I have a question about Exercise 4.54 from Section 4.3.3 of Structure and Interpretation of Computer Programs (http://mitpress.mit.edu/sicp/full-text/book/book-Z-H-28.html#%_sec_4.3.3). This exercise concerns the Amb evaluator.
The exercise is the following:
If we had not realized that
require
could be implemented as an ordinary procedure that usesamb
, to be defined by the user as part of a nondeterministic program, we would have had to implement it as a special form. This would require syntax procedures
(define (require? exp) (tagged-list? exp 'require))
(define (require-predicate exp) (cadr exp))
and a new clause in the dispatch in
analyze
((require? exp) (analyze-require exp))
as well the procedure
analyze-require
that handlesrequire
expressions. Complete the following definition ofanalyze-require
.(define (analyze-require exp) (let ((pproc (analyze (require-predicate exp)))) (lambda (env succeed fail) (pproc env (lambda (pred-value fail2) (if <??> <??> (succeed 'ok fail2))) fail))))
I completed it as follows:
(define (analyze-require exp)
(let ((pproc (analyze (require-predicate exp))))
(lambda (env succeed fail)
(pproc env
(lambda (pred-value fail2)
(if (false? pred-value)
(fail2) ;; or (fail)
(succeed 'ok fail2)))
fail))))
My doubt is the following:
I know that, during execution, when the predicate value pred-value
is false, require
should fail; that is, it should call a failure continuation procedure. But I'm a bit confused as to whether it should call (fail)
or (fail2)
. Which one is correct?
(fail2)
is the correct one. The procedure conforms to a continuation-passing style, and the correct continuation procedure in this case is fail2
.