Re: [boost] Keeping C++ feature discussion on topic for Boost

From: Karl Nelson (kenelson_at_[hidden])
Date: 2002-02-22 11:38:28

> From: "David Abrahams" <david.abrahams_at_[hidden]>
> > auto is a wierd edge case: it isn't, strictly speaking, needed for library
> > writing. However, libraries which use expression templates (e.g. lambda
> and
> > bind!) are a lot more usable if we have auto. If there were a way to
> provide
> > the same functionality using the existing language, you can be sure that
> > writers of such libraries would do so.
> 
> Yes, absolutely agreed, but we need to get our priorities right. I don't
> need auto to implement a real Boost.Bind (or a real Lambda), but I do need
> argument forwarding and typeof. This is my "library writer" perspective (
> i.e. the on-topic part. :-) ) (The pro-auto perspective is in the other
> post.)

I second this.  I could have done far more with sigc++ if I had 
argument forwarding and typeof.  (Yes, even sigc++ could meet boost
standards if rewritten with these extensions as late binding would
be possible.)  Without them the best the C++ can support is a large
template meta code to somewhat support signals/slots/bind (with nasty
limitations) or a crippled limited interface on like sigc++.

--Karl