Making Web Components Work
engineering.mixpanel.com
engineering.mixpanel.com
One major issue we've had though, which wasn't touched in this article, is inter-component communication among siblings. Sure you can just do a global event bus, and that works, but I've yet to see an elegant solution for this yet.
Tossing in Redux (or whatever Flux-like flavor of the day) is a low-effort, high-value add on top of React and obviates the possibility of some bat-shit data-passing situation that would make an OO inheritance junkie blush.
If you're the sensitive to boilerplate type, there are some simple libraries to solve that issue. Unless you're tossing together a three component TODO list demo, not using Redux will just bite you in the ass a year down the line in maintenance costs as your application grows.
This is really starting to remind me of the "Don't worry about the data model, just toss it in Mongo!" all over again.
When I see folks saying, "Oh just deal with it in the components" I have to assume either A) they're working on an extremely small, self-contained piece (which is fine) or B) they're not bothering to think about their data from the get-go.
Would Redux potentially be a good fit? It's a "global event bus" kind of thing, but with a pretty clear information flow model and an increasingly large user community. (Both of which are good things, IMHO.)
Redux feels like a step in the right direction conceptually with its centralized state bag, but I feel like it's a bit distracting (since it was primarily conceived to handle serializable actions for a very specific use case), and also doesn't have good patterns out of the box to organize side-effects (sagas kinda do, but people I've worked with are fairly divided on whether it works for them or not)
Recently I've been thinking that a lot of the complexity around React+Redux comes from wanting testable state, but as it turns out, Jest can mock imports making the entire context shenanigans somewhat pointless - just import a global state bag. I've had pretty decent success with just having a naive global state bag, would love to hear about similar approaches
Initially I was wondering why we didn't have a centralised store, but once free from that mindset it allows you to keep observable data locally to where it's needed. Much nicer for encapsulation and separation of concerns.
Far less boilerplate too, as MobX does some very smart usage & dirty checking for you.
mobx-react also provides an @observer class annotation, which ends up working pretty magically - things re-render when they need to.
If anyone's interested, here are some tutorials from the author: https://egghead.io/courses/manage-complex-state-in-react-app...
In a discreet component architecture there is basically no reason for arbitrary siblings to know anything about each other - if there is common state to be shared it should be inhereted.
Granted, in JS land that sometimes involves jumping through weird IoC hoops or cached callbacks, which are akward to say the least, but in the abstract an instance of a component should be concerned with its own behaviour, and possibly that of it's descendents by way of that behavior.
[0]: https://vuejs.org/v2/guide/components-custom-events.html#syn...
You can do better than template literals. You can use TSX, which gives you compile-time verification for tags and attributes, in addition to JavaScript expressions.
See: https://github.com/wisercoder/uibuilder
With TSX (which is JSX written using TypeScript, you get things like syntax coloring and indentation for tags, because HTML and JavaScript are both first-class citizens in TSX.
It’s pretty flexible and thin.
React components are brittle, that's not FUD. For example, CSS classes used by React components can clash with other React components, or with the page that includes them. What's FUD about that?
That's one of its limitations.
React is a component technology and doesn't solve the problem of clashing styles, element ids and so on. Web components is a component technology that does.
> clashing IDs
React solves that one, not through namespacing though. You never need to use IDs with React.
> clashing styles
React gives you power to solve this one by your preferred solution. You can use expressions in your JSX, and thus it's very easy to do one of the following:
- Generate optimal global CSS stylesheet with no code repetition, generate classnames on the fly (https://github.com/typestyle/typestyle)
- Import your CSS file as a module using Webpack - classnames are guaranteed to be unique; additionally, similar approach as TypeStyle (global optimal stylesheet) can be set up
- Use just "inline" styles (preferably not really inlined in your JSX) since React will keep it updated for you as well
If you don't care about accessibility then yes.l, but try using properly annotated aria attributes without appropriately scoped ids(which you can do in react, but it's not free).
