Crate oxinum_float

Expand description

Arbitrary-precision floating-point arithmetic for the OxiNum ecosystem.

Provides wrappers over dashu-float’s FBig/DBig with convenience functions for elementary math: exp, ln, sqrt, pow, trigonometric functions (sin, cos, tan, atan, atan2), hyperbolic functions (sinh, cosh, tanh), and high-precision constants (pi, e, ln2).

Precision-control helpers (precision::with_precision, precision::epsilon, precision::ulp) wrap the underlying dashu-float precision primitives in an ergonomic free-function API.

§Why `Hash` is not implemented

DBig (FBig<HalfAway, 10>) intentionally does not implement core::hash::Hash. In IEEE 754 (and by extension in essentially every numeric library that takes the standard seriously) floating point values defy the Hash contract: distinct representations can compare equal (e.g. 1.0 and 1.00 at different precisions, or signed zeros +0 and -0), special values like NaN are not equal to themselves, and the canonical form needed to make hashing well-defined varies by use case (totalOrder vs. mathematical equality vs. bitwise identity). Rust’s standard library follows the same convention: f32 and f64 deliberately do not implement Hash. Callers who need a hash key should either use a canonicalised representation (e.g. a tuple of (significand, exponent, precision)) or wrap DBig in a newtype with an explicit, documented hashing policy.

§Features

serde (off by default) — enables serde::Serialize / Deserialize for DBig via dashu-float’s own serde feature.

Modules§

native: Native binary-base arbitrary-precision floating-point implementation.
precision: Precision-control helpers for the DBig wrapper.
round: Rounding modes re-exported from dashu-float so callers do not need a direct dependency on dashu-float for basic rounding-mode selection.

Structs§

Context: The context containing runtime information for the floating point number and its operations.
FBig: An arbitrary precision floating point number with arbitrary base and rounding mode.

Enums§

OxiNumError: Errors from OxiNum operations.
RoundingMode: Rounding modes for arbitrary-precision arithmetic.

Functions§

atan: Compute atan(x) using dashu-float’s binary conversion and Taylor approach internally. For |x| <= 0.5 we use the series directly; for larger |x| <= 1 we use the half-angle formula; for |x| > 1 we use the identity atan(x) = pi/2 - atan(1/x).
atan2: Compute atan2(y, x) with proper quadrant handling.
compute_e: Compute Euler’s number e to the given number of decimal digits.
compute_ln2: Compute ln(2) to the given number of decimal digits.
compute_pi: Compute pi to the given number of decimal digits of precision.
cos: Compute cos(x) using Taylor series with argument reduction.
cosh: Compute cosh(x) = (e^x + e^(-x)) / 2.
exp: Compute e^x with the given number of significant decimal digits.
ln: Compute ln(x) (natural logarithm) with the given precision.
pow: Compute base^exp for arbitrary float exponents.
sin: Compute sin(x) using Taylor series with argument reduction.
sinh: Compute sinh(x) = (e^x - e^(-x)) / 2.
sqrt: Compute the square root of x with the given precision.
tan: Compute tan(x) = sin(x) / cos(x).
tanh: Compute tanh(x) = sinh(x) / cosh(x).

Type Aliases§

BigFloat: Type alias for decimal big-float.
DBig: Multi-precision float number with decimal exponent and HalfAway rounding mode
OxiNumResult: Convenience alias for Result<T, OxiNumError>.