Merge pull request #1339 from CastagnaIT/webvtt_segments

[DashTree] Cleanup subtitles stream/container detection
xbmc · Jul 29, 2023 · 37705b2 · 37705b2
2 parents b3bfc65 + 66e75e7
commit 37705b2
Show file tree

Hide file tree

Showing 4 changed files with 58 additions and 20 deletions.
diff --git a/src/parser/DASHTree.cpp b/src/parser/DASHTree.cpp
@@ -67,16 +67,18 @@ PLAYLIST::ContainerType DetectContainerType(std::string_view mimeType)
     return ContainerType::WEBM;
   if (STRING::Contains(mimeType, "/x-matroska"))
     return ContainerType::MATROSKA;
+  if (STRING::Contains(mimeType, "/ttml+xml") || STRING::Contains(mimeType, "vtt"))
+    return ContainerType::TEXT;
 
   return ContainerType::MP4;
 }
 
 std::string DetectCodecFromMimeType(std::string_view mimeType)
 {
   if (mimeType == "text/vtt")
-    return "wvtt";
+    return CODEC::FOURCC_WVTT;
   if (mimeType == "application/ttml+xml")
-    return "ttml";
+    return CODEC::FOURCC_TTML;
 
   return "";
 }
@@ -683,28 +685,24 @@ void adaptive::CDashTree::ParseTagRepresentation(pugi::xml_node nodeRepr,
   {
     StreamType streamType = DetectStreamType("", repr->GetMimeType());
     const auto& codecs = repr->GetCodecs();
-    if (streamType == StreamType::NOTYPE && (CODEC::Contains(codecs, CODEC::FOURCC_WVTT) ||
-                                             CODEC::Contains(codecs, CODEC::FOURCC_TTML) ||
-                                             CODEC::Contains(codecs, CODEC::FOURCC_STPP)))
+    if (streamType == StreamType::NOTYPE)
     {
-      streamType = StreamType::SUBTITLE;
+      // Try find stream type by checking the codec string
+      for (const std::string& codec : codecs)
+      {
+        if (CODEC::IsSubtitleFourCC(codec))
+        {
+          streamType = StreamType::SUBTITLE;
+          break;
+        }
+      }
     }
+
     adpSet->SetStreamType(streamType);
-  }
 
-  // Set properties for subtitles types
-  if (repr->GetMimeType() != "application/mp4") // Handle text type only, not ISOBMFF format
-  {
-    const auto& codecs = repr->GetCodecs();
-    if (repr->GetMimeType() == "application/ttml+xml" || repr->GetMimeType() == "text/vtt" ||
-        CODEC::Contains(codecs, CODEC::FOURCC_WVTT) ||
-        CODEC::Contains(codecs, CODEC::FOURCC_TTML) || CODEC::Contains(codecs, CODEC::FOURCC_STPP))
-    {
-      if (adpSet->SegmentTimelineDuration().IsEmpty())
-        repr->SetIsSubtitleFileStream(true); // Treat as single subtitle file
-      else
-        repr->SetContainerType(ContainerType::TEXT); // Segmented subtitles
-    }
+    if (streamType == StreamType::SUBTITLE &&
+        repr->GetMimeType() != "application/mp4") // Text format type only, not ISOBMFF
+      repr->SetContainerType(ContainerType::TEXT);
   }
 
   // ISA custom attribute
@@ -972,6 +970,15 @@ void adaptive::CDashTree::ParseTagRepresentation(pugi::xml_node nodeRepr,
   else if (adpSet->GetStreamType() == StreamType::AUDIO && repr->GetAudioChannels() == 0)
     repr->SetAudioChannels(2); // Fallback to 2 channels when no value is set
 
+  // For subtitles that are not as ISOBMFF format and where there is no timeline for segments
+  // we should treat them as a single subtitle file
+  if (repr->GetContainerType() == ContainerType::TEXT && repr->GetMimeType() != "application/mp4" &&
+      !adpSet->HasSegmentTimelineDuration() && !repr->HasSegmentTimeline())
+  {
+
+    repr->SetIsSubtitleFileStream(true);
+  }
+
   // Generate timeline segments
   if (!repr->HasSegmentTimeline() && repr->HasSegmentTemplate())
   {

