4 ms·
I'd say that not enough functional extremism killed Scala. It tried to be all things for all purposes, compatible with every approach. So it grew into a incompr
by nine_k 8y ago
I'd say that not enough functional extremism killed Scala. It tried to be all things for all purposes, compatible with every approach. So it grew into a incomprehensible monster of a language, full of subtle traps, like C++.
Making things compact and focused helps. Of mixed-approach languages, I see Kotlin's approach to integrating FP as more successful.
- kod 8y agoKotlin doesn't even have real pattern matching. Calling it integrating FP is a sad joke.
- dmitriid 8y agoPattern matching has nothing to do with FP.
- kod 8y agoOnly if you think algebraic types have nothing to do with FP, in which case you have a very idiosyncratic definition of FP.
- nine_k 8y agoIt's got algebraic types, and a typeclasses proposal (with a PR!) was actively worked on, last time I checked.
- kod 8y agoAlgebraic types without being able to conveniently destructure them is maybe half the benefit at best.
- ldng 8y agoI'm sorry, I respectfully disagree on the "not enough" :-) I think it is a matter of choice of audience. If you want pure functional languages, there are plenty that will rock your boat (Haskell, OCaml ...). But, to my mind, I see Rust as a somewhat "modernised" C/C++ with, like Python, the most interesting and accessible parts of FP backed in. It is a balance and it is fragile. So, as I said, I hope that Rust will keep that fine line. All that said, it only is my opinion, I might be wrong and the wider audience might be ready for "harder" FP. But I doubt it.
- nine_k 8y agoRust not only has some good parts from FP; they are added thoughtfully enough to keep them logically sound. So the fact that there are math-backed theories behind them still matters. With C++, I can't say the same is true. Some libraries try to alleviate that, of course.
- rujuladanh 8y agoWhat math-backed theories? Why do they matter? Do you mean for the compiler team or for the users?
- nine_k 8y agoHindley-Milner type inference and linear types (used indirectly for lifetimes) spring to mind.
- rujuladanh 8y agoThose are useful for compiler devs and language designers, not so much for the average user of the language.
- leshow 8y agoI'm not sure what parts of FP python has you're referring to but it seems pretty lacking. It has a few combinators, that's about it. The influence of "FP" on rust is really more like the influence of Haskell's type system. Rust is pretty far from a functional language but it does share a lot of type system features with Haskell: type classes, sum and product types, HM type inference