#ifndef MATRIXOPERATIONS_H_
#define MATRIXOPERATIONS_H_

#include <fsfw/retval.h>
#include <stdint.h>

#include <cmath>
#include <cstring>
#include <utility>

template <typename T1, typename T2 = T1, typename T3 = T2>
class MatrixOperations {
 public:
  // do not use with result == matrix1 or matrix2
  static void multiply(const T1 *matrix1, const T2 *matrix2, T3 *result, uint8_t rows1,
                       uint8_t columns1, uint8_t columns2) {
    if ((matrix1 == (T1 *)result) || (matrix2 == (T2 *)result)) {
      // SHOULDDO find an implementation that is tolerant to this
      return;
    }
    for (uint8_t resultColumn = 0; resultColumn < columns2; resultColumn++) {
      for (uint8_t resultRow = 0; resultRow < rows1; resultRow++) {
        result[resultColumn + columns2 * resultRow] = 0;
        for (uint8_t i = 0; i < columns1; i++) {
          result[resultColumn + columns2 * resultRow] +=
              matrix1[i + resultRow * columns1] * matrix2[resultColumn + i * columns2];
        }
      }
    }
  }

  static void transpose(const T1 *matrix, T2 *transposed, uint8_t size) {
    uint8_t row, column;
    transposed[0] = matrix[0];
    for (column = 1; column < size; column++) {
      transposed[column + size * column] = matrix[column + size * column];
      for (row = 0; row < column; row++) {
        T1 temp = matrix[column + size * row];
        transposed[column + size * row] = matrix[row + size * column];
        transposed[row + size * column] = temp;
      }
    }
  }

  // Overload transpose to support non symmetrical matrices
  // do not use with transposed == matrix && columns != rows
  static void transpose(const T1 *matrix, T2 *transposed, uint8_t rows, uint8_t columns) {
    uint8_t row, column;
    transposed[0] = matrix[0];
    if (matrix == transposed && columns == rows) {
      transpose(matrix, transposed, rows);
    } else if (matrix == transposed && columns != rows) {
      // not permitted
      return;
    }
    for (column = 0; column < columns; column++) {
      for (row = 0; row < rows; row++) {
        transposed[row + column * rows] = matrix[column + row * columns];
      }
    }
  }

  static void add(const T1 *matrix1, const T2 *matrix2, T3 *result, uint8_t rows, uint8_t columns) {
    for (uint8_t resultColumn = 0; resultColumn < columns; resultColumn++) {
      for (uint8_t resultRow = 0; resultRow < rows; resultRow++) {
        result[resultColumn + columns * resultRow] = matrix1[resultColumn + columns * resultRow] +
                                                     matrix2[resultColumn + columns * resultRow];
      }
    }
  }

  static void subtract(const T1 *matrix1, const T2 *matrix2, T3 *result, uint8_t rows,
                       uint8_t columns) {
    for (uint8_t resultColumn = 0; resultColumn < columns; resultColumn++) {
      for (uint8_t resultRow = 0; resultRow < rows; resultRow++) {
        result[resultColumn + columns * resultRow] = matrix1[resultColumn + columns * resultRow] -
                                                     matrix2[resultColumn + columns * resultRow];
      }
    }
  }

  static void addScalar(const T1 *matrix1, const T2 scalar, T3 *result, uint8_t rows,
                        uint8_t columns) {
    for (uint8_t resultColumn = 0; resultColumn < columns; resultColumn++) {
      for (uint8_t resultRow = 0; resultRow < rows; resultRow++) {
        result[resultColumn + columns * resultRow] =
            matrix1[resultColumn + columns * resultRow] + scalar;
      }
    }
  }

  static void multiplyScalar(const T1 *matrix1, const T2 scalar, T3 *result, uint8_t rows,
                             uint8_t columns) {
    for (uint8_t resultColumn = 0; resultColumn < columns; resultColumn++) {
      for (uint8_t resultRow = 0; resultRow < rows; resultRow++) {
        result[resultColumn + columns * resultRow] =
            matrix1[resultColumn + columns * resultRow] * scalar;
      }
    }
  }

  static bool isFinite(const T1 *inputMatrix, uint8_t rows, uint8_t cols) {
    for (uint8_t col = 0; col < cols; col++) {
      for (uint8_t row = 0; row < rows; row++) {
        if (not std::isfinite(inputMatrix[row * cols + cols])) {
          return false;
        }
      }
    }
    return true;
  }

