Dataframe Plugin disabling

[mattbaumann]

I've worked with the library and the plugin for the last few weeks. I think the library is great and I get the hang of it.

For the plugin I've a wish I'd like to let you know about.

Is it possible to disable the plugin automatically when the Datatype is DataFrame<*>. This is because, the type signifies that I want to use the string-based accessors anyhow. This is also a performance consideration, since

tablea = DataFrame.emptyOf<Any>() as DataFrame<*>
tableb = DataFrame.emptyOf<Any>() as DataFrame<*>
(df as DataFrame<*>).concat(tablea.rows(), tableb.rows())

is forbidden by the plugin. It leads to a compiler error as the plugin has created types for tablea and tableb. But

tablea = DataFrame.emptyOf<Any>() as DataFrame<*>
tableb = DataFrame.emptyOf<Any>() as DataFrame<*>
(df as DataFrame<*>).concat(tablea.rows()).concat(tableb.rows())

is compilable and executable. This is most certainly performance break and IMHO confusing difference.

Does AnyFrame have any magic in that regard with the plugin?

[koperagen]

Hi! For now i don't understand the idea for plugin fully
From these examples, in first case code doesn't compile because we don't have DataFrame.concat(vararg Iterable<DataRow<*>>)

dataframe/core/src/main/kotlin/org/jetbrains/kotlinx/dataframe/api/concat.kt

Line 31 in ddb2eb3

public fun <T> DataFrame<T>.concat(vararg frames: DataFrame<T>): DataFrame<T> = concatImpl(listOf(this) + frames)

tablea = DataFrame.emptyOf<Any>() as DataFrame<*>
tableb = DataFrame.emptyOf<Any>() as DataFrame<*>
(df as DataFrame<*>).concat(tablea.rows(), tableb.rows())

With vararg dataframes: DataFrame it should work:

(df as DataFrame<*>).concat(tablea, tableb)

Is it possible to disable the plugin automatically when the Datatype is DataFrame<*>

It can be useful to gradually introduce typing:

val rawDf = DataFrame.readJson(...)
rawDf
  .add("myCol") { "field1"<String>().toInt() }
  .myCol

But depending on what problems plugin-generated types cause we can figure something out