I've been building a memory safe language that transpiles to Zig with a Go-like runtime that can run interpreted (no GC) or compiled - high-level that feels like Ruby but with incremental typing like TypeScript.
The Zig team between 0.16 and this has really made me glad I chose Zig as the target instead of Rust - which probably would've been a lot easier to target (since it's already memory safe).
I believed it had the best build system design and was the best transpilation target, and I really believe that 6 months later.
The main reason I wanted no GC is because I think aliasing is the root of all evil, and I want a language with zero global complexity (but doesn't require a PhD to use).
Working on something kinda similar. No GC, Python feel, managed memory, performance approaching C. It's here: https://blorp-lang.org if you want to compare approaches.
Yeah, concurrency in blorp doesn't allow shared mutable references, so deadlocks aren't really a concern. Otherwise it's meant to be simple-ish -- virtual threads, channels, no async/await. Pure functions allow safe parallelism naturally, so that's fairly straightforward, though the API is still incomplete, for example the "Parallel" section here: https://blorp-lang.org/docs/lists/. It's still under heavy development (working on it right now).
1) I want to minimize global complexity, which by definition maximizes local reasoning.
2) I want to make the vast majority of bugs simply unrepresentable - taking it past Rust, and even past Pony - WHILE allowing shared mutable memory, but without requiring a PhD to use.
The goal is in EASY mode, it's barely harder to use than Ruby or Python (just the occasional pedant compiler error that has automatic options to fix itself most of the time). You don't even have to supply types or compile. It has a REPL, etc.
When you bump to DEFAULT mode and then to STRICT mode, all the annotation is automatic - your code just might look "ugly" if you like having no types anywhere etc.
But DEFAULT & STRICT mode give people and LLMs everything they need to know to understand the effects of an individual function.
I have some similar goals. Have you considered leaning more into inference than gradual typing? One pattern I like is allowing the compiler to develop a more complex mental model, but keeping it straightforward for users -- you can do that with inference, ownership, purity, effect types, etc. What I actually think is really tantalizing is using tooling to fill in some of those gaps -- for instance, the editor could know types, required capabilities etc, without the user ever needing to type anything, but when the user needs it, they can find it, query it, test against it.
Cool - it sounds pretty similar. It's interesting that it looks so different. I'll have to investigate more.
WRT to inference, yes. I infer everything in EASY mode.
And the compiler give the user autofix via choice when a type is ambiguous (I don't default to huge union types - I assume no one wants to do that and make them choose a type - may potentially allow AutoUnion to allow that).
I couldn't tell if you're using affine ownership, but I assume so if you don't have a GC. If they try to create an alias - they get a use after move error, and the compiler tells them they need to either COPY (auto-fix) or create a RefCount (usually auto-fix) - they pick.
No, blorp doesn't use affine types (one or zero uses). In blorp, ownership is not explicitly controlled by users at all, so it's opaque. Under the hood, it's perceus for compile-time ownership and borrowing and automatic reference counting with copy-on-write optimizations for the runtime. This is made reasonably easy for the compiler to reason about in blorp because semantically it doesn't _really_ have in-place mutation -- `var` really means "re-bindable" to a new value; and then under the hood we'll mutate in place where we can.
If you want some feedback, I'd recommend putting some concurrent benchmarks on your home page (and if you have some already - I'd clearly separate them).
When I originally scanned it, I just assumed this was another predominately sequential language with no good concurrency story.
If you're actually competitive with Rust/Tokio/Crossbeam and Go on the most common concurrent patterns, then you've got a really compelling project!
I suspect if you don't cherry-pick benchmarks, you're going to run into some performance problems with not allowing shared mutable memory - though maybe you can avoid most of that if you have some type of built-in actor pattern.
But if you're actually competitive across the board with Go & Rust/Tokio/Crossbeam - I'd love to take a deeper look, because that is NOT easy to accomplish.
I didn't see any of that from a cursory look at the language, though.
Still incomplete there. But yeah performance should be decent. Not aiming for parity either rust or go entirely but in the same ball park is the expectation.
How is compiling to zig? I considered doing it but chose C instead because of how much zig is still changing. I've considered using it's C compiler for targeting multiple platforms locally.
Comptime is very powerful and awesome to use for a transpilation target.
Also, Io and how Zig handles allocation and gives you full control of the allocator make it super easy to do things you probably want to do in your language.
The Zig team between 0.16 and this has really made me glad I chose Zig as the target instead of Rust - which probably would've been a lot easier to target (since it's already memory safe).
I believed it had the best build system design and was the best transpilation target, and I really believe that 6 months later.
The main reason I wanted no GC is because I think aliasing is the root of all evil, and I want a language with zero global complexity (but doesn't require a PhD to use).