[Reactor] Bug when sending MQTT boolean payloads

[tunnus]

Is the current manual (incl. examples) up to date with how retention value is handled in time series configuration? Referring to this post

[toggledbits]

@tunnus Not entirely. I need to spend some time on that. As the post suggested, comment out or remove your retention configuration value. It will be computed from interval and depth.

Simple moving average:

        -
          id: temp_sma
          name: "Outdoor Temperature Average"
          capabilities:
            value_sensor:
              attributes:
                value:
                  model: time series
                  entity: "weather>home"
                  attribute: "wx.temperature"
                  aggregate: sma
                  interval: 60  # minutes
                  depth: 5
                  precision: 2
          primary_attribute: value_sensor.value
          type: ValueSensor

Rate:

        -
          id: garage_humidity_rate
          name: "Garage Humidity Change Rate"
          capabilities:
            value_sensor:
              attributes:
                value:
                  model: time series
                  entity: "mqtt>garage_handt"
                  attribute: "humidity_sensor.value"
                  aggregate: rate
                  interval: 30  # minutes
                  depth: 3  # last three samples, t-0, t-30, t-60
                  precision: 2
                units: "%/min"
          primary_attribute: value_sensor.value
          type: ValueSensor

[tunnus]

@toggledbits I cannot get "accel" (acceleration) method to work. It used to work at some point, but at least with the latest builds (now using 25328) I'm getting "null" as a value. Entity values below:

value_sensor.units=null
value_sensor.value=null
x_virtualentity.last_request_time=null
x_virtualentity.request_failures=null
x_virtualentity.template=null
x_virtualentity.timeseries_aggregate="accel"
x_virtualentity.timeseries_debug=[{"time":1766248080000,"value":694},{"time":1766248380000,"value":694},{"time":1766248680000,"value":694},{"time":1766248980000,"value":701},{"time":1766249280000,"value":701},{"time":1766249580000,"value":691},{"time":1766249880000,"value":691},{"time":1766250180000,"value":691},{"time":1766250480000,"value":696},{"time":1766250780000,"value":696},{"time":1766251080000,"value":696},{"time":1766251380000,"value":707},{"time":1766251680000,"value":707}]
x_virtualentity.timeseries_depth=null
x_virtualentity.timeseries_interval=5
x_virtualentity.timeseries_length=13
x_virtualentity.timeseries_nextsample=1766251980000
x_virtualentity.timeseries_retention=60

And configuration from reactor.yaml:

 id: virtual9
          name: "Netatmo CO2 change rate"
          capabilities:
            value_sensor:
              attributes:
                value:
                  model: time series
                  entity: "hass>sensor_carbon_dioxide"
                  attribute: "value_sensor.value"
                  interval: 5
                  aggregate: accel
                  retention: 60 
                  precision: 2
          primary_attribute: value_sensor.value
          type: ValueSensor

When I tried to drop "retention" and used "depth" instead, got the following error:

Unable to configure time series "value_sensor.value" on "virtual9": missing required configuration 'retention'
Controller: VirtualEntityController#virtual
Last 15:56:17

Looking at the entity values, there's "x_virtualentity.timeseries_depth=null", and if depth is needed in the calculation, that could be the culprit?

[toggledbits]

You have to wait for the time series to fill before acceleration can be computed. depth is used in the calculation, as well as retention... that's a peculiarity of this aggregate (accel). There need to be at least 2*depth-1 samples collected before the calculation can occur.

I have updated the documentation in the online version just yesterday, so check here.

[tunnus]

@toggledbits this configuration has been running for hours (retention being 60 min), and still "null"

[toggledbits]

@tunnus OK. Let's start here: copy-paste the device attributes for that VEC entity (remember to use fenced code block formatting when pasting in reply).

[tunnus]

@toggledbits those were already in my previous post, or was something missing?

[toggledbits]

The previous post shows depth as null or not provided. Please specify both depth and retention, restart and wait at least one full interval, then copy-paste the attributes if a value hasn't been calculated.

[tunnus]

@toggledbits ok, when I get back home I’ll try that. But wasn’t depth optional, i.e. defaults to 2 if not specified?

[toggledbits]

In the version you are using, it is still required; that's the docs leading the released code by a step (sorry). The version you are using also doesn't enforce limits well, so it allows values that could produce no result, or store more data than needed.

For example, given your retention of 60 and interval of 5, there would be r/i+1 = 13 samples in the series. But if you specify depth of 2, then only the most recent 2d-1 = 3 samples are considered... the other 10 samples don't contribute to the result value. A depth of 7 would be the best match for the given interval and retention (just mathematically, ignoring your semantic goals).

[tunnus]

@toggledbits adding depth did the trick, thanks again!