Re: About modules

[Clark McGrew <mcgrew at ale dot physics dot sunysb dot edu>]

>>>>> "Jost" == Jost Boekemeier <jostobfe@calvados.zrz.TU-Berlin.DE> writes:

    Jost> Clark McGrew <mcgrew@ale.physics.sunysb.edu> writes:

    >>  From my perspective, the gh_new_procedure stuff is the only
    >> thing that is likely to break many things.  Would it be trivial
    >> to have gh_new_procedure always add to the (guile-user) module

    Jost> Yes of course.  Next question: Which symbols should be
    Jost> visible in (guile user); pure R5rs or almost everything?  Do
    Jost> you need the old C module API also?

As far as I'm concerned, kill the old C module API, but others may
feel differently. 

 R5RS defines interaction-environment, null-environment and
scheme-report-environment.  The entire problem might be solved by
providing them as gh_ routines (ie gh_interaction_environment).

From R5RS: "The intent is that this procedure [interaction-environment]
will return the environment in which the implementation would evaluate
expressions dynamically typed by the user."

That sounds tailor made for the default environment.  Again, only if
it's *easy*: I'd suggest that adding gh_interaction_environment that
creates returns a sort of global environment a little like guile-user.
By default interaction environment should include scheme-interaction
environment, and the "usual" guile extensions.  The routines
gh_new_procedure, gh_define, and gh_load all work in this environment,
and be marked as obsolete.  Then the new preferred routines (say
gh_add_procedure, gh_bind, and gh_load_file) would all take an
environment.

<aside>
Does anybody use gh_define, gh_load and friends?  Maybe they can just
change.
</aside>

Thanks for working on this stuff.  I think it's going to really
improve guile.

Cheers,

Clark