Re: [wsrf] Discussion regarding issue 48

[Steve Graham <sggraham@us.ibm.com>]

Interesting.

So, lets say I have the following resource
property definition:

<xs:schema ... targetNamespace="foo"
... xmlns:foo="foo" >
...
  <xs:element name="bar"
type="xs:String" nillable="true"/>

  <xs:element name="RPDocDecl">
    <xs:sequence>
      <xs:element
ref="bar" minOccurs="1" maxOccurs="unbounded"
/>
...
    </xs:sequence>
  </xs:element>
...
</xs:schema>

I have a WS-Resource defined on a portType
that references foo:RPDocDecl as its resource properties document schema,
and this particular WS-Resource has only one element value of "bar"
that is nil. That is, a serialization of the resource property document
could appear as:
  
<foo:RPDocDecl>
  <foo:bar xsi:nil="true"
/>
 ...
</foo:RPDocDecl>

So, if a requestor issued the following
GetResource Properties request:

  <wsrp:GetResourceProperty>
    foo:bar
  </wsrp:GetResourceProperty>

I would expect:

  <wsrp:GetResourcePropertyResponse>
    <foo:bar xsi:nil="true"
/>
  </wsrp:GetResourcePropertyResponse>

Now a couple of other "variants":

There are several values of bar, that
is the resource properties document looks like:
<foo:RPDocDecl>
  <foo:bar xsi:nil="true"
/>
  <foo:bar>someString</foo:bar>
 ...
</foo:RPDocDecl>

I would expect the getResourceProperties
response to look like:

  <wsrp:GetResourcePropertyResponse>
    <foo:bar xsi:nil="true"
/>
    <foo:bar>someString</foo:bar>
  </wsrp:GetResourcePropertyResponse>

and if there are no values for foo:bar,
that is the resource properties document looks like:
<foo:RPDocDecl>
...
</foo:RPDocDecl>

I would expect the getResourceProperties
response to look like:

  <wsrp:GetResourcePropertyResponse>
  </wsrp:GetResourcePropertyResponse>

What aspect of the scenario am I missing
that suggests the invention of some conventional element like "nillablerp"?

sgg
++++++++
Steve Graham
(919)254-0615 (T/L 444)
STSM, On Demand Architecture
Member, IBM Academy of Technology
<Soli Deo Gloria/>
++++++++

Samuel Meder <meder@mcs.anl.gov> 06/15/2004 02:05 PM Please respond to meder

To:        "Murray, Bryan P." <bryan.murray@hp.com>         cc:        wsrf-oasis <wsrf@lists.oasis-open.org>, Jarek Gawor <gawor@mcs.anl.gov>         Subject:        Re: [wsrf] Discussion regarding issue 48

I've talked to Jarek about issue 48. What we would like to see is a
clarification of what to return when a property is declared nillable and
it is nil. Our suggestion is to return something like 

<nillablerp xsi:nil="true"/>

Other alternatives are to return nothing or to throw a fault. The other
issue that should be clarified is the precedence of

"Note: in the case where the resource property element is defined
with
minOccurs="0" and the resource properties document does not contain
any
value for that resource property, the response MUST be an empty
wsrpw:GetResourcePropertyResponse element."

over whatever behavior we decide on for the nillable case. Our
suggestion is to give the minOccurs="0" rule precedence.

More comments below:

On Thu, 2004-06-10 at 14:31, Murray, Bryan P. wrote:
> The minutes indicate I have an action item to start an email discussion
> to clarify issue 48. This issue was originally raised by Jarek Gawor.
> 
> I am not sure whether the original intent is with reference only to
> properties that are declared in schema as nillable or if it is about
any
> property which can possibly have a null value.
> 
> As an implementer I can comment on properties that can have a null
> value. For instance, in most languages a string can have a non-null
> value, can be empty and non-null, and can have a null value. If the
> property is not declared as nillible in schema and minOccurs >
0, what
> is the correct value to return if the string has a null value? The
> choices in this case are 1) return property with no value, 2) don't
> return the property at all but don't fault, 3) fault. I don't think
we
> can return the property with nil=true. My preference is #1 which is
the
> same if the string has a non-null value that has 0 length.

I think we should return a fault since it is not declared as nillable.

> Another case I can think of that is related to having a null value
is a
> property that is declared with minOccurs=0 and maxOccurs > 1. This
may
> sometimes be implemented as an array. What should be done if the array
> is null, empty, or non-null and non-empty? It seems to me that if
the
> array is null no property elements should be returned just as for
an
> array with no elements, but this is an implementation detail.
> 

It seems to me that this should depend on whether the property is
declared to be nillable or not. My suggestion would be:

      * Nillable
              * if array == null or
empty return <rpelement
                xsi:nil=true>
              * if array is not empty
do the usual
      * Non-nillable
              * if array == null or
empty throw fault
              * if array is not empty
do the usual 

/Sam

> If someone has comments about properties that are declared as nillable
> in the schema or comments about properties having null values please
> reply to the list.
> 
> Bryan
-- 
Sam Meder <meder@mcs.anl.gov>
The Globus Alliance - University of Chicago
630-252-1752