The Joy of Mixing Custom Elements, Web Components, and Markdown | Hacker News Reader

The Joy of Mixing Custom Elements, Web Components, and Markdown

deanebarker.net

116 pointsdeanebarker41 comments

I recently discovered a static site generator called Astro, which supports many syntaxes but the .astro is a nice mix of TypeScript and JSX-like syntax. Content can use MDX which is like Markdown but with {JSX} style markup for variables and etc. The static components are used very similar to React, with familiar import statements and <ComponentName props=etc> patterns. It is extremely easy to pick up. Best of all, it has plugins to support all sorts of other interactivity, so you can create interactive 'islands' of content using React or Preact or SolidJS or Vue etc. That way you have most of your content statically generated, and then the dynamic parts can be done from the client side.

Best of all, if you use simple unchanged files for other dynamic stuff like JSON etc, you can just generate those on build and serve those files in the host directly as the 'response' to a simple REST request, which is sometimes overlooked despite being the most fundamental form of a REST API.

https://astro.build/

I came across this after researching various options for a website that had, mostly for my own entertainment, restrictions on wanting to be mostly statically generated but customizable easily without learning a lot of new syntax / etc, something JSX-like with Markdown support etc, and MDX was an immediate find - and astro was the easiest SSG I found for it after trying with 11ty and several others. Actually felt like a delight playing with it.

> if you use simple unchanged files for other dynamic stuff like JSON etc, you can just generate those on build and serve those files in the host directly

Other web frameworks support this too, if you look for "static export" options. Next.js, for example, supports this via the getInitialProps function.

What I like especially about Astro, that you perform this data loading during build time from any component/file on your page. With Next.js, this is only possible via the top level Page component.

That is legacy, the current way with page router model is to use getStaticProps or getServerSideProps.

And if using app router model, that is part of React server components.

I love astro. I use it for one of game curation projects. https://exhibitplay.com/

Built a few reusable templates and JSON for storing game data (maybe not ideal at scale, but seems to be working for now). JavaScript, CSS, and MDX. Hosted on Netlify which is an 'Official Deployment Partner'. It's light and simple. So far has been a joy to work with.

>In the end, your document is now fully an HTML document, not a Markdown document that becomes an HTML document. It’s a minor perspective shift, but might have some cascading effects on things I’ve written above.

But this style of custom-elements requires successful javascript program execution to achieve that "HTML" document. Just like markdown requires some parser program to turn it in to HTML. It's not really fully an HTML document.

It's a good idea. It just would be a better one to write the custom-elements as wrappers for actual HTML elements. Like how https://blog.jim-nielsen.com/2023/html-web-components-an-exa... shows instead of trying to do it SPA style and requiring perfect JS execution for anything to show properly.

HTML mark-up really isn't that heavy. The avoidance of it seems mostly to be because it's considered "old" and "old" is bad, or at least not useful on a resume. But it's old because it's so good it's stuck around for a long time. Only machine generated HTML is bulky. Hand written can be just as neat and readable as any Markdown.

> It just would be a better one to write the custom-elements as wrappers for actual HTML elements.

pandoc has an extension for this:

https://pandoc.org/demo/example33/8.18-divs-and-spans.html

KeenWrite, my (R) Markdown editor, supports pandoc annotations:

https://youtu.be/7icc4oZB2I4?list=PLB-WIt1cZYLm1MMx2FBG9KWzP...

> Just like markdown requires some parser program to turn it in to HTML.

Or XHTML, which is XML, which can then be transformed into TeX macros, and then typeset into a PDF file with a theme (much like CSS stylizes HTML).

https://youtu.be/3QpX70O5S30?list=PLB-WIt1cZYLm1MMx2FBG9KWzP...

This allows separating content from presentation, allowing them to vary independently.

JavaScript is one of the three core file types of the web. You can rely on it as much as HTML and CSS. I don't get the unique derision of JS compared to the other files types.

That's an understandable take in nearly all commercial and institutional contexts. But in others just involving human people, no. Many times JS does fail or isn't available. So building progressively enhanced web documents preserves utility across the spectrum of human visitors (and maintains accessibility). But if you only have a profit motive, then yes, there's no need for robust solutions. The amount of people that can't do JS well won't eat into profits or cause enough complaints to get you in trouble.

