On Sep 6, 2009, at 2:57 AM, Brian Harvey wrote:
> If there's a REPL, there aren't any phases. The REPL keeps shifting
> back
> and forth between syntax and semantics.
Can you please explain to me how the REPL involves phase-shifting in a
way that an ordinary top-level program does not in the case when only
`syntax-rules' is allowed on the right side of a `define-syntax' or
`let-syntax' form?
More generally, can you explain why forms which are input at the REPL
can't be treated identically to the forms of a top-level program, even
when a lower-level macro system is involved? As long as the user is
given the opportunity to reinitialize the top-level REPL environment
by starting a new top-level program with a fresh set of library
imports, I don't see why the semantics of the REPL need be any more
confusing than the semantics of top-level programs themselves. Such a
Scheme (excuse the pun) could be used to directly translate the
semantics of R6RS top-level programs to REPL semantics:
* When the REPL is initialized, a new top-level program is created
with an implementation-chosen set of initial bindings, preferably
containing at least (rnrs (6)) at phase (meta 0).
* When a certain token is encountered at the REPL, a fresh top-level
environment is created, and the REPL waits to read an `import' form to
define the bindings available in this environment. The natural choice
for this token is "#!r6rs".
* To aid in meta-programming at the REPL, the top-level environment
previously created at the REPL can be assigned a name and be made
available as a library for import at this point. In an implicitly-
phased implementation, the bindings are available directly. A separate-
instantiation implementation must record the top-level forms entered
at the REPL for future instantiations of the library.
This adds no more phase-shifting to the REPL semantics beyond that
which is already present in the semantics of top-level programs. All
phase boundaries are explicitly marked as such.
Here's an example of how this would work:
$ hypothetical-r6rs
> (define a 42)
> (define-syntax b (lambda (stx) a))
WARNING: reference to unbound identifier `a' in phase 1.
> #!r6rs
(import (rnrs base (6))
(rnrs syntax-case (6))
(for (rnrs base (6)) (meta -1)))
> (define foo (lambda (stx) (syntax-case stx () ((_ a body ...)
(identifier? #'a) #'(let ((a 42)) body ...)))))
> (define bar (display "hi"))
OUTPUT: hi
> #!r6rs
(import (rnrs base (6))
(top-level 1 ())
(for (top-level 0 ()) (meta 1)))
OUTPUT: hi ; in a separate-instantiation implementation only
> (define-syntax baz foo)
> (baz n (+ a n))
84
> bar
ERROR: reference to unbound identifier `bar' in phase 0.
--
Brian Mastenbrook
[email protected]
http://brian.mastenbrook.net/
_______________________________________________
r6rs-discuss mailing list
[email protected]
http://lists.r6rs.org/cgi-bin/mailman/listinfo/r6rs-discuss
