This cheatsheet aims to succinctly cover the most important aspects of F# 7.0.
The Microsoft F# Documentation is complete and authoritative and has received a lot of love in recent years; it's well worth the time investment to read. Only after you've got the lowdown here of course ;)
If you have any comments, corrections, or suggested additions, please open an issue or send a pull request to https://github.com/fsprojects/fsharp-cheatsheet. Questions are best addressed via the F# slack or the F# discord.
Block comments are placed between (* and *). Line comments start from // and continue until the end of the line.
(* This is block comment *)
// And this is line comment
XML doc comments come after /// allowing us to use XML tags to generate documentation.
/// The `let` keyword defines an (immutable) value
let result = 1 + 1 = 2
F# string type is an alias for System.String type.
// Create a string using string concatenation
let hello = "Hello" + " World"
Use verbatim strings preceded by @ symbol to avoid escaping control characters (except escaping " by "").
let verbatimXml = @"<book title=""Paradise Lost"">"
We don't even have to escape " with triple-quoted strings.
let tripleXml = """<book title="Paradise Lost">"""
Backslash strings indent string contents by stripping leading spaces.
let poem =
"The lesser world was daubed\n\
By a colorist of modest skill\n\
A master limned you in the finest inks\n\
And with a fresh-cut quill."
String Slicing is supported by using [start..end] syntax.
let str = "Hello World"
let firstWord = str[0..4] // "Hello"
let lastWord = str[6..] // "World"
String Interpolation is supported by prefixing the string with $ symbol. All of these will output "Hello" \ World!:
let expr = "Hello"
printfn " \"%s\" \\ World! " expr
printfn $" \"{expr}\" \\ World! "
printfn $" \"%s{expr}\" \\ World! " // using a format specifier
printfn $@" ""{expr}"" \ World! "
printfn $@" ""%s{expr}"" \ World! "
See Strings (MS Learn) for more on escape characters, byte arrays, and format specifiers.
Integer Prefixes for hexadecimal, octal, or binary
let numbers = (0x9F, 0o77, 0b1010) // (159, 63, 10)
Literal Type Suffixes for integers, floats, decimals, and ascii arrays
let ( sbyte, byte ) = ( 55y, 55uy ) // 8-bit integer
let ( short, ushort ) = ( 50s, 50us ) // 16-bit integer
let ( int, uint ) = ( 50, 50u ) // 32-bit integer
let ( long, ulong ) = ( 50L, 50uL ) // 64-bit integer
let bigInt = 9999999999999I // System.Numerics.BigInteger
let float = 50.0f // signed 32-bit float
let double = 50.0 // signed 64-bit float
let scientific = 2.3E+32 // signed 64-bit float
let decimal = 50.0m // signed 128-bit decimal
let byte = 'a'B // ascii character; 97uy
let byteArray = "text"B // ascii string; [|116uy; 101uy; 120uy; 116uy|]
Primes (or a tick ' at the end of a label name) are idiomatic to functional languages and are included in F#. They are part of the identifier's name and simply indicate to the developer a variation of an existing value or function. For example:
let x = 5
let x' = x + 1
let x'' = x' + 1
See Literals (MS Learn) for complete reference.
The let keyword also defines named functions.
let negate x = x * -1
let square x = x * x
let print x = printfn "The number is: %d" x
Pipe operator |> is used to chain functions and arguments together. Double-backtick identifiers are handy to improve readability especially in unit testing:
let ``square, negate, then print`` x =
x |> square |> negate |> print
This operator can assist the F# type checker by providing type information before use:
let sumOfLengths (xs : string []) =
xs
|> Array.map (fun s -> s.Length)
|> Array.sum
Composition operator >> is used to compose functions:
let squareNegateThenPrint' =
square >> negate >> print
The rec keyword is used together with the let keyword to define a recursive function:
let rec fact x =
if x < 1 then 1
else x * fact (x - 1)
Mutually recursive functions (those functions which call each other) are indicated by and keyword:
let rec even x =
if x = 0 then true
else odd (x - 1)
and odd x =
if x = 0 then false
else even (x - 1)
Pattern matching is often facilitated through match keyword.
