Add NewEncodedReeader API

Changes:
  * Add NewEncodedReeader method to Track interface
  * Add video archival example

README.md

@@ -13,9 +13,7 @@
 </p>
 <br>
 
-[MediaDevices](https://developer.mozilla.org/en-US/docs/Web/API/MediaDevices) provides access to connected media input devices like cameras and microphones, as well as screen sharing. It can also be used to encode your video/audio stream to various codec selections. 
-
-The focus of the project has been to seek out a **simple** and **elegant design** for writing media pipelines.
+`mediadevices` provides access to media input devices like cameras, microphones, and screen capture. It can also be used to encode your video/audio stream to various codec selections. `mediadevices` abstracts away the complexities of interacting with things like hardware and codecs allowing you to focus on building appilcations, interacting only with an amazingly simple, easy, and elegant API!
 
 ![](img/demo.gif)
 
@@ -81,6 +79,7 @@ func main() {
 * [Face Detection](/examples/facedetection) - Use a machine learning algorithm to detect faces in a camera stream
 * [RTP Stream](examples/rtp) - Capture camera stream, encode it in H264/VP8/VP9, and send it to a RTP server
 * [HTTP Broadcast](/examples/http) - Broadcast camera stream through HTTP with MJPEG
+* [Archive](/examples/archive) - Archive H264 encoded video stream from a camera
 
 ## Available Media Inputs
 
@@ -90,8 +89,49 @@ func main() {
 | Microphone |  ✔️   | ✔️  |   ✔️    |
 |   Screen   |  ✔️   | ✖️  |   ✖️    |
 
+By default, there's no media input registered. This decision was made to allow you to pay what you need. Therefore, you need to import the associated packages for the media inputs. For example, if you want to use a camera, you need to import the camera package as a side effect:
+
+```go
+import (
+	...
+	_ "github.com/pion/mediadevices/pkg/driver/camera"
+)
+```
+
 ## Available Codecs
 
+In order to encode your video/audio, `mediadevices` needs to know what codecs that you want to use and their parameters. To do this, you need to import the associated packages for the codecs, and add them to the codec selector that you'll pass to `GetUserMedia`:
+
+```go
+package main
+
+import (
+	"github.com/pion/mediadevices"
+	"github.com/pion/mediadevices/pkg/codec/x264"      // This is required to use H264 video encoder
+	_ "github.com/pion/mediadevices/pkg/driver/camera" // This is required to register camera adapter
+)
+
+func main() {
+	// configure codec specific parameters
+	x264Params, _ := x264.NewParams()
+	x264Params.Preset = x264.PresetMedium
+	x264Params.BitRate = 1_000_000 // 1mbps
+
+	codecSelector := mediadevices.NewCodecSelector(
+		mediadevices.WithVideoEncoders(&x264Params),
+	)
+	
+	mediaStream, _ := mediadevices.GetUserMedia(mediadevices.MediaStreamConstraints{
+		Video: func(c *mediadevices.MediaTrackConstraints) {},
+		Codec: codecSelector, // let GetUsermedia know available codecs
+	})
+}
+```
+
+Since `mediadevices` doesn't implement the video/audio codecs, it needs to call the codec libraries from the system through cgo. Therefore, you're required to install the codec libraries before you can use them in `mediadevices`. In the next section, it shows a list of available codecs, where the packages are defined (documentation linked), and installation instructions.
+
+Note: we do not provide recommendations on choosing one codec or another as it is very complex and can be subjective.
+
 ### Video Codecs
 
 #### x264
@@ -143,28 +183,28 @@ A totally open, royalty-free, highly versatile audio codec.
 
 ## Benchmark
 
-Result as of Nov 4, 2020 with Go 1.14 on a Raspberry pi 3, `mediadevices` can produce a **720p at 30 fps with <500ms latency video**.  
+Result as of Nov 4, 2020 with Go 1.14 on a Raspberry pi 3, `mediadevices` can produce video, encode, send across network, and decode at **720p, 30 fps with < 500 ms latency**.  
 
-The test was taken by capturing a camera stream, decode raw frames, encode the video stream with mmal to H264, and send the stream through Webrtc.
+The test was taken by capturing a camera stream, decoding the raw frames, encoding the video stream with mmal, and sending the stream through Webrtc.
 
