Search code examples
juliadeep-copy

Alternative to deepcopy() in Julia

I have the following code in Julia:

using LightGraphs
using LinearAlgebra
prop=[1 rand() rand() rand() rand(); rand() 1 rand() rand() rand(); rand() rand()  1 rand() rand(); rand() rand() rand() 1 rand(); rand() rand() rand() rand() 1]
prop=Symmetric(prop)
thresh=sort(unique(prop)
g=[SimpleGraph(0) for a in 1:length(thresh)]
for j=1:length(thresh)
    g[j]=Graph(Int.(prop.>=thresh[j])-I)
end

I need to create a copy of g, say g2, on which I need to use the function add_edge! but without modifying the original object g. I know that g2=deepcopy(g) will do perfectly the trick, but if g is very large (it's not the case in this simple example), the function deepcopy() takes a large amount of time to create the independent copy. I measured it with @time. Isn't there an alternative way in Julia to do what I want without taking a large amount of time for just making the copy of g? I have to run the code several times and want avoid to consume time just for copying g.

Solution

  • I don't think you want an alternative to deepcopy, which is already doing what you want. It's just that heap memory allocation (as done by deepcopy) takes time. What you can do is make a single copy of g, then update that every time you'd need the deepcopy function, it might look like this

    g_copy = deepcopy(g)
add_edge!(g_copy[i], edge)
# ... do stuff
g_copy .= g

    In that way you'll just change the contents of g_copy but never release the memory, so should be faster. I should mention this is untested (as I can't naturally expand your MWE, which doesn't use deepcopy or add_edge! at all).