[racket] Behavior of continuation-mark-set->list and continuation-mark-set-first with respect to key values

Thanks Robby.

I'm attempting to implement the behavior described in Section 3.2 of 
your 2007 paper, specifically

"each continuation frame can have any number of marks with distinct 
keys"

by defining a new 'with-continuation-marks' syntax which will set up 
arbitrary number of nested 'with-continuation-mark' calls in response to 
a list of keys

As Eli B. picked up on earlier, I'm having trouble with Chapter 16 of 
the Racket Guide and its a mess at the moment.



Is there a basic example of pattern matching for a list of arbitrary 
number of elements (including zero)? I


; usage (with-continuation-marks '(id1 id2 id3) body)

;this doesn't work because of bad pattern matching in the 
with-continuation-marks macro.  (make-binding ...) just sets up storage.
;its the '(id1 id2 id3) in the arguments that's causing the problem.


(define-syntax with-continuation-marks
   (syntax-rules ()
     [(with-continuation-marks (id) body) (with-continuation-mark id 
(make-binding id) body)]
     [(with-continuation-marks (id1 id2 ...) body) 
(with-continuation-mark id1 (make-binding id1) (with-continuation-marks 
(id2 ...) body ))]))


On Sun, May 6, 2012 at 9:56 AM, Robby Findler wrote:

> continuation-mark-set->list always does return all of the bindings,
> but continuation marks are carefully designed to avoid breaking tail
> recursion (they let you understand the tail behavior of your program,
> more accurately) so that's why you see only one binding.
>
> continuation-mark-set-first can be more efficient than
> continuation-mark-set->list and it is more likely to work in futures.
> That's the only reason it exists.
>
> Robby
>
> On Sat, May 5, 2012 at 11:02 PM, Galler <lzgaller at optonline.net> 
> wrote:
>> BACKGROUND:
>>
>> Per the documentation:
>>
>> with-continuation-mark  causes a key to be associated with a binding 
>> within
>> a continuation frame
>>
>> continuation-mark-set->list returns the binding(s) associated with 
>> that
>> particular key, for any given continuation-mark-set
>>
>> while
>>
>> continuation-mark-set-first returns the first (innermost) binding 
>> associated
>> with a particular key
>>
>> QUESTIONABLE  BEHAVIOR
>>
>> The issue is:
>>
>> if a key is reused, the only binding continuation-mark-set->list 
>> returns is
>> the innermost binding.
>>
>> So continuation-mark-set->list could only possibly return **one** 
>> value
>>
>> and
>>
>> continuation-mark-set-first  appears redundant
>>
>> Is that the desired behavior? I would have expected
>> continuation-mark-set->list to return all the bindings for a 
>> particular key.
>>
>> Additionally, I suspect continuation-mark-set->context has to be 
>> implemented
>> using the 'return a list of all bindings for a single key' behavior, 
>> since
>> the documentation indicates it uses a single private key.
>>
>>
>> EXAMPLE
>>
>> #lang racket
>> (with-continuation-mark 'global  'outermost-value
>>  (let/cc k-outer
>>    (with-continuation-mark 'global 'middle-value
>>      (let/cc k-inner
>>        (with-continuation-mark 'global 'inner-value (begin
>>                                                       (display
>> (continuation-mark-set->list (current-continuation-marks) 'global))
>>                                                       (newline)
>>                                                       (display
>> (continuation-mark-set->list (continuation-marks k-inner) 'global))
>>                                                       (newline)
>>                                                       (display
>> (continuation-mark-set->list (continuation-marks k-outer) 'global))
>>         ))))))
>>
>> ;Returns
>>
>> ;(inner-value)
>>
>> ;(middle-value)
>>
>> ;(outermost-value)
>>
>> but I would have expected:
>>
>> ;Returns
>>
>> ;(inner-value) (middle-value) (outermost-value)
>>
>> ;(middle-value) (outermost-value)
>>
>> ;(outermost-value)
>>
>>
>> Thanks very much for looking at this.
>>
>> R/
>>
>> Zack
>> ____________________
>>  Racket Users list:
>>  http://lists.racket-lang.org/users