/* -*- C++ -*-

 *

 * This file is a part of LEMON, a generic C++ optimization library

 *

 * Copyright (C) 2003-2007

 * Egervary Jeno Kombinatorikus Optimalizalasi Kutatocsoport

 * (Egervary Research Group on Combinatorial Optimization, EGRES).

 *

 * Permission to use, modify and distribute this software is granted

 * provided that this copyright notice appears in all copies. For

 * precise terms see the accompanying LICENSE file.

 *

 * This software is provided "AS IS" with no warranty of any kind,

 * express or implied, and with no claim as to its suitability for any

 * purpose.

 *

 */

#include <iostream>

#include<lemon/lp_cplex.h>

///\file

///\brief Implementation of the LEMON-CPLEX lp solver interface.

namespace lemon {

  LpCplex::LpCplex() : LpSolverBase() {

    //    env = CPXopenCPLEXdevelop(&status);     

    env = CPXopenCPLEX(&status);     

    lp = CPXcreateprob(env, &status, "LP problem");

  }

  LpCplex::~LpCplex() {

    CPXfreeprob(env,&lp); 

    CPXcloseCPLEX(&env); 

  }

  LpSolverBase &LpCplex::_newLp() 

  {

    //The first approach opens a new environment

    LpCplex* newlp=new LpCplex();

    return *newlp;

  }

  LpSolverBase &LpCplex::_copyLp() {

    ///\bug FixID data is not copied!

    //The first approach opens a new environment

    LpCplex* newlp=new LpCplex();

    //The routine CPXcloneprob can be used to create a new CPLEX problem 

    //object and copy all the problem data from an existing problem 

    //object to it. Solution and starting information is not copied.

    newlp->lp = CPXcloneprob(env, lp, &status);

    return *newlp;

  }

  int LpCplex::_addCol()

  {

    int i = CPXgetnumcols(env, lp);

    Value lb[1],ub[1];

    lb[0]=-INF;//-CPX_INFBOUND;

    ub[0]=INF;//CPX_INFBOUND;

    status = CPXnewcols(env, lp, 1, NULL, lb, ub, NULL, NULL);

    return i;

  }

  int LpCplex::_addRow() 

  {

    //We want a row that is not constrained

    char sense[1];

    sense[0]='L';//<= constraint

    Value rhs[1];

    rhs[0]=INF;

    int i = CPXgetnumrows(env, lp);

    status = CPXnewrows(env, lp, 1, rhs, sense, NULL, NULL);

    return i;

  }

  void LpCplex::_eraseCol(int i) {

    CPXdelcols(env, lp, i, i);

  }

  void LpCplex::_eraseRow(int i) {

    CPXdelrows(env, lp, i, i);

  }

  void LpCplex::_getColName(int col, std::string &name) const

  {

    ///\bug Untested

    int storespace;

    CPXgetcolname(env, lp, 0, 0, 0, &storespace, col, col);

    storespace *= -1;

    char buf[storespace];

    char *names[1];

    int dontcare;

    ///\bug return code unchecked for error

    CPXgetcolname(env, lp, names, buf, storespace, &dontcare, col, col);

    name = names[0];

  }

  void LpCplex::_setColName(int col, const std::string &name)

  {

    ///\bug Untested

    char *names[1];

    names[0] = const_cast<char*>(name.c_str());

    ///\bug return code unchecked for error

    CPXchgcolname(env, lp, 1, &col, names);    

  }

  int LpCplex::_colByName(const std::string& name) const

  {

    int index;

    if (CPXgetcolindex(env, lp, 

                       const_cast<char*>(name.c_str()), &index) == 0) {

      return index;

    }

    return -1; 

  }

  ///\warning Data at index 0 is ignored in the arrays.

  void LpCplex::_setRowCoeffs(int i, ConstRowIterator b, ConstRowIterator e)

  {

    std::vector<int> indices;

    std::vector<int> rowlist;

    std::vector<Value> values;

    for(ConstRowIterator it=b; it!=e; ++it) {

      indices.push_back(it->first);

      values.push_back(it->second);

      rowlist.push_back(i);

    }

    status = CPXchgcoeflist(env, lp, values.size(), 

			    &rowlist[0], &indices[0], &values[0]); 

