lemon/lp_base.h
author alpar
Tue, 18 Apr 2006 13:25:41 +0000
changeset 2062 35e0355c6d9b
parent 1993 2115143eceea
child 2085 1970a93dfaa8
permissions -rw-r--r--
Also check mingw compatibility
     1 /* -*- C++ -*-
     2  *
     3  * This file is a part of LEMON, a generic C++ optimization library
     4  *
     5  * Copyright (C) 2003-2006
     6  * Egervary Jeno Kombinatorikus Optimalizalasi Kutatocsoport
     7  * (Egervary Research Group on Combinatorial Optimization, EGRES).
     8  *
     9  * Permission to use, modify and distribute this software is granted
    10  * provided that this copyright notice appears in all copies. For
    11  * precise terms see the accompanying LICENSE file.
    12  *
    13  * This software is provided "AS IS" with no warranty of any kind,
    14  * express or implied, and with no claim as to its suitability for any
    15  * purpose.
    16  *
    17  */
    18 
    19 #ifndef LEMON_LP_BASE_H
    20 #define LEMON_LP_BASE_H
    21 
    22 #include<vector>
    23 #include<map>
    24 #include<limits>
    25 #include<cmath>
    26 
    27 #include<lemon/bits/utility.h>
    28 #include<lemon/error.h>
    29 #include<lemon/bits/invalid.h>
    30 
    31 ///\file
    32 ///\brief The interface of the LP solver interface.
    33 ///\ingroup gen_opt_group
    34 namespace lemon {
    35   
    36   ///Internal data structure to convert floating id's to fix one's
    37     
    38   ///\todo This might be implemented to be also usable in other places.
    39   class _FixId 
    40   {
    41   protected:
    42     std::vector<int> index;
    43     std::vector<int> cross;
    44     int first_free;
    45   public:
    46     _FixId() : first_free(-1) {};
    47     ///Convert a floating id to a fix one
    48 
    49     ///\param n is a floating id
    50     ///\return the corresponding fix id
    51     int fixId(int n) const {return cross[n];}
    52     ///Convert a fix id to a floating one
    53 
    54     ///\param n is a fix id
    55     ///\return the corresponding floating id
    56     int floatingId(int n) const { return index[n];}
    57     ///Add a new floating id.
    58 
    59     ///\param n is a floating id
    60     ///\return the fix id of the new value
    61     ///\todo Multiple additions should also be handled.
    62     int insert(int n)
    63     {
    64       if(n>=int(cross.size())) {
    65 	cross.resize(n+1);
    66 	if(first_free==-1) {
    67 	  cross[n]=index.size();
    68 	  index.push_back(n);
    69 	}
    70 	else {
    71 	  cross[n]=first_free;
    72 	  int next=index[first_free];
    73 	  index[first_free]=n;
    74 	  first_free=next;
    75 	}
    76 	return cross[n];
    77       }
    78       ///\todo Create an own exception type.
    79       else throw LogicError(); //floatingId-s must form a continuous range;
    80     }
    81     ///Remove a fix id.
    82 
    83     ///\param n is a fix id
    84     ///
    85     void erase(int n) 
    86     {
    87       int fl=index[n];
    88       index[n]=first_free;
    89       first_free=n;
    90       for(int i=fl+1;i<int(cross.size());++i) {
    91 	cross[i-1]=cross[i];
    92 	index[cross[i]]--;
    93       }
    94       cross.pop_back();
    95     }
    96     ///An upper bound on the largest fix id.
    97 
    98     ///\todo Do we need this?
    99     ///
   100     std::size_t maxFixId() { return cross.size()-1; }
   101   
   102   };
   103     
   104   ///Common base class for LP solvers
   105   
   106   ///\todo Much more docs
   107   ///\ingroup gen_opt_group
   108   class LpSolverBase {
   109 
   110   public:
   111 
   112     ///Possible outcomes of an LP solving procedure
   113     enum SolveExitStatus {
   114       ///This means that the problem has been successfully solved: either
   115       ///an optimal solution has been found or infeasibility/unboundedness
   116       ///has been proved.
   117       SOLVED = 0,
   118       ///Any other case (including the case when some user specified limit has been exceeded)
   119       UNSOLVED = 1
   120     };
   121       
   122       ///\e
   123     enum SolutionStatus {
   124       ///Feasible solution has'n been found (but may exist).
   125 
   126       ///\todo NOTFOUND might be a better name.
