make sure aggregation uses fast path if a single column is used

[bkamins]

I will fix it. Currently :x => sum uses a fast path, but [:x] => sum does not.

[sa-]

Looking into this. Pasting some of your Slack messages here in case the messages get deleted:

fast path is implemented here: https://github.com/JuliaData/DataFrames.jl/blob/main/src/groupeddataframe/fastaggregates.jl

10:30
In the fast path you do not have to calculate grouping views. You can just pre-allocate the target container (whose eltype you know upfront) and just iterate groups and perform reduction.

[bkamins]

The key thing to do here is to rewrite code like:

@chain df begin
    groupby(:a)
    combine([:b] => sum)
end

to:

@chain df begin
    groupby(:a)
    combine(:b => sum)
end

before executing it.

We should check in the syntax src => fun and src => fun => dst if src is Union{AbstractVector{<:AbstractString}}, AbstractVector{Symbol}, AbstractVector{<:Union{Signed, Unsigned}} of length 1 and if this is so its first element should be fetched with only before further processing.

Or even better - check that src after preprocessing the column list using a standard mechanism of src preprocessing is a single element vector (it then should be Vector{Int} as this is what Index should ensure).

[sa-]

Can this be closed now?

[bkamins]

yes