smartstring VS cats-effect

Your use case sounds like it will involve a lot of small strings that use a subset of UTF-8. If you’re concerned about performance, you could look into something like smartstring. Sixbit also looks interesting, but it looks like it won’t give you any more characters and it’d probably require additional computation to do the conversion (and they’d have to be converted back out).

> If you have a long-running async function, then pass parameters by value! If you have a polymorphic async function, then return your result in a Box.

I've taken to making heavy use of the smallvec and smartstring crates for this. Most lists and strings are small in practice. Using smallvec / smartstring lets you keep most clone() calls allocation-free. This in turn lets you use owned objects, which are easier to reason about - for you and the borrow checker. And you keep a lot of the performance of just passing around references.

I tried to use async rust a couple of years ago, and fell on my face in the process. Most of my rust at the moment is designed to compile to wasm - and then I'm leaning on nodejs for networking and IO. Writing async networked code is oh so much easier to reason about in javascript. When GAT, TAIT and some other language features to fix async land I'll muster up the courage to make another attempt. But rust's progress at fixing these problems feels painfully slow.

https://crates.io/crates/smallvec / https://crates.io/crates/smartstring

Just to point out, smartstring no longer assumes String memory layout. From the changelog:

There's some really good crates that implement SSO floating around - eg, SmartString. But I agree - its a pity they're needed. Swift built this into the core string type in the language. I think that was the right call.

Comparatively: * SmolStr can inline up to 22 bytes but does not adjust down for 32-bit architectures, meaning it's potentially wasting memory on 32-bit archs. Similarly though it's immutable and Clone is O(1) * SmartString can inline up to 23 bytes, but it's mutable and Clone is O(n). Also this crate makes assumptions about the memory layout of a String, which in theory should be fine, but is a slight caveat.

SmartString is used to store identifiers (which tends to be short, fewer than 23 characters, and ASCII-based) because they can usually be stored inline. Map keys now also use SmartString.

I’ve been using smartstrings, which is both excellent and maintained. https://github.com/bodil/smartstring

Those benchmarks are using a snapshot version of cats-effect. I don't know where that one comes from, but previously they were using a snapshot from https://github.com/typelevel/cats-effect/pull/3332 which had some issues (3.5-6581dc4, 70% performance degradation), which have since been resolved (see that PR for more info and comparative benchmarks).

Recently, CE3 has had similar issues reported across multiple repositories, almost an epidemic of reports!

The one advantage Rust has over Scala is that it detects data races at compile time, and that's a big time saver if you use low level thread synchronization. However, if you write pure FP code with ZIO or Cats Effect that's basically a non-issue anyway.

See also cats-effect and fs2. cats-effect gives you your IO Monad (and IOApp to run it with on supported platforms). fs2 is the ecosystem’s streaming library, which is much more pervasive in functional Scala than in Haskell. For example, http4s and Doobie are both based on fs2.

On the contrary, switching to the functional mindset, with something like Typelevel Scala3 and respective cats and cats-effect fs2 frameworks, helps to rethink a lot of designs and development approaches.

I think "better Haskell on JVM" (in contrast to "worse Haskell") is a good identity for Scala to have. (Please note that this is an intentional hyperbole.)

Of course, there are areas where Haskell is stronger than Scala (hint: modularity, crucial for good Software Engineering, is not one of them). And Scala has its own way of doing things, so just imitating Haskell won't work well.

Examples of this "better Haskell" are https://typelevel.org/cats-effect/ and https://zio.dev/ .

All together, Scala may be a better choice for you if you want to do Pure Functional Programming. And is definitely less risky (runs on JVM, Java libraries interop, IntelliJ, easy debugging, etc...).

None of the other languages you mentioned are viable in this sense (if also you want a powerful type system, which rules out Clojure).

I agree that Rust's identity is pretty clear: a modern language for use cases where only C or C++ could have been used before.

I would use Scala. I like FP and Scala comes with some awesome libraries for concurrent/async programming like Cats Effect or ZIO. Good choice for creating modern style micro-services to be run in the cloud (or even macro-services, Scala has a powerful module system, so it's made to handle large codebases).

https://typelevel.org/cats-effect/

https://zio.dev/

The language, the community and customs are great. You don't have to worry about nulls, things are immutable by default, domain modelling with ADTs and patter matching is pure joy.

The tooling available is from good to great and Scala is big enough that there are good libraries for typical if not vast majority of stuff and Java libs as a reliable fallback.

What took my interest is this (for both JVM and future multithreaded Scala native): https://github.com/typelevel/cats-effect/discussions/3070 Having the same threads poll available IO events and execute callbacks should improve performance greatly

The author is the creator of Monix and implemented the first version of cats-effect. He knows what he is doing.

You want Kernmantle, which quite honestly shouldn't be hard to implement around Cats and cats-effect. In particular, although Kernmantle doesn't require the use of the Arrow typeclass, there happen to be Arrow (actually ArrowChoice) instances for both Function1 from the standard library and Kleisli from Cats itself, given a Monad instance for the Kleilsi's F[_] type parameter. In other words, we should be able to port Kernmantle from Haskell to Scala (with the Typelevel ecosystem) and instantly be able to use pretty much anything else from the Typelevel ecosystem, or wrapped with it, in our workflow graphs. Pure functions, monadic functions, applicative functions, GADTs with hand-written interpreters, any of it. I think this would be eminently worth doing.