fix tests and move seekmode to metadata.h

Author: Molly Xu

src/torchcodec/_core/Metadata.cpp

@@ -14,12 +14,8 @@ std::optional<double> StreamMetadata::getDurationSeconds(
     case SeekMode::custom_frame_mappings:
     case SeekMode::exact:
       // In exact mode, use the scanned content value
-      if (endStreamPtsSecondsFromContent.has_value() &&
-          beginStreamPtsSecondsFromContent.has_value()) {
-        return endStreamPtsSecondsFromContent.value() -
-            beginStreamPtsSecondsFromContent.value();
-      }
-      return std::nullopt;
+      return endStreamPtsSecondsFromContent.value() -
+          beginStreamPtsSecondsFromContent.value();
     case SeekMode::approximate:
       if (durationSecondsFromHeader.has_value()) {
         return durationSecondsFromHeader.value();
@@ -38,10 +34,7 @@ double StreamMetadata::getBeginStreamSeconds(SeekMode seekMode) const {
   switch (seekMode) {
     case SeekMode::custom_frame_mappings:
     case SeekMode::exact:
-      if (beginStreamPtsSecondsFromContent.has_value()) {
-        return beginStreamPtsSecondsFromContent.value();
-      }
-      return 0.0;
+      return beginStreamPtsSecondsFromContent.value();
     case SeekMode::approximate:
       return 0.0;
   }
@@ -53,10 +46,7 @@ std::optional<double> StreamMetadata::getEndStreamSeconds(
   switch (seekMode) {
     case SeekMode::custom_frame_mappings:
     case SeekMode::exact:
-      if (endStreamPtsSecondsFromContent.has_value()) {
-        return endStreamPtsSecondsFromContent.value();
-      }
-      return getDurationSeconds(seekMode);
+      return endStreamPtsSecondsFromContent.value();
     case SeekMode::approximate:
       return getDurationSeconds(seekMode);
   }
@@ -67,10 +57,7 @@ std::optional<int64_t> StreamMetadata::getNumFrames(SeekMode seekMode) const {
   switch (seekMode) {
     case SeekMode::custom_frame_mappings:
     case SeekMode::exact:
-      if (numFramesFromContent.has_value()) {
-        return numFramesFromContent.value();
-      }
-      return std::nullopt;
+      return numFramesFromContent.value();
     case SeekMode::approximate: {
       if (numFramesFromHeader.has_value()) {
         return numFramesFromHeader.value();

src/torchcodec/_core/Metadata.h

@@ -10,8 +10,6 @@
 #include <string>
 #include <vector>
 
-#include "SeekMode.h"
-
 extern "C" {
 #include <libavcodec/avcodec.h>
 #include <libavutil/avutil.h>
@@ -20,6 +18,8 @@ extern "C" {
 
 namespace facebook::torchcodec {
 
+enum class SeekMode { exact, approximate, custom_frame_mappings };
+
 struct StreamMetadata {
   // Common (video and audio) fields derived from the AVStream.
   int streamIndex;

src/torchcodec/_core/SeekMode.h

@@ -16,7 +16,7 @@
 #include "DeviceInterface.h"
 #include "FFMPEGCommon.h"
 #include "Frame.h"
-#include "SeekMode.h"
+#include "Metadata.h"
 #include "StreamOptions.h"
 #include "Transform.h"
 

src/torchcodec/_core/SingleStreamDecoder.h

@@ -64,10 +64,10 @@ class SingleStreamDecoderTest : public testing::TestWithParam<bool> {
 
       auto contextHolder = std::make_unique<AVIOFromTensorContext>(tensor);
       return std::make_unique<SingleStreamDecoder>(
-          std::move(contextHolder), SingleStreamDecoder::SeekMode::approximate);
+          std::move(contextHolder), SeekMode::approximate);
     } else {
       return std::make_unique<SingleStreamDecoder>(
-          filepath, SingleStreamDecoder::SeekMode::approximate);
+          filepath, SeekMode::approximate);
     }
   }
 

test/VideoDecoderTest.cpp

@@ -186,7 +186,7 @@ def test_repr():
   bit_rate: 64000.0
   codec: mp3
   stream_index: 0
-  duration_seconds: 13.013
+  duration_seconds: {expected_duration_seconds_from_header}
   begin_stream_seconds: 0.0
   sample_rate: 8000
   num_channels: 2