Module precision_boundary 

Precision boundary between FinancialValue (Decimal) and StatisticalValue (f64).

Every crossing between the two representations passes through one of the two checked helpers in this module. Silent fallbacks (d.to_f64().unwrap_or(0.0), Decimal::try_from(f).unwrap_or(Decimal::ONE)) are forbidden by the specs/precision/financial-precision-boundary.md published-language spec in qbot-core — every caller must propagate the error or surface it.

When to use

• decimal_to_f64_checked: entering a statistical kernel (covariance, regression, HRP, z-score comparison). The kernel consumes f64 because Decimal lacks the required math (sqrt, ln, eigendecomposition).
• f64_to_decimal_checked: exiting a statistical kernel back into any booked value (position size, hedge ratio, stop multiplier). The guard ensures NaN / ±Inf — which every well-behaved estimator produces on degenerate inputs — refuses to become a silent zero or one.

The functions return Result because the boundary is the correct place to raise a domain error: a covariance collapse, a regression that failed to converge, a rate-limiter returning ∞ backoff — all legitimate events the domain layer must decide how to handle. Converting them to arbitrary defaults hides real failures.

Reference

• Spec: specs/precision/financial-precision-boundary.md (qbot-core).
• Historical scars: hedge_sizing.rs:109 (NaN β → 1:1 hedge), pair_simulation/cost.rs:34 (silent zero z-score).

Enums

NonFiniteKind

Classification of a rejected non-finite f64 at a precision boundary.

PrecisionBoundaryError

Error returned by the precision-boundary conversion helpers.

Constants

PRECISION_SCALE

The canonical rounding scale for f64 → Decimal crossings.

Functions

decimal_to_f64_checked

Canonical Decimal → f64 conversion at a precision boundary.

f64_to_decimal_checked

Canonical f64 → Decimal conversion at a precision boundary.