Use new *Dirty traits from the digest crate
#153
Conversation
Updates all of the hash implementations in this repo to use the new `*Dirty` traits which support blanket impls of the original traits, which either consume the hasher instance or can be reset. This will also provide a marginal efficiency boost, at least until placement return lands (which, as it were, may be soon).
tarcieri
force-pushed
the
dirty-digest-crate
branch
from
4980227
to
9d48b6d
Compare
| fn finalize_fixed(mut self) -> GenericArray<u8, U32> { | ||
| { | ||
| let self_state = &mut self.state; | ||
| fn finalize_into_dirty(&mut self, out: &mut GenericArray<u8, U32>) { |
There was a problem hiding this comment.
I think it could be worth to be consistent and use digest::Output<Self> across all crates.
There was a problem hiding this comment.
The reason I couldn't here is because it doesn't impl Default, and digest::Output (presently, at least) uses the Digest trait:
https://github.com/RustCrypto/traits/blob/master/digest/src/digest.rs#L96
It could potentially use FixedOuput instead.
There was a problem hiding this comment.
Hrmm, changing it to use FixedOutput is not trivial... it means you can't use it with Digest because not all Digests necessarily impl FixedOutput, and that kind of defeats the point.
I think we should leave it as-is. That said, I think we can fix the version used in the macro.
|
Dirty traits fit even better than I expected. |
|
Indeed, surprisingly easy refactor for what it is. I'll go ahead and merge and circle back on changing |
Updates all of the hash implementations in this repo to use the new
*Dirtytraits which support blanket impls of the original traits, which either consume the hasher instance or can be reset.This will also provide a marginal efficiency boost, at least until placement return lands (which, as it were, may be soon).