I am confused about the following behaviour of Julia when using array comprehensions that have arrays/matrices in them. They seemingly seem to bury the other dimensions of the inside array. I am looking to get an array of dimension 2,4,3 in the end.
myarr = Array{Float64}(undef,4,3)
broken_off_dim = [myarr for i in 1:2]
size(broken_off_dim)
Somehow, I get (2,) and I wonder what happened to the inner dimensions? Can you help modify the above code so that (2,4,3) comes out? I tried reshaping but that throws an error.
Functions that are applied over a specific margin also do not work properly then.
If you want a comprehension to make one array rather than an array of arrays, then you probably want stack.
julia> vv = [[i, 2i, 3i] for i in 1:4]
4-element Vector{Vector{Int64}}:
[1, 2, 3]
[2, 4, 6]
[3, 6, 9]
[4, 8, 12]
julia> mat = stack(vv) # inner dimension(s) go first
3×4 Matrix{Int64}:
1 2 3 4
2 4 6 8
3 6 9 12
julia> mat == stack([i, 2i, 3i] for i in 1:4)
true
julia> mat == stack((i, 2i, 3i) for i in 1:4)
true
Notice that it accepts an un-collected comprehension generator too, without making the Vector{Vector{T}} first. In fact almost any iterator of iterators.
With more dimensions, all dimensions are preserved. (I think the example above probably wants stack(myarr(i) for i in 1:2; dims=1).)
julia> vm = [[i 10i; 2i 21i] for i in 1:4]
4-element Vector{Matrix{Int64}}:
[1 10; 2 21]
[2 20; 4 42]
[3 30; 6 63]
[4 40; 8 84]
julia> vm[end]
2×2 Matrix{Int64}:
4 40
8 84
julia> stack(vm) |> summary
"2×2×4 Array{Int64, 3}"
julia> stack(vm; dims=2) |> summary
"2×4×2 Array{Int64, 3}"
On Julia 1.8 or earlier, you will need using Compat for this.