Add RTPVideoHeader::GetAsMetadata().

In preparation of adding RTPVideoHeader::SetFromMetadata() method, the
VideoFrameMetadata construct-from-RTPVideoHeader is replaced by
RTPVideoHeader::GetAsMetadata(). This serves two purposes:
1. Having "GetAs" and "SetFrom" in the same file reduces the risk of
   these two methods getting out of sync as we expand its usage.
2. This is necessary to avoid a circular dependency that would
   otherwise be introduced by RTPVideoHeader::SetFromMetadata().

Bug: webrtc:14709
Change-Id: I127b3d15f9a8c6af210449a5a50d414c9ba79930
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/285080
Reviewed-by: Tony Herre <herre@google.com>
Reviewed-by: Harald Alvestrand <hta@webrtc.org>
Commit-Queue: Henrik Boström <hbos@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#38735}

api/video/BUILD.gn

@@ -333,7 +333,7 @@ rtc_source_set("video_frame_metadata") {
   ]
   deps = [
     "..:array_view",
-    "../../modules/rtp_rtcp:rtp_video_header",
+    "../../rtc_base/system:rtc_export",
     "../transport/rtp:dependency_descriptor",
   ]
   absl_deps = [

api/video/video_frame_metadata.cc

@@ -10,19 +10,69 @@
 
 #include "api/video/video_frame_metadata.h"
 
-#include "modules/rtp_rtcp/source/rtp_video_header.h"
-
 namespace webrtc {
 
-VideoFrameMetadata::VideoFrameMetadata(const RTPVideoHeader& header)
+VideoFrameMetadata::VideoFrameMetadata() = default;
-    : width_(header.width), height_(header.height) {
+
-  if (header.generic) {
+uint16_t VideoFrameMetadata::GetWidth() const {
-    frame_id_ = header.generic->frame_id;
+  return width_;
-    spatial_index_ = header.generic->spatial_index;
+}
-    temporal_index_ = header.generic->temporal_index;
+
-    frame_dependencies_ = header.generic->dependencies;
+void VideoFrameMetadata::SetWidth(uint16_t width) {
-    decode_target_indications_ = header.generic->decode_target_indications;
+  width_ = width;
-  }
+}
+
+uint16_t VideoFrameMetadata::GetHeight() const {
+  return height_;
+}
+
+void VideoFrameMetadata::SetHeight(uint16_t height) {
+  height_ = height;
+}
+
+absl::optional<int64_t> VideoFrameMetadata::GetFrameId() const {
+  return frame_id_;
+}
+
+void VideoFrameMetadata::SetFrameId(absl::optional<int64_t> frame_id) {
+  frame_id_ = frame_id;
+}
+
+int VideoFrameMetadata::GetSpatialIndex() const {
+  return spatial_index_;
+}
+
+void VideoFrameMetadata::SetSpatialIndex(int spatial_index) {
+  spatial_index_ = spatial_index;
+}
+
+int VideoFrameMetadata::GetTemporalIndex() const {
+  return temporal_index_;
+}
+
+void VideoFrameMetadata::SetTemporalIndex(int temporal_index) {
+  temporal_index_ = temporal_index;
+}
+
+rtc::ArrayView<const int64_t> VideoFrameMetadata::GetFrameDependencies() const {
+  return frame_dependencies_;
+}
+
+void VideoFrameMetadata::SetFrameDependencies(
+    rtc::ArrayView<const int64_t> frame_dependencies) {
+  frame_dependencies_.assign(frame_dependencies.begin(),
+                             frame_dependencies.end());
+}
+
+rtc::ArrayView<const DecodeTargetIndication>
+VideoFrameMetadata::GetDecodeTargetIndications() const {
+  return decode_target_indications_;
+}
+
+void VideoFrameMetadata::SetDecodeTargetIndications(
+    rtc::ArrayView<const DecodeTargetIndication> decode_target_indications) {
+  decode_target_indications_.assign(decode_target_indications.begin(),
+                                    decode_target_indications.end());
 }
 
