Hacker News Reader

Comment by tiniuclx | Hacker News Reader

I have been paying attention to Rust for the past year or so and I think it'll be the next language I will end up learning. I do have a question: what is const fn supposed to be used for? Is it for denoting pure functions? That would be really nice, considering that not many non-functional languages explicitly recognise pure functions.

View on HN

When a const fn is called from a const context, like assigning a value to a constant, the compiler interprets the function at compile time.

If you're interested in more info the docs [1] and the New Rustacean 1.31 Part 2 podcast [2] are helpful.

[1] https://github.com/rust-lang-nursery/reference/blob/master/s...

[2] https://newrustacean.com/show_notes/news/rust_1_31/part_2/in...

const fn is used for compile-time programming. "pure" is a bit over-loaded of a term, but you can sort of think of it that way, sure.

You can of course also run these functions at runtime; they're a more restricted kind of function. You cannot even currently use if inside of them, for example. We're working on expanding it out, but we're going slow because it's important for soundness reasons.

Should we think of them like the equivalent of constexpr functions in C++?

Yes, but hopefully without the amusing issue of non-constant constexpr functions in C++ :)

http://b.atch.se/posts/non-constant-constant-expressions/

So they're more like C++20's consteval?

http://wg21.link/p1073r3

As I understand it, not really no.

The conceptual difference between constexpr and consteval is that consteval is guaranteed to be evaluated at compile-time, so unlike constexpr code, consteval code will never actually end up as a function in your binary.

Rust's const fns on the other hand can be called at runtime, so in that respect they'd be closer to constexpr.

I am not familiar enough to declare an equivalence, but roughly the same, yes.

Ah thank you, that makes a lot of sense! It's good that this sort of stuff can be made explicit... a lot of C++ compile-time evaluation optimisations feel like magic to me.

Reply on news.ycombinator.com