Is evaluating an object without knowing the proto definitions conceptually intended?

[google]

Is evaluating an object without knowing the proto definitions conceptually intended?

[TristonianJones]

Hi David,

You are always welcome to reach out to [email protected], especially if there's some of the Go code involved.

The short answer to your question is that evaluation over protos which are not linked into the environment and added as cel.Types (or cel.TypeDescs) will result in an error. Generally the evaluation library is embedded in the binary where the evaluation occurs, but the parse / check can remain separate. For example, you could parse / check via gRPC, but evaluate the result with https://github.com/google/cel-cpp or https://github.com/google/cel-go.

However, you should be able build a JSON object in CEL with a top-level @type set to proper type url and get the protobuf.Any unpack methods to properly deserialize the JSON to a protobuf known to your environment.

Hope that helps,

-Tristan

[drx777]

Thank you for your response and explanation. The problem is I'm in Java/Kotlin code here, so using the go library is not a possibility. I would have to make the protobuf definitions known to the CEL service, which somewhat defies the purpose of a general evaluation service. Still, thank you for the clarification.
Feel free to close the issue.

[TristonianJones]

Here are a few things to consider:

• There is a chance we will open source a Java version of CEL.
• The JSON trick I mentioned will make the object compatible with Any deserialization, and might work for you.
• The conformance service was really only meant for handling a uniform set of test cases, and Go in particular doesn't support dynamic proto construction like some of the other proto implementations do. However, if you know the set of proto message types, you can modify the service and register the types in Go.