Possible to convert ND arrays to nested arrays "in place"?

New to Julia
1

I know I can reinterpret an array of SVector as an ND array as follows:

julia> SA=[@SVector rand(2) for _=1:100];

julia> A=reinterpret(reshape,Float64,SA);

julia> @show SA[1],A[:,1];
(SA[1], A[:, 1]) = ([0.01565087031449286, 0.4977468415798615], [0.01565087031449286, 0.4977468415798615])

I can even modify the SVector through the reinterpretation:

julia> A[:,1].=0.0;

julia> @show SA[1],A[:,1];
(SA[1], A[:, 1]) = ([0.0, 0.0], [0.0, 0.0])

My question is, given a 2D array like A above, is there a way to convert to a 1D array of SVectors in-place, like SA above, i.e., sharing the same memory ?

2

Just use reinterpret:

julia> A = rand(3,6)
3×6 Matrix{Float64}:
 0.781215  0.864153  0.190954  0.513353   0.684802  0.304443
 0.717076  0.248486  0.394957  0.231284   0.572842  0.174061
 0.605211  0.814295  0.173915  0.0166475  0.597283  0.482926

julia> B = reinterpret(SVector{3,Float64}, vec(A))
6-element reinterpret(SVector{3, Float64}, ::Vector{Float64}):
 [0.7812150216015193, 0.7170764937769242, 0.6052113147390563]
 [0.8641531429329602, 0.24848604689820653, 0.8142954282406346]
 [0.1909536286428969, 0.3949565597637975, 0.17391456890697454]
 [0.5133528560285338, 0.23128368035325675, 0.016647532659139097]
 [0.6848016310407239, 0.5728415428873161, 0.5972827074627644]
 [0.3044430234630139, 0.17406109329315256, 0.4829260227587332]

julia> A[1,1] = 3.14159
3.14159

julia> B[1]
3-element SVector{3, Float64} with indices SOneTo(3):
 3.14159
 0.7170764937769242
 0.6052113147390563
1 Like