Gil Tayar

What did I miss? What is `signal.addAbortCallback()`?

Probably. But I'm using Neon as my postgres db so I don't know. And for my use, neon is free!

iirc, neon does the same thing for the free tier - shuts it down when not in use.

But in both cases, the services go up almost instantly!

It's not server oriented. It's server less. I have one service and I pay 5$ but that's only because it's the minimum amount. Railway says the actual cowt is 1$.

But railway shuts it down when not in use so I don't know how much real time it is using. It does restart very quickly. <1s.

You pay for compute and ram, but only when used. So if more and more clients are not a significant load on your cpu, then you should be good.

I'm using @railway.com and loving it!

In this case, `expect.poll` isn't really necessary because you could just write `await expect(page.getByRole(...)).toBeVisisble()`.

`expect.poll` is useful in cases where you want to retry an expectation where what you're testing is not a locator.

@amitsheen.bsky.social please!

Also, there is _no way_ to have a variable that is scoped to only the `if`, whereas there is a way to have a variable that is scoped to both if and else.

So I would go with scoping only to the if also because of this - it's the only way to do it!

I think most developers would be _surprised_ to know that it's scoped to the else too.

Yes, thinking about it a bit more, they would agree that it _has its uses_, but they'd still be surprised and lean toward it not being scoped to the else.

The majority of the answers here be like that.

I have a web app on their platform, which is not used a lot, and my usage is around $1 a month. I still pay 5$ a month because that's the minimum. This is perfect for a hobby app or (in my case) a backoffice CRM for my partner.

I am LOVING the UI and easy setup, and everything works as advertised!

You should be proud!

Looks like you've been slashdotted! (or whatever the modern equivalent is) - the blog post is stuck on loading. 🥲

And this still doesn't take care of chunking, which would still be a manual process. While I can't think of a solution that makes the browser take care of downloading only what's needed. I wish there was such a thing.

Not just try entrypoints - the whole module graph. That's a lot of text. It could be thousands of modules.

And I seem to remember an alternative proposal that had some kind of bundle that had a directory of files up front for the browser to use...? Tried looking it uo and couldn't find it.

I never understood why web bundles were abandoned. They seemed like the perfect solution to enable browsers to deal with when to load modules instead of having developers and bundlers deal (mostly badly) with chunking.

I believe @yoav.ws would have LOTs more to say on this subject than I can.

What Maestro said is true - we don't want that overhead for small projects. But even for bundlers - after they do DCE, I believe even there job would be less complex if they didn't have to take care of chunking and just supplied the browser with the bundle and let the b take care of loading.

What happened to the alternative suggestion that didn't use a new display type but rather created new properties under the grid display? If I remember correctly it was suggested by Apple.

Interesting! It's why I found the move from Java and Scala to JavaScript so refreshing (TypeScript was in its early stages then).

I was wondering what discussions in the TS community you were referring to.

Oooh! You work at Neon?! Yay! I am just trying out the Neon db for a home project of mine, and so far it's working nicely and the UX is really nice!

Battlestar Galactica

Because you were born in the wrong generation 😉

1 for frontend projects
4 for backend

What about frameworks like htmx, unpoly, turbolink? Not sure that they constitute "HTML reuse strategies" but usually when those other options are mentioned (e.g. PHP, React, Web Compnnents), then so are they.

(The importance of preserving import execution is why I don't like how Vitest and Jest hoist some code above the imports, but that's another discussion entirely 😁)

While I agree with @rich-harris.dev that relying on that order is VERY problematic, I still think that being spec-compliant in this fundamental issue is important.

I have to agree with Devon here. The ES spec around ESM is very clear about execution order, and bundlers should preserve that order so that they are compliant around that, and are compliant with all the runtimes that are compliant in that regard.

Yes! That is exactly how I use it.

Thanks, everybody!

I'm writing a "TypeScript in Node.js" talk for NodeTLV (assuming it happens 😢) and this helped me a lot.