@webqit/webflo

v0.11.61-0

Published

a year ago

The universal, fullstack JavaScript framework.

Downloads

295

0High
0Medium
0Low

ox-harris

git-workflow self-hosting automatic-deployment command-line-interface progressive-web-apps single-page-apps vanila-javascript

Webflo

Webflo is a universal web, mobile, and API backend framework based on vanilla HTML, CSS, and JavaScript! It's a powerful little thing written to facilitate building more authentic, web-native applications!

Here, we've put all of that up for a 20mins straight read!

Heads Up...

Webflo is a framework on its own track - of working and thinking in vanilla HTML, CSS and JavaScript! Instead of trying to follow certain patterns that have been touted as "norms", it takes a plunge to draw on native web platform features - plus some more futurisric, fascinating new stuff from across a few web paltforn proposals! This means that to happily meet Webflo, you also have to be excited about the idea of going one step ahead!

Note that Webflo is still evolving and some things may change quickly! Careful with production! Or, just remember to keep tabs with progress here!

The Wins...

Much of what eludes the web today!

Introducing...

the path of least engineering - with an all-new HTML-first thinking!
a focused, standards-based philosophy for building more authentic, web-native applications!

Natively supporting notable new possibilities...

a new approach to reactivity that's based on no syntax at all but plain old JavaScript!
a new "imports" feature for HTML that makes HTML more reusable!
and much more.

Documentation

(A bit of a long read!)

Overview

Here's a glimpse of your Webflo app.

For when your application is a static site, or has static files to serve:

The public directory for static files.

my-app
  ├── public/index.html ----------------- http://localhost:3000/index.html | http://localhost:3000
  └── public/logo.png ------------------- http://localhost:3000/logo.png

For when your application requires dynamic request handling on the server:

The server directory for server-side routing.

my-app
  └── server/index.js

In which case a typical index.js route handler has the following anatomy:

/**
server
 ├── index.js
 */
export default function(event, context, next) {
    if (next.pathname) {
        return next();  // <--------------------------------- http://localhost:3000/logo.png (or other non-root URLs)
    }
    return { title: 'Hello from Server' };  // <------------- http://localhost:3000/ (root URL)
}

In which case data is returned either as a JSON (API) response, or as a rendered page response where there is an index.html file in the public directory that pairs with the route.

my-app
  ├── server/index.js
  └── public/index.html

In which case a typical index.html page has the following anatomy:

<!--
public
  ├── index.html
-->
<!DOCTYPE html>
<html>
    <head>
        <link rel="stylesheet" href="/style.css" />   <!-- ---------------------- Application CSS -->
        <script type="module" src="/bundle.js"></script>   <!-- ----------------- Application JS bundle -->
        <template name="routes" src="/bundle.html"></template>   <!-- ------------- Reusable HTML Templates and partials (Details ahead) -->
    </head>
    <body>...</body>
</html>

For when your application requires dynamic request handling on the client (the browser):

The client directory for client-side routing,

And, optionally, the worker directory for an all-new Service Worker based routing! (As detailed ahead.)

my-app
  ├── client/index.js
  └── worker/index.js

Where in both cases, a typical index.js route handler has the same familiar anatomy:

/**
[client|worker]
 ├── index.js
 */
export default function(event, context, next) {
    if (next.pathname) {
        return next();  // <--------------------------------- http://localhost:3000/logo.png (or other non-root URLs)
    }
    return { title: 'Hello from [Browser|Worker]' };  // <--- http://localhost:3000/ (root URL)
}

This and much more - ahead!

Here's a glimpse of the standards-based stack you get of Webflo!

For when your application involves routing:

The Fetch Standard, comprising of the Request, Response, and Headers interfaces - used for all things requests and responses - across client, server, and Service Worker environments. (Details ahead)
This paves the way to using other native APIs as-is, when handling requests and responses. For example, if you sent an instance of the native FormData, Blob, File, or ReadableStream object from the browser side of your application, you'd be getting the equivalent instance on the server side!
WHATWG URL and WHATWG URLPattern - used for all things URL and URL pattern matching, respectively - across client, server, and Service Worker environments. (Details ahead)

For when your application involves pages and a UI:

The HTML Standard - held for all things markup - across client, server, and Service Worker environments! You'll absolutely ❤ love it that your pages and page components can live as plain .html files! The browser already does!
This HTML-first approach is new! And you can get away with a "zero-JavaScript" proposition, or something more of a Progressive Enhancement proposition that makes do with "just-enough JavaScript"!
WHATWG DOM - available universally - not only on the client-side, but also on the server-side via OOHTML-SSR - all of which lets us have dynamic pages.
And you get to have Custom Elements as powerful building blocks.

Same web standards everwhere you look! You come off with a web-native app!

Installation

Every Webflo project starts on an empty directory that you can create on your machine. The command below makes a new directory my-app from the terminal and navigates into it.

mkdir my-app
cd my-app

With npm available on your terminal, run the following command to install Webflo to your project:

System Requirements: Node.js 18.0 (having stable Fetch API support) or later

npm i @webqit/webflo

The installation automatically creates a package.json file at project root, containing @webqit/webflo as a project dependency.

{
  "dependencies": {
    "@webqit/webflo": "..."
  }
}

Other important definitions like project name, package type, and aliases for common Webflo commands will now also belong here.

{
  "name": "my-app",
  "type": "module",
  "scripts": {
    "start": "webflo start::server --mode=dev",
    "generate": "webflo generate::client --compression=gz --auto-embed"
  },
  "dependencies": {
    "@webqit/webflo": "..."
  }
}

And that gets it all ready! The commands npm start and npm run generate will be coming in often during development.

