Sam and I have been developing a prototype language using redex and
#lang for about a year now, and that's been a really great experience (I
don't believe we could have done this research and gotten it right
without this tool support).
But on my way to OOPSLA I re-implemented some of the core parts of our
model to try to boil things back to their essential parts. This
involved a model for a PCF-like language (ie, simply typed lambda
calculus + numbers + if0 + some primitive operations). Because I wanted
to experiment by writing #lang pcf programs, I just wrote a trivial
module-begin and top-interaction wrapper to call
apply-reduction-relation* (LaTeX can't do that!). This is what Sam and
I had previously done for our much larger model.
Then I thought, PCF is typed, so let's get compile-time type checking.
That turned out to be a couple lines of code. Now as I write PCF
programs I get real-time feedback of type errors thanks to online check
syntax.
Then I thought, PCF's binding structure is a lot like Racket's: wouldn't
it be nice to get syntax coloring, syntax arrows, renaming, etc. That
was about another 30 lines of code.
In the end, with very little effort, I was able to whip up a language
with a static and dynamic semantics written in Redex and then leverage
much of the DrRacket environment to make programming in those languages
a much more pleasant experience than plain ol' s-expressions.
All of which is just to say: very cool. Racket should be considered as
the tool of choice for discriminating language hackers.
If you'd like to try it out or steal some of the ideas, the code is very
simple and located here:
https://github.com/dvanhorn/pcf
If you have Sam's raco git package, you can install with:
raco git --github dvanhorn pcf
David
____________________
Racket Users list:
http://lists.racket-lang.org/users
