ANSCORE/MediaClient/media/video_player.h

#ifndef VIDEO_PLAYER_H
#define VIDEO_PLAYER_H

#include "sys_inc.h"
#include "hqueue.h"
#include "video_decoder.h"
#include "audio_decoder.h"
#include "audio_play.h"
#include "avi_write.h"
#include "media_util.h"
#include <list>
#include <string>
#include <opencv2/imgproc.hpp>
#include <opencv2/highgui.hpp>
#include <opencv2/opencv.hpp>
#include <turbojpeg.h>
#include <chrono>

typedef struct
{
    uint32          SyncTimestamp;
    struct timeval  SyncTime;
} HTCLOCK;

typedef std::list<AVFrame*> FRAMELIST;

// Thread-safe queue for AVFrame
class FrameQueue {
private:
    std::queue<AVFrame*> frameQueue;
    std::mutex queueMutex;
    const size_t maxFrames = 20;  // Increased buffer size for better performance
    uint64_t m_frameSeq = 0;     // Sequence counter — increments on each new frame push

public:
    // Push a new frame to the queue (removes oldest if full)
    void pushFrame(AVFrame* newFrame) {
        if (!newFrame) return;

        std::lock_guard<std::mutex> lock(queueMutex);

        // Clone the frame to ensure proper ownership transfer
        AVFrame* frameCopy = av_frame_clone(newFrame);
        if (!frameCopy) {
            std::cerr << "Failed to clone AVFrame!" << std::endl;
            return;
        }

        frameQueue.push(frameCopy);
        m_frameSeq++;  // New frame arrived

        // If queue exceeds max size, remove the oldest frame
        if (frameQueue.size() > maxFrames) {
            AVFrame* oldFrame = frameQueue.front();
            frameQueue.pop();
            av_frame_free(&oldFrame);
        }
    }

    // Get current sequence number (check if new frames arrived)
    uint64_t getSequence() {
        std::lock_guard<std::mutex> lock(queueMutex);
        return m_frameSeq;
    }

    // Retrieve latest frame (returns a clone for thread safety)
    AVFrame* getLatestFrame() {
        std::lock_guard<std::mutex> lock(queueMutex);

        if (frameQueue.empty()) {
            return nullptr;  // No frames available
        }

        // Clone the latest frame before returning it
        return av_frame_clone(frameQueue.back());
    }

    // Retrieve and remove the oldest frame from the queue
    AVFrame* popFrame() {
        std::lock_guard<std::mutex> lock(queueMutex);

        if (frameQueue.empty()) {
            return nullptr;
        }

        AVFrame* frontFrame = frameQueue.front();
        frameQueue.pop();
        return frontFrame;  // Caller must free the returned frame
    }

    // Check if the queue is empty
    bool isEmpty() {
        std::lock_guard<std::mutex> lock(queueMutex);
        return frameQueue.empty();
    }

    // Clear the queue (e.g., when shutting down)
    void clearQueue() {
        std::lock_guard<std::mutex> lock(queueMutex);
        while (!frameQueue.empty()) {
            AVFrame* frame = frameQueue.front();
            frameQueue.pop();
            av_frame_free(&frame);
        }
        m_frameSeq = 0;
    }
};

class CVideoPlayer
{

public:
    CVideoPlayer();
    virtual ~CVideoPlayer();
    virtual BOOL    open(std::string fileName);
    virtual BOOL    open(std::string username, std::string password, std::string url);
    virtual BOOL    play() = 0;
    virtual void    stop() = 0;
    virtual BOOL    pause() = 0;
    virtual void    close();
    virtual BOOL    seek(int pos) { return FALSE; }
    virtual BOOL    isPlaying() { return m_bPlaying; }
    virtual BOOL    isPaused() { return m_bPaused; }
    virtual void    setVolume(int volume);
    virtual void    snapshot(int videofmt);
    virtual BOOL    record(std::string baseName);
    virtual void    stopRecord();
    virtual BOOL    isRecording() { return m_bRecording; }
    virtual BOOL    onRecord() { return FALSE; }
    virtual int     getVideoClock() { return 1000; }
    virtual int     getAudioClock() { return 1000; }
    //virtual void    setWindowSize(int size);
    virtual int64   getElapse() { return 0; }
    virtual int64   getDuration() { return 0; }
    virtual int     getVideoCodec() { return m_nVideoCodec; }
    virtual int     getAudioCodec() { return m_nAudioCodec; }

    virtual void    setAuthInfo(std::string acct, std::string pass) { m_acct = acct; m_pass = pass; }
    //virtual void    setRenderMode(int mode) { m_nRenderMode = mode; }
    virtual void    setHWDecoding(int mode, int preferredGpu = -1) { m_nHWDecoding = mode; m_nPreferredGpu = preferredGpu; }
    virtual bool    isHWDecodingActive() const {
        return m_pVideoDecoder && m_pVideoDecoder->isHardwareDecoderEnabled();
    }
    virtual int     getHWDecodingGpuIndex() const {
        return m_pVideoDecoder ? m_pVideoDecoder->getHWGpuIndex() : -1;
    }
    // Image quality mode: 0=fast (OpenCV BT.601, ~2ms), 1=quality (sws BT.709+range, ~12ms)
    virtual void    setImageQuality(int mode) { m_nImageQuality = mode; }
    void            setTargetFPS(double intervalMs);  // Set minimum interval between processed frames in ms (0 = no limit, 100 = ~10 FPS)
    virtual void    setRtpMulticast(BOOL flag) {}
    virtual void    setRtpOverUdp(BOOL flag) {}

