I am a big fan of Vite. But I have zero clue what those other tools are. I swear to God, I just put my head down to do some work and all the sudden, frontend tooling has evolved. I wonder if there is a push towards a "boring but works" stack.
Fear not, engineer! Recently some smart folks invented something called BackRub that’ll soon solve all your problems with not being able to find information in even less than the time it takes you to type a comment and out yourself as clueless!
> But I have zero clue what those other tools are.
The incorporated tools are actually really amazing:
- vitest, an ultra fast test runner. After using a lot of others, including jest and node's built in one, I love vitest.
- oxlint, replaces eslint but is compatible with its file format and ultra fast, since it isn't written JavaScript. I tried biome, but I found oxlint to have more rules and the eslint compatibility was nice.
- oxfmt, replaces prettier and is faster since it isn't written in JavaScript.
- rolldown, replaces rollup and is compatible with it but it is much faster since it isn't written in JavaScript.
The above are my go-tos in new projects anyhow since they are killer good and fast.
At work we've tried switching to vitest, and it's 1.5-2x slower than Jest (I think it's related to our very large and circular module graph), so performance is very much a your-mileage-may-vary thing.
The initial selling point was performance, but then they gradually realized that a lot of the slow cruft in Jest was necessary for correctness, and now it's about the same performance as Jest (obviously may vary in some specific situation).
However vitest is still great! Selling points now are stuff like:
- shares config with vite
- works with ESM out of the box (I think Jest still doesn't)
- integrated browser testing mode that is very nice
- overall just has a ton of nicely integrated features
I do recommend turning off "isolate" for as much of your code base as possible when it makes sense. And I recommend ensuring "maxWorkers" is being used properly, I prefer something like 60% of my totals cores as the number of workers to use. And use a top level vitest start so it properly runs all the packages in a pipeline rather than as separate vitest runs (which would mess up the maxWorkers optimization anyhow.)
I have my 3000 test project suite completing in 15 seconds on my MacBook Air M3. It is pretty sweet with that setup.
I wish Oxfmt supported plugins. Prettier's plugin API is one of the worst APIs I've ever worked with. I'm eager to switch to a different formatter with a better plugin API (and I need plugins).
I was hoping for something more flexible. I suppose the whole point of Prettier is that it's not flexible but I find 90% of the times the code is actually uglier when using it.
the other tools are for testing, bundling, linting and formatting. Previously you would use different tools from very different open source projects for these things, with different configurations, update cycles etc. Now it's all covered by one simple toolchain. Vite+ is basically the "boring but works" stack, while also being more performant and with less configuration required.
> Vite, Vitest, Rolldown, tsdown, Oxlint, and Oxfmt
You already use the first, the second you should be using and has existed for years - it's a jest-compatible runner integrated to Vite.
Rolldown is a part of Vite already (since version 8, alongside other implementing tools like lightnincss).
The ox* suite are native replacements for eslint and prettier. The latter has existed for years. Both of them aim to be drop-in compatible soon.
This is a boring stack.
FWIR, prior to the purchase by Vercel (or was it Cloud flare?), Vite+ was going to be a convenient, no fuss, low movement way set up projects that they sold to businesses for funding.
I think the boring stack would skip the need for most of these.
That is 5 different tools to write TypeScript in order to get some HTML, CSS and JavaScript for things that are not covered by modern web standards. Not including the frameworks itself which mostly have the same goal.
A "push" towards a boring stack? That won't happen because the hype cycle trends towards new tools like water down a river. But if you're looking, I can't recommend Rails enough in 2026. Built on web standards, it's quietly pushing the framework forward and is so much less maintenance than modern JS apps.
I guess I'm just an old man pumping my fist and yelling at the clouds at this point, but I think compartmentalizing every possible aspect of development and then making each one of those as complex as possible lead AI in the bureaucratic deterioration of the trade.
I'm currently experimenting with a stack combining assistance, ASP.NET Razor Pages, and htmx, specifically to avoid dealing with the complexities of the modern frontend stack.
AI eliminates a lot of the tedium, grinding tasks, while HTML partials and htmx complement each other perfectly.
Not having to manage frontend framework, SSR states is incredibly liberating. Sprinkle in some vanillla js and redraw specific parts of the page as needed... done
I think they should find a better name for this project. I find it very confusing since it's not really a better Vite.
At the time Void Zero was probably looking to monetize the Vite brand but now that they've been acquired by Cloudflare they don't need to do that anymore.
