use ndarray::Array1;
use stochastic_rs_core::simd_rng::Deterministic;
use stochastic_rs_core::simd_rng::SeedExt;
use stochastic_rs_core::simd_rng::Unseeded;
use stochastic_rs_distributions::normal::SimdNormal;
use crate::traits::FloatExt;
use crate::traits::ProcessExt;
#[derive(Copy, Clone)]
pub struct Wn<T: FloatExt, S: SeedExt = Unseeded> {
pub n: usize,
pub mean: Option<T>,
pub std_dev: Option<T>,
pub seed: S,
}
impl<T: FloatExt> Wn<T> {
pub fn new(n: usize, mean: Option<T>, std_dev: Option<T>) -> Self {
Wn {
n,
mean,
std_dev,
seed: Unseeded,
}
}
}
impl<T: FloatExt> Wn<T, Deterministic> {
pub fn seeded(n: usize, mean: Option<T>, std_dev: Option<T>, seed: u64) -> Self {
Wn {
n,
mean,
std_dev,
seed: Deterministic::new(seed),
}
}
}
impl<T: FloatExt, S: SeedExt> ProcessExt<T> for Wn<T, S> {
type Output = Array1<T>;
fn sample(&self) -> Self::Output {
let mean = self.mean.unwrap_or(T::zero());
let std_dev = self.std_dev.unwrap_or(T::one());
let mut out = Array1::<T>::zeros(self.n);
if self.n == 0 {
return out;
}
let out_slice = out.as_slice_mut().expect("Wn output must be contiguous");
let normal = SimdNormal::<T>::from_seed_source(mean, std_dev, &self.seed);
normal.fill_slice_fast(out_slice);
out
}
}
py_process_1d!(PyWn, Wn,
sig: (n, mean=None, std_dev=None, seed=None, dtype=None),
params: (n: usize, mean: Option<f64>, std_dev: Option<f64>)
);