This year, the 9th of April marked two years since I started writing this blog. Since I missed the occasion, I want to celebrate it a little late and write about three interesting reactions that I saw some of my readers have (a FAQ post, if you wish). The comments/questions were more diverse, but these are basically what it boils down to.
By the way, the title is a quote from a brilliant movie and is what I usually think about when I hear programmers complaining about object orientation.
I never really knew how to call the first reaction, so let’s call it “habit” or “laziness”. They say “that’s not how it’s done” most of the times. To them, I say simply: there is no way of doing it, there is no cook book for OOP. When it comes to object oriented programming we only have a few guidelines, some design patterns and some silly interview questions.
When I hear the above complaint, I always hope they mean to say that you cannot incorporate the presented ideas into existing frameworks very easily. Indeed, the challenge is big, but who says you can’t design a better alternative, a more object-oriented one? Never be easily impressed, always question the big players (especially them) and their solutions! Don’t forget that the coolest platforms that you are using every day are written by developers who are just as bored as you are and not necessarily smarter or more experienced.
The second reaction is they start yelling about design principles, most commonly SRP, as if their life depended on it. They seem to forget that SRP applies to classes, not to objects. That is, a class can very well respect it, while an object, which is composed/decorated of many classes, can behave in many ways. Just because there is an interface, it doesn’t mean that you will find everything behind it – not at all, the abstraction should go further, becoming concrete only gradually.
Furthermore, I understand the Single Responsibility Principle by the definition: “A class should have only one reason to change”. Taking it literally as “just one responsibility” only leads to more chaos: what is one responsibility? Any action/concern, no matter how big or small, can be broken down. I like to think of it more as “Single Domain Principle” – that is, if the class doesn’t go out of its domain of expertise, it’s fine. As an example, I would expect
Students to have methods
worst() and others related to a class of students, and I would also expect it to be an
Iterable<Student>, so I can see exactly who are the pupils in my class (pay attention: the iteration logic wouldn’t be in the
Students implementation(s), it would be in the
Iterator returned by the
Nevertheless, what I find funny is that everybody seems to throw design principles at you, but nobody seems to complain about the fact that JDK interfaces/classes such as
File are offering dozens of methods.
The third reaction seems to be fear of abstraction: “You cannot hide that”, “what if this, or that?” etc. Especially when it comes to network access: quite a few complained about hiding stuff such as HTTP calls inside Iterators. Their arguments are usually well founded but they are all in the lines of this (really good) article.
Taking two examples from that article, even though TCP/IP and our cars are not perfect, we’re still using them because the advantages they offer are much bigger than their drawbacks. In the meantime, engineers out there are working to improve them while maintaining abstractions. Avoiding encapsulation in OOP is like telling those engineers to limit the cars’ speed to 50 km/h in order to avoid accidents or like telling your friends to stop using the Internet and use land mail because, that way, there will surely be no lag and the messages will be guaranteed to arrive intact.
To summarize, all I have to say to my lazy/concerned/scared readers is the following: it’s our job to move things forward, hide as many concerns from the user as possible! It’s our job, as engineers, to understand, criticize and improve the tools and principles we are using.
I don’t know about you, but I want to investigate new ideas and design cool, automated solutions, not just sit in an office and play with what others figured might be useful in solving “business issues” fast. I suppose you want the same thing, so I urge you to be curious, be fearless, hate comfort and always ask questions!
Use syntax highlighting in your comments, to make them more readable.