Borrows the same design from homeassistant.helpers.storage to avoid
rescheduling the timer every time async_schedule is called if a timer
is already running.
Instead of the timer fires too early it gets rescheduled for the time
we wanted it. This avoids 1000s of timer add/cancel during startup
* Increase MQTT broker socket buffer size
* Revert unrelated change
* Try to increase buffer size
* Set INITIAL_SUBSCRIBE_COOLDOWN back to 0.5 sec
* Sinplify and add test
* comments
* comments
---------
Co-authored-by: J. Nick Koston <nick@koston.org>
* Fix MQTT discovery cooldown too short with large setup
* Set to 5 sec
* Only change the discovery cooldown
* Fire immediatly when teh debouncing period is over
* Improve lists for MQTT integration
* Extra diagnostics tests
* Revert changes where the original version was probably faster
* Revert change to gather and await in series
* Debounce MQTT unsubscribes and merge to one call
* Make _async_unsubscribe a callback
* Make sure unsubscribes are processed
* Move debug log out of lock
* Reduce calls and raise outside lock
* Cancel any unsubscribe when queing
* Copy pending unsubscribes
* Only convert topics to list once
* No copy needed
* Typo in comment
* Cleanup config merging and adding defaults
* Optimize and update tests
* Do not mix entry and yaml config
* Make sure hass.data is initilized
* remove check on get_mqtt_data
* Tweaks to MQTT client
* Remove None assigment mqtt client and fix mock
* Debounce and group mqtt subscriptions
* Cleanup
* Do not cooldown on resubscribe
* Remove lock from task
Co-authored-by: Erik Montnemery <erik@montnemery.com>
* ruff
* Longer initial cool down. Manages unsubscribes
* Own lock for access to self._pending_subscriptions
* adjust
* Subscribe to highest QoS when sharing subscription
* do not block _pending_subscriptions_lock with io
* Test the highest qos is subscribed at
* Cleanup max qos
* Follow up comments part 1
* Make docstr more generic
* Make max qos update thread safe
* Add lock on clearing _max_qos when resubscribing
* Wait for linger task
* User copy
* Check for key before cleaning up
* Fix lingering task
* Do not use a lock
* do not await _async_queue_subscriptions
* Replace copy with assignment
* Update max qos before returning
* Do not iterate if max_qos == 0
* Do not ieterate subs if max qos == 0
* Set initial cooldown correctly
* Ensure discovery cooldown ends after subscribing
* plan last subscribe with debouncer timeout
* cooldown if self._pending_subscriptions is set
* Revert format changes
* Remove stale assingnment self._last_subscribe
* Remove not used property
* Also check while for pending subscriptions
* revert first added sleep()
* Optimize
---------
Co-authored-by: Erik Montnemery <erik@montnemery.com>
Co-authored-by: J. Nick Koston <nick@koston.org>