I love Vite, Vitest, Oxlint and Oxfmt and look in their direction for most of my new projects! I hope these folks manage to get a bunch of money and can fund the continued development for at least the next decade.
Sure beats opening some ancient project and seeing some mix of Gulp, Grunt, webpack and a bunch of other disjointed stuff (I migrated that one over to also use the newer stack).
> I hope these folks manage to get a bunch of money and can fund the continued development for at least the next decade.
I believe VoidZero has been acquired by Cloudflare [1], so money should not be an issue. Question is if Cloudflare will be willing to continue letting these people work on Vite and Vite+ features that benefit all cloud platforms, not just Cloudflare.
Making all this (for example) work nicely together can be tricky: Vite, ESLint, Prettier, Typescript and React, especially if it's full stack with SSR.
If you only focus on the front-end and remove Typescript from the equation it becomes easy enough. We'll have to see if Vite+ helps for the more complex cases.
> Sure beats opening some ancient project and seeing some mix of Vite, Vitest, Oxlint and Oxfmt and a bunch of other disjointed stuff (I migrated that one over to also use the newer stack).
I mean if I see those in N years, I'll be happier than with the older stack that came before them - the jank levels seem to generally be decreasing with every next attempt to get things right!
Did very little changes to tsconfig during past 6 months adoption
My day-to-day process - get the new package unless it some antd6, echart or some rendering engine or geo spatial lib, clean up with Claude, strict and unify type system and align it with my vite, tsconfig, oxlint tastes. The result - no need to follow libs bloat and supply chain attack issues. Easy to read, easy to fix.
Because VILT is dead and Livewire is now on version 1100?
I've worked on both stacks in the last few years across several clients. Honestly like with anything in tech it seems to mostly fall apart with half-regarded usage of the tools in growing teams that don't care about their quality in favour of "get ticket done".
Vite had five major version in the four years 2022-2026. Version 3 => 4 => 5 => 6 => 7 => 8. Each one of those had breaking changes and required devs to go through a migration. It's too much. And for what? It's not as if it is dramatically better now than it was in version 3.
I can't say I would really look forward to bringing this level of needless churn and constant disruption to the rest of my development toolchain. Anyway, Vite+ is really just wrapping existing tools into an abstracted command-line interface? And so I have more layers of indirection to wade through in order to get the thing to do what I want? So far I am not optimistic about this prospect...
I've followed all the main migrations and I've say they where really quite smooth, can't remember having any major issues and each time it tended to be worth it.
I did the migrations and they were not horrible. A few breaking changes but they were relatively isolated. And the speed and improvements over these versions had been quite significant.
The major change through all of those was solidifying Vite's concepts of "environments" for better SSR support. If you weren't using that, there wasn't much instability.
Surely lacking features also creates demand for more tooling.
uv removes all the pain from python packaging. and if you ever used it for anything more than one off scripts you know it used to be a lot of pain.
manually activating venv, inconsistent python and dep versions, duplicate files taking up space, slow and broken tools, fragmented configs, global state, all of that is gone now. if you never experienced it you have no idea how bad it was.
not all of it was uv specifically (pyproject.toml was a proposal for some time before) but they cleaned it up. its the only reason i even think about python as an option for new projects.
In my experience the bundling isn’t really the important aspect (though it also doesn’t harm anything), it’s more just having an ecosystem of plugins for code transpiling, static asset inclusion (e.g. text files) etc and a configuration format folks are already used to.
For me, the main benefit is deployment bundle/artifact size reduction. Mostly from dropping unneeded files from node_modules. Many packages include both esm and cjs builds, sources, docs, TS types, etc. stuff that you don’t need in prod.
This matters for lambdas, for example, because deployed code size has limits there.
Running typescript without compilation is still tricky with plain Node. `vite dev` has amazing DX not available for Node programs. I'm wondering if Vite+ tackles this problem.
Don't we have `tsx` and `nodemon` (or the native Node reloader) for that? What are the DX gaps you see on the server side out of on-the-fly transpilation and reload on watch?
Yes, I use tsx for Node programs. It's not great when sharing the same codebase for both client and server code, they have completely different dev workflows.
One advantage of precompilation is risk reduction. Say tsx gets hacked somehow (hardly unprecedented with Node modules!) you’ve got it running on your production server exposed to the internet. Precompilation on a CI pipeline is still a risk but a significantly lower one.
In theory, typescript doesn't need to be transpiled, you can run ts files using `node --experimental-strip-types file.ts` as long as you don't use any code that needs transpilation (like typescript enums).
