Re: [Boost-users] [Bind] arity of arbitrary object

From: Peter Dimov (pdimov_at_[hidden])
Date: 2007-01-17 11:00:57

Chris Uzdavinis wrote:
> On 1/16/07, Sohail Somani <s.somani_at_[hidden]> wrote:
>>  Ok the incomplete type was because there was no specialization of
>> _bi::value<T> which occurs when the user actually supplies a value.
>> Does this solve the problem then?
>
> Thanks, that is pretty clever!  I was, however, holding onto a little
> hope that the solution would have been a little easier or
> straight-forward.  Utilizing implementation details like that makes me
> simultaneously smile and cringe.  Watching this discussion unfold has
> been rather enlightening.  My thanks to you and Peter.
>
> Any possibility that the Bind library could make this available as
> part of its documented interface, or were there reasons for its
> omission?

A proper way to provide this functionality would be to expose R,F,L as 
result_type (already done), function_type and argument_list_type, then make 
_bi::listN<> a conforming MPL (fusion?) sequence.

One reason for not extending boost::bind further is that the extensions 
won't necessarily make their way into tr1::bind or std::bind. It's not clear 
whether this should stop us, of course.