[zengid]

Tangent but I'm holding out from learning Swift until I see how the "Cross platform UI" wars shake out. We've got at least Kotlin, Dart, and JavaScript all battling it out (via Kotlin Multiplatform, Flutter, React Native) to be the king of UI languages across all platforms. I really like Swift though (from the little I've learned), so it's too bad it doesn't really have a cross platform UI story yet.

(I could learn all four but I have kids, so it'll probably just be React Native for me.)

[MobileVet]

I've worked in mobile since 2004 (not a typo, '04). The 'one platform to rule them all' has been just around the corner the entire time. As was already said, don't hold your breath.

All of that said, if you are time constrained and really want to look at a single thing, I would go with Flutter. It impressed me when I cranked out a side project with it a year or so ago. Hopefully they continue to put effort towards it.

[rapsey]

From our experience Flutter apps are easy to start with but quite hard to maintain. The core system keeps changing so much you are constantly having to fix things and dealing with broken dependencies. Would never use it for anything important.

[CrimsonRain]

Our experience is opposite. We started with GetX because we were new. Then moved to Riverpod/Go-Router as we gained experience. We have a set-up of four modules (shared lib; shared ui; app1 and app2) for two apps targeting web+ios+android; about 50k LoC @ 160 line length.

Obviously there were a lot of friction when moving out of GetX, but upgrading things have always been easy. Usually at most one day worth of work (worst case). We don't upgrade as soon as a new flutter version is released. We give it a month at least and then upgrade.

During upgrade, most of the problems usually come from 3P dependencies which interact with native such as package info plus. Also, we don't add dependencies unless we really need to. I think we are at ~30 total between all modules.

On the other hand, our react projects are forever in dependency hell. Much larger project, yes but way too many dependencies and too many things break during upgrade.

[bartvk]

> Also, we don't add dependencies unless we really need to

I think this is the trick. I've worked for a client whose coding standards basically forbade 3rd party clients. Turns out, it often is not needed at all. Do you need to include a library for a bar chart? No, a bar chart is often just a bunch of rectangles and labels.

When you have stupid-simple project setups, you can have huge projects but they'll compile fast and upgrade quickly.

[dan1234]

I have the same problem with react native.

[com2kid]

I stopped supporting an open source React Native library because keeping the sample code working in Expo was taking too much of my time.

This was years ago, maybe things have calmed down since then, but Expo always felt like hanging onto the horns of a bull, so long as you didn't fall off it was a fun ride!

[timmeta]

That's similar to my experience. Not just React but across JS in general. Was super hard to keep everything at pace when it was and is moving so fast. I slightly envy the old COBOL people sometimes...

[kitsunesoba]

In my experience the issue with Flutter apps is though they aren't too bad on Android, they have an odd off-putting feel to them on iOS, more than is typically visible with competing cross-platform UI frameworks.

[martinald]

It's because unlike a lot of frameworks like React Native and Xamarin, it doesn't use native UI components under the hood (unless this has changed recently). It's a bit like using a canvas element in HTML and drawing your own toolkit, or old Java GUI apps.

[__m]

Especially when you need a date picker that doesn't scream Android

[lockhouse]

The biggest problem with Flutter is Dart. Not the language specifically (it’s a nice enough flavor of C-family language), but the ecosystem. Even Swift is better in this regard. Flutter ported to TypeScript and sufficiently distanced from Google would actually be very compelling.

[danielvaughn]

Yep, I just commented the same thing before I read this. I'd say it's both Dart and Material Design. Dart doesn't seem bad (I've only glanced), it's just not very compelling or interesting.

[madeofpalk]

More specifically, there are a bunch of efforts, but the Swift ecosystem doesn't seem too interested in this.

Even Apple can barely get it with SwiftUI being a dumpster fire on macOS.

[Vt71fcAqt7]

I have to agree. Also I feel like those complaining about dart just have a bad taste in their mouth from dart 1.0. Dart has named parameters (so better than java) and null saftey. It's a pretty standard GCed OOP language. I also don't understand when people complain about the dart ecosystem. Seems fine to me.

