libcarla/include/system/boost/math/tools/luroth_expansion.hpp
2024-10-18 13:19:59 +08:00

139 lines
4.0 KiB
C++

// (C) Copyright Nick Thompson 2020.
// Use, modification and distribution are subject to the
// Boost Software License, Version 1.0. (See accompanying file
// LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_0.txt)
#ifndef BOOST_MATH_TOOLS_LUROTH_EXPANSION_HPP
#define BOOST_MATH_TOOLS_LUROTH_EXPANSION_HPP
#include <vector>
#include <ostream>
#include <iomanip>
#include <cmath>
#include <limits>
#include <stdexcept>
namespace boost::math::tools {
template<typename Real, typename Z = int64_t>
class luroth_expansion {
public:
luroth_expansion(Real x) : x_{x}
{
using std::floor;
using std::abs;
using std::sqrt;
using std::isfinite;
if (!isfinite(x))
{
throw std::domain_error("Cannot convert non-finites into a Luroth representation.");
}
d_.reserve(50);
Real dn1 = floor(x);
d_.push_back(static_cast<Z>(dn1));
if (dn1 == x)
{
d_.shrink_to_fit();
return;
}
// This attempts to follow the notation of:
// "Khinchine's constant for Luroth Representation", by Sophia Kalpazidou.
x = x - dn1;
Real computed = dn1;
Real prod = 1;
// Let the error bound grow by 1 ULP/iteration.
// I haven't done the error analysis to show that this is an expected rate of error growth,
// but if you don't do this, you can easily get into an infinite loop.
Real i = 1;
Real scale = std::numeric_limits<Real>::epsilon()*abs(x_)/2;
while (abs(x_ - computed) > (i++)*scale)
{
Real recip = 1/x;
Real dn = floor(recip);
// x = n + 1/k => lur(x) = ((n; k - 1))
// Note that this is a bit different than Kalpazidou (examine the half-open interval of definition carefully).
// One way to examine this definition is better for rationals (it never happens for irrationals)
// is to consider i + 1/3. If you follow Kalpazidou, then you get ((i, 3, 0)); a zero digit!
// That's bad since it destroys uniqueness and also breaks the computation of the geometric mean.
if (recip == dn) {
d_.push_back(static_cast<Z>(dn - 1));
break;
}
d_.push_back(static_cast<Z>(dn));
Real tmp = 1/(dn+1);
computed += prod*tmp;
prod *= tmp/dn;
x = dn*(dn+1)*(x - tmp);
}
for (size_t i = 1; i < d_.size(); ++i)
{
// Sanity check:
if (d_[i] <= 0)
{
throw std::domain_error("Found a digit <= 0; this is an error.");
}
}
d_.shrink_to_fit();
}
const std::vector<Z>& digits() const {
return d_;
}
// Under the assumption of 'randomness', this mean converges to 2.2001610580.
// See Finch, Mathematical Constants, section 1.8.1.
Real digit_geometric_mean() const {
if (d_.size() == 1) {
return std::numeric_limits<Real>::quiet_NaN();
}
using std::log;
using std::exp;
Real g = 0;
for (size_t i = 1; i < d_.size(); ++i) {
g += log(static_cast<Real>(d_[i]));
}
return exp(g/(d_.size() - 1));
}
template<typename T, typename Z2>
friend std::ostream& operator<<(std::ostream& out, luroth_expansion<T, Z2>& scf);
private:
const Real x_;
std::vector<Z> d_;
};
template<typename Real, typename Z2>
std::ostream& operator<<(std::ostream& out, luroth_expansion<Real, Z2>& luroth)
{
constexpr const int p = std::numeric_limits<Real>::max_digits10;
if constexpr (p == 2147483647)
{
out << std::setprecision(luroth.x_.backend().precision());
}
else
{
out << std::setprecision(p);
}
out << "((" << luroth.d_.front();
if (luroth.d_.size() > 1)
{
out << "; ";
for (size_t i = 1; i < luroth.d_.size() -1; ++i)
{
out << luroth.d_[i] << ", ";
}
out << luroth.d_.back();
}
out << "))";
return out;
}
}
#endif