Access inner function variables in Javascript

You are still missing the point.

The point of unit testing is to verify that the object's public interface does what is expected of it. Unit tests show how the code works.

The only thing that should be tested is the object's public interface. That way, when you the developer want to change how the object is implemented, you only worry about if the object being tested still does what it is expected of it.

If you feel that the object that is inside that closure needs testing, then test that, but do it externally and then pass it into the closure.

var Raphael= function(listIterator) {
  listIterator.method();
}(new ListIterator());

Spurious hacks, such as shown below, are totally inappropriate (in unit tests or anywhere).

Test functions should be simple, test one thing only, and have one assertion. This can usually happen in three to ten lines of test code.

When you get to the point where your test functions are complicated, as they would be following the approach you are asking about, then either (1) realize that your design might not be what you want it to be and change it so that it is, or (2) change your expectations in the test.

Regarding the code you posted, you forgot var, missed a semicolon, and used two reserved words as identifiers: private and public.

The consequence of not using var is the potential to trigger errors and problems related to various implementations of nonstandard GlobalScopePolluter-type objects ("Object doesn't support this property or method" seen in IE). The consequence of using a FutureReservedWord is SyntaxError. Implementation may provide a syntax extesion to allow FutureReservedWord as identifier, and indeed many do, however it is best to not rely on such extensions and if you got an error, it would be completely your fault.

You mentioned delivering code to users. I suggest that you not do that until you get some more experience and understanding with what you are doing.

// DO NOT USE THIS CODE.
var Raphael = (function(){
    var _private = function(a,b) {return a+b;};
    var _public = function(a) {return _private(a,a);};
    var object = {mult2:_public};
    return object;
})();

var leakedFunction;

// Spurious hack:
//   Give valueOf a side effect of leaking function.
//   valueOf is called by the _private function as a
//   side effect of primitive conversion, where 
//   ToPrimitive(input argument, hint Number) results 
//   in calling valueOf.

function valueOfSnoop(){ 
    leakedFunction = leakedFunction || valueOfSnoop.caller || function(){};
    return 2;
}

var a = {
  valueOf : valueOfSnoop
};

Raphael.mult2(a, 3);
var privateMathod = leakedFunction;
alert(leakedFunction(1, 2));

That example code is only as a demonstration that such thing is possible. Given the choice, it is a poor alternative to the alternatives mentioned earlier; either change your design or change your tests.

Try this:

var adder = function(a,b) {
    return a + b;
}

Raphael = function(fn){
    var _private = function(a,b) {
        fn(a,b);
    }

    var _public = function(a) {
        return _private(a,a);
    }

    var object = {doubleIt: _public};

    return object;
}(adder);

Just a little function injection

The best solution I came up with:

In source Javascript file use

Raphael = (function(){
// start linked_list
    var linked_list = function() {/*...*/};
// end linked_list
    var object = {mult2:_public};
    return object;
})();

Now, use a script to extract objects between // start ([a-zA-Z_]*) and // end ([a-zA-Z_]*), and unit test the extracted code.

Apparently it is impossible to access variables in the inner scope of a function from an outer scope. As written in the SO question Jason linked to in the comments.

var Raphael;
var test = true; //or false;

Raphael = (function(){
    var private = function(a,b) {return a+b;};
    var public = function(a) {return private(a,a);}
    var object = {mult2:public};

    if (test) Raphael.private = private;

    return object;
})();