On 8/4/22 12:39 PM, Marc Nieper-Wißkirchen wrote:
> While the `vector-copy` solution seems to be correct with respect to the
> R[67]RS spec, its behavior is not optimal, I think, because it differs
> from the canonical `map`.  In some sense, `vector-map` is underspecified
> in R[67]RS.

That's what I was trying to say in

https://srfi-email.schemers.org/srfi-231/msg/20379060/

where I gave essentially the same proposed vector-map implementation as
yours (screwy-vector-map).

It's not enough that capturing and reinvoking a continuation not affect
previously returned results, it has to not affect future results
returned from invoking a different captured continuation.