June 21, 2016software engineering complexity abstractions
For as long as I’ve been a developer I have been reminded of the rule of three for refactoring and every time I hear it, I die a little inside.
If building a new abstraction is so painful, that it is easier to copy and paste large chunks of code, you need to rethink the way you are building your abstractions. Building the right abstractions isn’t just a time consuming distraction from building software, it is software engineering.
With that out of the way, let me try and explain how I decide what to abstract so that years from now when I’m older some obnoxious young know it all can tell me how I’ve been doing it wrong :-)
Before I do that, I’d like to first try and paint a picture of a software as something that has a life of its own. When it is born, it is only a small cluster of abstract concepts that fit nicely into a solution for the user. But the days of clear cut simplicity are few, as the application grows it is fed features, design changes and bug fixes. Depending on the process your team used to digest all your application’s food, you can end up with a beauty or a beast.
Now let’s pair back this analogy to something more mathematical that we can better understand as engineers, let’s think of our application as a growing crystal. Your core abstractions sit on the inside, and all the features, changes and fixes it is fed cause new abstractions to crystalize around the old ones.
Each addition to the crystal should only contain the thing that makes that abstraction unique, and for the parts that are not unique, it should attach itself to the crystal on which it grows.
What do I mean by attach itself? Well, say I have a user info form and a user info dialog. If I’m following the rule of three, I would just copy the form and be on with my day. Then, two weeks later, developer John Smith joins the project and has no clue about the user info forms. He gets a story to add a user info sidebar, so he looks for a user info form and all he has to do is copy it into the sidebar and presto chango! Meanwhile, it is time to add a field to the user info form (it is required, sorry). The change was easy, but with it, we just broke a dialog and sidebar.
If you were to attach it to the abstractions that are already there, you would compose the form into your new dialog abstraction, and then two weeks later, when John Smith joins the project, you tell him, we go by this philosophy and you link him here. Then, when he needs to add the form to the sidebar, he knows that copy pasting is not okay, and he builds a sidebar component that composes the form. The time comes to add the field to the form, you don’t even need to think about anything on the crystal further out than your abstraction because as long as you are meeting the basic expectations of all the users of your abstraction, all of your changes should successfully propagate throughout your app without you having to break a sweat.
Using this technique, so long as you maintain clear interfaces for using your abstraction, when you make a change to it, you only need to worry about the part of your crystal that is directly adjacent towards the inside. This allows you to think in the context of the abstraction you are working on and not have to worry about fringe complexity that was added in after the fact.