ArrayAssemblers offers two methods to assemble arrays and higher-order tensors out of arrays-of-arrays.
The motivation for this package comes mostly from two use patterns:
block(f, a)is an alternative to aflatmapfunctor. It works essentially the same as[y for x in c1 for y in f(x)]. It produces an array and supports do-syntax.lay(f, a)works as a shortcut forhcat(vector_of_vectors...)orreduce(hcat, vector_of_vectors), but it also supports do-syntax and can take generators.
julia> using ArrayAssemblers
julia> a1 = [1 2; 3 4]
2×2 Matrix{Int64}:
1 2
3 4
julia> a2 = [5 6; 7 8]
2×2 Matrix{Int64}:
5 6
7 8
julia> block([a1 a2])
2×4 Matrix{Int64}:
1 2 5 6
3 4 7 8
julia> block([a1, a2])
4×2 Matrix{Int64}:
1 2
3 4
5 6
7 8
julia> block([a1;;; a2])
2×2×2 Array{Int64, 3}:
[:, :, 1] =
1 2
3 4
[:, :, 2] =
5 6
7 8
julia> lay(a:a+1 for a in 1:2:6)
2×3 Matrix{Int64}:
1 3 5
2 4 6
## flatmap behavior with `block`
julia> Zn = block(1:3) do n
-n:2:n
end
9-element Vector{Int64}:
-1
1
-2
0
2
-3
-1
1
3
julia> Zn == [x for n in 1:3 for x in -n:2:n]
true
## Assembling a tensor with recursive calls to `lay` and `map` using do-syntax
julia> lay(0:1) do c
lay(0:2) do b
map(1:2) do a
(c * 3 + b) * 2 + a
end
end
end
2×3×2 Array{Int64, 3}:
[:, :, 1] =
1 3 5
2 4 6
[:, :, 2] =
7 9 11
8 10 12
Image montage (tiling):
using TestImages, ImageView
myimages = ["cameraman" "plastic_bubbles_he_512" "woman_darkhair"; "resolution_test_512" "pirate" "walkbridge"]
imshow(block(testimage.(myimages)))
