Community
Modules 3

System.Random

stable

This library deals with the common task of pseudo-random number generation.

Introduction

This module provides type classes and instances for the following concepts:

Pure pseudo-random number generators
RandomGen is an interface to pure

pseudo-random number generators.

StdGen, the standard pseudo-random number generator provided in this library, is an instance of RandomGen. It uses the SplitMix implementation provided by the splitmix package. Programmers may, of course, supply their own instances of RandomGen.

Usage

In pure code, use uniform and uniformR to generate pseudo-random values with a pure pseudo-random number generator like StdGen.

>>> :{
let rolls :: RandomGen g => Int -> g -> [Word]
rolls n = fst . uniformListR n (1, 6)
pureGen = mkStdGen 137
in
rolls 10 pureGen :: [Word]
:}
[4,2,6,1,6,6,5,1,1,5]

To run use a monadic pseudo-random computation in pure code with a pure pseudo-random number generator, use runStateGen and its variants.

>>> :{
let rollsM :: StatefulGen g m => Int -> g -> m [Word]
rollsM n = uniformListRM n (1, 6)
pureGen = mkStdGen 137
in
runStateGen_ pureGen (rollsM 10) :: [Word]
:}
[4,2,6,1,6,6,5,1,1,5]

Pure number generator interface

Pseudo-random number generators come in two flavours: pure and monadic.

RandomGen: pure pseudo-random number generators
These generators produce

a new pseudo-random value together with a new instance of the pseudo-random number generator.

Pure pseudo-random number generators should implement split if they are splittable, that is, if there is an efficient method to turn one generator into two. The pseudo-random numbers produced by the two resulting generators should not be correlated. See [1] for some background on splittable pseudo-random generators.

StatefulGen: monadic pseudo-random number generators

See System.Random.Stateful module

class RandomGen g

RandomGen is an interface to pure pseudo-random number generators.

StdGen is the standard RandomGen instance provided by this library.

genWord8 :: g -> (Word8, g)

Returns a Word8 that is uniformly distributed over the entire Word8 range.

genWord16 :: g -> (Word16, g)

Returns a Word16 that is uniformly distributed over the entire Word16 range.

genWord32 :: g -> (Word32, g)

Returns a Word32 that is uniformly distributed over the entire Word32 range.

genWord64 :: g -> (Word64, g)

Returns a Word64 that is uniformly distributed over the entire Word64 range.

genWord32R :: Word32 -> g -> (Word32, g)

genWord32R upperBound g returns a Word32 that is uniformly distributed over the range [0, upperBound].

genWord64R :: Word64 -> g -> (Word64, g)

genWord64R upperBound g returns a Word64 that is uniformly distributed over the range [0, upperBound].

unsafeUniformFillMutableByteArray :: MutableByteArray s -> Int -> Int -> g -> ST s g

Fill in the supplied `MutableByteArray` with uniformly generated random bytes. This function is unsafe because it is not required to do any bounds checking. For a safe variant use `System.Random.Sateful.uniformFillMutableByteArrayM` instead.

Default type class implementation uses `defaultUnsafeUniformFillMutableByteArray`.

Mutable array to fill with random bytes

Int

Offset into a mutable array from the beginning in number of bytes. Offset must be non-negative, but this will not be checked

Int

Number of randomly generated bytes to write into the array. Number of bytes must be non-negative and less then the total size of the array, minus the offset. This also will be checked.

split :: g -> (g, g)

In favor of `splitGen`

Returns two distinct pseudo-random number generators.

Implementations should take care to ensure that the resulting generators are not correlated. Some pseudo-random number generators are not splittable. In that case, the split implementation should fail with a descriptive error message.

class RandomGen g => SplitGen g

Pseudo-random generators that can be split into two separate and independent psuedo-random generators should provide an instance for this type class.

Historically this functionality was included in the `RandomGen` type class in the `split` function, however, few pseudo-random generators possess this property of splittability. This lead the old `split` function being usually implemented in terms of `error`.

splitGen :: g -> (g, g)

Returns two distinct pseudo-random number generators.

