lemon/lp_base.h
author deba
Fri, 27 Jan 2006 14:18:11 +0000
changeset 1915 f1f523d39d32
parent 1900 b16ca599472f
child 1956 a055123339d5
permissions -rw-r--r--
Add new ItemSetTraits for ANode and BNode
     1 /* -*- C++ -*-
     2  * lemon/lp_base.h - Part of LEMON, a generic C++ optimization library
     3  *
     4  * Copyright (C) 2006 Egervary Jeno Kombinatorikus Optimalizalasi Kutatocsoport
     5  * (Egervary Research Group on Combinatorial Optimization, EGRES).
     6  *
     7  * Permission to use, modify and distribute this software is granted
     8  * provided that this copyright notice appears in all copies. For
     9  * precise terms see the accompanying LICENSE file.
    10  *
    11  * This software is provided "AS IS" with no warranty of any kind,
    12  * express or implied, and with no claim as to its suitability for any
    13  * purpose.
    14  *
    15  */
    16 
    17 #ifndef LEMON_LP_BASE_H
    18 #define LEMON_LP_BASE_H
    19 
    20 #include<vector>
    21 #include<map>
    22 #include<limits>
    23 #include<cmath>
    24 
    25 #include<lemon/utility.h>
    26 #include<lemon/error.h>
    27 #include<lemon/invalid.h>
    28 
    29 ///\file
    30 ///\brief The interface of the LP solver interface.
    31 ///\ingroup gen_opt_group
    32 namespace lemon {
    33   
    34   ///Internal data structure to convert floating id's to fix one's
    35     
    36   ///\todo This might be implemented to be also usable in other places.
    37   class _FixId 
    38   {
    39   protected:
    40     std::vector<int> index;
    41     std::vector<int> cross;
    42     int first_free;
    43   public:
    44     _FixId() : first_free(-1) {};
    45     ///Convert a floating id to a fix one
    46 
    47     ///\param n is a floating id
    48     ///\return the corresponding fix id
    49     int fixId(int n) const {return cross[n];}
    50     ///Convert a fix id to a floating one
    51 
    52     ///\param n is a fix id
    53     ///\return the corresponding floating id
    54     int floatingId(int n) const { return index[n];}
    55     ///Add a new floating id.
    56 
    57     ///\param n is a floating id
    58     ///\return the fix id of the new value
    59     ///\todo Multiple additions should also be handled.
    60     int insert(int n)
    61     {
    62       if(n>=int(cross.size())) {
    63 	cross.resize(n+1);
    64 	if(first_free==-1) {
    65 	  cross[n]=index.size();
    66 	  index.push_back(n);
    67 	}
    68 	else {
    69 	  cross[n]=first_free;
    70 	  int next=index[first_free];
    71 	  index[first_free]=n;
    72 	  first_free=next;
    73 	}
    74 	return cross[n];
    75       }
    76       ///\todo Create an own exception type.
    77       else throw LogicError(); //floatingId-s must form a continuous range;
    78     }
    79     ///Remove a fix id.
    80 
    81     ///\param n is a fix id
    82     ///
    83     void erase(int n) 
    84     {
    85       int fl=index[n];
    86       index[n]=first_free;
    87       first_free=n;
    88       for(int i=fl+1;i<int(cross.size());++i) {
    89 	cross[i-1]=cross[i];
    90 	index[cross[i]]--;
    91       }
    92       cross.pop_back();
    93     }
    94     ///An upper bound on the largest fix id.
    95 
    96     ///\todo Do we need this?
    97     ///
    98     std::size_t maxFixId() { return cross.size()-1; }
    99   
   100   };
   101     
   102   ///Common base class for LP solvers
   103   
   104   ///\todo Much more docs
   105   ///\ingroup gen_opt_group
   106   class LpSolverBase {
   107 
   108   public:
   109 
   110     ///Possible outcomes of an LP solving procedure
   111     enum SolveExitStatus {
   112       ///This means that the problem has been successfully solved: either
   113       ///an optimal solution has been found or infeasibility/unboundedness
   114       ///has been proved.
   115       SOLVED = 0,
   116       ///Any other case (including the case when some user specified limit has been exceeded)
   117       UNSOLVED = 1
   118     };
   119       
   120       ///\e
   121     enum SolutionStatus {
   122       ///Feasible solution has'n been found (but may exist).
   123 
   124       ///\todo NOTFOUND might be a better name.
   125       ///
   126       UNDEFINED = 0,
   127       ///The problem has no feasible solution
   128       INFEASIBLE = 1,
   129       ///Feasible solution found
   130       FEASIBLE = 2,
   131       ///Optimal solution exists and found
   132       OPTIMAL = 3,
   133       ///The cost function is unbounded
   134 
   135       ///\todo Give a feasible solution and an infinite ray (and the
   136       ///corresponding bases)
   137       INFINITE = 4
   138     };
   139 
   140     ///\e The type of the investigated LP problem
   141     enum ProblemTypes {
   142       ///Primal-dual feasible
   143       PRIMAL_DUAL_FEASIBLE = 0,
   144       ///Primal feasible dual infeasible
   145       PRIMAL_FEASIBLE_DUAL_INFEASIBLE = 1,
   146       ///Primal infeasible dual feasible
   147       PRIMAL_INFEASIBLE_DUAL_FEASIBLE = 2,
   148       ///Primal-dual infeasible
   149       PRIMAL_DUAL_INFEASIBLE = 3,
   150       ///Could not determine so far
   151       UNKNOWN = 4
   152     };
   153 
   154     ///The floating point type used by the solver
   155     typedef double Value;
   156     ///The infinity constant
   157     static const Value INF;
   158     ///The not a number constant
   159     static const Value NaN;
   160     
   161     ///Refer to a column of the LP.
   162 
   163     ///This type is used to refer to a column of the LP.
   164     ///
   165     ///Its value remains valid and correct even after the addition or erase of
   166     ///other columns.
   167     ///
   168     ///\todo Document what can one do with a Col (INVALID, comparing,
   169     ///it is similar to Node/Edge)
   170     class Col {
   171     protected:
   172       int id;
   173       friend class LpSolverBase;
   174     public:
   175       typedef Value ExprValue;
   176       typedef True LpSolverCol;
   177       Col() {}
   178       Col(const Invalid&) : id(-1) {}
   179       bool operator< (Col c) const  {return id< c.id;}
   180       bool operator> (Col c) const  {return id> c.id;}
   181       bool operator==(Col c) const  {return id==c.id;}
   182       bool operator!=(Col c) const  {return id!=c.id;}
   183     };
   184 
   185     ///Refer to a row of the LP.
   186 
   187     ///This type is used to refer to a row of the LP.
