Developing Open Source Software Libraries

As I started reinventing the wheel in Clojure (writing a few libraries for backend web development), I actively became aware that I should think about the philosophy behind the libraries. By that, I mean the general direction and goals of the development process, something that should guide and be the goal of all kinds of software libraries 1.

I don’t think enough is written about the subject of creating software libraries, even though it might be the most popular artifact on github. It is simply expected that the thoughts are present when the code is written and that the result follows.

So here am I, trying to change that 2.

The Characteristics of a Good Software Library

A software library is a reusable piece of software that someone is going to import into a software project. We use libraries because it saves us the trouble of writing code that can be abstracted by a third party and, thus, be reused.

I will be describing sets of characteristics that libraries should have to be considered “good” 3.

User Developers and Programming Developers

I begin by making a distinction between the two extremes of developers. It is by these points of view that we shall focus.

Programming Developers should be viewed as (mainly) a superset of User Developers.

From a User Developer’s POV

When searching for a library, users will be interested in the one that solves the problem and delivers the most bang for the buck. Remember that the goal is to have the user not write the code for the functionality that the library is supposed to provide.

As such, the four characteristics that I deem as the most meaningful are:

Remember, most people using your library will fall into this category. Attack these points and you’ll make a lot of people happy.

From a Programming Developer’s POV

A programming developer will always dive into the code. They are interested in knowing how things work. So how can we help them with a smooth dive?

In practice, it’s all about following good programming practices. A software library is just like any other medium/large software artifact. One should aim for a codebase that is easy to read and understand and easy to change. Apply modularity, low coupling, high cohesion, comment your code, write documentation - the whole shebang.

While commenting your code helps, when trying to understand a new codebase, it helps a lot to have architecture documentation, i.e., something explaining how the pieces are tied together. This is something most open source software projects fail at, perhaps understandably, due to the boring nature of writing documentation.

So the best way to help programming developers is to write good code and documentation. Worried? Don’t be.

Nature vs Nurture

I don’t think that it should be expected for a software library to be “good” at v0.1 or < ~100 commits 5. A software library might not be born good, but it can surely grow into a good library. And that’s what we should strive for.

Conception date: June 1st, 2014

  1. While I was unconsciously aware of the requirements and goals, this passive kind of thinking is fuzzy and can get dilluted along the way. 

  2. Even though there are way more qualified people to do that. 

  3. This doesn’t mean that libraries that fail to satisfy these characteristics are not useful. 

  4. I am actually curious and interested in seeing at cases that are opposite of the UNIX way. 

  5. Why 100? Sounds like a reasonable number. YMMV. 

comments powered by Disqus