Random Numbers
Random number generation in Julia uses the Xoshiro256++ algorithm by default, with per-Task
state. Other RNG types can be plugged in by inheriting the AbstractRNG
type; they can then be used to obtain multiple streams of random numbers.
The PRNGs (pseudorandom number generators) exported by the Random
package are:
TaskLocalRNG
: a token that represents use of the currently active Task-local stream, deterministically seeded from the parent task, or byRandomDevice
(with system randomness) at program startXoshiro
: generates a high-quality stream of random numbers with a small state vector and high performance using the Xoshiro256++ algorithmRandomDevice
: for OS-provided entropy. This may be used for cryptographically secure random numbers (CS(P)RNG).MersenneTwister
: an alternate high-quality PRNG which was the default in older versions of Julia, and is also quite fast, but requires much more space to store the state vector and generate a random sequence.
Most functions related to random generation accept an optional AbstractRNG
object as first argument. Some also accept dimension specifications dims...
(which can also be given as a tuple) to generate arrays of random values. In a multi-threaded program, you should generally use different RNG objects from different threads or tasks in order to be thread-safe. However, the default RNG is thread-safe as of Julia 1.3 (using a per-thread RNG up to version 1.6, and per-task thereafter).
The provided RNGs can generate uniform random numbers of the following types: Float16
, Float32
, Float64
, BigFloat
, Bool
, Int8
, UInt8
, Int16
, UInt16
, Int32
, UInt32
, Int64
, UInt64
, Int128
, UInt128
, BigInt
(or complex numbers of those types). Random floating point numbers are generated uniformly in $[0, 1)$. As BigInt
represents unbounded integers, the interval must be specified (e.g. rand(big.(1:6))
).
Additionally, normal and exponential distributions are implemented for some AbstractFloat
and Complex
types, see randn
and randexp
for details.
To generate random numbers from other distributions, see the Distributions.jl package.
Because the precise way in which random numbers are generated is considered an implementation detail, bug fixes and speed improvements may change the stream of numbers that are generated after a version change. Relying on a specific seed or generated stream of numbers during unit testing is thus discouraged - consider testing properties of the methods in question instead.
Random numbers module
Random.Random
— ModuleRandom
Support for generating random numbers. Provides rand
, randn
, AbstractRNG
, MersenneTwister
, and RandomDevice
.
Random generation functions
Base.rand
— Functionrand([rng=default_rng()], [S], [dims...])
Pick a random element or array of random elements from the set of values specified by S
; S
can be
an indexable collection (for example
1:9
or('x', "y", :z)
)an
AbstractDict
orAbstractSet
objecta string (considered as a collection of characters), or
a type from the list below, corresponding to the specified set of values
concrete integer types sample from
typemin(S):typemax(S)
(exceptingBigInt
which is not supported)concrete floating point types sample from
[0, 1)
concrete complex types
Complex{T}
ifT
is a sampleable type take their real and imaginary components independently from the set of values corresponding toT
, but are not supported ifT
is not sampleable.all
<:AbstractChar
types sample from the set of valid Unicode scalarsa user-defined type and set of values; for implementation guidance please see Hooking into the
Random
APIa tuple type of known size and where each parameter of
S
is itself a sampleable type; return a value of typeS
. Note that tuple types such asTuple{Vararg{T}}
(unknown size) andTuple{1:2}
(parameterized with a value) are not supporteda
Pair
type, e.g.Pair{X, Y}
such thatrand
is defined forX
andY
, in which case random pairs are produced.
S
defaults to Float64
. When only one argument is passed besides the optional rng
and is a Tuple
, it is interpreted as a collection of values (S
) and not as dims
.
See also randn
for normally distributed numbers, and rand!
and randn!
for the in-place equivalents.
Support for S
as a tuple requires at least Julia 1.1.
Support for S
as a Tuple
type requires at least Julia 1.11.
