clipper/backend/main.go

package main

import (
	"bytes"
	"encoding/binary"
	"encoding/json"
	"errors"
	"fmt"
	"io"
	"log"
	"net/http"
	"os"
	"os/exec"
	"strconv"
	"strings"
	"time"

	"github.com/kkdai/youtube/v2"
)

const (
	SizeOfInt16                = 2
	ContentTypeAudioM4A        = "audio/m4a"
	ContentTypeApplicationJSON = "application/json"
	ItagM4AAudio               = 140

	DefaultFormat       = "s16le"
	DefaultAudioCodec   = "pcm_s16le"
	DefaultSampleRate   = 48000
	DefaultHTTPBindAddr = "0.0.0.0:8888"
	DefaultTimeout      = 30 * time.Second
)

type AudioFileMetadata struct {
	Bytes      int64 `json:"bytes"`
	Channels   int   `json:"channels"`
	Frames     int64 `json:"frames"`
	SampleRate int   `json:"sample_rate"`
}

type AudioFile struct {
	AudioFileMetadata
	videoID string
	exists  bool
}

func NewAudioFile(videoID string) (*AudioFile, error) {
	f := &AudioFile{videoID: videoID}

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

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

		f.exists = true
	}

	return f, nil
}

func (f *AudioFile) rawAudioPath() string     { return fmt.Sprintf("cache/%s.raw", f.videoID) }
func (f *AudioFile) encodedAudioPath() string { return fmt.Sprintf("cache/%s.m4a", f.videoID) }
func (f *AudioFile) metadataPath() string     { return fmt.Sprintf("cache/%s.json", f.videoID) }

func (f *AudioFile) Exists() bool {
	if f.exists {
		return true
	}
	if _, err := os.Stat(f.metadataPath()); err == nil {
		f.exists = true
		return true
	}
	return false
}

func (f *AudioFile) Download() error {
	log.Println("fetching audio stream from youtube...")
	var ytClient youtube.Client

	video, err := ytClient.GetVideo(f.videoID)
	if err != nil {
		return fmt.Errorf("error fetching video: %v", err)
	}

	var format *youtube.Format
	for _, candidate := range video.Formats.WithAudioChannels() {
		if format == nil || (candidate.ContentLength > 0 && candidate.ContentLength < format.ContentLength) {
			format = &candidate
		}
	}
	if format == nil {
		return errors.New("error selecting format: no format available")
	}

	stream, _, err := ytClient.GetStream(video, format)
	if err != nil {
		return fmt.Errorf("error fetching stream: %v", err)
	}

	rawAudioFile, err := os.Create(f.rawAudioPath())
	if err != nil {
		return fmt.Errorf("error creating raw audio file: %v", err)
	}

	encodedAudioFile, err := os.Create(f.encodedAudioPath())
	if err != nil {
		return fmt.Errorf("error creating encoded audio file: %v", err)
	}
	streamReader := io.TeeReader(stream, encodedAudioFile)

	var errOut bytes.Buffer
	cmd := exec.Command("ffmpeg", "-i", "-", "-f", DefaultFormat, "-ar", strconv.Itoa(DefaultSampleRate), "-acodec", DefaultAudioCodec, "-")
	cmd.Stdin = streamReader
	cmd.Stdout = rawAudioFile
	cmd.Stderr = &errOut

	if err = cmd.Run(); err != nil {
		log.Println(errOut.String())
		return fmt.Errorf("error processing audio: %v", err)
	}

	if err = rawAudioFile.Close(); err != nil {
		return fmt.Errorf("error writing file: %v", err)
	}

	rawAudioFile, err = os.Open(f.rawAudioPath())
	if err != nil {
		return fmt.Errorf("error reading file: %v", err)
	}

	fi, err := rawAudioFile.Stat()
	if err != nil {
		return fmt.Errorf("error reading file info: %v", err)
	}

	numFrames := fi.Size() / int64(SizeOfInt16) / int64(format.AudioChannels)
	sampleRate, err := strconv.Atoi(format.AudioSampleRate)
	if err != nil {
		return fmt.Errorf("invalid samplerate: %s", format.AudioSampleRate)
	}

	f.AudioFileMetadata = AudioFileMetadata{
		Bytes:      fi.Size(),
		Channels:   format.AudioChannels,
		Frames:     numFrames,
		SampleRate: sampleRate,
	}

	metadataFile, err := os.Create(f.metadataPath())
	if err != nil {
		return fmt.Errorf("error opening metadata file: %v", err)
	}

	if err = json.NewEncoder(metadataFile).Encode(f.AudioFileMetadata); err != nil {
		return fmt.Errorf("error encoding metadata: %v", err)
	}

	if err = metadataFile.Close(); err != nil {
		return fmt.Errorf("error writing metadata file: %v", err)
	}

	return nil
}

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

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

	numChannels := f.Channels

	startByte := start * int64(numChannels) * 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, numChannels)
	for i := 0; i < numChannels; i++ {
		peaks[i] = make([]int16, numBins)
	}

	samples := make([]int16, framesPerBin*int64(numChannels))