Implementations should take care to ensure that the resulting generators are not correlated.

uniform :: (Uniform a, RandomGen g) => g -> (a, g)

Generates a value uniformly distributed over all possible values of that type.

This is a pure version of uniformM.

====Examples

>>> import System.Random
>>> let pureGen = mkStdGen 137
>>> uniform pureGen :: (Bool, StdGen)
(True,StdGen {unStdGen = SMGen 11285859549637045894 7641485672361121627})

You can use type applications to disambiguate the type of the generated numbers:

>>> :seti -XTypeApplications
>>> uniform @Bool pureGen
(True,StdGen {unStdGen = SMGen 11285859549637045894 7641485672361121627})

uniformR :: (UniformRange a, RandomGen g) => (a, a) -> g -> (a, g)

Generates a value uniformly distributed over the provided range, which is interpreted as inclusive in the lower and upper bound.

  • uniformR (1 :: Int, 4 :: Int) generates values uniformly from the set \{1,2,3,4\}

  • uniformR (1 :: Float, 4 :: Float) generates values uniformly from the set \{x\;|\;1 \le x \le 4\}

The following law should hold to make the function always defined:

uniformR (a, b) = uniformR (b, a)

This is a pure version of uniformRM.

====Examples

>>> import System.Random
>>> let pureGen = mkStdGen 137
>>> uniformR (1 :: Int, 4 :: Int) pureGen
(4,StdGen {unStdGen = SMGen 11285859549637045894 7641485672361121627})

You can use type applications to disambiguate the type of the generated numbers:

>>> :seti -XTypeApplications
>>> uniformR @Int (1, 4) pureGen
(4,StdGen {unStdGen = SMGen 11285859549637045894 7641485672361121627})

class Random a

The class of types for which random values can be generated. Most instances of `Random` will produce values that are uniformly distributed on the full range, but for those types without a well-defined "full range" some sensible default subrange will be selected.

Random exists primarily for backwards compatibility with version 1.1 of this library. In new code, use the better specified Uniform and UniformRange instead.

randomR :: RandomGen g => (a, a) -> g -> (a, g)

Takes a range (lo,hi) and a pseudo-random number generator g, and returns a pseudo-random value uniformly distributed over the closed interval [lo,hi], together with a new generator. It is unspecified what happens if lo>hi, but usually the values will simply get swapped.

>>> let gen = mkStdGen 26
>>> fst $ randomR ('a', 'z') gen
'z'
>>> fst $ randomR ('a', 'z') gen
'z'

For continuous types there is no requirement that the values lo and hi are ever produced, but they may be, depending on the implementation and the interval.

There is no requirement to follow the Ord instance and the concept of range can be defined on per type basis. For example product types will treat their values independently:

>>> fst $ randomR (('a', 5.0), ('z', 10.0)) $ mkStdGen 26
('z',5.22694980853051)

In case when a lawful range is desired `uniformR` should be used instead.

random :: RandomGen g => g -> (a, g)

The same as randomR, but using a default range determined by the type:

  • For bounded types (instances of Bounded, such as Char), the range is normally the whole type.

  • For floating point types, the range is normally the closed interval [0,1].

  • For Integer, the range is (arbitrarily) the range of Int.

randomRs :: RandomGen g => (a, a) -> g -> [a]

Plural variant of randomR, producing an infinite list of pseudo-random values instead of returning a new generator.

randoms :: RandomGen g => g -> [a]

Plural variant of random, producing an infinite list of pseudo-random values instead of returning a new generator.

class Uniform a

The class of types for which a uniformly distributed value can be drawn from all possible values of the type.

class UniformRange a

The class of types for which a uniformly distributed value can be drawn from a range.

class Finite a

A type class for data with a finite number of inhabitants. This type class is used in the default implementation of Uniform.

Users are not supposed to write instances of Finite manually. There is a default implementation in terms of Generic instead.

