Re: [boost] Re: std::complex design (Was: N-Dimensional array library submission...)

From: David Abrahams (david.abrahams_at_[hidden])
Date: 2002-01-11 13:36:58

----- Original Message -----
From: "rwgk" <rwgk_at_[hidden]>

> --- In boost_at_y..., "David Abrahams" <david.abrahams_at_r...> wrote:
> > Don't you want a constructor? ;-)
> >
> > Check the POD requirements I posted again. Pretty restrictive.
>
> David,
> Is the bottom line that a core language change is required
> to support C or FORTRAN compatibility for a type
> like std::complex? E.g., the definition of an "enhanced POD"
> which allows a constructor "as a notational convenience"
> but has the same guarantees as a POD?

Yes.

> This sounds like a big political mission for fixing a relatively
> theoretical problem that might only begin to bite for real
> when someone comes up with radically new hardware.

It's not just theoretical at all. There are all kinds of ways in which this
comes up. I have similar problems with Boost.Python when I want to embed a
C++ object directly in a Python object. Believe me, your arguments will
carry weight with the committee... you just need to make sure they're
accurate ;-)

> Yet, from a language designers viewpoint, I would say the
> current situation is very unsatisfactory. Are there precedences
> for core language changes like the one that would be required
> here?

In some sense we're now in the realm of "no precedent" because we're working
on the first revision after standardization. There have been no major core
language changes yet. Still, I think this is one area that would find a
receptive audience.

-Dave