On Aug 1, 8:11 pm, Jason Stelzer <men...@[EMAIL PROTECTED]
> wrote:
> Hello, I'm rather new to corba so if this is a faq, I'll happily go
> read the relevant docs or url. I just need a ****ge.
>
> My situation is this. The company I'm at has a java app running in
> jboss that they need to integrate into a legacy distributed perl app. I
> chose corba because this is going to be used in a performance critical
> piece of code and web services simply have too much overhead. So at
> this point I'm using jacorb on the java side and omniORB on the C/perl
> side.
>
> Since I'm using jboss and the way they sup****t corba is via annotations
> in the java class files combined with reflection at deployment time to
> expose things, this means I need to generate IDL from the class files
> then create C++ stub code. I've done this and have things working.
>
> I've written a C++ library with C wrappers and a perl module in C to
> use them. I have a fairly complicated data structure to return, so I
> created a custom type but I've come into a minor bit of difficulty. I'm
> not clear on how to map a 'traditional' pojo in corba. Something like:
>
> public class Foo {
> private String val;
> public String getVal(){
> return val;
> }
> public void setVal(String s){
> val = s;
> }
> ...
>
> }
>
> When I used idl generated by rmic for this type of bean, I would get
> back an object ref that would consistently return an empty
> CORBA::WStringValue object. After examining the objects in gdb, I'm
> reasonably sure what is happening is that the 'private' string isn't
> being returned by getVal(). Its there. I see it marshaled, the accessor
> just isn't returning what I thought it should. The problem is, I'm not
> sure what the generated IDL should look like as this is the first time
> I've done something like this.
>
> In the interim, what I've done is create a class in java that is
> essentially a glorified C style struct.
>
> public class Foo {
> public String val;
> ...
>
> }
>
> I populate the member variables of a Foo with all the values I need to
> return and on the C++ side it is unmarshaled exactly as I'd expect.
> After that its pretty boring. I just need to copy the data I need into
> perl side data structures, release references and do cleanup.
>
> I'd really like to be able to use the former style pojo as the latter
> makes the other java developers twitch.
>
> I've only spent a few days working on this, and I'm still reading quite
> a bit and learning how stuff gets marshalled. If anyone has example IDL
> I could look at that'd be great. I've done ok with fixing up a lot of
> the busted crap that rmic generates, but I seriously lack experience at
> this stuff.
>
> As it stands, I'm at least feature complete from a prototype
> perspective. I can make the calls I need to from perl and get the right
> data back. I'm sure I have a bit of tweaking to do to ensure there are
> no leaks, but things work.
>
> --
> J.
Hi,
you can use "attribute's" to represent ur simple bean.
Ur IDL will look something like this:
sample.idl :
interface Sample
{
attribute String val;
};
when this idl file is compiled by a idl compiler it will generate
automatically a getter and a setter (like your bean).
I have used omniORB its pretty cool and easy to use,
but I have one question for u - howcome ur using JacORB (on java side)
instead the std SunORB (that comes as part of the JDK), any specific
reason?
Regards,
Arun
|
