alt Hacker News> React is a relatively well-designed library that isn't really that bloated.
I am one of the people who irreconcilably disagree. I think svelte and solidjs are - obviously - technically superior. But it leaves the question of why react is still so popular. I think the biggest reason is all the non-technical aspects of react:
- They have excellent documentation. And have, from day 1.
- They produced videos, sample projects, and all sorts of "getting started" documentation.
- They ran react conferences, teaching everyone who would listen about "1 way data flows" and pretending like they invented FP.
The amount of hype around it made it really feel like the next big thing. Between the very well funded react team and the outside developer community, there was real momentum. People learned it in droves. Taught students. Built websites with it. When react's poor design choices caused issues (and there were a lot of issues), then you were blamed for holding it wrong. (Component classes, state, CSS, hooks, webpack and babel taking ages, big bundle sizes, slow re-renders, and so on.)
By the time the next generation of JS frameworks broke onto the scene, there was a collective moan from the community. "Oh no, not again - we just relearned how to make websites." React was the wave, and in its wake we all had Framework fatigue.
Software doesn't get popular without a lot of work by dedicated people. I really admire the work standards bodies do. But they rarely bother to take the time to produce documentation, videos, tutorials, starter projects, blogs and podcasts and all the rest of that work.
Replies
There really isn't anything in react that is inherently "bloated", most companies were indeed using it wrong, or at least in a naive way. Over-reliance on useEffects, re-render issues, non-memoized components etc. When used correctly, react 18 provides really fast web apps on whatever scale (small one pager app to massive enterprise apps).
I have to step in here to say that, at least documentation wise, this is revisionist history.
The original react documentation is fine, but focused almost entirely on class components. What was there on hooks was either outdated or outright harmful.
It's not until 2023, 4 years after react hooks launched, that the new react docs which focused on them were released.
I haven’t found a front end library I like more than React except maybe Astro.
I didn’t really like Svelte and I am pretty opposed to htmx. It is certainly better than Angular and Vue, though Vue 3 is tolerable. Maybe there are more popular options nowadays but tbh I am not really looking to move from React. I have zero complaints other than maybe the push towards server rendering and partnering with Next.js
React is simple and I like that it feels a bit like FP. I love that I can write just about my entire application in TypeScript with all of the benefits that incurs. It requires some knowledge/discipline e.g. around useEffect.
If you care to use native browser APIs and features you can have lightweight, performant applications that behave well in the browser.
Most devs don’t do this, even with vibe coding. It’s a very easy way to filter for those who care about what they are building.
> I am one of the people who irreconcilably disagree. I think svelte and solidjs are - obviously - technically superior.
I don't really think that React is necessarily the best possible library, but there is certainly more to a UI library than simply compiling out the need for diffing. For one thing, I find JSX to be a relatively unobtrusive addition to the language. It is implemented by a variety of things and is a pretty simple transform as far as things go; basically pure sugar, it's possible to avoid it if you want. It could be better designed than it is, but I think it's at least sufficiently general - it is feasible to say, use an alternate library with JSX, like Preact.
Svelte on the other hand by its nature just simply requires a more complicated compiler step to work and deliver on its promises. That's a trade-off. Whether you think it's worth it is up to you, but presenting it as an objectively technically superior design is disingenuous. Superior how? Runtime costs? Sure, but it does beg the question how much. Will a well-optimized Svelte app feel much better than a well-optimized React app?
> The amount of hype around it made it really feel like the next big thing.
We are a really long way away from the hype of React. I would argue that hype stopped carrying React a long time ago. Momentum? Sure, but if something was truly dramatically better, then I think it would have a fair shake at beating out React.
The problem is that the web isn't microbenchmarks and developer experience does matter to some degree, so the trade-offs that seem so good on paper don't always pan out.
I'm not a hater of Svelte. On paper, it is a very elegant idea. However, when I actually tried to use it, I genuinely came out feeling that it was simply not for me, and the benefits it has are not enticing enough for me to continue experimenting. My React apps are not particularly slow or bad at handling huge amounts of data. (One of my apps has no problem handling a file grid of over 1 million items, and in fact, I run into issues with browsers not being able to handle a large enough scroll area long before performance of my React code is a problem. That is good enough for me.)
React has an incredible ecosystem, it's almost synonymous with web development. The available render backends alone, namely React Native, are currently unbeaten.
As interested as I may be in the alternatives, they simply don't have this.
I don't think React pretended like they've invented FP, and I've always heard them be pretty open about FRP work existing in the past and them leaning on it.
I mean, prior to React the big Framework at the time was AngularJS 1 - a hellscape of two way data binding.
Talking about one-way data flows was an excellent way to speak to a lot of tortured souls at the time (who had just found out they were going to be forced to migrate their projects to something else anyway due to Google's absolutely insane "we're deprecating this, we'll have a replacement for it at... some point, good luck").
> They ran react conferences, teaching everyone who would listen about "1 way data flows" and pretending like they invented FP.
When you have to resort to outright lying about something it's usually a good sign your argument is flawed.
> pretending like they invented FP
More like trying to explain to people who have only done imperative programming (and that also in JS of all languages) why a render function was necessary. The creator of React, Jordan Walke, was big into OCaml and one might even say he got some of the ideas of React through working in OCaml, so much so that he tried to bridge both by making ReasonML which is an alternative syntax for OCaml which has JSX and React bindings.