How to write a arithmetic expression using a discriminated union in F#? - f#

Recently, I started learning F# and I am a bit struggling with discriminated unions and function signatures.
I am trying to define an arithmetic expression (for sum, product, multiply, average etc) as a discriminate union and write a function that evaluates it.
However, I can't figure out what I am doing wrong. Could anyone point me in the right direction? This what I've tried so far:
Attempt 1
type Expr =
| Sum of int * int
| Avg of int * int
| Mul of int * int
let Evaluate (input : Expr) =
match input with
| Sum(a,b) -> a + b
printfn "%A" (Sum(5,10))
My output is :
Sum (5,10)
Attempt 2
I also tried something like :
type Expr = int -> int -> int
let evaluate (a : Expr) (b : Expr) =
match a,b with
| a,b -> (fun a -> a + b)
printfn "%A" (evaluate 5 10)
since all common arithmetic expressions (like sum, product, multiply, average etc) take two integer inputs and output a single integer.
I am getting errors for: 'This expression was expected to have type Expr but here has type Int'.
edit 1
let evaluate (input : Expr) =
match input with
| Sum(a,b) -> a + b
| Avg(a,b) -> a + b / 2
| Mul(a,b) -> a * b
let evaluate = function
| Sum(a,b) -> a + b
| Avg(a,b) -> a + b / 2
| Mul(a,b) -> a * b

Your first attempt comes close.
type Expr =
| Sum of int * int
| Avg of int * int
| Mul of int * int
let evaluate = function
| Sum(a,b) -> a + b
| Avg(a,b) -> a + b / 2
| Mul(a,b) -> a * b
As commented you left out the evaluation of the expression.
Also there's no reason for using printfn "%A", since we know the return type.
Sum(5,10) // Expr
|> evaluate // int
|> printfn "%i" // unit
In your second attempt you're mixing things up a bit.
type Expr = int -> int -> int
As Lee (again) correctly points out, this signature represents a function with 2 arguments.
let add : Expr = fun a b -> a + b
Or shorthand
let add : Expr = (+)
As Expr represents an arithmetic operator, following evaluate function combines them.
let evaluate (a : Expr) (b : Expr) =
match a,b with
| a,b -> (fun a -> a + b)
If you're intent was to sum two input integers, a and b should've been of type int.
If you want to combine multiple expressions, FuneSnabel's proposal is the way to go.
For example: (1 + 2) * (3 + 4)
type Expr =
| Cte of int
| Add of Expr * Expr
| Mul of Expr * Expr
let rec eval = function
| Cte(i) -> i
| Add(a,b) -> eval a + eval b
| Mul(a,b) -> eval a * eval b
Mul( Add(Cte 1,Cte 2) , Add(Cte 3,Cte 4) )
|> eval
|> printfn "%i"

Attempt 1 looks about right but limited as it doesn't allow you to create expressions like 1+2+3.
Instead you could try something like this:
type Expr =
| Constant of int
| Add of Expr*Expr
| Sub of Expr*Expr
| Mul of Expr*Expr
| Div of Expr*Expr
A bit more flexible and with support for bindings like x, y and so on:
type Expr =
| Constant of int
| Binding of string
| BinaryOp of string*(int -> int -> int)*Expr*Expr
| UnaryOp of string*(int -> int)*Expr

Related

Can one set default values for Discriminated Union types?