Examples
julia> rand(Int, 2)
2-element Array{Int64,1}:
1339893410598768192
1575814717733606317
julia> using Random
julia> rand(Xoshiro(0), Dict(1=>2, 3=>4))
3 => 4
julia> rand((2, 3))
3
julia> rand(Float64, (2, 3))
2×3 Array{Float64,2}:
0.999717 0.0143835 0.540787
0.696556 0.783855 0.938235
The complexity of rand(rng, s::Union{AbstractDict,AbstractSet})
is linear in the length of s
, unless an optimized method with constant complexity is available, which is the case for Dict
, Set
and dense BitSet
s. For more than a few calls, use rand(rng, collect(s))
instead, or either rand(rng, Dict(s))
or rand(rng, Set(s))
as appropriate.
Random.rand!
— Functionrand!([rng=default_rng()], A, [S=eltype(A)])
Populate the array A
with random values. If S
is specified (S
can be a type or a collection, cf. rand
for details), the values are picked randomly from S
. This is equivalent to copyto!(A, rand(rng, S, size(A)))
but without allocating a new array.
Examples
julia> rand!(Xoshiro(123), zeros(5))
5-element Vector{Float64}:
0.521213795535383
0.5868067574533484
0.8908786980927811
0.19090669902576285
0.5256623915420473
Random.bitrand
— Functionbitrand([rng=default_rng()], [dims...])
Generate a BitArray
of random boolean values.
Examples
julia> bitrand(Xoshiro(123), 10)
10-element BitVector:
0
1
0
1
0
1
0
0
1
1
Base.randn
— Functionrandn([rng=default_rng()], [T=Float64], [dims...])
Generate a normally-distributed random number of type T
with mean 0 and standard deviation 1. Given the optional dims
argument(s), generate an array of size dims
of such numbers. Julia's standard library supports randn
for any floating-point type that implements rand
, e.g. the Base
types Float16
, Float32
, Float64
(the default), and BigFloat
, along with their Complex
counterparts.
(When T
is complex, the values are drawn from the circularly symmetric complex normal distribution of variance 1, corresponding to real and imaginary parts having independent normal distribution with mean zero and variance 1/2
).
See also randn!
to act in-place.
Examples
Generating a single random number (with the default Float64
type):
julia> randn()
-0.942481877315864
Generating a matrix of normal random numbers (with the default Float64
type):
julia> randn(2,3)
2×3 Matrix{Float64}:
1.18786 -0.678616 1.49463
-0.342792 -0.134299 -1.45005
Setting up of the random number generator rng
with a user-defined seed (for reproducible numbers) and using it to generate a random Float32
number or a matrix of ComplexF32
random numbers:
julia> using Random
julia> rng = Xoshiro(123);
julia> randn(rng, Float32)
-0.6457307f0
julia> randn(rng, ComplexF32, (2, 3))
2×3 Matrix{ComplexF32}:
-1.03467-1.14806im 0.693657+0.056538im 0.291442+0.419454im
-0.153912+0.34807im 1.0954-0.948661im -0.543347-0.0538589im
Random.randn!
— Functionrandn!([rng=default_rng()], A::AbstractArray) -> A
Fill the array A
with normally-distributed (mean 0, standard deviation 1) random numbers. Also see the rand
function.
Examples
julia> randn!(Xoshiro(123), zeros(5))
5-element Vector{Float64}:
-0.6457306721039767
-1.4632513788889214
-1.6236037455860806
-0.21766510678354617
0.4922456865251828
Random.randexp
— Functionrandexp([rng=default_rng()], [T=Float64], [dims...])
Generate a random number of type T
according to the exponential distribution with scale 1. Optionally generate an array of such random numbers. The Base
module currently provides an implementation for the types Float16
, Float32
, and Float64
(the default).
Examples
julia> rng = Xoshiro(123);
julia> randexp(rng, Float32)
1.1757717f0
julia> randexp(rng, 3, 3)
3×3 Matrix{Float64}:
1.37766 0.456653 0.236418
3.40007 0.229917 0.0684921
0.48096 0.577481 0.71835
Random.randexp!
— Functionrandexp!([rng=default_rng()], A::AbstractArray) -> A
Fill the array A
with random numbers following the exponential distribution (with scale 1).
