revid: added chunkSize to audioInput and bufferSize to lex functions

The audio lexers need to know how much data they will be receiving unlike video which has a fixed buffer size.
This means that all the lex function will need to be given a buffer size since they are used as a function pointer with the same signature.

cmd/revid-cli/main.go

@@ -135,10 +135,12 @@ func handleFlags() revid.Config {
 		saturationPtr         = flag.Int("Saturation", 0, "Set Saturation. (100-100)")
 		exposurePtr           = flag.String("Exposure", "auto", "Set exposure mode. ("+strings.Join(revid.ExposureModes[:], ",")+")")
 		autoWhiteBalancePtr   = flag.String("Awb", "auto", "Set automatic white balance mode. ("+strings.Join(revid.AutoWhiteBalanceModes[:], ",")+")")
-		sampleRatePtr         = flag.Int("SampleRate", 48000, "Sample rate of recorded audio")
-		channelsPtr           = flag.Int("Channels", 1, "Record in Mono or Stereo (1 or 2)")
-		recPeriodPtr          = flag.Int("recPeriod", 5, "How many seconds to record at a time")
-		bitDepthPtr           = flag.Int("bitDepth", 16, "Bit Depth to record audio at.")
+
+		// Audio specific flags.
+		sampleRatePtr = flag.Int("SampleRate", 48000, "Sample rate of recorded audio")
+		channelsPtr   = flag.Int("Channels", 1, "Record in Mono or Stereo (1 or 2)")
+		recPeriodPtr  = flag.Int("recPeriod", 5, "How many seconds to record at a time")
+		bitDepthPtr   = flag.Int("bitDepth", 16, "Bit Depth to record audio at.")
 	)
 
 	var outputs flagStrings

codec/lex/lex.go

@@ -36,10 +36,6 @@ import (
 	"time"
 )
 
-const (
-	audioChunkSize = 16000
-)
-
 var noDelay = make(chan time.Time)
 
 func init() {
@@ -52,7 +48,7 @@ var h264Prefix = [...]byte{0x00, 0x00, 0x01, 0x09, 0xf0}
 // 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 H264(dst io.Writer, src io.Reader, delay time.Duration) error {
+func H264(dst io.Writer, src io.Reader, delay time.Duration, bufSize int) error {
 	var tick <-chan time.Time
 	if delay == 0 {
 		tick = noDelay
@@ -62,7 +58,7 @@ func H264(dst io.Writer, src io.Reader, delay time.Duration) error {
 		tick = ticker.C
 	}
 
-	const bufSize = 8 << 10
+	bufSize = 8 << 10 //TODO(Trek): Pass this in rather than set it in here.
 
 	c := newScanner(src, make([]byte, 4<<10)) // Standard file buffer size.
 
@@ -207,7 +203,7 @@ func (c *scanner) reload() error {
 
 // MJPEG parses MJPEG frames read from src into separate writes to dst with
 // successive writes being performed not earlier than the specified delay.
-func MJPEG(dst io.Writer, src io.Reader, delay time.Duration) error {
+func MJPEG(dst io.Writer, src io.Reader, delay time.Duration, bufSize int) error {
 	var tick <-chan time.Time
 	if delay == 0 {
 		tick = noDelay
@@ -252,7 +248,7 @@ func MJPEG(dst io.Writer, src io.Reader, delay time.Duration) error {
 
 // PCM reads from the given source and breaks the PCM into chunks that
 // are an appropriate size for mts and pes packets.
-func PCM(dst io.Writer, src io.Reader, delay time.Duration) error {
+func PCM(dst io.Writer, src io.Reader, delay time.Duration, bufSize int) error {
 	var tick <-chan time.Time
 	if delay == 0 {
 		tick = noDelay
@@ -263,7 +259,7 @@ func PCM(dst io.Writer, src io.Reader, delay time.Duration) error {
 	}
 
 	for {
-		buf := make([]byte, 0, audioChunkSize)
+		buf := make([]byte, 0, bufSize)
 		_, err := src.Read(buf)
 		if err != nil {
 			return err
@@ -281,7 +277,7 @@ func PCM(dst io.Writer, src io.Reader, delay time.Duration) error {
 // are an appropriate size for mts and pes packets.
 // Since PCM and ADPCM are not any different when it comes to how they are
 // transmitted, ADPCM is just a wrapper for PCM.
-func ADPCM(dst io.Writer, src io.Reader, delay time.Duration) error {
-	err := PCM(dst, src, delay)
+func ADPCM(dst io.Writer, src io.Reader, delay time.Duration, bufSize int) error {
+	err := PCM(dst, src, delay, bufSize)
 	return err
 }

revid/revid.go

@@ -103,7 +103,7 @@ type Revid struct {
 	cmd *exec.Cmd
 
 	// lexTo, encoder and packer handle transcoding the input stream.
-	lexTo func(dest io.Writer, src io.Reader, delay time.Duration) error
+	lexTo func(dest io.Writer, src io.Reader, delay time.Duration, bufSize int) error
 
 	// buffer handles passing frames from the transcoder
 	// to the target destination.
@@ -620,7 +620,7 @@ func (r *Revid) startRaspivid() error {
 	}
 
 	r.wg.Add(1)
-	go r.processFrom(stdout, 0)
+	go r.processFrom(stdout, 0, 0)
 	return nil
 }
 
@@ -662,7 +662,7 @@ func (r *Revid) startV4L() error {
 	}
 
 	r.wg.Add(1)
-	go r.processFrom(stdout, time.Duration(0))
+	go r.processFrom(stdout, time.Duration(0), 0)
 	return nil
 }
 
@@ -678,21 +678,20 @@ func (r *Revid) setupInputForFile() error {
 
 	// TODO(kortschak): Maybe we want a context.Context-aware parser that we can stop.
 	r.wg.Add(1)
-	go r.processFrom(f, time.Second/time.Duration(r.config.FrameRate))
+	go r.processFrom(f, time.Second/time.Duration(r.config.FrameRate), 0)
 	return nil
 }
 
 // startAudioInput is used to start capturing audio from an audio device and processing it.
 func (r *Revid) startAudioInput() error {
 	ai := NewAudioInput(&r.config)
-
-	go r.processFrom(ai, time.Second/time.Duration(r.config.WriteRate))
+	go r.processFrom(ai, time.Second/time.Duration(r.config.WriteRate), ai.ChunkSize())
 	return nil
 }
 
-func (r *Revid) processFrom(read io.Reader, delay time.Duration) {
+func (r *Revid) processFrom(read io.Reader, delay time.Duration, bufSize int) {
 	r.config.Logger.Log(logger.Info, pkg+"reading input data")
-	r.err <- r.lexTo(r.buffer, read, delay)
+	r.err <- r.lexTo(r.buffer, read, delay, bufSize)
 	r.config.Logger.Log(logger.Info, pkg+"finished reading input data")
 	r.wg.Done()
 }