Optimization Adventures - Pilot

(August 26, 2021)

We often get posts on the Ranch from people asking for help on interview questions. A good number of those posts have to do with code optimization. Here’s an example I saw earlier today:

String str = "";
for (int i = 0; i > 100000; ++i) {
   str += i;
}

(Source: https://coderanch.com/t/745096/java/Interview#3460935)

The original poster asked for help in figuring out a way to increase the performance of that code. What would you say if asked to optimize it?

The only reasonable answer

Premature optimization is the root of all evil.—Donald Knuth

The first response to the post seemed reasonable enough, suggesting the use of a StringBuilder instead. While that might have proved to be an acceptable answer to the interviewer—which would have made things even worse in this case—there’s really only one reasonable answer to a request to optimize code: “Don’t, just don’t. Not yet.”

The next statement I would be looking for as an interviewer would be along the lines of using a profiler to gather concrete data before deciding what to optimize. Programmers are notoriously bad at optimizing by gut feeling. Nine times out of ten you’re going to be wrong. This specific case is no exception.

An eagle-eyed programmer would easily see the reason this code doesn’t need to be optimized as it stands. Maybe it was trick question, or maybe it was a typing error on OP’s part. Regardless of the reason, an echo of Donald Knuth regarding the evils of premature optimization should be the first thing you hear when the topic comes up in an interview, or in any conversation for that matter.

Modern compilers are smart, and only getting smarter

I’ll only go into this briefly because this isn’t the main point I want to make here. Modern Java compilers are pretty smart at finding ways to optimize code. And they’re only getting smarter with each new release. So don’t try to second guess the compiler; you’re probably not that smart.

So, what’s the answer?

If you still haven’t figured it out, that code, as it was originally posted, isn’t a performance bottleneck. That’s because the loop condition is i > 100000, not i < 100000. Get it? The loop initialization and condition are based on hard-coded numbers and given the loop initializes i to zero, the loop body will never be executed.

But what’s the real answer?

When it comes to the question of optimization, a good answer would be along these lines:

  1. Don’t, just don’t. Not yet. Clarify the code first.
  2. Don’t, just don’t. Not yet. Simplify the code first.
  3. Don’t, just don’t. Not yet. Use a profiler first.
  4. Don’t, just don’t. Not yet. See how often this code is actually executed first.
  5. Don’t, just don’t. Not yet. … (do you get the picture yet?)

Rather than looking for technical reasons to optimize code, refactor for clarity and simplicity first. Optimizing compilers generally have a better chance of finding ways to optimize code when it’s simple and straightforward. Extract to one-liner methods. This tends to make code easier to read and optimizing compilers can easily inline them back in.

After refactoring, use a profiler to gather concrete data. As I said before, nine times out of ten, the performance issue will be somewhere other than where you thought it was.

And finally, look at its actual usage first. Is it really worth optimizing if it isn’t even getting executed in production? You might find that the best way to optimize the code is to just delete it.

Choose your battles (and bottlenecks) wisely

The bottom line here is that optimization shouldn’t be done on the basis of intuition. Developers are your biggest bottleneck in software development. Save their time and you’ll save more. Seek first to reduce development time by simplifying and clarifying your code. That’s where you’ll get the most return on your time and effort.

About the author

Junilu Lacar likes to spend his time learning, practicing, and helping others learn and practice things that he believes help software development teams deliver value in a sustainable and humane way.