反射或运行时类声明？

Question

well I have a simple question.

we are working on a simple application server(like), and this server accepts clients business at runtime without restarting the server.

When user implements it's business and deploy it to the server, server just try to find out the archive descriptor, and load the modules, and it works good.

some operations need to much reflection calls, and for each call they called again and again. for example there is a method which accepts any object, then search for a certain field witch has signed by a annotation and do some business with it, so if we call this method 1000 times with one same object, this is going to reflect 1000 times.

my question is, is it efficient? I mean doesn't it eat up the CPU?! the only possible solution I'm thinking is that create a class and compile it for each object(maybe wrapper) and the method will just find out the wrapper class. but I know this may make the system complex, and hard to debug.

current solution is working, but I think doing a work 1000 times is kinda not logical even it's simple and easy.

Thanks in advanced.

Answer 1

The use of reflection to dynamically load classes at runtime is not a bad choice per-se. Based on your description, you should provide an extensible framework that allows your client to make an implementation, and run their business logic based on that instead of some implicit run-time annotated magic.

A good real-world example for this off the top of my head is The Servlet API .

Answer 2

for example there is a method which accepts any object, then search for a certain field witch has signed by a annotation and do some business with it, so if we call this method 1000 times with one same object, this is going to reflect 1000 times.

In this case I suggest you to use caching. After reflection is finished you'll know class name and the field name. Store them in a HashMap with Class type key and a Method as a value. Next time you invoke "the method" check cache first.