Is there business value in elegant code?

February 18, 2013 at 9:38 pm | Posted in Programming | Leave a comment

Recently I run into the following Java code (actual variable names replaced by x and y to anonymize):

if (x != 0) {
  y = 4 - x;
} else {
  y = 4;
}

So what’s the problem with this code? There is the use of the magic number 4, maybe we could test the x for equality to zero and swap the if and the else condition. And of course the whole test is nonsense and this 5 line code snippet could be just as well be replaced by:

  y = 4 - x;

But again, is there a problem with the original code? The code works as expected. The compiler will probably optimize this code anyhow and get rid of that test statement. The unit test and functional tests passed. So at least from a business (or end user if you will) point of view no problem at all.

And of course there are costs involved with refactoring the above code. If this happens during active development costs are probably quite low and the benefits will outweigh those costs. But if this is a maintenance project the situation might be different. The actual code is already in production and might be mission critical for the company. And if there is no (fully automated) continuous delivery in place the cost of this seemingly small code improvement can be one or two orders of magnitude higher than during development.

24093_old_measuring_instrument_for_navigation_isolated

Now lets try to make a business case for the above code change in a maintenance project. If you did an MBA you learned that business case = benefits – costs. An MBA is hardly more than that. Usually this formula results in overly optimistic hockeystick curves with an ROI of less than a year. But I digress. So we have to do two things here: we have to minimize the costs and maximize the benefits.

To start with the benefits. You can make a case that refactoring your code will usually result in less code and certainly in higher quality. If the code becomes more pleasant to work with it will also result in more motivated and productive developers instead of the attitude “I am not going to touch this mess that John left behind unless I really have to.” In general: cleaning up your code base and reducing it to half its size will leave you with about 1/4 of the maintenance costs. That might sound ambitious but over the last 20 years I have seen a lot of non-trivial pieces of software (ranging from 10k – 500k LOCs) for which this was no problem at all.

On the costs side of the equation you need a couple of things to minimize these costs:

  • Continuous delivery: automated building, unit testing, functional testing, integration testing, performance testing, version control, automated deployment, etc. Push hard to get these into place in your organisation. If you really can’t have continuous delivery on short notice, at least make sure that you have tests covering the code that you are going to change. Avoid the temptation to do ‘easy changes’ without them.
  • Any modern IDE will help for trivial refactoring. I call them trivial, because most if not all IDE’s only support refactoring on the syntax level. You can extract methods, rename variables, etc. But the IDE has no clue of what the code actually means. So refactoring on a semantic or design level is mostly not supported.
  • Automate your refactoring! This seems to contradict the previous point but there are tools available (usually as plug-ins for Java IDE’s) that can take refactoring one step further by looking at the AST (Abstract Syntax Tree) of your code and recognise patterns. I wouldn’t be surprised if they could detect and fix the example I started with.

In a next blogpost I will go into detail on what levels of refactoring there exist and how they could (at least theoretically) be automated. Until then:

Have fun refactoring your legacy code base!

(You might be wondering why I inserted the image of an old instrument in this post. My point is that in the past design of many things went far beyond pure functionality. For example scientific devices were often real pieces of art. I like code to be more than ‘just functional’ and have a certain elegancy about it.)

Leave a Comment »

RSS feed for comments on this post. TrackBack URI

Leave a comment

Create a free website or blog at WordPress.com.
Entries and comments feeds.