Blanket implementation for DomainType

[greg-szabo]

Closes #535 again.

This change implements default functions for the DomainType trait, instead of implementing the functionality in the DomainType macro. The macro becomes syntactic sugar for impl DomainType<RawType> for MyDomainStruct {}.

It also fixes a "feature" that wasn't correctly implemented: all functions in the trait that relate to Prost::Message (encode, decode, encode_len, etc) now work exactly as intended and they don't consume the DomainType struct.

The trait now requires additional traits to be implemented on the struct (notably Clone) but it's a small price to pay.

It also enhances the DomainType trait with direct encoding/decoding using Vec<u8>. This functionality was available in the AminoMessage trait before.

Thanks to Ismail for pointing out that this can be done.

• Referenced an issue explaining the need for the change
• Updated all relevant documentation in docs
• Updated all code comments where relevant
• Wrote tests
• Updated CHANGELOG.md

[liamsi]

Not entirely. I was also wondering if we need the whole proc_macro at all.
You can replace all these two lines that look like this

#[derive(DomainType)]
#[rawtype(RawPartSetHeader)]

with:
impl DomainType<MyRawType> for ActualType {}.
As a consequence you can delete the whole macro code:

tendermint-rs/proto-derive/src/lib.rs

Lines 10 to 56 in e892067

	#[proc_macro_derive(DomainType, attributes(rawtype))]
	pub fn domaintype(input: TokenStream) -> TokenStream {
	let input = syn::parse_macro_input!(input as syn::DeriveInput);
	expand_domaintype(&input)
	}
	fn expand_domaintype(input: &syn::DeriveInput) -> TokenStream {
	let ident = &input.ident;
	// Todo: Make this function more robust and easier to read.
	let rawtype_attributes = &input
	.attrs
	.iter()
	.filter(|&attr| attr.path.is_ident("rawtype"))
	.collect::<Vec<&syn::Attribute>>();
	if rawtype_attributes.len() != 1 {
	return syn::Error::new(
	rawtype_attributes.first().span(),
	"exactly one #[rawtype(RawType)] expected",
	)
	.to_compile_error()
	.into();
	}
	let rawtype_tokens = rawtype_attributes[0]
	.tokens
	.clone()
	.into_iter()
	.collect::<Vec<quote::__private::TokenTree>>();
	if rawtype_tokens.len() != 1 {
	return syn::Error::new(rawtype_attributes[0].span(), "#[rawtype(RawType)] expected")
	.to_compile_error()
	.into();
	}
	let rawtype = match &rawtype_tokens[0] {
	proc_macro2::TokenTree::Group(group) => group.stream(),
	_ => {
	return syn::Error::new(
	rawtype_tokens[0].span(),
	"#[rawtype(RawType)] group expected",
	)
	.to_compile_error()
	.into()
	}
	};

And in fact the whole proto-derive crate can be deleted, without losing any functionality or any further consequences. I would argue whole code becomes easier to navigate as the macro creates some unnecessary indirection for a single line that explains what is happing.

[greg-szabo]

Yes, true. Annotations seem cleaner to me, because I don't expect to look at the derive function. Serde does the same thing and it worked out pretty well there.

Also, we need to do some kind of JSON serialization for these domain types too and I didn't have time to get into the details of how we're going to do it. Maybe the derive macro helps, maybe not. Well, currently not really, it's just syntactic sugar.

I'll open this up to the team, I'm happy to do it either way, I just really wanted to do annotations. :)

[liamsi]

Great work 👍

[liamsi]

I was too excited. It needs some cleanup in the documentation but everything else looks 👌

[liamsi]

Actually some of the comments need updating now.

[greg-szabo]

I'll do that.

[greg-szabo]

Minor updates in the documentation: removed references to the DomainType macro. I hope this is it. :)

[liamsi]

LGTM