Ooh, nice method, I've had this problem before.

To add to that, you could ease debugging (or at least, tracing of calls to 
EventAdmin) by attaching the method that calls this postEvent or sendEvent 
method, for instance by using
    result.append(Thread.currentThread().getStackTrace()[2])

I am aware that this will fail in certain environments, but in most situations 
it gives you an idea of 'where this message came from'.

Angelo

On Jan 22, 2011, at 3:18 PM, Marcel Offermans wrote:

> For ACE-112 [1] I had to debug the events flowing through the system to see 
> if the events being sent were created and received correctly. Most events (in 
> ACE) are being sent asynchronously, which makes debugging by setting 
> breakpoints a bit more difficult, as setting breakpoints in handleEvent() 
> will not reveal the sender.
> 
> One way to get more information is to "intercept" all calls to EventAdmin, 
> and the technique to do this is what I wanted to share with you in this 
> e-mail.
> 
> What we can do is use the Dependency Manager to create an aspect service on 
> top of EventAdmin. An aspect can intercept all calls to services and in your 
> interceptor you can still delegate the calls to the original service. 
> Technically, this is being implemented using service rankings, which means 
> that consumers will automatically react to aspects.
> 
> What I did was add the following to the Activator of the VaadinClient:
> 
>        manager.add(createAspectService(EventAdmin.class, null, 10, null)
>            .setImplementation(new EventAdmin() {
>                volatile EventAdmin m_eventAdmin;
> 
>                private String print(Event event) {
>                    StringBuffer result = new StringBuffer();
>                    result.append("Event[ ");
>                    String[] names = event.getPropertyNames();
>                    for (String name : names) {
>                        result.append(name);
>                        result.append('=');
>                        result.append(event.getProperty(name));
>                        result.append(' ');
>                    }
>                    result.append(']');
>                    return result.toString();
>                }
> 
>                public void postEvent(Event event) {
>                    System.out.println("postEvent: " + print(event));
>                    m_eventAdmin.postEvent(event);
>                }
> 
>                public void sendEvent(Event event) {
>                    System.out.println("sendEvent: " + print(event));
>                    m_eventAdmin.sendEvent(event);
>                }
>            }));
> 
> When we create the aspect service, the first two parameters are the service 
> class and a filter condition (null here) we want to create aspects on top of. 
> For each service that matches the service name and filter, and aspect will be 
> instantiated of the class specified below. The third parameters is the 
> service ranking for this aspect. This is there because you can also chain 
> aspects on top of each other. The final option (null here) is the name of the 
> member that should be injected into the aspect service (null means any member 
> of the right class).
> 
> The implementation is just one that intercepts all calls, prints the event, 
> and then delegates them to the real EventAdmin (which is injected 
> automatically).
> 
> So it's quite easy to debug a service this way, and it's fully dynamic. Even 
> in a running production system, you can install a new bundle that adds an 
> aspect on top of an existing service and you can instantly intercept all 
> calls, and stop or uninstall the bundle again once you're done.
> 
> Greetings, Marcel
> 
> [1] https://issues.apache.org/jira/browse/ACE-112
> 

Reply via email to