Add the @rename macro

[metanoid]

It's not strictly necessary, because DataFrames.rename exists, but it is nice to have the keyword argument form.

[pdeffebach]

Thanks!

This will be good to add pre 1.0.

[pdeffebach]

Bumping this. Current struggling to rename something programmatically in dplyr. Would be great to have this in DataFramesMeta.jl

[MatthewRGonzalez]

@pdeffebach would something silly like this be acceptable?

using DataFrames
using Chain

macro testrename(x, arg)
    esc(testrename_helper(x, (arg)))
end

function testrename_helper(x, arg)
    t = arg
    quote
        $DataFrames.rename($x, $t)
    end
end

df = DataFrame(x=rand(10), y=rand(10))
df[!, Symbol("Y 2")] = randn(10)


@chain df begin
    @testrename [:x, :y] .=> [:apple, :banana]
    @testrename Symbol("Y 2") => :pineapple
end

[MatthewRGonzalez]

Hi, @pdeffebach. I want to bump this to discuss the design. Should the final version of @rename be something like

df(old_col1 = 1:10, old_col2 = 1:10)

@rename(df :new_col = :old_col)

@rename df begin  
     :new_col1 = :old_col1
     :new_col2 = :old_col2
end

Please let me know what you think!

[pdeffebach]

Yes. I guess rename should allow arbitrary expressions in the RHS so that you can do

@rename df :newcol = begin
    "old_col" * "1"
end

and of course `@rename df $(:x => :y)

[MatthewRGonzalez]

Yes. I guess rename should allow arbitrary expressions in the RHS so that you can do

To make sure I'm understanding this correctly, would the DataFrames version of achieving this be something like

df = DataFrame(a_col = rand(10))
rename(x->x * "1", df)

[pdeffebach]

No. see the :newcol = ... at the top, it would be equivalent to

rename(df, ("old_col" * "1") => :new_col)

@rename should only ever create a rename(df, a => b, c => d) type-of expression. No need to bother with the function stuff.

[MatthewRGonzalez]

Got it, thanks. I currently have the following:

@rename df :new1 = begin 
    $("old_col" * "1")
end

however, arbitrary RHS expressions need to be escaped. Further, arbitrary RHS expressions can not be mixed with other standard expressions. For example

@rename(df, :new1 = $("old_col" * "1"), :new2 = :old_col2)

errors in the same manner as

rename(df, ("old_col" * "1") => :new1, :old_col2 => :new2)
ERROR: MethodError: no method matching rename!(::DataFrame, ::Vector{Pair{A, Symbol} where A})

Which I think makes sense if this implementation is to rely on rename. Is this sufficient?

[pdeffebach]

The error above is because all the old names need to be the same type and all the new names need to be the same type. Just cast everything to string in the final expression. i.e., produce

rename(df, string(("old_col" * "1")) => string(:new1), string(:old_col2) => string(:new2))