VideoStreamContainer.hpp

Go to the documentation of this file.

#pragma once
 
#include "MayaFlux/Kakshya/StreamContainer.hpp"
 
#include "MayaFlux/Kakshya/Utils/PixelStorage.hpp"
 
#include "MayaFlux/Transitive/Memory/RingBuffer.hpp"
#include "MayaFlux/Transitive/Memory/SeqLock.hpp"
 
namespace MayaFlux::Registry::Service {
struct IOService;
}
 
namespace MayaFlux::Kakshya {
 
/**
 * @class VideoStreamContainer
 * @brief Concrete base implementation for streaming video containers.
 *
 * VideoStreamContainer provides a complete, concrete implementation of all
 * StreamContainer functionality for decoded video frame data. It serves as:
 * 1. A standalone streaming container for real-time video processing
 * 2. A base class for specialized containers like VideoFileContainer
 *
 * Data is stored in the DataVariant alternative matching the container's
 * ImageFormat: uint8_t for 8-bit formats, uint16_t for 16-bit, float for
 * 32-bit. Each frame is width * height * bytes_per_pixel bytes. All frames
 * are stored contiguously in a single DataVariant.
 *
 * Dimensions follow the format's modality:
 *   dims[0] → TIME (frame count)
 *   dims[1] → SPATIAL_Y (height)
 *   dims[2] → SPATIAL_X (width)
 *   dims[3] → CHANNEL, or DEPTH for range formats (component count)
 *
 * Reader model follows WindowContainer's pattern: a simple atomic reader
 * count rather than per-dimension/per-channel tracking. Video frames are
 * atomic spatial units — channel-level access is a processor concern,
 * not a container concern.
 *
 * Uses virtual inheritance to support diamond inheritance when used as a
 * base for FileContainer-derived classes.
 */
class MAYAFLUX_API VideoStreamContainer : public virtual StreamContainer {
public:
    /**
     * @brief Construct a VideoStreamContainer with specified parameters.
     * @param width       Frame width in pixels.
     * @param height      Frame height in pixels.
     * @param format      Pixel format; determines channel count, bytes per
     *                    pixel, the DataVariant element type, and whether
     *                    the structure is VIDEO_COLOR or VIDEO_DEPTH.
     * @param frame_rate  Temporal rate in frames per second.
     */
    VideoStreamContainer(uint32_t width = 0,
        uint32_t height = 0,
        Portal::Graphics::ImageFormat format = Portal::Graphics::ImageFormat::RGBA8,
        double frame_rate = 0.0);
 
    ~VideoStreamContainer() override = default;
 
    // =========================================================================
    // NDDimensionalContainer
    // =========================================================================
 
    [[nodiscard]] std::vector<DataDimension> get_dimensions() const override;
    [[nodiscard]] uint64_t get_total_elements() const override;
    [[nodiscard]] MemoryLayout get_memory_layout() const override { return m_structure.memory_layout; }
    void set_memory_layout(MemoryLayout layout) override;
 
    [[nodiscard]] uint64_t get_frame_size() const override;
    [[nodiscard]] uint64_t get_num_frames() const override;
 
    std::vector<DataVariant> get_region_data(const Region& region) const override;
    void set_region_data(const Region& region, const std::vector<DataVariant>& data) override;
 
    std::vector<DataVariant> get_region_group_data(const RegionGroup& group) const override;
    std::vector<DataVariant> get_segments_data(const std::vector<RegionSegment>& segment) const override;
 
    [[nodiscard]] std::type_index value_element_type() const override;
 
    [[nodiscard]] uint64_t coordinates_to_linear_index(const std::vector<uint64_t>& coordinates) const override;
    [[nodiscard]] std::vector<uint64_t> linear_index_to_coordinates(uint64_t linear_index) const override;
 
    void clear() override;
 
    [[nodiscard]] const void* get_raw_data() const override;
    [[nodiscard]] bool has_data() const override;
 
    ContainerDataStructure& get_structure() override { return m_structure; }
    const ContainerDataStructure& get_structure() const override { return m_structure; }
    void set_structure(ContainerDataStructure structure) override { m_structure = structure; }
 
    // =========================================================================
    // Ring buffer streaming API
    // =========================================================================
 
