Writing PureScript Bindings for a jQuery Plugin


You’ve started a new project at work with a tight deadline. You’re in charge of picking the stack and after assessing technologies best for the project you’ve decided to use PureScript on the front-end. One factor determining the success of the project is that it be delivered on time.

PureScript is the technology I chose for a new project at work and one of the ways it’s helping my company deliver our project on time is by allowing us to leverage existing JavaScript libraries using its foreign function interface (FFI), whilst still allowing us to exploit strong typing. The experience of writing FFI bindings has been so positive that I wanted to share with you how you might write your own bindings for an existing jQuery plugin.

In this post, together we’ll be writing bindings for a subset of the functionality of jQuery-steps: a lightweight wizard UI component written for jQuery. This won’t be a deep dive into PureScript’s FFI but will give you an insight into some of the practical considerations involved in using JavaScript code from PureScript. It assumes familiarity with JavaScript, PureScript and jQuery.


PureScript’s FFI allows communication from PureScript code to JavaScript code. It’s straight-forward and flexible, making it possible to write bindings for even the most involved JavaScript code. Provided, that is, that you understand the underlying runtime representation of what you’re dealing with.


As is typical for jQuery plugins, jQuery-steps takes a configuration object to tell it which DOM elements it should bind to. We begin by defining the types we’ll need to safely represent the fields of this configuration object in PureScript.

We use newtype wrappers to represent the values of the fields in our configuration object. We do this because newtypes have the same runtime representation as the underlying type, whilst being distinct from the perspective of the type checker. This makes newtypes an easy win for providing an extra layer of type safety when defining FFI bindings.

We use a type synonym to represent our configuration rather than an algebraic data type (ADT). Much like a newtype, the type synonym’s runtime representation will have a direct correspondence with what jQuery-steps expects, whereas an ADT’s constructors would present an extra level of indirection that we’d have to handle in our JavaScript wrapper.

We define a default configuration object to let consumers of our bindings use record update syntax to override the defaults we’ve set.

Here’s what defaultConfig will look like at runtime. You can see that despite adding an extra layer of type safety when working with our configuration object in PureScript, we don’t pay any extra cost for the abstraction at runtime. Practically speaking, this means that we can pass our configuration object directly to jQuery-steps.

We define our export function that wraps jQuery-steps by providing a curried interface. It’s worth noting that we’re using the FFI style introduced in PureScript 0.7.

And, finally, we write a foreign import declaration and we’re done. It’s really as simple as that. Well, almost.

Callbacks and Dishonesty

Binding to DOM elements is all well and good but very soon you’ll want to add dynamic behaviour such as client-side validation. This means delving into into the unprincipled world of JavaScript callbacks and exposing the plugin’s event handler API.

We begin by extending Config to include two new fields: onFinished and onFinishing. The type constructor Callback2 comes from purescript-foreign-callbacks and represents an effectful computation that takes two arguments, whilst the type constructor Fn3 comes from purescript-functions and represents a pure function that takes three arguments.

Something seems a bit odd here - why is onFinished effectful, whilst onFinishing is pure? Well, the truth is that they are both effectful, but since JavaScript offers no means of distinguishing between effectful and pure computations, we need to circumvent the type checker to ensure functions we pass to onFinishing have the correct runtime representation.

mkFinishing transforms our function into an appropriate runtime representation by uncurrying it and by unsafely pulling the inner value out of the Eff constructor. We uncurry our function because JavaScript functions aren’t curried by default and we pull the inner value out of Eff because jQuery-steps expects a callback that returns a boolean value.

If at this point you’re feeling a little uncomfortable, that’s OK. Once a value is inside Eff you shouldn’t really be pulling it back out without a very good reason. But we’re dealing with JavaScript code here and managing the impedance mismatch, so consumers of our bindings don’t have to, is probably reason enough. In any case, the callback function wouldn’t be very useful if it was pure!

mkFinished, on the other hand, is much simpler. It uses callback2 to take our effectful computation of two arguments and transforms it into an appropriate runtime representation for jQuery-steps.

And, finally, here’s what the resulting defaultConfig looks like. Now we’re really done. That wasn’t so bad, was it?


PureScript lets us easily reuse existing JavaScript libraries with help from its FFI. Being able to reuse code can help us prototype more rapidly by allowing us to stand on the shoulders of giants. Combining this property with a strongly typed language can help us when that rapid prototype needs to make its way into a production setting.