New Architecture is here(reactnative.dev) |
New Architecture is here(reactnative.dev) |
It was a mixed experience. When I migrated to expo two years ago, many problems were solved but not all.
But I still encounters bugs and problems with many common dependencies. It is not uncommon to have bugs on certain Android brands, with the community on github reporting the bug but waiting months for it to be fixed.
iOS is by far better and more stable than Android.
Performance is great on iOS, but less great on Android.
Our apps are animation heavy using react-native-reanimated and react-native-skia. Everything went perfect on iOS but we had to remove some animations or simplify them on Android.
Upgrading your dependencies every four months will probably break something somewhere : deep links stop working, some animations stop working, or maybe it's another dependency from the JS world. Sometime the fix is easy, other time an issue with the regression can be found on Github, other time we have no data.
Overall I'd say react-native is perfectly servicable and is easy to learn for anyone, which is a big plus. I'd recommend react-native because it is easy, have a big JS ecosystem, but I am now on the Flutter side.
They can be great as long as you and your customers stay on their most well-trodden path. But as these frameworks grow and become more byzantine, and as your project requirements start reaching for more rarified features and your customers start using new and differing runtime platforms, maintenance overhead starts to dominate and you find yourself running into invisible walls that make it hard for you to deliver on your project roadmap or satisfy the support standards you want for your customers.
This has always been the case for these frameworks, going back many decades, but especially since the explosion of efforts to build them around web stacks, which are easier for developers to use but harder for framework designers to keep sufficiently robust and capable as they age.
But I still feel in the end that for many CRUD style apps it's worth it to deal with react native's problems, especially if you can also have significant code sharing with your web app.
If you're trying to build the next snapchat or tiktok you'd better go full native though.
Flutter just feels like a much more polished and stable platform built explicitly for purpose and I've never had any performance issues.
Also how's the accessibility on Flutter? I'm pretty skeptical that it's going to have decent accessibility given the game-engine style rendering.
Which similar frameworks' developers have subscribed to the Android ecosystem and do much better work there?
Welcome to Android. That's not just a React Native issue
https://github.com/M66B/FairEmail/blob/master/app/src/main/j...
1 person talking about how they wish react native didn't exist
1 person asking about Capacitor
1 person complaining about Expo
1 person saying that they wouldn't use react native and recommending Kotlin Multiplatform instead
1 person complaining about the quality of the discussion (Me)
0 people talking about the new architecture
I still love Hacker News but the discussions are becoming increasingly pointless.
All that's missing is:
1 person complaining about the style in which the article was written
When they emerged, the mobile development scene was completely different than today.
Today, we have Swift UI and Compose, both are pretty solid. I'm not sure if it's the consensus amongst mobile developers, but I believe that on the mid/long run you will be better off - even if you write things twice. In terms of end user experience, developer experience, and in the business sense, everywhere.
Sure if you have an Flutter / RN app that has years of engineering efforts invested into it, go ahead and continue (duh), but I wouldn't start new apps with them.
It's often times fine on iOS and then incredibly slow on Android. Hermes is very exciting but still requires many polyfills to make simple NPM packages work. I hope one day, the web (and embedding web apps on mobile) makes React Native fully obsolete.
We don't use Expo, either. It's very painless.
That's not what react native is for. Its for building native applications with a react-like syntax for the views. To me it seems like you used hammer to put in screws, no?
That's interesting though, I would've expected iOS to be slower with android largely leveraging chrome because Google.
The team has been able to progressively target the different platforms where needed with native modules and TS files targeting the arch. Expo's build plugins have also saved our bacon.
We've been pretty excited for the new architecture. Our early tests show a lot of performance benefits on android, and so far the conversion process has been pretty good.
Use "npx create-expo-app@latest --template default@beta" to install new version. Start with production mode, using android device, it takes about 300ms to render 2000 Views. The same test on React web is still much faster.
Can't wait to try out the new arch when 0.76 lands in expo.
Most mobile apps (at least the ones that get you paid), though, are different, usually need good quality integration with native APIs, keyboard, etc...