Examples
julia> randexp!(Xoshiro(123), zeros(5))
5-element Vector{Float64}:
1.1757716836348473
1.758884569451514
1.0083623637301151
0.3510644315565272
0.6348266443720407
Random.randstring
— Functionrandstring([rng=default_rng()], [chars], [len=8])
Create a random string of length len
, consisting of characters from chars
, which defaults to the set of upper- and lower-case letters and the digits 0-9. The optional rng
argument specifies a random number generator, see Random Numbers.
Examples
julia> Random.seed!(3); randstring()
"Lxz5hUwn"
julia> randstring(Xoshiro(3), 'a':'z', 6)
"iyzcsm"
julia> randstring("ACGT")
"TGCTCCTC"
chars
can be any collection of characters, of type Char
or UInt8
(more efficient), provided rand
can randomly pick characters from it.
Subsequences, permutations and shuffling
Random.randsubseq
— Functionrandsubseq([rng=default_rng(),] A, p) -> Vector
Return a vector consisting of a random subsequence of the given array A
, where each element of A
is included (in order) with independent probability p
. (Complexity is linear in p*length(A)
, so this function is efficient even if p
is small and A
is large.) Technically, this process is known as "Bernoulli sampling" of A
.
Examples
julia> randsubseq(Xoshiro(123), 1:8, 0.3)
2-element Vector{Int64}:
4
7
Random.randsubseq!
— Functionrandsubseq!([rng=default_rng(),] S, A, p)
Like randsubseq
, but the results are stored in S
(which is resized as needed).
Examples
julia> S = Int64[];
julia> randsubseq!(Xoshiro(123), S, 1:8, 0.3)
2-element Vector{Int64}:
4
7
julia> S
2-element Vector{Int64}:
4
7
Random.randperm
— Functionrandperm([rng=default_rng(),] n::Integer)
Construct a random permutation of length n
. The optional rng
argument specifies a random number generator (see Random Numbers). The element type of the result is the same as the type of n
.
To randomly permute an arbitrary vector, see shuffle
or shuffle!
.
In Julia 1.1 randperm
returns a vector v
with eltype(v) == typeof(n)
while in Julia 1.0 eltype(v) == Int
.
Examples
julia> randperm(Xoshiro(123), 4)
4-element Vector{Int64}:
1
4
2
3
Random.randperm!
— Functionrandperm!([rng=default_rng(),] A::Array{<:Integer})
Construct in A
a random permutation of length length(A)
. The optional rng
argument specifies a random number generator (see Random Numbers). To randomly permute an arbitrary vector, see shuffle
or shuffle!
.
Examples
julia> randperm!(Xoshiro(123), Vector{Int}(undef, 4))
4-element Vector{Int64}:
1
4
2
3
Random.randcycle
— Functionrandcycle([rng=default_rng(),] n::Integer)
Construct a random cyclic permutation of length n
. The optional rng
argument specifies a random number generator, see Random Numbers. The element type of the result is the same as the type of n
.
Here, a "cyclic permutation" means that all of the elements lie within a single cycle. If n > 0
, there are $(n-1)!$ possible cyclic permutations, which are sampled uniformly. If n == 0
, randcycle
returns an empty vector.
randcycle!
is an in-place variant of this function.
In Julia 1.1 and above, randcycle
returns a vector v
with eltype(v) == typeof(n)
while in Julia 1.0 eltype(v) == Int
.
Examples
julia> randcycle(Xoshiro(123), 6)
6-element Vector{Int64}:
5
4
2
6
3
1
Random.randcycle!
— Functionrandcycle!([rng=default_rng(),] A::Array{<:Integer})
Construct in A
a random cyclic permutation of length n = length(A)
. The optional rng
argument specifies a random number generator, see Random Numbers.
Here, a "cyclic permutation" means that all of the elements lie within a single cycle. If A
is nonempty (n > 0
), there are $(n-1)!$ possible cyclic permutations, which are sampled uniformly. If A
is empty, randcycle!
leaves it unchanged.
randcycle
is a variant of this function that allocates a new vector.
