> Their usage of upsert appears different than I was used to:
> Them: Upsert = Get or Insert
I agree that their choice of labeling the proposal as "upsert" is less than ideal. However, this functionality is reminiscent of a very useful Perl capability known as autovivification[0] as described in the motivation section:
A common problem when using a Map or WeakMap is how to
handle doing an update when you're not sure if the key
already exists in the map.
I do this all the time, getOrInsert would come really handy: you need something from a Map-backed storage, but the value may be unset, so you first check if it's undefined, set the default value, and then use that.
This API is inherently unsafe. Invalid pointers, incorrect signatures, or accessing memory after it has been freed can crash the process or corrupt memory.
Absolutely great idea to expose such "features" to the web dev world!
All the JS devs that are already struggling with mildly complicated language features will love the giant new field of bugs they only dreamed of.
The arrival of the first very hyped tool that will make activating FFI support a requirement will be a great moment in JS history. Happily an army of mildly educated web devs will activate a feature which potential risks they do not even understand.
Luckily nowadays supply chain attacks are a thing of the past in the JS world, oh, wait...
Criticism noted! But it is experimental, unconstrained FFI is inherently unsafe, and not all JavaScript developers are "mildly educated web devs".
I doubt we'll see this go into mainstream dependencies likely to be used by such developers – it hasn't in Ruby which has had easily accessible FFI mechanisms for years.
Oh hey, they're the people behind Oxlint and Oxfmt: https://oxc.rs/
I moved some projects over to those from ESLint + Prettier and while the compatibility isn't 100% (I didn't need that), and the time to process a codebase went from like way over a minute with the old tools to a few seconds with theirs.
After getting burned so many times on libraries, frameworks, services and platforms, even entire languages - one learns to be wary of critical dependencies. Every new project offers convenience in exchange for you giving up control of part of the software stack, and the power dynamic is often exploited sooner or later as revenue source. You can't trust anything that becomes irreplaceable, or that you can't write it (or at least understand it) yourself.
I mostly agree. But without argument, I can point out that a modern webapp requires tooling for capabilities like testing, linting, formatting, and bundling. Vite (and its ecosystem) has proven its mettle, and when it comes to being able to understand your dependencies, I'll take fewer, and simpler, and way faster, and more coherent, and more independent of misaligned corporate influence, every time. It's not even a trade-off, it's just better. I have deep expertise in wrangling eslint plugins and prettier configs and webpack, and am so grateful that's all in the rear-view mirror. An astonishing percentage of the world's most popular websites are built on a fragile and nearly-incomprehensible stack which no sane developer would choose. VoidZero (and TanStack, FWIW) are a breath of fresh air in making it possible to reason about your frontend tooling and architecture, and stepping away from unnecessary complexity and/or vendor lock-in. Of course it will eventually change. But as someone who's been building and improving web-based experiences for a living since the late 90's (for tiny startups and F500 enterprises and everything between), this is as good as it's ever been, and I recommend it without reservation.
VoidZero's business model is in Void, their deployment platform. Open source projects will always stay open source. This was announced at the very beginning.
Yes, nothing different from any other VC dev tool startup. When the community fractures people simply move on to something else. See rome -> biome for a very recent example.
Not all corporations are the same. I've worked in multiple regulated environments over the years and always been able to vet and experiment with tools. Also, your second sentence contradicts your first.
I'm really looking forward to the temporal api being universally available. Moment and Luxon are fairly good but sensible date/time handling is something that really ought to be baked into the platform ootb.
I always thought the old Date is kind of elegant... increment anything with an overflow and it all wraps around correctly, like `d.setDate(d.getDate() + 100)` to advance a date 100 days. "March 208th" is interpreted like you'd expect, as are the hours and minutes and such.
Of course, complete lack of non-local non-GMT time zones is a huge downside.
i'm pretty sure all that stuff works w/ Temporal... Temporal is extremely well-designed, in my experience. the js date object, on the other hand, has insane pitfalls, and i say this as someone who thinks not understanding JS ASI is a "skill issue", among other happily-un-"ergonomic" worldviews...
`d.add({ days: 100 })` also wraps like you'd expect. `d.with({ day: 208 })` becomes the last day of the month instead but "March 208th" is kinda nonsense anyway so whatever. You could emulate it with `d.add({ days: 208 - d.day })`
I thought this was the release where the built in sqlite got its experimental tag removed, but I don't see it in the release notes. THAT'S got me excited more than Temporal. A stable API, huge utility and one less dependency.
This is a surprise. And they still haven't included corepack as an official instruction on the nodejs.org download page. Is corepack a failed experiment?
Honest question, what isn't compatible? Where I work we've simply replaced node with bun across a lot of overcomplicated + crappy projects, and on my work+personal computers I alias bun/bunx to node/npx with seemingly no issues at all
Maybe if you start from scratch with a new project, but when migrating an old project it's definitely not a drop-in replacement. I try once or twice per year, but it's not worth the effort when the upside isn't that big.
In my testing Bun wasn't much faster most of the time, usually on par for all non-IO related stuff, and there were some cases with scheduling where Bun was noticable slower.
Their usage of upsert appears different than I was used to:
Me: Upsert = Update or Insert
Them: Upsert = Get or Insert