There is a very strange behavior of the Ikarus REPL that I have discovered
a few months ago, but which I found difficult to reduce to a simple test case,
so that I have suffered it. I have now the simple enough test case.
Consider the following x.sls library, which is a greatly simplified version
of my sweet-macros library:
$ cat x.sls
(library (x)
(export syntax-match def-syntax)
(import (rnrs))
(define-syntax syntax-match
(lambda (y)
(syntax-case y (sub)
((syntax-match x (literal ...) (sub patt skel) ...)
(for-all identifier? #'(literal ...))
#'(syntax-case x (literal ...)
(patt skel)
...))
)))
(define-syntax def-syntax
(lambda (y)
(syntax-case y ()
((def-syntax name transformer)
#'(define-syntax name
(lambda (x)
(syntax-case x (<source>)
((name <source>) #''(... (... transformer))); comment-me!
(x (transformer #'x))))))
)))
)
The library defines a ``syntax-match`` form which is a simplified version
of syntax-case (no fenders), and such that each clause is
introduced by a literal identifier ``sub``. The library also
defines a ``def-syntax`` form which stores the source code
of the transformer (this happens in the line marked as ``comment-me!``).
The code as it stands is - I believe - correct; you can test it
with the following script:
$ cat xx.ss
(import (rnrs) (x))
(def-syntax macro (lambda (y) (syntax-match y () (sub (ctx x) #'x))))
(display (macro <source>))
Running it, you get the source code of the macro, as you would expect:
$ scheme-script xx.ss
(lambda (y) (syntax-match y () (sub (ctx x) #'x)))
The strange issue is that the script *does not work* at the REPL:
$ ikarus xx.ss
Ikarus Scheme version 0.0.4-rc1+ (revision 1767, build 2009-04-13)
Copyright (c) 2006-2009 Abdulaziz Ghuloum
Unhandled exception:
Condition components:
1. &message: "invalid syntax"
2. &syntax:
form: (syntax-match y () (sub (ctx x) #'x))
subform: #f
3. &source-position:
file-name: "xx.ss"
character: 51
4. &trace: #<syntax (syntax-match y () (sub (ctx x) #'x)) [char 51 of xx.ss]>
It looks like ``syntax-match`` is unable to recognize the literal identifier
``sub`` at the REPL.
I have tracked down the problem at the line marked as ``comment-me!``.
If I remove it (and I lose the feature of storing the source code
of the macro) there are no differences between the REPL and script
mode. I am baffled about what could be the origin of the problem,
which is quite disturbing, since it means that I cannot define any
macro at the REPL using sweet-macros, and I can test things only
non-interactively, with R6RS scripts :-(
Is this a bug of the REPL, or what? This has caused me lots of
pain and a head scratching ...
