Subj : Re: spidermonkey: reserved slot vs. private
To   : Jens Thiele
From : Brendan Eich
Date : Fri Sep 17 2004 06:43 pm

Jens Thiele wrote:
> Are JS_SetPrivate / JS_GetPrivate functionally "equal"
> to
> JS_SetReservedSlot / JS_GetReservedSlot
> ?
> 
> to use JS_SetPrivate / JS_GetPrivate one has to specify
> JSCLASS_HAS_PRIVATE in the JSClass structure
> 
> to use JS_SetReservedSlot / JS_GetReservedSlot one has to specify
> JSCLASS_HAS_RESERVED_SLOTS(xxx)


Slots are slot, if that's what you mean by "equal", but 
JSCLASS_HAS_PRIVATE predates JSCLASS_HAS_RESERVED_SLOTS(n) by many 
years, and backward API compatibility is king.


> are reserved slots handled like "normal" properties?
> (they are jsval's, they are garbage collected?, ...)
> but hidden


They're unnamed by any property, scanned by the GC, got and set by 
native code using the API.


> => reserved slots could sometimes be an alternative to private data
> 
> What I want to do:
> i use SetPrivate/GetPrivate to store some pointer to an native object
> (always making sure the class is correct)
> 
> the js object finalizer should only free the native object referenced by
> the pointer if the js object owns this native object
> => i must store that ownership information somewhere
> of course i could wrap the pointer and the ownership information in
> another struct (or store this information somewhere completely
> different) but if i could use the reserved slot to store the ownership
> information this would be much simpler.


You may use a reserved slot for this flag.

Under what conditions would the JSObject *not* own the private data?  In 
those cases, what protects the JSObject from having a dangling (or null, 
if you JS_SetPrivate(cx, obj, NULL)) private?

/be

.