The situation is changing in the CLR land, now that they can just say that legacy codebases can stay on .NET 4.x, and .NET Core is where all the fancy development is happening. There are already some C# language features that are Core-only because they require the corresponding runtime changes.

The other thing is - people always forget that, unlike JVM, CLR has more than just the object layer. By design, it has enough low-level constructs to compile a language like C++, complete with multiple inheritance, unions, varargs etc. Obviously, you can do pretty much anything on top of that - the only problem is that you won't be able to interop with other .NET code, except through C-style FFI. But, hypothetically, it would be possible to establish a higher-level ABI without baking it into the VM as an object model, thus allowing a reboot without throwing everything away.

> By design, it has enough low-level constructs to compile a language like C++, complete with multiple inheritance, unions, varargs etc

But the CLR doesn't support that. If you compile C++ code to IL then you'll get a compiler error if you use any types that use multiple-inheritance. The CLR's underlying type system is a huge limitation when it comes to using even simple modern ADTs.

For example, in F# you can define a union type, and the union subtypes can contain normal library types, but you cannot define a library type as a union subtype, whereas you can in TypeScript (and Rust too, I think?).

You're confusing managed C++ with C++/CLI. The latter tries to introduce additional constructs to C++, so that it can partake in the CLR object model - and there you get all those limitations like no multiple inheritance. But you can, in fact, compile any random C++ code to IL - just run cl.exe with /clr:pure. The only thing that doesn't work in that case is setjmp/longjmp; everything else is available.

If you want to see it for yourself, take some .cpp file, and compile it with cl.exe /clr:pure /O2 /FAs. The latter switch will dump the IL assembly into the corresponding .asm file. Or you can inspect the output with ILSpy etc. You'll see that it compiles native C++ types down to CLR structs with no fields, but with explicitly set size (via StructLayout.Explicit); and then uses pointer arithmetic to access field values.

The /clr:pure option was removed in VisualC++ 2017 - it was last available in 2015.

It was deprecated and will be removed eventually, but for now, it's still working fine.

Anyway, the point is that it's possible within CLR constraints. There's just not much demand for it, and it's effectively a whole separate CPU arch for the compiler to support, in addition to x86/x64/ARM.

It surprises me as well. And to belabor my own point some more: there's a reason the #1 question about Rust from newbies seems to be "Is there a language like Rust, but with a garbage collector?" - sometimes reworded as "Is there a way to turn off the borrow checker?"

To be fair, I think that a decent amount is possible on JVM and CLR. Scala, for as much hate as it gets, has a much stronger type system than Kotlin/Java.

And Swift has its own backwards baggage! It has to work with Objective-C. There's actually a few weird things in Swift that I've bumped into that I'm pretty sure are only there because of Objective-C.

The Rust devs certainly take backwards compatibility very seriously and they guarantee backwards compatibility forever. There will be no breaking changes in Rust except in extreme cases of finding unsoundness or whatever.

There are currently ZERO good high level app-building languages, IMO.

Rust isn't it because of the lack of garbage collection (which is not a point against Rust- just for the "domain" of optimizing for user-space app development).

Swift would probably be it, but it's pretty much Apple-only and I don't expect that to change.

TypeScript is close, but it's held back by needing to work with JavaScript and the JavaScript standard library and ecosystem suck. It also can't do threads.

Kotlin/Java are decent, but not good at concurrency, and having things like different sized integer types is really silly for a high-level language where everything goes on the heap anyway...

Python is slow and can't do threads.

Some languages like Lisps and MLs might be good for building apps if they just had the ecosystem around it. Haskell would be a giant pain the ass because real apps are full of "IO".

So, yeah. It's kind of a miracle that we can get anything done. I guess that's why we get paid so well...

Kotlin is the application language you're looking for (and Scala 3 to a lesser extent). Contrary to what you say it is the best language I've ever used for concurrency. It has it all, structured concurrency, cancelation, Flow, transparency (no await), etc. Regarding your second point the JVM is increasingly using the stack and with the soon complete generics you'll be able to avoid the boxed versions of the primitive types (though they remain useful when you want identity)

> Contrary to what you say it is the best language I've ever used for concurrency. It has it all, structured concurrency, cancelation, Flow, transparency (no await), etc.

I disagree.

Have you ever tried to actually implement something non-trivial that takes advantage of structured concurrency with cancellation? It's pretty hard to do correctly. Can you really tell me off the top of your head what the difference is between `withContext(coroutineContext) {}` and `coroutineScope {}` from within a suspend function?

