> I want to (1) build blindingly fast, low-latency, super performant UX for users, … (2) maintain developer velocity, (3) have access to nice UX primitives and widgets … (4) have it look nice and modern.

Find me when you find this, because I actually think it is impossible.

I think there is fundamentally too much inherent complexity to abstract away to keep 2 and not sacrifice 1. Specifically for something properly cross platform.

If you are only targeting MacOS or windows then I think it’s absolutely possible but then you are using the wrong language, nothing against rust at all on that, the platform defaults are just swift / C# for those use cases.

And I’m not sure but unless you are just using a gtk/Qt style framework you would absolutely run into a11y problems that you would need to build around yourself.

Sounds like you probably want egui though… if your primary UI is a big canvas that needs to be hardware accelerated and interaction is just to assist that use case egui is probably a good bet for you. But you wouldn’t be hiring web devs to do that.

I believe it’s possible, but it’s going to take a little bit of outside of the box thinking in that it’d be most practical if the UI toolkit isn’t bound strictly to a single paradigm.

Both imperative and declarative UIs have serious problems. Imperative toolkits can get hairy with their boilerplate and can make staying in sync with data state a real challenge, while declarative toolkits have a strong tendency towards rigidity that makes for awkward DX and ugly code in all but the simplest todo app kind of examples and don’t lend themselves to fine grained control.

I think there’s a happy medium to be found between the two in a well-designed hybrid. This hypothetical framework would allow full declarative in situations where that’s highly suitable, but would also allow the dev to do widget configuration in a more traditional imperative style or if necessary fall back to full imperative. Support for reactivity on both sides of the coin would be robust so staying in lockstep with data is simplified.

This would bring down boilerplate substantially since the broader layout of most screens could be done declaratively, but it wouldn’t come at the cost of more advanced functionality, flexibility, and developer control on the individual widget level.

I’d like to dig further into these ideas at some point if I get time.

I’ve been thinking about this myself and really UI development is mostly about design up front, many times with a well built design system UI changes can be trivial.

But as you alluded to, data binding can be difficult. For some things, and even most web pages fall into this, you effectively load the data up front render something and then don’t touch it again, on change you can absolutely just rerender everything because it is truly trivial to render.

But for other things you need extremely complex data binding with constraint systems and things are extremely complex to render (CAD, Games, etc). For that you need cacheing and partial updates etc

These is a spectrum between that and I just don’t see the ability to abstract over it.

And taking a11y into consideration no matter how you choose to render you also need to construct a tree or graph for screen readers, this also needs data binding. You will probably also be tying keyboard navigation to that graph (other than for a game but even then maybe)

I feel like though a lot of performance problems on web is 1) people not understanding fundamentally how expensive certain operations are 2) lack of hardware acceleration for some vector operations in CEF / browsers

For 1 just think about some UI element, a rounded button or something. A mesh had to be dynamically generated for that UI element and then drawn, if you for some reason are causing that mesh to be regenerated every frame that becomes expensive very quickly.

Text is generally generated on the fly and cached as well, if its not software rendered.

Once you get down to the nuts and bolts modern UI is a complex problem, and to build a framework for that, you need to get into the nuts and bolts.

We’re not even discussing rendering APIs yet…

This is part of why I believe that smart hybridization is the key. In a complex application, you need qualities from all major schools of UI framework design, but broadly speaking we’ve been obsessed with trying to find a “silver bullet” one true paradigm that fits every use case, when no such thing exists.

I also think that UI frameworks try to hard too be “magic” and unopinionated where they should instead be highly communicative and naturally guide developers into the happy path. So following your example of the excessively re-rendered button, the framework should really be detecting the unneeded redraws and complaining to the dev about them and potentially even presenting as compiler errors in production builds until they’re fixed (which the compiler offers hints on how to do).

I'm no UX expert, but I regularly try out new (and old) toolkits to understand the problem space.

It really sounds like you want an immediate mode toolkit. Retained mode will never be "super-snappy", there's an entire sandwich between your code and the pixels. Look at Blender or Reaper, this is the kind of "feel" you'd be getting.

If you want retained mode + "true" native widgets on all platforms, investigate: Toga (Python), WxWidgets (C++), and Tk (Tcl). The native toolkits are often best in class on each platform, and these wrappers are about as thin as they can reasonably get (e.g. Toga uses pyobjc). Integration with Rust is left as an exercise to the reader ;)

A rich widget library is nice, but consider the depth as well. Egui went to great lengths to integrate assistive technologies, which depending on your target audience may be impactful. (Also: accessibility is for everyone. https://shortcat.app/)

If you want easy hiring, you have to go with mainstream. We've +/- named all of the options by now. Otherwise, hire a talent who's worked on the next closest thing to what you're building, and trust them to decide the direction.

Looks and beauty are in the eye of the beholder. There are many apps that have a distinct, sometimes quirky, but appreciable style. Reaper looks out of place on every platform, but I prefer it over Logic or Ableton.

I'm super biased in favor of iced.

Yes, there are missing features (it's not even version 1.0 yet!) but I think the number is very small now and the solution is usually to fill in the gaps yourself—which is possible because iced is totally modular

I've made a spreadsheet editor and a slideshow editor with it so "slimmed down photoshop" seems feasible although admittedly you will likely need to get deep into the renderer (possibly write your own renderer traits and there), but I suppose you're comfortable with that given your goal.

