[ELE-1443] Adding a hardcoded lower (or upper) bounds to the anomaly bracket.

[ari-nz]

Hi Team,

Is your feature request related to a problem? Please describe.

I find, sometimes, due to the stochasticity of data, that the anomaly checks sometime provide 'nonsense' results.

For example, here is a volume row count check

You can see that the data would only be anomalous if there were negative rows in the table.

Describe the solution you'd like It would be awesome if it was possible to specify sensible limits (and maybe even have them default?)

Something like

- elementary.volume_anomalies:
    anomaly_sensitivity: 2.5
    lower_bound: 0
    upper_bound: 100000

And the display and/or alerts would take the value that makes sense. In the case of lower bounds either 0 or a positive integer (if that's what is calculated). For upper bounds, you'd display 100,000 or a lower value if that's what was calculated.

Fundamentally there are some values that should auto-default to these sorts of limits:

• Counts >=0
• 0 <= Percentages <= 1
• Variance/StdDev >= 0

Would you be willing to contribute this feature? Happy to help try to implement something if it doesn't go against any key principles you have :)

_{From SyncLinear.com | ELE-1443}

[Maayan-s]

I totally agree with you that we have a problem there @ari-nz, we have this issue that we didn't address yet.

I agree that hardcoded bounds are the simple solution, but I still think that we might be missing a way that fits our approach that these tests are not assertions and shouldn't require you to configure harcoded limits.

[ari-nz]

861 must have slipped through my search net, apologies.

I agree that allowing setting arbitrary limits sort of goes against the philosophy; however for those fundamental values that are checked, would it make sense to instead have some sort of flag that could be invoked as a setting to at least indicate that 'logical settings' should apply?

Counts >=0
0 <= Percentages <= 1
Variance/StdDev >= 0

[Hadarsagiv]

This could also be solved in ways like adding a trend check.

Slack link: https://elementary-community.slack.com/archives/C02CTC89LAX?cid=C02CTC89LAX&thread_ts=1682076795.443979

[elongl]

I agree that allowing setting arbitrary limits sort of goes against the philosophy; however for those fundamental values that are checked, would it make sense to instead have some sort of flag that could be invoked as a setting to at least indicate that 'logical settings' should apply?

Counts >=0
0 <= Percentages <= 1
Variance/StdDev >= 0

Hi @ari-nz , we've applied the changes you suggested in this PR! They'll be released in the upcoming version.