   188     ///
   189     ///Its value remains valid and correct even after the addition or erase of
   190     ///other rows.
   191     ///
   192     ///\todo Document what can one do with a Row (INVALID, comparing,
   193     ///it is similar to Node/Edge)
   194     class Row {
   195     protected:
   196       int id;
   197       friend class LpSolverBase;
   198     public:
   199       typedef Value ExprValue;
   200       typedef True LpSolverRow;
   201       Row() {}
   202       Row(const Invalid&) : id(-1) {}
   203 
   204       bool operator< (Row c) const  {return id< c.id;}
   205       bool operator> (Row c) const  {return id> c.id;}
   206       bool operator==(Row c) const  {return id==c.id;}
   207       bool operator!=(Row c) const  {return id!=c.id;} 
   208    };
   209     
   210     ///Linear expression of variables and a constant component
   211     
   212     ///This data structure strores a linear expression of the variables
   213     ///(\ref Col "Col"s) and also has a constant component.
   214     ///
   215     ///There are several ways to access and modify the contents of this
   216     ///container.
   217     ///- Its it fully compatible with \c std::map<Col,double>, so for expamle
   218     ///if \c e is an Expr and \c v and \c w are of type \ref Col, then you can
   219     ///read and modify the coefficients like
   220     ///these.
   221     ///\code
   222     ///e[v]=5;
   223     ///e[v]+=12;
   224     ///e.erase(v);
   225     ///\endcode
   226     ///or you can also iterate through its elements.
   227     ///\code
   228     ///double s=0;
   229     ///for(LpSolverBase::Expr::iterator i=e.begin();i!=e.end();++i)
   230     ///  s+=i->second;
   231     ///\endcode
   232     ///(This code computes the sum of all coefficients).
   233     ///- Numbers (<tt>double</tt>'s)
   234     ///and variables (\ref Col "Col"s) directly convert to an
   235     ///\ref Expr and the usual linear operations are defined, so  
   236     ///\code
   237     ///v+w
   238     ///2*v-3.12*(v-w/2)+2
   239     ///v*2.1+(3*v+(v*12+w+6)*3)/2
   240     ///\endcode
   241     ///are valid \ref Expr "Expr"essions.
   242     ///The usual assignment operations are also defined.
   243     ///\code
   244     ///e=v+w;
   245     ///e+=2*v-3.12*(v-w/2)+2;
   246     ///e*=3.4;
   247     ///e/=5;
   248     ///\endcode
   249     ///- The constant member can be set and read by \ref constComp()
   250     ///\code
   251     ///e.constComp()=12;
   252     ///double c=e.constComp();
   253     ///\endcode
   254     ///
   255     ///\note \ref clear() not only sets all coefficients to 0 but also
   256     ///clears the constant components.
   257     ///
   258     ///\sa Constr
   259     ///
   260     class Expr : public std::map<Col,Value>
   261     {
   262     public:
   263       typedef LpSolverBase::Col Key; 
   264       typedef LpSolverBase::Value Value;
   265       
   266     protected:
   267       typedef std::map<Col,Value> Base;
   268       
   269       Value const_comp;
   270   public:
   271       typedef True IsLinExpression;
   272       ///\e
   273       Expr() : Base(), const_comp(0) { }
   274       ///\e
   275       Expr(const Key &v) : const_comp(0) {
   276 	Base::insert(std::make_pair(v, 1));
   277       }
   278       ///\e
   279       Expr(const Value &v) : const_comp(v) {}
   280       ///\e
   281       void set(const Key &v,const Value &c) {
   282 	Base::insert(std::make_pair(v, c));
   283       }
   284       ///\e
   285       Value &constComp() { return const_comp; }
   286       ///\e
   287       const Value &constComp() const { return const_comp; }
   288       
   289       ///Removes the components with zero coefficient.
   290       void simplify() {
   291 	for (Base::iterator i=Base::begin(); i!=Base::end();) {
   292 	  Base::iterator j=i;
   293 	  ++j;
   294 	  if ((*i).second==0) Base::erase(i);
   295 	  j=i;
   296 	}
   297       }
   298 
   299       ///Removes the coefficients closer to zero than \c tolerance.
   300       void simplify(double &tolerance) {
   301 	for (Base::iterator i=Base::begin(); i!=Base::end();) {
   302 	  Base::iterator j=i;
   303 	  ++j;
   304 	  if (std::fabs((*i).second)<tolerance) Base::erase(i);
   305 	  j=i;
   306 	}
   307       }
   308 
   309       ///Sets all coefficients and the constant component to 0.
   310       void clear() {
   311 	Base::clear();
   312 	const_comp=0;
   313       }
   314 
   315       ///\e
   316       Expr &operator+=(const Expr &e) {
   317 	for (Base::const_iterator j=e.begin(); j!=e.end(); ++j)
   318 	  (*this)[j->first]+=j->second;
   319 	const_comp+=e.const_comp;
   320 	return *this;
   321       }
   322       ///\e
   323       Expr &operator-=(const Expr &e) {
   324 	for (Base::const_iterator j=e.begin(); j!=e.end(); ++j)
   325 	  (*this)[j->first]-=j->second;
   326 	const_comp-=e.const_comp;
   327 	return *this;
   328       }
   329       ///\e
   330       Expr &operator*=(const Value &c) {
   331 	for (Base::iterator j=Base::begin(); j!=Base::end(); ++j)
   332 	  j->second*=c;
   333 	const_comp*=c;
   334 	return *this;
   335       }
   336       ///\e
   337       Expr &operator/=(const Value &c) {
   338 	for (Base::iterator j=Base::begin(); j!=Base::end(); ++j)
   339 	  j->second/=c;
   340 	const_comp/=c;
   341 	return *this;
   342       }
   343     };
   344     
   345     ///Linear constraint
   346 
   347     ///This data stucture represents a linear constraint in the LP.
   348     ///Basically it is a linear expression with a lower or an upper bound
   349     ///(or both). These parts of the constraint can be obtained by the member
   350     ///functions \ref expr(), \ref lowerBound() and \ref upperBound(),
   351     ///respectively.
   352     ///There are two ways to construct a constraint.
   353     ///- You can set the linear expression and the bounds directly
   354     ///  by the functions above.
   355     ///- The operators <tt>\<=</tt>, <tt>==</tt> and  <tt>\>=</tt>
   356     ///  are defined between expressions, or even between constraints whenever
   357     ///  it makes sense. Therefore if \c e and \c f are linear expressions and
   358     ///  \c s and \c t are numbers, then the followings are valid expressions
   359     ///  and thus they can be used directly e.g. in \ref addRow() whenever
   360     ///  it makes sense.
