Add vararg groupby

[CameronBieganek]

Sorry if this has been discussed before, but I couldn't find any related issues.

It would be nice if we could write

groupby(df, :a, :b)

instead of

groupby(df, [:a, :b])

This is just me being greedy. :) But off the top of my head I can't think of any reason why it wouldn't work.

[bkamins]

It was not possible before because of by. However, since we will drop it in 0.22 I think we can add it. Let us wait for others to comment (in particular @pdeffebach in connection with DataFramesMeta.jl - but I think there should be no problem there).

[pdeffebach]

I don't think it's that important, tbh. It would be mildly annoying for DataFramesMeta because we will keep @by because users like it.

I don't think its a great idea to open up a pandora's box of adding vararg everywhere we can.

[CameronBieganek]

Well, it seems a little inconsistent that you can do select(df, :a, :b) but not groupby(df, :a, :b).

To elaborate on the inconsistency, the docstring for groupby says that the second argument can be any column selector. However, it leaves out one type of column selector, namely the vararg column selector, which is available in select. For select, these are all equivalent:

julia> df = DataFrame(a=1:2, b=3:4, c=5:6);

julia> select(df, All(:a, :b))
2×2 DataFrame
│ Row │ a     │ b     │
│     │ Int64 │ Int64 │
├─────┼───────┼───────┤
│ 1   │ 1     │ 3     │
│ 2   │ 2     │ 4     │

julia> select(df, [:a, :b])
2×2 DataFrame
│ Row │ a     │ b     │
│     │ Int64 │ Int64 │
├─────┼───────┼───────┤
│ 1   │ 1     │ 3     │
│ 2   │ 2     │ 4     │

julia> select(df, :a, :b)
2×2 DataFrame
│ Row │ a     │ b     │
│     │ Int64 │ Int64 │
├─────┼───────┼───────┤
│ 1   │ 1     │ 3     │
│ 2   │ 2     │ 4     │

[bkamins]

"column selector" in the docstring means any value accepted by getindex function of AbstractIndex.

Please note that we allow select(df, :a, :b) only because in select we allow multiple operations to be performed in a chain. Actually in this specific case it is rewritten to select(df, :a => identity => :a, :b => identity => :b) before being processed and is handled differently than e.g. select(df, [:a, :b]).

For instance select(df, [:a, :b], :a => ByRow(sin) => :b) will work, while select(df, :a, :b, :a => ByRow(sin) => :b) will error.

Having said that:

1. @pdeffebach is right that it is not that obvious as it looks like
2. still it would be doable of course and some rules can be established

Since it is non-breaking there is no rush to make this decision and let us wait for more feedback.