Code for this chapter available here.
In this section we are going to create the server that will render our web app. We will also set up a development mode and a production mode for this server.
💡 Express is by far the most popular web application framework for Node. It provides a very simple and minimal API, and its features can be extended with middleware.
Let's set up a minimal Express server to serve an HTML page with some CSS.
- Delete everything inside
src
Create the following files and folders:
- Create a
public/css/style.cssfile containing:
body {
width: 960px;
margin: auto;
font-family: sans-serif;
}
h1 {
color: limegreen;
}-
Create an empty
src/client/folder. -
Create an empty
src/shared/folder.
This folder is where we put isomorphic / universal JavaScript code – files that are used by both the client and the server. A great use case of shared code is routes, as you will see later in this tutorial when we'll make an asynchronous call. Here we simply have some configuration constants as an example for now.
- Create a
src/shared/config.jsfile, containing:
// @flow
export const WEB_PORT = process.env.PORT || 8000
export const STATIC_PATH = '/static'
export const APP_NAME = 'Hello App'If the Node process used to run your app has a process.env.PORT environment variable set (that's the case when you deploy to Heroku for instance), it will use this for the port. If there is none, we default to 8000.
- Create a
src/shared/util.jsfile containing:
// @flow
// eslint-disable-next-line import/prefer-default-export
export const isProd = process.env.NODE_ENV === 'production'That's a simple util to test if we are running in production mode or not. The // eslint-disable-next-line import/prefer-default-export comment is because we only have one named export here. You can remove it as you add other exports in this file.
- Run
yarn add express compression
compression is an Express middleware to activate Gzip compression on the server.
- Create a
src/server/index.jsfile containing:
// @flow
import compression from 'compression'
import express from 'express'
import { APP_NAME, STATIC_PATH, WEB_PORT } from '../shared/config'
import { isProd } from '../shared/util'
import renderApp from './render-app'
const app = express()
app.use(compression())
app.use(STATIC_PATH, express.static('dist'))
app.use(STATIC_PATH, express.static('public'))
app.get('/', (req, res) => {
res.send(renderApp(APP_NAME))
})
app.listen(WEB_PORT, () => {
// eslint-disable-next-line no-console
console.log(`Server running on port ${WEB_PORT} ${isProd ? '(production)' : '(development)'}.`)
})Nothing fancy here, it's almost Express' Hello World tutorial with a few additional imports. We're using 2 different static file directories here. dist for generated files, public for declarative ones.
- Create a
src/server/render-app.jsfile containing:
// @flow
import { STATIC_PATH } from '../shared/config'
const renderApp = (title: string) =>
`<!doctype html>
<html>
<head>
<title>${title}</title>
<link rel="stylesheet" href="${STATIC_PATH}/css/style.css">
</head>
<body>
<h1>${title}</h1>
</body>
</html>
`
export default renderAppYou know how you typically have templating engines on the back-end? Well these are pretty much obsolete now that JavaScript supports template strings. Here we create a function that takes a title as a parameter and injects it in both the title and h1 tags of the page, returning the complete HTML string. We also use a STATIC_PATH constant as the base path for all our static assets.
It might be possible to get syntax highlighting working for HTML code inside template strings depending on your editor. In Atom, if you prefix your template string with an html tag (or any tag that ends with html, like ilovehtml), it will automatically highlight the content of that string. I sometimes use the html tag of the common-tags library to take advantage of this:
import { html } from `common-tags`
const template = html`
<div>Wow, colors!</div>
`I did not include this trick in the boilerplate of this tutorial, since it seems to only work in Atom, and it's less than ideal. Some of you Atom users might find it useful though.
Anyway, back to business!
- In
package.jsonchange yourstartscript like so:"start": "babel-node src/server",
🏁 Run yarn start, and hit localhost:8000 in your browser. If everything works as expected you should see a blank page with "Hello App" written both on the tab title and as a green heading on the page.
Note: Some processes – typically processes that wait for things to happen, like a server for instance – will prevent you from entering commands in your terminal until they're done. To interrupt such processes and get your prompt back, press Ctrl+C. You can alternatively open a new terminal tab if you want to keep them running while being able to enter commands. You can also make these processes run in the background but that's out of the scope of this tutorial.