Examples
julia> randcycle!(Xoshiro(123), Vector{Int}(undef, 6))
6-element Vector{Int64}:
5
4
2
6
3
1
Random.shuffle
— Functionshuffle([rng=default_rng(),] v::AbstractArray)
Return a randomly permuted copy of v
. The optional rng
argument specifies a random number generator (see Random Numbers). To permute v
in-place, see shuffle!
. To obtain randomly permuted indices, see randperm
.
Examples
julia> shuffle(Xoshiro(123), Vector(1:10))
10-element Vector{Int64}:
5
4
2
3
6
10
8
1
9
7
Random.shuffle!
— Functionshuffle!([rng=default_rng(),] v::AbstractArray)
In-place version of shuffle
: randomly permute v
in-place, optionally supplying the random-number generator rng
.
Examples
julia> shuffle!(Xoshiro(123), Vector(1:10))
10-element Vector{Int64}:
5
4
2
3
6
10
8
1
9
7
Generators (creation and seeding)
Random.default_rng
— FunctionRandom.default_rng() -> rng
Return the default global random number generator (RNG), which is used by rand
-related functions when no explicit RNG is provided.
When the Random
module is loaded, the default RNG is randomly seeded, via Random.seed!()
: this means that each time a new julia session is started, the first call to rand()
produces a different result, unless seed!(seed)
is called first.
It is thread-safe: distinct threads can safely call rand
-related functions on default_rng()
concurrently, e.g. rand(default_rng())
.
The type of the default RNG is an implementation detail. Across different versions of Julia, you should not expect the default RNG to always have the same type, nor that it will produce the same stream of random numbers for a given seed.
This function was introduced in Julia 1.3.
Random.seed!
— Functionseed!([rng=default_rng()], seed) -> rng
seed!([rng=default_rng()]) -> rng
Reseed the random number generator: rng
will give a reproducible sequence of numbers if and only if a seed
is provided. Some RNGs don't accept a seed, like RandomDevice
. After the call to seed!
, rng
is equivalent to a newly created object initialized with the same seed. The types of accepted seeds depend on the type of rng
, but in general, integer seeds should work.
If rng
is not specified, it defaults to seeding the state of the shared task-local generator.
Examples
julia> Random.seed!(1234);
julia> x1 = rand(2)
2-element Vector{Float64}:
0.32597672886359486
0.5490511363155669
julia> Random.seed!(1234);
julia> x2 = rand(2)
2-element Vector{Float64}:
0.32597672886359486
0.5490511363155669
julia> x1 == x2
true
julia> rng = Xoshiro(1234); rand(rng, 2) == x1
true
julia> Xoshiro(1) == Random.seed!(rng, 1)
true
julia> rand(Random.seed!(rng), Bool) # not reproducible
true
julia> rand(Random.seed!(rng), Bool) # not reproducible either
false
julia> rand(Xoshiro(), Bool) # not reproducible either
true
Random.AbstractRNG
— TypeAbstractRNG
Supertype for random number generators such as MersenneTwister
and RandomDevice
.
Random.TaskLocalRNG
— TypeTaskLocalRNG
The TaskLocalRNG
has state that is local to its task, not its thread. It is seeded upon task creation, from the state of its parent task, but without advancing the state of the parent's RNG.
As an upside, the TaskLocalRNG
is pretty fast, and permits reproducible multithreaded simulations (barring race conditions), independent of scheduler decisions. As long as the number of threads is not used to make decisions on task creation, simulation results are also independent of the number of available threads / CPUs. The random stream should not depend on hardware specifics, up to endianness and possibly word size.
Using or seeding the RNG of any other task than the one returned by current_task()
is undefined behavior: it will work most of the time, and may sometimes fail silently.
When seeding TaskLocalRNG()
with seed!
, the passed seed, if any, may be any integer.
Seeding TaskLocalRNG()
with a negative integer seed requires at least Julia 1.11.
Task creation no longer advances the parent task's RNG state as of Julia 1.10.
