Factory for SoundBuffer?

[labsound]

Factory for SoundBuffer?

[ddiakopoulos]

SoundBuffer is a real outlier to the rest of the codebase. I think it reveals an underlying problem there's no concept of an asset. To this end, I agree about the principled use of an asset loader / factory that owns and loans out the underlying resource: a library!

There might be a hybrid approach here. Assuming the use of a string id as a handle:

static std::map<std::string, std::shared_ptr<audio_asset>> asset_library;

I started thinking about this for another project and have a half-working impl here: https://github.com/ddiakopoulos/sandbox/blob/master/virtual-reality/assets.hpp ... I would then add a loader lambda that produces an underlying AudioBuffer instead of the simplistic register_asset(...) that uses an rvalue reference. In theory get_asset(...) could call the lamba to even enable lazy loading (although this is probably not that desirable in the context of realtime audio!)

The asset_library could even be templated on input type (ogg, wav, etc). This would make it maximally-platform independent and even libnyquist could be considered a non-core dependency.

[meshula]

That's promising. Your approach is very similar to Pixar's.

https://github.com/PixarAnimationStudios/USD/blob/master/pxr/usd/lib/ar/resolver.h

A resolver class like this could work cleanly with an asset library.

[meshula]

This issue is obsoleted by 501ef6c