Re: Interaction of dictionaries and platform objects is a bit weird

A somewhat related issue is that there's no way for a developer to know
whether the object being passed in is a dictionary or an object that is
actually held onto.

var obj = {
    handleEvent: function() { alert('noclick') }
};
document.body.addEventListener('click', obj);
obj.handleEvent = function() { alert('click') };
// Alerts 'click' when clicking on the body.

var obj = {
    bubbles: true
}
var myEvent = new Event('customevent', obj);
obj.bubbles = false;
// myEvent.bubbles == true

This is kind of gross. I'm not really sure what we can do to make it better.

On Wed, Oct 17, 2012 at 7:08 PM, Boris Zbarsky <bzbarsky@mit.edu> wrote:

> Consider the following IDL:
>
>   dictionary Dict {
>     long a;
>   };
>   [Constructor()]
>   interface Iface {
>     void foo(long x, optional Dict y);
>     void bar(optional Dict z);
>     void bar(Element w);
>   };
>
> and this script, running in a web page:
>
>   var x = new Iface();
>   x.foo(5, document);
>   x.bar(document);
>
> What happens?
>
> As far as I can tell, the foo() call succeeds.  The "a" property of the
> document, if any, is placed in the dictionary, and then the number 5 and
> the dictionary are passed to the WebIDL method implementation.
>
> Also as far as I can tell the call bar() call throws.  This is because we
> enter the overload resolution algorithm, set argcount to 1, throw our the
> 0-argument overload, set d to 0, examine arg 0, determine that it does not
> satisfy either overload (because document does not implement Element and
> because document is a platform object, so never enters step 13.6 of the
> overload resolution algorithm), and a TypeError is thrown.
>
> This seems a bit odd.  I think we should either always throw a TypeError
> when converting platform objects to dictionaries or allow any object to
> enter step 13.6 of overload resolution... at least for dictionaries; not
> sure about callbacks.  Maybe dictionaries need to be split out into a
> separate step here?
>
> -Boris
>
>

Received on Tuesday, 23 October 2012 01:19:23 UTC