On 05/22/2018 02:51 PM, John Cowan wrote:
The SRFI requires that a c64/c128vector be disjoint from an f32/f64vector, as everything must be disjoint from everything else except u8vectors, which must be the same as bytevectors. However, the former can be implemented using the latter with a trivial record wrapper, and that's how the sample implementation is going to work.
Right - but I suggest adding some wording to that effect. Something like:
The real and imaginary parts, respectively, of the elements of a c64vector
*should* have the same range and precision as the elements of a f32vector,
and similarly for c128vector/f64vector.
Or replace *should* by "typically" or "in most implmentation".