TinyJS – Shorten JavaScript QuerySelect with $ and $ | Hacker News Reader

TinyJS – Shorten JavaScript QuerySelect with $ and $$

github.com

102 pointssynergy2084 comments

I've been using this for years:

    const $id = new Proxy({}, {
        // get element from cache, or from DOM
        get: (tgt, k, r) => (tgt[k] || ((r = document.getElementById(k)) && (tgt[k] = r))),
        
        // prevent programming errors
        set: () => $throw(`Attempt to overwrite id cache key!`)
    });

It's nice to be able to refer to elements by property name, so:

    <div id="thing"></div>

Is reachable with:

    $id.thing

And, since the underlying structure is just an object, you can still enumerate it with Object.keys, which can sometimes be a useful debugging aid and general catalog of accessed elements.

Anyways.. Proxy is a wildly underappreciated and used class in JavaScript.

I also put together a Proxy microlib a few years ago. I agree it's very powerful.

You can use it like:

    $('.my-elements').className = 'Important'; 
    $('.my-elements').style.color = 'Red';     
    $('.my-elements').classList.add('Another'); 


    const $ = (() => {
        function listProxy(arr) {
            return new Proxy(arr, {
                set: (t, p, v, r) => {
                    for(let x of t) {
                        x[p] = v;
                    }
                },
                get: (t, p, r) => {
                    if(t.length > 0 && t[0][p] instanceof Function) {
                        return (...args) => { Array.prototype.map.call(t, (x) => x[p](args)) };
                    } else {
                        return listProxy( Array.prototype.map.call(t, (x) => x[p]) );
                    }
                }
            })
        }

        return (sel, root) => {
            if(root === undefined) root = document;
            return listProxy(root.querySelectorAll(sel));
        }
    })();

demo: https://codepen.io/anon/pen/RxGgNR

I built a Proxy-based microlib for making fluent REST calls just on a lark a couple years ago. It was never anything production-ready, but it was so handy that I used it in all of my JS projects until I moved away from webdev. The API was basically:

    const api = new Proxy({route: baseRoute}, handler); // handler is the microlib's export
    const result = await api.get.some.route.invoke();   // GET {baseRoute}/some/route

invoke() is just how it finally fires the call. I didn't feel like spending too much time making it automatic, the benefits just weren't large enough to justify compared to just calling invoke().

Elysia Eden [1] uses the same thing, except putting `get` to the place of `invoke`

`api.some.route.get()`

[1] https://elysiajs.com/eden/treaty/overview.html

Element IDs are automatically attached to the window object

  <script>
  document.addEventListener('DOMContentLoaded', () => {
    window.thing.innerHTML = 'Hello, World!';
  });
  </script>
  <div id="thing"></div>

That's (much) slower and has surprising edge cases: https://news.ycombinator.com/item?id=32997636

    <div id="parent"></div>

Whoops.

Does caching help? If so, shouldn't getElementById do the same thing under the hood?

It used to more in the past.

The native call does usually include a cache and will use it optimistically; however, the strategies around invalidating this cache are varied and some surprisingly basic DOM actions can trigger it. Then browsers usually fallback to the full DOM query case.

The cache eliminates this variability, which prior to flexbox, could be very useful in highly nested site designs particularly in mobile contexts.

If you like brittle things, the id attribute is already made into an attribute on the window for legacy reasons

Edit: My tone may have indicated that parent's solution was brittle. It's not!

The id attribute can take on values that are already present or reserved in window. "fetch", "opener", etc..

The reason to have a separate system that correctly calls getElementById() is to avoid this issue.

So, it's actually a _less_ brittle mechanism that doesn't rely on legacy mechanisms and lacks the surprises that come with that.

Sorry, I didn't mean to suggest your solution was brittle -- I actually quite like it and want to adopt it!

But I do think the legacy browser behavior with the ID attribute as window properties is very brittle for the reasons you suggest

My fault, I tend to "rapid fire" sometimes. Yours was next to another reply on the identical subject and that caused me to mistake the meaning of the comma in your sentence.

Reading it more slowly, I see it now, and, thank you!

Nah, i rather skip brittle :) But what's not brittle in JS? Elm? (TS to some extend)

This is awesome! Proxies can be so cool. I made this proxy many years ago for tracking stats/buffs/debuffs/gear for game things.

https://github.com/MrLeap/Statsi

Should that be `Object.create(null)` to avoid problems with `<div id="toString"/>`?

A call to `$id.toString` actually yields the native unbound function whether or not there is such an element on the page. So get is not invoked in this case.

