Marko – a Declarative, Html‑based Language

Don't blame Marko for this type of abomination. This is basically fancy react JSX.

ITS just bizzare people want to parse JavaScript at the same instance they're parsing html.

Also, LLMs are going to destroy any new framework. Someone's gonna need to figure out how to integrate these things into new tools. LLMs suck but it'll be much worse if they freeze innovations cause they're too expensive to chase the new hotness.

How would you prefer to write those examples?

  <script src="https://unpkg.com/alpinejs@3.x.x/dist/cdn.min.js" defer></script>
  <div x-data="{ 
    today: new Date().toDateString(),
    randomNum: Math.floor(Math.random() \* 100)
  }">
    <p>Today is <span x-text="today"></span></p>
    <p>Random number: <span x-text="randomNum"></span></p>
  </div>

Is there any significant difference between that and

    element.innerHTML=`
      <p>Today is ${new Date().toDateString()}</p>
      <p>Random number: ${Math.floor(Math.random() * 100)}</p>
    `

(ideally .setHTML() when it's available)

At that point I think I'd have a skeleton html file that fetches a JS that does it all. I'd take JS with embedded HTML over HTML with embedded JS.

this comment really proves that people do not consider the information presented to them

marko is not comparable to php

it is much closer to svelte

i used to sympathize with people complaining about js-fatigue, but at some point its a skill issue

  p “Today is {Date.today}”,
  p “Random number: {(^100).pick}”

Thought I would share this example as written in Raku using https://harcstack.org

> I already disliked that immensely in PHP. Not going back to that spaghetti mesh-up.

That looks like a pretty normal template to me and nothing like plain PHP templates? What do you mean by "spaghetti mesh-up"?

Oh wow! So I guess Marko has been around for while. Despite that, this is my first time hearing about it.

I think I managed to combine three languages in one with Mint (https://mint-lang.com/):

1. There is HTML (tags) with, but without interpolation {...} you can put string literals, variables and everything that type checks as HTML children.

2. There is CSS but only in style blocks where you can interpolate any expression you need and put in if and case expressions too.

3. There is the normal Mint code you write the logic in (this would be the JavaScript in other languages).

Here is an example which have all three: https://mint-lang.com/examples/7guis/flight-booker

The challenge was to make it seamless enough that so it doesn't look like that we tried to mash languages up, but to make them form a different language that is consistent and simple at the same time.

Isn't syntax pretty much just compression? We could write down the AST itself in some generic notation but that would be orders of magnitudes larger, so invent clever tricks to compress it, which we call syntax.

Edit: typo

almost all code looks like nonsense when you're unfamiliar with it

There’s a Marko Pollo joke, too, but I’m too chicken to say it.

Someone posted here, thanks for that!! It immediately reminded me of HAML: https://harcstack.org/

I like HAML a lot, it was the most pleasant to develop with. And it shares a lot in common with Stylus. They both shared things in common.

NO NEED FOR: Curly braces, parentheses and semicolons. The cool thing, it was all optional and I wasn't forced to make use of all shortcuts!

I developed my own CSS Framework in 2003, shared with some UX guy at Yahoo, who incorporated it into YUI mostly as is, after I waived all rights. Most of that became internet standard. Later I had my own PHP based CSS scaffolding framework in 2005 that could also generate grids (before flex-box). SCASS/LESS was really similar to my framework, when it came out.

But I disliked it, it just looked like PHP mixed with CSS. I thought why accept the ugly syntax, despite a compiler being available?

The best ever existed is: HAML + Stylus + (HARC?)

See the beauty of it: https://stylus-lang.com/docs/selectors.html

Now compare with HAML https://haml.info/

I think https://harcstack.org/ makes a good successor.

vue allows other languages in .vue files for both css & html similar to js

Svelte already has a Pug preprocessor :)

I used to use it years ago. So much nicer than HTML.