   127       ///
   128       UNDEFINED = 0,
   129       ///The problem has no feasible solution
   130       INFEASIBLE = 1,
   131       ///Feasible solution found
   132       FEASIBLE = 2,
   133       ///Optimal solution exists and found
   134       OPTIMAL = 3,
   135       ///The cost function is unbounded
   136 
   137       ///\todo Give a feasible solution and an infinite ray (and the
   138       ///corresponding bases)
   139       INFINITE = 4
   140     };
   141 
   142     ///\e The type of the investigated LP problem
   143     enum ProblemTypes {
   144       ///Primal-dual feasible
   145       PRIMAL_DUAL_FEASIBLE = 0,
   146       ///Primal feasible dual infeasible
   147       PRIMAL_FEASIBLE_DUAL_INFEASIBLE = 1,
   148       ///Primal infeasible dual feasible
   149       PRIMAL_INFEASIBLE_DUAL_FEASIBLE = 2,
   150       ///Primal-dual infeasible
   151       PRIMAL_DUAL_INFEASIBLE = 3,
   152       ///Could not determine so far
   153       UNKNOWN = 4
   154     };
   155 
   156     ///The floating point type used by the solver
   157     typedef double Value;
   158     ///The infinity constant
   159     static const Value INF;
   160     ///The not a number constant
   161     static const Value NaN;
   162 
   163     static inline bool isNaN(const Value& v) { return v!=v; }
   164     
   165     ///Refer to a column of the LP.
   166 
   167     ///This type is used to refer to a column of the LP.
   168     ///
   169     ///Its value remains valid and correct even after the addition or erase of
   170     ///other columns.
   171     ///
   172     ///\todo Document what can one do with a Col (INVALID, comparing,
   173     ///it is similar to Node/Edge)
   174     class Col {
   175     protected:
   176       int id;
   177       friend class LpSolverBase;
   178     public:
   179       typedef Value ExprValue;
   180       typedef True LpSolverCol;
   181       Col() {}
   182       Col(const Invalid&) : id(-1) {}
   183       bool operator< (Col c) const  {return id< c.id;}
   184       bool operator> (Col c) const  {return id> c.id;}
   185       bool operator==(Col c) const  {return id==c.id;}
   186       bool operator!=(Col c) const  {return id!=c.id;}
   187     };
   188 
   189     ///Refer to a row of the LP.
   190 
   191     ///This type is used to refer to a row of the LP.
   192     ///
   193     ///Its value remains valid and correct even after the addition or erase of
   194     ///other rows.
   195     ///
   196     ///\todo Document what can one do with a Row (INVALID, comparing,
   197     ///it is similar to Node/Edge)
   198     class Row {
   199     protected:
   200       int id;
   201       friend class LpSolverBase;
   202     public:
   203       typedef Value ExprValue;
   204       typedef True LpSolverRow;
   205       Row() {}
   206       Row(const Invalid&) : id(-1) {}
   207 
   208       bool operator< (Row c) const  {return id< c.id;}
   209       bool operator> (Row c) const  {return id> c.id;}
   210       bool operator==(Row c) const  {return id==c.id;}
   211       bool operator!=(Row c) const  {return id!=c.id;} 
   212    };
   213     
   214     ///Linear expression of variables and a constant component
   215     
   216     ///This data structure strores a linear expression of the variables
   217     ///(\ref Col "Col"s) and also has a constant component.
   218     ///
   219     ///There are several ways to access and modify the contents of this
   220     ///container.
   221     ///- Its it fully compatible with \c std::map<Col,double>, so for expamle
   222     ///if \c e is an Expr and \c v and \c w are of type \ref Col, then you can
   223     ///read and modify the coefficients like
   224     ///these.
   225     ///\code
   226     ///e[v]=5;
   227     ///e[v]+=12;
   228     ///e.erase(v);
   229     ///\endcode
   230     ///or you can also iterate through its elements.
   231     ///\code
   232     ///double s=0;
   233     ///for(LpSolverBase::Expr::iterator i=e.begin();i!=e.end();++i)
   234     ///  s+=i->second;
   235     ///\endcode
   236     ///(This code computes the sum of all coefficients).
   237     ///- Numbers (<tt>double</tt>'s)
   238     ///and variables (\ref Col "Col"s) directly convert to an
   239     ///\ref Expr and the usual linear operations are defined, so  
   240     ///\code
   241     ///v+w
   242     ///2*v-3.12*(v-w/2)+2
   243     ///v*2.1+(3*v+(v*12+w+6)*3)/2
   244     ///\endcode
   245     ///are valid \ref Expr "Expr"essions.
   246     ///The usual assignment operations are also defined.
   247     ///\code
   248     ///e=v+w;
   249     ///e+=2*v-3.12*(v-w/2)+2;
   250     ///e*=3.4;
   251     ///e/=5;
   252     ///\endcode
   253     ///- The constant member can be set and read by \ref constComp()
   254     ///\code
   255     ///e.constComp()=12;
   256     ///double c=e.constComp();
   257     ///\endcode
   258     ///
   259     ///\note \ref clear() not only sets all coefficients to 0 but also
   260     ///clears the constant components.
