Strange, the handler isn't called at all. Here is the code of
constructor of example JS object:
---- 8< ----
v8::Handle<v8::Value> jsnative_myobj(const v8::Arguments& args) {
if( args.IsConstructCall() ) {
// Create object template here (well, this is just a test,
performance doesn't matter)
Handle<ObjectTemplate> objTpl = ObjectTemplate::New();
objTpl->SetInternalFieldCount( 1 );
objTpl->SetNamedPropertyHandler(
MyNamedPropertyGetter, MyNamedPropertySetter,
MyNamedPropertyQuery, MyNamedPropertyDeleter,
MyNamedPropertyEnumerator );
// Create new object instance
Local<v8::Object> obj = objTpl->NewInstance();
// Store an application-specific pointer in an internal field
as a persistent handle
Persistent< External > hm( External::New( new
map<string,string>() ) ); // Set handle data
hm.MakeWeak( 0, RemoveMyObj ); // Make the handle weak,
specify the callback
obj->SetInternalField( 0, hm ); // Store handle in internal
field
// Return instance just created, which should cause the
current This object to be discarded.
return obj;
}
else
return v8::Undefined();
}
---- 8< ----
The corresp. function object is created using a function template and
set as a global property:
---- 8< ----
Local<FunctionTemplate> f_myobj =
FunctionTemplate::New(jsnative_myobj);
f_myobj->SetClassName( String::New("MY_OBJ") );
context->Global()->Set( String::New("myobj"), f_myobj-
>GetFunction() );
---- 8< ----
Then I execute the following script:
---- 8< ----
for( i=1; i<1000000; ++i ) {
o = new myobj
if( i % 100 == 0 ) print( i );
o.p1 = "123";
o.p2 = i.toString();
}
---- 8< ----
I expect that GC would remove instances created in a loop, which
causes weak handles in internal fields of my instances to die, which
in turn causes the invocation of the RemoveMyObj() handle.
Unfortunally, this doesn't happen; task manager shows that memory
usage constantly grows. What am I doing wrong?
Regards,
Deadmorous
On 22 фев, 00:54, deadmorous <[email protected]> wrote:
> Yes, exactly; I have smart pointers with reference counting in my app,
> but this doesn't matter.
>
> Actually, as wou warned, my callback hasn't been called in a small
> example, even though the engine should have shut down (I called
> Persistent<Context>::Dispose() for the context, and V8::Dispose()
> afterwards). Still my handler has not been called. Well, if this is
> just becasue GC hasn't worked a single time, I can of course kill all
> of my instances after engine shutdown. Will also try running a more
> memory intensive test script to find out what's going on...
>
> On 21 фев, 23:32, Alfred Rossi <[email protected]> wrote:
>
> > You should hold your object in a weak persistent reference from the
> > start.
>
> > If I understand you correctly, I think what you actually want to do is
> > create a Persistent::External for holding your wrapped class pointer,
> > and call MakeWeak on that. The handler given to MakeWeak should cast the
> > pointer back to it's original type and delete it. The rest about
> > deleting after removal is automatic and happens as deemed necessary by
> > the garbage collector.
>
> > On Sun, 2010-02-21 at 12:28 -0800, deadmorous wrote:
> > > Well, the most important to me is to take some action in responce to
> > > adding/removing properties, which is done using interceptors. Once a
> > > JS property is removed, I can reflect that in my own property tree,
> > > and also call Persistent::MakeWeak() on the object referred to by the
> > > property that's being removed. Then at some time GC would remove JS
> > > instance, which in turn would dereference my instance, which in turn
> > > would typically remove it. At least this is how I see that now - will
> > > try to implement. V8's API is not obvious, though SpiderMonkey's is
> > > harder to understand and use)
>
> > > On 21 фев, 20:44, Alfred Rossi <[email protected]> wrote:
> > > > I think global here is referring to the global HandleScope and not the
> > > > owning object. I suspect that by holding it with a persistent handle
> > > > you globalize (free it's ties to C++ scope dependent handles like
> > > > Local) the object. According to the Embedder's Guide:
>
> > > > "Persistent handles are not held on a stack and are deleted only when
> > > > you specifically remove them. Just like a local handle, a persistent
> > > > handle provides a reference to a heap-allocated object. Use a
> > > > persistent handle when you need to keep a reference to an object for
> > > > more than one function call, or when handle lifetimes do not
> > > > correspond to C++ scopes...."
