// =============================================
//  Aster: model_render.cpp
//  Copyright (c) 2020-2024 Anish Bhobe
// =============================================

#include "buffer.h"
#include "constants.h"
#include "context.h"
#include "device.h"
#include "global.h"
#include "image.h"
#include "physical_device.h"
#include "pipeline.h"
#include "swapchain.h"
#include "window.h"

#include "frame.h"
#include "helpers.h"

#include "model_loader.h"
#include "pipeline_utils.h"
#include "render_resource_manager.h"

#include <EASTL/array.h>
#include <stb_image.h>
#include <tiny_gltf.h>
#include <filesystem>

constexpr u32 MAX_FRAMES_IN_FLIGHT = 3;
constexpr auto MODEL_FILE = "model/OrientationTest.glb";

struct ImageFile
{
    u8 *m_Data = nullptr;
    u32 m_Width = 0;
    u32 m_Height = 0;
    u32 m_NumChannels = 0;
    bool m_Constant = false;

    bool Load(cstr fileName);
    bool Load(vec4 color);
    [[nodiscard]] usize GetSize() const;
    ~ImageFile();
};

struct Camera
{
    mat4 m_Model;
    mat4 m_View;
    mat4 m_Perspective;
};

int
main(int, char **)
{
    MIN_LOG_LEVEL(Logger::LogType::eInfo);

    Context context = {"ModelRender [WIP]", VERSION};
    Window window = {"ModelRender [WIP] (Aster)", &context, {640, 480}};

    PhysicalDevices physicalDevices = {&window, &context};
    PhysicalDevice deviceToUse = FindSuitableDevice(physicalDevices);

    INFO("Using {} as the primary device.", deviceToUse.m_DeviceProperties.deviceName.data());

    Features enabledDeviceFeatures = {
        .m_Vulkan10Features = {.samplerAnisotropy = true},
        .m_Vulkan12Features =
            {
                .descriptorIndexing = true,
                .shaderSampledImageArrayNonUniformIndexing = true,
                .shaderStorageBufferArrayNonUniformIndexing = true,
                .shaderStorageImageArrayNonUniformIndexing = true,
                .descriptorBindingUniformBufferUpdateAfterBind = true, // Not related to Bindless
                .descriptorBindingSampledImageUpdateAfterBind = true,
                .descriptorBindingStorageImageUpdateAfterBind = true,
                .descriptorBindingStorageBufferUpdateAfterBind = true,
                .descriptorBindingPartiallyBound = true,
                .runtimeDescriptorArray = true,
            },
        .m_Vulkan13Features = {.dynamicRendering = true},
    };

    QueueAllocation queueAllocation = FindAppropriateQueueAllocation(&deviceToUse);
    Device device = {&context, &deviceToUse, &enabledDeviceFeatures, {queueAllocation}, "Primary Device"};
    vk::Queue commandQueue = device.GetQueue(queueAllocation.m_Family, 0);
    Swapchain swapchain = {&window, &device, "Primary Chain"};
    GpuResourceManager resourceManager = {&device, 1000};

    ModelLoader modelLoader = {&resourceManager, commandQueue, queueAllocation.m_Family, queueAllocation.m_Family};

    auto model = modelLoader.LoadModel(MODEL_FILE);

    Pipeline pipeline = CreatePipeline(&device, &swapchain, &resourceManager);

    Camera camera = {
        .m_Model = {1.0f},
        .m_View = glm::lookAt(vec3(0.0f, 12.0f, 12.0f), vec3(0.0f), vec3(0.0f, 1.0f, 0.0f)),
        .m_Perspective = glm::perspective(
            70_deg, Cast<f32>(swapchain.m_Extent.width) / Cast<f32>(swapchain.m_Extent.height), 0.1f, 100.0f),
    };

    vk::DescriptorPool descriptorPool;
    vk::DescriptorSet descriptorSet;