The trick is, some frameworks smooth over the rough spots and throw a few sweet perks your way, while others give you a taste of both the good and the gritty.
- Easier in the long run
- You're able to ship a far more featureful application if you deal with media and/or VoIP
- Passing apples reviews seem to be faster if you ship the modules yourself in my experience
Can you expand on this? How does Expo prevent or make it harder for you to ship media- and VOIP-related features?
When react-strict-dom is totally ready for prime time it's going to be a game changer and react native will become an absolute juggernaut.
Right now react-native-web is not nearly as good as iOS and Android react-native. react-strict-dom is a massive leap forward for universal apps.
I think React Native was the go to place in the past but it has been surpassed now.
Also, I remember transferring lots of data through the bridge could be a bottleneck for some use cases. Is that effectively solved with this architecture?
Can I comprehend this as a new Virtual Native UI immutable tree running in native space?
And react native mounts and updates basically synchronally updating this immutable tree and reconciliation being done in native space, dynamically updating the app layout?
Reconciliation is still done in React on the JS thread, similar to React DOM.
How does the renderer ensure consistency in case o multiple immutable tree references?
1 person complaining about the amount of JavaScript loaded just to display this one article
One way to address this might be to compute some relevance metrics using embeddings (they’re the new hotness after all) and downrank low relevance discussions. I assume it’d be especially effective for the “ads and JavaScript” discussions.
I would simply say a lot of contrarian viewpoints, regardless of merit, are downvoted into oblivion. Once you get hammered on something you simply had a different viewpoint on? Users tend to stop commenting for fear of reprisals. Your "karma" on here is not easy to obtain so when people downvote you simply for a differing opinion, it makes it less likely they will wade into a discussion again and find themselves on the wrong side of some Hacker News diva having a bad day.
This means you have a lot of people (like myself) simply opting out of a lot of discussions because if you're not on the right side, your comment will get downvoted immediately. There is no data point on people like myself, so there's no way to tell people that the quality on HN is declining, everything is just fine and normal when in reality, you have a lot of users who aren't engaging for fear of getting downvoted into oblivion.
And hostile.
Point out that someone is wrong (even with hard evidence at hand) and people will still try to push their deluded conclusions nonetheless.
Same thing for expressing an opinion outside what the hivemind deems acceptable.
Btw, I think this phenomenon is a widespread cultural thing, not HN specific. Happens irl so much that it is now almost impossible to have an actual conversation with anybody.
Meanwhile, solitude and suicides are skyrocketing and people do not see the correlation ...
I think it's a combination of:
1. Ragebait being so useful at generating engagement that it's become a standard form of human interaction
2. The Internet making people feel safe from physical repercussions which makes people feel comfortable with treating others badly
3. Internet communities quickly becoming echo chambers where you're forced to pick a side if you want a sense of belonging
So we've been programmed and manipulated to be angry, to be tribal, and to act without fear of retaliation, and all for what? For ads and followers.
But, somehow devs _still_ expect a 100% perfect DX while maintaining the ability to publish to mind-boggling different targets such as iOS, Android, and Web
¯\_(ツ)_/¯
Congrats on being the person comparing HN to Reddit.
If this was on programming or even web dev you'd just see "react bad" or "embedding a browser for an app is blablabla" (when react native doesn't even do that)
People have believed this the whole time and also a ton of people didn't then and don't now. What has actually changed?
That said, it is absolutely up there with the best choices for mobile app dev these days.
I don't see any reason not to use it there, and the industry seems to be thinking in the same direction atm
I don't think most apps are like that, though. If it is worth having a desktop app (instead of just, you know, having a web app), than that app is probably relying heavily on native integration. Also, mobile apps and desktop apps are usually quite different (as they should), those are two completely different interfaces.
About web, I'm not sure about RN, but Flutter IMO is so terrible on web, that it's so rare that it would make sense to use it, that my default advice would be that write the web in something different, even if you use Flutter already on mobile and desktop.