 }  // namespace webrtc

api/video/video_frame_metadata.h

@@ -17,37 +17,44 @@
 #include "absl/types/optional.h"
 #include "api/array_view.h"
 #include "api/transport/rtp/dependency_descriptor.h"
+#include "rtc_base/system/rtc_export.h"
 
 namespace webrtc {
 
-struct RTPVideoHeader;
-
 // A subset of metadata from the RTP video header, exposed in insertable streams
 // API.
-class VideoFrameMetadata {
+class RTC_EXPORT VideoFrameMetadata {
  public:
-  explicit VideoFrameMetadata(const RTPVideoHeader& header);
+  VideoFrameMetadata();
   VideoFrameMetadata(const VideoFrameMetadata&) = default;
   VideoFrameMetadata& operator=(const VideoFrameMetadata&) = default;
 
-  uint16_t GetWidth() const { return width_; }
+  uint16_t GetWidth() const;
-  uint16_t GetHeight() const { return height_; }
+  void SetWidth(uint16_t width);
-  absl::optional<int64_t> GetFrameId() const { return frame_id_; }
-  int GetSpatialIndex() const { return spatial_index_; }
-  int GetTemporalIndex() const { return temporal_index_; }
 
-  rtc::ArrayView<const int64_t> GetFrameDependencies() const {
+  uint16_t GetHeight() const;
-    return frame_dependencies_;
+  void SetHeight(uint16_t height);
-  }
+
+  absl::optional<int64_t> GetFrameId() const;
+  void SetFrameId(absl::optional<int64_t> frame_id);
+
+  int GetSpatialIndex() const;
+  void SetSpatialIndex(int spatial_index);
+
+  int GetTemporalIndex() const;
+  void SetTemporalIndex(int temporal_index);
+
+  rtc::ArrayView<const int64_t> GetFrameDependencies() const;
+  void SetFrameDependencies(rtc::ArrayView<const int64_t> frame_dependencies);
 
   rtc::ArrayView<const DecodeTargetIndication> GetDecodeTargetIndications()
-      const {
+      const;
-    return decode_target_indications_;
+  void SetDecodeTargetIndications(
-  }
+      rtc::ArrayView<const DecodeTargetIndication> decode_target_indications);
 
  private:
-  int16_t width_;
+  int16_t width_ = 0;
-  int16_t height_;
+  int16_t height_ = 0;
   absl::optional<int64_t> frame_id_;
   int spatial_index_ = 0;
   int temporal_index_ = 0;

api/video/video_frame_metadata_unittest.cc

@@ -23,14 +23,14 @@ using ::testing::IsEmpty;
 TEST(VideoFrameMetadata, GetWidthReturnsCorrectValue) {
   RTPVideoHeader video_header;
   video_header.width = 1280u;
-  VideoFrameMetadata metadata(video_header);
+  VideoFrameMetadata metadata = video_header.GetAsMetadata();
   EXPECT_EQ(metadata.GetWidth(), video_header.width);
 }
 
 TEST(VideoFrameMetadata, GetHeightReturnsCorrectValue) {
   RTPVideoHeader video_header;
   video_header.height = 720u;
-  VideoFrameMetadata metadata(video_header);
+  VideoFrameMetadata metadata = video_header.GetAsMetadata();
   EXPECT_EQ(metadata.GetHeight(), video_header.height);
 }
 
@@ -39,13 +39,13 @@ TEST(VideoFrameMetadata, GetFrameIdReturnsCorrectValue) {
   RTPVideoHeader::GenericDescriptorInfo& generic =
       video_header.generic.emplace();
   generic.frame_id = 10;
-  VideoFrameMetadata metadata(video_header);
+  VideoFrameMetadata metadata = video_header.GetAsMetadata();
   EXPECT_EQ(metadata.GetFrameId().value(), 10);
 }
 
