JIRA issue TUSCANY-8 raises the issue of mapping WSDL operation names to
Java method names.
Let's imagine a scenario with an SCA entry point. The entry point uses a
WSDL service definition and a Web Service binding to publish a POJO
component. The POJO component implements a Java interface. This is a
top-down scenario where you start with the WSDL file describing the Web
Service (let's say it's a standard WSDL, or you have agreed on that WSDL
with one of your business partners), and you try to write or generate
the corresponding Java interface for your POJO component. The problem is
that your WSDL operation names won't always translate nicely to nice
Java method names. You'll run into WSDL name to Java name mapping issues
in a number of cases, for example:
- WSDL operation "GetQuote" -> do u map that to a getQuote method? or to
a GetQuote method (which doesn't follow the Java naming best practices)
- WSDL operation "get-quote" -> you need to map this to a valid Java
method name without the '-' like getQuote
- WSDL operation "continue" or "import" -> these are Java keywords so
again you can't just use the operation name as your method name.
We could come up with a name mangling algorithm but I don't really like
that. If we generate the interface we'll end up with ugly names for
these cases, if I'm an app developer and I want to write the interface
by hand I'll have to understand the name mangling algorithm and compute
the mangled name in my head. Such an algorithm may also cause
ambiguities / undeterministic behavior (if for example we look for both
getQuote and GetQuote Java methods for a GetQuote WSDL operation).
Instead of doing that I'd like to propose the following:
- a one-to-one mapping to cover all the cases that do not cause this
issue (i.e. a getQuote WSDL operation simply maps to a getQuote method)
- the use of annotations to define the WSDL to Java mapping, we would
allow you to annotate a WSDL operation with an XML element to drive the
generation of a Java method with the name you want (this is similar to
what SDO does with its commonj.sdo/java annotations)
- the use of Java 5 annotations to indicate which WSDL operation a given
Java method maps to.
The same principles would also apply to the WSDL portType name to Java
interface name mapping (which presents very similar issues).
Any thoughts?
--
Jean-Sebastien
