* Add counter entities to the ZHA coordinator device
* rework to prepare for non coordinator device counters
* Initial scaffolding to support quirks v2 entities
* update for zigpy changes
* add assertion error message
* clean up test
* update group entity discovery kwargs
* constants and clearer names
* apply custom device configuration
* quirks switches
* quirks select entities
* quirks sensor entities
* update discovery
* move call to super
* add complex quirks v2 discovery test
* remove duplicate replaces
* add quirks v2 button entity support
* add quirks v2 binary sensor entity support
* fix exception in counter entitiy discovery
* oops
* update formatting
* support custom on and off values
* logging
* don't filter out entities quirks says should be created
* fix type alias warnings
* sync up with zigpy changes and additions
* add a binary sensor test
* button coverage
* switch coverage
* initial select coverage
* number coverage
* sensor coverage
* update discovery after rebase
* coverage
* single line
* line lengths
* fix double underscore
* review comments
* set category from quirks in base entity
* line lengths
* move comment
* imports
* simplify
* simplify
* Cache device triggers on startup
* reorg zha init
* don't reuse gateway
* don't nuke yaml configuration
* review comments
* Add unit tests
* Do not cache device and entity registries
* [WIP] Wrap ZHA data in a dataclass
* [WIP] Get unit tests passing
* Use a helper function for getting the gateway object to fix annotations
* Remove `bridge_id`
* Fix typing issues with entity references in group websocket info
* Use `Platform` instead of `str` for entity platform matching
* Use `get_zha_gateway` in a few more places
* Fix flaky unit test
* Use `slots` for ZHA data
Co-authored-by: J. Nick Koston <nick@koston.org>
---------
Co-authored-by: David F. Mulcahey <david.mulcahey@icloud.com>
Co-authored-by: J. Nick Koston <nick@koston.org>
* Always use `Light` for lights, including subclasses
* Clean up other platforms
* Add a unit test to ensure all future entity classes have names
* Remove stale `_name`
* Address review feedback and rename `Open` to `Opening`
* rename channel -> cluster handler
* remove refs to channels and create endpoint class
* remove remaining references to channels
* fix filter
* take in latest changes from #91403
* missed one
* missed a reference
* Initial "group members assume state" implementation for ZHA
* Remove left-over debug flag (where polling was disabled)
* Implement _send_member_assume_state_event() method and also use after turn_off
* Only assume updated arguments from service call to group
* Make code more readable and change checks slightly
* Move "send member assume state" events to LightGroup on/off calls
* Include new config option in tests
* Check that member is available before updating to assumed state
* Lower "update group from child delay" for debouncer to basically 0 when using assumed member state
* Allow "child to group" updates regardless of config option
This is not needed, as group members will not update their state, as long as they're transitioning. (If a group transitions, it also sets its members to transitioning mode)
This fixes multiple issues. Previously, the state of a group was completely wrong when:
- turn on group with 10 second transition
- turn on members individually
- turn off members individually
- group state would not update correctly
* Move "default update group from child delay" constant
* Change to new constant name in test
* Also update fan test to new constant name
* Decrease "update group from child delay" to 10ms
In my testing, 0.0 also works without any issues and correctly de-bounces child updates when using the "assume state option".
This is just for avoiding multiple state changes when changing the group -> children issue individual updates.
With 2 children in a group and delay 0, both child updates only cause one group re-calculation and state change.
0.01 (10ms) should be plenty for very slow systems to de-bounce the update (and in the worst case, it'll cause just another state change but nothing breaks)
* Also implement "assuming state" for effect
Not sure if anybody even uses this, but this one is a bit special because the effect is always deactivated if it's not provided in the light.turn_on call.
* Move shortened delay for "assuming members" to a constant
* Add basic test to verify that group members assume on/off state
* Move _assume_group_state function declaration out of async_added_to_hass
* Fix rare edge-case when rapidly toggling lights and light groups at the same time
This prevents an issue where either the group transition would unset the transition flag or the single light would unset the group transition status midst-transition.
Note: When a new individual transition is started, we want to unset the group flag, as we actually cancel that transition.
