Add GroupBy variable converter in Jupyter #663
Conversation
| public data class CodeWithConverter(val declarations: Code, val converter: (VariableName) -> Code = EmptyConverter) { | ||
| public data class CodeWithConverter<T : CodeConverter>( | ||
| val declarations: Code, | ||
| val converter: T |
There was a problem hiding this comment.
maybe it's an idea to rename this variable or rework it a bit, because looking at this with new eyes, even with the documentation, I have no idea what's going on
There was a problem hiding this comment.
Idk, hard to tell more than expressed in the code itself. Basically CodeWithConverter is all the information that dataframe integration needs to generate except name of the variable being updated. Variable name is provided by the caller due to separation of concerns. So there's hardly any special "domain" meaning behind it
| @@ -7,14 +7,19 @@ import org.jetbrains.kotlinx.jupyter.api.VariableName | |||
| * Class representing generated code declarations for a [Marker]. | |||
| * | |||
| * @param declarations The generated code. | |||
| * @param converter Needs to provide additional info (name) from org.jetbrains.dataframe.impl.codeGen.CodeGenerator to its callers | |||
There was a problem hiding this comment.
there are two @param converters now
| @@ -27,3 +32,19 @@ public data class CodeWithConverter(val declarations: Code, val converter: (Vari | |||
| else -> declarations + "\n" + converter(name) | |||
| } | |||
| } | |||
|
|
|||
| public sealed interface CodeConverter : (VariableName) -> Code | |||
There was a problem hiding this comment.
This probably needs docs to explain what it does/should do. As it's a foreign concept and essentially just a String->String, which doesn't tell much. Maybe it could steal some docs from CodeWithConverter
| fun `groupBy`() { | ||
| """ | ||
| val groupBy = dataFrameOf("a")("1", "11", "2", "22").groupBy { expr { "a"<String>().length } named "k" } | ||
| groupBy.keys.k |
There was a problem hiding this comment.
wait, this works in the same cell?
There was a problem hiding this comment.
No :) The test executes each line individually. It's like 2 separate cells here
| * @param converter Optional converter for the [Marker], such as a [org.jetbrains.kotlinx.dataframe.api.cast], often used for Jupyter. | ||
| */ | ||
| public data class CodeWithConverter(val declarations: Code, val converter: (VariableName) -> Code = EmptyConverter) { | ||
| public data class CodeWithConverter<T : CodeConverter>( |
There was a problem hiding this comment.
I see most functions return a CodeWithWithConverter<*> and since both CodeConverterImpl and ProvidedCodeConverter are essentially just a (VariableName) -> Code, do we need the generic type T?
There was a problem hiding this comment.
For GroupBy i need to combine names of two generated interfaces together for cast, and so CodeWithConverter provides this additional info. This parameter doesn't make sense for other usages, so there we use more generic CodeWithWithConverter<*>
|
I'll also check whether it influences #662 or not |
|
what needs to be done about this pr? |
This was a long standing minor inconvenience. Having this variable in a cell1 you couldn't access
kin keys in the cell2cell1:
cell2:
Other use case is adding a new column to
GroupBygroupBy.aggregate { newCol into "newCol" }