Random.Xoshiro
— TypeXoshiro(seed::Union{Integer, AbstractString})
Xoshiro()
Xoshiro256++ is a fast pseudorandom number generator described by David Blackman and Sebastiano Vigna in "Scrambled Linear Pseudorandom Number Generators", ACM Trans. Math. Softw., 2021. Reference implementation is available at https://prng.di.unimi.it
Apart from the high speed, Xoshiro has a small memory footprint, making it suitable for applications where many different random states need to be held for long time.
Julia's Xoshiro implementation has a bulk-generation mode; this seeds new virtual PRNGs from the parent, and uses SIMD to generate in parallel (i.e. the bulk stream consists of multiple interleaved xoshiro instances). The virtual PRNGs are discarded once the bulk request has been serviced (and should cause no heap allocations).
If no seed is provided, a randomly generated one is created (using entropy from the system). See the seed!
function for reseeding an already existing Xoshiro
object.
Passing a negative integer seed requires at least Julia 1.11.
Examples
julia> using Random
julia> rng = Xoshiro(1234);
julia> x1 = rand(rng, 2)
2-element Vector{Float64}:
0.32597672886359486
0.5490511363155669
julia> rng = Xoshiro(1234);
julia> x2 = rand(rng, 2)
2-element Vector{Float64}:
0.32597672886359486
0.5490511363155669
julia> x1 == x2
true
Random.MersenneTwister
— TypeMersenneTwister(seed)
MersenneTwister()
Create a MersenneTwister
RNG object. Different RNG objects can have their own seeds, which may be useful for generating different streams of random numbers. The seed
may be an integer, a string, or a vector of UInt32
integers. If no seed is provided, a randomly generated one is created (using entropy from the system). See the seed!
function for reseeding an already existing MersenneTwister
object.
Passing a negative integer seed requires at least Julia 1.11.
Examples
julia> rng = MersenneTwister(123);
julia> x1 = rand(rng, 2)
2-element Vector{Float64}:
0.37453777969575874
0.8735343642013971
julia> x2 = rand(MersenneTwister(123), 2)
2-element Vector{Float64}:
0.37453777969575874
0.8735343642013971
julia> x1 == x2
true
Random.RandomDevice
— TypeRandomDevice()
Create a RandomDevice
RNG object. Two such objects will always generate different streams of random numbers. The entropy is obtained from the operating system.
Hooking into the Random
API
There are two mostly orthogonal ways to extend Random
functionalities:
- generating random values of custom types
- creating new generators
The API for 1) is quite functional, but is relatively recent so it may still have to evolve in subsequent releases of the Random
module. For example, it's typically sufficient to implement one rand
method in order to have all other usual methods work automatically.
The API for 2) is still rudimentary, and may require more work than strictly necessary from the implementor, in order to support usual types of generated values.
Generating random values of custom types
Generating random values for some distributions may involve various trade-offs. Pre-computed values, such as an alias table for discrete distributions, or “squeezing” functions for univariate distributions, can speed up sampling considerably. How much information should be pre-computed can depend on the number of values we plan to draw from a distribution. Also, some random number generators can have certain properties that various algorithms may want to exploit.
The Random
module defines a customizable framework for obtaining random values that can address these issues. Each invocation of rand
generates a sampler which can be customized with the above trade-offs in mind, by adding methods to Sampler
, which in turn can dispatch on the random number generator, the object that characterizes the distribution, and a suggestion for the number of repetitions. Currently, for the latter, Val{1}
(for a single sample) and Val{Inf}
(for an arbitrary number) are used, with Random.Repetition
an alias for both.
The object returned by Sampler
is then used to generate the random values. When implementing the random generation interface for a value X
that can be sampled from, the implementor should define the method
rand(rng, sampler)
for the particular sampler
returned by Sampler(rng, X, repetition)
.
Samplers can be arbitrary values that implement rand(rng, sampler)
, but for most applications the following predefined samplers may be sufficient:
SamplerType{T}()
can be used for implementing samplers that draw from typeT
(e.g.rand(Int)
). This is the default returned bySampler
for types.SamplerTrivial(self)
is a simple wrapper forself
, which can be accessed with[]
. This is the recommended sampler when no pre-computed information is needed (e.g.rand(1:3)
), and is the default returned bySampler
for values.SamplerSimple(self, data)
also contains the additionaldata
field, which can be used to store arbitrary pre-computed values, which should be computed in a custom method ofSampler
.