#ifdef OVER_HTTP
    virtual void    setRtspOverHttp(int flag, int port) {}
#endif

#ifdef OVER_WEBSOCKET
    virtual void    setRtspOverWs(int flag, int port) {}
#endif

#ifdef BACKCHANNEL
    virtual int     getBCFlag() { return 0; }
    virtual void    setBCFlag(int flag) {}
    virtual int     getBCDataFlag() { return 0; }
    virtual void    setBCDataFlag(int flag) {}
    virtual void    setAudioDevice(int index) {}
#endif

#ifdef REPLAY
    virtual int     getReplayFlag() { return 0; }
    virtual void    setReplayFlag(int flag) {}
    virtual void    setScale(double scale) {}
    virtual void    setRateControlFlag(int flag) {}
    virtual void    setImmediateFlag(int flag) {}
    virtual void    setFramesFlag(int flag, int interval) {}
    virtual void    setReplayRange(time_t start, time_t end) {}
#endif

    BOOL            openVideo(int codec, uint8* extradata = NULL, int extradata_size = 0);
    BOOL            openVideo(enum AVCodecID codec, uint8* extradata = NULL, int extradata_size = 0);
    void            closeVideo();
    BOOL            openAudio(int codec, int samplerate, int channels, int bitpersample = 0);
    BOOL            openAudio(enum AVCodecID codec, int samplerate, int channels, int bitpersample = 0);
    void            closeAudio();
    void            enableAudio(bool status);
    void            playVideo(uint8* data, int len, uint32 ts, uint16 seq);
    void            playAudio(uint8* data, int len, uint32 ts, uint16 seq);

    void            recordVideo(uint8* data, int len, uint32 ts, uint16 seq);
    void            recordAudio(uint8* data, int len, uint32 ts, uint16 seq);

    int             getVideoWidth();
    int             getVideoHeight();
    int             getOriginalWidth() { return m_nv12OrigWidth; }   // Full NV12 resolution (before display downscale)
    int             getOriginalHeight() { return m_nv12OrigHeight; } // Full NV12 resolution (before display downscale)
    double          getFrameRate();
    int             getSampleRate() { return m_nSampleRate; }
    int             getChannel() { return m_nChannel; }

    void            onVideoFrame(AVFrame* frame);
    void            onAudioFrame(AVFrame* frame);
    void            StartVideoDecoder();
    void            StopVideoDecoder();


    cv::Mat         getImage(int& width, int& height, int64_t& pts);// Get image
    std::string     getJpegImage(int& width, int& height, int64_t& pts);// Get image
    AVFrame*        getNV12Frame();  // Transfers ownership of NV12/YUV frame for GPU fast-path (caller must av_frame_free)
    AVFrame*        getCudaHWFrame(); // Returns clone of CUDA HW frame (device ptrs); caller must av_frame_free
    bool            isCudaHWAccel() const;
    void            setBbox(cv::Rect bbox);
    void            setCrop(bool crop);
protected:
    void            updateClock(HTCLOCK* clock, uint32 ts, int frequency);
    BOOL            initFrame(AVFrame*& frame, int width, int height, AVPixelFormat pixfmt);
    BOOL            doSnapshot(AVFrame* srcframe);
    AVFrame* convertFrame(AVFrame* srcframe, AVFrame* dstframe, BOOL updown);
    void            recordVideoEx(uint8* data, int len, uint32 ts, uint16 seq);
    BOOL            recordSwitchCheck();
    void            recordFileSwitch();
    AVFrame* cropFrame(const AVFrame* srcFrame, cv::Rect bBox, bool cropFlag);

    cv::Mat         avframeAnyToCvmat(const AVFrame* frame);
    cv::Mat         avframeNV12ToCvMat(const AVFrame* frame);
    cv::Mat         avframeYUVJ420PToCvmat(const AVFrame* frame);
    cv::Mat         avframeToCVMat(const AVFrame* frame);

    // YUVJ420P to JPEG
    std::string     avframeYUVJ420PToJpegString(const AVFrame* pFrame);
    std::string     avframeYUVJ420PToJpegStringUsingTurboJPEG(const AVFrame* pFrame);
    std::string     avframeYUVJ420PToJpegStringUsingFFMpeg(const AVFrame* pFrame);
    AVFrame* convertNV12ToYUVJ420P(const AVFrame* nv12Frame);
    std::string     encodeYUVJ420PToJPEG(AVFrame* frame, int quality = 90);

    //Direct conversion from NV12 to JPEG
    std::string     avframeToJpegString(const AVFrame* pFrame);
    std::string     encodeNV12ToJPEG_TurboJPEG(const AVFrame* frame, int quality = 90);
    std::string     encodeNV12ToJPEG_FFmpeg(const AVFrame* pFrame, int quality = 90);