Yes. Mostly because:

• JSX is well understood by a lot of developers • support is already built in to text editors • it is understood by typescript

Ryan was working on Solid before he joined eBay/Marko. Both projects have benefited from the shared knowledge and approaching a similar solution space from different angles.

He eventually got the opportunity to work on Solid in a more full-time capacity and decided to take it, but still talks with the Marko team from time to time

JS inside HTML! Groundbreaking! It's nothing like Netscape ever conceived in 1995...

Honestly I don't know... I'm somewhat skeptical about these "next big thing that will fix all your pains in web development". There is so much fragmentation in JS libraries / frameworks. Angular, React, Vue, Angular, Asto, SolidJS, NextJS, Nuxt, Qwik... The list is so overwhelming. Almost each one claims that it fixes a problem in other framework, a year later the other framework fixes an issue... I think it's better to stick to a big old player, such as Svelte.

Marko has been around for over a decade at this point and powers most of eBay. It's not the oldest or the largest, but it's got a pretty solid track record

Why are you copy pasting the same comment, under different usernames everywhere?

FWIW, marko comes from Ebay. So, it scales - and being primarily SSR, its a better UX than your preferred frameworks

React has a clear separation of concern? Excuse me? I only see do-it-all pre-styled components in the real world.

I remain convinced that RSC and the SSR craze was a result of someone (or multiple) people needing a raise and their friends wanting to start a company selling abstract compute. Statically hydrated, minimal React was pretty great when served over good CDN infrastructure. Then I watched the bundle sizes and lock-in balloon. That second article is a dragon slayer. It really lays out the problem with React. In marrying itself to Next.js and embracing the server, it's betrayed the platform. Meanwhile, the platform itself has matured. React practically built my career, and I just don't have a reason to choose it anymore.

That’s interesting. I’ve always held SvelteKit in high regard for greenfield projects because it balances capability, developer experience, and performance, but I’ll have to give Marko a look. I’d love to see a similar deep dive into Electron style desktop frameworks since that space still feels underexplored compared to mobile. I honestly wouldn’t know where to start for a video game interface, and that bothers me.

File-based routing is fundamentally flawed and it cannot be fixed. A number of libraries opt for it since it's easier for newcomers to pick up, but eventually you run into all of the cases where you do need something else. This in turn leads you to a hybrid system of multiple things where there's no single source of truth and everything is spaghetti.

Let’s not pretend that useState() is plain TypeScript either. It’s a DSL in disguise.

JSX is amazing for stateless rendering of templates. Not so much for state management. That should really have been given a dedicated DSL. Here I think Marko did the right thing, why they then made even for-loops a dsl is more questionable.

404

Some interesting context from the podcast:

> A lot of people are coming to eBay from a link that someone shared or from a search engine... This whole "amortized cost savings" you get from a Single-Page App (SPA) you don't necessarily get with eBay. Or people might go to eBay and open up ten [browser] tabs... If that's ten SPAs you're opening you're not really saving that much.

> At the same time in 2012 people are coming out with React, Angular... the question was "can we just use these tools?" and the answer was "kinda no"... Initially React was considered but the things we needed right out of the gate was streaming [sending as much HTML as... available without waiting for services responding with loaded data for the specific page]... With streaming you can send out stuff to the browser and have the browser [start] showing content to user without having to wait for your slowest service. At eBay there are a lot of services... Essentially if we were to adopt React or Angular the fact that there wasn't streaming would essentially mean that we're throwing away two seconds or so... which is not acceptable.

I mostly agree with you but React isn’t just JavaScript. JSX is not JavaScript. It’s just that we’re so used to it we don’t consider it notable any more. Worth keeping in mind when you’re looking at a brand new framework.

