vector_ta/indicators/moving_averages/

wma.rs

#[cfg(all(feature = "python", feature = "cuda"))]
use crate::cuda::cuda_available;
#[cfg(all(feature = "python", feature = "cuda"))]
use crate::cuda::moving_averages::wma_wrapper::DeviceArrayF32Py;
#[cfg(all(feature = "python", feature = "cuda"))]
use crate::cuda::moving_averages::CudaWma;
#[cfg(feature = "python")]
use crate::utilities::kernel_validation::validate_kernel;
#[cfg(feature = "python")]
use numpy::{IntoPyArray, PyArray1, PyArrayMethods, PyReadonlyArray1, PyReadonlyArray2};
#[cfg(feature = "python")]
use pyo3::exceptions::PyValueError;
#[cfg(feature = "python")]
use pyo3::prelude::*;
#[cfg(feature = "python")]
use pyo3::types::PyDict;
#[cfg(all(target_arch = "wasm32", feature = "wasm"))]
use serde::{Deserialize, Serialize};
#[cfg(all(target_arch = "wasm32", feature = "wasm"))]
use wasm_bindgen::prelude::*;

use crate::utilities::data_loader::{source_type, Candles};
use crate::utilities::enums::Kernel;
use crate::utilities::helpers::{
    alloc_with_nan_prefix, detect_best_batch_kernel, detect_best_kernel, init_matrix_prefixes,
    make_uninit_matrix,
};
use aligned_vec::{AVec, CACHELINE_ALIGN};
#[cfg(all(feature = "nightly-avx", target_arch = "x86_64"))]
use core::arch::x86_64::*;
#[cfg(not(target_arch = "wasm32"))]
use rayon::prelude::*;
use std::convert::AsRef;
use std::error::Error;
use std::mem::MaybeUninit;
use thiserror::Error;

impl<'a> AsRef<[f64]> for WmaInput<'a> {
    #[inline(always)]
    fn as_ref(&self) -> &[f64] {
        match &self.data {
            WmaData::Slice(slice) => slice,
            WmaData::Candles { candles, source } => source_type(candles, source),
        }
    }
}

#[derive(Debug, Clone)]
pub enum WmaData<'a> {
    Candles {
        candles: &'a Candles,
        source: &'a str,
    },
    Slice(&'a [f64]),
}

#[derive(Debug, Clone)]
pub struct WmaOutput {
    pub values: Vec<f64>,
}

#[derive(Debug, Clone)]
#[cfg_attr(
    all(target_arch = "wasm32", feature = "wasm"),
    derive(Serialize, Deserialize)
)]
pub struct WmaParams {
    pub period: Option<usize>,
}

impl Default for WmaParams {
    fn default() -> Self {
        Self { period: Some(30) }
    }
}

#[derive(Debug, Clone)]
pub struct WmaInput<'a> {
    pub data: WmaData<'a>,
    pub params: WmaParams,
}

impl<'a> WmaInput<'a> {
    #[inline]
    pub fn from_candles(c: &'a Candles, s: &'a str, p: WmaParams) -> Self {
        Self {
            data: WmaData::Candles {
                candles: c,
                source: s,
            },
            params: p,
        }
    }
    #[inline]
    pub fn from_slice(sl: &'a [f64], p: WmaParams) -> Self {
        Self {
            data: WmaData::Slice(sl),
            params: p,
        }
    }
    #[inline]
    pub fn with_default_candles(c: &'a Candles) -> Self {
        Self::from_candles(c, "close", WmaParams::default())
    }
    #[inline]
    pub fn get_period(&self) -> usize {
        self.params.period.unwrap_or(30)
    }
}

#[derive(Copy, Clone, Debug)]
pub struct WmaBuilder {
    period: Option<usize>,
    kernel: Kernel,
}

impl Default for WmaBuilder {
    fn default() -> Self {
        Self {
            period: None,
            kernel: Kernel::Auto,
        }
    }
}

impl WmaBuilder {
    #[inline(always)]
    pub fn new() -> Self {
        Self::default()
    }
    #[inline(always)]
    pub fn period(mut self, n: usize) -> Self {
        self.period = Some(n);
        self
    }
    #[inline(always)]
    pub fn kernel(mut self, k: Kernel) -> Self {
        self.kernel = k;
        self
    }

    #[inline(always)]
    pub fn apply(self, c: &Candles) -> Result<WmaOutput, WmaError> {
        let p = WmaParams {
            period: self.period,
        };
        let i = WmaInput::from_candles(c, "close", p);
        wma_with_kernel(&i, self.kernel)
    }

    #[inline(always)]
    pub fn apply_slice(self, d: &[f64]) -> Result<WmaOutput, WmaError> {
        let p = WmaParams {
            period: self.period,
        };
        let i = WmaInput::from_slice(d, p);
        wma_with_kernel(&i, self.kernel)
    }

    #[inline(always)]
    pub fn into_stream(self) -> Result<WmaStream, WmaError> {
        let p = WmaParams {
            period: self.period,
        };
        WmaStream::try_new(p)
    }
}

#[derive(Debug, Error)]
pub enum WmaError {
    #[error("wma: Input data slice is empty.")]
    EmptyInputData,

    #[error("wma: All values are NaN.")]
    AllValuesNaN,

    #[error("wma: Invalid period: period = {period}, data length = {data_len}")]
    InvalidPeriod { period: usize, data_len: usize },

    #[error("wma: Not enough valid data: needed = {needed}, valid = {valid}")]
    NotEnoughValidData { needed: usize, valid: usize },

    #[error("wma: Non-batch kernel passed to batch path: {0:?}")]
    InvalidKernelForBatch(crate::utilities::enums::Kernel),

    #[error("wma: Output length mismatch: expected = {expected}, got = {got}")]
    OutputLengthMismatch { expected: usize, got: usize },

    #[error("wma: Invalid range expansion: start = {start}, end = {end}, step = {step}")]
    InvalidRange {
        start: usize,
        end: usize,
        step: usize,
    },

    #[error("wma: invalid input: {0}")]
    InvalidInput(String),
}

#[inline]
pub fn wma(input: &WmaInput) -> Result<WmaOutput, WmaError> {
    wma_with_kernel(input, Kernel::Auto)
}

pub fn wma_with_kernel(input: &WmaInput, kernel: Kernel) -> Result<WmaOutput, WmaError> {
    let (data, period, first, chosen) = wma_prepare(input, kernel)?;
    let len = data.len();
    let warm = first + period - 1;
    let mut out = alloc_with_nan_prefix(len, warm);

    wma_compute_into(data, period, first, chosen, &mut out);

    Ok(WmaOutput { values: out })
}

#[inline]
pub fn wma_into_slice(dst: &mut [f64], input: &WmaInput, kern: Kernel) -> Result<(), WmaError> {
    let (data, period, first, chosen) = wma_prepare(input, kern)?;

    if dst.len() != data.len() {
        return Err(WmaError::OutputLengthMismatch {
            expected: data.len(),
            got: dst.len(),
        });
    }

    wma_compute_into(data, period, first, chosen, dst);

    let warmup_end = first + period - 1;
    for v in &mut dst[..warmup_end] {
        *v = f64::NAN;
    }