No, because of ESM import resolution rules. Typescript suggests extensionless imports, making it incompartible with ESM and therefore Node. Luckly, `node --import=tsx file.ts` handles imports well.
This is especially hairy when making a typescript library that is distributed non-compiled (without dist/) and is supposed to run in both browser and Node.
This allows you to use fully-specified imports in TypeScript files, which works basically everywhere — NodeJS, TypeScript, bundlers, etc.
The exceptions are browsers (obviously, only normal JS syntax there), and packages inside `node_modules`, which NodeJS will not do any type stripping for. So if you're writing a library, you'll probably still need to distribute the compiled sources, rather than distributing the raw TypeScript files alone. Or you use the JSDoc syntax for TypeScript, which can do everything that .ts files can do, but is more verbose and idiosyncratic.
I transpile for prod, but use --strip-types when running in dev, and all I had to do was to make a 10-line ESM register hook that rewrites .js to .ts if the .js import fails, and then a one-liner import register trampoline script. Not sure I'd do that in prod, but works fine in dev at least.
This way I could just use node --watch instead of tsx or nodemon.
In cases where startup time matters and you have a slow disk, bundling can drastically reduce the number of filesystem calls you have to make for large dependency graphs.
I’ve found if you want interop ts esm and js cjs you need to compile your code - and then `tsc` doesn’t bundle your dependencies for you and outputs incomplete code.
Yeah, not only the name: they’re also going with various semiotic signs that are strongly associated with a subscription service, including their website design, choice of typography, and even the press release–style copy.
Looks like they have been acquired by Cloudflare, and pivoted to fully open source, but they haven’t really tweaked their messaging to make that fully land with unsuspecting visitors.
It’s kinda like the reverse situation of open source projects that switch to a source available license, but keep the aesthetics of an open source project. Kinda funny!
This is just what modern languages have out of box. (Like rust and go.) it’s a true shame that web isn’t actually unified behind a type safe language with a single solid toolchain. It’s a huge pain to manage and I’m curious how much money it’s cost the industry. “Vite+” isn’t a true solution to that. There are many competing toolchains. And no default standardized one.
I'm not very familiar with Rust, but doesn't cargo pull a lot of external dependencies for most projects? I really like how Go can do everything with just the standard library, but I wasn't aware Rust was similar. For typescript we've moved our stuff to bun. It has it's own risk management perspective compared to node, but at least it's now possible to build web services without having to rely on a bunch of external dependencies. Which in our highly regulated business would require security policies for each dependency explaining the risks, why we accept them and how we mitigate them.
> without having to rely on a bunch of external dependencies. Which in our highly regulated business would require security policies for each dependency explaining the risks, why we accept them and how we mitigate them.
How about the dependencies Bun is pulling? How did you ever managed to pass security policies with Bun which has so many segfaults that nobody even bothers to write CVEs for them.
Cargo itself doesn't pull the dependencies, but yes to Rust's standard library being much more lean than Go. Bring your own HTTP, text templating, and such, but core data structures are provided.
Go gives you a bunch of goodies in the standard library.
Rust provides things like your build system, testing, and package management all together, which is what I assume OP meant.
Vite+ isn't a layer, it's "just" a high-performance suite of excellent tools that work well together to provide a great DX for developers.
Vite+ can improve and simplify what developers are already doing with ad-hoc collections of tools. Vite is already an industry standard, and Vite+ has a good chance of achieving that status as well.
Some barbarian without a grasp of the Dutch language knee-jerked the down-vote button so I'll add a Swedish version which adds an important attribute.
Imse vimse spindel klättra upp för trå'n.
Ner faller regnet, spolar spindeln bort.
Upp stiger solen, torkar bort allt regn.
Imse vimse spindel klättrar upp igen.
Here's how to interpret this saga of the ever-climbing little spider in the context of web development. It climbs up its tread (klättra upp för trå'n) 'cause that new framework will sure make catching those flies (clicks/jobs/likes/whatevers) easier. And then the rain starts (the CVEs start piling up, the corrupted packages come flooding in) and the hapless spider gets thrown off its web (Pwned!) until the sun comes back and dries away the rain (a new framework, yay, this will solve all problems) upon which the spider climbs up its thread again.
I think web development does not need that many layers. Usually there is a clear purpose for each layer. I think most problems in web are self-created.
It's all great to leverage until something breaks in a middle layer and you can't reproduce without submitting your entire project in the GitHub issue.
