Documentation for VariantArray and VariantNames say their VARIANTS constants are arrays, but they're actually slices

[IndigoLily]

The incorrect documentation:

strum/strum_macros/src/lib.rs

Line 169 in 2056939

/// Implements `Strum::VariantNames` which adds an associated constant `VARIANTS` which is an array of discriminant names.

strum/strum_macros/src/lib.rs

Line 215 in 2056939

/// Adds a static array with all of the Enum's variants.

strum/strum/src/lib.rs

Line 220 in 2056939

/// A trait for retrieving a static array containing all the variants in an Enum.

The actual types:

strum/strum/src/lib.rs

Line 217 in 2056939

const VARIANTS: &'static [&'static str];

strum/strum/src/lib.rs

Line 227 in 2056939

const VARIANTS: &'static [Self];

I personally think these constants should be actual arrays (because it's much easier to get a slice from an array than vice versa), but I'm not sure how feasible that is in current stable Rust.

[T0mstone]

Wow, I was about to write the same issue. What a coincidence!

You're right, making them true arrays won't work yet, since associated types aren't allowed to depend on associated constants. A workaround would be to make the array length a generic parameter of the trait, but that's pretty unwieldy, in addition to being semantically completely wrong.

I would be in favor of the macros also generating an inherent associated constant which is a true array. That way we can at least use the constant length in non-generic code.

[Peternator7]

There was discussion when the features were added about slices vs arrays. Like @T0mstone said, associated constants still aren't great to use. It also makes adding a new variant to an enum a breaking change because the length of the array changes, but slices don't have that issue.