Why am I writing a JavaScript toolchain in Zig?
injuly.in
injuly.in
For every 1000 JS developers and every 1000 Zig developers there are 25 that know both. Only those developers can modify the system.
Further, you double the dependencies. JS pulls in library to do X that requires 50 dependencies, Zig pulls in a library to do X that requires 50 dependencies.
You also have 2 toolchains to maintain.
And you have 2 dev environments to maintain. Get all your editors/IDEs to understand both language. Get double your checkers, double your testing, etc etc.
You effectively increase the tech debt exponentially.
The fact uv is not in python means it's immune to the many bootstrapping problems the language suffers, which in itself is a huge win.
Granted, it works only because it's well engineered. The python integration is fluid, it respects standards, you never actually feels you are dealing with a separate tech.
This is rare.
Usually the experience is more something like asdf which gets 70% there then left you hanging.
However, when that works, it brings many benefits. An unconspiscious one for uv is that we now have a packaging tool compatible with pip that you can call from any other language with no need for a python runtime, and no risk of breaking if other runtimes exist and move.
The thing about the tooling is often that most users do not have the ability to contribute even if it is in a language they are proficient in.
Oh and until there is a good enough AOT compiler for JS there will always be a lot of tooling that is not in JS.
When step 1 of your tool is "install ruby/python/nodejs/JVM"[2] it means you do not care about your users. Either bundle[3] the runtime with your tool or make it in a statically linked binary.
Like why am I, as a go/nodejs developer, need to figure out how to install python2 and pip and make them work just to have the privilege of paying you money to host my code.
[1]: Some runtimes come pre-installed in some OS (like Ruby in Macos) so tools often rely on it, that is NOT okay either. OS gets updated, versions change and tools break.
[2]: The only runtime under this sun that is acceptable to rely on is Bash and even that it is not great because Windows.
[3]: I think the cloud providers specifically don't bundle like that because the CLI gets too big to download which makes it hard to use in automated process (like CI pipelines). But this is no excuse, just make your tool in a different language.
Countless people before me have tried, and I myself have written and maintained a (proprietary) JS toolchain that has caused a some trouble with users over the years.
In the end, using a systems language is what I've settled on.
> Further, you double the dependencies
I'm acutely aware of this problem. Which is why if you look at Jam now, it has zero dependencies. Not even libc (which is mostly thanks to Zig).
The only "dependency" is a Unicode property detection library I wrote for Jam itself. And it is currently the fastest implementation for Unicode identifiers to exist. This simply isn't possible in JavaScript.
Same story with the Parser, etc. Parsers for JS already exist, but it's already known how far one can go, stitching together existing tools that aren't meant to work with each other.
Some dependencies can't be avoided in the long run, But I try to be very cautious, and vet every dependency thoroughly before considering it an option.
Ultimately, the dependence on two languages is a tradeoff that I've accepted, and the problem you mentioned with dev environments is one I'm going have to live with.
You're generally right, but the JS eco system has true culture problems, a barrier like Zig brings lots of virtue to such a project, there is space for a good tradeoff on that front.
No JS written tool will ever be faster than, for example, esbuild (Go).
IMHO, if JS had struct types, the difference could be much smaller. Dead proposal here: https://github.com/rbuckton/proposal-struct
Not sure if it started with Bun but there's good momentum.
But JS needs a better compiled extension story for that to happen. The absence of a standard like wheels makes everything difficult.
And they went for python, not js, because of this.
I could swear I saw this exact comment on this topic quite recently, but not as recently as the last few hours.
Examples:
- High-fructose corn syrup is eating the world. . You can fight it like so many people before, but you will lose like everyone else
- Cancer is eating the world. You can fight it like so many people before, but you will lose like everyone else
Etc.
I think bad things could be replaced with better things where we have this luxury. Taking a passive stance on this is pointless and lazy.
JS is no HFCS, what an absurd comparison.
JS is being tolerated. Not adored.
Frontenders are living in a bubble and don't see it.