    Ok(())
}

#[cfg(not(all(target_arch = "wasm32", feature = "wasm")))]
#[inline]
pub fn wma_into(input: &WmaInput, out: &mut [f64]) -> Result<(), WmaError> {
    wma_into_slice(out, input, Kernel::Auto)
}

fn wma_prepare<'a>(
    input: &'a WmaInput,
    kernel: Kernel,
) -> Result<(&'a [f64], usize, usize, Kernel), WmaError> {
    let data: &[f64] = input.as_ref();
    let len = data.len();
    if len == 0 {
        return Err(WmaError::EmptyInputData);
    }

    let first = data
        .iter()
        .position(|x| !x.is_nan())
        .ok_or(WmaError::AllValuesNaN)?;
    let period = input.get_period();

    if period < 2 || period > len {
        return Err(WmaError::InvalidPeriod {
            period,
            data_len: len,
        });
    }
    if len - first < period {
        return Err(WmaError::NotEnoughValidData {
            needed: period,
            valid: len - first,
        });
    }

    let chosen = match kernel {
        Kernel::Auto => Kernel::Scalar,
        k => k,
    };

    Ok((data, period, first, chosen))
}

#[inline(always)]
fn wma_compute_into(data: &[f64], period: usize, first: usize, kernel: Kernel, out: &mut [f64]) {
    unsafe {
        match kernel {
            Kernel::Scalar | Kernel::ScalarBatch => wma_scalar(data, period, first, out),
            #[cfg(all(feature = "nightly-avx", target_arch = "x86_64"))]
            Kernel::Avx2 | Kernel::Avx2Batch => wma_avx2(data, period, first, out),
            #[cfg(all(feature = "nightly-avx", target_arch = "x86_64"))]
            Kernel::Avx512 | Kernel::Avx512Batch => wma_avx512(data, period, first, out),
            _ => unreachable!(),
        }
    }
}

#[inline]
pub fn wma_scalar(data: &[f64], period: usize, first_val: usize, out: &mut [f64]) {
    debug_assert_eq!(out.len(), data.len());
    let lookback = period - 1;
    let period_f = period as f64;

    let weights = period_f * (period_f + 1.0) * 0.5;

    unsafe {
        let base = data.as_ptr().add(first_val);
        let end = data.as_ptr().add(data.len());

        let mut sum = 0.0_f64;
        let mut weight_sum = 0.0_f64;

        let mut k = 0usize;
        while k < lookback {
            let v = *base.add(k);
            weight_sum += v * (k as f64 + 1.0);
            sum += v;
            k += 1;
        }

        let mut in_new = base.add(lookback);
        let mut in_old = base;
        let mut out_ptr = out.as_mut_ptr().add(first_val + lookback);

        while in_new < end {
            let v = *in_new;
            weight_sum += v * period_f;
            sum += v;

            *out_ptr = weight_sum / weights;

            weight_sum -= sum;
            sum -= *in_old;

            in_new = in_new.add(1);
            in_old = in_old.add(1);
            out_ptr = out_ptr.add(1);
        }
    }
}

#[cfg(all(feature = "nightly-avx", target_arch = "x86_64"))]
#[inline]
pub fn wma_avx2(data: &[f64], period: usize, first_valid: usize, out: &mut [f64]) {
    wma_scalar(data, period, first_valid, out)
}

#[cfg(all(feature = "nightly-avx", target_arch = "x86_64"))]
#[inline]
pub fn wma_avx512(data: &[f64], period: usize, first_valid: usize, out: &mut [f64]) {
    wma_scalar(data, period, first_valid, out)
}

#[inline]
pub fn wma_avx512_short(data: &[f64], period: usize, first_valid: usize, out: &mut [f64]) {
    wma_scalar(data, period, first_valid, out)
}

#[inline]
pub fn wma_avx512_long(data: &[f64], period: usize, first_valid: usize, out: &mut [f64]) {
    wma_scalar(data, period, first_valid, out)
}

#[inline(always)]
pub fn wma_with_kernel_batch(
    data: &[f64],
    sweep: &WmaBatchRange,
    k: Kernel,
) -> Result<WmaBatchOutput, WmaError> {
    let kernel = match k {
        Kernel::Auto => detect_best_batch_kernel(),
        other if other.is_batch() => other,
        _ => return Err(WmaError::InvalidKernelForBatch(k)),
    };

    let simd = match kernel {
        #[cfg(all(feature = "nightly-avx", target_arch = "x86_64"))]
        Kernel::Avx512Batch => Kernel::Avx512,
        #[cfg(all(feature = "nightly-avx", target_arch = "x86_64"))]
        Kernel::Avx2Batch => Kernel::Avx2,
        Kernel::ScalarBatch => Kernel::Scalar,
        _ => Kernel::Scalar,
    };
    wma_batch_par_slice(data, sweep, simd)
}

#[derive(Clone, Debug)]
pub struct WmaStream {
    period: usize,
    buffer: Vec<f64>,
    head: usize,
    filled: bool,

    plain_sum: f64,
    weighted_sum: f64,
    inv_div: f64,
    p_f64: f64,
}

impl WmaStream {
    pub fn try_new(params: WmaParams) -> Result<Self, WmaError> {
        let period = params.period.unwrap_or(30);
        if period < 2 {
            return Err(WmaError::InvalidPeriod {
                period,
                data_len: 0,
            });
        }

        let sum_of_weights = (period * (period + 1)) as f64 * 0.5;
        Ok(Self {
            period,
            buffer: vec![f64::NAN; period],
            head: 0,
            filled: false,
            plain_sum: 0.0,
            weighted_sum: 0.0,
            inv_div: 1.0 / sum_of_weights,
            p_f64: period as f64,
        })
    }

    #[inline(always)]
    pub fn update(&mut self, value: f64) -> Option<f64> {
        let write_idx = self.head;
        self.buffer[write_idx] = value;

        self.head += 1;
        if self.head == self.period {
            self.head = 0;
        }

        if !self.filled {
            if self.head == 0 {
                let mut wsum = 0.0;
                let mut ssum = 0.0;

                let mut idx = self.head;
                for w in 1..=self.period {
                    let v = self.buffer[idx];
                    ssum += v;
                    wsum += (w as f64) * v;
                    idx += 1;
                    if idx == self.period {
                        idx = 0;
                    }
                }
                let out = wsum * self.inv_div;

                self.weighted_sum = wsum - ssum;
                let oldest_next = self.buffer[self.head];
                self.plain_sum = ssum - oldest_next;

                self.filled = true;
                Some(out)
            } else {
                None
            }
        } else {
            let oldest = self.buffer[self.head];
            self.weighted_sum += self.p_f64 * value;
            self.plain_sum += value;

            let out = self.weighted_sum * self.inv_div;

            self.weighted_sum -= self.plain_sum;
            self.plain_sum -= oldest;

            Some(out)
        }
    }
}

#[derive(Clone, Debug)]
pub struct WmaBatchRange {
    pub period: (usize, usize, usize),
}

impl Default for WmaBatchRange {
    fn default() -> Self {
        Self {
            period: (2, 251, 1),
        }
    }
}

#[derive(Clone, Debug, Default)]
pub struct WmaBatchBuilder {
    range: WmaBatchRange,
    kernel: Kernel,
}

impl WmaBatchBuilder {
    pub fn new() -> Self {
        Self::default()
    }
    pub fn kernel(mut self, k: Kernel) -> Self {
        self.kernel = k;
        self
    }