Coroutines use unchecked exceptions for control flow. Kotlin also uses unchecked exceptions for fatal and non-fatal error handling. Figuring out how all these things interplay when it comes to coroutines and suspend functions has some subtleties that, IMO, are very difficult to figure out from just documentation and blog posts.

Also, Kotlin's standard types are entirely unsafe to use concurrently. The fact that MutableList inherits from List means that a function that accepts a List parameter CANNOT assume that the list wont change while the function is executing. So if you write `if (list.isNotEmpty()) { doSomething(list.first()) }` - that's a race condition because the list can literally become empty between the if clause and the body.

"But, wait! You should have just been smart enough to make a copy of your List before sending it between threads/coroutines." Okay, great. Let's do full copies of potentially-large collections. Thank goodness Kotlin is so concurrency ready that the standard collection types are persistent .. colle..ctions... oh.

Kotlin's concurrency story is really not that awesome. Scala is better, but still not perfect. Clojure is better still. Rust is good. Elixir (or anything with some kind of actor framework, I guess) is good. Haskell is good.

But I agree, overall, that if I had to pick a best app language today, it's either Kotlin or Scala, or Swift if you're writing for Apple stuff. I'll admit that I have a glaring experience gap with .NET languages, so I can't honestly say anything about C# and F#.

> I'll admit that I have a glaring experience gap with .NET languages, so I can't honestly say anything about C# and F#.

C#/.NET comes with language-level mutexes (`lock`) and the .NET library has thread-safe generic collections (ConcurrentDictionary, ConcurrentBag) and true immutable collections (ImmutableArray, ImmutableList, ImmutableDictionary) with optimized copy operations (e.g. ImmutableList.Add is O(1), but ImmutableArray.Add is O(n)). It's a nice addition to the library with only a few warts.

That's good to hear.

Java/Kotlin also have mutexes- just not as language built-ins (well, it does have `synchronized`).

They also have a ConcurrentFoo set of collections as well. And actually an ImmutableMap (but I don't see ImmutableList, etc. Why?).

The "problem" is that they're opt-in.

I spent years writing multi-threaded C++. But I've become very spoiled with modern languages that make concurrency safe(r)-by-default, such as Rust, Clojure, and Elixir (I haven't actually used concurrent Haskell).

Honestly, it's not anywhere near as bad as it used to be. The ecosystem(s) have embraced immutability everywhere that it's possible, so you don't have to worry quite as much about accidents.

> but I don't see ImmutableList, etc. Why

.NET doesn't have an ImmutableList either (the ImmutableBag type is an unordered collection). This is because unlike with hash-tables you always need to lock the entire structure when mutating a List/Vector (with hashtables you only need to lock the specific bin/bucket).

Ah! Good point.

> Coroutines use unchecked exceptions for control flow

That's... horrible.

Why can't Kotlin do it the way C# does with heap-allocated state-machines?

Honestly, I truly don't know and I don't feel qualified to judge the merits of the two approaches. There could be performance or behavior implications that they just prioritized differently from how C# does it. Or maybe something to do with the Java underpinnings. Are C# promises/whatever cancellable?

But, as an "end user", the exception thing drives me absolutely nuts. It wouldn't be nearly as bad if Kotlin either didn't use exceptions for ALL error handling, or even if it had checked exceptions so that "normal" errors would be, mentally, separate from fatal errors and/or coroutine cancellation.

> Are C# promises/whatever cancellable?

Yes, but it's opt-in (it requires the author of the async method to add a `CancellationToken` parameter and to respect `IsCancellationRequested`).

> Kotlin/Java are decent, but not good at concurrency

Since when JVM languages aren’t good at concurrency?

Also, primitives are usually not heap-allocated, so I don’t see what’s the problem with it. It makes the JVM a beast when it comes to number crunching.

> Since when JVM languages aren’t good at concurrency?

Since it's trivially easy to accidentally mutate data across concurrent contexts (threads). You have to actively remember to reach for Mutexes. You also have do understand how `synchronized` works and remember to actually use it. If you use a class that someone else wrote, you have to dig into their code (if available) to make sure they made their class thread-safe.

> Also, primitives are usually not heap-allocated, so I don’t see what’s the problem with it. It makes the JVM a beast when it comes to number crunching.

You don't need a number crunching beast to write most application-level software. That's my point. It's great for number crunching that you have byte, short, int, long as separate primitives. You know what most applications actually want? A number that wont magically wrap around to negative-LARGE_NUMBER when you guess the maximum size wrong. Java has fixed-size arrays, too, for super-performance mode. But applications just use List<> that has no size limit that you have to guess. It should do that same for Integers.