>>> :seti -XDeriveGeneric -XDeriveAnyClass
>>> import GHC.Generics (Generic)
>>> data MyBool = MyTrue | MyFalse deriving (Generic, Finite)
>>> data Action = Code MyBool | Eat (Maybe Bool) | Sleep deriving (Generic, Finite)

Seed

mkSeed :: forall g m. (SeedGen g, F.MonadFail m) => ByteArray -> m (Seed g)

Construct a `Seed` from a `ByteArray` of expected length. Whenever `ByteArray` does not match the `SeedSize` specified by the pseudo-random generator, this function will `F.fail`.

mkSeedFromByteString :: (SeedGen g, F.MonadFail m) => BS.ByteString -> m (Seed g)

Just like `mkSeed`, but uses `ByteString` as argument. Results in a memcopy of the seed.

nonEmptyFromSeed :: forall g. SeedGen g => Seed g -> NonEmpty Word64

Convert a `Seed` to a list of 64bit words.

nonEmptyToSeed :: forall g. SeedGen g => NonEmpty Word64 -> Seed g

Construct a seed from a list of 64-bit words. At most `SeedSize` many bytes will be used.

seedGenTypeName :: forall g. SeedGen g => String

This is a function that shows the name of the generator type, which is useful for error reporting.

seedSize :: forall g. SeedGen g => Int

Get the expected size of the `Seed` in number bytes

seedSizeProxy :: forall proxy g. SeedGen g => proxy g -> Int

Just like `seedSize`, except it accepts a proxy as an argument.

unSeed :: Seed g -> ByteArray

Unwrap the `Seed` and get the underlying `ByteArray`

unSeedToByteString :: Seed g -> BS.ByteString

Just like `unSeed`, but produced a `ByteString`. Results in a memcopy of the seed.

withSeed :: SeedGen g => Seed g -> (g -> (a, g)) -> (a, Seed g)

Helper function that allows for operating directly on the `Seed`, while supplying a function that uses the pseudo-random number generator that is constructed from that `Seed`.

====Example

>>> :set -XTypeApplications
>>> import System.Random
>>> withSeed (nonEmptyToSeed (pure 2024) :: Seed StdGen) (uniform @Int)
(1039666877624726199,Seed [0xe9, 0x07, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00])

withSeedFile :: (SeedGen g, MonadIO m) => FilePath -> (Seed g -> m (a, Seed g)) -> m a

Read the seed from a file and use it for constructing a pseudo-random number generator. After supplied action has been applied to the constructed generator, the resulting generator will be converted back to a seed and written to the same file.

withSeedM :: (SeedGen g, Functor f) => Seed g -> (g -> f (a, g)) -> f (a, Seed g)

Same as `withSeed`, except it is useful with monadic computation and frozen generators.

See `System.Random.Stateful.withSeedMutableGen` for a helper that also handles seeds for mutable pseduo-random number generators.

newtype Seed g

This is a binary form of pseudo-random number generator's state. It is designed to be safe and easy to use for input/output operations like restoring from file, transmitting over the network, etc.

Constructor is not exported, becasue it is important for implementation to enforce the invariant of the underlying byte array being of the exact same length as the generator has specified in `System.Random.Seed.SeedSize`. Use `System.Random.Seed.mkSeed` and `System.Random.Seed.unSeed` to get access to the raw bytes in a safe manner.

class (KnownNat (SeedSize g), 1 <= SeedSize g, Typeable g) => SeedGen g

Interface for converting a pure pseudo-random number generator to and from non-empty sequence of bytes. Seeds are stored in Little-Endian order regardless of the platform it is being used on, which provides cross-platform compatibility, while providing optimal performance for the most common platform type.

Conversion to and from a `Seed` serves as a building block for implementing serialization for any pure or frozen pseudo-random number generator.

It is not trivial to implement platform independence. For this reason this type class has two alternative ways of creating an instance for this class. The easiest way for constructing a platform indepent seed is by converting the inner state of a generator to and from a list of 64 bit words using `toSeed64` and `fromSeed64` respectively. In that case cross-platform support will be handled automaticaly.