We provide examples for each of these. We assume here that the choice of algorithm is independent of the RNG, so we use AbstractRNG
in our signatures.
Random.Sampler
— TypeSampler(rng, x, repetition = Val(Inf))
Return a sampler object that can be used to generate random values from rng
for x
.
When sp = Sampler(rng, x, repetition)
, rand(rng, sp)
will be used to draw random values, and should be defined accordingly.
repetition
can be Val(1)
or Val(Inf)
, and should be used as a suggestion for deciding the amount of precomputation, if applicable.
Random.SamplerType
and Random.SamplerTrivial
are default fallbacks for types and values, respectively. Random.SamplerSimple
can be used to store pre-computed values without defining extra types for only this purpose.
Random.SamplerType
— TypeSamplerType{T}()
A sampler for types, containing no other information. The default fallback for Sampler
when called with types.
Random.SamplerTrivial
— TypeSamplerTrivial(x)
Create a sampler that just wraps the given value x
. This is the default fall-back for values. The eltype
of this sampler is equal to eltype(x)
.
The recommended use case is sampling from values without precomputed data.
Random.SamplerSimple
— TypeSamplerSimple(x, data)
Create a sampler that wraps the given value x
and the data
. The eltype
of this sampler is equal to eltype(x)
.
The recommended use case is sampling from values with precomputed data.
Decoupling pre-computation from actually generating the values is part of the API, and is also available to the user. As an example, assume that rand(rng, 1:20)
has to be called repeatedly in a loop: the way to take advantage of this decoupling is as follows:
rng = Xoshiro()
sp = Random.Sampler(rng, 1:20) # or Random.Sampler(Xoshiro, 1:20)
for x in X
n = rand(rng, sp) # similar to n = rand(rng, 1:20)
# use n
end
This is the mechanism that is also used in the standard library, e.g. by the default implementation of random array generation (like in rand(1:20, 10)
).
Generating values from a type
Given a type T
, it's currently assumed that if rand(T)
is defined, an object of type T
will be produced. SamplerType
is the default sampler for types. In order to define random generation of values of type T
, the rand(rng::AbstractRNG, ::Random.SamplerType{T})
method should be defined, and should return values what rand(rng, T)
is expected to return.
Let's take the following example: we implement a Die
type, with a variable number n
of sides, numbered from 1
to n
. We want rand(Die)
to produce a Die
with a random number of up to 20 sides (and at least 4):
struct Die
nsides::Int # number of sides
end
Random.rand(rng::AbstractRNG, ::Random.SamplerType{Die}) = Die(rand(rng, 4:20))
# output
Scalar and array methods for Die
now work as expected:
julia> rand(Die)
Die(5)
julia> rand(Xoshiro(0), Die)
Die(10)
julia> rand(Die, 3)
3-element Vector{Die}:
Die(9)
Die(15)
Die(14)
julia> a = Vector{Die}(undef, 3); rand!(a)
3-element Vector{Die}:
Die(19)
Die(7)
Die(17)
A simple sampler without pre-computed data
Here we define a sampler for a collection. If no pre-computed data is required, it can be implemented with a SamplerTrivial
sampler, which is in fact the default fallback for values.
In order to define random generation out of objects of type S
, the following method should be defined: rand(rng::AbstractRNG, sp::Random.SamplerTrivial{S})
. Here, sp
simply wraps an object of type S
, which can be accessed via sp[]
. Continuing the Die
example, we want now to define rand(d::Die)
to produce an Int
corresponding to one of d
's sides:
julia> Random.rand(rng::AbstractRNG, d::Random.SamplerTrivial{Die}) = rand(rng, 1:d[].nsides);
julia> rand(Die(4))
1
julia> rand(Die(4), 3)
3-element Vector{Any}:
2
3
3
Given a collection type S
, it's currently assumed that if rand(::S)
is defined, an object of type eltype(S)
will be produced. In the last example, a Vector{Any}
is produced; the reason is that eltype(Die) == Any
. The remedy is to define Base.eltype(::Type{Die}) = Int
.
