[Webkit-unassigned] [Bug 64572] Initial implementation of calling WebScriptObject::callWebScriptMethod() via COM

bugzilla-daemon at webkit.org bugzilla-daemon at webkit.org
Thu Jul 14 21:24:22 PDT 2011


--- Comment #3 from Anthony Johnson <anthony.johnson at flexsim.com>  2011-07-14 21:24:22 PST ---
Overall design summary:
1. I created static methods on WebScriptObject, primarily for sub-classes to access.

a. The main method that's doing most of the work is WebScriptObject::callWebScriptMethodWithJSDOMWrapper(). This method is primarily a port from Objective-C at Source/WebCore/bindings/objc/WebScriptObject.mm. I found that most c++ interaction requires a JSDOMWrapper around a DOM element, so I created the method on WebScriptObject, and subclasses then call that method, passing their associated wrappers to the method.

b. Secondarily, I created two static methods on WebScriptObject:  convertVariantToValue() and convertValueToVariant(), for translating from VARIANT parameters to Javascript values, and back. At this point, these methods are only partially filled out as I want to get feedback on the overall design before I go ahead with the rest of it.

2. I've implemented callWebScriptMethod() methods on the DOMElement, and the DOMNode. Each of these essentially gets its wrapper, and then calls the WebScriptObject static method.

3. I also filled out DOMHTMLElement::innerHTML() and DOMHTMLElement::setInnerHTML(), as I use these in my tests. These are primarily just translated from DOMHTMLElement::innerText() and DOMHTMLElement::setInnerText().

4. Regarding translation from VARIANT to JSValue, most of the primitive types are pretty simple, but I want feedback on my decision for array types. I would like to give the user a "fastest possible" option so that the user can pass large amounts of data between C++/COM and javascript with the least amount of overhead. Ideally this would simply be passing a pointer to a memory buffer, and javascript would somehow properly create the right wrappers around that buffer to make it usable. However, as far as I can tell, this isn't feasible. So, the next fastest option would be to have just one memory copy of the entire buffer. This is what I've implemented. For a VARIANT array of primitive types, I create a javascript ArrayBufferView, which essentially effects a one-time copy of the VARIANT's buffer. I want to create type-specific ArrayBufferView's, i.e. for a VARIANT array of int's I would create an Int32Array. However, I haven't yet figured out how to get these properly translat
 ed to javascript. I can create the Int32Array in c++, but when I look at it in javascript, it just looks like an ArrayBufferView, i.e. the length attribute as well as getters and setters are not present (any help here would be greatly appreciated). For now, I think it's fine as it is since it will be forward-compatible once we get that piece working, and the I (the user) can do work-around's in javascript until we get to that point. 

I later plan on translating a VARIANT array of VARIANT's into a javascript array type (yet to be implemented).

Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

More information about the webkit-unassigned mailing list