[pkg/ottl] Improve slice and maps dynamic indexing validation

[edmocosta]

Component(s)

pkg/ottl

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

We've recently introduced support for indexing maps and slices with dynamic values (#36644), allowing statements to access elements using Int, String, Converter, and Path, which was a great addition to OTTL.

Given slices and maps are indexed using different data types (slices w/ integers, and maps w/ strings), we should improve the OTTL contexts/parser to validate the path's key's types and avoid raising errors at runtime:

ERROR	failed processing logs	{"error": "failed to execute statement: set(attributes[\"test\"], attributes[cache[\"key\"]]), unable to resolve a string index in map: could not resolve key for map/slice, expecting 'string' but got 'int64'"}

Describe the solution you'd like

When users configure statements using invalid key types, for example, slices with string keys, or maps with integer keys, we should raise an error at the collector startup:

log_statements:
 - context: log
   statements:
    - set(cache["key"], 1)
    - set(attributes["test"], attributes[cache["key"]]) # maps indexes must be a string, not int

Describe alternatives you've considered

No response

Additional context

No response

[github-actions]

Pinging code owners:

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

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

[bacherfl]

I'd like to look into implementing this if this gets accepted

[bacherfl]

@edmocosta can this actually be done during the collector startup? In most cases I think we will not know what will be the type of the values we are going to index during the statement execution. So in the example from the issue description, we will only know the type of attributes["array.attribute"] once we execute the statement for a given log context.

[edmocosta]

I'm sorry, that was a bad example! I'll update it.
You're correct, we don't know the dynamic attributes values at startup, what I meant was validating it for the paths we already know the types, such as attributes, cache, etc.

[bacherfl]

Thank you for the clarification, @edmocosta! I will look into how this could be achieved. Another limitation however may be that we can only do this for statements that consist of known paths, like attributes, and literals, as something like the following might also not be possible until we have a context object:

log_statements:
 - context: log
   statements:
    - set(cache["key"], attributes["foo"]) # type of attributes["foo"] is not known at startup
    - set(attributes["test"], attributes[cache["key"]]) # maps indexes must be a string, not int

[edmocosta]

One possible idea would be delegating the validation to the context's path parsers, as they're already responsible for validating/creating the path's get & setters and know their types beforehand (mostly of the time).

Let's see what other folks says, I think it would be useful to have such validation, but I'm not completely sure if it would worth the extra complexity of adding it (at least for now).

[bacherfl]

After looking further into this, i think we will have to actually execute the statement sequence upon validation, to know the types of the values being used for indexing. Like in the above example there would otherwise no way of knowing what type cache["key"] would be within the second statement, without executing the previous statement first. Then we would still have the limitation of this only working with literals, as we do not know the types of attributes from e.g. a log entry that will be sent to the collector until we receive n actual entry during runtime.

[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 @edmocosta

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