    {
        vk::DescriptorSetLayout descriptorSetLayout = pipeline.m_SetLayouts[1];
        eastl::array poolSizes = {
            vk::DescriptorPoolSize{
                .type = vk::DescriptorType::eUniformBuffer,
                .descriptorCount = 1,
            },
        };
        vk::DescriptorPoolCreateInfo descriptorPoolCreateInfo = {
            .maxSets = 1, .poolSizeCount = Cast<u32>(poolSizes.size()), .pPoolSizes = poolSizes.data()};
        AbortIfFailed(device.m_Device.createDescriptorPool(&descriptorPoolCreateInfo, nullptr, &descriptorPool));

        vk::DescriptorSetAllocateInfo descriptorSetAllocateInfo = {
            .descriptorPool = descriptorPool,
            .descriptorSetCount = 1,
            .pSetLayouts = &descriptorSetLayout,
        };
        AbortIfFailed(device.m_Device.allocateDescriptorSets(&descriptorSetAllocateInfo, &descriptorSet));
    }

    UniformBuffer ubo;
    ubo.Init(&device, sizeof camera, "Camera UBO");
    ubo.Write(&device, 0, sizeof camera, &camera);
    vk::DescriptorBufferInfo descriptorBufferInfo = {
        .buffer = ubo.m_Buffer,
        .offset = 0,
        .range = ubo.GetSize(),
    };
    eastl::array writeDescriptors = {
        vk::WriteDescriptorSet{
            .dstSet = descriptorSet,
            .dstBinding = 0,
            .dstArrayElement = 0,
            .descriptorCount = 1,
            .descriptorType = vk::DescriptorType::eUniformBuffer,
            .pBufferInfo = &descriptorBufferInfo,
        },
    };
    device.m_Device.updateDescriptorSets(Cast<u32>(writeDescriptors.size()), writeDescriptors.data(), 0, nullptr);

    resourceManager.Update();

    // Persistent variables
    vk::Viewport viewport = {
        .x = 0,
        .y = Cast<f32>(swapchain.m_Extent.height),
        .width = Cast<f32>(swapchain.m_Extent.width),
        .height = -Cast<f32>(swapchain.m_Extent.height),
        .minDepth = 0.0,
        .maxDepth = 1.0,
    };

    vk::Rect2D scissor = {
        .offset = {0, 0},
        .extent = swapchain.m_Extent,
    };

    auto fnResizeViewportScissor = [&viewport, &scissor](vk::Extent2D extent) {
        viewport.y = Cast<f32>(extent.height);
        viewport.width = Cast<f32>(extent.width);
        viewport.height = -Cast<f32>(extent.height);
        scissor.extent = extent;
    };
    swapchain.RegisterResizeCallback(fnResizeViewportScissor);

    vk::ImageSubresourceRange subresourceRange = {
        .aspectMask = vk::ImageAspectFlagBits::eColor,
        .baseMipLevel = 0,
        .levelCount = 1,
        .baseArrayLayer = 0,
        .layerCount = 1,
    };
    vk::ImageMemoryBarrier topOfThePipeBarrier = {
        .oldLayout = vk::ImageLayout::eUndefined,
        .newLayout = vk::ImageLayout::eColorAttachmentOptimal,
        .srcQueueFamilyIndex = queueAllocation.m_Family,
        .dstQueueFamilyIndex = queueAllocation.m_Family,
        .subresourceRange = subresourceRange,
    };
    vk::ImageMemoryBarrier renderToPresentBarrier = {
        .oldLayout = vk::ImageLayout::eColorAttachmentOptimal,
        .newLayout = vk::ImageLayout::ePresentSrcKHR,
        .srcQueueFamilyIndex = queueAllocation.m_Family,
        .dstQueueFamilyIndex = queueAllocation.m_Family,
        .subresourceRange = subresourceRange,
    };

