Subj : Re: JSClass callbacks
To   : netscape.public.mozilla.jseng
From : Jamie Nordmeyer
Date : Fri Mar 25 2005 11:21 am

Oh, OK.  So for example, if I wanted to create my own version of the 'new
ActiveXObject' call used in Internet Explorer's implementation of Javascript
to attach to a late-bound COM object, the Resolve hook is where I'd do this
then.  For example:

var obj = new ActiveXObject("SomeObject.SomeClass");
obj.someLateBoundFunc();

I get it now.  Cool.  Thanks a ton, Brenden!

"Brendan Eich" <brendan@meer.net> wrote in message
news:4244613F.8010708@meer.net...
> Jamie Nordmeyer wrote:
> > Is there a good internet resource, or can somebody provide information,
on
> > when and why to use the Resolve hook?  In the simplest example I've
seen,
> > the resolve hook is simply JS_ResolveStandardClass.  In more complex
> > examples, it's actually defining new properties for an object.
>
>
> JS_ResolveStandardClass defines new properties too, so there's no
> difference in principle between those cases.  You write "simply" and
> "more complex", and you're right that if you have to define the id
> yourself, the API call is a little more complicated than the API call to
> JS_ResolveStandardClass.
>
> But the idea's the same: when resolve is called, id does not name a
> property directly contained in obj (it might name a property in a
> prototype of obj; we don't know yet, and resolve could check if it needs
> to know).  If the object's class "knows" about id somehow -- it's on a
> list of well-known properties that should be resolved lazily, not
> defined eagerly when obj was created, e.g. -- then resolve should define
> a property for id.
>
> The most common application is when reflecting properties in obj from a
> "peer" object system implemented in C, C++, Java, etc., where the bridge
> between JS and its peer must not eagerly reflect all objects and
> properties for obvious performance reasons, but must reflect those (obj,
> id) pairs demanded by running scripts.
>
> /be

.