@bin_float.BinFloat

This page tracks the current repository implementation and is written as the 0.2.0 API baseline.

Representation

Finite values are stored as:

significand * 2^exponent2

with an attached working precision.

Constructors and Stored Form

• BinFloat::make
• BinFloat::zero
• BinFloat::one
• BinFloat::inf
• BinFloat::nan
• BinFloat::from_int
• BinFloat::from_bigint
• BinFloat::from_float
• BinFloat::from_double

Notes:

• Public finite constructors normalize the stored representation.
• compare aborts on NaN.
• sign() currently returns Sign::Zero for NaN.

Access, Normalization, and Comparison

• classify
• precision
• sign
• significand
• exponent2
• is_zero
• normalized
• with_precision
• ulp
• compare
• min
• max
• clamp

Notes:

• clamp aborts if the bounds are unordered or NaN.
• ulp() returns NaN on non-finite inputs.

Arithmetic and Conversion

• neg
• abs
• add
• sub
• mul
• div

Supported operators:

• +
• -
• *
• /
• unary -

Special-value notes:

• NaN generally propagates.
• inf - inf with opposite signs becomes NaN.
• Division by zero produces inf or NaN depending on the numerator class.

Checked Arithmetic API

Direct exported helpers:

• sqrt_bounds_for_precision
• sqrt_for_precision
• compare_checked
• div_checked
• sqrt
• pow_int

Checked-behavior notes:

• sqrt_bounds_for_precision and sqrt_for_precision require non-negative finite inputs.
• compare_checked returns a structured unordered-comparison error on NaN.
• div_checked returns a structured division-by-zero error.
• pow_int returns a structured division-by-zero error for negative powers of zero.

Trait Surface

BinFloat currently implements:

• @def.Floating
• @arithmetic.SqrtChecked
• @arithmetic.DivChecked
• @arithmetic.CompareChecked
• @arithmetic.PowNatChecked
• @arithmetic.PowIntChecked
• Eq, Add, Sub, Mul, Div, Neg, Show