My use case: 1. `ProcessorA` generates class `a.b.ClassA` 1. `ProcessorB` enumerates declarations ...

# ksp

[Norbi]

My use case:
1. ProcessorA generates class a.b.ClassA
2. ProcessorB enumerates declarations in package a.b using com.google.devtools.ksp.processing.Resolver#getDeclarationsFromPackage and generates one aggregated class a.b.Aggregated from them.
If ProcessorB is executed before ProcessorA then ProcessorB is called one more time for a round#2.
I tried to cache data about the classes processed in round#1 in the ProcessorB instance but when I try to regenerate a.b.Aggregated in round#2 then a kotlin.io.FileAlreadyExistsException: .../Aggregated.kt exception is thrown, despite using aggregating = true.

Maybe do you have an idea what should I try? Thanks.

[Norbi]

I use ksp 1.9.10-1.0.13 - maybe I should try to upgrade?
EDIT: I upgraded to 1.9.21-1.0.15 but it still fails with FileAlreadyExistsException.

[ephemient]

aggregating doesn't mean you can overwrite previous outputs

[Norbi]

As I understand, my use-case cannot be implemented with KSP currently...

[ephemient]

you should be able to gather all the names as process is called through multiple rounds (make sure not to cache the symbols themselves, https://kotlinlang.org/docs/ksp-multi-round.html#information-consistent-cross-rounds) and then write out the aggregation in finish

[Norbi]

I haven't noticed finish() until now (probably simply because there was no need for it 🙂 ).
Thanks for the very useful idea.