Yes, the #:when clause has to be last. I got my reasoning backwards. Do try both and report back. I am curious how expensive the ec is compared to the #:when test.
On Jul 25, 2012, at 4:37 AM, Jens Axel Søgaard wrote: > 2012/7/24 Vincent St-Amour <[email protected]>: >> You could capture the continuation outside `for*' check the index in the >> body and jump out when it gets out of range. >> >> (let/ec: done : Void >> (for*: (for:-clause ...) >> (unless (< i m*n) (done (void))) >> (define x (let () . defs+exprs)) >> (vector-set! v i x) >> (set! i (assert (+ i 1) index?)))) > > That works. > > Matthias: >> But if Jens is happy, I am happy :-) > > Only part happy :-) > > Using let/ec feels like cheating in the context of the for-framework. > I dug up the for for*/vector. It uses for*/fold with an #:when clause > as the last (phew) clause: > > (define-syntax (for*/vector stx) > (syntax-case stx () > [(for*/vector (for-clause ...) body ...) > (syntax/loc stx > (list->vector > (for*/list (for-clause ...) body ...)))] > [(for*/vector #:length length-expr (for-clause ...) body ...) > (syntax/loc stx > (let ([len length-expr]) > (unless (exact-nonnegative-integer? len) > (raise-argument-error 'for*/vector > "exact-nonnegative-integer?" len)) > (let ([v (make-vector len)]) > (for*/fold ([i 0]) > (for-clause ... #:when (< i len)) > (vector-set! v i (let () body ...)) > (add1 i)) > v)))])) > > I'll try both the let/ec and the for*/fold solution and see what's faster. > > /Jens Axel > ____________________ > Racket Users list: > http://lists.racket-lang.org/users ____________________ Racket Users list: http://lists.racket-lang.org/users
