little_tools/stat/latex_table.cpp

// =====================================================================================
//
//       Filename:  latex_table.cpp
//
//    Description:  generate latex table
//
//        Version:  1.0
//        Created:  2018年 02月 19日 星期一 16:42:17 CST
//       Revision:  none
//       Compiler:  g++
//
//         Author:  Jinkun Lin, jkunlin@gmail.com
//   Organization:  School of EECS, Peking University
//
// =====================================================================================
#include <algorithm>
#include <fstream>
#include <iostream>
#include <limits>
#include <sstream>
#include <vector>
using namespace std;

bool SMALL_BETTER = false;
int FAILED_SIZE = numeric_limits<int>::min();
double PENALTY_TIME = 100;
double DELTA = 0.0001;

string basename(string name) {
  size_t begin = name.find_last_of('/');
  return name.substr(begin + 1, name.size() - begin);
}

bool equal_for_double(double d1, double d2, double delta = DELTA) {
  return fabs(d1 - d2) < delta;
}

template <class T1, class T2> bool better(T1 a, T2 b) {
  return SMALL_BETTER ? a < b : a > b;
}

int main(int argc, char const *argv[]) {
  int solver_count = argc - 1;

  // =====================================================================================
  //      read data from file
  // =====================================================================================
  vector<vector<int>> size_matrix(solver_count);
  vector<vector<double>> time_matrix(solver_count);
  vector<vector<char>> exact_matrix(solver_count);

  for (int i = 0; i < solver_count; ++i) {
    ifstream in_file(argv[i + 1]);

    in_file.seekg(0, in_file.end);
    int length = in_file.tellg();
    in_file.seekg(0, in_file.beg);
    if (!in_file || length == 0) {
      std::cout << argv[i + 1] << '\t' << i + 1 << " file error" << std::endl;
      for (int j = 0; j < 10; ++j) {
        size_matrix[i].push_back(FAILED_SIZE);
        time_matrix[i].push_back(PENALTY_TIME);
        exact_matrix[i].push_back('h');
      }
      continue;
    }

    int seed, size;
    double solver_time;
    char exact;
    while (in_file >> seed >> size >> solver_time >> exact) {
      if (size == 0) {
        size_matrix[i].push_back(FAILED_SIZE);
        time_matrix[i].push_back(PENALTY_TIME);
        exact_matrix[i].push_back('h');
      } else {
        size_matrix[i].push_back(size);
        time_matrix[i].push_back(solver_time);
        exact_matrix[i].push_back(exact);
      }
    }
    in_file.close();
  }

  // =====================================================================================
  //       caculate data
  // =====================================================================================
  vector<int> best_size(solver_count);
  vector<double> avg_size(solver_count);
  vector<double> avg_time(solver_count);
  vector<int> is_exact(solver_count, 0);

  int g_best_size = FAILED_SIZE;
  double g_avg_best_size = FAILED_SIZE;
  double g_min_avg_t = std::numeric_limits<double>::max();

  for (int i = 0; i < solver_count; ++i) {
    int okay_count = 0;

    int best_s = FAILED_SIZE;
    double avg_s = 0;
    double avg_t = 0;
    for (int j = 0; j < size_matrix[i].size(); j++) {
      if (size_matrix[i][j] != FAILED_SIZE) {
        if (best_s == FAILED_SIZE || better(size_matrix[i][j], best_s)) {
          best_s = size_matrix[i][j];
        }
        avg_s += size_matrix[i][j];
        avg_t += time_matrix[i][j];
        if (exact_matrix[i][j] == 'x') {
          is_exact[i] = 1;
        }
        okay_count++;
      } else {
        avg_t += PENALTY_TIME;
      }
    }

    if (okay_count != 0) {
      best_size[i] = best_s;
      avg_size[i] = round(avg_s / okay_count * 100) / 100;
      avg_time[i] = round(avg_t / size_matrix[i].size() * 100) / 100;

      if (better(best_size[i], g_best_size)) {
        g_best_size = best_size[i];
      }
      if (better(avg_size[i], g_avg_best_size)) {
        g_avg_best_size = avg_size[i];
      }
      if (g_min_avg_t > avg_time[i]) {
        g_min_avg_t = avg_time[i];
      }
    } else {
      best_size[i] = FAILED_SIZE;
      avg_size[i] = FAILED_SIZE;
      avg_time[i] = PENALTY_TIME;
    }
  }