    FrameManager frameManager = {&device, queueAllocation.m_Family, MAX_FRAMES_IN_FLIGHT};
    eastl::fixed_vector<DepthImage, MAX_FRAMES_IN_FLIGHT> depthImages(frameManager.m_FramesInFlight);
    {
        u32 depthImageIdx = 0;
        for (auto &depthImage : depthImages)
        {
            auto name = fmt::format("Depth {}", depthImageIdx++);
            depthImage.Init(&device, swapchain.m_Extent, name.c_str());
        }
    }

    auto fnRecreateDepthBuffers = [&device, &depthImages](vk::Extent2D extent) {
        u32 depthImageIdx = 0;
        for (auto &depthImage : depthImages)
        {
            auto name = fmt::format("Depth {}", depthImageIdx++);
            depthImage.Destroy(&device);
            depthImage.Init(&device, extent, name.c_str());
        }
    };
    swapchain.RegisterResizeCallback(fnRecreateDepthBuffers);

    Time::Init();

    struct ModelData
    {
        BufferHandle m_VertexBuffer;
        BufferHandle m_Color;
        BufferHandle m_UvBuffer;
        BufferHandle m_Materials;
        BufferHandle m_Nodes;
    };

    ModelData modelData = {
        .m_VertexBuffer = model.m_VertexPositionHandle,
        .m_Color = model.m_VertexColorHandle,
        .m_UvBuffer = model.m_TexCoord0Handle,
        .m_Materials = model.m_MaterialsHandle,
        .m_Nodes = model.m_NodeHandle,
    };

    INFO("Starting loop");
    while (window.Poll())
    {
        Time::Update();

        camera.m_Model *= rotate(mat4{1.0f}, Cast<f32>(45.0_deg * Time::m_Delta), vec3(0.0f, 1.0f, 0.0f));
        ubo.Write(&device, 0, sizeof camera, &camera);

        Frame *currentFrame = frameManager.GetNextFrame(&swapchain, &window);

        u32 imageIndex = currentFrame->m_ImageIdx;
        vk::ImageView currentImageView = swapchain.m_ImageViews[imageIndex];
        vk::Image currentImage = swapchain.m_Images[imageIndex];
        vk::CommandBuffer cmd = currentFrame->m_CommandBuffer;
        vk::ImageView currentDepthImageView = depthImages[currentFrame->m_FrameIdx].m_View;

        topOfThePipeBarrier.image = currentImage;
        renderToPresentBarrier.image = currentImage;

        vk::CommandBufferBeginInfo beginInfo = {.flags = vk::CommandBufferUsageFlagBits::eOneTimeSubmit};
        AbortIfFailed(cmd.begin(&beginInfo));

        cmd.pipelineBarrier(vk::PipelineStageFlagBits::eTopOfPipe, vk::PipelineStageFlagBits::eColorAttachmentOutput,
                            {}, 0, nullptr, 0, nullptr, 1, &topOfThePipeBarrier);

        // Render
        eastl::array attachmentInfos = {
            vk::RenderingAttachmentInfo{
                .imageView = currentImageView,
                .imageLayout = vk::ImageLayout::eColorAttachmentOptimal,
                .resolveMode = vk::ResolveModeFlagBits::eNone,
                .loadOp = vk::AttachmentLoadOp::eClear,
                .storeOp = vk::AttachmentStoreOp::eStore,
                .clearValue = vk::ClearColorValue{0.0f, 0.0f, 0.0f, 1.0f},
            },
        };

        vk::RenderingAttachmentInfo depthAttachment = {
            .imageView = currentDepthImageView,
            .imageLayout = vk::ImageLayout::eDepthAttachmentOptimal,
            .resolveMode = vk::ResolveModeFlagBits::eNone,
            .loadOp = vk::AttachmentLoadOp::eClear,
            .storeOp = vk::AttachmentStoreOp::eDontCare,
            .clearValue = vk::ClearDepthStencilValue{.depth = 1.0f, .stencil = 0},
        };

        vk::RenderingInfo renderingInfo = {
            .renderArea = {.extent = swapchain.m_Extent},
            .layerCount = 1,
            .colorAttachmentCount = Cast<u32>(attachmentInfos.size()),
            .pColorAttachments = attachmentInfos.data(),
            .pDepthAttachment = &depthAttachment,
        };

        cmd.beginRendering(&renderingInfo);