   361     ///\code
   362     ///  e<=s
   363     ///  e<=f
   364     ///  e==f
   365     ///  s<=e<=t
   366     ///  e>=t
   367     ///\endcode
   368     ///\warning The validity of a constraint is checked only at run time, so
   369     ///e.g. \ref addRow(<tt>x[1]\<=x[2]<=5</tt>) will compile, but will throw a
   370     ///\ref LogicError exception.
   371     class Constr
   372     {
   373     public:
   374       typedef LpSolverBase::Expr Expr;
   375       typedef Expr::Key Key;
   376       typedef Expr::Value Value;
   377       
   378 //       static const Value INF;
   379 //       static const Value NaN;
   380 
   381     protected:
   382       Expr _expr;
   383       Value _lb,_ub;
   384     public:
   385       ///\e
   386       Constr() : _expr(), _lb(NaN), _ub(NaN) {}
   387       ///\e
   388       Constr(Value lb,const Expr &e,Value ub) :
   389 	_expr(e), _lb(lb), _ub(ub) {}
   390       ///\e
   391       Constr(const Expr &e,Value ub) : 
   392 	_expr(e), _lb(NaN), _ub(ub) {}
   393       ///\e
   394       Constr(Value lb,const Expr &e) :
   395 	_expr(e), _lb(lb), _ub(NaN) {}
   396       ///\e
   397       Constr(const Expr &e) : 
   398 	_expr(e), _lb(NaN), _ub(NaN) {}
   399       ///\e
   400       void clear() 
   401       {
   402 	_expr.clear();
   403 	_lb=_ub=NaN;
   404       }
   405 
   406       ///Reference to the linear expression 
   407       Expr &expr() { return _expr; }
   408       ///Cont reference to the linear expression 
   409       const Expr &expr() const { return _expr; }
   410       ///Reference to the lower bound.
   411 
   412       ///\return
   413       ///- \ref INF "INF": the constraint is lower unbounded.
   414       ///- \ref NaN "NaN": lower bound has not been set.
   415       ///- finite number: the lower bound
   416       Value &lowerBound() { return _lb; }
   417       ///The const version of \ref lowerBound()
   418       const Value &lowerBound() const { return _lb; }
   419       ///Reference to the upper bound.
   420 
   421       ///\return
   422       ///- \ref INF "INF": the constraint is upper unbounded.
   423       ///- \ref NaN "NaN": upper bound has not been set.
   424       ///- finite number: the upper bound
   425       Value &upperBound() { return _ub; }
   426       ///The const version of \ref upperBound()
   427       const Value &upperBound() const { return _ub; }
   428       ///Is the constraint lower bounded?
   429       bool lowerBounded() const { 
   430 	using namespace std;
   431 	return finite(_lb);
   432       }
   433       ///Is the constraint upper bounded?
   434       bool upperBounded() const {
   435 	using namespace std;
   436 	return finite(_ub);
   437       }
   438     };
   439     
   440     ///Linear expression of rows
   441     
   442     ///This data structure represents a column of the matrix,
   443     ///thas is it strores a linear expression of the dual variables
   444     ///(\ref Row "Row"s).
   445     ///
   446     ///There are several ways to access and modify the contents of this
   447     ///container.
   448     ///- Its it fully compatible with \c std::map<Row,double>, so for expamle
   449     ///if \c e is an DualExpr and \c v
   450     ///and \c w are of type \ref Row, then you can
   451     ///read and modify the coefficients like
   452     ///these.
   453     ///\code
   454     ///e[v]=5;
   455     ///e[v]+=12;
   456     ///e.erase(v);
   457     ///\endcode
   458     ///or you can also iterate through its elements.
   459     ///\code
   460     ///double s=0;
   461     ///for(LpSolverBase::DualExpr::iterator i=e.begin();i!=e.end();++i)
   462     ///  s+=i->second;
   463     ///\endcode
   464     ///(This code computes the sum of all coefficients).
   465     ///- Numbers (<tt>double</tt>'s)
   466     ///and variables (\ref Row "Row"s) directly convert to an
   467     ///\ref DualExpr and the usual linear operations are defined, so
   468     ///\code
   469     ///v+w
   470     ///2*v-3.12*(v-w/2)
   471     ///v*2.1+(3*v+(v*12+w)*3)/2
   472     ///\endcode
   473     ///are valid \ref DualExpr "DualExpr"essions.
   474     ///The usual assignment operations are also defined.
   475     ///\code
   476     ///e=v+w;
   477     ///e+=2*v-3.12*(v-w/2);
   478     ///e*=3.4;
   479     ///e/=5;
   480     ///\endcode
   481     ///
   482     ///\sa Expr
   483     ///
   484     class DualExpr : public std::map<Row,Value>
   485     {
   486     public:
   487       typedef LpSolverBase::Row Key; 
   488       typedef LpSolverBase::Value Value;
   489       
   490     protected:
   491       typedef std::map<Row,Value> Base;
   492       
   493     public:
   494       typedef True IsLinExpression;
   495       ///\e
   496       DualExpr() : Base() { }
   497       ///\e
   498       DualExpr(const Key &v) {
   499 	Base::insert(std::make_pair(v, 1));
   500       }
   501       ///\e
   502       void set(const Key &v,const Value &c) {
   503 	Base::insert(std::make_pair(v, c));
   504       }
   505       
   506       ///Removes the components with zero coefficient.
   507       void simplify() {
   508 	for (Base::iterator i=Base::begin(); i!=Base::end();) {
   509 	  Base::iterator j=i;
   510 	  ++j;
   511 	  if ((*i).second==0) Base::erase(i);
   512 	  j=i;
   513 	}
   514       }
   515 
   516       ///Removes the coefficients closer to zero than \c tolerance.
   517       void simplify(double &tolerance) {
   518 	for (Base::iterator i=Base::begin(); i!=Base::end();) {
   519 	  Base::iterator j=i;
   520 	  ++j;
   521 	  if (std::fabs((*i).second)<tolerance) Base::erase(i);
   522 	  j=i;
   523 	}
   524       }
   525 
   526 
   527       ///Sets all coefficients to 0.
