On 26/12/2012, at 5:01 PM, Dobes Vandermeer wrote:
> I think the #true thing is not bad. It may be worth considering treating
> special cases like that as if "true" were a boolean literal rather than an
> identifier, and doesn't require a parameter. But that said, #true isn't that
> bad. In LISP they always write #t and it doesn't bother people much.
>
> I've been playing around with Stratego/XT and they decided to use a rule "if
> it's undefined, bind it - if it has a value already, match it" It's bitten
> me many times where I thought I was binding a new variable but actually there
> was already a variable bound with that name so it tried to match instead and
> failed. So ... that's an error-prone path to avoid despite allowing even
> less syntax. Having it syntactically clear whether one is binding or
> matching is definitely better for readability.
Yeah. And, somewhat generalised, your argument applies to other things in Felix
too. Felix uses quite a lot of sugar and magic and I'm not sure it is so sweet
as when I first implemented it.
The canonical example is:
var x = 1,2,3;
That's an array. originally you had to write:
var x = [< 1,2,3 >];
to convert a tuple to an array. Another more subtle one is:
fun add: int * int -> int = "$1+$2";
Yep, add looks like a function. But it isn't. Its a macro.
Arguments are usually lazily evaluated, that is, they're
just plugged in to $1 and $2.
however if you use it like a closure:
var f = add of (int * int);
then the arguments are eagerly evaluated.
How can Felix make a macro into a closure?
Answer: the compiler notices and generates a wrapper:
fun closure (x:int, y:int) => add (x,y);
which really is a function.
And .. it looks like the function takes a tuple argument.
But it doesn't. It takes two separate arguments.
So if you call it like:
var x = 1,2;
var y = add x;
it works, but only because Felix unpacks the tuple.
A lot of the "magic" is justified in terms of saying Felix is a scripting
language.
We cannot afford to be too precise, or we'd clutter code with a whole lot
of structural conversions with no semantic importance to the user.
ATS is a more precise language. Actually I quite like it because of that.
I'm thinking to "undo" some of the magic in Felix, to make it a bit
more "deterministic".
Also to lift some of the magic out of the compiler into the library.
One of these is operator dot (.). I am close to the point where I can
probably say:
operator dot is just reverse application
And when I can say that, I can throw out all the hacks in the
compiler that first do field check then reverse application,
and just put operator dot in the library. (At the moment
structure-value . field
does lookup first in the structure definition before trying
field structure-value
by the reverse application rule: I cannot tell the compiler to do this
in the library at the moment).
--
john skaller
skal...@users.sourceforge.net
http://felix-lang.org
------------------------------------------------------------------------------
LogMeIn Rescue: Anywhere, Anytime Remote support for IT. Free Trial
Remotely access PCs and mobile devices and provide instant support
Improve your efficiency, and focus on delivering more value-add services
Discover what IT Professionals Know. Rescue delivers
http://p.sf.net/sfu/logmein_12329d2d
_______________________________________________
Felix-language mailing list
Felix-language@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/felix-language
