#include "gpu_shader_preset.hpp"
#include "gpu_texture.hpp"

#include <SDL3/SDL_log.h>
#include <fstream>
#include <sstream>
#include <algorithm>
#include <memory>

// ============================================================================
// Helpers
// ============================================================================

static std::vector<uint8_t> readFile(const std::string& path) {
    std::ifstream f(path, std::ios::binary | std::ios::ate);
    if (!f) return {};
    std::streamsize sz = f.tellg();
    f.seekg(0, std::ios::beg);
    std::vector<uint8_t> buf(static_cast<size_t>(sz));
    if (!f.read(reinterpret_cast<char*>(buf.data()), sz)) return {};
    return buf;
}

static std::string trim(const std::string& s) {
    size_t a = s.find_first_not_of(" \t\r\n");
    if (a == std::string::npos) return {};
    size_t b = s.find_last_not_of(" \t\r\n");
    return s.substr(a, b - a + 1);
}

// ============================================================================
// GpuShaderPreset
// ============================================================================

bool GpuShaderPreset::parseIni(const std::string& ini_path) {
    std::ifstream f(ini_path);
    if (!f) {
        SDL_Log("GpuShaderPreset: cannot open %s", ini_path.c_str());
        return false;
    }

    int num_passes = 0;
    std::string line;
    while (std::getline(f, line)) {
        // Strip comments
        auto comment = line.find(';');
        if (comment != std::string::npos) line = line.substr(0, comment);
        line = trim(line);
        if (line.empty()) continue;

        auto eq = line.find('=');
        if (eq == std::string::npos) continue;

        std::string key   = trim(line.substr(0, eq));
        std::string value = trim(line.substr(eq + 1));
        if (key.empty() || value.empty()) continue;

        if (key == "name") {
            name_ = value;
        } else if (key == "passes") {
            num_passes = std::stoi(value);
        } else {
            // Try to parse as float parameter
            try {
                params_[key] = std::stof(value);
            } catch (...) {
                // Non-float values stored separately (pass0_vert etc.)
            }
        }
    }

    if (num_passes <= 0) {
        SDL_Log("GpuShaderPreset: no passes defined in %s", ini_path.c_str());
        return false;
    }

    // Second pass: read per-pass file names
    f.clear();
    f.seekg(0, std::ios::beg);
    descs_.resize(num_passes);
    while (std::getline(f, line)) {
        auto comment = line.find(';');
        if (comment != std::string::npos) line = line.substr(0, comment);
        line = trim(line);
        if (line.empty()) continue;
        auto eq = line.find('=');
        if (eq == std::string::npos) continue;
        std::string key   = trim(line.substr(0, eq));
        std::string value = trim(line.substr(eq + 1));

        for (int i = 0; i < num_passes; ++i) {
            std::string vi = "pass" + std::to_string(i) + "_vert";
            std::string fi = "pass" + std::to_string(i) + "_frag";
            if (key == vi) descs_[i].vert_name = value;
            if (key == fi) descs_[i].frag_name = value;
        }
    }

    // Validate
    for (int i = 0; i < num_passes; ++i) {
        if (descs_[i].vert_name.empty() || descs_[i].frag_name.empty()) {
            SDL_Log("GpuShaderPreset: pass %d missing vert or frag in %s", i, ini_path.c_str());
            return false;
        }
    }
    return true;
}

SDL_GPUGraphicsPipeline* GpuShaderPreset::buildPassPipeline(SDL_GPUDevice* device,
                                                              const std::string& vert_spv_path,
                                                              const std::string& frag_spv_path,
                                                              SDL_GPUTextureFormat target_fmt) {
#ifdef __APPLE__
    constexpr SDL_GPUShaderFormat kFmt = SDL_GPU_SHADERFORMAT_MSL;
    constexpr const char* kEntry = "main0";
#else
    constexpr SDL_GPUShaderFormat kFmt = SDL_GPU_SHADERFORMAT_SPIRV;
    constexpr const char* kEntry = "main";
#endif

    auto vert_spv = readFile(vert_spv_path);
    auto frag_spv = readFile(frag_spv_path);
    if (vert_spv.empty()) {
        SDL_Log("GpuShaderPreset: cannot read %s", vert_spv_path.c_str());
        return nullptr;
    }
    if (frag_spv.empty()) {
        SDL_Log("GpuShaderPreset: cannot read %s", frag_spv_path.c_str());
        return nullptr;
    }
#ifdef __APPLE__
    vert_spv.push_back(0);
    frag_spv.push_back(0);
#endif

    SDL_GPUShaderCreateInfo vert_info = {};
    vert_info.code               = vert_spv.data();
    vert_info.code_size          = vert_spv.size();
    vert_info.entrypoint         = kEntry;
    vert_info.format             = kFmt;
    vert_info.stage              = SDL_GPU_SHADERSTAGE_VERTEX;
    vert_info.num_samplers       = 0;
    vert_info.num_uniform_buffers = 0;

