> Meh, an overrated minor point and it is not particularly elegant in either > Rust or Swift anyway.
My opinion differs. I came to Nim from OCaml and found OCaml's ADTs much more elegant than Nim's way. Efficient, expressive, _elegant_. That's what it says on the box! Yes, I'm well aware that elegance, like beauty, is in the eye of the beholder. I've yet to meet a programmer who has used a language with real ADTs who hasn't found that more elegant. > Sum types have inherent problems with serialization. Nim's way of doing it is > ok. Protobufs at least, are implemented for Rust (serde too), and support unions/sum types. In what way are these deficient? That "no known C# or Python shops ban exceptions" is a silly point. C#, Python and many other languages wouldn't even be considered at all in those exception eschewing C++ shops because of their heavy runtimes. [This](http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2019/p0709r4.pdf) is the longish document that gets into the problem, and I could have sworn that you yourself posted it once. Anyways, I'm very glad you agreed that some changes to Nim to accommodate C++ are worthwhile. I'm not a fan of OO in general, or C++ in particular, but I can tell you that some amazingly knowledgeable C++ programmers (I am not even close to being in their league) now believe that C++ is beyond saving, and are looking for other ways forward. That way has to take into account the existing code. > The standard library should embrace .push: raises: [] (that means, exceptions > need to be annotated explicitly) and then we have the best solution around. > IMHO. People will say that was tried with Java, and considered a failure, and ask "how is it different this time around?". @deech I suppose if there is an official pattern matching library that would be better but I still think building it into the core would be better. Nim has multiple pattern matching libraries, but I haven't seen a single macro library in Rust to express things the way Nim does, only some union stuff added to aid C binding. @Sixte, @deech: Yes, I'd support C++ style vtables, introduce classes and multiple inheritance. @shirleyquirk was disgusted at my proposing that before, and **rightly so** IMO, but Nim might be the best way forward for C++. Java => Kotlin, JS => TypeScript, C++ => ???. Nim could be that ???. Also, CRTP is used in the wild, at real companies with lots of code. @deech if there is a more minimal Nim extension that you think is sufficient, I'm all ears. BTW, I think you meant "gin up". There is a British expression "gen up" but it means something different.