[mikhael28]

Dart and Flutter are totally fine, and great in production. I really don't understand the hate.

[M4v3R]

Flutter is mostly fine. Until you need to implement text input in your app. And it turns out that autocomplete doesn’t work properly on iOS for non-English phones and you discover there’s a bug filled on their tracker for 5 years already and it’s still open [1]. Or you encounter another similar issue that is niche enough that the developers ignore it completely but is frustrating for your users and there’s nothing you can do.

Don’t get me wrong, I would really love Flutter to be THE solution to the cross platform mobile app problem, but sadly it only gets you 95% there and if you really care about the last 5% you’re out of luck.

[1] https://github.com/flutter/flutter/issues/12920

[jononomo]

Isn't getting 95% of the way there better than getting 0% of the way there? What are you going to do instead?

[waboremo]

Your equation isn't complete. You can't only care about getting 95% of the way there if the final 5% takes substantially more time than it normally would or worse halts development entirely.

[1123581321]

You can get 100% of the way there with Kotlin and Swift for UIs and a multiplatform library. Or web.

[M4v3R]

It depends. For some use cases it’s definitely enough and if you’re on tight budget it might get you far enough. But eventually it might fall short and you’ll have to either fight the framework or switch to a native solution.

I believe it’s good for knocking out an MVP if the mobile app is not critical to your business. I would not choose it otherwise though.

[cush]

I'm still holding out for Xamarin

[afavour]

I’ve been dabbling in this space for a long time and I’ve become convinced there will be no good cross-platform mobile UI story. The UIs are too different for there to be a one size fits all solution.

(as an aside I really like Swift too and have been enjoying noodling around with it as a server language. Probably rarely a wise choice given that it isn’t their focus but there is some attention given: https://swift.org/server/)

[waboremo]

Seconding this. The cross platform experience has been so abysmal for so long, that I question whether the benefits of "easier to support" are even real considering the generally worse experience, longstanding bugs that are harder to nail down, and the added time of having to switch mental models between cross platform and the native code you're going to have to eventually write.

I'm at the point now where native apps are still back in heavy consideration, as it's far more straight forward to cascade launch your products on different platforms than it is to half-ass everywhere.

[criddell]

> there will be no good cross-platform mobile UI story

Generally (IMHO) there are no good cross-platform stories, mobile or otherwise with the notable exception of some game engines. The closest you can get is choosing the web browser as your platform and writing a web app (Electron).

If you want to make the best app you can, use the native toolkit directly or via a framework for your UI.

[koito17]

For web apps on desktop you're not even restricted to Electron anymore. Thanks to Tauri, I feel a lot more comfortable shipping web apps beyond the browser (and React Native) since I don't have to pay the cost of bundling Chromium and Node.js. The Rust ecosystem is large enough that any backend-related task I would use Node.js for has an equivalent in Rust.

[pjmlp]

We were never restricted to Electron.

There are browsers already installed, and MSHTML apps and XUL predate it for a decade.

[Delphiki]

> Probably rarely a wise choice

I think you'd be surprised! There are some big names using it at scale[1].

[1] https://youtu.be/Te0aCoenCMg?t=710

[wiseowise]

There’s a good cross platform mobile UI story, it’s called browser.

[afavour]

I believe in the superiority of the web in general, but absolutely not in UI.

[wiseowise]

Why?

[afavour]

It has an extreme lack of UI primitives. Look how seamlessly iOS native views transition to each other, the consistent navigation bar, the modal views, etc etc. There’s no equivalent of UITableView letting you render long lists efficiently.

Any equivalent of the above has to be hacked together using JS, which ties up the main thread and can never fully match the host OS.

[aflukasz]

Exactly.

At this point, the most straightforward technical road to cross platform UI seems to be building high level UI toolkit that runs on browser runtime (ONE specific browser run time), and "platforms" slowly integrating this runtime as their native and only behavior. Note that I say "most straightforward", not "easy".