  }

  void LpSoplex::_getRowCoeffs(int i, RowIterator b) const {

    /// \todo implement

  }

  void LpCplex::_setColCoeffs(int i, ConstColIterator b, ConstColIterator e)

  {

    std::vector<int> indices;

    std::vector<int> collist;

    std::vector<Value> values;

    for(ConstColIterator it=b; it!=e; ++it) {

      indices.push_back(it->first);

      values.push_back(it->second);

      collist.push_back(i);

    }

    status = CPXchgcoeflist(env, lp, values.size(), 

			    &indices[0], &collist[0], &values[0]); 

  }

  void LpSoplex::_getColCoeffs(int i, ColIterator b) const {

    /// \todo implement

  }

  void LpCplex::_setCoeff(int row, int col, Value value) 

  {

    CPXchgcoef(env, lp, row, col, value);

  }

  LpCplex::Value LpCplex::_getCoeff(int row, int col) const

  {

    LpCplex::Value value;

    CPXgetcoef(env, lp, row, col, &value);

    return value;

  }

  void LpCplex::_setColLowerBound(int i, Value value)

  {

    int indices[1];

    indices[0]=i;

    char lu[1];

    lu[0]='L';

    Value bd[1];

    bd[0]=value;

    status = CPXchgbds(env, lp, 1, indices, lu, bd);

  }

  LpCplex::Value LpCplex::_getColLowerBound(int i) const

  {

    LpCplex::Value x;

    CPXgetlb (env, lp, &x, i, i);

    return x;

  }

  void LpCplex::_setColUpperBound(int i, Value value)

  {

    int indices[1];

    indices[0]=i;

    char lu[1];

    lu[0]='U';

    Value bd[1];

    bd[0]=value;

    status = CPXchgbds(env, lp, 1, indices, lu, bd);

  }

  LpCplex::Value LpCplex::_getColUpperBound(int i) const

  {

    LpCplex::Value x;

    CPXgetub (env, lp, &x, i, i);

    return x;

  }

  //This will be easier to implement

  void LpCplex::_setRowBounds(int i, Value lb, Value ub)

  {

    //Bad parameter

    if (lb==INF || ub==-INF) {

      //FIXME error

    }

    int cnt=1;

    int indices[1];

    indices[0]=i;

    char sense[1];

    if (lb==-INF){

      sense[0]='L';

      CPXchgsense(env, lp, cnt, indices, sense);

      CPXchgcoef(env, lp, i, -1, ub);

    }

    else{

      if (ub==INF){

	sense[0]='G';

	CPXchgsense(env, lp, cnt, indices, sense);

	CPXchgcoef(env, lp, i, -1, lb);

      }

      else{

	if (lb == ub){

	  sense[0]='E';

	  CPXchgsense(env, lp, cnt, indices, sense);

	  CPXchgcoef(env, lp, i, -1, lb);

	}

	else{

	  sense[0]='R';

	  CPXchgsense(env, lp, cnt, indices, sense);

	  CPXchgcoef(env, lp, i, -1, lb);

	  CPXchgcoef(env, lp, i, -2, ub-lb);	  

	}

      }

    }

  }

//   void LpCplex::_setRowLowerBound(int i, Value value)

//   {

//     //Not implemented, obsolete

//   }

//   void LpCplex::_setRowUpperBound(int i, Value value)

//   {

//     //Not implemented, obsolete

// //     //TODO Ezt kell meg megirni

// //     //type of the problem

// //     char sense[1];

// //     status = CPXgetsense(env, lp, sense, i, i);

// //     Value rhs[1];

// //     status = CPXgetrhs(env, lp, rhs, i, i);

// //     switch (sense[0]) {

// //     case 'L'://<= constraint

// //       break;

// //     case 'E'://= constraint

// //       break;

// //     case 'G'://>= constraint

// //       break;

// //     case 'R'://ranged constraint

// //       break;

// //     default: ;

// //       //FIXME error

// //     }

// //     status = CPXchgcoef(env, lp, i, -2, value_rng);

//   }

  void LpCplex::_getRowBounds(int i, Value &lb, Value &ub) const

  {

    char sense;

    CPXgetsense(env, lp, &sense,i,i);

    lb=-INF;

    ub=INF;

    switch (sense)

      {

      case 'L':

	CPXgetcoef(env, lp, i, -1, &ub);

	break;

      case 'G':

