Add time seek similar as ov_time_seek

[mrDIMAS]

Time seeking similar to ov_time_seek would be very useful. Or maybe it is already there?

[est31]

There is seeking support in lewton! The seek_absgp_pg function provides page-granularity seeking in terms of absolute granule position which is IIRC just the sample number, so you use the sample rate to convert to timestamps and back. As the function is in page granularity, you'll need to advance the decoder until you get to your desired position inside the page.

[mrDIMAS]

Hm, I tried it as seek_absgp_pg(0) and next calls to read_dec_packet_itl just returns zero-sized vectors, I tried it few times in a row (because first packet seems to be empty everytime) and all of them were empty.

[mrDIMAS]

As the function is in page granularity, you'll need to advance the decoder until you get to your desired position inside the page.

Maybe it not working because of this. I'm trying to "rewind" decoder to beginning of data after I read all the contents in source by just calling seek_absgp_pg(0). What does this: you'll need to advance the decoder mean exactly? How to do that?

[est31]

What does this: you'll need to advance the decoder mean exactly? How to do that?

I think currently if you use the ogg reader module, the only way is to call the packet decoding functions enough times while keeping count of the number of their samples.

[mrDIMAS]

Not sure if I get you right, do I need to track total amount of samples I read and then when I need to seek beginning I need to do something like (pseudocode)

while current_pos > 0 {
   current_pos -= current_packet_size
   seek_absgp_pg(current_pos)
}

Also when I do seek_absgp_pg(0) I get BadAudio(AudioIsHeader) error.

[est31]

No you need to track the number of samples read after you call seek_absgp_pg. Seeking always brings you to the page boundary, no matter where you were before. The seek_absgp_pg function brings you roughly into the area, but the sample precise fine tuning is up to you.

Also when I do seek_absgp_pg(0) I get BadAudio(AudioIsHeader) error.

Hmm yeah it makes sense that this happens. It might be a good idea to let lewton be responsible for this so maybe I should fix it. In the meantime, you can create a completely new OggStreamReader when you want to rewind completely.

[mrDIMAS]

Yeah, that would be awesome if it would be fixed on library side. New OggStreamReader instance is ok for now, but in future I want to be able to seek at any time pos I need. Simplest use case: set playback position for streaming sound after loaded a saved game, this is useful for long ambient sounds that eats too much memory when fully decoded.

I'm curious why seeking produces BadAudio(AudioIsHeader) error? Because internal reader was set to a header position, not a packet position?

[mrDIMAS]

Ok, this solution works (I wish rewind be less ugly)

pub struct OggDecoder {
    reader: Option<OggStreamReader<DataSource>>,
    samples: vec::IntoIter<i16>,
    channel_count: usize,
    sample_rate: usize,
}

fn i16_to_f32_sample(sample: i16) -> f32 {
    sample as f32 / 32767.0
}

impl Iterator for OggDecoder {
    type Item = f32;

    #[inline]
    fn next(&mut self) -> Option<Self::Item> {
        if let Some(sample) = self.samples.next() {
            Some(i16_to_f32_sample(sample))
        } else {
            self.samples = self.reader
                .as_mut()
                .unwrap()
                .read_dec_packet_itl()
                .ok()
                .and_then(|data| data)
                .unwrap_or_default()
                .into_iter();
            self.samples
                .next()
                .and_then(|v| Some(i16_to_f32_sample(v)))
        }
    }
}

impl OggDecoder {
    pub fn new(source: DataSource) -> Result<Self, SoundError> {
        let mut reader = OggStreamReader::new(source)?;
        Ok(Self {
            samples: reader.read_dec_packet_itl()?
                .ok_or(SoundError::UnsupportedFormat)?
                .into_iter(),
            channel_count: reader.ident_hdr.audio_channels as usize,
            sample_rate: reader.ident_hdr.audio_sample_rate as usize,
            reader: Some(reader),
        })
    }

    fn get_channel_count(&self) -> usize {
        self.channel_count
    }

    fn rewind(&mut self) -> Result<(), SoundError> {
        let mut source = self.reader
            .take()
            .unwrap()
            .into_inner()
            .into_inner();
        source.seek(SeekFrom::Start(0))?;
        *self = Self::new(source)?;
        Ok(())
    }

    fn get_sample_rate(&self) -> usize {
        self.sample_rate
    }
}

[est31]

Wonderful that you found a solution!

Note: you can avoid doing the conversion in i16_to_f32_sample by using the read_dec_packet_generic function and InterleavedSamples<f32> instead of read_dec_packet_itl

[mrDIMAS]

Great, thanks! Still API of OggStreamReader requires some improvements, my very first thought at beginning was - "wow, it is cool, I hope it will have similar API as libvorbis", but then I realized that is lacking high level functionality like seeking at time pos, which was a bit frustrating. Anyway, thanks for the great library!

Should this function be included into the library's API instead of having the user implement it themselves?
If it should, I'll try to make a PR.

[TonalidadeHidrica]

@JackRedstonia In my opinion (I'm not an owner) It'd be extremely useful, but it's not as simple as it may seem. I've been working on it for while, but I don't have time recently so it's not clear I'll make it. If you have time go ahead, and I can help you time to time.

At least, I really want that future.