Examples:
* Python preferred over Ruby
* TypeScript preferred over Dart or even JavaScript (which is fine and, as a bonus, doesn't require compilation step like TS)
* Go is preferred over Crystal and D.
While Python, TypeScript and Go are quite alright, there is no doubt in my mind that their alternatives are absolutely superior as languages. Yes, in case of Dart, Crystal and D the ecosystem doesn't have the abundance of well-tested libraries, but as languages they are simply better. The Go argument that it's popular because it's simpler is absurd in the sense that no one really forces you to write complex code and use classes or other advanced OOP features in D.I actually liked D very much, and WB had been a personal hero of mine when I was in college. But I am not betting my career on an ecosystem built around by a single brilliant guy. For high-stakes projects, a wise decision is building on a platform with several deep-pocketed backers.
And for toy/personal projects... do you even need a language anymore? Just ask your favorite LLM to generate you an executable which does what you want (partially joking here).
It's not perfect, as some people cannot resist using the C preprocessor for some bizarre constructions.
I used to write those bizarre things myself in C, and was proud of my work. But one day I decided to remove them all, and the code was better.
To the parent's point of startups, betting the farm on something like a particular language out of some sense of superiority might mean you are not focusing on the right problems. But if the founders happen to know a less widely used tool it doesn't seem inappropriate either. The type of employee that can drive a startup or a big tech project forward is not going to be thwarted by a language, and they might find something new to learn fun.
For C# Microsoft eventually embraced NuGet their package manager, and used it to put core packages that don't need to be fully available OOTB but can then benefit from frequent updates on a per project basis as opposed to updating the entire language runtime.
For Go it was the out of the box packages, like if I want to make a website, I can pull in net/http and their templating packages that come out of the box with Go, maybe a reasonably simple core maintainer package or packages that go into Dub would be a strong selling point. Right now Vibe.d is the only option for web dev, but there's no reason a much simpler web server couldn't exist.
For Rust, I just really love Cargo, I think its one of the nicest package managers I've ever used.
The other thing that would really help D is if something significant is built around D, whether it be a framework (like what Rails did for Ruby) or some major application that needs D to function at its core and is used by many, this could be a groundbreaking modern IDE, or anything really, a database that uses the best bits and pieces of D to scale, or even a really rich cross-platform GUI stack (my kingdom for std.gui to be a thing in D, and reasonably exhaustive).
I wish I had unlimited time and money, I would invest it in D. Alas, I'm not a language maintainer, just a guy who loves really good tools.
Maybe it exists and I am just ignorant but it doesn't seem to be in the list of supported languages.
... Perhaps what you're describing is having a niche opinion. If you had some opinions, like a preference for "Everything must be done in as many ways as possible with funky characters" or "I hate indentation", it would certainly seem that the world is against you. But, perhaps, you just really smart and can remember the intention of all the complicated code you wrote a year ago, so you don't even need to write comments, and thats great. However, being special does not mean that some technolgoy is objectively inferior, unless you can actually come up with a provably objective metric.
Overall, the technology that is there, solving most of the problems for most of the cases is the technology that is superior, by the law of the universe, not the other way around.
I don't agree with any of your examples, but I have my own, like Pascal is a better language than C, by many metrics. I can also accept that C, is what people who invented unix, also invented. And that makes Pascal inferior to C, now, as choice for any project that requires that you hire embedded software developers. That's what the universe decided.
Are you sure you're not talking about Perl here? Because there are very few "funky characters" in Ruby and code written in it tends to be very readable, more so than Python in many cases.
I agree with OP. While Python is not a bad language, Ruby is a better language in general, and I'm reminded of it every time I have to work in Python (which is pretty often nowadays, due to Python's dominance in ML ).
I can give many examples as to why, but here's a quick toy example to show off the difference in philosophy between Ruby and Python. You have a list of numbers, you want to keep only odd numbers, sort them, and convert them to a string where each number is separated by comma. Here's the Ruby code:
xs = [12, 3, 5, 8, 7, 10, 1, 4]
ys = xs.filter { |x| x.odd? }.sort.join(", ")
Now let's do the same in Python: xs = [12, 3, 5, 8, 7, 10, 1, 4]
ys = [x for x in xs if x % 2 != 0]
ys.sort()
ys = ", ".join(str(y) for y in ys)
Or alternatively in a single line (but in more complex cases this gets extremely unwieldy since, unlike Ruby, Python forces you to nest these, so you can't write nice pipelines that read top-to-bottom and left-to-right): xs = [12, 3, 5, 8, 7, 10, 1, 4]
ys = ", ".join(str(y) for y in sorted(x for x in xs if x % 2 != 0))
And this matches my experience pretty well. Things I can do in Ruby in one line usually take two or three lines in Python, are more awkward to write and are less readable.This all makes sense to a person who knows the language a lot, but wrinkles the brain of a newcomer. Too many concepts to juggle in order to understand. On the other hand, the Python one reads quite easily, even if you may have to go right to left.
Questions such as
> why does it not have parentheses around it but the `", "` passed to `join` does?
would be exactly the same for JavaScript, Go or D. Ruby has the best syntax with regards to blocks/lambdas/closures.
I don't understand this argument. You are a beginner only for a tiny fraction of your time using a given programming language. Why are we optimizing a programming language for people who don't know it, instead of optimizing it for people who actually program in it?