    /**
     * @brief Allocate m_data[0] as a ring of ring_capacity frames.
     *
     * Switches the container from flat mode to ring mode. m_data[0] is
     * resized to ring_capacity x frame_byte_size. m_num_frames is set to
     * total_frames so processors see the full temporal extent. Pixel data
     * is indexed by frame_index % ring_capacity.
     *
     * @param total_frames  Total frames in the source (file, stream, etc).
     * @param ring_capacity Number of frame slots (must be power of 2).
     * @param width         Frame width in pixels.
     * @param height        Frame height in pixels.
     * @param format        Pixel format for the ring's frames.
     * @param frame_rate    Frame rate in fps.
     * @param refill_threshold Frames of look-ahead below which refill callback fires.
     * @param reader_id The current class ID registered at the stream/file-read source
     */
    void setup_ring(uint64_t total_frames,
        uint32_t ring_capacity,
        uint32_t width,
        uint32_t height,
        Portal::Graphics::ImageFormat format,
        double frame_rate,
        uint32_t refill_threshold,
        uint64_t reader_id = 0);
 
    /**
     * @brief Mutable pointer into m_data[0] for the decode thread to write into.
     * @param frame_index Absolute frame index; mapped to slot via modulo.
     * @return Pointer into the pixel vector, or nullptr if not in ring mode.
     */
    [[nodiscard]] uint8_t* mutable_slot_ptr(uint64_t frame_index);
 
    /**
     * @brief Publish a decoded frame. Sets validity, pushes to ready queue,
     *        notifies any thread blocked in get_frame_pixels().
     * @param frame_index Absolute frame index just written.
     */
    void commit_frame(uint64_t frame_index);
 
    /**
     * @brief Invalidate all ring slots. Called before seek.
     */
    void invalidate_ring();
 
    /**
     * @brief Check if a frame is currently valid in the ring.
     * @param frame_index Absolute frame index.
     */
    [[nodiscard]] bool is_frame_available(uint64_t frame_index) const;
 
    /**
     * @brief True if the container is operating in ring mode.
     */
    [[nodiscard]] bool is_ring_mode() const { return m_ring_capacity > 0; }
 
    [[nodiscard]] uint32_t get_ring_capacity() const { return m_ring_capacity; }
    [[nodiscard]] uint64_t get_total_source_frames() const { return m_total_source_frames; }
 
    /**
     * @brief Set the number of frames below which the refill callback fires.
     *        Called by the reader before or immediately after setup_ring().
     * @param threshold Frames of look-ahead below which notification fires.
     */
    void set_refill_threshold(uint32_t threshold)
    {
        m_refill_threshold = threshold;
    }
 
    /**
     * @brief Advance the container's view of how many frames have been decoded.
     *        Called by the decode thread (via VideoFileReader) after commit_frame().
     *        Monotonically increasing; never decremented (seek resets via setup_ring).
     * @param frame_index The highest frame index just committed.
     */
    void advance_cache_head(uint64_t frame_index)
    {
        uint64_t prev = m_cache_head.load(std::memory_order_relaxed);
        while (frame_index > prev
            && !m_cache_head.compare_exchange_weak(prev, frame_index,
                std::memory_order_release, std::memory_order_relaxed)) { }
    }
 
    /**
     * @brief Total frame count known at construction / setup_ring() time.
     *        Non-zero even before any frames are decoded.
     */
    [[nodiscard]] uint64_t get_cache_head() const
    {
        return m_cache_head.load(std::memory_order_acquire);
    }
 
    /**
     * @brief Processed frame at @p frame_index as a normalised float span.
     *
     * Valid after FrameAccessProcessor has written processed_data.
     * uint8_t source values are divided by 255.0f. float source is
     * zero-copy. Returns empty span if frame_index is out of range or
     * the variant holds a non-pixel type.
     *
     * The cache covers the last requested frame_index only. A call with
     * a different index invalidates and recomputes.
     *
     * @param frame_index Zero-based index into processed_data. Defaults to 0.
     * @return Normalised float span, w * h * channels elements.
     */
    [[nodiscard]] std::span<const float> processed_frame_as_float(
        uint64_t frame_index = 0) const;
 
    // =========================================================================
    // RegionGroup management
    // =========================================================================
 
    void add_region_group(const RegionGroup& group) override;
    RegionGroup get_region_group(const std::string& name) const override;
    std::unordered_map<std::string, RegionGroup> get_all_region_groups() const override;
    void remove_region_group(const std::string& name) override;
 
    bool is_region_loaded(const Region& region) const override;
    void load_region(const Region& region) override;
    void unload_region(const Region& region) override;
 
    // =========================================================================
    // Read position and looping
    // =========================================================================
 
    void set_read_position(const std::vector<uint64_t>& position) override;
    void update_read_position_for_channel(size_t channel, uint64_t frame) override;
    [[nodiscard]] const std::vector<uint64_t>& get_read_position() const override;
    void advance_read_position(const std::vector<uint64_t>& frames) override;
    [[nodiscard]] bool is_at_end() const override;
    void reset_read_position() override;
 