        cmd.setViewport(0, 1, &viewport);
        cmd.setScissor(0, 1, &scissor);
        cmd.bindPipeline(vk::PipelineBindPoint::eGraphics, pipeline.m_Pipeline);
        cmd.bindDescriptorSets(vk::PipelineBindPoint::eGraphics, pipeline.m_Layout, 0, 1,
                               &resourceManager.m_DescriptorSet, 0, nullptr);
        cmd.bindDescriptorSets(vk::PipelineBindPoint::eGraphics, pipeline.m_Layout, 1, 1, &descriptorSet, 0, nullptr);

        cmd.bindIndexBuffer(model.m_IndexBuffer.m_Buffer, 0, vk::IndexType::eUint32);
        cmd.pushConstants(pipeline.m_Layout, vk::ShaderStageFlagBits::eAll, 0, sizeof modelData, &modelData);

        for (auto &prim : model.m_MeshPrimitives)
        {
            cmd.pushConstants(pipeline.m_Layout, vk::ShaderStageFlagBits::eAll, sizeof modelData,
                              sizeof prim.m_MaterialIdx, &prim.m_MaterialIdx);
            cmd.pushConstants(pipeline.m_Layout, vk::ShaderStageFlagBits::eAll, sizeof modelData + sizeof prim.m_MaterialIdx,
                              sizeof prim.m_TransformIdx, &prim.m_TransformIdx);
            cmd.drawIndexed(prim.m_IndexCount, 1, prim.m_FirstIndex, Cast<i32>(prim.m_VertexOffset), 0);
        }

        cmd.endRendering();

        cmd.pipelineBarrier(vk::PipelineStageFlagBits::eColorAttachmentOutput, vk::PipelineStageFlagBits::eBottomOfPipe,
                            {}, 0, nullptr, 0, nullptr, 1, &renderToPresentBarrier);

        AbortIfFailed(cmd.end());

        vk::PipelineStageFlags waitDstStage = vk::PipelineStageFlagBits::eColorAttachmentOutput;
        vk::SubmitInfo submitInfo = {
            .waitSemaphoreCount = 1,
            .pWaitSemaphores = &currentFrame->m_ImageAcquireSem,
            .pWaitDstStageMask = &waitDstStage,
            .commandBufferCount = 1,
            .pCommandBuffers = &cmd,
            .signalSemaphoreCount = 1,
            .pSignalSemaphores = &currentFrame->m_RenderFinishSem,
        };
        AbortIfFailed(commandQueue.submit(1, &submitInfo, currentFrame->m_FrameAvailableFence));

        currentFrame->Present(commandQueue, &swapchain, &window);
    }

    AbortIfFailed(device.m_Device.waitIdle());

    for (auto &depthImage : depthImages)
    {
        depthImage.Destroy(&device);
    }
    ubo.Destroy(&device);
    device.m_Device.destroy(descriptorPool, nullptr);

    return 0;
}

bool
ImageFile::Load(vec4 color)
{
    constexpr usize size = 512llu * 512llu * 4llu;
    u8 *pData = new u8[size];

    const vec4 color255 = 255.999f * color;
    const glm::vec<4, u8> color8 = color255;

    for (usize i = 0; i < size; i += 4)
    {
        memcpy(pData + i, &color8, sizeof color8);
    }

    m_Data = pData;
    m_Constant = true;
    m_Height = 512;
    m_Width = 512;
    m_NumChannels = 4;

    return true;
}

bool
ImageFile::Load(cstr fileName)
{
    int width, height, nrChannels;
    m_Data = stbi_load(fileName, &width, &height, &nrChannels, 4);
    ERROR_IF(!m_Data, "Could not load {}", fileName);

    if (!m_Data)
    {
        return false;
    }

    m_Width = width;
    m_Height = height;
    m_NumChannels = 4;
    m_Constant = false;

    return true;
}

usize
ImageFile::GetSize() const
{
    return Cast<usize>(m_Width) * m_Height * m_NumChannels;
}

ImageFile::~ImageFile()
{
    if (m_Constant)
    {
        delete[] Cast<u8 *>(m_Data);
    }
    else
    {
        stbi_image_free(m_Data);
    }
    m_Data = nullptr;
}