The truth has been staring you in the face. You’ve put off dealing with it for too long, afraid of what you’ll discover. All the symptoms are there: the continued unexplainable problems, the mood swings, the fatigue, the headaches, and the exhaustion when faced with simple tasks.
It’s time to face the facts: your app has cancer.
It didn’t start this way, of course. When the project began your team was moving swiftly and silently. Morale was high and so was the velocity. But, over time, things started to change: more and more bugs crept in and remained, velocity slowed, and developer morale tanked. To make matters worse, your team spends more time fixing old features than adding new ones.
It could be that this scenario doesn’t describe your project and your situation. Or maybe it just doesn’t describe it yet. Wherever you are in your project’s lifecycle, you need to be aware of the symptoms, causes, and dangers of cancer in you software.
Unlike real life cancer, there isn’t a blood test or biopsy which can be performed to easily diagnose cancer in an application. You can’t put your code under a microscope to see what ails it. It’s only by paying attention to the symptoms that you will ever know how sick your software is.
Coined by Kent Beck while helping Martin Fowler write “Refactoring,” the term “code smell” refers to certain identifiable patterns in code which are “…a surface indication that usually corresponds to a deeper problem in the system” (CodeSmell).
Patterns such as duplicated code, comments, conditional complexity, and uncommunicative names are all examples of code smells. Individually, code smells don’t guarantee a deeper underlying problem, but collectively code smell indicates something rotten in the product’s development.
One of the most obvious symptoms of software cancer is an increasing number of bugs. Not just an increase, but a persistence of bugs. These persistent bugs can’t be squashed and more bugs arise in their place just as soon as others are put down.
The primary cause of these types of bugs are classes, methods, and functions which try to take on too much responsibility. These blocks of code are easy to spot: they take up more than one screen and are filled with indented code inside indented code. The scary thing is a lot of this code probably started off tight. But, as changes in requirements came in and deadlines loomed shortcuts were taken and pretty soon everything was a right mess.
Now, because the code is responsible for so much, any change, any flaw, affects every other piece of code that’s dependent on it. This is why bugs seem to appear out of completely unrelated blocks of code. What once should have taken hours to build or fix now takes days or weeks.
Computers are amazing at following deeply nested logic structures, but humans aren’t. In fact, we kinda suck at it. For a programmer to debug a deeply nested conditional like what was mentioned previously, and which can receive any number of inputs, he or she will likely fall back to drawing out the flow on paper, trying to trace the route of the logic. It takes a lot of time and mental energy to not only trace the logic, but keep as much of it in your head as possible.
The result? Instead of adding new features or improving efficiency of the application, developers burn hours just trying to understand the code and keeping the system running.
As you can imagine, the previous scenario isn’t a lot of fun. In fact, unless you’re one of the rare developers who actively enjoys working with legacy code, you’ll probably be miserable.
Decreased morale means decreased productivity followed shortly by a decrease in staff. The best developers will leave first, followed by everyone else. This leaves the developers who either can’t find another job or don’t care enough to. These will be the ones maintaining the system.
Why does cancer grow in a healthy app? As with cancer in the human body, there is rarely one single cause. For humans, it may be a combination of environmental factors, stress, and what we put into our bodies. For software, however, the reasons are quite different.
Much like craftsmen in other creative fields, programmers have developed collections of best practices for programming in specific languages, their use of specific tools, deployment, source code management, and more.
Best practices are “procedures which have been shown by research and experience to produce optimal results”. Test Driven Development (TDD), Continuous Integration (CI), rigorous refactoring, and code reviews are just a handful of practices which enable the production of superior software.
As with other creative fields, when programmers stray from best practices the resulting code and project will contain the artifacts of that deviation. These artifacts, while not visible to anyone other than developers, are “felt” by the users of the software in application errors, slow application performance, and longer release cycles.
Another cause of cancerous software is often due to nothing more than inexperience. Junior developers, even those who are exceptionally skilled, lack the necessary experience which guides more seasoned developers. This experience is knowledge only attainable through doing the work and choosing to learn from it.
Unfortunately, unless there are senior developers on a team who can provide some level of mentoring, the inexperienced developer will primarily learn through mistakes and at the expense of the project.
While inexperienced developers can wreak their own level of havoc on a project, potentially more damaging to a project is the influence of inexperienced management. Inexperienced managers and project managers can seriously reduce the effectiveness of even the most seasoned developer by involving them in too many unnecessary meetings, not providing enough of a buffer between development and the business, not providing clear direction, and being indecisive in the decisions.