1. What is the next thing that the team wants to focus on improving?
2. What are the performance differences between the old architecture & new one?
3. What are your thoughts on the fragmented state of rn wrt react-native-web/react-native-windows/react-native-macos?
4. It is quite difficult to know what supports RN vs what relies on react-dom. Is there any thought to create some ecosystem focused around RN? Or if something like that is too cumbersone, perhaps even just adding some badge to github pages for "Supports RN"?
5. I forget what it was called, but the creator of react-native-web stated that they wanted to start winding down support in favor of an alternate approach which attempts to bring web apis to native instead of trying to make the native api work on web. I.e. instantiate div elements in native instead of view. What are your thoughts on this?
6. React (and IMO Meta as a whole) seems to generally have had the tech philosophy of take what you want, leave what you dont. With the dropping of create-react-app and endorsement of frameworks like Expo, it seems like its getting harder to just take the pieces we want. Is there any thought about this trend?
7. Related: as for the upgrade process: it would be cool if there were a way to "opt-in" to auto upgrades. E.g. what if there were a package which contained a base class controlled by the RN team so that a client side upgrade could be as simple as updating the version of the library the base class is in? (customization would be simple extending the class and doing w/e else needed there)
Again, thanks for all the work!
The next thing is to continue building on this foundation and fix some long standing issues things like scroll perf and text input. A lot of our focus has been on the gradual migration strategy for the new arch, so now we'll have more capacity to work on other things.
For perf differences, we shared some benchmarks here: https://github.com/reactwg/react-native-new-architecture/dis...
But perf alone doesn't really tell the whole story. In raw perf terms, flashing empty content for just one frame is only a few milliseconds, but user is disproportionally impacted by that flicker. The new arch allows us to fix those types of issue in addition to the raw perf wins.
Is there a plan to fix this flakiness that I experience every 3 months or so?
As the core library, we need to support all the different ways React Native can be added to an app (from fully react native to adding react native to an existing app) and all the different build tools an existing app may use. So it's hard for us to be opinionated about the setup in a way that would make upgrades seamless, but a framework can solve this for you.
Use Expo because you like the extra features it ships with, but not because you have problems with native dependencies. The React Native built-in experience is pretty much perfect to start with.
The native dependencies are 100% solved by prebuilds, CNG, and config plugins.
I've only used Expo during hackathons and university to quickly get something running, but for actual production apps it's should be avoided imo.
expo has solved the native dependencies issue, and it's a fantastic way to build an app.
this is one of the few hills i'm willing to die on. expo is great for the RN community.
also you can download my RN expo app here: www.shopcats.app.
But when it comes to performance. None of them beats RN. You can make some good looking continous animation of most of them (not cordova or ionic). But when it comes to lag when moving between screens, RN on mobile comes on top without doubt.
Indeed, it is definitely the case that this kind of behavior/content gets amplified. When you log in, in some sense, any reality could be crafted just for you, for good or for bad. The overwhelming majority of people are vulnerable to this. What they see == what they think it's real, me included, btw.
I once read an article about how the vast majority of dating now begins through an online interaction (say , Tinder), and how also the vast majority of these apps are controlled by 2-3 companies. Think about the massive power they have over everyone else's lives. You want to encourage interracial relationships? Suppress matches within the same race and encourage matches outside of it. (And the opposite could be done, I'm not making a political statement here). These people have the power to completely change the demographic landscape of a country in a couple decades(!). They should be heavily regulated, but far from it, no one is even aware of this.
It will only get worse with "AI", unfortunately.
Seriously, thanks for all your (and team's) hard work, I look forward to all my packages being on the new architecture and upgrading!
Was thinking of learning Flutter or even RN, but now not so sure. On the other hand, it could mean writing it three times.
That is all too before they have even touched bringing WebGPU in so I think in a medium term scenario it will be an extremely safe bet.
Just by way of example Google Earth just did a major rewrite to put Flutter at the heart of their UI precisely because it was a better experience including on web so I offer that anecdote to suggest your info is kind of out of date.
