API

AbstractPythonRNG <: Random.AbstractRNG

Supertype of random number generators that delegate sampling to a Python object through PythonCall.jl.

Concrete subtypes:

• PythonRandom: backed by Python's standard random.Random.
• NumPyRandomDefaultRNG: backed by numpy.random.default_rng.
• NumPyRandom: backed by the legacy numpy.random.RandomState.

PythonRandom([seed])

Create an AbstractRNG that draws random numbers from Python's standard random.Random.

rand(rng, Float64) returns exactly the value that random.Random(seed).random() would return on the same draw. When seed is omitted the generator is seeded from the operating system entropy.

NumPyRandomDefaultRNG([seed])

Create an AbstractRNG that draws random numbers from NumPy's numpy.random.default_rng.

rand(rng, Float64) returns exactly the value that default_rng(seed).random() would return on the same draw. When seed is omitted the generator is seeded from the operating system entropy.

NumPyRandom([seed])

Create an AbstractRNG that draws random numbers from NumPy's legacy numpy.random.RandomState, i.e. the generator behind np.random.seed(...) and np.random.random(...).

rand(rng, Float64) returns exactly the value that RandomState(seed).random_sample() would return on the same draw. When seed is omitted the generator is seeded from the operating system entropy.

Random.seed!(rng::AbstractPythonRNG, [seed])

Reseed rng. With an integer seed, the wrapped Python generator is replaced by a fresh one created from seed, so the stream restarts exactly as it would from a newly constructed generator. When seed is omitted (or nothing), the generator is seeded from operating system entropy.

copy(rng::AbstractPythonRNG)

Return an independent copy of rng that starts from the same state, using Python's copy.deepcopy on the wrapped generator.

NotSupportedError(msg)

Thrown by PythonRandom, NumPyRandomDefaultRNG, or NumPyRandom when a requested draw has no equivalent in the wrapped Python backend.