Re: Lightweight Languages Workshop (DDJ Article)

[Friedrich Domincus <address@hidden>]

"Scott McKay" <swm@mediaone.net> writes:

> "Eric Kidd" <eric.kidd@pobox.com> wrote in message
> news:200201160037.TAA29849@life.ai.mit.edu...
> > On Tue, 2002-01-15 at 07:47, Rob Myers wrote:
> > > IIRC this isn't the case with modern Lisp if-you-know-what-you're-doing,
> > > but it is made explicit with Dylan's type constraints and sealed/open
> > > dynamism control. Fun-Dev's optimization syntax colouring makes this
> > > easy to learn and manage.
> >
> > Dylan still has a number of primitive operations with hard-to-predict
> > timing behaviors.  Method dispatch requires O(n log n) time for n
> > methods with simple argument types, and even greater time for (say)
> > type-unions of limited integer types.
> 
> "Virtual" method dispatch in any language that supports it has the
> same performance, though -- C++ and Java, e.g.
Can't confirm or deny it. But the Eiffel implementors insist that they
have found a solution with add just a constant factor to each
Methodinvocation. Well one have to admit Eiffel ist not very dynamic
and the compiler has to do a Full System check each time. Maybe than
it's possible to come to such a solution.

Regards
Friedrich