 ## FAQ
 
 ### Failed to find the best driver that fits the constraints
 
-`mediadevices` provides an automated driver discovery through `GetUserMedia` and `GetDisplayMedia`. In an oversimplified explanation, the discovery algorithm as followed:
+`mediadevices` provides an automated driver discovery through `GetUserMedia` and `GetDisplayMedia`. The driver discover algorithm works something like:
 
 1. Open all registered drivers
 2. Get all properties (property describes what a driver is capable of, e.g. resolution, frame rate, etc.) from opened drivers
 3. Find the best property that meets the criteria
 
 So, when `mediadevices` returns `failed to find the best driver that fits the constraints` error, one of the following conditions might have occured:
-* In your program, the driver has never been imported as a side effect, e.g. `import _ github.com/pion/mediadevices/pkg/driver/camera`
-* Your constraint is too strict that there's no driver can fullfil your requirements. In this case, you can try to turn up the debug level by specifying the following environment variable: `export PION_LOG_DEBUG=all`
-* Your driver is not supported/implemented. In this case, you can either wait for the maintainers to implement it. Or, you can implement it yourself and register it through `RegisterDriverAdapter`
+* Driver was not imported as a side effect in your program, e.g. `import _ github.com/pion/mediadevices/pkg/driver/camera`
+* Your constraint is too strict that there's no driver can fullfil your requirements. In this case, you can try to turn up the debug level by specifying the following environment variable: `export PION_LOG_DEBUG=all` to see what was too strict and tune that.
+* Your driver is not supported/implemented. In this case, you can either let us know (file an issue) and wait for the maintainers to implement it. Or, you can implement it yourself and register it through `RegisterDriverAdapter`
 
 ### Failed to find vpx/x264/mmal/opus codecs
 
-Since `mediadevices` uses cgo to access video/audio codecs, it needs to find these libraries from the system. To do that, `mediadevices` uses [pkg-config](https://www.freedesktop.org/wiki/Software/pkg-config/) for library discovery.
+Since `mediadevices` uses cgo to access video/audio codecs, it needs to find these libraries from the system. To accomplish this, [pkg-config](https://www.freedesktop.org/wiki/Software/pkg-config/) is used for library discovery.
 