    #[inline]
    pub fn period_range(mut self, start: usize, end: usize, step: usize) -> Self {
        self.range.period = (start, end, step);
        self
    }
    #[inline]
    pub fn period_static(mut self, p: usize) -> Self {
        self.range.period = (p, p, 0);
        self
    }

    pub fn apply_slice(self, data: &[f64]) -> Result<WmaBatchOutput, WmaError> {
        wma_with_kernel_batch(data, &self.range, self.kernel)
    }

    pub fn with_default_slice(data: &[f64], k: Kernel) -> Result<WmaBatchOutput, WmaError> {
        WmaBatchBuilder::new().kernel(k).apply_slice(data)
    }

    pub fn apply_candles(self, c: &Candles, src: &str) -> Result<WmaBatchOutput, WmaError> {
        let slice = source_type(c, src);
        self.apply_slice(slice)
    }

    pub fn with_default_candles(c: &Candles) -> Result<WmaBatchOutput, WmaError> {
        WmaBatchBuilder::new()
            .kernel(Kernel::Auto)
            .apply_candles(c, "close")
    }
}

#[derive(Clone, Debug)]
pub struct WmaBatchOutput {
    pub values: Vec<f64>,
    pub combos: Vec<WmaParams>,
    pub rows: usize,
    pub cols: usize,
}
impl WmaBatchOutput {
    pub fn row_for_params(&self, p: &WmaParams) -> Option<usize> {
        self.combos
            .iter()
            .position(|c| c.period.unwrap_or(30) == p.period.unwrap_or(30))
    }

    pub fn values_for(&self, p: &WmaParams) -> Option<&[f64]> {
        self.row_for_params(p).map(|row| {
            let start = row * self.cols;
            &self.values[start..start + self.cols]
        })
    }
}

#[inline(always)]
fn expand_grid(r: &WmaBatchRange) -> Vec<WmaParams> {
    fn axis_usize((start, end, step): (usize, usize, usize)) -> Vec<usize> {
        if step == 0 || start == end {
            return vec![start];
        }
        if start < end {
            return (start..=end).step_by(step.max(1)).collect();
        }

        let mut out = Vec::new();
        let mut x = start as isize;
        let end_i = end as isize;
        let st = (step as isize).max(1);
        while x >= end_i {
            out.push(x as usize);
            x -= st;
        }
        if out.is_empty() {
            return out;
        }
        if *out.last().unwrap() != end {
            out.push(end);
        }
        out
    }

    let periods = axis_usize(r.period);
    let mut out = Vec::with_capacity(periods.len());
    for &p in &periods {
        out.push(WmaParams { period: Some(p) });
    }
    out
}

#[inline(always)]
pub fn wma_batch_slice(
    data: &[f64],
    sweep: &WmaBatchRange,
    kern: Kernel,
) -> Result<WmaBatchOutput, WmaError> {
    wma_batch_inner(data, sweep, kern, false)
}

#[inline(always)]
pub fn wma_batch_par_slice(
    data: &[f64],
    sweep: &WmaBatchRange,
    kern: Kernel,
) -> Result<WmaBatchOutput, WmaError> {
    wma_batch_inner(data, sweep, kern, true)
}

#[inline(always)]
fn wma_batch_inner(
    data: &[f64],
    sweep: &WmaBatchRange,
    kern: Kernel,
    parallel: bool,
) -> Result<WmaBatchOutput, WmaError> {
    let combos = expand_grid(sweep);
    let rows = combos.len();
    let cols = data.len();
    if cols == 0 {
        return Err(WmaError::EmptyInputData);
    }

    rows.checked_mul(cols)
        .ok_or_else(|| WmaError::InvalidInput("rows*cols overflow".into()))?;

    let mut buf_mu = make_uninit_matrix(rows, cols);

    let first = data
        .iter()
        .position(|x| !x.is_nan())
        .ok_or(WmaError::AllValuesNaN)?;
    let warm: Vec<usize> = combos
        .iter()
        .map(|c| first + c.period.unwrap() - 1)
        .collect();
    init_matrix_prefixes(&mut buf_mu, cols, &warm);

    let mut guard = core::mem::ManuallyDrop::new(buf_mu);
    let out: &mut [f64] =
        unsafe { core::slice::from_raw_parts_mut(guard.as_mut_ptr() as *mut f64, guard.len()) };

    wma_batch_inner_into(data, sweep, kern, parallel, out)?;

    let values = unsafe {
        Vec::from_raw_parts(
            guard.as_mut_ptr() as *mut f64,
            guard.len(),
            guard.capacity(),
        )
    };

    Ok(WmaBatchOutput {
        values,
        combos,
        rows,
        cols,
    })
}

#[inline(always)]
fn wma_batch_inner_into(
    data: &[f64],
    sweep: &WmaBatchRange,
    kern: Kernel,
    parallel: bool,
    out: &mut [f64],
) -> Result<Vec<WmaParams>, WmaError> {
    let combos = expand_grid(sweep);
    if combos.is_empty() {
        let (start, end, step) = sweep.period;
        return Err(WmaError::InvalidRange { start, end, step });
    }

    let first = data
        .iter()
        .position(|x| !x.is_nan())
        .ok_or(WmaError::AllValuesNaN)?;
    let max_p = combos.iter().map(|c| c.period.unwrap()).max().unwrap();
    if data.len() - first < max_p {
        return Err(WmaError::NotEnoughValidData {
            needed: max_p,
            valid: data.len() - first,
        });
    }

    let rows = combos.len();
    let cols = data.len();

    let needed = rows
        .checked_mul(cols)
        .ok_or_else(|| WmaError::InvalidInput("rows*cols overflow".into()))?;
    if out.len() != needed {
        return Err(WmaError::OutputLengthMismatch {
            expected: needed,
            got: out.len(),
        });
    }

    let warm: Vec<usize> = combos
        .iter()
        .map(|c| first + c.period.unwrap() - 1)
        .collect();

    let out_uninit = unsafe {
        std::slice::from_raw_parts_mut(out.as_mut_ptr() as *mut MaybeUninit<f64>, out.len())
    };

    unsafe { init_matrix_prefixes(out_uninit, cols, &warm) };

    let cols = data.len();
    let mut pref_a = AVec::<f64>::with_capacity(CACHELINE_ALIGN, cols + 1);
    let mut pref_b = AVec::<f64>::with_capacity(CACHELINE_ALIGN, cols + 1);
    unsafe {
        pref_a.set_len(cols + 1);
        pref_b.set_len(cols + 1);
    }
    pref_a[0] = 0.0;
    pref_b[0] = 0.0;
    for i in 0..cols {
        let x = if i < first { 0.0 } else { data[i] };
        pref_a[i + 1] = pref_a[i] + x;
        pref_b[i + 1] = pref_b[i] + (i as f64) * x;
    }

    let do_row = |row: usize, dst_mu: &mut [MaybeUninit<f64>]| unsafe {
        let period = combos[row].period.unwrap();
        let denom = (period * (period + 1)) as f64 / 2.0;
        let inv_div = 1.0 / denom;
        let warm_end = first + period - 1;

        let out_row =
            core::slice::from_raw_parts_mut(dst_mu.as_mut_ptr() as *mut f64, dst_mu.len());

        for i in warm_end..cols {
            let s_a = pref_a[i + 1] - pref_a[i + 1 - period];
            let s_b = pref_b[i + 1] - pref_b[i + 1 - period];
            let wsum = s_b - ((i + 1 - period) as f64 - 1.0) * s_a;
            out_row[i] = wsum * inv_div;
        }
    };

    if parallel {
        #[cfg(not(target_arch = "wasm32"))]
        {
            out_uninit
                .par_chunks_mut(cols)
                .enumerate()
                .for_each(|(row, slice)| do_row(row, slice));
        }

        #[cfg(target_arch = "wasm32")]
        {
            for (row, slice) in out_uninit.chunks_mut(cols).enumerate() {
                do_row(row, slice);
            }
        }
    } else {
        for (row, slice) in out_uninit.chunks_mut(cols).enumerate() {
            do_row(row, slice);
        }
    }

