319 lines
8.0 KiB
Go
319 lines
8.0 KiB
Go
package media
|
|
|
|
import (
|
|
"bytes"
|
|
"context"
|
|
"errors"
|
|
"fmt"
|
|
"io"
|
|
"log"
|
|
"os/exec"
|
|
"sort"
|
|
"strconv"
|
|
"strings"
|
|
"time"
|
|
|
|
"github.com/aws/aws-sdk-go-v2/service/s3"
|
|
youtubev2 "github.com/kkdai/youtube/v2"
|
|
)
|
|
|
|
const s3Bucket = "clipper-development"
|
|
|
|
const (
|
|
rawAudioCodec = "pcm_s16le"
|
|
rawAudioFormat = "s16le"
|
|
rawAudioSampleRate = 48_000
|
|
)
|
|
|
|
// progressReader is a reader that prints progress logs as it reads.
|
|
type progressReader struct {
|
|
io.Reader
|
|
label string
|
|
total, exp int
|
|
}
|
|
|
|
func (pw *progressReader) Read(p []byte) (int, error) {
|
|
n, err := pw.Reader.Read(p)
|
|
pw.total += n
|
|
|
|
log.Printf("[ProgressReader] [%s] Read %d of %d (%.02f%%) bytes from the provided reader", pw.label, pw.total, pw.exp, (float32(pw.total)/float32(pw.exp))*100.0)
|
|
|
|
return n, err
|
|
}
|
|
|
|
// S3Client stubs the AWS S3 service client.
|
|
type S3Client interface {
|
|
CreateMultipartUpload(context.Context, *s3.CreateMultipartUploadInput, ...func(*s3.Options)) (*s3.CreateMultipartUploadOutput, error)
|
|
UploadPart(context.Context, *s3.UploadPartInput, ...func(*s3.Options)) (*s3.UploadPartOutput, error)
|
|
AbortMultipartUpload(ctx context.Context, params *s3.AbortMultipartUploadInput, optFns ...func(*s3.Options)) (*s3.AbortMultipartUploadOutput, error)
|
|
CompleteMultipartUpload(context.Context, *s3.CompleteMultipartUploadInput, ...func(*s3.Options)) (*s3.CompleteMultipartUploadOutput, error)
|
|
}
|
|
|
|
// YoutubeClient stubs the youtube.Client client.
|
|
type YoutubeClient interface {
|
|
GetVideoContext(context.Context, string) (*youtubev2.Video, error)
|
|
GetStreamContext(context.Context, *youtubev2.Video, *youtubev2.Format) (io.ReadCloser, int64, error)
|
|
}
|
|
|
|
type MediaSet2 struct {
|
|
id string
|
|
}
|
|
|
|
func NewMediaSet2(id string) *MediaSet2 {
|
|
return &MediaSet2{
|
|
id: id,
|
|
}
|
|
}
|
|
|
|
// VideoFetchService fetches a video via an io.Reader.
|
|
type VideoFetchService struct {
|
|
youtube YoutubeClient
|
|
s3 S3Client
|
|
}
|
|
|
|
func NewVideoFetchService(youtubeClient YoutubeClient, s3Client S3Client) *VideoFetchService {
|
|
return &VideoFetchService{
|
|
youtube: youtubeClient,
|
|
s3: s3Client,
|
|
}
|
|
}
|
|
|
|
// Fetch handles the entire process to fetch and process the audio and video
|
|
// parts of a MediaSet.
