frame.h

Go to the documentation of this file.

#pragma once
 
#include <vector>
#include <memory>
#include <stdexcept>
 
#include "avcompat.h"
 
#if AVCPP_CXX_STANDARD >= 20
#if __has_include(<span>)
#include <span>
#endif
#endif
 
#include "ffmpeg.h"
#include "rational.h"
#include "timestamp.h"
#include "pixelformat.h"
#include "sampleformat.h"
#include "avutils.h"
 
extern "C" {
#include <libavutil/imgutils.h>
#include <libavutil/attributes.h>
}
 
namespace av
{
 
namespace frame {
 
namespace priv {
void channel_layout_copy(AVFrame &dst, const AVFrame &src);
}
 
static inline int64_t get_best_effort_timestamp(const AVFrame* frame) {
#if AVCPP_AVUTIL_VERSION_MAJOR < 56 // < FFmpeg 4.0
    return av_frame_get_best_effort_timestamp(frame);
#else
    return frame->best_effort_timestamp;
#endif
}
 
// Based on db6efa1815e217ed76f39aee8b15ee5c64698537
static inline uint64_t get_channel_layout(const AVFrame* frame) {
#if AVCPP_AVUTIL_VERSION_MAJOR < 56 // < FFmpeg 4.0
    return static_cast<uint64_t>(av_frame_get_channel_layout(frame));
#elif AVCPP_API_NEW_CHANNEL_LAYOUT
    return frame->ch_layout.order == AV_CHANNEL_ORDER_NATIVE ? frame->ch_layout.u.mask : 0;
#else
    return frame->channel_layout;
#endif
}
 
static inline void set_channel_layout(AVFrame* frame, uint64_t layout) {
#if AVCPP_AVUTIL_VERSION_MAJOR < 56 // < FFmpeg 4.0
    av_frame_set_channel_layout(frame, static_cast<int64_t>(layout));
#elif AVCPP_API_NEW_CHANNEL_LAYOUT
    av_channel_layout_uninit(&frame->ch_layout);
    av_channel_layout_from_mask(&frame->ch_layout, layout);
#else
    frame->channel_layout = layout;
#endif
}
 
 
static inline int get_channels(const AVFrame* frame) {
#if AVCPP_AVUTIL_VERSION_MAJOR < 56 // < FFmpeg 4.0
    return av_frame_get_channels(frame);
#elif AVCPP_API_NEW_CHANNEL_LAYOUT
    return frame->ch_layout.nb_channels;
#else
    return frame->channels;
#endif
}
 
static inline bool is_valid_channel_layout(const AVFrame *frame)
{
#if AVCPP_API_NEW_CHANNEL_LAYOUT
    return av_channel_layout_check(&frame->ch_layout);
#else
    return frame->channel_layout;
#endif
}
 
static inline int get_sample_rate(const AVFrame* frame) {
#if AVCPP_AVUTIL_VERSION_MAJOR < 56 // < FFmpeg 4.0
    return av_frame_get_sample_rate(frame);
#else
    return frame->sample_rate;
#endif
}
 
static inline void set_sample_rate(AVFrame* frame, int sampleRate) {
#if AVCPP_AVUTIL_VERSION_MAJOR < 56 // < FFmpeg 4.0
    av_frame_set_sample_rate(frame, sampleRate);
#else
    frame->sample_rate = sampleRate;
#endif
}
 
 
} // ::frame
 
template<typename T>
class Frame : public FFWrapperPtr<AVFrame>
{
protected:
    T& assignOperator(const T &rhs) {
        if (this == &rhs)
            return static_cast<T&>(*this);
        T(rhs).swap(static_cast<T&>(*this));
        return static_cast<T&>(*this);
    }
 
    T& moveOperator(T &&rhs) {
        if (this == &rhs)
            return static_cast<T&>(*this);
        T(std::move(rhs)).swap(static_cast<T&>(*this));
        return static_cast<T&>(*this);
    }
 
public:
    Frame() {
        m_raw = av_frame_alloc();
        m_raw->opaque = this;
    }
 
    ~Frame() {
        av_frame_free(&m_raw);
    }
 
    Frame(const AVFrame *frame) {
        if (frame) {
            m_raw = av_frame_alloc();
            m_raw->opaque = this;
            av_frame_ref(m_raw, frame);
        }
    }
 
    // Helper ctors to implement move/copy ctors
    Frame(const T& other) : Frame(other.m_raw) {
        if (m_raw)
            copyInfoFrom(other);
    }
 
