av/revid/audio-input.go

/*
NAME
  audio-input.go

AUTHOR
  Trek Hopton <trek@ausocean.org>

LICENSE
  audio-input.go is Copyright (C) 2019 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 in gpl.txt.
  If not, see [GNU licenses](http://www.gnu.org/licenses).
*/

package revid

import (
	"bytes"
	"errors"
	"fmt"
	"io"
	"sync"
	"time"

	"github.com/yobert/alsa"

	"bitbucket.org/ausocean/av/codec/adpcm"
	"bitbucket.org/ausocean/av/codec/pcm"
	"bitbucket.org/ausocean/utils/logger"
	"bitbucket.org/ausocean/utils/ring"
)

const (
	rbTimeout     = 100 * time.Millisecond
	rbNextTimeout = 100 * time.Millisecond
	rbLen         = 200
)

const (
	running = iota
	paused
	stopped
)

// Rates contains the audio sample rates used by revid.
var Rates = [8]int{8000, 16000, 32000, 44100, 48000, 88200, 96000, 192000}

// AudioDevice holds everything we need to know about the audio input stream.
type AudioDevice struct {
	l      Logger
	mu     sync.Mutex
	source string // Name of audio source, or empty for the default source.
	// Operating mode, either running, paused, or stopped.
	// "running" means the input goroutine is reading from the ALSA device and writing to the ringbuffer.
	// "paused" means the input routine is sleeping until unpaused or stopped.
	// "stopped" means the input routine is stopped and the ALSA device is closed.
	mode uint8

	dev       *alsa.Device // Audio input device.
	ab        alsa.Buffer  // ALSA's buffer.
	rb        *ring.Buffer // Our buffer.
	chunkSize int          // This is the number of bytes that will be stored at a time.

	*AudioConfig
}

// AudioConfig provides parameters used by AudioDevice.
type AudioConfig struct {
	SampleRate int
	Channels   int
	BitDepth   int
	RecPeriod  float64
	Codec      uint8
}

// NewAudioDevice initializes and returns an AudioDevice which can be started, read from, and stopped.
func NewAudioDevice(cfg *AudioConfig, l Logger) (*AudioDevice, error) {
	a := &AudioDevice{}
	a.AudioConfig = cfg
	a.l = l

	// Open the requested audio device.
	err := a.open()
	if err != nil {
		a.l.Log(logger.Error, "failed to open audio device", "error", err.Error())
		return nil, errors.New("failed to open audio device")
	}

	// Setup ring buffer to capture audio in periods of a.RecPeriod seconds and buffer rbDuration seconds in total.
	a.ab = a.dev.NewBufferDuration(time.Duration(a.RecPeriod * float64(time.Second)))
	cs := (float64((len(a.ab.Data)/a.dev.BufferFormat().Channels)*a.Channels) / float64(a.dev.BufferFormat().Rate)) * float64(a.SampleRate)
	if cs < 1 {
		a.l.Log(logger.Error, "given AudioConfig parameters are too small", "error", err.Error())
		return nil, errors.New("given AudioConfig parameters are too small")
	}
	if a.Codec == ADPCM {
		a.chunkSize = adpcm.EncBytes(int(cs))
	} else {
		a.chunkSize = int(cs)
	}
	a.rb = ring.NewBuffer(rbLen, a.chunkSize, rbTimeout)

	a.mode = paused
	go a.input()

	return a, nil
}

// Start will start recording audio and writing to the ringbuffer.
func (a *AudioDevice) Start() error {
	a.mu.Lock()
	mode := a.mode
	a.mu.Unlock()
	switch mode {
	case paused:
		a.mu.Lock()
		a.mode = running
		a.mu.Unlock()
		return nil
	case stopped:
		// TODO(Trek): Make this reopen device and start recording.
		return errors.New("device is stopped")
	case running:
		return nil
	default:
		return errors.New("invalid mode")
	}
	return nil
}

// Stop will stop recording audio and close the device.
func (a *AudioDevice) Stop() {
	a.mu.Lock()
	a.mode = stopped
	a.mu.Unlock()
}

// ChunkSize returns the number of bytes written to the ringbuffer per a.RecPeriod.
func (a *AudioDevice) ChunkSize() int {
	return a.chunkSize
}

// open the recording device with the given name and prepare it to record.
// If name is empty, the first recording device is used.
func (a *AudioDevice) open() error {
	// Close any existing device.
	if a.dev != nil {
		a.l.Log(logger.Debug, "closing device", "source", a.source)
		a.dev.Close()
		a.dev = nil
	}

	// Open sound card and open recording device.
	a.l.Log(logger.Debug, "opening sound card")
	cards, err := alsa.OpenCards()
	if err != nil {
		a.l.Log(logger.Debug, "failed to open sound card")
		return err
	}
	defer alsa.CloseCards(cards)

	a.l.Log(logger.Debug, "finding audio device")
	for _, card := range cards {
		devices, err := card.Devices()
		if err != nil {
			continue
		}
		for _, dev := range devices {
			if dev.Type != alsa.PCM || !dev.Record {
				continue
			}
			if dev.Title == a.source || a.source == "" {
				a.dev = dev
				break
			}
		}
	}
	if a.dev == nil {
		a.l.Log(logger.Debug, "failed to find audio device")
		return errors.New("no audio device found")
	}

	a.l.Log(logger.Debug, "opening audio device", "source", a.dev.Title)
	err = a.dev.Open()
	if err != nil {
		a.l.Log(logger.Debug, "failed to open audio device")
		return err
	}

	// 2 channels is what most devices need to record in. If mono is requested,
	// the recording will be converted in formatBuffer().
	_, err = a.dev.NegotiateChannels(2)
	if err != nil {
		return err
	}

