Re: [boost] Re: is_base_and_derived

From: Brian McNamara (lorgon_at_[hidden])
Date: 2003-10-27 17:08:00

On Mon, Oct 27, 2003 at 10:06:05AM -1000, David Abrahams wrote:
> Brian McNamara <lorgon_at_[hidden]> writes:
> > Presumably MPL has
> > some magical way to make
> >
> >    is_base_of<B,_>
> > and
> >    is_base_of<_,D>
> >
> > turn into the corresponding metafunctions that test if "T is a subtype
> > of B" or "T is a supertype of D", respectively, yes?
> 
> Well, yes it does.  It's just what you quoted above about visual
> binding:
> 
>         is_base_of<D,_>
> 
> Looks like a way of testing whether something is a base of  D.  How
> would you name a metafunction which tests if something is a base of
> some fixed class Foo?  Probably "is_base_of_Foo", right?

Yes.  I'm not sure where the boundary between error-prone-ness and
common sense lies.  One can say that

   fcpp::minus(3)

looks like a function to subtract 3 from a number, but a couple moments'
thought should reveal the correct thing to say is

   fcpp::minus(_,3)

I don't think this reveals a flaw in the construction/naming of minus(),
but rather it just reveals a bit of human psychology, namely that we are
prone to slip up when currying a subset of a function's arguments.  (In
Haskell, sugar like "infix syntax" and "operator sections" save the
day, but such tools don't exist for C++ template meta-programming.)

IMO this isn't worth the amount of discussion it's receiving, though.
Just give me a "proper" (in the sense of "subset" v "proper subset")
version of is_base_and_derived, as well as the current version, and
call them whatever you like, and I will make sure that my code using it
works.

-- 
-Brian McNamara (lorgon_at_[hidden])