Julia: columns, rows & slices views

[StefanKarpinski]

https://github.com/JuliaLang/julia/pull/32310 made me think of a discussion I'd had about data frames and wanting to be able to write columns(df) to access a data frame as a collection of columns. We could make eachcol(df) work like that but the each in the name gets a bit weird. So perhaps we might want to rename (or alias until 2.0) eachcol to columns and make it act more like a view of the underlying collection than as just an iterator. There would be corresponding rows and slices view wrappers. Thoughts?

[bkamins]

In DataFrames.jl eachcol and eachrow already produce AbstractArray objects. eachcol and eachrow were simply legacy names (and consistent with Base).

Therefore, at least from DataFrames.jl perspective it is not a problem if in Base the following decision were made:

• start using columns and rows as target function names and define const eachcol = columns and const eachrow = rows temporarily;
• in some future time drop eachrow and eachcol aliases

[nalimilan]

Indeed the "each" prefix sounds a bit off when the function returns (or could return) an AbstractVector, so +1 to switching to rows, columns and slices.

[StefanKarpinski]

The approach would be to make rows, columns and slices the primary names and leave the each* versions around but document them as being deprecated and then actually deprecate them in 2.0.

[AzamatB]

+1 for this change. Although this was originally discussed in #14491 (comment), where it was decided against rows, columns, slices and in favor of each*.