Playing the same sound several times requires allocating a new buffer each time

[Darksonn]

I've been looking at the various Rodio sources for a while and I'm unable to find any sources that allow reusing a buffer of sound.

The closest solution appears to be using SamplesBuffer and cloning the vector each time, but this is inefficient as it requires an allocation and deallocation for each played sound.

I believe this crate should contain a source that wraps a vector of samples in an Arc, allowing the allocated memory to be reused.

I've attached a sample implementation of such a source: main.rs.

[Xaeroxe]

@Darksonn Hey! We had a similar problem in the Amethyst game engine. What I chose to do was create an Arc of the raw encoded file data (Arc<Vec<u8>>) and then implement std::convert::AsRef<[u8]> for a wrapper structure over that Arc, then when I needed to play the sound I'd initialize a new std::io::Cursor over that structure and feed it into a new rodio::Decoder. I like this approach because it allows the in-memory representation to stay compressed with the original audio compression algorithm, audio compression algorithms can reduce the memory footprint of a sound by up to 10x in some instances.

[sinesc]

I think this is a minimal example of the above approach. Thanks @Xaeroxe

use rodio;
use std::io;
use std::convert::AsRef;

pub struct Sound (Arc<Vec<u8>>);

impl AsRef<[u8]> for Sound {
    fn as_ref(&self) -> &[u8] {
        &self.0
    }
}

impl Sound {
    pub fn load(filename: &str) -> io::Result<Sound> {
        use std::fs::File;
        let mut buf = Vec::new();
        let mut file = File::open(filename)?;
        file.read_to_end(&mut buf)?;
        Ok(Sound(Arc::new(buf)))
    }
    pub fn cursor(self: &Self) -> io::Cursor<Sound> {
        io::Cursor::new(Sound(self.0.clone()))
    }
    pub fn decoder(self: &Self) -> rodio::Decoder<io::Cursor<Sound>> {
        rodio::Decoder::new(self.cursor()).unwrap()
    }
}

...

rodio::play_raw(&device, sound.decoder().convert_samples());