 If you see the following error message at compile time:
 ```

examples/archive/README.md

@@ -0,0 +1,36 @@
+## Instructions
+
+### Install required codecs
+
+In this example, we'll be using x264 as our video codec. Therefore, we need to make sure that these codecs are installed within our system. 
+
+Installation steps:
+
+* [x264](https://github.com/pion/mediadevices#x264)
+
+### Download archive examplee
+
+```
+git clone https://github.com/pion/mediadevices.git
+```
+
+### Run archive example
+
+Run `cd mediadevices/examples/archive && go build && ./archive recorded.h264`
+
+### Playback recorded video
+
+Install GStreamer and run:
+```
+gst-launch-1.0 playbin uri=file://${PWD}/recorded.h264
+```
+
+Or run VLC media plyer:
+```
+vlc recorded.h264
+```
+
+A video should start playing in your GStreamer or VLC window.
+
+Congrats, you have used pion-MediaDevices! Now start building something cool
+

examples/archive/main.go

@@ -0,0 +1,82 @@
+package main
+
+import (
+	"fmt"
+	"image"
+	"io"
+	"os"
+	"os/signal"
+	"syscall"
+
+	"github.com/pion/mediadevices"
+	"github.com/pion/mediadevices/pkg/codec/x264"      // This is required to use H264 video encoder
+	_ "github.com/pion/mediadevices/pkg/driver/camera" // This is required to register camera adapter
+	"github.com/pion/mediadevices/pkg/frame"
+	"github.com/pion/mediadevices/pkg/io/video"
+	"github.com/pion/mediadevices/pkg/prop"
+)
+
+func must(err error) {
+	if err != nil {
+		panic(err)
+	}
+}
+
+func main() {
+	if len(os.Args) != 2 {
+		fmt.Printf("usage: %s <path/to/file.h264>\n", os.Args[0])
+		return
+	}
+	dest := os.Args[1]
+
+	sigs := make(chan os.Signal, 1)
+	signal.Notify(sigs, syscall.SIGINT)
+
+	x264Params, err := x264.NewParams()
+	must(err)
+	x264Params.Preset = x264.PresetMedium
+	x264Params.BitRate = 1_000_000 // 1mbps
+
+	codecSelector := mediadevices.NewCodecSelector(
+		mediadevices.WithVideoEncoders(&x264Params),
+	)
+
+	mediaStream, err := mediadevices.GetUserMedia(mediadevices.MediaStreamConstraints{
+		Video: func(c *mediadevices.MediaTrackConstraints) {
+			c.FrameFormat = prop.FrameFormat(frame.FormatYUY2)
+			c.Width = prop.Int(640)
+			c.Height = prop.Int(480)
+		},
+		Codec: codecSelector,
+	})
+	must(err)
+
+	videoTrack := mediaStream.GetVideoTracks()[0].(*mediadevices.VideoTrack)
+	defer videoTrack.Close()
+
+	videoTrack.Transform(video.TransformFunc(func(r video.Reader) video.Reader {
+		return video.ReaderFunc(func() (img image.Image, release func(), err error) {
+			// we send io.EOF signal to the encoder reader to stop reading. Therefore, io.Copy
+			// will finish its execution and the program will finish
+			select {
+			case <-sigs:
+				return nil, func() {}, io.EOF
+			default:
+			}
+
+			return r.Read()
+		})
+	}))
+
+	reader, err := videoTrack.NewEncodedReader(x264Params.RTPCodec().Name)
+	must(err)
+	defer reader.Close()
+
+	out, err := os.Create(dest)
+	must(err)
+
+	fmt.Println("Recording... Press Ctrl+c to stop")
+	_, err = io.Copy(out, reader)
+	must(err)
+	fmt.Println("Your video has been recorded to", dest)
+}

examples/webrtc/README.md

@@ -39,4 +39,4 @@ Copy the text that `./webrtc` just emitted and copy into second text area
 
 A video should start playing in your browser above the input boxes, and will continue playing until you close the application.
 
-Congrats, you have used pion-WebRTC! Now start building something cool
+Congrats, you have used pion-MediaDevices! Now start building something cool

ioreader.go

@@ -0,0 +1,14 @@
+package mediadevices
+
+type encodedReadCloserImpl struct {
+	readFn  func([]byte) (int, error)
+	closeFn func() error
+}
+
+func (r *encodedReadCloserImpl) Read(b []byte) (int, error) {
+	return r.readFn(b)
+}
+
+func (r *encodedReadCloserImpl) Close() error {
+	return r.closeFn()
+}

mediastream_test.go

@@ -1,6 +1,7 @@
 package mediadevices
 
 import (
+	"io"
 	"testing"
 
 	"github.com/pion/webrtc/v2"
@@ -37,6 +38,10 @@ func (track *mockMediaStreamTrack) NewRTPReader(codecName string, mtu int) (RTPR
 	return nil, nil
 }
 
+func (track *mockMediaStreamTrack) NewEncodedReader(codecName string) (io.ReadCloser, error) {
+	return nil, nil
+}
+
 func TestMediaStreamFilters(t *testing.T) {
 	audioTracks := []Track{
 		&mockMediaStreamTrack{AudioInput},

pkg/codec/vpx/vpx.go