|
|
func (s *VideoFetchService) Fetch(ctx context.Context, id string) (*MediaSet, error) {
|
|
video, err := s.youtube.GetVideoContext(ctx, id)
|
|
if err != nil {
|
|
return nil, fmt.Errorf("error fetching video: %v", err)
|
|
}
|
|
|
|
if len(video.Formats) == 0 {
|
|
return nil, errors.New("no format available")
|
|
}
|
|
|
|
// just the audio for now
|
|
|
|
// grab an audio stream from youtube
|
|
// TODO: avoid possible panic
|
|
format := sortAudio(video.Formats)[0]
|
|
|
|
sampleRate, err := strconv.Atoi(format.AudioSampleRate)
|
|
if err != nil {
|
|
return nil, fmt.Errorf("invalid samplerate: %s", format.AudioSampleRate)
|
|
}
|
|
|
|
approxDurationMsecs, err := strconv.Atoi(format.ApproxDurationMs)
|
|
if err != nil {
|
|
return nil, fmt.Errorf("could not parse audio duration: %s", err)
|
|
}
|
|
approxDuration := time.Duration(approxDurationMsecs) * time.Millisecond
|
|
approxFrames := int64(approxDuration/time.Second) * int64(sampleRate)
|
|
|
|
mediaSet := MediaSet{
|
|
ID: id,
|
|
Audio: Audio{
|
|
// we need to decode it to be able to know bytes and frames exactly
|
|
ApproxFrames: approxFrames,
|
|
Channels: format.AudioChannels,
|
|
SampleRate: sampleRate,
|
|
},
|
|
}
|
|
|
|
return &mediaSet, nil
|
|
}
|
|
|
|
type PeakIterator interface {
|
|
Next() (FetchPeaksProgress, error)
|
|
Close() error
|
|
}
|
|
|
|
func (s *VideoFetchService) FetchPeaks(ctx context.Context, id string) (PeakIterator, error) {
|
|
mediaSet := NewMediaSet(id)
|
|
if !mediaSet.Exists() {
|
|
// TODO check if audio uploaded already, don't bother again
|
|
return nil, errors.New("no media set found")
|
|
}
|
|
|
|
if err := mediaSet.Load(); err != nil {
|
|
return nil, fmt.Errorf("error loading media set: %v", err)
|
|
}
|
|
|
|
video, err := s.youtube.GetVideoContext(ctx, id)
|
|
if err != nil {
|
|
return nil, fmt.Errorf("error fetching video: %v", err)
|
|
}
|
|
|
|
if len(video.Formats) == 0 {
|
|
return nil, errors.New("no format available")
|
|
}
|
|
|
|
// TODO: avoid possible panic
|
|
format := sortAudio(video.Formats)[0]
|
|
|
|
stream, _, err := s.youtube.GetStreamContext(ctx, video, &format)
|
|
if err != nil {
|
|
return nil, fmt.Errorf("error fetching stream: %v", err)
|
|
}
|
|
|
|
// wrap it in a progress reader
|
|
progressStream := &progressReader{Reader: stream, label: "audio", exp: int(format.ContentLength)}
|
|
|
|
ffmpegReader, err := newFfmpegReader(ctx, progressStream, "-i", "-", "-f", rawAudioFormat, "-ar", strconv.Itoa(rawAudioSampleRate), "-acodec", rawAudioCodec, "-")
|
|
if err != nil {
|
|
return nil, fmt.Errorf("error creating ffmpegreader: %v", err)
|
|
}
|
|
|
|
// set up uploader, this is writer 1
|
|
uploader, err := newMultipartUploadWriter(
|
|
ctx,
|
|
s.s3,
|
|
s3Bucket,
|
|
fmt.Sprintf("media_sets/%s/audio.webm", id),
|
|
"application/octet-stream",
|
|
)
|
|
if err != nil {
|
|
return nil, fmt.Errorf("error creating uploader: %v", err)
|
|
}
|
|
|
|
peakIterator := newFetchPeaksIterator(
|
|
mediaSet.Audio.ApproxFrames,
|
|
format.AudioChannels,
|
|
100,
|
|
)
|
|
|
|
state := fetchPeaksState{
|
|
fetchPeaksIterator: peakIterator,
|
|
ffmpegReader: ffmpegReader,
|
|
uploader: uploader,
|
|
}
|
|
go state.run(ctx) // pass ctx?
|
|
|
|
return &state, nil
|
|
}
|
|
|
|
type fetchPeaksState struct {
|
|
*fetchPeaksIterator
|
|
|
|
ffmpegReader *ffmpegReader
|
|
uploader *multipartUploadWriter
|
|
err error
|
|
}
|
|
|
|
// run copies the audio data from ffmpeg, waits for termination and then cleans
|
|
// up appropriately.
