How about creating a "helper" to handle recursion.
"Object a" on delete sends a call to "helper". Says, "Hey, I'm fixing to
off myself!" Can you take care of a little referencial trash when I'm
gone?"
then "helper" finishes the reference clipping on behalf of deceased "object
a".
----- Original Message -----
From: "Robert Rainwater" <[EMAIL PROTECTED]>
To: "Raymond Smith" <[EMAIL PROTECTED]>
Sent: Sunday, March 11, 2001 9:07 PM
Subject: Re: [Dynapi-Dev] Garbage Collection
>
> This is where Mozilla messes everything up. Try doing any type of
> recursiving deleting in Mozilla and it will hack up since it follows
> all pointers to objects and since we use recursive pointers like:
> dlyr.doc.lyrobj = dlyr it goes crazy. I would love to see a solution
> for Mozilla.
>
> --
> Robert Rainwater
>
>
> On 3/11/2001, 11:24:23 PM EST, Raymond wrote about "[Dynapi-Dev] Garbage
Collection":
>
> > I've been looking at Bob's code in more detail and the insight is
revealing
> > when compared to the base constructs of the DynAPI. The first thing I
> > noticed is his use of interfaces, inner classes and upcasting (very
Java,
> > C++ ish) in the creation of Objects. This allows a little distance
between
> > the object constructor and it's parent.
>
> >> http://www.mindspring.com/~bobclary/base/docs/xbObject.html site of
> > xbObject.
>
> > Then he leverages this by explicitly creating object deconstruction that
is
> > callable and reverse chains up to it's parent to allow full nulling of
the
> > object to allow for garbage collection. In other words, he created a
very
> > clever form of garbage collection within the class constructors
themselves.
>
> > With the DynAPI we tend to be a level flatter and without the ability to
> > explicitly call "object destroy" since we assume that the JS GC will
> > naturally clean up our mess, probably a mistake on our part. Also,
since we
> > are a level shorter in hierarchical height we also are having problems
> > creating a natural "object clip point" to allow us to effectively "throw
the
> > used object away" and have the GC actually pick it up.
>
> > This "added object tier" structure had been tickling the back of my
brain
> > lately as I consumed mass quantities of Java. The question now is how
easy
> > will it be for us to reverse engineer explicit garbage collection (on a
> > clipable reference object level) into the current DynAPI as this will
> > probably solve the current GC issues.
>
> > Laters
>
> > Ray
>
>
>
> > _______________________________________________
> > Dynapi-Dev mailing list
> > [EMAIL PROTECTED]
> > http://lists.sourceforge.net/lists/listinfo/dynapi-dev
>
>
>
> _______________________________________________
> Dynapi-Dev mailing list
> [EMAIL PROTECTED]
> http://lists.sourceforge.net/lists/listinfo/dynapi-dev
>
_______________________________________________
Dynapi-Dev mailing list
[EMAIL PROTECTED]
http://lists.sourceforge.net/lists/listinfo/dynapi-dev