   528       void clear() {
   529 	Base::clear();
   530       }
   531 
   532       ///\e
   533       DualExpr &operator+=(const DualExpr &e) {
   534 	for (Base::const_iterator j=e.begin(); j!=e.end(); ++j)
   535 	  (*this)[j->first]+=j->second;
   536 	return *this;
   537       }
   538       ///\e
   539       DualExpr &operator-=(const DualExpr &e) {
   540 	for (Base::const_iterator j=e.begin(); j!=e.end(); ++j)
   541 	  (*this)[j->first]-=j->second;
   542 	return *this;
   543       }
   544       ///\e
   545       DualExpr &operator*=(const Value &c) {
   546 	for (Base::iterator j=Base::begin(); j!=Base::end(); ++j)
   547 	  j->second*=c;
   548 	return *this;
   549       }
   550       ///\e
   551       DualExpr &operator/=(const Value &c) {
   552 	for (Base::iterator j=Base::begin(); j!=Base::end(); ++j)
   553 	  j->second/=c;
   554 	return *this;
   555       }
   556     };
   557     
   558 
   559   protected:
   560     _FixId rows;
   561     _FixId cols;
   562 
   563     //Abstract virtual functions
   564     virtual LpSolverBase &_newLp() = 0;
   565     virtual LpSolverBase &_copyLp(){
   566       ///\todo This should be implemented here, too,  when we have problem retrieving routines. It can be overriden.
   567 
   568       //Starting:
   569       LpSolverBase & newlp(_newLp());
   570       return newlp;
   571       //return *(LpSolverBase*)0;
   572     };
   573 
   574     virtual int _addCol() = 0;
   575     virtual int _addRow() = 0;
   576     virtual void _eraseCol(int col) = 0;
   577     virtual void _eraseRow(int row) = 0;
   578     virtual void _getColName(int col,       std::string & name) = 0;
   579     virtual void _setColName(int col, const std::string & name) = 0;
   580     virtual void _setRowCoeffs(int i, 
   581 			       int length,
   582                                int  const * indices, 
   583                                Value  const * values ) = 0;
   584     virtual void _setColCoeffs(int i, 
   585 			       int length,
   586                                int  const * indices, 
   587                                Value  const * values ) = 0;
   588     virtual void _setCoeff(int row, int col, Value value) = 0;
   589     virtual void _setColLowerBound(int i, Value value) = 0;
   590     virtual void _setColUpperBound(int i, Value value) = 0;
   591 //     virtual void _setRowLowerBound(int i, Value value) = 0;
   592 //     virtual void _setRowUpperBound(int i, Value value) = 0;
   593     virtual void _setRowBounds(int i, Value lower, Value upper) = 0;
   594     virtual void _setObjCoeff(int i, Value obj_coef) = 0;
   595     virtual void _clearObj()=0;
   596 //     virtual void _setObj(int length,
   597 //                          int  const * indices, 
   598 //                          Value  const * values ) = 0;
   599     virtual SolveExitStatus _solve() = 0;
   600     virtual Value _getPrimal(int i) = 0;
   601     virtual Value _getDual(int i) = 0;
   602     virtual Value _getPrimalValue() = 0;
   603     virtual bool _isBasicCol(int i) = 0;
   604     virtual SolutionStatus _getPrimalStatus() = 0;
   605     virtual SolutionStatus _getDualStatus() = 0;
   606     ///\todo This could be implemented here, too, using _getPrimalStatus() and
   607     ///_getDualStatus()
   608     virtual ProblemTypes _getProblemType() = 0;
   609 
   610     virtual void _setMax() = 0;
   611     virtual void _setMin() = 0;
   612     
   613     //Own protected stuff
   614     
   615     //Constant component of the objective function
   616     Value obj_const_comp;
   617     
   618 
   619 
   620     
   621   public:
   622 
   623     ///\e
   624     LpSolverBase() : obj_const_comp(0) {}
   625 
   626     ///\e
   627     virtual ~LpSolverBase() {}
   628 
   629     ///Creates a new LP problem
   630     LpSolverBase &newLp() {return _newLp();}
   631     ///Makes a copy of the LP problem
   632     LpSolverBase &copyLp() {return _copyLp();}
   633     
   634     ///\name Build up and modify the LP
   635 
   636     ///@{
   637 
   638     ///Add a new empty column (i.e a new variable) to the LP
   639     Col addCol() { Col c; c.id=cols.insert(_addCol()); return c;}
   640 
   641     ///\brief Adds several new columns
   642     ///(i.e a variables) at once
   643     ///
   644     ///This magic function takes a container as its argument
   645     ///and fills its elements
   646     ///with new columns (i.e. variables)
   647     ///\param t can be
   648     ///- a standard STL compatible iterable container with
   649     ///\ref Col as its \c values_type
   650     ///like
   651     ///\code
   652     ///std::vector<LpSolverBase::Col>
   653     ///std::list<LpSolverBase::Col>
   654     ///\endcode
   655     ///- a standard STL compatible iterable container with
   656     ///\ref Col as its \c mapped_type
   657     ///like
   658     ///\code
   659     ///std::map<AnyType,LpSolverBase::Col>
   660     ///\endcode
   661     ///- an iterable lemon \ref concept::WriteMap "write map" like 
   662     ///\code
   663     ///ListGraph::NodeMap<LpSolverBase::Col>
   664     ///ListGraph::EdgeMap<LpSolverBase::Col>
   665     ///\endcode
   666     ///\return The number of the created column.
   667 #ifdef DOXYGEN
   668     template<class T>
   669     int addColSet(T &t) { return 0;} 
   670 #else
   671     template<class T>
   672     typename enable_if<typename T::value_type::LpSolverCol,int>::type
   673     addColSet(T &t,dummy<0> = 0) {
   674       int s=0;
   675       for(typename T::iterator i=t.begin();i!=t.end();++i) {*i=addCol();s++;}
   676       return s;
   677     }
   678     template<class T>
   679     typename enable_if<typename T::value_type::second_type::LpSolverCol,
   680 		       int>::type
   681     addColSet(T &t,dummy<1> = 1) { 
   682       int s=0;
   683       for(typename T::iterator i=t.begin();i!=t.end();++i) {
   684 	i->second=addCol();
   685 	s++;
   686       }
   687       return s;
   688     }
   689     template<class T>
   690     typename enable_if<typename T::MapIt::Value::LpSolverCol,
   691 		       int>::type
   692     addColSet(T &t,dummy<2> = 2) { 
   693       int s=0;
   694       for(typename T::MapIt i(t); i!=INVALID; ++i)
   695 	{
   696 	  i.set(addCol());
   697 	  s++;
   698 	}
   699       return s;
   700     }
   701 #endif
   702 
   703     ///Set a column (i.e a dual constraint) of the LP
   704 
   705     ///\param c is the column to be modified
   706     ///\param e is a dual linear expression (see \ref DualExpr)
   707     ///a better one.