I have been doing "modern web" things since essentially day zero (you kids with your fancy JIT compiled javascript interpreters!)
SvelteKit, and by extension, Vite, has been the single most productive webstack I have ever used. If this offers anything on top of that, I welcome it with open arms.
Is everyone project so simple that it can fit in these "vp check" / "vp dev" commands? Like even for my amateurish web app, I have a custom web server with a self-signed certificate with an "/etc/hosts" domain; and for checks I need to do custom checks for GraphQL and a couple of cloned NPM packages.
I have removed vite because dev build and reload is noticable slower than just esbuild and browser refresh. Vite does nothing for me that an LLM can not just trivially rebuild in a bespoke manner.
I am actually pushing our frontend devs to remove more and more dependencies and leverage LLMs to just write the code instead of all the dumbass packages in hellscape of supply chain attacks via node/npm.
You are signing up for another hellscape of unmaintainable slop. Enable package cooldowns and only whitelist internal packages and you are better off than 90%
You act like the existing packages being published in this ecosystem aren’t already slop or quickly getting there.
We already do cooldowns and disable preinstall and postinstall scripts on all packages except for ones that actually require it.
I bet if you looked at 70% of your dependencies pulled in, you would be horrified. I would rather have that capabilities via code in my repos at this point.
How do you bundle web workers that import dependencies? iirc the issue in esbuild for that is still open and users are manually building their workers as separate entry points, which is very fragile.
I am a big fan of Vite. But I have zero clue what those other tools are. I swear to God, I just put my head down to do some work and all the sudden, frontend tooling has evolved. I wonder if there is a push towards a "boring but works" stack.
Fear not, engineer! Recently some smart folks invented something called BackRub that’ll soon solve all your problems with not being able to find information in even less than the time it takes you to type a comment and out yourself as clueless!
> But I have zero clue what those other tools are.
The incorporated tools are actually really amazing:
- vitest, an ultra fast test runner. After using a lot of others, including jest and node's built in one, I love vitest.
- oxlint, replaces eslint but is compatible with its file format and ultra fast, since it isn't written JavaScript. I tried biome, but I found oxlint to have more rules and the eslint compatibility was nice.
- oxfmt, replaces prettier and is faster since it isn't written in JavaScript.
- rolldown, replaces rollup and is compatible with it but it is much faster since it isn't written in JavaScript.
The above are my go-tos in new projects anyhow since they are killer good and fast.
I don't get how a test runner can be "ultra fast". Surely all the time is taken by the tests, not calling the test functions?
Say this again when you have worked with Jest, one of the worst and slowest pieces of software I've ever worked with.
At work we've tried switching to vitest, and it's 1.5-2x slower than Jest (I think it's related to our very large and circular module graph), so performance is very much a your-mileage-may-vary thing.
The initial selling point was performance, but then they gradually realized that a lot of the slow cruft in Jest was necessary for correctness, and now it's about the same performance as Jest (obviously may vary in some specific situation).
However vitest is still great! Selling points now are stuff like:
- shares config with vite
- works with ESM out of the box (I think Jest still doesn't)
- integrated browser testing mode that is very nice
- overall just has a ton of nicely integrated features
I do recommend turning off "isolate" for as much of your code base as possible when it makes sense. And I recommend ensuring "maxWorkers" is being used properly, I prefer something like 60% of my totals cores as the number of workers to use. And use a top level vitest start so it properly runs all the packages in a pipeline rather than as separate vitest runs (which would mess up the maxWorkers optimization anyhow.)
I have my 3000 test project suite completing in 15 seconds on my MacBook Air M3. It is pretty sweet with that setup.
> Surely all the time is taken by the tests, not calling the test functions?
Calling tests has overhead. Also knowing how to schedule and parallelize tests with dependencies is not as "simple".
I wish Oxfmt supported plugins. Prettier's plugin API is one of the worst APIs I've ever worked with. I'm eager to switch to a different formatter with a better plugin API (and I need plugins).
I was excited about Oxfmt until I tried it and found that it's mostly intended to be a Prettier replacement. Ugh.
I'm curious to know what you were hoping for. Are there features you're looking for in a formatter that prettier/oxfmt don't offer?
I was hoping for something more flexible. I suppose the whole point of Prettier is that it's not flexible but I find 90% of the times the code is actually uglier when using it.
Have you tried contributing a system for that?
Rolldown is used in Vite (non plus) though.
the other tools are for testing, bundling, linting and formatting. Previously you would use different tools from very different open source projects for these things, with different configurations, update cycles etc. Now it's all covered by one simple toolchain. Vite+ is basically the "boring but works" stack, while also being more performant and with less configuration required.