      <b><MyComponent attr=“yes” /></b>

      <b>{MyComponent({ attr: “yes” })}</b>

   <b>{jsx(MyComponent, { attr: "yes" })</b>

    <{tagName} />

    <span {someStringWithAttributes} />

    <div>{stringContainingHTML}</div>

      var ListComponent = () => {
        let count = 0, selected = null
        return {
          view: ({attrs: {items}}) =>
            m("div", [
              m("p", "Clicked: " + count + " times"),
              m("ul", items.map(item =>
                m("li", {
                  onclick: () => { count++; selected = item },
                  style: {cursor: "pointer", color: item === selected ? "blue" : "black"}
                }, item)
              )),
              selected && m("p", "Selected: " + selected)
            ])
        }
      }

    const Hello = () => div(
      p("Hello"),
      ul(
        li("World"),
        li(a({href: "https://vanjs.org/"}, "VanJS")),
      ),
    )
    van.add(document.body, Hello())

    import { createElement as m } from "your-jsx-compatible-library";
    
    var ListComponent = () => {
      let count = 0, selected = null;
      return {
        view: ({ attrs: { items }}) =>
          m("div", null,
            m("p", null, "Clicked: " + count + " times"),
            m("ul", null, items.map((item) =>
              m("li", {
                onclick: () => { count++; selected = item; },
                style: { cursor: "pointer", color: item === selected ? "blue" : "black" },
              }, item)
            )),
            selected && m("p", null, "Selected: " + selected)
          )
      };
    };

  <MyComponent prop={value}></MyComponent>

  jsx(MyComponent, { prop: value })

To each their own. This syntax actially resonates with some people, which is why template-based frameworks like Vue and Svelte are also popular. In fact, at first glance this reminds me a lot Vue in some of its approach and syntax.

BTW - with Vue you can use entirely JSX of you dislike HTML component syntax (I don’t know enough about Svelte to know if it allows the same).

Do you really believe React is just javascript?

Agreed on the syntax part. We’ve had good syntax for templates before:

  <ul>
    {% for user in users %}
      <li>{{ user.firstName }}</li>
    {% endfor %}
  </ul>

And we have good syntax for templates now:

  <ul>
    {#each users as user}
      <li>{user.firstName}</li>
    {/each}
  </ul>

Why do we have to squish everything into HTML-like-but-not-quite blocks?

But no, JSX isn’t that great either:

  <ul>
    {users.map(user => (
      <li>{user.firstName}</li>
    ))}
  </ul>

    <ul v-if={users}>
      <li v-for={some in users}>{some.name}
    </ul>

  <ul #if={users}>
    {#each users as user}
      <li>{user.name}</li>
    {/each}
  </ul>

No Vue, is just JavaScript. React is JSX.

React has not been just JavaScript for a long time. The react DSL just keeps getting more and more bloated as they add more hooks for more and more edge cases (useFormStatus, useActionState, etc…). It’s becoming just another bloated mess now. And I used to love react! This looks promising though. The syntax looks very straightforward. Even though it’s not “just JavaScript” it is very easily understood by most programmers. I’ve glanced at it for all of 2 minutes and it all makes perfect sense. Functions look like functions. Variables look like variables. I think it looks cool!

In a past life we did try it... Marko has been around for years.

It started as a project at eBay and then got spun out to an open project around 2015

Really, this point can't be stated often enough.

It was my reason for switching to React when I learned TypeScript after getting more into JS frameworks via Vue.JS.

My starting point was Vue 2.7, and I started out using string templates haha :)

Even wrote some reactive custom code (without Vue, just regular DOM code) in a customer widget that utilized Object.defineProperty et al, inspired by Vue.

And today, while I'm using React at $job, I also think Vue 3 is probably a solid framework as well.

Last time I checked, they improved on DX of their component and templating system. But I'm not even sure whether they still recommend the v-if etc helper tags.

For what it's worth, even Vue 2 always also supported JSX and later TSX

Yes, we go in circles, but there are subtle (and sometimes not so subtle) improvements every iteration. Of course sometimes there are also dead ends.

It is exciting to see what the ingenuity of the next group brings, even though some existing things are lost, but hopefully not forgotten.