I implemented a Discriminated Union type that would be used to select a function:
type BooleanCombinator =
| All
| Some
| None
| AtLeast of int
| MoreThan of int
| NotMoreThan of int
| LessThan of int
| ExactlyOne
| ExactlyTwo
| AllButOne
| AllButTwo
let boolToInt (b: bool) : int = if b then 1 else 0
let combineBooleans (combinator : BooleanCombinator)
(bools : bool list)
: bool =
let n = List.sumBy boolToInt bools
match combinator with
| BooleanCombinator.All -> List.forall id bools
| BooleanCombinator.Some -> bools |> List.exists id
| BooleanCombinator.None -> bools |> List.exists id |> not
| BooleanCombinator.AtLeast i -> n >= i
| BooleanCombinator.MoreThan i -> n > i
| BooleanCombinator.NotMoreThan i -> n <= i
| BooleanCombinator.LessThan i -> n < i
| BooleanCombinator.ExactlyOne -> n = 1
| BooleanCombinator.ExactlyTwo -> n = 2
| BooleanCombinator.AllButOne -> n = bools.Length - 1
| BooleanCombinator.AllButTwo -> n = bools.Length - 2
This looked Ok to me but the compiler started to look at all instances of Some and None as belonging to this DU, instead of the Option DU.
I do not want to go through all of my code replacing Some with Option.Some and None with Option.None.
Is there a way to tell the compiler that unqualified Some and None are actually Option.Some and Option.None?
Or should I just give different names to these DU cases, like AtLeastOne and ExactlyZero
The general rule for resolving name collisions in F# is "last declaration wins". Because your custom DU is declared after Option, its constructors Some and None win over those of Option.
But this rule offers a way to fix the problem: you just need to "reassert" the declarations after your custom DU:
type Bogus = Some of int | None
let g = function Some _ -> 42 | None -> 5
let x = Some 42
let inline Some a = Option.Some a
let inline None<'a> = Option.None : 'a option
let (|Some|None|) = function | Option.Some a -> Some a | Option.None -> None
let f = function Some _ -> 42 | None -> 5
let y = Some 42
If you inspect the types of g, x, f, and y in the above code:
> g
g : Bogus -> int
> f
f : 'a option -> int
> x
Bogus
> y
int option
The function g and value x were inferred to have type Bogus -> int and Bogus respectively, because Some and None in their bodies refer to Bogus.Some and Bogus.None.
The function f and value y were inferred to have Option-related types, because Some and None in their bodies refer to the Some function and the (|Some|None|) active pattern that I defined just above.
Of course, this is a rather hacky way to restore status quo. This will convince the compiler, but humans will still have a hard time reading your code. I suggest you rename the cases of your DU instead.
You can mark your DU with [<RequireQualifiedAccess>] attribute.
This means that you will be required to qualify the case name with the type whenever you use it in the code - which is something you do now anyway in your match expression.
That way an unqualified Some would still be resolved to mean Option.Some, despite the fact that you reuse the name.
It's a useful technique to know when you want to use a snappy name for a DU case - like None, Yes, Failure etc. - that by itself would be ambiguous or confusing to the reader (or the compiler, for that matter).

Simplifying nested pattern matching F#

I am writing a simple expressions parser in F# and for each operator I want only to support a certain number of operands (e.g. two for Modulo, three for If). Here is what I have:
type Operator =
| Modulo
| Equals
| If
let processOperator operands operator =
match operator with
| Modulo ->
match operands with
| [ a:string; b:string ] -> (Convert.ToInt32(a) % Convert.ToInt32(b)).ToString()
| _ -> failwith "wrong number of operands"
| Equals ->
match operands with
| [ a; b ] -> (a = b).ToString()
| _ -> failwith "wrong operands"
| If ->
match operands with
| [ a; b; c ] -> (if Convert.ToBoolean(a) then b else c).ToString()
| _ -> failwith "wrong operands"
I would like to get rid of or simplify the inner list matches. What is the best way to accomplish this? Should I use multiple guards ?
open System
type Operator =
| Modulo
| Equals
| If
let processOperator operands operator =
match (operator, operands) with
| Modulo, [a: string; b] -> string ((int a) % (int b))
| Equals, [a; b] -> string (a = b)
| If, [a; b; c] -> if Convert.ToBoolean(a) then b else c
| _ -> failwith "wrong number of operands"
But I would suggest to move this logic of the operands to the parser, this way you get a clean operator expression, which is more idiomatic and straight forward to process, at the end you'll have something like this:
open System
type Operator =
| Modulo of int * int
| Equals of int * int
| If of bool * string * string
let processOperator = function
| Modulo (a, b) -> string (a % b)
| Equals (a, b) -> string (a = b)
| If (a, b, c) -> if a then b else c
Fold in the operands matching:
let processOperator operands operator =
match operator, operands with
| Modulo, [a; b] -> (Convert.ToInt32(a) % Convert.ToInt32(b)).ToString()
| Equals, [a; b] -> (a = b).ToString()
| If, [ a; b; c ] -> (if Convert.ToBoolean(a) then b else c).ToString()
| _ -> failwith "wrong number of operands"
Better yet, if you can, change the datatype to the following.
type Operator =
| Modulo of string * string
| Equals of string * string
| If of string * string * string
Then in the match, you can no longer fail.