"Hello World!"

To be sure that Webflo is listening, run npx webflo help on the terminal. An overview of available commands should be shown.

If you can't wait to say Hello World! 😅, you can have an HTML page say that right away!

Create an index.html file in a new subdirectory public.

public
  └── index.html

<!DOCTYPE html>
<html>
    <head>
        <title>My App</title>
    </head>
    <body>
        <h1>Hello World!</h1>
        <p>This is <b>My App</b></p>
    </body>
</html>

Start the Webflo server and visit http://localhost:3000 on your browser to see your page. 😃
```
npm start
```
Welcome to Webflo!

Concepts

Handler Functions and Layout

Functions come in in Webflo when you need to dynamically handle requests.

Whether building a server-based, browser-based, or universal application, Webflo gives you one consistent way to handle requests and navigation: using handler functions!

You just define an index.js file with a function that gets called to handle a request!

/**
[server|client|worker]
 ├── index.js
 */
export default function(event, context, next, fetch) {
}

The following function handles only GET requests:

export function GET(event, context, next, fetch) {
}

Function names take after HTTP methods: GET, POST, PUT, PATCH, DELETE, OPTIONS, HEAD, etc.

Function names are lower case for Webflo version <= 0.11.23, in which case delete is del.

The following function can simply await asynchronous stuff:

export async function GET(event, context, next, fetch) {
    let a = await b();
}

Each function receives an event object representing details about the request - e.g. event.request, event.url, event.session. (Details ahead.)

While the context and next parameters are discussed below, fetch is a fetch-equivalent function passed in for convenience - for initiating remote requests.

Functions that will respond to requests on the server-side go into a directory named server.

/**
server
 ├── index.js
 */
export default function(event, context, next) {
    if (next.pathname) return next();
    return {
        title: 'Home | FluffyPets',
        source: 'server',
    };
}

The above function will respond on starting the server - npm start on your terminal - and visiting http://localhost:3000.

Funtions that will respond to requests on the client-side (from right within the browser) go into a directory named client.

/**
client
 ├── index.js
 */
export default function(event, context, next) {
    if (next.pathname) return next();
    return {
        title: 'Home | FluffyPets',
        source: 'in-browser',
    };
}

The above function is built as part of your application's client-side script from the npm run generate command. It is typically bundled to the file ./public/bundle.js. And the --auto-embed flag in that command gets it automatically embedded on your ./public/index.html page as <script type="module" src="/bundle.js"></script>. Then it responds from right in the browser on visiting http://localhost:3000.

For browser-based applications that want to employ Service-Workers (typically, Progressive Web Apps), Webflo allows for equivalent request handlers as part of the Service Worker. These worker-based functions go into a directory named worker.

/**
worker
 ├── index.js
 */
export default function(event, context, next) {
    if (next.pathname) return next();
    return {
        title: 'Home | FluffyPets',
        source: 'service-worker',
    };
}

The above function is built as part of your application's Service Worker script from the npm run generate command. It is typically bundled to the file ./public/worker.js, and the main application bundle automatically registers this as the application's Service Worker. Now, our function responds from within the Service Worker on visiting http://localhost:3000. (More details ahead.)

So, depending on where requests are best handled for your type of application, handler functions may be placed as below:

client
  ├── index.js

worker
  ├── index.js

server
  ├── index.js

Static files, e.g. images, stylesheets, etc, have their place in a files directory named public.

public
  └── logo.png

Step Functions and Workflows

Whether routing in the /client, /worker, or /server directory above, nested URLs follow the concept of Step Functions! These are parent-child layout of handlers that model your application's URL structure.

server
  ├── index.js --------------------------------- http://localhost:3000
  └── products/index.js ------------------------ http://localhost:3000/products
        └── stickers/index.js ------------------ http://localhost:3000/products/stickers

Each step calls a next() function to forward the current request to the next step of the given URL.

/**
server
 ├── index.js
 */
export default async function(event, context, next) {
    if (next.stepname) return next();
    return { title: 'Home | FluffyPets' };
}

/**
server
 ├── products/index.js
 */
export default function(event, context, next) {
    if (next.stepname) return next();
    return { title: 'Products' };
}

We get a step-based workflow that helps to decomplicate routing and lets us build out each route horizontally!

Here, a parent step can pass a context object to a child step, and can recompose its return value.

/**
server
 ├── index.js
 */
export default async function(event, context, next) {
    if (next.stepname) {
        let childContext = { user: { id: 2 }, };
        let childResponse = await next(childContext);
        return { ...childResponse, title: childResponse.title + ' | FluffyPets' };
    }
    return { title: 'Home | FluffyPets' };
}

It's versatile, so the next() function can be used to re-route the current request to a different handler - using a relative or absolute URL.

/**
server
 ├── index.js
 */
export default async function(event, context, next) {
    if (next.stepname === 'products') {
        return next(context, '/api/products?params=allowed'); // With an absolute URL
    }
    return { title: 'Home | FluffyPets' };
}

/**
server
 ├── products/index.js
 */
export default async function(event, context, next) {
    if (next.stepname) return next();
    return next(context, '../api/products?params=allowed'); // With a relative URL
}

And you can make that a full-fledged in-app request - passing in fetch-equivalent parameters.

/**
server
 ├── index.js
 */
export default async function(event, context, next) {
    if (next.stepname === 'products') {
        return next(context, '/api/products?params=allowed', {
            method: 'get', { headers: { Authorization: 'djjdd' } } 
        });
    }
    return { title: 'Home | FluffyPets' };
}

These requests are received at destination route - /api/products above - as regular HTTP requests!

