RescriptReact comes with a router! We've leveraged the language and library features in order to create a router that's:

  • The simplest, thinnest possible.

  • Easily pluggable anywhere into your existing code.

  • Performant and tiny.

How does it work?

The available methods are listed here:

  • RescriptReactRouter.push(string): takes a new path and update the URL.

  • RescriptReactRouter.replace(string): like push, but replaces the current URL.

  • RescriptReactRouter.watchUrl(f): start watching for URL changes. Returns a subscription token. Upon url change, calls the callback and passes it the RescriptReactRouter.url record.

  • RescriptReactRouter.unwatchUrl(watcherID): stop watching for URL changes.

  • RescriptReactRouter.dangerouslyGetInitialUrl(): get url record outside of watchUrl. Described later.

  • RescriptReactRouter.useUrl(~serverUrl): returns the url record inside a component.

If you want to know more about the low level details on how the router interface is implemented, refer to the RescriptReactRouter implementation.

Match a Route

There's no API! watchUrl gives you back a url record of the following shape:

ReScriptJS Output
type url = {
  /* path takes window.location.pathname, like "/book/title/edit" and turns it into `list{"book", "title", "edit"}` */
  path: list<string>,
  /* the url's hash, if any. The # symbol is stripped out for you */
  hash: string,
  /* the url's query params, if any. The ? symbol is stripped out for you */
  search: string

So the url is given back as:

ReScriptJS Output
  path: list{"book", "10", "edit"},
  hash: "author",
  search: "name=Jane"

Basic Example

Let's start with a first example to see how a ReScript React Router looks like:

ReScriptJS Output
// App.res
let make = () => {
  let url = RescriptReactRouter.useUrl()
  switch url.path {
    | list{"user", id} => <User id />
    | list{} => <Home/>
    | _ => <PageNotFound/>

Directly Get a Route

In one specific occasion, you might want to take hold of a url record outside of watchUrl. For example, if you've put watchUrl inside a component's didMount so that a URL change triggers a component state change, you might also want the initial state to be dictated by the URL.

In other words, you'd like to read from the url record once at the beginning of your app logic. We expose dangerouslyGetInitialUrl() for this purpose.

Note: the reason why we label it as "dangerous" is to remind you not to read this url in any arbitrary component's e.g. render, since that information might be out of date if said component doesn't also contain a watchUrl subscription that re-renders the component when the URL changes. Aka, please only use dangerouslyGetInitialUrl alongside watchUrl.

Push a New Route

From anywhere in your app, just call e.g. RescriptReactRouter.push("/books/10/edit#validated"). This will trigger a URL change (without a page refresh) and watchUrl's callback will be called again.

We might provide better facilities for typed routing + payload carrying in the future!

Note: because of browser limitations, changing the URL through JavaScript (aka pushState) cannot be detected. The solution is to change the URL then fire a "popState" event. This is what Router.push does, and what the event watchUrl listens to. So if, for whatever reason (e.g. incremental migration), you want to update the URL outside of RescriptReactRouter.push, just do window.dispatchEvent(new Event('popState')).

Design Decisions

We always strive to lower the performance and learning overhead in RescriptReact, and our router design's no different. The entire implementation, barring browser features detection, is around 20 lines. The design might seem obvious in retrospect, but to arrive here, we had to dig back into ReactJS internals & future proposals to make sure we understood the state update mechanisms, the future context proposal, lifecycle ordering, etc. and reject some bad API designs along the way. It's nice to arrive at such an obvious solution!

The API also doesn't dictate whether matching on a route should return a component, a state update, or a side-effect. Flexible enough to slip into existing apps.

Performance-wise, a JavaScript-like API tends to use a JS object of route string -> callback. We eschewed that in favor of pattern-matching, since the latter in Rescript does not allocate memory, and is compiled to a fast jump table in C++ (through the JS JIT). In fact, the only allocation in the router matching is the creation of the url record!