   261     ///
   262     ///\sa Constr
   263     ///
   264     class Expr : public std::map<Col,Value>
   265     {
   266     public:
   267       typedef LpSolverBase::Col Key; 
   268       typedef LpSolverBase::Value Value;
   269       
   270     protected:
   271       typedef std::map<Col,Value> Base;
   272       
   273       Value const_comp;
   274   public:
   275       typedef True IsLinExpression;
   276       ///\e
   277       Expr() : Base(), const_comp(0) { }
   278       ///\e
   279       Expr(const Key &v) : const_comp(0) {
   280 	Base::insert(std::make_pair(v, 1));
   281       }
   282       ///\e
   283       Expr(const Value &v) : const_comp(v) {}
   284       ///\e
   285       void set(const Key &v,const Value &c) {
   286 	Base::insert(std::make_pair(v, c));
   287       }
   288       ///\e
   289       Value &constComp() { return const_comp; }
   290       ///\e
   291       const Value &constComp() const { return const_comp; }
   292       
   293       ///Removes the components with zero coefficient.
   294       void simplify() {
   295 	for (Base::iterator i=Base::begin(); i!=Base::end();) {
   296 	  Base::iterator j=i;
   297 	  ++j;
   298 	  if ((*i).second==0) Base::erase(i);
   299 	  j=i;
   300 	}
   301       }
   302 
   303       ///Removes the coefficients closer to zero than \c tolerance.
   304       void simplify(double &tolerance) {
   305 	for (Base::iterator i=Base::begin(); i!=Base::end();) {
   306 	  Base::iterator j=i;
   307 	  ++j;
   308 	  if (std::fabs((*i).second)<tolerance) Base::erase(i);
   309 	  j=i;
   310 	}
   311       }
   312 
   313       ///Sets all coefficients and the constant component to 0.
   314       void clear() {
   315 	Base::clear();
   316 	const_comp=0;
   317       }
   318 
   319       ///\e
   320       Expr &operator+=(const Expr &e) {
   321 	for (Base::const_iterator j=e.begin(); j!=e.end(); ++j)
   322 	  (*this)[j->first]+=j->second;
   323 	const_comp+=e.const_comp;
   324 	return *this;
   325       }
   326       ///\e
   327       Expr &operator-=(const Expr &e) {
   328 	for (Base::const_iterator j=e.begin(); j!=e.end(); ++j)
   329 	  (*this)[j->first]-=j->second;
   330 	const_comp-=e.const_comp;
   331 	return *this;
   332       }
   333       ///\e
   334       Expr &operator*=(const Value &c) {
   335 	for (Base::iterator j=Base::begin(); j!=Base::end(); ++j)
   336 	  j->second*=c;
   337 	const_comp*=c;
   338 	return *this;
   339       }
   340       ///\e
   341       Expr &operator/=(const Value &c) {
   342 	for (Base::iterator j=Base::begin(); j!=Base::end(); ++j)
   343 	  j->second/=c;
   344 	const_comp/=c;
   345 	return *this;
   346       }
   347     };
   348     
   349     ///Linear constraint
   350 
   351     ///This data stucture represents a linear constraint in the LP.
   352     ///Basically it is a linear expression with a lower or an upper bound
   353     ///(or both). These parts of the constraint can be obtained by the member
   354     ///functions \ref expr(), \ref lowerBound() and \ref upperBound(),
   355     ///respectively.
   356     ///There are two ways to construct a constraint.
   357     ///- You can set the linear expression and the bounds directly
   358     ///  by the functions above.
   359     ///- The operators <tt>\<=</tt>, <tt>==</tt> and  <tt>\>=</tt>
   360     ///  are defined between expressions, or even between constraints whenever
   361     ///  it makes sense. Therefore if \c e and \c f are linear expressions and
   362     ///  \c s and \c t are numbers, then the followings are valid expressions
   363     ///  and thus they can be used directly e.g. in \ref addRow() whenever
   364     ///  it makes sense.
   365     ///\code
   366     ///  e<=s
   367     ///  e<=f
   368     ///  e==f
   369     ///  s<=e<=t
   370     ///  e>=t
   371     ///\endcode
   372     ///\warning The validity of a constraint is checked only at run time, so
   373     ///e.g. \ref addRow(<tt>x[1]\<=x[2]<=5</tt>) will compile, but will throw a
   374     ///\ref LogicError exception.
   375     class Constr
   376     {
   377     public:
   378       typedef LpSolverBase::Expr Expr;
   379       typedef Expr::Key Key;
   380       typedef Expr::Value Value;
   381       
   382 //       static const Value INF;
   383 //       static const Value NaN;
   384 
   385     protected:
   386       Expr _expr;
   387       Value _lb,_ub;
   388     public:
   389       ///\e
   390       Constr() : _expr(), _lb(NaN), _ub(NaN) {}
   391       ///\e
   392       Constr(Value lb,const Expr &e,Value ub) :
   393 	_expr(e), _lb(lb), _ub(ub) {}
   394       ///\e
   395       Constr(const Expr &e,Value ub) : 
   396 	_expr(e), _lb(NaN), _ub(ub) {}
   397       ///\e
   398       Constr(Value lb,const Expr &e) :
   399 	_expr(e), _lb(lb), _ub(NaN) {}
   400       ///\e
   401       Constr(const Expr &e) : 
   402 	_expr(e), _lb(NaN), _ub(NaN) {}
   403       ///\e
   404       void clear() 
   405       {
   406 	_expr.clear();
   407 	_lb=_ub=NaN;
   408       }
   409 
   410       ///Reference to the linear expression 
   411       Expr &expr() { return _expr; }
   412       ///Cont reference to the linear expression 
   413       const Expr &expr() const { return _expr; }
   414       ///Reference to the lower bound.