D is safer and more productive. It's a joy to write in D because most of the time it feels like whatever you think, you code. This is unlike C++ where you fight the language all the time. C++ is not a productive programming language. I say this with experience: I coded in C++ as an "expert" for many many years, including these last couple of years. It's not fun to write in C++, which translates to another kind of loss of productivity.
C++ is a burden and liability for companies but no CTO will be blamed for chosing it because it's popular. I can list so many popular things and persons that worth nothing but I will refrain from getting political.
Yes, on paper, there are way more C++ programmers out there than D programmers. But I interview these C++ programmers occasionally. Most of them don't even have an inkling that they don't know C++ at all.
How about engineering with C++? That is such a difficult task. I went over header file hygiene with a colleague a couple of months ago. The number of points that you should pay attention to is mind boggling: Don't #include unnecessarily, do forward declare as much as possible (but what can be forward declared is hard to understand even for experienced programmers), #include your own API header first to prove that it's complete (and good luck!), don't forget header guards, don't reuse header guards, etc. etc. This is just efficient header file usage! We haven't started coding yet!
My friends, the emperor doesn't have clothes. C++ simply is not a tool that is designed well. People who choose it do so because they have to or they are masochists. (True story: I asked a relatively young Google meetup presenter once why he was using C++ instead of a modern language and he said "because it is hard".) C++ separates the elite from the masses; I used to strive to be a C++ elite; I am not interested a bit anymore; I want to write useful programs with D; and I do.
D is niche only because humans are populists. We are not encouraged to use tools (or products) that are designed better. We follow popular leaders. It takes one some time to find his or her own voice to reject bad products and use only good ones. I am extremely lucky to work for a company that allows me to use D to write useful products.
I still take the same joy from programming that I did when I first learned it.
Then there is the human aspect of it: I want to be associated with real people isntead of snobby elites. (Remember how C++ was marketed at around 2000? "Yes, C++ is hard but it was never meant to be for normal programmers anyway." Ha ha ha! I am old enough now to reject that mentality. Bad design is bad design my friends; you can't defend it by blaming the user for not being elite.)
I can go on and on...
Now it's my turn to ask: Why would anyone choose C++ for their projects despite the production costs that it brings? None of your programmers really know it; they introduce hidden liabilities in the projects, their source code become non-refactorable monsters. Why waste that money on C++ when you can produce products easily. Products that just work...
Familiarity, and all the libraries and tools available for C++. I see that D has a section on C++ interop,[0] but it looks about as painful as FFI usually is, and even more painful given how template-heavy C++ code tends to be.
(Completely unrelated: I can't mention FFI without also mentioning how amazing LuaJIT's C FFI is. The developer(s) really nailed it.)
0: https://gist.github.com/ske2004/336d8cce8cd9db59d61ceb13c1ed...
IMHO, Zig is the closest thing to being D-esk (like with comptime), but it's still not a mainstream option yet.
A fix is on master/beta but will still take some time to be released.
That was fixed within the week, with a notification given that it had been sent to the reporter.
One of my projects has a really simple server written in nodejs that's basically (in terms of complexity) just an auth'd chatroom, and I wanted to switch it from using raw tcp sockets to websockets. And since the server is so simple, why not refactor it to another language and see if there's no some performance gains from that? I ended up doing something pretty similar to that "Comparing 10 programming languages. I built the same app in all of them." video from Tom Delalande (https://www.youtube.com/watch?v=-MbTj8DGOP0). I had several working versions of the server in:
- Bun, using Bun APIs (https://bun.sh/docs/api/websockets)
- Dart, using Dart APIs (https://api.dart.dev/stable/latest/dart-io/WebSocket-class.h...)
- Java, using Java-WebSocket (https://github.com/TooTallNate/Java-WebSocket)
- Kotlin, using Ktor (https://start.ktor.io/p/ktor-websockets)
- Rust, using tokio-tungstenite (https://docs.rs/tokio-tungstenite/latest/tokio_tungstenite/i...)
- Zig, using websocket.zig (https://zigistry.dev/packages/karlseguin/websocket.zig/)
- D, using serverino (https://code.dlang.org/packages/serverino)
And Dlang was, by far, the worst experience out of the lot. Firstly is the lack of adequate, comprehensive, and centralised tooling. I almost gave up when dmd could not even compile a freshly init'd project. The impression I got is that you're not really meant to use dmd directly, you're meant to use dub, like how you compile Java projects with Maven/Gradle, not javac. Except that there's also apparently three competing compilers (https://wiki.dlang.org/Compilers)? And good luck remembering the names of the tooling because they're all some random three-letter combination.
Serverino makes heavy use of mixins and attributes (think Java annotations), which is not ideal. But what really killed the deal was (despite using the recommended intellij plugin (https://wiki.dlang.org/IDEs) with the recommended tools installed and setup) not being able to inspect[1] serverino's mixin or its attributes. So I look at serverino's source code, except its source also has mixins... which I can't inspect. I'm not going to use something when I cannot easily ascertain its control flow. And while, yes, I probably should have gone with vibe-d (https://code.dlang.org/packages/vibe-d%3Ahttp) in the first place, mixins and attributes are nonetheless part of the language and the tooling should be able to tell me about them.
- [1] When I say "inspect" I mean requesting the IDE to show me the source/definition so I can see what it is, what it does, and where it's known to be used.
Yep this also happened to me when I tried D. I love the idea of the language and the syntax is great, but I really don't want to fight my tools when I'm working on a project.