Composable, transformable, controllable randomness.

• Motivation
• Examples
• Installation
• Interested in learning more?
• License

Swift's randomness API is powerful and simple to use. It allows us to create random values from many basic types, such as booleans and numeric types, and it allows us to randomly shuffle arrays and pluck random elements from collections.

However, it does not make it easy for us to extend the randomness API, nor does it provide an API that is composable, which would allow us to create complex types of randomness from simpler pieces.

Gen is a lightweight wrapper over Swift's randomness APIs that makes it easy to build custom generators of any kind of value.

Gen's namesake type, Gen, is responsible for producing random values. Most often you will reach for one of the static variables inside Gen to get access to a Gen value:

Gen.bool
// Gen<Bool>

Rather than immediately producing a random value, Gen describes a random value that can be produced by calling its run method:

let myGen = Gen.bool
// Gen<Bool>

myGen.run() // true
myGen.run() // true
myGen.run() // false

Every random function that comes with Swift is also available as a static function on Gen:

The reason it is powerful to wrap randomness in the Gen type is that we can make the Gen type composable. For example, a generator of integers can be turned into a generator of numeric strings with a simple application of the map function:

let digit = Gen.int(in: 0...9)           // Gen<Int>
let stringDigit = digit.map(String.init) // Gen<String>

stringDigit.run() // "7"
stringDigit.run() // "1"
stringDigit.run() // "3"

Already this is a form of randomness that Swift's API's do not provide out of the box.

Gen provides many operators for generating new types of randomness, such as map, flatMap and zip, as well as helper functions for generating random arrays, sets, dictionaries, strings, distributions and more! A random password generator, for example, is just a few operators away.

// Take a generator of random letters and numbers.
let password = Gen.letterOrNumber
  // Generate 6-character strings of them.
  .string(of: .always(6))
  // Generate 3 segments of these strings.
  .array(of: .always(3))
  // And join them.
  .map { $0.joined(separator: "-") }

password.run() // "9BiGYA-fmvsOf-VYDtDv"
password.run() // "dS2MGr-FQSuC4-ZLEicl"
password.run() // "YusZGF-HILrCo-rNGfCA"

This kind of composition makes it simple to generate random values of anything.

// Use `zip` to combine generators together and build structures.

let randomPoint = zip(.int(in: -10...10), .int(in: -10...10))
  .map(CGPoint.init(x:y:))
// Gen<CGPoint>

But composability isn't the only reason the Gen type shines. By delaying the creation of random values until the run method is invoked, we allow ourselves to control randomness in circumstances where we need determinism, such as tests. The run method has an overload that takes a RandomNumberGenerator value, which is Swift's protocol that powers their randomness API. By default it uses the SystemRandomNumberGenerator, which is a good source of randomness, but we can also provide a seedable "pseudo" random number generator, so that we can get predictable results in tests:

var xoshiro = Xoshiro(seed: 0)
Gen.int(in: 0...9).run(using: &xoshiro) // "1"
Gen.int(in: 0...9).run(using: &xoshiro) // "0"
Gen.int(in: 0...9).run(using: &xoshiro) // "4"

xoshiro = Xoshiro(seed: 0)
Gen.int(in: 0...9).run(using: &xoshiro) // "1"
Gen.int(in: 0...9).run(using: &xoshiro) // "0"
Gen.int(in: 0...9).run(using: &xoshiro) // "4"

This means you don't have to sacrifice testability when leveraging randomness in your application.

For more examples of using Gen to build complex randomness, see our blog post on creating a Zalgo generator and our two-part video series (part 1 and part 2) on creating generative art.

If you want to use Gen in a project that uses SwiftPM, it's as simple as adding a dependencies clause to your Package.swift:

dependencies: [
  .package(url: "https://github.com/pointfreeco/swift-gen.git", from: "0.4.0")
]

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 design of this library was explored in the following Point-Free episodes:

• Episode 30: Composable Randomness
• Episode 31: Decodable Randomness: Part 1
• Episode 32: Decodable Randomness: Part 2
• Episode 47: Predictable Randomness: Part 1
• Episode 48: Predictable Randomness: Part 2
• Episode 49: Generative Art: Part 1 🆓
• Episode 50: Generative Art: Part 2 🆓

All modules are released under the MIT license. See LICENSE for details.