Ctrl-C exception in REPL

[ilicog]

Is there a snippet of code / example of throwing a CTRL-C exception that would gracefully halt an NLopt optimization in julia REPL?

[stevengj]

throw(InterruptException())?

[ilicog]

somewhat confused, so I was ideally looking for a MWE (of something that works in REPL/terminal/for multiple processes)

[stevengj]

Just throw the exception from your objective function.

[ilicog]

In this dummy example (that will just run for 10 sec)

using NLopt

function myfunc(x::Vector, grad::Vector)
    global count
    count::Int += 1
    
    res = sqrt(x[2])

    return res
end

function run_opt()
    opt = Opt(:GN_ISRES, 2)
    upper_bounds!(opt, [100.0, 100.0])
    lower_bounds!(opt, [0.0, 0.0])
    min_objective!(opt, myfunc)
    maxtime!(opt, 10.0)

    (minf,minx,ret) = optimize(opt, [1.234, 5.678])
    return (minf,minx,ret)
end

would throw(InterruptException()) go inside myfunc, with a forced_stop!(opt) in run_opt() and if so how exactly? Tried several things, can't seem to get it right.

[mzaffalon]

using NLopt

function myfunc(x::Vector, grad::Vector)
    throw(InterruptException())
    0.0
end

opt = Opt(:GN_ISRES, 2)
min_objective!(opt, myfunc)
maxtime!(opt, 10.0)

When I run it, I get

julia> optimize(opt, [1.234, 5.678])
in callback catch
ERROR: InterruptException:
Stacktrace:
 [1] myfunc(::Array{Float64,1}, ::Array{Float64,1}) at ./REPL[2]:2
 [2] nlopt_callback_wrapper(::UInt32, ::Ptr{Float64}, ::Ptr{Float64}, ::Ptr{Void}) at /home/michele/.julia/v0.6/NLopt/src/NLopt.jl:427
 [3] optimize!(::NLopt.Opt, ::Array{Float64,1}) at /home/michele/.julia/v0.6/NLopt/src/NLopt.jl:526
 [4] optimize(::NLopt.Opt, ::Array{Float64,1}) at /home/michele/.julia/v0.6/NLopt/src/NLopt.jl:532

[avonmoll]

I read the README and also found this thread, but I never seem to get exceptions back when I call optimize. The snippet in the previous post just causes the optimizer to return :FORCED_STOP. So, for real code, the optimizer will run for a while and then hit a case where there is an exception and immediately returns :FORCED_STOP and I am left with no information on what went wrong.

[odow]

Closing because this seems to work fine.

julia> count = 0
0

julia> using NLopt

julia> function myfunc(x::Vector, grad::Vector)
           global count
           count::Int += 1
           
           res = sqrt(x[2])

           return res
       end
myfunc (generic function with 1 method)

julia> function run_opt()
           opt = Opt(:GN_ISRES, 2)
           upper_bounds!(opt, [100.0, 100.0])
           lower_bounds!(opt, [0.0, 0.0])
           min_objective!(opt, myfunc)
           maxtime!(opt, 10.0)

           (minf,minx,ret) = optimize(opt, [1.234, 5.678])
           return (minf,minx,ret)
       end
run_opt (generic function with 1 method)

julia> run_opt()
^C(0.0, [6.676949509795536, 0.0], :FORCED_STOP)

The forced stop issue is covered by #156.