851726 - Implement support for arrays in WebIDL

Status: RESOLVED INVALID

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

Comment 1

[Boris Zbarsky [:bzbarsky]]

Reuben, which arrays do we need here?  Readonly, fixed-length, or variable-length?  Also, what's the inner type?  I'd be interested in seeing the IDL we're looking at here; is there a spec for it?

Note also 850159.

Comment 2

[Reuben Morais [:reuben]]

(In reply to Boris Zbarsky (:bz) from comment #1)
> Reuben, which arrays do we need here?  Readonly, fixed-length, or
> variable-length? Also, what's the inner type?

Variable-length, DOMString, and ContactFields.

> I'd be interested in seeing the IDL we're looking at here;
> is there a spec for it? Note also 850159.

https://mxr.mozilla.org/mozilla-central/source/dom/interfaces/contacts/nsIDOMContactProperties.idl

Comment 3

[Boris Zbarsky [:bzbarsky]]

That's xpidl, not WebIDL.

What I'm trying to understand is whether the right thing is in fact IDL arrays or sequences.  And that depends on what the interface is actually meant to behave like, hence the question about a spec.

Comment 4

[Reuben Morais [:reuben]]

(In reply to Boris Zbarsky (:bz) from comment #3)
> That's xpidl, not WebIDL.
> 
> What I'm trying to understand is whether the right thing is in fact IDL
> arrays or sequences.  And that depends on what the interface is actually
> meant to behave like, hence the question about a spec.

Oh, yea, I meant to also link to https://wiki.mozilla.org/WebAPI/ContactsAPI. It's not a spec but it explains the reasons behind the arrays. I just looked at the DAP Contacts spec and it uses sequences, so maybe I can use that.

Comment 5

[Boris Zbarsky [:bzbarsky]]

Well, here are some litmus tests.  If I have a ContactProperties in myContactProperties, what should the following snippets of JS alert?

1)  alert(myContactProperties.name === myContactProperties.name)

2)  myContactProperties.name.length = 100000;
    alert(myContactProperties.name.length);
    alert(myContactProperties.name[99999]);

3)  myContactProperties.name[0] = "foopy";
    alert(mycontactProperties.name[0]);

4)  var contact = new Contact(myContactProperties);
    alert(contact.name === myContactProperties.name);
    contact.name[0] = "Hey";
    alert(myContactProperties.name[0]);

Comment 6

[Reuben Morais [:reuben]]

1) true

2) 100000
   undefined

3) foopy (ReferenceError: mycontactProperties is not defined :)

4) false
   foopy

Comment 7

[Boris Zbarsky [:bzbarsky]]

OK.  What about:

5)  var contact = new Contact(myContactProperties);
    contact.name[0] = "Hey";
    alert(contact.name[0]);

Comment 8

[Reuben Morais [:reuben]]

(In reply to Boris Zbarsky (:bz) from comment #7)
> OK.  What about:
> 
> 5)  var contact = new Contact(myContactProperties);
>     contact.name[0] = "Hey";
>     alert(contact.name[0]);

"Hey"

Comment 9

[Reuben Morais [:reuben]]

(I'm receiving delayed bugmail every now and then, feel free to ping me on IRC for these questions.)

Comment 10

[Boris Zbarsky [:bzbarsky]]

OK.  So it sounds like we really do want IDL arrays here, writable and variable-length ones, but want to make a copy on construction of the Contact?

Note that the WebIDL behavior for DOMString[] on length being increased is not what you might want here (it'll fill it with "undefined" strings); might be worth raising a WebIDL spec issue on that.

But OK, it sounds like we need to do a bunch of work to make this work.

Comment 12

[Jan-Ivar Bruaroey [:jib] (needinfo? me)]

Would be useful in Bug 882145, but I think we can work around it, so I'm not marking it as dependent.

Our current take on http://dev.w3.org/2011/webrtc/editor/getusermedia.html#dictionary-mediatrackconstraints-members

 typedef object MediaTrackConstraint;

 dictionary MediaTrackConstraints {
     MediaTrackConstraintSet mandatory;
 // Bug 882145
 //    MediaTrackConstraint[] _optional;
     MediaTrackConstraint _optional;
 };

Given how the spec is written, there doesn't seem to be a better webidl construct for MediaTrackConstraint than object, and since the array is of objects, treating the whole array as an object is not much of a hardship. FYI.

Comment 13

[Jan-Ivar Bruaroey [:jib] (needinfo? me)]

Sorry, make that:

 typedef object MediaTrackConstraint;

 dictionary MediaTrackConstraints {
     MediaTrackConstraintSet mandatory;
 // Bug 851726
 //    MediaTrackConstraint[] _optional;
     object _optional;
 };

Comment 14

[Boris Zbarsky [:bzbarsky]]

It really makes no sense to put an IDL array in a dictionary.  The point of IDL arrays is to pass by reference, but dictionaries are always passed by value anyway...

You could happily make that sequence<object>, I think.

Comment 15

[Jan-Ivar Bruaroey [:jib] (needinfo? me)]

Of course, that worked fine - Thanks, I keep forgetting that (as does the WebRTC working group which makes me feel better). Definitely not a blocker then. Please ignore.

Comment 16

[Boris Zbarsky [:bzbarsky]]

IDL arrays no longer exist in the spec.