    Ok(combos)
}

#[inline(always)]
unsafe fn wma_row_scalar(data: &[f64], first: usize, period: usize, out: &mut [f64]) {
    wma_scalar(data, period, first, out)
}

#[cfg(all(feature = "nightly-avx", target_arch = "x86_64"))]
#[inline(always)]
unsafe fn wma_row_avx2(data: &[f64], first: usize, period: usize, out: &mut [f64]) {
    wma_row_scalar(data, first, period, out)
}

#[cfg(all(feature = "nightly-avx", target_arch = "x86_64"))]
#[inline(always)]
pub unsafe fn wma_row_avx512(data: &[f64], first: usize, period: usize, out: &mut [f64]) {
    if period <= 32 {
        wma_row_avx512_short(data, first, period, out);
    } else {
        wma_row_avx512_long(data, first, period, out);
    }
}

#[cfg(all(feature = "nightly-avx", target_arch = "x86_64"))]
#[inline(always)]
unsafe fn wma_row_avx512_short(data: &[f64], first: usize, period: usize, out: &mut [f64]) {
    wma_row_scalar(data, first, period, out)
}

#[cfg(all(feature = "nightly-avx", target_arch = "x86_64"))]
#[inline(always)]
unsafe fn wma_row_avx512_long(data: &[f64], first: usize, period: usize, out: &mut [f64]) {
    wma_row_scalar(data, first, period, out)
}

#[cfg(test)]
mod tests {
    use super::*;
    use crate::skip_if_unsupported;
    use crate::utilities::data_loader::read_candles_from_csv;
    use paste::paste;

    #[cfg(not(all(target_arch = "wasm32", feature = "wasm")))]
    #[test]
    fn test_wma_into_matches_api() -> Result<(), Box<dyn Error>> {
        let mut data = Vec::with_capacity(256);

        data.extend_from_slice(&[f64::NAN, f64::NAN, f64::NAN, f64::NAN]);
        for i in 0..252u32 {
            let v = 0.5 * (i as f64) + ((i % 7) as f64);
            data.push(v);
        }

        let params = WmaParams { period: Some(30) };
        let input = WmaInput::from_slice(&data, params);

        let baseline = wma(&input)?.values;

        let mut out = vec![0.0; data.len()];
        wma_into(&input, &mut out)?;

        assert_eq!(baseline.len(), out.len());
        for (i, (&a, &b)) in baseline.iter().zip(out.iter()).enumerate() {
            let equal = (a.is_nan() && b.is_nan()) || (a == b);
            assert!(equal, "Mismatch at index {}: api={} into={}", i, a, b);
        }

        Ok(())
    }

    fn check_wma_partial_params(test_name: &str, kernel: Kernel) -> Result<(), Box<dyn Error>> {
        skip_if_unsupported!(kernel, test_name);
        let file_path = "src/data/2018-09-01-2024-Bitfinex_Spot-4h.csv";
        let candles = read_candles_from_csv(file_path)?;

        let default_params = WmaParams { period: None };
        let input = WmaInput::from_candles(&candles, "close", default_params);
        let output = wma_with_kernel(&input, kernel)?;
        assert_eq!(output.values.len(), candles.close.len());

        let params_period_14 = WmaParams { period: Some(14) };
        let input2 = WmaInput::from_candles(&candles, "hl2", params_period_14);
        let output2 = wma_with_kernel(&input2, kernel)?;
        assert_eq!(output2.values.len(), candles.close.len());

        let params_custom = WmaParams { period: Some(20) };
        let input3 = WmaInput::from_candles(&candles, "hlc3", params_custom);
        let output3 = wma_with_kernel(&input3, kernel)?;
        assert_eq!(output3.values.len(), candles.close.len());
        Ok(())
    }

    fn check_wma_accuracy(test_name: &str, kernel: Kernel) -> Result<(), Box<dyn Error>> {
        skip_if_unsupported!(kernel, test_name);
        let file_path = "src/data/2018-09-01-2024-Bitfinex_Spot-4h.csv";
        let candles = read_candles_from_csv(file_path)?;
        let data = &candles.close;
        let default_params = WmaParams::default();
        let input = WmaInput::from_candles(&candles, "close", default_params);
        let result = wma_with_kernel(&input, kernel)?;

        let expected_last_five = [
            59638.52903225806,
            59563.7376344086,
            59489.4064516129,
            59432.02580645162,
            59350.58279569892,
        ];
        assert!(result.values.len() >= 5, "Not enough WMA values");
        assert_eq!(
            result.values.len(),
            data.len(),
            "WMA output length should match input length"
        );
        let start_index = result.values.len().saturating_sub(5);
        let last_five = &result.values[start_index..];
        for (i, &value) in last_five.iter().enumerate() {
            assert!(
                (value - expected_last_five[i]).abs() < 1e-6,
                "WMA value mismatch at index {}: expected {}, got {}",
                i,
                expected_last_five[i],
                value
            );
        }
        let period = input.params.period.unwrap_or(30);
        for val in result.values.iter().skip(period - 1) {
            if !val.is_nan() {
                assert!(val.is_finite(), "WMA output should be finite");
            }
        }
        Ok(())
    }

    fn check_wma_default_candles(test_name: &str, kernel: Kernel) -> Result<(), Box<dyn Error>> {
        skip_if_unsupported!(kernel, test_name);
        let file_path = "src/data/2018-09-01-2024-Bitfinex_Spot-4h.csv";
        let candles = read_candles_from_csv(file_path)?;
        let input = WmaInput::with_default_candles(&candles);
        match input.data {
            WmaData::Candles { source, .. } => assert_eq!(source, "close"),
            _ => panic!("Expected WmaData::Candles"),
        }
        let output = wma_with_kernel(&input, kernel)?;
        assert_eq!(output.values.len(), candles.close.len());
        Ok(())
    }

    fn check_wma_zero_period(test_name: &str, kernel: Kernel) -> Result<(), Box<dyn Error>> {
        skip_if_unsupported!(kernel, test_name);
        let input_data = [10.0, 20.0, 30.0];
        let params = WmaParams { period: Some(0) };
        let input = WmaInput::from_slice(&input_data, params);
        let res = wma_with_kernel(&input, kernel);
        assert!(
            res.is_err(),
            "[{}] WMA should fail with zero period",
            test_name
        );
        Ok(())
    }

    fn check_wma_period_exceeds_length(
        test_name: &str,
        kernel: Kernel,
    ) -> Result<(), Box<dyn Error>> {
        skip_if_unsupported!(kernel, test_name);
        let data_small = [10.0, 20.0, 30.0];
        let params = WmaParams { period: Some(10) };
        let input = WmaInput::from_slice(&data_small, params);
        let res = wma_with_kernel(&input, kernel);
        assert!(
            res.is_err(),
            "[{}] WMA should fail with period exceeding length",
            test_name
        );
        Ok(())
    }

    fn check_wma_very_small_dataset(test_name: &str, kernel: Kernel) -> Result<(), Box<dyn Error>> {
        skip_if_unsupported!(kernel, test_name);
        let single_point = [42.0];
        let params = WmaParams { period: Some(9) };
        let input = WmaInput::from_slice(&single_point, params);
        let res = wma_with_kernel(&input, kernel);
        assert!(
            res.is_err(),
            "[{}] WMA should fail with insufficient data",
            test_name
        );
        Ok(())
    }

