A (de)serializer for RLP encoding in ETH

Cargo.toml

serlp = "0.2.2" serde = { version = "1.0", features = ['derive'] }

Not Supported Types

We do not support enum when deserializing because we lost some information (i.e. variant index) about the original value when serializing. However, in some specific cases you can derive Deserialize trait for a enum with the help of RlpProxy, which will be discussed later.

We have to choose this approach because there is no enums in Golang while ETH is written in go. Treating enums as a transparent layer can make our furture implementation compatiable with ETH.

Design principle

Accroding to the ETH Yellow Paper, all supported data structure can be represented with either recursive list of byte arrays or byte arrays . So we can transform all Rust's compound types, for example, tuple, struct and list, into lists. And then encode them as exactly described in the paper

For example, the structure in example code, can be internally treated as the following form:

[ "This is a tooooooooooooo loooooooooooooooooooong tag", [ 114514, [191, -9810], [ [[], [[]], [[], [[]]]] ] ], "哼.啊啊啊啊啊啊啊啊啊啊啊啊啊啊啊啊啊啊" ]

Features

ZST serialization

In Rust, we can represent 'empty' in many ways, for example:

[], (), "", b"", struct Empty, Variant::Empty, None, PhantomData<T>

In our implementation:

  1. [] and () are considered empty list, thus should be serialized into 0xc0
  2. All other ZSTs are considered empty, thus should be serialized into 0x80

To better understand ZSTs' behavior when serializing, try this code:

```rust

[test]

fn testcompoundzst() { #[derive(Serialize, Debug, PartialEq, Eq)] struct ZST;

#[derive(Serialize, Debug, PartialEq, Eq)]
enum Simple {
    Empty(ZST),
    #[allow(dead_code)]
    Int((u32, u64))
}

#[derive(Serialize, Debug, PartialEq, Eq)]
struct ContainZST(Simple);

#[derive(Serialize, Debug, PartialEq, Eq)]
struct StructZST {
    zst: Simple
}

let zst = Simple::Empty(ZST);
let zst_res = to_bytes(&zst).unwrap();
assert_eq!(zst_res, [0x80]);

let with_zst = ContainZST(Simple::Empty(ZST));
let with_zst_res = to_bytes(&with_zst).unwrap();
// the container is transparent because its a newtype
assert_eq!(with_zst_res, [0x80]);

let with_zst = StructZST { zst: Simple::Empty(ZST) };
let with_zst_res = to_bytes(&with_zst).unwrap();
// the container is a list, to this is equivlent to [""]
assert_eq!(with_zst_res, [0xc1, 0x80]);

} ```

RLP Proxy

We have a RlpProxy struct that implemented Deserialize trait, which just stores the original rlp encoded data after deserialization (no matter what type it is). You can gain more control over the deserialization process with it.

Here is an example:

```rust

[derive(Serialize, Deserialize, PartialEq, Eq, Debug)]

[serde(from = "RlpProxy")]

enum Classify { Zero(u8), One(u8), Ten((u8, u8)) }

impl From for Classify { fn from(proxy: RlpProxy) -> Self { let raw = proxy.raw(); let mut tree = proxy.rlptree(); if tree.valuecount() == 2 { return Classify::Ten(from_bytes(raw).unwrap()) }

    let val = tree.next().unwrap()[0];
    match val {
        0 => Classify::Zero(0),
        1 => Classify::One(1),
        _ => panic!("Value Error.")
    }
}

} ```

Example code

You can find more examples here

```rust use serlp::rlp::{frombytes, tobytes}; use serde::{Serialize, Deserialize}; use serde_bytes;

[derive(Serialize, Debug, PartialEq, Eq, Deserialize)]

struct Third { inner: T }

[derive(Serialize, Debug, PartialEq, Eq, Deserialize)]

struct Embeding<'a> { tag: &'a str, ed: Embedded, #[serde(with = "serde_bytes")] bytes: Vec }

[derive(Serialize, Debug, PartialEq, Eq, Deserialize)]

struct Embedded { time: u64, out: (u8, i32), three: Third<((), ((),), ((), ((),)))> }

fn main() { let embed = Embeding { tag: "This is a tooooooooooooo loooooooooooooooooooong tag", ed: Embedded { time: 114514, out: (191, -9810), three: Third { inner: ((), ((),), ((), ((),))) } }, bytes: "哼.啊啊啊啊啊啊啊啊啊啊啊啊啊啊啊啊啊啊".asbytes().tovec() };

let encode = to_bytes(&embed).unwrap();
let origin: Embeding = from_bytes(&encode).unwrap();

println!("encode result: {:?}", encode);

assert_eq!(origin, embed);

} ```