Surplus SSR

This library exposes a relatively simple implementation of SSR using Surplus.

Installation

Either install through npm:

npm i --save surplus-ssr

or clone this repository and import index.js.

Usage

The default export of the library is a function that produces a middleware function for use with server applications (e.g. express).

ssr(rootPath, getState, [options])

Return a middleware function for SSR using the given root path (path to the root of the content to server) and the given getState function, which should return the current state given the request object (described more in Writing Content).

Options may be an object with the following keys:

• clientJS (boolean) - whether to emit client-side javascript (default: true)
• pageRoot (string) - the root, relative to rootPath, of pages and directories to serve (default: "pages")
• compile (array) - the steps to take when compiling files (default: [ssr.COMPILE_SURPLUS]). Should be an array of functions that take the file source as an argument and returns the modified/compiled source. The special value ssr.COMPILE_SURPLUS can also be specified, which will use the surplus compiler installed in rootPath. Compiled results are cached.

The middleware function returned is of the typical form (taking two arguments, a request and a response). It does not use the common third argument (a chaining function), so technically isn't middleware per-se. This could easily be changed if requested.

The function uses the version of surplus installed in rootPath to compile the surplus expressions, and runs any code within rootPath (node modules are loaded from rootPath).

f.middleware

The function returned by the above also has a property called middleware that acts as true middleware and adds a respondWithPage function to the response object. This function makes less assumptions about the request path to resource/entrypoint mapping, and is useful if you don't like the default way that routing is done. It must be called with the path to the entrypoint as an argument, where the path is relative to rootPath. Any additional arguments are forwarded to the function exported by the page, if possible. This will run the code at the given path and appropriately bundle and send the response. For example:

var srv = express();

srv.use(ssr("path/to/root", getState).middleware);

srv.get('/my-path', (req, res) => {
    res.respondWithPage("my-path-page.js", "page-argument");
});

rootPath setup

All modules and content needed to render pages should appear within rootPath. For this reason, it's almost always appropriate to initialize a node project within rootPath, and within that project install s-js and surplus. These are required for the SSR to function properly.

If using the default ssr middleware function, page entry-points are expected in the options.pageRoot directory (default "pages"). So this directory must exist, and should contain the request path layout that you expect. The default middleware does the following mapping:

your_url.com/page -> {options.pageRoot}/page.js | {options.pageRoot}/page/index.js

In words: if the page exists as a javascript file (with the .js extension) it is used, otherwise if it exists as a directory and index.js exists in the directory, that is used. Otherwise the request fails.

Writing Content

All code that is used for the webpage must use CommonJS/UMD modules. The loader has not yet been extended to support the experimental node ES6 modules. The code must also be appropriate to run directly on the client side. I have not used pre-processing in my workflows, but in theory that should be possible. Importing modules must be done using require() just like other nodejs code.

One important difference right now is that require() is always relative to rootPath; relative paths from the current file are not yet supported. In my opinion, this makes some things much clearer. Importing external code from node_modules works as usual (e.g. require("d3")), but, for instance, if you have a components directory in rootPath, then importing files from there should always be done with require("components/path-to-file").

The following global variables will be available, both on the server side and the client side:

• S - the loaded S.js module.
• Surplus - the loaded Surplus module.
• STATE - the state returned from the getState function passed to the middleware constructor. The getState function is passed a single argument, the request object.
• isServer (boolean) - Whether the code is running on the server or not (client).

This means that your pages don't all need to require("s-js") and require("surplus"); you may just use them as if they were already imported.

Pages must export an object, or a function that returns an object, with the following keys:

• body - the body of the page, which must be a DOM element supporting outerHTML.
• head (optional) - an array of DOM elements (e.g. <link />) that should be in the head of the response.

If the page exports a function, that function will be called with any additional arguments passed to the middleware respondWithPage function.

I've used d3 for (fairly complex) server-side SVG rendering, and it seems to work well.

Client JS

If options.clientJS is true (the default), all code for the page is bundled with the page, and upon load (if the page supports javascript) the content of the page will be replaced with a 'live' surplus/s-js version. All dependencies of a page are tracked and bundled together, but no minification is done. If it is false, no javascript will be bundled, which can save on response size.

I've been able to use websockets and REST APIs without any issue here. It's fairly simple to create S.js computations around the original server state, and then update those with data that comes over websockets or from other events. I've found this approach to work quite well even with high update rates and fairly large state objects (on the order of tens of thousands of individual pieces of state, however you may define that...).

Compiling Code

By default, the surplus compiler is run on all source files. If options.compile is specified, those compile steps will be used on source files. This is useful, for instance, to run the Typescript compiler or Babel stages on source files when developing. Typically for release builds, you'll want to set options.compile to an empty array (or false) and pre-process/compile the files as part of the production build.

Example Usage

Server Code (using express)

const express = require('express');
const ssr = require('surplus-ssr');
const path = require('path');
const babel = require('babel-core');

const listen_port = 8080;
// Root is in the 'public' folder
const publicRoot = path.join(__dirname, "public");

// Create server
const srv = express();

// Create view state to be used to render the page
// This state provides an object defining the number of overall page loads and
// the load time.
let page_loads = 0;
const getState = () => {
    page_loads++;
    return {
        load_time: new Date().toString(),
        page_loads: page_loads
    };
};

// Babel compiler, using babel-preset-env to target a set of browsers and
// perform some minification.
const compile_babel = s => {
    return babel.transform(s, {
        ast: false,
        babelrc: false,
        comments: false,
        minified: true,
        presets: [
            ["env", {
                "targets": {
                    "browsers": [">0.5%", "not op_mini all", "not dead", "last 1 version"]
                }
            }]
        ]
    }).code;
};

// Middleware options
const ssropts = {
    compile: [ssr.COMPILE_SURPLUS, compile_babel]
};

// Render and serve page contents
// Could omit 'middleware', in which case the following srv.get() calls could be
// left out as well, and pages/index.js would work as expected. But pages/getId.js
// wouldn't get arguments in that situation.
srv.use(ssr(publicRoot, getState, ssropts).middleware);

srv.get('/', (req, res) => { res.respondWithPage("pages/index.js"); });
srv.get('/getId/:id', (req, res) => {
    res.respondWithPage("pages/getId.js", req.params.id);
});

// Run server
const inst = srv.listen(listen_port);

Client/Website Code

public/pages/index.js

Shows the number of page loads, the server-side page load time, and an updating time based on the client's clock (if JS is enabled).

const now = S.value(new Date());

if (!isServer) {
    // Update displayed time every second on the client
    setInterval(() => {
        now(new Date());
    }, 1000);
}

module.exports = {
    body: S.root(() => (
        <div>
            <h1>Hello, World</h1>
            <p>Number of loads: {STATE.page_loads}</p>
            <p>Loaded at: {STATE.load_time}</p>
            <p>Right now, it is {now()}</p>
        </div>
    ))
};

public/pages/getId.js

Echoes whatever the parameter to the page is.

module.exports = (id) => {
    return {
        body: <div><p>Id is {id}</p></div>
    };
};

Improvements

• Partially pre-render pages - it would be useful and powerful to pre-render all pages as much as possible, such that when a request comes in the minimal amount of processing needs to be done (i.e. things that depend on state are rendered). This will require changes to or a custom implementation of the server-side DOM.
• Re-hydrate on the client side rather than replacing all content - with some changes to the surplus compiler, it should be possible to use the DOM that was rendered on the server side with live S.js functions on the client side.
• Rewrite in Typescript - to better match the S.js and Surplus code.