replace_pattern() OTTL function should allow referencing capturing groups

[sirianni]

Component(s)

pkg/ottl

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

I'd like to rename metrics scraped from certain destinations to better align to naming conventions (for example, replacing dashes with dots, etc.)

My attempt was to use the transform processor with OTTL for this capability

  transform/kube_state_rename:
    metric_statements:
    - context: metric
      statements:
      - replace_pattern(name, "kube_([0-9A-Za-z]+_)", "k8s.$$1")

However, the replace_pattern() OTTL function does not support expanding references to capturing groups.

Describe the solution you'd like

I would like replace_pattern() OTTL to not support expanding references to capturing groups (using replace_pattern() OTTL function does not support expanding references to capturing groups.).

Describe alternatives you've considered

No response

Additional context

Seems tangentially related to #18084

[github-actions]

Pinging code owners:

• pkg/ottl: @TylerHelmuth @kentquirk @bogdandrutu @evan-bradley

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

[TylerHelmuth]

This is an interesting use case and doesn't add too much complexity. Using ReplaceAllString instead of ReplaceAllLiteralString still preserves all the existing use cases except for literal uses of $, which is probably a low percentage. Although a breaking change, those use cases can use $$ instead.

[sirianni]

Although a breaking change, those use cases can use $$ instead.

Actually it looks like the env var interpolation already eats a single $, so you would need $$1 to refer to a capturing group and $$$ to get a literal $.

[TylerHelmuth]

hmmmm. Thats pretty annoying and confusing. Did you test this out to confirm?

[sirianni]

hmmmm. Thats pretty annoying and confusing. Did you test this out to confirm?

Yes, it took me a while debugging it to realize that the single $ was not even making it through to the OTTL parser.

[TylerHelmuth]

I guess we just have to document that behavior in the function's README entry.