[pkg/ottl] Update statements to reflect their context

[TylerHelmuth]

Component(s)

pkg/ottl

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

Currently OTTL statements are independent of any context. Users must tell OTTL the context in which the statement is should be run.

This results in OTTL statements not being directly portable between components. Unless the config implementing OTTL is the same between components, the OTTL statement may not be able to be copied from one component to another.

If the OTTL statement expresses the context itself then there is no ambiguity and the statement would mean the same thing in any component.

Describe the solution you'd like

Update OTTL statements to express the context, probably via the path names. When parsing the statement, infer the context to be the lowest context defined in the statement.

The Parser needs updated to be told which contexts to consider. To start, one implementation would be to try parsing the statement in each context, starting with the "highest" context. The first context to pass parsing is used.

Describe alternatives you've considered

No response

Additional context

No response

[github-actions]

This issue has been inactive for 60 days. It will be closed in 60 days if there is no activity. To ping code owners by adding a component label, see Adding Labels via Comments, or if you are unsure of which component this issue relates to, please ping @open-telemetry/collector-contrib-triagers. If this issue is still relevant, please ping the code owners or leave a comment explaining why it is still relevant. Otherwise, please close it.

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.

[jsuereth]

For this issue, I was thinking of proposing the following syntax:

on <context>
(where <expr>) ?
(yield <expr> | drop)

You could then allow statements to either be what they are today, or this new syntax which carries the context.

Example:

on metric
where metric.attributes["environment"] == "qa"
drop

Note: the yield syntax is part of a larger proposal. For now you could ignore that instead just have the last part of the grammar be existing statements like Route(), Drop(), set(...).

[TylerHelmuth]

@jsuereth for context, during a working session at Kubecon we imagine the solution would be to include the context in all paths. So something like:

transform:
  error_mode: ignore
  trace_statements:
    - context: resource
      statements:
        - keep_keys(attributes, ["service.name", "service.namespace", "cloud.region", "process.command_line"])
    - context: span
      statements:
        - set(status.code, 1) where attributes["http.path"] == "/health"

becomes

transform:
  error_mode: ignore
  traces_statements:
      - keep_keys(resource.attributes, ["service.name", "service.namespace", "cloud.region", "process.command_line"])
      - set(span.status.code, 1) where span.attributes["http.path"] == "/health"

And the parser could infer the context based on the "lowest" specified context in the statement. A big benefit to this solution is the ability to build the new statements from existing config. I believe there is a way to implement this solution without incurring a breaking change, leaving both configurations available until OTTL goes stable.