routility

0.1.1 • Public • Published

Routility

npm version Circle CI

Sauce Test Status

A generic routing utility navigate users in browser and convert URLs to objects, so your application logic can consume them without wrestling with a specify API, which uncovers the endless possibilities of doing awesome stuff. Because all you need to do now is managing data instead of learning 3rd party libraries. This works exceptionally great with architectures that are reactive or has a single source of truth, e.g. React, Redux or Flux in general.

Before reaching 1.0, API and semantic will change between minor versions.

Usage

var Routility = require('routility');
var r = Routility.r;
var redirect = Routility.redirect;
 
var routes = ( // Parentheses are not required, it looks nice to align all the routes
  r('/', 'root', [
    redirect('/', '/login'),
    r('/login', 'login'),
    r('/user/:id', 'user', [
      r('/', 'index'), // If you don't define "/", "/user/:id" won't be handled
      r('/profile', 'profile')
    ])
  ])
);
 
var navTo = Routility.start(routes, function (state) {
  // This handler will be called as soon as you call "navTo"
});
 
// Will first change url to "/user/123" and return new state
// Those will happen synchronously
navTo('/user/123');
// This returns current state:
// {
//   root: {
//     user: {
//       id: '123',
//       index: {}
//     }
//   },
//   queryParams: {}
// }

Why

Many front end frameworks and view libraries have to manage state in their application. There are many solutions to manage state. But integrated route management is often overlooked. Most libraries will have a router abstraction to deal with routing in the browser. A common mistake of those abstractions is that they didn't treat routes as data. This caused their API is be highly coupled with specific use cases.

A route to a browser application should just be some states. Nothing more. When we interpret route at this level, the application logic can be easily fully integrated with the rest of state management layer, which enables lots of possibilities.

API

r(path, name, subRoutes) -> RouteDefinition

Define route hierarchy. (The ( and ) are not necessary, it just help align the r function calls, makes them nice to read.) name argument is related to the structure of state object.

var routes = ( // Parentheses are not required
  r('/', 'root', [
    redirect('/', '/login'),
    r('/login', 'login'),
    r('/user/:id', 'user', [
      r('/', 'index'), // If you don't define "/", "/user/:id" won't be handled
      r('/profile', 'profile')
    ])
  ])
);

redirect(path, targetPath)

path has the same semantic as the path argument of r function. targetPath is an absolute path to the new path.

start(definition, handler, [opts]) -> navTo

Start routing for browser. Return a navTo helper. handler will only be called when user use browser back and forward button to navigate. Refer to Structure of state section to learn what the state object will look like.

var navTo = start(definition, function (state) {
  // You can notify the application that state has changed
});

handler(state)

state is same as state object returned by navTo function

options

  • opts.browserHistory = false Use HTML5 history. If false, will use url hashtag. If true, will use the entire path.

navTo([path]) -> state|null

Navigate to target URL.

  • If called without arguments, will return current state.
  • If hit a redirect route, it will go directly to the redirect target path, and only return the final state object. The final state object will contain two additional properties, redirectTo and redirectFrom. (Refer to Structure of state section to learn what the state object will look like.)
navTo(); // return current state, no effect on url
navTo('/');
navTo('/login');
navTo('/user/123');
navTo('/user/123/profile?q=abc');

parse(definition, path) -> state|null

A helper to take a definition and a path, and generate the state object for that path. This function has no effect on browser url. It just return the state object, and that's it! This can be used to run on Node.js (server) environment. It will return the final state object if it the path hits a redirect route. It will return null if path matches no route (both when directly or after redirect).

Structure of state

// A example route definition
r('/', 'root', [
  r('/', 'index'),
  redirect('/sign-in', '/login'),
  r('/login', 'login'),
  r('/user/:id', 'user', [
    r('/', 'index'),
    r('/profile', 'profile')
  ])
]);
 
// For '/'
{
  "root": {
    "index": {}
  },
  "queryParams": {}
}
 
// For '/login'
{
  "root": {
    "login": {}
  },
  "queryParams": {}
}
 
// For '/sign-in'
{
  "root": {
    "login": {} // Show login instead because it hits a redirect route
  },
  "redirectFrom": "/sign-in", // Redirect route property
  "redirectTo": "/login", // Redirect route property
  "queryParams": {}
}
 
// For '/user/123'
{
  "root": {
    "user": {
      "id": "123" // Notice "id" will always be a string
    }
  },
  "queryParams": {}
}
 
// For '/user/123/profile?q=abc&name=one'
{
  "root": {
    "user": {
      "id": "123",
      "profile": {}
    }
  },
  "queryParams": {
    "q": "abc",
    "name": "one"
  }
}

Dependencies (2)

Dev Dependencies (21)

Package Sidebar

Install

npm i routility

Weekly Downloads

0

Version

0.1.1

License

MIT

Last publish

Collaborators

  • daiweilu