This is the mail archive of the
kawa@sourceware.cygnus.com
mailing list for the Kawa project.
Re: Best way to interface Java?
- To: "Nic Ferrier" <nferrier at tapsellferrier dot co dot uk>
- Subject: Re: Best way to interface Java?
- From: Per Bothner <per at bothner dot com>
- Date: 21 Jun 2000 13:10:43 -0700
- Cc: kawa at sourceware dot cygnus dot com
- References: <s95129c2.072@tapsellferrier.co.uk>
"Nic Ferrier" <nferrier@tapsellferrier.co.uk> writes:
> Do you consider that more "schemey" than using the environment test?
I don't know about "schemey" but I don't consider environment tests
to be in the spirit of Kawa. I recommend modules that can be pre-compiled;
each module has a fixed set of names. Client modules check the names
exported by moudles they use at *compile time*. Environment tests
don't work well in such a framwork. (While Kawa doesn't have really usable
compile-time type-checking, I'm a big believer in it, at least as an
option. What Kawa does have is compile-time name-resolution, and that
is even more important than type-checking. You can't do type-checking
without name-resolution. Run-time environment tests should generally
be avoided.)
> One of the things I was trying to do was offer a bit of type
> protection (or at least argument number protection) because the procs
> used each have different number of arguments (and they are of
> different types).
My proposal using (object ...) does have the type protection
*and* is extensible *and* it fits well with Java. Whether it is
"schemey" is a matter for debate though ...
--
--Per Bothner
per@bothner.com http://www.bothner.com/~per/