scale-encode

[v0.8.0] - 2024-10-21

• chore(deps): bump derive_more from 0.99 to 1.0 (#26)
• chore(deps): bump primitive-types from 0.12.0 to 0.13.1 (#27)

Full Changelog: v0.7.2...v0.8.0

[v0.7.2] - 2024-10-21

• Update dependencies syn 2.0, darling 0.20 and proc-macro-crate 3.1 (#24)

[v0.7.1] - 2024-05-17

• Implement EncodeAsFields for pointer types like Arc and Box (#22)

[v0.7.0] - 2024-04-29

Update the scale-type-resolver dependency to 0.2.0 (and bump scale-bits for the same reason).

The main change here is that type IDs are now passed by value, rather than reference.

[v0.6.0] - 2024-02-16

Up until now, scale-info has been the library that gives us the information needed to know how to SCALE encode values to the correct shape. In this release, we remove it from our dependency tree and replace it with scale-type-resolver, which provides a generic TypeResolver trait whose implementations are able to provide the information needed to encode/decode types. So now, rather than taking in a scale_info::PortableRegistry, the EncodeAsType and EncodeAsFields traits take a generic R: scale_type_resolver::TypeResolver value. scale_info::PortableRegistry implements TypeResolver, and so it can continue to be used similarly to before (though now, type_id is passed as a reference), but now we are generic over where the type information we need comes from.

To be more concrete, EncodeAsType used to look roughly like this:

pub trait EncodeAsType {
    fn encode_as_type_to(
        &self,
        type_id: u32,
        types: scale_info::PortableRegistry,
        out: &mut Vec<u8>,
    ) -> Result<(), Error>;
}

And now it looks like this:

pub trait EncodeAsType {
    fn encode_as_type_to<R: TypeResolver>(
        &self,
        type_id: &R::TypeId,
        types: &R,
        out: &mut Vec<u8>,
    ) -> Result<(), Error>;
}

One effect that this has is that EncodeAsType and EncodeAsFields are no longer object safe (since the method they expose accepts a generic type now). Internally this led us to also change how scale_encode::Composite works slightly (see the docs for that for more information). if you need object safety, and know the type resolver that you want to use, then you can make a trait + blanket impl like this which is object safe and is implemented for anything which implements EncodeAsType:

trait EncodeAsTypeWithResolver<R: TypeResolver> {
    fn encode_as_type_with_resolver_to(
        &self,
        type_id: &R::TypeId,
        types: &R,
        out: &mut Vec<u8>,
    ) -> Result<(), Error>;
}
impl<T: EncodeAsType, R: TypeResolver> EncodeAsTypeWithResolver<R> for T {
    fn encode_as_type_with_resolver_to(
        &self,
        type_id: &R::TypeId,
        types: &R,
        out: &mut Vec<u8>,
    ) -> Result<(), Error> {
        self.encode_as_type_to(type_id, types, out)
    }
}

We can now have &dyn EncodeAsTypeWithResolver<SomeConcreteResolver> instances.

The full PR is here:

• Enable generic type encoding via TypeResolver and remove dependency on scale-info (#19).

[v0.5.0] - 2023-08-02

• Improve custom error handling: custom errors now require Debug + Display on no_std or Error on std.
  Error::custom() now accepts anything implementing these traits rather than depending on Into<Error>
  (#13).
• Enable using #[codec(skip)] or #[encode_as_type(skip)] to ignore fields when using the EncodeAsType macro.
  Skipping isn't generally necessary, but can be useful in edge cases (such as allowing a multi-field struct to be
  encoded to a number if all but one numeric field is skipped) (#16).

Prep to release 0.4.0 (#12)

Prep for 0.3.0 release, and restore object safety of EncodeAsFields t…

• Update scale-info to latest, removing deprecated method calls.
• Change EncodeAsFields to accept an iterator over fields, to allow more flexibility in how we provide fields to encode.

Fix a silly typo in the scale-encode-derive README.

Improve the documentation with more examples, and tweak BTreeMap encoding and Composite encoding to be more tolerant.

Initial release.