Move to(Start) keeping inside group

[CarloMariaProietti]

About #1255
It comes with docs and a test case

[Jolanrensen]

Oh, I see we also have these shortcut functions:

• public fun <T> DataFrame<T>.moveTo(newColumnIndex: Int, columns: ColumnsSelector<T, *>)
• public fun <T> DataFrame<T>.moveTo(newColumnIndex: Int, vararg columns: String): DataFrame<T>
• moveToStart(), moveToEnd()

Could you do the same there as with .to()? :)

[CarloMariaProietti]

Good Morning, I implemented a new logic as suggested in #1489 (comment).
Each test works, I'll proceed to add shortcut functions as requested in
#1489 (comment).

[CarloMariaProietti]

about #1489 (comment) ,

Now there are both toStart() and toEnd(),

Oh, I see we also have these shortcut functions:

• public fun <T> DataFrame<T>.moveTo(newColumnIndex: Int, columns: ColumnsSelector<T, *>)
• public fun <T> DataFrame<T>.moveTo(newColumnIndex: Int, vararg columns: String): DataFrame<T>

.to() implements yet these shortcuts because it is possible to create a MoveClause with both ColumnsSelector<T, *>
and vararg String

[Jolanrensen]

oh, please keep

@Refine
@Interpretable("MoveToStart0")

I will add compiler plugin for this variant of the function. It can use the same interpreter as toStart(), but it needs to be merged first and published as dev version so we can test it works in the compiler plugin in the Kotlin repo.

[Jolanrensen]

@CarloMariaProietti I'm not sure what you mean,
We can write

public fun <T> DataFrame<T>.moveTo(newColumnIndex: Int, columns: ColumnsSelector<T, *>): DataFrame<T> =
    move(columns).to(newColumnIndex)

public fun <T> DataFrame<T>.moveTo(newColumnIndex: Int, insideGroup: Boolean, columns: ColumnsSelector<T, *>): DataFrame<T> =
    move(columns).to(newColumnIndex, insideGroup)

and

public fun <T> DataFrame<T>.moveTo(newColumnIndex: Int, vararg columns: String): DataFrame<T> =
    moveTo(newColumnIndex) { columns.toColumnSet() }

public fun <T> DataFrame<T>.moveTo(newColumnIndex: Int, insideGroup: Boolean, vararg columns: String): DataFrame<T> =
    moveTo(newColumnIndex, insideGroup) { columns.toColumnSet() }

right?

If you want, we could also add these in another PR of course.

[Jolanrensen]

please specify "insideGroup", as it would be fine to move them with insideGroup = false

[Jolanrensen]

if all columns here have the same parent... isn't their "reduced path" simply their name?

[Jolanrensen]

It's great you made it work! But I have a feeling this is way more complicated than it needs to be.

[Jolanrensen]

My feeling was right, I managed to pass all your tests with this logic:

val columnsToMoveNames = columnsToMove.map { it.name() }
return df.replace { parentOfFirst.asColumnGroup() }.with {
    it.asDataFrame()
        .move { columnsToMoveNames.toColumnSet() }.to(columnIndex)
        .asColumnGroup(it.name())
}

Let me explain what's done here:

• we take the column names of those we want to move. This is enough, because we already checked they have the same parent.
• since we now only need to move columns around on the same level, this is just as easy as moveTo() on the top level. We can use replace { parent } to get access to just the parent group
• then, treating the column group as a DataFrame, we can simply call move {}.to()
• and convert it back to a column group with the right name

[CarloMariaProietti]

Thank you for the explanation, your solution is much more smooth. I ignored replace { } so i had to remove the level and reinsert it... If you agree, I'll paste this in the code.

[Jolanrensen]

Suggestion from my colleague: Can you create your own exception that implements DataFrameError? that way we can catch it in the compiler plugin later :)

[CarloMariaProietti]

Done!

[Jolanrensen]

make sure columns is the last argument, else the lambda outside parentheses doesn't work. For consistency, you can do the same in the vararg String overloads

[Jolanrensen]

that way people can write:

df.moveToStart { cols }
// and
df.moveToStart(insideGroup = true) { cols }

[CarloMariaProietti]

Done, I fixed also moveToEnd