In fact, the criticism that react components can have style clashes is precisely one of not having enough separation of concerns. The fact that there are numerous libraries to address that deficiency further indicates that react as a platform is severely affected by the underlying level of abstraction when it comes to styling.
It's fine to defend a technology you like but calling a weakness a strength does nobody any favors.
Are you using some template tag function that handles escaping? Seems risky to rely on manual escaping everywhere.
The way we typically handle something like this is by connecting components through a common ancestor. The default naive behavior of Panel apps is for linked components to share a "centralized state bag" (as one of the other comments calls it re: Redux), and updates to that state automatically cascade to all the components. It's a blunt, easy and lightweight way to propagate changes arbitrarily throughout an app, but it does get unwieldy as the state object gets big so we've ended up with some mechanisms for partial state-sharing to address this issue.
Most communication I solve by custom event dispatching or attribute callbacks, but there are the odd cases, especially regarding more advanced drag/drop, that still requires state at some level. Do you have any concrete example of inter-component communication to look at?
Having a component tree (because the rendering model of a browser requires it) does not mean state management needs to be cramped into the same tree or use a tree in the first place!
A DOM tree itself doesn't do any kind of binding and it doesn't do inter-component communication. (it merely offers events) Why should web components? Keep them as dumb as DOM nodes please.
A lone H1 element doesn't register itself on some event bus, doesn't observe attribute changes and doesn't create odd 'actions' tied to some obscure state management system. And it shouldn't.
Zone js implements zones and by that they apparently mean exactly 1 zone.. making a supposed abstraction where you can't safely add 2 implementations of web components (or apps?) to 1 page without planning for that with 1 global setup for them..
My initial impression of shadow DOM is that I would get the same scope separation as iframes with a nicer method of passing data and describing the layout.. Then I assumed web components were this missing piece finally implemented.. now I'm wondering if I completely misjudged and everyone else really wanted a complicated footgun where seeming builtins are colliding with your namespace?
The result is your can never just use the basic DOM APIs and devtools to inspect and interact with your application's components. Instead, you always have to get at them through some other API, and understand how the UI and state you see in the DOM is produced and updated by some mediating layer (virtual DOM, view hierarchy, etc.). It feels like trying to get at the truth by talking with two different people, who each only possess a subset of the facts.
Perhaps there should be. Has there been any discussion on standardizing such a thing?
In reality, browser engineers are not that stupid. Mutating the DOM will queue a layout operation, but it will not actually occur until the current JS task has finished executing. Within a JS task, such as an event handler, an XHR completion, or a setTimeout, you can mutate the DOM as many times as you like, and it will only result in a single layout pass.
(The exception is if you try to read back some measurement from the DOM, such as an element's bounding box, after mutating it. In this case, the browser does have to block while it performs layout, but that is not something can be solved with virtual DOM).
So, what is the purpose of the virtual DOM? It was invented to allow React to provide the illusion of full re-rendering. React's authors wanted to a provide an experience similar to that found on the server, where an entire HTML page is re-rendered from scratch for every load. In that way, there is never any possibility of part of the HTML becoming stale, as it all gets recreated from scratch each time.
However, the browser DOM is not designed to be blown away and recreated from scratch all the time. Nodes are expensive objects, spanning the JS and C++ worlds. Recreating the whole DOM tree each time any part of it needed updating would be too slow. So, instead, they created the virtual DOM as an intermediate data structure. React renders the virtual DOM. The virtual DOM is diffed against its previous state, and then the changes are applied to the actual DOM tree. In that way, every component's render() method can be executed, but only those parts of the DOM that have actually changed will be updated.
It's a nifty optimisation, but it's not about avoiding reflow, it's just another method of dirty checking, similar to that done in other frameworks like Angular or Ember. It's just that React chooses to diff the data structure produced by render(), rather than diffing the model data that is later used for rendering.
The draw for me is that I don't need to manage the state of a bunch of DOM objects anymore. I wouldn't even mind if React were slower- the performance isn't really the point.
I don't and will never want web components for this reason.
It goes into some details how different approaches in different frameworks work, and why. Moreover, it shows how you basically need to re-implement Virtual DOM and other tricks in vanilla JS code to approach the same speed.
