CODEWITHSUNDEEP
arraysjulia
BoZenKhaa
BoZenKhaa

Reputation: 941

Efficient way to build julia array element by element

What's the best way to build a Julia array element by element when I am not sure of the final array size? In my problem, I am building the array in a single for loop.

So far, I have figured out I can either push into the array or index into a pre-allocated array:

using BenchmarkTools

function push_to_array!(array, N)
    for i in 1:N
        # some computation
        push!(array, rand(Int64))
    end
end

function fill_array!(array, N)
    for i in 1:N
        # some computation
        array[i]=rand(Int64)
    end
end

N = 100_000_000 # unknown in the real problem

empty = Vector{Int64}()
preallocated = Vector{Int64}(undef, 2*N) # 2*N represents some upper bound on N

@btime push_to_array!(empty, N)
#  28.272 s (6 allocations: 0 bytes)
@btime fill_array!(preallocated, N)
#  2.449 s (0 allocations: 0 bytes)

So filling a pre-allocated array is a lot faster than pushing, however, it's a bit cumbersome as I need to trim the output with correct_size = preallocated[1:N].

Is there a faster/better way of doing this?

Upvotes: 2

Views: 709

Answers (1)

Przemyslaw Szufel
Przemyslaw Szufel

Reputation: 42214

In Julia you have views - no trimming required. Simply do:

correct_size = @view preallocated[1:1_000_000]

or

correct_size = view(preallocated, 1:1_000_000)

Now this operation is very cheap (BTW note also that you are not bencharking correctly as @time measures both compile time and execution time):

julia> using BenchmarkTools

julia> @btime correct_size = @view $preallocated[1:1_000_000];
  1.800 ns (0 allocations: 0 bytes)

In conclusion making a view is basically free.

Finally, note that you can just resize! the array (the time cost is the same as in case of a @view):

resize!(preallocated, 1_000_000)

Upvotes: 5

Related Questions