This is the latest emerging "boring but works" stack.
"Latest emerging boring but works" sounds like an oxymoron.
Or sarcasm.
I’ve bounced off Vite because I don’t see a big advantage over esbuild. (I also use Deno.) What do you find useful?
That's what I liked from Bun's proposition. A single binary that just works. Hopefully the others take notice.
> Vite, Vitest, Rolldown, tsdown, Oxlint, and Oxfmt
You already use the first, the second you should be using and has existed for years - it's a jest-compatible runner integrated to Vite.
Rolldown is a part of Vite already (since version 8, alongside other implementing tools like lightnincss).
The ox* suite are native replacements for eslint and prettier. The latter has existed for years. Both of them aim to be drop-in compatible soon.
This is a boring stack.
FWIR, prior to the purchase by Vercel (or was it Cloud flare?), Vite+ was going to be a convenient, no fuss, low movement way set up projects that they sold to businesses for funding.
> This is a boring stack.
I think the boring stack would skip the need for most of these. That is 5 different tools to write TypeScript in order to get some HTML, CSS and JavaScript for things that are not covered by modern web standards. Not including the frameworks itself which mostly have the same goal.
Cloudflare.
A "push" towards a boring stack? That won't happen because the hype cycle trends towards new tools like water down a river. But if you're looking, I can't recommend Rails enough in 2026. Built on web standards, it's quietly pushing the framework forward and is so much less maintenance than modern JS apps.
I guess I'm just an old man pumping my fist and yelling at the clouds at this point, but I think compartmentalizing every possible aspect of development and then making each one of those as complex as possible lead AI in the bureaucratic deterioration of the trade.
I'm currently experimenting with a stack combining assistance, ASP.NET Razor Pages, and htmx, specifically to avoid dealing with the complexities of the modern frontend stack.
AI eliminates a lot of the tedium, grinding tasks, while HTML partials and htmx complement each other perfectly.
Not having to manage frontend framework, SSR states is incredibly liberating. Sprinkle in some vanillla js and redraw specific parts of the page as needed... done
I think they should find a better name for this project. I find it very confusing since it's not really a better Vite.
At the time Void Zero was probably looking to monetize the Vite brand but now that they've been acquired by Cloudflare they don't need to do that anymore.
Its vite... plus a bunch of stuff. Plus can mean different things
> I think they should find a better name for this project.
Need another plus? Vite++
I love Vite, Vitest, Oxlint and Oxfmt and look in their direction for most of my new projects! I hope these folks manage to get a bunch of money and can fund the continued development for at least the next decade.
Sure beats opening some ancient project and seeing some mix of Gulp, Grunt, webpack and a bunch of other disjointed stuff (I migrated that one over to also use the newer stack).
> I hope these folks manage to get a bunch of money and can fund the continued development for at least the next decade.
I believe VoidZero has been acquired by Cloudflare [1], so money should not be an issue. Question is if Cloudflare will be willing to continue letting these people work on Vite and Vite+ features that benefit all cloud platforms, not just Cloudflare.
1. https://blog.cloudflare.com/voidzero-joins-cloudflare/
Making all this (for example) work nicely together can be tricky: Vite, ESLint, Prettier, Typescript and React, especially if it's full stack with SSR.
If you only focus on the front-end and remove Typescript from the equation it becomes easy enough. We'll have to see if Vite+ helps for the more complex cases.
> Sure beats opening some ancient project and seeing some mix of Vite, Vitest, Oxlint and Oxfmt and a bunch of other disjointed stuff (I migrated that one over to also use the newer stack).
I mean if I see those in N years, I'll be happier than with the older stack that came before them - the jank levels seem to generally be decreasing with every next attempt to get things right!
Extremely happy user of Vite, Vitest, Rolldown, tsdown, Oxlint, and Oxfmt.
I do have lot of hardforked packages, and dont want to look back. Everything just works.
If you confused by the naming, start from Oxlint https://oxc.rs/docs/guide/usage/linter Rolldown https://rolldown.rs/
Did very little changes to tsconfig during past 6 months adoption
My day-to-day process - get the new package unless it some antd6, echart or some rendering engine or geo spatial lib, clean up with Claude, strict and unify type system and align it with my vite, tsconfig, oxlint tastes. The result - no need to follow libs bloat and supply chain attack issues. Easy to read, easy to fix.
