docsearch/docs/README.md

# Documentation website

This subdirectory holds the documentation website content as well as scripts to
generate it.

## Main commands

### Building the website

You can build a new version of the website by running `yarn run build` in this
directory or `yarn run docs:build` at the repository root.

It will read all source files in `./src` and build the final static website in
`./dist`.

### Local development

You can run a local copy of the documentation website by running
`yarn run serve`. This is an alias for running `yarn run docs:serve` at the
repository root.

This will build the website in `./dist` and expose it on `localhost`, along with
live-reload.

### Deploying the website manually

You can deploy the website manually by running `yarn run deploy` in this
directory or `yarn run docs:deploy` at the repository root.

This will build the website and then commit the content of the `./dist` folder
to the `gh-pages` branch and push it to GitHub.

### Auto-deploying

Any new commit on `master` that modifies the `./docs` folder will automatically
trigger a build and deploy it.

It works by having Netlify listen to any new commit on `master` and running
`./scripts/netlify-master` in response (see `netlify.toml` for details). Note
that the website is **not** hosted on Netlify, but on GitHub Pages (more on that
later).

The script compares the date of the last commit in `./docs` with the date of the
last deploy. If no new commits were added, it will stop. Otherwise, it will
continue and build the website.

To make this comparison, both this script and the manual deploy script add a
`last_update` file to the `./dist` folder containing the timestamp of the last
deploy. This file is then used to see if changes are present and should be
deployed.

Once the build is complete, the `./dist` folder is committed to the `gh-pages`
branch and pushed to GitHub. This part requires some non-trivial `git` and `ssh`
configuration commands to push data from Netlify to GitHub pages on our behalf
(check `./scripts/netlify-master` for more details).

### Deploy previews

Any new Pull Request to the documentation will trigger a deploy preview build.

Netlify is configured to run `./scripts/netlify-deploy-preview` on each new PR
(check `netlify.toml` for details).

This script will first check if changes were made to the `./docs` subfolder in
the PR. If no change were made, the preview will not be generated (this will
make processing time faster).

Whenever the preview is ready, a message from Algobot will be added to the PR,
along with the link to the preview. This is configured in Netlify UI in _Build
and Deploy > Deploy notifications > Comment on GitHub pull request when deploy
succeeds_. It uses a GitHub token from Algobot to post on its behalf. To
generate such a token, login to Netlify with Algobot and pretend to create such
a notification on any project, generate a token, and then copy-paste it in the
real DocSearch account in Netlify.

## Internals

The documentation generation is not using any existing static websites
generators, but custom JavaScript scripts.

The main two entry points are `./scripts/build.js` and `./scripts/serve.js`. The
second one adds a webserver with live-reload on top of the first one.

### HTML

All Markdown files situated in `./src` will be transformed into `.html` files in
`./dist`. They will be wrapped into the layout defined in their front-matter.

All the headers will be converted to their respective `<hX>` tag, along with a
unique `#id` to allow for easy anchoring.

You can use plain HTML inside those Markdown files if you need more advanced
styling. The custom `{my-class}` syntax is also possible if to add CSS classes
to elements.

```markdown
This is my paragraph. {p-2}

![Pretty image](./img.jpg) {mt-2}
```

```html
<p class="p-2">This is my paragraph</p>

<p class="mt-2">
  <img src="./img.jpg" alt="Pretty image" />
</p>
```

### Layouts

All layouts are saved in the `./src/_layouts` folder.

All config options defined into `config.json` are passed to the layouts and can
be used there.

You can also use mixins or include other files from the layouts.

_Note that the current layout logic is simple and might not handle complex
recursive cases, but should be enough for simple cases._

### CSS

CSS is processed through PostCSS. It expects an entry file in `./src/style.css`.

We are using `postcss-import`, allowing you to `@import` files from the
`./src/_styles/` directory to better split your CSS code in logical chunks.

Most of the styling based on tailwind.css, with the config file behing
`tailwind.config.js`. It contains default sizing and coloring to follow the
Algolia brand guidelines.

The final CSS files is then compressed through PurgeCSS (to keep CSS classes
that are actually used) and CleanCSS (to minify it).

### JavaScript

JavaScript code is processed through Babel. It will compile all files situated
in `./src/js`.

_Note that it compiles JS, and does not bundle it. We might add Webpack/Parcel
support later._

### Assets

Any file with the following extensions found in the `./src` folder will be
automatically copied to the `./dist` folder with the same folder structure:
`gif`, `jpg`, `png`, `ico`, `html`, `svg` and `woff`.

### Placeholders

Values defined in the `placeholders` key of the `config.json` file can be used
in JavaScript and Markdown files by using the `{{key}}` syntax.

For example if you have:

```json
// config.json
{
  "placeholders": {
    "projectVersion": "1.4.2"
  }
}
```

Every occurrence of `{{projectVersion}}` in any `.md` or `.js` file will be
replaced with `1.4.2`.

### Sidebar

The left sidebar of the documentation is generated based on the `sidebar` key of
the `config.json`. The key should contain an array where each key is a part of
the sidebar, with a `title` and a list of `pages`. Each of those pages in turn
is an object with a `title` and `url` value.

The layout will then automatically create all the links and color the active
page. Subsections inside the current page will also be added for every `h2`
element extracted from the markup of the current page.

### Redirects

The `config.json` file can hold a `redirects` array, with object containing both
a `from` and a `to` key. For each `from`, it will create a plain HTML page at
this location, that will redirect anyone visiting it to the page defined in
`to`.

_Note that both those links must be defined as relative to the `site.url`
value._