Such as?
> JS is being tolerated. Not adored.
Can you prove it? I’ve seen enough people of JS being adored. I am one of them (maybe because I use it mostly for hobby projects).
There is such a variation between js, perl, python, rust, C#, lisp, bash, haskell, erlang and sql that the blind spot is huge.
It's like eating hot dogs all your life and debating food merits with people that tasted japanese, indian and french food.
It's a bit off-topic but if you liked Elixir it's kind of like Elixir but fast, and like OCaml but widely usable in production. It could also be familiar after Python due to mostly indentation-based scoping and not having to define signatures (except you go from something terrible and untyped to Hindley-Milner).
Spend several years with LISP, Elixir, Haskell, OCaml, then you'd be in a better position to say you keep loving JS.
You are sticking to the same type of languages. This warps your point of view.
What I said still applies, the guy seems to have a familiarity bias and said nothing to show otherwise.
As for being condescending, your reply seems like a textbook example.
Lost on your way to Reddit, I see.
You are free to interpret it any flame-bait ways. I ain't biting though.
Peace.
and of course even more pertinent - which languages did he even criticize other than to point out things that are not possible in Scala/Rust (certainly not a critique, each language has its pros/cons...)
I forgot. ¯\_(ツ)_/¯
I do remember I've seen one website emulating parts of Photoshop that was made with WASM but I lost it since I have no interest in the area.
You can take a look here though: https://madewithwebassembly.com/
And don't try to pick those apart one by one. Point was that WASM works and people do stuff with it. I've seen it several times and since it's not my main area of interest plus I don't get paid for it, I am not exactly keeping up to date every week.
> Can you prove it? I’ve seen enough people of JS being adored. I am one of them (maybe because I use it mostly for hobby projects).
Can I prove it? How would I even do that?
We all have anecdotal evidence, plus I am not secretly one of the sysadmins of Stackoverflow or Reddit. People literally have no choice but to ultimately use JS when coding for the browser, and that's a fact of life. There's Typescript nowadays and I hear from frontenders that it's definitely better but they also say every now and then you have to drop to JS and this is where they start sighing.
So no, I can't prove it. But the mere fact that JS is a monopoly of sorts, plus it has a huge and well-documented number of problems should incline you to believe as I do.
Finally, if you allow me, you have answered your own question by saying that you use JS mostly for hobby projects. That changes things SEVERELY; you are enthusiastic about the thing you are working on and the potential idiosyncrasies of the chosen tech (in this case JS) are seen as small and only mildly annoying hurdles. And if you get tired or annoyed you can just drop the hobby project until your enthusiasm gets naturally rejuvenated.
Start working with it every day and having to integrate with a lot of 3rd party libraries and other people's code and then tell me you love JS.
...And I've heard that too btw (that people work with JS every day and love it) but as a sibling comment pointed out, you likely don't have the perspective of working with other languages with very different features and tradeoffs.
JavaScript on the server, only when I can't say no for whatever reason, like being forced to search for another job as consequence of it.
Instead of rewriting JavaScript tools in Zig, Rust, Go, whatever cool compiled language of the month, use the said language directly.
I do agree that it's time to write frontend not in JS. It has been time for 10+ years even but oh well.
...Oh? What's that? I am hearing something faint in the distance, like... "their devs are not using JS well", maybe?
Tell you what. If a technology can be held wrong, then it's a bad technology. Which, yes, encompasses most technology nowadays. You heard me.
Can't really comment on Slack and Teams - I've obviously used both, but haven't been a heavy user of Teams and never was in a large "corporation-wide" Slack instance. I have, however, been a heavy user and developer of a Slack competitor. Not surprisingly, it had a lot of problems, but most of them stemmed from a) clumsy handling of megabytes of chat logs, b) a codebase that grew organically out of a startup project and needed a major rewrite, c) misallocation of resources (for the longest time, everything Electron-related was handled by a single developer in a satellite office). In my opinion, we would have had the same performance issues and bugs if it had been developed in any other language. We actually had some Qt apps that we deprecated and rewrote in JS as part of the main client - there was no user-noticeable performance drop, but significantly fewer crashes.