Whereas inexperience leads developers to make poor decisions out of ignorance, an apathetic developer makes poor programming decisions because he or she stopped caring. There are a number of reasons which can lead to this:
Apathy fuels the cancer so it affects not only software and development staff, but every area of your business. It should dealt with as soon as it’s discovered.
Deadlines and budget constraints are very real factors facing any organization. The have an especially strong impact on software development. As pressures mount, different organizations respond differently: some respond by increasing efforts and hours worked, while others seek to take shortcuts. The software produced in either case is usually saddled with technical debt.
Like a financial debt, the technical debt incurs interest payments, which come in the form of the extra effort that we have to do in future development because of the quick and dirty design choice - Martin Fowler
As with financial debt, there is interest associated with technical debt revealing itself as reduced team productivity and lower software performance. Technical debt is a byproduct of software cancer. Continuing to incur technical debt only exacerbates these issues and feeds the cancer. Eventually like financial debt, the debt must either be paid or the project liquidated.
In 1967, Melvin Conway submitted a paper, “How Do Committees Invent?” to the Harvard Business Review. His thesis was, “Any organization that designs a system (defined broadly) will produce a design whose structure is a copy of the organization’s communication structure.”
This can be applied to many types of systems and processes within a company, but software, more than anything else, is the most susceptible to an organization’s communication structure. Organizations with very siloed groups tend to have software which doesn’t interact well with other systems. Those which tend only to make decisions in meetings have software which is bloated and overly complex. Companies with a very top-down communication structure tend to have very tightly-coupled, inflexible code bases.
Many of the projects I’ve been called in to rescue were in the state they were in because of how those organizations communicated. Discovering these issues not only allowed me to rescue the projects, but also showed my customers how they could correct other, more foundational issues, by changing how they communicated.
In 1982, a criminological theory dubbed the “broken windows theory” was introduced. The authors of the theory concluded that when communities are maintained and cared for, crime can’t gain a foothold. Their example: even the simple act of not repairing a broken window can lead to an increase in vandalism and crime.
The theory is not without its detractors, but in the software industry it’s recognized as an inevitability:
One broken window—a badly designed piece of code, a poor management decision that the team must live with for the duration of the project—is all it takes to start the decline. If you find yourself working on a project with quite a few broken windows, it’s all too easy to slip into the mindset of “All the rest of this code is crap, I’ll just follow suit.” It doesn’t matter if the project has been fine up to this point. — Andy Hunt and David Thomas, “The Pragmatic Programmer”
Bad code is exactly like a malignant tumor: it grows and spreads. And while bad code spreads, it becomes increasingly more complex and can couple itself to areas outside its purview.
Complexity in a software project is par for the course. Every new feature means the business logic must necessarily get more complicated, but that doesn’t mean the underlying code has to be harder to understand. You can have very complicated business logic described by easy to understand programming.
When we allow bad code to gain a foothold, the added complexity only makes understanding it more difficult. The computer doesn’t have a problem with it – although it may perform slower – but people aren’t designed to keep large amounts of complicated logic in their head at one time. We perform much better with smaller, simpler instruction sets.
The primary danger of complex code then, isn’t the code itself, but the programmer’s ability to correctly understand all the necessary logic. The more logic that needs to to be understood, the greater likelihood that defects and increased complexities are introduced. This results in even greater complexity, further increasing the difficulty of understanding the logic, and further increasing the likelihood of introducing more errors.
The cure many companies take to combat the ever-increasing number of bugs and decreasing productivity of the programmers is to throw more programmers – or worse, managers – at the problem. Too bad that’s the wrong solution.
Different methods are employed to treat cancer in people. Sometimes a single type of therapy is required, but oftentimes multiple methods are used together to put cancer into remission. Medical treatments are not the only weapons used to combat cancer. Patients also need support from friends and family, maintain a healthy lifestyle, and keep a positive attitude. Cancer patients have the most success when they combat their disease holistically.
Treating cancer in software is no different. To effectively treat cancer in your software, a holistic approach must be taken. Correcting the most troubling areas in the code can provide temporary relief. Unless the problems which led to the issues are addressed, it’s only a matter of time before the cancer begins to grow and spread again.
If you’re serious about tackling the issues plaguing your software projects, the most important thing you can do is support your development staff. You must trust them to not only perform their job to the best of their ability, but also trust their recommendations.
