Matt's anti patterns
_A while back I had the pleasure of working with an occasionally grumpy developer. Nothing frustrated Matt as much as badly written code. He used to sit at his desk and from time to time you could hear ‘WTF?!’, ‘What were they thinking?!’, ‘Who wrote this?!!’ and ‘What the hell is this doing?’. I enjoyed pair programming with Matt, not only because he was an excellent developer, but also because he used to come with the funniest terms for how not to do things. Every time he came up with a new phrase for an anti pattern, I opened a text file and wrote it down. Here is some of the stuff he came up with.
The Ultimate Unit Test of Doom
We have all seen this many times. It’s surprising how often developers manage to fall into this trap. When you come across a unit test starting with an initialisation of a spring context file or even connecting to a database, and then doing 10s of assertions in one test case, you know you’ve come across the Ultimate Unit Test of Doom --insert dramatic music here--
Such tests usually contain only one test case and it might span a few hundred lines of code. It might check if the stored proc exists and then call a service method and expect a table in the database to be populated. It is testing a whole subsystem instead of a unit of behaviour. The next code snippet shows one such example
_The evangelists of unit testing preach that unit testing is not about testing at all. Unit testing is all about specifying behaviour. When you’re writing your unit test if you’re asking yourself ‘Is the system working as it should?’ you know you’re going down the route of the ultimate unit test of doom. Instead you should be asking yourself questions such as ‘If I put these inputs in this interface, once I have an implementation for it, it really ought to be returning me these kind of outputs’. More than a testing framework, unit testing is a thought process which makes you think ahead about how to build the small units of your software and how they interact together.
Pass the Banana
We came across this anti pattern when we were trying to understand how a system property was propagated through the system. The ‘Banana’ was a boolean flag that was passed as a startup main argument. The flag was passed from one method to the other, from class to class until it got read and used at the bottom of the stack. The following code illustrates an example of a ‘Pass the Banana’.
_Most developers would see the above and ask ‘Why isn’t startThreadPool a system property and whatever class needs to access it can just read it directly?’. The thing is, such bad practises aren’t usually written by just one person. They usually evolve over time through incremental changes by different developers. The above example most likely started with simple implementation where the main method started the thread pool directly. As the system got more complex and larger, the call to start the thread pool moved further down the layers. Each developer kept passing the banana every time he added a new class or method.
Solving these kind of issues and keeping code complexity down is not an easy task especially when you have a large team and tight schedules. Allocating time for refactoring and improving code quality in each sprint is key. A code analyser tool helps, but its not enough. It also comes down to each developer’s attitude and willingness to improve code quality.
If each developer took one step in the right direction, instead of a ‘pass the banana’ pattern, it would he the ‘happy hour’ pattern... The one where developers are down the pub instead of debugging horrible code.
Martin Fowler wrote a good article on opportunistic refactoring at: http://martinfowler.com/bliki/OpportunisticRefactoring.html
The Monolithic method of Death
This is quite simple to explain. It’s a method that takes 7 arguments or more, is called ‘process’ or ‘execute’,is about 200 or more lines of code and has horrible multiple nested ifs and loops. You know what I mean... we’ve all come across one of these..._
_In my experience people underestimate the impact on productivity of bad code and tech dept. It’s also very hard to explain to non technical business people that a part of the system needs redeveloping. Why should I pay you to write code that is going to do exactly what the current code already does? It is a very valid argument. In an ideal world it should never get to at point where a system needs to be reimplemented. Coding should be an evolution. Small incremental improvements that make the code more modular, readable and easy to scale. Unfortunately this is hardly ever the case. Schedules are tight, pressure builds up and corners are cut. Again there is no technical fix for this. It all boils down to methodology and individual attitudes and some of the team dynamics. Are the correct development practices being followed? Is there enough time in each sprint for refactoring and reducing tech debt? Is there good team communication?
Poo in a File
This is not really an anti pattern. It’s more something Matt did and made some people laugh. One morning I updated my code from subversion and I noticed a new file called 'pooinafile.txt'. When I opened this I found the text 'poo' in it.
It turned out that Matt was testing why his bit of code wasn't reading from the file system due to some permission issues. He had created the file and committed by mistake. Since then we kept on using the phrase every time someone committed a file that didn't belong in the repository. Matt, wish you had named your file something a bit more exciting...
6/6/2019 04:52:55 pm
Hello James, I took your class from udemy.com for codility exams for java! Thanks for your time making the videos are very helpful. I enter to your website to see the stack and queue problem but I wanted to check a little more. I am in the "process" method to become a java developer at Telus International but I need to pass the tests. Thanks!
6/7/2019 03:21:01 am
Thank you Diego! Enjoy the course and good luck for test!
Leave a Reply.
James Cutajar is a software developer, with interests in high performance computing, algorithms design and distributed data structures.