   415 
   416       ///\return
   417       ///- \ref INF "INF": the constraint is lower unbounded.
   418       ///- \ref NaN "NaN": lower bound has not been set.
   419       ///- finite number: the lower bound
   420       Value &lowerBound() { return _lb; }
   421       ///The const version of \ref lowerBound()
   422       const Value &lowerBound() const { return _lb; }
   423       ///Reference to the upper bound.
   424 
   425       ///\return
   426       ///- \ref INF "INF": the constraint is upper unbounded.
   427       ///- \ref NaN "NaN": upper bound has not been set.
   428       ///- finite number: the upper bound
   429       Value &upperBound() { return _ub; }
   430       ///The const version of \ref upperBound()
   431       const Value &upperBound() const { return _ub; }
   432       ///Is the constraint lower bounded?
   433       bool lowerBounded() const { 
   434 	using namespace std;
   435 	return finite(_lb);
   436       }
   437       ///Is the constraint upper bounded?
   438       bool upperBounded() const {
   439 	using namespace std;
   440 	return finite(_ub);
   441       }
   442     };
   443     
   444     ///Linear expression of rows
   445     
   446     ///This data structure represents a column of the matrix,
   447     ///thas is it strores a linear expression of the dual variables
   448     ///(\ref Row "Row"s).
   449     ///
   450     ///There are several ways to access and modify the contents of this
   451     ///container.
   452     ///- Its it fully compatible with \c std::map<Row,double>, so for expamle
   453     ///if \c e is an DualExpr and \c v
   454     ///and \c w are of type \ref Row, then you can
   455     ///read and modify the coefficients like
   456     ///these.
   457     ///\code
   458     ///e[v]=5;
   459     ///e[v]+=12;
   460     ///e.erase(v);
   461     ///\endcode
   462     ///or you can also iterate through its elements.
   463     ///\code
   464     ///double s=0;
   465     ///for(LpSolverBase::DualExpr::iterator i=e.begin();i!=e.end();++i)
   466     ///  s+=i->second;
   467     ///\endcode
   468     ///(This code computes the sum of all coefficients).
   469     ///- Numbers (<tt>double</tt>'s)
   470     ///and variables (\ref Row "Row"s) directly convert to an
   471     ///\ref DualExpr and the usual linear operations are defined, so
   472     ///\code
   473     ///v+w
   474     ///2*v-3.12*(v-w/2)
   475     ///v*2.1+(3*v+(v*12+w)*3)/2
   476     ///\endcode
   477     ///are valid \ref DualExpr "DualExpr"essions.
   478     ///The usual assignment operations are also defined.
   479     ///\code
   480     ///e=v+w;
   481     ///e+=2*v-3.12*(v-w/2);
   482     ///e*=3.4;
   483     ///e/=5;
   484     ///\endcode
   485     ///
   486     ///\sa Expr
   487     ///
   488     class DualExpr : public std::map<Row,Value>
   489     {
   490     public:
   491       typedef LpSolverBase::Row Key; 
   492       typedef LpSolverBase::Value Value;
   493       
   494     protected:
   495       typedef std::map<Row,Value> Base;
   496       
   497     public:
   498       typedef True IsLinExpression;
   499       ///\e
   500       DualExpr() : Base() { }
   501       ///\e
   502       DualExpr(const Key &v) {
   503 	Base::insert(std::make_pair(v, 1));
   504       }
   505       ///\e
   506       void set(const Key &v,const Value &c) {
   507 	Base::insert(std::make_pair(v, c));
   508       }
   509       
   510       ///Removes the components with zero coefficient.
   511       void simplify() {
   512 	for (Base::iterator i=Base::begin(); i!=Base::end();) {
   513 	  Base::iterator j=i;
   514 	  ++j;
   515 	  if ((*i).second==0) Base::erase(i);
   516 	  j=i;
   517 	}
   518       }
   519 
   520       ///Removes the coefficients closer to zero than \c tolerance.
   521       void simplify(double &tolerance) {
   522 	for (Base::iterator i=Base::begin(); i!=Base::end();) {
   523 	  Base::iterator j=i;
   524 	  ++j;
   525 	  if (std::fabs((*i).second)<tolerance) Base::erase(i);
   526 	  j=i;
   527 	}
   528       }
   529 
   530 
   531       ///Sets all coefficients to 0.
