RandomKit is a Swift framework that makes random data generation simple and easy.
- Build Status
- Installation
- Benchmark
- Usage
- Extra
- License
Branch | Status |
---|---|
master |
- Platforms:
- macOS 10.9+
- iOS 8.0+
- watchOS 2.0+
- tvOS 9.0+
- Linux
- Xcode 8.0+
- Swift 3.0.2+ & 4.0
RandomKit is possibly also compatible with FreeBSD, Android, and Windows (under Cygwin) but has not been tested for those platforms.
The Swift Package Manager is a decentralized dependency manager for Swift.
-
Add the project to your
Package.swift
.import PackageDescription let package = Package( name: "MyAwesomeProject", dependencies: [ .Package(url: "https://github.com/nvzqz/RandomKit.git", majorVersion: 5) ] )
-
Import the RandomKit module.
import RandomKit
CocoaPods is a centralized dependency manager for Objective-C and Swift. Go here to learn more.
-
Add the project to your Podfile.
use_frameworks! pod 'RandomKit', '~> 5.2.3'
If you want to be on the bleeding edge, replace the last line with:
pod 'RandomKit', :git => 'https://github.com/nvzqz/RandomKit.git'
-
Run
pod install
and open the.xcworkspace
file to launch Xcode. -
Import the RandomKit framework.
import RandomKit
Carthage is a decentralized dependency manager for Objective-C and Swift.
-
Add the project to your Cartfile.
github "nvzqz/RandomKit"
-
Run
carthage update
and follow the additional steps in order to add RandomKit to your project. -
Import the RandomKit framework.
import RandomKit
Various components of RandomKit can be easily benchmarked by running benchmark.sh
.
./benchmark.sh [FLAGS] [PROTOCOLS]
Use the --help
flag for information regarding how to use it.
Note: The default count is 10000000, which is A LOT if using the --array
flag.
This can be changed by passing an argument into --count
or -c
.
Try it out for yourself! Download the repo and open 'RandomKit.playground'.
The RandomGenerator
protocol defines basic methods for generating primitive
values and randomizing a buffer.
All provided types that conform to RandomGenerator
have a static default
value that can be passed as an inout
argument to generation functions.
let value = Int.random(using: &Xoroshiro.default)
-
ARC4Random
- Because the symbols for the
arc4random
family of functions aren't exported with Foundation on Linux and other platforms, they're dynamically loaded at runtime.
- Because the symbols for the
-
DeviceRandom
- Reads from "/dev/random" or "/dev/urandom" as its source.
-
MersenneTwister
-
Xoroshiro
-
Xorshift
-
XorshiftStar
-
ChaCha
SeedableRandomGenerator
is for types that can be seeded with some associated
Seed
type.
The RandomBytesGenerator
protocol is for types that specialize in generating a
specific type that fills up a number of bytes. For example, MersenneTwister
specializes in generating UInt64
while Xorshift
generates UInt32
values.
For single-threaded programs, it is safe to use a global generator instance such
as Xoroshiro.default
as a source of randomness.
For multi-threaded programs, the thread-local instances should be used. This allows for different threads to use their own separate random generators without a shared mutable state.
In the following example, randomGenerator
is unique to each thread.
let randomBytes = Xoroshiro.withThreadLocal { randomGenerator in
return [UInt8](randomCount: 1000, using: &randomGenerator)
}
Thread-local generators are deallocated upon thread exit, so there's no need to worry about cleanup.
It's recommended to not call withThreadLocal(_:)
or get the threadLocal
pointer each individual time it's needed. Retrieving the thread-local instance
incurs avoidable overhead.
// Bad
let value = Int.random(using: &Xoroshiro.threadLocal.pointee)
array.shuffle(using: &Xoroshiro.threadLocal.pointee)
// Good
let threadLocal = Xoroshiro.threadLocal
let value = Int.random(using: &threadLocal.pointee)
array.shuffle(using: &threadLocal.pointee)
// Better
Xoroshiro.withThreadLocal { randomGenerator in
let value = Int.random(using: &randomGenerator)
array.shuffle(using: &randomGenerator)
}
As a shortcut, you can even apply a function directly as a parameter.
let value = Xoroshiro.withThreadLocal(Int.random)
Prior to v4.4.0,
thread safety could be achieved by instantiating a new seeded instance of a
given RandomGenerator
type. The problem with this is that unnecessary seeding
occurs each time. With this, the generator is seeded once and can then be reused
at later points.
