I have a bad habit of using novel combinations of tools on every side project. And then 80% of the time spent on the project is just fighting to make those things play nicely together.
Posts by Aaron Pettengill
Silksong coming out less than a week after the next Path of Exile 2 update puts me in a tough spot haha.
I'm falling down the "before I can start this project, I need to make this other thing" rabbit hole and now I'm seriously considering writing my own CRDT library.
Send help
I like my auto-complete with superpowers because it's still basically what I want to write, just faster. But not going anywhere near the full-on hands-off just-say-what-you-want-and-it-writes-the-code-for-you features.
Gah! I use that thing all the time T.T
This is always my first thought when library/framework authors say they want to be model-first, model-driven, etc. Optimizing for what LLMs spit out seems like a great way to stagnate.
The reason I've heard is it makes it hard to use multiple of these hooks on the same element.
But with a mergeRefs helper, that justification doesn't hold water for me.
I've lost count of the number of times I've said "have you tried doing what it says in the error?" and it was the right answer haha.
I like the direction! I feel like this will let me make the zod-form-data API feel a lot more zod-native.
I think in a perfect world enum values would have their own semantics. Something like Ocaml's polymorphic variants would feel javascripty, IMO.
That's a good point about serialization. Defaulting to symbols would help, I think. As long a we won't default to numbers starting from 0 haha.
IMO, if I was writing the v1 for this proposal I'd probably opt to start auto-initialized only and use Symbols as the default initializer.
The fact that `MyEnum.foo` just returns the value of the initializer has always made it feel like unnecessary sugar. And you can always just skip the enum and use the initializer directly if you want. My dream enum would be nominally typed or at least return some sort of special enum value.
The semantics of TS enums is part of why I don't like them, so it's kind of annoying that the proposal just wants to replicate that as closely as possible. ๐ฎโ๐จ
At least they're not introducing auto-initializing as numbers right now so that's better.
Is there anything like shadcn/ui but using Ariakit?
Tried SST on a project recently and now I'm not sure there's a reason to use anything else haha.
Valibot looks pretty great.
If StandardSchema sees enough adoption, I'm totally gonna write my own validation library.
I guess that last part is obvious actually. Shows how little time I've spent with server components ๐ญ
I haven't tried the React compiler so that's totally possible. But since the promise has a consistent identity on every render this seems like it should be okay.
Also worth noting that this is in a SPA so it's client-side code.
Javascript code reading: const someModule = import("some-module"); const Component = () => { const { useLazyHook } = use(someModule); useLazyHook(); }
I just realized you can lazy load hooks like this.
White cat perched above the lamp so he looks menacing.
I'm tempted to make this my pfp.
I've recently started using `useSuspenseQuery` and managing loading states feels way nicer to me.
And even without prefetching, I think it's more natural to avoid request waterfalls.
That just feels so wrong haha. It's not subtle but it isn't flashy either. Is just ๐คข
Oh and
Device: Galaxy S21 ultra
Browser: chrome opened within Bsky
Goose: 90
Toucan: 150
Puffin: 20
Also, the app froze when I opened puffin the first time because it defaulted to what I set for Toucan lol.
I usually go with Button and ButtonLink components in the same file and the styles are just a top-level variable in that file.
Screenshot of GitHub wrapped showing 24,000 commits.
Mine got a little over-inflated by the automation in my GH profile repo lol.
git-wrapped.com/profiles/air...
Ooh, neat! Maybe now every form library won't need to write it's own validation adapters.