Strange behaviour using hcat with arrays obtained from arraysplit

New to Julia
1

Hi all,

I experience a strange behaviour using hcat with arrays obtained from arraysplit (DSP package). My issue is explained in the code snippet below:

using DSP

a = 1:10
q = arraysplit(a,5,3)

A = hcat(q[1],q[2]) # Leads to 2 identical columns ??

For some reason A results in

julia> A
5×2 Array{Float64,2}:
 3.0  3.0
 4.0  4.0
 5.0  5.0
 6.0  6.0
 7.0  7.0

However, I would have expected A becomes as follows

julia> A
5×2 Array{Float64,2}:
 1.0  3.0
 2.0  4.0
 3.0  5.0
 4.0  6.0
 5.0  7.0

What am I misunderstanding here ?

Regards

2

Yeah I’ve noticed that as well, it’s probably due to the implementation of arraysplit to minimize allocations. You may use copy of the splits to get what you want.

https://github.com/JuliaDSP/DSP.jl/blob/8edec5c75a9461869c324ff6db4eefd7fd789a35/src/periodograms.jl#L42
Here you see how it returns the same array each iteration, it just has different values in it. Probably because views allocate?

3

Ok. Thank you for the hint. It works with

A = hcat(collect(q[1]),collect(q[2]))

Could you help me with another issue related to that? What if I want to concatenate many more arrays as given in the example below.

using DSP
a = 1:50
q = arraysplit(a,5,3)

A = hcat(collect(q[1]),collect(q[2]) ... collect(q[23])) #how do I achieve that with an e.g. for loop ?

Thank you again for your help…

4

You can try collect(copy.(arraysplit)). I’m not by a computer so can not try it out.

Or potentially reduce(hcat, copy.(arraysplit))

5

Thanks again. Gave up trying it with hcat. Solved it now as follows that gives me exactly what I need.

a = 1:10

q = arraysplit(a,5,3)
A = zeros(Float64,5,length(q))

for i = 1:length(q)
    A[:,i] = q[i]
end