   532       void clear() {
   533 	Base::clear();
   534       }
   535 
   536       ///\e
   537       DualExpr &operator+=(const DualExpr &e) {
   538 	for (Base::const_iterator j=e.begin(); j!=e.end(); ++j)
   539 	  (*this)[j->first]+=j->second;
   540 	return *this;
   541       }
   542       ///\e
   543       DualExpr &operator-=(const DualExpr &e) {
   544 	for (Base::const_iterator j=e.begin(); j!=e.end(); ++j)
   545 	  (*this)[j->first]-=j->second;
   546 	return *this;
   547       }
   548       ///\e
   549       DualExpr &operator*=(const Value &c) {
   550 	for (Base::iterator j=Base::begin(); j!=Base::end(); ++j)
   551 	  j->second*=c;
   552 	return *this;
   553       }
   554       ///\e
   555       DualExpr &operator/=(const Value &c) {
   556 	for (Base::iterator j=Base::begin(); j!=Base::end(); ++j)
   557 	  j->second/=c;
   558 	return *this;
   559       }
   560     };
   561     
   562 
   563   protected:
   564     _FixId rows;
   565     _FixId cols;
   566 
   567     //Abstract virtual functions
   568     virtual LpSolverBase &_newLp() = 0;
   569     virtual LpSolverBase &_copyLp(){
   570       ///\todo This should be implemented here, too,  when we have problem retrieving routines. It can be overriden.
   571 
   572       //Starting:
   573       LpSolverBase & newlp(_newLp());
   574       return newlp;
   575       //return *(LpSolverBase*)0;
   576     };
   577 
   578     virtual int _addCol() = 0;
   579     virtual int _addRow() = 0;
   580     virtual void _eraseCol(int col) = 0;
   581     virtual void _eraseRow(int row) = 0;
   582     virtual void _getColName(int col,       std::string & name) = 0;
   583     virtual void _setColName(int col, const std::string & name) = 0;
   584     virtual void _setRowCoeffs(int i, 
   585 			       int length,
   586                                int  const * indices, 
   587                                Value  const * values ) = 0;
   588     virtual void _setColCoeffs(int i, 
   589 			       int length,
   590                                int  const * indices, 
   591                                Value  const * values ) = 0;
   592     virtual void _setCoeff(int row, int col, Value value) = 0;
   593     virtual void _setColLowerBound(int i, Value value) = 0;
   594     virtual void _setColUpperBound(int i, Value value) = 0;
   595 //     virtual void _setRowLowerBound(int i, Value value) = 0;
   596 //     virtual void _setRowUpperBound(int i, Value value) = 0;
   597     virtual void _setRowBounds(int i, Value lower, Value upper) = 0;
   598     virtual void _setObjCoeff(int i, Value obj_coef) = 0;
   599     virtual void _clearObj()=0;
   600 //     virtual void _setObj(int length,
   601 //                          int  const * indices, 
   602 //                          Value  const * values ) = 0;
   603     virtual SolveExitStatus _solve() = 0;
   604     virtual Value _getPrimal(int i) = 0;
   605     virtual Value _getDual(int i) = 0;
   606     virtual Value _getPrimalValue() = 0;
   607     virtual bool _isBasicCol(int i) = 0;
   608     virtual SolutionStatus _getPrimalStatus() = 0;
   609     virtual SolutionStatus _getDualStatus() = 0;
   610     ///\todo This could be implemented here, too, using _getPrimalStatus() and
   611     ///_getDualStatus()
   612     virtual ProblemTypes _getProblemType() = 0;
   613 
   614     virtual void _setMax() = 0;
   615     virtual void _setMin() = 0;
   616     
   617     //Own protected stuff
   618     
   619     //Constant component of the objective function
   620     Value obj_const_comp;
   621     
   622 
   623 
   624     
   625   public:
   626 
   627     ///\e
   628     LpSolverBase() : obj_const_comp(0) {}
   629 
   630     ///\e
   631     virtual ~LpSolverBase() {}
   632 
   633     ///Creates a new LP problem
   634     LpSolverBase &newLp() {return _newLp();}
   635     ///Makes a copy of the LP problem
   636     LpSolverBase &copyLp() {return _copyLp();}
   637     
   638     ///\name Build up and modify the LP
   639 
   640     ///@{
   641 
   642     ///Add a new empty column (i.e a new variable) to the LP
   643     Col addCol() { Col c; c.id=cols.insert(_addCol()); return c;}
   644 
   645     ///\brief Adds several new columns
   646     ///(i.e a variables) at once
   647     ///
   648     ///This magic function takes a container as its argument
   649     ///and fills its elements
   650     ///with new columns (i.e. variables)
   651     ///\param t can be
   652     ///- a standard STL compatible iterable container with
   653     ///\ref Col as its \c values_type
   654     ///like
   655     ///\code
   656     ///std::vector<LpSolverBase::Col>
   657     ///std::list<LpSolverBase::Col>
   658     ///\endcode
   659     ///- a standard STL compatible iterable container with
   660     ///\ref Col as its \c mapped_type
   661     ///like
   662     ///\code
   663     ///std::map<AnyType,LpSolverBase::Col>
   664     ///\endcode
   665     ///- an iterable lemon \ref concept::WriteMap "write map" like 
   666     ///\code
   667     ///ListGraph::NodeMap<LpSolverBase::Col>
   668     ///ListGraph::EdgeMap<LpSolverBase::Col>
   669     ///\endcode
   670     ///\return The number of the created column.
