#include "core/image.hpp"

#include <gdk-pixbuf/gdk-pixbuf.h>

#include <algorithm>
#include <atomic>
#include <cctype>
#include <cstring>
#include <memory>
#include <mutex>
#include <stdexcept>
#include <thread>

namespace effector {
namespace {

// The plain C pixbuf API is used on purpose: it needs no gtkmm wrapper
// registration and no display, so the core renders headlessly.
using PixbufPtr = std::unique_ptr<GdkPixbuf, decltype(&g_object_unref)>;

PixbufPtr wrap(GdkPixbuf* pixbuf) { return PixbufPtr(pixbuf, &g_object_unref); }

std::string takeError(GError* error) {
    std::string message = error != nullptr && error->message != nullptr ? error->message
                                                                       : "unknown error";
    if (error != nullptr) g_error_free(error);
    return message;
}

std::string lowerExtension(const std::string& path) {
    const auto dot = path.find_last_of('.');
    if (dot == std::string::npos) return {};
    std::string ext = path.substr(dot + 1);
    std::transform(ext.begin(), ext.end(), ext.begin(),
                   [](unsigned char c) { return static_cast<char>(std::tolower(c)); });
    return ext;
}

}  // namespace

Image::Image(int width, int height)
    : width_(width), height_(height),
      pixels_(static_cast<std::size_t>(std::max(width, 0)) * 3 *
              static_cast<std::size_t>(std::max(height, 0))) {}

Image Image::load(const std::string& path, std::vector<std::uint8_t>* alphaOut) {
    GError* error = nullptr;
    PixbufPtr pixbuf = wrap(gdk_pixbuf_new_from_file(path.c_str(), &error));
    if (!pixbuf) throw std::runtime_error(takeError(error));

    if (GdkPixbuf* oriented = gdk_pixbuf_apply_embedded_orientation(pixbuf.get())) {
        pixbuf = wrap(oriented);
    }

    const int width = gdk_pixbuf_get_width(pixbuf.get());
    const int height = gdk_pixbuf_get_height(pixbuf.get());
    const int channels = gdk_pixbuf_get_n_channels(pixbuf.get());
    const int srcStride = gdk_pixbuf_get_rowstride(pixbuf.get());
    const guint8* src = gdk_pixbuf_get_pixels(pixbuf.get());

    Image image(width, height);
    if (alphaOut != nullptr) {
        alphaOut->assign(static_cast<std::size_t>(std::max(width, 0)) *
                             static_cast<std::size_t>(std::max(height, 0)),
                         255);
    }
    for (int y = 0; y < height; ++y) {
        const guint8* in = src + static_cast<std::size_t>(y) * srcStride;
        std::uint8_t* out = image.row(y);
        if (channels == 3) {
            std::memcpy(out, in, static_cast<std::size_t>(width) * 3);
        } else if (alphaOut != nullptr) {
            // Kept apart rather than flattened: the caller does the compositing.
            std::uint8_t* alpha = alphaOut->data() + static_cast<std::size_t>(y) * width;
            for (int x = 0; x < width; ++x) {
                for (int c = 0; c < 3; ++c) out[x * 3 + c] = in[x * channels + c];
                alpha[x] = in[x * channels + 3];
            }
        } else {
            // Flatten alpha onto white, matching what a viewer would show.
            for (int x = 0; x < width; ++x) {
                const int a = in[x * channels + 3];
                for (int c = 0; c < 3; ++c) {
                    const int value = in[x * channels + c] * a + 255 * (255 - a);
                    out[x * 3 + c] = static_cast<std::uint8_t>((value + 127) / 255);
                }
            }
        }
    }
    image.setPath(path);
    return image;
}

void Image::save(const std::string& path) const {
    if (empty()) throw std::runtime_error("nothing to save");

    PixbufPtr pixbuf = wrap(gdk_pixbuf_new(GDK_COLORSPACE_RGB, FALSE, 8, width_, height_));
    if (!pixbuf) throw std::runtime_error("out of memory");

    const int dstStride = gdk_pixbuf_get_rowstride(pixbuf.get());
    guint8* dst = gdk_pixbuf_get_pixels(pixbuf.get());
    for (int y = 0; y < height_; ++y) {
        std::memcpy(dst + static_cast<std::size_t>(y) * dstStride, row(y), stride());
    }

    const std::string ext = lowerExtension(path);
    std::string type = "png";
    if (ext == "jpg" || ext == "jpeg") type = "jpeg";
    else if (ext == "webp") type = "webp";
    else if (ext == "tif" || ext == "tiff") type = "tiff";
    else if (ext == "bmp") type = "bmp";

    GError* error = nullptr;
    const gboolean saved =
        type == "jpeg"
            ? gdk_pixbuf_save(pixbuf.get(), path.c_str(), type.c_str(), &error, "quality", "95",
                              nullptr)
            : gdk_pixbuf_save(pixbuf.get(), path.c_str(), type.c_str(), &error, nullptr);
    if (saved == FALSE) throw std::runtime_error(takeError(error));
}

std::vector<Image> loadImages(const std::vector<std::string>& paths,
                              std::vector<LoadFailure>& failures,
                              const LoadProgress* progress) {
    // Decoding is the dominant cost of a headless run, and the files are
    // independent, so they are decoded in parallel. Threads take the next path
    // from a shared cursor rather than a fixed slice, so one huge file cannot
    // stall the files behind it and progress arrives at a steady rate. Slots
    // keep the result in the order the paths were given.
    std::vector<Image> slots(paths.size());
    std::vector<std::string> errors(paths.size());
    std::atomic<std::size_t> cursor{0};
    std::atomic<std::size_t> done{0};

    unsigned hardware = std::thread::hardware_concurrency();
    if (hardware == 0) hardware = 1;
    const std::size_t workers = std::min<std::size_t>(hardware, std::max<std::size_t>(paths.size(), 1));

    const auto decode = [&] {
        for (;;) {
            if (progress != nullptr && progress->cancel != nullptr &&
                progress->cancel->load(std::memory_order_relaxed)) {
                return;
            }
            const std::size_t i = cursor.fetch_add(1, std::memory_order_relaxed);
            if (i >= paths.size()) return;
            try {
                slots[i] = Image::load(paths[i]);
            } catch (const std::exception& error) {
                errors[i] = error.what();
            }
            if (progress != nullptr && progress->onFile) {
                progress->onFile(done.fetch_add(1, std::memory_order_relaxed) + 1, paths.size(),
                                 paths[i]);
            }
        }
    };

    std::vector<std::thread> threads;
    threads.reserve(workers > 0 ? workers - 1 : 0);
    for (std::size_t w = 1; w < workers; ++w) threads.emplace_back(decode);
    decode();
    for (auto& thread : threads) thread.join();

    std::vector<Image> images;
    images.reserve(paths.size());
    for (std::size_t i = 0; i < paths.size(); ++i) {
        if (!errors[i].empty()) {
            failures.push_back({paths[i], errors[i]});
        } else if (!slots[i].empty()) {
            images.push_back(std::move(slots[i]));
        }
    }
    return images;
}

}  // namespace effector