I just found this discussion on SO and it was very helpful:
http://stackoverflow.com/questions/28271308/avoid-memory-allocation-when-indexing-an-array-in-julia
Julia allocates a new array on indexing every time.
Here is how to achieve it if someone is trying to figure the same thing out
:
julia> a = [1,2,3,4,5,6,7,8,9,0]
10-element Array{Int64,1}:
1
2
3
4
5
6
7
8
9
0
julia> b = sub(a,[1,4,7])
3-element SubArray{Int64,1,Array{Int64,1},(Array{Int64,1},),0}:
1
4
7
julia> b[1] = 2
2
julia> a
10-element Array{Int64,1}:
2
2
3
4
5
6
7
8
9
0
सोमवार, 4 मई 2015 को 3:04:53 पूर्व UTC+5:30 को, Maco Anshu ने लिखा:
>
> Hi,
>
> When I index over a Julia array:
>
> julia> x = reshape(1:16, 4, 4)
> 4x4 Array{Int64,2}:
> 1 5 9 13
> 2 6 10 14
> 3 7 11 15
> 4 8 12 16
>
> julia> b = x[2:3, 2:end-1]
> 2x2 Array{Int64,2}:
> 6 10
> 7 11
>
> julia> b[1,1] = 3
> 3
>
> julia> b
> 2x2 Array{Int64,2}:
> 3 10
> 7 11
>
> julia> x
> 4x4 Array{Int64,2}:
> 1 5 9 13
> 2 6 10 14
> 3 7 11 15
> 4 8 12 16
>
> When does the indexed part of the array gets copied to a new block of
> memory ? Is b produced as a different array just after indexing or is it
> copied when one of its elements are changed or does Julia manage this using
> some other technique ?
> Another question over this is if we have a very huge array (which somehow
> just fits in RAM), is it possible to do indexing over it without using more
> and more RAM (avoiding copies) ?
> Something like Python :
> >>> x = np.array([[1,2,3],[4,5,6],[7,8,9]])
> >>> b = x[1:2,:]
> >>> b
> array([[4, 5, 6]])
> >>> b[0] = 0
> >>> x
> array([[1, 2, 3],
> [0, 0, 0],
> [7, 8, 9]])
>
> I am not aware of how Julia actually works while indexing and it might be
> doing something cool like clojure to manage such indexes and avoid copies,
> so please forgive if the question is really stupid.
>