Yes, browser engineers are not that stupid. But they don't batch operations as efficiently as a proper Virtual DOM implementation would (including efficiently handling event listeners, looking for clues like `keys` on repeating DOM elements etc. etc.).
So the takeaway is that DOM mutation can be slower or faster depending on the techniques you use. Using a Virtual DOM library can help you improve performance, because these techniques are often baked in, but they don't intrinsically result from using a Virtual DOM: You could write a V-DOM library that didn't use them, and was slow, and you could write, for example, a static template library that did, and was fast.
In theory, a compiled, static template approach should be faster than V-DOM (or anything other than a set of totally bespoke and optimised vanilla JS functions for each app operation). Because while a compiled template does not have the turing-complete flexibility of something like JSX, it has the advantage of knowing in advance exactly which parts of the DOM can change and how, which eliminates the need to build and diff the Virtual DOM tree.
Maybe faster in basic microbenchmarks. As soon as you start optimizing UI library for complex applications, there are many other optimization goals you should focus on, like the size of the generated code per component type, the amount of different code paths(increase probability that executed code is JITed), etc.
Virtual DOM solves many issues, it is not just diffing, it is also a lightweight component model, possibility to add lightweight synthetic events, single code path for creating/updating/destroying DOM nodes, the simplest algo that handles complex tree transformations, etc.
Surplus [1] is the fastest in most benchmarks, and it doesn't use a virtual DOM.
1. https://github.com/localvoid/js-framework-benchmark/commit/1...
2. http://rawgit.com/localvoid/js-framework-benchmark/sandbox/w...
Inferno here is also slightly patched[1], instead of using stateless components, it is using stateful components to demonstrate how fast it can go down when you start using high-level abstractions in a library that focuses only on low-level primitives.
And if you really want to understand the fundamental flaw in libraries with fine-grained direct data bindings, try to reimplement this[2] 70 lines of React code with such library.
1. https://github.com/localvoid/js-framework-benchmark/commit/2...
2. https://github.com/localvoid/uibench-react/blob/master/js/fc...
Anyway, I have nothing else to say on the matter. Clearly anything with only 10% overhead over vanilla JS is definitely fast enough, and Inferno is there now if you want something React-like.
As an aside, do you have any experience with Ivi? I'm looking to learn something else and wondering if I should dig into Web Components or something like Ivi. Web component performance was terrible last I checked.
And now is the main question :) If virtual dom is competitive in benchmark that is super biased towards direct data bindings libraries, what is the point of using direct data binding solutions when they won't be able to handle even basic use cases that involve client-server communications when server sends data snapshot. There won't be any information about data changes, and you'll end up with reimplementing tree diffing algo, so that you can apply it to the data.
I'm not sure what you mean. Surplus is built on S.js, in which all data is lifted into reactive expressions. S.js already handles that for you but at the data model level, where it arguably should be, not at the UI model level.
It only works as long as it is able to track changes, many client-server applications doesn't send you list of changes that you should apply to your data, they just send you data snapshots. With virtual dom library I'll just update my data and rerender everything that depends on this data, with direct data bindings good luck figuring out solution to such simple problem.
So you have to visit every data node, and then also visit every changed UI node. Where if you have deltas, you only visit changed data nodes and then changed UI nodes.
Re: data snapshots, it's easy to design your own service to use a delta protocol. But even when you can't, you can separate the code used to construct your reactive objects from the code that initializes them. This is just basic function abstraction, and it doesn't really add any work. From the example on S.js site:
const //
a = S.data(1), // a() | 1 3 3 5
b = S.data(2), // b() | 2 2 4 6
c = S(() => a() + b()), // c() | 3 5 7 11
d = S(() => c() * a()); // t0 // d() | 3 15 21 55
a(3); // t1 // +------------------------>
b(4); // t2 // t0 t1 t2 t3
S.freeze(() => { //
a(5); //
b(6); //
}); // t3 //
const
a = S.data(1),
b = S.data(2),
c = S(() => a() + b()),
d = S(() => c() * a());
update(3, 4);
S.freeze(() => update(5,6));
function update(aval, bval) {
a(aval);
b(bval);
}
Or maybe I can just ask you one more time to reimplement this[1] 70 lines of React code with Surplus, but you are probably will just ignore it :)
1. https://github.com/localvoid/uibench-react/blob/master/js/fc...
