/* Example inspired by the toy example in https://www.r-bloggers.com/2020/06/understanding-lasso-and-ridge-regression-2/
 * blog post by Dr. Atakan Ekiz.
 * Here we wish to predict the number of shark attacks (that is, this number is our response variable),
 * using the following predictor variables:
 * - percentage of swimmers who watched the movie Jaws
 * - the number of swimmers in the water
 * - the average temperature of the day
 * - the price of your favorite tech stock of the day (totally uncorrelated variable) */

static char help[] = "Tests basic creation and destruction of PetscRegressor objects.\n\n";

#include <petscregressor.h>

int main(int argc, char **args)
{
  PetscRegressor regressor;
  PetscMPIInt    rank;
  Mat            X;
  Vec            y, y_predicted, coefficients;
  PetscScalar    intercept;

  PetscScalar y_array[20] = {98, 53, 39, 127, 73, 42, 71, 61, 83, 74, 85, 82, 62, 60, 43, 69, 67, 69, 85, 3}; // Number of shark attacks

  PetscScalar X_array[80] = {37.92934, 513, 92.89899, 137.2139, // % watched Jaws, #swimmers, temperature, stock price
                             52.77429, 451, 87.86271, 145.7987, //
                             60.84441, 456, 88.28927, 149.7299, //
                             26.54302, 546, 89.43875, 147.1180, //
                             54.29125, 431, 88.01132, 124.3068, //
                             55.06056, 355, 88.06297, 114.1730, //
                             44.25260, 557, 87.78536, 112.5773, //
                             44.53368, 398, 87.49603, 125.1628, //
                             44.35548, 498, 88.95234, 124.8483, //
                             41.09962, 406, 89.00630, 115.9223, //
                             45.22807, 610, 86.38794, 148.1111, //
                             40.01614, 452, 88.83585, 131.7050, //
                             42.23746, 429, 87.78222, 106.3717, //
                             50.64459, 450, 87.97008, 121.1523, //
                             59.59494, 337, 89.67538, 145.7158, //
                             48.89715, 383, 91.12611, 123.3896, //
                             44.88990, 282, 93.29563, 145.4085, //
                             40.88805, 366, 88.45329, 129.8872, //
                             41.62828, 471, 93.21182, 131.5871, //
                             74.15835, 453, 87.68438, 143.4579};

  PetscInt rows_ix[20] = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19};
  PetscInt cols_ix[4]  = {0, 1, 2, 3};

  PetscCall(PetscInitialize(&argc, &args, (char *)0, help));
  PetscCallMPI(MPI_Comm_rank(PETSC_COMM_WORLD, &rank));

  PetscCall(VecCreate(PETSC_COMM_WORLD, &y));
  PetscCall(VecSetSizes(y, PETSC_DECIDE, 20));
  PetscCall(VecSetFromOptions(y));
  PetscCall(VecDuplicate(y, &y_predicted));
  PetscCall(MatCreate(PETSC_COMM_WORLD, &X));
  PetscCall(MatSetSizes(X, PETSC_DECIDE, PETSC_DECIDE, 20, 4));
  PetscCall(MatSetFromOptions(X));
  PetscCall(MatSetUp(X));

  if (!rank) {
    PetscCall(VecSetValues(y, 20, rows_ix, y_array, INSERT_VALUES));
    PetscCall(MatSetValues(X, 20, rows_ix, 4, cols_ix, X_array, ADD_VALUES));
  }
  PetscCall(VecAssemblyBegin(y));
  PetscCall(VecAssemblyEnd(y));
  PetscCall(MatAssemblyBegin(X, MAT_FINAL_ASSEMBLY));
  PetscCall(MatAssemblyEnd(X, MAT_FINAL_ASSEMBLY));

  PetscCall(PetscRegressorCreate(PETSC_COMM_WORLD, &regressor));
  PetscCall(PetscRegressorSetType(regressor, PETSCREGRESSORLINEAR));
  PetscRegressorSetFromOptions(regressor);
  PetscCall(PetscRegressorFit(regressor, X, y));
  PetscCall(PetscRegressorPredict(regressor, X, y_predicted));
  PetscCall(PetscRegressorLinearGetIntercept(regressor, &intercept));
  PetscCall(PetscRegressorLinearGetCoefficients(regressor, &coefficients));

  PetscCall(PetscPrintf(PETSC_COMM_WORLD, "Intercept is %lf\n", intercept));
  PetscCall(PetscPrintf(PETSC_COMM_WORLD, "Coefficients are\n"));
  PetscCall(VecView(coefficients, PETSC_VIEWER_STDOUT_WORLD));
  PetscCall(PetscPrintf(PETSC_COMM_WORLD, "Predicted values are\n"));
  PetscCall(VecView(y_predicted, PETSC_VIEWER_STDOUT_WORLD));

  PetscCall(PetscRegressorDestroy(&regressor));

  PetscCall(PetscFinalize());
  return 0;
}