Google Earth is almost like a video game so flutter makes perfect sense for that. If you're building a fintech CRUD app flutter makes absolutely no sense on the web.
With react native web you can share your code with next.js in a monorepo and absolutely blow flutter away in terms of page weight and time to first interaction.
Why do fake internet points matter so much? Yes, karma is difficult to attain, but it also does (almost) nothing other than indicate how active a person is here.
> This means you have a lot of people (like myself) simply opting out of a lot of discussions because if you're not on the right side, your comment will get downvoted immediately.
In short, so what? Does it really matter if one thing you said one time gets downvoted? We're all (supposed to be) adults here. All that happens is fewer people read your comment. Big whoop.
To use your own extensions you have to eject/or start a new application with their own native libraries. For most of what expo does you can use the unimodules libraries, they're quite good in my experience.
Unless I'm misunderstanding what you're saying, this isn't true at all. At $day_job we have an Expo app with a custom native library and it works just fine; you just have to write an Expo-specific adapter for it and can't use Expo Go in that case.
But another commentator wrote "yet another SaaS with questionable payment tiers" about Expo (https://news.ycombinator.com/item?id=41937886) which makes me already not want to use it if I ever touch React Native again. Not sure why a SaaS would be involved at all in this process.
The anti expo sentiment is almost all based on the old days where you had to eject to use native dependencies. Expo is amazing now.
These folks should be working together on just a few open-source projects :-D. Trying to support every mainstream platform is a huge undertaking and as mentioned in this thread, few succeed, and even fewer over the long term.
By opting for Kotlin and Compose UI, in addition to being a first-class citizen on Android, I gain the ability to share code with many other platforms to any degree I need. That is, just the app logic can be done in Kotlin while keeping the actual UI in whatever each platform prefers (UIKit on Apple platforms, Swing on JVM, etc.), or it can be entirely done with Compose UI (which usually means being rendered via something like Skia). And even if you adopt Compose UI, there are different degrees to which that can be applied; Compose UI can output either DOM elements or render to a canvas for the Web target, for example. (It's actually curious you mentioned Qt. As a day-to-day KDE user, I'm interested in playing around with the possibility of wrapping Qt components with Kotlin/Compose as well.)
No matter what degree of that I choose, it's still going to be more "native" than something like RN. Depending on the target platform, Kotlin will compile down to JVM/Dalvik bytecode, LLVM bitcode/machine code, JavaScript, WASM, etc. RN (last I checked, at least) still relies on running in a JavaScript engine at the end of the day. Flutter will get you closer than RN, but it's still not quite native to any one platform in particular and you're still introducing Dart as another language you need to know. A .NET-based solution suffers a similar problem, and IMO it feels even more out of place on Linux than the rest.
(To be clear, Compose UI has been able to learn a lot from Flutter, as they both originated at Google and both take a declarative UI approach; I largely consider Compose UI to be Flutter's successor.)
Ultimately, it can boil down to this: if you want to support multiple platforms, you need expertise in each. Adding most traditional cross-platform frameworks such as Flutter or RN also means you need expertise in that framework as well. In the worst case, my expertise with Compose UI will remain good for the work I need to do on Android. The worst case for most of these other frameworks is that it becomes irrelevant and you need to learn something else anyway.
Did much of the JPC team get let go like the Flutter team in the last big round of layoffs?
The problem at the time was multifaceted. The devices couldn't handle it, the network couldn't handle it, and webkit was still relatively limited. But it did work. https://9to5mac.com/2021/06/03/remembering-apples-sweet-solu...
Performance isn’t incredible, but neither is the “native” app (95% sure it’s react native or similar), which frequently dips below 60fps.
MacOS is getting there kind of lately with safari being able to turn web pages into apps. But there isn't really built in mechanisms for that: https://www.macrumors.com/2023/06/14/how-web-apps-work-macos...
The part about being slower is still true though.
People just don’t know it’s an option. And as you say, there are fewer native APIs available to it.