For even more flexibility, workflows may be designed with wildcard steps using a hyphen - as step name. At runtime, a wildcard step matches any URL segment at its level in the hierarchy! A this.stepname property is always available to tell which URL segment has been matched.

/**
server
 ├── -/index.js
 */
export default function(event, context, next) {
    if (this.stepname === 'products') {
        return { title: 'Products' };
    }
    return { title: 'Untitled' };
}

Every handler function has the following contextual properties:
this.stepname - The name of the current step.
this.pathname - The URL pathname of the current step.
next.stepname - The name of the next step.
next.pathname - The remaining URL pathname after the current step. Server-side handlers have the following in addition:
this.dirname - The filesystem pathname of the current step.

Additionally, workflows may be designed with as many or as few step functions as necessary; the flow control parameters next.stepname and next.pathname are always available at any point to help with the remaining part of the given URL.

This means that it is even possible to handle all URLs from the root handler alone.

/**
server
 ├── index.js
 */
export default function(event, context, next) {
    // For http://localhost:3000/products
    if (next.pathname === 'products') {
        return { title: 'Products' };
    }

    // For http://localhost:3000/products/stickers
    if (next.pathname === 'products/stickers') {
        return { title: 'Stickers' };
    }
    
    // Should we later support other URLs like static assets http://localhost:3000/logo.png
    if (next.pathname) {
        return next();
    }
    
    // For the root URL http://localhost:3000
    return { title: 'Home' };
}

The next() function still plays an important role above because Webflo takes a default action when next() is called at the edge of the workflow - this point where there are no more step functions as there are URL segments.

For workflows in the /server directory, the default action of next()ing at the edge is to go match and return a static file in the public directory.

So, above, should our handler receive static file requests like http://localhost:3000/logo.png, the statement return next() would get Webflo to match and return the logo at public/logo.png, if any; a 404 response otherwise.

my-app
  ├── server/index.js ------------------------- http://localhost:3000, http://localhost:3000/products, http://localhost:3000/products/stickers, etc
  └── public/logo.png ------------------------- http://localhost:3000/logo.png

Note Obviously, the root handler effectively becomes the single point of entry to the application - being that it sees even requests for static files!

For workflows in the /worker directory, the default action of next()ing at the edge is to send the request through the network to the server. (But Webflo will check to see whether to (and how to) resolve the request from the application cache.)

So, above, if we defined handler functions in the /worker directory, we could selectively handle specific requests while next()ing others to the server.

/**
worker
 ├── index.js
 */
export default async function(event, context, next) {
    // For http://localhost:3000/about
    if (next.pathname === 'about') {
        return {
            name: 'FluffyPets',
            version: '1.0',
        };
    }
    
    // For http://localhost:3000/logo.png
    if (next.pathname === 'logo.png') {
        let response = await next();
        console.log( 'Logo file size:', response.headers.get('Content-Length') );
        return response;
    }
    
    // For every other URL
    return next();
}

Now we get the following handler-to-URL mapping for our application:

my-app
  ├── worker/index.js ------------------------- http://localhost:3000/about, http://localhost:3000/logo.png
  ├── server/index.js ------------------------- http://localhost:3000, http://localhost:3000/products, http://localhost:3000/products/stickers, etc
  └── public/logo.png ------------------------- http://localhost:3000/logo.png

Handlers in the /worker directory see only Same-Origin requests, being that Cross-Origin URLs like https://auth.example.com/oauth do not belong in the application's layout! But as detailed later, these external URLs may be may configured for strategic caching by the Service Worker.

For workflows in the /client directory, the default action of next()ing at the edge is to send the request through the network to the server. But where there is a Service Worker layer, then that becomes the next destination.

So, above, if we defined handler functions in the /client directory, we could selectively handle specific navigation requests in-browser while next()ing others down to the server, or first, the Service Worker layer.

/**
client
 ├── index.js
 */
export default async function(event, context, next) {
    // For http://localhost:3000/login
    if (next.pathname === 'login') {
        return {
            name: 'John Doe',
            role: 'owner',
        };
    }
    
    // For every other URL
    return next();
}

Our overall handler-to-URL mapping for the hypothetical application in context now becomes:

my-app
  ├── client/index.js ------------------------- http://localhost:3000/login
  ├── worker/index.js ------------------------- http://localhost:3000/about, http://localhost:3000/logo.png
  ├── server/index.js ------------------------- http://localhost:3000, http://localhost:3000/products, http://localhost:3000/products/stickers, etc
  └── public/logo.png ------------------------- http://localhost:3000/logo.png

If there's anything we have now, it is the ability to break work down[i], optionally across step functions, optionally between layers!

Pages, Layout and Templating

HTML files in the public directory, just like every other public file, are served statically when accessed directly - e.g. http://localhost:3000/index.html. But index.html files, specifically, are treated as pages by Webflo. They are, therefore, also accessible with path URLs like http://localhost:3000.

my-app
  └── public/index.html ----------------------- http://localhost:3000/index.html, http://localhost:3000

But, where an index.html file pairs with a route...

my-app
  ├── server/index.js
  └── public/index.html

...the route handler determines what happens.

