I'm assuming based on the "Future improvements" section on this KEEP <https://github.com/Kotlin/KEEP...

# announcements

[Chris Miller]

I'm assuming based on the "Future improvements" section on this KEEP https://github.com/Kotlin/KEEP/blob/master/proposals/data-class-inheritance.md that the following isn't possible? Any suggestions for the most elegant alternative?

sealed class TimedEvent(val eventTime: Long)
data class FxRateEvent(val currency: String, val rate: Double, eventTime: Long) : TimedEvent(eventTime)

[adimit]

Try

sealed class A(open val f: String)
data class B(val foo: String, override val f: String): A(f)

[Chris Miller]

Ah nice, thanks!

[Chris Miller]

Hmm, doesn't work for my use case unfortunately:

org.apache.avro.AvroRuntimeException: avro.shaded.com.google.common.util.concurrent.UncheckedExecutionException: org.apache.avro.AvroTypeException:
class pojo.TimedEvent contains two fields named: private long pojo.TimedEvent.eventTime

[adimit]

That's weird, we've been using this in production a lot. Maybe some problem with reflection in Javaland? I don't know what these value types compile to.

[Chris Miller]

Well looking at the bytecode, looks like the "override val" creates a new field, getter and setter, so the field is duplicated rather than reused. I've changed to using normal (non-data) classes instead which behaves as expected

[adimit]

You'll have to implement .equals and .hashCode though 😕 I'm surprised the byte code is this simple.

[Chris Miller]

I often end up having to implement .equals and .hashCode on data classes anyway (or just not use them) because I tend to only want to test for equality based on a single ID field rather than all of them (which in some cases can end up quite expensive)