	CPXgetcoef(env, lp, i, -1, &lb);

	break;

      case 'E':

	CPXgetcoef(env, lp, i, -1, &lb);

	ub=lb;

	break;

      case 'R':

	CPXgetcoef(env, lp, i, -1, &lb);

	Value x;

	CPXgetcoef(env, lp, i, -2, &x);

	ub=lb+x;

	break;

      }

  }

  void LpCplex::_setObjCoeff(int i, Value obj_coef)

  {

    CPXchgcoef(env, lp, -1, i, obj_coef);

  }

  LpCplex::Value LpCplex::_getObjCoeff(int i) const

  {

    Value x;

    CPXgetcoef(env, lp, -1, i, &x);

    return x;

  }

  void LpCplex::_clearObj()

  {

    for (int i=0;i< CPXgetnumcols(env, lp);++i){

      CPXchgcoef(env, lp, -1, i, 0);

    }

  }

  // The routine returns zero unless an error occurred during the

  // optimization. Examples of errors include exhausting available

  // memory (CPXERR_NO_MEMORY) or encountering invalid data in the

  // CPLEX problem object (CPXERR_NO_PROBLEM). Exceeding a

  // user-specified CPLEX limit, or proving the model infeasible or

  // unbounded, are not considered errors. Note that a zero return

  // value does not necessarily mean that a solution exists. Use query

  // routines CPXsolninfo, CPXgetstat, and CPXsolution to obtain

  // further information about the status of the optimization.

  LpCplex::SolveExitStatus LpCplex::_solve()

  {

    //CPX_PARAM_LPMETHOD 

    status = CPXlpopt(env, lp);

    //status = CPXprimopt(env, lp);

#if CPX_VERSION >= 800

    if (status)

    {

      return UNSOLVED;

    }

    else

    {

      switch (CPXgetstat(env, lp))

      {

        case CPX_STAT_OPTIMAL:

        case CPX_STAT_INFEASIBLE:

        case CPX_STAT_UNBOUNDED:

          return SOLVED;

        default:

          return UNSOLVED;

      }

    }

#else

    if (status == 0){

      //We want to exclude some cases

      switch (CPXgetstat(env, lp)){

      case CPX_OBJ_LIM:

      case CPX_IT_LIM_FEAS:

      case CPX_IT_LIM_INFEAS:               

      case CPX_TIME_LIM_FEAS:

      case CPX_TIME_LIM_INFEAS:

	return UNSOLVED;

      default:

	return SOLVED; 

      }

    }

    else{

      return UNSOLVED;

    }

#endif

  }

  LpCplex::Value LpCplex::_getPrimal(int i) const

  {

    Value x;

    CPXgetx(env, lp, &x, i, i);

    return x;

  }

  LpCplex::Value LpCplex::_getDual(int i) const

  {

    Value y;

    CPXgetpi(env, lp, &y, i, i);

    return y;

  }

  LpCplex::Value LpCplex::_getPrimalValue() const

  {

    Value objval;

    //method = CPXgetmethod (env, lp);

    //printf("CPXgetprobtype %d \n",CPXgetprobtype(env,lp));

    status = CPXgetobjval(env, lp, &objval);

    //printf("Objective value: %g \n",objval);

    return objval;

  }

  bool LpCplex::_isBasicCol(int i) const

  {

    int cstat[CPXgetnumcols(env, lp)];

    CPXgetbase(env, lp, cstat, NULL);

    return (cstat[i]==CPX_BASIC);

  }  

//7.5-os cplex statusai (Vigyazat: a 9.0-asei masok!)

// This table lists the statuses, returned by the CPXgetstat() routine, for solutions to LP problems or mixed integer problems. If no solution exists, the return value is zero.

