[r6rs-discuss] multiple value returns to non-tail positions in `begin' et al

[Abdulaziz Ghuloum]

On Nov 1, 2008, at 3:07 PM, Andy Wingo wrote:

> Greetings wranglers of the lambda calculus,
>
> Quoth the R6:
>
>     The continuations of all non-final expressions within a  
> sequence of
>     expressions, such as in lambda, begin, let, let*, letrec, letrec*,
>     let-values, let*-values, case, and cond forms, usually take an
>     arbitrary number of values.
>
> Question 1: "Usually"?

Looks like an editorial oversight.  Cross "usually".

> Question 2: Why the change from R5, which sayeth:
>
>     Except for continuations created by the `call-with-values'
>     procedure, all continuations take exactly one value.

Because such continuations (e.g., in a begin) ignore their values.
Is there a reason to restrict them to a single value which they
ignore anyways, and force them to signal an error when given anything
other than a single value?

> Question 3: Does anyone's Scheme, R5 or R6, error with the following:
>
>     (define (nothing) (values))
>     ((lambda () (nothing) 1))
>
> This seems to be equal to 1 with R6, and undefined in R5. Of special
> interest are those schemes that do not construct compound objects for
> multiple value returns.

Ikarus does not construct compound objects for multiple values;
It also does not signal an error on that correct program.  Chez
is similar on both counts.  I don't know about the internals of
others.

Aziz,,,