v0.1.1

Changed

• Renamed to contenir/contenir-queue, to match its repository. It replaces contenir/queue, so a project that still
  requires the old name gets this package, and never both. The namespaces are unchanged.

v0.1.0

Added

• Durable jobs in a queue_job table on MySQL or SQLite, created by a peptolab/phpdb-migration migration in
  schema/: job class, payload, queue, state (waiting, running, done, failed), attempts, next attempt, lock and last
  error.
• QueueInterface::push() to store a job, now or for later, on a named queue.
• JobHandlerInterface for handlers, resolved from a PSR-11 container by the job's class.
• JobRunner, which runs a claimed job and records the result. Failures retry with Backoff (1 minute doubling to
  24 hours, then failed). RetryLaterException sets the next attempt, including from an HTTP Retry-After, and
  UnrecoverableJobException fails a job at once, as does a handler missing from the container.
• PhpDbJobStore, which claims jobs atomically so several workers and web requests can share the table, and only
  finishes a job while it still holds the lock it was claimed with.
• bin/contenir-queue, the worker cron runs every minute. It takes a lock in a queue_worker table so only one runs,
  renews it as a heartbeat, and runs due jobs until nothing is due or 50 seconds (configurable) have passed.
  --daemon runs it with no time limit, waiting for work, until SIGTERM, its memory limit or a job limit, for a
  process supervisor.
• Worker, which the command runs. Before taking jobs it retries jobs whose worker died mid-run (locked for over an
  hour) and deletes done jobs older than seven days.
• Drainer::drain(), for a host to call after the response has been sent: it runs up to three due jobs within two
  seconds, unless the worker's heartbeat is under two minutes old.
• Schedules: Schedule with a Frequency (everyMinutes(), or cron() in a time zone, read by the package's own
  standard five-field cron parser), added to the Schedules service or configured under contenir_queue.schedules.
  Scheduler::tick(), run by the worker on every turn and by the drain, pushes the job of each schedule that has
  fallen due. A run whose last job is still waiting or running is skipped. Each run is recorded in
  queue_schedule_run, and the schedule's state is kept in queue_schedule.
• QueueMonitor, the data for a host's screens and checks: job counts, jobs by state, retry and delete, schedules
  with their last and next runs, a schedule's recent runs, and QueueHealth with the heartbeat's age, what is
  waiting and what has failed.
• ConfigProvider and PSR-11 factories, configured under contenir_queue.
• Mago, PHPUnit unit and integration suites (SQLite, and MySQL in CI), Infection (MSI 100%) and Codecov in CI.