I like D, but it's a bit of a mess at the moment with the DIP-1000 (Rust-inspired stuff for memory management) half-baked changes, the new experimental allocators (probably coming from Zig), it keeps breaking (on MacOS at least). With the very advanced metaprogramming, it also suffers from lack of good tooling, just like Nim... Nim was going to "solve" that a few years ago, but as far as I can see from the outside, nothing really "shipped" recently to improve the situation. Without good tooling, I am quite sure I am more productive even in Rust (and I am no expert on it), let alone OCaml (tooling is decent, and the language is just really powerful) and Dart (which has absolutely awesome tooling), for example.
I wanted to check those languages at some point, didn't have time, but I doubt they offer a performance for the stuff I am doing. It is also another reason I discard Java and Go (C# seems slightly better than those).
I deal primarily with scientific/engineering calculations, mostly ODE simulations, so I have some demands regarding performance. Python is of course important, but not for the core production code. Fortran is nice for math stuff, but not much else. C/C++ are as performant as awful. Rust is viable, has nice tooling and ecosystem, but introduces too much hassle that is not that important in my domain. Julia has nice ideas, but poor execution, low quality tooling, AOT delegated to some package that may or may not work with your code etc.
I played with D and it strikes a nice balance. With LDC and careful allocation strategies it can be as performant as Rust, with DMD it compiles much faster, and outside performance-critical code parts it is more ergonomic because of GC. Tooling is decent. Because contrary to sentiment popular on this website I actually think Visual Studio is still the best IDE out there, I am glad that Visual-D exists.
My main issue with D is exactly that - it didn't take off. As a consequence the knowledge-base is small, a lot of things are poorly documented etc. With Nim it is similar, but worse.
I ask because I've seen Dart beat Java/Go/D (very high memory throughput, no idea what Dart is doing differently but it's amazingly good at it)... but I can imagine that if you only perform numerical computation, Dart may miss optimisations that will make it slower (normally, it will be optimised to approximately the same as you'd written in C, but in some cases it can hit the "slow path" - those are getting less and less every release, and if you provide an example the Dart team would likely be happy to fix that).
Some very good frameworks/service to look at for backend development:
* Serverpod - https://serverpod.dev/
* DartFrog - https://dart-frog.dev/
* DartStream - https://dartstream-prod.frb.io/
* Appwrite - https://appwrite.io/docs/quick-starts/dart
It's just a really good language to work on, try the tooling and you'll see it for yourself (either with Flutter or just Dart):
type StatusCodes int
const (
Success StatusCodes = iota
Ongoing
Done
)
enum StatusCodes {Success, Ongoing, Done}
type
StatusCodes = (Success, Ongoing, Done)the last thing one can say about Go is that it is expressive