Awesome D: Curated list of Dlang documents, frameworks, libraries and software | Hacker News Reader

Awesome D: Curated list of Dlang documents, frameworks, libraries and software

github.com

19 pointsteleforce11 comments

I’ve been seeing a lot of D stuff lately. Is there a reason for it, or is it just the cycle of someone posts a D article, which causes someone else to get interested in D, which leads to another article post, repeat?

Could be a Baader-Meinhof phenomenon?

I think one or two D-related articles get upvoted on HN per month pretty consistently for as long as I can remember. (Usually full of comments like "What a shame.", "This language could have been great." That sort of thing.)

mostly i see these mentions in threads about zig, odin and vale langs (might be others, don't watch that closely), where people look at C, C++, golang arena and dive into pre 1.0 releases and then speculate about these langs getting marquee apps, funding, industry penetration in e.g. gaming, crypto, ML or trading and staying power.

Before Java was announced to the world, DDJ had an issue about successor programing languages to C and C++, they were Actor (Smalltalk inspired), C@+ (talk about naming things), Eiffel, and a couple others I cannot remember.

None of them survived, instead replaced by the "All must be Java" hype of the 2000's.

While we keep trying to get an established successor to C and C++, 30 years later.

It is not that we can't but we won't. Rust and D are two of the best shots so far for replacing C++, and based on the TIOBE 2024 Index, they are on the 19th and 21st place respectively. But time will tell whether any of these promising languages will eventually succeed, and become C++ successor that is currently ranked 3rd. Personally I'd love D to success thus we can replace a single language for the cumbersome and the mess of the C++ and Python programming languages duality in data science and machine learning implementation for examples.

And still, as long as Rust depends on GCC and LLVM for its reference implementations, that alone will be enough to keep C++ around.

Same applies to D, as dmd's optimizer isn't as rich, nor its backend supports as many CPUs, as do ldc and gdc, built on top of LLVM and GCC respectively.

It's a bit rich complaining on D dmd optimization when D compilation speed performance is much faster than both C++ and Rust. The backend supports could be better but I'm afraid it's a catch-22 situation until D is more widely used. Like I mentioned before for Data Science and Machine Learning eco-system, system programmers (B-type programmer) generally won't replace C++ not that they can't, while on the other hand analyst programmers (A-type) are avoiding C++ like a plague and prefer user friendly programming languages for examples Python and Matlab [1].

[1] There are two types of data scientists — and two types of problems to solve:

https://medium.com/@jamesdensmore/there-are-two-types-of-dat...

A few weeks ago, some D developers decided to fork the language/runtime/stdlib into OpenD with the goal of fixing some deficits in the language. I guess this spurred some new interest in D from new and older (ex-) users to look again into it. See https://news.ycombinator.com/item?id=38986546

I'm an avid user of Rust. What things does D bring to the table that Rust does not? My uninformed guess is:

- smaller binary sizes

- faster compile times

- easier learning curve

That's not a very compelling list for my use cases. Perhaps I am missing some killer features of D.

Not dealing with the borrow checker, unless it really matters, enjoying the productivity of a tracing GC.

Then take advantage of @nogc and simplified ownership rules to optimize that piece that actually makes a dent on the profiling report.

Machine learning system implementation is hard, and according to this article it is 4D rather 2D in approach [1]. It is always helpful and much easier to perform debugging on a single uniform programming language rather the cumbersome and the mess of the popular Python and C++ duality in current implementation approach.

[1]Why is machine learning 'hard'? (2016)

https://news.ycombinator.com/item?id=39109481

Reply on news.ycombinator.com