Formal comment #111 (simplification)

syntax-rules should evaluate to a transformer
Reported by:	David Van Horn

Component:	macros
Version:	5.91

Summary

A syntax-rules form should evaluate to a transformer procedure at
expansion time.

Description

For simplicity and consistency, syntax-rules should evaluate to a
transformer. This simplifies Section 9.3, Syntax definitions, by
making the right hand side of define-syntax to be a transformer,
rather than a transformer or a syntax-rules form. Likewise, Section
9.20, Binding constructs for syntactic keywords, should simplify
transformer spec to transformer. Current language such as "the
specified transformers" already blurs the distinction and the obvious
implementation of syntax-rules in terms of syntax-case provides this
property.

This also enhances Section 17.3, Transformers, where it is said that a
binding for a syntactic keyword must be a transformer spec, and a
transformer spec must be an expression that evaluates to a transformer
(overriding the previous definition of a transformer spec). This
implicitly prohibits syntax-rules from being used in (r6rs
syntax-case) macros. Making syntax-rules evaluate to a transformer
eliminates this restriction. It also enhances this section by allowing
syntax-rules forms as inputs to make-variable-transformer.

RESPONSE:

The next report draft will describe syntax-rules as expanding to a
transformer.

Note:  The motivation for treating syntax-rules essentially as part of the
syntactic binding syntax was that it would allow implementations to
determine statically when only rewrite-style macros are used by observing
that no libraries are imported "for expand".  This will be addressed in the
next report draft by having the base library export both and only
syntax-rules and identifier-syntax, so that if no other bindings are
imported "for expand", the only valid transformer-creating expressions
will be syntax-rules and identifier-syntax.