Re: primitive-load returning result of last evaluation?

[Dirk Herrmann]

On Mon, 22 Apr 2002, Rob Browning wrote:

> Dirk Herrmann <address@hidden> writes:
> 
> > I can't think of any problems, except that the standard 'load itself
> > does not provide a return value.  And, implementing a load* that
> > returns a value is quite simple.  However, in order to spare people
> > some work (I once also needed such a load function), we should
> > provide it, but, I'd rather go for a standard implementation and
> > provide such extensions as optional packages.
> 
> So you'd like to see a separate function providing this feature
> i.e. load-with-result and primitive-load-with-result (or whatever)?

Yes, in a loadable module.  We could also make it the default and provide
the pure standard versions as an extension.  Both is possible.  Which way
we go depends on how people prefer to use guile:  1) Code in RnRS scheme
most of the time, being aware of using extensions, or 2) Code in
guile-extended scheme most of the time, only from time to time caring
about portability.

Personally, I prefer to have my code as portable as possible, but others
may think differently.  It's not worth to fight about it, but we should
think about which way we would like to go.  If we go for guile-extended
scheme, we should, however, have an easy way to start guile in an
as-strict-as-possible RnRS mode.

IMO, the best solution would be a minimum RnRS guile, including only
functionality for loading modules.  Even optional RnRS features should be
loaded on demand.  This way, so I hope, we could make guile's default size
very small, much smaller than today.

Best regards
Dirk Herrmann