I’d venture to say that the idea of a “correct paradigm” is based on a false premise. Why would there be one paradigm to rule them all? Maybe there is more nuance. Maybe certain paradigms are better for certain applications.

We also had JSF, which was even cooler - being able to reconstruct the state server-side. It was ridiculously fast to write complex form-driver websites with that! No DTOs/schemas in different languages, no worry about how the client calls the server, what happens if it fails, etc.

The only problem is that it won't necessarily scale to some insane numbers without some care.

(Not sure why the past tense, it does work and developed still)

> It was the correct paradigm all along

Debateable.

> It was the correct paradigm all along but millennials wouldn't be caught dead working with such a "lame" technology so they bestowed SPA on us and now they are slowly walking it back.

Oh man, I wish people would stop attributing picking SPA's to not wanting to use "lame" technology. It makes them sound myopic and naive. Really naive.

You may not have been around at the time, but I certainly was. And the idea that SPAs don't (or didn't) have their place is just plain absurd.

Like "Tell me you're a backend dev who is made they have to learn new stuff to stay current" without telling me.

In fact, I'm not even sure what your argument actually is. Is it MSP vs SPA? Is it JSP vs any of the other backend languages you can still use with React? Is it templates vs JSX? What are you actually trying to argue?

Are your rose-colored glasses ignoring what a pain a decent UX was to build in those technologies? That mobile apps pushed backends to an api-only version and SPAs took over for the frontend? Are you saying people should pick old technologies wtih old tooling just because you didn't get on board with it the first time?

It's not swinging back to JSP, it's finding a middle-ground between the 2. THAT'S what progress is.

Could you please try to be a bit more substantive with your comments?

E.g., Why do you think JSX is the best? What limitations did you hit with those other template languages?

  // data
  const numbered = true // or false
  const strings = ["a", "b", "c"]

  // render
  const items = strings.map(i => <li key={i}>Item {i}</li>)

  if (numbered)
    return (<ol>{items}</ol>)
  else
    return (<ul>{items}</ul>)

I written lots of Vue/angular/react (like more than 3 major projects with each) and I'm a firm believer that:

1. Jsx is nicer to write but that's a non issue

2. vue's and angular's directives and bindings lend themselves to much saner and lighter rendering (performance does matter)

3. Vue is much easier to tame, it's reactivity model does not require a PhD in hooks or a deep understanding of React's internals. It just works out of the box as it should and is much easier to control the rendering lifecycle.

At the end of the day, after many years, my preference goes with Vue and Nuxt especially which is tremendously better than the monsters of Next or RR. That's what pays the bill the easiest and is eventually easier to maintain and optimize.

Have you ever given lit html a whirl?

I've been liking the model of the Python library Dominate [1]. You write your HTML as regular Python code, and you render() once at the end, having full control over the formatting. Well, at least in theory; in practice the formatting is brittle and the library otherwise makes some choices I don't like.

I wrote a Rust library with a more restricted/verbose API, and I've been enjoying using that. Unfortunately, I find it really hard to make it as fast as I want. It's really the perfect use case for arena allocation, but doing that and keeping the API as function calls mirroring HTML is not trivial, and probably requires macros to rewrite the tree as a series of stack pushes.

1. https://pypi.org/project/dominate/

I dunno, to me that seems like all YAML's mistakes all over again. I quite like the conciseness, and significant whitespace seems like a good match here, but the double hyphen thing really seems odd to me. And the syntax is so hard to parse, apparently, that their own example is syntax highlighted incorrect, coloring content as if it's tags.

I was looking at Marko a few years ago because of the concise syntax. I have always thought highly of Pug and would have loved a framework that integrated that sort of elegant, minimal syntax. Unfortunately, Marko doesn’t even get the syntax highlighting right in its own docs for this style.

The example on that page with leading commas to separate tag attributes, and a number of other choices across the framework are also a turn off for me personally.

I’ve mostly been using Svelte for the past half-decade instead but still hope for something more elegant to come along.