Anton,
I have run a few experiments using an ObjectTemplate in place of a
FunctionTemplate, and it looks like it falls short of a number of
ways:
I have an Object with a CallAsFunctionHandler named 'Foo' in this
scope that just returns a plain object.
Foo.prototype = {bar: 'baz'}
var f = new Foo()
print("foo: ", Foo) //=> [object Object], should be function Foo() {}.
print("typeof Foo: ", typeof Foo) //=> 'object', should be 'function'
print("f.constructor == Foo: ", f.constructor == Foo) //=> false,
should be true
print("f.bar: ", f.bar) //=> undefined, should be 'baz'
print("f instanceof Foo", f instanceof Foo) //=> TypeError: Expecting
a function in instanceof check, but got #<an Object>
you can see the full program here:
http://gist.github.com/527745
As you can see SetCallAsFunctionHandler() doesn't really do all that
well as a constructor.
cheers,
Charles
On Aug 10, 3:05 pm, Charles Lowell <[email protected]> wrote:
> It might, provided typeof is 'function' and the constructor property
> is setup correctly. The other problem is that I currently invoke these
> constructor functions from C++ via NewInstance() and I can't find the
> equivalent of CallAsFunction() or NewInstanceAsFunction() on Object.
>
> I may be able to find a way around this by wiring up the ruby and js
> peers without calling constructors from C++, but IIRC that was a very
> clean solution to make sure all the peered objects, whether
> constructed from javascript or injected directly into the context went
> through the same coupling code. I'll give it a try and let you know
> what I find.
>
> cheers,
> Charles
>
> On Aug 10, 2:31 pm, Anton Muhin <[email protected]> wrote:
>
>
>
> > I don't know for sure, but won't SetCallAsFunctionHandler help you?
>
> > yours,
> > anton.
>
> > On Tue, Aug 10, 2010 at 11:11 PM, Charles Lowell
>
> > <[email protected]> wrote:
>
> > > On Aug 10, 8:09 am, Anton Muhin <[email protected]> wrote:
> > >> Sorry, wrong address again.
>
> > >> On Tue, Aug 10, 2010 at 5:06 PM, Anton Muhin <[email protected]> wrote:
> > >> > Charles,
>
> > >> > I probably miss something here, but why you need FunctionTemplates at
> > >> > all, could not you use only ObjectTemplates?
>
> > > How do I implement constructor functions with ObjectTemplate ?
>
> > >> > yours,
> > >> > anton.
>
> > >> > On Mon, Aug 9, 2010 at 10:58 PM, Charles Lowell
> > >> > <[email protected]> wrote:
> > >> >> Anton,
>
> > >> >> Thanks for the reply. Perhaps it will make it clearer if I provide
> > >> >> some context. I'm asking in the context of The Ruby Racer, which
> > >> >> allows for embedding V8 into Ruby, and in particular what ruby classes
> > >> >> look like from javascript.
>
> > >> >> Currently, we model ruby instances reflected into javascript as
> > >> >> instances created off of a FunctionTemplate which corresponds to that
> > >> >> instance's class. The FunctionTemplate corresponding to the ruby class
> > >> >> has property interceptors which it sets via
> > >> >> FunctionTemplate::InstanceTemplate()->Set{Named/Indexed}
> > >> >> PropertyHandler.
>
> > >> >> The problem comes when we actually embed the ruby classes themselves
> > >> >> into javascript. This is done by having the class's FunctionTemplate's
> > >> >> CallHandler invoke the ruby constructor of the class, and then
> > >> >> associat it with a javascript instance created by calling
> > >> >> GetFunction().NewInstance(). That way, the constructor can be
> > >> >> (optionally) called directly from javascript, and the object's
> > >> >> 'constructor' will point to the right place.
>
> > >> >> However, ruby classes (like javascript constructors) are objects too
> > >> >> with their own properties and methods and so I would like to be able
> > >> >> to intercept access to properties against the constructor function
> > >> >> itself in the same way that I intercept property access on objects. In
> > >> >> one sentence: I want to delegate Function object properties to a ruby
> > >> >> object in the same way that I delegate regular object properties to a
> > >> >> ruby object.
>
> > >> >> Perhaps a little code would shed some light.
