279021 - controls not bound to external instance data

Status: RESOLVED FIXED

(Core Graveyard :: XForms, defect)

Description

[aaronr]

Binding to data in an external instance document doesn't work.

Comment 1

[aaronr]

Attachment: shows the problem

both input fields should show names, data brought in from the external instance
data.

Comment 2

[aaronr]

Attachment: the external instance data for the testcase

the externtal instance data

Comment 3

[Allan Beaufour]

Changed the title to match the problem. The external data _is_ loaded, the
problem is that the external instance document is not ready when a control tries
to find the context node for a bind in nsXFormsUtils::FindBindContext().

Comment 4

[Allan Beaufour]

Attachment: Quick patch

This is kind of hackish... I assume that if a model is returned, its valid even
though the node binding fails.

That makes the control bind to the model, even though the node binding fails in
the first place, and thus it gets bound correctly when it is refreshed after
the external instances are loaded.

Comment 5

[Olli Pettay [:smaug][[e-mail]]]

Comment on attachment 172965 [details] [diff] [review]
Quick patch

I wonder why this patch hasn't been checked in. 
(or sr'd)
I'd say it is anyway doing the right thing, it is not too hackish.

Comment 6

[aaronr]

This patch will work for most cases.  But there are a few edge cases, like if
the model is AFTER the controls, then it still won't work.

If we look at: http://www.w3.org/TR/xforms/index-all.html#evt-revalidate in the
'note:' part, it really makes it sound like we shouldn't bind any controls at
all until the xforms-ready is handled.  So maybe we should just build a list of
controls that need binding to something (i.e. they have single node binding
attrs) and then process this list of controls when the document is ready?  That
would solve other inefficiencies, too, like recalcs that happen before the form
is ready.  But by doing this, we pretty much give up all hope of allowing the
user to  start filling in the form while we are loading and, indeed, we could
write over whatever the user might fill in early when our bindings finally hit.

Comment 7

[Allan Beaufour]

(In reply to comment #5)
> (From update of attachment 172965 [details] [diff] [review] [edit])
> I wonder why this patch hasn't been checked in. 
> (or sr'd)
> I'd say it is anyway doing the right thing, it is not too hackish.

I would say that we check this one in, and then create a "model after controls"-bug.

Comment 8

[Allan Beaufour]

(In reply to comment #6)
> This patch will work for most cases.  But there are a few edge cases, like if
> the model is AFTER the controls, then it still won't work.

You are probably right here.
 
> If we look at: http://www.w3.org/TR/xforms/index-all.html#evt-revalidate in the
> 'note:' part, it really makes it sound like we shouldn't bind any controls at
> all until the xforms-ready is handled.

But probably not here :) xforms-ready is fired when all controls are
initialized, after the xforms-model-construct-done event. So
xforms-model-construct-done is the correct one to listen to
(http://www.w3.org/TR/xforms/slice4.html#evt-modelConstructDone)

So as I wrote above: We check this one in and create a new bug, ok?

Comment 9

[aaronr]

It is fine with me if this patch goes in.  But we still need to worry about:

1) what happens if model is later
2) inefficiencies with recalcs, revalidates, etc. when controls aren't bound,
yet.  I'll post a bug when I figure out exactly what I mean :=)

Comment 10

[Allan Beaufour]

Checked in.