Truly have so much trouble keeping up with the frontend (or JavaScript?) ecosystem. I so miss working in laravel. Wish more jobs paid well to use it.
Trust me you don't want to work with Laravel Livewire and Alpine.js, that would still require keeping up and for a less than satisfying result.
Because VILT is dead and Livewire is now on version 1100?
I've worked on both stacks in the last few years across several clients. Honestly like with anything in tech it seems to mostly fall apart with half-regarded usage of the tools in growing teams that don't care about their quality in favour of "get ticket done".
you actually don't have to keep up, whatever you were using still works
I feel ya, we're slowly phasing out our Laravel monolith for python lambdas. I miss those beautiful Laravel 6 days!
Vite had five major version in the four years 2022-2026. Version 3 => 4 => 5 => 6 => 7 => 8. Each one of those had breaking changes and required devs to go through a migration. It's too much. And for what? It's not as if it is dramatically better now than it was in version 3.
I can't say I would really look forward to bringing this level of needless churn and constant disruption to the rest of my development toolchain. Anyway, Vite+ is really just wrapping existing tools into an abstracted command-line interface? And so I have more layers of indirection to wade through in order to get the thing to do what I want? So far I am not optimistic about this prospect...
I've followed all the main migrations and I've say they where really quite smooth, can't remember having any major issues and each time it tended to be worth it.
> Version 3 => 4 => 5 => 6 => 7 => 8.
I did the migrations and they were not horrible. A few breaking changes but they were relatively isolated. And the speed and improvements over these versions had been quite significant.
Echoing another person, but we did all those major version migrations without any breakage / churn. What were the breaks you ran into?
> It's not as if it is dramatically better now than it was in version 3
Adding stuff for SSR was a major improvement.
I went from 4 to 8 in a single step. I had to change 5 lines of config.
Can we please stop the incessant belly-aching over problems that don't exist. I have to question if you even use these tools?
also i'm pretty sure claude can one shot this these days. a non-issue.
The churn is the product.
This makes absolutely zero sense. If you're going to post cynically, at least try to have some sort of coherent point?
Tooling instability creates the demand for more tooling.
The major change through all of those was solidifying Vite's concepts of "environments" for better SSR support. If you weren't using that, there wasn't much instability.
Surely lacking features also creates demand for more tooling.
it worked for uv so i can imagine a competent team can do the same thing for javascript!
Surprised to see this is the only uv reference in the comments!
Feels like an obvious comparison to me, and a very welcome development for the JS ecosystem.
uv made me actually _enjoy_ working in Python again.
a single tool was the enabler of enjoyment? It seems enjoyment is a fleeting thing these days if that's the case
uv removes all the pain from python packaging. and if you ever used it for anything more than one off scripts you know it used to be a lot of pain.
manually activating venv, inconsistent python and dep versions, duplicate files taking up space, slow and broken tools, fragmented configs, global state, all of that is gone now. if you never experienced it you have no idea how bad it was.
not all of it was uv specifically (pyproject.toml was a proposal for some time before) but they cleaned it up. its the only reason i even think about python as an option for new projects.
Can it be used for Node builds or browser-only same as Vite?
It uses Vite so the same limitations as Vite. However, I have been using Vite for my NestJS servers without any problem with `vite-plugin-node`. See example at https://github.com/leosuncin/nest-vite-example/blob/master/v...
Here is my particular incantation for targeting node that is working very well: https://pastebin.com/ynz4B5X0
Essentially you pretend to be a library
I am using Vite+ for CLIs as well, yes. You don't use Vite as dev server then but lint, format, task running and caching is still there!
I think tsdown is much better suited for CLI and library builds. Vite has many web-isms that don't matter for these.
I'd be interested in seeing this implementation if it's publicly available. Do you have a GitHub link? Thanks!
I'm always curious of the use case when someone proposes Node code bundling. What's the advantage? Obfuscation in SEA?
In my experience the bundling isn’t really the important aspect (though it also doesn’t harm anything), it’s more just having an ecosystem of plugins for code transpiling, static asset inclusion (e.g. text files) etc and a configuration format folks are already used to.
For me, the main benefit is deployment bundle/artifact size reduction. Mostly from dropping unneeded files from node_modules. Many packages include both esm and cjs builds, sources, docs, TS types, etc. stuff that you don’t need in prod. This matters for lambdas, for example, because deployed code size has limits there.
Running typescript without compilation is still tricky with plain Node. `vite dev` has amazing DX not available for Node programs. I'm wondering if Vite+ tackles this problem.
