I excluded "being ridiculously wasteful or using poorly designed algorithms". Of course, we need to be careful but not in the way we used to fret over optimal coding. When specifying an for-loop counter, how many people load the size of an ArrayList into a local variable rather than call size() each iteration (I do but most people don't).
----- Original Message ----- From: "Dan Armbrust" <[EMAIL PROTECTED]> To: <axis-user@ws.apache.org> Sent: Thursday, June 02, 2005 5:47 PM Subject: Re: web service error handling design issue > > > > > >programmers no longer need to worry about performance. In fact, we stopped > >worrying about performance once we abandoned C++ for Java! > > > > Speak for yourself... Maybe you work in a world where the datasets are > small, and hardware budgets grow quickly.... > > I remember digging into a third party package that we were using - they > had recently ported from c to java, and the performance of the new > version sucked on certain operations. Examination of the code revealed > that there was one method that got its result by opening a random file > reader on a 200K file, and finding the appropriate bits. The algorithm > called this method several hundred times per operation. A simple change > to a hashtable based approach with a one-time read of the file resulted > in an operation that ran about 1000 times faster. Multiple that by the > 500,000,000 times I needed to call that method, and we start talking > years instead of hours to process the same data.... > > Dan > > > > >
