[r6rs-discuss] A different take on libraries

From: MichaelL_at_frogware.com <MichaelL>
Date: Wed Nov 15 12:46:01 2006

> > I think there's another way to go, though. Imagine, for example, that
> > libraries were designed the way that records are. Imagine that there
was a
> > procedural layer, and that the syntax was built on top of the
procedural
> > layer.
>
> I don't think that would work. Libraries are similar to "lambda" in the

> sense that they have to control compile-time visibility of bindings.
This
> cannot be done with procedures, just like "lambda" cannot be simulated
with
> procedures.

With syntax-case procedures can run at compile time, and they can
manipulate objects in the compile-time environment. If you look at
portable syntax-case you'll see that that's how module bits are
implemented. For example, here's the map that adds prefixes to the exports
of a library followed by the definition of the prefix-adding procedure:

        (map (prefix-add (syntax prefix-id)) exports)

        (define prefix-add
          (lambda (prefix-id)
            (let ((prefix (symbol->string (syntax-object->datum
prefix-id))))
              (lambda (id)
                (datum->syntax-object id
                  (string->symbol
                    (string-append prefix
                      (symbol->string (syntax-object->datum id)))))))))

So it's workable in some sense.

But there are undoubtedly some deeper concepts that would be better
expressed in primitive forms. That was kind of my point. The design
philosophy of Scheme is to identify and address such weaknesses. That's
why the language has tail calls and continuations; they're building blocks
that can be used to solve more than just one problem.
Received on Wed Nov 15 2006 - 12:46:33 UTC