283376 - GetTypeForControl API needs to return default type for node

Status: RESOLVED FIXED

(Core Graveyard :: XForms, defect)

Description

[aaronr]

nsXFormsModelElement::GetTypeForControl calls nsXFormsUtils::ParseTypeFromNode.
 Neither return a type of 'string' which is the default type for an instance
node if there is no @type or type MIP bound to the instance node. 
nsXFormsUtils::ParseTypeFromNode is probably the right place to do this work.

Things to make sure about is handling type="" and that we return the proper NS
prefix for the document that the node is in.

Comment 1

[aaronr]

Attachment: proposed fix

This patch will need to change to accept properties once bug 283004 goes in

Comment 2

[aaronr]

Attachment: same patch with dos2unix run on it

Comment 3

[aaronr]

merged this patch with 283004 and ran into problems.  lookupPrefix failing if
the document doesn't have the http://www.w3.org/2001/XMLSchema defined to a
prefix and it sometimes fails even if it is set.  We'll have to change the code
in ValidateNode and GetTypeForControl to accept a failure from
nsXFormsUtils::ParseTypeFromNode and pass in 'http://www.w3.org/2001/XMLSchema'
and 'string' to nsXFormsSchemaValidator::GetType()

Comment 4

[aaronr]

Attachment: 2nd attempt

Changed ParseTypeFromNode to return NS_ERROR_NOT_AVAILABLE if type attribute
not found.  Also changed ValidateNode and GetTypeForControl to use this return
to set up 'xsd:string' as the type to use in those situations.

Doing this patch seperately from 283004 since that bug seems to be at an impass
at the moment -> waiting for the ability to put properties on attributes, I
believe.  Allan, should we just put this patch as part of 283004, or put it in
ahead of 283004?

Comment 5

[Allan Beaufour]

(In reply to comment #4)
> Created an attachment (id=175492) [edit]
> 2nd attempt

Just glanced at the code ... having the exact code at two places, does not seem
that good?

> Allan, should we just put this patch as part of 283004, or put it in
> ahead of 283004?

Just go ahead with this one.

Comment 6

[aaronr]

Attachment: 3rd attempt

consolidated as much as I could so that code exists in one place
(GetTypeAndNSFromNode) and called from two places (ValidateNode and
GetTypeForControl)

Comment 7

[Doron Rosenberg (IBM)]

Comment on attachment 175571 [details] [diff] [review]
3rd attempt

nits given over IRC.

Comment 8

[aaronr]

Attachment: fixed Doron's nits

Comment 9

[Doron Rosenberg (IBM)]

Comment on attachment 175576 [details] [diff] [review]
fixed Doron's nits

>+  if(!typeExists) {
>+    return NS_ERROR_NOT_AVAILABLE;
>+  }

should be if (), other than that, r=me

Comment 10

[Olli Pettay [:smaug][[e-mail]] (holiday -> Aug30)]

Comment on attachment 175576 [details] [diff] [review]
fixed Doron's nits


>+  PRBool typeExists = PR_FALSE;
>+  nodeElem->HasAttributeNS(NS_LITERAL_STRING(NS_NAMESPACE_XML_SCHEMA_INSTANCE),
>+                           NS_LITERAL_STRING("type"), &typeExists);
>+  if(!typeExists) {
>+    return NS_ERROR_NOT_AVAILABLE;
>+  }
>+
>   nsAutoString typeAttribute;
>   nodeElem->GetAttributeNS(NS_LITERAL_STRING(NS_NAMESPACE_XML_SCHEMA_INSTANCE),
>                            NS_LITERAL_STRING("type"), typeAttribute);
> 

You could use named literal strings for NS_NAMESPACE_XML_SCHEMA_INSTANCE and
"type".
With those r=me

Comment 11

[aaronr]

Attachment: final patch

I fixed Doron and Olli's suggestions.  Doron, please check in for me.

Comment 12

[Doron Rosenberg (IBM)]

checked in.