/**
server
 ├── index.js
 */
export default async function(event, context, next) {
    // For http://localhost:3000/index.html, etc
    if (next.pathname) {
        return next();
    }
    // For http://localhost:3000 specifically
    return { ... };
}

Now, we are able to access the data component of a route differently from its HTML component!

my-app
  └── server/index.js ------------------------- http://localhost:3000 -------------------- application/json
  └── public/index.html ----------------------- http://localhost:3000/index.html --------- text/html

But, we can also access the route in a way that gets the data rendered into the automatically-paired index.html file for a dynamic page response. We'd simply set the Accept header of the request to a value of text/html, and Webflo will automatically perform Server-Side Rendering to give a page response.

Note The Accept header hint is already how browsers make requests on every page load. Here, it just works!

This automatic pairing of an index.html file with a route works the same for nested routes! But top-level index.html files are implicitly inherited down the hierarchy. That means that subroutes do not need to have their own index.html document, unless necessary.

Layout and Templating Overview

In a Single Page Application, all pages are based off a single index.html document. In a Multi Page Application, pages are individual index.html documents - ideally. But, Server-Side Rendering makes it possible to serve the same, but dynamically-rendered, index.html document across page loads - essentially an SPA architecture hiding on the server. But, here, lets take Multi Page Applications for an individual-page architecture.

In a Multi Page Application (with an individual-page architecture), each page is its own index.html document, and it is often necessary to have certain page sections - e.g. site header, footer, and sidebar, etc. - stay consistent across pages. These sections can be defined once and imported on every page.

my-app
  └── public
      ├── about/index.html ------------------------- <!DOCTYPE html>
      ├── products/index.html ---------------------- <!DOCTYPE html>
      ├── index.html ------------------------------- <!DOCTYPE html>
      ├── header.html ------------------------------ <header></header> <!-- To appear at top of each index.html page -->
      └── footer.html ------------------------------ <footer></footer> <!-- To appear at bottom of each index.html page -->

In a Single Page Application, each page is the same index.html document, and it is often necessary to have the main page sections change on each route. These sections can be defined per-route and imported to the document on navigating to their respective route.

my-app
  └── public
      ├── about/main.html -------------------------- <main></main> <!-- To appear at main area of index.html -->
      ├── products/main.html ----------------------- <main></main> <!-- To appear at main area of index.html -->
      ├── main.html -------------------------------- <main></main> <!-- To appear at main area of index.html -->
      └── index.html ------------------------------- <!DOCTYPE html>

This, in both cases, is templating - the ability to define HTML partials once, and have them reused multiple times. Webflo just concerns itself with templating, and the choice of a Multi Page Application or Single Page Application becomes yours! And heck, you can even have the best of both worlds in the same application - with an architecture we'll call Multi SPA! It's all a layout thing!

Now, with pages in Webflo being DOM-based (both client-side and server-side), documents can be manipulated directly with DOM APIs, e.g. to replace or insert nodes, attributes, etc. But even better, templating in Webflo is based on the HTML Modules and HTML Imports features in OOHTML - unless disabled in config. These features provide a powerful declarative templating system on top of the standard HTML <template> element - all in a module, export and import paradigm.

Here, you are able to define reusable contents in a <template> element...

<head>
    <template name="routes">
        <header exportgroup="header.html">Header Area</header>
        <main exportgroup="main.html">Main Area</main>
    </template>
</head>

...and have them imported anywhere on the root document using an <import> element:

<body>
    <import template="routes" name="header.html"></import>
    <import template="routes" name="main.html"></import>
</body>

The module element - <template> - is also able to load its contents from a remote .html file that serves as a bundle:

<!--
public
 ├── bundle.html
-->
<header exportgroup="header.html">Header Area</header>
<main exportgroup="main.html">Main Area</main>

<!--
public
 ├── index.html
-->
<head>
    <template name="routes" src="/bundle.html"></template>
</head>

What we'll see shortly is how multiple standalone .html files - e.g. those header.html, footer.html, main.html files above - come together into one bundle.html file for an application.

In a Multi Page Layout

In a Multi Page layout (as seen earlier), generic contents - e.g. header and footer sections, etc. - are typically bundled into one bundle.html file that can be embedded on each page of the application.

<!--
public
 ├── index.html
-->
<!DOCTYPE html>
<html>
    <head>
        <script type="module" src="/bundle.js"></script>
        <template name="routes" src="/bundle.html"></template>
    </head>
    <body>
        <import template="routes" name="header.html"></import>
        <main>Welcome to our Home Page</main>
        <import template="routes" name="footer.html"></import>
    </body>
</html>

<!--
public/about
 ├── index.html
-->
<!DOCTYPE html>
<html>
    <head>
        <script type="module" src="/bundle.js"></script>
        <template name="routes" src="/bundle.html"></template>
    </head>
    <body>
        <import template="routes" name="header.html"></import>
        <main>Welcome to our About Page</main>
        <import template="routes" name="footer.html"></import>
    </body>
</html>

<!--
public/products
 ├── index.html
-->
<!DOCTYPE html>
<html>
    <head>
        <script type="module" src="/bundle.js"></script>
        <template name="routes" src="/bundle.html"></template>
    </head>
    <body>
        <import template="routes" name="header.html"></import>
        <main>Welcome to our Products Page</main>
        <import template="routes" name="footer.html"></import>
    </body>
</html>

In this architecture, navigation is traditional - a new page loads each time. The bundle.js script comes with the appropriate OOHTML support level required for the imports to function.

In a Single Page Layout

In a Single Page layout (as seen earlier), page-specific contents - e.g. main sections - are typically bundled together into one bundle.html file that can be embedded on the document root. Notice how nested routes end up as nested <template> elements that form the equivalent of the application's URL structure.

