Open arbales opened 13 years ago
This would be great but would require a bit of re-architecture on how we store songs and grab them. I'll keep it in mind when I refactor the backend a bit this week.
Was thinking about this — it'd be cool if we could eventually switch to a datastore thats a little more relational. If we wanted to automagically fill the queue from things like a list of users or other factors, we'd want to be able to run joins.
Yes, Redis was unnecessary and overkill. I just wanted to play with it. PostgreSQL or even sqlite are better choices.
The backend is closer to getting this sort of functionality implemented. master now uses PostgreSQL for storing everything. Redis is just used for the playlist, background worker queues, and pub-sub to socket.io.
Currently, you're given a limited view of your Pandora stations. We're showing the current state of the station. Instead, we could simply allow a user to start a station, and periodically pull from the station. To emphasize the ever-building local cache of the station, the sourcelist will show a mixture of recent pulls and previously cached content — this will also reduce our Pandora API load as time progresses.
This way, we can move to a more abstracted music display model, that allows us to view the caches by artist, album, source. Over time, this will help reduce the barrier to adding new services.