Well, IMO if we try to make the RPC MR schema aware, then what we are doing is almost what the code generated MR does, so why not use the code generation for that. Second if try to make it schema aware then the complexity become high and will have a performance impact too. I think let's get the issue resolve first and then let's see what we can do.
Anyway we can not do any code changes until we fix the build failures. Thank you! Deepal > Hi Glen & Deepal, > > As I understood from the discussion we need to support following > (correct me if this is wrong), > > Type Schema Result if missing > --------------------------------------------------------------------- > primitive minOc=1 exception > primitive !( minOc=1) exception > Object minOc=0 && nillable=true null > Object !(minOc=0 && nillable=true) exception > > > Making RPC MR schema aware shouldn't be that difficult. > What do you think? > > Thanks, > Irantha > > > > > > ----- Original Message ----- From: "Deepal jayasinghe" > <deep...@gmail.com> > To: <axis-dev@ws.apache.org> > Sent: Saturday, December 13, 2008 12:18 AM > Subject: Re: Looking into open issue 3364 > > >> Hi Glen, >>> >>>> As I mentioned in the JIRA, I did a local fix and get that working, >>>> however most of the java2wsdl and some runtime code generation tests >>>> failed. So I did not commit the changes. My changes only involve in >>>> SchemaGenerator, and it is very simple fix. Only thing is to make sure >>>> you get the test cases working. Currently for any input type it sets >>>> minoccurs to 0 , so you just need to change that to 1. >>> >>> So you're saying we shouldn't support optional parameters, and >>> therefore we should be throwing an exception for missing parameters? >>> And we shouldn't be able to pass null values? Hmm... >> Well I was referring to generate WSDL to cope with RPC MR. I agree with >> you, but to handle those we need to fix the RPC MR. So my suggestion was >> to get what working right, and then address the improvements later. >>> >>> I think we should be able to be as expressive and simple as possible >>> in making our schemas align with the Java structure - and supporting >>> missing wire parameters which will become null values for Object types >>> certainly seems reasonable toward that end. >> Agreed. But if you put minoccurs=0 , that's mean it is possible that >> parameter not appear in the request. Of course we handle nillable >> correctly. In my understanding nillable and minoccurs are two different >> things (please correct me if I am wrong) >>> >>> In other words, IMO we should support on the server side: >>> >>> Type Result if missing Schema >>> ---------------------------------------------------------- >>> primitive exception minOc=1 >> +1, but we need to burn schema aware passing logic into RPC MR. >>> Object null minOc=0 >> Nope, here we need to add additional "nillable=true" attribute for >> complex types. >>> >>> Also, this way if the client wants to send a null there's a way to do >>> it (don't include the param). >>> >>> This doesn't seem that hard to do, and I think it makes for nicer REST >>> URLs and SOAP messages in some cases. >>> >>> Thoughts? >>> >>> --Glen >>> >> >> >> -- >> Thank you! >> >> >> http://blogs.deepal.org >> > -- Thank you! http://blogs.deepal.org