<!--
public
 ├── bundle.html
-->
<template name="about">
    <main exportgroup="main.html">Welcome to our About Page</main>
</template>
<template name="products">
    <main exportgroup="main.html">Welcome to our Products Page</main>
</template>
<main exportgroup="main.html">Welcome to our Home Page</main>

Now, the <main> elements are each imported on navigating to their respective routes. This time, Webflo takes care of setting the URL path as a global template attribute on the <body> element such that <import> elements that inherit this global attribute are resolved from its current value.

<!--
public
 ├── index.html
-->
<!DOCTYPE html>
<html>
    <head>
        <script type="module" src="/bundle.js"></script>
        <template name="routes" src="/bundle.html"></template>
    </head>
    <body template="routes/"> <!-- This "template" attribute automatically changes to routes/about or routes/products as we navigate to http://localhost:3000/about and http://localhost:3000/products respectively -->
        <header></header>
        <import name="main.html"></import> <!-- This import element omits its "template" attribute so as to inherit the global one -->
        <footer></footer>
    </body>
</html>

In this architecture, navigation is instant and sleek - Webflo prevents a full page reload, obtains and sets data at document.state.data for the new URL, then sets the template attribute on the <body> element to the new URL path. The bundle.js script comes with the appropriate OOHTML support level required for the imports to function.

In a Multi SPA Layout

It's all a layout thing, so a hybrid of the two architectures above is possible in one application, to take advantage of the unique benefits of each! Here, you are able to have routes that are standalone index.html documents (MPA), which in turn, are able to act as a single document root for their subroutes (SPA).

my-app
  └── public
      ├── about/index.html ------------------------- <!DOCTYPE html> <!-- Document root 1 -->
      ├── products
      │     ├── free/main.html --------------------------- <main></main> <!-- To appear at main area of document root 2 -->
      │     ├── paid/main.html --------------------------- <main></main> <!-- To appear at main area of document root 2 -->
      │     ├── main.html -------------------------------- <main></main> <!-- To appear at main area of document root 2 -->
      │     └── index.html ------------------------------- <!DOCTYPE html> <!-- Document root 2, (doubles as an SPA) -->
      ├── index.html ------------------------------- <!DOCTYPE html> <!-- Document root 0 -->
      ├── header.html ------------------------------ <header></header> <!-- To appear at top of each document root -->
      └── footer.html ------------------------------ <footer></footer> <!-- To appear at bottom of each document root -->

The above gives us three document roots: /index.html, /about/index.html, /products/index.html. The /products route doubles as a Single Page Application such that visiting the /products route loads the document root /products/index.html and lets Webflo SPA routing determine which of /products/main.html, /products/free/main.html, /products/paid/main.html is imported on a given URL.

Webflo ensures that only the amount of JavaScript for a document root is actually loaded! So, above, a common JavaScript build is shared across the three document roots alongside an often tiny root-specific build.

<!--
public
 ├── products/index.html
-->
<!DOCTYPE html>
<html>
    <head>
        <script type="module" src="/webflo.bundle.js"></script>
        <script type="module" src="/products/bundle.js"></script>
        <template name="pages" src="/bundle.html"></template>
    </head>
    <body>...</body>
</html>

<!--
public
 ├── about/index.html
-->
<!DOCTYPE html>
<html>
    <head>
        <script type="module" src="/webflo.bundle.js"></script>
        <script type="module" src="/about/bundle.js"></script>
        <template name="pages" src="/bundle.html"></template>
    </head>
    <body>...</body>
</html>

<!--
public
 ├── index.html
-->
<!DOCTYPE html>
<html>
    <head>
        <script type="module" src="/webflo.bundle.js"></script>
        <script type="module" src="/bundle.js"></script>
        <template name="pages" src="/bundle.html"></template>
    </head>
    <body>...</body>
</html>

The Webflo generate command automatically figures out a given architecture and generates the appropriate scripts for the application! It also factors into the generated scripts the location of each document root so that all navigations to these roots are handled as a regular page load.

Bundling

Template .html files are bundled from the filesystem into a single file using the OOHTML CLI utility. On installing this utility, you may want to add the following to your npm scripts in package.json.

"scripts": {
    "generate:templates": "oohtml bundle --recursive --auto-embed=routes"
}

The --recursive flag gets the bundler to recursively bundle subroots in a Multi SPA layout - where certain subdirectories have their own index.html document. (These subroots would be ignored otherwise.)

The --auto-embed flag gets the bundler to automatically embed the generated bundle.html file on the matched index.html document. A value of routes for the flag ends up as the name of the embed template: <template name="routes" src="/bundle.html"></template>.

Note If your HTML files are actually based off the public directory, you'll need to tell the above command to run in the public directory, either by configuring the bundler, or by rewriting the command with a prefix: cd public && oohtml bundle --recursive --auto-embed=routes.

Client and Server-Side Rendering

With pages in Webflo being DOM-based (both client-side and server-side), we are able to access and manipulate documents and elements using familiar DOM APIs - e.g. to replace or insert contents, attributes, etc. Rendering in Webflo is based on this concept!

Here, Webflo simply makes sure that the data obtained from each route is available as part of the document object as document.state.data, making it accessible to our rendering logic.

We are able embed a script on our page and render this data on the relevant parts of the document.

<!--
public
 ├── index.html
-->
<!DOCTYPE html>
<html>
    <head>
        <title></title>
        <script>
         setTimeout(() => {
             console.log( document.state.data ); // { title: 'Home | FluffyPets' }
             let { title } = document.state.data;
             document.title = title;
         }, 0);
        </script>
    </head>
    <body></body>
</html>

Where your rendering logic is an external script, your <script> element would need to have an ssr Boolean attribute to get the rendering engine to fetch and run your script on the server.

