Why is deepcopy() -ing Sets slower than Arrays of the same size?

General Usage Performance
2

Yes, this is a case where reading the implementation (which is very readable Julia code) makes clear that it must strictly be slower since the Set implementation contains an Array inside of it.

If you read the Set definition, you’ll see:

struct Set{T} <: AbstractSet{T}
    dict::Dict{T,Nothing}

    Set{T}() where {T} = new(Dict{T,Nothing}())
    Set{T}(s::Set{T}) where {T} = new(Dict{T,Nothing}(s.dict))
end

At which point you might jump to the Dict definition to see:

mutable struct Dict{K,V} <: AbstractDict{K,V}
    slots::Array{UInt8,1}
    keys::Array{K,1}
    vals::Array{V,1}
    ndel::Int
    count::Int
    age::UInt
    idxfloor::Int  # an index <= the indices of all used slots
    maxprobe::Int

    function Dict{K,V}() where V where K
        n = 16
        new(zeros(UInt8,n), Vector{K}(undef, n), Vector{V}(undef, n), 0, 0, 0, 1, 0)
    end
    function Dict{K,V}(d::Dict{K,V}) where V where K
        new(copy(d.slots), copy(d.keys), copy(d.vals), d.ndel, d.count, d.age,
            d.idxfloor, d.maxprobe)
    end
    function Dict{K, V}(slots, keys, vals, ndel, count, age, idxfloor, maxprobe) where {K, V}
        new(slots, keys, vals, ndel, count, age, idxfloor, maxprobe)
    end
end
2 Likes