[racket] Style. for/fold for/list for/lists

 I think, that in semantic of for/..., I mean (for/... ... (values ...)), it is impossible.

Possible API could be 

(for/collect ([a list] [b sum])
                    ([i (in-list '(1 2 3 4)])
   (cond 
      [(oddp? i) (a i) (b i)]
      [(< (b) 10) (b 1)]))

Returns (values '(2 4) 8)

Inside body of (for/collect ([id option ...] ...) body)
(id val) = Add val to collection
(id) = return current collection value

For for/fold-like one may make
(define-collector fold [fld #:bind v #:inner v])

Then
(for/fold ([sum 0]) ([i (in-list '(1 2 3 4)]) ... (values (+ sum i)))

Become
(for/collect ([sum fold :initial 0])  ([i (in-list '(1 2 3 4)]) ... (sum (+ (sum) i)) ...)

But I cannot imagine, how to make this without closures and set!.

Суббота, 18 января 2014, 13:51 -05:00 от "J. Ian Johnson" <ianj at ccs.neu.edu>:
>Suffice it to say, my package does not have that expressive power. I'd be happy to collaborate with you to design a good extension to my package so that this kind of accumulation is easier.
>-Ian
>----- Original Message -----
>From: "Roman Klochkov" < kalimehtar at mail.ru >
>To: "J. Ian Johnson" < ianj at ccs.neu.edu >
>Cc:  users at racket-lang.org
>Sent: Saturday, January 18, 2014 1:13:38 PM GMT -05:00 US/Canada Eastern
>Subject: Re[2]: [racket] Style. for/fold for/list for/lists
>
>It is great generalization for for/fold 
>
>But still missed something like this: 
>
>(iter (for i in l) 
>(if (oddp i) (collect i into a) (collect i into b)) 
>(when (> i 10) 
>(collect i into a)) 
>(finally (return a b))) 
>
>Or in your for/acc 
>(for/acc ([:type append] [:type append]) 
>([i (in-list l)]) 
>(if (odd? i) (values (if (> i 10) (list i i) (list i)) null) 
>(values (if (> i 10) (list i) null) (list i)))) 
>
>I had to double condition and the result is not so transparent. 
>
>Simple solution is: 
>
>(define (make-collector init add post) 
>(let ([a init]) 
>(case-lambda 
>[(val) (set! a (add val a)] 
>[() a]))) 
>
>(define (make-list-collector [init '()]) (make-collector init cons reverse)) 
>
>(let ([a (make-list-collector)] 
>[b (make-list-collector)]) 
>(for ([i (in-list l)]) 
>(if (odd? i) (a i) (b i)) 
>(when (> i 10) (a i))) 
>(values (a) (b))) 
>
>But they say, that purity is the best and one should avoid set!... so this way is the dark way, isn't it? 
>
>
>Суббота, 18 января 2014, 8:02 -05:00 от "J. Ian Johnson" < ianj at ccs.neu.edu >: 
>
>
>
>
>
>
>You have hit limitations of the for/fold macros. I wrote an extension for controlling the way values get accumulated, along with post-processing like list reversal. Let me know what you think! 
>
>https://github.com/ianj/nifty-macros/tree/master/for-acc
>-Ian 
>----- Original Message ----- 
>From: Roman Klochkov <  kalimehtar at mail.ru > 
>To:  users at racket-lang.org
>Sent: Sat, 18 Jan 2014 06:25:10 -0500 (EST) 
>Subject: [racket] Style. for/fold for/list for/lists 
>
>Where can I find style guide for these 
>
>for/lists: 
>
>I have to duplicate values names: 
>
>(define-values ([(desc-args lengths add-data) 
>(for/lists (desc-args lengths add-data) ([arg (in-list %args)]) 
>....)) is it ok? Or better 
>(define-values ([(desc-args lengths add-data) 
>(for/lists (a b c) (...) ...)) 
>
>? 
>
>for/fold has no "return" clause. So I have to do 
>(call-with-values 
>(for/fold (acc1 acc2 sum) (...) ...) 
>(lambda (acc1 acc2 sum) (reverse acc1) (reverse acc2) sum)) 
>
>And for/list doesn't allow to add several items to resulting list in one round. 
>Even with for/fold it gives 
>
>(reverse 
>(for/fold (acc) ([item (in-list list)]) 
>(values (cons (get-val1 item) 
>(cons (get-val2 item) acc))))) 
>
>Awful, especially for reading. 
>
>It is clumsy and unobvious. 
>In Common Lisp iterate I could do 
>(iter (...) 
>(collect (foo) into acc1) 
>(collect (bar) into acc2) 
>(summing ... into sum) 
>(finally (return acc1 acc2 sum))) 
>
>http://common-lisp.net/project/iterate/doc/index.html
>
>Is there something similar? 
>
>-- 
>Roman Klochkov 
>
>
>-- 
>Roman Klochkov 


-- 
Roman Klochkov
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.racket-lang.org/users/archive/attachments/20140119/852635e5/attachment-0001.html>