>>> :set -XDataKinds -XTypeFamilies
>>> import Data.Word (Word8, Word32)
>>> import Data.Bits ((.|.), shiftR, shiftL)
>>> import Data.List.NonEmpty (NonEmpty ((:|)))
>>> data FiveByteGen = FiveByteGen Word8 Word32 deriving Show
>>> :{
instance SeedGen FiveByteGen where
type SeedSize FiveByteGen = 5
fromSeed64 (w64 :| _) =
FiveByteGen (fromIntegral (w64 `shiftR` 32)) (fromIntegral w64)
toSeed64 (FiveByteGen x1 x4) =
let w64 = (fromIntegral x1 `shiftL` 32) .|. fromIntegral x4
in (w64 :| [])
:}
>>> FiveByteGen 0x80 0x01020304
FiveByteGen 128 16909060
>>> fromSeed (toSeed (FiveByteGen 0x80 0x01020304))
FiveByteGen 128 16909060
>>> toSeed (FiveByteGen 0x80 0x01020304)
Seed [0x04, 0x03, 0x02, 0x01, 0x80]
>>> toSeed64 (FiveByteGen 0x80 0x01020304)
549772722948 :| []

However, when performance is of utmost importance or default handling of cross platform independence is not sufficient, then an adventurous developer can try implementing conversion into bytes directly with `toSeed` and `fromSeed`.

Properties that must hold:

> fromSeed (toSeed gen) == gen
> fromSeed64 (toSeed64 gen) == gen

Note, that there is no requirement for every `Seed` to roundtrip, eg. this proprty does not even hold for `StdGen`:

>>> let seed = nonEmptyToSeed (0xab :| [0xff00]) :: Seed StdGen
>>> seed == toSeed (fromSeed seed)
False

SeedSize :: SeedSize g :: Nat

Number of bytes that is required for storing the full state of a pseudo-random number generator. It should be big enough to satisfy the roundtrip property:

> fromSeed (toSeed gen) == gen

fromSeed :: Seed g -> g

Convert from a binary representation to a pseudo-random number generator

toSeed :: g -> Seed g

Convert to a binary representation of a pseudo-random number generator

fromSeed64 :: NonEmpty Word64 -> g

Construct pseudo-random number generator from a list of words. Whenever list does not have enough bytes to satisfy the `SeedSize` requirement, it will be padded with zeros. On the other hand when it has more than necessary, extra bytes will be dropped.

For example if `SeedSize` is set to 2, then only the lower 16 bits of the first element in the list will be used.

toSeed64 :: g -> NonEmpty Word64

Convert pseudo-random number generator to a list of words

In case when `SeedSize` is not a multiple of 8, then the upper bits of the last word in the list will be set to zero.

Generators for sequences of pseudo-random bytes

** Lists

uniforms :: (Uniform a, RandomGen g) => g -> [a]

Produce an infinite list of pseudo-random values. Integrates nicely with list fusion. Naturally, there is no way to recover the final generator, therefore either use `split` before calling `uniforms` or use `uniformList` instead.

Similar to `randoms`, except it relies on `Uniform` type class instead of `Random`

====Examples

>>> let gen = mkStdGen 2023
>>> import Data.Word (Word16)
>>> take 5 $ uniforms gen :: [Word16]
[56342,15850,25292,14347,13919]

uniformRs :: (UniformRange a, RandomGen g) => (a, a) -> g -> [a]

Produce an infinite list of pseudo-random values in a specified range. Same as `uniforms`, integrates nicely with list fusion. There is no way to recover the final generator, therefore either use `split` before calling `uniformRs` or use `uniformListR` instead.

Similar to `randomRs`, except it relies on `UniformRange` type class instead of `Random`.

====Examples

>>> let gen = mkStdGen 2023
>>> take 5 $ uniformRs (10, 100) gen :: [Int]
[32,86,21,57,39]

uniformList :: (Uniform a, RandomGen g) => Int -> g -> ([a], g)

Produce a list of the supplied length with elements generated uniformly.

