NetNews Usenet Archive 1992 #30/NN_1992_30.iso/spool/comp/lang/scheme/2758

Internet Message Format | 1992-12-15 | 3.2 KB

Path: sparky!uunet!cs.utexas.edu!usc!news.aero.org!Aero.org!doner
From: [e-mail] (John Doner)
Newsgroups: comp.lang.scheme
Subject: Re: Something I don't understand about bindings
Date: 15 Dec 1992 21:23:15 GMT
Organization: The Aerospace Corporation
Lines: 73
Distribution: world
Message-ID: <[e-mail]>
References: <a2vogler.724431792@rzdspc23> <[e-mail]> <[e-mail]>
NNTP-Posting-Host: armadillo.aero.org

In article <[e-mail]>, [e-mail] (Harald Hanche-Olsen) writes:
|> >>>>> On 15 Dec 1992 17:03:04 GMT, [e-mail] (John Doner) said:
|> John> NNTP-Posting-Host: armadillo.aero.org
|>
|> John> In article <a2vogler.724431792@rzdspc23>, [e-mail] (Christian Vogler) writes:
|>
...
|> John> |> (define (timed-prime-test n)
|> John> |>   (define start-time (runtime))
|> John> |>   (define found-prime? (prime? n))
|> John> |>   (define elapsed-time (- (runtime) start-time))
|> John> |>   (print n)
|> John> |>   (cond (found-prime?
|> John> |>          (print " *** ")
|> John> |>          (print elapsed-time))))
|> John> |>
|> John> |> The procedure then correctly prints the elapsed time for finding
|> John> |> a prime number, but why does this work?  [...]
|>
|> John> You are probably interpreting the define's as defining
|> John> functions, whereas they are actually binding variables to new
|> John> locations and assigning numeric values to those locations.  The
|> John> defines are evaluated sequentially, and all three of start-time,
|> John> found-prime?, and elapsed-time have their values before the
|> John> printing begins.
|>
|> That is not what r4rs says.  Indeed the above code is completely
|> equivalent to
|>
|>   (define (timed-prime-test n)
|>     (letrec ((start-time (runtime))
|>          (found-prime? (prime? n))
|>          (elapsed-time (- (runtime) start-time)))
|>       (print n)
|>       (cond (found-prime?
|>          (print " *** ")
|>          (print elapsed-time)))))
|>
|> which should not work, in standard r4rs scheme.  Not only is the order
|> of evaluation in a letrec unspecified, but as it says in the section
|> of internal defines:
|>
|> Just as for the equivalent \ide{letrec} expression, it must be
|> possible to evaluate each \hyper{expression} of every internal
|> definition in a \hyper{body} without assigning or referring to
|> the value of any \hyper{variable} being defined.
|>
|> I can only conclude that the original poster uses a non-r4rs
|> conforming scheme.
|>
|> - Harald

You're right, it says there is an equivalent letrec expression, but isn't very
precise about how to determine that.  And indeed, the letrec order of evaluation is
unspecified.  On the other hand, his original define (again according to r4rs) is
equivalent to

    (define timed-prime-test
        (lambda (n)
            (define start-time (runtime))
            (define found-prime? (prime? n))
            .
            .
            .
                ))

So the sequence of expressions beginning with the inner defines is the body of
a lambda expression.  And r4rs says the expressions in the body of a lambda
expression are to be evaluated sequentially.

Now I don't know what to think.

John