RE: [sdo] ISSUE 41: an API for elementForm and attributeForm

["Barack, Ron" <ron.barack@sap.com>]

Hi Blaise,

 

Right now, as I understand the spec, the default behavior is both elements and the attributes are rendered “unqualified”, so I’m worried that your suggestion for generating metadata from DataObjects is a breaking change.  If we could make the choice freely, I would have preferred your rule as the default behavior, since this seems to be the more standard XSD usage.  Also, is your suggestion “in addition to” or “instead of” the ability to control the qualitification through @xmlPropertyUri?    

 

Regarding generating metadata from Java Annotations, I think this is a nice idea, but we should also be aware that the mapping from namespace to schema is 1:n, and that, if we were writing the schema per hand, we could give each schema different “formDefaults”.  How common this is in practice, I don’t know.

 

Similarly, even when the metadata comes from XSD, we have no guarantee that we will be splitting up the name space in the same way when the XSD is rendered.  It could be that 2 schemas are used to define the same namespace, and each schema has different values for “elementFormDefault”.  When we later generate XSD from all the types in the namespace, which value should we use?

 

The fact is, certain details get lost when converting from XSD to SDO… this is intentional, otherwise, SDO would be just as crazily complex as XSD.  Obviously, though, we want to be as accurate as possible.  But here we have a change that doesn’t improve the quality of the mapping at all.  Here we have a detail that only has to do with how humanly readable the generated schema is,  with whether or not we need to write the “form=’qualified’” on every property or have a default value. 

 

Maybe we should add a XSDHelper.generate method that takes some options, this would give at least a framework for implementing this functionality.  I think we may need these options anyway, so that we can supply something like a reverse-schema-resolver, to generate the import statements for referenced types (e.g., in another namespace).

 

Best Regards,

Ron

 

From: Blaise Doughan [mailto:blaise.doughan@oracle.com]
Sent: Montag, 14. Juni 2010 17:57
To: sdo@lists.oasis-open.org
Subject: Re: [sdo] ISSUE 41: an API for elementForm and attributeForm

 

Hi Ron,

The approach you describe will result in properly qualified XML documents, but does not sufficiently handle the schema generation aspect which I feel is an important use case.

Metadata generation from XML schema:

• Namespace qualification rules for properties are clear
• Types could remember the schema qualification used on the define when the generate schema is performed

Metadata generation from Java annotations:

• We could add a package level annotation called @SdoXmlSchema that would allow the settings for elementFormDefault/attributeFormDefault to be set.  This would avoid the need for most XML namespace qualification needing to be done through annotations (assuming most users want elementFormDefault="qualified" and attributeFormDefault="unqualified").
• Types could remember the schema qualification used on the define when generate schema is performed.

Metadata generation from DataObjects:

• elementFormDefault=qualfied & attributeFormDefault=unqualified could be assumed when defining metadata this way.

-Blaise

Barack, Ron wrote:

http://www.osoa.org/jira/browse/SDO-41

 

Hi Everyone,

 

Since we don’t have a construct that represents the namespace, much less the schema, I view would like to defer setting the elementFormDefault and attributeFormDefaults  until we have the appropriate targets on which to set them.  And anyway, we don’t have a requirement to produce schema that looks like a human wrote it, so that I think it is 100% satisfactory to specify on the property themselves whether they should be rendered in XML “qualified” or “unqualified”.  This is a real requirement, IMO, since it has to do with producing valid XML.

 

Whether or not a property is rendered with a namespace is currently determined by the schema (ie, when the type is defined through XSD) or by looking at the annotations on the Java getter method (@SdoXmlPropery.propertyUri).   It seems to me that the only real hole here is that

a.      There is no way to specify the property URI though the API.

b.      There is no way to inspect the value.

Both of these issues can be addressed by adding a new open content property, xmlPropertyUri.  We already use this approach eg in xmlElement.

 

I was going to make a concrete proposal with wording, but on reviewing the core spec, I think the description of xmlElement is extremely poor, and needs to be completely rewritten.  For instance, on reading the spec, it seems like xmlElement can only be used on openContent properties.

 

Perhaps we can get some consensus on the approach during this week’s call, and I can try for a concrete proposal for next week.

 

 

Ron