rust-fr

'rust-fr' (aka rust for real) is a simple, non-self-describing data-interchange format.

installation

You can use either of these methods.

• Add via cargo add rust-fr
• Add via Cargo.toml

[dependencies]
rust-fr = "1"

usage.

use serde::{Serialize, Deserialize};
use rust_fr::{serializer, deserializer};

// define some data
#[derive(Debug, Serialize, Deserialize, PartialEq, Eq)]
struct Human {
    name: String,
    age: u8
};
let human = Human {
    name: "Ayush".to_string(),
    age: 19
};

// serialize the data to bytes (Vec<u8>)
let human_bytes = serializer::to_bytes(&human).unwrap();

// deserialize the data from serialized bytes.
let deserialized_human = deserializer::from_bytes::<Human>(&human_bytes).unwrap();

assert_eq!(human, deserialized_human);

benchmark.

• Run cargo test -- --nocapture --ignored to run the benchmark tests.

running 3 tests
---- Small Data ----
rust_fr:        218 bytes
serde_json:     332 bytes
rmp_serde:      146 bytes
ciborium:       170 bytes
test tests::length_test_small_data ... ok
---- Medium Data ----
rust_fr:        14264 bytes
serde_json:     30125 bytes
rmp_serde:      10731 bytes
ciborium:       18347 bytes
test tests::length_test_medium_data ... ok
---- Large Data ----
rust_fr:        139214 bytes
serde_json:     367595 bytes
rmp_serde:      157219 bytes
ciborium:       198277 bytes
test tests::length_test_large_data ... ok

test result: ok. 3 passed; 0 failed; 0 ignored; 0 measured; 4 filtered out; finished in 0.01s

why?

The goal was to learn/understand. I wrote this so I can learn how serde internally works and how to encode data into bytes that can ultimately be transferred over the wire or elsewhere.

format specification.

• The format is non-self-describing.
• Primitive types are serialized as is.
  • bool: 0 -> false, 1 -> true (1 bit)
  • i8, i16, i32, i64: as is.
  • u8, u16, u32, u64: as is.
  • f32, f64: as is.
  • char: as u32 (4 bytes)
• Delimiters are used to separate different types of data.
• String, Byte and Map Delimiters are 1 byte long while all other delimiters are 3 bits long.
• Delimiters:
  • String = 134; 0b10000110
  • Byte = 135; 0b10000111
  • Unit = 2; 0b010
  • Seq = 3; 0b011
  • SeqValue = 4; 0b100
  • Map = 139; 0b10001011
  • MapKey = 6; 0b110
  • MapValue = 7; 0b111
• String, Bytes, Unit, Option are serialized as:
  • str: bytes + STRING_DELIMITER
  • bytes: bytes + BYTE_DELIMITER
  • unit: UNIT (null)
  • option: None -> unit(), Some -> self
• Structs are serialized as:
  • unit_struct: unit()
  • newtype_struct: self
  • tuple_struct: seq()
• Enums are serialized as:
  • unit_variant: variant_index
  • newtype_variant: variant_index + self
  • tuple_variant: variant_index + tuple()
  • struct_variant: variant_index + struct()
• seq(): Sequences are serialized as:
  • SEQ_DELIMITER + value_1 + SEQ_VALUE_DELIMITER + value_2 + SEQ_VALUE_DELIMITER + ... + SEQ_DELIMITER
• map(): Maps are serialized as:
  • key_1 + MAP_KEY_DELIMITER + value_1 + MAP_VALUE_DELIMITER + key_2 + MAP_KEY_DELIMITER + value_2 + MAP_VALUE_DELIMITER + ... + MAP_DELIMITER
• Tuples and Structs are serialized as:
  • tuple: seq()
  • struct: map()

license.

It's MIT so you can do whatever you want. You can still read it here.