<!--
public
 ├── index.html
-->
<!DOCTYPE html>
<html>
    <head>
        <title></title>
        <script src="app.js" ssr></script>
    </head>
    <body></body>
</html>

From here, even the most-rudimentary form of rendering (using vanilla HTML and native DOM methods) becomes possible, and this is a good thing: you get away with less tooling until you absolutely need to add up on tooling!

However, since the document objects in Webflo natively support OOHTML - unless disabled in config, we are able to write reactive UI logic! Here, OOHTML makes it possible to embed reactive <script> elements (called Subscript) right within HTML elements - where each expression automatically self-updates whenever references to data, or its properties, get an update!

 <!--
 public
  ├── index.html
 -->
 <!DOCTYPE html>
 <html>
     <head>
         <title></title>
     </head>
     <body>
         <h1></h1>
         <script type="subscript">
          let { title } = document.state.data;
          document.title = title;
          let h1Element = this.querySelector('h1');
          h1Element.innerHTML = title;
         </script>
     </body>
 </html>

So, this is simple to think about: HTML already just let's us embed <script> elements for UI logic, and so be it! What OOHTML does further is simply to extend the plain old <script> element with the subscript type which gets any JavaScript code to be reactive! Compared with other syntax alternatives, this uniquely enables us to do all things logic in the actual language for logic - JavaScript.

Note that because these scripts are naturally reactive, we do not require any setTimeout() construct like we required earlier in the case of the classic <script> element. These expressions self-update as the values they depend on become available, removed, or updated - i.e. as document.state gets updated.

From here, we are also able to write more succinct code! Using the Namespaced HTML feature in OOHTML, we could do without those querySelector() calls up there. Also, we could go on to use any DOM manipulation library of our choice; e.g jQuery, or even better, the jQuery-like Play UI library.

 <!--
 public
  ├── index.html
 -->
 <!DOCTYPE html>
 <html>
     <head>
         <title></title>
         <script src="/jquery.js"></script>
     </head>
     <body namespace>
         <h1 id="headline1"></h1>
         <script type="subscript">
          let { title } = document.state.data;
          document.title = title;
          let { headline1, headline2 } = this.namespace;
          $(headline1).html(title);
          if (headline2) {
              $(headline2).html(title);
          }
         </script>
     </body>
 </html>

Above, we've also referenced some currently non-existent element headline2 - ahead of when it becomes added in the DOM! This should give a glimpse of the powerful reactivity we get with having OOHTML around on our document!

setTimeout(() => {
    let headline2 = document.createElement('h2');
    headline2.id = 'headline2';
    document.body.append(headline2);
}, 1000);

Taking things further, it is possible to write class-based components that abstract away all logic! You can find a friend in Custom Elements! Plus, your Custom Elements can function reactively using SubscriptElement as base class!

Custom Render Functions

Custom render functions can be defined on a route (export function render() {}) to entirely handle, or extend, rendering.

/**
server
 ├── index.js
 */
export default async function(event, context, next) {
    return { title: 'Home | FluffyPets' };
}
export async function render(event, data, next) {
    return `
    <!DOCTYPE html>
    <html>
        <head><title>FluffyPets</title></head>
        <body>
            <h1>${ data.title }</h1>
        </body>
    </html>
    `;
}

/**
server
 ├── index.js
 */
export async function render(event, data, next) {
    // For render callbacks at child step
    if (next.stepname) {
        return next();
    }
    return `
    <!DOCTYPE html>
    <html>
        <head><title>FluffyPets</title></head>
        <body>
            <h1>${ data.title }</h1>
        </body>
    </html>
    `;
}

Note Typically, though, child steps do not always need to have an equivalentrender callback being that they automatically inherit rendering from their parent or ancestor.

But, custom render functions do not always need to do as much as entirely handle rendering. It is possible to get them to trigger Webflo's native rendering and simply modify the documents being rendered. Here, you would simply call the next() function to advance the render workflow into Webflo's default rendering. A window instance is returned containing the document being rendered.

/**
server
 ├── index.js
 */
export default async function(event, context, next) {
    return { title: 'Home | FluffyPets' };
}
export async function render(event, data, next) {
    let window = await next( data );
    let { document } = window;
    console.log( document.state.data ); // { title: 'Home | FluffyPets' }
    return window;
}

Custom render functions must return a value, and window objects are accepted. (Actually, any object that has a toString() method can be returned.)

The Idea of State

There often needs to be a central point in an application where things are stored and managed. You could think of it as having a global object initialized window.store = {} on which different parts of an application can store and retrieve values. This is the basic idea of state. But it also doesn't go without the idea of observability - something that lets the different parts of the application observe and respond to changes made on this object!

State and Observability in Webflo applications come down to the same basic form:

first, an object...

document.state = {};
// and for elements: element.state = {};

...and then, a way to observe property changes on it...

Observer.observe(document.state, changes => {
    changes.forEach(change => {
        console.log(change.name, change.value);
    });
});

Observer.observe(document.state, propertyName, change => {
    console.log(change.name, change.value);
});

...and with incredible syntax benefits - where all references to the object and its properties from within embedded Subscript code are reactive.

<script type="subscript">
    // Always log the value of this property in realtime
    console.log(document.state.propertyName);
</script>

Look, this covers it all: state management and reactive UIs, with none of a state machine!

Interestingly, this idea of "state" is all of OOHTML's: the State API that's natively available in OOHTML-based documents - both client-side and server-side! Webflo simply just leverages it!

