Home

Awesome

kee-frame

re-frame with batteries included.

Build Status

Clojars Project

cljdoc badge

project chat

Project status (August 2021)

The API and functionality of kee-frame is stable and proven to work. Pull requests are welcome!

Quick walkthrough

(require '[kee-frame.core :as k])

(k/start!  {:routes         [["/" :live]
                             ["/league/:id/:tab" :league]]
            :app-db-spec    :my-app/db-spec
            :initial-db     {:some-prop true}
            :root-component [my-root-reagent-component]
            :debug?         true})

(k/reg-controller :league
                  {:params (fn [route-data]
                             (when (-> route-data :data :name (= :league))
                               (-> route-data
                                   :path-params
                                   :id)))
                   :start  (fn [ctx id] [:league/load id])})

(k/reg-chain :league/load
            
             (fn [ctx [id]]
               {:http-xhrio {:method          :get
                             :uri             (str "/leagues/" id)}})
            
             (fn [{:keys [db]} [_ league-data]]
               {:db (assoc db :league league-data)}))

(k/path-for [:league {:id 14 :tab :fixtures}]) => "/league/14/fixtures"

(reg-event-fx :todo-added
              (fn [_ [todo]]
                {:navigate-to [:todo {:id (:id todo)}]]}))

(defn main-view []
  [k/switch-route (fn [route] (-> route :data :name))
   :index [index-page] 
   :orders [orders-page]
   nil [:div "Loading..."])

Benefits of leaving the URL in charge

Kee-frame wants you to focus on the URL and let it contain all data necessary to load a view. When you let the URL guide you app architecture like this, strange things start to happen:

That's it!

You've reached the end of the quick summary. Keep reading for a more in-depth guide!

Articles

Learning kee-frame in 5 minutes

Introduction and background for kee-frame controllers

Controller tricks

Demo application

I made a simple demo app showing football results. Have a look around, and observe how all data loading just works while navigating and refreshing the page.

Online demo app

Demo app source

Feel free to clone the demo app and do some figwheelin' with it!

Support

Contact the author on Twitter or join the discussion on Slack. Don't be afraid to create issues. Lack of user friendliness is also a bug!

Installation

There are 2 simple options for bootstrapping your project:

1. Manual installation

Add the following dependency to your project.clj file:

[kee-frame "1.2.0"]

2. Luminus template

Luminus is a framework that makes it easy to get started with web app development in clojure. It comes with kee-frame if you do this:

lein new luminus your-app-name-here +kee-frame

API stability

This library tries hard conform to the high standards of many Clojure libraries, by not breaking backwards compatibility. I believe this is very important, an application made several years ago should be able to upgrade with close to zero effort.

The kee-frame API has remained stable since the launch in early 2018. Here is a list of important/breaking changes:

Getting started

The kee-frame.core namespace contains the public API. It also contains wrapped versions of reg-event-db and reg-event-fx.

(require '[kee-frame.core :as k :refer [reg-controller reg-chain reg-event-db reg-event-fx]])

Routes

Kee-frame uses reitit for routing (since 0.3.0). Read the reitit docs for more details on the syntax, here are the routes from the demo app:

(def routes [["/" :live]
             ["/league/:id/:tab" :league]])

Starting your app

The start! function starts the router and configures the application.

(k/start!  {:routes         my-routes
            :app-db-spec    :my-app/db-spec
            :initial-db     your-blank-db-map
            :root-component [my-root-reagent-component]
            :debug?         true})

Subsequent calls to start are not a problem, so call this function as often as you want. Typically on every figwheel reload.

The routes property causes kee-frame to wire up the browser to navigate by those routes. Skip this property if you want to do your own routing. See the "Introducing kee-frame into an existing app" section.

You can set the hash-routing? property to true for /#/todos/1 style urls. Otherwise kee-frame defaults to using the browser history without the hash. The hash bit should not be included in your route definition, kee-frame strips it off before matching the route.

By default an unknown URL/route causes an error. You can provide a string URL under config key :not-found as your default 404 when no route is found.

If you provide :root-component, kee-frame will render that component in the DOM element with id "app". Make sure you have such an element in your index.html. You are free to do the initial rendering yourself if you want, just skip this setting. If you use this feature, make sure that k/start! is called every time figwheel reloads your code.

The :log option accepts a timbre log configuration map. See the Logging section for more details.

The debug? and debug-config options were replaced by the :log option in 0.4.1.

If you provide an app-db-spec, the framework will let you know when a bug in your event handler is trying to corrupt your DB structure. This is incredibly useful, so you should put down the effort to spec up your db!

