Hi there, can you help me with the following examp...

# datascience

[holgerbrandl]

Hi there, can you help me with the following example:

data class Car(val type: String, val model: String)

val cars: DataFrame<*> = dataFrameOf("owner", "car")(
    "Max", Car("audi", "a8"),
    "Tom", Car("toyota", "corolla")
)

// Unfold the 'car' column into separate columns for each property of the Car class
val unfolded = cars.unfold("car")

I believe the intent is clear from the code, however this fails with (kotilin-datframe v0.14.1) with

> Exception in thread "main" org.jetbrains.kotlinx.dataframe.exceptions.UnequalColumnSizesException: Unequal column sizes. Expected rows count: 2. Actual column sizes:
> owner: 2
> car: 0
> at org.jetbrains.kotlinx.dataframe.impl.DataFrameImpl.init(DataFrameImpl.kt:45)
> at org.jetbrains.kotlinx.dataframe.api.ConstructorsKt.dataFrameOf(constructors.kt:267)
> at org.jetbrains.kotlinx.dataframe.api.ToDataFrameKt.toDataFrameAnyColumn(toDataFrame.kt:65)
I believe the problem is the lack of type information for the column car , but it's unclear to me how to provide it.

[roman.belov]

Oh, looks like some weird bug for me
I don't understand why, ATM, but it depends on the placement of the data class. And the code below perfectly works for me, while yours fails:
cc @Nikita Klimenko [JB]

[Jolan Rensen [JB]]

@holgerbrandl actually, this is a bug, something with detecting visibility in our type guessing system. If you write a class inside a function body, they are not considered "public" and will be cast to Any. Then unfold has no way to guess the properties of the class anymore. You can see this if you print the .schema() of cars.

I'll make an issue for it. In the meantime you can sorta "force" the df to recognize the "car" column as `Car`:

val cars: DataFrame<*> = dataFrameOf("owner", "car")(
    "Max", Car("audi", "a8"),
    "Tom", Car("toyota", "corolla"),
).convert("car").with { it as Car }

then unfold works fine

[Jolan Rensen [JB]]

https://github.com/Kotlin/dataframe/issues/928 thanks for letting us know 🙂

[holgerbrandl]

Thanks for your kind support.