Shortcuts to the reseeding version of a generator are also available:
Xoroshiro.withThreadLocalReseeding {
...
}
Which is way better than writing:
ReseedingRandomGenerator.withThreadLocal(createdWith: { Xoroshiro.reseeding }) {
...
}
RandomKit is very protocol-oriented, which gives it the ability to be very flexible and modular.
A protocol for types that can generate random values using a RandomGenerator
.
A protocol for types that can generate optional random values within a range
using a RandomGenerator
.
Int.random(in: 0 ..< 0, using: &randomGenerator) // nil
A protocol for types that can generate random values within a closed range
using a RandomGenerator
.
Int.random(in: -100 ... 100, using: &randomGenerator) // -79
A protocol for types that can generate random values from a base value to another value, noninclusive.
The base value for integers is 0. This means that calling random(to:using:)
on
a negative value will yield a random negative value or zero whereas a positive
value will yield a random positive value or zero.
If value
== randomBase
, value
will be returned for random(to:using:)
.
Int.random(to: 2, using: &randomGenerator) // Either 0 or 1
Int.random(to: 0, using: &randomGenerator) // Always 0
Int.random(to: 32, using: &randomGenerator) // 15
Int.random(to: -5, using: &randomGenerator) // -3
A protocol for types that can generate random values from a base value through another value, inclusive.
The same rules regarding the base value of RandomToValue
apply to
RandomThroughValue
.
A protocol for types whose instances can have random elements retrieved.
["Bob", "Cindy", "May", "Charles", "Javier"].random(using: &randomGenerator) // "Charles"
"Hello".characters.random(using: &randomGenerator) // "e"
Some Foundation types like NSArray
conform to this protocol.
A protocol for types whose instances can have random elements retrieved from
within a Range<Index>
.
[20, 37, 42].random(in: 1 ..< 3, using: &randomGenerator) // Either 37 or 42
A protocol for types whose elements can be shuffled.
// Array
[1, 2, 3, 4, 5].shuffled(using: &randomGenerator) // [3, 4, 1, 5, 2]
// Dictionary
["a": 1, "b": 2, "c": 3].shuffled(using: &randomGenerator) // ["a": 3, "b": 1, "c": 2]
The mutable counterpart of shuffled(using:)
is shuffle(using:)
.
For better Array
shuffling performance, consider shuffling in-place with
shuffle(using:)
.
Similar to Shuffleable
, except no element is ever in its initial position.
All of Swift's native integer types conform to the Random-
protocols.
The random(using:)
function creates an integer of any value. As a result,
negative values can result for signed integers.
Int.random(using: &randomGenerator) // An Int within Int.min and Int.max
Int.random(in: 10...20, using: &randomGenerator) // An Int within 10 and 20
To create a positive signed integer, use random(to:using:)
or random(through:using:)
.
Int.random(to: 1000, using: &randomGenerator) // 731
Int.random(through: 10, using: &randomGenerator) // 4
Signed integers can be created from any range, without danger of overflow.
Int.random(in: (.min + 1000)...(.max - 200), using: &randomGenerator) // 5698527899712144154
Generate a random floating point value from within a range or 0.0...1.0
by
default.
Double.random(using: &randomGenerator) // 0.9813615573117475
Double.random(in: -10...10, using: &randomGenerator) // -4.03042337718197
Float.random(in: -10...10, using: &randomGenerator) // 5.167088
Float80.random(in: -10...10, using: &randomGenerator) // -3.63204542399198874
All FloatingPoint
types can also conform to RandomInClosedRange
out-of-the-box.
Bool.random(using:)
has a 50/50 chance of being true
.
If you need different probability, there's also random(withWeight:using:)
,
which has 1 in weight
chance of being true
.
String
, Character
, and UnicodeScalar
generate values within " "..."~"
by
default.
String.random(ofLength: 10, using: &randomGenerator) // "}+[=Ng>$w1"
String.random(ofLength: 10, in: "A"..."z", using: &randomGenerator) // "poUtXJIbv["
Character.random(using: &randomGenerator) // "#"
Character.random(in: "A"..."z", using: &randomGenerator) // "s"
An array of random values can be generated for types conforming to Random
with
init(randomCount:using:)
.
