Add new video codec factories

webrtc/api/video_codecs/video_decoder_factory.h

Index: webrtc/api/video_codecs/video_decoder_factory.h
diff --git a/webrtc/api/video_codecs/video_decoder_factory.h b/webrtc/api/video_codecs/video_decoder_factory.h
new file mode 100644
index 0000000000000000000000000000000000000000..591b16114d1aa7e0978728d010ac42fff671cc81
--- /dev/null
+++ b/webrtc/api/video_codecs/video_decoder_factory.h
@@ -0,0 +1,42 @@
+/*
+ *  Copyright (c) 2017 The WebRTC project authors. All Rights Reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS.  All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#ifndef WEBRTC_API_VIDEO_CODECS_VIDEO_DECODER_FACTORY_H_
+#define WEBRTC_API_VIDEO_CODECS_VIDEO_DECODER_FACTORY_H_
+
+#include <memory>
+#include <vector>
+
+namespace cricket {
+struct VideoCodec;
+}  // namespace cricket

[stefan-webrtc, 2017/09/06 12:40:42]

What do you think of the fit of VideoCodec in the cricket vs the webrtc
namespace?

[magjed_webrtc, 2017/09/10 15:27:49]

It's unfortunate. cricket::VideoCodec represents the codec in SDP and is very
simple and just contains a name as a string and parameters as a string key/value
map. I introduced a new struct SdpVideoFormat to carry this information instead
so that we don't have to depend on cricket::VideoCodec. This is similar to how
they did it for audio.

[kwiberg-webrtc, 2017/09/10 19:00:21]

Pardon the tangent, but do the two of you think that the separate “cricket”
namespace is useful? If we had just one namespace, outsiders would have only one
WebRTC namespace to remember (likely “webrtc”), and we ourselves would almost
never need to mention the namespace at all since all our stuff would be in the
same namespace.

My opinion is that the “rtc” namespace could be folded into “webrtc”, but I have
much less personal experience with the “cricket” namespace.

[magjed_webrtc, 2017/09/12 12:08:48]

It's still often a sign that we are mixing two code bases and probably making
the dependency graph worse when we are using cricket in the webrtc namespace.
It's an interesting discussion, let's file an issue or start an email thread.

+
+namespace webrtc {
+
+class VideoDecoder;
+
+// A factory that creates VideoDecoders.
+// NOTE: This class is still under development and may change without notice.
+class VideoDecoderFactory {
+ public:
+  // Returns a list of supported video codecs in order of preference, to use for
+  // signaling etc.
+  virtual std::vector<cricket::VideoCodec> GetSupportedCodecs() const = 0;
+
+  // Creates a VideoDecoder for the specified codec.
+  virtual std::unique_ptr<VideoDecoder> CreateVideoDecoder(

[andersc, 2017/09/06 15:00:06]

Do we also need a version of this method that takes a VideoDecoderParams
argument?

[magjed_webrtc, 2017/09/10 15:27:49]

I will investigate a bit, but hopefully we don't need to add VideoDecoderParams
here.

[magjed_webrtc, 2017/09/12 12:08:48]

I looked it up and we don't need to add VideoDecoderParams.

+      const cricket::VideoCodec& codec) = 0;
+
+  virtual ~VideoDecoderFactory() {}
+};
+
+}  // namespace webrtc
+
+#endif  // WEBRTC_API_VIDEO_CODECS_VIDEO_DECODER_FACTORY_H_