   708     void col(Col c,const DualExpr &e) {
   709       std::vector<int> indices;
   710       std::vector<Value> values;
   711       indices.push_back(0);
   712       values.push_back(0);
   713       for(DualExpr::const_iterator i=e.begin(); i!=e.end(); ++i)
   714 	if((*i).second!=0) {
   715 	  indices.push_back(rows.floatingId((*i).first.id));
   716 	  values.push_back((*i).second);
   717 	}
   718       _setColCoeffs(cols.floatingId(c.id),indices.size()-1,
   719 		    &indices[0],&values[0]);
   720     }
   721 
   722     ///Add a new column to the LP
   723 
   724     ///\param e is a dual linear expression (see \ref DualExpr)
   725     ///\param obj is the corresponding component of the objective
   726     ///function. It is 0 by default.
   727     ///\return The created column.
   728     Col addCol(const DualExpr &e, Value obj=0) {
   729       Col c=addCol();
   730       col(c,e);
   731       objCoeff(c,obj);
   732       return c;
   733     }
   734 
   735     ///Add a new empty row (i.e a new constraint) to the LP
   736 
   737     ///This function adds a new empty row (i.e a new constraint) to the LP.
   738     ///\return The created row
   739     Row addRow() { Row r; r.id=rows.insert(_addRow()); return r;}
   740 
   741     ///\brief Add several new rows
   742     ///(i.e a constraints) at once
   743     ///
   744     ///This magic function takes a container as its argument
   745     ///and fills its elements
   746     ///with new row (i.e. variables)
   747     ///\param t can be
   748     ///- a standard STL compatible iterable container with
   749     ///\ref Row as its \c values_type
   750     ///like
   751     ///\code
   752     ///std::vector<LpSolverBase::Row>
   753     ///std::list<LpSolverBase::Row>
   754     ///\endcode
   755     ///- a standard STL compatible iterable container with
   756     ///\ref Row as its \c mapped_type
   757     ///like
   758     ///\code
   759     ///std::map<AnyType,LpSolverBase::Row>
   760     ///\endcode
   761     ///- an iterable lemon \ref concept::WriteMap "write map" like 
   762     ///\code
   763     ///ListGraph::NodeMap<LpSolverBase::Row>
   764     ///ListGraph::EdgeMap<LpSolverBase::Row>
   765     ///\endcode
   766     ///\return The number of rows created.
   767 #ifdef DOXYGEN
   768     template<class T>
   769     int addRowSet(T &t) { return 0;} 
   770 #else
   771     template<class T>
   772     typename enable_if<typename T::value_type::LpSolverRow,int>::type
   773     addRowSet(T &t,dummy<0> = 0) {
   774       int s=0;
   775       for(typename T::iterator i=t.begin();i!=t.end();++i) {*i=addRow();s++;}
   776       return s;
   777     }
   778     template<class T>
   779     typename enable_if<typename T::value_type::second_type::LpSolverRow,
   780 		       int>::type
   781     addRowSet(T &t,dummy<1> = 1) { 
   782       int s=0;
   783       for(typename T::iterator i=t.begin();i!=t.end();++i) {
   784 	i->second=addRow();
   785 	s++;
   786       }
   787       return s;
   788     }
   789     template<class T>
   790     typename enable_if<typename T::MapIt::Value::LpSolverRow,
   791 		       int>::type
   792     addRowSet(T &t,dummy<2> = 2) { 
   793       int s=0;
   794       for(typename T::MapIt i(t); i!=INVALID; ++i)
   795 	{
   796 	  i.set(addRow());
   797 	  s++;
   798 	}
   799       return s;
   800     }
   801 #endif
   802 
   803     ///Set a row (i.e a constraint) of the LP
   804 
   805     ///\param r is the row to be modified
   806     ///\param l is lower bound (-\ref INF means no bound)
   807     ///\param e is a linear expression (see \ref Expr)
   808     ///\param u is the upper bound (\ref INF means no bound)
   809     ///\bug This is a temportary function. The interface will change to
   810     ///a better one.
   811     ///\todo Option to control whether a constraint with a single variable is
   812     ///added or not.
   813     void row(Row r, Value l,const Expr &e, Value u) {
   814       std::vector<int> indices;
   815       std::vector<Value> values;
   816       indices.push_back(0);
   817       values.push_back(0);
   818       for(Expr::const_iterator i=e.begin(); i!=e.end(); ++i)
   819 	if((*i).second!=0) { ///\bug EPSILON would be necessary here!!!
   820 	  indices.push_back(cols.floatingId((*i).first.id));
   821 	  values.push_back((*i).second);
   822 	}
   823       _setRowCoeffs(rows.floatingId(r.id),indices.size()-1,
   824 		    &indices[0],&values[0]);
   825 //       _setRowLowerBound(rows.floatingId(r.id),l-e.constComp());
   826 //       _setRowUpperBound(rows.floatingId(r.id),u-e.constComp());
   827        _setRowBounds(rows.floatingId(r.id),l-e.constComp(),u-e.constComp());
   828     }
   829 
   830     ///Set a row (i.e a constraint) of the LP
   831 
   832     ///\param r is the row to be modified
   833     ///\param c is a linear expression (see \ref Constr)
   834     void row(Row r, const Constr &c) {
   835       row(r,
   836 	     c.lowerBounded()?c.lowerBound():-INF,
   837 	     c.expr(),
   838 	     c.upperBounded()?c.upperBound():INF);
   839     }
   840 
   841     ///Add a new row (i.e a new constraint) to the LP
   842 
   843     ///\param l is the lower bound (-\ref INF means no bound)
   844     ///\param e is a linear expression (see \ref Expr)
   845     ///\param u is the upper bound (\ref INF means no bound)
   846     ///\return The created row.
   847     ///\bug This is a temportary function. The interface will change to
   848     ///a better one.
   849     Row addRow(Value l,const Expr &e, Value u) {
   850       Row r=addRow();
   851       row(r,l,e,u);
   852       return r;
   853     }
   854 
   855     ///Add a new row (i.e a new constraint) to the LP
   856 
   857     ///\param c is a linear expression (see \ref Constr)
   858     ///\return The created row.
