#include <getopt.h>
#include <iomanip>

#include "fits.hpp"

int main(int argc, char* argv[]) {
  unsigned Nend = 1;
  unsigned M = 10;
  unsigned nBatches = 5;
  Real σ = 0.2;
  Real iniVar = 0.0;
  long unsigned maxSteps = 1e12;

  int opt;

  while ((opt = getopt(argc, argv, "M:s:S:B:i:N:")) != -1) {
    switch (opt) {
    case 'N':
      Nend = (unsigned)atof(optarg);
      break;
    case 'M':
      M = (unsigned)atof(optarg);
      break;
    case 's':
      σ = atof(optarg);
      break;
    case 'S':
      maxSteps = (long unsigned)atof(optarg);
      break;
    case 'B':
      nBatches = (unsigned)atof(optarg);
      break;
    case 'i':
      iniVar = atof(optarg);
      break;
    default:
      exit(1);
    }
  }

  Rng r;

  Data data = generateData([](Real x) {return std::sin(2 * M_PI * x);}, M, σ, r);

  std::cout << std::setprecision(15);

  for (std::tuple<Real, Real> datum : data) {
    std::cout << std::get<0>(datum) << " " << std::get<1>(datum) << " ";
  }
  std::cout << std::endl;

  for (unsigned N = 1; N <= M; N++) {
    Vector a = underSolve(data, N);

    for (Real ai : a) {
      std::cout << ai << " ";
    }
    std::cout << std::endl;
  }

  for (unsigned N = Nend; N > M; N--) {
    Vector a₀ = Vector::Zero(N);
    for (Real& aa : a₀) {
      aa = r.variate<Real, std::normal_distribution>(0, iniVar);
    }

    Vector a = stochasticGradientDescent(data, a₀, nBatches, maxSteps);

    for (Real ai : a) {
      std::cout << ai << " ";
    }
    std::cout << std::endl;
  }

  return 0;
}