To non-programmers, programming can seem very intimidating. There are funny looking characters, weird terminology, an endless number of abbreviations, and it just looks hard. This fear of the unknown has led to some seriously wrong conclusions about programming:
Programmers are not the 21st century’s line workers. They can’t be swapped in and out as if they are all capable of stamping out the same code. Just as in other fields such as law, art, medicine, and management, there are different levels of skill and ability among programmers. That ability level makes an enormous difference in the quality of software produced.
When your development staff pushes to rework a section of an app, to add automated tests, or stop development for a period of time to upgrade libraries, it is not out of a desire to slack off, to navel-gaze, or be a burden to the budget. Programmers want to produce an exceptional product for the organization. They need your permission and support to do what they need to do.
Men and months are interchangeable commodities only when a task can be partitioned among many workers with no communication among them. This is true of reaping wheat or picking cotton; it is not even approximately true of systems programming. — Frederick P. Brooks “The Mythical Man-Month”
Trusting your development staff that they know how to solve the problem is like admitting you’re sick in the first place: it doesn’t cure the disease, but it allows treatment to begin. To address a cancerous project, the problematic code needs to be refactored.
Refactoring (noun): a change made to the internal structure of software to make it easier to understand and cheaper to modify without changing its observable behavior. — Martin Fowler, “Refactoring”
When many people hear this, they balk, asking, “What’s the point of changing the underlying code if the app’s behavior doesn’t change?” These people see refactoring only as an expense, but during a field study performed at Microsoft on the benefits of refactoring, one developer asked in response:
The value of refactoring is difficult to measure. How do you measure the value of a bug that never existed, or the time saved on a later undetermined feature? How does this value bubble up to management? Because there’s no way to place immediate value on the practice of refactoring, it makes it difficult to justify to management. — “A Field Study of Refactoring Challenges and Benefits”
In this study, developers reported the following results:
On the surface, refactoring appears to be an expense, but based on these results, it’s easy to see that it’s an expense providing a major return (and healing).
A further benefit provided by refactoring, and one which is even more difficult to measure, is improved developer morale. Developers don’t enjoy writing poor code, but when deadlines are moved up, and new features are added to the project list, it’s the hidden mandate. Giving developers time to go back and fix the shortcuts, workarounds, and hacks is immeasurably satisfying to a programmer. It’s giving them permission to do things the right way.
It is uncommon for a person to discover they have cancer without any warning signs. There are usually symptoms which precede the eventual diagnosis. A person can tell when something’s wrong with their body, because their nervous system informs them of the pain or discomfort. Without a nervous system, we would never know we are sick or injured without seeing it or being told by someone else.
Automated tests perform the same role for software as the nervous system performs in a person; they alert the programmer to problems in the code.
It’s not uncommon for development shops to include a testing department to ensure new features perform correctly and don’t break existing functionality. However, it is exceedingly difficult for those departments to systematically test every aspect of an app every time a change is made to the underlying code. Automated testing can do that.
An automated test is code written to ensure its product code counterpart always behaves the same way. When it doesn’t, the test suite alerts the programmer of the error. Unlike manual testing, automated testing can perform thousands of tests in seconds allowing the development staff to know immediately if any change or refactor affects any other part of the product. It really is like giving your application a nervous system.
The idea of rewriting an application is popular among programmers, especially those who’ve never done it before. The idea is to take all good business logic from the old application and create a new application from it. The new version can be architected better so that all the technical debt, bugs, deprecated functionality, and outdated thinking is left behind. That’s the idea. Unfortunately, the idea usually falls far short of reality.
Joel Spolsky gave four reasons why you should never rewrite software from scratch:
I am not a proponent of rewriting applications. There are exceptions, of course (we’ll get to those later), but in my twenty years of programming, I’ve never seen a major rewrite keep the promises which were made about it.
With that said, here are my recommendations for when to rewrite a piece of software:
Software rewrites almost never address the real issues. It’s like giving a chain smoker a lung transplant to cure his cancer.
For cancer survivors, being in remission “means that tests, physical exams, and scans show that all signs of your cancer are gone.” It’s different from being cured, because there is a possibility that the patient can experience a recurrence (i.e., the cancer comes back).
Even after cancer has been purged from an application, it too can experience a recurrence. Unlike real cancer, however, there are steps both management and developers can take to ensure the cancer stays cured.
If you’re concerned about increasing the quality of your software, the simplest solution is to improve the quality of your development staff. The simplest way to do that is training.