   859     Row addRow(const Constr &c) {
   860       Row r=addRow();
   861       row(r,c);
   862       return r;
   863     }
   864     ///Erase a coloumn (i.e a variable) from the LP
   865 
   866     ///\param c is the coloumn to be deleted
   867     ///\todo Please check this
   868     void eraseCol(Col c) {
   869       _eraseCol(cols.floatingId(c.id));
   870       cols.erase(c.id);
   871     }
   872     ///Erase a  row (i.e a constraint) from the LP
   873 
   874     ///\param r is the row to be deleted
   875     ///\todo Please check this
   876     void eraseRow(Row r) {
   877       _eraseRow(rows.floatingId(r.id));
   878       rows.erase(r.id);
   879     }
   880 
   881     /// Get the name of a column
   882     
   883     ///\param c is the coresponding coloumn 
   884     ///\return The name of the colunm
   885     std::string ColName(Col c){
   886       std::string name;
   887       _getColName(cols.floatingId(c.id), name);
   888       return name;
   889     }
   890     
   891     /// Set the name of a column
   892     
   893     ///\param c is the coresponding coloumn 
   894     ///\param name The name to be given
   895     void ColName(Col c, const std::string & name){
   896       _setColName(cols.floatingId(c.id), name);
   897     }
   898     
   899     /// Set an element of the coefficient matrix of the LP
   900 
   901     ///\param r is the row of the element to be modified
   902     ///\param c is the coloumn of the element to be modified
   903     ///\param val is the new value of the coefficient
   904 
   905     void Coeff(Row r, Col c, Value val){
   906       _setCoeff(rows.floatingId(r.id),cols.floatingId(c.id), val);
   907     }
   908 
   909     /// Set the lower bound of a column (i.e a variable)
   910 
   911     /// The lower bound of a variable (column) has to be given by an 
   912     /// extended number of type Value, i.e. a finite number of type 
   913     /// Value or -\ref INF.
   914     void colLowerBound(Col c, Value value) {
   915       _setColLowerBound(cols.floatingId(c.id),value);
   916     }
   917     
   918     ///\brief Set the lower bound of  several columns
   919     ///(i.e a variables) at once
   920     ///
   921     ///This magic function takes a container as its argument
   922     ///and applies the function on all of its elements.
   923     /// The lower bound of a variable (column) has to be given by an 
   924     /// extended number of type Value, i.e. a finite number of type 
   925     /// Value or -\ref INF.
   926 #ifdef DOXYGEN
   927     template<class T>
   928     void colLowerBound(T &t, Value value) { return 0;} 
   929 #else
   930     template<class T>
   931     typename enable_if<typename T::value_type::LpSolverCol,void>::type
   932     colLowerBound(T &t, Value value,dummy<0> = 0) {
   933       for(typename T::iterator i=t.begin();i!=t.end();++i) {
   934 	colLowerBound(*i, value);
   935       }
   936     }
   937     template<class T>
   938     typename enable_if<typename T::value_type::second_type::LpSolverCol,
   939 		       void>::type
   940     colLowerBound(T &t, Value value,dummy<1> = 1) { 
   941       for(typename T::iterator i=t.begin();i!=t.end();++i) {
   942 	colLowerBound(i->second, value);
   943       }
   944     }
   945     template<class T>
   946     typename enable_if<typename T::MapIt::Value::LpSolverCol,
   947 		       void>::type
   948     colLowerBound(T &t, Value value,dummy<2> = 2) { 
   949       for(typename T::MapIt i(t); i!=INVALID; ++i){
   950 	colLowerBound(*i, value);
   951       }
   952     }
   953 #endif
   954     
   955     /// Set the upper bound of a column (i.e a variable)
   956 
   957     /// The upper bound of a variable (column) has to be given by an 
   958     /// extended number of type Value, i.e. a finite number of type 
   959     /// Value or \ref INF.
   960     void colUpperBound(Col c, Value value) {
   961       _setColUpperBound(cols.floatingId(c.id),value);
   962     };
   963 
   964     ///\brief Set the lower bound of  several columns
   965     ///(i.e a variables) at once
   966     ///
   967     ///This magic function takes a container as its argument
   968     ///and applies the function on all of its elements.
   969     /// The upper bound of a variable (column) has to be given by an 
   970     /// extended number of type Value, i.e. a finite number of type 
   971     /// Value or \ref INF.
   972 #ifdef DOXYGEN
   973     template<class T>
   974     void colUpperBound(T &t, Value value) { return 0;} 
   975 #else
   976     template<class T>
   977     typename enable_if<typename T::value_type::LpSolverCol,void>::type
   978     colUpperBound(T &t, Value value,dummy<0> = 0) {
   979       for(typename T::iterator i=t.begin();i!=t.end();++i) {
   980 	colUpperBound(*i, value);
   981       }
   982     }
   983     template<class T>
   984     typename enable_if<typename T::value_type::second_type::LpSolverCol,
   985 		       void>::type
   986     colUpperBound(T &t, Value value,dummy<1> = 1) { 
   987       for(typename T::iterator i=t.begin();i!=t.end();++i) {
   988 	colUpperBound(i->second, value);
   989       }
   990     }
   991     template<class T>
   992     typename enable_if<typename T::MapIt::Value::LpSolverCol,
   993 		       void>::type
   994     colUpperBound(T &t, Value value,dummy<2> = 2) { 
   995       for(typename T::MapIt i(t); i!=INVALID; ++i){
   996 	colUpperBound(*i, value);
   997       }
   998     }
   999 #endif
  1000 
  1001     /// Set the lower and the upper bounds of a column (i.e a variable)
  1002 
  1003     /// The lower and the upper bounds of
  1004     /// a variable (column) have to be given by an 
  1005     /// extended number of type Value, i.e. a finite number of type 
  1006     /// Value, -\ref INF or \ref INF.
  1007     void colBounds(Col c, Value lower, Value upper) {
  1008       _setColLowerBound(cols.floatingId(c.id),lower);
  1009       _setColUpperBound(cols.floatingId(c.id),upper);
  1010     }
  1011     
  1012     ///\brief Set the lower and the upper bound of several columns
  1013     ///(i.e a variables) at once
  1014     ///
  1015     ///This magic function takes a container as its argument
  1016     ///and applies the function on all of its elements.
  1017     /// The lower and the upper bounds of
  1018     /// a variable (column) have to be given by an 
  1019     /// extended number of type Value, i.e. a finite number of type 
  1020     /// Value, -\ref INF or \ref INF.
  1021 #ifdef DOXYGEN
  1022     template<class T>
  1023     void colBounds(T &t, Value lower, Value upper) { return 0;} 
  1024 #else
  1025     template<class T>
  1026     typename enable_if<typename T::value_type::LpSolverCol,void>::type
  1027     colBounds(T &t, Value lower, Value upper,dummy<0> = 0) {
  1028       for(typename T::iterator i=t.begin();i!=t.end();++i) {
  1029 	colBounds(*i, lower, upper);
  1030       }
  1031     }
  1032     template<class T>
  1033     typename enable_if<typename T::value_type::second_type::LpSolverCol,
  1034 		       void>::type
  1035     colBounds(T &t, Value lower, Value upper,dummy<1> = 1) { 
  1036       for(typename T::iterator i=t.begin();i!=t.end();++i) {
  1037 	colBounds(i->second, lower, upper);
  1038       }
  1039     }
  1040     template<class T>
  1041     typename enable_if<typename T::MapIt::Value::LpSolverCol,
  1042 		       void>::type
  1043     colBounds(T &t, Value lower, Value upper,dummy<2> = 2) { 
  1044       for(typename T::MapIt i(t); i!=INVALID; ++i){
  1045 	colBounds(*i, lower, upper);
  1046       }
  1047     }
  1048 #endif
  1049     
  1050 //     /// Set the lower bound of a row (i.e a constraint)
  1051 
  1052 //     /// The lower bound of a linear expression (row) has to be given by an 
  1053 //     /// extended number of type Value, i.e. a finite number of type 
  1054 //     /// Value or -\ref INF.
