1y ago

Are we ready for javascript without a build step on the front end in 2023?

On the current typescript / anti-typescript internet drama I saw someone mention javascript without a build step.

Do you think we're already there?

Last time I attempted it:

there were too many libraries I couldn't import
JSX (using babel) had a warning saying you shouldn't do it in the browser for production
there was some advice against not using a bundler, because several requests for different .js files is slower and bigger than a bundled package

48 comments

I dont get the third point. Its weirdly written.
Imagine not using typescript. The python community gaslighting the js community. Imagine a big project website, and all in plain js, + multiple devs. Thats is going to be a mess. Never going back from TS.
- About the third point, the performance of your JavaScript code can be worse if it's broken down into several small files rather than a single, bundled file. When a browser encounters a script tag linking to an external JavaScript file, it makes an HTTP request to fetch that file. This process occurs for each separate file. Each HTTP request involves time for network latency, server processing, and data transfer.
  I'm usually preferring typescript too, but this point got me curious. I'm guessing it wasn't an honest point, almost everywhere I look people are still using a build step, and I didn't notice any move in a different direction
  
  Http/1.0 would require serial connections, and the "multiple files bad" absolutely applies. Bundling and minification into a single JS file was common, even required - and I would consider that a build step. Otherwise you are dealing with all your code in a single massive file.
  On http/1.1 browsers would open 6-8 concurrent connections to fetch files simultaneously. At this point, code splitting had benefits.
  Most webservers now run http/2 which can fetch multiple files at the same time over the same connection. I believe it is "virtually unlimited" and the initial connection setup - which is often the largest performance hit - only happens once. At which point code splitting has such little impact on the transport layer, that it is more perfomant than transferring and loading all the code.
  https://blog.vespa.ai/http2/ Has an more details as well as some load testing against http/1.0, http/1.1 and http/2
  
  There is also a limit to the number of files the browser can download in parallel, so if many files have to be fetched, they have to wait until the previous downloads are finished. This slows down performance even more
- I think it’s a double negative: “against not using”
  As in, use a bundler.
- The python community gaslighting the js community.
  As a Python developer, I am absolutely giving ya'all a hard time for your gross bolted-on JavaScript type solution. /s
  I do this primarily because I find it cathartic while I work with our gross bolted-on Python type solution. /s (mostly sarcasm, some truth here)
  I do this secondarily because I'm jealous that Typescript seems to be a good bit more mature than MyPy. (No sarcasm here. I really am jealous on this point.)
  
  I recommend Pyright over Mypy if you don't mind it being owned by Microsoft. It has far fewer bugs, and if you do stumble on one, you don't have to fix it yourself because Microsoft's paid devs will fix it in a couple of working days (at least for the small bugs I've reported).
  
  Yeah i programm a lot in python and typescript too. Even if i type something in python its still very weakly typed.
- If you want type safety and no build step you do like svelte did and use jsdoc instead. You can run the typescript type checker on those annotations so if you care about not having a build step you can still have type safety.
  
  That sounds like spending a lot of extra effort just to avoid a little up-front effort.
  
  its like js but with extra steps WHY NOT USING TS then?

Why though? I think I am missing the point, but I don't see the problem with having a build step in your projects. Especially for frontend it is not just JavaScript, but things like Sass/SCSS to consider etc.
- When there's no build step, all the time is spent coding. None of it is spent configuring or setting up.
  The hardest part of any software class in my experience is the triple combo of:
  installation
  "well it works on my machine" and
  "well this code worked for last semester's class"
  When I have students start off editing one html file using pinned URL imports, the reliability is just insane. People might claim "installing typescript is reliable" but doesn't even come close to the reliability of not having a build step at all. No build step Just Works™, no M1 Mac edgecases, no npm audit, no rm rf node_modules.
  
  You always have linter steps, testing etc and a competent developer should be able to deal with all that. Of course you don't start with all this with new students, but I don't think that is what this post is about.
  
