Update column operations documentation

[koperagen]

column arithmetics are essentially obsolete IMO, doesn't make sense to keep this page around

[Jolanrensen]

related issues: #530, #265

I wouldn't say they're obsolete actually. I've seen AI recommend colA + colB on multiple occasions. If we can, why shouldn't we support such a thing?

[koperagen]

I think they are obsolete in a sense that we don't have cases where we'd recommend using it. At least i don't. dataframe is not pandas, and to be useful in Kotlin most of the time you need to operate on individual objects / values. Which it the default for majority of our operations, except for rare groupBy { s.length() }. But even there i'd recommend groupBy { expr { s.length } } because we do have DataColumn<String>.length(), but not DataColumn<List>.size().

We can keep these functions around (i don't suggest to deprecate them), but i prefer to migrate examples from using it whenever i see. If you have specific examples when you see it to be useful, please share

[Jolanrensen]

@koperagen I think most useful examples are from Kandy:

Let's say you have some data with values and corrections of those values:

val df = dataFrameOf(
    "ID" to columnOf(1L, 2L, 3L),
    "score" to columnOf(10, 20, 30),
    "correction" to columnOf(-0.5, 0.0, 0.5),
)

It would be nice to plot the immediate results of the scores with correction without having to add a new "temporary" column, just for plotting:

df.plot { 
    bars { 
        x(ID)
        y(score + correction)
    }
}

(this is currently not possible between 2 columns, and there is no easy way to zip two columns... There's add {} but this requires an extra cell in notebooks (for now))

"column operator value" or "value operator value" can usually be replaced with something like expr { column + value } or a column.map { it + value }. But it may be nice to be able to do a x(columnA / 100.0) at once in Kandy.

[koperagen]

@koperagen I think most useful examples are from Kandy:

True! I can easily imagine that, i remember writing similar code myself.
I suggest to remove the page, and when we decide to introduce it back - let's provide "goal" of that API

1. That most of the time column operations are not needed in dataframe unlike pandas, because we provide row-based APIs
2. Recommended "adapter" functions like map or expr
3. Example where it might still be useful, and only then what APIs we provide

[Jolanrensen]

@koperagen good plan :) Certainly better to remove it than have a // TODO on a page