Skip to content

CHANGELOG.md

Latest commit

 

History

History
73 lines (60 loc) · 3.79 KB

CHANGELOG.md

File metadata and controls

73 lines (60 loc) · 3.79 KB
Raw

Changelog

All notable changes to PlayolaPlayer are documented here. This project follows Semantic Versioning. Versions correspond to git tags, which Swift Package Manager consumers pin to.

0.19.0

Added

  • Bounded retry/backoff on the initial schedule fetch. play(stationId:) now retries a failed initial GET /v1/stations/{id}/schedule up to 3 times with exponential backoff (0.5s / 1s / 2s) before giving up, matching the recovery behavior the player already used for spin loading and the ongoing schedule poll. Only transient failures are retried — server 5xx responses and an explicit allow-list of connectivity URLErrors (timeouts, host/DNS, connection-lost, etc.). Permanent failures (404, decode errors, empty schedules, and non-connectivity URLErrors such as .badURL) still fail fast. This means a transient backend outage no longer instantly fails a station start with no automatic recovery.

Changed

  • PlayolaStationPlayer.State gained a terminal .error(StationPlayerError) case. A failed play(...) now emits .loading(0) when the attempt begins and .error(_) when it fails terminally (instead of throwing without ever changing state, which could leave consumers stuck on a loading spinner with no signal). play(...) still throws as before — the new state is emitted in addition to the thrown error. Cancellation does not produce an .error state.

    Source-breaking for consumers that switch exhaustively over State: add a case .error (or default) arm. A typical handler treats .error as a recoverable, retry-able state (show the message, let the user tap play again). StationPlayerError is now Sendable.

  • State transitions are now supersession-safe (last play() wins). Each play(...)/stop() takes a new internal generation; work from a prior attempt (its scheduling loop, or a spin whose audio starts later) can no longer publish state into a newer attempt. This closes a race where a lingering session could overwrite a freshly-emitted .error (or .idle) with .playing. Behavioral note: starting a new play() supersedes the previous session immediately, so a failed station switch lands on .error rather than rolling back to the previously-playing station.

  • playNow(from:to:) and schedulePlay(at:) are now async. Both public methods changed from synchronous to async so their audio work can run off the main thread. Source-breaking for direct callers: add await at the call site (callers must already be in an async context, e.g. a Task).

0.18.0

Added

  • Network logging seam for observing the library's JSON API traffic (binary audio downloads are intentionally excluded). New public API in PlayolaCore:

    • PlayolaNetworkLogEvent — a Sendable value describing a single request/response (method, url, request headers/body, status code, response body, duration, and any error description). Events are raw and unredacted; the consuming app is responsible for redacting sensitive values such as Authorization headers before storing them.
    • PlayolaNetworkLogger.handler — a thread-safe, Sendable hook the consuming app sets at startup to receive events. nil (the default) disables logging, leaving behavior and performance unchanged.
    • PlayolaNetworkLoggingSession — a transparent URLSessionProtocol wrapper that times each call and emits an event on both success and thrown error without altering return values or rethrown errors.

    The schedule fetch (PlayolaStationPlayer) and the listening-session reporter (ListeningSessionReporter) now route their JSON API calls through this wrapper by default. No breaking API changes: existing initializers and call sites compile unchanged.