  1055 //     void rowLowerBound(Row r, Value value) {
  1056 //       _setRowLowerBound(rows.floatingId(r.id),value);
  1057 //     };
  1058 //     /// Set the upper bound of a row (i.e a constraint)
  1059 
  1060 //     /// The upper bound of a linear expression (row) has to be given by an 
  1061 //     /// extended number of type Value, i.e. a finite number of type 
  1062 //     /// Value or \ref INF.
  1063 //     void rowUpperBound(Row r, Value value) {
  1064 //       _setRowUpperBound(rows.floatingId(r.id),value);
  1065 //     };
  1066 
  1067     /// Set the lower and the upper bounds of a row (i.e a constraint)
  1068 
  1069     /// The lower and the upper bounds of
  1070     /// a constraint (row) have to be given by an 
  1071     /// extended number of type Value, i.e. a finite number of type 
  1072     /// Value, -\ref INF or \ref INF.
  1073     void rowBounds(Row c, Value lower, Value upper) {
  1074       _setRowBounds(rows.floatingId(c.id),lower, upper);
  1075       // _setRowUpperBound(rows.floatingId(c.id),upper);
  1076     }
  1077     
  1078     ///Set an element of the objective function
  1079     void objCoeff(Col c, Value v) {_setObjCoeff(cols.floatingId(c.id),v); };
  1080     ///Set the objective function
  1081     
  1082     ///\param e is a linear expression of type \ref Expr.
  1083     ///\bug Is should be called obj()
  1084     void setObj(Expr e) {
  1085       _clearObj();
  1086       for (Expr::iterator i=e.begin(); i!=e.end(); ++i)
  1087 	objCoeff((*i).first,(*i).second);
  1088       obj_const_comp=e.constComp();
  1089     }
  1090 
  1091     ///Maximize
  1092     void max() { _setMax(); }
  1093     ///Minimize
  1094     void min() { _setMin(); }
  1095 
  1096     
  1097     ///@}
  1098 
  1099 
  1100     ///\name Solve the LP
  1101 
  1102     ///@{
  1103 
  1104     ///\e Solve the LP problem at hand
  1105     ///
  1106     ///\return The result of the optimization procedure. Possible values and their meanings can be found in the documentation of \ref SolveExitStatus.
  1107     ///
  1108     ///\todo Which method is used to solve the problem
  1109     SolveExitStatus solve() { return _solve(); }
  1110     
  1111     ///@}
  1112     
  1113     ///\name Obtain the solution
  1114 
  1115     ///@{
  1116 
  1117     /// The status of the primal problem (the original LP problem)
  1118     SolutionStatus primalStatus() {
  1119       return _getPrimalStatus();
  1120     }
  1121 
  1122     /// The status of the dual (of the original LP) problem 
  1123     SolutionStatus dualStatus() {
  1124       return _getDualStatus();
  1125     }
  1126 
  1127     ///The type of the original LP problem
  1128     ProblemTypes problemType() {
  1129       return _getProblemType();
  1130     }
  1131 
  1132     ///\e
  1133     Value primal(Col c) { return _getPrimal(cols.floatingId(c.id)); }
  1134 
  1135     ///\e
  1136     Value dual(Row r) { return _getDual(rows.floatingId(r.id)); }
  1137 
  1138     ///\e
  1139     bool isBasicCol(Col c) { return _isBasicCol(cols.floatingId(c.id)); }
  1140 
  1141     ///\e
  1142 
  1143     ///\return
  1144     ///- \ref INF or -\ref INF means either infeasibility or unboundedness
  1145     /// of the primal problem, depending on whether we minimize or maximize.
  1146     ///- \ref NaN if no primal solution is found.
  1147     ///- The (finite) objective value if an optimal solution is found.
  1148     Value primalValue() { return _getPrimalValue()+obj_const_comp;}
  1149     ///@}
  1150     
  1151   };  
  1152 
  1153   ///\e
  1154   
  1155   ///\relates LpSolverBase::Expr
  1156   ///
  1157   inline LpSolverBase::Expr operator+(const LpSolverBase::Expr &a,
  1158 				      const LpSolverBase::Expr &b) 
  1159   {
  1160     LpSolverBase::Expr tmp(a);
  1161     tmp+=b;
  1162     return tmp;
  1163   }
  1164   ///\e
  1165   
  1166   ///\relates LpSolverBase::Expr
  1167   ///
  1168   inline LpSolverBase::Expr operator-(const LpSolverBase::Expr &a,
  1169 				      const LpSolverBase::Expr &b) 
  1170   {
  1171     LpSolverBase::Expr tmp(a);
  1172     tmp-=b;
  1173     return tmp;
  1174   }
  1175   ///\e
  1176   
  1177   ///\relates LpSolverBase::Expr
  1178   ///
  1179   inline LpSolverBase::Expr operator*(const LpSolverBase::Expr &a,
  1180 				      const LpSolverBase::Value &b) 
  1181   {
  1182     LpSolverBase::Expr tmp(a);
  1183     tmp*=b;
  1184     return tmp;
  1185   }
  1186   
  1187   ///\e
  1188   
  1189   ///\relates LpSolverBase::Expr
  1190   ///
  1191   inline LpSolverBase::Expr operator*(const LpSolverBase::Value &a,
  1192 				      const LpSolverBase::Expr &b) 
  1193   {
  1194     LpSolverBase::Expr tmp(b);
  1195     tmp*=a;
  1196     return tmp;
  1197   }
  1198   ///\e
  1199   
  1200   ///\relates LpSolverBase::Expr
  1201   ///
  1202   inline LpSolverBase::Expr operator/(const LpSolverBase::Expr &a,
  1203 				      const LpSolverBase::Value &b) 
  1204   {
  1205     LpSolverBase::Expr tmp(a);
  1206     tmp/=b;
  1207     return tmp;
  1208   }
  1209   
  1210   ///\e
  1211   
  1212   ///\relates LpSolverBase::Constr
  1213   ///
  1214   inline LpSolverBase::Constr operator<=(const LpSolverBase::Expr &e,