Training has changed a lot over the past couple of decades. There was a time when you’d have to actually leave the office and attend a class in person. Today, most training can be accomplished through webinars, screencasts, e-learning, and web-based training. Some prefer to learn through reading and experimentation; that should be encouraged as well.
The advantages of training developers extend beyond the benefits seen in the quality of the work they produce. When they know there is an actual interest in their growth, it can spur them on to excel.
Our chief want is someone who will inspire us to be what we know we could be. – Ralph Waldo Emerson
The following is just a sample of the benefits experienced through an intentional training program for your programming staff:
Yes, increased ability and skill matter, but not nearly as much as having motivated and driven employees who want to be working on your project, and who want to produce their best work.
Another solution to slowing the recurrence of cancer is performing regular code reviews. Code reviews are very much like peer reviews in the scientific community. There are a number of different times when they can be performed, but in general, a code review should be performed before any branch or feature is allowed to be merged into the master branch. During the process the reviewer looks for logic errors and simple improvements; ensures the code matches the organization’s style and meets the expected level of quality; and, of course, ensures the code actually works.
In his book, “Code Complete,” Steve McConnell provides the results from numerous case studies highlighting the effectiveness of code reviews.
… software testing alone has limited effectiveness – the average defect detection rate is only 25 percent for unit testing, 35 percent for function testing, and 45 percent for integration testing. In contrast, the average effectiveness of design and code inspections are 55 and 60 percent. Case studies of review results have been impressive:
- In a software-maintenance organization, 55 percent of one-line maintenance changes were in error before code reviews were introduced. After reviews were introduced, only 2 percent of the changes were in error. When all changes were considered, 95 percent were correct the first time after reviews were introduced. Before reviews were introduced, under 20 percent were correct the first time.
- In a group of 11 programs developed by the same group of people, the first 5 were developed without reviews. The remaining 6 were developed with reviews. After all the programs were released to production, the first 5 had an average of 4.5 errors per 100 lines of code. The 6 that had been inspected had an average of only 0.82 errors per 100. Reviews cut the errors by over 80 percent.
- The Aetna Insurance Company found 82 percent of the errors in a program by using inspections and was able to decrease its development resources by 20 percent.
- IBM’s 500,000 line Orbit project used 11 levels of inspections. It was delivered early and had only about 1 percent of the errors that would normally be expected.
- A study of an organization at AT&T with more than 200 people reported a 14 percent increase in productivity and a 90 percent decrease in defects after the organization introduced reviews.
- Jet Propulsion Laboratories estimates that it saves about $25,000 per inspection by finding and fixing defects at an early stage.
h/t: Jeff Atwood - Code Reviews: Just Do It
But beyond mere financial and quality gains, consider these other important benefits to code reviews:
In the end, ensuring cancer remains in remission requires vigilance, a willingness to change, and commitment. It’s not a war to be won, but one which is continually fought. Cancer can grow in an app regardless of how old it is. I’ve rescued apps which began life diseased, but I’ve also seen ancient software (i.e., more than three years old) which was the picture of health. This all depends on the team developing the app, and their commitment to making great software.
Be alert for symptoms when they arise: code smells, increasing numbers of defects, and decreased productivity and morale among your staff. These are all signs that your software may no longer be in remission.
Remember to adhere to best practices such as Test Driven Development (TDD) and code reviews. Be sure to practice refactorings as a part of your process. These three practices, more than anything, will act as antibodies to any cancer which could start to grow in your app.
Finally, trust your developers to do what is right for the success and longevity of the product. Encourage their ownership of the software by supporting code reviews and considering their recommendations. No programmer wants to work in convoluted and confusing code, and unless they’ve been regularly shut down in the past, will seek solutions to make the code base something they’re proud of.
Not every organization has the resources to properly fight the cancer which exists in their software. In those cases, it makes sense to hire outside consultants who specialize in repairing legacy software. I pride myself on rehabilitating diseased and ailing software, and bringing new life and value to dying projects.
“Cancer” is an ugly word and it’s an even uglier disease. In software, cancer results in bugs, slower development times, reduced performance, and lower profits. It can spread throughout a system, causing major delays, low morale, and even result in employee turnover. In people, cancer is even worse.
Every day, more than 1,500 people die of cancer in the US alone. That’s 1,500 sons, daughters, mothers, fathers, friends, and family. It’s 1,500 what might have beens, and 1,500 lives which will never shine their light again.
It happens every day.