See `uniformListM` for a stateful counterpart.

====Examples

>>> let gen = mkStdGen 2023
>>> import Data.Word (Word16)
>>> uniformList 5 gen :: ([Word16], StdGen)
([56342,15850,25292,14347,13919],StdGen {unStdGen = SMGen 6446154349414395371 1920468677557965761})

uniformListR :: (UniformRange a, RandomGen g) => Int -> (a, a) -> g -> ([a], g)

Produce a list of the supplied length with elements generated uniformly.

See `uniformListM` for a stateful counterpart.

====Examples

>>> let gen = mkStdGen 2023
>>> uniformListR 10 (20, 30) gen :: ([Int], StdGen)
([26,30,27,24,30,25,27,21,27,27],StdGen {unStdGen = SMGen 12965503083958398648 1920468677557965761})

uniformShuffleList :: RandomGen g => [a] -> g -> ([a], g)

Shuffle elements of a list in a uniformly random order.

====Examples

>>> uniformShuffleList "ELVIS" $ mkStdGen 252
("LIVES",StdGen {unStdGen = SMGen 17676540583805057877 5302934877338729551})

Bytes

uniformByteArray :: RandomGen g => Bool -> Int -> g -> (ByteArray, g)

Efficiently generates a sequence of pseudo-random bytes in a platform independent manner.

Bool

Should byte array be allocated in pinned or unpinned memory.

Int

Number of bytes to generate

g

Pure pseudo-random numer generator

uniformByteString :: RandomGen g => Int -> g -> (ByteString, g)

Generates a ByteString of the specified size using a pure pseudo-random number generator. See uniformByteStringM for the monadic version.

====Examples

>>> import System.Random
>>> import Data.ByteString (unpack)
>>> let pureGen = mkStdGen 137
>>> unpack . fst $ uniformByteString 10 pureGen
[51,123,251,37,49,167,90,109,1,4]

uniformShortByteString :: RandomGen g => Int -> g -> (ShortByteString, g)

Same as `uniformByteArray` `False`, but for `ShortByteString`.

Returns a ShortByteString of length n filled with pseudo-random bytes.

====Examples

>>> import System.Random
>>> import Data.ByteString.Short (unpack)
>>> let pureGen = mkStdGen 137
>>> unpack . fst $ uniformShortByteString 10 pureGen
[51,123,251,37,49,167,90,109,1,4]

uniformFillMutableByteArray :: RandomGen g => MutableByteArray s -> Int -> Int -> g -> ST s g

Fill in a slice of a mutable byte array with randomly generated bytes. This function does not fail, instead it clamps the offset and number of bytes to generate into a valid range.

Mutable array to fill with random bytes

Int

Offset into a mutable array from the beginning in number of bytes. Offset will be clamped into the range between 0 and the total size of the mutable array

Int

Number of randomly generated bytes to write into the array. This number will be clamped between 0 and the total size of the array without the offset.

Deprecated

genByteString :: RandomGen g => Int -> g -> (ByteString, g)

In favor of `uniformByteString`

Generates a ByteString of the specified size using a pure pseudo-random number generator. See uniformByteStringM for the monadic version.

====Examples

>>> import System.Random
>>> import Data.ByteString
>>> let pureGen = mkStdGen 137
>>> :seti -Wno-deprecations
>>> unpack . fst . genByteString 10 $ pureGen
[51,123,251,37,49,167,90,109,1,4]

genShortByteString :: Int -> g -> (ShortByteString, g)

In favor of `System.Random.uniformShortByteString`

Same as `uniformByteArray` `False`, but for `ShortByteString`.

genShortByteString n g returns a ShortByteString of length n filled with pseudo-random bytes.

Note - This function will be removed from the type class in the next major release as it is no longer needed because of `unsafeUniformFillMutableByteArray`.