    SDL_GPUShaderCreateInfo frag_info = {};
    frag_info.code               = frag_spv.data();
    frag_info.code_size          = frag_spv.size();
    frag_info.entrypoint         = kEntry;
    frag_info.format             = kFmt;
    frag_info.stage              = SDL_GPU_SHADERSTAGE_FRAGMENT;
    frag_info.num_samplers       = 1;
    frag_info.num_uniform_buffers = 1;

    SDL_GPUShader* vert_shader = SDL_CreateGPUShader(device, &vert_info);
    SDL_GPUShader* frag_shader = SDL_CreateGPUShader(device, &frag_info);

    if (!vert_shader || !frag_shader) {
        SDL_Log("GpuShaderPreset: shader creation failed for %s / %s: %s",
                vert_spv_path.c_str(), frag_spv_path.c_str(), SDL_GetError());
        if (vert_shader) SDL_ReleaseGPUShader(device, vert_shader);
        if (frag_shader) SDL_ReleaseGPUShader(device, frag_shader);
        return nullptr;
    }

    // Full-screen triangle: no vertex input, no blend
    SDL_GPUColorTargetBlendState no_blend = {};
    no_blend.enable_blend            = false;
    no_blend.enable_color_write_mask = false;

    SDL_GPUColorTargetDescription ct_desc = {};
    ct_desc.format      = target_fmt;
    ct_desc.blend_state = no_blend;

    SDL_GPUVertexInputState no_input = {};

    SDL_GPUGraphicsPipelineCreateInfo pipe_info = {};
    pipe_info.vertex_shader   = vert_shader;
    pipe_info.fragment_shader = frag_shader;
    pipe_info.vertex_input_state = no_input;
    pipe_info.primitive_type  = SDL_GPU_PRIMITIVETYPE_TRIANGLELIST;
    pipe_info.target_info.num_color_targets         = 1;
    pipe_info.target_info.color_target_descriptions = &ct_desc;

    SDL_GPUGraphicsPipeline* pipeline = SDL_CreateGPUGraphicsPipeline(device, &pipe_info);

    SDL_ReleaseGPUShader(device, vert_shader);
    SDL_ReleaseGPUShader(device, frag_shader);

    if (!pipeline)
        SDL_Log("GpuShaderPreset: pipeline creation failed: %s", SDL_GetError());

    return pipeline;
}

bool GpuShaderPreset::load(SDL_GPUDevice* device,
                            const std::string& dir,
                            SDL_GPUTextureFormat swapchain_fmt,
                            int w, int h) {
    dir_           = dir;
    swapchain_fmt_ = swapchain_fmt;

    // Parse ini
    if (!parseIni(dir + "/preset.ini"))
        return false;

    int n = static_cast<int>(descs_.size());
    passes_.resize(n);

    // Intermediate render target format (signed float to handle NTSC signal range)
    SDL_GPUTextureFormat inter_fmt = SDL_GPU_TEXTUREFORMAT_R16G16B16A16_FLOAT;

    for (int i = 0; i < n; ++i) {
        bool is_last = (i == n - 1);
        SDL_GPUTextureFormat target_fmt = is_last ? swapchain_fmt : inter_fmt;

#ifdef __APPLE__
        const char* ext = ".spv.msl";
#else
        const char* ext = ".spv";
#endif
        std::string vert_spv = dir + "/" + descs_[i].vert_name + ext;
        std::string frag_spv = dir + "/" + descs_[i].frag_name + ext;

        passes_[i].pipeline = buildPassPipeline(device, vert_spv, frag_spv, target_fmt);
        if (!passes_[i].pipeline) {
            SDL_Log("GpuShaderPreset: failed to build pipeline for pass %d", i);
            return false;
        }

        if (!is_last) {
            // Create intermediate render target
            auto tex = std::make_unique<GpuTexture>();
            if (!tex->createRenderTarget(device, w, h, inter_fmt)) {
                SDL_Log("GpuShaderPreset: failed to create intermediate target for pass %d", i);
                return false;
            }
            passes_[i].target = tex.get();
            targets_.push_back(std::move(tex));
        }
        // Last pass: target = null (caller binds swapchain)
    }

    SDL_Log("GpuShaderPreset: loaded '%s' (%d passes)", name_.c_str(), n);
    return true;
}

void GpuShaderPreset::recreateTargets(SDL_GPUDevice* device, int w, int h) {
    SDL_GPUTextureFormat inter_fmt = SDL_GPU_TEXTUREFORMAT_R16G16B16A16_FLOAT;
    for (auto& tex : targets_) {
        tex->destroy(device);
        tex->createRenderTarget(device, w, h, inter_fmt);
    }
}

void GpuShaderPreset::destroy(SDL_GPUDevice* device) {
    for (auto& pass : passes_) {
        if (pass.pipeline) {
            SDL_ReleaseGPUGraphicsPipeline(device, pass.pipeline);
            pass.pipeline = nullptr;
        }
    }
    for (auto& tex : targets_) {
        if (tex) tex->destroy(device);
    }
    targets_.clear();
    passes_.clear();
    descs_.clear();
    params_.clear();
}

float GpuShaderPreset::param(const std::string& key, float default_val) const {
    auto it = params_.find(key);
    return (it != params_.end()) ? it->second : default_val;
}