diff --git a/src/test/TestDASHTree.cpp b/src/test/TestDASHTree.cpp
@@ -409,46 +409,57 @@ TEST_F(DASHTreeAdaptiveStreamTest, subtitles)
   EXPECT_EQ(adpSets[1]->GetStreamType(), PLAYLIST::StreamType::SUBTITLE);
   EXPECT_EQ(adpSets[1]->GetRepresentations()[0]->IsSubtitleFileStream(), true);
   EXPECT_EQ(CODEC::Contains(adpSets[1]->GetRepresentations()[0]->GetCodecs(), CODEC::FOURCC_TTML), true);
+  EXPECT_EQ(adpSets[1]->GetRepresentations()[0]->GetContainerType(), PLAYLIST::ContainerType::TEXT);
 
   EXPECT_EQ(adpSets[2]->GetStreamType(), PLAYLIST::StreamType::SUBTITLE);
   EXPECT_EQ(adpSets[2]->GetRepresentations()[0]->IsSubtitleFileStream(), true);
   EXPECT_EQ(CODEC::Contains(adpSets[2]->GetRepresentations()[0]->GetCodecs(), CODEC::FOURCC_TTML), true);
+  EXPECT_EQ(adpSets[2]->GetRepresentations()[0]->GetContainerType(), PLAYLIST::ContainerType::TEXT);
 
   EXPECT_EQ(adpSets[3]->GetStreamType(), PLAYLIST::StreamType::SUBTITLE);
   EXPECT_EQ(adpSets[3]->GetRepresentations()[0]->IsSubtitleFileStream(), true);
   EXPECT_EQ(CODEC::Contains(adpSets[3]->GetRepresentations()[0]->GetCodecs(), CODEC::FOURCC_TTML), true);
+  EXPECT_EQ(adpSets[3]->GetRepresentations()[0]->GetContainerType(), PLAYLIST::ContainerType::TEXT);
 
   EXPECT_EQ(adpSets[4]->GetStreamType(), PLAYLIST::StreamType::SUBTITLE);
   EXPECT_EQ(adpSets[4]->GetRepresentations()[0]->IsSubtitleFileStream(), true);
   EXPECT_EQ(CODEC::Contains(adpSets[4]->GetRepresentations()[0]->GetCodecs(), CODEC::FOURCC_TTML), true);
+  EXPECT_EQ(adpSets[4]->GetRepresentations()[0]->GetContainerType(), PLAYLIST::ContainerType::TEXT);
 
   EXPECT_EQ(adpSets[5]->GetStreamType(), PLAYLIST::StreamType::SUBTITLE);
   EXPECT_EQ(adpSets[5]->GetRepresentations()[0]->IsSubtitleFileStream(), true);
   EXPECT_EQ(CODEC::Contains(adpSets[5]->GetRepresentations()[0]->GetCodecs(), CODEC::FOURCC_WVTT), true);
+  EXPECT_EQ(adpSets[5]->GetRepresentations()[0]->GetContainerType(), PLAYLIST::ContainerType::TEXT);
 
   EXPECT_EQ(adpSets[6]->GetStreamType(), PLAYLIST::StreamType::SUBTITLE);
   EXPECT_EQ(adpSets[6]->GetRepresentations()[0]->IsSubtitleFileStream(), true);
   EXPECT_EQ(CODEC::Contains(adpSets[6]->GetRepresentations()[0]->GetCodecs(), CODEC::FOURCC_WVTT), true);
+  EXPECT_EQ(adpSets[6]->GetRepresentations()[0]->GetContainerType(), PLAYLIST::ContainerType::TEXT);
 
   EXPECT_EQ(adpSets[7]->GetStreamType(), PLAYLIST::StreamType::SUBTITLE);
   EXPECT_EQ(adpSets[7]->GetRepresentations()[0]->IsSubtitleFileStream(), true);
   EXPECT_EQ(CODEC::Contains(adpSets[7]->GetRepresentations()[0]->GetCodecs(), CODEC::FOURCC_WVTT), true);
+  EXPECT_EQ(adpSets[7]->GetRepresentations()[0]->GetContainerType(), PLAYLIST::ContainerType::TEXT);
 
