Front-end infrastructure and code to complement mozilla/addons-server.

This code and its associated production website are included in Mozilla’s web and services bug bounty program. If you find a security vulnerability, please submit it via the process outlined in the program and FAQ pages. Further technical details about this application are available from the Bug Bounty Onramp page.

Please submit all security-related bugs through Bugzilla using the web security bug form.

Never submit security-related bugs through a Github Issue or by email.

• You need Node 10.x which is the current LTS (long term support) release.
• Install yarn to manage dependencies and run scripts.

The easiest way to manage multiple node versions in development is to use nvm.

• type yarn to install all dependencies
• type yarn amo:stage to start a local server that connects to a hosted staging server

Here are some commands you can run:

You can enter the interactive jest mode by typing yarn test. This is the easiest way to develop new features.

Here are a few tips:

• When you start yarn test, you can switch to your code editor and begin adding test files or changing existing code. As you save each file, jest will only run tests related to the code you change.
• If you had typed a when you first started then jest will continue to run the full suite even when you change specific files. Type o to switch back to the mode of only running tests related to the files you are changing.
• Sometimes running tests related to your file changes is slow. In these cases, you can type p or t to filter tests by name while you working fixing a specific test suite. More info.
• If you see something like Error watching file for changes: EMFILE on Mac OS then brew install watchman might fix it. See jestjs/jest#1767

By default, yarn test will only run a subset of tests that relate to the code you are working on.

To explicitly run a subset of tests, you can type t or p which are explained in the jest watch usage.

Alternatively, you can start the test runner with a specific file or regular expression, like:

yarn test tests/unit/amo/components/TestAddon.js

If you want to run all tests and exit, type:

yarn test-once

As you run tests you will see a report of Eslint errors at the end of the test output:

yarn test

If you would like to run tests without Eslint checks, set an environment variable:

NO_ESLINT=1 yarn test

There is limited support for using Flow to validate the intention of our program.

As you run tests you will see a report of Flow errors at the end of the test output:

yarn test

If you would like to run tests without Flow checks, set an environment variable:

NO_FLOW=1 yarn test

To only check for Flow issues during development while you edit files, run:

yarn flow:dev

If you are new to working with Flow, here are some tips:

• Check out the getting started guide.
• Read through the web-ext guide for hints on how to solve common Flow errors.

To add flow coverage to a source file, put a /* @flow */ comment at the top. The more source files you can opt into Flow, the better.

Here is our Flow manifesto:

• We use Flow to declare the intention of our code and help others refactor it with confidence. Flow also makes it easier to catch mistakes before spending hours in a debugger trying to find out what happened.
• Avoid magic Flow declarations for any internal code. Just declare a type alias next to the code where it's used and export/import it like any other object.
• Never import a real JS object just to reference its type. Make a type alias and import that instead.
• Never add more type annotations than you need. Flow is really good at inferring types from standard JS code; it will tell you when you need to add explicit annotations.
• When a function like getAllAddons takes object arguments, call its type object GetAllAddonsParams. Example:

type GetAllAddonsParams = {|
  categoryId: number,
|};

function getAllAddons({ categoryId }: GetAllAddonsParams = {}) {
  ...
}

• Use Exact object types via the pipe syntax ({| key: ... |}) when possible. Sometimes the spread operator triggers an error like 'Inexact type is incompatible with exact type' but that's a bug. You can use the Exact<T> workaround from src/core/types/util if you have to. This is meant as a working replacement for $Exact.
• Add a type hint for components wrapped in HOCs (higher order components) so that Flow can validate calls to the component. We need to add a hint because we don't yet have decent type coverage for all the HOCs we rely on. Here is an example:

// Imagine this is something like components/ConfirmButton/index.js
import { compose } from 'redux';
import * as React from 'react';

// This expresses externally used props, i.e. to validate how the app would use <ConfirmButton />
type Props = {|
  prompt?: string | null,
|};

