> Really? The examples I have seen all use the stack trace in some way. > How are you accessing the current method?
Hehe... to be honest, I've not yet tried it on Java, but I've used a similar approach about a year ago on a .NET application and I've used something like Method.getCurrentMethod(). This was pretty fast. I don't know how this method internally works, but it was definitely not as slow as the way I've done it in my current Java project via Stack traces. I've thought that there is a similar method in Java too... but it seems that this assumption was wrong. Dominik -----Ursprüngliche Nachricht----- Von: Jeff Bischoff [mailto:[EMAIL PROTECTED] Gesendet: Dienstag, 7. November 2006 20:12 An: MyFaces Discussion Betreff: Re: AW: AW: AW: AW: [O/T] JSF Best Practices for Authentication/Authorization Dominik, Reponses inline: [EMAIL PROTECTED] wrote: >> Hmm, I wonder what kind of overhead that incurs. Of course, if you are >> only checking it once per HTTP request, I don't suppose it would matter. >> Sounds like it would be roughly equivalent to creating an exception (due >> to the stack trace manipulation). > > You are totally right ;). At the beginning I've added this check to many > methods (which is not necessary at all) and because the generation of the > StackTrace is a very expensive operation, some pages needed long time to get > generated (about 1-2 seconds) (The check method was called more than a few > thousand times). > Heh... I knew all those years toying with chess engines would pay off somehow - your brain gets wired to search for expensive code. ;p This is also why logging frameworks like Log4j have big warning signs next to the output options that rely on the stack trace (e.g. to print out the method name, fully-qualified class name, line number). Those stack traces are best served in moderation. Then again, nothing exceeds like excess. ;) > I've ended up in just adding the annotations to methods, which did not get > called thousand times per request ;). Mostly I am checking the access rights > via the annotations on bean creation (e.g.: It does not make sense to allow > a customer to instantiate an administration bean). > > But you can easily pass the current method as parameter to the check method, > so no stack trace has to be generated, which is solving the performance > problem anyway. > Really? The examples I have seen all use the stack trace in some way. How are you accessing the current method? Regards, Jeff Bischoff Kenneth L Kurz & Associates, Inc.
