#pragma once
#include <vector>
#include <cmath>
#include <algorithm>
#include <cstdint>

namespace FDNReverb {

     // -----------------------------------------------------------------------------
     // memory pool (Single-Large Buffer)
     // -----------------------------------------------------------------------------
    class DelayMemoryPool {
    public:
        void allocate(size_t totalSamples) {
            buffer.assign(totalSamples, 0.0f);
            allocOffset = 0;
        }

         // pointer sized up to the next power of two (also outputs an index mask)
        float* requestMemory(size_t samplesNeeded, int& outMask) {
            size_t powerOfTwoSize = 1;
            while (powerOfTwoSize < samplesNeeded) powerOfTwoSize *= 2;

            if (allocOffset + powerOfTwoSize > buffer.size()) return nullptr;

            float* ptr = buffer.data() + allocOffset;
            outMask = static_cast<int>(powerOfTwoSize - 1);
            allocOffset += powerOfTwoSize;

            return ptr;
        }

        void clear() { std::fill(buffer.begin(), buffer.end(), 0.0f); }

    private:
        std::vector<float> buffer;
        size_t allocOffset{ 0 };
    };

     // -----------------------------------------------------------------------------
     // interpolation
     // -----------------------------------------------------------------------------
    class LinearDelayLine {
    public:
        void init(float* memory, int bitmask) {
            buffer = memory;
            mask = bitmask;
            writeIndex = 0;
        }

         // linear interpolation ( high band natural Air Absorption )
        inline float read(float delayInSamples) const noexcept {
            int id = static_cast<int>(delayInSamples);
            float frac = delayInSamples - static_cast<float>(id);

             // bitwise ops undefined behavior completely , uint32_t
            uint32_t uWrite = static_cast<uint32_t>(writeIndex);
            uint32_t uId = static_cast<uint32_t>(id);
            uint32_t uMask = static_cast<uint32_t>(mask);

            int readIdx1 = static_cast<int>((uWrite - uId) & uMask);
            int readIdx2 = static_cast<int>((uWrite - uId - 1) & uMask);

            return buffer[readIdx1] + frac * (buffer[readIdx2] - buffer[readIdx1]);
        }

        inline void write(float input) noexcept {
            buffer[writeIndex] = input;
            writeIndex = (writeIndex + 1) & mask;
        }

    private:
        float* buffer{ nullptr };
        int mask{ 0 };
        int writeIndex{ 0 };
    };

     // -----------------------------------------------------------------------------
     // Thiran allpass interpolation (preserves the phase response)
     // linear interpolation would dull high-band decay (sinc(pi*f) rolloff), so use a Thiran allpass
     // which keeps |H(w)| = 1, preserving high-band clarity in the FDN feedback loops.
     // -----------------------------------------------------------------------------
    class ThiranDelayLine {
    public:
        void init(float* memory, int bitmask) {
            buffer = memory;
            mask = bitmask;
            writeIndex = 0;
            thiranX1 = 0.0f;
            thiranY1 = 0.0f;
        }

        void resetState() noexcept {
            thiranX1 = 0.0f;
            thiranY1 = 0.0f;
        }

         // Thiran first-order allpass: y[n] = a*x[n] + x[n-1] - a*y[n-1]
        // a = (1-D)/(1+D), D = fractional delay
        inline float read(float delayInSamples) noexcept {
            int id = static_cast<int>(delayInSamples);
            float frac = delayInSamples - static_cast<float>(id);

             // clamp below to avoid instability as frac->0, a->1
            frac = std::max(frac, 0.1f);
            const float a = (1.0f - frac) / (1.0f + frac);

            uint32_t uWrite = static_cast<uint32_t>(writeIndex);
            uint32_t uId = static_cast<uint32_t>(id);
            uint32_t uMask = static_cast<uint32_t>(mask);

            float xn = buffer[static_cast<int>((uWrite - uId) & uMask)];

            float yn = a * xn + thiranX1 - a * thiranY1;
            thiranX1 = xn;
            thiranY1 = yn;

            return yn;
        }

        inline void write(float input) noexcept {
            buffer[writeIndex] = input;
            writeIndex = (writeIndex + 1) & mask;
        }

    private:
        float* buffer{ nullptr };
        int mask{ 0 };
        int writeIndex{ 0 };
        float thiranX1{ 0.0f };
        float thiranY1{ 0.0f };
    };

} // namespace FDNReverb