Broken with Guile Scheme 2.0.7

[zilti]

Output from REPL:

scheme@(guile-user) [1]> (load "projects/logic/mk.scm")
scheme@(guile-user) [1]> (define f# fail)
scheme@(guile-user) [1]> (define s# succeed)
scheme@(guile-user) [1]> (run* (q) (== 1 q))
/home/zilti/projects/logic/mk.scm:216:2: In procedure subsume:
/home/zilti/projects/logic/mk.scm:216:2: In procedure module-lookup: Unbound variable: remp

Entering a new prompt.  Type `,bt' for a backtrace or `,q' to continue.
scheme@(guile-user) [2]> ,bt
In /home/zilti/projects/logic/mk.scm:
     69:7  4 (take #f #<procedure 2669f20 at <current input>:15:0 ()>)
In current input:
     15:0  3 (#<procedure 2669f20 at <current input>:15:0 ()>)
In /home/zilti/projects/logic/mk.scm:
   252:11  2 (#<procedure 2657ac0 at /home/zilti/projects/logic/mk.scm:247:4 (S+)> ((#(q) . 1)))
    305:6  1 (#<procedure 2668600 at /home/zilti/projects/logic/mk.scm:304:4 (A)> ())
    216:2  0 (subsume () ())

[webyrd]

Hi! Looks like remp isn't built into Guile.

(remp even? '(1 2 3 4 5)) => (1 3 5)

If you add this defn, do you get any other errors?

(define remp
(lambda (p ls)
(cond
((null? ls) '())
((p (car ls)) (remp p (cdr ls)))
(else (cons (car ls) (remp p (cdr ls)))))))

Thanks.

--Will

[drcz]

Hi!
It seems that guile2.0 lacks remp, exists and call-with-string-output-port, and sort takes arguments in opposite order -- appending the following [re]definitions to mk should do the job:

(define (remp p ls)
(cond ((null? ls) '())
((p (car ls)) (remp p (cdr ls)))
(else (cons (car ls) (remp p (cdr ls))))))

(define (exists p ls)
(cond ((null? ls) #f)
((p (car ls)) #t)
(else (exists p (cdr ls)))))

(define sorter (lambda (ls) (sort ls lex<=?)))

(define datum->string
(lambda (x)
(with-output-to-string (lambda () (display x)))))

Cheers,
d.

[enn]

This is still broken in chicken scheme.

The fix provided works as long as you delete the original sorter (which uses those args backwards).

[webyrd]

Hmm. The version of Vicare Scheme I'm using doesn't even define 'sort', although the code works if 'sort' is defined to be 'list-sort'.

I wonder if implementation-specific preludes are a better way to handle these incompatibilities, as was done with earlier versions of full Kanren. Any thoughts?

[RubyTuesdayDONO]

+1 for preludes if they can reliably adapt miniKanren to specific implementations … so many fun things to do with guile!

[RubyTuesdayDONO]

better yet: can miniKanren somehow generate the translations needed to implement it for a given scheme engine, provided abstract definitions of the external functions it requires? for example, "i need a procedure remp such that (equal? (remp 'b '(a b c)) '(a c))