From: Douglas Gregor (gregod_at_[hidden])
Date: 2002-05-11 10:15:09
On Saturday 11 May 2002 09:12 am, you wrote:
> I appreciate your input, and this is indeed another way to look at the
> problem. But I don't really thing it is easier to write ArietyTrait than
> it is to write make_function. (It might be more reusable though???)
>
> Also, I forgot to mention that I'm using MSVC6 so no partial
> specialization.
>
> :(
>
> Dirk Gerrits
You can do the same without partial specialization using function overloading:
template<typename R>
inline boost::function0<R> make_function(R (*f)())
{ return boost::function0<R>(f); }
template<typename R, typename T1>
inline boost::function1<R, T1> make_function(R (*f)(T1))
{ return boost::function1<R, T1>(f); }
template<typename R, typename T1>
inline boost::function1<R, T1> make_function(R (T1::*f)())
{ return boost::function1<R, T1>(f); }
// etc...
Of course, this style only works for free and member functions. We'd need
something like ArityTraits to handle arbitrary function objects. That's the
reason that make_function never really made it: we don't have a good set of
function object traits to work with.
If we _did_ have a working set of function traits, it would be possible to
make Boost.Function objects implicitly constructed only from function objects
that are callable by the Boost.Function object.
Doug