// For Simplex, Barrier  

// 1  	CPX_OPTIMAL  

// 	 Optimal solution found  

// 2  	CPX_INFEASIBLE  

// 	 Problem infeasible  

// 3    CPX_UNBOUNDED  

// 	 Problem unbounded  

// 4  	CPX_OBJ_LIM  

// 	 Objective limit exceeded in Phase II  

// 5  	CPX_IT_LIM_FEAS  

// 	 Iteration limit exceeded in Phase II  

// 6  	CPX_IT_LIM_INFEAS  

// 	 Iteration limit exceeded in Phase I  

// 7  	CPX_TIME_LIM_FEAS  

// 	 Time limit exceeded in Phase II  

// 8  	CPX_TIME_LIM_INFEAS  

// 	 Time limit exceeded in Phase I  

// 9  	CPX_NUM_BEST_FEAS  

// 	 Problem non-optimal, singularities in Phase II  

// 10 	CPX_NUM_BEST_INFEAS  

// 	 Problem non-optimal, singularities in Phase I  

// 11 	CPX_OPTIMAL_INFEAS  

// 	 Optimal solution found, unscaled infeasibilities  

// 12 	CPX_ABORT_FEAS  

// 	 Aborted in Phase II  

// 13 	CPX_ABORT_INFEAS  

// 	 Aborted in Phase I  

// 14  	CPX_ABORT_DUAL_INFEAS  

// 	 Aborted in barrier, dual infeasible  

// 15  	CPX_ABORT_PRIM_INFEAS  

// 	 Aborted in barrier, primal infeasible  

// 16  	CPX_ABORT_PRIM_DUAL_INFEAS  

// 	 Aborted in barrier, primal and dual infeasible  

// 17  	CPX_ABORT_PRIM_DUAL_FEAS  

// 	 Aborted in barrier, primal and dual feasible  

// 18  	CPX_ABORT_CROSSOVER  

// 	 Aborted in crossover  

// 19  	CPX_INForUNBD  

// 	 Infeasible or unbounded  

// 20   CPX_PIVOT

//       User pivot used

//

//     Ezeket hova tegyem:

// ??case CPX_ABORT_DUAL_INFEAS           

// ??case CPX_ABORT_CROSSOVER             

// ??case CPX_INForUNBD                   

// ??case CPX_PIVOT              

//Some more interesting stuff:

// CPX_PARAM_LPMETHOD  1062  int  LPMETHOD

// 0 Automatic 

// 1 Primal Simplex 

// 2 Dual Simplex 

// 3 Network Simplex 

// 4 Standard Barrier 

// Default: 0 

// Description: Method for linear optimization. 

// Determines which algorithm is used when CPXlpopt() (or "optimize" in the Interactive Optimizer) is called. Currently the behavior of the "Automatic" setting is that CPLEX simply invokes the dual simplex method, but this capability may be expanded in the future so that CPLEX chooses the method based on problem characteristics 

  void statusSwitch(CPXENVptr env,int& stat){

#if CPX_VERSION < 900

    int lpmethod;

    CPXgetintparam (env,CPX_PARAM_LPMETHOD,&lpmethod);

    if (lpmethod==2){

      if (stat==CPX_UNBOUNDED){

	stat=CPX_INFEASIBLE;

      }

      else{

	if (stat==CPX_INFEASIBLE)

	  stat=CPX_UNBOUNDED;

      }

    }

#endif

  }

  LpCplex::SolutionStatus LpCplex::_getPrimalStatus() const

  {

    //Unboundedness not treated well: the following is from cplex 9.0 doc

    // About Unboundedness

    // The treatment of models that are unbounded involves a few

    // subtleties. Specifically, a declaration of unboundedness means that

    // ILOG CPLEX has determined that the model has an unbounded

    // ray. Given any feasible solution x with objective z, a multiple of

    // the unbounded ray can be added to x to give a feasible solution

    // with objective z-1 (or z+1 for maximization models). Thus, if a

    // feasible solution exists, then the optimal objective is

    // unbounded. Note that ILOG CPLEX has not necessarily concluded that

    // a feasible solution exists. Users can call the routine CPXsolninfo

    // to determine whether ILOG CPLEX has also concluded that the model

    // has a feasible solution.

    int stat = CPXgetstat(env, lp);

#if CPX_VERSION >= 800

    switch (stat)

    {

      case CPX_STAT_OPTIMAL:

        return OPTIMAL;

      case CPX_STAT_UNBOUNDED:

        return INFINITE;

      case CPX_STAT_INFEASIBLE:

        return INFEASIBLE;

      default:

        return UNDEFINED;

    }

#else

    statusSwitch(env,stat);

    //CPXgetstat(env, lp);

