Traits overview

The library's organising principle is a small set of traits. Each trait captures one abstraction, with concrete types implementing the trait generically over the float type via FloatExt.

The trait map

FloatExt

The trait every numerical type is generic over:

pub trait FloatExt: Float + …  {
    // arithmetic, comparison, conversions
    // closed-form math: erf, erfc, gamma, lgamma, beta, …
    // RNG-friendly: ZERO, ONE, TWO, PI, TAU, EPSILON, …
}

f32 and f64 are blanket-impl'd; user code rarely names the trait explicitly except as a bound:

pub fn my_pricer<T: FloatExt>(s0: T, k: T, t: T) -> T { /* … */ }

The SimdFloatExt subset is what bulk samplers rely on for f32x8 / f64x4 operations.

ProcessExt<T>

Every stochastic process implements ProcessExt<T>:

pub trait ProcessExt<T: FloatExt> {
    fn sample(&self) -> Array1<T>;
    fn sample_par(&self, m: usize) -> Array2<T>;
    // optional: malliavin-style methods via MalliavinExt
}

sample() produces a single path of length n (the steps); sample_par(m) produces m independent paths in parallel via Rayon.

DistributionExt<T>

Closed-form moments / pdf / cdf / characteristic function — implemented by 18/19 distributions in the library (5 named no-closed-form unimplemented!() cases on specific moments per the project status memo).

pub trait DistributionExt<T: FloatExt> {
    fn pdf(&self, x: T) -> T;
    fn cdf(&self, x: T) -> T;
    fn characteristic_fn(&self, t: T) -> Complex<T>;
    fn mean(&self) -> T;
    fn variance(&self) -> T;
    fn skewness(&self) -> T { unimplemented!("not implemented for {}", type_name) }
    fn kurtosis(&self) -> T { unimplemented!("not implemented for {}", type_name) }
}

Default behaviour: unimplemented moments panic with a helpful message — never silently return zero. This is a hard project rule.

Calibrator<T> and CalibrationResult

Calibration is Result-based with associated types for the parameter struct and error:

pub trait Calibrator<T: FloatExt> {
    type Params;
    type Error;
    fn calibrate(&self, market: &MarketData<T>) -> Result<CalibrationResult<Self>, Self::Error>;
}

CalibrationResult carries the calibrated Params, the final RMSE, iteration count, and a success flag. See the Quant catalog for the calibrators that ship today.

GreeksExt<T> and the Greeks aggregator

All pricers that derive from ModelPricer<T> automatically get a default greeks() implementation via finite differences. MC pricers override the default with a single-pass computation that re-uses the path bundle.

pub struct Greeks<T> {
    pub delta: T, pub gamma: T,
    pub vega: T,  pub theta: T, pub rho: T,
    pub vanna: T, pub charm: T,
    pub volga: T, pub veta: T,
}

See Quant for the pricers that expose Greeks.