// This expresses internally used props, such as i18n which is injected by translate()
type InternalProps = {|
  ...Props,
  i18n: I18nType,
|};

export class ConfirmButtonBase extends React.Component<InternalProps> {
  render() {
    const prompt = this.props.prompt || this.props.i18n.gettext('Confirm');
    return <button>{prompt}</button>;
  }
}

// This provides a type hint for the final component with its external props.
// The i18n prop is not in external props because it is injected by translate() for internal use only.
const ConfirmButton: React.ComponentType<Props> = compose(translate())(
  ConfirmButtonBase,
);

export default ConfirmButton;

• Try to avoid loose types like Object or any but feel free to use them if you are spending too much time declaring types that depend on other types that depend on other types, and so on.
• You can add a $FLOW_FIXME comment to skip a Flow check if you run into a bug or if you hit something that's making you bang your head on the keyboard. If it's something you think is unfixable then use $FLOW_IGNORE instead. Please explain your rationale in the comment and link to a GitHub issue if possible.
• If you're stumped on why some Flow annotations aren't working, try using the yarn flow type-at-pos ... command to trace which types are being applied to the code. See yarn flow -- --help type-at-pos for details.

We use Prettier to automatically format our JavaScript code and stop all the on-going debates over styles. As a developer, you have to run it (with yarn prettier-dev) before submitting a Pull Request.

To see a report of code coverage, type:

yarn test-coverage-once

This will print a table of files showing the percentage of code coverage. The uncovered lines will be shown in the right column but you can open the full report in a browser:

open coverage/lcov-report/index.html

A proxy server is provided for running the AMO app with the API on the same host as the frontend. This mimics our production setup.

Start developing against a hosted API like this:

yarn amo:dev

This configures the proxy to use https://addons-dev.allizom.org for API data. This command is the most common way to develop new frontend features. See the table of commands up above for similar ways to run the server.

To use a local API server running in Docker, you can use the yarn amo command. However, this is currently not working. See issue-7196.

Authentication will work when initiated from addons-frontend and will persist to addons-server but it will not work when logging in from an addons-server page. See mozilla/addons-server#4684 for more information on fixing this.

If you need to override any settings while running yarn amo, yarn amo:dev, or yarn amo:stage, first create a local config file named exactly like this:

touch config/local-development-amo.js

Make any config changes. For example:

module.exports = {
  trackingEnabled: true,
};

Restart the server to see it take affect.

Consult the config file loading order docs to learn more about how configuration is applied.

When running yarn disco, your local server will be configured for a hosted development API. If you want to run your own addons-server API or make any other local changes, you'll need to create a custom config file named exactly like this:

touch config/local-development-disco.js

Here's what local-development-disco.js would look like when overriding the apiHost parameter so that it points to your docker container:

module.exports = {
  apiHost: 'http://olympia.test',
};

Restart the server to see it take affect.

If you want to access your local server on an Android device you will need to change a few settings. Let's say your local machine is accessible on your network at the IP address 10.0.0.1. You could start your server like this:

API_HOST=http://10.0.0.1:3000 \
    SERVER_HOST=10.0.0.1 \
    WEBPACK_SERVER_HOST=10.0.0.1 \
    yarn amo:dev

On your Android device, you could then access the development site at http://10.0.0.1:3000.

NOTE: At this time, it is not possible to sign in with this configuration because the Firefox Accounts client redirects to localhost:3000. You may be able to try a different approach by editing /etc/hosts on your device so that localhost points to your development machine but this has not been fully tested.

When developing locally with a webpack server, the randomly generated asset URL will fail our Content Security Policy (CSP) and clutter your console with errors. You can turn off all CSP errors by settings CSP to false in any local config file, such as local-development-amo.js. Example:

module.exports = {
  CSP: false,
};

The documentation you are reading right now lives inside the source repository as Github flavored Markdown. When you make changes to these files you can create a pull request to preview them or, better yet, you can use grip to preview the changes locally. After installing grip, run it from the source directory like this:

grip .

