Awesome
Ntha Programming Language
a tiny statically typed functional programming language.
Installation
- brew install z3
- cabal install ntha
Features
- Global type inference with optional type annotations.
- Lisp flavored syntax with Haskell like semantic inside.
- Support basic types: Integer, Character, String, Boolean, Tuple, List and Record.
- Support unicode keywords.
- Support destructuring.
- ADTs and pattern matching.
- Haskell like type signature for type checking.
- Refined types (still in early stage, just support basic arithmetic operations and propositinal logic, here is some examples), based on z3-encoding
- Module system (still in early stage, lack of namespace control).
- Support pattern matching on function parameters.
- Lambdas and curried function by default.
- Global and Local let binding.
- Recursive functions.
- If-then-else / Cond control flow.
- Type alias.
- Do notation.
- Begin block.
Future Works
- Atoms (need to handle mutable state in evaluation procedure, reference to the implementation of Clea Programming Language).
- error propagation (try / catch).
- Lazyness.
- JIT backend.
- Type-classes (desuger to Records).
- Rank-N types (a naive implementation of First-Class Polymorphism).
- λπ
- Fully type checked lisp like macros (comply with the internal design of Template Haskell).
- TCO.
Screenshot
Example
(type Name String)
(type Env [(Name . Expr)])
(data Op Add Sub Mul Div Less Iff)
(data Expr
(Num Number)
(Bool Boolean)
(Var Name)
(If Expr Expr Expr)
(Let [Char] Expr Expr)
(LetRec Name Expr Expr)
(Lambda Name Expr)
(Closure Expr Env)
(App Expr Expr)
(Binop Op (Expr . Expr)))
(let op-map {:add +
:sub -
:mul *
:div /
:less <
:iff =})
(arith-eval : (α → (β → Z)) → ((α × β) → (Maybe Expr)))
(ƒ arith-eval [fn (v1 . v2)]
(Just (Num (fn v1 v2))))
(logic-eval : (α → (β → B)) → ((α × β) → (Maybe Expr)))
(ƒ logic-eval [fn (v1 . v2)]
(Just (Bool (fn v1 v2))))
(let eval-op
(λ op v1 v2 ⇒
(match (v1 . v2)
(((Just (Num v1)) . (Just (Num v2))) ⇒
(match op
(Add ⇒ (arith-eval (:add op-map) (v1 . v2)))
(Sub ⇒ (arith-eval (:sub op-map) (v1 . v2)))
(Mul ⇒ (arith-eval (:mul op-map) (v1 . v2)))
(Div ⇒ (arith-eval (:div op-map) (v1 . v2)))
(Less ⇒ (logic-eval (:less op-map) (v1 . v2)))
(Iff ⇒ (logic-eval (:iff op-map) (v1 . v2)))))
(_ ⇒ Nothing))))
(eval : [(S × Expr)] → (Expr → (Maybe Expr)))
(ƒ eval [env expr]
(match expr
((Num _) ⇒ (Just expr))
((Bool _) → (Just expr))
((Var x) ⇒ (do Maybe
(val ← (lookup x env))
(return val)))
((If condition consequent alternative) →
(match (eval env condition)
((Just (Bool true)) → (eval env consequent))
((Just (Bool false)) → (eval env alternative))
(_ → (error "condition should be evaluated to a boolean value"))))
((Lambda _ _) → (Just (Closure expr env)))
((App fn arg) → (let [fnv (eval env fn)]
(match fnv
((Just (Closure (Lambda x e) innerenv)) →
(do Maybe
(argv ← (eval env arg))
(eval ((x . argv) :: innerenv) e)))
(_ → (error "should apply arg to a function")))))
((Let x e1 in-e2) ⇒ (do Maybe
(v ← (eval env e1))
(eval ((x . v) :: env) in-e2)))
;; use fix point combinator to approach "Turing-complete"
((LetRec x e1 in-e2) → (eval env (Let "Y" (Lambda "h" (App (Lambda "f" (App (Var "f") (Var "f")))
(Lambda "f" (App (Var "h")
(Lambda "n" (App (App (Var "f") (Var "f"))
(Var "n")))))))
(Let x (App (Var "Y") (Lambda x e1))
in-e2))))
((Binop op (e1 . e2)) => (let [v1 (eval env e1)
v2 (eval env e2)]
(eval-op op v1 v2)))))
(begin
(print "start")
(let result (match (eval [] (LetRec "fact" (Lambda "n" (If (Binop Less ((Var "n") . (Num 2)))
(Num 1)
(Binop Mul ((Var "n") . (App (Var "fact")
(Binop Sub ((Var "n") . (Num 1))))))))
(App (Var "fact") (Num 5))))
((Just (Num num)) ⇒ (print (int2str num)))
(Nothing ⇒ (error "oops"))))
(print result)
(print "finish"))
License
Copyright © 2016 zjhmale