Subj : Re: Enumeration of extended properties of Array and Object prototypes
To   : netscape.public.mozilla.jseng
From : Chris Zumbrunn <chris@deployzone.com>
Date : Sat Mar 27 2004 02:21 am

NNTP-Posting-Host: adsl-212-90-218-6.cybernet.ch
User-Agent: MT-NewsWatcher/3.3b1 (PPC Mac OS X)
Xref: newsmst01.news.prodigy.com netscape.public.mozilla.jseng:8138

Brendan Eich writes:

> Chris Zumbrunn wrote:
> 
> > One real-world example that might serve well to illustrate what I mean 
> > would be getter/setter methods for data conversion of Javascript objects 
> > (for example to and from XML).
> 
> If you mean someone calling Object.prototype.__defineGetter__('id', f), 
> or likewise for a setter, then the property that results in that case is 
> not enumerable in SpiderMonkey.
> 
> This looks to me like a bug introduced when the old getter= form was 
> deprecated and its code copied into new __defineGetter__ and 
> __defineSetter__ methods of Object.prototype, as the getter= code 
> includes JSPROP_ENUMERATE among the attributes.  But, you can take 
> advantage of it for now.

I didn't know about __defineSetter__ and __defineGetter__ at all. I've 
had some trouble getting this to work in the environment where I use 
Spidermonkey - but that's just our embedding that will need to be fixed.

> The problem hinted at by my last posting is that it's hard to change JS 
> now, given all the implementations; harder still, as you point out, to 
> change it in incompatible ways.  Your best bet is to lobby ECMA people, 
> if you can contact them, about getting Edition 4 changed.

Regarding ECMA4/JS2, I just found a list of changes Douglas Crockford 
suggests at http://www.crockford.com/javascript/recommend.html 
and "dontenum myClass.prototype.myMethod;" is on that list - I would 
think he sent these suggestions to the ECMA people. 

> Adding a way to define a non-enumerable property could be done in 
> SpiderMonkey -- is that something you could use well, or would you need 
> standard support to get other implementations to pay attention to the 
> change and implement the extension?

No, I wouldn't need standard support. That's why I originally also asked 
if somebody had already done or attempted a patch of this kind in their 
implementations.

However, I would also need a way of doing this in Rhino. Does Rhino 
provide anything similar to Spidermonkey's __defineSetter__ and 
__defineGetter__?

Could a dontenum statement be added to both Spidermonkey and Rhino, 
even-though it isn't part of the ECMA specs (yet)? That would be ideal!

Chris

.