FScheme - Scheme in F#

Bài viết
09/24/2010

One of my New Year’s goals is to re-read Lisp in Small Pieces and implement all 11 interpreters and 2 compilers. As much as I like the "Lisp in Lisp" idea and enjoyed the eureka moment in SICP when Sussman writes the metacircular interpreter on the board to the music from Space Odyssey, I don't want to do Lisp in Lisp itself. Lisp in F# sounds like more fun.

As a warm-up, I’m going through Bill Hails’ absolutely awesome book where he builds “PScheme”; a Scheme interpreter in Perl. Of course, doing it in F#. Here’s v0.0.0. It turned out pretty small (about 100 lines):

Tokenizer

I kind of hate these kind of state machine parsers. Later I’ll come back and redo this as parser combinators or lex/yacc, etc. but for now it’s simple enough (just strings, numbers and symbols):

open System open
System.Numerics

let tokenize source =
    let rec string acc = function
        | '\\' :: '"' :: t -> string (acc + "\"") t // escaped quote becomes quote
| '"' :: t -> acc, t // closing quote terminates
| c :: t -> string (acc + (c.ToString())) t // otherwise accumulate chars
        | _ -> failwith "Malformed string."
let rec token acc = function
        | (')' :: _) as t -> acc, t // closing paren terminates
        | w :: t when Char.IsWhiteSpace(w) -> acc, t // whitespace terminates
        | [] -> acc, [] // end of list terminates
        | c :: t -> token (acc + (c.ToString())) t // otherwise accumulate chars
    let rec tokenize' acc = function
        | w :: t when Char.IsWhiteSpace(w) -> tokenize' acc t // skip whitespace
| '(' :: t -> tokenize' (Open :: acc) t
        | ')' :: t -> tokenize' (Close :: acc) t
| '"' :: t -> // start of string
            let s, t' = string "" t
            tokenize' (Token.String(s) :: acc) t'
        | '-' :: d :: t when Char.IsDigit(d) -> // start of negative number
            let n, t' = token ("-" + d.ToString()) t
            tokenize' (Token.Number(n) :: acc) t'
        | '+' :: d :: t | d :: t when Char.IsDigit(d) -> // start of positive number
            let n, t' = token (d.ToString()) t
            tokenize' (Token.Number(n) :: acc) t'
| s :: t -> // otherwise start of symbol
            let s, t' = token (s.ToString()) t
            tokenize' (Token.Symbol(s) :: acc) t'
        | [] -> List.rev acc // end of list terminates
    tokenize' [] source

Parser

The near-equivalence between syntax and semantics in Scheme makes the parser trivial. There’s almost a 1:1 correspondence between tokens and expressions. One exception is the Open/Close tokens denote Expression lists:

let parse source =
    let map = function
        | Token.Number(n) -> Expression.Number(BigInteger.Parse(n))
        | Token.String(s) -> Expression.String(s)
        | Token.Symbol(s) -> Expression.Symbol(s)
        | _ -> failwith "Syntax error."
    let rec parse' acc = function
| Open :: t – >
            let e, t' = parse' [] t
            parse' (List(e) :: acc) t'
| Close :: t -> (List.rev acc), t
        | h :: t -> parse' ((map h) :: acc) t
| [] -> (List.rev acc), []
    let result, _ = parse' [] (tokenize source) result

Printer

For printing expressions to the console:

let rec print = function
    | List(list) -> "(" + String.Join(" ", (List.map print list)) + ")"
    | String(s) – >
        let escape = String.collect (function '"' -> "\\\"" | c -> c.ToString()) // escape quotes
        "\"" + (escape s) + "\""
    | Symbol(s) –> s
    | Number(n) -> n.ToString() | Function(_) | Special(_) -> "Function"

Primitives

We begin with just a few primitives (which are simply Expression –> Expression functions) for doing basic math and conditionals and seed a global environment with them:

let Multiply args =
let prod a = function Number(b) -> a * b | _ -> failwith "Malformed multiplication argument."
Number(List.fold prod 1I args)

let Subtract = function
    | [] -> Number(0I) // (-) == 0
    | [Number(n)] -> Number(-n) // (- a) == –a
    | Number(n) :: ns -> // (- a b c) == a - b – c
        let sub a = function Number(b) -> a - b | _ -> failwith "Malformed subtraction argument."
        Number(List.fold sub n ns) | _ -> failwith "Malformed subtraction."

