How can i put a function into a function in Julia?-CodePudding

I want to run a code from a paper by Julia. There is a function about data storage. The function need to insert a function called oracle. I create a c vector and put it into its code but there comes an error. Here is my code :

function portfolio(c::Vector{Float64})
    d = length(c)

    mod = Model(with_optimizer(Gurobi.Optimizer, OutputFlag = 0))
    @variable(mod, 0<= w[1:d] <= 1)
    @constraint(mod, sum(w[i] for i = 1:d) <= 400)

    @objective(mod, Min, dot(c, w))

    optimize!(mod)
    z_ast = objective_value(mod)
    w_ast = value.(w)

    return (z_ast, w_ast);
end

c_df = vec(df[2])
X_df = df[1]
function oracle_dataset(c,oracle)
    (d, n) = size(c)
    z_star_data = zeros(n)
    w_star_data = zeros(d, n)
    for i = 1:n
        (z_i, w_i) = oracle(c[:,i])
        z_star_data[i] = z_i
        w_star_data[:,i] = w_i
    end
    return (z_star_data, w_star_data)
end
oracle_dataset(c_df, portfolio)

The error is:

BoundsError: attempt to access Tuple{Int64} at index [2]

Stacktrace:
 [1] indexed_iterate
   @ .\tuple.jl:86 [inlined]
 [2] oracle_dataset(c::Vector{Float64}, oracle::typeof(portfolio))
   @ Main .\In[15]:63
 [3] top-level scope
   @ In[15]:73
 [4] eval
   @ .\boot.jl:360 [inlined]
 [5] include_string(mapexpr::typeof(REPL.softscope), mod::Module, code::String, filename::String)
   @ Base .\loading.jl:1116

Why does this happen? I have tried another easier example and there is no error in putting a function inside a function. My easier code is :

function foo(x)
    g = x*3
    return g;
end

function bar(c,func)
    result = func(c)
    return result;
end

num = 3
bar(num,foo)

Thank you for your help...I really want to know what mistake i had done in here. Thank you~

CodePudding user response：

It looks like you try to access second element in tuple, while there is only one element, probably here:

c_df = vec(df[2])

CodePudding user response：

The c_df which you pass to oracle_dataset is actually a Vector (which you can see in the 6th line of the stacktrace, starting with [2]), which is one-dimensional. This means that size(c) in the first line of the oracle_dataset function will return a 1-tuple. However, the author of that function seems to be assuming that c will be a Matrix, which is 2-dimensional. Calling size on that would return a 2-tuple, which is deconstructed to (d, n) in the function.

The error seems to specifically trigger on the line :

    (d, n) = size(c)

meaning that you would likely need to supply a Matrix as c.

I can't speak to what that matrix would look like or how you would construct it. Perhaps df[2] is a matrix itself?