I think progressive enhancement is a cool approach to building stuff.

I also think “turn JS on” is a fairly reasonable ask these days. A lot of the web tends to break when CSS is disabled or fails, too.

If your HTML or CSS fails to load, you're going to have a hard time too. Web pages have many critical resources.

Failing to load is not a problem. Failing to execute is.

Usually because the web dev have used some new Javascript feature only $latest JS engines support. HTML and CSS if they're there they're there. Sure, there's caniuse for HTML and CSS. But they only have to load. The text/images/etc will be there. JS both has to be loaded and executed. If the later doesn't happen just right then the text and other multi-media won't be there. It's a very big difference.

Generally i agree.

I think there are a group of people who are salty that js became “the lang” for the web. Another group of people loath the framework insanity of webdev. I count myself among the ladder not the former. I equally hate all languages.

Js is heavily overused but the “web” of today is not the web of the 90s or 2010s which some people cannot get over.

No, you can't really rely on it. Welcome to Performance Inequality Gap: https://infrequently.org/2024/01/performance-inequality-gap-...

One additional thing that article fails to mention: you should not test your device in a context where it can cool itself easily. Test on your devices when they are wrapped in a blanket, and while there's another program using 100% CPU.

Your conclusion is not the same as the article you link. Js is fine but it should be used relative to the targeted use case.

That astonishingly long and researched read loses impact when it draws a primarily moral based argument in the thesis. Being fast is better for both the privileged and underprivileged.

Moral handwringing rarely moves people to action.

> Your conclusion is not the same as the article you link.

My conclusion directly derives from the article. If your app relies on Javascript, it will be non-functional/broken/unusable for a huge number of people while their devices struggle to download, unzip, parse and run your JS bundles.

BTW. It's worse with web components built with default assumptions (without bundling). Since `import` statements will cause a long waterfall as each component loads its dependencies.

Custom elements are really great for editors and developers. You can provide a rich set of primitives that editors can use to display certain content. In the past, I used MDX [1] extensively so non-technical writers can create a rich UI for a documentation site.

- [1] https://mdxjs.com/

I just wanted to throw some extra :hearts: towards https://mdxjs.com/ it's really the best of both worlds.

May I ask what you are using instead nowadays?

MDX is great and I like to use it.

At Stripe, we built Markdoc to solve many of the issues mentioned!

https://stripe.com/blog/markdoc

This feels like Markdown + Liquid.

I wrote quite a bit about Fluid, which is the C# implementation of Liquid. https://deanebarker.net/tech/fluid/

Nice. But reminds us that most templating languages support markdown by default.

> I didn’t find the need for blank lines in the CommonMark spec,

It's in #6 and #7

https://spec.commonmark.org/0.31.2/#html-blocks

I too ran a site on markdown with HTML. Originally I had my own hacky way of using HTML which was effectively to (1) replace all HTML with placeholder_number (2) format the markdown (3) replace placeholder_number with the html that used to be there. Not as simple as that but close.

Eventually i switched to a commonmark spec formatter and then tried to fix any old pages that had spaces in the HTML. Some were basically hard to fix like a pre section with code inside so I added some other hacks to do the same as above for those sections by surrounding them with {{#html}}....stuff..with...blank...lines{{/html}}.

So now my solution is handlebars.js meets markdown-it.

It certainly is a joy! I'd generalise it to any reasonable plaintext format.

Org mode keeps me locked in for this very reason.

No need for Emacs... pandoc (which I use) builds my static site. It supports enough Org markup to let me compile header metadata, latex, footnotes, markup etc. as well as inline live JavaScript code and use it in-page, just like I'd do in a regular stand-alone HTML page.

For example: a post with inlined static HTML and JavaScript, freely mixed-and-matched with org-native source block syntax. No plugins needed.

Raw: https://raw.githubusercontent.com/adityaathalye/shite/refs/h...

Pandoc-rendered live site: Compare Raw source with this section: https://www.evalapply.org/posts/animate-text-art-javascript/...

