- Improve responsiveness by downloading the smallest possible chunk
size when seeking or first loading.
- Improve download time and decrease CPU usage by downloading the
largest possible chunk size as throughput allows, still allowing
for reasonable seek responsiveness (~1 second).
- As a result, take refactoring opportunities: simplify prefetching
logic, download threading, command sending, and some ergonomics.
- Fix disappearing controls in the Spotify mobile UI while loading.
- Fix handling of seek, pause, and play commands while loading.
- Fix download rate calculation (don't use the Mercury rate).
- Fix ping time calculation under lock contention.
Special thanks to @eladyn for all of their help and suggestions.
* Add all player events to `player_event_handler.rs`
* Move event handler code to `player_event_handler.rs`
* Add session events
* Clean up and de-noise events and event firing
* Added metadata support via a TrackChanged event
* Add `event_handler_example.py`
* Handle invalid track start positions by just starting the track from the beginning
* Add repeat support to `spirc.rs`
* Add `disconnect`, `set_position_ms` and `set_volume` to `spirc.rs`
* Set `PlayStatus` to the correct value when Player is loading to avoid blanking out the controls when `self.play_status` is `LoadingPlay` or `LoadingPause` in `spirc.rs`
* Handle attempts to play local files better by basically ignoring attempts to load them in `handle_remote_update` in `spirc.rs`
* Add an event worker thread that runs async to the main thread(s) but sync to itself to prevent potential data races for event consumers.
* Get rid of (probably harmless) `.unwrap()` in `main.rs`
* Ensure that events are emited in a logical order and at logical times
* Handle invalid and disappearing devices better
* Ignore SpircCommands unless we're active with the exception of ShutDown
- Keep using the same hyper client instead of building a new one for
each request
- This allows the client to reuse connections and improves the
performance of multiple requests by almost 2x.
- The playlist_tracks example takes 38 secs before and 20 secs after the
change to enumerate a 180 track playlist
- To avoid carrying the hyper Client generics through the whole project,
`ProxyConnector` is always used as the Connector, but disabled when
not using a proxy.
- The client creation is done lazily to keep the `HttpClient::new`
without a `Result` return type
- Fixed resolve function panicking when resolving endpoint type with no
AP in the list
- Fixed fallback APs not being applied when only some of the AP types
were missing
- Switch container type from `Vec` to `VecDeque` for the `AccessPoints`
- Remove the note about fallback AP being used even if the port is not
matching the configured `ap_port`
- Also fix an overflow panic when a token cannot be parsed.
- Getting tokens always requires the keymaster client ID;
passing the actual client ID yields a HashCash challenge.
Some fields were wrongly parsed as `SpotifyId`s, although they do not
always encode exactly 16 bytes in practice. Also, some optional fields
caused `[]` to be parsed as `SpotifyId`, which obviously failed as well.