let rec fib n =
match n with
| 0 -> 0
| 1 -> 1
| _ -> fib (n - 1) + fib (n - 2)
In order to match sophisticated inputs, one can use when to create filters or guards on patterns:
let sign x =
match x with
| 0 -> 0
| x when x < 0 -> -1
| x -> 1
Pattern matching can be done directly on arguments:
let fst' (x, _) = x
or implicitly via function keyword:
/// Similar to `fib`; using `function` for pattern matching
let rec fib' = function
| 0 -> 0
| 1 -> 1
| n -> fib' (n - 1) + fib' (n - 2)
For more complete reference visit Pattern Matching (MS Learn).
A list is an immutable collection of elements of the same type.
// Lists use square brackets and `;` delimiter
let list1 = [ "a"; "b" ]
// :: is prepending
let list2 = "c" :: list1
// @ is concat
let list3 = list1 @ list2
// Recursion on list using (::) operator
let rec sum list =
match list with
| [] -> 0
| x :: xs -> x + sum xs
Arrays are fixed-size, zero-based, mutable collections of consecutive data elements.
// Arrays use square brackets with bar
let array1 = [| "a"; "b" |]
// Indexed access using dot
let first = array1.[0]
A sequence is a logical series of elements of the same type. Individual sequence elements are computed only as required, so a sequence can provide better performance than a list in situations in which not all the elements are used.
// Sequences can use yield and contain subsequences
let seq1 =
seq {
// "yield" adds one element
yield 1
yield 2
// "yield!" adds a whole subsequence
yield! [5..10]
}
The same list [ 1; 3; 5; 7; 9 ] or array [| 1; 3; 5; 7; 9 |] can be generated in various ways.
Using range operator ..
let xs = [ 1..2..9 ]
Using list or array comprehensions
let ys = [| for i in 0..4 -> 2 * i + 1 |]
Using init function
let zs = List.init 5 (fun i -> 2 * i + 1)
Lists and arrays have comprehensive sets of higher-order functions for manipulation.
fold starts from the left of the list (or array) and foldBack goes in the opposite direction
let xs' =
Array.fold (fun str n ->
sprintf "%s,%i" str n) "" [| 0..9 |]
reduce doesn't require an initial accumulator
let last xs = List.reduce (fun acc x -> x) xs
map transforms every element of the list (or array)
let ys' = Array.map (fun x -> x * x) [| 0..9 |]
iterate through a list and produce side effects
let _ = List.iter (printfn "%i") [ 0..9 ]
All these operations are also available for sequences. The added benefits of sequences are laziness and uniform treatment of all collections implementing IEnumerable<'T>.
let zs' =
seq {
for i in 0..9 do
printfn "Adding %d" i
yield i
}
A tuple is a grouping of unnamed but ordered values, possibly of different types:
// Tuple construction
let x = (1, "Hello")
// Triple
let y = ("one", "two", "three")
// Tuple deconstruction / pattern
let (a', b') = x
The first and second elements of a tuple can be obtained using fst, snd, or pattern matching:
let c' = fst (1, 2)
let d' = snd (1, 2)
let print' tuple =
match tuple with
| (a, b) -> printfn "Pair %A %A" a b
Records represent simple aggregates of named values, optionally with members:
// Declare a record type
type Person = { Name : string; Age : int }
// Create a value via record expression
let paul = { Name = "Paul"; Age = 28 }
// 'Copy and update' record expression
let paulsTwin = { paul with Name = "Jim" }
Records can be augmented with properties and methods:
type Person with
member x.Info = (x.Name, x.Age)
Records are essentially sealed classes with extra topping: default immutability, structural equality, and pattern matching support.
let isPaul person =
match person with
| { Name = "Paul" } -> true
| _ -> false
Discriminated unions (DU) provide support for values that can be one of a number of named cases, each possibly with different values and types.
