On 2009-09-18, at 06:49, Rami Ojares / AMG Oy wrote:
Hi,
I noticed the following difference between swf8 and swf9 runtimes.
Code:
var functionVar = someObject["method_name"]
functionVar(args)
This works in swf9 but not in swf8
But you can still call the function in swf8 using construct
someObject["method_name"](args)
Does anybody know what is going on in here?
You are seeing a feature that is added by ActionScript 3 (AS3 is the
variant of Javascript supported by the swf9 player) that we cannot
emulate in the LZX compiler called "method closures"[1]. In AS3 when
you write an expression that returns a method of an instance, `this`
in the method is a free variable, and the runtime will create a
closure binding `this` to the instance used in computing the method.
(Another equivalent model is that you are creating a partial
evaluation of the method, binding the -1th argument `this` to the
instance.) In Javascript (swf8, DHTML), no method closure is
created. When you evaluate functionVar(args), it is as if you called
the method with `this` unbound. In Javascript, when you call a
function without binding `this`, it will be bound to the global object
[2]. So, you can see that if your method refers to `this` at all, it
will not work as expected.
The way to achieve the same effect in LZX is to say:
/* Helper function we should probably build in to LZX */
function createMethodClosure (instance, methodName) {
return function () { return instance[methodName].apply(arguments); }
}
var functionVar = createMethodClosure(someObject, 'method_name');
functionVar(args);
---
[1]: It would be nearly impossible for the LZX compiler to make this
transformation for you automatically, because Javascript is such a
dynamic language. You really don't know until runtime that the
expression `someObject["method_name"]` evaluates to a method and hence
should create a method closure. In AS3, this is handled by the SWF
runtime.
[2]: The designers of Javascript realize that this decision was a bad
one. It is both functionally less useful, and a security flaw. It
should be improved in EcmaScript 5 (and actually work in the "Harmony"
release!)