Standard pseudo-random number generator

newtype StdGen

The standard pseudo-random number generator.

mkStdGen :: Int -> StdGen

Constructs a StdGen deterministically from an `Int` seed. See `mkStdGen64` for a `Word64` variant that is architecture agnostic.

mkStdGen64 :: Word64 -> StdGen

Constructs a StdGen deterministically from a `Word64` seed.

The difference between `mkStdGen` is that `mkStdGen64` will work the same on 64-bit and 32-bit architectures, while the former can only use 32-bit of information for initializing the psuedo-random number generator on 32-bit operating systems

initStdGen :: MonadIO m => m StdGen

Initialize StdGen using system entropy (i.e. /dev/urandom) when it is available, while falling back on using system time as the seed.

Global standard pseudo-random number generator

There is a single, implicit, global pseudo-random number generator of type StdGen, held in a global mutable variable that can be manipulated from within the IO monad. It is also available as globalStdGen, therefore it is recommended to use the new System.Random.Stateful interface to explicitly operate on the global pseudo-random number generator.

It is initialised with initStdGen, although it is possible to override its value with setStdGen. All operations on the global pseudo-random number generator are thread safe, however in presence of concurrency they are naturally become non-deterministic. Moreover, relying on the global mutable state makes it hard to know which of the dependent libraries are using it as well, making it unpredictable in the local context. Precisely of this reason, the global pseudo-random number generator is only suitable for uses in applications, test suites, etc. and is advised against in development of reusable libraries.

It is also important to note that either using StdGen with pure functions from other sections of this module or by relying on runStateGen from stateful interface does not only give us deterministic behaviour without requiring IO, but it is also more efficient.

getStdRandom :: MonadIO m => (StdGen -> (a, StdGen)) -> m a

Uses the supplied function to get a value from the current global random generator, and updates the global generator with the new generator returned by the function. For example, rollDice produces a pseudo-random integer between 1 and 6:

>>> rollDice = getStdRandom (randomR (1, 6))
>>> replicateM 10 (rollDice :: IO Int)
[1,1,1,4,5,6,1,2,2,5]

This is an outdated function and it is recommended to switch to its equivalent applyAtomicGen instead, possibly with the globalStdGen if relying on the global state is acceptable.

>>> import System.Random.Stateful
>>> rollDice = applyAtomicGen (uniformR (1, 6)) globalStdGen
>>> replicateM 10 (rollDice :: IO Int)
[2,1,1,5,4,3,6,6,3,2]

getStdGen :: MonadIO m => m StdGen

Gets the global pseudo-random number generator. Extracts the contents of globalStdGen

setStdGen :: MonadIO m => StdGen -> m ()

Sets the global pseudo-random number generator. Overwrites the contents of globalStdGen

newStdGen :: MonadIO m => m StdGen

Applies split to the current global pseudo-random generator globalStdGen, updates it with one of the results, and returns the other.

randomIO :: (Random a, MonadIO m) => m a

A variant of randomM that uses the global pseudo-random number generator globalStdGen.

>>> import Data.Int
>>> randomIO :: IO Int32
114794456

This function is equivalent to getStdRandom random and is included in this interface for historical reasons and backwards compatibility. It is recommended to use uniformM instead, possibly with the globalStdGen if relying on the global state is acceptable.

>>> import System.Random.Stateful
>>> uniformM globalStdGen :: IO Int32
-1768545016

randomRIO :: (Random a, MonadIO m) => (a, a) -> m a

A variant of randomRM that uses the global pseudo-random number generator globalStdGen

>>> randomRIO (2020, 2100) :: IO Int
2028

Similar to randomIO, this function is equivalent to @getStdRandom randomR@ and is included in this interface for historical reasons and backwards compatibility. It is recommended to use uniformRM instead, possibly with the globalStdGen if relying on the global state is acceptable.

>>> import System.Random.Stateful
>>> uniformRM (2020, 2100) globalStdGen :: IO Int
2044

Compatibility and reproducibility

** Backwards compatibility and deprecations

genRange :: g -> (Int, Int)

No longer used

Yields the range of values returned by next.

