Simon

stop doomscrolling. put down your phone and take a doomwalk outside. doomdrink some water. doombreathe.

crates.io: Rust Package Registry

Ever found yourself allocating altogether too many strings, just to get human-friendly output? crates.io/crates/liter...

*linker flags

Ah right. Yeah, the only way I’ve gotten that to work has been by severing the other dependency. It *should* work to have two shim crates both reexporting all symbols, then use some custom line flags like `whole-archive` etc to actually include everything in the static variant. Maybe you tried this?

Normally I would say it smells like an X/Y problem, but since it’s you, I’m curious… what linker crimes are you up to?

But if you do it anyway, I salute you for not using AI to “polish” your work — it makes it much easier for other non-natives to detect that you are human, and therefore may have interesting things to say.

As a non-native casual user of the English language, I feel like I can probably say this: If communication and writing are not skills you have or want to practice, how about… not writing blog posts? What is this obsession with creating an online portfolio of mediocre “articles”?

`@include` …??? 👀

Very early work in progress atm, but I’m optimistic!

Here’s a wild thought: if facet provides a way to invoke arbitrary function pointers, this could even automatically wrap them so they’re callable from JS. 🤯

A very thoughtful post

The only thing I miss from C++ is the drama.

Oh, and specialization.

The obvious workaround is to keep using good old em-dash, but introduce a light sprinkling of spelling errors or non-idiomatic phrasing.

Non-native speakers finally have an advantage! The Uncanny Valley of language is our only hope.

WASM is so close to being actually good. Just give me any kind of dynamic linking. Please. 🥺

Passing data structures between Rust and V8.

Mine is named werk.

Hard agree 💖

Been trying out unsynn for my own projects. I have to say, it’s actually quite pleasant, but I really wish the library would commit to a bit more stability. The current design seems fine.

Most hated by people who actually use it: C++
Most hated by people who don’t use it: Rust
Least hated by people who actually use it: Rust
Least hated by people who don’t use it: C++

I miss when this was what America was about.

I made a new ultra-efficient string interning library for fun and profit. Pretty proud of the amount of magic here, not gonna lie.
simonask.github.io/introducing-...

c.tenor.com/yttVK0I9LaIA...

I think the point isn’t “this was possible for all”, but rather “this was possible at all”.

Build scripts and proc macros are not rebuilt on every check. R-a today caches them quite aggressively (and often actually needs a full restart when they change, at least in VS Code). They do need to *run* though, and that’s a good point.

More generally, LSP servers can be structured very differently from a compiler (and r-a is, to my knowledge), and can choose very different tradeoffs, because interactivity is more important, and passing stuff to codegen is a non-issue.

I mean, for sure there’s a lot of overlap, but codegen is very, very slow (kind of intrinsically).

I’m willing to bet that these issues would be way less important to people if rust-analyzer was actually fast.

I ended up going with the classic “MIT OR Apache-2.0” 🙂

There’s probably a niche for a less ambitious rand crate, but should anybody use it?

Thought: If the goal was stated as “generate a pseudo-random number”, I would agree. But it does a lot more: crypto-suitability, max performance, choices of distributions… hard things to get right. And still not suitable for something like games.

Thanks! Yeah, I’ve been debating it, so it’s kind of “intentionally left blank” at the moment. Since it is a tool, not a library, I’m considering something a little less permissive, but I’m still not sure.