Ralf,
that just reminds me that this is not only related to Hibernate, but to
Castor JDO as well, as we are using CGLIB to create proxies for 1:1
relations that are flagged as lazy-loaded.
Testing for the term CGLIB in a class name does not appear to me like an
ideal solution, but I cannot think about any other way to go about this
problem.
Regards
Werner
Ralf Joachim wrote:
> Hi Stephen,
>
> that CGLIB stuff is a bit tricky. Maybe I can give soem informations
> that can help you to find a solution.
>
> If you create a CGLIB-Proxy for class A, the proxy will behave as being
> of a class that extends A and may also implement some additional
> interfaces. There for if 'proxy is the CGLIB-Proxy of class A you get:
>
> assertTrue(proxy instanceof A);
> assertTrue(proxy.getClass() != A.class);
> assertTrue(proxy.getClass().getName().equals("A$$ProxiedByCGLIB"));
> // (not sure on the extension of A but it includs '$$' and 'CGLIB')
> assertTrue(proxy.getClass().getSuperClass() == A.class);
>
> Depending on the construction of the proxy, the following may also be
> valid:
>
> assertTrue(proxy instanceof AnAdditionalInterface);
> assertFalse(new A() instanceof AnAdditionalInterface);
>
> The proxy itself behaves similar to java.lang.reflect.InvocationHandler.
> Each method call is passed to an invoce method where the implementer of
> the proxy can define the behaviour of the proxy. In most cases the proxy
> will add some additional functionality to the original class. At
> hibernate I expect that on every invocation of a method of A, the proxy
> will load the required object and call the same method on it. For
> internal use I also expect hibernate to implement an interface that
> offers additional methods.
>
> The problem that Paul has, and others had before, is that Castor calls
> getClass().getName() on every object it have to marshal. As this call
> does not return the name of the original calls but the one of the proxy
> (e.g. A$$ProxiedByCGLIB) Castor is not able to find the right mapping
> and uses intorspection instead. Introspection also finds the hidden
> interface methods and tries to marshal their result.
>
> The only 2 solution I can see are:
>
> 1. test for the hidden interfaces with instanceof. as this would require
> us to rever to a class of hibernate I'll refuse that.
>
> 2. test if getClass().getName() includes 'CGLIB'. if that's true call
> getClass().getSuperClass().getName() instead and use this one to search
> for a mapping.
>
> Just my 0,02 cent or maybe a bit more ;-)
>
> Regards
> Ralf
>
>
>
> Stephen Bash schrieb:
>> Paul-
>>
>> I've been pondering your problem on and off most of the day. Recently
>> someone else had a different problem also related to the CGLIB
>> proxies, so there may be some merit in looking into them.
>>
>> Just to make sure I understand the problem, the proxy classes have
>> extra get methods that return objects you don't want to marshal? And
>> since Castor validates the mapping file against the real classes, not
>> the proxy classes, marking the "field" as transient causes a
>> MappingException?
>>
>> I'll try to think about it a little and see if we can find a simple
>> solution.
>>
>> Stephen
>>
>>
>> On 3/2/06, Paul Grillo <[EMAIL PROTECTED]> wrote:
>>
>>> I have a nasty problem. Been using Castor for a couple of years for
>>> all our
>>> classes. Basically we have a large set of mapping files that
>>> generally use
>>> auto-complete="true" and we specify what is transient.
>>>
>>> We are now implementing Hibernate to perist. However, when we load
>>> up an
>>> object from the DB and then go to marshal it we see the following
>>> problem.
>>>
>>> Hibernate uses CGLIB and proxies for the class. At run time there is a
>>> "getxxxx" that returns an object that we definately do not want to
>>> marshal.
>>>
>>> So, i can'd set it transient because Castor complains that it cannot
>>> find it
>>> in the unproxied/mapped class.
>>>
>>> It will be a large problem for us to crawl through all our mapping
>>> files and
>>> specify each attribute for mapping. Is there a way around this?
>>>
>>> Are there any hooks in Castor where i can indicate to it to not
>>> marshal a
>>> particular attribute? Any code i can inject any where to handle this?
>>>
>>> thanks.
>>>
>>> oh, the decision to use Hibernate has been made. Not sure i can
>>> switch to
>>> anything else at this point.
>>
>>
>
> -------------------------------------------------
> If you wish to unsubscribe from this list, please send an empty message
> to the following address:
>
> [EMAIL PROTECTED]
> -------------------------------------------------
>
>
-------------------------------------------------
If you wish to unsubscribe from this list, please
send an empty message to the following address:
[EMAIL PROTECTED]
-------------------------------------------------
