Re: [boost] Metaprogramming observations

From: Peter Dimov (pdimov_at_[hidden])
Date: 2001-12-15 13:04:07

From: "David Abrahams" <david.abrahams_at_[hidden]>
> From: "Peter Dimov" <pdimov_at_[hidden]>
>
> > However our particular domain (compile-time metaprogramming) is very
close
> > to the "ideal", academic pure functional language, since the "impure",
> > "real world" features are more than adequately implemented by the
> > traditional C++.
> > So it makes sense to base a compile-time metaprogramming library on
> > "classic" Lisp.
>
> I don't buy that argument: "traditional C++" doesn't give us anything at
> compile time. I would think a stronger argument would be that we should
use
> a "pure functional" style because template programs are "pure functional"
by
> constraint. I'm not sure whether I buy that one either, but I think it's a
> proposition worth considering.

I think that you're reading more in that "argument" than I intended. ;-)

I just wanted to justify my use of "Lisp" meaning the abstract, pure
functional Lisp machine, rather than a real world Lisp dialect that cannot
afford the luxury to be so pure and abstract.

My main concern is that the MPL attempts to hide the pair-based binary tree
from me, presenting me with an iterator-based sequence concept instead. It's
not the concept that I don't like, but its opacity. I know that pair-based
binary trees work; they are a proven idiom; I'm familiar with them.

--
Peter Dimov
Multi Media Ltd.