In any case, there's no inherent difference between basic values and ordered lists and data transforms. The latter two are just recursive application of operations on basic values.
It won't be so hard to solve use cases that I am talking about. Just create an index(HashMap), implement simple diffing algo that updates hashmap index and doesn't care about item positions, then from the ordered list generate a new list with values from the hashmap index, and then apply this inefficient diff algo that uses object identity to track objects[2].
1. https://github.com/ismail-codar/surplus-material/blob/3dce38...
2. https://github.com/adamhaile/S-array/blob/1046fca3032691d4ef...
There's also a lot of "implicit state" not captured in such a virtual DOM, like the user's current selection, the scroll position, form values, and so on. If these are blown away, it creates a really poor experience. The virtual DOM isn't just an optimization, it is actually a correctness fix.
The only real missing piece is the concept of props over attributes. If I want to pass an array of ids to a sub-component declaratively, there's not really an easy way to do it. Hopefully there will be some consensus about this soon. The best thing is that this is all part of modern browsers without libraries, or even precompilation. Not every web app needs massive libraries, so having web components as a built-in, light-weight component system is fantastic.
Plenty of HTML templating languages allow for setting properties declaratively. Polymer, lit-html, Vue, Angular, all do.
For instance, in lit-extended (an extension of core lit-html), properties are set by default, attributes have a '$' suffix, and event handlers have an 'on-' prefix:
html`
<my-element
aProp=${foo}
an-attribute$="${}"
on-click=${clickHandler}>
</my-element>
`;I'm not sure that HTML itself will get a standard way to set properties, even with the Template Instantiation proposal. Encoding the intent to set a property is hard given that all valid attribute names are already valid attribute names, and could have current semantics. Adding new support for currently invalid attribute names sounds like a very difficult thing to do.
This can stay at the level of opinionated libraries. You learn how to use that library and it's method for setting properties, and you don't have to worry about other components choices for templating.
(BTW, lit-html is really well done. I look forward to seeing its progress - LitElement could really be the underpinning to a standards-based competitor to React, which is a very welcome thing in my opinion.)
This seems like a framework layer feature and not something that should be baked into the standard. I don't think web components are supposed to act like a framework and optimize a particular workflow or programming style.
I think the way snabbdom implements props just assigns Javascript properties to the DOM object.
Yes, there is. See https://github.com/wisercoder/uibuilder
Polymer has always been a set of helpers for making Web Components. There are others like SkateJS and Stencil.
Vue is not normally Web Components, but they have an option in CLI 3 to compile to Web Components.
React and React native have basically nothing to do with Web Components, though React can use Web Components because Web Components are just HTML elements.
Yup. And you can't use Polymer elements in React because React passes data as attributes whereas Polymer expects properties.
And then there's the issue of event handlers etc. etc. etc.
"Web Components" are now a largely non-intersecting collection of mutually incompatible frameworks and libs: https://custom-elements-everywhere.com
Good luck using Polymer's <custom-element binding="{{a.b}}"> anywhere
This is mostly React's existing opinion that elements don't even need properties set, and is not the fault of Polymer or Web Components. Most Polymer elements are fine with simple types like strings or numbers being set as properties or attributes.
> "Web Components" are now a largely non-intersecting collection of mutually incompatible frameworks and libs: https://custom-elements-everywhere.com
That is a site that says how well _frameworks_ work with Custom Elements, ie do they make closed-world assumptions about HTML that interfere with using Custom Elements. It says nothing much about what a Web Component is, so I'm not sure how Web Components could be "non-intersecting".
