A library for function composition.
We work with functions all the time, but function composition is hiding in plain sight!
For instance, we work with functions when we use higher-order methods, like map
on arrays:
[1, 2, 3].map { $0 + 1 }
// [2, 3, 4]
If we wanted to modify this simple closure to square our value after incrementing it, things begin to get messy.
[1, 2, 3].map { ($0 + 1) * ($0 + 1) }
// [4, 9, 16]
Functions allow us to identify and extract reusable code. Let's define a couple functions that make up the behavior above.
func incr(_ x: Int) -> Int {
return x + 1
}
func square(_ x: Int) -> Int {
return x * x
}
With these functions defined, we can pass them directly to map
!
[1, 2, 3]
.map(incr)
.map(square)
// [4, 9, 16]
This refactor reads much better, but it's less performant: we're mapping over the array twice and creating an intermediate copy along the way! While we could use lazy
to fuse these calls together, let's take a more general approach: function composition!
[1, 2, 3].map(pipe(incr, square))
// [4, 9, 16]
The pipe
function glues other functions together! It can take more than two arguments and even change the type along the way!
[1, 2, 3].map(pipe(incr, square, String.init))
// ["4", "9", "16"]
Function composition lets us build new functions from smaller pieces, giving us the ability to extract and reuse logic in other contexts.
let computeAndStringify = pipe(incr, square, String.init)
[1, 2, 3].map(computeAndStringify)
// ["4", "9", "16"]
computeAndStringify(42)
// "1849"
The function is the smallest building block of code. Function composition gives us the ability to fit these blocks together and build entire apps out of small, reusable, understandable units.
The most basic building block in Overture. It takes existing functions and smooshes them together. That is, given a function (A) -> B
and a function (B) -> C
, pipe
will return a brand new (A) -> C
function.
let computeAndStringify = pipe(incr, square, String.init)
computeAndStringify(42)
// "1849"
[1, 2, 3].map(computeAndStringify)
// ["4", "9", "16"]
The with
and update
functions are useful for applying functions to values. They play nicely with the inout
and mutable object worlds, wrapping otherwise imperative configuration statements in an expression.
class MyViewController: UIViewController {
let label = updateObject(UILabel()) {
$0.font = .systemFont(ofSize: 24)
$0.textColor = .red
}
}
And it restores the left-to-right readability we're used to from the method world.
with(42, pipe(incr, square, String.init))
// "1849"
Using an inout
parameter.
update(&user, mut(\.name, "Blob"))
The concat
function composes with single types. This includes composition of the following function signatures:
(A) -> A
(inout A) -> Void
<A: AnyObject>(A) -> Void
With concat
, we can build powerful configuration functions from small pieces.
let roundedStyle: (UIView) -> Void = {
$0.clipsToBounds = true
$0.layer.cornerRadius = 6
}
let baseButtonStyle: (UIButton) -> Void = {
$0.contentEdgeInsets = UIEdgeInsets(top: 12, left: 16, bottom: 12, right: 16)
$0.titleLabel?.font = .systemFont(ofSize: 16, weight: .medium)
}
let roundedButtonStyle = concat(
baseButtonStyle,
roundedStyle
)
let filledButtonStyle = concat(roundedButtonStyle) {
$0.backgroundColor = .black
$0.tintColor = .white
}
let button = with(UIButton(type: .system), filledButtonStyle)
These functions make up the Swiss army knife of composition. They give us the power to take existing functions and methods that don't compose (e.g, those that take zero or multiple arguments) and restore composition.
For example, let's transform a string initializer that takes multiple arguments into something that can compose with pipe
.
String.init(data:encoding:)
// (Data, String.Encoding) -> String?
We use curry
to transform multi-argument functions into functions that take a single input and return new functions to gather more inputs along the way.
curry(String.init(data:encoding:))
// (Data) -> (String.Encoding) -> String?
And we use flip
to flip the order of arguments. Multi-argument functions and methods typically take data first and configuration second, but we can generally apply configuration before we have data, and flip
allows us to do just that.
flip(curry(String.init(data:encoding:)))
// (String.Encoding) -> (Data) -> String?
Now we have a highly-reusable, composable building block that we can use to build pipelines.
let stringWithEncoding = flip(curry(String.init(data:encoding:)))
// (String.Encoding) -> (Data) -> String?
let utf8String = stringWithEncoding(.utf8)
// (Data) -> String?
Swift also exposes methods as static, unbound functions. These functions are already in curried form. All we need to do is flip
them to make them more useful!
String.capitalized
// (String) -> (Locale?) -> String
let capitalized = flip(String.capitalized)
// (Locale?) -> (String) -> String
["hello, world", "and good night"]
.map(capitalized(Locale(identifier: "en")))
// ["Hello, World", "And Good Night"]
And zurry
restores composition for functions and methods that take zero arguments.