type Tree<'T> =
| Node of Tree<'T> * 'T * Tree<'T>
| Leaf
let rec depth = function
| Node(l, _, r) -> 1 + max (depth l) (depth r)
| Leaf -> 0
F# Core has built-in discriminated unions for error handling, e.g., option and Result.
let optionPatternMatch input =
match input with
| Some i -> printfn "input is an int=%d" i
| None -> printfn "input is missing"
Single-case discriminated unions are often used to create type-safe abstractions with pattern matching support:
type OrderId = Order of string
// Create a DU value
let orderId = Order "12"
// Use pattern matching to deconstruct single-case DU
let (Order id) = orderId
A statically resolved type parameter is a type parameter that is replaced with an actual type at compile time instead of at run time. They are primarily useful in conjunction with member constraints.
let inline add x y = x + y
let integerAdd = add 1 2
let floatAdd = add 1.0f 2.0f // without `inline` on `add` function, this would cause a type error
type RequestA = { Id: string; StringValue: string }
type RequestB = { Id: string; IntValue: int }
let requestA: RequestA = { Id = "A"; StringValue = "Value" }
let requestB: RequestB = { Id = "B"; IntValue = 42 }
let inline getId<'t when 't : (member Id: string)> (x: 't) = x.Id
let idA = getId requestA // "A"
let idB = getId requestB // "B"
See Statically Resolved Type Parameters (MS Learn) and Constraints (MS Learn) for more examples.
An illustrative example with: custom F# exception creation, all exception aliases, raise() usage, and an exhaustive demonstration of the exception handler patterns:
open System
exception MyException of int * string // (1)
let guard = true
try
failwith "Message" // throws a System.Exception (aka exn)
nullArg "ArgumentName" // throws a System.ArgumentNullException
invalidArg "ArgumentName" "Message" // throws a System.ArgumentException
invalidOp "Message" // throws a System.InvalidOperation
raise(NotImplementedException("Message")) // throws a .NET exception (2)
raise(MyException(0, "Message")) // throws an F# exception (2)
true // (3)
with
| :? ArgumentNullException -> printfn "NullException"; false // (3)
| :? ArgumentException as ex -> printfn $"{ex.Message}"; false // (4)
| :? InvalidOperationException as ex when guard -> printfn $"{ex.Message}"; reraise() // (5,6)
| MyException(num, str) when guard -> printfn $"{num}, {str}"; false // (5)
| MyException(num, str) -> printfn $"{num}, {str}"; reraise() // (6)
| ex when guard -> printfn $"{ex.Message}"; false
| ex -> printfn $"{ex.Message}"; false
(1) define your own F# exception types with exception, a new type that will inherit from System.Exception;
(2) use raise() to throw an F# or .NET exception;
(3) the entire try..with expression must evaluate to the same type, in this example: bool;
(4)ArgumentNullException inherits from ArgumentException, so ArgumentException must follow after;
(5) support for when guards;
(6) use reraise() to re-throw an exception; works with both .NET and F# exceptions
The difference between F# and .NET exceptions is how they are created and how they can be handled.
The try..finally expression enables you to execute clean-up code even if a block of code throws an exception. Here's an example that also defines custom exceptions.
exception InnerError of string
exception OuterError of string
let handleErrors x y =
try
try
if x = y then raise (InnerError("inner"))
else raise (OuterError("outer"))
with
| InnerError str -> printfn "Error1 %s" str
finally
printfn "Always print this."
Note that finally does not follow with. try..with and try..finally are separate expressions.
This example is a basic class with (1) local let bindings, (2) properties, (3) methods, and (4) static members.
type Vector(x : float, y : float) =
let mag = sqrt(x * x + y * y) // (1)
member _.X = x // (2)
member _.Y = y
member _.Mag = mag
member _.Scale(s) = // (3)
Vector(x * s, y * s)
static member (+) (a : Vector, b : Vector) = // (4)
Vector(a.X + b.X, a.Y + b.Y)
Call a base class from a derived one.
type Animal() =
member _.Rest() = ()
type Dog() =
inherit Animal()
member _.Run() =
base.Rest()
Upcasting is denoted by :> operator.
