define-syntax

define-syntax

[Pach Roman (DGS-EC/ESG3)]

[-- Attachment #1: Type: text/plain, Size: 779 bytes --]

I've performed following two short tests.

test 1:
-------

(use-syntax (ice-9 syncase))

(define-syntax my-macro-1
  (syntax-rules ()
    ((_ par1 par2 par3)
     (begin
	   (string-concatenate (list par1 par2 par3))
	   ))))

(define (dummy)
  (my-macro-1 "a" "b" "c"))

(format #t "dummy => ~s\n" (procedure-source dummy))

result:
dummy => (lambda () (string-concatenate (list "a" "b" "c")))

test 2:
-------
(define-macro (my-macro-2 par1 par2 par3)
  (string-concatenate (list par1 par2 par3)))

(define (dummy)
  (my-macro-2 "a" "b" "c")
  )
(dummy)
(format #t "dummy => ~s\n" (procedure-source dummy))

result:
dummy => (lambda () "abc")


It seems to me the define-syntax is broken in version "1.8.2"
or I've done something wrong.

[-- Attachment #2: Type: text/html, Size: 3011 bytes --]

Re: define-syntax

[Neil Jerram]

2008/12/9 Pach Roman (DGS-EC/ESG3) <Roman.Pach@de.bosch.com>:
> I've performed following two short tests.
>
> test 1:
> -------
>
> (use-syntax (ice-9 syncase))
>
> (define-syntax my-macro-1
>   (syntax-rules ()
>     ((_ par1 par2 par3)
>      (begin
>            (string-concatenate (list par1 par2 par3))
>            ))))
>
> (define (dummy)
>   (my-macro-1 "a" "b" "c"))
>
> (format #t "dummy => ~s\n" (procedure-source dummy))
>
> result:
> dummy => (lambda () (string-concatenate (list "a" "b" "c")))
>
> test 2:
> -------
> (define-macro (my-macro-2 par1 par2 par3)
>   (string-concatenate (list par1 par2 par3)))
>
> (define (dummy)
>   (my-macro-2 "a" "b" "c")
>   )
> (dummy)
> (format #t "dummy => ~s\n" (procedure-source dummy))
>
> result:
> dummy => (lambda () "abc")
>
> It seems to me the define-syntax is broken in version "1.8.2"
> or I've done something wrong.

These examples both look good to me.  Which one do you think is wrong?

It may help if I point out that define-macro usage usually involves
backquoting.  The define-macro version of your define-syntax example
would normally be:

 (define-macro (my-macro-2 par1 par2 par3)
   `(string-concatenate (list ,par1 ,par2 ,par3)))

Regards,
       Neil

Re: define-syntax

[Neil Jerram]

2008/12/9 Pach Roman (DGS-EC/ESG3) <Roman.Pach@de.bosch.com>:
>
> take look at the results
>
> (1) dummy => (lambda () (string-concatenate (list "a" "b" "c")))
> (2) dummy => (lambda () "abc")
>
> in the case of define-syntax (1) the string "abc" will be computed at
> the run time,
> the macro version (2) hat produced the "abc" already by reading the
> macro definition.

Agreed.

> I've expected both of the be computed during the reading of scheme
> source.

That's not what define-syntax is supposed to do, as I understand it.
If you think we've got this wrong, can you point to something specific
in R5RS?

Regards,
      Neil

define-syntax

[Ludovic Courtès]

Hello,

From R6RS Section 10:

  define-syntax form  The expander expands and evaluates the
    right-hand-side expression and binds the keyword to the resulting
    transformer.

Thus I think the following should work:

  (define-syntax +
    (let ((plus +))  ;; `+' should resolve to whatever `+' is bound to
                     ;; before this definition
      (lambda (stx)
        (syntax-case stx ()
          ((_ args ...)
           (apply plus (map syntax->datum #'(args ...))))))))

But it fails when expanding the macro:

  scheme@(guile-user)> (+ 2 3)
  Throw to key `wrong-type-arg':
  ERROR: In procedure vm-debug-engine:
  ERROR: Wrong type to apply: #<syntax-transformer +>

Thoughts?

Thanks,
Ludo’.

Re: define-syntax

[Andy Wingo]

Hi,

On Tue 15 Jun 2010 22:48, ludo@gnu.org (Ludovic Courtès) writes:

> From R6RS Section 10:
>
>   define-syntax form  The expander expands and evaluates the
>     right-hand-side expression and binds the keyword to the resulting
>     transformer.
>
> Thus I think the following should work:
>

An interesting issue, and an interesting example. The problem boils down
to eval-when.