If you're not allergic to Discord, come join us. It's a helpful, awesome tight community IMHO

Link for the lazy: https://discord.gg/3xZJ65GAhd

Project link: https://iced.rs

No support for RTL or CJK text, IMEs, or accessibility tools leaves it pretty firmly in the "toy" category for me for now, sadly.

This isn't meant to bash on the project specifically, basically the only tools that meet that criteria are platform native UIs, the web stack (including Electron), Qt and Flutter.

AccessKit is promising to change that but I haven't checked in a while how it's going. I did see GTK looking into it as a cross platform option as GTK's accessibility support only works on the Linux desktop currently

> No support for RTL or CJK text, IMEs, or accessibility tools leaves it pretty firmly in the "toy" category for me for now, sadly.

I18n, L10n and A11y tend to be very challenging features, making your wishlist a tall order. Solving that will require a larger developer mindshare. (I'm not sure how much of the existing frameworks can be reused here).

> basically the only tools that meet that criteria are platform native UIs ...

Here's the good news! Iced is already a platform native UI toolkit. It powers the Cosmic desktop and a few important applications are based on it. Iced is sure to receive a lot of attention once Cosmic stabilizes. Hopefully, you will get your wish sooner than later.

In that case, I'm happy to let you know your info is outdated. CJK/IME support has landed.

https://github.com/iced-rs/iced/pull/2777

すごい！Japanese is one of our localization targets, so this is amazing!

I'm building a graphically heavy app, so two more questions:

1) Is iced appropriate for building a canvas-type drawing app with image inclusions? Right now we're using Konva.js and we'd be losing a tremendous amount of out of the box flexibility in porting to iced, but the performance wins would be a major selling point.

2) Can you include Bevy panes? We're also doing some 3D visualization, and we currently use Three.js. But we have a lot of Bevy experience too.

Awesome! Glad to hear that and I hope you do give it another try.

1. Yes but not necessarily a "resounding yes" because images right now are drawn on the top of each layer, so if you want to write on top of images you'll need to create different layers for each of them which is more expensive than not doing that. You may be OK with that for now (I am, for my slideshow editor), but there's pending work to be done there to remove the requirement of explicit layering and thus improve performance.

2. I'm not familiar enough with Bevy to opine. I know there's https://github.com/tasgon/bevy_iced but I think that might do the opposite of what you want. Having said that, iced lets you write custom shaders so you can go as deep in the stack as you'd like for any 3D visualization. The caveat then being that you're implementing all that logic yourself, but there's no stopping you.

Importantly, I think it's worth joining the Discord and asking around, seeing what people are building, trying the library, and pointing out where you think you need help. This helps the core team understand use cases better and in some ways guides development IMHO. I'm not part of the core team, so I could be wrong, but that's my perception.

Is it possible to force iced to use vulkan exclusively?

Yes, if you write a custom vulkan Renderer and use it instead of the provided wgpu one

iced is generic over the Renderer type

I know it’s usually taken as a given around here that Electron is slow, and many of the big-name apps using it are cited as examples with good reason.

From working on a Tauri app myself for a few years (video editor) I just think the blame is misattributed. These things are not inherently slow. Slower than native? Maybe, probably, at the level of milliseconds. Visibly laggy? No, that’s the badly-written UI code’s fault. (see also: the latest iterations of the macOS System Settings UI, where the search box lags like crazy)

A webview can be extremely responsive. It won’t be if you treat it like a web page (where clicking buttons fires off HTTP requests) or if you let the JS framework code get out of hand, but those are not the fault of the wrapper.

If you like building with HTML/CSS/JS then I’d recommend doing some perf experiments to see how far these tools can take you. Of course if you don’t want to use that stack then pick something else :)

If you’re building photoshop, the main UI will probably be canvas anyway, where drawing is fully under your control, no matter which framework you go with. That stuff can be very fast or very not-fast depending on how the code is written.

Slint recently added Bevy support. I’ve been keeping an eye on it since I’ve used Qt and love working in Qml.

Slint cites Javascript. Is it another Electron/Tauri-like?

Slint does not use a browser. Instead, it has its own runtime written in rust and uses a custom DSL to describe the UI.

It has API for different programming language.

For Javascript, it uses node or deno for the application logic, and then spawn the UI with its own runtime without the browser.

In a way it is the opposite which took the JS runtime out of electron to replace it with a Rust API, while Slint keeps the JS runtime but swaps the browser for its own runtime (for the JS dev point of view)

No, it's like Qt and QML.

What?! No. It's Rust GUI toolkit. I swear people see JS support and think Electron.

Makepad is neither gtk generation nor browser based. Might check the "just text/button widget" box though, I'd certainly place it on the minimalistic end of the spectrum. (haven't worked with makepad, just enjoyed the demo)

Makepad [0] is indeed quite impressive. You can directly try Makepad Studio in your browser [1]. Don't know if that's the latest or the demo version. Their demo vids are a good watch.

[0] https://makepad.nl

[1] https://makepad.dev

> Is Dioxus (or Leptos) much more performant than Tauri/Electron?

For now it's largely the same. Both Dioxus and Leptos render using Tauri (or a web browser). For Dioxus, a Blitz-based renderer (Dioxus Native ) is in development which will change the story slightly.

I would suggest Iced if you're looking for efficient (I don't think it's any less featureful than any other Rust-based GUI). With honourable mentions for Slint and Vizia.

Certainly not a "Rust UX _expert_", but I do find GPUI interesting. Some nice examples here of not-"just text/button widget programs": https://github.com/longbridge/gpui-component

Check out https://www.gpui.rs/ too

Does it use Vulkan exclusively?