Yari

Quickstart

Before you can begin with Yari, you need Content. See its README which basically, says something like this:

git clone https://github.com/mdn/content.git mdn/content

Now, you just need to note where that folder is before you can start Yari.

To run Yari locally, you'll first need to install git, Node.js (>= 12.0.0) and Yarn 1. After that, run these commands in your bash:

git clone https://github.com/mdn/yari.git
cd yari
yarn
echo CONTENT_ROOT=/path/to/mdn/content/files >> .env
yarn dev
open http://localhost:3000

Make sure you point to the /files folder inside your clone of the content repo.

If you prefer, you can fork the repo first and do the git clone with your fork instead of the mdn one.

The yarn dev command will compile and prepare certain files. This always takes a little extra time. If you prefer you can use yarn start which will re-use any previously compiled files which is "riskier" but faster. The yarn start command will also start a server which doesn't automatically reload when its source code files change, so use with caution.

See also our reviewing guide for information on how to review Yari changes.

How to stay up-to-date

Periodically, the code and the content changes. Make sure you're staying up-to-date with these commands:

git pull origin master
yarn
yarn dev

These are also good steps to always take when you embark on making a change. Then, the only extra command needed is git checkout -b my-new-branch (or however you prefer to create new git branches)

License

All source code is MPL-2.0.

For content, see its license in the mdn/content repository.

How it works

Yari is multiple things but at its core is the MDN content as index.html files, in git, that contain the metadata (as front-matter) and the bulk of the document.

The builder converts these "source files" into "build files" using a CLI tool that iterates over the files, builds the HTML, and lastly packages it up with the front-end code, ready to be served as static files.

Development

First of all, development on yari can mean the source code (e.g. the styling of the header) or it can mean the content, since it's all one repo. This document doesn't distinguish between the two. In the future we might expand with more documentation specifically for contributing to the content exclusively.

The yarn start command encapsulates the front-end dev server (on http://localhost:3000) and the server (on http://localhost:5000).

All the sub-commands of yarn start can be broken down and run individually if you want to work more rapidly.

Setting up $EDITOR

If you configure an environment variable called EDITOR, either on your system as a whole or in the root .env file, it can be used in the development server to link to sources which, when clicked, opens in your preferred editor/IDE. For example, in the root:

echo 'EDITOR=code' >> .env

Now clicking certain links will open files directly in the currently open VSCode IDE. To test it, view any document on http://localhost:3000 and click the "Open in your editor" button.

How the server works

The server has two main jobs:

1. Simulate serving the site (e.g. from a server, S3 or a CDN).
2. Trigger builds of documents that haven't been built, by URL.

Linting

All JavaScript and TypeScript code needs to be formatted with prettier and it's easy to test this with:

yarn prettier-check

And conveniently, if you're not even interested in what the flaws were, run:

yarn prettier-format

But automatically when you ran yarn the first time (yarn is an alias for yarn install) it set up a git pre-commit hook that uses pretty-quick which is a wrapper on prettier that checks only the files in the git commit.

If in doubt about formatting, you can create a pull request and if you have formatting flaws, the pull request checks should catch it.

Upgrading Packages

We maintain the dependencies using Dependabot in GitHub but if you want to manually upgrade some you can use:

yarn upgrade-interactive --latest

Sharing your dev environment with ngrok

ngrok is a great tool for starting a HTTP proxy server from the Internet into your Yari server. This can be useful for testing your current build on external tools like BrowserStack, WebPageTest, Google Translate, or to simply show a friend what you're up to. Obviously it'll never be faster than your uplink Internet connection but it should be fairly feature complete.

1. Create in account on Ngrok.com
2. Download the executable
3. Start your Yari server with yarn start in one terminal
4. Start the ngrok executable with: /path/to/your/ngrok http 5000

This will display something like this:

Session Status                online
Account                        (Plan: Free)
Version                       2.3.35
Region                        United States (us)
Web Interface                 http://127.0.0.1:4040
Forwarding                    http://920ba2108da8.ngrok.io -> http://localhost:5000
Forwarding                    https://920ba2108da8.ngrok.io -> http://localhost:5000

Connections                   ttl     opn     rt1     rt5     p50     p90
                            0       0       0.00    0.00    0.00    0.00

Now, take that "Forwarding" URL https://920ba2108da8.ngrok.io (in this example) and share it.

Building

The server builds content automatically (on-the-fly) when you're viewing pages. But if you want to you can pre-emptively build all the content in advance. One potential advantage is that you can get a more complete list of all possible "flaws" across all documents before you even visit them. The most fundamental CLI command is:

yarn build

What gets built

Every index.html becomes two files:

• index.html fully formed and complete HTML file
• index.json the React needed state to build the page in the client

Flaw checks

When building you can enable specific "flaw checks" and their level of handling. Some flaws are "cosmetic" and some are more severe but they should never block a full build.

More information about how to set flaws can be found in docs/envvars.md.

Essentially, the default is to warn about any flaw and you can see those flaws when using http://localhost:3000. But for completed builds, all flaws are ignored. This makes the build faster and there's also no good place to display the flaws in a production-grade build.

In the future, we might make the default flaw level error instead. That means that any new edits to (or creation of) any document will break in continuous integration if there's a single flaw and the onus will be on you to fix it.

Icons and logos

The various formats and sizes of the favicon is generated from the file mdn-web-docs.svg in the repository root. This file is then converted to favicons using realfavicongenerator.net. To generate new favicons, edit or replace the mdn-web-docs.svg file and then re-upload that to realfavicongenerator.net.

Troubleshooting

Some common issues and how to resolve them.

Error: ENOSPC: System limit for number of file watchers reached

There are two options to resolve this.

1. Disable the watcher via REACT_APP_NO_WATCHER

   echo REACT_APP_NO_WATCHER=true >> .env

2. Increase max_user_watches:
   See https://github.com/guard/listen/wiki/Increasing-the-amount-of-inotify-watchers

Error: Cannot find module 'levenary'

We can't know for sure what's causing this error but speculate a bug in how yarn fails to resolve if certain @babel helper libs should install its own sub-dependencies. A sure way to solve it is to run:

rm -fr node_modules
yarn install