    fn check_wma_reinput(test_name: &str, kernel: Kernel) -> Result<(), Box<dyn Error>> {
        skip_if_unsupported!(kernel, test_name);
        let file_path = "src/data/2018-09-01-2024-Bitfinex_Spot-4h.csv";
        let candles = read_candles_from_csv(file_path)?;
        let first_params = WmaParams { period: Some(14) };
        let first_input = WmaInput::from_candles(&candles, "close", first_params);
        let first_result = wma_with_kernel(&first_input, kernel)?;
        let second_params = WmaParams { period: Some(5) };
        let second_input = WmaInput::from_slice(&first_result.values, second_params);
        let second_result = wma_with_kernel(&second_input, kernel)?;
        assert_eq!(second_result.values.len(), first_result.values.len());
        for val in &second_result.values[50..] {
            assert!(!val.is_nan());
        }
        Ok(())
    }

    fn check_wma_nan_handling(test_name: &str, kernel: Kernel) -> Result<(), Box<dyn Error>> {
        skip_if_unsupported!(kernel, test_name);
        let file_path = "src/data/2018-09-01-2024-Bitfinex_Spot-4h.csv";
        let candles = read_candles_from_csv(file_path)?;
        let params = WmaParams { period: Some(14) };
        let input = WmaInput::from_candles(&candles, "close", params);
        let result = wma_with_kernel(&input, kernel)?;
        assert_eq!(result.values.len(), candles.close.len());
        if result.values.len() > 50 {
            for i in 50..result.values.len() {
                assert!(!result.values[i].is_nan());
            }
        }
        Ok(())
    }

    fn check_wma_streaming(test_name: &str, kernel: Kernel) -> Result<(), Box<dyn Error>> {
        skip_if_unsupported!(kernel, test_name);
        let file_path = "src/data/2018-09-01-2024-Bitfinex_Spot-4h.csv";
        let candles = read_candles_from_csv(file_path)?;
        let period = 30;
        let input = WmaInput::from_candles(
            &candles,
            "close",
            WmaParams {
                period: Some(period),
            },
        );
        let batch_output = wma_with_kernel(&input, kernel)?.values;

        let mut stream = WmaStream::try_new(WmaParams {
            period: Some(period),
        })?;
        let mut stream_values = Vec::with_capacity(candles.close.len());
        for &price in &candles.close {
            match stream.update(price) {
                Some(val) => stream_values.push(val),
                None => stream_values.push(f64::NAN),
            }
        }
        assert_eq!(batch_output.len(), stream_values.len());
        for (i, (&b, &s)) in batch_output.iter().zip(stream_values.iter()).enumerate() {
            if b.is_nan() && s.is_nan() {
                continue;
            }
            let diff = (b - s).abs();
            assert!(
                diff < 1e-8,
                "[{}] WMA streaming f64 mismatch at idx {}: batch={}, stream={}, diff={}",
                test_name,
                i,
                b,
                s,
                diff
            );
        }
        Ok(())
    }

    #[cfg(debug_assertions)]
    fn check_wma_no_poison(test_name: &str, kernel: Kernel) -> Result<(), Box<dyn Error>> {
        skip_if_unsupported!(kernel, test_name);

        let file_path = "src/data/2018-09-01-2024-Bitfinex_Spot-4h.csv";
        let candles = read_candles_from_csv(file_path)?;

        let test_periods = vec![2, 5, 10, 14, 20, 30, 50, 100, 200];

        for &period in &test_periods {
            if period > candles.close.len() {
                continue;
            }

            let input = WmaInput::from_candles(
                &candles,
                "close",
                WmaParams {
                    period: Some(period),
                },
            );
            let output = wma_with_kernel(&input, kernel)?;

            for (i, &val) in output.values.iter().enumerate() {
                if val.is_nan() {
                    continue;
                }

                let bits = val.to_bits();

                if bits == 0x11111111_11111111 {
                    panic!(
						"[{}] Found alloc_with_nan_prefix poison value {} (0x{:016X}) at index {} with period {}",
						test_name, val, bits, i, period
					);
                }

                if bits == 0x22222222_22222222 {
                    panic!(
						"[{}] Found init_matrix_prefixes poison value {} (0x{:016X}) at index {} with period {}",
						test_name, val, bits, i, period
					);
                }

                if bits == 0x33333333_33333333 {
                    panic!(
						"[{}] Found make_uninit_matrix poison value {} (0x{:016X}) at index {} with period {}",
						test_name, val, bits, i, period
					);
                }
            }
        }

        Ok(())
    }

    #[cfg(not(debug_assertions))]
    fn check_wma_no_poison(_test_name: &str, _kernel: Kernel) -> Result<(), Box<dyn Error>> {
        Ok(())
    }

    macro_rules! generate_all_wma_tests {
        ($($test_fn:ident),*) => {
            paste! {
                $(
                    #[test]
                    fn [<$test_fn _scalar_f64>]() {
                        let _ = $test_fn(stringify!([<$test_fn _scalar_f64>]), Kernel::Scalar);
                    }
                )*
                #[cfg(all(feature = "nightly-avx", target_arch = "x86_64"))]
                $(
                    #[test]
                    fn [<$test_fn _avx2_f64>]() {
                        let _ = $test_fn(stringify!([<$test_fn _avx2_f64>]), Kernel::Avx2);
                    }
                    #[test]
                    fn [<$test_fn _avx512_f64>]() {
                        let _ = $test_fn(stringify!([<$test_fn _avx512_f64>]), Kernel::Avx512);
                    }
                )*
            }
        }
    }

    #[cfg(feature = "proptest")]
    #[allow(clippy::float_cmp)]
    fn check_wma_property(
        test_name: &str,
        kernel: Kernel,
    ) -> Result<(), Box<dyn std::error::Error>> {
        use proptest::prelude::*;
        skip_if_unsupported!(kernel, test_name);

        let strat = (2usize..=100).prop_flat_map(|period| {
            (
                prop::collection::vec(
                    (-1e6f64..1e6f64).prop_filter("finite", |x| x.is_finite()),
                    period..400,
                ),
                Just(period),
            )
        });

        proptest::test_runner::TestRunner::default()
            .run(&strat, |(data, period)| {
                let params = WmaParams {
                    period: Some(period),
                };
                let input = WmaInput::from_slice(&data, params.clone());

                let WmaOutput { values: out } = wma_with_kernel(&input, kernel).unwrap();

                let WmaOutput { values: ref_out } =
                    wma_with_kernel(&input, Kernel::Scalar).unwrap();

                let first = data.iter().position(|x| !x.is_nan()).unwrap_or(0);
                let warmup_end = first + period - 1;

                for i in 0..warmup_end.min(out.len()) {
                    prop_assert!(
                        out[i].is_nan(),
                        "Expected NaN during warmup at index {}, got {}",
                        i,
                        out[i]
                    );
                }

                for i in warmup_end..out.len() {
                    prop_assert!(
                        out[i].is_finite(),
                        "Expected finite value after warmup at index {}, got {}",
                        i,
                        out[i]
                    );
                }

                for i in warmup_end..data.len() {
                    let window_start = i + 1 - period;
                    let window = &data[window_start..=i];

                    let lo = window.iter().cloned().fold(f64::INFINITY, f64::min);
                    let hi = window.iter().cloned().fold(f64::NEG_INFINITY, f64::max);
                    let y = out[i];

                    prop_assert!(
                        y >= lo - 1e-9 && y <= hi + 1e-9,
                        "WMA at index {} = {} is outside window bounds [{}, {}]",
                        i,
                        y,
                        lo,
                        hi
                    );

                    if window.windows(2).all(|w| (w[0] - w[1]).abs() < 1e-12) {
                        prop_assert!(
                            (y - window[0]).abs() <= 1e-9,
                            "Constant input should produce constant output: {} vs {}",
                            y,
                            window[0]
                        );
                    }

                    let r = ref_out[i];
                    if y.is_finite() && r.is_finite() {
                        let y_bits = y.to_bits();
                        let r_bits = r.to_bits();
                        let ulp_diff = y_bits.abs_diff(r_bits);