 TEST(VideoFrameMetadata, HasNoFrameIdForHeaderWithoutGeneric) {
   RTPVideoHeader video_header;
-  VideoFrameMetadata metadata(video_header);
+  VideoFrameMetadata metadata = video_header.GetAsMetadata();
   ASSERT_FALSE(video_header.generic);
   EXPECT_EQ(metadata.GetFrameId(), absl::nullopt);
 }
@@ -55,13 +55,13 @@ TEST(VideoFrameMetadata, GetSpatialIndexReturnsCorrectValue) {
   RTPVideoHeader::GenericDescriptorInfo& generic =
       video_header.generic.emplace();
   generic.spatial_index = 2;
-  VideoFrameMetadata metadata(video_header);
+  VideoFrameMetadata metadata = video_header.GetAsMetadata();
   EXPECT_EQ(metadata.GetSpatialIndex(), 2);
 }
 
 TEST(VideoFrameMetadata, SpatialIndexIsZeroForHeaderWithoutGeneric) {
   RTPVideoHeader video_header;
-  VideoFrameMetadata metadata(video_header);
+  VideoFrameMetadata metadata = video_header.GetAsMetadata();
   ASSERT_FALSE(video_header.generic);
   EXPECT_EQ(metadata.GetSpatialIndex(), 0);
 }
@@ -71,13 +71,13 @@ TEST(VideoFrameMetadata, GetTemporalIndexReturnsCorrectValue) {
   RTPVideoHeader::GenericDescriptorInfo& generic =
       video_header.generic.emplace();
   generic.temporal_index = 3;
-  VideoFrameMetadata metadata(video_header);
+  VideoFrameMetadata metadata = video_header.GetAsMetadata();
   EXPECT_EQ(metadata.GetTemporalIndex(), 3);
 }
 
 TEST(VideoFrameMetadata, TemporalIndexIsZeroForHeaderWithoutGeneric) {
   RTPVideoHeader video_header;
-  VideoFrameMetadata metadata(video_header);
+  VideoFrameMetadata metadata = video_header.GetAsMetadata();
   ASSERT_FALSE(video_header.generic);
   EXPECT_EQ(metadata.GetTemporalIndex(), 0);
 }
@@ -87,13 +87,13 @@ TEST(VideoFrameMetadata, GetFrameDependenciesReturnsCorrectValue) {
   RTPVideoHeader::GenericDescriptorInfo& generic =
       video_header.generic.emplace();
   generic.dependencies = {5, 6, 7};
-  VideoFrameMetadata metadata(video_header);
+  VideoFrameMetadata metadata = video_header.GetAsMetadata();
   EXPECT_THAT(metadata.GetFrameDependencies(), ElementsAre(5, 6, 7));
 }
 
 TEST(VideoFrameMetadata, FrameDependencyVectorIsEmptyForHeaderWithoutGeneric) {
   RTPVideoHeader video_header;
-  VideoFrameMetadata metadata(video_header);
+  VideoFrameMetadata metadata = video_header.GetAsMetadata();
   ASSERT_FALSE(video_header.generic);
   EXPECT_THAT(metadata.GetFrameDependencies(), IsEmpty());
 }
@@ -103,7 +103,7 @@ TEST(VideoFrameMetadata, GetDecodeTargetIndicationsReturnsCorrectValue) {
   RTPVideoHeader::GenericDescriptorInfo& generic =
       video_header.generic.emplace();
   generic.decode_target_indications = {DecodeTargetIndication::kSwitch};
-  VideoFrameMetadata metadata(video_header);
+  VideoFrameMetadata metadata = video_header.GetAsMetadata();
   EXPECT_THAT(metadata.GetDecodeTargetIndications(),
               ElementsAre(DecodeTargetIndication::kSwitch));
 }
@@ -111,7 +111,7 @@ TEST(VideoFrameMetadata, GetDecodeTargetIndicationsReturnsCorrectValue) {
 TEST(VideoFrameMetadata,
      DecodeTargetIndicationsVectorIsEmptyForHeaderWithoutGeneric) {
   RTPVideoHeader video_header;
-  VideoFrameMetadata metadata(video_header);
+  VideoFrameMetadata metadata = video_header.GetAsMetadata();
   ASSERT_FALSE(video_header.generic);
   EXPECT_THAT(metadata.GetDecodeTargetIndications(), IsEmpty());
 }

