I'm referring to the hidden classes that is talked about here:
http://code.google.com/apis/v8/design.html#prop_access

So, are the hidden classes mentioned there actually the object
Boilerplates, the __proto__, or something else?  Is there perhaps some
more detailed documentation about the design of v8 internals and data
structures (particularly about object layout) that I can read to
better understand this issue?

Thanks.

On Nov 7, 9:27 am, Feng Qian <[EMAIL PROTECTED]> wrote:
> Do you mean 'hidden prototype object'? It is like a normal object
> except that not accessible by JS code '__proto__'.
>
> I am not sure if object cloning can help much because you need to fix
> whole prototype chain for a cloned object. V8 is doing something like
> that for object literals (Boilerplate).
>
> On Thu, Nov 6, 2008 at 10:11 PM, [EMAIL PROTECTED] <[EMAIL PROTECTED]> wrote:
>
> > I've got a few questions regarding the representation of hidden
> > classes:
>
> > 1. Which v8 data structure is the hidden class that the v8 literature
> > talks about?
> > 2. Are hidden classes supposed to be immutable?  Or are they mutable
> > like any JS Objects?
> > 3. If immutable, is it possible to share the hidden classes across
> > execution contexts?
> > 4. If mutable and not shareable across execution contexts, are there
> > any data structures that make up the object that is static meta-data
> > that can be shared?
>
> > The reason for my asking is because I'm wondering if it is possible to
> > use object cloning to speed up the process of populating a new context
> > with objects.  I'm looking for a mechanism (either an existing one or
> > to write one myself) that is similar to snapshots.
>
> > The difference is that the snapshot mechanism seems to serialize the
> > entire heap into a data stream that is later reloaded in entirety into
> > a new launch of v8.  What I want to do instead is to only clone one
> > set of objects (as defined by a reference execution context) into a
> > new context.  Serializing/deserializing this set of objects can be
> > done as an advanced exercise later.
>
> > I'm also hoping that the objects in the new context can reuse all the
> > hidden classes from the reference context.  This is why I'm asking if
> > the hidden classes (which I presume are just collections of metadata)
> > are immutable.
>
> > Please advise.  Thanks.
--~--~---------~--~----~------------~-------~--~----~
v8-dev mailing list
[email protected]
http://groups.google.com/group/v8-dev
-~----------~----~----~----~------~----~------~--~---

Reply via email to