Make it easy to access the contents of any hash (newtype) #81
Conversation
|
I'm curious what kind of situations you need this for.. |
If rust-bitcoin/rust-secp256k1#211 is not accepted, then the One could still argue that it is good use of the standard library to provide these conversion functions if the conversion exists, just to allow a more flexible use of the types :) |
|
concept ACK. I wonder if this would work now with 1.29 as the MSRV. |
thomaseizinger
force-pushed
the
impl-into
branch
from
ac9db7b
to
2a0ced0
Compare
I rebased onto latest master and pushed again, let's see what CI thinks. |
Hashes can be represented as fixed length arrays. That is in fact what they are internally. There is value in using Rust's visibility rules to enforce that no every byte array is a valid hash. However, every hash is certainly a valid byte array, hence it should be easy to access their contents as such. We add a `From` impl for any hash newtype for the inner byte array. This makes it possible to pass hash-newtypes the way they are to any function that needs a byte-array. We also add a `From` impl for any inner hash type that allows us to convert to the inner representation. Both of these changes are especially useful from a broader ecosystem perspective: Because `From` and `Into` are part of the standard library, any application or crate depends on them and can write functions that support a form of method overloading by accepting anything that is `Into<[u8; 32]>`. Implementing these conversion traits on our types makes them compatible with this pattern which results in less verbose APIs.
thomaseizinger
force-pushed
the
impl-into
branch
from
2a0ced0
to
13ab752
Compare
Nope. We will need Rust 1.41 for parts of this: https://blog.rust-lang.org/2020/01/30/Rust-1.41.0.html#relaxed-restrictions-when-implementing-traits Closing then. |
Hashes can be represented as fixed length arrays. That is in fact
what they are internally.
There is value in using Rust's visibility rules to enforce that
no every byte array is a valid hash.
However, every hash is certainly a valid byte array, hence it
should be easy to access their contents as such.
We add a
Fromimpl for any hash newtype for the inner byte array.This makes it possible to pass hash-newtypes the way they are to
any function that needs a byte-array.
We also add a
Fromimpl for any inner hash type that allows usto convert to the inner representation.
Both of these changes are especially useful from a broader ecosystem
perspective:
Because
FromandIntoare part of the standard library, anyapplication or crate depends on them and can write functions that
support a form of method overloading by accepting anything that is
Into<[u8; 32]>.Implementing these conversion traits on our types makes them
compatible with this pattern which results in less verbose APIs.