Electrobun: Cross-platform desktop applications written in TypeScript
electrobun.dev323 points by CharlesW 7 months ago
323 points by CharlesW 7 months ago
Hey hn! Author here:
Thanks to whoever submitted.
Electrobun is in the very early stages. The roadmap is a bit out of date. Here are some highlights:
- it’s like Tauri in that it uses the system webview and does not bundle chromium
- it’s like Electron in that it bundles Bun instead of Node for the main process so you just write typescript for both the main and browser contexts
- native bindings written in c/objc with a zig wrapper that the bun process runs
- it will be cross platform, but working on stability first
- the cli handles updates code signing and everything else you need and outputs build artifacts that you just need to upload to S3 or any static file host.
- it has a custom optimized bsdiff implementation in zig that autogenerates updates for your app as small as 4KB
- it has a custom zig self extracting mechanism that uses zstd so your initial download is also as small as can be.
- it has a custom encrypted easy to use RPC mechanism between bun and browser processes where you just write method signatures and handlers.
- it has a custom OOPIF implementation so you can embed isolated browser contexts with your html <electrobun-webview> element that are performant and easy to use so you could build a multi tab web browser with it if you wanted.
> Thanks to whoever submitted.
That would be https://news.ycombinator.com/user?id=CharlesW aka Charles Wiltgen.
Tauri also uses the system webview and does not bundle Chromium, right?
"it’s like Tauri in that it uses the system webview and does not bundle chromium"
[flagged]
> You should pick a different name.
Can we please stop telling people to change their names because something else exists with a similar name?
If we take this stance seriously, then Google would have to change their name as San Diego Colab has existed for a year before Jupyter was created, therefore existed before Google Colab.
These comments don't solve the annoyance of false assumptions on HN, because they often contain a false assumption of their own.
Google didn't have the same issue with confusion because of San Diego Colab that colab.sh will have if they keep their name and get popular. That's because San Diego Colab wasn't nearly as well known as Google Colab is now.
So I can say "Can we please stop" to your comment, and I'm sure you can find something with my comment and get in a loop of "can we please stop".
And then when this project explodes and becomes as well known as Electron is eclipsing Google Colab (I never heard of it until today) then your argument is invalid.
And you're right, we can keep getting into this loops, so lets nip it in the bud where it all begins and shocker, not tell people to change the name of their stuff because something else exists with a similar name but completely unrelated.
So this aspires to be something like Tauri, but with Zig for the fast/native bits, and leaning harder on and more opinionated about the frontend/UI part?
That's a pretty interesting proposal, but also a staggeringly huge amount of work.
It’s only been a few months but yes it’s a huge undertaking.
I’m only picking up steam though and my whole career has been taking on these kinds of huge projects so it’s just getting started.
Awesome. :) Good luck.
I think it's a really attractive package for an early-stage one-person project. Nice docs, and cute bunny.
FWIW, it also seems super-sensible to focus on Mac only for now, and get the basics working well there — a good Mac DX could provide the traction/momentum to get more contributors to help build out the obvious cross-platform promise.
> it also seems super-sensible to focus on Mac only for now
I’m the first guy to agree to reduced scope, but unfortunately the reality of cross platform is that the subtle details affect the APIs, and once you add more platforms, you can realize issues too late. As a simple example, MacOS has a concept of “showing/activating the app” whereas Windows has… well individual windows. They add up, especially if you add mobile to the mix.
I think the most sensible is to keep the api surface small, and also to peek at electron, tauri etc to learn and try not making expensive mistakes.
Ya, I kind of hate to say it, but it feels like because of Window's backwards compatibility and variation in UI styles over the years, people expect Mac apps to be comparatively pretty consistent, and consistently good. There's some wiggle room, but usually it's some niche open source app that's just too generally useful to prioritize mac
This is awesome! I built a similar tool as an experiment while at Recurse: https://github.com/zephraph/webview. Didn’t really do any heavy lifting though, just reused some of Tauri’s crates. Does Bun run on the same process as the GUI binding? OSX steals the main thread when rendering a native window which made me lean towards separating the processes. Still wonder if there’s a better way.
I’m using bun for the main process. Bun runs a zig binary which can call objc/c methods. So the “main native application thread” is technically the zig process.
Then there’s all kinds of fancy rpc between bun and zig and between bun and browser contexts.
Ah, cool, that’s essentially what I’m doing too. Rust binary does the system interactions. I’m just using a pretty simple json RPC over stdio though.
How do they manage the OS-specific stuff with pure TypeScript? How does it compare in benchmarks against Tauri for size and speed?
I dug through it a tiny bit because I was curious as well. To me it looks like OS-specifics aren't dealt with in the project. It uses bun to run the TypeScript. From the getting started docs the project just opens a browser window for the UI. I didn't look to see if this meant the user's default browser or a bundled browser.
Tangential but I hadn't heard of Bun before and their performance claims had me interested, sadly seems you just can't beat V8 in the real world. I threw some poorly optimized art code I wrote yesterday at it and it's almost 2x as slow as when I run the same code in node.