Personally, I don't see any viable *business* road for this, but happy to read some fan fiction on that.

[danielvaughn]

Yep. The only real option is that each platform agrees to abide by a common protocol for describing GUIs. I don't see that happening.

[snyp]

I’ve been writing mobile apps since 2013. Cross Platform UI is a myth. Just use a framework and language you are familiar or comfortable with. I prefer native so I just stick to that. You are sure to be most productive in languages you are familiar with rather than any new/old cross platform UI framework. Just be cognizant of the fact that all of them have trade offs.

[nwienert]

A lot of people saying it doesn’t exist or work, but as the creator of Tamagui, I have to vouch for it. I think it works pretty amazingly, and we have some really beautiful apps in every App Store (plus web).

[rudedogg]

Sorry but what you say just isn't possible. Especially with a browser-based cross-platform solution, you're gonna pay on the performance and native integration front.

What you built looks nice and I'm sure it works well for many classes of apps though.

[nwienert]

React Native renders to native views on iOS, Android, Windows and Mac!

[waboremo]

It certainly exists/works, but is it productive? With my time using Tamagui I would go with no.

There are just so many traps everywhere. 99% of it isn't your fault, just the name of the fragile RN game. Install another package here to handle one use case on one platform, avoid this component on one platform because RN doesn't support this thing, entire blocks of very simple styles commented out due to performance cost. List goes on. It's a never-ending game of your codebase feeling horribly fragile, and that's if you remain within the RN = Android and iOS realm. As soon as you start trying to get it to be truly cross platform with desktop apps - endless nightmares.

I so desperately want cross-platform development to feel as smooth as people proclaim, but it's really like playing russian roulette with every new component added. At the end, you question if it was really worth it due to how much you've had to rewrite anyways. You become sort of delirious after working on a project that uses such solutions, wondering if you traded user experience for developer experience and wound up messing both up in the long run.

Beautiful site/docs though.

[nwienert]

Good to hear about your experience, and we’ve steadily improved the onboarding costs. Also have a new starter kit coming soon that goes all in on batteries included.

When did you last try it? Because it’s improved dramatically since January.

But of course you have to compare it to the alternatives.

Aka have you tried to build a SwiftUI app? There’s endless problems. And then you have to rewrite your whole app three times.

Building any app will leave you delirious, but I’ve found myself far more productive in React Native than the alternatives - and I’ve tried them. I would happily challenge anyone to ship a cross platform app using their framework of choice and compare on time to ship and ultimate UX. I think there’s a clear winner.

[zengid]

Hey that looks pretty cool! https://github.com/tamagui/tamagui

[marcellus23]

Do you have a link to an example iOS app built in Tamagui to check out?

[nwienert]

The beta https://beta.aidungeon.com is pretty nice, not sure if the app is public yet or still on TestFlight.

[krembanan]

> Cross Platform UI is a myth

How can you say this when many of the worlds most used apps are written in React Native? Same with Electron and desktop apps. I find it hard to believe you can be more productive writing native code when you need two entirely different codebases, compared to a single one, no matter how good of an iOS and Android developer you are.

[wk_end]

I don't know if I was using React Native "wrong" or what, but when I tried it out I found it to be a miserable experience. There's almost no usable components "built-in"...there seems to be an expectation that you cobble your UI together using one of a variety of different third-party components or component libraries, and every single one I tried was incomplete, buggy, poorly-maintained, poorly-documented...

Based on that trial: I'd much rather work on two sane codebases instead of one in React Native that's a maintenance disaster. As a bonus, since your apps will be native, they'll probably be faster, feel higher quality, integrate better with the platforms, etc.

[tentacleuno]

You were using it wrong IMO. The platform comes with a lot of built-in components (just look at the docs!) just from importing the React Native package.

[waboremo]

You do realize you still have to write a lot of native code in cross platform apps, right? It's what all the most used apps do.

