Liking cljdoc? Tell your friends :D

Initial Code Walk-through

At this point, you are about 50% of the way to understanding re-frame. You are armed with:

  • a high level understanding of the 6 domino process (from re-frame's README)
  • an understanding of application state (from the previous tutorial)

By the end of this tutorial, you'll be at 70%, which is good enough to start coding by yourself.

In this tutorial, we'll look at re-frame code.

Table Of Contents

What Code?

This repo contains an example application called "simple", which has around 70 lines of code. We'll look at every line of the file.

This app:

  • displays the current time in a nice big, colourful font
  • provides a single text input field, into which you can type a hex colour code, like "#CCC", used for the time display

Here's what it looks like:

Example App image

To run the code:

Then:

  1. git clone https://github.com/Day8/re-frame.git
  2. cd re-frame/examples/simple
  3. lein do clean, figwheel
  4. open http://localhost:3449/example.html

Namespace

Because this example is tiny, the code is in a single namespace which you can find here: https://github.com/Day8/re-frame/blob/master/examples/simple/src/simple/core.cljs

Within this namespace, we'll need access to both reagent and re-frame. So, at the top, we start like this:

(ns simple.core
  (:require [reagent.core :as reagent]
            [re-frame.core :as rf]))

Data Schema

Now, normally, I'd strongly recommended you write a quality schema for your application state (the data stored in app-db). But, here, to minimise cognitive load, we'll cut that corner.

But ... we can't cut it completely. You'll still need an informal description, and here it is ... for this app app-db will contain a two-key map like this:

{:time       (js/Date.)  ;; current time for display
 :time-color "#f88"}     ;; the colour in which the time should be be shown

re-frame itself owns/manages app-db (see FAQ #1), and it will supply the value within it (a two-key map in this case) to your various handlers as required.

Events (domino 1)

Events are data. You choose the format.

re-frame uses a vector format for events. For example:

[:delete-item 42]

The first element in the vector identifies the kind of event. The further elements are optional, and can provide additional data associated with the event. The additional value above, 42, is presumably the id of the item to delete.

Here are some other example events:

[:admit-to-being-satoshi false]
[:set-spam-wanted false :continue-harassment-nevertheless-flag]
[:some-ns/on-GET-success response] 

The kind of event is always a keyword, and for non-trivial applications it tends to be namespaced.

Rule: events are pure data. No sneaky tricks like putting callback functions on the wire. You know who you are.

dispatch

To send an event, call dispatch with the event vector as argument:

   (rf/dispatch [:event-id  value1 value2])

In this "simple" app, a :timer event is dispatched every second:

(defn dispatch-timer-event
  []
  (let [now (js/Date.)]
    (rf/dispatch [:timer now])))  ;; <-- dispatch used

;; call the dispatching function every second
(defonce do-timer (js/setInterval dispatch-timer-event 1000))

This is an unusual source of events. Normally, it is an app's UI widgets which dispatch events (in response to user actions), or an HTTP POST's on-success handler, or a websocket which gets a new packet.

After dispatch

dispatch puts an event into a queue for processing.

So, an event is not processed synchronously, like a function call. The processing happens later - asynchronously. Very soon, but not now.

The consumer of the queue is a router which looks after the event's processing.

The router:

  1. inspects the 1st element of an event vector
  2. looks in a registry for the event handler which is registered for this kind of event
  3. calls that event handler with the necessary arguments

As a re-frame app developer, your job, then, is to write and register a handler for each kind of event.

Event Handlers (domino 2)

Collectively, event handlers provide the control logic in a re-frame application.

In this application, 3 kinds of event are dispatched: :initialise :time-color-change :timer

3 events means we'll be registering 3 event handlers.

Two ways To register

Event handlers can be registered via either reg-event-db or reg-event-fx (-db vs -fx).

Handler functions take coeffects (input args) and return effects, however reg-event-db allows you to write simpler handlers. The handler functions it registers (1) take just one coeffect - the current app state, and (2) return only one effect - the updated app state.

Whereas reg-event-fx registered handlers are more flexible.

Because of its simplicity, we'll be using the former here.

reg-event-db

We register event handlers using re-frame's reg-event-db.

reg-event-db is used like this:

(rf/reg-event-db
  :the-event-id
  the-event-handler-fn)

The handler function you provide should expect two parameters:

  • db the current application state (contents of app-db)
  • v the event vector

So, your function will have a signature like this: (fn [db v] ...).

