Remove `Index` trait bound from `Hash`

[tcharding]

The Index trait is not required if we provide the AsRef trait. Users can just call as_ref()[0] to get access to the 0th element of a Hash.

For times when we take a reference to the whole slice [..] using as_ref() is capable and arguably more ergonomic.

From the Hash trait remove the trait bound on Index and its associated traits, instead use the trait bound AsRef (we already have a trait bound of Borrow).

The first patch removes Deref implementations, I did this because the docs say so - not sure if that is enough justification or if I'm missing something.

https://doc.rust-lang.org/std/ops/trait.Deref.html

This work was inspired by https://github.com/rust-bitcoin/rust-miniscript/pull/450#discussion_r932704196

Verification

This PR is verified to not break usage of the lib by patching rust-secp256k1 and rust-bitcoin.

• https://github.com/rust-bitcoin/rust-secp256k1/pull/482
• https://github.com/rust-bitcoin/rust-bitcoin/pull/1173

[tcharding]

hmm, the CI fail appears to be on master right now. Will investigate more.

[apoelstra]

Yeah, I can get behind this. When I added Index I think it [..] was the fashionale way to convert slice-like objects to slices. Or maybe I just saw it done this way once and copied it.

I do think that .as_ref is more explicit about what's going on, and a bit less less likely to happen by accident/confusingly.

[tcharding]

CI needs https://github.com/rust-bitcoin/bitcoin_hashes/pull/167 to go green

[apoelstra]

Ok, merged #167. Can you rebase this?

[tcharding]

Rebased!

[Kixunil]

I meant removing bound not impl. But since indexing into hash is kinda weird, I'm not against removing impl. Could be done in two steps.

[apoelstra]

cc @TheBlueMatt are you opposed to this? It's a reduction in the API surface which arguably needlessly breaks people (in a mechanically fixable way, but still annoying). But OTOH it makes the separation between hashes and byteslices a little cleaner.

[sanket1729]

ACK removing the bound.

[tcharding]

Rebase only, no other changes.

[apoelstra]

I've ACKed but I would still like a (concept) ACK from @TheBlueMatt since we're removing API functions that are fairly widely used.

[Kixunil]

For consideration: some of these methods to avoid inference issues and possibly get better clarity than with as_ref:

• as_bytes(&self) -> &[u8; LEN]
• as_byte_array(&self) -> &[u8; LEN]
• as_bytes(&self) -> &[u8]
• as_byte_slice(&self) -> &[u8]
• as_mut_*

[tcharding]

I've added as_bytes(&self) -> &[u8; LEN] for all hash types as well as for types created by hash_newtype!.

[tcharding]

I've added as_mut_bytes as well, as a separate patch. I'm not understanding the utility of this method though - why would one want to modify a hash making the preimage unknown? If we want a hash with an unknown pre-image we can just use Hash::all_zeros.

[Kixunil]

@tcharding the only utility I can think of is consistency and I think three more lines are fine just for consistency.

[tcharding]

Cool, cheers.

[tcharding]

Closing in preparation for archiving this repo. See https://github.com/rust-bitcoin/rust-bitcoin/pull/1284