   671 #ifdef DOXYGEN
   672     template<class T>
   673     int addColSet(T &t) { return 0;} 
   674 #else
   675     template<class T>
   676     typename enable_if<typename T::value_type::LpSolverCol,int>::type
   677     addColSet(T &t,dummy<0> = 0) {
   678       int s=0;
   679       for(typename T::iterator i=t.begin();i!=t.end();++i) {*i=addCol();s++;}
   680       return s;
   681     }
   682     template<class T>
   683     typename enable_if<typename T::value_type::second_type::LpSolverCol,
   684 		       int>::type
   685     addColSet(T &t,dummy<1> = 1) { 
   686       int s=0;
   687       for(typename T::iterator i=t.begin();i!=t.end();++i) {
   688 	i->second=addCol();
   689 	s++;
   690       }
   691       return s;
   692     }
   693     template<class T>
   694     typename enable_if<typename T::MapIt::Value::LpSolverCol,
   695 		       int>::type
   696     addColSet(T &t,dummy<2> = 2) { 
   697       int s=0;
   698       for(typename T::MapIt i(t); i!=INVALID; ++i)
   699 	{
   700 	  i.set(addCol());
   701 	  s++;
   702 	}
   703       return s;
   704     }
   705 #endif
   706 
   707     ///Set a column (i.e a dual constraint) of the LP
   708 
   709     ///\param c is the column to be modified
   710     ///\param e is a dual linear expression (see \ref DualExpr)
   711     ///a better one.
   712     void col(Col c,const DualExpr &e) {
   713       std::vector<int> indices;
   714       std::vector<Value> values;
   715       indices.push_back(0);
   716       values.push_back(0);
   717       for(DualExpr::const_iterator i=e.begin(); i!=e.end(); ++i)
   718 	if((*i).second!=0) {
   719 	  indices.push_back(rows.floatingId((*i).first.id));
   720 	  values.push_back((*i).second);
   721 	}
   722       _setColCoeffs(cols.floatingId(c.id),indices.size()-1,
   723 		    &indices[0],&values[0]);
   724     }
   725 
   726     ///Add a new column to the LP
   727 
   728     ///\param e is a dual linear expression (see \ref DualExpr)
   729     ///\param obj is the corresponding component of the objective
   730     ///function. It is 0 by default.
   731     ///\return The created column.
   732     Col addCol(const DualExpr &e, Value obj=0) {
   733       Col c=addCol();
   734       col(c,e);
   735       objCoeff(c,obj);
   736       return c;
   737     }
   738 
   739     ///Add a new empty row (i.e a new constraint) to the LP
   740 
   741     ///This function adds a new empty row (i.e a new constraint) to the LP.
   742     ///\return The created row
   743     Row addRow() { Row r; r.id=rows.insert(_addRow()); return r;}
   744 
   745     ///\brief Add several new rows
   746     ///(i.e a constraints) at once
   747     ///
   748     ///This magic function takes a container as its argument
   749     ///and fills its elements
   750     ///with new row (i.e. variables)
   751     ///\param t can be
   752     ///- a standard STL compatible iterable container with
   753     ///\ref Row as its \c values_type
   754     ///like
   755     ///\code
   756     ///std::vector<LpSolverBase::Row>
   757     ///std::list<LpSolverBase::Row>
   758     ///\endcode
   759     ///- a standard STL compatible iterable container with
   760     ///\ref Row as its \c mapped_type
   761     ///like
   762     ///\code
   763     ///std::map<AnyType,LpSolverBase::Row>
   764     ///\endcode
   765     ///- an iterable lemon \ref concept::WriteMap "write map" like 
   766     ///\code
   767     ///ListGraph::NodeMap<LpSolverBase::Row>
   768     ///ListGraph::EdgeMap<LpSolverBase::Row>
   769     ///\endcode
   770     ///\return The number of rows created.
   771 #ifdef DOXYGEN
   772     template<class T>
   773     int addRowSet(T &t) { return 0;} 
   774 #else
   775     template<class T>
   776     typename enable_if<typename T::value_type::LpSolverRow,int>::type
   777     addRowSet(T &t,dummy<0> = 0) {
   778       int s=0;
   779       for(typename T::iterator i=t.begin();i!=t.end();++i) {*i=addRow();s++;}
   780       return s;
   781     }
   782     template<class T>
   783     typename enable_if<typename T::value_type::second_type::LpSolverRow,
   784 		       int>::type
   785     addRowSet(T &t,dummy<1> = 1) { 
   786       int s=0;
   787       for(typename T::iterator i=t.begin();i!=t.end();++i) {
   788 	i->second=addRow();
   789 	s++;
   790       }
   791       return s;
   792     }
   793     template<class T>
   794     typename enable_if<typename T::MapIt::Value::LpSolverRow,
   795 		       int>::type
   796     addRowSet(T &t,dummy<2> = 2) { 
   797       int s=0;
   798       for(typename T::MapIt i(t); i!=INVALID; ++i)
   799 	{
   800 	  i.set(addRow());
   801 	  s++;
   802 	}
   803       return s;
   804     }
   805 #endif
   806 
   807     ///Set a row (i.e a constraint) of the LP
   808 
   809     ///\param r is the row to be modified
   810     ///\param l is lower bound (-\ref INF means no bound)
   811     ///\param e is a linear expression (see \ref Expr)
   812     ///\param u is the upper bound (\ref INF means no bound)
   813     ///\bug This is a temportary function. The interface will change to
   814     ///a better one.
