New to KSP here so I got a question. I am building...

# ksp

[Stylianos Gakis]

New to KSP here so I got a question.
I am building something in KSP to generate some code according to some annotations. I got the KSP code in a separate module called :kspProcessors inside the project, and I am importing it into my module (let’s call it :app) where I want the code generated with add("kspJs", project(":kspProcessors")) , add("kspAndroid", project(":kspProcessors")) and add("kspIosX64", project(":kspProcessors")).

Thing is, inside my KSP, code, it’d be very convenient to have access to the classes that exist inside my :app module, to more easily reference them in the code I generate. But would I not then introduce a circular dependency if :kspProcessors was depending on :app with implementation(project(":app")) and had :app depend on the ksp module as I am describing above?

Would I need to keep the KSP code inside the same module? Maybe there’s no circular dep if one depends on the other using implementation and the other using ksp? I do not know if I am wording my question as well as I could’ve since I am totally not familiar with how to work with KSP in general, so please ask clarifying questions if I can make something more understandable.

[wasyl]

I think you’d have the same problem with regular annotation processing, and that the typical solution is to have processor-api project with annotations/types that you’d like to share, and just depend on it from both processor and app code

[Stylianos Gakis]

Ah yeah that makes sense, I guess whenever there’s such problems the solution is always a new module with just the API which can be shared 😄
Currently in my generated code I’m literally manually putting some package and class names, which does work but is so brittle in case of changes.
Thanks a lot for your input, at least now I know what my options are, I really appreciate it 🙌

[wasyl]

Hardcoding class names is that terrible if the processor is in the same build anyway, I think IJ is usually smart enough to recognize the string references when refactoring. But if you want safety then yeah, I’d just have a separate API module

[Stylianos Gakis]

Yeah that is true, it’s a simple processor for our own internal use, not necessarily something that others would use, so in the end I may never introduce the new API module and rely on IntelliJ magic to hold my hand so I don’t break stuff 😅