bool BufferSinkFilterContext::getAudioSamples(AudioSamples &samples, size_t samplesCount, OptionalErrorCode ec)
{
if (m_type != FilterMediaType::Audio) {
throws_if(ec, Errors::IncorrectBufferSinkMediaType);
return false;
}
return getSamples(samples.raw(), samplesCount, ec);
}
bool BufferSinkFilterContext::getAudioFrame(AudioSamples &samples, int flags, OptionalErrorCode ec)
{
if (m_type != FilterMediaType::Audio) {
throws_if(ec, Errors::IncorrectBufferSinkMediaType);
return false;
}
return getFrame(samples.raw(), flags, ec);
}
void AudioResampler::push(const AudioSamples &src, error_code &ec)
{
if (!m_raw)
{
fflog(AV_LOG_ERROR, "SwrContext does not inited\n");
throws_if(ec, Errors::ResamplerNotInited);
return;
}
// Null samples is allowed
if (src)
{
if (src.sampleRate() != srcSampleRate() ||
src.sampleFormat() != srcSampleFormat() ||
src.channelsCount() != srcChannels() ||
src.channelsLayout() != srcChannelLayout())
{
throws_if(ec, Errors::ResamplerInputChanges);
return;
}
}
auto sts = swr_convert_frame(m_raw, nullptr, src.raw());
if (sts < 0)
{
fflog(AV_LOG_DEBUG, "Src is null: %d, payload: %p\n", src.isNull(), src.data());
throws_if(ec, sts, ffmpeg_category());
return;
}
// TODO: need protection if we still work in scheme: One Resampler Per Channel
m_streamIndex = src.streamIndex();
// Need to restore PTS in output frames
if (m_prevPts > src.pts()) // Reset case
m_nextPts = Timestamp();
m_prevPts = src.pts();
//auto result = swr_get_delay(m_raw, m_dstRate);
//clog << " delay [push]: " << result << endl;
}
AudioSamples AudioDecoderContext::decode(const Packet &inPacket, size_t offset, OptionalErrorCode ec)
{
clear_if(ec);
AudioSamples outSamples;
int gotFrame = 0;
auto st = decodeCommon(outSamples, inPacket, offset, gotFrame, avcodec_decode_audio_legacy);
if (get<1>(st))
{
throws_if(ec, get<0>(st), *get<1>(st));
return AudioSamples();
}
if (!gotFrame)
{
outSamples.setComplete(false);
}
// Fix channels layout
if (outSamples.channelsCount() && !outSamples.channelsLayout())
av_frame_set_channel_layout(outSamples.raw(), av_get_default_channel_layout(outSamples.channelsCount()));
return outSamples;
}
int
AudioResampler :: resample(IAudioSamples * pOutSamples,
IAudioSamples* pInSamples,
unsigned int numSamples)
{
int retval = -1;
AudioSamples* outSamples = dynamic_cast<AudioSamples*>(pOutSamples);
AudioSamples* inSamples = dynamic_cast<AudioSamples*>(pInSamples);
unsigned int sampleSize=0;
try {
if (!outSamples)
throw std::invalid_argument("no output samples");
if (outSamples == inSamples)
throw std::invalid_argument("resampling into the same IAudioSamples is not allowed");
// null out the output samples.
outSamples->setComplete(false, 0, mOSampleRate, mOChannels,
mOFmt, Global::NO_PTS);
if (inSamples)
{
if (!inSamples->isComplete())
throw std::invalid_argument("input samples are not complete");
if (inSamples->getSampleRate() != mISampleRate)
throw std::invalid_argument("unexpected input sample rate");
if (inSamples->getChannels() != mIChannels)
throw std::invalid_argument("unexpected # of input channels");
if (inSamples->getFormat() != mIFmt)
throw std::invalid_argument("unexpected sample format");
if (numSamples == 0)
numSamples = inSamples->getNumSamples();
else
numSamples = FFMIN(numSamples, inSamples->getNumSamples());
sampleSize = inSamples->getSampleBitDepth()/8;
} else {
numSamples = 0;
sampleSize = IAudioSamples::findSampleBitDepth(mIFmt)/8;
}
int32_t neededSamples = getMinimumNumSamplesRequiredInOutputSamples(numSamples);
int32_t bytesPerOutputSample = mOChannels*IAudioSamples::findSampleBitDepth(mOFmt)/8;
int32_t neededBytes = neededSamples * bytesPerOutputSample;
// This causes a buffer resize to occur if needed
if (outSamples->ensureCapacity(neededBytes) < 0)
throw std::runtime_error("attempted to resize output buffer but failed");
uint32_t outBufSize = outSamples->getMaxBufferSize();
int32_t gap = (neededSamples*bytesPerOutputSample)-outBufSize;
if (gap > 0) {
// VS_LOG_ERROR("maxBufferSize: %d; neededSampleRoom: %d; sampleSize: %d; numSamples: %d; conversionRatio: %f;",
// (int32_t)outSamples->getMaxBufferSize(),
// neededSampleRoom,
// sampleSize,
// numSamples,
// conversionRatio);
(void) sampleSize; // to avoid a -Werror error
throw std::invalid_argument("not enough room in output buffer");
}
short * inBuf = inSamples ? inSamples->getRawSamples(0) : 0;
short *outBuf = outSamples->getRawSamples(0);
if (!outBuf)
throw std::invalid_argument("could not get output bytes");
VS_ASSERT(mContext, "Should have been set at initialization");
if (!mContext)
throw std::invalid_argument("programmer error");
// Now we should be far enough along that we can safely try a resample.