@@ -237,10 +237,19 @@ func (e *encoder) Read() ([]byte, func(), error) {
 		e.raw.d_w, e.raw.d_h = C.uint(width), C.uint(height)
 	}
 
+	duration := t - e.tLastFrame
+	// VPX doesn't allow 0 duration. If 0 is given, vpx_codec_encode will fail with VPX_CODEC_INVALID_PARAM.
+	// 0 duration is possible because mediadevices first gets the frame meta data by reading from the source,
+	// and consequently the codec will read the first frame from the buffer. This means the first frame won't
+	// have a pause to the second frame, which means if the delay is <1 ms (vpx duration resolution), duration
+	// is going to be 0.
+	if duration == 0 {
+		duration = 1
+	}
 	var flags int
 	if ec := C.encode_wrapper(
 		e.codec, e.raw,
-		C.long(t-e.tStart), C.ulong(t-e.tLastFrame), C.long(flags), C.ulong(e.deadline),
+		C.long(t-e.tStart), C.ulong(duration), C.long(flags), C.ulong(e.deadline),
 		(*C.uchar)(&yuvImg.Y[0]), (*C.uchar)(&yuvImg.Cb[0]), (*C.uchar)(&yuvImg.Cr[0]),
 	); ec != C.VPX_CODEC_OK {
 		return nil, func() {}, fmt.Errorf("vpx_codec_encode failed (%d)", ec)

track.go

@@ -3,6 +3,7 @@ package mediadevices
 import (
 	"errors"
 	"image"
+	"io"
 	"math/rand"
 	"sync"
 
@@ -57,6 +58,8 @@ type Track interface {
 	// NewRTPReader creates a new reader from the source. The reader will encode the source, and packetize
 	// the encoded data in RTP format with given mtu size.
 	NewRTPReader(codecName string, mtu int) (RTPReadCloser, error)
+	// NewEncodedReader creates a new Go standard io.ReadCloser that reads the encoded data in codecName format
+	NewEncodedReader(codecName string) (io.ReadCloser, error)
 }
 
 type baseTrack struct {
@@ -182,6 +185,31 @@ func (track *baseTrack) unbind(pc *webrtc.PeerConnection) error {
 	return nil
 }
 
+func (track *baseTrack) newEncodedReader(reader codec.ReadCloser) (io.ReadCloser, error) {
+	var encoded []byte
+	release := func() {}
+	return &encodedReadCloserImpl{
+		readFn: func(b []byte) (int, error) {
+			var err error
+
+			if len(encoded) == 0 {
+				release()
+				encoded, release, err = reader.Read()
+				if err != nil {
+					reader.Close()
+					track.onError(err)
+					return 0, err
+				}
+			}
+
+			n := copy(b, encoded)
+			encoded = encoded[n:]
+			return n, nil
+		},
+		closeFn: reader.Close,
+	}, nil
+}
+
 func newTrackFromDriver(d driver.Driver, constraints MediaTrackConstraints, selector *CodecSelector) (Track, error) {
 	if err := d.Open(); err != nil {
 		return nil, err
@@ -298,6 +326,21 @@ func (track *VideoTrack) NewRTPReader(codecName string, mtu int) (RTPReadCloser,
 	}, nil
 }
 
+func (track *VideoTrack) NewEncodedReader(codecName string) (io.ReadCloser, error) {
+	reader := track.NewReader(false)
+	inputProp, err := detectCurrentVideoProp(track.Broadcaster)
+	if err != nil {
+		return nil, err
+	}
+
+	encodedReader, _, err := track.selector.selectVideoCodecByNames(reader, inputProp, codecName)
+	if err != nil {
+		return nil, err
+	}
+
+	return track.newEncodedReader(encodedReader)
+}
+
 // AudioTrack is a specific track type that contains audio source which allows multiple readers to access, and
 // manipulate.
 type AudioTrack struct {
@@ -380,7 +423,7 @@ func (track *AudioTrack) NewRTPReader(codecName string, mtu int) (RTPReadCloser,
 		return nil, err
 	}
 
-	sample := newVideoSampler(selectedCodec.ClockRate)
+	sample := newAudioSampler(selectedCodec.ClockRate, inputProp.Latency)
 
 	// FIXME: not sure the best way to get unique ssrc. We probably should have a global keeper that can generate a random ID and does book keeping?
 	packetizer := rtp.NewPacketizer(mtu, selectedCodec.PayloadType, rand.Uint32(), selectedCodec.Payloader, rtp.NewRandomSequencer(), selectedCodec.ClockRate)
@@ -402,3 +445,18 @@ func (track *AudioTrack) NewRTPReader(codecName string, mtu int) (RTPReadCloser,
 		closeFn: encodedReader.Close,
 	}, nil
 }
+
+func (track *AudioTrack) NewEncodedReader(codecName string) (io.ReadCloser, error) {
+	reader := track.NewReader(false)
+	inputProp, err := detectCurrentAudioProp(track.Broadcaster)
+	if err != nil {
+		return nil, err
+	}
+
+	encodedReader, _, err := track.selector.selectAudioCodecByNames(reader, inputProp, codecName)
+	if err != nil {
+		return nil, err
+	}
+
+	return track.newEncodedReader(encodedReader)
+}