We learn as we go, we write as we learn.

About michelada.io


By joining forces with your team, we can help with your Ruby on Rails and Javascript work. Or, if it works better for you, we can even become your team! From e-Commerce to Fintech, for years we have helped turning our clients’ great ideas into successful business.

Go to our website for more info.

Tags


How do I write React

3rd February 2020

I've write a lot of Javascript through my short career, and one the many things that I have learned is that, bad code still works on Javascript.

What I like about Javascript is how flexible it is, you can split your code into different files, put them anywhere you want, pass the path to the require() function and you are good to go, you don't need a strict folder structure for the app to be working.

This is good when you are been introduced to programming, but this is also one of the reason why people don't get along with Javascript.

React doesn't help with this problem, in fact it could complicate things even more, that's why I've decided to follow some rules when writing React.

I describe them in this blog post, some of them might seem odd to you, some of them not, take what works for you and pass it to the next!

Javascript linter

Even though this post is about React, we can't write clean React code, without writing clean Javascript code first.

Every time I write Javascript I like to use a linter to check some basic code style rules.

I used to use ESLint, which is pretty common among JS developers, but then I found StandardJS.

StandardJS has some special rules (it doesn't use semi-colon, you can't ignore a rule), that make it look like really strict but it's precisely this set of standard rules what makes it great.

It's really easy to use, all you have to do is add it to your project:

yarn add standard
// or
// npm install --save standard
  • You don't need a configuration file to start using the linter. There's a lot of discussions about programming style in js (tabs vs spaces, semi-colons) that's why I like standard, because it picks some 'standard' rules and that's all you have, no more, no less therefore no more discussions.

  • Automatically fix your problems (just use the --fix flag). I always create 2 npm scripts: npm run lint and npm run lint:fix.The first script is what I use more often, that shows all the errors with information about the line and file where they occurs.The second script is just to automatically fix the common errors, but I still try to manually fix as much as possible.

  • Git pre-commit hook.Sometimes when I'm more strict about the programming style, I create a pre-commit hook for the project, that can save some review time.

Initializing a react project

The npm registry is one of the biggest databases of public and private libraries for javascript. It offers a command line client to use all the features like, download, upload and do some other things, that allow you to interact with several javascript packages.

There's one particular package that I use a lot when creating a react app: npx.

This lib allows you to execute the package binaries files, it executes the <command> you passed in, following an specific order:

  • Local node_modules/.bin folder
  • A central cache
  • Install it and execute

Why npx instead of npm?

Honestly, I just don't like to struggle with all the versions and deprecations stuff, with npx I ensure that the library that I'm using is on the latest stable version.

npx <command>

You can use npm to download React and start your project from scratch (configure Webpack to do all the fancy stuff you are used to have 'out-of-the-box') or you can use create-react-app (CRA).

This library allow you to bootstrap a React project. It has all the configurations needed for a rapid development with React (like hot reload, ES6 support, etc...)

One of the good things about CRA, is the ability to have all the latest patches and features with a single bump version of your react-scripts dependency (which is what CRA uses), so you don't have to worry for this anymore. CRA also allows you to have your custom setup by ejecting your application, this action will give you the full control of the Webpack configuration, so you can twist it and do whatever you want with it.

npx & create-react-app

So now you know all the benefits of using npx and create-react-app, you can start figuring out how we can mix up these 2 libraries to simplify the creation of a React application.

Each time I start a new React project I just run:

npx create-react-app awesomeApp

That command will download (the latest-stable version) CRA and executes it, that's why we need to pass the name of the project we want to use (awesomeApp).

Organize app structure

CRA offers a very basic folder structure for your app:

 awesomeApp/
  |__public/
  |__src/
    |__App.css
    |__App.js
    |__App.test.js
    |__index.css
    |__index.js
    |__logo.svg
    |__serviceWorker.js
    |__setupTests.js
  |__.gitignore
  |__package.json
  |__README.md
Some apps that I have in production are so simple that this folder structure is enough to work.

When I know that a project will be a little more complicated than that, I change the folder structure so that it's easy for me or any other developer make changes.