Most of the speed ups seem to be in optimizations to the native code and porting js libraries to it. I'd expect pure javascript code to not have much of a difference. 2x sounds like you hit some limitations of JavaScriptCore itself.
So is it a yet another webview-based framework like NeutralinoJS (https://neutralino.js.org), Electrino (https://github.com/pojala/electrino) or DeskGap (https://deskgap.com)? What's their advantage apart from using Bun instead of Node?
For relly lightweight cross-platform desktop apps better use a non-webview-based native framework like Qt, GTK, wxWidgets or even recently released FLTK 1.4.
Modern apps these days often end up needing to render complex layouts and rich text, so you end up needing QWebview anyway. At least that's been my experience creating desktop apps lately.
You can get pretty much any layout using Qt/QML or Gtk.Builder, no need for bloated webview.
> Modern apps these days often end up needing to render complex layouts and rich text
/end up needing/choosing/
Couldn't agree more, QML with C++ for the logic (or Rust or whatever other bindings you want to use), is the best imo.
If you want to use Rust, Slint[0] may be something to consider, seeing as the existing bindings for Qt and QML are not very pleasing to use (mostly talking about qmetaobject-rs and cxx-qt). If I'm not mistaken, Slint is developed by ex-QML people.
[0]: https://slint.dev
One of my favorite things about Tauri is that I can write all the "back end" code in Rust. I think this has a lot of advantages for performance and correctness. For that reason alone I'm more inclined to reach for it than Electron or something like this when I have options.
I love to see projects like this, the more alternatives we have for creating cross-platform apps, the better.
As an alternative to electron, using bun as a base sounds nice. But being better than electron is a low bar when it's the source of the laziest, most bloated programs on my system.
Also, still waiting for bun to work on freebsd. Patiently! But it's on my xmas wishlist :)
Basically, this is an Electron that replaces Node.js with Bun and Chromium with WebView?
It seems it bundles Chromium
It doesn't yet.
"Use the built-in system Webview (or bundle a 3rd party webview like Chromium: coming soon...)"
https://electrobun.dev/docs/guides/Getting%20Started/What%20...
Yoav, you are a genius. Keep up the good work.
What rendering engine does it use? Project name suggests Electron, but they never mention it
> The current Electrobun Playground app is 50.4MB in size (most of this is the bun runtime)
Seems to be more than just bun runtime.
Electron is not a rendering engine. It’s an application bundler, which itself bundles Chromium and NodeJS.
This project uses native web views, like Tauri. They wrote that they might provide the option of bundling your own engine, ie like Electron, which I personally think it’s a bad idea. Tauri proved that you don’t need it.
But now that you mention it, agreed that 50MB is a lot.. maybe just standard JS dep bloat? That could be clarified.
> Tauri proved that you don’t need it
Are there any major cross-platform apps based on it?
Last I checked they did a pretty bad job showcasing the apps. But search for “awesome-tauri”, it’s pretty standard these days. Way past the initial hipster phase, almost boring.
A bit of self promo: you can also check out https://payload.app/ which is the project I’ve been working on for way too long now.
>Security and Performance with isolation between the main and webview processes
That's one of the performance characteristics I'm afraid will hinder certain applications.
It sounds like you need to use a IPC bridge to share data between the main process and renderor. Which means copying all the shared data. Like if I wanted to use ffmpeg for decoding video then each frame I'm waiting for a the decoded image to be copied before rendering it.
Originally I had it going via browser -> postmessage -> zig -> bun via named pipes and then bun -> zig via named pipe -> js via evaluate js.
I’m building https://colab.sh/ with electrobun and that rpc was pretty slow when trying to load multi MB files in the code editor.
Last week I added an encrypted socket rpc option directly between bun and browser.
Loading a 2MB file went from a few seconds to 400ms.
I made it so that in contexts where CORS allows it automatically upgrades to the socket based faster RPC.
That’s via RPC though. electrobun also exposes a custom views:// scheme in the browser context that can load files directly from specific sandboxed places in the file system. More to improve there but for a very big file you’d be better off writing it to the file system that would be much faster.
Sure but "fast communication" is always going to be prohibitively slow if you are used to "reading the same chunk of memory".
I don't really think that's an issue for the kinds of apps people use this sort of thing for and at first blush the IPC model here is at least better than Electron which is an absolute trash fire.
Eh, it’ll hinder certain applications but very few, really. Assuming the webview is intended for UI and very little else.
Interesting to watch it grow. I'm not going to jump in right away, but this is a great project!
Never understood why these things aren't done by running a local http server and use a regular browser.
Developing for Electron you usually had a regular webpack server in the background to handle hot-reload cases and technically you could have a regular browser.
In production however you don't want a browser for 2 reasons, first should a local app really expose internals to the network(And get mucked up by firewalls)? secondly the deployment is easier the more it's self-contained.
Also since there are internal communication channels between "browser" and native/"server" parts (that are far faster than going over the network) you don't want to diverge the production and development environment to avoid having hard to debug things.
