field_values.hh

Go to the documentation of this file.

/*!
 *
 * Copyright (C) 2015 Technical University of Liberec.  All rights reserved.
 * 
 * This program is free software; you can redistribute it and/or modify it under
 * the terms of the GNU General Public License version 3 as published by the
 * Free Software Foundation. (http://www.gnu.org/licenses/gpl-3.0.en.html)
 * 
 * This program is distributed in the hope that it will be useful, but WITHOUT
 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
 * FOR A PARTICULAR PURPOSE.  See the GNU General Public License for more details.
 *
 * 
 * @file    field_values.hh
 * @brief   
 */

#ifndef FIELD_VALUES_HH_
#define FIELD_VALUES_HH_

#include <string.h>                                    // for memcpy
#include <boost/core/explicit_operator_bool.hpp>       // for optional::oper...
#include <boost/optional/optional.hpp>                 // for get_pointer
#include <cmath>                                       // for abs
#include <cstdlib>                                     // for abs
#include <limits>                                      // for numeric_limits
#include <new>                                         // for operator new[]
#include <ostream>                                     // for operator&, ope...
#include <string>                                      // for basic_string
#include <system/exceptions.hh>                        // for THROW, ExcStream
#include <type_traits>                                 // for is_floating_point
#include <vector>                                      // for vector
#include <armadillo>

#include "system/fmt/format.h"
#include "input/accessors.hh"                          // for Array, Iterator
#include "input/accessors_impl.hh"                     // for Array::size
#include "input/input_exception.hh"                    // for ExcFV_Input::~...
#include "input/type_base.hh"                          // for Array, String
#include "input/type_generic.hh"                       // for Parameter
#include "input/type_selection.hh"                     // for Selection
#include "system/armor.hh"                             // for Armor::Array

namespace IT=Input::Type;

/**
 * @file
 *
 * This file contains various dispatch classes to simplify implementation of Fields. Essential is class template
 * @p FieldValues_  which provides unified access and initialization to scalar, vector and matrix type object in Armadillo library
 */

TYPEDEF_ERR_INFO( EI_InputMsg, const string );
DECLARE_INPUT_EXCEPTION( ExcFV_Input, << "Wrong field value input: " << EI_InputMsg::val );

typedef unsigned int FieldEnum;



namespace internal {

// Helper functions to get scalar type name
/*
std::string type_name_(double);
std::string type_name_(int);
std::string type_name_(FieldEnum);
*/

/**
 * InputType dispatch from elementary type @p ET of FieldValues_ to elementary Input::Type, i.e. descendant of Input::Type::Scalar.
 */
template <class ET>
struct InputType { typedef Input::Type::Double type; };

template <>
struct InputType<int> { typedef Input::Type::Integer type; };

template <>
struct InputType<FieldEnum> { typedef Input::Type::Selection type; };


////////////////////////////////////////////////////////////////////////////////////////////////////////////////
// resolution of Value::return_type

// general element type
template<int NRows, int NCols, class ET>
struct ReturnType { typedef typename arma::Mat<ET>::template fixed<NRows, NCols> return_type; };

template<class ET>
struct ReturnType<1,1,ET> { typedef ET return_type; };

template <class ET>
struct ReturnType<0,1,ET> { typedef arma::Col<ET>  return_type; };

template <int NRows, class ET>
struct ReturnType<NRows,1,ET> { typedef typename arma::Col<ET>::template fixed<NRows> return_type; };


// FiledEnum element type - this just returns types with ET=unsigned int, however input should be different
template<int NRows, int NCols>
struct ReturnType<NRows, NCols, FieldEnum> { typedef typename arma::Mat<unsigned int>::template fixed<NRows, NCols> return_type; };

template<>
struct ReturnType<1,1, FieldEnum> { typedef unsigned int return_type; };

template <>
struct ReturnType<0,1, FieldEnum> { typedef arma::Col<unsigned int> return_type; };

template <int NRows>
struct ReturnType<NRows,1, FieldEnum> { typedef typename arma::Col<unsigned int>::template fixed<NRows> return_type; };


