[racket] call-with-composable-continuation and DrRacket

OK I think I grok it.

When I write in the definitions window of DrRacket
#lang racket
(+ 100
    (call-with-composable-continuation
    (λ (k) (+  (k 1) 1000))))

what actually gets executed is at least:

(call-with-values (+ 100
    (call-with-composable-continuation
    (λ (k) (+  (k 1) 1000)))) print-values)

and the continuation captured by cwcc is at least:
(λ (x) (call-with-values (lambda () (+ 100 x)) print-values))

so what gets executed is:
((call-with-values
 (+ 100
    (+ 1000
       (call-with-values ((λ (x) (+ 100 x)) 1)
                         print-values)))
 print-values)


printing out
101
1201

When I actually tried to execute this, I got an unbound identifier
error for print-values and I couldn't find a reference to print-values
in the racket reference manual.

Harry Spier



On Thu, Aug 30, 2012 at 6:11 PM, Matthias Felleisen
<matthias at ccs.neu.edu> wrote:
>
> On Aug 30, 2012, at 5:48 PM, Harry Spier wrote:
>
>> When I run this in the definitions window of DrRacket
>> #lang racket
>> (+ 100
>>   (call-with-composable-continuation
>>    (λ (k) (+ 1000 (k 1)))))
>>
>> it prints
>> 101
>> 1201
>> in the interactions window.
>> Why doesn't it just print 1201 ?
>
>
> As John says, the meaning of this program isn't "on its sleeve" as semanticists used to say in the 1980s. If you use the Macro Stepper to expand the above program (- racket, -library), you get
>
> (module anonymous-module racket
>   (#%module-begin
>    (#%app
>     call-with-values
>     (lambda ()
>       (#%app
>        +
>        (quote 100)
>        (#%app call-with-composable-continuation (lambda (k) (#%app + (quote 1000) (#%app k (quote 1)))))))
>     print-values)))
>
> The key is to note that an expression is wrapped in a (call-with-values . print-values), plus an outer #%module-begin. Hence the meaning of cwcc is (at least)
>
>  (lambda (x) (call-with-values (lambda () (+ 100 x)) print-values))
>
> ;; ---
>
> In contrast, the same expression in the repl gets expanded via #%top, which does not wrap the print-values around it -- after all the "printer" is a part of the REPL.
>
> ;; ---
>
> Perhaps this is somewhat ironic, because the idea is really that each expression in the def window should be wrapped in a "prompt" -- but it means only "control delimiter". Something to consider is that we take the meaning of the word "prompt" literally and have the two areas behave identically.
>
> -- Matthias
>
>
>
>