Add docs on how to handle pretty URLS.

[gauntface]

Taken from #640

So I went offline, and things weren't working as expected. I eventually realized that the links throughout my page all use "pretty URLs", whereas WB cached the full file names, including the .html extension. Seems like a pretty common scenario, maybe WB should have a flag that handles this?

At this point, I realized that I probably need to my own service worker with a fetch listener that handles the pretty URLs. It was surprisingly hard to find out how to do this on the docs site. This seems like a core user-journey, I think it should have an entry on the "How To" page. I think I need to use the injectManifest() method, but my first attempt hasn't worked yet.

[gauntface]

Thinking on this a little more, this is messy for precaching.

Let's say you have a precache of the following:

• /index.html
• /about.html
• /something/nested.html

At the moment (Out of the box) we will do the following:

• Match for the URLs out right (i.e. '/index.html', '/about.html', '/something/nested.html).
• If that fails, we match minus search params
• If that fails and the URL DOES end in '/', we add index.html and see if that matches.

This means we'll handle '/'. However, we won't handle the pretty URL case of '/about' or '/something/nested'.

@jeffposnick would you be happy to add this into precaching's cache matching? This isn't an issue on template matching or runtime routes.

[jeffposnick]

I'm okay with that additional "magic" being added in, and perhaps on by default, with the option to disable it via prettyUrls: false.

FWIW, the other option is to create a recipe that they used in their build process along the lines of:

buildConfig.templatedUrls = glob.sync('**/*.html').map((htmlFile) => {
  const url = path.basename(htmlFile, '.html');
  return {[url]: htmlFile};
});

That's pretty gnarly, and wouldn't work well for developers who are just using the CLI or webpack plugin. So we can support this by default instead.

[raviqqe]

Is there any workaround so far? I really want this feature so that my websites work offline as it is without rewriting any link URLs.

[jeffposnick]

A workaround is to use a config like the one in #903 (comment), where you explicitly set templatedUrls to a mapping of "pretty" URL => the underlying .html file.

Have you had any luck with that?

[raviqqe]

I forgot to mention that I use the CLI version.

[jeffposnick]

You could actually do this with the CLI version, too 😄

(I totally understand that this is ugly and not what we'd ideally want developers to have to do, but I want to help unblock you given what's currently possible.)

// Inside workbox.config.js:

// Run `npm install -g glob` to make sure it's globally available:
const glob = require('glob');

const templatedUrls = glob.sync('**/*.html').map((htmlFile) => {
  const url = path.basename(htmlFile, '.html');
  return {[url]: htmlFile};
});

module.exports = {
  templatedUrls,
  // Your existing config:
  globDirectory: '...',
  globPatterns: ['...']
};

[raviqqe]

Thank you @jeffposnick! I ended up this script. Note that I use Jekyll here and compiled HTML files of my website is located in _site directory.

const glob = require('glob')
const path = require('path')

module.exports = {
  templatedUrls: Object.assign(...glob.sync('_site/**/*.html').map((filename) => {
    if (path.basename(filename) === 'index.html') {
      return {}
    }

    return {[filename.slice(6, -5)]: filename}
  })),
  globDirectory: '_site/',
  globPatterns: [
    '**/*.{css,html,js,json,png}'
  ],
  globIgnores: [
    'register-service-worker.js',
    'service-worker.js'
  ],
  swDest: 'service-worker.js'
}

[gauntface]

This is actually documented under clean URLs here: https://developers.google.com/web/tools/workbox/modules/workbox-precaching

[raviqqe]

For all users who reached to this issue, now it's no need for configuration of clean URLs. With the latest version of workbox-cli, the script I posted above is almost equivalent to:

module.exports = {
  globDirectory: '_site/',
  globPatterns: [
    '**/*.{css,html,js,json,png}'
  ],
  globIgnores: [
    'register-service-worker.js',
    'service-worker.js'
  ],
  swDest: 'service-worker.js'
}

[Stanzilla]

So that works for precaching but how would I use a normal route to cache them? I currently precache (using the html files) and the cache never refreshes.