It was already mentioned that people can do better apps with Electron than others. Not surprising.
My objection is that the possibility to make laggy and bad apps exists. It should not exist.
As for what problems a Slack competitor faces, well, I am not interested. That's the dev team's problem and not mine. I want a lagless Slack.
I see. Then the only way forward is to stop all software development altogether and enjoy memories of apps that we used 20 years ago when we were much younger, and fantasies of apps that we could build with newly acquired wisdom, but will not.
Progress is way too slow and blindly follows local maximii. That's not OK. That's the part I am expressing displeasure with.
I also created a native block editor (like Notion's) in QML and C++: https://get-notes.com
I really hope that I can show that Qt with QML is a viable alternative to the inefficient resource hogs that are Electron/web apps.
Electron/JS is a herd just running in one direction, personified.
The fact that something is popular does not imply quality, not at all. Most of the time it implies that everybody settled on it and does not want to hear otherwise for various reasons (in this case: money, as employers are never going to give you the requisite time to build stuff with Qt or others).
The kind of argument you're making would be much more convincing if there were relevant examples of Qt-based apps that provided a superior user experience to the Electron apps you're criticising.
I don't personally find Electron apps laggy in general. VSCode is very responsive, for example.
1. High RAM usage.
2. The bloat will accelerate as more abstractions are added to web frameworks.
3. Battery usage - the most noticeable issue for me. I've yet to empirically test this, but my guess would be that Electron apps use substantially more power than native apps. I will test this theory eventually.
> The kind of argument you're making would be much more convincing if there were relevant examples of Qt-based apps that provided a superior user experience to the Electron apps you're criticising.
That's arguing in circles and it's the chicken-and-egg problem. Toolkits like Qt get less and less mindshare because Electron is desired and is encouraged by employers. I can't speak for Qt in particular btw, just using it as an example here, but my point is that it could be the best UI toolkit ever but since it takes a little more time to spin up an app with it then the business will want you to not use it.
Hence, I can't show you those Qt apps.
I feel that in these discussions people just blindly trust popularity to be an indicator of quality. I have a huge problem with that mindset and my career has seen the exact opposite, many times. Hence my comments. And if you want to take away just one thing from our discussion, it's this: please don't mistake Electron's popularity with the idea that Electron is the best UI toolkit.
>Toolkits like Qt get less and less mindshare because Electron is desired and is encouraged by employers.
But why is it desired? I've used Qt before (a long time ago) and know enough C++ to be dangerous. I'd still avoid using Qt simply because I already know how to make UI's in HTML+CSS. I can't say that replacing those tools with a manually-constructed C++ object graph – or with an unfamiliar markup language (QML) – strikes me as an appealing option. Most of the people carping about Electron apps on HN don't want to make a UI in Qt either! It is becoming one of these legendary magical technologies (like Lisp) that is much praised and yet too little used for its flaws to be common knowledge.
Past experiences. I've seen much smoother and faster apps more than 20 years ago. Surely they weren't using black magic.
> But why is it desired?
Because it is slightly quicker to start an app with it and because people want to have a web variant of it (runnable in the browser) if/when necessary. Those are actual selling points and I agree with them wholeheartedly -- I object to Electron's lagginess however. They can and they should do better.
> I'd still avoid using Qt simply because I already know how to make UI's in HTML+CSS.
Well yes, that's one of the factors -- people default to what they already know. But zero mention is given to the numerous problems with the DOM model of making UIs as well... a topic way too big for me to have any desire to delve in, again.
Inertia. A bad thing starts being used widely because of 2-3 desirable traits, and when it gains enough critical mass nobody wants to criticize it.
Well, I do.
Me too, but they didn't have all the whizzbang UI features that you get for free from a web stack. I could live without most of those, but users do expect all that stuff to work these days.
