Spec of syntax-case

["jocelyn paine" <ireson_paine at hotmail dot com>]

I think I'm finding that syntax-case doesn't act as I'd expect it to from 
the Dybvig specification. I tried loading a source file which
contains
  (define-syntax test-or
    (lambda (x)
      (syntax-case x ()
        ((_)
         (syntax #f)
        )
        ((_ e)
         (syntax e)
        )
        ((_ e1 e2 e3 ...)
         (syntax (let ((t e1))
                    (if t t (test-or e2 e3 ...))
                      )
                 )
         )
        )
    )
  )
which is Dybvig's example of how to implement 'or', and got
  #|kawa:1|# (load "test.scm")
  test.scm:12:28: evaluating syntax transformer `test-or' threw
   java.lang.NullPointerException:

I also tried
  (define-syntax test
    (lambda (x)
      (syntax-case x ()
        ((_ e)
         (identifier e)
         (syntax 'id)
        )
      )
      (syntax-case x ()
        ((_ e)
         (integer? e)
         (syntax 'int)
        )
      )
    )
  )

This loads OK, but when I try expanding it applied to an identifier, doesn't 
recognise the 'identifier?' function called in the "fender":
  #|kawa:1|# (load "t.scm")
  #|kawa:2|# (test a)
  <msg_stdin>:2:2: evaluating syntax transformer `test' threw
    gnu.mapping.UnboundSymbol: Unbound symbol identifier?

Removing the branch with identifier? in it (lines 3 to 8 in the definition) 
and expanding the macro applied to an integer gives:
  #|kawa:1|# (load "t.scm")
  #|kawa:2|# (test 1)
  <msg_stdin>:2:2: evaluating syntax transformer `test' threw
    gnu.mapping.UnboundSymbol: Unbound symbol syntax
So the call to integer? in its "fender" has succeeded, but then the macro 
isn't generating the result it should.

What I want to do is use macros to implement a pattern-matcher. Some of my 
patterns will be lists that may contain identifiers to be bound, or 
constants to be matched literally, e.g.
  (pattern 1 x 1 y 'a)
should match any 5-element list with 1 in 1st and 3rd positions and 'a in 
the 5th. syntax-rules is insufficient for this, because my macro has to 
recognise whether each element is a constant, variable, etc., and generate 
code appropriately. (Hence the experimenting with the type tests in the 
definition above.) Can I do this with Kawa's current implementation of 
syntax-case, and where should I look to find out what's implemented?

Jocelyn Paine
http://www.ifs.org.uk/~popx/







_________________________________________________________________
Send and receive Hotmail on your mobile device: http://mobile.msn.com