av/container/mts/encoder.go

/*
NAME
  encoder.go

AUTHOR
  Dan Kortschak <dan@ausocean.org>
  Saxon Nelson-Milton <saxon@ausocean.org>

LICENSE
  encoder.go is Copyright (C) 2017-2018 the Australian Ocean Lab (AusOcean)

  It is free software: you can redistribute it and/or modify them
  under the terms of the GNU General Public License as published by the
  Free Software Foundation, either version 3 of the License, or (at your
  option) any later version.

  It is distributed in the hope that it will be useful, but WITHOUT
  ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
  FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
 for more details.

  You should have received a copy of the GNU General Public License
  along with revid in gpl.txt. If not, see http://www.gnu.org/licenses.
*/

package mts

import (
	"fmt"
	"io"
	"strconv"
	"time"

	"bitbucket.org/ausocean/av/codec/h264"
	"bitbucket.org/ausocean/av/codec/h264/h264dec"
	"bitbucket.org/ausocean/av/container/mts/meta"
	"bitbucket.org/ausocean/av/container/mts/pes"
	"bitbucket.org/ausocean/av/container/mts/psi"
	"bitbucket.org/ausocean/utils/realtime"
)

// Stream IDs as per ITU-T Rec. H.222.0 / ISO/IEC 13818-1 [1], tables 2-22 and 2-34.
const (
	H264ID        = 27
	H265ID        = 36
	MJPEGID       = 28
	audioStreamID = 0xc0 // ADPCM audio stream ID.
)

// Constants used to communicate which media codec will be packetized.
const (
	EncodeH264 = iota
	EncodeH265
	EncodeMJPEG
	EncodeAudio
)

// The program IDs we assign to different types of media.
const (
	VideoPid = 256
	AudioPid = 210
)

// Time-related constants.
const (
	// ptsOffset is the offset added to the clock to determine
	// the current presentation timestamp.
	ptsOffset = 700 * time.Millisecond

	// PCRFrequency is the base Program Clock Reference frequency in Hz.
	PCRFrequency = 90000

	// PTSFrequency is the presentation timestamp frequency in Hz.
	PTSFrequency = 90000

	// MaxPTS is the largest PTS value (i.e., for a 33-bit unsigned integer).
	MaxPTS = (1 << 33) - 1
)

// If we are not using NAL based PSI intervals then we will send PSI every 7 packets.
const psiSendCount = 7

const (
	hasPayload         = 0x1
	hasAdaptationField = 0x2
)

const (
	hasDTS = 0x1
	hasPTS = 0x2
)

// Some common manifestations of PSI.
var (
	// StandardPAT is a minimal PAT.
	StandardPAT = psi.PSI{
		Pf:  0x00,
		Tid: 0x00,
		Ssi: true,
		Pb:  false,
		Sl:  0x0d,
		Tss: &psi.TSS{
			Tide: 0x01,
			V:    0,
			Cni:  true,
			Sn:   0,
			Lsn:  0,
			Sd: &psi.PAT{
				Pn:    0x01,
				Pmpid: 0x1000,
			},
		},
	}

	// Base PMT is a minimal PMT without specific data.
	BasePMT = psi.PSI{
		Pf:  0x00,
		Tid: 0x02,
		Ssi: true,
		Sl:  0x12,
		Tss: &psi.TSS{
			Tide: 0x01,
			V:    0,
			Cni:  true,
			Sn:   0,
			Lsn:  0,
		},
	}
)

// Meta allows addition of metadata to encoded mts from outside of this pkg.
// See meta pkg for usage.
//
// TODO: make this not global.
var Meta *meta.Data

// This will help us obtain a realtime for timestamp meta encoding.
var RealTime = realtime.NewRealTime()

var (
	patTable = StandardPAT.Bytes()
	pmtTable []byte
)

// Encoder encapsulates properties of an MPEG-TS generator.
type Encoder struct {
	dst io.WriteCloser

	clock       time.Duration
	lastTime    time.Time
	writePeriod time.Duration
	ptsOffset   time.Duration
	tsSpace     [PacketSize]byte
	pesSpace    [pes.MaxPesSize]byte

	continuity map[uint16]byte

	nalBasedPSI  bool
	pktCount     int
	psiSendCount int
	mediaPid     uint16
	streamID     byte
}

// NewEncoder returns an Encoder with the specified media type and rate eg. if a video stream
// calls write for every frame, the rate will be the frame rate of the video.
func NewEncoder(dst io.WriteCloser, rate float64, mediaType int, options ...func(*Encoder) error) (*Encoder, error) {
	var mPID uint16
	var sID byte
	nbp := true
	switch mediaType {
	case EncodeAudio:
		mPID = AudioPid
		sID = audioStreamID
		nbp = false
	case EncodeH265:
		mPID = VideoPid
		sID = H265ID
	case EncodeH264:
		mPID = VideoPid
		sID = H264ID
	case EncodeMJPEG:
		mPID = VideoPid
		sID = MJPEGID
	}

	pmt := BasePMT
	pmt.Tss.Sd = &psi.PMT{
		Pcrpid: 0x0100,
		Pil:    0,
		Essd: &psi.ESSD{
			St:   byte(sID),
			Epid: mPID,
			Esil: 0x00,
		},
	}
	pmtTable = pmt.Bytes()

	e := &Encoder{
		dst: dst,

		writePeriod: time.Duration(float64(time.Second) / rate),
		ptsOffset:   ptsOffset,

		nalBasedPSI: nbp,

		pktCount: 8,

		mediaPid: mPID,
		streamID: sID,

		continuity: map[uint16]byte{
			PatPid: 0,
			PmtPid: 0,
			mPID:   0,
		},
	}

	for _, option := range options {
		err := option(e)
		if err != nil {
			return nil, fmt.Errorf("option failed with error: %w", err)
		}
	}
	return e, nil
}