  1215 					 const LpSolverBase::Expr &f) 
  1216   {
  1217     return LpSolverBase::Constr(-LpSolverBase::INF,e-f,0);
  1218   }
  1219 
  1220   ///\e
  1221   
  1222   ///\relates LpSolverBase::Constr
  1223   ///
  1224   inline LpSolverBase::Constr operator<=(const LpSolverBase::Value &e,
  1225 					 const LpSolverBase::Expr &f) 
  1226   {
  1227     return LpSolverBase::Constr(e,f);
  1228   }
  1229 
  1230   ///\e
  1231   
  1232   ///\relates LpSolverBase::Constr
  1233   ///
  1234   inline LpSolverBase::Constr operator<=(const LpSolverBase::Expr &e,
  1235 					 const LpSolverBase::Value &f) 
  1236   {
  1237     return LpSolverBase::Constr(e,f);
  1238   }
  1239 
  1240   ///\e
  1241   
  1242   ///\relates LpSolverBase::Constr
  1243   ///
  1244   inline LpSolverBase::Constr operator>=(const LpSolverBase::Expr &e,
  1245 					 const LpSolverBase::Expr &f) 
  1246   {
  1247     return LpSolverBase::Constr(-LpSolverBase::INF,f-e,0);
  1248   }
  1249 
  1250 
  1251   ///\e
  1252   
  1253   ///\relates LpSolverBase::Constr
  1254   ///
  1255   inline LpSolverBase::Constr operator>=(const LpSolverBase::Value &e,
  1256 					 const LpSolverBase::Expr &f) 
  1257   {
  1258     return LpSolverBase::Constr(f,e);
  1259   }
  1260 
  1261 
  1262   ///\e
  1263   
  1264   ///\relates LpSolverBase::Constr
  1265   ///
  1266   inline LpSolverBase::Constr operator>=(const LpSolverBase::Expr &e,
  1267 					 const LpSolverBase::Value &f) 
  1268   {
  1269     return LpSolverBase::Constr(f,e);
  1270   }
  1271 
  1272   ///\e
  1273   
  1274   ///\relates LpSolverBase::Constr
  1275   ///
  1276   inline LpSolverBase::Constr operator==(const LpSolverBase::Expr &e,
  1277 					 const LpSolverBase::Expr &f) 
  1278   {
  1279     return LpSolverBase::Constr(0,e-f,0);
  1280   }
  1281 
  1282   ///\e
  1283   
  1284   ///\relates LpSolverBase::Constr
  1285   ///
  1286   inline LpSolverBase::Constr operator<=(const LpSolverBase::Value &n,
  1287 					 const LpSolverBase::Constr&c) 
  1288   {
  1289     LpSolverBase::Constr tmp(c);
  1290     ///\todo Create an own exception type.
  1291     if(!isnan(tmp.lowerBound())) throw LogicError();
  1292     else tmp.lowerBound()=n;
  1293     return tmp;
  1294   }
  1295   ///\e
  1296   
  1297   ///\relates LpSolverBase::Constr
  1298   ///
  1299   inline LpSolverBase::Constr operator<=(const LpSolverBase::Constr& c,
  1300 					 const LpSolverBase::Value &n)
  1301   {
  1302     LpSolverBase::Constr tmp(c);
  1303     ///\todo Create an own exception type.
  1304     if(!isnan(tmp.upperBound())) throw LogicError();
  1305     else tmp.upperBound()=n;
  1306     return tmp;
  1307   }
  1308 
  1309   ///\e
  1310   
  1311   ///\relates LpSolverBase::Constr
  1312   ///
  1313   inline LpSolverBase::Constr operator>=(const LpSolverBase::Value &n,
  1314 					 const LpSolverBase::Constr&c) 
  1315   {
  1316     LpSolverBase::Constr tmp(c);
  1317     ///\todo Create an own exception type.
  1318     if(!isnan(tmp.upperBound())) throw LogicError();
  1319     else tmp.upperBound()=n;
  1320     return tmp;
  1321   }
  1322   ///\e
  1323   
  1324   ///\relates LpSolverBase::Constr
  1325   ///
  1326   inline LpSolverBase::Constr operator>=(const LpSolverBase::Constr& c,
  1327 					 const LpSolverBase::Value &n)
  1328   {
  1329     LpSolverBase::Constr tmp(c);
  1330     ///\todo Create an own exception type.
  1331     if(!isnan(tmp.lowerBound())) throw LogicError();
  1332     else tmp.lowerBound()=n;
  1333     return tmp;
  1334   }
  1335 
  1336   ///\e
  1337   
  1338   ///\relates LpSolverBase::DualExpr
  1339   ///
  1340   inline LpSolverBase::DualExpr operator+(const LpSolverBase::DualExpr &a,
  1341 				      const LpSolverBase::DualExpr &b) 
  1342   {
  1343     LpSolverBase::DualExpr tmp(a);
  1344     tmp+=b;
  1345     return tmp;
  1346   }
  1347   ///\e
  1348   
  1349   ///\relates LpSolverBase::DualExpr
  1350   ///
  1351   inline LpSolverBase::DualExpr operator-(const LpSolverBase::DualExpr &a,
  1352 				      const LpSolverBase::DualExpr &b) 
  1353   {
  1354     LpSolverBase::DualExpr tmp(a);
  1355     tmp-=b;
  1356     return tmp;
  1357   }
  1358   ///\e
  1359   
  1360   ///\relates LpSolverBase::DualExpr
  1361   ///
  1362   inline LpSolverBase::DualExpr operator*(const LpSolverBase::DualExpr &a,
  1363 				      const LpSolverBase::Value &b) 
  1364   {
  1365     LpSolverBase::DualExpr tmp(a);
  1366     tmp*=b;
  1367     return tmp;
  1368   }
  1369   
  1370   ///\e
  1371   
  1372   ///\relates LpSolverBase::DualExpr
  1373   ///
  1374   inline LpSolverBase::DualExpr operator*(const LpSolverBase::Value &a,
  1375 				      const LpSolverBase::DualExpr &b) 
  1376   {
  1377     LpSolverBase::DualExpr tmp(b);
  1378     tmp*=a;
  1379     return tmp;
  1380   }
  1381   ///\e
  1382   
  1383   ///\relates LpSolverBase::DualExpr
  1384   ///
  1385   inline LpSolverBase::DualExpr operator/(const LpSolverBase::DualExpr &a,
  1386 				      const LpSolverBase::Value &b) 
  1387   {
  1388     LpSolverBase::DualExpr tmp(a);
  1389     tmp/=b;
  1390     return tmp;
  1391   }
  1392   
  1393 
  1394 } //namespace lemon
  1395 
  1396 #endif //LEMON_LP_BASE_H