AOP的实现机制

根据你上边说的我大胆猜测性能应该是这个顺序的吧，instrument>javassit>cglib>jdk 动态代理

但是为什么sping默认实现是jdk的动态代理呢，我的册数数据确实jdk的动态代理性能更好点啊，不知lz有没有测试数据，可以比较一下性能差别，这样也好以后有个比对吧。

以下是你附件的性能测试数据，我加了个javassit的

public class PerformanceTest {
    public static void main(String[] args) throws NotFoundException, CannotCompileException, InstantiationException, IllegalAccessException {
        
    	long start = System.currentTimeMillis();
        JavassistAopDemo.aop();
        long  end = System.currentTimeMillis();
        long javassistNeedTime = end - start; 
    	 start = System.currentTimeMillis();
        DynamicProxyDemo.aop();
         end = System.currentTimeMillis();
        long dynamicProxyNeedTime = end - start;

        start = System.currentTimeMillis();
        ProxyBusiness.staticDynamic();
        end = System.currentTimeMillis();
        long staticProxyNeedTime = end - start;

        start = System.currentTimeMillis();
        CglibAopDemo.byteCodeGe();
        end = System.currentTimeMillis();
        long byteCodeGeNeedTime = end - start;
        
        
        System.out.println("dynamicProxy need time:" + dynamicProxyNeedTime + "ms");
        System.out.println("staticDynamic need time:" + staticProxyNeedTime + "ms");
        System.out.println("byteCodeGe need time:" + byteCodeGeNeedTime + "ms");
        System.out.println("javassist need time:" + javassistNeedTime + "ms");
    }

}

dynamicProxy need time:16ms
staticDynamic need time:0ms
byteCodeGe need time:78ms
javassist need time:93ms

代码已经更新，跟我们想想的不一样啊，嘿嘿

从性能上讲，instrument秒杀其它实现

期待lz的性能测试数据，任何东西没有测试数据做依据就是再好的理论都是乏味的，用事实说话是最直接的，嘿嘿