av/codec/h264/lex.go

/*
NAME
  lex.go

DESCRIPTION
  lex.go provides a lexer to lex h264 bytestream into access units.

AUTHOR
  Dan Kortschak <dan@ausocean.org>

LICENSE
  lex.go is Copyright (C) 2017 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.
*/

// lex.go provides a lexer to lex h264 bytestream into access units.

package h264

import (
	"bytes"
	"encoding/binary"
	"fmt"
	"io"
	"time"

	"bitbucket.org/ausocean/av/codec/codecutil"
	"bitbucket.org/ausocean/av/protocol/rtp"
)

// NAL types.
const (
	// Single nal units bounds.
	typeSingleNALULowBound  = 1
	typeSingleNALUHighBound = 23

	// Single-time aggregation packets.
	typeSTAPA = 24
	typeSTAPB = 25

	// Multi-time aggregation packets.
	typeMTAP16 = 26
	typeMTAP24 = 27

	// Fragmentation packets.
	typeFUA = 28
	typeFUB = 29
)

var noDelay = make(chan time.Time)

func init() {
	close(noDelay)
}

var h264Prefix = [...]byte{0x00, 0x00, 0x01, 0x09, 0xf0}

// LexFromBytestream lexes H.264 NAL units read from src into separate writes
// to dst with successive writes being performed not earlier than the specified
// delay. NAL units are split after type 1 (Coded slice of a non-IDR picture), 5
// (Coded slice of a IDR picture) and 8 (Picture parameter set).
func LexFromBytestream(dst io.Writer, src io.Reader, delay time.Duration) error {
	var tick <-chan time.Time
	if delay == 0 {
		tick = noDelay
	} else {
		ticker := time.NewTicker(delay)
		defer ticker.Stop()
		tick = ticker.C
	}

	const bufSize = 8 << 10

	c := codecutil.NewByteScanner(src, make([]byte, 4<<10)) // Standard file buffer size.

	buf := make([]byte, len(h264Prefix), bufSize)
	copy(buf, h264Prefix[:])
	writeOut := false
outer:
	for {
		var b byte
		var err error
		buf, b, err = c.ScanUntil(buf, 0x00)
		if err != nil {
			if err != io.EOF {
				return err
			}
			break
		}

		for n := 1; b == 0x0 && n < 4; n++ {
			b, err = c.ReadByte()
			if err != nil {
				if err != io.EOF {
					return err
				}
				break outer
			}
			buf = append(buf, b)

			if b != 0x1 || (n != 2 && n != 3) {
				continue
			}

			if writeOut {
				<-tick
				_, err := dst.Write(buf[:len(buf)-(n+1)])
				if err != nil {
					return err
				}
				buf = make([]byte, len(h264Prefix)+n, bufSize)
				copy(buf, h264Prefix[:])
				buf = append(buf, 1)
				writeOut = false
			}

			b, err = c.ReadByte()
			if err != nil {
				if err != io.EOF {
					return err
				}
				break outer
			}
			buf = append(buf, b)

			// http://www.itu.int/rec/dologin_pub.asp?lang=e&id=T-REC-H.264-200305-S!!PDF-E&type=items
			// Table 7-1 NAL unit type codes
			const (
				nonIdrPic   = 1
				idrPic      = 5
				suppEnhInfo = 6
				paramSet    = 8
			)
			switch nalTyp := b & 0x1f; nalTyp {
			case nonIdrPic, idrPic, paramSet, suppEnhInfo:
				writeOut = true
			}
		}
	}
	if len(buf) == len(h264Prefix) {
		return nil
	}
	<-tick
	_, err := dst.Write(buf)
	return err
}

// Buffer sizes.
const (
	maxAUSize  = 100000 // Max access unit size in bytes.
	maxRTPSize = 1500   // Max ethernet transmission unit in bytes.
)

// RTPLexer is a lexer for lexing H264 from RTP packets.
type RTPLexer struct {
	buf  *bytes.Buffer // Holds the current access unit.
	frag bool          // Indicates if we're currently dealing with a fragmentation packet.
}

