Solid 2.0

20 points by bio a day ago on lobsters | 19 comments

natkr | 22 hours ago

Solid 1 was pretty much my go-to example of what a web (front-end) framework should look like, so it's really sad to lose it to the slop machines too. (Even if it's been a long term coming.)

I guess at this point I'm pretty much resigned to the idea that any big new web app project would have to start with yak-shaving some form of Solid 1 clone.

…at least it's fun to go back to squeezing everything into the server-rendered box in the meantime. (And dealing with vanilla JS where it's not too bad.)

conartist6 | 21 hours ago

Same reaction.

I recently forked the Solid 1 signals library and read through all the code in it. When I took a look at the changes to signals in Solid 2, I was immediately hit with a massive heap of rotting trash comments.

It makes me sad and angry because there's probably enough good ideas there that I'll have to implement them, but if I want a long-term-maintenance-quality implementation it seems like I'll have to write it myself.

My favorite bit is that into their effects code they put a bunch of logic about "affects". Public Service Announcement: "an affect" (the noun usage) is basically "putting on airs", like pretending to have a lisp or a southern accent. Its noun form is not related to cause and effect as is the verb "to affect". The noun form of that verb is "effect".

conartist6 | 21 hours ago

It'd be kinda like if they wrote a codebase about tasks and introduced a new type of object called "tesk" that is obviously just a misspelling of "task" but which nevertheleess is used everywhere as if was really a separate conceptual entity.

nemin | 21 hours ago

So like thunks?

alloyed | 17 hours ago

having both "effect" and "affect" as important but entirely separate concepts seems especially mean to anybody who wants to discuss their code in person, potentially with people with vowel shifted accents

did they say effect, or did they say affect....?

kaspar | 19 hours ago

I thought Solid 1 signals are very much tied into Solid itself. I am excited about Solid 2 because it's a separate package there. You split out a signals lib from Solid 1?

Agree with sibling comments that affect is probably short for async effect.

conartist6 | 19 hours ago

I have a work in progress signals lib split out from Solid 1, but I'm not aiming for 1:1 compatibility.

oneirine | 20 hours ago

My favorite bit is that into their effects code they put a bunch of logic about "affects".

I haven't looked at how they're using this term, but PureScript has an Aff type for describing async effects; maybe that's the precedent?

tentacloids | 18 hours ago

The docs for that use "effect" (noun) quite generously. I'd guess it's a portmanteau of async + effects.

What I see of affects in solid docs makes me think it's the cutesy declarative "fluent" verbiage popular in test frameworks. There's also ample mention of effect elsewhere.

altano | 13 hours ago

I must be missing something because that linked doc is the correct use of the verb “affects.” The subject is implicitly “this function.”

tentacloids | 3 hours ago

Yeah, you're not missing anything, it's all grammatically correct, I was just saying I don't think Aff is the precedent. However, "this function" is implicitly the subject (ish) of literally all statements in imperative programming, so it should just be "affect" (verb). I would hazard that that incorrect mood is part of the original issue.

altano | 3 hours ago

You’re missing that this function doesn’t do anything itself, it’s marking x as affected. So if you fetch something imperatively, you say “this.fetch(x)” but if you are just recording the fetch you say “this.fetches(x)”. It feels correct as-is to me.

The comments here seems more intent on nitpicking correct grammar than they do discussing the technical aspects of this major release of Solid. Kinda sad what’s become of this place.

tentacloids | 3 hours ago

Yeah but look at the example, it's not going subject.verbs(object)

const reload = action(function* () {
  affects(thing); // `thing` and its derivations pend…
  refresh(thing); // …over this otherwise-quiet re-ask
  yield api.done();
});

Indicative followed by imperative. It's inconsistent, that's all.

Kinda sad what’s become of this place.

ok??? I'm literally just trying to resolve a misunderstanding for everyone's benefit?

"But, we aren't writing the code anymore!" We hear you.

People really say this?

I'm so sad. I clicked this article expecting something fun and exciting about a framework I've been hoping to use in a future project, and instead I've discovered that I'm going to have to find something else. I don't think I really enjoy being a developer anymore…

srcrip | 7 hours ago

Why is it lost to the slop just cause they care about making it work well with AI? Are you saying they’re also developing it with AI or is this must about the feature in the article? I don’t think it’s realistic to expect developers not to care about supporting AI coding in their tools. It just seems to me like you can still choose to write it by hand. It’s an amazing framework with amazing ergonomics, written for humans.

That being said if there’s some proof of AI slop in the solid codebase I understand your complaint but I hadn’t seen it.

conartist6 | 2 hours ago

No, I'm referring to what I found when I opened the signals code to read it.

Ryan is using Cursor to write it and there is about 10 lines of low value slop comments on top of each and every method

altano | 17 hours ago

I’m glad to see the unbearable slop prose of previous communication is now just annoyingly corrected prose. Genuinely a huge improvement.

If you're sensitive to bundle sizes (which you almost certainly aren't), a heads up: the Solid 2 bundles are much bigger. My counter component bundle test is:

 4.8 KiB (brotli'd) in Solid v1.9.18
20.3 KiB (brotli'd) in Solid v2.0.0

vincent-uden | an hour ago

Agreed. For some reason Ryan has had quite a strange relationship with LLMs. I used to enjoy his writing a lot but he somehow became convinced that his writing was bad. I remember him talking about this quite a lot on streams a while back (maybe about a year). He tried posting some LLM-written/edited articles which just read as total slop. I would every single day of the week rather read a grammatically bad article from someone so technically competent as Ryan than to perceive him through some LLM filter which smooths out the content, both its qualitative peaks and valleys.

Similarly I'm surprised this is the direction his development has gone too. He was hesitant and slow to AI-adoption for quite a while. I wasn't expecting a swing over to the other side due to that.

liberty | 19 hours ago

I initially thought this was Tim Berners-Lee's Solid. Disappointing that it's another Web framework, especially one that's repositioning itself to be used by LLMs.