Don't we have `tsx` and `nodemon` (or the native Node reloader) for that? What are the DX gaps you see on the server side out of on-the-fly transpilation and reload on watch?
Yes, I use tsx for Node programs. It's not great when sharing the same codebase for both client and server code, they have completely different dev workflows.
One advantage of precompilation is risk reduction. Say tsx gets hacked somehow (hardly unprecedented with Node modules!) you’ve got it running on your production server exposed to the internet. Precompilation on a CI pipeline is still a risk but a significantly lower one.
If only the whole JavaScript wasn't as dependency hell of single function packages....
In theory, typescript doesn't need to be transpiled, you can run ts files using `node --experimental-strip-types file.ts` as long as you don't use any code that needs transpilation (like typescript enums).
Still need tsx to do type checking
No, because of ESM import resolution rules. Typescript suggests extensionless imports, making it incompartible with ESM and therefore Node. Luckly, `node --import=tsx file.ts` handles imports well.
This is especially hairy when making a typescript library that is distributed non-compiled (without dist/) and is supposed to run in both browser and Node.
https://github.com/nodejs/node/issues/46006
https://github.com/microsoft/TypeScript/issues/16577
You can use one of the following:
`allowImportingTsExtensions: true` (https://www.typescriptlang.org/tsconfig/#allowImportingTsExt..., useful if you're running `tsc` in noEmit mode as a linter)
`rewriteRelativeImportExtensions: true` (https://www.typescriptlang.org/tsconfig/#rewriteRelativeImpo..., useful if you're using `tsc` to compile TS files to JS.
This allows you to use fully-specified imports in TypeScript files, which works basically everywhere — NodeJS, TypeScript, bundlers, etc.
The exceptions are browsers (obviously, only normal JS syntax there), and packages inside `node_modules`, which NodeJS will not do any type stripping for. So if you're writing a library, you'll probably still need to distribute the compiled sources, rather than distributing the raw TypeScript files alone. Or you use the JSDoc syntax for TypeScript, which can do everything that .ts files can do, but is more verbose and idiosyncratic.
I transpile for prod, but use --strip-types when running in dev, and all I had to do was to make a 10-line ESM register hook that rewrites .js to .ts if the .js import fails, and then a one-liner import register trampoline script. Not sure I'd do that in prod, but works fine in dev at least.
This way I could just use node --watch instead of tsx or nodemon.
Mind sharing the implementation? I think it's basically what tsx is doing when used in `node --import tsx`.
@afavour if you need precompilation in CI can't you simply use... tsc?
In cases where startup time matters and you have a slow disk, bundling can drastically reduce the number of filesystem calls you have to make for large dependency graphs.
I’ve found if you want interop ts esm and js cjs you need to compile your code - and then `tsc` doesn’t bundle your dependencies for you and outputs incomplete code.
Is there a subscription with this?
I'm just wary about anything with a '+' and I assume there is a subscription attached to it.
Looking at this it doesn't look like it.
My first thought too. "$name+" is strongly coded now as "subscription service for $name"
Yeah, not only the name: they’re also going with various semiotic signs that are strongly associated with a subscription service, including their website design, choice of typography, and even the press release–style copy.
Looks like they have been acquired by Cloudflare, and pivoted to fully open source, but they haven’t really tweaked their messaging to make that fully land with unsuspecting visitors.
It’s kinda like the reverse situation of open source projects that switch to a source available license, but keep the aesthetics of an open source project. Kinda funny!
Says:
"It is fully open source under the MIT license"
I think that used to be the idea but then they got acqui-hired
Naming is worrisome!
Toolchain Grand Vitesse
Doug McIlroy once said: "Make each program do one thing well".
Then he'd love this. Like Unix, Vite+ is a collection of programs that do one thing well.
Layer on layer on layer on layer on layer.... Web development is just a meme by now
This is just what modern languages have out of box. (Like rust and go.) it’s a true shame that web isn’t actually unified behind a type safe language with a single solid toolchain. It’s a huge pain to manage and I’m curious how much money it’s cost the industry. “Vite+” isn’t a true solution to that. There are many competing toolchains. And no default standardized one.
I'm not very familiar with Rust, but doesn't cargo pull a lot of external dependencies for most projects? I really like how Go can do everything with just the standard library, but I wasn't aware Rust was similar. For typescript we've moved our stuff to bun. It has it's own risk management perspective compared to node, but at least it's now possible to build web services without having to rely on a bunch of external dependencies. Which in our highly regulated business would require security policies for each dependency explaining the risks, why we accept them and how we mitigate them.