This idea that individually you can be more productive on a cross platform app has no basis in reality. You're still having to concern yourself with platform specific aspects, except now you're also throwing in another layer into the mix for your shared aspects. These shared aspects tend to also not be up to user expectation most of the time, so you're having to rewrite things that comes easy for native apps, to ensure consistency and accessibility.

Such moves makes sense for Facebook. Given who they hire, what they work on. Same for Microsoft, but not sure using them is a good example considering their app experiences are universally terrible. For a lot of other places though? There has been a grand total of zero proven demonstration of increased productivity or dramatic savings. You still have to hire android/ios/windows/mac/linux devs respectfully whenever you eventually want to expand to those platforms.

[wiseowise]

> You do realize you still have to write a lot of native code in cross platform apps, right? It's what all the most used apps do.

You asked from cross platform UI framework, not for abstracted away mobile development framework.

[pjmlp]

Because when problems come up you need an iOS and Android developer to sort them for you.

[brailsafe]

It's just reps no matter how you do it.

[jb1991]

No, it won't be Reactive Native or any other. There are no cross-platform UI wars any more. There are efforts at this, all with huge compromises, but working with a native UI toolkit on any platform will always offer a superior developer experience and a superior user experience; that's just the way it has always been, still is, and I'd predict always will be.

[jeremy_wiebe]

The problem is that developer experience (and often user experience) aren’t always the highest priority. For many businesses that want to ship a mobile app, cost is by far the biggest constraint. And “writing once, shipping everywhere” pushes out native development.

We at Khan Academy have moved much of our app to React Native and overall it’s a great experience and much easier to maintain than two native apps.

Of course, the constraints and requirements vary by business and so what’s right for us may not be right for you.

[no_butterscotch]

How much of KA's web react code (I'm assuming?) can be re-used in RN land?

Or vice versa, do you use RN-Web or plan to?

[jeremy_wiebe]

Currently we only share Perseus (our exercise and article npm library). We do hope to begin work to share more code at some point but it hasn’t risen up in the priority list just yet.

[tarkin2]

I was put off my React Native and Flutter since they don't give a native experience and some native features. Good and fast at the start, but after a while they'll be lacking.

Kotlin multiplatform--with the business logic shared but UI native--seemed like the best bet, backed by an impressive stack and company. And KM has gained traction in quite a few companies, slowly moving shared logic to Kotlin, while keeping separate iOS, Android and web teams.

Yet I feel more companies value Flutter, or even React Native, because it allows companies to reduce the number of employees they need, and many companies will value that over than having a perfect native experience.

I can't see there being a winner anytime soon. Web-savvy companies will favour React Native, companies with native developers will move some things to Kotlin Multiplatform, companies looking to reduce head counts more than improve user experience will favour Flutter or React Native.

[dlivingston]

I wouldn't hold your breath. It's very explicitly tied to Apple APIs and OS conventions; I doubt it will ever be cross-platform.

If you want a cross-platform SwiftUI, QtQuick/QML is probably your best bet.

[nicoburns]

Add Rust to your list. The Rust UI ecosystem is still very immature, but it has a tendency to implement things very well once they are implemented and the potential is huge.

[zengid]

Speaking of Rust UI, do you have any favorites? I've been casually watching for a long time but haven't seen a clear winner there besides Iced or egui.

[nicoburns]

Iced is still the most capable atm, but Vizia and Freya are also worth looking at. The great thing is that the ecosystem is mostly made up of modular composable libraries, so to a large extent once one framework gains a functionality they all do.

[zengid]

That's another one I'm watching for sure!

[inopinatus]

I’ve written a few command-line utilities in Swift lately and really enjoyed working with it. The type system I find especially appealing, being an enabler far more than an impediment, and coming from writing tons of C and Ruby there’s a lot that is familiar. Thus making the current multi-platform story a real shame, it’s a massive barrier to ever choosing Swift for a significant project.

It feels like Apple’s reflexive parochialism is limiting their influence, and that is something only leadership could overturn. I doubt they feel any financial incentive to do so.