Github-rendered in repo (not github pages): It is (surprisingly) good enough for most Org content, including generating a nice Table of Contents... Except for any footnotes, and inlined or hosted JavaScript source (understandably, security may be a concern). https://github.com/adityaathalye/shite/blob/master/sources/p...

"...Pre-processing and the initial excitement of a domain-specific language is a siren that might lead you to your doom.

(I swear, I have never tried to do something this foolish… I swear…)..."

I will say, that almost didn't make me laugh, bur, I will not swear.

I've recently come off from one such rabbit hole - A CLI tool for Pre-processing vue project templates. It was over-engineered.

Will probably go back to re-build it when I get the free time.

└── Dey well; Be well

I use https://github.com/nuxt-modules/mdc with a Nuxt Content project. Also looked at https://markdoc.dev

Unfortunately, also not-joy of completely unbreakable website ... https://usercontent.irccloud-cdn.com/file/TPJZ2AeN/100001246...

Did you do something to font color?

only opened the page on mobile network, and something choked a CSS somewhere apparently. Then it stayed this way.

Always curious why no one has made Markdown + pugjs-style[0] html parsing system.

[0] https://pugjs.org/api/getting-started.html

We did this a decade or two ago when Markdown was Markdown and pug was Jade. It was great.

Apart from that, also back in the day people were still building pages using e.g. Dreamweaver, so we also put together a thing that could leverage a designer's HTML that rendered as a complete placeholder page. That "works as plain .HTML" became the template, used YAML as data, and Enlive (a Clojure lib) to turn the HTML with its placeholders into a template and then stuff YAML sidecar data into it to generate new HTML.

Not long after, Caddy's markdown support had promise, but agree with another commenter here that a few years ago astro.build started hitting a more interesting spot. In the middle ages between those, yay pandoc.

Pug (previously Jade) was a precursor to web frameworks. Technologies we use and enjoy today were built by previous decades of stumbling in the dark with one feature and hammering it into formidable shape. A nod to Handlebars while we're on the topic.

Recently found myself saying in conversations that, "Markdown is arguably one of the most valuable file type we have; media, documents, code blocks of almost any language, ... all in one place for further processing."

└── Dey well; Be well

My tool of choice for what TFA describes is Vitepress. Markdown plus about a dozen useful plugins and understands Vue3 components and treats each *.md as an SFC.

It hits a sweet spot for me.

I'm currently using Markdig in Razor Components (Blazor). I wonder if I could get Razor Components rendered in the markdown.

I would say it's possible. I just write my own Markdig plugins for all the custom elements I need, though.

Ha! Small world! I just started building a documentation editor using markdown, built as a custom element[0]. It's still deeply in alpha, but there is a working demo, at least!

[0] https://github.com/catapart/magnitce-docs-editor

Nice to see pro-Markdown literature that extends its application, but not to the extent that (that I’ve found) its detractors use against it.

There are a lot of nooks in crannies in this article.

What's the advantage of using Web Components here? Why not process the tag on the server side aswell and replace it with Html?

Because if you're relying on server-side software to process your finished product, then your finished product is vulnerable to dependency rot and overall general software rot.

But if you process your markup at build time into it's finished HTML, then you can host it anywhere - anytime and it'll work.

I think you misunderstood the comment you’re replying to. Its proposition is: you’re already doing a build process for Markdown to HTML, and even perhaps adding some preprocessing on that, so if you’re only using Custom Elements as a simple template engine, why not apply that in your build process too, and emit more reliable plain HTML?

I've recently hit a very similar problem. There's various problems with the web components (like he mentions) and I wanted a better way to write plain html with components. I ended up making a tiny python build script to process fake html components into proper HTML: https://github.com/shminge/builder

I've designed it for myself, deliberately to work around the flaws with web components mentioned. It hugely speeds up my process of writing pages for my blog (after trying everything, I've ended up just writing plain html) because I can define a component that holds all the boilerplate of the page (header, css, etc) and then write my pages as `<pagelayout> <p> body content here </p> </pagelayout>` and have that be expanded into proper valid HTML

Reply on news.ycombinator.com