169000 - Write a wrapper library to access XPConnect native wrapped objects from within chrome JS

Status: RESOLVED FIXED

(SeaMonkey :: UI Design, defect)

Description

[Christopher Aillon (sabbatical, not receiving bugmail)]

I've fussed over this making it better and better.  I think we're ready for
prime time.  Patch to follow.

Comment 1

[Christopher Aillon (sabbatical, not receiving bugmail)]

Attachment: nativeDOM.js v.1.0

Comment 2

[jag (Peter Annema)]

Looks good. Get input from some other folks, perhaps blake, I'm sure Phoenix can
(and should) use this.

Comment 3

[Christopher Aillon (sabbatical, not receiving bugmail)]

Attachment: nativeDOM.js v.1.1

I had rginda have a look at this from a JS standpoint.	This fixes his issues:

- Checks the exception for the one we want to eat, throws all others.
- return null instead of undefined; undefined can be assigned to, null can't.
- small optimization in getNativePropertyProxyFor()
- remove unnecessary consts in _getNativesForProp()

Comment 4

[jag (Peter Annema)]

Comment on attachment 99450 [details] [diff] [review]
nativeDOM.js  v.1.1

sr=jag

Comment 5

[Robert Ginda]

Talking with jag on IRC, we came to the conclusion that an XPCOMProxy object
would be a better idea.

Clients would say something like...

var proxy = new XPCOMProxy (obj);
// create getter/setter pairs for these property names.
proxy.fetch ("foo", "bar", ...);

// for..in over this object, creating properties for each key found.
property.fetch (Components.interfaces.nsIFoo);

Comment 6

[Christopher Aillon (sabbatical, not receiving bugmail)]

Comment on attachment 99450 [details] [diff] [review]
nativeDOM.js  v.1.1

I have this working, but waiting on changes in bug 169321 to land first, since
this relies on that.

Comment 7

[Christopher Aillon (sabbatical, not receiving bugmail)]

Oh, there is a workaround for that bug that brendan mentioned.  I'll attach a
patch anyway either later tonight or tomorrow morning that implements rginda's
suggestion.  I'd prefer to do this though with the fix in bug 169321.

Comment 8

[Brendan Eich [:brendan]]

I'm missing the motivation here -- can someone give a pep-talk?

/be

Comment 9

[Christopher Aillon (sabbatical, not receiving bugmail)]

Responded to Brendan via private e-mail.

Comment 10

[Christopher Aillon (sabbatical, not receiving bugmail)]

Attachment: Updated patch

Implements the syntax proposal from rginda and jag.

Comment 11

[Johnny Stenback (:jst)]

Comment on attachment 107789 [details] [diff] [review]
Updated patch

sr=jst

Comment 12

[[e-mail]]

+      var ary = aName.match(/([^(]*)(\(\))?$/);
+      var name = ary[1];
+
+      // If we are passed a string like "foo()", ary[2] will be of type string,
+      // and contain "()".  If passed "foo", then it will be of type undefined.
+      if (typeof ary[2] === "string")
+        this._doImportMethod(name);
+      else
+        this._doImportProperty(name);

IMHO this is very ugly.

if (name.slice(-2) == "()")
  this._doImportMethod(name.slice(0, -2));
else
  this._doImportProperty(name);

Comment 13

[jag (Peter Annema)]

Comment on attachment 107789 [details] [diff] [review]
Updated patch

Fix Neil's nit and r=jag

Comment 14

[jag (Peter Annema)]

Unless of course the regexp is faster than two slices.

Comment 15

[[e-mail]]

Fix checked in.