                        prop_assert!(
                            (y - r).abs() <= 1e-9 || ulp_diff <= 4,
                            "Kernel mismatch at index {}: {} vs {} (ULP={})",
                            i,
                            y,
                            r,
                            ulp_diff
                        );
                    }
                }

                if period == 2 && out.len() >= 2 {
                    let idx = warmup_end;
                    if idx < out.len() {
                        let expected = (data[idx - 1] + 2.0 * data[idx]) / 3.0;
                        prop_assert!(
                            (out[idx] - expected).abs() <= 1e-9,
                            "Period=2 calculation mismatch: {} vs expected {}",
                            out[idx],
                            expected
                        );
                    }
                }

                if period <= 5 && warmup_end < out.len() {
                    let idx = warmup_end;
                    let window_start = idx + 1 - period;

                    let mut weighted_sum = 0.0;
                    let mut weight_sum = 0.0;
                    for (j, &val) in data[window_start..=idx].iter().enumerate() {
                        let weight = (j + 1) as f64;
                        weighted_sum += weight * val;
                        weight_sum += weight;
                    }
                    let expected = weighted_sum / weight_sum;

                    prop_assert!(
                        (out[idx] - expected).abs() <= 1e-9,
                        "Weight formula verification failed at index {}: {} vs expected {}",
                        idx,
                        out[idx],
                        expected
                    );
                }

                if data.len() >= period * 2 {
                    let mid = data.len() / 2;
                    if mid > warmup_end {
                        let mut step_data = vec![10.0; data.len()];
                        for i in mid..step_data.len() {
                            step_data[i] = 100.0;
                        }

                        let step_input = WmaInput::from_slice(&step_data, params.clone());
                        let WmaOutput { values: step_out } =
                            wma_with_kernel(&step_input, kernel).unwrap();

                        if mid + period < step_out.len() {
                            let wma_after_step = step_out[mid + period - 1];
                            let distance_to_new = (wma_after_step - 100.0).abs();
                            let distance_to_old = (wma_after_step - 10.0).abs();
                            prop_assert!(
								distance_to_new < distance_to_old,
								"WMA should respond more to recent values: {} should be closer to 100 than 10",
								wma_after_step
							);
                        }
                    }
                }

                if data.len() == period {
                    let valid_count = out.iter().filter(|x| x.is_finite()).count();
                    prop_assert!(
                        valid_count == 1,
                        "With data.len() == period, should have exactly 1 valid output, got {}",
                        valid_count
                    );

                    prop_assert!(
                        out[data.len() - 1].is_finite(),
                        "Last value should be valid when data.len() == period"
                    );
                }

                let is_monotonic_increasing = data.windows(2).all(|w| w[1] >= w[0] - 1e-12);
                if is_monotonic_increasing && out.len() > warmup_end + 1 {
                    for i in (warmup_end + 1)..out.len() {
                        prop_assert!(
                            out[i] >= out[i - 1] - 1e-9,
                            "Monotonic input should produce monotonic WMA: {} < {} at index {}",
                            out[i],
                            out[i - 1],
                            i
                        );
                    }
                }

                #[cfg(debug_assertions)]
                {
                    for (i, &val) in out.iter().enumerate() {
                        if !val.is_nan() {
                            let bits = val.to_bits();
                            prop_assert!(
                                bits != 0x11111111_11111111
                                    && bits != 0x22222222_22222222
                                    && bits != 0x33333333_33333333,
                                "Found poison value at index {}: {} (0x{:016X})",
                                i,
                                val,
                                bits
                            );
                        }
                    }
                }

                Ok(())
            })
            .unwrap();

        Ok(())
    }

    generate_all_wma_tests!(
        check_wma_partial_params,
        check_wma_accuracy,
        check_wma_default_candles,
        check_wma_zero_period,
        check_wma_period_exceeds_length,
        check_wma_very_small_dataset,
        check_wma_reinput,
        check_wma_nan_handling,
        check_wma_streaming,
        check_wma_no_poison
    );

    #[cfg(feature = "proptest")]
    generate_all_wma_tests!(check_wma_property);

    fn check_invalid_kernel_error(test: &str) -> Result<(), Box<dyn Error>> {
        let data = vec![1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 10.0];
        let sweep = WmaBatchRange { period: (2, 5, 1) };

        let non_batch_kernels = vec![Kernel::Scalar, Kernel::Avx2, Kernel::Avx512];
        for kernel in non_batch_kernels {
            let result = wma_with_kernel_batch(&data, &sweep, kernel);
            assert!(
                matches!(result, Err(WmaError::InvalidKernelForBatch(_))),
                "[{}] Expected InvalidKernelForBatch error for {:?}, got {:?}",
                test,
                kernel,
                result
            );
        }

        let batch_kernels = vec![Kernel::Auto, Kernel::ScalarBatch];
        #[cfg(all(feature = "nightly-avx", target_arch = "x86_64"))]
        let batch_kernels = vec![
            Kernel::Auto,
            Kernel::ScalarBatch,
            Kernel::Avx2Batch,
            Kernel::Avx512Batch,
        ];

        for kernel in batch_kernels {
            let result = wma_with_kernel_batch(&data, &sweep, kernel);
            assert!(
                result.is_ok(),
                "[{}] Expected success for batch kernel {:?}, got error: {:?}",
                test,
                kernel,
                result.err()
            );
        }

        Ok(())
    }

    fn check_batch_default_row(test: &str, kernel: Kernel) -> Result<(), Box<dyn Error>> {
        skip_if_unsupported!(kernel, test);
        let file = "src/data/2018-09-01-2024-Bitfinex_Spot-4h.csv";
        let c = read_candles_from_csv(file)?;

        let output = WmaBatchBuilder::new()
            .kernel(kernel)
            .apply_candles(&c, "close")?;

        let def = WmaParams::default();
        let row = output.values_for(&def).expect("default row missing");

        assert_eq!(row.len(), c.close.len());

        let expected = [
            59638.52903225806,
            59563.7376344086,
            59489.4064516129,
            59432.02580645162,
            59350.58279569892,
        ];
        let start = row.len() - 5;
        for (i, &v) in row[start..].iter().enumerate() {
            assert!(
                (v - expected[i]).abs() < 1e-6,
                "[{test}] default-row mismatch at idx {i}: {v} vs {expected:?}"
            );
        }
        Ok(())
    }

