recursive descent parser and functional programming

Question

So lately I have been working on writing a simple compiler to better understand compiler concepts. Being a diligent reader of stackoverfolow, it seems there is a consensus that

Answer 1

One strategy for functional parsing is monadic parser combinators. You can read some about it here (and follow links) or use a library like FParsec. I do not recommend this approach if you're just learning/starting F#/compilers, though.

Another approach with F# is to use FsLex/FsYacc (in the PowerPack). I kinda loathe Lex/Yacc technology, so I also don't recommend this.

I think you should write a recursive decent parser by hand. I don't have strong feelings regarding a tokenizer, but simply tokeninize the entire file into a(n immutable) list of tokens and then doing recursive descent (and leveraging some pattern-matching) is a good way to to deal with parsing. And of course, you'll want to use discrimated unions to represent the AST output of the parser (a la here).

I haven't read the dragon book in a long time, but I'm apparently the only person on the planet who doesn't like it. I would consider abandoning that text in favor of a book that discusses compilers using some ML-based language, though I can't recommend one offhand.

EDIT

I haven't done one of these in a while, so I took a few minutes to code a small sample.

// AST for tiny language
type Op = 
    | Plus 
    | Minus 
type Expr = 
    | Literal of int 
    | BinaryOp of Expr * Op * Expr // left, op, right 
type Stmt =
    | IfThenElse of Expr * Stmt * Stmt // cond, then, else; 0=false in cond 
    | Print of Expr

// sample program
let input = @"
    if 1+1-1 then 
        print 42 
    else 
        print 0"

// expected AST
let goal = 
    IfThenElse(
        BinaryOp( BinaryOp(Literal(1),Plus,Literal(1)), Minus, Literal(1)), 
        Print(Literal(42)), 
        Print(Literal(0))) 

////////////////////////////////////////////////////////////////////////////
// Lexer

type Token =
    | IF
    | THEN
    | ELSE
    | PRINT
    | NUM of int  // non-negative
    | PLUS
    | MINUS
    | EOF

let makeTokenizer (s:string) =
    let i = ref 0
    let keywords = [
        "if", IF 
        "then", THEN
        "else", ELSE
        "print", PRINT
        "+", PLUS
        "-", MINUS ]
    let rec getNextToken() =
        if !i >= s.Length then
            EOF
        elif System.Char.IsWhiteSpace(s.[!i]) then
            incr i
            getNextToken()
        elif System.Char.IsDigit(s.[!i]) then
            let mutable j = !i
            while j < s.Length && System.Char.IsDigit(s.[j]) do
                j <- j + 1
            let numStr = s.Substring(!i, j - !i)
            i := j
            NUM(System.Int32.Parse(numStr)) // may throw, e.g. if > MAXINT
        else 
            let keyword = keywords |> List.tryPick (fun (kwStr,kwTok) ->
                if s.IndexOf(kwStr, !i) = !i then
                    i := !i + kwStr.Length
                    Some(kwTok)
                else
                    None)
            match keyword with
            | Some k -> k
            | None -> 
                failwith "unexpected char '%c' at position %d" s.[!i] !i
    getNextToken

let tokens = 
    let nextToken = makeTokenizer input
    let t = ref(nextToken())
    [ 
        yield !t
        while !t <> EOF do
            t := nextToken()
            yield !t
    ]

printfn "%A" tokens // sanity check our tokenizer works

/////////////////////////////////////////////////////////////////////////
// Parser

let parseExpr toks =
    match toks with
    | NUM x :: rest ->
        let mutable rest = rest
        let mutable expr = Literal x
        while rest.Head = PLUS || rest.Head = MINUS do
            let op,y,r = 
                match rest with
                | PLUS::NUM y::t -> Plus, Literal y, t
                | MINUS::NUM y::t -> Minus, Literal y, t
                | _ -> 
                    failwith "parse error in expression, expected number"
            expr <- BinaryOp(expr, op, y)
            rest <- r
        expr, rest
    | _ -> failwith "parse error in expression, expected number"
let rec parseStmt toks =
    match toks with
    | PRINT :: rest -> 
        let e,rest = parseExpr(rest)
        Print(e), rest
    | IF :: rest ->
        let e,rest = parseExpr(rest)
        match rest with
        | THEN :: rest ->
            let s1,rest = parseStmt(rest)
            match rest with
            | ELSE :: rest ->
                let s2,rest = parseStmt(rest)
                IfThenElse(e,s1,s2), rest
            | _ -> 
                failwith "parse error after if branch, espected 'else'"
        | _ -> 
            failwith "parse error after if expression, expected 'then'"
    | _ -> failwith "parse error, expected statement"
let parseProgram toks =
    let s,rest = parseStmt toks
    match rest with
    | [EOF] -> s
    | _ -> failwith "parse error after statement, expected EOF"

let p = parseProgram tokens
printfn "%A" p
assert( p = goal )

(Hopefully there are no egregious bugs.)