   815     ///\todo Option to control whether a constraint with a single variable is
   816     ///added or not.
   817     void row(Row r, Value l,const Expr &e, Value u) {
   818       std::vector<int> indices;
   819       std::vector<Value> values;
   820       indices.push_back(0);
   821       values.push_back(0);
   822       for(Expr::const_iterator i=e.begin(); i!=e.end(); ++i)
   823 	if((*i).second!=0) { ///\bug EPSILON would be necessary here!!!
   824 	  indices.push_back(cols.floatingId((*i).first.id));
   825 	  values.push_back((*i).second);
   826 	}
   827       _setRowCoeffs(rows.floatingId(r.id),indices.size()-1,
   828 		    &indices[0],&values[0]);
   829 //       _setRowLowerBound(rows.floatingId(r.id),l-e.constComp());
   830 //       _setRowUpperBound(rows.floatingId(r.id),u-e.constComp());
   831        _setRowBounds(rows.floatingId(r.id),l-e.constComp(),u-e.constComp());
   832     }
   833 
   834     ///Set a row (i.e a constraint) of the LP
   835 
   836     ///\param r is the row to be modified
   837     ///\param c is a linear expression (see \ref Constr)
   838     void row(Row r, const Constr &c) {
   839       row(r,
   840 	     c.lowerBounded()?c.lowerBound():-INF,
   841 	     c.expr(),
   842 	     c.upperBounded()?c.upperBound():INF);
   843     }
   844 
   845     ///Add a new row (i.e a new constraint) to the LP
   846 
   847     ///\param l is the lower bound (-\ref INF means no bound)
   848     ///\param e is a linear expression (see \ref Expr)
   849     ///\param u is the upper bound (\ref INF means no bound)
   850     ///\return The created row.
   851     ///\bug This is a temportary function. The interface will change to
   852     ///a better one.
   853     Row addRow(Value l,const Expr &e, Value u) {
   854       Row r=addRow();
   855       row(r,l,e,u);
   856       return r;
   857     }
   858 
   859     ///Add a new row (i.e a new constraint) to the LP
   860 
   861     ///\param c is a linear expression (see \ref Constr)
   862     ///\return The created row.
   863     Row addRow(const Constr &c) {
   864       Row r=addRow();
   865       row(r,c);
   866       return r;
   867     }
   868     ///Erase a coloumn (i.e a variable) from the LP
   869 
   870     ///\param c is the coloumn to be deleted
   871     ///\todo Please check this
   872     void eraseCol(Col c) {
   873       _eraseCol(cols.floatingId(c.id));
   874       cols.erase(c.id);
   875     }
   876     ///Erase a  row (i.e a constraint) from the LP
   877 
   878     ///\param r is the row to be deleted
   879     ///\todo Please check this
   880     void eraseRow(Row r) {
   881       _eraseRow(rows.floatingId(r.id));
   882       rows.erase(r.id);
   883     }
   884 
   885     /// Get the name of a column
   886     
   887     ///\param c is the coresponding coloumn 
   888     ///\return The name of the colunm
   889     std::string ColName(Col c){
   890       std::string name;
   891       _getColName(cols.floatingId(c.id), name);
   892       return name;
   893     }
   894     
   895     /// Set the name of a column
   896     
   897     ///\param c is the coresponding coloumn 
   898     ///\param name The name to be given
   899     void ColName(Col c, const std::string & name){
   900       _setColName(cols.floatingId(c.id), name);
   901     }
   902     
   903     /// Set an element of the coefficient matrix of the LP
   904 
   905     ///\param r is the row of the element to be modified
   906     ///\param c is the coloumn of the element to be modified
   907     ///\param val is the new value of the coefficient
   908 
   909     void Coeff(Row r, Col c, Value val){
   910       _setCoeff(rows.floatingId(r.id),cols.floatingId(c.id), val);
   911     }
   912 
   913     /// Set the lower bound of a column (i.e a variable)
   914 
   915     /// The lower bound of a variable (column) has to be given by an 
   916     /// extended number of type Value, i.e. a finite number of type 
   917     /// Value or -\ref INF.
   918     void colLowerBound(Col c, Value value) {
   919       _setColLowerBound(cols.floatingId(c.id),value);
   920     }
   921     
   922     ///\brief Set the lower bound of  several columns
   923     ///(i.e a variables) at once
   924     ///
   925     ///This magic function takes a container as its argument
   926     ///and applies the function on all of its elements.
