The icon for Withings sleep sensor was using `mdi:bed` (which is correct for MDI v5.0.45 and later). However Home Assistant still uses an older version of Material Design Icons (4.9.95), so this `mdi:bed` icon was not displaying at all. It should be`mdi:hotel` instead, which you can see here: https://cdn.materialdesignicons.com/4.9.95/
- Bumps bravia-tv lib to 1.0.6 which fixes is_connected() to actually
return True only when API is connected, instead of just returning whether
or not cookies are cached (regardless if they actually worked).
- Wrap is_connected() because it now performs io.
- Remove unnecessary logic to refresh cookies. Now that
is_connected() works, the bravia instance only needs to be
reconnected when is_connected is False and TV is not off.
* Move logbook continuous domain filtering to sql
sensors tend to generate a significant amount of states
that are filtered out by logbook. In testing 75% of
states can be filtered away in sql to avoid the
sqlalchemy ORM overhead of creating objects that will
be discarded.
* remove un-needed nesting
* Switch homekit to use async_track_state_change_event
Calling async_track_state_change_event directly
is faster than async_track_state_change and has
slightly lower latency triggering state updates
in homekit
* check for deleted entities
* Update additional tests for linked sensor removals
* Ensure removing entities does not result in an exception
* Cleanup async_track_state_change and augment docstrings.
Skip from_state and to_state matching in
async_track_state_change when they are None
Optimize the state change listener for the most
common use case: no to_state and from_state
matching.
* Update benchmark to be more realistic (previously we assumed only one entity was present in the whole instance)
* Add more tests to ensure behavior is preserved
* Ensure new behavior matches test
* remove MATCH_ALL from zone automation since its the default anyways
* Might as well use async_track_state_change_event instead since MATCH_ALL is removed
* Switch to dispatcher
* Adjust tests for dispatcher
* Store device in sensor
* Move state application into binary sensor entity class
* Move more specifics into per platform classes
* Should not apply event in init of sensor
* Switch to call_later
* Make apply_event public
* No point in slugifying debug logs
* Adjust error in off delay and event
* Make sure we match with masked id
* Migrate Guardian to use the DataUpdateCoordinator
* Finish work
* Cleanup
* Don't use UpdateFailed error
* Code cleanup
* Code cleanup
* Remove unnecessary change
* Code review
* Code review
* Use a subclass of DataUpdateCoordinator
* Make sure to pop client upon unload
* Adjust coverage
* Switch tests to use hass objects instead of direct
* Make sure sensor update state
* Add some initial binary sensor tests
* Add initial binary sensor tests
* Add tests for pt2262
* Add test for off delay
Calling async_track_state_change_event directly
is faster than async_track_state_change (see #37251) and has
slightly lower latency triggering state updates