Untyped lambda calculus to Grass compiler.

Grass is an esoteric language based on the untyped lambda calculus, and its program is written by using only three characters: w, W, and v.

npm i -g @susisu/grasspiler

grasspile <srcfile> -o <outfile>

For more information about the CLI, see grasspile -h.

The source language is (of course) based on the untyped lambda calculus, and has a simple ML-like syntax.

The syntax is defined as below (though this is very informal and incomplete):

<program> ::= <definition>+
<definition> ::= let <ident>+ = <term>
<term> ::= <ident>                            variable
           <term> <term>                      application
           λ<ident>+. <term>                  lambda abstraction
           let <ident>+ = <term> in <term>    local binding

A lambda abstraction can be written in several kinds of notation, and also mixture of them.

(* lambda calculus-like *)
let S = λx y z. x z (y z)
(* ML-like *)
let S = fun x y z -> x z (y z)
(* Haskell-like *)
let S = \x y z -> x z (y z)
(* mixture of something *)
let S = fun x y z => x z (y z)
(* Of course you can also write it like this *)
let S x y z = x z (y z)

The primitives Out, Succ, w and In are exposed to the environment, and the value defined at the end of the source program is called by the Grass VM with itself as an argument.

Here is a simple program, which takes a character as input, and output it repeatedly.

(* fixed-point operator *)
let fix = fun f ->
    (fun x -> f (fun y -> x x y))
    (fun x -> f (fun y -> x x y))
(* infinite loop *)
let loop = fix (fun loop c ->
    let _ = Out c in
    loop c
)
let main _ = loop (In w)

Save this source program as resonance.ml (it is not an ML program, though) and compile it with grasspile.

$ grasspile resonance.ml -o resonance.w
$ cat resonance.w
wwWWwwWwwvwwWWWwWWWwvwWWwWWWwwWwwvwwWWWWWWwWWWwwvWWwvwWWWWWWWWWWwwwwwwwwwWWWw

Then you can run the compiled program with a Grass compiler/interpreter (for instance, Grass-JS).

$ grass resonance.w
A
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA...

Stop it if you are satisfied.

MIT License

Susisu (GitHub, Twitter)