    #[cfg(debug_assertions)]
    fn check_batch_no_poison(test: &str, kernel: Kernel) -> Result<(), Box<dyn Error>> {
        skip_if_unsupported!(kernel, test);

        let file = "src/data/2018-09-01-2024-Bitfinex_Spot-4h.csv";
        let c = read_candles_from_csv(file)?;

        let batch_configs = vec![
            (2, 10, 1),
            (5, 25, 5),
            (10, 30, 10),
            (20, 100, 10),
            (30, 150, 30),
            (50, 200, 50),
            (2, 5, 1),
        ];

        for (start, end, step) in batch_configs {
            if start > c.close.len() {
                continue;
            }

            let output = WmaBatchBuilder::new()
                .kernel(kernel)
                .period_range(start, end, step)
                .apply_candles(&c, "close")?;

            for (idx, &val) in output.values.iter().enumerate() {
                if val.is_nan() {
                    continue;
                }

                let bits = val.to_bits();
                let row = idx / output.cols;
                let col = idx % output.cols;
                let period = output.combos[row].period.unwrap_or(0);

                if bits == 0x11111111_11111111 {
                    panic!(
                        "[{}] Found alloc_with_nan_prefix poison value {} (0x{:016X}) at row {} col {} (flat index {}) for period {} in range ({}, {}, {})",
                        test, val, bits, row, col, idx, period, start, end, step
                    );
                }

                if bits == 0x22222222_22222222 {
                    panic!(
                        "[{}] Found init_matrix_prefixes poison value {} (0x{:016X}) at row {} col {} (flat index {}) for period {} in range ({}, {}, {})",
                        test, val, bits, row, col, idx, period, start, end, step
                    );
                }

                if bits == 0x33333333_33333333 {
                    panic!(
                        "[{}] Found make_uninit_matrix poison value {} (0x{:016X}) at row {} col {} (flat index {}) for period {} in range ({}, {}, {})",
                        test, val, bits, row, col, idx, period, start, end, step
                    );
                }
            }
        }

        Ok(())
    }

    #[cfg(not(debug_assertions))]
    fn check_batch_no_poison(_test: &str, _kernel: Kernel) -> Result<(), Box<dyn Error>> {
        Ok(())
    }

    macro_rules! gen_batch_tests {
        ($fn_name:ident) => {
            paste! {
                #[test] fn [<$fn_name _scalar>]()      {
                    let _ = $fn_name(stringify!([<$fn_name _scalar>]), Kernel::ScalarBatch);
                }
                #[cfg(all(feature = "nightly-avx", target_arch = "x86_64"))]
                #[test] fn [<$fn_name _avx2>]()        {
                    let _ = $fn_name(stringify!([<$fn_name _avx2>]), Kernel::Avx2Batch);
                }
                #[cfg(all(feature = "nightly-avx", target_arch = "x86_64"))]
                #[test] fn [<$fn_name _avx512>]()      {
                    let _ = $fn_name(stringify!([<$fn_name _avx512>]), Kernel::Avx512Batch);
                }
                #[test] fn [<$fn_name _auto_detect>]() {
                    let _ = $fn_name(stringify!([<$fn_name _auto_detect>]), Kernel::Auto);
                }
            }
        };
    }
    gen_batch_tests!(check_batch_default_row);
    gen_batch_tests!(check_batch_no_poison);

    #[test]
    fn test_invalid_kernel_error() {
        let _ = check_invalid_kernel_error("test_invalid_kernel_error");
    }
}

#[cfg(feature = "python")]
#[pyfunction(name = "wma")]
#[pyo3(signature = (data, period, kernel=None))]

pub fn wma_py<'py>(
    py: Python<'py>,
    data: numpy::PyReadonlyArray1<'py, f64>,
    period: usize,
    kernel: Option<&str>,
) -> PyResult<Bound<'py, numpy::PyArray1<f64>>> {
    use numpy::{IntoPyArray, PyArrayMethods};

    let slice_in = data.as_slice()?;

    let kern = validate_kernel(kernel, false)?;

    let params = WmaParams {
        period: Some(period),
    };
    let wma_in = WmaInput::from_slice(slice_in, params);

    let result_vec: Vec<f64> = py
        .allow_threads(|| wma_with_kernel(&wma_in, kern).map(|o| o.values))
        .map_err(|e| PyValueError::new_err(e.to_string()))?;

    Ok(result_vec.into_pyarray(py))
}

#[cfg(feature = "python")]
#[pyclass(name = "WmaStream")]
pub struct WmaStreamPy {
    stream: WmaStream,
}

#[cfg(feature = "python")]
#[pymethods]
impl WmaStreamPy {
    #[new]
    fn new(period: usize) -> PyResult<Self> {
        let params = WmaParams {
            period: Some(period),
        };
        let stream =
            WmaStream::try_new(params).map_err(|e| PyValueError::new_err(e.to_string()))?;
        Ok(WmaStreamPy { stream })
    }

    fn update(&mut self, value: f64) -> Option<f64> {
        self.stream.update(value)
    }
}

#[cfg(feature = "python")]
#[pyfunction(name = "wma_batch")]
#[pyo3(signature = (data, period_range, kernel=None))]

pub fn wma_batch_py<'py>(
    py: Python<'py>,
    data: numpy::PyReadonlyArray1<'py, f64>,
    period_range: (usize, usize, usize),
    kernel: Option<&str>,
) -> PyResult<Bound<'py, pyo3::types::PyDict>> {
    use numpy::{IntoPyArray, PyArray1, PyArrayMethods};
    use pyo3::types::PyDict;

    let slice_in = data.as_slice()?;

    let sweep = WmaBatchRange {
        period: period_range,
    };

    let kern = validate_kernel(kernel, true)?;

    let combos = expand_grid(&sweep);
    let rows = combos.len();
    let cols = slice_in.len();

    let needed = rows
        .checked_mul(cols)
        .ok_or_else(|| PyValueError::new_err("rows*cols overflow"))?;
    let out_arr = unsafe { PyArray1::<f64>::new(py, [needed], false) };
    let slice_out = unsafe { out_arr.as_slice_mut()? };

    let combos = py
        .allow_threads(|| {
            let kernel = match kern {
                Kernel::Auto => detect_best_batch_kernel(),
                k => k,
            };
            let simd = match kernel {
                #[cfg(all(feature = "nightly-avx", target_arch = "x86_64"))]
                Kernel::Avx512Batch => Kernel::Avx512,
                #[cfg(all(feature = "nightly-avx", target_arch = "x86_64"))]
                Kernel::Avx2Batch => Kernel::Avx2,
                Kernel::ScalarBatch => Kernel::Scalar,
                _ => Kernel::Scalar,
            };

            wma_batch_inner_into(slice_in, &sweep, simd, true, slice_out)
        })
        .map_err(|e| PyValueError::new_err(e.to_string()))?;

    let dict = PyDict::new(py);
    dict.set_item("values", out_arr.reshape((rows, cols))?)?;
    dict.set_item(
        "periods",
        combos
            .iter()
            .map(|p| p.period.unwrap() as u64)
            .collect::<Vec<_>>()
            .into_pyarray(py),
    )?;
    dict.set_item("rows", rows)?;
    dict.set_item("cols", cols)?;

    Ok(dict)
}

#[cfg(all(feature = "python", feature = "cuda"))]
#[pyfunction(name = "wma_cuda_batch_dev")]
#[pyo3(signature = (data_f32, period_range, device_id=0))]
pub fn wma_cuda_batch_dev_py(
    py: Python<'_>,
    data_f32: numpy::PyReadonlyArray1<'_, f32>,
    period_range: (usize, usize, usize),
    device_id: usize,
) -> PyResult<DeviceArrayF32Py> {
    if !cuda_available() {
        return Err(PyValueError::new_err("CUDA not available"));
    }

    let slice_in = data_f32.as_slice()?;
    let sweep = WmaBatchRange {
        period: period_range,
    };

    let (inner, ctx, dev_id) = py.allow_threads(|| {
        let cuda = CudaWma::new(device_id).map_err(|e| PyValueError::new_err(e.to_string()))?;
        let ctx = cuda.context_arc();
        let dev_id = cuda.device_id();
        let arr = cuda
            .wma_batch_dev(slice_in, &sweep)
            .map_err(|e| PyValueError::new_err(e.to_string()))?;
        Ok::<_, pyo3::PyErr>((arr, ctx, dev_id))
    })?;

    Ok(DeviceArrayF32Py::new_from_rust(inner, ctx, dev_id))
}

#[cfg(all(feature = "python", feature = "cuda"))]
#[pyfunction(name = "wma_cuda_many_series_one_param_dev")]
#[pyo3(signature = (data_tm_f32, period, device_id=0))]
pub fn wma_cuda_many_series_one_param_dev_py(
    py: Python<'_>,
    data_tm_f32: numpy::PyReadonlyArray2<'_, f32>,
    period: usize,
    device_id: usize,
) -> PyResult<DeviceArrayF32Py> {
    use numpy::PyUntypedArrayMethods;

    if !cuda_available() {
        return Err(PyValueError::new_err("CUDA not available"));
    }

    let flat_in = data_tm_f32.as_slice()?;
    let rows = data_tm_f32.shape()[0];
    let cols = data_tm_f32.shape()[1];
    let params = WmaParams {
        period: Some(period),
    };

    let (inner, ctx, dev_id) = py.allow_threads(|| {
        let cuda = CudaWma::new(device_id).map_err(|e| PyValueError::new_err(e.to_string()))?;
        let ctx = cuda.context_arc();
        let dev_id = cuda.device_id();
        let arr = cuda
            .wma_multi_series_one_param_time_major_dev(flat_in, cols, rows, &params)
            .map_err(|e| PyValueError::new_err(e.to_string()))?;
        Ok::<_, pyo3::PyErr>((arr, ctx, dev_id))
    })?;

    Ok(DeviceArrayF32Py::new_from_rust(inner, ctx, dev_id))
}

#[cfg(all(target_arch = "wasm32", feature = "wasm"))]
#[derive(Serialize, Deserialize)]
pub struct WmaBatchConfig {
    pub period_range: (usize, usize, usize),
}

#[cfg(all(target_arch = "wasm32", feature = "wasm"))]
#[derive(Serialize, Deserialize)]
pub struct WmaBatchJsOutput {
    pub values: Vec<f64>,
    pub combos: Vec<WmaParams>,
    pub rows: usize,
    pub cols: usize,
}

#[cfg(all(target_arch = "wasm32", feature = "wasm"))]
#[wasm_bindgen(js_name = wma_batch)]
pub fn wma_batch_unified_js(data: &[f64], config: JsValue) -> Result<JsValue, JsValue> {
    let cfg: WmaBatchConfig = serde_wasm_bindgen::from_value(config)
        .map_err(|e| JsValue::from_str(&format!("Invalid config: {e}")))?;
    let sweep = WmaBatchRange {
        period: cfg.period_range,
    };

    let out = wma_batch_inner(data, &sweep, detect_best_kernel(), false)
        .map_err(|e| JsValue::from_str(&e.to_string()))?;

    let js = WmaBatchJsOutput {
        values: out.values,
        combos: out.combos,
        rows: out.rows,
        cols: out.cols,
    };
    serde_wasm_bindgen::to_value(&js)
        .map_err(|e| JsValue::from_str(&format!("Serialization error: {e}")))
}

#[cfg(all(target_arch = "wasm32", feature = "wasm"))]
#[wasm_bindgen]
pub fn wma_js(data: &[f64], period: usize) -> Result<Vec<f64>, JsValue> {
    let params = WmaParams {
        period: Some(period),
    };
    let input = WmaInput::from_slice(data, params);

    let mut output = vec![0.0; data.len()];
    wma_into_slice(&mut output, &input, Kernel::Auto)
        .map_err(|e| JsValue::from_str(&e.to_string()))?;

    Ok(output)
}

#[cfg(all(target_arch = "wasm32", feature = "wasm"))]
#[wasm_bindgen]
pub fn wma_batch_js(
    data: &[f64],
    period_start: usize,
    period_end: usize,
    period_step: usize,
) -> Result<Vec<f64>, JsValue> {
    let sweep = WmaBatchRange {
        period: (period_start, period_end, period_step),
    };

    wma_batch_inner(data, &sweep, Kernel::Auto, false)
        .map(|output| output.values)
        .map_err(|e| JsValue::from_str(&e.to_string()))
}

#[cfg(all(target_arch = "wasm32", feature = "wasm"))]
#[wasm_bindgen]
pub fn wma_batch_metadata_js(
    period_start: usize,
    period_end: usize,
    period_step: usize,
) -> Result<Vec<f64>, JsValue> {
    let sweep = WmaBatchRange {
        period: (period_start, period_end, period_step),
    };

    let combos = expand_grid(&sweep);
    let mut metadata = Vec::with_capacity(combos.len());

    for combo in combos {
        metadata.push(combo.period.unwrap() as f64);
    }

    Ok(metadata)
}

#[cfg(all(target_arch = "wasm32", feature = "wasm"))]
#[wasm_bindgen]
pub fn wma_alloc(len: usize) -> *mut f64 {
    let mut vec = Vec::<f64>::with_capacity(len);
    let ptr = vec.as_mut_ptr();
    std::mem::forget(vec);
    ptr
}

#[cfg(all(target_arch = "wasm32", feature = "wasm"))]
#[wasm_bindgen]
pub fn wma_free(ptr: *mut f64, len: usize) {
    if !ptr.is_null() {
        unsafe {
            let _ = Vec::from_raw_parts(ptr, len, len);
        }
    }
}

#[cfg(all(target_arch = "wasm32", feature = "wasm"))]
#[wasm_bindgen]
pub fn wma_into(
    in_ptr: *const f64,
    out_ptr: *mut f64,
    len: usize,
    period: usize,
) -> Result<(), JsValue> {
    if in_ptr.is_null() || out_ptr.is_null() {
        return Err(JsValue::from_str("Null pointer provided"));
    }

    unsafe {
        let data = std::slice::from_raw_parts(in_ptr, len);
        let params = WmaParams {
            period: Some(period),
        };
        let input = WmaInput::from_slice(data, params);

        if in_ptr == out_ptr {
            let mut temp = vec![0.0; len];
            wma_into_slice(&mut temp, &input, Kernel::Auto)
                .map_err(|e| JsValue::from_str(&e.to_string()))?;
            let out = std::slice::from_raw_parts_mut(out_ptr, len);
            out.copy_from_slice(&temp);
        } else {
            let out = std::slice::from_raw_parts_mut(out_ptr, len);
            wma_into_slice(out, &input, Kernel::Auto)
                .map_err(|e| JsValue::from_str(&e.to_string()))?;
        }
        Ok(())
    }
}

#[cfg(all(target_arch = "wasm32", feature = "wasm"))]
#[wasm_bindgen]
pub fn wma_batch_into(
    in_ptr: *const f64,
    out_ptr: *mut f64,
    len: usize,
    period_start: usize,
    period_end: usize,
    period_step: usize,
) -> Result<usize, JsValue> {
    if in_ptr.is_null() || out_ptr.is_null() {
        return Err(JsValue::from_str("Null pointer provided"));
    }

    unsafe {
        let data = std::slice::from_raw_parts(in_ptr, len);

        let sweep = WmaBatchRange {
            period: (period_start, period_end, period_step),
        };

        let combos = expand_grid(&sweep);
        let rows = combos.len();
        let total_size = rows * len;

        let out = std::slice::from_raw_parts_mut(out_ptr, total_size);

        wma_batch_inner_into(data, &sweep, Kernel::Auto, false, out)
            .map_err(|e| JsValue::from_str(&e.to_string()))?;

        Ok(rows)
    }
}