scalars

Minimal numeric traits for Rust. Zero dependencies.

Replacement for num-traits with single-purpose traits instead of monolithic kitchen-sink traits.

Traits

Identities & Operations

• Zero — fn zero(), fn is_zero()
• One — fn one(), fn is_one()
• Inv — fn inv() with associated Output type
• Sqrt — fn sqrt()

Exponential

• Exp — fn exp(), fn exp2() with associated Output type (supports Bivector → Rotor)

Trigonometry

• Trigonometry — fn sin_cos() -> [Self; 2], fn sin(), fn cos(), fn tan()
• InverseTrigonometry — fn asin(), fn acos(), fn atan(), fn atan2()

Ordering

• Clamp — fn min(), fn max(), fn clamp()

Parsing

• FromStrRadix — fn from_str_radix() with associated Error type

Supertraits

• Numeric — Copy + PartialEq + Add + Sub + Mul + Zero + One
• Integer — Numeric + Eq + Ord + Div + Rem
• Real — Numeric + PartialOrd + Div + Neg + Inv + Sqrt + Exp + Trigonometry + InverseTrigonometry + Clamp

All supertraits have blanket implementations.

Implementations

• f32, f64 — all traits including Real
• i8..i128, u8..u128, isize, usize — Zero, One, Clamp, FromStrRadix, Integer

Example

use scalars::{Real, Zero};

fn normalize<T: Real>(values: &mut [T]) {
    let sum: T = values.iter().copied().fold(T::zero(), |accumulator, value| accumulator + value);
    if !sum.is_zero() {
        for value in values {
            *value = *value / sum;
        }
    }
}

AI-assisted development

This crate is designed for use with AI coding assistants. Each trait covers exactly one concept, all supertraits use blanket implementations, and there is no implicit behavior. The Exp trait's associated Output type enables geometric algebra types (Bivector → Rotor) alongside scalar floats.