$ cat > foo.scm
(define-syntax +
  (let ((plus +))  ;; `+' should resolve to whatever `+' is bound to
                   ;; before this definition
    (lambda (stx)
      (syntax-case stx ()
        ((_ args ...)
         (apply plus (map syntax->datum #'(args ...))))))))
^D
$ guile -l foo.scm -c '(begin (display (+ 1 2 3)) (newline))'
6

Indeed + does resolve to whatever + was bound to before the definition;
it's just that when you define the + macro it usually defines at
compile-time too! By compiling ahead of time and exiting we leave + in
its pristine state. See also the discussion of
eval-syntax-expanders-when in psyntax.scm or in
http://www.scheme.com/csug8/system.html#./system:s78.

Cheers,

Andy
-- 
http://wingolog.org/

Re: define-syntax

[Ludovic Courtès]

Hey!

Andy Wingo <wingo@pobox.com> writes:

> Indeed + does resolve to whatever + was bound to before the definition;
> it's just that when you define the + macro it usually defines at
> compile-time too! By compiling ahead of time and exiting we leave + in
> its pristine state. See also the discussion of
> eval-syntax-expanders-when in psyntax.scm or in
> http://www.scheme.com/csug8/system.html#./system:s78.

Ooh, interesting.  Thanks for the explanation!

(I eventually found out about the problem in my constant folding macro,
where bits of ‘eval-when’ in the right place did the trick.)

Ludo’.

Re: define-syntax

[Marco Maggi]

"Ludovic Courtès" wrote:
>  (define-syntax +
>    (let ((plus +))

I am assuming you are starting the program with:

(import (rnrs))

or you are importing at level 1 a library which exports "+";
if this is  the case, and you want  R6RS compatibility, IMHO
this should fail because  you are redefining the binding for
"+";  many R6RS  implementations agree  with this  (with the
exception of Ypsilon which has hygiene problems and must not
be taken as model).

  Notice that  bindings from  "(rnrs)" are imported  at both
levels 0 and 1[1]:

    For  the libraries  defined in  the library  report, the
    export  level   is  0  for  nearly   all  bindings.  The
    exceptions are syntax-rules, identifier-syntax, ..., and
    _ from  the (rnrs base (6)) library,  which are exported
    with  level 1, set!  from the  (rnrs base  (6)) library,
    which is exported with levels  0 and 1, and all bindings
    from  the  composite  (rnrs  (6)) library  (see  library
    chapter on “Composite library”), which are exported with
    levels 0 and 1.

  As  a side  note: the  existence  of the  binding for  the
keyword can be recorded by  SYNTAX in the right-hand side of
a DEFINE-SYNTAX, so  that the binding itself is  can be used
to refer to the context of the definition:

  (import (rnrs))

  (define ciao 123)

  (define-syntax this
    (lambda (stx)
      (syntax-case stx ()
        ((_)
         (datum->syntax #'this 'ciao)))))

  (write (this))
  (newline)

and also:

  (import (rnrs))

  (define ciao 123)

  (define-syntax this
    (let ((ctx #'this))
      (lambda (stx)
        (syntax-case stx ()
          ((_)
           (datum->syntax ctx 'ciao))))))

  (write (this))
  (newline)

for this kind  of things, I suggest taking  the behaviour of
Larceny as model: if it works with it, it will probably work
with all the other R6RS implementations.

HTH

[1] <http://www.r6rs.org/final/html/r6rs/r6rs-Z-H-10.html#node_sec_7.2>
-- 
Marco Maggi

Re: define-syntax

[Andy Wingo]

On Wed 16 Jun 2010 09:35, Marco Maggi <marco.maggi-ipsu@poste.it> writes:

> "Ludovic Courtès" wrote:
>>  (define-syntax +
>>    (let ((plus +))
>
> I am assuming you are starting the program with:
>
> (import (rnrs))

Probably not a good assumption to make on this list :)

Andy
-- 
http://wingolog.org/

Re: define-syntax

[Marco Maggi]

"Andy Wingo" wrote:
> Probably not a good assumption to make on this list :)

The message from the OP started with:

> From R6RS Section 10:
>
>  define-syntax form  The expander expands and evaluates the
>    right-hand-side expression and binds the keyword to the resulting
>    transformer.
-- 
Marco Maggi