sort samples by DTS

go.mod

@@ -9,7 +9,7 @@ require (
 	github.com/alecthomas/kong v0.9.0
 	github.com/bluenviron/gohlslib v1.3.0
 	github.com/bluenviron/gortsplib/v4 v4.8.0
-	github.com/bluenviron/mediacommon v1.9.3-0.20240408205450-b2b6c85fb0e0
+	github.com/bluenviron/mediacommon v1.9.2
 	github.com/datarhei/gosrt v0.6.0
 	github.com/fsnotify/fsnotify v1.7.0
 	github.com/gin-gonic/gin v1.9.1

go.sum

@@ -24,8 +24,8 @@ github.com/bluenviron/gohlslib v1.3.0 h1:I9t1Nba6VJKg5rLoXSzQFPkZZYBUwBqCU2Divp0
 github.com/bluenviron/gohlslib v1.3.0/go.mod h1:wD8ysO6HB90d17sxoIQXGHINo2KYj/mZirMnPtKLJZQ=
 github.com/bluenviron/gortsplib/v4 v4.8.0 h1:nvFp6rHALcSep3G9uBFI0uogS9stVZLNq/92TzGZdQg=
 github.com/bluenviron/gortsplib/v4 v4.8.0/go.mod h1:+d+veuyvhvikUNp0GRQkk6fEbd/DtcXNidMRm7FQRaA=
-github.com/bluenviron/mediacommon v1.9.3-0.20240408205450-b2b6c85fb0e0 h1:GkvT15WaWrShOkl+XkYtBWoHaf7hacT/h8qQBh0UFQg=
-github.com/bluenviron/mediacommon v1.9.3-0.20240408205450-b2b6c85fb0e0/go.mod h1:0z/KHiSTlaAB8FoyW+mYulZRG70Kupcy/0yZZkgLe5M=
+github.com/bluenviron/mediacommon v1.9.2 h1:EHcvoC5YMXRcFE010bTNf07ZiSlB/e/AdZyG7GsEYN0=
+github.com/bluenviron/mediacommon v1.9.2/go.mod h1:lt8V+wMyPw8C69HAqDWV5tsAwzN9u2Z+ca8B6C//+n0=
 github.com/bytedance/sonic v1.5.0/go.mod h1:ED5hyg4y6t3/9Ku1R6dU/4KyJ48DZ4jPhfY1O2AihPM=
 github.com/bytedance/sonic v1.9.1 h1:6iJ6NqdoxCDr6mbY8h18oSO+cShGSMRGCEo7F2h0x8s=
 github.com/bytedance/sonic v1.9.1/go.mod h1:i736AoUSYt75HyZLoJW9ERYxcy6eaN6h4BZXU064P/U=

internal/playback/mp4/mp4_writer.go

@@ -0,0 +1,83 @@
+package mp4
+
+import (
+	"io"
+
+	"github.com/abema/go-mp4"
+)
+
+type mp4Writer struct {
+	w *mp4.Writer
+}
+
+func newMP4Writer(w io.WriteSeeker) *mp4Writer {
+	return &mp4Writer{
+		w: mp4.NewWriter(w),
+	}
+}
+
+func (w *mp4Writer) writeBoxStart(box mp4.IImmutableBox) (int, error) {
+	bi := &mp4.BoxInfo{
+		Type: box.GetType(),
+	}
+	var err error
+	bi, err = w.w.StartBox(bi)
+	if err != nil {
+		return 0, err
+	}
+
+	_, err = mp4.Marshal(w.w, box, mp4.Context{})
+	if err != nil {
+		return 0, err
+	}
+
+	return int(bi.Offset), nil
+}
+
+func (w *mp4Writer) writeBoxEnd() error {
+	_, err := w.w.EndBox()
+	return err
+}
+
+func (w *mp4Writer) writeBox(box mp4.IImmutableBox) (int, error) {
+	off, err := w.writeBoxStart(box)
+	if err != nil {
+		return 0, err
+	}
+
+	err = w.writeBoxEnd()
+	if err != nil {
+		return 0, err
+	}
+
+	return off, nil
+}
+
+func (w *mp4Writer) rewriteBox(off int, box mp4.IImmutableBox) error {
+	prevOff, err := w.w.Seek(0, io.SeekCurrent)
+	if err != nil {
+		return err
+	}
+
+	_, err = w.w.Seek(int64(off), io.SeekStart)
+	if err != nil {
+		return err
+	}
+
+	_, err = w.writeBoxStart(box)
+	if err != nil {
+		return err
+	}
+
+	err = w.writeBoxEnd()
+	if err != nil {
+		return err
+	}
+
+	_, err = w.w.Seek(prevOff, io.SeekStart)
+	if err != nil {
+		return err
+	}
+
+	return nil
+}