// Resolution of helper functions for raw constructor
template <class RT> inline RT & set_raw_scalar(RT &, double *raw_data) { return *raw_data;}
template <class RT> inline RT & set_raw_scalar(RT &, int *raw_data) { return *raw_data;}
template <class RT> inline RT & set_raw_scalar(RT &, FieldEnum *raw_data) { return *raw_data;}

template <class RT> inline RT & set_raw_vec(RT &val, double *raw_data) { arma::access::rw(val.mem) = raw_data; return val;}
template <class RT> inline RT & set_raw_vec(RT &val, int *raw_data) { arma::access::rw(val.mem) = raw_data; return val;}
template <class RT> inline RT & set_raw_vec(RT &val, FieldEnum *raw_data) { arma::access::rw(val.mem) = raw_data; return val;}

template <class RT> inline RT & set_raw_fix(RT &val, double *raw_data) {  val = RT(raw_data); return val;}
template <class RT> inline RT & set_raw_fix(RT &val, int *raw_data) { val = RT(raw_data); return val;}
template <class RT> inline RT & set_raw_fix(RT &val, FieldEnum *raw_data) { val = RT(raw_data); return val;}



// Resolution of input accessor for vector values of enums,
template <class ET>
struct AccessTypeDispatch { typedef ET type;};
template <>
struct AccessTypeDispatch<unsigned int> { typedef Input::Enum type; };




/**
 * Initialize an Armadillo matrix from the input.
 * Since only limited methods are used we can use this template to initialize StringTensor as well.
 * Used methodas are: at(row,col), zeros()
 */
template<class MatrixType>
void init_matrix_from_input( MatrixType &value, Input::Array rec ) {
    typedef typename MatrixType::elem_type ET;
    unsigned int nrows = value.n_rows;
    unsigned int ncols = value.n_cols;

    Input::Iterator<Input::Array> it = rec.begin<Input::Array>();
    if (it->size() == 1 && nrows == ncols) {
        // square tensor
        // input = 3  expands  to [ [ 3 ] ]; init to  3 * (identity matrix)
        // input = [1, 2, 3] expands to [[1], [2], [3]]; init to diag. matrix
        // input = [1, 2, 3, .. , (N+1)*N/2], ....     ; init to symmetric matrix [ [1 ,2 ,3], [2, 4, 5], [ 3, 5, 6] ]
        if (rec.size() == 1)  {// scalar times identity
            value.zeros();
            ET scalar=*(it->begin<ET>());
            for(unsigned int i=0; i< nrows; i++) value.at(i,i)=scalar;
        } else if (rec.size() == nrows) { // diagonal vector
            value.zeros();
            for(unsigned int i=0; i< nrows; i++, ++it) value.at(i,i)=*(it->begin<ET>());
        } else if (rec.size() == (nrows+1)*nrows/2) { // symmetric part
            for( unsigned int row=0; row<nrows; row++)
                for( unsigned int col=0; col<ncols; col++)
                    if (row <= col) {
                        value.at(row,col) = *(it->begin<ET>());
                        ++it;
                    } else value.at(row,col) = value.at(col,row);
        } else {
            THROW( ExcFV_Input()
                    << EI_InputMsg(
                            fmt::format("Initializing symmetric matrix {:d}x{:d} by vector of wrong size {:d}, "
                                    "should be 1, {:d}, or {:d}.",
                                    nrows, ncols, rec.size(), nrows, (nrows+1)*nrows/2))
                    << rec.ei_address()

                 );
        }

    } else {
        // accept only full tensor
        if (rec.size() == nrows && it->size() == ncols) {

            for (unsigned int row = 0; row < nrows; row++, ++it) {
                if (it->size() != ncols)
                    THROW( ExcFV_Input() << EI_InputMsg("Wrong number of columns.")
                                         << rec.ei_address());
                Input::Iterator<ET> col_it = it->begin<ET>();
                for (unsigned int col = 0; col < ncols; col++, ++col_it)
                    value.at(row, col) = *col_it;
            }
        } else {
            THROW( ExcFV_Input()
                    << EI_InputMsg(
                            fmt::format("Initializing symmetric matrix {:d}x{:d} by vector of wrong size {:d}x{:d}.",
                                    nrows, ncols, rec.size(), it->size()))
                    << rec.ei_address()
                 );
        }
    }
}

/**
 * Initialize an Armadillo vector from the input.
 * Since only limited methods are used we can use this template to initialize StringTensor as well.
 * Used methodas are: at(row,col), zeros()
 */
template<class VectorType>
void init_vector_from_input( VectorType &value, Input::Array rec ) {
    typedef typename VectorType::elem_type ET;
    unsigned int nrows = value.n_rows;

    typedef typename AccessTypeDispatch<ET>::type InnerType;
    Input::Iterator<InnerType> it = rec.begin<InnerType>();

    if ( rec.size() == 1 ) {
        for(unsigned int i=0; i< nrows; i++)
            value.at(i)=ET(*it);
    } else if ( rec.size() == nrows ) {
        for(unsigned int i=0; i< nrows; i++, ++it) {
            value.at(i)=ET(*it);
        }
    } else {
        THROW( ExcFV_Input()
                << EI_InputMsg(
                        fmt::format("Initializing vector of size {:d} by vector of size {%d.}",
                            nrows, rec.size()))
                << rec.ei_address()
             );
    }
}



