Broken documentation: What on earth is "GenericArray"

[VorpalBlade]

https://docs.rs/sha2/latest/sha2/trait.Digest.html indicates I need to finalise into something called an "GenericArray". Unfortunately I have no idea what that is, as unlike most types this is not actually a link to the type. Usually rustdoc is pretty good about this, so I assume you did something unusual.

Hm, maybe it is some sort of trait. Lets try:

let mut hasher = Sha256::new();
// ...
let mut actual: [u8; 32] = [0; 32];
let hash = hasher.finalize_into(&mut actual);

Nope!

error[E0308]: mismatched types
   --> crates/paketkoll/src/arch.rs:191:45
    |
191 |             let hash = hasher.finalize_into(&mut actual);
    |                               ------------- ^^^^^^^^^^^ expected `&mut GenericArray<u8, UInt<..., ...>>`, found `&mut [u8; 32]`
    |                               |
    |                               arguments to this method are incorrect
    |
    = note: expected mutable reference `&mut GenericArray<u8, UInt<UInt<UInt<UInt<UInt<UInt<UTerm, B1>, B0>, B0>, B0>, B0>, B0>>`
               found mutable reference `&mut [u8; 32]`
note: method defined here

Also that is a really strange type. And it seems to confuse rust-analyzer in vscode such that I don't actually get any help for it.

All your examples in the sha2 crate only shows using finalize, not finalize_into. I want a stack allocated array in the end, and I need to compare it to an [u8; 32] that I get from elsewhere. It doesn't seem convertible. E.g. this doesn't work either:

let mut hasher = Sha256::new();
// ...
let mut actual = Default::default();
let hash = hasher.finalize_into(&mut actual);

if actual.into() != expected { // where expected is [u8; 32]
    // ...
}

(Nor does it help adding actual.into::<[u8; 32]>).

[VorpalBlade]

Apparently I can do actual[..] != expected[..], but that will have runtime bounds checks, so not an acceptable solution.

[tarcieri]

Usually rustdoc is pretty good about this, so I assume you did something unusual.

It actually seems like a bug in rustdoc. For whatever reason it's not linking to the relevant type, which is:

https://docs.rs/generic-array/0.14.7/generic_array/struct.GenericArray.html

Note that in the next release we'll be migrating to hybrid-array, although strangely the rustdoc issues seem to persist, e.g.:

https://docs.rs/digest/0.11.0-pre.8/digest/type.Output.html

I'm not sure why rustdoc is having an issue with this, but perhaps we should file an upstream bug, and in the meantime add manual links to the relevant types.

Note that hybrid-array will address your concerns about PartialEq impls.

[tarcieri]

This seems like a recent rustdoc regression. I've filed an upstream issue: rust-lang/rust#120983

[VorpalBlade]

It also seems your types (e.g. Sha256 or Md5) confuses Rust Analyzer enough that I don't get any completion suggestions on them. Could it be related (e.g. your code is too confusing for the tools for some reason).

[tarcieri]

That's probably related to RustCrypto/traits#1069

[jssblck]

It would be incredibly helpful if we could get a finalize_into_boxed_slice or something.
GenreicArray has been a pain the whole time I've worked with this crate :(

[VorpalBlade]

It would be incredibly helpful if we could get a finalize_into_boxed_slice or something. GenreicArray has been a pain the whole time I've worked with this crate :(

That wouldn't solve my problem. I don't want to do extra allocations. I want to finalize into a statically sized array on the stack. That would be the ideal solutions for me.

[tarcieri]

Digest::finalize_into can be used for that

[newpavlov]

Closing in favor of rust-lang/docs.rs#2528