let rec If = function
    | [condition; t; f] – >
        match eval condition with
        | List([]) | String("") -> eval f // empty list or empty string is false
        | Number(n) when n = 0I -> eval f // zero is false
        | _ -> eval t // everything else is true
    | _ -> failwith "Malformed 'if'."

and environment =
    Map.ofList [
        "*", Function(Multiply)
        "-", Function(Subtract) "if", Special(If)]

Eval/Apply

The classic yin/yang of eval/apply are very easy to implement. Literals eval to themselves, symbols are looked up in the environment, and lists are applied. Special forms are, well, special in that they don’t eval their arguments up front; leaving it up to the callee (e.g. If will eval one of two expressions depending on the conditional).

Notice that If and environment above along with eval and apply below are all mutually recursive:

and eval expression =
    match expression with
    | Number(_) as lit –> lit
    | String(_) as lit –> lit
    | Symbol(s) -> environment.[s]
    | List([]) –
> List([])     | List(h :: t) – >
match eval h with
        | Function(f) -> apply f t
        | Special(f) -> f t
        | _ -> failwith "Malformed expression."
    | _ -> failwith "Malformed expression."

and apply fn args = fn (List.map eval args)

REPL

The REPL now just reads a line from the console, parses, evals and prints it, then rinses and repeats.

let rep = List.ofSeq >> parse >> List.head >> eval >> print

let rec repl output =
printf "%s\n> " output try
Console.ReadLine() |> rep |> repl with ex -> repl ex.Message

The whole deal is kicked off with:

repl "Welcome to FScheme"

Tests

Gotta have tests:

let test () =
    let case source expected =
try
            let output = rep source
            if output <> expected then
                printf "TEST FAILED: %s (Expected: %s, Actual: %s)" source expected output
        with _ -> printf "TEST CRASHED: %s" source
    case "1" "1" // numbers
    case "+1" "1" // explicit positive numbers
    case "-1" "-1" // negative numbers
    case "\"hello\"" "\"hello\"" // strings
    case "(*)" "1" // multiplication
    case "(* 2 3)" "6" // multiplication
    case "(* 2 3 4)" "24" // multiplication
    case "(-)" "0" // strange subtraction case
    case "(- 10)" "-10" // negation
    case "(- 10 2)" "8" // subtraction
    case "(- 10 2 3)" "5" // subtraction
    case "(if (* 0 1) 10 20)" "20" // if
    case "(if (* 1 1) 10 20)" "10" // if
    case "(if (* 1 1) 10 bomb)" "10" // if (special form)
    case "(* 1234567890987654321 1234567890987654321)" "1524157877457704723228166437789971041" // bigint math

Next>

Comments

Anonymous
January 17, 2010
HiThis is great stuff! I know another person looking to do this (to learn more F#), and it will provide good insight.One question:After finishing LiSP, do you intend to write a proper R5RS Scheme in F#? Or maybe a statically typed one?Looking forward to more updates!Cheersleppie
Anonymous
January 17, 2010
The comment has been removed
Anonymous
January 18, 2011
Hi!I'm just taking my first steps in F# and I love you're examples. Now, one thing I don't quite get is the inner signature for the tokenizer' function:let rec tokenize' acc = functionWhen you call this function from the outer function, the seed-call, you do so with :tokenize' [] sourcethat is, you are calling it by supplying two arguments, two lists. But to me, the signature seems only to allow for one argument, the "acc" argument, that is, it only takes one list. Could you explain how it's implied that it actually takes two lists?Cheers!
Anonymous
February 03, 2011
The comment has been removed
Anonymous
July 14, 2012
I am having trouble building this in VS2012, ,.......are there some compatibility issues?
Anonymous
August 01, 2012
@Will, please try the latest here: github.com/.../FSchemeI just tested in VS 2012 RC.
Anonymous
July 25, 2014
that github link is dead for methis works github.com/.../FScheme

Chia sẻ qua