Each event handler must compute and return the new state of the application, which means it normally returns a modified version of db.

:initialize

On startup, application state must be initialised. We want to put a sensible value into app-db which will otherwise contain {}.

So a (dispatch [:initialize]) will happen early in the apps life (more on this below), and we need to write an event handler for it.

Now this event handler is slightly unusual because it doesn't much care about the existing value in db - it just wants to plonk a completely new value.

Like this:

(rf/reg-event-db              ;; sets up initial application state
  :initialize                 
  (fn [_ _]                   ;; the two parameters are not important here, so use _
    {:time (js/Date.)         ;; What it returns becomes the new application state
     :time-color "#f88"}))    ;; so the application state will initially be a map with two keys

This particular handler fn ignores the two parameters (usually called db and v) and simply returns a map literal, which becomes the application state.

Here's an alternative way of writing it which does pay attention to the existing value of db:

(rf/reg-event-db
  :initialize              
  (fn [db _]                 ;; we use db this time, so name it        
    (-> db          
      (assoc :time (js/Date.))
      (assoc :time-color "#f88")))

:timer

Earlier, we set up a timer function to (dispatch [:timer now]) every second.

Here's how we handle it:

(rf/reg-event-db                 ;; usage:  (rf/dispatch [:timer a-js-Date])
  :timer                         
  (fn [db [_ new-time]]          ;; <-- de-structure the event vector
    (assoc db :time new-time)))  ;; compute and return the new application state

Notes:

  1. the event will be like [:timer a-time], so the 2nd v parameter destructures to extract the a-time value
  2. the handler computes a new application state from db, and returns it

:time-color-change

When the user enters a new colour value (via an input text box):

(rf/reg-event-db                
  :time-color-change            ;; usage:  (rf/dispatch [:time-color-change 34562])
  (fn [db [_ new-color-value]]  
    (assoc db :time-color new-color-value)))   ;; compute and return the new application state

Effect Handlers (domino 3)

Domino 3 actions/realises the effects returned by event handlers.

In this "simple" application, our event handlers are implicitly returning only one effect: "update application state".

This particular effect is actioned by a re-frame supplied effect handler. So, there's nothing for us to do for this domino. We are using a standard re-frame effect handler.

And this is not unusual. You'll seldom have to write effect handlers, but we'll understand more about them in a later tutorial.

Subscription Handlers (domino 4)

Subscription handlers take application state as an argument, and they compute a query over it, returning something of a "materialised view" of that application state.

When the application state changes, subscription functions are re-run by re-frame, to compute new values (a new materialised view).

Ultimately, the data returned by query functions is used in the view functions (Domino 5).

One subscription can source data from other subscriptions. So it is possible to create a tree of dependencies.

The Views (Domino 5) are the leaves of this tree The tree's root is app-db and the intermediate nodes between the two are computations being performed by the query functions of Domino 4.

Now, the two examples below are trivial. They just extract part of the application state and return it. So, there's virtually no computation. A more interesting tree of subscriptions and more explanation can be found in the todomvc example.

reg-sub

reg-sub associates a query id with a function that computes that query, like this:

(rf/reg-sub
  :some-query-id  ;; query id (used later in subscribe)
  a-query-fn)     ;; the function which will compute the query

If, later, a view function subscribes to a query like this: (subscribe [:some-query-id]). Note use of :some-query-id then a-query-fn will be used to perform the query over the application state.

Each time application state changes, a-query-fn will be called again to compute a new materialised view (a new computation over app state) and that new value will be given to any view function which is subscribed to :some-query-id. This view function, itself, will then also be called again to compute new DOM (because it depends on a query value which changed).

Along this reactive chain of dependencies, re-frame will ensure the necessary calls are made, at the right time.

Here's the code:

(rf/reg-sub
  :time
  (fn [db _]     ;; db is current app state. 2nd unused param is query vector
    (:time db))) ;; return a query computation over the application state

(rf/reg-sub
  :time-color
  (fn [db _]
    (:time-color db)))

Like I said, both of these queries are trivial. See todomvc.subs.clj for more interesting ones.

View Functions (domino 5)

view functions turn data into DOM. They are "State in, Hiccup out" and they are Reagent components.

Any SPA will have lots of viewfunctions, and collectively, they render the app's entire UI.

Hiccup

Hiccup is a data format for representing HTML.