Incomplete match with AND patterns

I've defined an expression tree structure in F# as follows:
type Num = int
type Name = string
type Expr =
| Con of Num
| Var of Name
| Add of Expr * Expr
| Sub of Expr * Expr
| Mult of Expr * Expr
| Div of Expr * Expr
| Pow of Expr * Expr
| Neg of Expr
I wanted to be able to pretty-print the expression tree so I did the following:
let (|Unary|Binary|Terminal|) expr =
match expr with
| Add(x, y) -> Binary(x, y)
| Sub(x, y) -> Binary(x, y)
| Mult(x, y) -> Binary(x, y)
| Div(x, y) -> Binary(x, y)
| Pow(x, y) -> Binary(x, y)
| Neg(x) -> Unary(x)
| Con(x) -> Terminal(box x)
| Var(x) -> Terminal(box x)
let operator expr =
match expr with
| Add(_) -> "+"
| Sub(_) | Neg(_) -> "-"
| Mult(_) -> "*"
| Div(_) -> "/"
| Pow(_) -> "**"
| _ -> failwith "There is no operator for the given expression."
let rec format expr =
match expr with
| Unary(x) -> sprintf "%s(%s)" (operator expr) (format x)
| Binary(x, y) -> sprintf "(%s %s %s)" (format x) (operator expr) (format y)
| Terminal(x) -> string x
However, I don't really like the failwith approach for the operator function since it's not compile-time safe. So I rewrote it as an active pattern:
let (|Operator|_|) expr =
match expr with
| Add(_) -> Some "+"
| Sub(_) | Neg(_) -> Some "-"
| Mult(_) -> Some "*"
| Div(_) -> Some "/"
| Pow(_) -> Some "**"
| _ -> None
Now I can rewrite my format function beautifully as follows:
let rec format expr =
match expr with
| Unary(x) & Operator(op) -> sprintf "%s(%s)" op (format x)
| Binary(x, y) & Operator(op) -> sprintf "(%s %s %s)" (format x) op (format y)
| Terminal(x) -> string x
I assumed, since F# is magic, that this would just work. Unfortunately, the compiler then warns me about incomplete pattern matches, because it can't see that anything that matches Unary(x) will also match Operator(op) and anything that matches Binary(x, y) will also match Operator(op). And I consider warnings like that to be as bad as compiler errors.
So my questions are: Is there a specific reason why this doesn't work (like have I left some magical annotation off somewhere or is there something that I'm just not seeing)? Is there a simple workaround I could use to get the type of safety I want? And is there an inherent problem with this type of compile-time checking, or is it something that F# might add in some future release?
If you code the destinction between ground terms and complex terms into the type system, you can avoid the runtime check and make them be complete pattern matches.
type Num = int
type Name = string
type GroundTerm =
| Con of Num
| Var of Name
type ComplexTerm =
| Add of Term * Term
| Sub of Term * Term
| Mult of Term * Term
| Div of Term * Term
| Pow of Term * Term
| Neg of Term
and Term =
| GroundTerm of GroundTerm
| ComplexTerm of ComplexTerm
let (|Operator|) ct =
match ct with
| Add(_) -> "+"
| Sub(_) | Neg(_) -> "-"
| Mult(_) -> "*"
| Div(_) -> "/"
| Pow(_) -> "**"
let (|Unary|Binary|) ct =
match ct with
| Add(x, y) -> Binary(x, y)
| Sub(x, y) -> Binary(x, y)
| Mult(x, y) -> Binary(x, y)
| Div(x, y) -> Binary(x, y)
| Pow(x, y) -> Binary(x, y)
| Neg(x) -> Unary(x)
let (|Terminal|) gt =
match gt with
| Con x -> Terminal(string x)
| Var x -> Terminal(string x)
let rec format expr =
match expr with
| ComplexTerm ct ->
match ct with
| Unary(x) & Operator(op) -> sprintf "%s(%s)" op (format x)
| Binary(x, y) & Operator(op) -> sprintf "(%s %s %s)" (format x) op (format y)
| GroundTerm gt ->
match gt with
| Terminal(x) -> x
also, imo, you should avoid boxing if you want to be type-safe. If you really want both cases, make two pattern. Or, as done here, just make a projection to the type you need later on. This way you avoid the boxing and instead you return what you need for printing.
I think you can make operator a normal function rather than an active pattern. Because operator is just a function which gives you an operator string for an expr, where as unary, binary and terminal are expression types and hence it make sense to pattern match on them.
let operator expr =
match expr with
| Add(_) -> "+"
| Sub(_) | Neg(_) -> "-"
| Mult(_) -> "*"
| Div(_) -> "/"
| Pow(_) -> "**"
| Var(_) | Con(_) -> ""
let rec format expr =
match expr with
| Unary(x) -> sprintf "%s(%s)" (operator expr) (format x)
| Binary(x, y) -> sprintf "(%s %s %s)" (format x) (operator expr) (format y)
| Terminal(x) -> string x
I find the best solution is to restructure your original type defintion:
type UnOp = Neg
type BinOp = Add | Sub | Mul | Div | Pow
type Expr =
| Int of int
| UnOp of UnOp * Expr
| BinOp of BinOp * Expr * Expr
All sorts of functions can then be written over the UnOp and BinOp types including selecting operators. You may even want to split BinOp into arithmetic and comparison operators in the future.
For example, I used this approach in the (non-free) article "Language-oriented programming: The Term-level Interpreter
" (2008) in the F# Journal.