* Check that effect list exists, add return type
* Re-trigger CI due to timeout
* Increase ASSUME_UPDATE_GROUP_FROM_CHILD_DELAY slightly
The debouncer is used when updating group member states either by assuming them (in which case we want to barely have any delay), or between the time we get the results back from polling (where we want a slightly longer time).
As it's not easily possible to distinguish if a group member was updated via assuming the state of the group or by the polling that follows, 50 ms seems to be a good middle point.
* Add debug print for when updating group state
* Fix issues with "off brightness" when switching between group/members
This fixes a bunch of issues with "off brightness" and passes it down to the members correctly.
For example, if a light group is turned off with a transition (so bulbs get their level set to 1), this will also set the "off brightness" of all individual bulbs to the last level that they were at.
(It really fixes a lot of issues when using the "member assume group state" option. It's not really possible to fix them without that.)
Furthermore, issues where polling was previously needed to get the correct state after "playing with transitions", should now get be resolved and get correct state when using the "members assume group state" option.
Note: The only case which still can't be fixed is the following:
If individual lights have off_with_transition set, but not the group, and the group is then turned on without a level, individual lights might fall back to brightness level 1 (<- at least now shows correctly in UI even before polling).
Since all lights might need different brightness levels to be turned on, we can't use one group call. But making individual calls when turning on a ZHA group would cause a lot of traffic and thereby be counter-productive.
In this case, light.turn_on should just be called with a level (or individual calls to the lights should be made).
Another thing that was changed is to reset off_with_transition/off_brightness for a LightGroup when a member is turned on (even if the LightGroup wasn't turned on using its turn_on method).
off_with_transition/off_brightness for individual bulbs is now also turned off when a light is detected to be on during polling.
Lastly, the waiting for polled attributes could previously cause "invalid state" to be set (so mid-transition levels).
This could happen when group and members are repeatedly toggled at similar times. These "invalid states" could cause wrong "off brightness" levels if transitions are also used.
To fix this, we check after waiting for the polled attributes in async_get_state to see if a transition has started in the meanwhile. If so, the values can be discarded. A new poll will happen later and if using the "members assume group state" config option, the values should already be correct before the polling.
* Enable "group members assume state" config option by default
The config tests are also updated to expect the config option be enabled by default.
For all tests, the config option is generally disabled though:
There are only two group related tests. The one that tests this new feature overrides the config option to be enabled anyway.
The other tests works in a similar way but also "sends" attribute reports, so we want to disable the feature for that test.
(It would also run with it enabled (if the correct CHILD_UPDATE value is patched), but then it would test the same stuff as the other test, hence we're disabling the config option for that test.)
* Drop CALLABLE_T from zha
* Adjust .coveragerc
* Apply suggestions from code review
Co-authored-by: Marc Mueller <30130371+cdce8p@users.noreply.github.com>
* Add TypeVar
* Apply suggestions from code review
Co-authored-by: Marc Mueller <30130371+cdce8p@users.noreply.github.com>
* One more
Co-authored-by: Marc Mueller <30130371+cdce8p@users.noreply.github.com>
* Flake8
Co-authored-by: Marc Mueller <30130371+cdce8p@users.noreply.github.com>
* Make unit tests pass
* Flip response type check to not rely on it being a list
https://github.com/zigpy/zigpy/pull/716#issuecomment-1025236190
* Bump zigpy and quirks versions to ZCLR8 releases
* Fix renamed zigpy cluster attributes
* Handle the default response for ZLL `get_group_identifiers`
* Add more error context to `stage failed` errors
* Fix unit test returning lists as ZCL request responses
* Always load quirks when testing ZHA
* Bump zha-quirks to 0.0.69
* Delay ZHA group updates to ensure all members are updated first
After turning off a group, when the first device reports "off", the
other devices may still be "on". If HA processes the group state update
quickly enough, the group will see that some devices are on, so the
state of the group will revert back to "on", and then "off" when the
remaining devices all report "off". That would cause the UI toggle to go
back and forward quickly, and automations that trigger with "state: on"
to fire when the user turns the group off.
This PR fixes that by delaying the group state update, giving time for
all the devices to report their states first.
* Fix zha group tests
* Reorder sleeping.
* Update tests/components/zha/common.py
Co-authored-by: Alexei Chetroi <lexoid@gmail.com>
