Re: [webkit-dev] about hashmap staticDOMObjects

2009-04-23 Thread Maciej Stachowiak 


Re-adding list (I assume it was omitted by mistake).

On Apr 22, 2009, at 11:12 PM, ZHOU Xiao-bo wrote:


thanks, but I have another question:

Is every DOM wrapper that is inserted into the hashtable been  
protected, I mean whether they are added in

the protectset?
If yes, when are they inserted, if not, I think they may become wild  
pointer after Collector::collect is called.


Objects in this hashtable are not protected from garbage collection.  
Instead, all DOM object wrappers that are eligible to be in this hash  
table will automatically remove themselves when collected. This  
ensures that the hashtable does not contain bad pointers after a GC.


 - Maciej




2009/4/23 Maciej Stachowiak m...@apple.com

On Apr 22, 2009, at 12:27 AM, ZHOU Xiao-bo wrote:

hi all:
I still don't understand the purpose of the HashMap:

static DOMObjectMap domObjects()
{
   // Don't use malloc here. Calling malloc from a mark function can  
deadlock.

   static DOMObjectMap staticDOMObjects;
   return staticDOMObjects;
}

what kind of DOMObjects should be stored in it? And why?
I searched the source codes and I found that these classes below  
will do that:


Document, Event, HTMLCollection, XMLHttpRequest, CanvasGradient,  
CanvasPattern, CanvasRenderingContext2D, DOMCoreException,  
DOMImplementation, DOMParser, EventException, History, NamedNodeMap,  
NodeFilter, NodeIterator, NodeList, Range, RangeException,  
TreeWalker, XMLHttpRequestException, XMLSerializer, Clipboard


but what's the reason? Is it because these classes are essentially  
simple and just acting as a tool?


The hash map is to ensure that DOM wrappers are unique but created  
on demand, without consuming a pointer in the underlying C++ object.  
For DOM Nodes, we use a hash table in the Node's owner document, but  
in the case of non-Node objects and the Document itself (which has  
no owner document), we need this global hash table.


Regards,
Maciej




___
webkit-dev mailing list
webkit-dev@lists.webkit.org
http://lists.webkit.org/mailman/listinfo.cgi/webkit-dev

[webkit-dev] about hashmap staticDOMObjects

2009-04-22 Thread ZHOU Xiao-bo 
hi all:
 I still don't understand the purpose of the HashMap:

static DOMObjectMap domObjects()
{
// Don't use malloc here. Calling malloc from a mark function can
deadlock.
static DOMObjectMap staticDOMObjects;
return staticDOMObjects;
}

what kind of DOMObjects should be stored in it? And why?
I searched the source codes and I found that these classes below will do
that:

Document, Event, HTMLCollection, XMLHttpRequest, CanvasGradient,
CanvasPattern, CanvasRenderingContext2D, DOMCoreException,
DOMImplementation, DOMParser, EventException, History, NamedNodeMap,
NodeFilter, NodeIterator, NodeList, Range, RangeException, TreeWalker,
XMLHttpRequestException, XMLSerializer, Clipboard

but what's the reason? Is it because these classes are essentially simple
and just acting as a tool?

thanks a lot
___
webkit-dev mailing list
webkit-dev@lists.webkit.org
http://lists.webkit.org/mailman/listinfo.cgi/webkit-dev

Re: [webkit-dev] about hashmap staticDOMObjects

2009-04-22 Thread ZHOU Xiao-bo 
oh, I got an idea:
Is it because these domobjects doesn't belong to a Document, so they can't
be
stored in

static NodePerDocMap domNodesPerDocument()
{
ASSERT(JSLock::lockCount());
static NodePerDocMap staticDOMNodesPerDocument;
return staticDOMNodesPerDocument;
}




2009/4/22 ZHOU Xiao-bo zhxb.u...@gmail.com

 hi all:
  I still don't understand the purpose of the HashMap:

 static DOMObjectMap domObjects()
 {
 // Don't use malloc here. Calling malloc from a mark function can
 deadlock.
 static DOMObjectMap staticDOMObjects;
 return staticDOMObjects;
 }

 what kind of DOMObjects should be stored in it? And why?
 I searched the source codes and I found that these classes below will do
 that:

 Document, Event, HTMLCollection, XMLHttpRequest, CanvasGradient,
 CanvasPattern, CanvasRenderingContext2D, DOMCoreException,
 DOMImplementation, DOMParser, EventException, History, NamedNodeMap,
 NodeFilter, NodeIterator, NodeList, Range, RangeException, TreeWalker,
 XMLHttpRequestException, XMLSerializer, Clipboard

 but what's the reason? Is it because these classes are essentially simple
 and just acting as a tool?

 thanks a lot

___
webkit-dev mailing list
webkit-dev@lists.webkit.org
http://lists.webkit.org/mailman/listinfo.cgi/webkit-dev

Re: [webkit-dev] about hashmap staticDOMObjects

2009-04-22 Thread Maciej Stachowiak 


On Apr 22, 2009, at 12:27 AM, ZHOU Xiao-bo wrote:


hi all:
 I still don't understand the purpose of the HashMap:

static DOMObjectMap domObjects()
{
// Don't use malloc here. Calling malloc from a mark function  
can deadlock.

static DOMObjectMap staticDOMObjects;
return staticDOMObjects;
}

what kind of DOMObjects should be stored in it? And why?
I searched the source codes and I found that these classes below  
will do that:


Document, Event, HTMLCollection, XMLHttpRequest, CanvasGradient,  
CanvasPattern, CanvasRenderingContext2D, DOMCoreException,  
DOMImplementation, DOMParser, EventException, History, NamedNodeMap,  
NodeFilter, NodeIterator, NodeList, Range, RangeException,  
TreeWalker, XMLHttpRequestException, XMLSerializer, Clipboard


but what's the reason? Is it because these classes are essentially  
simple and just acting as a tool?


The hash map is to ensure that DOM wrappers are unique but created on  
demand, without consuming a pointer in the underlying C++ object. For  
DOM Nodes, we use a hash table in the Node's owner document, but in  
the case of non-Node objects and the Document itself (which has no  
owner document), we need this global hash table.


Regards,
Maciej

___
webkit-dev mailing list
webkit-dev@lists.webkit.org
http://lists.webkit.org/mailman/listinfo.cgi/webkit-dev