} // namespace internal



/**
 * Template for class representing all possible Filed values. It is just common interface to
 * scalar (double, int) values, vector and tensor values with fixed size and vector/tensor values with variable size.
 *
 * ET is type of elements, n_cols and n_rows gives fixed dimensions of the tensor value (nx1 is vector, 1x1 is scalar,
 * 0x1 is variable size vector, 0x0 is variable size tensor (not implemented yet) )
 *
 *
 */
template <int NRows, int NCols, class ET>
class FieldValue_ {
public:
    typedef ET element_type;
    typedef typename internal::ReturnType<NRows, NCols, ET>::return_type return_type;
    typedef typename internal::InputType<ET>::type ElementInputType;
    typedef Input::Array AccessType;
    const static int NRows_ = NRows;
    const static int NCols_ = NCols;
    const static int rank_ = 2;

    static std::string type_name() { return fmt::format("R[{:d},{:d}]", NRows, NCols); }
    static IT::Array get_input_type() {
        if (NRows == NCols) {
            // for square tensors allow initialization by diagonal vector, etc.
            return IT::Array( IT::Array( IT::Parameter("element_input_type"), 1), 1 );
        }
        else {
            return IT::Array( IT::Array( IT::Parameter("element_input_type"), NCols, NCols), NRows, NRows );
        }

    }
    static constexpr bool is_scalable() {
        return std::is_floating_point<element_type>::value;
    }


    inline FieldValue_(return_type &val) : value_(val) {}
    inline static const return_type &from_raw(return_type &val, ET *raw_data) {return internal::set_raw_fix(val, raw_data);}
    const ET * mem_ptr() const {
        return value_.memptr();
    }

    /// Casts value stored in Armor::Array to return type.
    inline static return_type get_from_array(const Armor::Array<element_type> &arr, uint idx) {
        return arr.template mat<NRows, NCols>(idx);
    }

    void init_from_input( AccessType rec ) {
        internal::init_matrix_from_input(value_, rec);
    }

    void set_n_comp(unsigned int) {};
    inline unsigned int n_cols() const
        { return NCols; }
    inline unsigned int n_rows() const
        { return NRows; }
    inline ET &operator() ( unsigned int i, unsigned int j)
        { return value_.at(i,j); }
    inline ET operator() ( unsigned int i, unsigned int j) const
        { return value_.at(i,j); }
    inline operator return_type() const
        { return value_;}

    // Set value to matrix of zeros.
    void zeros() {
        value_.zeros();
    }
    // Set value to identity matrix.
    void eye() {
        value_.eye();
    }
    // Set value to matrix of ones.
    void ones() {
        value_.ones();
    }
    // Elementwise equivalence.
    bool equal_to(const  return_type &other) {
        return arma::max(arma::max(arma::abs(value_ - other))) < 4*std::numeric_limits<ET>::epsilon();
    }
    // Multiplied value_ by double coefficient
    void scale(double scale_coef) {
        if (is_scalable())
            for( unsigned int row=0; row<n_rows(); row++)
                for( unsigned int col=0; col<n_cols(); col++)
                    value_.at(row,col) = scale_coef * value_.at(row,col);
    }

private:
    return_type &value_;
};





/// **********************************************************************
/// Specialization for scalars
template <class ET>
class FieldValue_<1,1,ET> {
public:
    typedef ET element_type;
    typedef typename internal::ReturnType<1, 1, ET>::return_type return_type;
    typedef typename internal::InputType<ET>::type ElementInputType;
    typedef typename internal::AccessTypeDispatch<ET>::type AccessType;
    const static int NRows_ = 1;
    const static int NCols_ = 1;
    const static int rank_ = 0;

    static std::string type_name() { return "R"; }
    static IT::Parameter get_input_type()
    {
        return IT::Parameter("element_input_type");
    }
    static constexpr bool is_scalable() {
        return std::is_floating_point<element_type>::value;
    }

    inline FieldValue_(return_type &val) : value_(val) {}

    /**
     * Returns reference to the return_type (i.e. double, or arma::vec or arma::mat); with data provided by the parameter @p raw_data.
     * A reference to a work space @p val has to be provided for efficient work with vector and matrix values.
     */
    inline static const return_type &from_raw(return_type &val, ET *raw_data) {return internal::set_raw_scalar(val, raw_data);}
    const ET * mem_ptr() const { return &(value_); }

