We don't need this. We have a plain old dynamic proxy on the front end with interceptors so it can be configured at runtime. At the end of the interceptor chain we do need a reflection call, but that may be in another vm or a different class loader.

-dain

On Friday, September 5, 2003, at 03:38 AM, Sagenschneider, Daniel A wrote:

Then can we extend that through code generation (i.e. switch statements automatically generated specific to the particular EJB) for the mapping of the appropriate EJB implemented method on the server? The generated code could be created at deployment and would be something as following:


class ParticularEnterpriseBeanMapper extends EnterpriseBeanMethodMapper {


void callImplementationMethod(ParticularEnterpriseBean ejbImplementationInstance
, int methodIndex
, Object[] arguments
, /* potential objects for callback as statically required */) {
// generic potential hookin calls
...
switch(methodIndex) {


                case 1:
                        // method specific potential hookin calls

ejbImplementationInstance.method((ArgumentTypeOne)agruments[0]
, (ArgumentTypeTwo)arguments[1]); // hardcoded linkage - reducing memory and computation


                ...

        }
        ...
        // generic potential hookin calls
}

}


This could be coded to be thread safe and thus have both a lower computation and memory overhead. We could even identify appropriate number ranges or better a separate namespace for the different EJB methods (ie create, remove and so forth).



Reply via email to