This API exposes an application-wide document.state object and a per-element element.state object. And these are live read/write objects that can be observed for property changes using the Observer API. It comes off as the simplest approach to state and reactivity!

Note The State API is available as long as the OOHTML support level in config is left as full, or set to scripting.

The `document.state.data` Object

This property reperesents the application data at any point in time - obtained from route handers on each navigation. Webflo simply updates this property and lets the page's rendering logic, or other parts of the application, take over.

console.log(document.state.data) // { title: 'Home | FluffyPets' }

Observer.observe(document.state, 'data', e => {
    console.log('Current page data is: ', e.value);
});

<script type="subscript">
 let { data: { title } } = document.state;
 document.title = title;
</script>

The `document.state.url` Object

This is a live object that reperesents the properties of the application URL at any point in time. The object exposes the same URL properties as of a standard URL object, but, here, as live properties that can be observed as navigation happens, and modified to initiate navigation - all using the Observer API.

let { url } = document.state;
console.log(url) // { hash, host, hostname, href, origin, password, pathname, port, protocol, search, searchParams, username }

Observer.observe(url, 'hash', e => {
    console.log(url.hash === e.value); // true
});

// Navigates to "/login#form" as if a link was clicked
document.addEventListener('synthetic-navigation', e => {
    Observer.set(url, 'href', '/login#form');
});

// Or...
document.addEventListener('synthetic-navigation', e => {
    Observer.set(url, { pathname: '/login', hash: '#form' });
});

console.log(url.hash); // #form

There is also the convenience query property that offers the URL parameters as a live object.

// For URL: http://localhost:3000/login?as=student
console.log(url.query.as) // student

// Re-rewrite the URL and initiate navigation by simply modifying a query parameter
document.addEventListener('synthetic-navigation', e => {
    Observer.set(url.query, 'as', 'business');
});

<script type="subscript">
 let { query: { as: role } } = url;
 document.title = 'Login as ' + role;
</script>

Requests and Responses

On each request, the event object passed to route handlers exposes the incoming request as event.request. This is an instance of event.Request - an extension of the WHATWG Request class. The event object also exposes event.Response - an extension of the WHATWG Response class, for returning instance-based responses. You enjoy routing that is based on standard interfaces!

Routes in Webflo can be designed for different types of request/response scenarios. Here are some important ones:

Scenario 1: Static File Requests and Responses

Static file requests like http://localhost:3000/logo.png are automatically responded to by Webflo when next()ed forward by route handlers, or where there are no route handlers.

On the server, Webflo serves files from the public directory. File contents along with the appropriate headers like Content-Type, Content-Length, etc. are returned as an instance of event.Response. Where a request has an Accept-Encoding header set (e.g. gzip, br) and there exists a matching compressed version of the said file on the file system (e.g. ./public/logo.png.gz, ./public/logo.png.br), the compressed version is served and the appropriate Content-Encoding response header is set.

Scenario 2: API Requests and Responses

JSON (API) requests - requests that expect to get a JSON response (i.e. Content-Type: application/json) - are automatically satisfied by Webflo with a valid JSON response. Here, Webflo simply jsonfies workflow return values - which are usually plain objects, or other jsonfyable types - string, number, boolean, array.

These requests need not have an Accept header; but if they should, it must be the value of application/json.
Routes intended to be accessed this way are expected to return a jsonfyable value (or an instance of event.Response containing same) from the workflow.
Workflow responses that are an instance of event.Response with a Content-Type header already set are sent as-is.

Scenario 3: Page Requests and Responses

HTML page requests - requests that expect to get an HTML response (i.e. Content-Type: text/html) - are automatically satisfied by Webflo with a valid HTML response. Workflow return values that are objects are automatically used for Server-Side Rendering.

These requests need to have an Accept header of text/html, or something that resolves to text/html - e.g. text/*, */html, */*.
Routes intended to be accessed this way are expected to return a plain object (or an instance of event.Response containing same) from the workflow in order to be renderable.
Workflow responses that are an instance of event.Response with a Content-Type header already set are sent as-is, and not rendered.

Scenario 4: Single Page Navigation Requests and Responses

In a Single Page Application layout, every navigation event (page-to-page navigation, history back and forward navigation, and form submissions) is expected to initiate a request/response flow without a full page reload, since the destination URLs are often based off the already loaded document. The Webflo client JS intercepts these navigation events and generates the equivalent request object with an Accept header of application/json, so that data can be obtained as a JSON object (scenerio 2 above) for Client-Side Rendering.

The generated request also hints the server on how to return cross-SPA redirects (redirects that will point to another origin, or to another SPA root (in a Multi SPA layout)) so that it can be handled manually by the client. The following headers are set: X-Redirect-Policy: manual-when-cross-spa, X-Redirect-Code: 200.

Same-SPA redirects are sent as-is, and the Webflo client JS receives and renders the final data and updates the address bar with the final URL.
Cross-SPA/cross-origin redirects are communicated back, as hinted, and the destination URL is opened as a fresh page load.

Scenario 5: Range Requests and Responses

In all cases, where a request specifies a Range header, Webflo automatically slices the response body to satisfy the range, and the appropriate Content-Range response header is set.

Workflow responses that are an instance of event.Response with a Content-Range header already set are sent as-is.

Other Requests and Responses

Workflows may return any other data type, e.g. an instance of the native FormData, Blob, File, or ReadableStream, etc., or an instance of event.Response containing same - usually on routes that do not double as a page route. Webflo tries to send these along with the appropriate response headers.

Note The fact that all requests, even static file requests, are seen by route handlers, where defined, means that they get a chance to dynamically generate the responses that the client sees!