    // Check if frame are identical
    bool areFramesIdentical(AVFrame* frame1, AVFrame* frame2);
protected:
    // Use a constant for the maximum queue size
    //const int MAX_VIDEO_FRAMES = 10;
    //const int MAX_AUDIO_FRAMES = 2;
    BOOL            m_bVideoInited;
    BOOL            m_bAudioInited;
    tjhandle        _tjInstance;

    std::recursive_mutex		_mutex;

    std::unique_ptr<CVideoDecoder> m_pVideoDecoder = std::make_unique<CVideoDecoder>();
    std::unique_ptr<CAudioDecoder> m_pAudioDecoder = std::make_unique<CAudioDecoder>();
    std::unique_ptr<CAudioPlay> m_pAudioPlay = nullptr;

    cv::Mat         m_currentImage;
    AVFrame*        m_currentNV12Frame = nullptr;  // Preserved NV12/YUV frame for GPU fast-path
    AVFrame*        m_currentCudaHWFrame = nullptr; // CUDA HW frame with device ptrs for zero-copy
    int             m_Width;
    int             m_Height;
    int             m_nv12OrigWidth = 0;   // Original NV12 frame width (before display resize)
    int             m_nv12OrigHeight = 0;  // Original NV12 frame height (before display resize)
    int64_t         m_pts;
    uint64_t        m_lastFrameSeq = 0;  // Last processed frame sequence number
    bool            m_bWaitingForKeyframe = true;  // Skip frames until first keyframe after start/restart
    int             m_cleanFrameCount = 0;         // Count of clean frames after keyframe
    static const int SETTLE_FRAME_COUNT = 5;       // Number of clean frames before delivering new frames

    // Frame rate limiting — skip post-decode processing for frames beyond target interval
    double m_targetIntervalMs = 100.0;                     // default 100ms (~10 FPS), 0 = no limit (process all frames)
    std::chrono::steady_clock::time_point m_lastProcessedTime;  // timestamp of last processed frame
    bool m_targetFPSInitialized = false;                   // first-frame flag

    BOOL            m_bPlaying;
    BOOL            m_bPaused;

    std::string     m_acct;
    std::string     m_pass;
    std::string     m_sFileName;
    std::string     m_sBaseName;
    std::string     m_jpegImage;
    std::string     m_lastJpegImage;

    int             m_nHWDecoding;
    int             m_nPreferredGpu = -1; // -1=auto (least-loaded), >=0 = prefer this GPU for NVDEC
    int             m_nImageQuality = 0;  // 0=fast (default), 1=quality BT.709
    //AVPixelFormat   m_nDstVideoFmt;
    BOOL            m_bUpdown;
    BOOL            m_bSnapshot;
    int             m_nSnapVideoFmt;

    H26XParamSets   m_h26XParamSets;

    int             m_nVideoCodec;
    int             m_nAudioCodec;
    int             m_nSampleRate;
    int             m_nChannel;
    int             m_nBitPerSample;

    AVFrame* m_pSnapFrame;
    FrameQueue      g_frameQueue;
    FrameQueue      a_frameQueue;

    BOOL            m_bRecording;
    BOOL            m_bNalFlag;
    AVICTX* m_pAviCtx;
    void* m_pRecordMutex;

    HTCLOCK         m_audioClock;
    void* m_pAudioListMutex;
    //FRAMELIST       m_audioFrameList;
    BOOL            m_audioPlayFlag;
    //pthread_t       m_audioPlayThread;

    HTCLOCK         m_videoClock;
    void* m_pVideoListMutex;
    //FRAMELIST       m_videoFrameList;
    BOOL            m_videoPlayFlag;
    //pthread_t       m_videoPlayThread;

    uint64          m_nLastAudioPts;
    time_t          m_lastAudioTS;
    cv::Rect        m_Bbox;
    bool			m_bCrop;
    SwsContext* swsCtx = nullptr;  // Shared SwsContext
    int lastWidth = 0, lastHeight = 0;
    AVPixelFormat lastPixFmt = AV_PIX_FMT_NONE;
    AVPixelFormat lastOutPixFmt = AV_PIX_FMT_NONE;

    // Dedicated NV12→BGR context with correct color space (BT.709 for HD/4K)
    SwsContext* m_nv12SwsCtx = nullptr;
    int m_nv12LastWidth = 0, m_nv12LastHeight = 0;
    int m_nv12LastColorspace = -1;
    int m_nv12LastRange = -1;

    void initSwsContext(int width, int height, AVPixelFormat pixFmt, AVPixelFormat outputPixFmt = AV_PIX_FMT_YUVJ420P);
    void initNV12SwsContext(const AVFrame* frame);
    void cropPlane(const AVFrame* srcFrame, AVFrame* croppedFrame, int planeIndex, int offsetX, int offsetY, int subsampleX, int subsampleY);
    bool cropFrameData(const AVFrame* srcFrame, AVFrame* croppedFrame, const cv::Rect& bBox);
};

#endif // end of VIDEO_PLAYER_H