	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 % numChannels
			if samp > peaks[chanIndex][binNum] {
				peaks[chanIndex][binNum] = samp
			}
		}
	}

	return peaks, nil
}

func handleRequest(w http.ResponseWriter, r *http.Request) {
	if r.Method != http.MethodGet {
		w.WriteHeader(http.StatusMethodNotAllowed)
		w.Write([]byte("method not allowed"))
		return
	}

	w.Header().Add("Content-Type", ContentTypeApplicationJSON)
	w.Header().Add("Access-Control-Allow-Origin", "*")

	if strings.HasPrefix(r.URL.Path, "/api/download") {
		videoID := r.URL.Query().Get("video_id")
		if videoID == "" {
			w.WriteHeader(http.StatusBadRequest)
			w.Write([]byte(`{"error": "no video ID provided"}`))
			return
		}

		audioFile, err := NewAudioFile(videoID)
		if err != nil {
			log.Printf("error building audio file: %v", err)
			w.WriteHeader(http.StatusInternalServerError)
			w.Write([]byte(`{"error": "could not download audio"}`))
			return
		}

		if !audioFile.Exists() {
			if err = audioFile.Download(); err != nil {
				log.Printf("error downloading audio file: %v", err)
				w.WriteHeader(http.StatusInternalServerError)
				w.Write([]byte(`{"error": "could not download audio"}`))
				return
			}
		}

		w.WriteHeader(http.StatusOK)

		err = json.NewEncoder(w).Encode(audioFile)
		if err != nil {
			log.Printf("error encoding audio file: %v", err)
		}

		return
	}

	if strings.HasPrefix(r.URL.Path, "/api/audio") {
		log.Printf("got headers for audio request: %+v", r.Header)
		videoID := r.URL.Query().Get("video_id")
		if videoID == "" {
			w.WriteHeader(http.StatusBadRequest)
			w.Write([]byte(`{"error": "no video ID provided"}`))
			return
		}

		audioFile, err := NewAudioFile(videoID)
		if err != nil {
			log.Printf("error building audio file: %v", err)
			w.WriteHeader(http.StatusInternalServerError)
			w.Write([]byte(`{"error": "could not download audio"}`))
			return
		}
		if !audioFile.Exists() {
			if err = audioFile.Download(); err != nil {
				log.Printf("error downloading audio file: %v", err)
				w.WriteHeader(http.StatusInternalServerError)
				w.Write([]byte(`{"error": "could not download audio"}`))
				return
			}
		}

		http.ServeFile(w, r, audioFile.encodedAudioPath())

		return
	}

	if strings.HasPrefix(r.URL.Path, "/api/peaks") {
		videoID := r.URL.Query().Get("video_id")
		if videoID == "" {
			w.WriteHeader(http.StatusBadRequest)
			w.Write([]byte(`{"error": "no video ID provided"}`))
			return
		}

		start, err := strconv.ParseInt(r.URL.Query().Get("start"), 0, 64)
		if err != nil {
			w.WriteHeader(http.StatusBadRequest)
			w.Write([]byte(`{"error": "invalid start parameter provided"}`))
			return
		}
		end, err := strconv.ParseInt(r.URL.Query().Get("end"), 0, 64)
		if err != nil {
			w.WriteHeader(http.StatusBadRequest)
			w.Write([]byte(`{"error": "invalid end parameter provided"}`))
			return
		}
		numBins, err := strconv.Atoi(r.URL.Query().Get("bins"))
		if err != nil {
			w.WriteHeader(http.StatusBadRequest)
			w.Write([]byte(`{"error": "invalid bins parameter provided"}`))
			return
		}

		audioFile, err := NewAudioFile(videoID)
		if err != nil {
			log.Printf("error building audio file: %v", err)
			w.WriteHeader(http.StatusInternalServerError)
			w.Write([]byte(`{"error": "could not download audio"}`))
			return
		}

		if !audioFile.Exists() {
			log.Println("audio file does not exists, cannot compute peaks")
			w.WriteHeader(http.StatusInternalServerError)
			w.Write([]byte(`{"error": "audio file not available"}`))
			return
		}

		peaks, err := audioFile.Peaks(start, end, numBins)
		if err != nil {
			log.Printf("error generating peaks: %v", err)
			w.WriteHeader(http.StatusInternalServerError)
			w.Write([]byte(`{"error": "could not generate peaks"}`))
		}

		w.WriteHeader(http.StatusOK)

		err = json.NewEncoder(w).Encode(peaks)
		if err != nil {
			log.Printf("error encoding peaks: %v", err)
		}

		return
	}

	w.WriteHeader(http.StatusNotFound)
	w.Write([]byte("page not found"))
}

func main() {
	srv := http.Server{
		ReadTimeout:  DefaultTimeout,
		WriteTimeout: DefaultTimeout,
		Addr:         DefaultHTTPBindAddr,
		Handler:      http.HandlerFunc(handleRequest),
	}

	log.Fatal(srv.ListenAndServe())
}