Skip to content

Consistently use ptsToSeconds #784

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
Jul 29, 2025
Merged

Consistently use ptsToSeconds #784

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
d83491d
consistently use ptsToSeconds
Jul 23, 2025
fffd499
use EXPECT_DOUBLE_EQ for ptsSeconds
Jul 25, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
23 changes: 11 additions & 12 deletions src/torchcodec/_core/SingleStreamDecoder.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,7 +18,7 @@ namespace facebook::torchcodec {
namespace {

double ptsToSeconds(int64_t pts, const AVRational& timeBase) {
return static_cast<double>(pts) * timeBase.num / timeBase.den;
return static_cast<double>(pts) * av_q2d(timeBase);
}

int64_t secondsToClosestPts(double seconds, const AVRational& timeBase) {
Expand Down Expand Up @@ -129,11 +129,11 @@ void SingleStreamDecoder::initializeDecoder() {

if (avStream->duration > 0 && avStream->time_base.den > 0) {
streamMetadata.durationSecondsFromHeader =
av_q2d(avStream->time_base) * avStream->duration;
ptsToSeconds(avStream->duration, avStream->time_base);
}
if (avStream->start_time != AV_NOPTS_VALUE) {
streamMetadata.beginStreamSecondsFromHeader =
av_q2d(avStream->time_base) * avStream->start_time;
ptsToSeconds(avStream->start_time, avStream->time_base);
}

if (avStream->codecpar->codec_type == AVMEDIA_TYPE_VIDEO) {
Expand Down Expand Up @@ -304,13 +304,12 @@ void SingleStreamDecoder::scanFileAndUpdateMetadataAndIndex() {
streamInfos_[streamIndex].allFrames.size();

if (streamMetadata.beginStreamPtsFromContent.has_value()) {
streamMetadata.beginStreamPtsSecondsFromContent =
*streamMetadata.beginStreamPtsFromContent *
av_q2d(avStream->time_base);
streamMetadata.beginStreamPtsSecondsFromContent = ptsToSeconds(
*streamMetadata.beginStreamPtsFromContent, avStream->time_base);
}
if (streamMetadata.endStreamPtsFromContent.has_value()) {
streamMetadata.endStreamPtsSecondsFromContent =
*streamMetadata.endStreamPtsFromContent * av_q2d(avStream->time_base);
streamMetadata.endStreamPtsSecondsFromContent = ptsToSeconds(
*streamMetadata.endStreamPtsFromContent, avStream->time_base);
}
}

Expand Down Expand Up @@ -344,11 +343,11 @@ void SingleStreamDecoder::readCustomFrameMappingsUpdateMetadataAndIndex(
all_frames[-1].item<int64_t>() + duration[-1].item<int64_t>();

auto avStream = formatContext_->streams[streamIndex];
streamMetadata.beginStreamPtsSecondsFromContent =
*streamMetadata.beginStreamPtsFromContent * av_q2d(avStream->time_base);
streamMetadata.beginStreamPtsSecondsFromContent = ptsToSeconds(
*streamMetadata.beginStreamPtsFromContent, avStream->time_base);

streamMetadata.endStreamPtsSecondsFromContent =
*streamMetadata.endStreamPtsFromContent * av_q2d(avStream->time_base);
streamMetadata.endStreamPtsSecondsFromContent = ptsToSeconds(
*streamMetadata.endStreamPtsFromContent, avStream->time_base);

streamMetadata.numFramesFromContent = all_frames.size(0);
for (int64_t i = 0; i < all_frames.size(0); ++i) {
Expand Down
16 changes: 8 additions & 8 deletions test/VideoDecoderTest.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -267,9 +267,9 @@ TEST_P(SingleStreamDecoderTest, SeeksCloseToEof) {
ourDecoder->addVideoStream(-1);
ourDecoder->setCursorPtsInSeconds(388388. / 30'000);
auto output = ourDecoder->getNextFrame();
EXPECT_EQ(output.ptsSeconds, 388'388. / 30'000);
EXPECT_DOUBLE_EQ(output.ptsSeconds, 388'388. / 30'000);
output = ourDecoder->getNextFrame();
EXPECT_EQ(output.ptsSeconds, 389'389. / 30'000);
EXPECT_DOUBLE_EQ(output.ptsSeconds, 389'389. / 30'000);
EXPECT_THROW(ourDecoder->getNextFrame(), std::exception);
}

Expand Down Expand Up @@ -300,7 +300,7 @@ TEST_P(SingleStreamDecoderTest, GetsFramePlayedAtTimestamp) {
// Sanity check: make sure duration is strictly positive.
EXPECT_GT(kPtsPlusDurationOfLastFrame, kPtsOfLastFrameInVideoStream);
output = ourDecoder->getFramePlayedAt(kPtsPlusDurationOfLastFrame - 1e-6);
EXPECT_EQ(output.ptsSeconds, kPtsOfLastFrameInVideoStream);
EXPECT_DOUBLE_EQ(output.ptsSeconds, kPtsOfLastFrameInVideoStream);
}