Generating values for an AbstractFloat
type
AbstractFloat
types are special-cased, because by default random values are not produced in the whole type domain, but rather in [0,1)
. The following method should be implemented for T <: AbstractFloat
: Random.rand(::AbstractRNG, ::Random.SamplerTrivial{Random.CloseOpen01{T}})
An optimized sampler with pre-computed data
Consider a discrete distribution, where numbers 1:n
are drawn with given probabilities that sum to one. When many values are needed from this distribution, the fastest method is using an alias table. We don't provide the algorithm for building such a table here, but suppose it is available in make_alias_table(probabilities)
instead, and draw_number(rng, alias_table)
can be used to draw a random number from it.
Suppose that the distribution is described by
struct DiscreteDistribution{V <: AbstractVector}
probabilities::V
end
and that we always want to build an alias table, regardless of the number of values needed (we learn how to customize this below). The methods
Random.eltype(::Type{<:DiscreteDistribution}) = Int
function Random.Sampler(::Type{<:AbstractRNG}, distribution::DiscreteDistribution, ::Repetition)
SamplerSimple(distribution, make_alias_table(distribution.probabilities))
end
should be defined to return a sampler with pre-computed data, then
function rand(rng::AbstractRNG, sp::SamplerSimple{<:DiscreteDistribution})
draw_number(rng, sp.data)
end
will be used to draw the values.
Custom sampler types
The SamplerSimple
type is sufficient for most use cases with precomputed data. However, in order to demonstrate how to use custom sampler types, here we implement something similar to SamplerSimple
.
Going back to our Die
example: rand(::Die)
uses random generation from a range, so there is an opportunity for this optimization. We call our custom sampler SamplerDie
.
import Random: Sampler, rand
struct SamplerDie <: Sampler{Int} # generates values of type Int
die::Die
sp::Sampler{Int} # this is an abstract type, so this could be improved
end
Sampler(RNG::Type{<:AbstractRNG}, die::Die, r::Random.Repetition) =
SamplerDie(die, Sampler(RNG, 1:die.nsides, r))
# the `r` parameter will be explained later on
rand(rng::AbstractRNG, sp::SamplerDie) = rand(rng, sp.sp)
It's now possible to get a sampler with sp = Sampler(rng, die)
, and use sp
instead of die
in any rand
call involving rng
. In the simplistic example above, die
doesn't need to be stored in SamplerDie
but this is often the case in practice.
Of course, this pattern is so frequent that the helper type used above, namely Random.SamplerSimple
, is available, saving us the definition of SamplerDie
: we could have implemented our decoupling with:
Sampler(RNG::Type{<:AbstractRNG}, die::Die, r::Random.Repetition) =
SamplerSimple(die, Sampler(RNG, 1:die.nsides, r))
rand(rng::AbstractRNG, sp::SamplerSimple{Die}) = rand(rng, sp.data)
Here, sp.data
refers to the second parameter in the call to the SamplerSimple
constructor (in this case equal to Sampler(rng, 1:die.nsides, r)
), while the Die
object can be accessed via sp[]
.
Like SamplerDie
, any custom sampler must be a subtype of Sampler{T}
where T
is the type of the generated values. Note that SamplerSimple(x, data) isa Sampler{eltype(x)}
, so this constrains what the first argument to SamplerSimple
can be (it's recommended to use SamplerSimple
like in the Die
example, where x
is simply forwarded while defining a Sampler
method). Similarly, SamplerTrivial(x) isa Sampler{eltype(x)}
.
Another helper type is currently available for other cases, Random.SamplerTag
, but is considered as internal API, and can break at any time without proper deprecations.
Using distinct algorithms for scalar or array generation
In some cases, whether one wants to generate only a handful of values or a large number of values will have an impact on the choice of algorithm. This is handled with the third parameter of the Sampler
constructor. Let's assume we defined two helper types for Die
, say SamplerDie1
which should be used to generate only few random values, and SamplerDieMany
for many values. We can use those types as follows:
Sampler(RNG::Type{<:AbstractRNG}, die::Die, ::Val{1}) = SamplerDie1(...)
