#fixed-point #math #soroban #smart-contracts #operations #traits #wasm

no-std soroban-fixed-point-math

Fixed point math support for Soroban smart contracts

4 stable releases

1.2.0 Nov 1, 2024
1.1.1 Jun 25, 2024
1.1.0 Apr 17, 2024
1.0.0 Dec 6, 2023

#506 in Magic Beans

Download history 25/week @ 2024-07-26 7/week @ 2024-08-02 4/week @ 2024-08-09 1/week @ 2024-08-16 29/week @ 2024-09-13 43/week @ 2024-09-20 56/week @ 2024-09-27 18/week @ 2024-10-04 22/week @ 2024-10-11 7/week @ 2024-10-18 1/week @ 2024-10-25 118/week @ 2024-11-01

148 downloads per month

MIT license

62KB
1.5K SLoC

soroban-fixed-point-math

A fixed-point math library for Soroban smart contacts.

Safety

This is experimental software and is provided on an "as is" and "as available" basis.

We do not give any warranties and will not be liable for any loss incurred through any use of this codebase.

Usage

Add this to your Cargo.toml:

[dependencies]
soroban_fixed_point_math = "<desired version>"

FixedPoint Trait

The FixedPoint trait is implemented for the Rust generic integers i64, u64, i128, and u128. Any overflow or divide by zero that occurs during the fixed point math calculation will return a value of None, in line with Rust's existing checked math functions.

Phantom overflows are errors that result from an overflow during an intermediate calculation, but the expected result would be within bounds. This trait manages this differently for each implemented integer type:

  • i64 and u64
    • In the event of a phantom overflow, the intermediary computation gets scaled to i128 / u128 and is retried.
  • i128 and u128
    • In the event of a phantom overflow, the function will terminate and return None. If you need larger numbers to be handled, please use the SorobanFixedPoint trait.

SorobanFixedPoint Trait

The SorobanFixedPoint trait is implemented for the Soroban host types I256 and U256. The trait will panic if any I256 or U256 arithmetic error occurs, as host function calls panic by default.

Additional implementations of the SorobanFixedPoint trait are included for i128 and u128 integer types to support larger intermediary computations. This removes the majority of phantom overflow events for 128 bit fixed point math.

Examples

For any supported type implementing FixedPoint, you can perform fixed-point operations like this:

use soroban_fixed_point_math::FixedPoint;

let x: u64 = 1_5000000;
let y: u64 = 2_0000000;
assert_eq!(x.fixed_mul_floor(y, 1_0000000).unwrap(), 3_0000000);

For any support type implementing SorobanFixedPoint, you can perform fixed-point operations like this:

use soroban_fixed_point_math::SorobanFixedPoint;
use soroban_sdk::{Env, I256};

let env: Env //... from your contract, or Env::default() in a test
let x: I256 = I256::from_i32(&env, 1_5000000);
let y: I256 = I256::from_i32(&env, 2_0000000);
assert_eq!(x.fixed_mul_ceil(&x, &y, I256::from_i32(&env, 1_0000000)), 3_0000000);

Acknowledgements

This library was inspired by or directly modified from many sources, primary:

WASM

The WASM target wasm32-unknown-unknown is supported.

Contributions

Contributions are welcome. Please check out the contribution guide (TODO)!

License

This library is released under the MIT License.

Dependencies

~11–15MB
~314K SLoC