I split my components in two types:

  • Components
  • Containers

Following this idea, the folder structure that I use, looks like pretty much like this:

 awesomeApp/
  |__public/
  |__src/
    |__components/
      |__ui/
    |__containers/
    |__utils/
    |__App.css
    |__App.js
    |__App.test.js
    |__index.css
    |__index.js
    |__logo.svg
    |__serviceWorker.js
    |__setupTests.js
  |__.gitignore
  |__package.json
  |__README.md
Components

This is where I put all my UI components, that means, components that doesn't do to much of a logic, they are just there to present some information to the user and depend a little bit from the props that we passed in.

The ui/ folder holds all the components that are related to the User Interface (i.e. custom component for commons elements like <CustomInput /> instead of <input /> or <CustomImg /> instead of <img />)

Containers

This is where I put the smart components. A smart component is the one that controls the state of a particular part of the app. I use this type of components to wrap the most of the base markdown of the pages.

Also I create a new folder called utils/, which I use for all the utility functions that I'd use across the app.

Note: When I work with React Native, I just rename the containers/ folder to pages/, seems better name for me.

Organize Code

The index.js file contains all the code that register the service working and also render your app. So this file is basically your entry point, I suggest not to touch this file unless you really have to.

Then we have the App.js file, which is the React component that's been rendered on the index.js file. I use this file as my main React file, and I try to keep it as simple as possible.

Most of my App.js file look like this:

import React from 'react'

import MainContainer from './containers/MainContainer'

function App() {
  return <MainContainer />
}

export default App

We can point up some things here:

  • 1) It's a functional component instead of a class component.
  • 2) It does nothing but render a main container component
  • 3) Export a default function which is the actual component
1) It's a functional component instead of a class component:

I used to use class components all the time, so that I could have an state and control everything with the lifecycles of React but since hooks came out, all my components started to shrink a lot, and I liked that, so I haven't needed a class component anymore.

2) It does nothing but render a main container component:

I always try to keep this component clean, unless I need some initial data that's coming from outside (i.e. API calls). So this will only return the main container, which will have all the business logic.

I often use this function to wrap my app with a High Order Component (HOC), like react router or any css theme, so that's available for any children component.

3) Export a default function which is the actual component

Each time I jump to an existing project and try to figure out all the imports a single file is doing, is really annoying to be searching if there's any export in a particular line, or if they default export a function that's been declare in line 128, that's why I prefer to have all my exports at the end of the file, so each time I wanted to see what's been exported, I just go the end of the file.

Props and State

I used to use class components for my containers/pages and functional components for all other component, this way I could separate the concerns for each type of component.

Now since hooks are live, I found myself writing cleaner components using functional components and hooks.

Class components

A simple class component of my own looks like this:

import React from 'react'

class HomeContainer extends React.Component {

  state = {}

  componentDidMount() {
    // Initialization of component's state
  }

  customMethod = () => {
    // 'this' is safe
  }

  render() {
    const { prop1, prop2 } = this.props

    // Render anything
  }
}

export default HomeContainer

First, I import React, some people use destructuring to import Component also, I use the React variable since Component is available as a property of the default export of react.

I also don't use constructor, I prefer to use class properties to define the state, and use the lifecycles of react to fetch initial data or to update state on renders.

I've always though that the use of this in javascript is really hardcore, but I liked though, seems to me like if you had all the javascript wisdom just because .bind is in your code.
I change all of that when working with React, (even tough I still think that using this is cool if that solves your problem) instead of the regular method declaration of the classes I use an arrow function assignment, so this keyword works as expected and looks cleaner.

The first line of the render() method is always the desctructuring of all the props of the component, so next time I come across this component, I know easily which props I'm using without having to dig into all the jsx code (which supposed to be clean).

And last but not least, I export the component at the end of the file.

Functional components

For functional components I follow kinda the same rules:

import React, { useEffect } from 'react'

function HomeContainer(props) {
  const { prop1, prop2 } = props

  // '[]' == no deps == one execution == componentDidMount
  useEffect(() => {
    // Initialization of component's 'state'

  }, []) 

  return (
    // All the render
  )
}

export default HomeContainer

So I still use the same destructuring-first technique for the props.