Sampler(RNG::Type{<:AbstractRNG}, die::Die, ::Val{Inf}) = SamplerDieMany(...)
Of course, rand
must also be defined on those types (i.e. rand(::AbstractRNG, ::SamplerDie1)
and rand(::AbstractRNG, ::SamplerDieMany)
). Note that, as usual, SamplerTrivial
and SamplerSimple
can be used if custom types are not necessary.
Note: Sampler(rng, x)
is simply a shorthand for Sampler(rng, x, Val(Inf))
, and Random.Repetition
is an alias for Union{Val{1}, Val{Inf}}
.
Creating new generators
The API is not clearly defined yet, but as a rule of thumb:
- any
rand
method producing "basic" types (isbitstype
integer and floating types inBase
) should be defined for this specific RNG, if they are needed; - other documented
rand
methods accepting anAbstractRNG
should work out of the box, (provided the methods from 1) what are relied on are implemented), but can of course be specialized for this RNG if there is room for optimization; copy
for pseudo-RNGs should return an independent copy that generates the exact same random sequence as the original from that point when called in the same way. When this is not feasible (e.g. hardware-based RNGs),copy
must not be implemented.
Concerning 1), a rand
method may happen to work automatically, but it's not officially supported and may break without warnings in a subsequent release.
To define a new rand
method for an hypothetical MyRNG
generator, and a value specification s
(e.g. s == Int
, or s == 1:10
) of type S==typeof(s)
or S==Type{s}
if s
is a type, the same two methods as we saw before must be defined:
Sampler(::Type{MyRNG}, ::S, ::Repetition)
, which returns an object of type saySamplerS
rand(rng::MyRNG, sp::SamplerS)
It can happen that Sampler(rng::AbstractRNG, ::S, ::Repetition)
is already defined in the Random
module. It would then be possible to skip step 1) in practice (if one wants to specialize generation for this particular RNG type), but the corresponding SamplerS
type is considered as internal detail, and may be changed without warning.
Specializing array generation
In some cases, for a given RNG type, generating an array of random values can be more efficient with a specialized method than by merely using the decoupling technique explained before. This is for example the case for MersenneTwister
, which natively writes random values in an array.
To implement this specialization for MyRNG
and for a specification s
, producing elements of type S
, the following method can be defined: rand!(rng::MyRNG, a::AbstractArray{S}, ::SamplerS)
, where SamplerS
is the type of the sampler returned by Sampler(MyRNG, s, Val(Inf))
. Instead of AbstractArray
, it's possible to implement the functionality only for a subtype, e.g. Array{S}
. The non-mutating array method of rand
will automatically call this specialization internally.
Reproducibility
By using an RNG parameter initialized with a given seed, you can reproduce the same pseudorandom number sequence when running your program multiple times. However, a minor release of Julia (e.g. 1.3 to 1.4) may change the sequence of pseudorandom numbers generated from a specific seed, in particular if MersenneTwister
is used. (Even if the sequence produced by a low-level function like rand
does not change, the output of higher-level functions like randsubseq
may change due to algorithm updates.) Rationale: guaranteeing that pseudorandom streams never change prohibits many algorithmic improvements.
If you need to guarantee exact reproducibility of random data, it is advisable to simply save the data (e.g. as a supplementary attachment in a scientific publication). (You can also, of course, specify a particular Julia version and package manifest, especially if you require bit reproducibility.)
Software tests that rely on specific "random" data should also generally either save the data, embed it into the test code, or use third-party packages like StableRNGs.jl. On the other hand, tests that should pass for most random data (e.g. testing A \ (A*x) ≈ x
for a random matrix A = randn(n,n)
) can use an RNG with a fixed seed to ensure that simply running the test many times does not encounter a failure due to very improbable data (e.g. an extremely ill-conditioned matrix).
The statistical distribution from which random samples are drawn is guaranteed to be the same across any minor Julia releases.