386388 - A for-in/for-each-in statement iterates over XML children

Status: RESOLVED WONTFIX

(Core :: JavaScript Engine, defect)

Description

[nanto_vi (TOYAMA Nao)]

ECMA-357 2nd 12.2/12.3 say, "When the Expression evaluates to an XML object, the for-in statement converts the XML object to an XMLList and iterates over the resulting XMLList."  But SpiderMonkey doesn't convert an XML to an XMLList.  Instead, it iterates over children of the XML object.

Actual:
js> var x = <a><b/><c/><d/></a>;
js> for (var i in x) print(i, x[i]);
0 <a>
  <b/>
  <c/>
  <d/>
</a>
1 undefined
2 undefined

Expected:
js> for (var i in x) print(i, x[i]);
0 <a>
  <b/>
  <c/>
  <d/>
</a>

Comment 1

[Jeff Walden [:Waldo]]

What's Rhino do here?

The spec mentions the specified behavior "blurs the distinction", but it's blurring the wrong way -- what should have happened is that if it's XML, you enumerate kids, if it's XMLList and length != 1, you enumerated members, and if it's XMLList and length == 1, you enumerate kids of the single contained XML item.

I think this is a spec bug.

Comment 2

[nanto_vi (TOYAMA Nao)]

Rhino 1.6 release 5 2006 11 18
js> var x = <a><b/><c/><d/></a>;
js> for (var i in x) print(i, x[i]);
0 <a>
  <b/>
  <c/>
  <d/>
</a>

I think the spec and the behavior of Rhino is proper.  If you enumerate kids for an XML and an XMLList whose length is 1, |for (var i in x) doSomethingWith(x[i]);| and |for each (var e in x) doSomethingWith(e);| come to have different meanings -- x[i] can be undefined in the former while e is always an XML in the latter.  It goes against the E4X design principle "consistent."  To enumerate kids, you should use |for (var i in x.*)|.

Comment 3

[Jason Orendorff [:jorendorff]]

(In reply to comment #1)

This isn't a spec bug.  You might say it's an annoying design decision by the E4X people, but even so, we should follow the spec.

Comment 5

[Biju]

Attachment: foreach_i_in_e4x.html

(from Bug 389129)
> if "x = <a> <b>x</b> <b>x</b> </a>;"
> then "for(j in x)" gives "0" and "1" as value of "j"
> 
> but "1 in x" is "false" and "x[1]" is "undefined"

That is an inconsistency.
And should be avoided.

Comment 6

[Biju]

Attachment: for_i_in_e4x.html

Comment 7

[Brendan Eich [:brendan]]

This will be easy to fix or indeed play around with via self-hosted E4X runtime objects implemented by "harmony proxies".

/be

Comment 8

[:Benjamin Peterson]

E4X will be leaving us soon.