Experimental Sounds Architecture

[CaptainLexington]

This PR is definitely just a suggestion/starting point. It changes the structure and semantics of the audio-state atom and the to-play function, in a way that I hope makes it more expressive but no less simple to use. :music and :effects are now sets instead of vecs, and to stop music or a sound effect, you just remove it from the set. When a sound effect finishes, it will automatically be removed from the set. Instead of an is-playing boolean that describes one piece of music, to-play receives the entire audio-state for the previous frame. The only difference between :music and :effects is that things in :music loop by default, using the .-loop property, but it is now easy to "loop" a sound effect too, by just keeping it in :effects as long as some condition holds.

Benefits

• Multiple pieces of music can play concurrently. Good if, e.g., you want background music and ambient sound, or to crossfade between two pieces of music.
• You can check on the status of, and interrupt, sound effects as well as music.
• Stopping/starting sounds feels more declarative and less imperative.

Drawbacks

• The set operations that run every frame may be expensive. Maybe there is a more sensible way to do this?
• There is no "API" way to pause a sound, rather than stop and restart it. The user would have to use the .pause function on the sound object to do that.
• Likewise, there is no way to store any "metadata" about a given sound, like its volume or duration. These properties are gettable and settable on the JS object, but that's not ideal and the user may want others. I am no sound expert.

While writing this it occurred to me that putting sounds in a set made it impossible to play a sound concurrently with itself, if e.g. two of the same laser guns go off at the same time. But then I realized that that's currently how it works. If you add the same sound effect object to the queue while it is already playing, nothing will happen. Either way, the user would need to create a copy of the sound to play multiple instances at once.

Please give me your thoughts!

[jarcane]

I really like this!

I think sets are probably fine, performance wise. The operation is actually faster than it would be on a vector, because sets are based on hash trees IIRC (they're just a specialization of maps).

The pause thing is probably still something that could probably be implemented on top of this as well, because sets also make object lookup easy, since sound objects are you unique. You're right it'd be less declarative to have to have a function for this but I think there's possibly some solution we just haven't thought of, and this is still more powerful than what we have. Same with the metadata.

I think as an improvement on something that was already kind of a crude hack that didn't get nearly as much time in the oven, this passes my approval at least. 👍 I'm happy to approve as is, unless you've got some more ideas you wanna add to this before it's complete. 😄

[CaptainLexington]

I updated the demo to introduce a hovering-type sound effect for when the ship is moving. There is, unfortunately, a slight gap when the effect loops, but some digging indicates this is a weakness of the HTMLMediaElement API (or at least implementation on Chrome). It seems like the best bet is to upgrade to using the newer but more complicated Web Audio API, but that sounds like a task for another PR.