|
Boost : |
From: Daniel Frey (daniel.frey_at_[hidden])
Date: 2002-02-18 11:54:16
Peter Dimov wrote:
>
> From: "Daniel Frey" <daniel.frey_at_[hidden]>
> >>>
> IMHO the question is: If we have your typeof, can we write a class
> 'auto' that doesn't require any further language support?
> <<<
>
> No, but we can write a macro. ;-)
>
> Frankly, I see nothing wrong with 'let', or 'auto', or 'dcl', or however it
> ends up (hopefully) being called. Why should we try to achieve the same
> effect with some other means? There's no point.
There is. Otherwise, we would end up with just another BASIC. (*Irony
on*) Why not start moving things from the STL to the language itself? It
would allow better optimizations, it would make 'static const string s =
"xyz";' an integral constant expression, inside of a class declaration,
etc. (*Irony off*) I don't thing this is very clever as it will result
in a language that doesn't give you options. I love the idea of
switching to another STL. Building a language with everything hidden in
the compiler, you can only switch between the compilers. And the
compilers themself become more error-prone (IMHO).
No, the language should be as minimalistic as possible but without
limiting you. This is IMHO the way it was designed and it is a Good
Thing(tm). We should add whatever is really needed and what makes the
language more consistent (e.g. thinking about that type-deduction for
constructor - if it works for functions, why shouldn't it be just the
same for ctor's?) but not just things that can also be implemented as a
class.
Regards, Daniel
-- Daniel Frey aixigo AG - financial training, research and technology Schloß-Rahe-Straße 15, 52072 Aachen, Germany fon: +49 (0)241 936737-42, fax: +49 (0)241 936737-99 eMail: daniel.frey_at_[hidden], web: http://www.aixigo.de
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk