// Copyright (c) 2021 Weird Constructor <weirdconstructor@gmail.com>
// This file is a part of HexoDSP. Released under GPL-3.0-or-later.
// See README.md and COPYING for details.

use std::cell::RefCell;
use num_traits::{Float, FloatConst, cast::FromPrimitive};

/// Logarithmic table size of the table in [fast_cos] / [fast_sin].
static FAST_COS_TAB_LOG2_SIZE : usize = 9;
/// Table size of the table in [fast_cos] / [fast_sin].
static FAST_COS_TAB_SIZE : usize      = 1 << FAST_COS_TAB_LOG2_SIZE; // =512
/// The wave table of [fast_cos] / [fast_sin].
static mut FAST_COS_TAB : [f32; 513] = [0.0; 513];

/// Initializes the cosine wave table for [fast_cos] and [fast_sin].
pub fn init_cos_tab() {
    for i in 0..(FAST_COS_TAB_SIZE+1) {
        let phase : f32 =
            (i as f32)
            * ((std::f32::consts::TAU)
               / (FAST_COS_TAB_SIZE as f32));
        unsafe {
            // XXX: note: mutable statics can be mutated by multiple
            //      threads: aliasing violations or data races
            //      will cause undefined behavior
            FAST_COS_TAB[i] = phase.cos();
        }
    }
}

/// Internal phase increment/scaling for [fast_cos].
const PHASE_SCALE : f32 = 1.0_f32 / (std::f32::consts::TAU);

/// A faster implementation of cosine. It's not that much faster than
/// Rust's built in cosine function. But YMMV.
///
/// Don't forget to call [init_cos_tab] before using this!
///
///```
/// use hexodsp::dsp::helpers::*;
/// init_cos_tab(); // Once on process initialization.
///
/// // ...
/// assert!((fast_cos(std::f32::consts::PI) - -1.0).abs() < 0.001);
///```
pub fn fast_cos(mut x: f32) -> f32 {
    x = x.abs(); // cosine is symmetrical around 0, let's get rid of negative values

    // normalize range from 0..2PI to 1..2
    let phase = x * PHASE_SCALE;

    let index = FAST_COS_TAB_SIZE as f32 * phase;

    let fract = index.fract();
    let index = index.floor() as usize;

    unsafe {
        // XXX: note: mutable statics can be mutated by multiple
        //      threads: aliasing violations or data races
        //      will cause undefined behavior
        let left         = FAST_COS_TAB[index as usize];
        let right        = FAST_COS_TAB[index as usize + 1];

        return left + (right - left) * fract;
    }
}

/// A faster implementation of sine. It's not that much faster than
/// Rust's built in sine function. But YMMV.
///
/// Don't forget to call [init_cos_tab] before using this!
///
///```
/// use hexodsp::dsp::helpers::*;
/// init_cos_tab(); // Once on process initialization.
///
/// // ...
/// assert!((fast_sin(0.5 * std::f32::consts::PI) - 1.0).abs() < 0.001);
///```
pub fn fast_sin(x: f32) -> f32 {
    fast_cos(x - (std::f32::consts::PI / 2.0))
}

/// A wavetable filled entirely with white noise.
/// Don't forget to call [init_white_noise_tab] before using it.
static mut WHITE_NOISE_TAB: [f64; 1024] = [0.0; 1024];

#[allow(rustdoc::private_intra_doc_links)]
/// Initializes [WHITE_NOISE_TAB].
pub fn init_white_noise_tab() {
    let mut rng = RandGen::new();
    unsafe {
        for i in 0..WHITE_NOISE_TAB.len() {
            WHITE_NOISE_TAB[i as usize] = rng.next_open01();
        }
    }
}

#[derive(Debug, Copy, Clone, PartialEq)]
/// Random number generator based on xoroshiro128.
/// Requires two internal state variables. You may prefer [SplitMix64] or [Rng].
pub struct RandGen {
    r: [u64; 2],
}

// Taken from xoroshiro128 crate under MIT License
// Implemented by Matthew Scharley (Copyright 2016)
// https://github.com/mscharley/rust-xoroshiro128
/// Given the mutable `state` generates the next pseudo random number.
pub fn next_xoroshiro128(state: &mut [u64; 2]) -> u64 {
    let s0: u64     = state[0];
    let mut s1: u64 = state[1];
    let result: u64 = s0.wrapping_add(s1);

    s1 ^= s0;
    state[0] = s0.rotate_left(55) ^ s1 ^ (s1 << 14); // a, b
    state[1] = s1.rotate_left(36); // c

    result
}

// Taken from rand::distributions
// Licensed under the Apache License, Version 2.0
// Copyright 2018 Developers of the Rand project.
/// Maps any `u64` to a `f64` in the open interval `[0.0, 1.0)`.
pub fn u64_to_open01(u: u64) -> f64 {
    use core::f64::EPSILON;
    let float_size         = std::mem::size_of::<f64>() as u32 * 8;
    let fraction           = u >> (float_size - 52);
    let exponent_bits: u64 = (1023 as u64) << 52;
    f64::from_bits(fraction | exponent_bits) - (1.0 - EPSILON / 2.0)
}

impl RandGen {
    pub fn new() -> Self {
        RandGen {
            r: [0x193a6754a8a7d469, 0x97830e05113ba7bb],
        }
    }

    /// Next random unsigned 64bit integer.
    pub fn next(&mut self) -> u64 {
        next_xoroshiro128(&mut self.r)
    }

    /// Next random float between `[0.0, 1.0)`.
    pub fn next_open01(&mut self) -> f64 {
        u64_to_open01(self.next())
    }
}

#[derive(Debug, Copy, Clone)]
/// Random number generator based on [SplitMix64].
/// Requires two internal state variables. You may prefer [SplitMix64] or [Rng].
pub struct Rng {
    sm: SplitMix64,
}

impl Rng {
    pub fn new() -> Self {
        Self { sm: SplitMix64::new(0x193a67f4a8a6d769) }
    }

    pub fn seed(&mut self, seed: u64) {
        self.sm = SplitMix64::new(seed);
    }

    #[inline]
    pub fn next(&mut self) -> f32 {
        self.sm.next_open01() as f32
    }

    #[inline]
    pub fn next_u64(&mut self) -> u64 {
        self.sm.next_u64()
    }
}

thread_local! {
    static GLOBAL_RNG: RefCell<Rng> = RefCell::new(Rng::new());
}

#[inline]
pub fn rand_01() -> f32 {
    GLOBAL_RNG.with(|r| r.borrow_mut().next())
}

#[inline]
pub fn rand_u64() -> u64 {
    GLOBAL_RNG.with(|r| r.borrow_mut().next_u64())
}

// Copyright 2018 Developers of the Rand project.
//
// Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or
// https://www.apache.org/licenses/LICENSE-2.0> or the MIT license
// <LICENSE-MIT or https://opensource.org/licenses/MIT>, at your
// option. This file may not be copied, modified, or distributed
// except according to those terms.
//- splitmix64 (http://xoroshiro.di.unimi.it/splitmix64.c) 
//
/// A splitmix64 random number generator.
///
/// The splitmix algorithm is not suitable for cryptographic purposes, but is
/// very fast and has a 64 bit state.
///
/// The algorithm used here is translated from [the `splitmix64.c`
/// reference source code](http://xoshiro.di.unimi.it/splitmix64.c) by
/// Sebastiano Vigna. For `next_u32`, a more efficient mixing function taken
/// from [`dsiutils`](http://dsiutils.di.unimi.it/) is used.
#[derive(Debug, Copy, Clone)]
pub struct SplitMix64(pub u64);

/// Internal random constant for [SplitMix64].
const PHI: u64 = 0x9e3779b97f4a7c15;

impl SplitMix64 {
    pub fn new(seed: u64) -> Self { Self(seed) }
    pub fn new_from_i64(seed: i64) -> Self {
        Self::new(u64::from_be_bytes(seed.to_be_bytes()))
    }

    pub fn new_time_seed() -> Self {
        use std::time::SystemTime;

        match SystemTime::now().duration_since(SystemTime::UNIX_EPOCH) {
            Ok(n) => Self::new(n.as_secs() as u64),
            Err(_) => Self::new(123456789),
        }
    }

    #[inline]
    pub fn next_u64(&mut self) -> u64 {
        self.0 = self.0.wrapping_add(PHI);
        let mut z = self.0;
        z = (z ^ (z >> 30)).wrapping_mul(0xbf58476d1ce4e5b9);
        z = (z ^ (z >> 27)).wrapping_mul(0x94d049bb133111eb);
        z ^ (z >> 31)
    }

    #[inline]
    pub fn next_i64(&mut self) -> i64 {
        i64::from_be_bytes(
            self.next_u64().to_be_bytes())
    }

    #[inline]
    pub fn next_open01(&mut self) -> f64 {
        u64_to_open01(self.next_u64())
    }
}

/// Linear crossfade.
///
/// * `v1` - signal 1, range -1.0 to 1.0
/// * `v2` - signal 2, range -1.0 to 1.0
/// * `mix` - mix position, range 0.0 to 1.0, mid is at 0.5
#[inline]
pub fn crossfade(v1: f32, v2: f32, mix: f32) -> f32 {
    v1 * (1.0 - mix) + v2 * mix
}

/// Constant power crossfade.
///
/// * `v1` - signal 1, range -1.0 to 1.0
/// * `v2` - signal 2, range -1.0 to 1.0
/// * `mix` - mix position, range 0.0 to 1.0, mid is at 0.5
#[inline]
pub fn crossfade_cpow(v1: f32, v2: f32, mix: f32) -> f32 {
    let s1 = (mix * std::f32::consts::FRAC_PI_2).sin();
    let s2 = ((1.0 - mix) * std::f32::consts::FRAC_PI_2).sin();
    v1 * s2 + v2 * s1
}

const CROSS_LOG_MIN : f32 = -13.815510557964274; // (0.000001_f32).ln();
const CROSS_LOG_MAX : f32 = 0.0;                 // (1.0_f32).ln();

/// Logarithmic crossfade.
///
/// * `v1` - signal 1, range -1.0 to 1.0
/// * `v2` - signal 2, range -1.0 to 1.0
/// * `mix` - mix position, range 0.0 to 1.0, mid is at 0.5
#[inline]
pub fn crossfade_log(v1: f32, v2: f32, mix: f32) -> f32 {
    let x =
        (mix * (CROSS_LOG_MAX - CROSS_LOG_MIN) + CROSS_LOG_MIN)
        .exp();
    crossfade(v1, v2, x)
}

/// Exponential crossfade.
///
/// * `v1` - signal 1, range -1.0 to 1.0
/// * `v2` - signal 2, range -1.0 to 1.0
/// * `mix` - mix position, range 0.0 to 1.0, mid is at 0.5
#[inline]
pub fn crossfade_exp(v1: f32, v2: f32, mix: f32) -> f32 {
    crossfade(v1, v2, mix * mix)
}

#[inline]
pub fn clamp(f: f32, min: f32, max: f32) -> f32 {
         if f < min { min }
    else if f > max { max }
    else            {   f }
}

pub fn square_135(phase: f32) -> f32 {
      fast_sin(phase)
    + fast_sin(phase * 3.0) / 3.0
    + fast_sin(phase * 5.0) / 5.0
}

pub fn square_35(phase: f32) -> f32 {
      fast_sin(phase * 3.0) / 3.0
    + fast_sin(phase * 5.0) / 5.0
}

// note: MIDI note value?
pub fn note_to_freq(note: f32) -> f32 {
    440.0 * (2.0_f32).powf((note - 69.0) / 12.0)
}

// Ported from LMMS under GPLv2
// * DspEffectLibrary.h - library with template-based inline-effects
// * Copyright (c) 2006-2014 Tobias Doerffel <tobydox/at/users.sourceforge.net>
//
// Original source seems to be musicdsp.org, Author: Bram de Jong
// see also: https://www.musicdsp.org/en/latest/Effects/41-waveshaper.html
// Notes:
//     where x (in [-1..1] will be distorted and a is a distortion parameter
//     that goes from 1 to infinity. The equation is valid for positive and
//     negativ values. If a is 1, it results in a slight distortion and with
//     bigger a's the signal get's more funky.
//     A good thing about the shaper is that feeding it with bigger-than-one
//     values, doesn't create strange fx. The maximum this function will reach
//     is 1.2 for a=1.
//
//     f(x,a) = x*(abs(x) + a)/(x^2 + (a-1)*abs(x) + 1)
/// Signal distortion by Bram de Jong.
/// ```text
/// gain:        0.1 - 5.0       default = 1.0
/// threshold:   0.0 - 100.0     default = 0.8
/// i:           signal
/// ```
#[inline]
pub fn f_distort(gain: f32, threshold: f32, i: f32) -> f32 {
    gain * (
        i * ( i.abs() + threshold )
        / ( i * i + (threshold - 1.0) * i.abs() + 1.0 ))
}

// Ported from LMMS under GPLv2
// * DspEffectLibrary.h - library with template-based inline-effects
// * Copyright (c) 2006-2014 Tobias Doerffel <tobydox/at/users.sourceforge.net>
//
/// Foldback Signal distortion
/// ```text
/// gain:        0.1 - 5.0       default = 1.0
/// threshold:   0.0 - 100.0     default = 0.8
/// i:           signal
/// ```
#[inline]
pub fn f_fold_distort(gain: f32, threshold: f32, i: f32) -> f32 {
    if i >= threshold || i < -threshold {
        gain
        * ((
            ((i - threshold) % threshold * 4.0).abs()
            - threshold * 2.0).abs()
           - threshold)
    } else {
        gain * i
    }
}

pub fn lerp(x: f32, a: f32, b: f32) -> f32 {
    (a * (1.0 - x)) + (b * x)
}

pub fn lerp64(x: f64, a: f64, b: f64) -> f64 {
    (a * (1.0 - x)) + (b * x)
}

pub fn p2range(x: f32, a: f32, b: f32) -> f32 {
    lerp(x, a, b)
}

pub fn p2range_exp(x: f32, a: f32, b: f32) -> f32 {
    let x = x * x;
    (a * (1.0 - x)) + (b * x)
}

pub fn p2range_exp4(x: f32, a: f32, b: f32) -> f32 {
    let x = x * x * x * x;
    (a * (1.0 - x)) + (b * x)
}


pub fn range2p(v: f32, a: f32, b: f32) -> f32 {
    ((v - a) / (b - a)).abs()
}

pub fn range2p_exp(v: f32, a: f32, b: f32) -> f32 {
    (((v - a) / (b - a)).abs()).sqrt()
}

pub fn range2p_exp4(v: f32, a: f32, b: f32) -> f32 {
    (((v - a) / (b - a)).abs()).sqrt().sqrt()
}

/// ```text
/// gain: 24.0 - -90.0   default = 0.0
/// ```
pub fn gain2coef(gain: f32) -> f32 {
    if gain > -90.0 {
        10.0_f32.powf(gain * 0.05)
    } else {
        0.0
    }
}

// quickerTanh / quickerTanh64 credits to mopo synthesis library:
// Under GPLv3 or any later.
// Little IO <littleioaudio@gmail.com>
// Matt Tytel <matthewtytel@gmail.com>
pub fn quicker_tanh64(v: f64) -> f64 {
    let square = v * v;
    v / (1.0 + square / (3.0 + square / 5.0))
}

#[inline]
pub fn quicker_tanh(v: f32) -> f32 {
    let square = v * v;
    v / (1.0 + square / (3.0 + square / 5.0))
}

// quickTanh / quickTanh64 credits to mopo synthesis library:
// Under GPLv3 or any later.
// Little IO <littleioaudio@gmail.com>
// Matt Tytel <matthewtytel@gmail.com>
pub fn quick_tanh64(v: f64) -> f64 {
    let abs_v = v.abs();
    let square = v * v;
    let num =
        v * (2.45550750702956
             + 2.45550750702956 * abs_v
             + square * (0.893229853513558
                         + 0.821226666969744 * abs_v));
    let den =
        2.44506634652299
        + (2.44506634652299 + square)
          * (v + 0.814642734961073 * v * abs_v).abs();

    num / den
}

pub fn quick_tanh(v: f32) -> f32 {
    let abs_v = v.abs();
    let square = v * v;
    let num =
        v * (2.45550750702956
             + 2.45550750702956 * abs_v
             + square * (0.893229853513558
                         + 0.821226666969744 * abs_v));
    let den =
        2.44506634652299
        + (2.44506634652299 + square)
          * (v + 0.814642734961073 * v * abs_v).abs();

    num / den
}

/// A helper function for exponential envelopes.
/// It's a bit faster than calling the `pow` function of Rust.
///
/// * `x` the input value
/// * `v' the shape value.
/// Which is linear at `0.5`, the forth root of `x` at `1.0` and x to the power
/// of 4 at `0.0`. You can vary `v` as you like.
///
///```
/// use hexodsp::dsp::helpers::*;
///
/// assert!(((sqrt4_to_pow4(0.25, 0.0) - 0.25_f32 * 0.25 * 0.25 * 0.25)
///          .abs() - 1.0)
///         < 0.0001);
///
/// assert!(((sqrt4_to_pow4(0.25, 1.0) - (0.25_f32).sqrt().sqrt())
///          .abs() - 1.0)
///         < 0.0001);
///
/// assert!(((sqrt4_to_pow4(0.25, 0.5) - 0.25_f32).abs() - 1.0) < 0.0001);
///```
#[inline]
pub fn sqrt4_to_pow4(x: f32, v: f32) -> f32 {
    if v > 0.75 {
        let xsq1 = x.sqrt();
        let xsq = xsq1.sqrt();
        let v = (v - 0.75) * 4.0;
        xsq1 * (1.0 - v) + xsq * v

    } else if v > 0.5 {
        let xsq = x.sqrt();
        let v = (v - 0.5) * 4.0;
        x * (1.0 - v) + xsq * v

    } else if v > 0.25 {
        let xx = x * x;
        let v = (v - 0.25) * 4.0;
        x * v + xx * (1.0 - v)

    } else {
        let xx = x * x;
        let xxxx = xx * xx;
        let v = v * 4.0;
        xx * v + xxxx * (1.0 - v)
    }
}

/// A-100 Eurorack states, that a trigger is usually 2-10 milliseconds.
const TRIG_SIGNAL_LENGTH_MS : f32 = 2.0;

#[derive(Debug, Clone, Copy)]
pub struct TrigSignal {
    length:     u32,
    scount:     u32,
}

impl TrigSignal {
    pub fn new() -> Self {
        Self {
            length: ((44100.0 * TRIG_SIGNAL_LENGTH_MS) / 1000.0).ceil() as u32,
            scount: 0,
        }
    }

    pub fn reset(&mut self) {
        self.scount = 0;
    }

    pub fn set_sample_rate(&mut self, srate: f32) {
        self.length = ((srate * TRIG_SIGNAL_LENGTH_MS) / 1000.0).ceil() as u32;
        self.scount = 0;
    }

    #[inline]
    pub fn trigger(&mut self) { self.scount = self.length; }

    #[inline]
    pub fn next(&mut self) -> f32 {
        if self.scount > 0 {
            self.scount -= 1;
            1.0
        } else {
            0.0
        }
    }
}

impl Default for TrigSignal {
    fn default() -> Self { Self::new() }
}

#[derive(Debug, Clone, Copy)]
pub struct Trigger {
    triggered:  bool,
}

impl Trigger {
    pub fn new() -> Self {
        Self {
            triggered: false,
        }
    }

    #[inline]
    pub fn reset(&mut self) {
        self.triggered = false;
    }

    #[inline]
    pub fn check_trigger(&mut self, input: f32) -> bool {
        if self.triggered {
            if input <= 0.25 {
                self.triggered = false;
            }

            false

        } else if input > 0.75 {
            self.triggered = true;
            true

        } else {
            false
        }
    }
}

#[derive(Debug, Clone, Copy)]
pub struct TriggerPhaseClock {
    clock_phase:    f64,
    clock_inc:      f64,
    prev_trigger:   bool,
    clock_samples:  u32,
}

impl TriggerPhaseClock {
    pub fn new() -> Self {
        Self {
            clock_phase:    0.0,
            clock_inc:      0.0,
            prev_trigger:   true,
            clock_samples:  0,
        }
    }

    #[inline]
    pub fn reset(&mut self) {
        self.clock_samples = 0;
        self.clock_inc     = 0.0;
        self.prev_trigger  = true;
        self.clock_samples = 0;
    }

    #[inline]
    pub fn sync(&mut self) {
        self.clock_phase = 0.0;
    }

    #[inline]
    pub fn next_phase(&mut self, clock_limit: f64, trigger_in: f32) -> f64 {
        if self.prev_trigger {
            if trigger_in <= 0.25 {
                self.prev_trigger = false;
            }

        } else if trigger_in > 0.75 {
            self.prev_trigger = true;

            if self.clock_samples > 0 {
                self.clock_inc =
                    1.0 / (self.clock_samples as f64);
            }

            self.clock_samples = 0;
        }

        self.clock_samples += 1;

        self.clock_phase += self.clock_inc;
        self.clock_phase = self.clock_phase % clock_limit;

        self.clock_phase
    }
}

#[derive(Debug, Clone, Copy)]
pub struct TriggerSampleClock {
    prev_trigger:   bool,
    clock_samples:  u32,
    counter:        u32,
}

impl TriggerSampleClock {
    pub fn new() -> Self {
        Self {
            prev_trigger:   true,
            clock_samples:  0,
            counter:        0,
        }
    }

    #[inline]
    pub fn reset(&mut self) {
        self.clock_samples = 0;
        self.counter       = 0;
    }

    #[inline]
    pub fn next(&mut self, trigger_in: f32) -> u32 {
        if self.prev_trigger {
            if trigger_in <= 0.25 {
                self.prev_trigger = false;
            }

        } else if trigger_in > 0.75 {
            self.prev_trigger  = true;
            self.clock_samples = self.counter;
            self.counter       = 0;
        }

        self.counter += 1;

        self.clock_samples
    }
}

/// Default size of the delay buffer: 5 seconds at 8 times 48kHz
const DEFAULT_DELAY_BUFFER_SAMPLES : usize = 8 * 48000 * 5;

macro_rules! fc {
    ($F: ident, $e: expr) => { F::from_f64($e).unwrap() }
}

#[derive(Debug, Clone, Default)]
pub struct DelayBuffer<F: Float> {
    data:   Vec<F>,
    wr:     usize,
    srate:  F,
}

impl<F: Float + FromPrimitive> DelayBuffer<F> {
    pub fn new() -> Self {
        Self {
            data:   vec![fc!(F,0.0); DEFAULT_DELAY_BUFFER_SAMPLES],
            wr:     0,
            srate:  fc!(F, 44100.0),
        }
    }

    pub fn new_with_size(size: usize) -> Self {
        Self {
            data:   vec![fc!(F, 0.0); size],
            wr:     0,
            srate:  fc!(F, 44100.0),
        }
    }

    pub fn set_sample_rate(&mut self, srate: F) {
        self.srate = srate;
    }

    pub fn reset(&mut self) {
        self.data.fill(0.0);
        self.wr = 0;
    }

    /// Feed one sample into the delay line and increment the write pointer.
    /// Please note: For sample accurate feedback you need to retrieve the
    /// output of the delay line before feeding in a new signal.
    #[inline]
    pub fn feed(&mut self, input: F) {
        self.data[self.wr] = input;
        self.wr = (self.wr + 1) % self.data.len();
    }

    /// Combines [DelayBuffer::cubic_interpolate_at] and [DelayBuffer::feed]
    /// into one convenient function.
    #[inline]
    pub fn next_cubic(&mut self, delay_time_ms: F, input: F) -> F {
        let res = self.cubic_interpolate_at(delay_time_ms);
        self.feed(input);
        res
    }

    /// Shorthand for [DelayBuffer::cubic_interpolate_at].
    #[inline]
    pub fn tap_c(&self, delay_time_ms: F) -> F {
        self.cubic_interpolate_at(delay_time_ms)
    }

    /// Shorthand for [DelayBuffer::cubic_interpolate_at].
    #[inline]
    pub fn tap_n(&self, delay_time_ms: F) -> F {
        self.nearest_at(delay_time_ms)
    }

    /// Shorthand for [DelayBuffer::cubic_interpolate_at].
    #[inline]
    pub fn tap_l(&self, delay_time_ms: F) -> F {
        self.linear_interpolate_at(delay_time_ms)
    }

    /// Fetch a sample from the delay buffer at the given time.
    ///
    /// * `delay_time_ms` - Delay time in milliseconds.
    pub fn linear_interpolate_at(&self, delay_time_ms: F) -> F {
        let data   = &self.data[..];
        let len    = data.len();
        let s_offs = (delay_time_ms * self.srate) / 1000.0;
        let offs   = s_offs.floor() as usize % len;
        let fract  = s_offs.fract();

        let i = (self.wr + len) - offs;
        let x0  = data[i       % len];
        let x1  = data[(i + 1) % len];

        let fract = fract as F;
        x0 * (1.0 - fract) + x1 * fract
    }

    /// Fetch a sample from the delay buffer at the given time.
    ///
    /// * `delay_time_ms` - Delay time in milliseconds.
    #[inline]
    pub fn cubic_interpolate_at(&self, delay_time_ms: F) -> F {
        let data   = &self.data[..];
        let len    = data.len();
        let s_offs = (delay_time_ms * self.srate) / 1000.0;
        let offs   = s_offs.floor() as usize % len;
        let fract  = s_offs.fract();

        let i = (self.wr + len) - offs;

        // Hermite interpolation, take from 
        // https://github.com/eric-wood/delay/blob/main/src/delay.rs#L52
        //
        // Thanks go to Eric Wood!
        //
        // For the interpolation code:
        // MIT License, Copyright (c) 2021 Eric Wood
        let xm1 = data[(i - 1) % len];
        let x0  = data[i       % len];
        let x1  = data[(i + 1) % len];
        let x2  = data[(i + 2) % len];

        let c     = (x1 - xm1) * 0.5;
        let v     = x0 - x1;
        let w     = c + v;
        let a     = w + v + (x2 - x0) * 0.5;
        let b_neg = w + a;

        let fract = fract as F;
        (((a * fract) - b_neg) * fract + c) * fract + x0
    }

    #[inline]
    pub fn nearest_at(&self, delay_time_ms: F) -> F {
        let len  = self.data.len();
        let offs = ((delay_time_ms * self.srate) / 1000.0).floor() as usize % len;
        let idx  = ((self.wr + len) - offs) % len;
        self.data[idx]
    }

    #[inline]
    pub fn at(&self, delay_sample_count: usize) -> F {
        let len  = self.data.len();
        let idx  = ((self.wr + len) - delay_sample_count) % len;
        self.data[idx]
    }
}

/// Default size of the delay buffer: 1 seconds at 8 times 48kHz
const DEFAULT_ALLPASS_COMB_SAMPLES : usize = 8 * 48000;

#[derive(Debug, Clone, Default)]
pub struct AllPass<F: Float> {
    delay: DelayBuffer<F>,
}

impl<F: Float> AllPass<F> {
    pub fn new() -> Self {
        Self {
            delay: DelayBuffer::new_with_size(DEFAULT_ALLPASS_COMB_SAMPLES),
        }
    }

    pub fn set_sample_rate(&mut self, srate: F) {
        self.delay.set_sample_rate(srate);
    }

    pub fn reset(&mut self) {
        self.delay.reset();
    }

    #[inline]
    pub fn delay_tap_n(&self, time_ms: F) -> F {
        self.delay.tap_n(time_ms)
    }

    #[inline]
    pub fn next(&mut self, time_ms: F, g: F, v: F) -> F {
        let s = self.delay.nearest_at(time_ms);
        let input = v + -g * s;
        self.delay.feed(input);
        input * g + s
    }
}

#[derive(Debug, Clone)]
pub struct Comb {
    delay: DelayBuffer<f32>,
}

impl Comb {
    pub fn new() -> Self {
        Self {
            delay: DelayBuffer::new_with_size(DEFAULT_ALLPASS_COMB_SAMPLES),
        }
    }

    pub fn set_sample_rate(&mut self, srate: f32) {
        self.delay.set_sample_rate(srate);
    }

    pub fn reset(&mut self) {
        self.delay.reset();
    }

    #[inline]
    pub fn delay_tap_c(&self, time_ms: f32) -> f32 {
        self.delay.tap_c(time_ms)
    }

    #[inline]
    pub fn delay_tap_n(&self, time_ms: f32) -> f32 {
        self.delay.tap_n(time_ms)
    }

    #[inline]
    pub fn next_feedback(&mut self, time: f32, g: f32, v: f32) -> f32 {
        let s = self.delay.cubic_interpolate_at(time);
        let v = v + s * g;
        self.delay.feed(v);
        v
    }

    #[inline]
    pub fn next_feedforward(&mut self, time: f32, g: f32, v: f32) -> f32 {
        let s = self.delay.next_cubic(time, v);
        v + s * g
    }
}

// one pole lp from valley rack free:
// https://github.com/ValleyAudio/ValleyRackFree/blob/v1.0/src/Common/DSP/OnePoleFilters.cpp
#[inline]
/// Process a very simple one pole 6dB low pass filter.
/// Useful in various applications, from usage in a synthesizer to
/// damping stuff in a reverb/delay.
///
/// * `input`  - Input sample
/// * `freq`   - Frequency between 1.0 and 22000.0Hz
/// * `israte` - 1.0 / samplerate
/// * `z`      - The internal one sample buffer of the filter.
///
///```
///    use hexodsp::dsp::helpers::*;
///
///    let samples  = vec![0.0; 44100];
///    let mut z    = 0.0;
///    let mut freq = 1000.0;
///
///    for s in samples.iter() {
///        let s_out =
///            process_1pole_lowpass(*s, freq, 1.0 / 44100.0, &mut z);
///        // ... do something with the result here.
///    }
///```
pub fn process_1pole_lowpass(input: f32, freq: f32, israte: f32, z: &mut f32) -> f32 {
    let b = (-std::f32::consts::TAU * freq * israte).exp();
    let a = 1.0 - b;
    *z = a * input + *z * b;
    *z
}

#[derive(Debug, Clone, Copy, Default)]
pub struct OnePoleLPF<F: Float> {
    israte:     F,
    a:          F,
    b:          F,
    freq:       F,
    z:          F,
}

impl<F: Float + FloatConst> OnePoleLPF<F> {
    pub fn new() -> Self {
        Self {
            israte: 1.0 / 44100.0,
            a:      0.0,
            b:      0.0,
            freq:   1000.0,
            z:      0.0,
        }
    }

    pub fn reset(&mut self) {
        self.z = 0.0;
    }

    #[inline]
    fn recalc(&mut self) {
        self.b = (-F::TAU * self.freq * self.israte).exp();
        self.a = 1.0 - self.b;
    }

    pub fn set_sample_rate(&mut self, srate: F) {
        self.israte = 1.0 / srate;
        self.recalc();
    }

    #[inline]
    pub fn set_freq(&mut self, freq: F) {
        if freq != self.freq {
            self.freq = freq;
            self.recalc();
        }
    }

    #[inline]
    pub fn process(&mut self, input: F) -> F {
        self.z = self.a * input + self.z * self.b;
        self.z
    }
}

// one pole hp from valley rack free:
// https://github.com/ValleyAudio/ValleyRackFree/blob/v1.0/src/Common/DSP/OnePoleFilters.cpp
#[inline]
/// Process a very simple one pole 6dB high pass filter.
/// Useful in various applications.
///
/// * `input`  - Input sample
/// * `freq`   - Frequency between 1.0 and 22000.0Hz
/// * `israte` - 1.0 / samplerate
/// * `z`      - The first internal buffer of the filter.
/// * `y`      - The second internal buffer of the filter.
///
///```
///    use hexodsp::dsp::helpers::*;
///
///    let samples  = vec![0.0; 44100];
///    let mut z    = 0.0;
///    let mut y    = 0.0;
///    let mut freq = 1000.0;
///
///    for s in samples.iter() {
///        let s_out =
///            process_1pole_highpass(*s, freq, 1.0 / 44100.0, &mut z, &mut y);
///        // ... do something with the result here.
///    }
///```
pub fn process_1pole_highpass(input: f32, freq: f32, israte: f32, z: &mut f32, y: &mut f32) -> f32 {
    let b  = (-std::f32::consts::TAU * freq * israte).exp();
    let a  = (1.0 + b) / 2.0;

    let v =
          a  * input
        - a  * *z
        + b  * *y;
    *y = v;
    *z = input;
    v
}

#[derive(Debug, Clone, Copy, Default)]
pub struct OnePoleHPF<F: Float> {
    israte:     F,
    a:          F,
    b:          F,
    freq:       F,
    z:          F,
    y:          F,
}

impl<F: Float + FloatConst> OnePoleHPF<F> {
    pub fn new() -> Self {
        Self {
            israte: 1.0 / 44100.0,
            a:      0.0,
            b:      0.0,
            freq:   1000.0,
            z:      0.0,
            y:      0.0,
        }
    }

    pub fn reset(&mut self) {
        self.z = 0.0;
        self.y = 0.0;
    }

    #[inline]
    fn recalc(&mut self) {
        self.b = (-F::TAU * self.freq * self.israte).exp();
        self.a = (1.0 + self.b) / 2.0;
    }


    pub fn set_sample_rate(&mut self, srate: F) {
        self.israte = 1.0 / srate;
        self.recalc();
    }

    #[inline]
    pub fn set_freq(&mut self, freq: F) {
        if freq != self.freq {
            self.freq = freq;
            self.recalc();
        }
    }

    #[inline]
    pub fn process(&mut self, input: F) -> F {
        let v =
              self.a * input
            - self.a * self.z
            + self.b * self.y;

        self.y = v;
        self.z = input;

        v
    }
}


// one pole from:
// http://www.willpirkle.com/Downloads/AN-4VirtualAnalogFilters.pdf
// (page 5)
#[inline]
/// Process a very simple one pole 6dB low pass filter in TPT form.
/// Useful in various applications, from usage in a synthesizer to
/// damping stuff in a reverb/delay.
///
/// * `input`  - Input sample
/// * `freq`   - Frequency between 1.0 and 22000.0Hz
/// * `israte` - 1.0 / samplerate
/// * `z`      - The internal one sample buffer of the filter.
///
///```
///    use hexodsp::dsp::helpers::*;
///
///    let samples  = vec![0.0; 44100];
///    let mut z    = 0.0;
///    let mut freq = 1000.0;
///
///    for s in samples.iter() {
///        let s_out =
///            process_1pole_tpt_highpass(*s, freq, 1.0 / 44100.0, &mut z);
///        // ... do something with the result here.
///    }
///```
pub fn process_1pole_tpt_lowpass(input: f32, freq: f32, israte: f32, z: &mut f32) -> f32 {
    let g = (std::f32::consts::PI * freq * israte).tan();
    let a = g / (1.0 + g);

    let v1 = a * (input - *z);
    let v2 = v1 + *z;
    *z = v2 + v1;

    // let (m0, m1) = (0.0, 1.0);
    // (m0 * input + m1 * v2) as f32);
    v2
}

// one pole from:
// http://www.willpirkle.com/Downloads/AN-4VirtualAnalogFilters.pdf
// (page 5)
#[inline]
/// Process a very simple one pole 6dB high pass filter in TPT form.
/// Useful in various applications.
///
/// * `input`  - Input sample
/// * `freq`   - Frequency between 1.0 and 22000.0Hz
/// * `israte` - 1.0 / samplerate
/// * `z`      - The internal one sample buffer of the filter.
///
///```
///    use hexodsp::dsp::helpers::*;
///
///    let samples  = vec![0.0; 44100];
///    let mut z    = 0.0;
///    let mut freq = 1000.0;
///
///    for s in samples.iter() {
///        let s_out =
///            process_1pole_tpt_lowpass(*s, freq, 1.0 / 44100.0, &mut z);
///        // ... do something with the result here.
///    }
///```
pub fn process_1pole_tpt_highpass(input: f32, freq: f32, israte: f32, z: &mut f32) -> f32 {
    let g  = (std::f32::consts::PI * freq * israte).tan();
    let a1 = g / (1.0 + g);

    let v1 = a1 * (input - *z);
    let v2 = v1 + *z;
    *z = v2 + v1;

    input - v2
}

/// The internal oversampling factor of [process_hal_chamberlin_svf].
const FILTER_OVERSAMPLE_HAL_CHAMBERLIN : usize = 2;
// Hal Chamberlin's State Variable (12dB/oct) filter
// https://www.earlevel.com/main/2003/03/02/the-digital-state-variable-filter/
// Inspired by SynthV1 by Rui Nuno Capela, under the terms of
// GPLv2 or any later:
/// Process a HAL Chamberlin filter with two delays/state variables that is 12dB.
/// The filter does internal oversampling with very simple decimation to
/// rise the stability for cutoff frequency up to 16kHz.
///
/// * `input` - Input sample.
/// * `freq` - Frequency in Hz. Please keep it inside 0.0 to 16000.0 Hz!
/// otherwise the filter becomes unstable.
/// * `res`  - Resonance from 0.0 to 0.99. Resonance of 1.0 is not recommended,
/// as the filter will then oscillate itself out of control.
/// * `israte` - 1.0 divided by the sampling rate (eg. 1.0 / 44100.0).
/// * `band` - First state variable, containing the band pass result
/// after processing.
/// * `low` - Second state variable, containing the low pass result
/// after processing.
///
/// Returned are the results of the high and notch filter.
///
///```
///    use hexodsp::dsp::helpers::*;
///
///    let samples  = vec![0.0; 44100];
///    let mut band = 0.0;
///    let mut low  = 0.0;
///    let mut freq = 1000.0;
///
///    for s in samples.iter() {
///        let (high, notch) =
///            process_hal_chamberlin_svf(
///                *s, freq, 0.5, 1.0 / 44100.0, &mut band, &mut low);
///        // ... do something with the result here.
///    }
///```
#[inline]
pub fn process_hal_chamberlin_svf(
    input: f32, freq: f32, res: f32, israte: f32, band: &mut f32, low: &mut f32)
    -> (f32, f32)
{
    let q      = 1.0 - res;
    let cutoff = 2.0 * (std::f32::consts::PI * freq * 0.5 * israte).sin();

    let mut high  = 0.0;
    let mut notch = 0.0;

    for _ in 0..FILTER_OVERSAMPLE_HAL_CHAMBERLIN {
        *low  += cutoff * *band;
        high = input - *low - q * *band;
        *band += cutoff * high;
        notch = high + *low;
    }

    //d// println!("q={:4.2} cut={:8.3} freq={:8.1} LP={:8.3} HP={:8.3} BP={:8.3} N={:8.3}",
    //d//     q, cutoff, freq, *low, high, *band, notch);

    (high, notch)
}

/// This function processes a Simper SVF with 12dB. It's a much newer algorithm
/// for filtering and provides easy to calculate multiple outputs.
///
/// * `input` - Input sample.
/// * `freq` - Frequency in Hz.
/// otherwise the filter becomes unstable.
/// * `res`  - Resonance from 0.0 to 0.99. Resonance of 1.0 is not recommended,
/// as the filter will then oscillate itself out of control.
/// * `israte` - 1.0 divided by the sampling rate (eg. 1.0 / 44100.0).
/// * `band` - First state variable, containing the band pass result
/// after processing.
/// * `low` - Second state variable, containing the low pass result
/// after processing.
///
/// This function returns the low pass, band pass and high pass signal.
/// For a notch or peak filter signal, please consult the following example:
///
///```
///    use hexodsp::dsp::helpers::*;
///
///    let samples   = vec![0.0; 44100];
///    let mut ic1eq = 0.0;
///    let mut ic2eq = 0.0;
///    let mut freq  = 1000.0;
///
///    for s in samples.iter() {
///        let (low, band, high) =
///            process_simper_svf(
///                *s, freq, 0.5, 1.0 / 44100.0, &mut ic1eq, &mut ic2eq);
///
///        // You can easily calculate the notch and peak results too:
///        let notch = low + high;
///        let peak  = low - high;
///        // ... do something with the result here.
///    }
///```
// Simper SVF implemented from
// https://cytomic.com/files/dsp/SvfLinearTrapezoidalSin.pdf
// Big thanks go to Andrew Simper @ Cytomic for developing and publishing
// the paper.
#[inline]
pub fn process_simper_svf(
    input: f32, freq: f32, res: f32, israte: f32, ic1eq: &mut f32, ic2eq: &mut f32
) -> (f32, f32, f32) {
    // XXX: the 1.989 were tuned by hand, so the resonance is more audible.
    let k = 2f32 - (1.989f32 * res);
    let w = std::f32::consts::PI * freq * israte;

    let s1  = w.sin();
    let s2  = (2.0 * w).sin();
    let nrm = 1.0 / (2.0 + k * s2);

    let g0 = s2 * nrm;
    let g1 = (-2.0 * s1 * s1 - k * s2) * nrm;
    let g2 = (2.0 * s1 * s1) * nrm;

    let t0 = input - *ic2eq;
    let t1 = g0 * t0 + g1 * *ic1eq;
    let t2 = g2 * t0 + g0 * *ic1eq;

    let v1 = t1 + *ic1eq;
    let v2 = t2 + *ic2eq;

    *ic1eq += 2.0 * t1;
    *ic2eq += 2.0 * t2;

    // low   = v2
    // band  = v1
    // high  = input - k * v1 - v2
    // notch = low + high            = input - k * v1
    // peak  = low - high            = 2 * v2 - input + k * v1
    // all   = low + high - k * band = input - 2 * k * v1

    (v2, v1, input - k * v1 - v2)
}

/// This function implements a simple Stilson/Moog low pass filter with 24dB.
/// It provides only a low pass output.
///
/// * `input` - Input sample.
/// * `freq` - Frequency in Hz.
/// otherwise the filter becomes unstable.
/// * `res`  - Resonance from 0.0 to 0.99. Resonance of 1.0 is not recommended,
/// as the filter will then oscillate itself out of control.
/// * `israte` - 1.0 divided by the sampling rate (`1.0 / 44100.0`).
/// * `b0` to `b3` - Internal values used for filtering.
/// * `delay` - A buffer holding other delayed samples.
///
///```
///    use hexodsp::dsp::helpers::*;
///
///    let samples  = vec![0.0; 44100];
///    let mut b0   = 0.0;
///    let mut b1   = 0.0;
///    let mut b2   = 0.0;
///    let mut b3   = 0.0;
///    let mut delay = [0.0; 4];
///    let mut freq = 1000.0;
///
///    for s in samples.iter() {
///        let low =
///            process_stilson_moog(
///                *s, freq, 0.5, 1.0 / 44100.0,
///                &mut b0, &mut b1, &mut b2, &mut b3,
///                &mut delay);
///
///        // ... do something with the result here.
///    }
///```
// Stilson/Moog implementation partly translated from here:
// https://github.com/ddiakopoulos/MoogLadders/blob/master/src/MusicDSPModel.h
// without any copyright as found on musicdsp.org
// (http://www.musicdsp.org/showone.php?id=24).
//
// It's also found on MusicDSP and has probably no proper license anyways.
// See also: https://github.com/ddiakopoulos/MoogLadders
// and https://github.com/rncbc/synthv1/blob/master/src/synthv1_filter.h#L103
// and https://github.com/ddiakopoulos/MoogLadders/blob/master/src/MusicDSPModel.h
#[inline]
pub fn process_stilson_moog(
    input: f32, freq: f32, res: f32, israte: f32,
    b0: &mut f32, b1: &mut f32, b2: &mut f32, b3: &mut f32,
    delay: &mut [f32; 4],
) -> f32 {

    let cutoff = 2.0 * freq * israte;

    let p = cutoff * (1.8 - 0.8 * cutoff);
    let k = 2.0 * (cutoff * std::f32::consts::PI * 0.5).sin() - 1.0;

    let t1 = (1.0 - p) * 1.386249;
    let t2 = 12.0 + t1 * t1;

    let res = res * (t2 + 6.0 * t1) / (t2 - 6.0 * t1);

	let x = input - res * *b3;

    // Four cascaded one-pole filters (bilinear transform)
    *b0 = x   * p + delay[0] * p - k * *b0;
    *b1 = *b0 * p + delay[1] * p - k * *b1;
    *b2 = *b1 * p + delay[2] * p - k * *b2;
    *b3 = *b2 * p + delay[3] * p - k * *b3;

    // Clipping band-limited sigmoid
    *b3 -= (*b3 * *b3 * *b3) * 0.166667;

    delay[0] = x;
    delay[1] = *b0;
    delay[2] = *b1;
    delay[3] = *b2;

    *b3
}

// translated from Odin 2 Synthesizer Plugin
// Copyright (C) 2020 TheWaveWarden
// under GPLv3 or any later
#[derive(Debug, Clone, Copy)]
pub struct DCBlockFilter<F: Float> {
    xm1:    F,
    ym1:    F,
    r:      F,
}

impl<F: Float> DCBlockFilter<F> {
    pub fn new() -> Self {
        Self {
            xm1: 0.0,
            ym1: 0.0,
            r:   0.995,
        }
    }

    pub fn reset(&mut self) {
        self.xm1 = 0.0;
        self.ym1 = 0.0;
    }

    pub fn set_sample_rate(&mut self, srate: F) {
        self.r = 0.995;
        if srate > 90000.0 {
            self.r = 0.9965;
        } else if srate > 120000.0 {
            self.r = 0.997;
        }
    }

    pub fn next(&mut self, input: F) -> F {
        let y = input as f64 - self.xm1 + self.r * self.ym1;
        self.xm1 = input as f64;
        self.ym1 = y;
        y as F
    }
}

// PolyBLEP by Tale
// (slightly modified)
// http://www.kvraudio.com/forum/viewtopic.php?t=375517
// from http://www.martin-finke.de/blog/articles/audio-plugins-018-polyblep-oscillator/
//
// default for `pw' should be 1.0, it's the pulse width
// for the square wave.
#[allow(dead_code)]
fn poly_blep_64(t: f64, dt: f64) -> f64 {
    if t < dt {
        let t = t / dt;
        2. * t - (t * t) - 1.

    } else if t > (1.0 - dt) {
        let t = (t - 1.0) / dt;
        (t * t) + 2. * t + 1.

    } else {
        0.
    }
}

fn poly_blep(t: f32, dt: f32) -> f32 {
    if t < dt {
        let t = t / dt;
        2. * t - (t * t) - 1.

    } else if t > (1.0 - dt) {
        let t = (t - 1.0) / dt;
        (t * t) + 2. * t + 1.

    } else {
        0.
    }
}

/// This is a band-limited oscillator based on the PolyBlep technique.
/// Here is a quick example on how to use it:
///
///```
/// use hexodsp::dsp::helpers::{PolyBlepOscillator, rand_01};
///
/// // Randomize the initial phase to make cancellation on summing less
/// // likely:
/// let mut osc =
///     PolyBlepOscillator::new(rand_01() * 0.25);
///
///
/// let freq   = 440.0; // Hz
/// let israte = 1.0 / 44100.0; // Seconds per Sample
/// let pw     = 0.2;   // Pulse-Width for the next_pulse()
/// let waveform = 0;   // 0 being pulse in this example, 1 being sawtooth
///
/// let mut block_of_samples = [0.0; 128];
/// // in your process function:
/// for output_sample in block_of_samples.iter_mut() {
///    *output_sample =
///        if waveform == 1 {
///             osc.next_saw(freq, israte)
///        } else {
///             osc.next_pulse(freq, israte, pw)
///        }
/// }
///```
#[derive(Debug, Clone)]
pub struct PolyBlepOscillator {
    phase:       f32,
    init_phase:  f32,
    last_output: f32,
}

impl PolyBlepOscillator {
    /// Create a new instance of [PolyBlepOscillator].
    ///
    /// * `init_phase` - Initial phase of the oscillator.
    /// Range of this parameter is from 0.0 to 1.0. Passing a random
    /// value is advised for preventing phase cancellation when summing multiple
    /// oscillators.
    ///
    ///```
    /// use hexodsp::dsp::helpers::{PolyBlepOscillator, rand_01};
    ///
    /// let mut osc = PolyBlepOscillator::new(rand_01() * 0.25);
    ///```
    pub fn new(init_phase: f32) -> Self {
        Self {
            phase:       0.0,
            last_output: 0.0,
            init_phase,
        }
    }

    /// Reset the internal state of the oscillator as if you just called
    /// [PolyBlepOscillator::new].
    #[inline]
    pub fn reset(&mut self) {
        self.phase       = self.init_phase;
        self.last_output = 0.0;
    }

    /// Creates the next sample of a sine wave.
    ///
    /// * `freq` - The frequency in Hz.
    /// * `israte` - The inverse sampling rate, or seconds per sample as in eg. `1.0 / 44100.0`.
    ///```
    /// use hexodsp::dsp::helpers::{PolyBlepOscillator, rand_01};
    ///
    /// let mut osc = PolyBlepOscillator::new(rand_01() * 0.25);
    ///
    /// let freq   = 440.0; // Hz
    /// let israte = 1.0 / 44100.0; // Seconds per Sample
    ///
    /// // ...
    /// let sample = osc.next_sin(freq, israte);
    /// // ...
    ///```
    #[inline]
    pub fn next_sin(&mut self, freq: f32, israte: f32) -> f32 {
        let phase_inc = freq * israte;

        let s = fast_sin(self.phase * 2.0 * std::f32::consts::PI);

        self.phase += phase_inc;
        self.phase = self.phase.fract();

        s as f32
    }

    /// Creates the next sample of a triangle wave. Please note that the
    /// bandlimited waveform needs a few initial samples to swing in.
    ///
    /// * `freq` - The frequency in Hz.
    /// * `israte` - The inverse sampling rate, or seconds per sample as in eg. `1.0 / 44100.0`.
    ///```
    /// use hexodsp::dsp::helpers::{PolyBlepOscillator, rand_01};
    ///
    /// let mut osc = PolyBlepOscillator::new(rand_01() * 0.25);
    ///
    /// let freq   = 440.0; // Hz
    /// let israte = 1.0 / 44100.0; // Seconds per Sample
    ///
    /// // ...
    /// let sample = osc.next_tri(freq, israte);
    /// // ...
    ///```
    #[inline]
    pub fn next_tri(&mut self, freq: f32, israte: f32) -> f32 {
        let phase_inc = freq * israte;

        let mut s =
            if self.phase < 0.5 { 1.0 }
            else                { -1.0 };

        s += poly_blep(self.phase, phase_inc);
        s -= poly_blep((self.phase + 0.5).fract(), phase_inc);

        // leaky integrator: y[n] = A * x[n] + (1 - A) * y[n-1]
        s = phase_inc * s + (1.0 - phase_inc) * self.last_output;
        self.last_output = s;

        self.phase += phase_inc;
        self.phase = self.phase.fract();

        // the signal is a bit too weak, we need to amplify it
        // or else the volume diff between the different waveforms
        // is too big:
        s * 4.0
    }

    /// Creates the next sample of a sawtooth wave.
    ///
    /// * `freq` - The frequency in Hz.
    /// * `israte` - The inverse sampling rate, or seconds per sample as in eg. `1.0 / 44100.0`.
    ///```
    /// use hexodsp::dsp::helpers::{PolyBlepOscillator, rand_01};
    ///
    /// let mut osc = PolyBlepOscillator::new(rand_01() * 0.25);
    ///
    /// let freq   = 440.0; // Hz
    /// let israte = 1.0 / 44100.0; // Seconds per Sample
    ///
    /// // ...
    /// let sample = osc.next_saw(freq, israte);
    /// // ...
    ///```
    #[inline]
    pub fn next_saw(&mut self, freq: f32, israte: f32) -> f32 {
        let phase_inc = freq * israte;

        let mut s = (2.0 * self.phase) - 1.0;
        s -= poly_blep(self.phase, phase_inc);

        self.phase += phase_inc;
        self.phase = self.phase.fract();

        s
    }

    /// Creates the next sample of a pulse wave.
    /// In comparison to [PolyBlepOscillator::next_pulse_no_dc] this
    /// version is DC compensated, so that you may add multiple different
    /// pulse oscillators for a unison effect without too big DC issues.
    ///
    /// * `freq` - The frequency in Hz.
    /// * `israte` - The inverse sampling rate, or seconds per sample as in eg. `1.0 / 44100.0`.
    /// * `pw` - The pulse width. Use the value 0.0 for a square wave.
    ///```
    /// use hexodsp::dsp::helpers::{PolyBlepOscillator, rand_01};
    ///
    /// let mut osc = PolyBlepOscillator::new(rand_01() * 0.25);
    ///
    /// let freq   = 440.0; // Hz
    /// let israte = 1.0 / 44100.0; // Seconds per Sample
    /// let pw     = 0.0; // 0.0 is a square wave.
    ///
    /// // ...
    /// let sample = osc.next_pulse(freq, israte, pw);
    /// // ...
    ///```
    #[inline]
    pub fn next_pulse(&mut self, freq: f32, israte: f32, pw: f32) -> f32 {
        let phase_inc = freq * israte;

        let pw = (0.1 * pw) + ((1.0 - pw) * 0.5); // some scaling
        let dc_compensation = (0.5 - pw) * 2.0;

        let mut s =
            if self.phase < pw { 1.0 }
            else { -1.0 };

        s += poly_blep(self.phase, phase_inc);
        s -= poly_blep((self.phase + (1.0 - pw)).fract(),
                            phase_inc);

        s += dc_compensation;

        self.phase += phase_inc;
        self.phase = self.phase.fract();

        s
    }

    /// Creates the next sample of a pulse wave.
    /// In comparison to [PolyBlepOscillator::next_pulse] this
    /// version is not DC compensated. So be careful when adding multiple
    /// of this or generally using it in an audio context.
    ///
    /// * `freq` - The frequency in Hz.
    /// * `israte` - The inverse sampling rate, or seconds per sample as in eg. `1.0 / 44100.0`.
    /// * `pw` - The pulse width. Use the value 0.0 for a square wave.
    ///```
    /// use hexodsp::dsp::helpers::{PolyBlepOscillator, rand_01};
    ///
    /// let mut osc = PolyBlepOscillator::new(rand_01() * 0.25);
    ///
    /// let freq   = 440.0; // Hz
    /// let israte = 1.0 / 44100.0; // Seconds per Sample
    /// let pw     = 0.0; // 0.0 is a square wave.
    ///
    /// // ...
    /// let sample = osc.next_pulse_no_dc(freq, israte, pw);
    /// // ...
    ///```
    #[inline]
    pub fn next_pulse_no_dc(&mut self, freq: f32, israte: f32, pw: f32) -> f32 {
        let phase_inc = freq * israte;

        let pw = (0.1 * pw) + ((1.0 - pw) * 0.5); // some scaling

        let mut s =
            if self.phase < pw { 1.0 }
            else { -1.0 };

        s += poly_blep(self.phase, phase_inc);
        s -= poly_blep((self.phase + (1.0 - pw)).fract(),
                            phase_inc);

        self.phase += phase_inc;
        self.phase = self.phase.fract();

        s
    }
}

// This oscillator is based on the work "VECTOR PHASESHAPING SYNTHESIS"
// by: Jari Kleimola*, Victor Lazzarini†, Joseph Timoney†, Vesa Välimäki*
// *Aalto University School of Electrical Engineering Espoo, Finland;
// †National University of Ireland, Maynooth Ireland
//
// See also this PDF: http://recherche.ircam.fr/pub/dafx11/Papers/55_e.pdf
/// Vector Phase Shaping Oscillator.
/// The parameters `d` and `v` control the shape of the sinus
/// wave. This leads to interesting modulation properties of those
/// control values.
///
///```
/// use hexodsp::dsp::helpers::{VPSOscillator, rand_01};
///
/// // Randomize the initial phase to make cancellation on summing less
/// // likely:
/// let mut osc =
///     VPSOscillator::new(rand_01() * 0.25);
///
///
/// let freq   = 440.0; // Hz
/// let israte = 1.0 / 44100.0; // Seconds per Sample
/// let d      = 0.5;   // Range: 0.0 to 1.0
/// let v      = 0.75;  // Range: 0.0 to 1.0
///
/// let mut block_of_samples = [0.0; 128];
/// // in your process function:
/// for output_sample in block_of_samples.iter_mut() {
///     // It is advised to limit the `v` value, because with certain
///     // `d` values the combination creates just a DC offset.
///     let v = VPSOscillator::limit_v(d, v);
///     *output_sample = osc.next(freq, israte, d, v);
/// }
///```
///
/// It can be beneficial to apply distortion and oversampling.
/// Especially oversampling can be important for some `d` and `v`
/// combinations, even without distortion.
///
///```
/// use hexodsp::dsp::helpers::{VPSOscillator, rand_01, apply_distortion};
/// use hexodsp::dsp::biquad::Oversampling;
///
/// let mut osc = VPSOscillator::new(rand_01() * 0.25);
/// let mut ovr : Oversampling<4> = Oversampling::new();
///
/// let freq   = 440.0; // Hz
/// let israte = 1.0 / 44100.0; // Seconds per Sample
/// let d      = 0.5;   // Range: 0.0 to 1.0
/// let v      = 0.75;  // Range: 0.0 to 1.0
///
/// let mut block_of_samples = [0.0; 128];
/// // in your process function:
/// for output_sample in block_of_samples.iter_mut() {
///     // It is advised to limit the `v` value, because with certain
///     // `d` values the combination creates just a DC offset.
///     let v = VPSOscillator::limit_v(d, v);
///
///     let overbuf = ovr.resample_buffer();
///     for b in overbuf {
///         *b = apply_distortion(osc.next(freq, israte, d, v), 0.9,  1);
///     }
///     *output_sample = ovr.downsample();
/// }
///```
#[derive(Debug, Clone)]
pub struct VPSOscillator {
    phase:       f32,
    init_phase:  f32,
}

impl VPSOscillator {
    /// Create a new instance of [VPSOscillator].
    ///
    /// * `init_phase` - The initial phase of the oscillator.
    pub fn new(init_phase: f32) -> Self {
        Self {
            phase:       0.0,
            init_phase,
        }
    }

    /// Reset the phase of the oscillator to the initial phase.
    #[inline]
    pub fn reset(&mut self) {
        self.phase = self.init_phase;
    }

    #[inline]
    fn s(p: f32) -> f32 {
        -(std::f32::consts::TAU * p).cos()
    }

    #[inline]
    fn phi_vps(x: f32, v: f32, d: f32) -> f32 {
        if x < d {
            (v * x) / d
        } else {
            v + ((1.0 - v) * (x - d))/(1.0 - d)
        }
    }

    /// This rather complicated function blends out some
    /// combinations of 'd' and 'v' that just lead to a constant DC
    /// offset. Which is not very useful in an audio oscillator
    /// context.
    ///
    /// Call this before passing `v` to [VPSOscillator::next].
    #[inline]
    pub fn limit_v(d: f32, v: f32) -> f32 {
        let delta = 0.5 - (d - 0.5).abs();
        if delta < 0.05 {
            let x = (0.05 - delta) * 19.99;
            if d < 0.5 {
                let mm = x * 0.5;
                let max = 1.0 - mm;
                if v > max && v < 1.0 {
                    max
                } else if v >= 1.0 && v < (1.0 + mm) {
                    1.0 + mm
                } else {
                    v
                }
            } else {
                if v < 1.0 {
                    v.clamp(x * 0.5, 1.0)
                } else {
                    v
                }
            }
        } else {
            v
        }
    }

    /// Creates the next sample of this oscillator.
    ///
    /// * `freq` - The frequency in Hz.
    /// * `israte` - The inverse sampling rate, or seconds per sample as in eg. `1.0 / 44100.0`.
    /// * `d` - The phase distortion parameter `d` which must be in the range `0.0` to `1.0`.
    /// * `v` - The phase distortion parameter `v` which must be in the range `0.0` to `1.0`.
    ///
    /// It is advised to limit the `v` using the [VPSOscillator::limit_v] function
    /// before calling this function. To prevent DC offsets when modulating the parameters.
    pub fn next(&mut self, freq: f32, israte: f32, d: f32, v: f32) -> f32 {
        let s = Self::s(Self::phi_vps(self.phase, v, d));

        self.phase += freq * israte;
        self.phase = self.phase.fract();

        s
    }
}

// Adapted from https://github.com/ValleyAudio/ValleyRackFree/blob/v1.0/src/Common/DSP/LFO.hpp
//
// ValleyRackFree Copyright (C) 2020, Valley Audio Soft, Dale Johnson
// Adapted under the GPL-3.0-or-later License.
/// An LFO with a variable reverse point, which can go from reverse Saw, to Tri
/// and to Saw, depending on the reverse point.
#[derive(Debug, Clone, Copy)]
pub struct TriSawLFO<F: Float> {
    /// The (inverse) sample rate. Eg. 1.0 / 44100.0.
    israte: F,
    /// The current oscillator phase.
    phase:  F,
    /// The point from where the falling edge will be used.
    rev:    F,
    /// Whether the LFO is currently rising
    rising: bool,
    /// The frequency.
    freq:   F,
    /// Precomputed rise/fall rate of the LFO.
    rise_r: F,
    fall_r: F,
    /// Initial phase offset.
    init_phase: F,
}

impl<F: Float> TriSawLFO<F> {
    pub fn new() -> Self {
        let mut this = Self {
            israte: 1.0 / 44100.0,
            phase:  0.0,
            rev:    0.5,
            rising: true,
            freq:   1.0,
            fall_r: 0.0,
            rise_r: 0.0,
            init_phase: 0.0,
        };
        this.recalc();
        this
    }

    pub fn set_phase_offs(&mut self, phase: F) {
        self.init_phase = phase;
        self.phase      = phase;
    }

    #[inline]
    fn recalc(&mut self) {
        self.rev    = self.rev.clamp(0.0001, 0.999);
        self.rise_r =  1.0 / self.rev;
        self.fall_r = -1.0 / (1.0 - self.rev);
    }

    pub fn set_sample_rate(&mut self, srate: F) {
        self.israte = 1.0 / (srate as F);
        self.recalc();
    }

    pub fn reset(&mut self) {
        self.phase  = self.init_phase;
        self.rev    = 0.5;
        self.rising = true;
    }

    #[inline]
    pub fn set(&mut self, freq: F, rev: F) {
        self.freq = freq as F;
        self.rev  = rev  as F;
        self.recalc();
    }

    #[inline]
    pub fn next_unipolar(&mut self) -> F {
        if self.phase >= 1.0 {
            self.phase -= 1.0;
            self.rising = true;
        }

        if self.phase >= self.rev {
            self.rising = false;
        }

        let s =
            if self.rising {
                self.phase * self.rise_r
            } else {
                self.phase * self.fall_r - self.fall_r
            };

        self.phase += self.freq * self.israte;

        s
    }

    #[inline]
    pub fn next_bipolar(&mut self) -> F {
        (self.next_unipolar() * 2.0) - 1.0
    }
}

#[macro_export]
macro_rules! fa_distort { ($formatter: expr, $v: expr, $denorm_v: expr) => { {
    let s =
        match ($v.round() as usize) {
            0  => "Off",
            1  => "TanH",
            2  => "B.D.Jong",
            3  => "Fold",
            _  => "?",
        };
    write!($formatter, "{}", s)
} } }

#[inline]
pub fn apply_distortion(s: f32, damt: f32, dist_type: u8) -> f32 {
    match dist_type {
        1 => (damt.clamp(0.01, 1.0) * 100.0 * s).tanh(),
        2 => f_distort(1.0, damt * damt * damt * 1000.0, s),
        3 => {
            let damt = damt.clamp(0.0, 0.99);
            let damt = 1.0 - damt * damt;
            f_fold_distort(1.0, damt, s) * (1.0 / damt)
        },
        _ => s,
    }
}

//pub struct UnisonBlep {
//    oscs: Vec<PolyBlepOscillator>,
////    dc_block: crate::filter::DCBlockFilter,
//}
//
//impl UnisonBlep {
//    pub fn new(max_unison: usize) -> Self {
//        let mut oscs = vec![];
//        let mut rng = RandGen::new();
//
//        let dis_init_phase = 0.05;
//        for i in 0..(max_unison + 1) {
//            // randomize phases so we fatten the unison, get
//            // less DC and not an amplified signal until the
//            // detune desyncs the waves.
//            // But no random phase for first, so we reduce the click
//            let init_phase =
//                if i == 0 { 0.0 } else { rng.next_open01() };
//            oscs.push(PolyBlepOscillator::new(init_phase));
//        }
//
//        Self {
//            oscs,
////            dc_block: crate::filter::DCBlockFilter::new(),
//        }
//    }
//
//    pub fn set_sample_rate(&mut self, srate: f32) {
////        self.dc_block.set_sample_rate(srate);
//        for o in self.oscs.iter_mut() {
//            o.set_sample_rate(srate);
//        }
//    }
//
//    pub fn reset(&mut self) {
////        self.dc_block.reset();
//        for o in self.oscs.iter_mut() {
//            o.reset();
//        }
//    }
//
//    pub fn next<P: OscillatorInputParams>(&mut self, params: &P) -> f32 {
//        let unison =
//            (params.unison().floor() as usize)
//            .min(self.oscs.len() - 1);
//        let detune = params.detune() as f64;
//
//        let mix = (1.0 / ((unison + 1) as f32)).sqrt();
//
//        let mut s = mix * self.oscs[0].next(params, 0.0);
//
//        for u in 0..unison {
//            let detune_factor =
//                detune * (((u / 2) + 1) as f64
//                          * if (u % 2) == 0 { 1.0 } else { -1.0 });
//            s += mix * self.oscs[u + 1].next(params, detune_factor * 0.01);
//        }
//
////        self.dc_block.next(s)
//        s
//    }
//}

#[cfg(test)]
mod tests {
    use super::*;

    #[test]
    fn check_range2p_exp() {
        let a = p2range_exp(0.5, 1.0, 100.0);
        let x = range2p_exp(a, 1.0, 100.0);

        assert!((x - 0.5).abs() < std::f32::EPSILON);
    }

    #[test]
    fn check_range2p() {
        let a = p2range(0.5, 1.0, 100.0);
        let x = range2p(a, 1.0, 100.0);

        assert!((x - 0.5).abs() < std::f32::EPSILON);
    }
}