  static void writeSubmatrix(T1 *mainMatrix, T1 *subMatrix, uint8_t subRows, uint8_t subCols,
                             uint8_t mainRows, uint8_t mainCols, uint8_t startRow,
                             uint8_t startCol) {
    if ((startRow + subRows > mainRows) or (startCol + subCols > mainCols)) {
      return;
    }
    for (uint8_t row = 0; row < subRows; row++) {
      for (uint8_t col = 0; col < subCols; col++) {
        mainMatrix[(startRow + row) * mainCols + (startCol + col)] = subMatrix[row * subCols + col];
      }
    }
  }

  static ReturnValue_t inverseMatrix(const T1 *inputMatrix, T1 *inverse, uint8_t size) {
    // Stopwatch stopwatch;
    T1 matrix[size][size], identity[size][size];
    // reformat array to matrix
    for (uint8_t row = 0; row < size; row++) {
      for (uint8_t col = 0; col < size; col++) {
        matrix[row][col] = inputMatrix[row * size + col];
      }
    }
    // init identity matrix
    std::memset(identity, 0.0, sizeof(identity));
    for (uint8_t diag = 0; diag < size; diag++) {
      identity[diag][diag] = 1;
    }
    // gauss-jordan algo
    // sort matrix such as no diag entry shall be 0
    for (uint8_t row = 0; row < size; row++) {
      if (matrix[row][row] == 0.0) {
        bool swaped = false;
        uint8_t rowIndex = 0;
        while ((rowIndex < size) && !swaped) {
          if ((matrix[rowIndex][row] != 0.0) && (matrix[row][rowIndex] != 0.0)) {
            for (uint8_t colIndex = 0; colIndex < size; colIndex++) {
              std::swap(matrix[row][colIndex], matrix[rowIndex][colIndex]);
              std::swap(identity[row][colIndex], identity[rowIndex][colIndex]);
            }
            swaped = true;
          }
          rowIndex++;
        }
        if (!swaped) {
          return returnvalue::FAILED;  // matrix not invertible
        }
      }
    }

    for (int row = 0; row < size; row++) {
      if (matrix[row][row] == 0.0) {
        uint8_t rowIndex;
        if (row == 0) {
          rowIndex = size - 1;
        } else {
          rowIndex = row - 1;
        }
        for (uint8_t colIndex = 0; colIndex < size; colIndex++) {
          std::swap(matrix[row][colIndex], matrix[rowIndex][colIndex]);
          std::swap(identity[row][colIndex], identity[rowIndex][colIndex]);
        }
        row--;
        if (row < 0) {
          return returnvalue::FAILED;  // Matrix is not invertible
        }
      }
    }
    // remove non diag elements in matrix (jordan)
    for (int row = 0; row < size; row++) {
      for (int rowIndex = 0; rowIndex < size; rowIndex++) {
        if (row != rowIndex) {
          double ratio = matrix[rowIndex][row] / matrix[row][row];
          for (int colIndex = 0; colIndex < size; colIndex++) {
            matrix[rowIndex][colIndex] -= ratio * matrix[row][colIndex];
            identity[rowIndex][colIndex] -= ratio * identity[row][colIndex];
          }
        }
      }
    }
    // normalize rows in matrix (gauss)
    for (int row = 0; row < size; row++) {
      for (int col = 0; col < size; col++) {
        identity[row][col] = identity[row][col] / matrix[row][row];
      }
    }
    std::memcpy(inverse, identity, sizeof(identity));
    return returnvalue::OK;  // successful inversion
  }

  static void inverseMatrixDimThree(const T1 *matrix, T1 *output) {
    int i, j;
    double determinant = 0;
    double mat[3][3] = {{matrix[0], matrix[1], matrix[2]},
                        {matrix[3], matrix[4], matrix[5]},
                        {matrix[6], matrix[7], matrix[8]}};

    for (i = 0; i < 3; i++) {
      determinant = determinant + (mat[0][i] * (mat[1][(i + 1) % 3] * mat[2][(i + 2) % 3] -
                                                mat[1][(i + 2) % 3] * mat[2][(i + 1) % 3]));
    }
    for (i = 0; i < 3; i++) {
      for (j = 0; j < 3; j++) {
        output[i * 3 + j] = ((mat[(j + 1) % 3][(i + 1) % 3] * mat[(j + 2) % 3][(i + 2) % 3]) -
                             (mat[(j + 1) % 3][(i + 2) % 3] * mat[(j + 2) % 3][(i + 1) % 3])) /
                            determinant;
      }
    }
  }

  static void skewMatrix(const T1 *vector, T2 *result) {
    // Input Dimension [3], Output [3][3]
    result[0] = 0;
    result[1] = -vector[2];
    result[2] = vector[1];
    result[3] = vector[2];
    result[4] = 0;
    result[5] = -vector[0];
    result[6] = -vector[1];
    result[7] = vector[0];
    result[8] = 0;
  }
};

#endif /* MATRIXOPERATIONS_H_ */