    [[nodiscard]] uint64_t get_temporal_rate() const override;
    [[nodiscard]] uint64_t time_to_position(double time) const override;
    [[nodiscard]] double position_to_time(uint64_t position) const override;
 
    void set_looping(bool enable) override;
    [[nodiscard]] bool is_looping() const override { return m_looping_enabled; }
    void set_loop_region(const Region& region) override;
    [[nodiscard]] Region get_loop_region() const override;
 
    [[nodiscard]] bool is_ready() const override;
    [[nodiscard]] std::vector<uint64_t> get_remaining_frames() const override;
    uint64_t read_sequential(std::span<double> output, uint64_t count) override;
    uint64_t peek_sequential(std::span<double> output, uint64_t count, uint64_t offset) const override;
 
    // =========================================================================
    // Processing state
    // =========================================================================
 
    [[nodiscard]] ProcessingState get_processing_state() const override { return m_processing_state.load(); }
    void update_processing_state(ProcessingState new_state) override;
 
    void register_state_change_callback(
        std::function<void(const std::shared_ptr<SignalSourceContainer>&, ProcessingState)> callback) override;
    void unregister_state_change_callback() override;
 
    [[nodiscard]] bool is_ready_for_processing() const override;
    void mark_ready_for_processing(bool ready) override;
 
    void create_default_processor() override;
    void process_default() override;
    void set_default_processor(const std::shared_ptr<DataProcessor>& processor) override;
    [[nodiscard]] std::shared_ptr<DataProcessor> get_default_processor() const override;
 
    std::shared_ptr<DataProcessingChain> get_processing_chain() override;
    void set_processing_chain(const std::shared_ptr<DataProcessingChain>& chain) override { m_processing_chain = chain; }
 
    // =========================================================================
    // Reader tracking (WindowContainer-style atomic counting)
    // =========================================================================
 
    uint32_t register_dimension_reader(uint32_t dimension_index) override;
    void unregister_dimension_reader(uint32_t dimension_index) override;
    [[nodiscard]] bool has_active_readers() const override;
    void mark_dimension_consumed(uint32_t dimension_index, uint32_t reader_id) override;
    [[nodiscard]] bool all_dimensions_consumed() const override;
 
    // =========================================================================
    // Processing token
    // =========================================================================
 
    void reset_processing_token() override { m_processing_token_channel.store(-1); }
 
    bool try_acquire_processing_token(int channel) override
    {
        int expected = -1;
        return m_processing_token_channel.compare_exchange_strong(expected, channel);
    }
 
    [[nodiscard]] bool has_processing_token(int channel) const override
    {
        return m_processing_token_channel.load() == channel;
    }
 
    void invalidate_float_frame_cache(uint32_t slot_index = 0);
 
    // =========================================================================
    // Data access
    // =========================================================================
 
    const std::vector<DataVariant>& get_data() override { return m_data; }
 
    DataAccess channel_data(size_t channel) override;
    std::vector<DataAccess> all_channel_data() override;
 
    std::vector<DataVariant>& get_processed_data() override { return m_processed_data; }
    const std::vector<DataVariant>& get_processed_data() const override { return m_processed_data; }
 
    void mark_buffers_for_processing(bool) override { }
    void mark_buffers_for_removal() override { }
 
    // =========================================================================
    // Video-specific accessors
    // =========================================================================
 
    [[nodiscard]] uint32_t get_width() const { return m_width; }
    [[nodiscard]] uint32_t get_height() const { return m_height; }
    [[nodiscard]] double get_frame_rate() const { return m_frame_rate; }
 
    /** @brief Pixel format governing storage type, channel count, and modality. */
    [[nodiscard]] Portal::Graphics::ImageFormat get_format() const { return m_format; }
 
    /** @brief Component channels per pixel. Not bytes per pixel. */
    [[nodiscard]] uint32_t get_channels() const { return m_channels; }
 
    /** @brief Bytes occupied by one pixel across all channels. */
    [[nodiscard]] size_t get_bytes_per_pixel() const { return m_bpp; }
 
    /** @brief Elements in one frame: width * height * channels. */
    [[nodiscard]] size_t get_frame_element_count() const;
 
    /**
     * @brief Get raw pixel data for a single frame as a byte span.
     * @param frame_index Zero-based frame index.
     * @return Span of pixel bytes for the frame, empty if out of range.
     */
    [[nodiscard]] std::span<const uint8_t> get_frame_pixels(uint64_t frame_index) const;
 