   EXPECT_EQ(adpSets[8]->GetStreamType(), PLAYLIST::StreamType::SUBTITLE);
   EXPECT_EQ(adpSets[8]->GetRepresentations()[0]->IsSubtitleFileStream(), true);
   EXPECT_EQ(CODEC::Contains(adpSets[8]->GetRepresentations()[0]->GetCodecs(), CODEC::FOURCC_WVTT), true);
+  EXPECT_EQ(adpSets[8]->GetRepresentations()[0]->GetContainerType(), PLAYLIST::ContainerType::TEXT);
 
   EXPECT_EQ(adpSets[9]->GetStreamType(), PLAYLIST::StreamType::SUBTITLE);
   EXPECT_EQ(adpSets[9]->GetRepresentations()[0]->IsSubtitleFileStream(), true);
   EXPECT_EQ(CODEC::Contains(adpSets[9]->GetRepresentations()[0]->GetCodecs(), "my_codec"), true);
+  EXPECT_EQ(adpSets[9]->GetRepresentations()[0]->GetContainerType(), PLAYLIST::ContainerType::TEXT);
 
   EXPECT_EQ(adpSets[10]->GetStreamType(), PLAYLIST::StreamType::SUBTITLE);
   EXPECT_EQ(adpSets[10]->GetRepresentations()[0]->IsSubtitleFileStream(), true);
   EXPECT_EQ(CODEC::Contains(adpSets[10]->GetRepresentations()[0]->GetCodecs(), CODEC::FOURCC_TTML), true);
+  EXPECT_EQ(adpSets[10]->GetRepresentations()[0]->GetContainerType(), PLAYLIST::ContainerType::TEXT);
 
   EXPECT_EQ(adpSets[11]->GetStreamType(), PLAYLIST::StreamType::SUBTITLE);
   EXPECT_EQ(STR(adpSets[11]->GetRepresentations()[0]->GetMimeType()), "application/mp4");
   EXPECT_EQ(CODEC::Contains(adpSets[11]->GetRepresentations()[0]->GetCodecs(), CODEC::FOURCC_STPP), true);
+  EXPECT_EQ(adpSets[11]->GetRepresentations()[0]->GetContainerType(), PLAYLIST::ContainerType::MP4);
 
   SetTestStream(NewStream(adpSets[11].get()));
   testStream->start_stream();
@@ -459,6 +470,7 @@ TEST_F(DASHTreeAdaptiveStreamTest, subtitles)
   EXPECT_EQ(adpSets[12]->GetStreamType(), PLAYLIST::StreamType::SUBTITLE);
   EXPECT_EQ(STR(adpSets[12]->GetMimeType()), "application/mp4");
   EXPECT_EQ(CODEC::Contains(adpSets[12]->GetRepresentations()[0]->GetCodecs(), "stpp.ttml.im1t"), true);
+  EXPECT_EQ(adpSets[12]->GetRepresentations()[0]->GetContainerType(), PLAYLIST::ContainerType::MP4);
 
   SetTestStream(NewStream(adpSets[12].get()));
   testStream->start_stream();

diff --git a/src/utils/Utils.cpp b/src/utils/Utils.cpp
@@ -400,3 +400,13 @@ bool UTILS::CODEC::IsAudio(std::string_view codec)
   }
   return false;
 }
+
+bool UTILS::CODEC::IsSubtitleFourCC(std::string_view codec)
+{
+  for (const auto fourcc : CODEC::SUBTITLES_FOURCC_LIST)
+  {
+    if (STRING::Contains(codec, fourcc))
+      return true;
+  }
+  return false;
+}
diff --git a/src/utils/Utils.h b/src/utils/Utils.h
@@ -115,6 +115,8 @@ constexpr char* FOURCC_TTML = "ttml";
 // for example "stpp.ttml.im1t", or only "stpp"
 constexpr char* FOURCC_STPP = "stpp";
 
+constexpr std::array SUBTITLES_FOURCC_LIST = {FOURCC_WVTT, FOURCC_TTML, FOURCC_STPP};
+
 /*!
  * \brief Make a FourCC code as unsigned integer value
  * \param fourcc The FourCC code (4 chars)
@@ -166,6 +168,13 @@ std::string GetVideoDesc(const std::set<std::string>& list);
  */
 bool IsAudio(std::string_view codec);
 
+/*!
+ * \brief Determines if the codec string contains a fourcc of type subtitles.
+ * \param codec The codec string
+ * \return True if contains a fourcc of type subtitles, otherwise false
+ */
+bool IsSubtitleFourCC(std::string_view codec);
+
 }
 
 } // namespace UTILS