Re: difference between using eval from the interactive prompt and from a file?

[Per Bothner <per at bothner dot com>]

On 11/10/2011 05:59 PM, alex mitchell wrote:
> I know that eval is generally discouraged, but I've been looking at using
> it for what (I think) is a valid reason. However, I've immediately run
> into a problem.
>
> I notice that this code:
>
> (define (myproc)
>    (display "myproc\n"))
>
> (if (environment-bound? (interaction-environment) 'myproc)
>      (eval '(myproc))
>      (display "not bound\n"))
>
> gives the result "myproc" if I type the code into the kawa prompt
> interactively, but gives the result "not bound" if I save it in a file,
> say "evaltest.scm", and then run "kawa evaltest.scm". Is there a reason
> why, when running from a file, eval can't see myproc? I'm guessing it has
> something to do with the way the file is compiled, and when top-level
> definitions are evaluated. Anyone have any insight or suggestions?

It works if you do:
  kawa -f evaltest.scm

The -f flag means to read and evaluate the file line-by-line (or
command-by-command), similar  to the way the load works.

If you just say kawa evaltest.scm it will compile the file as a single
module, and then evaluate it.  This causes different behavior in cases
like this.  The reason is that evaluating a module should not IMO implicitly
change the global environment - that contradicts the purpose of a module
system, which is to separate the namespaces.  Instead, when you load a
module into a REPL, the module's exported bindings are added to the REPL
*after* the load completes, not during.  See the evalModule2 method in
gnu/expr/ModuleExp.java.

This behavior is not ideal.  However, it does have some advantages.
For example compiling a file and then saying "kawa evaltest"
behaves compatibly, which I think is desirable.

On 11/11/2011 02:02 AM, Helmut Eller wrote:
> R7RS section 5.1 Programs:
>
>   Expressions occurring at the top level of a program are
>   interpreted imperatively; they are executed in order when
>   the program is invoked or loaded, and typically perform
>   some kind of initialization.
>
> So I think this is one more case where Kawa is not standard conform.

Ah, but when you say "kawa evaltest.scm" evaltest.scm is treated as a
module, not a program.  If you want evaltest.scm to be treated as a
R7RS "program" use the -f flag.  (Of course most of the time it
doesn't matter - in Kawa certain modules are often also programs,
or at least behave the same way when loaded.)

So this is not an example of Kawa non-conformance.
--
	--Per Bothner
per@bothner.com   http://per.bothner.com/