A brief update on ATS3 implementation

[gmhwxi]

 Hi, there,

I would like to make a brief announcement, telling those in this group

about some recent progress made in the implementation of ATS3. The

project itself is publicly accessible at the following site:

https://github.com/githwxi/ATS-Xanadu

Note that you need to have ATS2 installed in order to compile ATS3.

So far I have essentially finished the following parts, though changes and fixes

are definitely needed for a long time to come:

Part 1: Concrete syntax design + Parsing

Part 2: Binding resolution based on static scoping

Part 3: Type-inference + supporting for symbol overloading

Part 4: Type-based template code selection

I have also implemented a basic interpreter for testing.

Compiling ATS3 to C is scheduled in the next phase. After that. advanced type-checking

(for dependent types and linear types) needs to be supported.

Before starting the next phase, I plan to write some documentation and hope to get more

people on board. More on that later.

Cheers!

--Hongwei

[gmhwxi]

Here is some code implementing the Sieve of Erastosthenes.

Note that very little type annotation is needed (due to the much stronger support

for type-inference in ATS3 than in ATS2)

(* ****** ****** *)

fun
sieve
( xs
: stream(int)) =
$lazy
(
let
val xs = $eval(xs)
in
case- xs of
|
strmcon_cons(x0, xs) =>
strmcon_cons(x0, sieve(filter(xs, x0)))
end
) (* end of [sieve] *)

and
filter(xs, x0) =
$lazy
(
let
val xs = $eval(xs)
in
case- xs of
|
strmcon_cons(x1, xs) =>
(
  if
  (x1%x0=0)
  then $eval(filter(xs, x0))
  else strmcon_cons(x1, filter(xs, x0))
)
end // end of [filter]                                                                                                                                                                                                                                                        

(* ****** ****** *)

val
xs =
sieve(from(2)) where
{
fun
from
(n:int): stream(int) =
$lazy(strmcon_cons(n, from(n+1)))
}

(* ****** ****** *)

val-
strmcon_cons(x0, xs) = $eval(xs)
val-
strmcon_cons(x1, xs) = $eval(xs)
val-
strmcon_cons(x2, xs) = $eval(xs)
val-
strmcon_cons(x3, xs) = $eval(xs)
val-
strmcon_cons(x4, xs) = $eval(xs)
val-
strmcon_cons(x5, xs) = $eval(xs)

(* ****** ****** *)

(* end of [sieve.dats] *)

[Richard]

Well done! Looking forward to the future progress.

On Sunday, December 15, 2019 at 9:08:21 PM UTC-5, gmhwxi wrote:

[Brandon Barker]

Great to hear about improved type inference! This quite wild to see in comparison to ATS2 code. 

I just hope that everyone remembers to annotate with types at least at the top-level; sometimes Haskellers don't do this and it makes the code more difficult to read, IMO (luckily that is a rarity) ;-)

[Raoul Duke]

my usual usability rant: any language with inference must have a mode that inserts the ascii of what it inferred. 

--
You received this message because you are subscribed to the Google Groups "ats-lang-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ats-lang-user...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/ats-lang-users/cf089bb4-7a5e-4d46-a1c7-589c966ff246%40googlegroups.com.