internal/playback/mp4/presentation.go

@@ -0,0 +1,208 @@
+// Package mp4 contains a MP4 muxer.
+package mp4
+
+import (
+	"io"
+	"time"
+
+	"github.com/abema/go-mp4"
+	"github.com/bluenviron/mediacommon/pkg/formats/fmp4/seekablebuffer"
+)
+
+const (
+	globalTimescale = 1000
+)
+
+func durationMp4ToGo(v int64, timeScale uint32) time.Duration {
+	timeScale64 := int64(timeScale)
+	secs := v / timeScale64
+	dec := v % timeScale64
+	return time.Duration(secs)*time.Second + time.Duration(dec)*time.Second/time.Duration(timeScale64)
+}
+
+// Presentation is timed sequence of video/audio samples.
+type Presentation struct {
+	Tracks []*Track
+}
+
+// Marshal encodes a Presentation.
+func (p *Presentation) Marshal(w io.Writer) error {
+	/*
+		|ftyp|
+		|moov|
+		|    |mvhd|
+		|    |trak|
+		|    |trak|
+		|    |....|
+		|mdat|
+	*/
+
+	dataSize, sortedSamples := p.sortSamples()
+
+	err := p.marshalFtypAndMoov(w)
+	if err != nil {
+		return err
+	}
+
+	return p.marshalMdat(w, dataSize, sortedSamples)
+}
+
+func (p *Presentation) sortSamples() (uint32, []*Sample) {
+	sampleCount := 0
+	for _, track := range p.Tracks {
+		sampleCount += len(track.Samples)
+	}
+
+	processedSamples := make([]int, len(p.Tracks))
+	elapsed := make([]int64, len(p.Tracks))
+	offset := uint32(0)
+	sortedSamples := make([]*Sample, sampleCount)
+	pos := 0
+
+	for i, track := range p.Tracks {
+		elapsed[i] = int64(track.TimeOffset)
+	}
+
+	for {
+		bestTrack := -1
+		var bestElapsed time.Duration
+
+		for i, track := range p.Tracks {
+			if processedSamples[i] < len(track.Samples) {
+				elapsedGo := durationMp4ToGo(elapsed[i], track.TimeScale)
+
+				if bestTrack == -1 || elapsedGo < bestElapsed {
+					bestTrack = i
+					bestElapsed = elapsedGo
+				}
+			}
+		}
+
+		if bestTrack == -1 {
+			break
+		}
+
+		sample := p.Tracks[bestTrack].Samples[processedSamples[bestTrack]]
+		sample.offset = offset
+
+		processedSamples[bestTrack]++
+		elapsed[bestTrack] += int64(sample.Duration)
+		offset += sample.PayloadSize
+		sortedSamples[pos] = sample
+		pos++
+	}
+
+	return offset, sortedSamples
+}
+
+func (p *Presentation) marshalFtypAndMoov(w io.Writer) error {
+	var outBuf seekablebuffer.Buffer
+	mw := newMP4Writer(&outBuf)
+
+	_, err := mw.writeBox(&mp4.Ftyp{ // <ftyp/>
+		MajorBrand:   [4]byte{'i', 's', 'o', 'm'},
+		MinorVersion: 1,
+		CompatibleBrands: []mp4.CompatibleBrandElem{
+			{CompatibleBrand: [4]byte{'i', 's', 'o', 'm'}},
+			{CompatibleBrand: [4]byte{'i', 's', 'o', '2'}},
+			{CompatibleBrand: [4]byte{'m', 'p', '4', '1'}},
+			{CompatibleBrand: [4]byte{'m', 'p', '4', '2'}},
+		},
+	})
+	if err != nil {
+		return err
+	}
+
+	_, err = mw.writeBoxStart(&mp4.Moov{}) // <moov>
+	if err != nil {
+		return err
+	}
+
+	mvhd := &mp4.Mvhd{ // <mvhd/>
+		Timescale:   globalTimescale,
+		Rate:        65536,
+		Volume:      256,
+		Matrix:      [9]int32{0x00010000, 0, 0, 0, 0x00010000, 0, 0, 0, 0x40000000},
+		NextTrackID: uint32(len(p.Tracks) + 1),
+	}
+	mvhdOffset, err := mw.writeBox(mvhd)
+	if err != nil {
+		return err
+	}
+
+	stcos := make([]*mp4.Stco, len(p.Tracks))
+	stcosOffsets := make([]int, len(p.Tracks))
+
+	for i, track := range p.Tracks {
+		res, err := track.marshal(mw)
+		if err != nil {
+			return err
+		}
+
+		stcos[i] = res.stco
+		stcosOffsets[i] = res.stcoOffset
+
+		if res.presentationDuration > mvhd.DurationV0 {
+			mvhd.DurationV0 = res.presentationDuration
+		}
+	}
+
+	err = mw.rewriteBox(mvhdOffset, mvhd)
+	if err != nil {
+		return err
+	}
+
+	err = mw.writeBoxEnd() // </moov>
+	if err != nil {
+		return err
+	}
+
+	moovEndOffset, err := outBuf.Seek(0, io.SeekCurrent)
+	if err != nil {
+		return err
+	}
+
+	dataOffset := moovEndOffset + 8
+
+	for i := range p.Tracks {
+		for j := range stcos[i].ChunkOffset {
+			stcos[i].ChunkOffset[j] += uint32(dataOffset)
+		}
+
+		err = mw.rewriteBox(stcosOffsets[i], stcos[i])
+		if err != nil {
+			return err
+		}
+	}
+
+	_, err = w.Write(outBuf.Bytes())
+	return err
+}
+
+func (p *Presentation) marshalMdat(w io.Writer, dataSize uint32, sortedSamples []*Sample) error {
+	mdatSize := uint32(8) + dataSize
+
+	_, err := w.Write([]byte{byte(mdatSize >> 24), byte(mdatSize >> 16), byte(mdatSize >> 8), byte(mdatSize)})
+	if err != nil {
+		return err
+	}
+
+	_, err = w.Write([]byte{'m', 'd', 'a', 't'})
+	if err != nil {
+		return err
+	}
+
+	for _, sa := range sortedSamples {
+		pl, err := sa.GetPayload()
+		if err != nil {
+			return err
+		}
+
+		_, err = w.Write(pl)
+		if err != nil {
+			return err
+		}
+	}
+
+	return nil
+}

internal/playback/mp4/sample.go

@@ -0,0 +1,12 @@
+package mp4
+
+// Sample is a sample of a Track.
+type Sample struct {
+	Duration        uint32
+	PTSOffset       int32
+	IsNonSyncSample bool
+	PayloadSize     uint32
+	GetPayload      func() ([]byte, error)
+
+	offset uint32 // filled by sortSamples
+}