You can override kee-frame's behaviour on route change through the :route-change-event option. Just specify the id of the event you want to use. One possible case is to perform some gatekeeping before executing the controllers for that route. If route execution is ok, the event could dispatch to kee-frame's built in :kee-frame.router/route-changed.

Controllers

A controller is a connection between the route data and your event handlers. It is a map with two required keys (params and start), and one optional (stop).

The params function receives the route data every time the URL changes. Its only job is to return the part of the route that it's interested in. This value combined with the previous value decides the next state of the controller. I'll come back to that in more detail.

The start function accepts the full re-frame context and the value returned from params. It should return nil or an event vector to be dispatched.

The stop function receives the re-frame context and also returns nil or an event vector.

(reg-controller :league
                {:params (fn [{:keys [handler route-params]}]
                           (when (= handler :league)
                             (:id route-params)))
                 :start  (fn [_ id]
                           [:league/load id])})

For start and stop it's very common to ignore the parameters and just return an event vector, and for that you can use a vector instead of a function:

(reg-controller :leagues
                {:params (constantly true) ;; Will cause the controller to start immediately, but only once
                 :start  [:leagues/load]}) ;; The route params will be appended to this vector, as the first event param.

Controller state transitions

This rules of controller states are stolen entirely from Keechma.

NOTE: false is not treated as falsy in controllers, it is considered as any other value. The explicitly negative value that stops controllers is nil.

The controller rules are:

All controllers :params fns are called every time the route changes. Therefore :params should return nil for routes on which it should not be started.

Event chains

Kee-frame uses re-chain to chain event handlers together for increased readability less boilerplate for common cases. See the example below for how to use it, visit the re-chain page for details and documentation

(k/reg-chain :league/load
            
             (fn [ctx [id]]
               {:http-xhrio {:method          :get
                             :uri             (str "/leagues/" id)}})
            
             (fn [{:keys [db]} [_ league-data]]
               {:db (assoc db :league league-data)}))

Browser navigation

Using URL strings in your links and navigation is error prone and quickly becomes a maintenance problem. Therefore, kee-frame encourages you to only interact with route data instead of concrete URLs. It provides 2 abstractions to help you with that:

The kee-frame.core/path-for function accepts a reitit route and returns a URL string:

(k/path-for [:todos {:id 14}]) => "/todos/14"

Kee-frame also includes a re-frame effect for triggering a browser navigation, after all navigation is a side effect. The effect is :navigate-to and it accepts a reitit route. The example below shows a handler that receives some data and navigates to the view page for those data.

(reg-event-fx :todo-added
              (fn [_ [todo]]
                {:db          (update db :todos conj todo)
                 :navigate-to [:todo :id (:id todo)]]})) ;; "/todos/14"

See this issue for some hints on how to use query parameters in your browser navigation.

Routing in your views

Most apps need to different views for different URLs. This isn't too hard to solve in re-frame, just subscribe to your route and implement your dispatch logic like this:

(defn main-view []
  (let [route (subscribe [:kee-frame/route])]
    (fn []
      [:div
       (case (:handler @route)
         :index [index-page]
         :orders [orders-page])])))

Kee-frame provides a simple helper to do this:

(defn main-view []
  [k/switch-route (fn [route] (-> route :data :name))
   :index [index-page] ;; Explicit call to reagent component, ignoring route data
   :orders orders-page]) ;; Orders page will receive the route data as its parameter because of missing []

It looks pretty much the same, only more concise. But it does help you with a few subtle but important things:

Finite State Machines

The last year there has been several experiments in kee-frame for integrating FSMs. I have now reached a conclusion on this, and created re-statecharts. It is available in kee-frame through glimt.

If you have been using event chains to do your HTTP requests, glimt is clearly a better option. It handles more than just the happy path and gives you with the state of the request along the way. I highly recommend trying out FSMs in your apps!

Logging

https://github.com/ptaoussanis/timbre is included for logging. The kee-frame.core/start! function accepts an optional timbre config map. The default config logs to browser console at level :info. Here's an example config you could pass in:

{:log {:level        :debug
       :ns-blacklist ["kee-frame.event-logger"]}
...
}

Kee-frame uses debug logging as an aid in debugging applications. This means that turning on all logging will get quite noisy. You might want to include some of these namespaces in your blacklist if you don't need them:

IMPORTANT: For debug logging to show up in Chrome you need to enable "Verbose" logging in Chrome dev tools. See https://github.com/ptaoussanis/timbre/issues/249

Introducing kee-frame into an existing app

Several parts of kee-frame are designed to be opt-in. This means that you can include kee-frame in your project and start using parts of it.