let dog = Dog()
let animal = dog :> Animal
Dynamic downcasting (:?>) might throw an InvalidCastException if the cast doesn't succeed at runtime.
let shouldBeADog = animal :?> Dog
Declare IVector interface and implement it in Vector'.
type IVector =
abstract Scale : float -> IVector
type Vector'(x, y) =
interface IVector with
member __.Scale(s) =
Vector'(x * s, y * s) :> IVector
member _.X = x
member _.Y = y
Another way of implementing interfaces is to use object expressions.
type ICustomer =
abstract Name : string
abstract Age : int
let createCustomer name age =
{ new ICustomer with
member __.Name = name
member __.Age = age }
// Basic
let (|EmailDomain|) email =
let match' = Regex.Match(email, "@(.*)$")
if match'.Success
then match'.Groups[1].ToString()
else ""
let (EmailDomain emailDomain) = "yennefer@aretuza.org" // emailDomain = 'aretuza.org'
// As Parameters
open System.Numerics
let (|Real|) (x: Complex) =
(x.Real, x.Imaginary)
let addReal (Real (real1, _)) (Real (real2, _)) = // conversion done in the parameters
real1 + real2
let addRealOut = addReal Complex.ImaginaryOne Complex.ImaginaryOne
// Parameterized
let (|Default|) onNone value =
match value with
| None -> onNone
| Some e -> e
let (Default "random citizen" name) = None // name = "random citizen"
let (Default "random citizen" name) = Some "Steve" // name = "Steve"
Single-case active patterns can be thought of as a simple way to convert data to a new form.
let (|Even|Odd|) i =
if i % 2 = 0 then Even else Odd
let testNumber i =
match i with
| Even -> printfn "%d is even" i
| Odd -> printfn "%d is odd" i
let (|Phone|Email|) (s:string) =
if s.Contains '@' then Email $"Email: {s}" else Phone $"Phone: {s}"
match "yennefer@aretuza.org" with // output: "Email: yennefer@aretuza.org"
| Email email -> printfn $"{email}"
| Phone phone -> printfn $"{phone}"
let (|DivisibleBy|_|) by n =
if n % by = 0 then Some DivisibleBy else None
let fizzBuzz = function
| DivisibleBy 3 & DivisibleBy 5 -> "FizzBuzz"
| DivisibleBy 3 -> "Fizz"
| DivisibleBy 5 -> "Buzz"
| i -> string i
Partial active patterns share the syntax of parameterized patterns but their active recognizers accept only one argument.
Modules are key building blocks for grouping related code; they can contain types, let bindings, or (nested) sub modules.
Identifiers within modules can be referenced using dot notation, or you can bring them into scope via the open keyword.
Illustrative-only example:
module Money =
type CardInfo =
{ number: string
expiration: int * int }
type Payment =
| Card of CardInfo
| Cash of int
module Functions =
let validCard (cardNumber: string) =
cardNumber.Length = 16 && (cardNumber[0], ['3';'4';'5';'6']) ||> List.contains
If there is only one module in a file, the module name can be declared at the top, and all code constructs
within the file will be included in the modules definition (no indentation required).
module Functions // notice there is no '=' when at the top of a file
let sumOfSquares n = seq {1..n} |> Seq.sumBy (fun x -> x * x) // Functions.sumOfSquares
Namespaces are simply dotted names that prefix type and module declarations to allow for hierarchical scoping.
The first namespace directives must be placed at the top of the file. Subsequent namespace directives either:
(a) create a sub-namespace; or (b) create a new namespace.
namespace MyNamespace
module MyModule = // MyNamspace.MyModule
let myLet = ... // MyNamspace.MyModule.myLet
namespace MyNamespace.SubNamespace
namespace MyNewNamespace // a new namespace
A top-level module's namespace can be specified via a dotted prefix:
module MyNamespace.SubNamespace.Functions
The open keyword can be used on module, namespace, and type.
module Groceries =
type Fruit =
| Apple
| Banana
let fruit1 = Groceries.Banana
open Groceries // module
let fruit2 = Apple
open System.Diagnostics // namespace
let stopwatch = Stopwatch.StartNew() // Stopwatch is accessible
open type System.Text.RegularExpressions.Regex // type
let isHttp url = IsMatch("^https?:", url) // Regex.IsMatch directly accessible
Available to module declarations only, is the AutoOpen attribute, which alleviates the need for an open.
