[r6rs-discuss] an essay on language design

From: Thomas Lord <lord>
Date: Wed, 25 Jul 2007 12:24:34 -0700

John Tobey wrote:
>
>
> even if you all take you marbles and go home tomorrow.

Very unlikely: I lost my marbles years and years ago.


>
> Scheme already has an encapsulation mechanism: closures! It should be
> "(<type> 'make) ((obj 'is-a?) <type>) (obj 'field)" and either "(set!
> (obj 'field) value)" or "((obj 'set!) 'field value)" according to your
> tastes. [....]
> The only trick is to get a compiler smart enough to make something
> like C++ virtual tables out of it.


You'll want a fast, global property table with that, although you
might want to make it easy to limit which modules can access
that property table directly.

You still need a way to implement "(record? <datum>)", for
example.

Implementing a really good global property table like that
is interesting in two ways. First, if it is fast (and especially
if it is optimized for some special case property names like
small exact integers), then it instantly turns scheme into
an everything-is-an-object, prototype-based object oriented
programming language (albeit, with a higher level MOP left as an
exercise for the interested reader). Second, the implementation
burden of providing something like an efficient property table
is technically similar, in interesting ways, to the implementation
burden of adding user-defined record types. It's very similar
work logically, and often pragmatically -- which perhaps should
give us a hint about piling on features.

We could give it some original name.... oaklisp?

-t
Received on Wed Jul 25 2007 - 15:24:34 UTC