Passing variable from anonymous objective function to main workspace

Hello guys, I have a quick and simple question for you about passing variables from anonymous objective functions.

[x, fval] = fmincon(@(x)objfun(x,var1, var2),x0)

And in this objfun I am calculating and saving some information in variable data (cell). After optimization is done, I would like to have that variable into my workspace. Currently I am declaring it as a global variable and of course since it's bad practice. Is there an alternative way to do it?

Saving into variable data for each iteration seems a little bit tedious.

[Mario Malic]

+1 on both answers since they both solve the problem. It takes a bit of time for me to understand nested functions, especially the way they are written. I went for Stephen's answer, just because of simplicity.

Answer 1

[Stephen23]

You can do this easily with nargin:

function val = objfun(x,v1,v2)
persistent data
if ~nargin
    val = data;
    return
end
... your code
end

and then call it after the fmincon call:

data = objfun();

Answer 2

[Geoff Hayes]

Mario - consider nesting your optimization function within the main function and have it return the variable to the workspace. See Nested Functions for details. An example of this might be

function [data] = myMainFunction

    data = [];
    function myNestedFunction(x)
        % do something
        for k = 1:ceil(x)
            data(k) = k;
        end
    end

    % call the nested function
    myNestedFunction(42);
    
end

The above code would be saved to a file named myMainFunction.m. It would be called from the command line as

>> x = myMainFunction;

[Mario Malic]

Thanks Geoff, I'll try and report tomorrow.