[<AutoOpen>]
module Groceries =
type Fruit =
| Apple
| Banana
let fruit = Banana
However, AutoOpen should be used cautiously. When an open or AutoOpen is used, all declarations in the containing element
will be brought into scope. This can lead to shadowing; where the last
named declaration replaces all prior identically-named declarations. There is no error - or even a warning - in F#, when shadowing occurs.
A coding convention (MS Learn) exists for open
statements to avoid pitfalls; AutoOpen would sidestep this.
F# supports public, private (limiting access to its containing type or module) and internal (limiting access to its containing assembly).
They can be applied to module, let, member, type, new (MS Learn), and val (MS Learn).
With the exception of let bindings in a class type, everything defaults to public.
| Element | Example with Modifier |
|---|---|
| Module | module internal MyModule = |
Module .. let |
let private value = |
| Record | type internal MyRecord = { id: int } |
| Record ctor | type MyRecord = private { id: int } |
| Discriminated Union | type internal MyDiscUni = A \| B |
| Discriminated Union ctor | type MyDiscUni = private A \| B |
| Class | type internal MyClass() = |
| Class ctor | type MyClass private () = |
| Class Additional ctor | internal new() = MyClass("defaultValue") |
Class .. let |
Always private. Cannot be overridden |
type .. member |
member private _.TypeMember = |
type .. val |
val internal explicitInt : int |
Making a primary constructor (ctor) private or internal is a common convention for ensuring value integrity;
otherwise known as "making illegal states unrepresentable" (YouTube:Effective ML).
Example of Single-case Discriminated Union with a private constructor that constrains a quantity between 0 and 100:
type UnitQuantity =
private UnitQuantity of int
module UnitQuantity = // common idiom: type companion module
let tryCreate qty =
if qty < 1 || qty > 100
then None
else Some (UnitQuantity qty)
let value (UnitQuantity uQty) = uQty
let zero = UnitQuantity 0
...
let unitQtyOpt = UnitQuantity.tryCreate 5
let validQty =
unitQtyOpt
|> Option.defaultValue UnitQuantity.zero
F#'s type inference and name resolution runs in file and line order. By default, any forward references are considered errors.
This default provides a single benefit, which can be hard to appreciate initially: you never need to look beyond the current file for a dependency.
In general this also nudges toward more careful design and organisation of codebases,
which results in cleaner, maintainable code. However, in rare cases forward referencing might be needed.
To do this we have rec for module and namespace; and and for type and let (Recursive Functions) functions.
module rec CarModule
exception OutOfGasException of Car // Car not defined yet; would be an error
type Car =
{ make: string; model: string; hasGas: bool }
member self.Drive destination =
if not self.hasGas
then raise (OutOfGasException self)
else ...
type Person =
{ Name: string; Address: Address }
and Address =
{ Line1: string; Line2: string; Occupant: Person }
See Namespaces (MS Learn) and Modules (MS Learn) to learn more.
Load another F# source file into FSI.
#load "../lib/StringParsing.fs"
Reference a .NET assembly (/ symbol is recommended for Mono compatibility).
Reference a .NET assembly:
#r "../lib/FSharp.Markdown.dll"
Reference a nuget package
#r "nuget:Serilog.Sinks.Console" // latest production release
#r "nuget:FSharp.Data, 6.3.0" // specific version
#r "nuget:Equinox, *-*" // latest version, including `-alpha`, `-rc` version etc
Include a directory in assembly search paths.
#I "../lib"
#r "FSharp.Markdown.dll"
Other important directives are conditional execution in FSI (INTERACTIVE) and querying current directory (__SOURCE_DIRECTORY__).
#if INTERACTIVE
let path = __SOURCE_DIRECTORY__ + "../lib"
#else
let path = "../../../lib"
#endif