[Q] CCodeGenerator and inlining
Marcel Weiher
marcel at metaobject.com
Mon Jan 17 13:34:21 UTC 2000
> From: Henrik Gedenryd <Henrik.Gedenryd at lucs.lu.se>
>
> Another, and less important, question is why the TNode classes are
entirely
> separated from the Node classes of the regular Smalltalk parser,
instead of
> extending these. Is there a reason that I've missed?
Just came across the paper "Deltatalk: An Empirically and
Aesthetically Motivated Simplifcation of the Smalltalk-80 Language"
by Alan Borning (1987). One of the changes they recommend is a
simplification of the treating programs as data.
That should also make it easier to put new syntactic "skins" on the
core language of message passing objects, for scripting, constraints,
visual interpretations etc.
> (Btw, inlining is among other things the answer to macros in Lisp,
as asked
> about a while ago. I never saw why one would want macros with a
separate
> syntax when you could just have inlining of the regular code.)
Not knowing much if anything about LISP macros, the idea of
eliminating higher order programming constructs at compile-time (only
when desired!) is also quite attractive, in addition to AOP-like
stuff.
Marcel
More information about the Squeak-dev
mailing list
|