Andrew Branch

@mattpocockuk @KaziAhmedDev I think this was @andhaveaniceday, but I was under the impression that those short descriptions weren't actually used in the generated HTML

@KaziAhmedDev lolol cc @atcb was this you?

When it's your last day and you just give up on writing docs. 🤦‍♂️ This is why we need @mattpocockuk #typescript

@ssalbdivad @Malix_Labs @arktypeio @andhaveaniceday Yeah, that’s right.

@Malix_Labs @arktypeio Main takeaway is obviously 500% is still a huge improvement, but would be curious if that matches the team's intuitions as well CC @andhaveaniceday @atcb

We tested TS Go with ArkType's stress tests: 📈5x faster 📉30% less memory ✅99% correct Even the craziest types in our repo were perfect on day one🤯

@saltyAom Awesome! Tip: try "TypeScript Go: Restart Language Server" instead of "Reload Window" to see the squiggles appear even faster :)

I test it and it's real Here's what I did 1. Create 1,000 endpoints with body + response validation 2. Infers schema to type and check if return type is correct 3. It happens in order for 1,000 time 4. See how long it takes for 1,001th route 5. Reload IDE ts: 3.35s ts-go: 0.63s

The new TypeScript is *really* fast

Today we're thrilled to announce our effort to port the TypeScript compiler and language service to native code, gaining a 10x speed boost in build times and editor responsiveness! devblogs.microsoft.com/typescript/typ…

Node.js v23.6.0 is out! 🥳🎉 This release enables the flag --experimental-strip-types by default. Node.js will execute TypeScript files without additional configuration! ⚠️ This feature is experimental, has limitations and is subject to change. nodejs.org/en/blog/releas…

@ballingt In those cases, you kind of _want_ to have duplicated and unassignable TS types, because that accurately describes the situation. But for simple functions and values, things that aren't meant to be used nominally, there's no issue.

@ballingt We do this on DT a lot because it's usually just not worth actually duplicating the .d.ts content. The only caveat is if the ESM/CJS JS actually export things that exhibit bad dual-package behavior, like classes that will be checked with `instanceof` or duplicate Symbol()

@atcb I have the TypeScript version of Dual Package hazard: I have some types that can't be assigned structurally (usually recursive types) so I'd like to ensure only a single set of types is used.

@ballingt But I mean, for some import that results in TS loading `a/dist/esm-types`, is the same import meaningful at runtime, e.g. will it load a value in Node.js? In other words, do any of the types describe the exports of a JS module, or is it purely a collection of types?

@atcb yeah a/dist/esm-types/ is a types-only directory pointed to from a "types" in package.json, although I can change the structure or do whatever it takes to use the same .d.ts files on disk for ESM and CJS types, hoping that that way those types would be assignable to each other

@ballingt So this is a package entrypoint that has no corresponding JS? (Or empty stubs for JS?)

Any suggested reading on this? Since it's my library I can guarantee these types are the same, I'm avoiding default exports etc. I know you've usefully crusaded against this and were always contenting with people saying they wanted to do this. Any reading on potential issues?

@mitschabaude But Node doesn't do type stripping inside node_modules, so if you're writing a library, you're in this situation where you can run TS during development, but you need to publish JS files. That makes it impossible to satisfy our rule of "write the specifier that works at runtime."

@mitschabaude We already had support for that via --allowImportingTsExtensions, but you were forced into --noEmit with that. That felt good enough for folks targeting Bun and bundlers, because for those scenarios, you never really need to emit .js files at all.

#TypeScript team just merged a PR that brings a beautiful consolidation to the JS ecosystem: We can finally use .ts extensions when importing .ts files 😀🥳 github.com/microsoft/Type… Thanks @atcb!

@bradleymeck @devongovett This is possibly the thing I most wish people understood about TypeScript

@devongovett Conditions are the same as TSConfig, flags for daaaaays since people want tool interop but tools don't pin to non-magic stuff so you have to make config options to fit other tools magic

Oh no, ANOTHER exports condition?? 😩💀

@wooorm @robpalmer2 @remcohaszing @ssalbdivad @johnny_reilly Yeah. Put another way: compile .ts and/or .js files, publish .js and .d.ts files. If you’re skipping the compilation step, you’re authoring .js and .d.ts.

Finding that working with JSDoc is made massively more fun by defining any non trivial type code in an `types.d.ts` TypeScript file which can then be imported into the JSDoc. Your codebase is still 100% JS for running code, but the type syntax in TS is much easier than in JSDoc

@johnny_reilly @mattpocockuk 🤩

@mattpocockuk Awesome to see mention of @atcb's Are The Types Wrong - such a handy tool! arethetypeswrong.github.io

New NPM publishing guide is MASSIVE (hitting publish next week)

Editing with @britto_gab's work on region-prioritized diagnostics has been so great: > in our testing on TypeScript’s own checker.ts, a full semantic diagnostics response took 3330ms. In contrast, the response for the first region-based diagnostics response took 143ms! 👏

@ballingt That’s so kind, hope you’re having a good summer too!

@atcb I'm missing your content on Twitter, I know you probably have a another job besides tweeting but I hope you're having a great summer!

Dual publishing ESM and CJS is hard. Getting the types incorrect in your package.json is disappointingly easy. If you want something to help you with both, then maybe tsup and Are the Types Wrong can help. Thanks to @atcb for his assistance with this ❤️🌻 johnnyreilly.com/dual-publishin…