ppforest2::math Namespace Reference

Numeric comparison utilities. More...

Functions
template<typename T>
bool	collinear (types::Vector< T > const &a, types::Vector< T > const &b)
	Check whether two vectors are collinear (parallel or anti-parallel).
template<typename A, typename B>
bool	is_approx (A a, B b)
	Overload using the default APPROX_THRESHOLD.
template<typename A, typename B, typename T>
bool	is_approx (A a, B b, T threshold)
	Check whether two scalars are approximately equal.
template<typename A, typename B>
bool	is_module_approx (A a, B b)
	Check whether the absolute values of two scalars are approximately equal.
int	proportion_to_count (float p, unsigned int total)
	Convert a proportion of a total into a count.

Detailed Description

Numeric comparison utilities.

Function Documentation

collinear()

template<typename T>

bool ppforest2::math::collinear	(	types::Vector< T > const &	a,
		types::Vector< T > const &	b )

Check whether two vectors are collinear (parallel or anti-parallel).

Parameters

a	First vector.
b	Second vector (same dimension as a).

Returns

True if |cos(angle)| ≈ 1.

is_approx() [1/2]

template<typename A, typename B>

bool ppforest2::math::is_approx ( A a, B b )

inline

Overload using the default APPROX_THRESHOLD.

is_approx() [2/2]

template<typename A, typename B, typename T>

bool ppforest2::math::is_approx ( A a, B b, T threshold )

inline

Check whether two scalars are approximately equal.

Parameters

a	First value.
b	Second value.
threshold	Maximum allowed absolute difference.

Returns

True if |a − b| < threshold.

is_module_approx()

template<typename A, typename B>

bool ppforest2::math::is_module_approx ( A a, B b )

inline

Check whether the absolute values of two scalars are approximately equal.

proportion_to_count()

int ppforest2::math::proportion_to_count ( float p, unsigned int total )

inline

Convert a proportion of a total into a count.

Turns a feature proportion (e.g. p_vars = 0.5) into a variable-selection count, rounding half to even (banker's rounding) to match R's round(), then clamping to at least 1 so a tiny proportion still selects one item (matching R's max(1L, ...)). Shared by the CLI and the R package so both resolve identical counts for the same proportion.

Half-to-even is computed by hand rather than via std::nearbyint so the result never depends on the current floating-point rounding mode. std::floor, the subtraction, and integer parity are all mode-independent and exact for these inputs, keeping results reproducible across compilers and platforms.

Parameters

p	Proportion in (0, 1].
total	Total number of items (>= 1).

Returns

Count in [1, total].