[racket] Pure functional Racket

25 minutes ago, Matthias Felleisen wrote:
> 
> On Jul 6, 2011, at 9:16 PM, Eli Barzilay wrote:
> 
> > How would it be useful?  (Not a rhetoric question.)
> 
> I can think of two and a half ways: 
> 
> 1. Haskell has had a tremendous benefit in putting a stake in the
>    ground with "totally pure".

[You mean:

  Totally pure!
  *wink* *wink* *nudge* *nudge* *cough* unless you^W^W
  Nah, it's totally pure.  (Watch out! there's a giant tiger behind
  you!)
]


>    It's decision mechanism. It appears to be elegant. It is an
>    easily remembered design slogan.

But there you get the benefit of a compiler that can take advantage of
this, and we don't.  (Even if it does, it would require even more
subtle points like avoiding struct definintions, threads, exceptions,
continuations...)


>    Try to come up with a slogan like that for Racket. (I still think
>    we should say Racket is a programming language programming
>    language.)

(I don't think I've seen that before...  I remember Sam's "Racket is a
Programmable Language".)


> I suspect one would similar benefits from a call-by-value
> parenthesized macroized possibly typed Racket. I am sure monadic
> programming would be much easier with macros. I am also sure people
> will prefer call-by-value over lazy, especially if you bother with
> type classes or something like that.
> 
> In short, it is an unexplored point in the design space and you
> never know what you find.
> 
> 2. As you point correctly, it is a non-trivial task to ensure total
>    purity: require is the first problem that comes to mind but I
>    wouldn't be surprised if other reflection mechanism don't end up
>    being in the way. It looks easy as in "require Racket, export
>    only pure features" but for all we know, we may discover that we
>    need to support other ways of restricting languages.
> 
> 3. It would also be another playing field in exploring
>    interoperability.

Yeah, that's what I was implying -- it would be interesting to have a
way to talk to an unpure world.  In any case, if Cesar is aware of
this then that would definitely be an interesting experiment.


> The first exercise would be to define what purity is.
>   I/O anyone? Is the world model from big-bang acceptable? How about
>   files? CPS-IO? Monads?
> 
> The second exercise would be to implement this core language.
>   How many of the macros do you want to allow in, can you afford to
>   allow in?
> 
> The third one is probably to design a type system.
>   
> The fourth one calls for macros.
> 
> Someone go for it -- Matthias

-- 
          ((lambda (x) (x x)) (lambda (x) (x x)))          Eli Barzilay:
                    http://barzilay.org/                   Maze is Life!