  // =====================================================================================
  //       generate latex table
  // =====================================================================================
  ofstream out_latex("bold_table.tex", ios::app);
  if (!out_latex) {
    std::cout << "out file error" << std::endl;
    return 1;
  }
  auto bf_write = [&](double d, bool is_bf) {
    if (is_bf) {
      out_latex << "\\textbf{" << d << "}";
    } else {
      out_latex << d;
    }
  };
  auto bf_write_time = [&](double d, bool is_bf) {
    if (is_bf) {
      if (d < 0.01) {
        out_latex << "\\boldmath{$<$}\\textbf{0.01}";
      } else {
        out_latex << "\\textbf{" << d << "}";
      }
    } else {
      if (d < 0.01) {
        out_latex << "$<$0.01";
      } else {
        out_latex << d;
      }
    }
  };

  out_latex << basename(argv[1]) << " & ";
  bool has_double_winner = false;
  int s_win_count = 0;
  int avg_s_win_count = 0;
  int avg_t_win_count = 0;
  for (int i = 0; i < solver_count; ++i) {
    if (best_size[i] == g_best_size) {
      s_win_count++;
    }
    if (equal_for_double(avg_size[i], g_avg_best_size)) {
      avg_s_win_count++;
    }
    if (equal_for_double(avg_time[i], g_min_avg_t)) {
      avg_t_win_count++;
    }

    if (best_size[i] == g_best_size &&
        equal_for_double(avg_size[i], g_avg_best_size)) {
      has_double_winner = true;
    }
  }

  for (int i = 0; i < solver_count; ++i) {
    if (best_size[i] == FAILED_SIZE) {
      out_latex << "N/A"
                << " & "
                << "N/A";
      if (i == solver_count - 1) {
        out_latex << " \\\\" << std::endl;
      } else {
        out_latex << " & ";
      }
      continue;
    }
    // size

    if (equal_for_double(best_size[i], avg_size[i])) {
      if (has_double_winner) {
        if ((s_win_count != solver_count || avg_s_win_count != solver_count) &&
            best_size[i] == g_best_size &&
            equal_for_double(avg_size[i], g_avg_best_size)) {
          bf_write(best_size[i], true);
        } else {
          bf_write(best_size[i], false);
        }
      } else {
        bf_write(best_size[i],
                 best_size[i] == g_best_size ||
                     equal_for_double(avg_size[i], g_avg_best_size));
      }
    } else { // max size dosen't equal to avg size
      if (has_double_winner) {
        if ((s_win_count != solver_count || avg_s_win_count != solver_count) &&
            (best_size[i] == g_best_size &&
             equal_for_double(avg_size[i], g_avg_best_size))) {
          bf_write(best_size[i], true);
          out_latex << "(";
          bf_write(avg_size[i], true);
          out_latex << ")";
        } else {
          bf_write(best_size[i], false);
          out_latex << "(";
          bf_write(avg_size[i], false);
          out_latex << ")";
        }
      } else {
        if (s_win_count != solver_count) {
          bf_write(best_size[i], best_size[i] == g_best_size);
        } else {
          bf_write(best_size[i], false);
        }

        out_latex << "(";
        if (avg_s_win_count != solver_count) {
          bf_write(avg_size[i], equal_for_double(avg_size[i], g_avg_best_size));
        } else {
          bf_write(avg_size[i], false);
        }
        out_latex << ")";
      }
    }
    if (is_exact[i]) {
      out_latex << "$^*$";
    }

    out_latex << " & ";

    // time
    if (s_win_count == solver_count && avg_s_win_count == solver_count &&
        avg_t_win_count != solver_count) {
      bf_write_time(avg_time[i], equal_for_double(avg_time[i], g_min_avg_t));
    } else {
      bf_write_time(avg_time[i], false);
    }

    if (i == solver_count - 1) {
      out_latex << " \\\\" << std::endl;
    } else {
      out_latex << " & ";
    }
  }
  out_latex.close();

  // =====================================================================================
  //       generate compare file
  // =====================================================================================
  vector<bool> win(solver_count, false);
  // if (s_win_count != solver_count || avg_s_win_count != solver_count) {
  for (int i = 0; i < solver_count; ++i) {
    if (best_size[i] == g_best_size &&
        equal_for_double(avg_size[i], g_avg_best_size)) {
      // if (best_size[i] == g_best_size) {
      win[i] = true;
    }
  }
  // }
  ofstream out_arch("arch.result", ios::app);
  if (!out_arch) {
    std::cout << "arch file error" << std::endl;
    return 1;
  }

  // win
  out_arch << (win[0] ? 1 : 0);
  for (int i = 1; i < solver_count; ++i) {
    out_arch << '\t' << (win[i] ? 1 : 0);
  }
  out_arch << std::endl;

  // time
  out_arch << avg_time[0];
  for (int i = 1; i < solver_count; ++i) {
    out_arch << '\t' << avg_time[i];
  }
  out_arch << std::endl;

  // okay solver
  out_arch << (best_size[0] != FAILED_SIZE ? 1 : 0);
  for (int i = 1; i < solver_count; ++i) {
    out_arch << '\t' << (best_size[i] != FAILED_SIZE ? 1 : 0);
  }
  out_arch << endl;

  // exact
  out_arch << (is_exact[0] != 0 ? 1 : 0);
  for (int i = 1; i < solver_count; ++i) {
    out_arch << '\t' << (is_exact[i] != 0 ? 1 : 0);
  }
  out_arch << endl;

  out_arch.close();

  return 0;
}