retval = audio_resample(mContext, outBuf, inBuf, numSamples);
#if 0
if (retval >0){
char string[2048*16+1];
unsigned int i=0;
for (i = 0; i < sizeof(string)-1; i++)
string[i] = 'X';
bool allZero = true;
for (i=0; i< FFMIN(numSamples, 2000);i++)
{
snprintf(string+(5*i), sizeof(string)-5*i, "%04hX.", inBuf[i]);
if (inBuf[i] != 0)
allZero = false;
}
VS_LOG_DEBUG("Input Buffer (%d): %s", numSamples, string);
for (i=0; i< FFMIN((unsigned int)retval, 2000);i++)
{
snprintf(string+(9*i), sizeof(string)-9*i, "%04hX%04hX.",
outBuf[2*i], outBuf[2*i+1]);
if (outBuf[2*i] != 0 || outBuf[2*i+1] != 0)
allZero = false;
}
VS_LOG_DEBUG("Output Buffer (%d): %s", retval, string);
if (!allZero)
VS_LOG_DEBUG("Got an audio buffer with content");
}
#endif // 0
if (retval >= 0)
{
// copy the Pts
int64_t pts = Global::NO_PTS;
if (inSamples)
{
pts = inSamples->getPts();
mNextPts = pts + IAudioSamples::samplesToDefaultPts(retval, mOSampleRate);
}
else
{
pts = mNextPts;
}
if (pts != Global::NO_PTS)
pts += mPtsOffset;
outSamples->setComplete(true, retval,
mOSampleRate, mOChannels,
mOFmt,
pts);
int expectedSamples = 0;
if (inSamples)
{
double top = mOSampleRate;
double bottom = mISampleRate;
double sampleOnlyConverstionRatio = top / bottom;
expectedSamples = (int)(numSamples * sampleOnlyConverstionRatio);
}
else
{
VS_LOG_TRACE("Got null samples; outputted all cached and set pts offset from %lld to 0",
mPtsOffset);
expectedSamples = retval;
// and reset the offset
mPtsOffset = 0;
}
if (retval != expectedSamples)
{
// we got a different number of samples than expected; we need to update
// our pts offset
int sampleDelta = retval - expectedSamples;
int64_t ptsDelta = IAudioSamples::samplesToDefaultPts(sampleDelta, mOSampleRate);
mPtsOffset += ptsDelta;
}
}
}
catch (std::invalid_argument & e)
{
VS_LOG_DEBUG("invalid argument: %s", e.what());
retval = -1;
}
return retval;
}
void FormatContext::writeUncodedFrameDirect(AudioSamples &frame, size_t streamIndex, OptionalErrorCode ec)
{
writeFrame(frame.raw(), streamIndex, ec, av_write_uncoded_frame);
}
bool AudioResampler::pop(AudioSamples &dst, bool getall, error_code &ec)
{
clear_if(ec);
if (!m_raw) {
fflog(AV_LOG_ERROR, "SwrContext does not inited\n");
throws_if(ec, Errors::ResamplerNotInited);
return false;
}
if (dst.sampleRate() != dstSampleRate() ||
dst.sampleFormat() != dstSampleFormat() ||
dst.channelsCount() != dstChannels() ||
dst.channelsLayout() != dstChannelLayout())
{
throws_if(ec, Errors::ResamplerOutputChanges);
return false;
}
auto result = swr_get_delay(m_raw, m_dstRate);
//clog << " delay [pop]: " << result << endl;
// Need more data
if (result < dst.samplesCount() && getall == false)
{
return false;
}
auto sts = swr_convert_frame(m_raw, dst.raw(), nullptr);
if (sts < 0)
{
throws_if(ec, sts, ffmpeg_category());
return false;
}
dst.setTimeBase(Rational(1, m_dstRate));
dst.setStreamIndex(m_streamIndex);
dst.setComplete(true);
// Ugly PTS handling. More clean one can be done by user code
if (!m_nextPts.isValid()) {
m_nextPts = Timestamp(0, dst.timeBase());
}
dst.setPts(m_nextPts);
m_nextPts = dst.pts() + Timestamp{dst.samplesCount(), dst.timeBase()};
//result = swr_get_delay(m_raw, m_dstRate);
//clog << " delay [pop]: " << result << endl;
// When no data, samples count sets to zero
return dst.samplesCount() ? true : false;
}