Why cache? I expect that getElementById will be efficient enough on its own.

Why would this make any sense?

    const myDiv = div(
      {id: 'container', className: 'my-class'}, 
      h1('Hello World'), 
      p('This is a dynamically generated paragraph.')
    );
    document.body.appendChild(myDiv);

That's completely unnecessary these days with template strings. It's gonna be much faster as well to use browser's native parsing.

    const div = document.createElement('div');
    let text = 'This is a dynamically generated paragraph';
    
    div.innerHTML = `
      <div id="container" class="my-class">
        <h1>Hello world</h1>
        <p>${text}</p>
      </div>
    `;
    document.body.append(...div.children);

Keep it simple!

Generally, I would recommend avoiding directly setting the innerHTML property, since it's vulnerable to XSS and other injection attacks. If you do, make sure you HTML-escape each variable you interpolate.

Here's a way to do that with a tagged template function (named it `htmlFragment` to make it super clear what it's for):

    // a couple of functions to help
    
    const sanitizeHTML = (unsafeStr) => {
        const div = document.createElement('div');
        div.textContent = unsafeStr;
        return div.innerHTML;
    };
    
    const htmlFragment = (fragments, ...variables) => {
        const result = variables.map((variable, i) => fragments[i] + sanitizeHTML(variable));
        result.push(fragments[fragments.length-1]);
        return result.join('');
    };
    
    // updated your code here
    
    const div = document.createElement('div');
    let text = 'This is a dynamically generated paragraph';
    
    div.innerHTML = htmlFragment`
      <div id="container" class="my-class">
        <h1>Hello world</h1>
        <p>${text}</p>
      </div>
    `;
    document.body.append(...div.children);

Unfortunately, to my knowledge there isn't yet a close-to-the-metal solution for templating and data binding in HTML/JS, although several proposals are currently being discussed.

> Unfortunately, to my knowledge there isn't yet a close-to-the-metal solution for templating and data binding in HTML/JS

I'm hoping for this one:

https://github.com/WICG/webcomponents/issues/1069

Would you not just be able to do the following:

    const div = document.createElement('div');
    const text = 'This is a dynamically generated paragraph';


    div.insertAdjacentHTML("beforeend", `<div id="container" class="my-class">
        <h1>Hello world</h1>
        <p>${text}</p>
      </div>`);

    document.body.append(...div.children);

Edit, figured I'd add the docs: https://developer.mozilla.org/en-US/docs/Web/API/Element/ins...

That's certainly better than innerHTML, but it has the same security implications as setting innerHTML.

From the MDN page:

> When inserting HTML into a page by using insertAdjacentHTML(), be careful not to use user input that hasn't been escaped.

You can use tagged template functions to escape `${text}`. The result is pretty close to lit-html [1].

[1]: https://lit.dev/docs/v1/lit-html/introduction/

beautiful!

One big difference and why I've been experimenting with JSX[1] is that in that example, if the `text` comes from an untrusted source it can lead to XSS, which TinyJS prevents (I checked)!

[1] https://x.com/FPresencia/status/1838176000267452704

Even if it comes from a trusted source, you usually want a good distinction between html interpolation and just-text chunks.

This is exactly what I was thinking. I'm always trying to have fewer third-party dependencies in my codebase no matter how tiny, especially if it's solving problems that already have platform/system native solutions.

Makes sense to me, looks better. writing html strings like that is annoying

you should probably worry about untrusted data in text and use a tagged template function that sanitizes its inputs for insertion in html.

