332803 - elements with xsi:type are not validated on form load

Status: RESOLVED FIXED

(Core Graveyard :: XForms, defect, P1)

Description

[Allan Beaufour]

We do not seem to validate elements with types set directly by xsi:type on form load.

Comment 1

[Allan Beaufour]

Attachment: Testcase

Comment 2

[aaronr]

The problem here is that during initialization an instance node won't be inserted into the MDG unless it is bound via a xf:bind.  So when we rebuild, recalculate and revalidate during form initialization, the nodes that are bound via @ref and have xsi:type on them won't be in the MDG, won't be picked up by recalculate and thus won't be validated by revalidate.  In the testcase when the node value is changed to another invalid value, it will correctly display as invalid because SetValue will mark the node as changed and thus it will go into the MDG to be picked up by recalculate and then used by revalidate which will call ValidateNode on it.

I don't know what the right answer is, though.  Maybe ValidateDocument should return a list of invalid nodes and we should then mark all of the controls bound to those nodes as invalid?

Comment 3

[aaronr]

Or maybe in when an instance is through loading it should mark every node it contains as dirty?

Comment 4

[Allan Beaufour]

(In reply to comment #2)
> I don't know what the right answer is, though.  Maybe ValidateDocument should
> return a list of invalid nodes and we should then mark all of the controls
> bound to those nodes as invalid?

Afaik, ValidateDocument() just bails when it finds something which is not valid, does it not?

(In reply to comment #3)
> Or maybe in when an instance is through loading it should mark every node it
> contains as dirty?

That would add every instance data node to the MDG or to the list :(

I've never been really keen on our (my) type handling, but it does make sure that we only check types for nodes that change value -- which is nice :)

Maybe we should just make sure that all instance data nodes that have controls bound to them exist in the MDG?

Comment 5

[Allan Beaufour]

(In reply to comment #4)
> Maybe we should just make sure that all instance data nodes that have controls
> bound to them exist in the MDG?

Best plan so far, I'll chew on it.

Comment 6

[Allan Beaufour]

Attachment: Revised testcase

Use xsd:date instead of custom type, and add inline schema attached type too

Comment 7

[Allan Beaufour]

Attachment: Patch

Every time the MDG creates a new node, it will type check it. This only happens one time (per rebuild), which is the first time a node is accessed by a control.

Comment 8

[Allan Beaufour]

Fixed on trunk