I think this idea is worth exploring. Sam and Eli should implement it.
On Oct 5, 2012, at 8:39 AM, Eli Barzilay wrote: > Yesterday, Matthias Felleisen wrote: >> >> I know >> >> What I mean is an _abstraction mechanism_ inside of TR. Possibly >> something that nobody else has. > > +18.59 > > And a quick reminder -- below is the ancient hack I once did for such > things, maybe something like this would be easier now. Possibly much > easier: if a type is (foo ...) and `foo' is phase-1-bound to a macro > (anything that is not a bound type), then just expand it -- so that in > Neil's case the code could be: > > (define-syntax-rule (Xs->Xs T ...) (case-> (T -> T) ...)) > (: foo : (Xs->Xs Zero Flonum Real Float-Complex Complex)) > > (Either way, this means that it's easy to create huge types, and > therefore stressing TR's not-too-fast typechecker, but it sounds like > not having an abstraction doesn't stand in Neil's way to write huge > types anyway.) > > > ------------------------------------------------------------------------ > #lang typed-scheme > > (require (for-syntax scheme/base scheme/private/at-syntax)) > > ;; (define-syntax (STX stx) > ;; (syntax-case stx () > ;; [(_ expr) (at-syntax #'expr)])) > (define-syntax (:: stx) > (syntax-case stx (:) > [(_ name : x ...) > (with-syntax ([(x ...) (map at-syntax (syntax->list #'(x ...)))]) > #`(: name : x ...))])) > > (define-for-syntax Number #'Number) > (define-for-syntax (T1 . -> . T2) #`(#,T1 -> #,T2)) > > (:: twice : (let ([n (Number . -> . Number)]) > (n . -> . n))) > (define (twice f) (lambda (x) (f (f x)))) > ((twice add1) 10) > > -- > ((lambda (x) (x x)) (lambda (x) (x x))) Eli Barzilay: > http://barzilay.org/ Maze is Life! _________________________ Racket Developers list: http://lists.racket-lang.org/dev
