[connector/signaltometrics] Implement optional attribute filtering

[lahsivjar]

Component(s)

connector/signaltometrics

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

Current attribute filtering performed by signal to metrics connector is:

1. If default_value is specified then ALWAYS produce the metric for the defined configuration. If the attribute is not present in the incoming data then the default value is used else the attribute value from the incoming data is used.
2. If default_value is NOT specified then only produce the metric for the defined configuration if the incoming data as all the attributes, without default_value, available in the incoming data.

There are cases when we want a third option i.e. an optional attribute. If the attribute is present in the incoming data then include it, else, don't include it. This basically means that there would be another possible classification bucket with that attribute missing.

Describe the solution you'd like

Introduce a boolean configuration option to the attributes called passthrough (better names welcomed), the attribute will be false by default and in this case, the behavior of the component will be identical to the current behavior. With the value set to true the attribute will be kept as read from the input signal i.e.:

1. If the input signal has the attribute defined then that value will be used.
2. If the input signal has the attribute missing then a new datapoint will be created with missing attribute.

The passthrough and default_value will be mutually exclusive i.e. only one of the value must be defined - we will validate this as part of config validation.

Describe alternatives you've considered

No response

Additional context

#38494

[github-actions]

Pinging code owners:

• connector/signaltometrics: @ChrsMark @lahsivjar

See Adding Labels via Comments if you do not have permissions to add labels yourself.

[lahsivjar]

/label -needs-triage

[capiisco]

Hi @lahsivjar perfectly meets my need. Thx
Do you have any idea on which release this feature will be embedded? :)

[lahsivjar]

@capiisco Too early to say. I will try to prepare a PR early next week and then we can take it from there.

[axw]

There are cases when we want a third option i.e. an optional attribute. If the attribute is present in the incoming data then include it, else, don't include it. This basically means that there would be another possible classification bucket with that attribute missing.

(emphasis added)

Introduce a boolean configuration option to the attributes called passthrough (better names welcomed), the attribute will be false by default and in this case, the behavior of the component will be identical to the current behavior. With the value set to true the attribute will be kept as read from the input signal i.e.:

1. If the input signal has the attribute defined then that value will be used.
2. If the input signal has the attribute missing then a new datapoint will be created with missing attribute.

The passthrough and default_value will be mutually exclusive i.e. only one of the value must be defined - we will validate this as part of config validation.

I don't find the idea of "passthrough" to be very intuitive. Perhaps call it "optional" (defaults to false)?

[lahsivjar]

I don't find the idea of "passthrough" to be very intuitive. Perhaps call it "optional" (defaults to false)?

Thanks for the review, it makes sense to me now that I think about it - I was on the edge with both the names. I have updated the PR to use optional now.