    /// Casts value stored in Armor::Array to return type.
    inline static return_type get_from_array(const Armor::Array<element_type> &arr, uint idx) {
        return arr.scalar(idx);
    }

    void init_from_input( AccessType val ) { value_ = return_type(val); }

    void set_n_comp(unsigned int) {};
    inline unsigned int n_cols() const
        { return 1; }
    inline unsigned int n_rows() const
        { return 1; }
    inline ET &operator() ( unsigned int, unsigned int )
        { return value_; }
    inline ET operator() ( unsigned int, unsigned int) const
        { return value_; }
    inline operator return_type() const
        { return value_;}

    // Set value to matrix of zeros.
    void zeros() {
        value_=0;
    }
    // Set value to identity matrix.
    void eye() {
        value_=1;
    }
    // Set value to matrix of ones.
    void ones() {
        value_=1;
    }
    bool equal_to(const  return_type &other) {
        return std::abs((double)value_ - other) < 4*std::numeric_limits<ET>::epsilon();
    }
    // Multiplied value_ by double coefficient
    void scale(double scale_coef) {
        if (is_scalable())
            value_ = scale_coef * value_;
    }

private:
    return_type &value_;
};




/// **********************************************************************
/// Specialization for variable size vectors
template <class ET>
class FieldValue_<0,1,ET> {
public:
    typedef ET element_type;
    typedef typename internal::ReturnType<0, 1, ET>::return_type return_type;
    typedef typename internal::InputType<ET>::type ElementInputType;
    typedef Input::Array AccessType;
    const static int NRows_ = 0;
    const static int NCols_ = 1;
    const static int rank_ = 10;


    static std::string type_name() { return "R[n]"; }
    static IT::Array get_input_type() {
        return IT::Array( IT::Parameter("element_input_type"), 1);
    }
    static constexpr bool is_scalable() {
        return std::is_floating_point<element_type>::value;
    }
    inline static const return_type &from_raw(return_type &val, ET *raw_data) {return internal::set_raw_vec(val, raw_data);}
    const ET * mem_ptr() const { return value_.memptr(); }

    /// Casts value stored in Armor::Array to return type.
    inline static return_type get_from_array(const Armor::Array<element_type> &arr, uint idx) {
        return arr.template vec<NRows_>(idx);
    }

    inline FieldValue_(return_type &val) : value_(val) {}


    void init_from_input( AccessType rec ) {
        internal::init_vector_from_input(value_, rec);
    }

    void set_n_comp(unsigned int n_comp) { value_ = return_type(n_comp,1); };
    inline unsigned int n_cols() const
        { return 1; }
    inline unsigned int n_rows() const
        { return value_.n_rows; }
    inline ET &operator() ( unsigned int i, unsigned int )
        { return value_.at(i); }
    inline ET operator() ( unsigned int i, unsigned int ) const
        { return value_.at(i); }

    inline operator return_type() const
        { return value_;}

    // Set value to matrix of zeros.
    void zeros() {
        value_.zeros();
    }
    // Set value to identity matrix.
    void eye() {
        value_.ones();
    }
    // Set value to matrix of ones.
    void ones() {
        value_.ones();
    }
    // Elementwise equivalence.
    bool equal_to(const  return_type &other) {
        return arma::max(arma::abs(value_ - other)) < 4*std::numeric_limits<ET>::epsilon();
    }
    // Multiplied value_ by double coefficient
    void scale(double scale_coef) {
        if (is_scalable())
            for(unsigned int i=0; i< n_rows(); i++) {
                value_.at(i) = scale_coef * value_.at(i);
            }
    }


private:
    return_type &value_;
};

/// **********************************************************************
/// Specialization for fixed size vectors
template <int NRows, class ET>
class FieldValue_<NRows,1,ET> {
public:
    typedef ET element_type;
    typedef typename internal::ReturnType<NRows, 1, ET>::return_type return_type;
    typedef typename internal::InputType<ET>::type ElementInputType;
    typedef Input::Array AccessType;
    const static int NRows_ = NRows;
    const static int NCols_ = 1;
    const static int rank_ = 1;


    static std::string type_name() { return fmt::format("R[{:d}]", NRows); }
    static IT::Array get_input_type() {
        return IT::Array( IT::Parameter("element_input_type"), 1, NRows);
    }
    static constexpr bool is_scalable() {
        return std::is_floating_point<element_type>::value;
    }

    inline FieldValue_(return_type &val) : value_(val) {}
    inline static const return_type &from_raw(return_type &val, ET *raw_data) {return internal::set_raw_fix(val, raw_data);}
    const ET * mem_ptr() const { return value_.memptr(); }