> without having to rely on a bunch of external dependencies. Which in our highly regulated business would require security policies for each dependency explaining the risks, why we accept them and how we mitigate them.
How about the dependencies Bun is pulling? How did you ever managed to pass security policies with Bun which has so many segfaults that nobody even bothers to write CVEs for them.
Cargo itself doesn't pull the dependencies, but yes to Rust's standard library being much more lean than Go. Bring your own HTTP, text templating, and such, but core data structures are provided.
Go gives you a bunch of goodies in the standard library.
Rust provides things like your build system, testing, and package management all together, which is what I assume OP meant.
Vite+ isn't a layer, it's "just" a high-performance suite of excellent tools that work well together to provide a great DX for developers.
Vite+ can improve and simplify what developers are already doing with ad-hoc collections of tools. Vite is already an industry standard, and Vite+ has a good chance of achieving that status as well.
Deze vuist op deze vuist. Deze vuist op deze vuist. Deze vuist op deze vuist. En zo klim ik naar boven.
You probably need to see a video or gif to get it.
Some barbarian without a grasp of the Dutch language knee-jerked the down-vote button so I'll add a Swedish version which adds an important attribute.
Here's how to interpret this saga of the ever-climbing little spider in the context of web development. It climbs up its tread (klättra upp för trå'n) 'cause that new framework will sure make catching those flies (clicks/jobs/likes/whatevers) easier. And then the rain starts (the CVEs start piling up, the corrupted packages come flooding in) and the hapless spider gets thrown off its web (Pwned!) until the sun comes back and dries away the rain (a new framework, yay, this will solve all problems) upon which the spider climbs up its thread again.the swedish song has an english version: https://en.wikipedia.org/wiki/Itsy_Bitsy_Spider
Pretty much all software is built like that.
I think web development does not need that many layers. Usually there is a clear purpose for each layer. I think most problems in web are self-created.
It's all great to leverage until something breaks in a middle layer and you can't reproduce without submitting your entire project in the GitHub issue.
Don't be so negative nancy here!
I have been doing "modern web" things since essentially day zero (you kids with your fancy JIT compiled javascript interpreters!)
SvelteKit, and by extension, Vite, has been the single most productive webstack I have ever used. If this offers anything on top of that, I welcome it with open arms.
Far from being a meme!
I use vp with sveltekit for my personal projects now, and I must say it's a very nice experience.
vp + sv seems to work very well, when I tried it. And oxfmt supports svelte now too!
Is everyone project so simple that it can fit in these "vp check" / "vp dev" commands? Like even for my amateurish web app, I have a custom web server with a self-signed certificate with an "/etc/hosts" domain; and for checks I need to do custom checks for GraphQL and a couple of cloned NPM packages.
I appreciate the effort to bring things together in this but…
> Vite+ will manage your global Node.js runtime and package manager.
What? Why?
You’re really going all-in if you adopt this; and… for what? A bit of cozy tooling around existing standard ways of doing things?
Ok, sure; I like tools, like vite.
…but even for an opinionated tool, this is extraordinarily opinionated. Like next.js
Im skeptical.
The pitch of bringing things together seems strong, but did we go too far here?
Reading reviews of people using this didn't really convince me.
It seems to be running on the coat tails of the vite name, rather than its own merit.
It’s a great move for Cloudflare to have bought up voidzero.
Im tried boss
I have removed vite because dev build and reload is noticable slower than just esbuild and browser refresh. Vite does nothing for me that an LLM can not just trivially rebuild in a bespoke manner.
YMMV
I am actually pushing our frontend devs to remove more and more dependencies and leverage LLMs to just write the code instead of all the dumbass packages in hellscape of supply chain attacks via node/npm.
You are signing up for another hellscape of unmaintainable slop. Enable package cooldowns and only whitelist internal packages and you are better off than 90%
You act like the existing packages being published in this ecosystem aren’t already slop or quickly getting there.
We already do cooldowns and disable preinstall and postinstall scripts on all packages except for ones that actually require it.
I bet if you looked at 70% of your dependencies pulled in, you would be horrified. I would rather have that capabilities via code in my repos at this point.
How do you bundle web workers that import dependencies? iirc the issue in esbuild for that is still open and users are manually building their workers as separate entry points, which is very fragile.
Vite pumps out major versions -- that is, breaking changes -- at an incredible rate.
I don't want to be a vite upgrade engineer.
I'll try to pass on this if I can.