On 8/25/06, Klaus D. Witzel <klaus.witzel at cobss.com> wrote:
> On Fri, 25 Aug 2006 09:46:39 +0200, Pavel Krivanek wrote:
>
> > I have tested it on Squeak 3.8
>
> Yes indeed, it starts in Squeak 3.8 :)
>
> > without problems.
>
> A typical developer understatement ;-)
>
> >
>
> Pavel, I have a working compiler for slot objects in Squeak (even a
> runtime), works "without problems". It does recursive descent parsing, so
> the developer has control and can debug (versus: it works in theory but
> YACC and friends' support for error reporting is terrible).
>
> I am about to adapt my system to make use of the #closures messages
> category of CompiledMethod (Squeak 3.9).

I was creating the outliner with the possible support of various
prototype enhancements for Squeak so feel free to use it.

> What are your intentions with Marvin:
>
> a - support Smalltalk syntax only
> b - support implicit-self sends
> c - support message cascading with ; (I do so)
>
> and, besides explicit redirection,
>
> d - support just one superclass (parent*) slot
> e - support multiple parent* slots
>

This issues are already resolved and implemented. Marvin uses its own
syntax (intersection of Self and Squeak syntax) and compiler that
translates it directly to Squeak bytecodes. And it has little VM
modifications that enables to store slot definitions and values
directly in one object and enhances it with delegation support.

-- Pavel