Statistics change_second should have an option to function filter out negative change (similar to sum_differences_nonnegative) about core HOT 5 OPEN

Hey there @ThomDietrich, mind taking a look at this issue as it has been labeled with an integration (statistics) you are listed as a code owner for? Thanks!

_{^{(message by CodeOwnersMention)}}

statistics documentation
statistics source
_{^{(message by IssueLinks)}}

from core.

Hello @bakerkj,
thanks for the feedback!
The statistics component calculates change_second like this:

Did I understand you correctly: You would like a new characteristic which take sum_differences_nonnegative and divides it by the seconds between the oldest and newest sample?

from core.

Hi @ThomDietrich,

I am trying to use change_second to compute the rate of increase of a counter that is always increasing, except when the device is reset (which is when the counter starts again from zero).

I think instead of the existing code

    def _stat_change_second(self) -> StateType:
        if len(self.states) > 1:
            age_range_seconds = (self.ages[-1] - self.ages[0]).total_seconds()
            if age_range_seconds > 0:
                return (self.states[-1] - self.states[0]) / age_range_seconds
        return None

I think I want something like this

    def _stat_change_second_something(self) -> StateType:
        if len(self.states) > 1:
            age_range_seconds = (self.ages[-1] - self.ages[0]).total_seconds()
            change = self.states[-1] - self.states[0]
            if age_range_seconds > 0 and change >= 0:
                return change / age_range_seconds
        return None

So for example

time in seconds | counter | statistic
0               | 0       | None
10              | 10      | 1
20              | 30      | 2
30              | 40      | 1
40              | 0       | None <- counter reset
50              | 20      | 2
60              | 30      | 1

I think in this way, the statistic will not be computed in the case where the counter resets (the change is negative)?

Does that make sense?

Thanks,
Ken

from core.

I think it does make sense. Did you check this integration? I wonder if it already does what you want (not that I am saying this is strictly out of scope for the statistics integration)

https://www.home-assistant.io/integrations/derivative/

from core.

I think it does make sense. Did you check this integration? I wonder if it already does what you want (not that I am saying this is strictly out of scope for the statistics integration)

https://www.home-assistant.io/integrations/derivative/

If I understand correctly, the derivative sensor has the same challenge with counters that reset to zero:

For sensors that reset to zero after a power interruption and need a “non-negative derivative”, such as bandwidth counters in routers, or rain gauges, consider using the Utility Meter integration instead. Otherwise, each reset will register a significant change in the derivative sensor.

I don't think the Utility Meter integration does what I am looking to do.

from core.