Dear List,
as a lurker on this list, and generally interested in the topic at hand,
I am asking myself lately what are the next plans regarding idst. I am
interested in seeing this project progress!
Those who have worked extensively with idst I am asking whether it
wouldn't be easier to directly make the AST representable in the
jolt/function/coke system directly instead of having a compiler object
(model). From looking at clojure, for instance, I could imagine adding
more powerful first class constructs to the lisp/schenme-like jolt would
eliminate some of this differences between the compiler object and the
rest. For instance concise syntax for associative arrays/maps or
annotations are at least to me sort of interesting sugar. On the other
hand I like the low level notion of the system, which is for reasons to
not having to change to some other language (like C) when implementing
system code. This on the other hand also raises the question why to have
to switch languages for accessing the current AST objects...
For instance the simple example:
(syntax begin
(lambda (node compiler)
`(let () ,@[node copyFrom: '1])))
could be natively implemented via for instance scheme's define-syntax or
some other form of pattern matching on the AST nodes, here I show the
standard R5RS scheme implementation:
(define-syntax begin
(syntax-rules ()
((_ first ....)
(let () first ...))))
The above makes the node natively accessible via the ... symbol.
My point is that this would make the system more self contained and
hence easier to learn - no?
I am very interested in hearing from you.
Thanks in advance
Jakob
_______________________________________________
fonc mailing list
fonc@vpri.org
http://vpri.org/mailman/listinfo/fonc
