I stumbled across this problem while working on the explicitHeaderBranch.
I have a fix for the getOperationQName code and the runtime.
The fix is to have getOperationQName use the correct operation qname ...not
some embedded element as is the current case for document style. I then
changed MessageContext.getPossibleOperationsByQName to detect document
style and
get the operations that have a parameter that that matches the
qname....much safer.
So this fix will be applied when the explicitHeaderBranch is merged in.
Rich Scheuerle
IBM WebSphere & Axis Web Services Development
512-838-5115 (IBM TL 678-5115)
Chris Burnley
<[EMAIL PROTECTED] To: [EMAIL PROTECTED]
> cc:
Subject: Re: Correctness of
10/03/2002 07:42
org.apache.axis.wsdl.toJava.Utils.getOperation QName
PM
Please respond to
axis-dev
Tom, you are right. "type" should only be used for encoded, but
"element" may be used for literal. I've changed the example I gave
earlier to use literal, using rc2 now, wsdl2java still generating the
operation name as "anElement".
If I run the ticker example in the 1.1 spec, the operation name is still
not correct. ( Even though it is possible that the example actually
doesn't work, because I found an mistake:
binding="tns:StockQuoteBinding" should be
binding="tns:StockQuoteSoapBinding" )
I'm willing to submit a patch for it as long as we can sort out what the
correct behaviour should be. If what I've got is illegal, then I can
always change the schemas/wsdl to be legal.
BTW, this used to work fine with beta 2.
Tom Jordahl wrote:
>Chris,
>
>If I understand your question right, see section 3.5 of the WSDL 1.1
specification, the third to last paragraph:
> "If use is encoded, then each message part references an abstract type
using the type attribute."
>
>WSDL spec should be at http://www.w3.org/TR/wsdl
>--
>Tom Jordahl
>Macromedia Server Development
>
>
>
>-----Original Message-----
>From: Chris Burnley [mailto:[EMAIL PROTECTED]]
>Sent: Wednesday, October 02, 2002 8:58 PM
>To: [EMAIL PROTECTED]
>Subject: Correctness of
>org.apache.axis.wsdl.toJava.Utils.getOperationQName
>
>
>I have a situation where I a client is using an element instead of a
>type for a message parameter in my WSDL file (simplified example
>attached). When I generate the stubs, the operation name call looks like
>this:
>
> _call.setOperationName(new
>javax.xml.namespace.QName("http://www.xyz.com/schema/Elements";,
>"anElement"));
>
>rather than the expected:
>
>_call.setOperationName(new
>javax.xml.namespace.QName("http://www.xyz.com/soap-services";,
>"anOperation"));
>
>I read the comments for getOperationQName and someone wrote:
>
> // NOTE: it is possible for someone to define a part as an element
> // while using rpc/encoded, which is wrong and we might want to
>catch it
> // here.
>
>Is this actually true ? Could someone point me to the spec that points
>this out ? It seems that this could be easy to fix by adding
>
>@@ -566,6 +566,9 @@
> if (ns == null) {
> ns = "";
> }
>+ else {
>+ return new QName(ns, operationName);
>+ }
>
>to Utils.java, revision RC1
>
>regards,
>
>Chris Burnley
>
>
>