Custom Redirect Responses

It is possible to hint the server on how to serve redirect responses. The idea is to substitute the standard 302, 301 response code for these redirects with a non-rediret status code so that it can be recieved as a normal response and handled manually. The following pair of headers make this possible: X-Redirect-Code, X-Redirect-Policy.

The X-Redirect-Code can be any valid HTTP status code (often preferably, in the 2xx). This is the response code that you want Webflo to substitute the actual redirect code with.
The X-Redirect-Policy header can be any of:
- manual - which means "treat all redirects as manual"
- manual-if-cross-origin - which means "treat cross-origin redirects as manual"
- manual-if-cross-spa - which means "treat cross-SPA redirects (including cross-origin redirects) as manual"
In each case, the substituted, original redirect code is returned back in the response in a special X-Redirect-Code response header, alongside the standard Location header.

Failure Responses

Where workflows return undefined, a Not Found status is implied.

On the server side, a 404 HTTP response is returned.
On the client-side, the initiating document in the browser has its document.state.data emptied. The error is also exposed on the document.state.network.error property.

Where workflows throw an exception, an error status is implied.

On the server side, the error is logged and a 500 HTTP response is returned.
On the client-side, the initiating document in the browser has its document.state.data emptied. The error is also exposed on the document.state.network.error property.

Cookie Responses

Handlers can set response cookies via the standard Response constructor, or using the standard Headers.set() method.

let response = event.Response(data, { headers: { 'Set-Cookie': cookieString }});

response.headers.set('Set-Cookie', cookieString);

Webflo also offers a convenience method.

let response = event.Response(data, { headers: { cookies: cookieString }});

response.headers.cookies = { 'Cookie-1': cookieString, 'Cookie-2': cookie2String };

let cookieObject = { value: 'cookie-val', expires, maxAge, domain, path, secure, HttpOnly, sameSite };
let cookie2Object = { value: 'cookie2-val' };
response.headers.cookies = { 'Cookie-1': cookieObject };
response.headers.cookies = { 'Cookie-2': cookie2Object };

console.log(response.headers.cookies); // { 'Cookie-1': cookieObject, 'Cookie-2': cookie2Object };

Set cookies are accessed on the next request via request headers.

console.log(event.request.headers.get('Cookie')); // Cookie-1=cookie-val&Cookie-2=cookie2-val;

Webflo also offers a convenience method.

console.log(event.request.headers.cookies); // { 'Cookie-1': 'cookie-val', 'Cookie-2': 'cookie2-val' };

Webflo Applications

Webflo comes ready for any type of application!

And you can go hybrid!

Client-Side Applications

Web pages that embed the Webflo client JS deliver a great user experience. It's simple: the npm run generate command does both the building and embedding of the script (or scripts), for the document root (or document roots - in a Multi Page / Multi SPA layout)!

On being loaded, the state of the application is initialized, or is restored through hydration - where Server-Side Rendering was involved to optimize for first paint, and an app-like experience kicks in! For Single-Page Applications, Client-Side Rendering is performed on each navigation.

SPA Navigation

Unless disabled in config, it is factored-in at build time for the application client JS to be able to automatially figure out when to intercept a navigation event and prevent a full page reload, and when not to.

SPA Navigation follows the following rules:

When it ascertains that the destination URL is based on the current running index.html document in the browser (an SPA architecture), a full page reload is prevented for soft navigation. But where the destination URL points out of the current document root (a Multi SPA architecture), navigation is allowed as a normal page load, and a new page root is loaded.
If navigation is initiated with any of the following keys pressed: Meta Key, Alt Key, Shift Key, Ctrl Key, navigation is allowed to work the default way - regardless of the first rule above.
If navigation is initiated from a link element that has the target attribute, or the download attribute, navigation is allowed to work the default way - regardless of the first rule above.
If navigation is initiated from a form element that has the target attribute, navigation is allowed to work the default way - regardless of the first rule above.

{ "spa_navigation": true }

File: .webqit/webflo/client.json

Command: webflo config client spa_navigation=TRUE

SPA State

On the client side of a Webflo application, the idea of state also includes the following aspects of the client-side lifecycle that can be used to provide visual cues on the UI.

The `document.state.network` Object

This is a live object that exposes the network activity and network state of the application.

console.log(document.state.network) // { requesting, remote, error, redirecting, connectivity, }

This property tells when a request is ongoing, in which case it exposes the params object used to initiate the request.

On the UI, this could be used to hide a menu drawer that may have been open.

<menu-drawer>
    <script type="subscript">
    let { network: { requesting } } = document.state;
    if (requesting) {
        $(this).attr('open', false);
    }
    </script>
</menu-drawer>

This property tells when a remote request is ongoing - usually the same navigation requests as at network.requesting, but when not handled by any client-side route handlers, or when next()ed to this point by route handlers. The remote property also goes live when a route handler calls the special fetch() function that they recieve on their fourth parameter.

On the UI, this could be used to show/hide a spinner, or progress bar, to provide a visual cue.

<progress-bar>
    <script type="subscript">
    let { network: { remote } } = document.state;
    $(this).attr('hidden', !remote);
    </script>
</progress-bar>

This property tells when a request is errored in which case it contains an Error instance of the error. For requests that can be retried, the Error instance also has a custom retry() method.

On the UI, this could be used to show/hide cute error elements.

<nice-error>
    <script type="subscript">
    let { network: { error } } = document.state;
    $(this).attr('hidden', !error);
    </script>
</nice-error>

This property tells when a client-side redirect is ongoing - see