// NewRTPLexer returns a new RTPLexer.
func NewRTPLexer() *RTPLexer {
	return &RTPLexer{
		buf: bytes.NewBuffer(make([]byte, 0, maxAUSize))}
}

// Lex extracts H264 access units from an RTP stream. This function
// expects that each read from src will provide a single RTP packet.
func (l *RTPLexer) Lex(dst io.Writer, src io.Reader, delay time.Duration) error {
	buf := make([]byte, maxRTPSize)
	for {
		n, err := src.Read(buf)
		switch err {
		case nil: // Do nothing.
		case io.EOF:
			return nil
		default:
			return fmt.Errorf("source read error: %v\n", err)
		}

		// Get payload from RTP packet.
		payload, err := rtp.Payload(buf[:n])
		if err != nil {
			return fmt.Errorf("could not get RTP payload, failed with err: %v\n", err)
		}

		nalType := payload[0] & 0x1f

		// If not currently fragmented then we ignore current write.
		if l.frag && nalType != typeFUA {
			l.buf.Reset()
			l.frag = false
			continue
		}

		if nalType >= typeSingleNALULowBound && nalType <= typeSingleNALUHighBound {
			l.writeWithPrefix(payload)
		} else {
			switch nalType {
			case typeSTAPA:
				l.handleSTAPA(payload)
			case typeFUA:
				l.handleFUA(payload)
			case typeSTAPB:
				panic("STAP-B type unsupported")
			case typeMTAP16:
				panic("MTAP16 type unsupported")
			case typeMTAP24:
				panic("MTAP24 type unsupported")
			case typeFUB:
				panic("FU-B type unsupported")
			default:
				panic("unsupported type")
			}
		}

		markerIsSet, err := rtp.Marker(buf[:n])
		if err != nil {
			return fmt.Errorf("could not get marker bit, failed with err: %v\n", err)
		}

		if markerIsSet {
			l.buf.WriteTo(dst)
			l.buf.Reset()
		}
	}
	return nil
}

// handleSTAP parses NAL units from an aggregation packet and writes
// them to the Lexers buffer buf.
func (l *RTPLexer) handleSTAPA(d []byte) {
	for i := 1; i < len(d); {
		size := int(binary.BigEndian.Uint16(d[i:]))

		// Skip over NAL unit size.
		const sizeOfFieldLen = 2
		i += sizeOfFieldLen

		// Get the NALU.
		nalu := d[i : i+size]
		i += size
		l.writeWithPrefix(nalu)
	}
}

// handleFragmentation parses NAL units from fragmentation packets and writes
// them to the Lexer's buf.
func (l *RTPLexer) handleFUA(d []byte) {
	// Get start and end indiciators from FU header.
	const FUHeadIdx = 1
	start := d[FUHeadIdx]&0x80 != 0
	end := d[FUHeadIdx]&0x40 != 0

	// If start, form new header, skip FU indicator only and set first byte to
	// new header. Otherwise, skip over both FU indicator and FU header.
	if start {
		const FUIndicatorIdx = 0
		newHead := (d[FUIndicatorIdx] & 0xe0) | (d[FUHeadIdx] & 0x1f)
		d = d[1:]
		d[0] = newHead
	} else {
		d = d[2:]
	}

	if start {
		if end {
			panic("bad fragmentation packet")
		}
		l.frag = true
		l.writeWithPrefix(d)
	} else {
		if end {
			l.frag = false
		}
		l.writeNoPrefix(d)
	}
}

// write writes a NAL unit to the Lexer's buf in byte stream format using the
// start code.
func (l *RTPLexer) writeWithPrefix(d []byte) {
	const prefix = "\x00\x00\x00\x01"
	l.buf.Write([]byte(prefix))
	l.buf.Write(d)
}

// writeNoPrefix writes data to the Lexer's buf. This is used for non start
// fragmentations of a NALU.
func (l *RTPLexer) writeNoPrefix(d []byte) {
	l.buf.Write(d)
}