Re: Questions regarding modules

[Per Bothner <per@bothner.com>]

On 03/01/2014 07:52 AM, Marius Kjeldahl wrote:
> Question 1 - The line that does require HelloKawa; Unless I list the
> filename "HelloKawa.scm" as the second parameter, I will get a kawa
> compiler error; invalid specifier for 'require'. I have no idea why.
> I've experienced with setting the working dir prior to calling the
> kawa compiler, but it does not seem to matter much. It looks like kawa
> has it's own idea about the proper path for start searching for the
> source file for the module.

IIRC the filename "HelloKawa.scm" is needed in a require if
the file is not already-compiled *and* it is not currently being
compiled (i.e. on the kawa -C command line).

This may change if/when we implement R7RS define-library, or
we implement some kind of Scheme packaging mechanism, as has been discussed.

> Question 2 - When invoking (HelloKawa:hello), how come kawa knows
> about my HelloKawa class without the full class name? I haven't done
> any define-alias on it, so it's a bit unexpected.

When you (require MODULE) what happens is that all the public (exported)
names in MODULE are added to the current lexical scope.  The class HelloKawa
is a public exported Scheme name in the HelloKawa module - since they happen
to have the same name - and they happen to get compiled to the same class.

A (define-simple-class HelloKawa () ...) creates a definition of the
identifier HelloKawa in the current Scheme scope - and that binding
is by default exported.  This action is in addition and distinct from
the action of create a JCM class net.kjeldahl.kawatest.app.HelloKawa.

Conceptually the "module class" net.kjeldahl.kawatest.app.HelloKawa
is different from the define-simple-class HelloKawa.  They're just
implemented using a single class, with module-level functionality being
static, and define-simple-class-level stuff being (usually) non-static.

> Question 3 - The file defining HelloKawa uses define-simple-class, but
> I've also tried it with a straight up define instead (you'll see that
> commented out in the source file above). If I do that, I get two
> errors from kawa; no known slot hello in java.lang.Object, and no
> declaration seen for HelloKawa.

Hopefully the previous answers will explain this.

>Based on what I've read, this second
> method of defining a module should work as long as we stay within the
> kawa domain, but I may have interpreted that wrong.

There is no special magic for the "kawa" domain.  They is some special
magic for "kawa.lib" when using R7RS-style import.
-- 
	--Per Bothner
per@bothner.com   http://per.bothner.com/