>
> > >> >> class SomeClass
> > >> >> #instance method
> > >> >> def one
> > >> >> 1
> > >> >> end
> > >> >> end
>
> > >> >> #a class method
> > >> >> def SomeClass.two
> > >> >> 2
> > >> >> end
>
> > >> >> cxt = V8::Context.new
> > >> >> cxt['SomeClass'] = SomeClass
> > >> >> cxt.eval(<<--JS)
> > >> >> var s = new SomeClass()
> > >> >> s.one //=> 1
> > >> >> s.constructor == SomeClass //=> true
> > >> >> SomeClass.two //=> currently undefined, I would like this to invoke
> > >> >> SomeClass.two()
> > >> >> JS
>
> > >> >> It could be that there is an obvious solution which I've missed, but
> > >> >> does that help frame the question?
>
> > >> >> cheers,
> > >> >> Charles
>
> > >> >> On Aug 9, 11:18 am, Anton Muhin <[email protected]> wrote:
> > >> >>> Sorry, from the right address.
>
> > >> >>> On Mon, Aug 9, 2010 at 8:16 PM, Anton Muhin <[email protected]>
> > >> >>> wrote:
> > >> >>> > Charles, good day,
>
> > >> >>> > On Mon, Aug 9, 2010 at 7:56 PM, Charles Lowell
> > >> >>> > <[email protected]> wrote:
> > >> >>> >> While we're on the subject of property interceptors and API
> > >> >>> >> asymmetries, I can't figure out to add named and indexed getter/
> > >> >>> >> setter, etc.. to Function objects inside the javascript runtime.
> > >> >>> >> I've
> > >> >>> >> searched extensively for a way to do this, but nothing as
> > >> >>> >> reliable as
> > >> >>> >> ObjectTemplate::SetNamedPropertyHandler()
>
> > >> >>> > I don't quite understand what do you mean by 'inside the javascript
> > >> >>> > runtime'? Do you want to invoke JS function? Or add native
> > >> >>> > interceptor to JS function (to be invoked as a constructor)?
>
> > >> >>> > AFAIK currently one can only add interceptors (both named and
> > >> >>> > indexed)
> > >> >>> > to FunctionTemplates.
>
> > >> >>> > BTW, what are you trying to achieve? If you could tell us, maybe
> > >> >>> > there could be found some way out.
>
> > >> >>> > yours,
> > >> >>> > anton.
>
> > >> >>> >> On Aug 6, 12:49 pm, Anton Muhin <[email protected]> wrote:
> > >> >>> >>> From the right address.
>
> > >> >>> >>> On Fri, Aug 6, 2010 at 9:39 PM, Anton Muhin <[email protected]>
> > >> >>> >>> wrote:
> > >> >>> >>> > Charles,
>
> > >> >>> >>> > I just didn't have to time to implement the same functionality
> > >> >>> >>> > for
> > >> >>> >>> > indexed interceptor. Filing a bug and cc'ing you.
>
> > >> >>> >>> > I'll try to implement it, but it might require some time to
> > >> >>> >>> > roll in to
> > >> >>> >>> > make transition to new api smooth.
>
> > >> >>> >>> > yours,
> > >> >>> >>> > anton.
>
> > >> >>> >>> > On Fri, Aug 6, 2010 at 9:31 PM, Charles Lowell
> > >> >>> >>> > <[email protected]> wrote:
> > >> >>> >>> >> Hi,
>
> > >> >>> >>> >> I've noticed a small asymmetry in the interfaces for
> > >> >>> >>> >> IndexedPropertyQuery and NamedPropertyQuery. Specifically
> > >> >>> >>> >> that the
> > >> >>> >>> >> indexed version returns just a Boolean instead of and Integer
> > >> >>> >>> >> encoding
> > >> >>> >>> >> property attributes. Is this intentional? Rhino, for example
> > >> >>> >>> >> allows
> > >> >>> >>> >> for specifying attributes for indexed values as well.
>
> > >> >>> >>> >> cheers,
> > >> >>> >>> >> Charles
>
> > >> >>> >>> >> --
> > >> >>> >>> >> v8-users mailing list
> > >> >>> >>> >> [email protected]
> > >> >>> >>> >>http://groups.google.com/group/v8-users
>
> > >> >>> >> --
> > >> >>> >> v8-users mailing list
> > >> >>> >> [email protected]
> > >> >>> >>http://groups.google.com/group/v8-users
>
> > >> >> --
> > >> >> v8-users mailing list
> > >> >> [email protected]
> > >> >>http://groups.google.com/group/v8-users
>
> > > --
> > > v8-users mailing list
> > > [email protected]
> > >http://groups.google.com/group/v8-users
--
v8-users mailing list
[email protected]
http://groups.google.com/group/v8-users
