[r6rs-discuss] [Formal] identifier-syntax is not a derived form

[Andre van Tonder]

---
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        : Andre van Tonder
Email       : andre at het.brown.edu
Type        : defect
Priority    : minor
Component   : libraries
Version     : 5.91
Pages       : 114
Dependencies: Libraries (phase semantics) and free-identifier=? (p. 112)

Summary:
--------

Identifier-syntax is not a derived form.

Discussion:
-----------

Identifier-syntax cannot currently be portably expressed as a derived form.
I know of two conforming implementations in which it can be so expressed only
with en extension to negative import levels.  A third implementation
can express identifier-syntax without negative import levels, but only
by violating the stated semantics of free-identifier=?.

Here is essentially how it must be espressed in two of these implementations:

(library (core identifier-syntax)
   (export identifier-syntax)
   (import (for (r6rs syntax-case base) run expand)
           (for (r6rs base) expand
                            run
                            (meta -1)))  ;; FOR LAST OCCURRENCE OF SET!

   (define-syntax identifier-syntax
     (lambda (x)
       (syntax-case x (set!)
         ((_ e)
          (syntax (lambda (x)
                    (syntax-case x ()
                      (id (identifier? (syntax id)) (syntax e))
                      ((_ x (... ...))              (syntax (e x (... ...))))))))
         ((_ (id exp1)
             ((set! var val) exp2))
          (and (identifier? (syntax id))
               (identifier? (syntax var)))
          (syntax
           (make-variable-transformer
            (lambda (x)
              (syntax-case x (set!)   ;; THIS IS THE PROBLEMATIC OCCURRENCE
                ((set! var val)               (syntax exp2))
                ((id x (... ...))             (syntax (exp1 x (... ...))))
                (id (identifier? (syntax id)) (syntax exp1))))))))))
   )

A third implementation does not require the "(meta -1)".  It does so by
modifying the semantics of free-identifier=?.  In this implementation, the
current definition

   "The free-identifier=? procedure returns #t if and only if the two identifiers
    would resolve to the same binding if both were to appear in the output of a
    transformer outside of any bindings inserted by the transformer."

no longer holds.

Suggestion:
-----------

Consider refining the semantics of phases, import levels and/or 
free-identifier=? to make identifier-syntax expressible as a derived form.

Alternatively, put free-identifier=? somewhere else than the "derived forms"
section.