Similar initializers exist for all other Random-
protocols.
let randoms = Array<Int>(randomCount: 100, using: &randomGenerator) // [8845477344689834233, -957454203475087100, ...]
For types conforming to UnsafeRandom
, a faster alternative is init(unsafeRandomCount:using:)
.
This initializer fills the buffer directly rather than using random(using:)
.
let unsafeRandoms = Array<Int>(unsafeRandomCount: 100, using: &randomGenerator) // [759709806207883991, 4618491969012429761, ...]
A benchmark of generating 1000 random Int
arrays of 10000 count:
Generator | Time (in seconds) |
---|---|
Xoroshiro |
0.0271 |
Xorshift |
0.0568 |
XorshiftStar |
0.0319 |
ChaCha |
0.2027 |
MersenneTwister |
0.0432 |
ARC4Random |
0.2416 |
DeviceRandom |
5.3348 |
Note: Results may vary due to various factors.
This same benchmark can be run with:
./benchmark.sh --all-generators --array 10000 --count 1000
A random Date
can be generated between two Date
or TimeInterval
values.
The default random(using:)
function returns a Date
within Date.distantPast
and
Date.distantFuture
.
Date.random(using: &randomGenerator) // "Aug 28, 2006, 3:38 AM"
Date.random(in: Date.distantPast...Date(), using: &randomGenerator) // "Feb 7, 472, 5:40 AM"
The Decimal
type conforms to various Random-
protocols.
The random(using:)
function returns a Decimal
between 0 and 1 by default.
Decimal.random(using: &randomGenerator) // 0.87490000409886706715888973957833129437
Decimal.random(in: 0.0...10.0, using: &randomGenerator) // 6.5464639772070720738747790627821299859
A random number can be generated from within an integer or double range, or
0...100
by default.
NSNumber.random(using: &randomGenerator) // 79
NSNumber.random(in: -50...100, using: &randomGenerator) // -27
NSNumber.random(in: 100...200, using: &randomGenerator) // 149.6156950363926
A random color can be generated, with or without random alpha.
NSColor.random(using: &randomGenerator) // r 0.694 g 0.506 b 0.309 a 1.0
NSColor.random(alpha: true, using: &randomGenerator) // r 0.859 g 0.57 b 0.409 a 0.047
UIColor.random(using: &randomGenerator) // r 0.488 g 0.805 b 0.679 a 1.0
UIColor.random(alpha: true, using: &randomGenerator) // r 0.444 g 0.121 b 0.602 a 0.085
Because CGFloat
conforms to FloatingPoint
, it conforms to
RandomInClosedRange
just like how Double
and Float
do.
CGFloat.random(using: &randomGenerator) // 0.699803650379181
CGFloat.random(in: 0...100, using: &randomGenerator) // 43.27969591675319
A random point can be generated from within ranges for x and y.
CGPoint.random(using: &randomGenerator) // {x 70.093 y 95.721}
CGPoint.random(xRange: 0...200, yRange: 0...10, using: &randomGenerator) // {x 73.795 y 0.991}
A random size can be generated from within ranges for width and height.
CGSize.random(using: &randomGenerator) // {w 3.744 h 35.932}
CGSize.random(widthRange: 0...50, heightRange: 0...400, using: &randomGenerator) // {w 38.271 h 239.636}
A random rectangle can be generated from within ranges for x, y, width, and height.
CGRect.random(using: &randomGenerator) // {x 3.872 y 46.15 w 8.852 h 20.201}
CGRect.random(xRange: 0...50,
yRange: 0...100,
widthRange: 0...25,
heightRange: 0...10,
using: &randomGenerator) // {x 13.212 y 79.147 w 20.656 h 5.663}
A random vector can be generated from within ranges for dx and dy.
CGVector.random(using: &randomGenerator) // {dx 13.992 dy 89.376}
CGVector.random(dxRange: 0...50, dyRange: 0...10, using: &randomGenerator) // {dx 35.224 dy 13.463}
RandomKit extensions for Károly's BigInt library are available in RandomKitBigInt.
RandomKit and its assets are released under the MIT License. Assets
can be found in the assets
branch.
Parts of this project utilize code written by Matt Gallagher and, in conjunction with the MIT License, are licensed with that found here.