Julia Distributed - why the variable is not defined on worker after evaluating the expression?

2
votes
function pass_object(src::Int, target::Vector{Int8}, name::Symbol; from_mod=Main, to_mod=Main)
    r = RemoteChannel(src)
    @sync @spawnat(src, put!(r, getfield(from_mod, name)))
    @sync for to in target
        @spawnat(to, Core.eval(to_mod, Expr(:(=), name, fetch(r))))
    end
end

variable1 = some_data
@sync pass_object(1, pid, :variable1)
result = @spawnat pid Core.eval(Main, :(new_variable=variable1))

Julia 1.5.3

I tried multiple ways how to pass object to another worker and then use the assigned variable, but the result is always that it is not defined. There are no errors before I try to access its attribute, e.g. remotecall_fetch(println, pid, new_variable.label)

Any ideas why this does not work? Thank you for any help.

1
julia

1 Answers

2
votes

Use ParallelDataTransfer for that. For an example.

using ParallelDataTransfer, Distributed
addprocs(1)
@everywhere using ParallelDataTransfer, Distributed

variable1 = [1,2,3]
@passobj 1 2 variable1

Now some console tests:

julia> fetch(@spawnat 2 sum(variable1))
6

julia> fetch(@spawnat 2 variable1[1]=1000)
1000

julia> sum(variable1)
6

julia> fetch(@spawnat 2 sum(variable1))
1005