	// Try to negotiate a rate to record in that is divisible by the wanted rate
	// so that it can be easily downsampled to the wanted rate.
	// Note: if a card thinks it can record at a rate but can't actually, this can cause a failure.
	// Eg. the audioinjector sound card is supposed to record at 8000Hz and 16000Hz but it can't due to a firmware issue,
	// a fix for this is to remove 8000 and 16000 from the Rates slice.
	foundRate := false
	for i := 0; i < len(Rates) && !foundRate; i++ {
		if Rates[i] < a.SampleRate {
			continue
		}
		if Rates[i]%a.SampleRate == 0 {
			_, err = a.dev.NegotiateRate(Rates[i])
			if err == nil {
				foundRate = true
				a.l.Log(logger.Debug, "Sample rate set", "rate", Rates[i])
			}
		}
	}

	// If no easily divisible rate is found, then use the default rate.
	if !foundRate {
		a.l.Log(logger.Warning, "Unable to sample at requested rate, default used.", "rateRequested", a.SampleRate)
		_, err = a.dev.NegotiateRate(defaultSampleRate)
		if err != nil {
			return err
		}
		a.l.Log(logger.Debug, "Sample rate set", "rate", defaultSampleRate)
	}

	var aFmt alsa.FormatType
	switch a.BitDepth {
	case 16:
		aFmt = alsa.S16_LE
	case 32:
		aFmt = alsa.S32_LE
	default:
		return fmt.Errorf("unsupported sample bits %v", a.BitDepth)
	}
	_, err = a.dev.NegotiateFormat(aFmt)
	if err != nil {
		return err
	}

	// Either 8192 or 16384 bytes is a reasonable ALSA buffer size.
	_, err = a.dev.NegotiateBufferSize(8192, 16384)
	if err != nil {
		return err
	}

	if err = a.dev.Prepare(); err != nil {
		return err
	}
	a.l.Log(logger.Debug, "Successfully negotiated ALSA params")
	return nil
}

// input continously records audio and writes it to the ringbuffer.
// Re-opens the device and tries again if ASLA returns an error.
func (a *AudioDevice) input() {
	for {
		// Check mode.
		a.mu.Lock()
		mode := a.mode
		a.mu.Unlock()
		switch mode {
		case paused:
			time.Sleep(time.Duration(a.RecPeriod) * time.Second)
			continue
		case stopped:
			if a.dev != nil {
				a.l.Log(logger.Debug, "closing audio device", "source", a.source)
				a.dev.Close()
				a.dev = nil
			}
			return
		}

		// Read from audio device.
		a.l.Log(logger.Debug, "recording audio for period", "seconds", a.RecPeriod)
		err := a.dev.Read(a.ab.Data)
		if err != nil {
			a.l.Log(logger.Debug, "read failed", "error", err.Error())
			err = a.open() // re-open
			if err != nil {
				a.l.Log(logger.Fatal, "reopening device failed", "error", err.Error())
				return
			}
			continue
		}

		// Process audio.
		a.l.Log(logger.Debug, "processing audio")
		toWrite := a.formatBuffer()

		// Write audio to ringbuffer.
		n, err := a.rb.Write(toWrite.Data)
		switch err {
		case nil:
			a.l.Log(logger.Debug, "wrote audio to ringbuffer", "length", n)
		case ring.ErrDropped:
			a.l.Log(logger.Warning, "old audio data overwritten")
		default:
			a.l.Log(logger.Error, "unexpected ringbuffer error", "error", err.Error())
			return
		}
	}
}

// Read reads a full PCM chunk from the ringbuffer, returning the number of bytes read upon success.
// Any errors returned are unexpected and should be considered fatal.
func (a *AudioDevice) Read(p []byte) (n int, err error) {
	// Ready ringbuffer for read.
	_, err = a.rb.Next(rbNextTimeout)
	switch err {
	case nil:
	case ring.ErrTimeout:
		return 0, nil
	default:
		return 0, err
	}

	// Read from ring buffer.
	n, err = a.rb.Read(p)
	switch err {
	case nil:
	case io.EOF:
		return 0, nil
	default:
		return 0, err
	}

	return n, nil
}

// formatBuffer returns audio that has been converted to the desired format.
func (a *AudioDevice) formatBuffer() alsa.Buffer {
	var err error

	// If nothing needs to be changed, return the original.
	if a.ab.Format.Channels == a.Channels && a.ab.Format.Rate == a.SampleRate {
		return a.ab
	}

	formatted := alsa.Buffer{Format: a.ab.Format, Data: a.ab.Data}
	if a.ab.Format.Channels != a.Channels {
		// Convert channels.
		// TODO(Trek): Make this work for conversions other than stereo to mono.
		if a.ab.Format.Channels == 2 && a.Channels == 1 {
			formatted.Data, err = pcm.StereoToMono(a.ab)
			if err != nil {
				a.l.Log(logger.Fatal, "channel conversion failed", "error", err.Error())
			}
		}
	}

	if a.ab.Format.Rate != a.SampleRate {
		// Convert rate.
		formatted.Data, err = pcm.Resample(formatted, a.SampleRate)
		if err != nil {
			a.l.Log(logger.Fatal, "rate conversion failed", "error", err.Error())
		}
	}

	switch a.Codec {
	case PCM:
	case ADPCM:
		b := bytes.NewBuffer(make([]byte, 0, adpcm.EncBytes(len(formatted.Data))))
		enc := adpcm.NewEncoder(b)
		_, err = enc.Write(formatted.Data)
		if err != nil {
			a.l.Log(logger.Fatal, "unable to encode", "error", err.Error())
		}
		formatted.Data = b.Bytes()
	default:
		a.l.Log(logger.Error, "codec conversion failed, audio has remained original codec", "error", err.Error())
	}

	return formatted
}