|
|
func (s *fetchPeaksState) run(ctx context.Context) {
|
|
mw := io.MultiWriter(s.fetchPeaksIterator, s.uploader)
|
|
done := make(chan error)
|
|
var err error
|
|
|
|
go func() {
|
|
_, copyErr := io.Copy(mw, s.ffmpegReader)
|
|
done <- copyErr
|
|
}()
|
|
|
|
outer:
|
|
for {
|
|
select {
|
|
case <-ctx.Done():
|
|
err = ctx.Err()
|
|
break outer
|
|
case err = <-done:
|
|
break outer
|
|
}
|
|
}
|
|
|
|
if readerErr := s.ffmpegReader.Close(); readerErr != nil {
|
|
log.Printf("error closing ffmpegReader: %v", readerErr)
|
|
if err == nil {
|
|
err = readerErr
|
|
}
|
|
}
|
|
|
|
if err == nil {
|
|
if uploaderErr := s.uploader.Complete(); uploaderErr != nil {
|
|
log.Printf("error closing uploader: %v", uploaderErr)
|
|
err = uploaderErr
|
|
}
|
|
}
|
|
|
|
if err != nil {
|
|
newCtx, cancel := context.WithTimeout(context.Background(), time.Second*5)
|
|
defer cancel()
|
|
|
|
if abortUploadErr := s.uploader.Abort(newCtx); abortUploadErr != nil {
|
|
log.Printf("error aborting uploader: %v", abortUploadErr)
|
|
}
|
|
s.Abort(err)
|
|
return
|
|
}
|
|
|
|
if iterErr := s.Close(); iterErr != nil {
|
|
log.Printf("error closing peak iterator: %v", iterErr)
|
|
}
|
|
}
|
|
|
|
type ffmpegReader struct {
|
|
io.ReadCloser
|
|
|
|
cmd *exec.Cmd
|
|
}
|
|
|
|
func newFfmpegReader(ctx context.Context, input io.Reader, arg ...string) (*ffmpegReader, error) {
|
|
var stdErr bytes.Buffer
|
|
|
|
cmd := exec.CommandContext(ctx, "ffmpeg", arg...)
|
|
cmd.Stdin = input
|
|
cmd.Stderr = &stdErr // TODO: fix error handling
|
|
|
|
r, err := cmd.StdoutPipe()
|
|
if err != nil {
|
|
return nil, fmt.Errorf("error creating pipe: %v", err)
|
|
}
|
|
|
|
if err := cmd.Start(); err != nil {
|
|
return nil, fmt.Errorf("error starting ffmpeg: %v", err)
|
|
}
|
|
|
|
return &ffmpegReader{ReadCloser: r, cmd: cmd}, nil
|
|
}
|
|
|
|
func (r *ffmpegReader) Close() error {
|
|
state, err := r.cmd.Process.Wait()
|
|
|
|
if err != nil {
|
|
return fmt.Errorf("error returned from process: %v", err)
|
|
}
|
|
|
|
if state.ExitCode() != 0 {
|
|
return fmt.Errorf("command exited with code %d", state.ExitCode())
|
|
}
|
|
|
|
log.Println("returning from ffmpegreader.close")
|
|
return nil
|
|
}
|
|
|
|
// sortAudio returns the provided formats ordered in descending preferred
|
|
// order. The ideal candidate is opus-encoded stereo audio in a webm container,
|
|
// with the lowest available bitrate.
|
|
func sortAudio(inFormats youtubev2.FormatList) youtubev2.FormatList {
|
|
var formats youtubev2.FormatList
|
|
for _, format := range inFormats {
|
|
if format.FPS == 0 && format.AudioChannels > 0 {
|
|
formats = append(formats, format)
|
|
}
|
|
}
|
|
sort.SliceStable(formats, func(i, j int) bool {
|
|
isOpusI := strings.Contains(formats[i].MimeType, "opus")
|
|
isOpusJ := strings.Contains(formats[j].MimeType, "opus")
|
|
if isOpusI && isOpusJ {
|
|
isStereoI := formats[i].AudioChannels == 2
|
|
isStereoJ := formats[j].AudioChannels == 2
|
|
if isStereoI && isStereoJ {
|
|
return formats[i].ContentLength < formats[j].ContentLength
|
|
}
|
|
return isStereoI
|
|
}
|
|
return isOpusI
|
|
})
|
|
return formats
|
|
}
|