Re: define and modules

[Dirk Herrmann]

On Wed, 6 Nov 2002, Lynn Winebarger wrote:

> On Wednesday 06 November 2002 01:54, Dirk Herrmann wrote:
> > On 4 Nov 2002, Marius Vollmer wrote:
> > 
> > > Dirk Herrmann <address@hidden> writes:
> > > 
> > > > OK, talking about the right thing:  How should guile react to the
> > > > following code:
> > > > 
> > > > (define define-private define)
> > > > 
> > > > This is done in boot9.scm.  Should this be allowed?
> > > 
> > > No, since 'define' is a syntactic keyword... which is probably not the
> > > reason you were expecting, right?
> > 
> > It is exactly the reason.  The question is, how should guile react here?
> 
>     "Syntax error:  invalid use of define" or "Syntax error: invalid define 
> form"?

I think this is not general enough.  It works for the special case of
define, but the point is not, that the define form is broken, but that a
macro is used as an argument to something else.  Consider

  (call-foo define)

The point that we have to decide is, whether such a form is reported as a
syntax error during memoization, or whether we just don't dereference the
macro during memoization and let the executor do the lookup.  The second
choice will allow that a new definition of the macro may be provided in
the meantime.  For example:

  (define (foo) (call-foo and))
  (define and #t)
  (foo)

Should this be possible?  If so, then can't signal an error
_during_memoization_ if a macro is used at some illegal place.

Best regards,
Dirk