[pkg/ottl] Support dynamic indexing

[odubajDT]

Description

Link to tracking issue

Fixes #36644

[evan-bradley]

Thanks for taking this one on @odubajDT. I have a few questions on the implementation, but based on the tests passing, I think we're definitely on the right track.

[TylerHelmuth]

This is so exciting! We've wanted this feature for a long time and I'm glad to see that it fit so nicely into our Path/Key pattern.

Can you add e2e tests that use the new ottl-context scoped syntax? I want to ensure statements like

set(log.attributes[resource.attributes["flags"]], "something33")

work.

[odubajDT]

This is so exciting! We've wanted this feature for a long time and I'm glad to see that it fit so nicely into our Path/Key pattern.

Can you add e2e tests that use the new ottl-context scoped syntax? I want to ensure statements like

set(log.attributes[resource.attributes["flags"]], "something33")

work.

Should be done, please have a look

[edmocosta]

I think we should add some extra type validations. Considering the ottl.Key index can now be an expression (mathExprLiteral), statements like set(severity_text, attributes[1]) and set(attributes["slice"][0.0], "bar") would be considered valid and they would panic on the FetchValueFromExpression function. I know they're invalid usages of the OTTL syntax, but I'd prefer to not panic if that happens. WDYT?

[odubajDT]

I think we should add some extra type validations. Considering the ottl.Key index can now be an expression (mathExprLiteral), statements like set(severity_text, attributes[1]) and set(attributes["slice"][0.0], "bar") would be considered valid and they would panic on the FetchValueFromExpression function. I know they're invalid usages of the OTTL syntax, but I'd prefer to not panic if that happens. WDYT?

You got a point here. Added some additional checks and polished the error messages to be clearer where the problem lays. Also added some e2e tests.