We Need DOM APIs in Workers
We Need DOM APIs in Workers
This web dev discusses why he loves DOM APIs that allow him to efficiently parse and manipulate XML data and to output HTML.
Join the DZone community and get the full member experience.Join For Free
How to Transform Your Business in the Digital Age: Learn how organizations are re-architecting their integration strategy with data-driven app integration for true digital transformation.
I need DOM APIs in Workers for different reasons than most people. Many people would like the DOM in Workers to make updating the DOM not block the main thread. I need it so I can efficiently parse and manipulate XML data and to output HTML, and I suspect many other people do.
In a recent project, I wanted to share as much logic as possible between the server, service worker, and the client. The project is essentially a simple RSS feed reader, it takes RSS feeds, parses the data, and merges them into a nice set of columns (much like TweetDeck), and also a single merged list.
The project works with the RSS feed data in three places:
- On client - When the page loads for the very first time, it AJAX requests the RSS feed data from a proxy service that I run, and it then caches the raw data in the
window.cachesobject for later use before rendering it in the client.
- In the service worker -
- When the main page loads and the service worker is installed, the service worker loads the shell and merges in the RSS feed data so that no AJAX requests need to be made on the 2nd load - thus keeping the Time to interactive time high.
- When a request to the proxy is made from the client, the service worker, when installed, will intercept the request and serve the data from the
window.caches. This allows the site to work offline.
- On the server - When the page is requested, we can take some of the data that is cached on the server and merge it directly into the response that we send to the client. By rendering some of the content directly from the server we can have a stable viewport on the first load which is normally important for slower connections on mobile (and SpeedIndex).
In each instance, there is a simple process that takes the RSS data and maps it into a JSON object that I can then apply to a template to generate HTML. Keeping one template and unified logic across the client, server, and service worker was a critical requirement. Maintaining one set of templates means that the input data has to be consistent across all places that will render data.
Because I run a proxy server, there is a simple solution: just transform all the RSS feeds into a consistent JSON form on the server. I discounted this because:
- Data transforms can be intensive to process.
- Data transforms can be done on the client to reduce the shared burden on the service.
- Most importantly, if an RSS feed is on HTTPS and supports CORS there is no need to go through the proxy service. This is the state that I want to be in in the future because it allows the feed reader to render content that might require the authentication of the user.
Processing the data on the client is possible (and desired in my case) because browsers have a little-used API called
DOMParser. DOMParser is as the name suggests: a parser of raw XML and HTML that builds a DOM. Once you have a DOM, you can do anything with it that you would do with normal DOMs (appendChild, getAttribute etc., etc.).
let parser = new DOMParser(); let dom = parser.parseFromString('<a><b>hello</b></a>', 'application/xml'); let bString = dom.querySelector('b').textContent;
Pretty simple stuff and I use this to convert the RSS data into a simple JSON structure so that I can pass it to a templating function (It is here if you are interested in seeing the code.)
This works perfectly in the client, but there is no DOM in web workers, service workers, nor any native DOM in the server.
Luckily there is an npm library that works everywhere. XML-DOM is a Level 2 compliant implementation of the W3C DOM with some Level 3 features, and it works pretty much as expected. It's not the end of the world, but it seems silly to have to import 64kb of JS, for something that the browser already has built in.
I only ever see the 'VDOM' use-case for DOM APIs in workers, and whilst I think it is an important use-case, I think it gets in the way of another important uses case: data manipulation off the main thread. The fact that we can't use workers to process HTML and XML documents (something that nearly every app has to do) without having to import a huge chunk of polyfill that won't run at the same speed as a native implementation and that we rely on OSS contributors to maintain seems like something that should be fixed.
Thank you to the people who maintain
xml-dom. Heroes work.
Published at DZone with permission of Paul Kinlan , DZone MVB. See the original article here.
Opinions expressed by DZone contributors are their own.