  Svelte/Vue/React components need to be compiled
  JavaScript should be minified if the project has a significant size
  File names should have a content hash, so they can be cashed in the browser
  Even with HTTP/2, there's still a case to be made for bundling hundreds or thousands of JS modules into a single file for better performance
  Bundlers give you a dev server with live reload and hot module replacement for great developer experience
  Setting up Vite is really easy and requires minimal configuration (compared to Webpack, for example)
- I switched from sass to tailwind a couple month ago and fucking love it. Highly recommend to give it try. Still requires a build step to trim out unused utility classes.

The real answer to the question you didn't ask is: just don't involve yourself in internet drama's.
Tell me instead, what is the problem with a build step and typescript?
- Well it's an extra step that you're only doing to address this problem.
  
  Are you one of those "types are noobs" people?

Yes. You can have reasonable load times using JS with no bundler. There are some notes though:
JSX isn't JavaScript, you will need to process these files (note: You can write React fairly reasonably without JSX).
TypeScript won't work because it isn't JavaScript. I would only want to use raw JavaScript on the most basic of projects.
A bundler/compressor is still going to reduce load times. Especially if you like lots of small files.
Caching will be a nightmare if you aren't doing version-based naming.

~~Just use WASM~~
If only...

Probably the more important question is whether DHH is your boss.
It’s fine to look for people with real experience/opinions on the internet, but at the end of the day, you have to build your own product.
I also am going to just say that I’m betting the kinds of stuff rails does in JS doesn’t really need a lot of complex JS. My guess is a lot of it paints on behavior similarly to what htmx does now, which doesn’t really require a ton of js code anymore. I don’t much see the point removing TS for the vast majority of projects.

Isn't Babel part of the build step? I might not be misunderstanding with what you mean by build step here.
In general, you probably can't avoid a bundler for the browser without significant compromises on performance and developer experience.
- I'm asking mostly out of curiosity, but I had a use case that I would like to completely avoid a build step. At work we have a very old web interface, that when I attempt to sell the idea of any major improvement the answer is "this is end of life, we are rewriting it". But the rewrite will take a long time, and it is easier to make gradual improvements without introducing new tooling. This one is from the 90s, there the JS is in a folder and is shipped as is.

It's doable and although I would best describe the dev experience as "ok", it is improving over time.
Per your bullets
use newer module based js libraries, yes this is limiting but getting better support over time, and you still have to deal with issues cause by different library types when using a bundler
JSX will require a build step at some point, pushing that to runtime doesn't improve anything. Instead I would favor lit html
probably true, but I would start without and wrap a bundler around the project when it becomes necessary, smaller projects will have a negligible effect. You should benchmark the differences yourself, and if you use es modules everywhere wrapping a bundler around it will be easy.
With the above you can get all the usual niceties too: hot reloading, lazy loading, etc

What do you mean by "build step"? For example, does running a barreler count as building?

I don't do front end work so I don't know how all the terms are used. If by build step you mean any sort of CI process then I would say no, I think automated tests are important and should be ran before pushing.
If you just mean some sort of transformation, transpilation, bundling, etc. then I don't know enough to answer. My gut feeling is that the question is framed incorrectly. Many of these things were made as workarounds for various problems (or to make things easier). If you don't have those problems then you don't need those solutions. It doesn't have anything to do with progress (as in what year it is). Originally JavaScript was just plain old JavaScript. If it worked then it can work now. If you need the solutions people have made to get around limitations and short comings and vanilla JavaScript can't do it then you'll still need those solutions.
- That's a great point. In any sort of enterprise system, you should be unit-testing your front end when you commit, and you should be UI-testing your front end before you deploy. If you're in a CI/CD pipeline, that normally happens right after the build step. If you need to have the pipeline running anyways, you might as well build.

Does it still make sense to use javascript? I thought WASM would make JS obsolete and introduce web UI frameworks in other languages.
- WASM’s biggest holdback is that it cannot directly access the DOM. Until then, JS will still have a prominent place in building anything rendered in a browser.
  
  That's actually a myth and real world performance isn't affected by this. See this video from leptos creator which is one of the more popular wasm ui framework https://youtu.be/4KtotxNAwME?si=D_vWV1LPQI-C9j8G
  The biggest issue is actually the size of the payload since you need to ship the entire app and language runtime.
- A lot of the WASM ecosystem is still pretty rough
- Absolutely. WASM is still not that good or easy.

48 comments