modules/rtp_rtcp/BUILD.gn

@@ -425,6 +425,7 @@ rtc_library("rtp_video_header") {
     "../../api:rtp_headers",
     "../../api/transport/rtp:dependency_descriptor",
     "../../api/video:video_frame",
+    "../../api/video:video_frame_metadata",
     "../../api/video:video_frame_type",
     "../../api/video:video_rtp_headers",
     "../../modules/video_coding:codec_globals_headers",

modules/rtp_rtcp/source/rtp_sender_video_frame_transformer_delegate.cc

@@ -35,7 +35,7 @@ class TransformableVideoSenderFrame : public TransformableVideoFrameInterface {
       uint32_t ssrc)
       : encoded_data_(encoded_image.GetEncodedData()),
         header_(video_header),
-        metadata_(header_),
+        metadata_(header_.GetAsMetadata()),
         frame_type_(encoded_image._frameType),
         payload_type_(payload_type),
         codec_type_(codec_type),

modules/rtp_rtcp/source/rtp_video_header.cc

@@ -12,13 +12,27 @@
 
 namespace webrtc {
 
-RTPVideoHeader::RTPVideoHeader() : video_timing() {}
-RTPVideoHeader::RTPVideoHeader(const RTPVideoHeader& other) = default;
-RTPVideoHeader::~RTPVideoHeader() = default;
-
 RTPVideoHeader::GenericDescriptorInfo::GenericDescriptorInfo() = default;
 RTPVideoHeader::GenericDescriptorInfo::GenericDescriptorInfo(
     const GenericDescriptorInfo& other) = default;
 RTPVideoHeader::GenericDescriptorInfo::~GenericDescriptorInfo() = default;
 
+RTPVideoHeader::RTPVideoHeader() : video_timing() {}
+RTPVideoHeader::RTPVideoHeader(const RTPVideoHeader& other) = default;
+RTPVideoHeader::~RTPVideoHeader() = default;
+
+VideoFrameMetadata RTPVideoHeader::GetAsMetadata() const {
+  VideoFrameMetadata metadata;
+  metadata.SetWidth(width);
+  metadata.SetHeight(height);
+  if (generic) {
+    metadata.SetFrameId(generic->frame_id);
+    metadata.SetSpatialIndex(generic->spatial_index);
+    metadata.SetTemporalIndex(generic->temporal_index);
+    metadata.SetFrameDependencies(generic->dependencies);
+    metadata.SetDecodeTargetIndications(generic->decode_target_indications);
+  }
+  return metadata;
+}
+
 }  // namespace webrtc

modules/rtp_rtcp/source/rtp_video_header.h

@@ -21,6 +21,7 @@
 #include "api/video/color_space.h"
 #include "api/video/video_codec_type.h"
 #include "api/video/video_content_type.h"
+#include "api/video/video_frame_metadata.h"
 #include "api/video/video_frame_type.h"
 #include "api/video/video_rotation.h"
 #include "api/video/video_timing.h"
@@ -62,6 +63,9 @@ struct RTPVideoHeader {
 
   ~RTPVideoHeader();
 
+  // The subset of RTPVideoHeader that is exposed in the Insertable Streams API.
+  VideoFrameMetadata GetAsMetadata() const;
+
   absl::optional<GenericDescriptorInfo> generic;
 
   VideoFrameType frame_type = VideoFrameType::kEmptyFrame;

video/rtp_video_stream_receiver_frame_transformer_delegate.cc

@@ -26,7 +26,7 @@ class TransformableVideoReceiverFrame
   TransformableVideoReceiverFrame(std::unique_ptr<RtpFrameObject> frame,
                                   uint32_t ssrc)
       : frame_(std::move(frame)),
-        metadata_(frame_->GetRtpVideoHeader()),
+        metadata_(frame_->GetRtpVideoHeader().GetAsMetadata()),
         ssrc_(ssrc) {}
   ~TransformableVideoReceiverFrame() override = default;