    /// Casts value stored in Armor::Array to return type.
    inline static return_type get_from_array(const Armor::Array<element_type> &arr, uint idx) {
        return arr.template vec<NRows>(idx);
    }

    void init_from_input( AccessType rec ) {
        internal::init_vector_from_input(value_, rec);
    }

    void set_n_comp(unsigned int) {};
    inline unsigned int n_cols() const
        { return 1; }
    inline unsigned int n_rows() const
        { return NRows; }
    inline ET &operator() ( unsigned int i, unsigned int )
        { return value_.at(i); }
    inline ET operator() ( unsigned int i, unsigned int ) const
        { return value_.at(i); }

    inline operator return_type() const
        { return value_;}

    // Set value to matrix of zeros.
    void zeros() {
        value_.zeros();
    }
    // Set value to identity matrix.
    void eye() {
        value_.ones();
    }
    // Set value to matrix of ones.
    void ones() {
        value_.ones();
    }
    // Elementwise equivalence.
    bool equal_to(const  return_type &other) {
        return arma::max(arma::abs(value_ - other)) < 4*std::numeric_limits<ET>::epsilon();
    }
    // Multiplied value_ by double coefficient
    void scale(double scale_coef) {
        if (is_scalable())
            for(unsigned int i=0; i< n_rows(); i++) {
                value_.at(i) = scale_coef * value_.at(i);
            }
    }

private:
    return_type &value_;
};



//****************************************************************************
// string tensor (for FieldFormula)


/**
 * Mimics arma::mat<std::string>. Used in FieldFormula
 */
class StringTensor {
public:
    typedef std::string elem_type;

    StringTensor( unsigned int n_rows, unsigned int n_cols )
    : n_rows(n_rows),n_cols(n_cols), values_(n_rows*n_cols) {}

    std::string & at(unsigned int row) { return at(row,0); }
    std::string & at(unsigned int row, unsigned int col) { return values_[col*n_rows+row]; }

    void zeros() {
        for( auto &elem: values_) elem = "0.0";
    }
    unsigned int n_rows;
    unsigned int n_cols;
private:
    std::vector<std::string>  values_;

};


template<int NRows, int NCols>
struct StringTensorInput
{
    typedef Input::Array AccessType;
    static IT::Array get_input_type() {
        if (NRows == NCols) {
            // for square tensors allow initialization by diagonal vector, etc.
            return IT::Array( IT::Array( IT::String(), 1), 1 );
        }
        else {
            return IT::Array( IT::Array( IT::String(), NCols, NCols), NRows, NRows );
        }

    }

    static void init_from_input(StringTensor &tensor, AccessType input) {
        internal::init_matrix_from_input(tensor, input);
    }
};

template<>
struct StringTensorInput<1,1> {
    typedef std::string AccessType;
    static IT::String get_input_type() { return IT::String(); }
    static void init_from_input(StringTensor &tensor, AccessType input) {
        tensor.at(0,0) = input;
    }
};

template<int Nrows>
struct StringTensorInput<Nrows,1> {
    typedef Input::Array AccessType;
    static IT::Array get_input_type() {
        return IT::Array( IT::String(), 1);
    }
    static void init_from_input(StringTensor &tensor, AccessType input) {
        internal::init_vector_from_input(tensor, input);
    }
};




/**
 * Class that provides common template-less interface to all templated Field structes.
 *
 * Maybe we can delete this since common interface will be given by "Quantity".
 */

template <int spacedim>
struct FieldValue {
    // typedefs for possible field values
    typedef FieldValue_<1,1,int>            Integer;
    typedef FieldValue_<1,1, FieldEnum>     Enum;
    typedef FieldValue_<0,1, FieldEnum>     EnumVector;
    typedef FieldValue_<1,1,double>         Scalar;
    typedef FieldValue_<spacedim,1,double>  VectorFixed;
    typedef FieldValue_<0,1,double>         Vector;
    typedef FieldValue_<0,1,int>            IntVector;
    typedef FieldValue_<spacedim,spacedim,double> TensorFixed;
};










#endif /* FIELD_VALUES_HH_ */