    /**
     * @brief Total byte size of one frame: width * height * bytes_per_pixel.
     *
     * Not width * height * channels. Those agree only for formats with one
     * byte per component. Use get_frame_element_count() for element counts.
     */
    [[nodiscard]] size_t get_frame_byte_size() const;
 
protected:
    void setup_dimensions();
    void notify_state_change(ProcessingState new_state);
 
    uint32_t m_width = 0;
    uint32_t m_height = 0;
    uint32_t m_channels = 4;
    size_t m_bpp = 4;
    Portal::Graphics::ImageFormat m_format = Portal::Graphics::ImageFormat::RGBA8;
    double m_frame_rate = 0.0;
    uint64_t m_num_frames = 0;
 
    ContainerDataStructure m_structure;
 
    std::vector<DataVariant> m_data;
    std::vector<DataVariant> m_processed_data;
 
    mutable Memory::Seqlock m_data_lock;
    mutable Memory::Seqlock m_region_lock;
    mutable Memory::Seqlock m_cb_lock;
 
    std::atomic<ProcessingState> m_processing_state { ProcessingState::IDLE };
    std::atomic<int> m_processing_token_channel { -1 };
 
    std::function<void(const std::shared_ptr<SignalSourceContainer>&, ProcessingState)> m_state_callback;
    std::shared_ptr<DataProcessor> m_default_processor;
    std::shared_ptr<DataProcessingChain> m_processing_chain;
 
    std::unordered_map<std::string, RegionGroup> m_region_groups;
 
    std::atomic<uint64_t> m_read_position { 0 };
    bool m_looping_enabled {};
    Region m_loop_region;
 
    std::atomic<uint32_t> m_registered_readers { 0 };
    std::atomic<uint32_t> m_consumed_readers { 0 };
 
    // =========================================================================
    // Ring buffer state (inactive when m_ring_capacity == 0)
    // =========================================================================
 
    uint32_t m_ring_capacity { 0 };
    uint64_t m_total_source_frames { 0 };
 
    std::vector<std::atomic<uint64_t>> m_slot_frame;
 
    static constexpr uint32_t READY_QUEUE_CAPACITY = 256;
    Memory::LockFreeQueue<uint64_t, READY_QUEUE_CAPACITY> m_ready_queue;
 
    /**
     * @brief Highest frame index committed by the decode thread.
     *        Written by the decode thread via commit_frame(); read by
     *        update_read_position_for_channel() to compute buffered-ahead count.
     */
    std::atomic<uint64_t> m_cache_head { 0 };
 
    /**
     * @brief Trigger refill when (m_cache_head - read_position) drops below this.
     *        Set by the reader at load_into_container() time.
     *        A value of 0 disables threshold notification.
     */
    uint32_t m_refill_threshold { 0 };
 
    Registry::Service::IOService* m_io_service { nullptr }; // non-owning; owned by registry
    uint64_t m_io_reader_id { 0 };
 
    [[nodiscard]] uint32_t slot_for(uint64_t frame_index) const
    {
        return static_cast<uint32_t>(frame_index % m_ring_capacity);
    }
 
    [[nodiscard]] DataSpanVariant get_frame_span_impl(uint64_t frame_index) const override
    {
        return get_frame_typed(frame_index);
    }
 
    void get_frames_impl(
        void* output,
        size_t count,
        uint64_t start_frame,
        uint64_t num_frames,
        const std::type_info& type) const override;
 
    void get_value_impl(const std::vector<uint64_t>& coords,
        void* out, const std::type_info& type) const override;
 
    void set_value_impl(const std::vector<uint64_t>& coords,
        const void* in, const std::type_info& type) override;
 
private:
    [[nodiscard]] DataSpanVariant get_frame_typed(uint64_t frame_index) const;
 
    template <typename T>
    void get_frames_typed_as(std::span<T> output, uint64_t start_frame, uint64_t num_frames) const;
 
    /**
     * @brief Value range declared on the component dimension, if any.
     *
     * Resolves DEPTH first, then CHANNEL, matching the two roles
     * setup_dimensions() emits for the component axis.
     */
    [[nodiscard]] std::optional<DataDimension::ValueRange> component_range() const;
 
    void get_frames_typed(std::span<uint8_t> output, uint64_t start_frame, uint64_t num_frames) const;
 
    mutable std::vector<std::vector<float>> m_float_frame_cache;
    mutable std::vector<std::atomic<bool>> m_float_frame_dirty;
 
    void reset_float_frame_cache();
};
 
} // namespace MayaFlux::Kakshya