If you want controllers and routes, you need to replace your current routing with kee-frame's routing. If your current routing requires a lot of work to fit with the standard reitit routing solution, you may implement a custom router. See the next section for more details.

Alternatively, make your current router dispatch the event [:kee-frame.router/route-changed route-data] on every route change. That should enable what you need for the controllers.

Using a different router implementation

You may not like reitit, or you are already using a different router. In that case, all you have to do is implement your own version of the protocol kee-frame.api/Router and pass it in with the rest of your config:

(k/start!  {:router         (->BidiRouter bidi-style-routes)
            :root-component [my-root-reagent-component]
            ...})

Here are some example (not fully tested) router implementations. If you are upgrading from a pre 0.3.0 version of kee-frame, you probably want to keep your current bidi routes. The old bidi router implementation can be found here, just make a copy and use it as your :router.

If you choose to use a different router than reitit, you also need to use the corresponding routing data format when using path-for and the :navigate-to effect.

Server side routes

If you want to use links without hashes (/some-route instead of /#/some-route), you need a bit of server setup for it to work perfectly. A React SPA is typically loaded from the "app" element inside index.html served from the root / of your server. If the user navigates to some client route /leagues/465 and then hits refresh, the server will be unable to match that route as it exists only on the client. We will get a 404 instead of the index.html that we need. We want this to work, so that URLs can still be deterministic, even if they exist only on the client.

You can solve this in several ways, the simplest way is to include a wildcard route as the last route on the server. The server should serve index.html on any route not found on the server. This works, the downside is that you won't be able to serve a 404 page for non-matched URLs on the server.

In compojure, the wildcard route would look like this:

(GET "*" req {:headers {"Content-Type" "text/html"}
                  :status  200
                  :body    (index-handler req)})

Screen size breakpoints

Most web apps benefit from having direct access to information about the size and orientation of the screen. Kee-frame ships with the nice and simple breaking-points library that provides subscriptions for the screen properties you're interested in.

The screen breakpoints are completely configurable, you can pass your preferred ones to the start! function. The ones listed in the example below are the defaults, so if you're happy with those you can just pass true to the :screen parameter. If you omit it altogether, or pass false - the screen breakpoints will be disabled.

(k/start!  {:screen {:breakpoints 
                        [:mobile
                         768
                         :tablet
                         992
                         :small-monitor
                         1200
                         :large-monitor]
                     :debounce-ms 166}
              ;; Other settings here
              })

The subscriptions available are:

(rf/subscribe [:breaking-point.core/screen-width]) ;; will be an int
(rf/subscribe [:breaking-point.core/screen-height]) ;; will be an int
(rf/subscribe [:breaking-point.core/screen]) ;; will be one of the following: :mobile, :tablet, :small-monitor, :large-monitor

(rf/subscribe [:breaking-point.core/orientation]) ;; will be either :portrait or :landscape
(rf/subscribe [:breaking-point.core/landscape?]) ;; true if width is >= height
(rf/subscribe [:breaking-point.core/portrait?]) ;; true if height > width

;; these will be based on the breakpoint names that you provide
(rf/subscribe [:breaking-point.core/mobile?]) ;; true if screen-width is < 768
(rf/subscribe [:breaking-point.core/tablet?]) ;; true if screen-width is >= 768 and < 992
(rf/subscribe [:breaking-point.core/small-monitor?]) ;; true if window width is >= 992 and < 1200
(rf/subscribe [:breaking-point.core/large-monitor?]) ;; true if window width is >= 1200

Error messages

Helpful error messages are important to kee-frame. You should not get stuck because of "undefined is not a function". If you make a mistake, kee-frame should make it very clear to you what you did wrong and how you can fix it. If you find pain spots, please post an issue so we can find better solutions.

React error boundaries

If you are unfamiliar with error boundaries, you can read the docs. After reading this, I decided to include Will's code snippet in kee-frame. Usage:

[kee-frame.error/boundary
  [your-badly-behaving-component-here props]]

This will print JS errors inside the component on screen instead of breaking the whole rendering tree. You can optionally include your own error-handling component function as the first param.

Example usage from demo app

Scroll behavior on navigation

In a traditional static website, the browser handles the scrolling for you nicely. Meaning that when you navigate back and forward, the browser "remembers" how far down you scrolled on the last visit. This is convenient for many websites, so Kee-frame utilizes a third-party JS lib to get this behavior for a SPA. The only thing you need to do is this in your main namespace:

(:require [kee-frame.scroll])

Credits

The implementation of kee-frame is quite simple, building on rock solid libraries and other people's ideas. The main influence is the Keechma framework. It is a superb piece of thinking and work, go check it out! Apart from that, the following libraries make kee-frame possible:

Thank you!