Can rand generate random vectors with custom type entries?

General Usage
1

I’m trying to understand how to implement rand for custom types. After looking at the documentation, I’ve come up with the following:

using Random

struct A{T}
    x::T
end

Random.rand(rng::AbstractRNG, ::Type{A{T}}) where T = A(rand(rng, T))

This generates single elements as expected:

julia> rand(A{Bool})
A{Bool}(false)

I thought that generating random vectors would work automatically, but it doesn’t:

julia> rand(A{Bool}, 3)
ERROR: MethodError: no method matching Random.Sampler(::Type{TaskLocalRNG}, ::Random.SamplerType{A{Bool}}, ::Val{1})

Closest candidates are:
  Random.Sampler(::Type{<:AbstractRNG}, ::Random.Sampler, ::Union{Val{1}, Val{Inf}})
   @ Random /usr/local/julia-1.10.0-rc1/share/julia/stdlib/v1.10/Random/src/Random.jl:147
  Random.Sampler(::Type{<:AbstractRNG}, ::Any, ::Union{Val{1}, Val{Inf}})
   @ Random /usr/local/julia-1.10.0-rc1/share/julia/stdlib/v1.10/Random/src/Random.jl:183
  Random.Sampler(::Type{<:AbstractRNG}, ::BitSet, ::Union{Val{1}, Val{Inf}})
   @ Random /usr/local/julia-1.10.0-rc1/share/julia/stdlib/v1.10/Random/src/generation.jl:450
  ...

Stacktrace:
 [1] Random.Sampler(T::Type{TaskLocalRNG}, sp::Random.SamplerType{A{Bool}}, r::Val{1})
   @ Random /usr/local/julia-1.10.0-rc1/share/julia/stdlib/v1.10/Random/src/Random.jl:147
 [2] Random.Sampler(rng::TaskLocalRNG, x::Random.SamplerType{A{Bool}}, r::Val{1})
   @ Random /usr/local/julia-1.10.0-rc1/share/julia/stdlib/v1.10/Random/src/Random.jl:139
 [3] rand(rng::TaskLocalRNG, X::Random.SamplerType{A{Bool}})
   @ Random /usr/local/julia-1.10.0-rc1/share/julia/stdlib/v1.10/Random/src/Random.jl:255
 [4] rand!
   @ Random /usr/local/julia-1.10.0-rc1/share/julia/stdlib/v1.10/Random/src/Random.jl:273 [inlined]
 [5] rand!
   @ Random /usr/local/julia-1.10.0-rc1/share/julia/stdlib/v1.10/Random/src/Random.jl:269 [inlined]
 [6] rand
   @ Random /usr/local/julia-1.10.0-rc1/share/julia/stdlib/v1.10/Random/src/Random.jl:290 [inlined]
 [7] rand
   @ Random /usr/local/julia-1.10.0-rc1/share/julia/stdlib/v1.10/Random/src/Random.jl:291 [inlined]
 [8] rand(::Type{A{Bool}}, ::Int64)
   @ Random /usr/local/julia-1.10.0-rc1/share/julia/stdlib/v1.10/Random/src/Random.jl:294
 [9] top-level scope
   @ REPL[5]:1

I find this particularly surprising because the first closest candidate displayed above should match in my opinion because

julia> TaskLocalRNG <: AbstractRNG, Random.SamplerType{A{Bool}} <: Random.Sampler
(true, true)

(The second should also match, but it’s less specific.) Am I missing anything?

2

(Note that you can always just do [rand(T) for _=1:n].)

3

Sure. But should rand(A{Bool}, 3) work? (So my question is rather about the API of Random. I’ve clarified the topic title.)

4

https://docs.julialang.org/en/v1/stdlib/Random/#Hooking-into-the-Random-API

1 Like
5

It does seem like there should be a default Sampler for SamplerType{T} that simply calls rand(rng, T) …

cc @rfourquet

6

Just to make sure the solution is spelled out here: the one change you need is to replace Type with Random.SamplerType in your method definition.

using Random

struct A{T}
    x::T
end

Random.rand(rng::AbstractRNG, ::Random.SamplerType{A{T}}) where {T} = A(rand(rng, T))

julia> rand(A{Bool})
A{Bool}(false)

julia> rand(A{Bool}, 3)
3-element Vector{A{Bool}}:
 A{Bool}(true)
 A{Bool}(false)
 A{Bool}(false)
5 Likes