// PacketBasedPSI is an option that can be passed to NewEncoder to select
// packet based PSI writing, i.e. PSI are written to the destination every
// sendCount packets.
func PacketBasedPSI(sendCount int) func(*Encoder) error {
	return func(e *Encoder) error {
		e.nalBasedPSI = false
		e.psiSendCount = sendCount
		e.pktCount = e.psiSendCount
		return nil
	}
}

// Write implements io.Writer. Write takes raw video or audio data and encodes into MPEG-TS,
// then sending it to the encoder's io.Writer destination.
func (e *Encoder) Write(data []byte) (int, error) {
	if e.nalBasedPSI {
		nalType, err := h264.NALType(data)
		if err != nil {
			return 0, fmt.Errorf("could not get type from NAL unit, failed with error: %w", err)
		}

		if nalType == h264dec.NALTypeSPS {
			err := e.writePSI()
			if err != nil {
				return 0, err
			}
		}
	} else if e.pktCount >= e.psiSendCount {
		e.pktCount = 0
		err := e.writePSI()
		if err != nil {
			return 0, err
		}
	}

	// Prepare PES data.
	pesPkt := pes.Packet{
		StreamID:     e.streamID,
		PDI:          hasPTS,
		PTS:          e.pts(),
		Data:         data,
		HeaderLength: 5,
	}

	buf := pesPkt.Bytes(e.pesSpace[:pes.MaxPesSize])

	pusi := true
	for len(buf) != 0 {
		pkt := Packet{
			PUSI: pusi,
			PID:  uint16(e.mediaPid),
			RAI:  pusi,
			CC:   e.ccFor(e.mediaPid),
			AFC:  hasAdaptationField | hasPayload,
			PCRF: pusi,
		}
		n := pkt.FillPayload(buf)
		buf = buf[n:]

		if pusi {
			// If the packet has a Payload Unit Start Indicator
			// flag set then we need to write a PCR.
			pkt.PCR = e.pcr()
			pusi = false
		}
		_, err := e.dst.Write(pkt.Bytes(e.tsSpace[:PacketSize]))
		if err != nil {
			return len(data), err
		}
		e.pktCount++
	}

	e.tick()

	return len(data), nil
}

// writePSI creates MPEG-TS with pat and pmt tables - with pmt table having updated
// location and time data.
func (e *Encoder) writePSI() error {
	// Write PAT.
	patPkt := Packet{
		PUSI:    true,
		PID:     PatPid,
		CC:      e.ccFor(PatPid),
		AFC:     hasPayload,
		Payload: psi.AddPadding(patTable),
	}
	_, err := e.dst.Write(patPkt.Bytes(e.tsSpace[:PacketSize]))
	if err != nil {
		return err
	}
	e.pktCount++
	pmtTable, err = updateMeta(pmtTable)
	if err != nil {
		return err
	}

	// Create mts packet from pmt table.
	pmtPkt := Packet{
		PUSI:    true,
		PID:     PmtPid,
		CC:      e.ccFor(PmtPid),
		AFC:     hasPayload,
		Payload: psi.AddPadding(pmtTable),
	}
	_, err = e.dst.Write(pmtPkt.Bytes(e.tsSpace[:PacketSize]))
	if err != nil {
		return err
	}
	e.pktCount++
	return nil
}

// tick advances the clock one frame interval.
func (e *Encoder) tick() {
	e.clock += e.writePeriod
}

// pts retuns the current presentation timestamp.
func (e *Encoder) pts() uint64 {
	return uint64((e.clock + e.ptsOffset).Seconds() * PTSFrequency)
}

// pcr returns the current program clock reference.
func (e *Encoder) pcr() uint64 {
	return uint64(e.clock.Seconds() * PCRFrequency)
}

// ccFor returns the next continuity counter for pid.
func (e *Encoder) ccFor(pid uint16) byte {
	cc := e.continuity[pid]
	const continuityCounterMask = 0xf
	e.continuity[pid] = (cc + 1) & continuityCounterMask
	return cc
}

// updateMeta adds/updates a metaData descriptor in the given psi bytes using data
// contained in the global Meta struct.
func updateMeta(b []byte) ([]byte, error) {
	p := psi.PSIBytes(b)
	if RealTime.IsSet() {
		Meta.Add("ts", strconv.Itoa(int(RealTime.Get().Unix())))
	}
	err := p.AddDescriptor(psi.MetadataTag, Meta.Encode())
	return []byte(p), err
}

func (e *Encoder) Close() error {
	return e.dst.Close()
}