Computing with Data
Why should we care about code craftsmanship? The long answer is in this chapter. First, let's see what a French philosopher, by the name of Guillaume Ferrero, discovered: the Principle of Least Effort. He wrote about it in 1894 — long before a single line of computer code, as we know it today, was written. Nevertheless, we can't help but think that, had he been alive to witness how most code is written, he would have demonstrated coding as the epitome of the least effort principle:
[Coding] stops as soon as minimally acceptable results are found.
Sounds familiar? Well, we've all been there; once we find acceptable results, there's no immediate need to improve the current solution; we just want to get the task done, get that dose of instant gratification, and move on to the shiny new project. Putting out fires and plugging holes are good examples of tasks that we don't want to revisit once we found a good enough fix; and that notion is a perilous inclination, especially in the face of deadlines. It's also hard to resist the temptation of instant gratification; it takes discipline, self-control, and foresight to seek delayed gratification instead. A landmark psychological study, the Stanford marshmallow experiment, found a correlation between seeking delayed gratification and being more successful in life. I believe the same concept can be applied to code: We can stop at acceptable results and have a marshmallow to celebrate checking off a box that says we're done; or we can strive for more and have two marshmallows to celebrate the better results of going the extra mile.
But is completing a task as tempting as enjoying a sweet treat? You bet! It may even be more tempting than you think; thanks to dopamine: a neurotransmitter connected to certain feelings like reward (completing tasks) and pleasure (enjoying sweets). Usually, code requires multiple iterations to reach acceptable results; with each disappointing attempt, the level of dopamine that gets fired is reduced; when our code works — according to our acceptance criteria (test cases) — we get a pleasant surprise: increased levels of dopamine. But why are we inclined to seek a new task after finding minimally accepted results for the one at hand? According to Hilary Scarlett in Neuroscience for Organizational Change, "[dopamine] levels rise when we experience something new"; so it's no surprise that our brains seek more dopamine by making us want to move on to that shiny new task. Yet, experienced coders understand that building a successful and growing business is more about making sure existing solutions scale sustainably rather than creating new ones.
We should care for the code we inherited, and the code we are writing for others to inherit someday; we should exert more effort to keep on improving all code to reach a higher level of craftsmanship than what's minimally acceptable. Michael Feathers, author of Working Effectively with Legacy Code, eloquently stated:
Clean code always looks like it was written by someone who cares. There is nothing obvious that you can do to make it better.
In order to justify that extra effort though, we need to have a good reason; otherwise, our actions will fade away because we weren't inspired to fight the uphill battle against our natural tendency to be lazy (by sweeping the messy code under the rug). One good reason is that craftsmanship is vital for software to really thrive; we explain why in later sections of this chapter.
So how do we define software craftsmanship? This is a hard question to answer; one may borrow an analogy of hardware craftsmanship because it's tangible; a good example is that drawn by Walter Isaacson in his book about Steve Jobs:
Jobs had always indulged his obsession that the unseen parts of a product should be crafted as beautifully as its façade, just as his father had taught him when they were building a fence. This too he took to extremes when he found himself unfettered at NeXT. He made sure that the screws inside the machine had expensive plating. He even insisted that the matte black finish be coated onto the inside of the cube’s case, even though only repairmen would see it.
Similarly, software developers should pay attention to details that repairmen — other developers — will see when they read up the source code; reading code is much more frequent than writing it. Code is not only written for machines to execute, but also represents a means of communicating with the next person who will read your code (or your future self).
Besides code, big data represent a set of craftsmanship challenges that are shaped by the four V's of big data: Volume, Velocity, Variety, and Veracity. It's challenging to keep track and make sense of the colossal number of data points collected and processed at a mind-boggling throughput; the various data sets, formats, and sources we handle; and the high bar of quality that enables us to trust said data.
In this chapter, we explore ways to elevate the state of computing with data from minimally acceptable result to balancing craftsmanship and pragmatism.