Here's a trivial view function which returns hiccup-formatted data:

(defn greet
 []
 [:div "Hello viewers"])  ;; means <div>Hello viewers</div>

And if we call it:

(greet)
;; ==>  [:div "Hello viewers"]

(first (greet))
;; ==> :div

Yep, that's a vector with two elements: a keyword and a string.

Now,greet is pretty simple because it only has the "Hiccup Out" part. There's no "Data In".

Subscribing

To render the DOM representation of some-part-of app state, view functions must query for that part of app-db, and that means using subscribe.

subscribe is always called like this:

   (rf/subscribe  [query-id some optional query parameters])

There's only one (global) subscribe function and it takes one argument, assumed to be a vector.

The first element in the vector (shown above as query-id) identifies/names the query and the other elements are optional query parameters. With a traditional database a query might be:

select * from customers where name="blah"

In re-frame, that would be done as follows: (subscribe [:customer-query "blah"]) which would return a ratom holding the customer state (a value which might change over time!).

Because subscriptions return a ratom, they must always be dereferenced to obtain the value. This is a recurring trap for newbies.

The View Functions

This view function renders the clock:

(defn clock
  []
  [:div.example-clock
   {:style {:color @(rf/subscribe [:time-color])}}
   (-> @(rf/subscribe [:time])
       .toTimeString
       (clojure.string/split " ")
       first)])

As you can see, it uses subscribe twice to obtain two pieces of data from app-db. If either change, re-frame will re-run this view function.

And this view function renders the input field:

(defn color-input
  []
  [:div.color-input
   "Time color: "
   [:input {:type "text"
            :value @(rf/subscribe [:time-color])        ;; subscribe
            :on-change #(rf/dispatch [:time-color-change (-> % .-target .-value)])}]])  ;; <---

Notice how it does BOTH a subscribe to obtain the current value AND a dispatch to say when it has changed.

It is very common for view functions to render event-dispatching functions. The user's interaction with the UI is usually the largest source of events.

And then something more standard:

(defn ui
  []
  [:div
   [:h1 "Hello world, it is now"]
   [clock]
   [color-input]])

Note: view functions tend to be organized into a hierarchy, often with data flowing from parent to child via parameters. So, not every view function needs a subscription. Very often the values passed in from a parent component are sufficient.

Note: view functions should never directly access app-db. Data is only ever sourced via subscriptions.

Components Like Templates?

view functions are like the templates you'd find in Django, Rails, Handlebars or Mustache -- they map data to HTML -- except for two massive differences:

  1. you have the full power of ClojureScript available to you (generating a Clojure data structure). The downside is that these are not "designer friendly" HTML templates.
  2. these templates are reactive. When their input Signals change, they are automatically rerun, producing new DOM. Reagent adroitly shields you from the details, but the renderer of any component is wrapped by a reaction. If any of the the "inputs" to that render change, the render is rerun.

Kick Starting The App

Below, run is the called when the HTML page has loaded to kick off the application.

It has two tasks:

  1. load the initial application state
  2. "mount" the GUI onto an existing DOM element.
(defn ^:export run
  []
  (rf/dispatch-sync [:initialize])     ;; puts a value into application state
  (reagent/render [ui]              ;; mount the application's ui into '<div id="app" />'
                  (js/document.getElementById "app")))

After run is called, the app passively waits for events. Nothing happens without an event.

When it comes to establishing initial application state, you'll notice the use of dispatch-sync, rather than dispatch. This is something of cheat which ensures a correct structure exists in app-db before any subscriptions or event handlers run.

Summary

Your job, when building an app, is to:

  • design your app's information model (data and schema layer)
  • write and register event handler functions (control and transition layer) (domino 2)
  • (once in a blue moon) write and register effect and coeffect handler functions (domino 3) which do the mutative dirty work of which we dare not speak in a pure, immutable functional context. Most of the time, you'll be using standard, supplied ones.
  • write and register query functions which implement nodes in a signal graph (query layer) (domino 4)
  • write Reagent view functions (view layer) (domino 5)

Further Code

You should also look at the todomvc example application.


Previous: app-db (Application State)       Up: Index       Next: Mental Model Omnibus

Can you improve this documentation? These fine people already did:
Mike Thompson, Daniel Compton, Stuart Mitchell & Matthew Jaoudi
Edit on GitHub

cljdoc is a website building & hosting documentation for Clojure/Script libraries

× close