It is required that:

  • If (a, b) = genRange g, then a < b.

  • genRange must not examine its argument so the value it returns is determined only by the instance of RandomGen.

The default definition spans the full range of Int.

next :: g -> (Int, g)

No longer used

Returns an Int that is uniformly distributed over the range returned by genRange (including both end points), and a new generator. Using next is inefficient as all operations go via Integer. See [here](https:/alexey.kuleshevi.chblog20191221random-benchmarks) for more details. It is thus deprecated.

Version 1.2 mostly maintains backwards compatibility with version 1.1. This has a few consequences users should be aware of:

  • The type class Random is only provided for backwards compatibility. New code should use Uniform and UniformRange instead.

  • The methods next and genRange in RandomGen are deprecated and only provided for backwards compatibility. New instances of RandomGen should implement word-based methods instead. See below for more information about how to write a RandomGen instance.

  • This library provides instances for Random for some unbounded types for backwards compatibility. For an unbounded type, there is no way to generate a value with uniform probability out of its entire domain, so the random implementation for unbounded types actually generates a value based on some fixed range.

For Integer, random generates a value in the Int range. For Float and Double, random generates a floating point value in the range @[0, 1)@.

This library does not provide Uniform instances for any unbounded types.

Reproducibility

If you have two builds of a particular piece of code against this library, any deterministic function call should give the same result in the two builds if the builds are

  • compiled against the same major version of this library

  • on the same architecture (32-bit or 64-bit)

Notes for pseudo-random number generator implementors

** How to implement RandomGen

Consider these points when writing a RandomGen instance for a given pure pseudo-random number generator:

  • If the pseudo-random number generator has a power-of-2 modulus, that is, it natively outputs 2^n bits of randomness for some n, implement genWord8, genWord16, genWord32 and genWord64. See below for more details.

  • If the pseudo-random number generator does not have a power-of-2 modulus, implement next and genRange. See below for more details.

  • If the pseudo-random number generator is splittable, implement split. If there is no suitable implementation, split should fail with a helpful error message.

How to implement RandomGen for a pseudo-random number generator with power-of-2 modulus

Suppose you want to implement a [permuted congruential generator](https:/en.wikipedia.orgwiki/Permuted_congruential_generator).

>>> data PCGen = PCGen !Word64 !Word64

It produces a full Word32 of randomness per iteration.

>>> import Data.Bits
>>> :{
let stepGen :: PCGen -> (Word32, PCGen)
stepGen (PCGen state inc) = let
newState = state * 6364136223846793005 + (inc .|. 1)
xorShifted = fromIntegral (((state `shiftR` 18) `xor` state) `shiftR` 27) :: Word32
rot = fromIntegral (state `shiftR` 59) :: Word32
out = (xorShifted `shiftR` (fromIntegral rot)) .|. (xorShifted `shiftL` fromIntegral ((-rot) .&. 31))
in (out, PCGen newState inc)
:}
>>> fst $ stepGen $ snd $ stepGen (PCGen 17 29)
3288430965

You can make it an instance of RandomGen as follows:

>>> :{
instance RandomGen PCGen where
genWord32 = stepGen
split _ = error "PCG is not splittable"
:}

How to implement RandomGen for a pseudo-random number generator without a power-of-2 modulus

We do not recommend you implement any new pseudo-random number generators without a power-of-2 modulus.

Pseudo-random number generators without a power-of-2 modulus perform significantly worse than pseudo-random number generators with a power-of-2 modulus with this library. This is because most functionality in this library is based on generating and transforming uniformly pseudo-random machine words, and generating uniformly pseudo-random machine words using a pseudo-random number generator without a power-of-2 modulus is expensive.

The pseudo-random number generator from L’Ecuyer (1988) natively generates an integer value in the range [1, 2147483562]. This is the generator used by this library before it was replaced by SplitMix in version 1.2.

