Re: apply vs. @apply

[Maciej Stachowiak <mstachow@mit.edu>]

mdj@nada.kth.se writes:
> Maciej Stachowiak <mstachow@mit.edu> writes:
> 
> > I noticed that Guile has an `apply' primitive, however, when
> > boot-9.scm is loaded, it loads r4rs.scm which redefines `apply' in
> > terms of the `@apply' macro and `apply:nconc2last'. It seems to me
> > that at least one of these definitions of `apply' should be
> > removed.
> 
> The primitive is used for bootstrapping until it is overridden by the
> scheme version. 

Since the first thing boot-9.scm does is load r4rs.scm, and the first
thing r4rs.scm does is set! apply to the Scheme version, I can't
imagine what bootstrapping this would be.

> The primitive is, in addition, used internally in Guile in lots of
> places.  


Yes but that does not require exporting it to Scheme as far as I can
tell.
 
> > The scheme version is also needed since it is tail-recursive.
> 
> > Other points: if the `@apply' version is kept, is there a way
> > to remove the need to have it also use `apply:nconc2last'? It seems to
> > me that this is an implementation detail that should be left at the C
> > level.
> 
> apply:nconc2last can easily be called from within the evaluator
> instead.  That would also be slightly more efficient.
> 
 
That's what I was thinking, especially since that is what scm_apply
essentially does.

> Also, if the C-implemented `apply' is fine and the comments
> about having to implement it in Scheme in terms of a macro to make it
> properly tail-recursive no longer apply,
> 
> What made you think they wouldn't apply?  They do.
 
I have no idea why they would or wouldn't, I never understood the
meaning of that comment in the first place. I thought since there was
a C-implemented and exported `apply' and some pains are taken to
specially declare it (it's not done with a scm_proc) that it might
work for some unknown magical reason.
 
 - Maciej