> Good luck using Polymer's <custom-element binding="{{a.b}}"> anywhere
I know you troll every Web Components and Polymer discussion you can now, but this is just a non-sequitur. Polymer's template syntax is only used to implement elements with Polymer. `<custom-element binding="{{a.b}}">` would only ever occur in a Polymer template and is not intended to be used "everywhere", so no luck is needed, thank you.
This is not trolling. This is called hard truths that Web Component proponents are completely blind to.
Riddle me this: why is it that there are almost no resources, blog posts, projects, or comments that talk about and use vanilla Web Components? 99.9% of posts and videos about WebComponents always immediately use Polymer, or Stencil, or... or... or...
The answer is easy: Web Components are crap, pure and simple. They are unusable in their present form. You need to throw in a data binding library, a templating library, a virtual DOM, a callback handling library, a..., a... etc. Better still, throw in a full framework like Polymer or Angular.
Even the project we're commenting on, titled "Making Web Components Work" needs the following to make them work: "does so by providing an easy-to-use state management and rendering layer built on Virtual DOM (the basis of the core rendering technology of React). Through use of the Snabbdom Virtual DOM library and first-class support for multiple templating formats".
How the hell is that "Web Components", "standards" and "work"?
When you call out all these inconsistencies, half-truths and outright lies about Web Components, you're called a troll, go figure.
---
As an aside. I remember saying that templating in Web Components is also crap because you have to use horrible DOM APIs or `innerHTML` to do anything useful. Obviously I was called a troll etc. However, when Apple proposed Mustache-like template instantiation [1] the same very people who called me a troll pissed themselves with joy and wonder. You know Web Components APIs and specs are crap. You chose to bury your head in the sand, turn a blind eye and preach how "WebComponents are the future". Well, it took browsers 11 years to get to two-thirds of jQuery's capabilities. It will take another 50 to maybe get one-third of capabilities of Web Component frameworks or React, or Angular, or Vue (hell, of even Snabbdom).
[1] https://github.com/w3c/webcomponents/blob/gh-pages/proposals...
I'm struggling to see the use of Web Components. And how easily it's going to be to compose WebComponents in with your custom frameworks' components, I wonder how easy that is going to be (e.g. passing props down to a web component).
It seems that you'll still be better off just picking a framework for the next wee while anyway.
Still, I'll look forward to your updated article ;)
Not to be harse. But for a company as big as Mixpanel, the blog UI is terrible. No branding at all and seems unprofessional.
Even more ironic is they say this as the first line in the post: "Clean, attractive user interfaces and effective user experience have always been pillars of Mixpanel’s products."
They should look at companies like GitHub, Facebook, Twitter, and the likes for an example.
Check the Mixpanel blog (https://mixpanel.com/blog/) and you'll find it the brand consistency more similar to the other tech companies you mentioned.
I'm actually okay with engineers spending less time on shiny css and more time on the content of their posts.
Maybe it was due to how simple the example was, but having to click back many times based on how many times a user used the interface is just bad design.
I'm looking forward to a future where we use web components.
How is this "making web components work" any different from making DOM work via snabbdom or React?
or is it strictly a web-browser-only thing?
Check out https://github.com/Polymer/pwa-starter-kit/wiki for concrete examples using LitElement (https://github.com/Polymer/lit-element)
That's pretty far from "not happening"
It's not Web Components that are happening, it's the dozens of frameworks required to make them work that are happening.
Even mixpanel's library "does so by providing an easy-to-use state management and rendering layer built on Virtual DOM (the basis of the core rendering technology of React). Through use of the Snabbdom Virtual DOM library and first-class support for multiple templating formats" [2]
Web Components are a very low-level extremely limited crummy DOM-based API that no one in their right mind wants to use. They are not happening.
---
[1] https://fronteers.nl/congres/2011/sessions/web-components-an...
Edit: looks like you've been doing this a lot, and it's really not what the site is for. Would you mind reading https://news.ycombinator.com/newsguidelines.html and using HN as intended? The idea is: if you have a substantive point to make, make it thoughtfully; if you don't, please don't comment until you do.