Create discriminated union data from file/database

I have a discriminated union for expressions like this one (EQ =; GT >; etc)
(AND (OR (EQ X 0)
(GT X 10))
(OR (EQ Y 0)
(GT Y 10)))
I want to create instances of DU from such expressions saved in file/database.
How do i do it? If it is not feasible, what is the best way to approach it in F#?
Daniel: these expressions are saved in prefix format (as above) as text and will be parsed in F#. Thanks.
If you just want to know how to model these expressions using DUs, here's one way:
type BinaryOp =
| EQ
| GT
type Expr =
| And of Expr * Expr
| Or of Expr * Expr
| Binary of BinaryOp * Expr * Expr
| Var of string
| Value of obj
let expr =
And(
Or(
Binary(EQ, Var("X"), Value(0)),
Binary(GT, Var("X"), Value(10))),
Or(
Binary(EQ, Var("Y"), Value(0)),
Binary(GT, Var("Y"), Value(10))))
Now, this may be too "loose," i.e., it permits expressions like And(Value(1), Value(2)), which may not be valid according to your grammar. But this should give you an idea of how to approach it.
There are also some good examples in the F# Programming wikibook.
If you need to parse these expressions, I highly recommend FParsec.
Daniel's answer is good. Here's a similar approach, along with a simple top-down parser built with active patterns:
type BinOp = | And | Or
type Comparison = | Gt | Eq
type Expr =
| BinOp of BinOp * Expr * Expr
| Comp of Comparison * string * int
module private Parsing =
// recognize and strip a leading literal
let (|Lit|_|) lit (s:string) =
if s.StartsWith(lit) then Some(s.Substring lit.Length)
else None
// strip leading whitespace
let (|NoWs|) (s:string) =
s.TrimStart(' ', '\t', '\r', '\n')
// parse a binary operator
let (|BinOp|_|) = function
| Lit "AND" r -> Some(And, r)
| Lit "OR" r -> Some(Or, r)
| _ -> None
// parse a comparison operator
let (|Comparison|_|) = function
| Lit "GT" r -> Some(Gt, r)
| Lit "EQ" r -> Some(Eq, r)
| _ -> None
// parse a variable (alphabetical characters only)
let (|Var|_|) s =
let m = System.Text.RegularExpressions.Regex.Match(s, "^[a-zA-Z]+")
if m.Success then
Some(m.Value, s.Substring m.Value.Length)
else
None
// parse an integer
let (|Int|_|) s =
let m = System.Text.RegularExpressions.Regex.Match(s, #"^-?\d+")
if m.Success then
Some(int m.Value, s.Substring m.Value.Length)
else
None
// parse an expression
let rec (|Expr|_|) = function
| NoWs (Lit "(" (BinOp (b, Expr(e1, Expr(e2, Lit ")" rest))))) ->
Some(BinOp(b, e1, e2), rest)
| NoWs (Lit "(" (Comparison (c, NoWs (Var (v, NoWs (Int (i, Lit ")" rest))))))) ->
Some(Comp(c, v, i), rest)
| _ -> None
let parse = function
| Parsing.Expr(e, "") -> e
| s -> failwith (sprintf "Not a valid expression: %s" s)
let e = parse #"
(AND (OR (EQ X 0)
(GT X 10))
(OR (EQ Y 0)
(GT Y 10)))"

