Borsh JS and Borsh Rust slightly different serialized output

I am trying to send borsh serialized data from JS to rust application. However, when serialising data in javascript AND rust (to compare the outputs), I am getting 4 extra bytes in rust serialised output. Here is the code:

Borsh JS code

// class
class Poll {
  id: string = '';
  question: string = '';
  options: string[] = [];
  votes: number[] = [];

  constructor(fields?: {
    id: string;
    question: string;
    options: string[];
    votes: number[];
  }) {
    if (fields) {
      this.id = fields.id;
      this.question = fields.question;
      this.options = fields.options;
      this.votes = fields.votes;
    }
  }
}

// Schema
const schema: Schema = new Map([
  [
    Poll,
    {
      kind: 'struct',
      fields: [
        ['id', 'string'],
        ['question', 'string'],
        ['options', ['string']],
        ['votes', ['u32', 1]],
      ],
    },
  ],
]);


// class object
const testPoll = new Poll({
  id: '1',
  question: 'What is your favorite color?',
  options: ['a', 'b', 'c'],
  votes: [100],
});

//object serialization
let serializedPoll: Uint8Array = new Uint8Array();
serializedPoll = serialize(schema, testPoll); // this succeeds

// output

[1, 0, 0, 0, 49, 28, 0, 0, 0, 87, 104, 97, 116, 32, 105, 115, 32, 121, 111, 117, 114, 32, 102, 97, 118, 111, 114, 105, 116, 101, 32, 99, 111, 108, 111, 114, 63, 3, 0, 0, 0, 1, 0, 0, 0, 97, 1, 0, 0, 0, 98, 1, 0, 0, 0, 99, 100, 0, 0, 0]

Borsh Rust Code

#[derive(BorshDeserialize, BorshSerialize, Debug)]
pub struct Poll {
    pub id: String,
    pub question: String,
    pub options: Vec<String>,
    pub votes: Vec<u32>,
}

// poll object - with same values as that in JS code above
let p = Poll {
        id: "1".to_string(),
        question: "What is your favorite color?".to_string(),
        options: vec!["a".to_string(), "b".to_string(), "c".to_string()],
        votes: vec![100],
};

// serialization
let serialized_data = p.try_to_vec().unwrap(); // this succeeds

//output
[1, 0, 0, 0, 49, 28, 0, 0, 0, 87, 104, 97, 116, 32, 105, 115, 32, 121, 111, 117, 114, 32, 102, 97, 118, 111, 114, 105, 116, 101, 32, 99, 111, 108, 111, 114, 63, 3, 0, 0, 0, 1, 0, 0, 0, 97, 1, 0, 0, 0, 98, 1, 0, 0, 0, 99, 1, 0, 0, 0, 100, 0, 0, 0]

Comparing outputs of both

1. Borsh JS
2. Borsh Rust

[1, 0, 0, 0, 49, 28, 0, 0, 0, 87, 104, 97, 116, 32, 105, 115, 32, 121, 111, 117, 114, 32, 102, 97, 118, 111, 114, 105, 116, 101, 32, 99, 111, 108, 111, 114, 63, 3, 0, 0, 0, 1, 0, 0, 0, 97, 1, 0, 0, 0, 98, 1, 0, 0, 0, 99, 100, 0, 0, 0]

[1, 0, 0, 0, 49, 28, 0, 0, 0, 87, 104, 97, 116, 32, 105, 115, 32, 121, 111, 117, 114, 32, 102, 97, 118, 111, 114, 105, 116, 101, 32, 99, 111, 108, 111, 114, 63, 3, 0, 0, 0, 1, 0, 0, 0, 97, 1, 0, 0, 0, 98, 1, 0, 0, 0, 99, 1, 0, 0, 0, 100, 0, 0, 0]

There are extra 4 bytes (1, 0, 0, 0) in rust serialised output. I believe this is happening because of Vec<u32> used for votes field (it works fine for u32). However I am unable to understand why this is happening.

Any sort of help/insights are appreciated.

Thank you!

Answer

A Vec<u32> has to encode the length of the data since a Vec represents a variable size. Conversely, you’ve designed your schema in JS as ['u32', 1] which is an array of length 1, so it does not need to encode a length since it is a fixed size.

To resolve the difference, either set your schema to a variable size array: ['u32']. Or change your type in Rust to be a fixed-size array: votes: [u32; 1].