    Frame(T&& other) : FFWrapperPtr<AVFrame>(nullptr) {
        if (other.m_raw) {
            m_raw = av_frame_alloc();
            m_raw->opaque = this;
            av_frame_move_ref(m_raw, other.m_raw);
            copyInfoFrom(other);
        }
    }
 
    static T null() {
        return T(nullptr);
    }
 
    // You must implement operators in deveritive classes using assignOperator() and moveOperator()
    void operator=(const Frame&) = delete;
 
    void swap(Frame &other) {
        using std::swap;
#define FRAME_SWAP(x) swap(x, other.x)
        FRAME_SWAP(m_raw);
        FRAME_SWAP(m_timeBase);
        FRAME_SWAP(m_streamIndex);
        FRAME_SWAP(m_isComplete);
#undef FRAME_SWAP
    }
 
    void copyInfoFrom(const T& other) {
        m_timeBase    = other.m_timeBase;
        m_streamIndex = other.m_streamIndex;
        m_isComplete  = other.m_isComplete;
    }
 
    bool isReferenced() const {
        return m_raw && m_raw->buf[0];
    }
 
    int refCount() const {
        if (m_raw && m_raw->buf[0])
            return av_buffer_get_ref_count(m_raw->buf[0]);
        else
            return 0;
    }
 
    AVFrame* makeRef() const {
        return m_raw ? av_frame_clone(m_raw) : nullptr;
    }
 
    T clone(size_t align = 1) const {
        T result;
 
        // Setup data required for buffer allocation
        result.m_raw->format         = m_raw->format;
        result.m_raw->width          = m_raw->width;
        result.m_raw->height         = m_raw->height;
        result.m_raw->nb_samples     = m_raw->nb_samples;
 
        frame::priv::channel_layout_copy(*result.m_raw, *m_raw);
 
        result.copyInfoFrom(static_cast<const T&>(*this));
 
        av_frame_get_buffer(result.m_raw, align);
        av_frame_copy(result.m_raw, m_raw);
        av_frame_copy_props(result.m_raw, m_raw);
        return result;
    }
 
    Timestamp pts() const
    {
        return {RAW_GET(pts, av::NoPts), m_timeBase};
    }
 
    attribute_deprecated void setPts(int64_t pts, Rational ptsTimeBase)
    {
        RAW_SET(pts, ptsTimeBase.rescale(pts, m_timeBase));
    }
 
    void setPts(const Timestamp &ts)
    {
        if (m_timeBase == Rational())
            m_timeBase = ts.timebase();
        RAW_SET(pts, ts.timestamp(m_timeBase));
    }
 
    const Rational& timeBase() const { return m_timeBase; }
 
    void setTimeBase(const Rational &value) {
        if (m_timeBase == value)
            return;
 
        if (!m_raw)
            return;
 
        int64_t rescaledPts          = NoPts;
        int64_t rescaledBestEffortTs = NoPts;
 
        if (m_timeBase != Rational() && value != Rational()) {
            if (m_raw->pts != av::NoPts)
                rescaledPts = m_timeBase.rescale(m_raw->pts, value);
 
            if (m_raw->best_effort_timestamp != av::NoPts)
                rescaledBestEffortTs = m_timeBase.rescale(m_raw->best_effort_timestamp, value);
        } else {
            rescaledPts          = m_raw->pts;
            rescaledBestEffortTs = m_raw->best_effort_timestamp;
        }
 
        if (m_timeBase != Rational()) {
            m_raw->pts                   = rescaledPts;
            m_raw->best_effort_timestamp = rescaledBestEffortTs;
        }
 
        m_timeBase = value;
    }
 
    int streamIndex() const {
        return m_streamIndex;
    }
 
    void setStreamIndex(int streamIndex) {
        m_streamIndex = streamIndex;
    }
 
    void setComplete(bool isComplete) {
        m_isComplete = isComplete;
    }
 
    bool isComplete() const { return m_isComplete; }
 
    bool isValid() const {
         return (!isNull() &&
             ((m_raw->data[0] && m_raw->linesize[0]) ||
             ((m_raw->format == AV_PIX_FMT_VAAPI) && ((intptr_t)m_raw->data[3] > 0)))
         );
    }
 
    operator bool() const { return isValid() && isComplete(); }
 
    uint8_t *data(size_t plane = 0) {
        if (!m_raw || plane >= size_t(AV_NUM_DATA_POINTERS + m_raw->nb_extended_buf))
            return nullptr;
        return m_raw->extended_data[plane];
    }
 
    const uint8_t *data(size_t plane = 0) const {
        if (!m_raw || plane >= size_t(AV_NUM_DATA_POINTERS + m_raw->nb_extended_buf))
            return nullptr;
        return m_raw->extended_data[plane];;
    }
 
    size_t size(size_t plane) const {
        if (!m_raw || plane >= size_t(AV_NUM_DATA_POINTERS + m_raw->nb_extended_buf))
            return 0;
        AVBufferRef *buf = plane < AV_NUM_DATA_POINTERS ?
                               m_raw->buf[plane] :
                               m_raw->extended_buf[plane - AV_NUM_DATA_POINTERS];
        if (buf == nullptr)
            return 0;
        return size_t(buf->size);
    }
 
    size_t size() const {
        if (!m_raw)
            return 0;
        size_t total = 0;
        if (m_raw->buf[0]) {
            for (auto i = 0; i < AV_NUM_DATA_POINTERS && m_raw->buf[i]; i++) {
                total += m_raw->buf[i]->size;
            }
 
