clipper/backend/media/media_set.go

package media

import (
	"encoding/binary"
	"encoding/json"
	"errors"
	"fmt"
	"io"
	"log"
	"os"
	"time"

	"github.com/google/uuid"
)

const SizeOfInt16 = 2

type Audio struct {
	Bytes    int64 `json:"bytes"`
	Channels int   `json:"channels"`
	// ApproxFrames is used during initial processing when a precise frame count
	// cannot be determined. Prefer Frames in all other cases.
	ApproxFrames int64  `json:"approx_frames"`
	Frames       int64  `json:"frames"`
	SampleRate   int    `json:"sample_rate"`
	YoutubeItag  int    `json:"youtube_itag"`
	MimeType     string `json:"mime_type"`
}

type Video struct {
	Bytes    int64         `json:"bytes"`
	Duration time.Duration `json:"duration"`
	// not sure if this are needed any more?
	ThumbnailWidth  int    `json:"thumbnail_width"`
	ThumbnailHeight int    `json:"thumbnail_height"`
	YoutubeItag     int    `json:"youtube_itag"`
	MimeType        string `json:"mime_type"`
}

// MediaSet represents the media and metadata associated with a single media
// resource (for example, a YouTube video).
type MediaSet struct {
	Audio     Audio     `json:"audio"`
	Video     Video     `json:"video"`
	ID        uuid.UUID `json:"id"`
	YoutubeID string    `json:"youtube_id"`

	exists bool `json:"exists"`
}

// New builds a new MediaSet with the given ID.
func NewMediaSet(youtubeID string) *MediaSet {
	return &MediaSet{YoutubeID: youtubeID}
}

// TODO: pass io.Readers/Writers instead of strings.
func (m *MediaSet) RawAudioPath() string     { return fmt.Sprintf("cache/%s.raw", m.YoutubeID) }
func (m *MediaSet) EncodedAudioPath() string { return fmt.Sprintf("cache/%s.m4a", m.YoutubeID) }
func (m *MediaSet) VideoPath() string        { return fmt.Sprintf("cache/%s.mp4", m.YoutubeID) }
func (m *MediaSet) ThumbnailPath() string    { return fmt.Sprintf("cache/%s.jpg", m.YoutubeID) }
func (m *MediaSet) MetadataPath() string     { return fmt.Sprintf("cache/%s.json", m.YoutubeID) }

func (m *MediaSet) Exists() bool {
	if m.YoutubeID == "" {
		return false
	}
	if m.exists {
		return true
	}
	if _, err := os.Stat(m.MetadataPath()); err == nil {
		m.exists = true
		return true
	}
	return false
}

func (m *MediaSet) Load() error {
	if m.YoutubeID == "" {
		return errors.New("error opening mediaset with blank ID")
	}

	metadataFile, err := os.Open(m.MetadataPath())
	if err != nil {
		return fmt.Errorf("error opening metadata file: %v", err)
	}
	defer func() { _ = metadataFile.Close() }()

	if err := json.NewDecoder(metadataFile).Decode(m); err != nil {
		return fmt.Errorf("error decoding metadata: %v", err)
	}
	return nil
}

func (m *MediaSet) Peaks(start, end int64, numBins int) ([][]int16, error) {
	if !m.Exists() {
		return nil, errors.New("cannot compute peaks for non-existent MediaSet")
	}

	var err error
	fptr, err := os.Open(m.RawAudioPath())
	if err != nil {
		return nil, fmt.Errorf("audio open error: %v", err)
	}
	defer fptr.Close()

	startByte := start * int64(m.Audio.Channels) * SizeOfInt16
	if _, err = fptr.Seek(startByte, io.SeekStart); err != nil {
		return nil, fmt.Errorf("audio seek error: %v", err)
	}

	numFrames := end - start
	framesPerBin := numFrames / int64(numBins)

	peaks := make([][]int16, m.Audio.Channels)
	for i := 0; i < m.Audio.Channels; i++ {
		peaks[i] = make([]int16, numBins)
	}

	samples := make([]int16, framesPerBin*int64(m.Audio.Channels))

	for binNum := 0; binNum < numBins; binNum++ {
		if err := binary.Read(fptr, binary.LittleEndian, samples); err != nil {
			return nil, fmt.Errorf("error reading samples: %v", err)
		}
		for i, samp := range samples {
			if samp < 0 {
				samp = -samp
			}
			chanIndex := i % m.Audio.Channels
			if samp > peaks[chanIndex][binNum] {
				peaks[chanIndex][binNum] = samp
			}
		}
	}

	log.Println("finished generating peaks")
	return peaks, nil
}