214574 - form.elements['item'] returns a wrong object

Status: RESOLVED INVALID

(Core :: DOM: Core & HTML, defect)

Description

[tom yeh]

User-Agent:       Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.3) Gecko/20030312
Build Identifier: Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.3) Gecko/20030312

If a form contains an input element called 'item', the following statement
returns the item function (I guess it is document.all.item) rather than the element.

document.forms['myForm'].elements['item']

Reproducible: Always

Steps to Reproduce:
1.
2.
3.
Actual Results:  
The item function.

Expected Results:  
The element called item

Comment 1

[Boris Zbarsky [:bzbarsky]]

Over to DOM0; should the 'item' object really clobber the item() property of the
form.elements nodelist?

Comment 2

[Erik Fabert]

Attachment: testcase

document.forms[0].elements["item"] returns the function while
document.forms[0].elements.namedItem("item") returns the element.
DOM HTML Appendix D states: "This object can also be dereferenced using square
bracket notation (e.g. obj["foo"]). Dereferencing using a string index is
equivalent to invoking the namedItem function with that index.", so I'd say
Mozilla's behaviour is wrong.

Comment 3

[Boris Zbarsky [:bzbarsky]]

Erik, the square bracket notation returns the function because that's how
ECMAScript works (all properties of the object can be accessed via either square
bracket notation or '.' notation).  That happens before the DOM code is ever called.

The real problem here is that whoever wrote the ECMAScript bindings for DOM HTML
didn't stop to consider the fact that they were setting up two requirements that
could not both be satisfied at the same time.

Comment 4

[Erik Fabert]

Well, I understand it is impossible to meet both requirements, so maybe the real 
question is, is it acceptable to be incompatible with the other browsers? IE and 
Opera both return the element "item", not the function (Same is true for 
"namedItem", btw.). 

Since these methods are rather useless in JS anyway, shouldn't they be hidden by 
elements of the same name just like any other form property? I know the form and 
form.elements are two very different objects, but from a scripting point of view 
it's strange that I can hide the whole elements collection with <input 
name="elements"> while the elements' methods are protected.  

Comment 5

[Boris Zbarsky [:bzbarsky]]

> so maybe the real question is, is it acceptable to be incompatible with the
> other browsers

That's what this bug is about, yes.

Doing that involves adding a bunch of hacky code to the elements collection of
the sort that we had for the form element.

Comment 6

[Gervase Markham [:gerv]]

This is an automated message, with ID "auto-resolve01".

This bug has had no comments for a long time. Statistically, we have found that
bug reports that have not been confirmed by a second user after three months are
highly unlikely to be the source of a fix to the code.

While your input is very important to us, our resources are limited and so we
are asking for your help in focussing our efforts. If you can still reproduce
this problem in the latest version of the product (see below for how to obtain a
copy) or, for feature requests, if it's not present in the latest version and
you still believe we should implement it, please visit the URL of this bug
(given at the top of this mail) and add a comment to that effect, giving more
reproduction information if you have it.

If it is not a problem any longer, you need take no action. If this bug is not
changed in any way in the next two weeks, it will be automatically resolved.
Thank you for your help in this matter.

The latest beta releases can be obtained from:
Firefox:     http://www.mozilla.org/projects/firefox/
Thunderbird: http://www.mozilla.org/products/thunderbird/releases/1.5beta1.html
Seamonkey:   http://www.mozilla.org/projects/seamonkey/

Comment 7

[Hixie (not reading bugmail)]

Confirming to keep on the radar. bz, do we want to WONTFIX? What interfaces do
we do this to? What should I say in HTML5?

Comment 8

[Boris Zbarsky [:bzbarsky]]

So...  We're not so consistent about this.  In particular:

1)  On window or document or form objects, the DOM0 named item lookup overrides
    IDL-defined properties.
2)  On HTMLCollection objects, the DOM ["name"] lookup is overridden by
    IDL-defined properties.

Note that IE somehow guesses which object you want, or something; that is, it'll
return the element object for nodelist["item"] in this case, and for
nodelist.item, but nodelist.item() will call the right function...

If we decide to change this behavior to return the element, it's just a matter
of implementing an appropriate NewResolve hook for nsNamedArraySH.

I bet if we decide we _really_ want to duplicate what IE does we might be able
to do that by having a Call hook on the elements we return... The question is
how much we want to duplicate it.  And if we don't want to duplicate it fully,
the question is what breaks more sites -- our current behavior or returning the
element for ["item"].

Oh, if only the DOM WG had written a spec that could be implemented here... :(

Comment 9

[Hixie (not reading bugmail)]

Well,

   var m = nodelist.namedItem;
   var a = m('a');

I don't think we want to go down _that_ road.

So at the moment we're doing special lookups only on Window, Document and
HTMLFormElement? Interesting. I'd suggest we stick to that and I'll make the
HTML5 spec say that's what we should be doing.

Comment 10

[Boris Zbarsky [:bzbarsky]]

> So at the moment we're doing special lookups only on Window, Document and
> HTMLFormElement?

Lookups that can override DOM props, yes.  Note that I think your example
doesn't work anyway because it needs the right |this|... and using call() would
work right if we had a call hook.

Comment 11

[Hixie (not reading bugmail)]

Here's what I propose to do. Make Window, Document, and HTMLFormElement have [[Get]] implementations that override what the IDL says. Make the others have a [[Get]] implementation that defers to the IDL first. Make all of them have no special magic with respect to [[Call]]. Any objections?

Comment 12

[lists.bugzilla]

Is there a reason you proposed to go with the inconsistent approach? I would prefer making all the [[Get]] implementations be consistent, although I don't care much whether they override the IDL or not. Making them inconsistent is going to be confusing for both implementors and users.

The only advantage I can see to having Window override the IDL is to guard against future extensions to Window that conflict with named items on already-existing pages. (i.e. some spec defines window.cnnAd, and cnnAd is a named item on cnn.com causing it to break on a UA that implements the spec). However, the same is much less likely for Document and HTMLFormElement, so I think those at least should be changed to defer to the IDL.

Comment 13

[Boris Zbarsky [:bzbarsky]]

> Is there a reason you proposed to go with the inconsistent approach?

It's what browsers do right now.

Comment 14

[lists.bugzilla]

It's what *some* browsers do right now. In HTMLCollection, Opera overrides the IDL for functions but not properties. (i.e. document.forms.namedItem could return a HTMLFormElement instead of a function). document.cookie can be overriden by a <form name="cookie"> in Opera and Safari. Browsers generally seem to be all over the place on this one.

Comment 15

[Hixie (not reading bugmail)]

I ended up just deferring to WebIDL on this.

Comment 16

[:Ms2ger (he/him; ⌚ UTC+1/+2)]

If I read WebIDL (<http://dev.w3.org/2006/webapi/WebIDL/#getownproperty>) correctly, our behaviour is correct.