math.hpp

Go to the documentation of this file.

#pragma once
 
#include <Eigen/Core>
#include <limits>
#include <vector>
 
namespace proxsuite {
namespace nlp {
 
template <typename T>
constexpr bool is_eigen_dense_type =
    std::is_base_of<Eigen::DenseBase<T>, T>::value;
 
template <typename T>
constexpr bool is_eigen_matrix_type =
    std::is_base_of<Eigen::MatrixBase<T>, T>::value;
 
template <typename T, typename T2 = void>
using enable_if_eigen_dense = std::enable_if_t<is_eigen_dense_type<T>, T2>;

#define PROXSUITE_NLP_DYNAMIC_TYPEDEFS(Scalar)                                 \
  using VectorXs = Eigen::Matrix<Scalar, Eigen::Dynamic, 1>;                   \
  using MatrixXs = Eigen::Matrix<Scalar, Eigen::Dynamic, Eigen::Dynamic>;      \
  using VectorOfVectors = std::vector<VectorXs>;                               \
  using VectorRef = Eigen::Ref<VectorXs>;                                      \
  using MatrixRef = Eigen::Ref<MatrixXs>;                                      \
  using ConstVectorRef = Eigen::Ref<const VectorXs>;                           \
  using ConstMatrixRef = Eigen::Ref<const MatrixXs>;                           \
  using VectorOfRef = std::vector<VectorRef>;                                  \
  using Vector3s = Eigen::Matrix<Scalar, 3, 1>;                                \
  using Vector6s = Eigen::Matrix<Scalar, 6, 1>;                                \
  using Matrix3Xs = Eigen::Matrix<Scalar, 3, Eigen::Dynamic>;                  \
  using Matrix6Xs = Eigen::Matrix<Scalar, 6, Eigen::Dynamic>;                  \
  using Matrix6s = Eigen::Matrix<Scalar, 6, 6>

template <typename _Scalar> struct math_types {
  using Scalar = _Scalar;
  PROXSUITE_NLP_DYNAMIC_TYPEDEFS(_Scalar);
};

namespace math {
template <typename MatType>
typename MatType::Scalar infty_norm(const Eigen::MatrixBase<MatType> &z) {
  if (z.rows() == 0 || z.cols() == 0) {
    return 0.;
  } else {
    return z.template lpNorm<Eigen::Infinity>();
  }
}
 
template <typename MatType>
typename MatType::Scalar infty_norm(const std::vector<MatType> &z) {
  const std::size_t n = z.size();
  typename MatType::Scalar out = 0.;
  for (std::size_t i = 0; i < n; i++) {
    out = std::max(out, infty_norm(z[i]));
  }
  return out;
}

template <typename Scalar> inline bool check_scalar(const Scalar value) {
  return std::isnan(value) || std::isinf(value);
}

template <typename Scalar>
bool scalar_close(const Scalar a, const Scalar b,
                  const Scalar prec = std::numeric_limits<Scalar>::epsilon()) {
  return std::abs(a - b) < prec * (1 + std::max(std::abs(a), std::abs(b)));
}
 
template <typename T, typename = std::enable_if_t<std::is_scalar<T>::value>>
bool check_value(const T &x) {
  static_assert(std::is_scalar<T>::value, "Parameter T should be scalar.");
  return check_scalar(x);
}
 
template <typename MatrixType>
bool check_value(const Eigen::MatrixBase<MatrixType> &x) {
  return (x.hasNaN() || (!x.allFinite()));
}
 
template <typename T> T sign(const T &x) {
  static_assert(std::is_scalar<T>::value, "Parameter T should be scalar.");
  return T((x > T(0)) - (x < T(0)));
}
 
} // namespace math
} // namespace nlp
} // namespace proxsuite