Some questions regarding incremental compilation. ...

# ksp

[Oliver.O]

Some questions regarding incremental compilation. Details in thread.

[Oliver.O]

1. If the list of a generated file's dependencies is only known after creating the file, is using

codeGenerator.createNewFile(
        dependencies = Dependencies(aggregating = true),
        // ...
    ).use { output ->
        // ...
    }

followed by

codeGenerator.associateWithClasses(
        classes = sourceClassDeclarations,
        // ...
    )

the correct way to do it?

2. For aggregating dependencies, is it correct to assume that the deletion of a source file will only trigger KSP processing if that source file is from the list of dependencies?

3. Is preventing KSP re-processing when deleting a source non-dependency the only optimization available for aggregating dependencies (as any new file or changing any existing file – dependency or not – must always trigger KSP)?

[Ting-Yuan Huang]

1. Yes
2. Yes
3. Kind of. You are right that the only situation that an aggregating output is up-to-date is the deletion of unrelated sources (besides nothing touched at all). On the other hand, when triggering re-processing, sources that are known unrelated won't be passed to processors and reprocessing of those unrelated sources are avoided.

[Oliver.O]

Thanks for the insights! Your addition regarding point 3 makes dependency registration seem valuable even in the aggregation case where file deletions are a rare event.