JavaScript: Productivity over Proof

James Fenn |

TypeScript Truthy and Falsey - James Padolsey Destructuring Assignment - MDN

I'll keep this blog post short, but this is in response to a conversation about testing and proving code written in JavaScript as opposed to more so-called "conventional languages" such as SML, Haskell, or OCaml.

The preface is that dynamically typed programs and languages are inherently fallible by nature, especially in a large codebase, as it makes it nearly impossible to create a "proof of correctness". Dynamic typing creates failures that can only be seen upon execution of the code, not compilation, allowing errors to pass through that might not be caught during testing.

Since scripting languages have no compilation phase, they rely entirely on tests and linting to enforce proper use and prevent errors, but loose typing makes this inherently difficult to do. In a language where 2 == "2", concepts such as referential transparency are impossible to attain. In C, this is equivalent to every function returning a void* with no static method of enforcing its type or that it even exists.

Why This is Okay

Now, there are several obvious rebuttals that I could make here, but I'll start with this: don't use JavaScript for something that must never fail. Different languages can be used for different things, and they're all built with different purposes in mind. Nobody would dream of using JavaScript to pilot an aircraft. If you really need a proof of correctness for everything you write, then maybe it isn't for you.

The focus of languages like JavaScript that explicitly use dynamic typing is not prevention of user error, but better polymorphism and prototype-based data - promoting code re-use. Its purpose is not to reduce errors for people that don't know how to use it, but to improve productivity and make it faster and easier to write for those that do. Just because you can't implement space shuttle style in a program doesn't mean you shouldn't write it. Not everything has to be stable - code changes over time, and you should write it in a way which enables that, not restricts it.

With proper documentation and source control, code reviews and test driven development can catch the issues that would be caused by dynamic typing before they reach production. And in the event that they do, then - in a properly documented and organized codebase - the functionality can be changed and fixed quickly without causing any problems or requiring a rewrite of the entire codebase which one might experience in a language that is statically typed.

Dynamic typing doesn't just bring languages a lack of stability; it allows for useful functionality as well, which might not otherwise be possible:

  • The absence of a required compilation stage means that tests can be run as soon as your code is written, or even while you're writing it.

  • The lack of verbosity brought by dynamic typing means that common functionality can be written quickly and without context. For example: truthy values can be used to quickly determine whether something exists, regardless of what it is. The integer 0 is false, and so is an empty string. However, even though 0 == '0', the string '0' is a truthy value and will evaluate to true.

  • "Destructuring assignment" is also a cool example of the advantages that prototype-based data and a lack of verbosity can bring:

    let [a, b, c = 4, ...d] = [1,2, ,4,5];
    // a = 1, b = 2, c = 4, d = [4, 5]
    

Languages all have their own pros and cons, and "absolute stability" isn't always a good thing. JavaScript is fairly good at what it is used for, and its widespread adoption means that it has a well-developed community behind it and is easy to learn and use as a result. The fact that you can't write a formal inductive proof of a JavaScript function just goes to show why most programmers aren't mathematicians (and don't need to be).

Workarounds

JavaScript's "loose equality" and type-converting comparisons are confusing at times, but there is a clear option to just not use them. In most cases, loose equalities (2 == 2) can be replaced by strict comparisons (2 === 2) that only evaluate if the operands are of the same type.

If you need to use JavaScript, but still want enforced typing, check out the TypeScript project - it's essentially a typed version of JavaScript that can compile to vanilla JS, so you can get the best of both worlds.