>>> data LegacyGen = LegacyGen !Int32 !Int32
>>> :{
let legacyNext :: LegacyGen -> (Int, LegacyGen)
legacyNext (LegacyGen s1 s2) = (fromIntegral z', LegacyGen s1'' s2'') where
z' = if z < 1 then z + 2147483562 else z
z = s1'' - s2''
k = s1 `quot` 53668
s1'  = 40014 * (s1 - k * 53668) - k * 12211
s1'' = if s1' < 0 then s1' + 2147483563 else s1'
k' = s2 `quot` 52774
s2' = 40692 * (s2 - k' * 52774) - k' * 3791
s2'' = if s2' < 0 then s2' + 2147483399 else s2'
:}

You can make it an instance of RandomGen as follows:

>>> :{
instance RandomGen LegacyGen where
next = legacyNext
genRange _ = (1, 2147483562)
split _ = error "Not implemented"
:}

References

  1. Guy L. Steele, Jr., Doug Lea, and Christine H. Flood. 2014. Fast splittable pseudorandom number generators. In Proceedings of the 2014 ACM International Conference on Object Oriented Programming Systems Languages & Applications (OOPSLA '14). ACM, New York, NY, USA, 453-472. DOI: https://doi.org/10.1145/2660193.2660195

Instances

43
  • instance Random Integer

    Note - `random` generates values in the `Int` range

  • instance Random Int8
  • instance Random Int16
  • instance Random Int32
  • instance Random Int64
  • instance Random Int
  • instance Random Word
  • instance Random Word8
  • instance Random Word16
  • instance Random Word32
  • instance Random Word64
  • instance Random CBool
  • instance Random CChar
  • instance Random CSChar
  • instance Random CUChar
  • instance Random CShort
  • instance Random CUShort
  • instance Random CInt
  • instance Random CUInt
  • instance Random CLong
  • instance Random CULong
  • instance Random CPtrdiff
  • instance Random CSize
  • instance Random CWchar
  • instance Random CSigAtomic
  • instance Random CLLong
  • instance Random CULLong
  • instance Random CIntPtr
  • instance Random CUIntPtr
  • instance Random CIntMax
  • instance Random CUIntMax
  • instance Random CFloat

    Note - `random` produces values in the closed range [0,1].

  • instance Random CDouble

    Note - `random` produces values in the closed range [0,1].

  • instance Random Char
  • instance Random Bool
  • instance Random Double

    Note - `random` produces values in the closed range [0,1].

  • instance Random Float

    Note - `random` produces values in the closed range [0,1].

  • instance (Random a, Random b) => Random (a, b)

    Note - `randomR` treats a and b types independently

  • instance (Random a, Random b, Random c) => Random (a, b, c)

    Note - `randomR` treats a, b and c types independently

  • instance (Random a, Random b, Random c, Random d) => Random (a, b, c, d)

    Note - `randomR` treats a, b, c and d types independently

  • instance (Random a, Random b, Random c, Random d, Random e) => Random (a, b, c, d, e)

    Note - `randomR` treats a, b, c, d and e types independently

  • instance (Random a, Random b, Random c, Random d, Random e, Random f) => Random (a, b, c, d, e, f)

    Note - `randomR` treats a, b, c, d, e and f types independently

  • instance (Random a, Random b, Random c, Random d, Random e, Random f, Random g) => Random (a, b, c, d, e, f, g)

    Note - `randomR` treats a, b, c, d, e, f and g types independently

Maintainer
libraries@haskell.org
Copyright
(c) The University of Glasgow 2001
License
BSD-style (see the file LICENSE in the 'random' repository)
Build record and source files

Community contribution. The service has not verified this documentation.

Target: Linux on x86-64. Some inferred types and resolved links are not available.

Generator: aihc-haddock-858b1951-web-1

Metadata snapshot: 1012568133f474470681e07d72ee6260a4d8fe5ba2de42241172dfb36bf88370

Open the dependency plan · Download the documentation model

Import a package

Enter a package name and an exact version from Hackage. The site contains imported versions only.