I have started using Nim fairly recently (past month). I actually wanted to start using it about a year back, but it never happened. But my return to Nim was a good step; this time I might be here for a while as I have started to understand the nitty-gritty details of the language [1].
Couple of reasons why I think Nim is an awesome language:
1. Syntax: It's beautiful. If you like Python syntax, you'll like this. On the other hand, it's much more concise, without losing readability.
2. Static type check, with a sprinkle of type inference only when assigning the initial value.
3. Proc/operator overloading: Have procedures (functions) named the same handle different data types differently. And the same with operators.
4. Support: Great support over Gitter, GitHub issues, Nim Forum, etc. Great devs! (though, I incorrectly thought earlier that that wasn't the case :))
5. Makes me feel home: Hacking/coding in Nim comes the closest to coding in Emacs Lisp. In Emacs, if I want to understand what a function does, I look at its source code. I can do the same in Nim.
6. Editor support: The nim-mode Emacs major mode and ob-nim Org Babel package are awesome!
7. Documentation: Check out the Nim the index.html to search anything, and also Devdocs.
8. Whitespace: Tab not allowed, yay!
9. Not as easy as installing go/rust toolchains locally (without admin rights), but not difficult either.
10. The nimble tool (equivalent of pip) works great for building/distributing Nim projects.
.. and much more! Computer science experts can talk about more benefits of Nim from systems programming point of view (GC, etc.).
For my use case, where I use Nim as Python/Bash replacement, CLI utility development, it looks excellent so far.
I chose to learn nim coming from Python, because found C++ to be annoying and difficult. It feels like python but faster. I have compiled to to JS and integrated with complex C++ libraries. I like it so far. Metaprogramming is cool feature. Community is also really helpful.
It's hard to pick between those two in particular. Nim is more fluid syntactically; D is more precise. D is also older and I would generally consider it more mature.
(After trying both, I went with D. But in the future I want to use Zig for some of the lowest level code.)
What I want in a language of this class is some ability to transition code from a slow/dynamic/generalized/exploratory phase to fast/static/specialized/checked. Both D and Nim offer features of that nature. (Something like Rust omits itself by forcing memory into a central role, leading to a lot of annotation boilerplate - it's correct for writing Servo or similarly complex parallelized systems, but I am not writing Servo.)
When I left Nim, it was because I had serious concerns about the project being able to tame its rapid accumulation of features. I had no faith in all the things it advertised to actually work correctly. The dev process I witnessed was "cave in to toxic people who whinged a lot." D has a lot of warts from being around longer, but in the end, what I want to write code in anger with is something that can boast a well-engineered compiler, and D has not one compiler but more than one.
That was years ago. To win me back you'll have to market a demonstration of similar stability.
Zig is in a different category from any other option because its design allows me to tackle a piece of low-hanging fruit: a modernization of old C libraries that does not reinvent the universe, create a new ABI, introduce a runtime, or otherwise disrupt the original context. I can write basically the same kind of code in the same kind of style, but the language automatically pushes my attention to more edge cases and undefined conditions, and the compiler simplifies the build process, removing the preprocessor noise that tends to appear in substantial libraries.
That's fair. Can you name features in Nim that you feel are unnecessary? I have seen a similar sentiment in the past but rarely do people actually give examples of what they think could be removed (the biggest challenge for us right now is the removal of features in preparation for 1.0).
How's the GC in D like recently? I tried to use D around 5 years ago, and the key showstopper was its poor GC implementation. Even worse, it was practically impossible to disable GC then.
> Learning a very different language is fun and useful.
Yes, that's one of my main motivations.
I'm comfortable enough with Python, Java (comfortable doesn't mean I enjoy this one) and C.
> Do you have projects in mind?
Yes, both are solely for the purpose of enrichment of my current set of skills and knowledge:
- Source control system that is a "lite" version of git.
- A bit torrent client.
I'm looking for a well-rounded language, which is versatile in the same direction as, e.g., Python is.
Ideally, it would also enable expanding one's own set of programming skills and knowledge/practice of interesting paradigms (i.e., a mind-shift of sorts).
Nice to have's:
- a rich library ecosystem (like Python and Go, for instance)
- as close to C as possible, in terms of speed
I've been looking for something similar to Python and Go as well. I can never seem to get the performance from Python that the great libraries do. Go is great for simple types but I haven't quite got good patterns for larger projects and run into performance issues with channels. Pony is where my current search and learning is at. I'll queue up Nim as another possibility.
Nim is like a fast version of Python and Crystal is like a fast version of Ruby. Rust and Go also have some use in the same space, but they all differ in the details as in whether the GC exists and is enabled by default, is optional, or doesn't exist.
Coming from Python, I found Nim to be the best for my mild systems programming needs. I haven't tried Crystal as it doesn't work on Windows (last I checked). Go is too simple/verbose and doesn't have enough language constructs by design. Rust seems like it is changing too much and too low level. I really wanted to like D, but had trouble getting up to speed in it. I bet it's much easier if you come from C++.
Apart from some syntactic commonalities and expressiveness, Nim doesn't have a lot in common with Python.
The idiomatic approach to solving problems in Python and Nim are different. I came from a strong Python background and now I prefer Nim to Python as a general purpose language and especially for writing scripts.
Nim is highly productive and with excellent type safety.
The compiler is relatively fast though maybe not as fast as D and certainly not as fast as Go. The compiler is a lot faster than Rust/Crystal/Pony and there is current work being done on incremental builds which should make it even faster.
If you take a look at these benchmarks the Nim and D fastest versions are faster than Rust and on par with the fastest C++.
Significantly, the fastest D version has double the lines of code as the fastest garbage collected version of Nim though the Nim version performs slightly faster.
The fastest garbage collected version of Nim has exactly the same lines of code as Python.
> Apart from some syntactic commonalities and expressiveness, Nim doesn't have a lot in common with Python.
> The idiomatic approach to solving problems in Python and Nim are different
On a practical level, they are close enough to let you convert Python code to Nim with reformatting and replacing instead of doing a complete rewrite.
Thank you, this is what I meant when I said similar to Python. Someone with Python familiarity should have zero issues with writing a Nim program and will mostly feel comfortable with the syntax.
Couple of reasons why I think Nim is an awesome language:
1. Syntax: It's beautiful. If you like Python syntax, you'll like this. On the other hand, it's much more concise, without losing readability.
2. Static type check, with a sprinkle of type inference only when assigning the initial value.
3. Proc/operator overloading: Have procedures (functions) named the same handle different data types differently. And the same with operators.
4. Support: Great support over Gitter, GitHub issues, Nim Forum, etc. Great devs! (though, I incorrectly thought earlier that that wasn't the case :))
5. Makes me feel home: Hacking/coding in Nim comes the closest to coding in Emacs Lisp. In Emacs, if I want to understand what a function does, I look at its source code. I can do the same in Nim.
6. Editor support: The nim-mode Emacs major mode and ob-nim Org Babel package are awesome!
7. Documentation: Check out the Nim the index.html to search anything, and also Devdocs.
8. Whitespace: Tab not allowed, yay!
9. Not as easy as installing go/rust toolchains locally (without admin rights), but not difficult either.
10. The nimble tool (equivalent of pip) works great for building/distributing Nim projects.
.. and much more! Computer science experts can talk about more benefits of Nim from systems programming point of view (GC, etc.).
For my use case, where I use Nim as Python/Bash replacement, CLI utility development, it looks excellent so far.
In plain terms, it's joy coding in Nim :)
[1]: https://scripter.co/notes/nim/