[Question] Guidelines on Using OPA Java WASM Client

[kjhunjhunwala]

Hi,

I have a few questions regarding the use of the OPA Java WASM client:

Can it be used as a replacement for the OPA Go SDK? Other than the limited builtin support, are there any known limitations?
Currently, we are using the OPA Go SDK within a container and evaluating our policies by making network calls from our Java-based application.

Since the OpaPolicy object is not thread-safe, what are the recommended best practices for use in a highly multithreaded environment? Would it be advisable to implement an object pool for OpaPolicy instances and reuse them? Please suggest if there are any other recommended approaches.

Thanks,
Kushal

[andreaTP]

Hi Kushal, great questions and thank you for your interest in this project!

Can it be used as a replacement for the OPA Go SDK?

Yes, it is a full replacement for the evaluation functionality of the OPA SDK.

Other than the limited builtin support, are there any known limitations?

None that I am aware of. I encourage anyone to open Issues or PRs to help improve builtin support.

Currently, we are using the OPA Go SDK within a container and evaluating our policies by making network calls from our Java-based application.

This is exactly the kind of scenario this project is designed for.

Since the OpaPolicy object is not thread-safe, what are the recommended best practices for use in a highly multithreaded environment? Would it be advisable to implement an object pool for OpaPolicy instances and reuse them? Please suggest if there are any other recommended approaches.

To be completely transparent, I did most of the work in this repository, but I am not using this library in production myself.
My focus has been on getting the core building blocks correct, reliable, and easy to use.
I have not personally tested it in a highly multithreaded production environment.

Some users have reported using object pools or similar approaches (for example, see #101). I would be happy to collaborate on building an open-source layer on top of the core library to make it more scalable for real-world scenarios.

To do that, I would need to work with someone who has access to a production-like environment where we can gather real data.
If you are interested in pursuing this, feel free to continue the conversation here publicly or contact me directly by email (you can find my personal address in my GitHub profile).
cc. @jegan-git87 to see if he is interested in collaborating too.