>
> > > > "...A persistent handle can be made weak, using Persistent::MakeWeak,
> > > > to trigger a callback from the garbage collector when the only
> > > > references to an object are from weak persistent handles."
>
> > > > The comments and documentation leave me with the impression that when
> > > > one object holds another it does not do so using a weak persistent
> > > > handle, and thus the garbage collector is prevented from invoking the
> > > > callback.
>
> > > > This gets into internals which I am not too keen on (I am just a user
> > > > like yourself), but there are developers on this list that can jump in
> > > > should you like to know more.
>
> > > > Best,
> > > > Alfred
>
> > > > On Sun, Feb 21, 2010 at 12:17 PM, deadmorous <[email protected]>
> > > > wrote:
> > > > > Thank you very much, Alfred.
> > > > > This seems to be exactly what I asked for. :) One thing I still don't
> > > > > understand is the description of "object" parameter in
> > > > > WeakReferenceCallback, saying "the weak global object to be reclaimed
> > > > > by the garbage collector". Why "global"? What if the object was a
> > > > > property of some other object, and the property was then removed?
>
> > > > > Best regards,
> > > > > Deadmorous
>
> > > > > On 21 фев, 19:32, Alfred Rossi <[email protected]> wrote:
> > > > >> I apologize, that was supposed to say "The garbage collector is NOT
> > > > >> guaranteed..."
>
> > > > >> Best,
> > > > >> Alfred
>
> > > > >> On Sun, 2010-02-21 at 11:31 -0500, Alfred Rossi wrote:
> > > > >> > You're looking for Persistent::MakeWeak. You can use MakeWeak to
> > > > >> > specify
> > > > >> > a callback to be invoked when the object is about to be garbage
> > > > >> > collected.
>
> > > > >> > You should be careful about building in an alternate way to free
> > > > >> > your C
> > > > >> > ++ objects. The garbage collector is guaranteed to be run, ever,
> > > > >> > even
> > > > >> > after the context has been disposed.
>
> > > > >> > See this for more details:
> > > > >> >http://stackoverflow.com/questions/173366/how-do-you-free-a-wrapped-c...
>
> > > > >> > Best,
> > > > >> > Alfred
>
> > > > >> > On Sun, 2010-02-21 at 08:24 -0800, deadmorous wrote:
> > > > >> > > Hello,
> > > > >> > > I'm trying to use V8 to expose functionality of my app to JS.
> > > > >> > > In my app, there are objects with properties and methods,
> > > > >> > > organized as
> > > > >> > > a tree of properties;
> > > > >> > > object instances can be created and added to the tree.
> > > > >> > > I understand how to expose my properties using V8 interceptors,
> > > > >> > > and
> > > > >> > > think that it's correct to expose my
> > > > >> > > methods as properties of prototypes. Further, I would like
> > > > >> > > objects to
> > > > >> > > be creatable from JS code,
> > > > >> > > so I would expose my native constructors to JS. It happens that
> > > > >> > > I have
> > > > >> > > to have a duality between my instances
> > > > >> > > and JS instances. In my app, I would put a handle to JS instance
> > > > >> > > corresponding to my instance;
> > > > >> > > and vice versa, I can put a pointer to my instance into
> > > > >> > > corresponding
> > > > >> > > JS instance, by using
> > > > >> > > Object::SetInternalField(), and using a value of type External.
>
> > > > >> > > Now to my question. What I really don't understand - how can I
> > > > >> > > know
> > > > >> > > that a JS object is being destroyed?
> > > > >> > > I really need to know that, at least in order to destroy my
> > > > >> > > instance
> > > > >> > > attached to the one being destroyed in JS.
>
> > > > > --
> > > > > 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