[johnthuss]

The only cross-platform UI that is likely to be feasible to use long term is HTML/CSS/JS. That comes with significant downsides in terms of display fidelity. But that is going to be the case (to some degree) with any cross-platform toolkit.

[mch82]

This seems right, at least for the next decade.

Recently, I really enjoy running containerized apps locally and accessing the GUI as a website on localhost. It feels like there’s a lot of room to build on that model.

[danielvaughn]

I really like Flutter's approach, I just wish they weren't so wedded to (a) Dart and (b) Material Design.

[redsaber]

I'm still hoping that Kotlin Multiplatform succeeds. Since kotlin is now the language recommended by Google so should be able to natively write for Android while also being able to deploy to other OS.

[chasing]

Learning one will make learning the others easier. They have conceptual differences, of course, but there's also a lot of overlap.

Regardless, the "cross platform UI wars shaking out" will mean there are three or four options with high market share that each have plusses and minuses depending on the specifics of your project. There won't be a single winner.

[imachine1980_]

for what understand (im not a mobile developer) the best of all of this is react native, works well most time the ux of expo is great and you can "port it to web" like twitter, all the people who use xamarin cry, all the comments who hear about kotlin multi platform is the ecosystem isn't there, and flutter is lots harder when thing make more complicate and google is unreliable they already kind kill flutter desktop or at leas the cut most funding, i feel react native is the only really companies beat real money in it, you have multiple app in your phone using it discord, instagram,pinterest, microsoft teams, they aren’t perfect but they aren't bad, also the comunity is huge, also i hear lots of bad things about swift so there too, the lack of docs and that.

[037]

While we were waiting for the perfect cross-platform solution, we underestimated the web standards that we've always had in front of us, and are slowly becoming the perfect, powerful common denominator.

[dwaite]

sort of. From a user perspective, the app security model is brokered (by a store) while the web security model is a direct ad-hoc relationship to whoever [you _think_] owns the domain you are on. Apps tend to need reviews for republication, while a website could change owners (through acquisition, domain expiry or other exploit) at any given time.

Modern systems sandbox both application access for security exploit reasons as well as user privacy reasons. The web lets apps talk to one another arbitrarily, embed one another in frames, and so on.

Because of this, hardware access will always be significantly more constrained than store-reviewed apps.

[lobstrosity420]

Swift had no horse in that race whatsoever. It was made to interface with Cocoa and it excels at pretty much exclusively that.

[cush]

The Browser Company is currently working on a Swift on Windows

[Mertax]

Why not include C#/MAUI on this list to make it 5?

[sirwhinesalot]

The moment it went with Catalyst on macOS and forgot Linux exists entirely it was dead on arrival. It's Xamarin Forms but worse. Well done Microsoft slow clap.

[pjmlp]

Ignoring Linux I can even understand, given the Year of Linux Desktop meme, but the Catalyst shortcut, what were they thinking.

Not even the new VS4MAC UI uses MAUI, they used their own bindings.

[criddell]

> it was dead on arrival

I understand holding out hope for year of the Linux desktop, but ignoring Linux has never really hurt anybody, has it?

[sirwhinesalot]

Well, depends. If I'm evaluating cross-platform UI libraries, Linux support is a must have, even if the market might be small.

Electron, Qt, Anything Java, Avalonia, wxWidgets, all the native webview wrappers, etc. All have Linux support.

[dangerboysteve]

Tauri looks great (Rust, webkit & js)

[sirwhinesalot]

Oh yes, it is my preferred solution these days. Web crap is annoying as heck to work with though...

[zengid]

I'm a C# dev but I don't really have faith in Microsoft sticking with any UI library longer than 5 years. They even seem to have embraced React Native to make their own products. I'd go with Avalonia if I picked a C# UI, but I prefer Typescript and JSX over C# and XAML.

[Dennip]

From my experience with Xamarin i dont have high hopes. Every time i needed to some do something slightly interesting I had to write wrappers for both the iOS and Android API's to get anywhere.

[egberts1]

Let us not ignore Electron GUi kit.