how would you add listeners if you were doing an html string? the first approach allows you to do something like this

``` const myButton = button({ onclick: e => console.log(e), textContent: "Click me" }) ```

Use topElement.querySelector('button'), then add the listener programmatically.

I use this:

    export const $ = document.querySelector.bind(document);
    export const $$ = document.querySelectorAll.bind(document);

When using TypeScript the types for querySelectorAll are a bit hairy to map but by defining the consts like above the types "just work".

    for (const tag of $$<HTMLParagraphElement>("p")) ...

I don't use Array.from in the result of $$ because sometimes creating an Array is not necessary. The NodeList returned can be iterated directly or converted to an array later if really needed:

    [...$$("p")].map(p => ...)

Since I use TypeScript I lean on TSX for building HTML. I use preact's render-to-string package to convert it to a string [1].

---

1: https://github.com/preactjs/preact-render-to-string

You're a wizzard!

JS keeps amazing me. It reminds me of Perl somehow. If Perl and Lua had a child, with some browsers APIs sprinkled on top.

I do believe I've seen something like this under another name.

Using "$" to shorten JavaScript? That seems a lot like jQuery.

> This README was generated by ChatGPT

You don't need AI to explain this one.

That explains the pompous introductory tone. Ugh.

I think the point is to be a more lightweight version, same spirit as HTMX

If ya don't want to do includes,

` window.$ = document.querySelector.bind(document) window.$$ = document.querySelectorAll.bind(document) `

Thanks, this is the more useful bit of the code to me.

edit: how about this so you can use array functions too when you need?

window.$ = document.querySelector.bind(document);

window.$$ = document.querySelectorAll.bind(document);

window.$$$ = (selector) => Array.from(document.querySelectorAll(selector));

I’ve been doing this for years. Lightweight but still saves me a ton of time.

You didn’t need a library to do this. Just alias document.querySelector and document.querySelectorAll to something shorter. Polluting the global namespace with functions to document.createElement on every possible html tag is not a good idea.

may I ask what's the point of `const tagNames =`? removing it would make no difference, as far as I can tell - just make sure the previous line ends with a semicolon (or add it at the same line as the opening square bracket)

Well, it's polluting the global window scope with all those helpers.

That's a reason to never use this imo.

There are plenty of other jquery-lite helper libraries out there. 95 percent the utility with hardly any cost in terms of script size.

Converting querySelectorAll to an array is honestly most of the benefit of these libraries.

> Converting querySelectorAll to an array is honestly most of the benefit of these libraries.

Now that NodeList is iterable, I don't even know that converting to an array is that useful anymore.

[...document.querySelector('div')] is pretty easy.

Iterator.from(document.querySelector('div')) too (doesn't work yet in Safari).

You don't need to convert querySelectorAll to an array, you can directly iterate it, or use forEach.

Not if you want to map or reduce it.

Yes, exactly. The first two statements have no use in the last two lines that this library is for. Or may be I don't understand Javascript much.

Show HN: Tiny JS – A Tiny JavaScript Library for Easy DOM Creation - https://news.ycombinator.com/item?id=41462817 - Sept 2024 (4 comments)

Show HN: A better way of writing HTML via JavaScript? - https://news.ycombinator.com/item?id=41451559 - Sept 2024 (9 comments)

Code is actually tiny. Impressive!

You could make it even smaller by removing all the element names, and just have it be passed in as an argument. Would also reduce the amount of functions you need to declare, though function count is likely not a problem for modern JS engines.

Anyway, great job you did there!

I remember when we would throw this into our utilities.js scripts back in like 2007. We didn't have querySelector yet, though. We'd also walk to school with no shoes, in the snow, uphill, both ways.

Reminds me of some simple tools I used years ago when CoffeeScript was big: https://github.com/jawj/affogato/blob/master/functions.coffe...

Nowadays I tend to use Mithril instead, because immediate-mode UI is so much less hassle than retained-mode: https://mithril.js.org/

I wrote my first jQuery lite selector clone a bit ago, any feedback would be appreciated, especially since I don't see the need to separate out querySelector & querySelectorAll

``` const $ = (param) => { if (typeof param === "string" || param instanceof String) { const elm = document.querySelectorAll(param); return elm.length > 1 ? elm : document.querySelector(param); } else { return [param].length === 1 ? [param][0] : [param]; } }; ```

The title ("TinyJS – Shorten JavaScript QuerySelect with $ and $$") seems to be about these two lines:

    window['$'] = selector => document.querySelector(selector)
    window['$$'] = selector => Array.from(document.querySelectorAll(selector))

The other part is a small & naive (and I mean that in the best way) DOM abstraction.

I think the discussion here seems to be about one or the other, not the thing as a whole.

Calling those $ and $$ selector helper functions new is a bit disingenuous. They’ve been present as Chrome dev tools shortcuts for years. Let alone JQuery, I’ve also spent years seeing little articles about this neat trick on everything from medium to dev.to.

I recommend the author remove that or note inspiration from the README to avoid detracting from the rest of the work. I realize there’s more here than the traditional document.querySelector.bind(document) assignment.

Like jQuery but for 2024?

That would just be jQuery.

Yep. jQuery 4.0 is in it's second beta[0] and seems to clock in at 27kB. And the "slim" version is <20kB.

[0] https://blog.jquery.com/2024/07/17/second-beta-of-jquery-4-0...

jQuery? What are you, five? We use jjQuey.

I dig it

> This README was generated by ChatGPT

Damn good README. Also, good library.

Too wordy yet doesn’t even explain the tag part of the API. It just shows up in an example.

A readme should be more reference / show and less drivel.