When I need to do some state initialization of my functional components (i.e using useState hook) I use the useEffect hook, which is the replacement for the lifecycles on the class components.

Finally I export my component at the end of file.

Handle JSX

JSX is the syntax extension for javascript, it looks like html tags, and allows you to manipulate the UI of your components.

There are some rules when using JSX though, one of the most known rule is the use of className instead of class for the html tag property, this is because the special keyword class in Javascript represents a class declaration and it's reserved.

Another special rule for jsx is that it doesn't allow multiple elements to be rendered, something like this:

import React from 'react'

function CheckBox(props) {

  return (
    <label>
      Checkbox
    </label>
    <input type="checkbox" value="1" />
  )
}

This component is not jsx-valid, since you can't render multiple elements from a React component, instead you have to wrap all the content within a parent element. Most people use a div

import React from 'react'

function CheckBox(props) {

  return (
    <div>
      <label>
        Checkbox
      </label>
      <input type="checkbox" value="1" />
    </div>
  )
}

This works perfectly mostly of the time, but there are some special cases where this could be an issue (i.e. inside of a table row, you can't have a div element as a child), so for those cases, the React team
build Fragment.

With Fragment you can safely return multiple elements without worrying about the semantic of the html

import React from 'react'

function CheckBox(props) {

  return (
    <React.Fragment> // <>
      <label>
        Checkbox
      </label>
      <input type="checkbox" value="1" />
    </React.Fragment> // </>
  )
}

There's a shortcut for Fragment that you can use instead: <> ... </> but you have to choose when to use it, since this shortcut doesn't accept any prop while the Fragment component let you use the key prop, which is helpful when creating elements within a loop.

Organize your dependencies

When I started working with javascript, I loved how the community helps to solve any kind of issue. Almost anything you would need when creating an app is likely to have it's own library/sdk than can help you with that.
At first glance, that's good, but it can lead you to a laziness development, where you are used to find a library for nearly any feature you'd need, so when you don't find the library, you start thinking that it might be hard to do (at least that was what I though :sad:).

In order to remove that bad habit of depend a lot of my dependencies (that's what the name stands for ??), I started to take a look to the code that I included on my projects, and that's how I realize that some of it it's really simple that it might wont worth to be included, and can be just a new file in the utils/ folder.

I also try to think twice before install a dependency that's kinda small (I used to include momentjs on each project that I needed to present a simple formatted date) so my node_modules/ folder doesn't grow up too much.

Versioning on your dependencies

Versionin is a huge problem on the Javascript environment (I supposed that all languages have this problem). You install the version 9.x of a dependency, and it works perfectly on your React 16.3, but then after a few months (or even weeks) in production, a new version of that library came out, and you just deploy normally to production, then npm install do it's job. Next, you have a white screen (no matter how many times you reload the page) presented to your users, ugh!!

With npm, you can use Range versions to control the version of your dependencies, by default it's configure to use the caret range, that means patch and minor updates are allowed

^1.0.0 => 1.x.x
~1.0.0 => 1.x.0

So when you install a package, your package json looks pretty much like this:

"dependencies": {
  "react": "^16.3.1",
  "momentjs": "^4.3.1",
  // ...
}

Even though that minor and patch updates should not break your code, not everyone follow that rule, and sometimes you can struggle with that for a long time without notice that it's because of the library version.

That's why I lock the version of my dependencies (I just remove the caret or the tilde), so whenever I do a npm install again, the same version of the dependency will be downloaded.

Of course, doing this will require that you keep up-to-date with the dependencies that are likely to be updated often.

Wire up

One of the reasons why Javascript is well adopted, is the flexibility on how to write code, it doesn't have an explicit convention on how to do it, but that can lead to technical debt when doing it, that's why I stick to some rules when working with Javascript, and also why you should do it too, the future yourself (or any developer you work with), will thank you for that.

I presented you a small style guide that I follow when working with React, you can use it, or twist at your convenience, whatever makes you feel happy when programming!

Thanks for reading and happy coding!

View Comments