On top of it, in this case bun is also a bundler so you get the typescript transpilation for free.
So to summarize, if your app is really only a plain webapp distributed to desktop then sure you can developer everything with a local webserver and package with whatever is available, but as soon as you involve native off-browser parts you don't want to start exposing everything and using the embedding systems(be it electron or webview2/khtml,etc) built in browser<->native communication is simply the saner choice and the point of these projects is to abstract that from the developer.
Local file system access is one reason, any other native APIs needed for making a desktop app, etc
Any plan for supporting native toolkit access ala react-native, or is the plan to just rely on web tech?
You can use my Sciter if you need native widgets.
Sciter is an HTML/CSS/JS engine with the ability to attach native widgets to DOM elements. So you'll benefit from two worlds: HTML/CSS layout richness and native widgets.
The major problem though is Linux family of OSes in this respect. I have version of Sciter that works directly on X11 or Wayland ( https://sciter.com/sciter-glx-generation-4/ ). Windows and MacOS have clear concept of child windows/widgets but no such things on X11 or Wayland realistically speaking.
So react-native is practically iOS/Android only thing.
FWIW, I was pretty impressed with react-native for mac os: https://github.com/microsoft/react-native-macos
Granted, it doesn't have the same support or attention or community that react-native for iOS does. But it was certainly a refreshing way to build a native app. I think some of the shine has come off it now that a lot of the ideas have been adopted into SwiftUI.
Which webview does it use?
Any similar platforms using Golang?
Very promising project!
Why do so many projects fail to include any screenshots at all?
Maybe the authors don't think it's necessary, but step 3 under `Getting Started` is called `Creating UI`. Why would they not show what the result of the tutorial looks like?
Not only are screenshots often missing, but I’m also surprised by how many projects fail to describe the purpose and goals of their project in short, simple sentences. Many mistakenly assume that only users and developers visit the website or repository, not decision-makers. More ELI5 ("Explain Like I'm 5") can never hurt.
Writing good documentation is extremely hard. Even 10x engeneers can be basically illiterate.
> Even 10x engeneers can be basically illiterate.
I see what you did there
There's certainly a persistent issue of typos: (34 here that we found: https://triplechecker.com/s/689196/electrobun.dev).
Screenshots are harder to produce, maintain and publish.
I also suspect many technical authors are less “visual” so to speak, putting less emphasis on visual presentation.
> Screenshots are harder to produce, maintain and publish
What, really? What am I missing? If you tell me video, yes, but a screenshot...
I don't think it's that hard, but it's easier for screenshots to drift and become out of date compared to just written text documentation.
You’re just trying to give us the gist of things with images. It’s basically selling your thing to me as something I would want to try out. Usually if I connect with something I see in first few moments then my likelihood of trying the thing out is higher.
We don’t need it to be updated on every commit. Or illustrated docs but some idea of what I’m getting myself into is nice to have.
It is literally a few minutes of effort
If you cant do such basic thing, which would significantly improve an ability to understand your product, then wtf
i mean in this case it would be a window with any web app in it… not sure that would be particularly helpful.
"...and cross-platform desktop applications" the homepage says
But, Sadly, this is mac only[0] no windows or linux ? Do I understand it correctly ?
Looks like they're planning to only ever support building on Mac, but they plan to support running on Windows and Linux in the future
Wild that one would pick the only platform with a rubbish VMs and CI story as the OS to use for building software!
What's wild about it? It's one person, and they probably mainly use a Mac. If they used Windows, it would be Windows-only at this stage. I have lots of projects that only run on Linux. Clearly, if the project succeeds and gains tractions, the other platforms will come.
It assume it will severely limit its uptake.
If I had to pick one platform to build on, it would be linux. Pretty much every platform has the ability to run linux executables (via Docker, among others).
If this is a personal project, that's cool, but then I really don't care that much. On the other hand, I think an electron-like with something like Bun could be really useful.
At least here in the US, the majority of developers daily drive macs on their desktop. Even (especially!) the ones that deploy exclusively on linux.
Having to run a docker container to get the dev env setup would be even more likely to slow adoption. Nobody wants to do that shit.
> At least here in the US, the majority of developers daily drive macs on their desktop.
Not according to statista.com (I am not a registered user, but it looks like you could register for no cost, in order to check the source):
https://www.statista.com/statistics/869211/worldwide-softwar...
Windows is the majority OS for development, with Unix/Linux and Mac being close to equal, but both considerably less than Windows.
This is also consistent with the 2023 Stack Overflow Developer Survey (https://survey.stackoverflow.co/2023/):
Under the "Operating System" heading it says:
"Windows is the most popular operating system for developers, across both personal and professional use."
They said "in the US". The data you have provided says "Worldwide" which doesnt prove the point about the US.
fair point, however I suspect that if you narrow the scope of the data to the US (even the data from those surveys) it will not change the results much.
Out of curiosity, I'm looking for data that is limited to the US, and will respond with results if I find them.