    //printf("A primal status: %d, CPX_OPTIMAL=%d \n",stat,CPX_OPTIMAL);

    switch (stat) {

    case 0:

      return UNDEFINED; //Undefined

    case CPX_OPTIMAL://Optimal

      return OPTIMAL;

    case CPX_UNBOUNDED://Unbounded

      return INFEASIBLE;//In case of dual simplex

      //return INFINITE;

    case CPX_INFEASIBLE://Infeasible 

 //    case CPX_IT_LIM_INFEAS:

//     case CPX_TIME_LIM_INFEAS:

//     case CPX_NUM_BEST_INFEAS:             

//     case CPX_OPTIMAL_INFEAS:              

//     case CPX_ABORT_INFEAS:                

//     case CPX_ABORT_PRIM_INFEAS:           

//     case CPX_ABORT_PRIM_DUAL_INFEAS:      

      return INFINITE;//In case of dual simplex

      //return INFEASIBLE;

//     case CPX_OBJ_LIM:                    

//     case CPX_IT_LIM_FEAS:             

//     case CPX_TIME_LIM_FEAS:                

//     case CPX_NUM_BEST_FEAS:                

//     case CPX_ABORT_FEAS:                  

//     case CPX_ABORT_PRIM_DUAL_FEAS:        

//       return FEASIBLE;

    default:

      return UNDEFINED; //Everything else comes here

      //FIXME error

    }

#endif

  }

//9.0-as cplex verzio statusai

// CPX_STAT_ABORT_DUAL_OBJ_LIM

// CPX_STAT_ABORT_IT_LIM

// CPX_STAT_ABORT_OBJ_LIM

// CPX_STAT_ABORT_PRIM_OBJ_LIM

// CPX_STAT_ABORT_TIME_LIM

// CPX_STAT_ABORT_USER

// CPX_STAT_FEASIBLE_RELAXED

// CPX_STAT_INFEASIBLE

// CPX_STAT_INForUNBD

// CPX_STAT_NUM_BEST

// CPX_STAT_OPTIMAL

// CPX_STAT_OPTIMAL_FACE_UNBOUNDED

// CPX_STAT_OPTIMAL_INFEAS

// CPX_STAT_OPTIMAL_RELAXED

// CPX_STAT_UNBOUNDED

  LpCplex::SolutionStatus LpCplex::_getDualStatus() const

  {

    int stat = CPXgetstat(env, lp);

#if CPX_VERSION >= 800

    switch (stat)

    {

      case CPX_STAT_OPTIMAL:

        return OPTIMAL;

      case CPX_STAT_UNBOUNDED:

        return INFEASIBLE;

      default:

        return UNDEFINED;

    }

#else

    statusSwitch(env,stat);

    switch (stat) {

    case 0:

      return UNDEFINED; //Undefined

    case CPX_OPTIMAL://Optimal

      return OPTIMAL;

    case CPX_UNBOUNDED:

     return INFEASIBLE;

    default:

      return UNDEFINED; //Everything else comes here

      //FIXME error

    }

#endif

  }

  LpCplex::ProblemTypes LpCplex::_getProblemType() const

  {

    int stat = CPXgetstat(env, lp);

#if CPX_VERSION >= 800

    switch (stat)

    {

      case CPX_STAT_OPTIMAL:

	return PRIMAL_DUAL_FEASIBLE;

      case CPX_STAT_UNBOUNDED:

 	return PRIMAL_FEASIBLE_DUAL_INFEASIBLE;

      default:

        return UNKNOWN;

    }

#else

    switch (stat) {

    case CPX_OPTIMAL://Optimal

	return PRIMAL_DUAL_FEASIBLE;

    case CPX_UNBOUNDED:

 	return PRIMAL_FEASIBLE_DUAL_INFEASIBLE;

// 	return PRIMAL_INFEASIBLE_DUAL_FEASIBLE;

// 	return PRIMAL_DUAL_INFEASIBLE;

//Seems to be that this is all we can say for sure

    default:

	//In all other cases

	return UNKNOWN;

      //FIXME error

    }

#endif

  }

  void LpCplex::_setMax()

  {

    CPXchgobjsen(env, lp, CPX_MAX);

   }

  void LpCplex::_setMin()

  {

    CPXchgobjsen(env, lp, CPX_MIN);

   }

  bool LpCplex::_isMax() const

  {

    if (CPXgetobjsen(env, lp)==CPX_MAX)

      return true;

    else

      return false;

  }

} //namespace lemon