💡 Nodemon is a utility to automatically restart your Node server when file changes happen in the directory.
We are going to use Nodemon whenever we are in development mode.
-
Run
yarn add --dev nodemon -
Change your
scriptslike so:
"start": "yarn dev:start",
"dev:start": "nodemon --ignore lib --exec babel-node src/server",start is now just a pointer to an other task, dev:start. That gives us a layer of abstraction to tweak what the default task is.
In dev:start, the --ignore lib flag is to not restart the server when changes happen in the lib directory. You don't have this directory yet, but we're going to generate it in the next section of this chapter, so it will soon make sense. Nodemon typically runs the node binary. In our case, since we're using Babel, we can tell Nodemon to use the babel-node binary instead. This way it will understand all the ES6/Flow code.
🏁 Run yarn start and open localhost:8000. Go ahead and change the APP_NAME constant in src/shared/config.js, which should trigger a restart of your server in the terminal. Refresh the page to see the updated title. Note that this automatic restart of the server is different from Hot Module Replacement, which is when components on the page update in real-time. Here we still need a manual refresh, but at least we don't need to kill the process and restart it manually to see changes. Hot Module Replacement will be introduced in the next chapter.
💡 PM2 is a Process Manager for Node. It keeps your processes alive in production, and offers tons of features to manage them and monitor them.
We are going to use PM2 whenever we are in production mode.
- Run
yarn add --dev pm2
In production, you want your server to be as performant as possible. babel-node triggers the whole Babel transpilation process for your files at each execution, which is not something you want in production. We need Babel to do all this work beforehand, and have our server serve plain old pre-compiled ES5 files.
One of the main features of Babel is to take a folder of ES6 code (usually named src) and transpile it into a folder of ES5 code (usually named lib).
This lib folder being auto-generated, it's a good practice to clean it up before a new build, since it may contain unwanted old files. A neat simple package to delete files with cross platform support is rimraf.
- Run
yarn add --dev rimraf
Let's add the following prod:build task to our package.json:
"prod:build": "rimraf lib && babel src -d lib --ignore .test.js",-
Run
yarn prod:build, and it should generate alibfolder containing the transpiled code, except for files ending in.test.js(note that.test.jsxfiles are also ignored by this parameter). -
Add
/lib/to your.gitignore
One last thing: We are going to pass a NODE_ENV environment variable to our PM2 binary. With Unix, you would do this by running NODE_ENV=production pm2, but Windows uses a different syntax. We're going to use a small package called cross-env to make this syntax work on Windows as well.
- Run
yarn add --dev cross-env
Let's update our package.json like so:
"scripts": {
"start": "yarn dev:start",
"dev:start": "nodemon --ignore lib --exec babel-node src/server",
"prod:build": "rimraf lib && babel src -d lib --ignore .test.js",
"prod:start": "cross-env NODE_ENV=production pm2 start lib/server && pm2 logs",
"prod:stop": "pm2 delete all",
"test": "eslint src && flow && jest --coverage",
"precommit": "yarn test",
"prepush": "yarn test"
},🏁 Run yarn prod:build, then run yarn prod:start. PM2 should show an active process. Go to http://localhost:8000/ in your browser and you should see your app. Your terminal should show the logs, which should be "Server running on port 8000 (production).". Note that with PM2, your processes are run in the background. If you press Ctrl+C, it will kill the pm2 logs command, which was the last command our our prod:start chain, but the server should still render the page. If you want to stop the server, run yarn prod:stop
Now that we have a prod:build task, it would be neat to make sure it works fine before pushing code to the repository. Since it is probably unnecessary to run it for every commit, I suggest adding it to the prepush task:
"prepush": "yarn test && yarn prod:build"🏁 Run yarn prepush or just push your files to trigger the process.
Note: We don't have any test here, so Jest will complain a bit. Ignore it for now.
Next section: 04 - Webpack, React, HMR
Back to the previous section or the table of contents.