I see that the following evaluates successfully:
;;; this-file.sls --
(import (rnrs) (srfi lightweight-testing))
(check-set-mode! 'report-failed)
(check
(let ()
(define-syntax doit
(lambda (macro-use-stx)
(syntax-case macro-use-stx ()
((use)
(datum->syntax (syntax use) '(a))))))
(let ((a (lambda () 123)))
(doit)))
=> 123)
(check
(let ()
(define-syntax doit
(lambda (macro-use-stx)
(syntax-case macro-use-stx ()
((use)
(datum->syntax (syntax use)
'(list (a) (b) (c)))))))
(let ((a (lambda () 123)))
(let ((b (lambda () 456)))
(let ((c (lambda () 789)))
(doit)))))
=> '(123 456 789))
(check
(let ()
(define-syntax doit
(lambda (macro-use-stx)
(syntax-case macro-use-stx (alpha)
((use ?arg)
(datum->syntax (syntax ?arg) '(a))))))
(let ((a (lambda () 123)))
(doit alpha)))
=> 123)
(check-report)
;;; end of file
So it seems that macros at 'expand' time have access
to all the bindings that will be available at 'run'
time in the region of a macro use. Informations about
those bindings is stored somehow in MACRO-USE-STX.
But why the context that DATUM->SYNTAX needs to query
the available bindings has to be '(syntax use)' or
'(syntax ?arg)' when ?ARG is a literal and not
MACRO-USE-STX itself? And why the context is not
implicit, as if the <syntax-case clause>s are evaluated
in a:
(parameterize ((macro-use-context MACRO-USE-STX))
---)
form?
The pattern is used only for matching, it does not
introduce new bindings, so the bindings that
DATUM->SYNTAX has available to build the result
should be fixed once MACRO-USE-STX has been built.
Or not?