Based on the readme this is a subset of jQuery primarily targeted at creating dom nodes.

Kind of cool if this is what you need.

With that name I'd expect a tiny implementation of JavaScript.

I was just reading about this https://porffor.dev

> Porffor is a unique JS engine/compiler/runtime, compiling JS code to WebAssembly or native ahead-of-time.

Hi everyone, author here. Let me know if you have any questions.

Sheer pollution. Never in anything I deploy to production.

Pls no, im tired boss

jQuery II: The Quickening!

I recently built a toy Golang SSR project using Zepto, a jQuery-like library, and felt like I was 17 again (**EDIT: it's unmaintained - don't use it). Also of note is that "Highlander II" takes place in the year 2024 [1]. It's a sign! Everything old is new again! `$` is immortal!

[1] https://en.wikipedia.org/wiki/Highlander_II:_The_Quickening

Zepto is unmaintained, which isn’t good for security issues or bugs, unfortunately.

jQuery proper is both maintained and has actively been working on a 4.0 release and still gets security and bug fixes

Thanks for bringing this up. Edited my OP - there's no impression as such on their front page and it seemed fairly ergonomic. My bad on that one.

I also created almost 10 years ago Umbrella JS, which is a tiny jQuery replacement but with more array-like arguments in the functions:

https://umbrellajs.com/

https://www.bennadel.com/blog/4184-replacing-jquery-110kb-wi...

IIRC, I used Zepto with Phonegap around ~2012. What a blast from the past.

"There can be only one"

Someone will fork this with proper ES modules and no window clobbering.

This is the whole code:

    (() => {
    const assignDeep = (elm, props) => Object.entries(props).forEach(([key, value]) =>
    typeof value === 'object' ? assignDeep(elm[key], value) : Object.assign(elm, {[key]: value}))

    const tagNames = ['a', 'abbr', 'address', 'area', 'article', 'aside', 'audio', 'b', 'base', 'bdi', 'bdo', 'blockquote', 'body', 'br', 'button', 'canvas', 'caption',     'cite', 'code', 'col', 'colgroup', 'data', 'datalist', 'dd', 'del',     'details', 'dfn', 'dialog', 'div', 'dl', 'dt', 'em', 'embed', 'fieldset',     'figcaption', 'figure', 'footer', 'form', 'h1', 'h2', 'h3', 'h4', 'h5',     'h6', 'head', 'header', 'hr', 'html', 'i', 'iframe', 'img', 'input', 'ins',    'kbd', 'label', 'legend', 'li', 'link', 'main', 'map', 'mark', 'meta',     'meter', 'nav', 'noscript', 'object', 'ol', 'optgroup', 'option', 'output',     'p', 'param', 'picture', 'pre', 'progress', 'q', 'rp', 'rt', 'ruby', 's',     'samp', 'script', 'section', 'select', 'small', 'source', 'span', 'strong',     'style', 'sub', 'summary', 'sup', 'table', 'tbody', 'td', 'template',     'textarea', 'tfoot', 'th', 'thead', 'time', 'title', 'tr', 'track', 'u',     'ul', 'var', 'video', 'wbr'  ].forEach(tag => window[tag] = function(...args) { 
    const props = typeof args[0] == 'object' && !(args[0] instanceof HTMLElement) ? args.shift() : null
    const elm = document.createElement(tag)
    props && assignDeep(elm, props)
    elm.append(...args.map(a => typeof a == 'string' ? document.createTextNode(a) : a))
    return elm
    })
    window['$'] = selector => document.querySelector(selector)
    window['$$'] = selector => Array.from(document.querySelectorAll(selector))

})()

    window['$'] = selector => document.querySelector(selector)
    window['$$'] = selector => Array.from(document.querySelectorAll(selector))

is really what I want to have

I'm quite fond of a little helper like so:

    createElement({
        className: 'p-2 flex justify-end',
        contents: createElement({
          tag: 'img',
          src: '/os-logo-maps.svg'
        }),
      });

I got the idea from Lea Verou's blissfuljs. When you're creating a bunch of nested elements it comes in handy.

That’s React without the JSX sugar.

The idea is older than react and react contains much more cruft (both wanted and unwanted).

Yes, and?

What's the point of the rest of the code? Looks useless.

Looks difficult. I have no clue what I'm reading there tbh.

It's creating global helper funcs for creating elements. E.g. `a(...)` for an anchor

Why do you pollute the global scope with all this crap? It's 2024, dude. Just make an ESM module.

I'll just leave this here: https://www.youtube.com/watch?v=Uo3cL4nrGOk

Reply on news.ycombinator.com