Open its localhost URL and you will see the rendered README.md file. As you make edits, it will update automatically.

If you run yarn storybook, you can see storybook locally here: http://localhost:9001/.

You can find the source files under the ./stories directory.

Check it out on Github Pages: https://mozilla.github.io/addons-frontend/

More info coming soon :)

The following are scripts that are used in deployment - you generally won't need unless you're testing something related to deployment or builds.

The env vars are:

• NODE_APP_INSTANCE: the name of the app, e.g., amo or disco
• NODE_ENV: the node environment, e.g. production or development
• NODE_CONFIG_ENV: the name of the configuration to load, e.g., dev, stage, prod

Example: Building and running a production instance of the AMO app:

NODE_APP_INSTANCE=amo NODE_ENV=production NODE_CONFIG_ENV=prod yarn build
NODE_APP_INSTANCE=amo NODE_ENV=production NODE_CONFIG_ENV=prod yarn start

Note: To run the app locally in production mode you'll need to create a config file for local production builds. It must be saved as config/local-prod-amo.js and should look like:

const amoCDN = 'http://127.0.0.1:4000';

module.exports = {
  // CDN URL points to the Node server.
  amoCDN,

  // Configure CSP with 'self' since we serve the compiled files from `dist/`.
  CSP: {
    directives: {
      fontSrc: ["'self'"],
      imgSrc: ["'self'", 'data:', 'https://addons.cdn.mozilla.net'],
      scriptSrc: ["'self'"],
      styleSrc: ["'self'"],
    },
    reportOnly: true,
  },

  // Serve static via Node.
  staticHost: undefined,
  enableNodeStatics: true,

  // No need for Sentry.
  publicSentryDsn: null,
};

After this, re-build and restart using yarn build and yarn start as documented above. If you have used 127.0.0.1 before with a different configuration, be sure to clear your cookies. The application should be available at: http://127.0.0.1:4000/.

NOTE: At this time, it's not possible to sign in using this approach.

When implementing user interfaces you will need to refer to the Sketch-based mocks that are hosted on InVision: https://mozilla.invisionapp.com/share/YCCYEOFRH#/screens.

If you need to explore the Sketch source file, you will need a license to run Sketch and you also need to install some fonts (which are free). Install Fira Sans, Open Sans, Fira Mono and Chivo.

On MacOS, you can use the Homebrew tap Caskroom-fonts:

brew tap homebrew/cask-fonts
brew cask install font-fira-sans font-open-sans font-chivo font-fira-mono

You can check to see what commit of addons-frontend is deployed as well as which A/B experiments are running and which feature flags are enabled by making a request like this:

curl https://addons-dev.allizom.org/__frontend_version__
{
    "build": "https://circleci.com/gh/mozilla/addons-frontend/10333",
    "commit": "47edfa6f24e333897b25516c587f504e294e8fa9",
    "experiments": {
        "homeHero": true
    },
    "feature_flags": {
        "enableFeatureAMInstallButton": true,
        "enableFeatureExperienceSurvey": false,
        "enableFeatureStaticThemes": true
    },
    "source": "https://github.com/mozilla/addons-frontend",
    "version": ""
}

This will return a 415 response if a version.json file doesn't exist in the root directory. This file is typically generated by the deploy process.

For consistency with monitoring scripts, the same data can be retrieved at this URL:

curl https://addons-dev.allizom.org/__version__

💡 You can install the amo-info extension to easily view this information.

This project will hold distinct front-ends e.g:

• Discovery Pane
• AMO or addons.mozilla.org

We've made a conscious decision to avoid "premature modularization" and keep this all in one repository. This will help us build out the necessary tooling to support a universal front-end infrastructure without having to worry about cutting packages and bumping versions the entire time.

At a later date if we need to move things out into their own project we still can.

• Based on Redux + React
• Code written in ES2015+
• Universal rendering via node
• Unit tests with high coverage (aiming for 100%)