If the Slack devs are as sloppy as you say, then we ought to be thankful that they’re writing JS and not C++!
Lag is something that some people care about a lot. I actually think that most users don’t notice it or care very much, up to a certain point.
I specifically asked not to focus on examples but OK, I guess you really had to get it off your chest. ¯\_(ツ)_/¯
The topic was not "Electron vs. Qt" at all.
> I actually think that most users don’t notice it or care very much, up to a certain point.
Good for them, to me it's a mood and productivity killer. And no I'll not change my mind on this, ever.
Broader point is: the people who worked on Electron, Slack etc. should have really done better. But I have witnessed how job security almost always leads to sloppy work.
I didn't know they were the same. Fascinating actually, thanks for making me aware of it!
I'd argue that QML is an order of magnitude easier to develop than any other web frameworks (yes, you do have to get familiar with it, and if you made some apps before you could re-use some essential, battle-tested components). I hope to maybe share/sell some of mine some day.
I'm not sure what to answer regarding the browser experience. WASM isn't really there, isn't it? I haven't had a chance to test it on my own apps yet tho. I recently had a thought that maybe that Qt Company should also target HTML/CSS/JS generation of a QML file. Another, more interesting thought would be for browsers to be able to process and render QML files natively (but that won't happen without substential market share).
So again: the chicken-and-egg problem.
Basically, nothing is going to improve until somebody rolls up their sleeves and does all the hard work for free and do better than Electron -- or a corporation will push for it, it's always one of both. Then everybody else will flock to them.
That's what I observed several times during my life and career.
No reason for me to feel attacked, I pointed out that what you said doesn't make much sense in terms of how long have these languages been around.
"Instead of rewriting JavaScript tools in Zig, Rust, Go, whatever cool compiled language of the month, use the said language directly."
Meaning use Zig instead of JavaScript, use Rust instead of JavaScript, use Go instead of JavaScript, use "whatever cool compiled language of the month" instead of JavaScript, or if you prefer for the last example, use "whatever compiled language takes your fancy" instead of JavaScript.
Clear enough?
If not -- then clear. If yes, then this is getting even more confusing.
Given that boring languages like Java, C#, OCaml, Haskell are kind behind their times for cool kids to rewrite stuff, that leaves the unamed cool programming language of the month as option, whatever that language may be.
An enumeration of three choices followed by a generalization reads like the three choices are examples of the generalization, not disjoint.
And seeing that a writer commented something similar to my comments upthread, I am not very inclined to assume I can't read English.
Still, your point is taken, I just can't help but think I am sensing some slight... shall we call it non-preference... towards a few programming languages.
Here's Scheme in wasm calling JavaScript functions and DOM methods: https://spritely.institute/news/building-interactive-web-pag...
As time goes on WebAssembly will keep displacing JavaScript. We've entered the era of Zombie JavaScript.
I was benchmarking some dom manipulation code and saw large GC pauses corresponding to the number of `element.classList` accesses I was doing from rust. I think the answer is that performance isn't great (compared to if browsers provided a wasm-native api that didn't have to go through javascript), but if you're doing a lot of calculation in rust or are writing a typically small UI it's not a significant cost... may be more important for graphics and visualization work.
Just like Java has been dead for 20 years, now it is JS turn.
Delusional.
That's the thing about WebAssembly. It supports any language that compiles to it.
Like an old, worn-out horse, JavaScript will be taken to the knackery and reduced to glue. And then not even the glue will be needed anymore.
For great justice? You know what you doing.
Not sure what to make of that. Can you elaborate?
Also single bloody quotes. Why?
Because double quotes requires two keystrokes. Single quotes requires one. The only defense of double quotes is an appeal to history.
That's not to say everyone should use them, but reality is more nuanced than you make it seem.
When that happens, I have to escape the quote (if the language even allows escapes in single quoted strings) or change the whole string to double quotes.
Because typing a single quote will break parsing, it's also hard for tooling to assist. If I put a single quote unescaped, the auto formatter will probably barf.