Parsing grammars using OCaml

I have a task to write a (toy) parser for a (toy) grammar using OCaml and not sure how to start (and proceed with) this problem.
Here's a sample Awk grammar:
type ('nonterm, 'term) symbol = N of 'nonterm | T of 'term;;
type awksub_nonterminals = Expr | Term | Lvalue | Incrop | Binop | Num;;
let awksub_grammar =
(Expr,
function
| Expr ->
[[N Term; N Binop; N Expr];
[N Term]]
| Term ->
[[N Num];
[N Lvalue];
[N Incrop; N Lvalue];
[N Lvalue; N Incrop];
[T"("; N Expr; T")"]]
| Lvalue ->
[[T"$"; N Expr]]
| Incrop ->
[[T"++"];
[T"--"]]
| Binop ->
[[T"+"];
[T"-"]]
| Num ->
[[T"0"]; [T"1"]; [T"2"]; [T"3"]; [T"4"];
[T"5"]; [T"6"]; [T"7"]; [T"8"]; [T"9"]]);;
And here's some fragments to parse:
let frag1 = ["4"; "+"; "3"];;
let frag2 = ["9"; "+"; "$"; "1"; "+"];;
What I'm looking for is a rulelist that is the result of the parsing a fragment, such as this one for frag1 ["4"; "+"; "3"]:
[(Expr, [N Term; N Binop; N Expr]);
(Term, [N Num]);
(Num, [T "3"]);
(Binop, [T "+"]);
(Expr, [N Term]);
(Term, [N Num]);
(Num, [T "4"])]
The restriction is to not use any OCaml libraries other than List... :/
Here is a rough sketch - straightforwardly descend into the grammar and try each branch in order. Possible optimization : tail recursion for single non-terminal in a branch.
exception Backtrack
let parse l =
let rules = snd awksub_grammar in
let rec descend gram l =
let rec loop = function
| [] -> raise Backtrack
| x::xs -> try attempt x l with Backtrack -> loop xs
in
loop (rules gram)
and attempt branch (path,tokens) =
match branch, tokens with
| T x :: branch' , h::tokens' when h = x ->
attempt branch' ((T x :: path),tokens')
| N n :: branch' , _ ->
let (path',tokens) = descend n ((N n :: path),tokens) in
attempt branch' (path', tokens)
| [], _ -> path,tokens
| _, _ -> raise Backtrack
in
let (path,tail) = descend (fst awksub_grammar) ([],l) in
tail, List.rev path
Ok, so the first think you should do is write a lexical analyser. That's the
function that takes the ‘raw’ input, like ["3"; "-"; "("; "4"; "+"; "2"; ")"],
and splits it into a list of tokens (that is, representations of terminal symbols).
You can define a token to be
type token =
| TokInt of int (* an integer *)
| TokBinOp of binop (* a binary operator *)
| TokOParen (* an opening parenthesis *)
| TokCParen (* a closing parenthesis *)
and binop = Plus | Minus
The type of the lexer function would be string list -> token list and the ouput of
lexer ["3"; "-"; "("; "4"; "+"; "2"; ")"]
would be something like
[ TokInt 3; TokBinOp Minus; TokOParen; TokInt 4;
TBinOp Plus; TokInt 2; TokCParen ]
This will make the job of writing the parser easier, because you won't have to
worry about recognising what is a integer, what is an operator, etc.
This is a first, not too difficult step because the tokens are already separated.
All the lexer has to do is identify them.
When this is done, you can write a more realistic lexical analyser, of type string -> token list, that takes a actual raw input, such as "3-(4+2)" and turns it into a token list.
I'm not sure if you specifically require the derivation tree, or if this is a just a first step in parsing. I'm assuming the latter.
You could start by defining the structure of the resulting abstract syntax tree by defining types. It could be something like this:
type expr =
| Operation of term * binop * term
| Term of term
and term =
| Num of num
| Lvalue of expr
| Incrop of incrop * expression
and incrop = Incr | Decr
and binop = Plus | Minus
and num = int
Then I'd implement a recursive descent parser. Of course it would be much nicer if you could use streams combined with the preprocessor camlp4of...
By the way, there's a small example about arithmetic expressions in the OCaml documentation here.

Resources