Subj : Re: updated jsgen (maybe rc1?) - beta 7
To   : Marcello Bastea-Forte <marcello@cellosoft.com>
From : Brendan Eich <brendan@meer.net>
Date : Sun Jan 04 2004 10:46 am

Marcello Bastea-Forte wrote:

>> Also, you're not worrying about thread safety -- do you plan to 
>> handle multiple threads some day?  Threads are the main motivation 
>> for contexts; a single-threaded embedding might use more than one 
>> context (in fact, Mozilla's DOM does), but strictly speaking it need 
>> not, because there are no thread-safety hazards requiring more than 
>> one context.
>
>
> I'm not sure how I should be worrying about that.


Avoid mutable statics -- looks like you've done that (sorry I haven't 
had time to read your source).

>
>> But to get back to your question: there's no need to call 
>> JS_InitClass(cx, obj, ...) more than once per obj.
>
>
> I know that there's no need, but it's more of a matter that my code 
> *might* call initclass multiple times.  In beta 7, I'm now using the 
> following code:
>
> JSObject *JSTest::JSInit(JSContext *cx, JSObject *obj) {
>     if (obj==NULL)
>         obj = JS_GetGlobalObject(cx);
>     jsval oldobj;
>     if (JS_TRUE == JS_GetProperty(cx, obj, "JSTest", &oldobj) && 
> !JSVAL_IS_VOID(oldobj))
>         return JSVAL_TO_OBJECT(oldobj);
>     return JS_InitClass(cx, obj, NULL, &JSTest::_jsClass,
>                                          JSTest::JSConstructor, 0,
>                                          JSTest::_JSPropertySpec, 
> JSTest::_JSFunctionSpec,
>                                          
> JSTest::_JSPropertySpec_static, JSTest::_JSFunctionSpec_static);


This is quite like what XBL does.  One small improvement: use 
JS_LookupProperty instead of JS_GetProperty if your global object 
class's getter does any computation -- Lookup retrieves the slot-cached 
value of the property without calling the getter.  Another nit: use 
JSTest::_jsClass.name instead of repeating "JSTest".

Comments on JS_GetGlobalObject below.

>
>> Explicitly parameterize?  The JS API has many entry points whose 
>> parameter lists begin (JSContext *cx, JSObject *obj...) and obj is 
>> the parameter independent of cx telling in which object to define a 
>> class or property, e.g.
>
>
> I'm also using JS_GetGlobalObject to get around not having a JSObject 
> * all the time.  I should probably add JSObject *obj as an optional 
> parameter on getJSObject(JSContext *cx).


That's neither fish nor foul: better would be to avoid any dependence on 
JS_GetGlobalObject, or use only it, and insist that there be a 1:1 
relationship between JSContext and global object.  Mozilla's DOM does 
the latter, but it is heavier-weight than the former.

/be

.