TEST_P(SingleStreamDecoderTest, SeeksToFrameWithSpecificPts) {
Expand All @@ -311,7 +311,7 @@ TEST_P(SingleStreamDecoderTest, SeeksToFrameWithSpecificPts) {
ourDecoder->setCursorPtsInSeconds(6.0);
auto output = ourDecoder->getNextFrame();
torch::Tensor tensor6FromOurDecoder = output.data;
EXPECT_EQ(output.ptsSeconds, 180'180. / 30'000);
EXPECT_DOUBLE_EQ(output.ptsSeconds, 180'180. / 30'000);
torch::Tensor tensor6FromFFMPEG =
readTensorFromDisk("nasa_13013.mp4.time6.000000.pt");
EXPECT_TRUE(torch::equal(tensor6FromOurDecoder, tensor6FromFFMPEG));
Expand All @@ -327,7 +327,7 @@ TEST_P(SingleStreamDecoderTest, SeeksToFrameWithSpecificPts) {
ourDecoder->setCursorPtsInSeconds(6.1);
output = ourDecoder->getNextFrame();
torch::Tensor tensor61FromOurDecoder = output.data;
EXPECT_EQ(output.ptsSeconds, 183'183. / 30'000);
EXPECT_DOUBLE_EQ(output.ptsSeconds, 183'183. / 30'000);
torch::Tensor tensor61FromFFMPEG =
readTensorFromDisk("nasa_13013.mp4.time6.100000.pt");
EXPECT_TRUE(torch::equal(tensor61FromOurDecoder, tensor61FromFFMPEG));
Expand All @@ -347,7 +347,7 @@ TEST_P(SingleStreamDecoderTest, SeeksToFrameWithSpecificPts) {
ourDecoder->setCursorPtsInSeconds(10.0);
output = ourDecoder->getNextFrame();
torch::Tensor tensor10FromOurDecoder = output.data;
EXPECT_EQ(output.ptsSeconds, 300'300. / 30'000);
EXPECT_DOUBLE_EQ(output.ptsSeconds, 300'300. / 30'000);
torch::Tensor tensor10FromFFMPEG =
readTensorFromDisk("nasa_13013.mp4.time10.000000.pt");
EXPECT_TRUE(torch::equal(tensor10FromOurDecoder, tensor10FromFFMPEG));
Expand All @@ -364,7 +364,7 @@ TEST_P(SingleStreamDecoderTest, SeeksToFrameWithSpecificPts) {
ourDecoder->setCursorPtsInSeconds(6.0);
output = ourDecoder->getNextFrame();
tensor6FromOurDecoder = output.data;
EXPECT_EQ(output.ptsSeconds, 180'180. / 30'000);
EXPECT_DOUBLE_EQ(output.ptsSeconds, 180'180. / 30'000);
EXPECT_TRUE(torch::equal(tensor6FromOurDecoder, tensor6FromFFMPEG));
EXPECT_EQ(ourDecoder->getDecodeStats().numSeeksAttempted, 1);
// We cannot skip a seek here because timestamp=6 has a different keyframe
Expand All @@ -379,7 +379,7 @@ TEST_P(SingleStreamDecoderTest, SeeksToFrameWithSpecificPts) {
ourDecoder->setCursorPtsInSeconds(kPtsOfLastFrameInVideoStream);
output = ourDecoder->getNextFrame();
torch::Tensor tensor7FromOurDecoder = output.data;
EXPECT_EQ(output.ptsSeconds, 389'389. / 30'000);
EXPECT_DOUBLE_EQ(output.ptsSeconds, 389'389. / 30'000);
torch::Tensor tensor7FromFFMPEG =
readTensorFromDisk("nasa_13013.mp4.time12.979633.pt");
EXPECT_TRUE(torch::equal(tensor7FromOurDecoder, tensor7FromFFMPEG));
Expand Down
4 changes: 2 additions & 2 deletions test/test_ops.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -751,7 +751,7 @@ def test_next(self, asset):
frame, asset.get_frame_data_by_index(frame_index)
)
frame_info = asset.get_frame_info(frame_index)
assert pts_seconds == frame_info.pts_seconds
assert pts_seconds == pytest.approx(frame_info.pts_seconds)
assert duration_seconds == frame_info.duration_seconds
frame_index += 1

Expand Down Expand Up @@ -955,7 +955,7 @@ def test_pts(self, asset):
frames, asset.get_frame_data_by_index(frame_index)
)

assert pts_seconds == start_seconds
assert pts_seconds == pytest.approx(start_seconds)
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'm surprised we need to use approx so just to be on the safe side I removed it and we're getting:

FAILED test/test_ops.py::TestAudioDecoderOps::test_next[asset0] - assert 0.5760000000000001 == 0.576
FAILED test/test_ops.py::TestAudioDecoderOps::test_next[asset1] - assert 0.9359999999999999 == 0.936
FAILED test/test_ops.py::TestAudioDecoderOps::test_pts[asset0] - assert 0.5760000000000001 == 0.576
FAILED test/test_ops.py::TestAudioDecoderOps::test_pts[asset1] - assert 0.9359999999999999 == 0.936

Which is probably OK. But I still find it surprising considering the implementation of av_q2d is just

static inline double av_q2d(AVRational a){
     return a.num / (double) a.den;
 }


def test_sample_rate_conversion(self):
def get_all_frames(asset, sample_rate=None, stop_seconds=None):
Expand Down
Loading