            for (auto i = 0; i < m_raw->nb_extended_buf; ++i) {
                total += m_raw->extended_buf[i]->size;
            }
        } else if (m_raw->data[0]) {
            if (m_raw->width && m_raw->height) {
                uint8_t data[4] = {0};
                int     linesizes[4] = {
                    m_raw->linesize[0],
                    m_raw->linesize[1],
                    m_raw->linesize[2],
                    m_raw->linesize[3],
                };
                total = av_image_fill_pointers(reinterpret_cast<uint8_t**>(&data),
                                               static_cast<AVPixelFormat>(m_raw->format),
                                               m_raw->height,
                                               nullptr,
                                               linesizes);
            } else if (m_raw->nb_samples && frame::is_valid_channel_layout(m_raw)) {
                for (auto i = 0; i < m_raw->nb_extended_buf + AV_NUM_DATA_POINTERS && m_raw->extended_data[i]; ++i) {
                    // According docs, all planes must have same size
                    total += m_raw->linesize[0];
                }
            }
        }
        return total;
    }
 
    void dump() const {
        if (!m_raw)
            return;
        if (m_raw->buf[0]) {
            for (size_t i = 0; i < AV_NUM_DATA_POINTERS && m_raw->buf[i]; i++) {
                av::hex_dump(stdout, m_raw->buf[i]->data, m_raw->buf[i]->size);
            }
        } else if (m_raw->data[0]) {
            av_hex_dump(stdout, m_raw->data[0], size());
        }
    }
 
protected:
    Rational             m_timeBase{};
    int                  m_streamIndex {-1};
    bool                 m_isComplete  {false};
};
 
 
class VideoFrame : public Frame<VideoFrame>
{
public:
    using Frame<VideoFrame>::Frame;
 
    VideoFrame() = default;
    VideoFrame(PixelFormat pixelFormat, int width, int height, int align = 1);
    VideoFrame(const uint8_t *data, size_t size, PixelFormat pixelFormat, int width, int height, int align = 1);
 
    VideoFrame(const VideoFrame &other);
    VideoFrame(VideoFrame &&other);
 
    VideoFrame& operator=(const VideoFrame &rhs);
    VideoFrame& operator=(VideoFrame &&rhs);
 
    PixelFormat            pixelFormat() const;
    int                    width() const;
    int                    height() const;
 
    bool                   isKeyFrame() const;
    void                   setKeyFrame(bool isKey);
 
    int                    quality() const;
    void                   setQuality(int quality);
 
    AVPictureType          pictureType() const;
    void                   setPictureType(AVPictureType type = AV_PICTURE_TYPE_NONE);
 
    Rational               sampleAspectRatio() const;
    void                   setSampleAspectRatio(const Rational& sampleAspectRatio);
 
    size_t                 bufferSize(int align = 1, OptionalErrorCode ec = throws()) const;
    bool                   copyToBuffer(uint8_t *dst, size_t size, int align = 1, OptionalErrorCode ec = throws());
    bool                   copyToBuffer(std::vector<uint8_t>& dst, int align = 1, OptionalErrorCode ec = throws());
 
 
    static VideoFrame wrap(const void *data, size_t size, PixelFormat pixelFormat, int width, int height, int align = 1);
 
#ifdef __cpp_lib_span
    static VideoFrame wrap(std::span<const std::byte> data, PixelFormat pixelFormat, int width, int height, int align = 1);
 
    static VideoFrame wrap(std::span<const std::byte> data, void (*deleter)(void *opaque, uint8_t *data),
                           void *opaque, PixelFormat pixelFormat, int width, int height, int align = 1);
#endif
};
 
 
class AudioSamples : public Frame<AudioSamples>
{
public:
    using Frame<AudioSamples>::Frame;
 
    AudioSamples() = default;
    AudioSamples(SampleFormat sampleFormat, int samplesCount, uint64_t channelLayout, int sampleRate, int align = SampleFormat::AlignDefault);
    AudioSamples(const uint8_t *data, size_t size,
                  SampleFormat sampleFormat, int samplesCount, uint64_t channelLayout, int sampleRate, int align = SampleFormat::AlignDefault);
 
    AudioSamples(const AudioSamples &other);
    AudioSamples(AudioSamples &&other);
 
    AudioSamples& operator=(const AudioSamples &rhs);
    AudioSamples& operator=(AudioSamples &&rhs);
 
    int            init(SampleFormat sampleFormat, int samplesCount, uint64_t channelLayout, int sampleRate, int align = SampleFormat::AlignDefault);
 
    SampleFormat sampleFormat() const;
    int            samplesCount() const;
    int            channelsCount() const;
    uint64_t channelsLayout() const;
    int            sampleRate() const;
    size_t         sampleBitDepth(OptionalErrorCode ec = throws()) const;
    bool           isPlanar() const;
 
    std::string    channelsLayoutString() const;
};
 
 
} // ::av