Subj : Re: E4X: calling appendChild method with existing child as the argument
To   : Martin.Honnen
From : Brendan Eich
Date : Fri Apr 08 2005 03:51 pm

Martin Honnen wrote:
> 
> In E4X XML objects have a method called appendChild which behaves 
> differently in regardance to object identity when the method is called 
> with an existing child as the argument:
> 
> Spidermonkey:
> 
> js> var gods = <gods><god>Kibo</god><god>Xibo</god></gods>;
> js> gods
> <gods>
>   <god>Kibo</god>
>   <god>Xibo</god>
> </gods>
> js> gods.appendChild(gods.god[0])
> <gods>
>   <god>Kibo</god>
>   <god>Xibo</god>
>   <god>Kibo</god>
> </gods>
> js> gods.god[0] === gods.god[2]
> true

SpiderMonkey is correct if you follow ECMA-357 9.2.1.2 XMLList [[Put]], 
in particular 2(g)(ii)(2).

I say file a Rhino bug, although 9.2.1.2 is a confusingly complicated 
part of the spec, and in particular, it wastes cycles and space making a 
new element node to insert as gods.god[2], then replaces it in 2(g) with 
a reference to gods.god[0].  Bcc'ing some folks who may have an opinion 
on how the spec might be improved here.

/be


> 
> so here after the appendChild the XML objects has two identical child 
> elements at different positions.
> 
> Compare to Rhino 1.6 release 1 2004 11 30
> js> var gods = <gods><god>Kibo</god><god>Xibo</god></gods>;
> js> gods
> <gods>
>   <god>Kibo</god>
>   <god>Xibo</god>
> </gods>
> js> gods.appendChild(gods.god[0])
> <gods>
>   <god>Kibo</god>
>   <god>Xibo</god>
>   <god>Kibo</god>
> </gods>
> js> gods.god[0] === gods.god[2]
> false
> 
> Which behaviour is right, Spidermonkey's or Rhino's?
>

.