String.uppercased
// (String) -> () -> String
flip(String.uppercased)
// () -> (String) -> String
let uppercased = zurry(flip(String.uppercased))
// (String) -> String
["hello, world", "and good night"]
.map(uppercased)
// ["HELLO, WORLD", "AND GOOD NIGHT"]
The get
function produces getter functions from key paths.
get(\String.count)
// (String) -> Int
["hello, world", "and good night"]
.map(get(\.count))
// [12, 14]
We can even compose other functions into get
by using the pipe
function. Here we build a function that increments an integer, squares it, turns it into a string, and then gets the string's character count:
pipe(incr, square, String.init, get(\.count))
// (Int) -> Int
The prop
function produces setter functions from key paths.
let setUserName = prop(\User.name)
// ((String) -> String) -> (User) -> User
let capitalizeUserName = setUserName(capitalized(Locale(identifier: "en")))
// (User) -> User
let setUserAge = prop(\User.age)
let celebrateBirthday = setUserAge(incr)
// (User) -> User
with(User(name: "blob", age: 1), concat(
capitalizeUserName,
celebrateBirthday
))
// User(name: "Blob", age: 2)
The over
and set
functions produce (Root) -> Root
transform functions that work on a Value
in a structure given a key path (or setter function).
The over
function takes a (Value) -> Value
transform function to modify an existing value.
let celebrateBirthday = over(\User.age, incr)
// (User) -> User
The set
function replaces an existing value with a brand new one.
with(user, set(\.name, "Blob"))
The mprop
, mver
and mut
functions are mutable variants of prop
, over
and set
.
let guaranteeHeaders = mver(\URLRequest.allHTTPHeaderFields) { $0 = $0 ?? [:] }
let setHeader = { name, value in
concat(
guaranteeHeaders,
{ $0.allHTTPHeaderFields?[name] = value }
)
}
let request = update(
URLRequest(url: url),
mut(\.httpMethod, "POST"),
setHeader("Authorization", "Token " + token),
setHeader("Content-Type", "application/json; charset=utf-8")
)
This is a function that Swift ships with! Unfortunately, it's limited to pairs of sequences. Overture defines zip
to work with up to ten sequences at once, which makes combining several sets of related data a snap.
let ids = [1, 2, 3]
let emails = ["[email protected]", "[email protected]", "[email protected]"]
let names = ["Blob", "Blob Junior", "Blob Senior"]
zip(ids, emails, names)
// [
// (1, "[email protected]", "Blob"),
// (2, "[email protected]", "Blob Junior"),
// (3, "[email protected]", "Blob Senior")
// ]
It's common to immediately map
on zipped values.
struct User {
let id: Int
let email: String
let name: String
}
zip(ids, emails, names).map(User.init)
// [
// User(id: 1, email: "[email protected]", name: "Blob"),
// User(id: 2, email: "[email protected]", name: "Blob Junior"),
// User(id: 3, email: "[email protected]", name: "Blob Senior")
// ]
Because of this, Overture provides a zip(with:)
helper, which takes a tranform function up front and is curried, so it can be composed with other functions using pipe
.
zip(with: User.init)(ids, emails, names)
Overture also extends the notion of zip
to work with optionals! It's an expressive way of combining multiple optionals together.
let optionalId: Int? = 1
let optionalEmail: String? = "[email protected]"
let optionalName: String? = "Blob"
zip(optionalId, optionalEmail, optionalName)
// Optional<(Int, String, String)>.some((1, "[email protected]", "Blob"))
And zip(with:)
lets us transform these tuples into other values.
zip(with: User.init)(optionalId, optionalEmail, optionalName)
// Optional<User>.some(User(id: 1, email: "[email protected]", name: "Blob"))
Using zip
can be an expressive alternative to let
-unwrapping!
let optionalUser = zip(with: User.init)(optionalId, optionalEmail, optionalName)
// vs.
let optionalUser: User?
if let id = optionalId, let email = optionalEmail, let name = optionalName {
optionalUser = User(id: id, email: email, name: name)
} else {
optionalUser = nil
}
-
Should I be worried about polluting the global namespace with free functions?
Nope! Swift has several layers of scope to help you here.
- You can limit exposing highly-specific functions beyond a single file by using
fileprivate
andprivate
scope. - You can define functions as
static
members inside types. - You can qualify functions with the module's name (e.g.,
Overture.pipe(f, g)
). You can even autocomplete free functions from the module's name, so discoverability doesn't have to suffer!
- You can limit exposing highly-specific functions beyond a single file by using
-
Are free functions that common in Swift?
It may not seem like it, but free functions are everywhere in Swift, making Overture extremely useful! A few examples:
- Initializers, like
String.init
. - Unbound methods, like
String.uppercased
. - Enum cases with associated values, like
Optional.some
. - Ad hoc closures we pass to
map
,filter
, and other higher-order methods. - Top-level Standard Library functions like
max
,min
, andzip
.
- Initializers, like
You can add Overture to an Xcode project by adding it as a package dependency.
If you want to use Overture in a SwiftPM project, it's as simple as adding it to a dependencies
clause in your Package.swift
:
dependencies: [
.package(url: "https://github.com/pointfreeco/swift-overture", from: "0.5.0")
]
This library was created as an alternative to swift-prelude, which is an experimental functional programming library that uses infix operators. For example, pipe
is none other than the arrow composition operator >>>
, which means the following are equivalent:
xs.map(incr >>> square)
xs.map(pipe(incr, square))
We know that many code bases are not going to be comfortable introducing operators, so we wanted to reduce the barrier to entry for embracing function composition.
These concepts (and more) are explored thoroughly in Point-Free, a video series exploring functional programming and Swift hosted by Brandon Williams and Stephen Celis.
The ideas in this episode were first explored in Episode #11:
All modules are released under the MIT license. See LICENSE for details.