[r6rs-discuss] [Formal] Quasiquote should support unquotation within bytes objects.

[Felix Klock]

---
This message is a formal comment which was submitted to formal-comment at r6rs.org, following the requirements described at: http://www.r6rs.org/process.html
---
name: Felix Klock
email: pfr6rs at pnkfx.org
type of issue: Enhancement
priority: Minor
R6RS component: Bytes
version of the report: 5.91
summary: quasiquote should support unquotation within bytes objects.

text:

It seems natural and useful to support unquotation for the elements  
of a bytes object.

For example:

(let ((a 0) (b 1) (cs '(2 3)))
   `#vu8(,a ,b , at cs))
==> #vu8(0 1 2 3)

This is relevant to the section on quasiquotation (which is  
unnumbered in R^{5.91}RS but immediately follows section 9.19) on  
pages 52-53.

To support this:

1.) The grammar for quasiquotation would need to be extended with a  
template for bytes, something like the following:

<bytes qq template D> -> #vu8(<qq template or splice D>*)

2.) The report would need to specify whether spliced expressions can  
evaluate to either an octet or a byte at whim, or if the expressions  
must always evaluate to octets, or always to bytes.

This corresponds to deciding whether unquote within a #vu8(...) can  
expand unconditionally to bytes-u8-set! (or bytes-s8-set!), or if  
some new procedure will need to be introduced that handles both  
octets and bytes as arguments.

3.) The report would need to specify the behavior when an unquoted  
expression evaluates to a value that cannot be stored in a bytes  
object; one choice would be to raise a &contract exception.


One contentious aspect of this suggestion is that the bytes objects  
support a variety of -set! and -ref procedures, but quasiquote would  
presumably only accept unquoting of expressions that evaluate to  
single bytes (or, in the case of unquote-splicing, lists of bytes).