (compute_bases.cpp)
attached to my previous post. The output of that test code shows that
one of the superclasses of any<requirements_t> is:

  concept_interface
  < requirements_t
  .
  .
  .
  >

However, I can't think of a use case for providing such a superclass.
The unspecialized concept_interface definition:

template<class Concept, class Base, class ID, class Enable = void>
struct concept_interface : Base {};

has an empty body; hence, the only use of putting such a
concept_interface into one of the superclasses of an any (as opposed
to just using Base) is to allow for a specialization of it.  That's
completely unexpected to me, and I suspect, to others.  Also, the
Concepts in Depth tutorial:

http://steven_watanabe.users.sourceforge.net/type_erasure/libs/type_erasure/doc/html/boost_typeerasure/concept.html

provides no examples where the 1st template argument to
concept_interface is an mpl sequence of other concepts, leading the
casual reader to jump to the conclusion that the Concept arg to
concept_interface is *not* an mpl sequence.

Could you please provide a use case for such a concept_interface
specialization so users could understand when it might be useful?  The
specialization provided in the test code when:

  #define OVERLOAD_SIMPLE

certainly is not much use since it hides the intended simple member
function in concept_interface<::simple<C>,...>.

TIA.

-regards,
Larry