Re: new Scheme()

[Dean Ferreyra <dferreyra at igc dot org>]

Per Bothner wrote:
> Dean Ferreyra wrote:
>  > Up until recently this kind of construct worked for us:
>
> > public class Env {
> >     public static void main(String[] strings) {
> >         kawa.standard.Scheme.registerEnvironment();
> >         kawa.standard.Scheme scm = new kawa.standard.Scheme();
> >         scm.eval( "(require 'srfi-1)" );
>
>
> > but with the latest CVS pull I get a null pointer exception on the 
> > eval().  Adding a call to this:
> >
> >         scm.setEnvironment(scm.getNewEnvironment());
> >
> > before the eval() seems to be the fix.
> >
> > Is this right?
>
>
> No, I don't think so.  Creating a "new Scheme" is questionable;
> the changes I've been making are towards making langauges be
> "singleton classes" initialized at class initialization time.
>
> Instead, try:
>   kawa.standard.Scheme scm = kawa.standard.Scheme.getInstance();
>   ... scm.getNewEnvironment() ...
>
> This works, for now:
>   ... scm.getEnvironment() ...
> However, I'm not sure it will do so:  I think we want to move to a model
> where the current "user environment" is language-independent.  That
> implies that each language has an environment of predefined bindings,
> but the user environment is thread-specific, not language-specific.

Okay, thanks.

I do still have a point of confusion about environments, though I 
don't think it affects our current code.  Can users maintain separate 
environments explicitly?  This Java snippet generates a 
NullPointerException for me:

  kawa.standard.Scheme.registerEnvironment();
  kawa.standard.Scheme scm = kawa.standard.Scheme.getInstance();
  gnu.mapping.Environment env = scm.getNewEnvironment();
  scm.eval( "(define a 1)", env );

when trying to expand "(define ...)" or other macros.

Dean