Hi
Stephen
I’ve encountered
this problem a few times in the past while implementing not only CIQ, but
some other standards as well.
Normally strong
typing is best left to the implementation. You’ll always need to tweak the
schema a little bit to fit your requirements, be it the Java binding
library you’re using, various flavours of parsers etc.
I, for example,
had to combine all CIQ schemas into one in order to implement it in
webMethods!
Perhaps it would
be interesting to have versions of CIQ schemas ready for implementation
for different platforms/libraries (eg .NET, XMLBeans, webMethods,
whatnot). Having said this, I would be hesitant of putting any of these
implementation specifics into the standard itself. This is the main reason
why there are no strong-typed elements in the standard itself so
far.
As far as
upgrading to future versions of the standards, there always changes you
will have to implement anyway. If you stick to a consistent data typing
approach, you’ll minimise your migration headaches.
This is
definitely not a purist approach, but a practical and pragmatic one borne
out of necessity in dealing with lots of different tools and libraries out
there.
I hope my comment
is helpful.
Regards,
--
Hido
Hasimbegovic
eBusiness Consultant | CustomWare Asia Pacific | www.customware.net
T:
1300 889 850 (+61-3-9667-0124) | F: 1300 791 715 (+61-3-9663-2616) | M:
+61-416-174-453
Level 2,
222 Latrobe
Street
Melbourne VIC
3000
Australia
Improve the
quality of your services with WmUnit
Unit testing
framework for WebMethods
http://www.customware.net/wmunit
From: Ram
Kumar [mailto:RKumar@msi.com.au]
Sent: Tuesday, 18 January 2005 9:35
AM
To: Viens,
Stephen
Cc:
ciq@lists.oasis-open.org
Subject: [ciq] RE: CIQ XML Schema
Issue & Resolution
Thank you for your detailed
email.
Yes, this is a problem with
V2.0 and it is already in our
list of things to do. The
future versions of CIQ will be strongly data typed.
From:
Viens, Stephen [mailto:Stephen.Viens@LibertyMutual.com]
Sent: Tue 1/18/2005 9:33
AM
To: Ram
Kumar
Subject: FW: CIQ
XML Schema Issue & Resolution
[Please forgive
the length of this message. I do believe you'll find the subject of
interest]
Ram, CIQ TC,
We're leveraging
the current CIQ standard schema's to implement a web service for managing
client data. We've included the CIQ schemas in the WSDL for our
service to describe the data managed by the service. We're using the
"WSDL First" method of web service development which recommends developing
the service's WSDL first and generating service and client code
(Java, C#, C++, etc) from that using common WSDL2Java code generation
tools.
Unfortunately
we've encountered a problem when trying to generate Java code from the CIQ
schemas. We're using two industry standard WSDL2Java generation
tools - Apache Axis 1.1 and IBM WSAD 5.1 - and we've been presented with
the same problem with both. The purpose of this note is to inform
you of the problem and describe how we've worked around
it.
When type
information for an elements character data is not specified the code
generation tools will produce a Java class without a class variable to
contain character data. In other words, the Java generation tools
assume that if type information isn't supplied for an element's character
data then it must be an "attributes only" element. Here's an
example:
<xs:element name="CustomerID" minOccurs="0"
maxOccurs="unbounded">
<xs:complexType mixed="true">
<xs:attribute name="Type"/>
<xs:attributeGroup ref="grPostal"/>
<xs:anyAttribute namespace="##other"/>
</xs:complexType>
</xs:element>
Here's the Java
code generated from the schema snippet above . You'll note that there is
no class variable to hold a CustomerID value - only the "Type" attribute
is represented.
public class CustomerID
{
private java.lang.Object type;
public CustomerID() {
}
public java.lang.Object getType() {
return type;
}
public void setType(java.lang.Object type) {
this.type = type;
}
}
Only when a
simpleContent element is specified with type information for the element's
character data (xs:string) will the generation tools produce Java that
includes a class variable and getter/setter methods for a CustomerID
value. Below is an example. This is the same schema snippet
shown above with the addition of the "xs:simpleContent" and "xs:extension"
elements:
<xs:element name="CustomerID" minOccurs="0"
maxOccurs="unbounded">
<xs:complexType mixed="true">
<xs:simpleContent>
<xs:extension
base="xs:string">
<xs:attribute name="Type"/>
<xs:attributeGroup ref="grPostal"/>
<xs:anyAttribute namespace="##other"/>
</xs:extension>
</xs:simpleContent>
</xs:complexType>
</xs:element>
Here is the Java
code that is generated from the modified schema. You'll notice that a
class variable for the CustomerID value is now present as well as getValue
and setValue methods:
public class CustomerID
{
private
java.lang.String value;
private java.lang.Object type;
public CustomerID() {
}
public java.lang.String getValue() {
return value;
}
public void setValue(java.lang.String value) {
this.value = value;
}
public java.lang.Object getType() {
return type;
}
public void setType(java.lang.Object type) {
this.type = type;
}
}
The way the
WSDL2Java tools work makes sense in a way. How else would they be
able to know if an element includes (or does not include) character data
without it being explicitly stated in the schema?
Would the CIQ
team consider updating the schemas to include these changes or at least
incorporating these changes into future
versions?
We have been
forced to make many alterations to the CIQ schemas for our web service to
leverage them. We're concerned that we will have to make all of the
same alterations when (if) we decide to upgrade to a new version of CIQ in
the future.
Steve
Stephen
Viens
Enterprise
Architect
Liberty Mutual
Insurance Company
Enterprise IT Services
(EITS)
stephen.viens@libertymutual.com
603-245-3467
(phone)
603-245-5338
(fax)