   927     /// The lower bound of a variable (column) has to be given by an 
   928     /// extended number of type Value, i.e. a finite number of type 
   929     /// Value or -\ref INF.
   930 #ifdef DOXYGEN
   931     template<class T>
   932     void colLowerBound(T &t, Value value) { return 0;} 
   933 #else
   934     template<class T>
   935     typename enable_if<typename T::value_type::LpSolverCol,void>::type
   936     colLowerBound(T &t, Value value,dummy<0> = 0) {
   937       for(typename T::iterator i=t.begin();i!=t.end();++i) {
   938 	colLowerBound(*i, value);
   939       }
   940     }
   941     template<class T>
   942     typename enable_if<typename T::value_type::second_type::LpSolverCol,
   943 		       void>::type
   944     colLowerBound(T &t, Value value,dummy<1> = 1) { 
   945       for(typename T::iterator i=t.begin();i!=t.end();++i) {
   946 	colLowerBound(i->second, value);
   947       }
   948     }
   949     template<class T>
   950     typename enable_if<typename T::MapIt::Value::LpSolverCol,
   951 		       void>::type
   952     colLowerBound(T &t, Value value,dummy<2> = 2) { 
   953       for(typename T::MapIt i(t); i!=INVALID; ++i){
   954 	colLowerBound(*i, value);
   955       }
   956     }
   957 #endif
   958     
   959     /// Set the upper bound of a column (i.e a variable)
   960 
   961     /// The upper bound of a variable (column) has to be given by an 
   962     /// extended number of type Value, i.e. a finite number of type 
   963     /// Value or \ref INF.
   964     void colUpperBound(Col c, Value value) {
   965       _setColUpperBound(cols.floatingId(c.id),value);
   966     };
   967 
   968     ///\brief Set the lower bound of  several columns
   969     ///(i.e a variables) at once
   970     ///
   971     ///This magic function takes a container as its argument
   972     ///and applies the function on all of its elements.
   973     /// The upper bound of a variable (column) has to be given by an 
   974     /// extended number of type Value, i.e. a finite number of type 
   975     /// Value or \ref INF.
   976 #ifdef DOXYGEN
   977     template<class T>
   978     void colUpperBound(T &t, Value value) { return 0;} 
   979 #else
   980     template<class T>
   981     typename enable_if<typename T::value_type::LpSolverCol,void>::type
   982     colUpperBound(T &t, Value value,dummy<0> = 0) {
   983       for(typename T::iterator i=t.begin();i!=t.end();++i) {
   984 	colUpperBound(*i, value);
   985       }
   986     }
   987     template<class T>
   988     typename enable_if<typename T::value_type::second_type::LpSolverCol,
   989 		       void>::type
   990     colUpperBound(T &t, Value value,dummy<1> = 1) { 
   991       for(typename T::iterator i=t.begin();i!=t.end();++i) {
   992 	colUpperBound(i->second, value);
   993       }
   994     }
   995     template<class T>
   996     typename enable_if<typename T::MapIt::Value::LpSolverCol,
   997 		       void>::type
   998     colUpperBound(T &t, Value value,dummy<2> = 2) { 
   999       for(typename T::MapIt i(t); i!=INVALID; ++i){
  1000 	colUpperBound(*i, value);
  1001       }
  1002     }
  1003 #endif
  1004 
  1005     /// Set the lower and the upper bounds of a column (i.e a variable)
  1006 
  1007     /// The lower and the upper bounds of
  1008     /// a variable (column) have to be given by an 
  1009     /// extended number of type Value, i.e. a finite number of type 
  1010     /// Value, -\ref INF or \ref INF.
  1011     void colBounds(Col c, Value lower, Value upper) {
  1012       _setColLowerBound(cols.floatingId(c.id),lower);
  1013       _setColUpperBound(cols.floatingId(c.id),upper);
  1014     }
  1015     
  1016     ///\brief Set the lower and the upper bound of several columns
  1017     ///(i.e a variables) at once
  1018     ///
  1019     ///This magic function takes a container as its argument
  1020     ///and applies the function on all of its elements.
  1021     /// The lower and the upper bounds of
  1022     /// a variable (column) have to be given by an 
  1023     /// extended number of type Value, i.e. a finite number of type 
  1024     /// Value, -\ref INF or \ref INF.
  1025 #ifdef DOXYGEN
  1026     template<class T>
  1027     void colBounds(T &t, Value lower, Value upper) { return 0;} 
  1028 #else
  1029     template<class T>
  1030     typename enable_if<typename T::value_type::LpSolverCol,void>::type
  1031     colBounds(T &t, Value lower, Value upper,dummy<0> = 0) {
  1032       for(typename T::iterator i=t.begin();i!=t.end();++i) {
  1033 	colBounds(*i, lower, upper);
  1034       }
  1035     }
  1036     template<class T>
  1037     typename enable_if<typename T::value_type::second_type::LpSolverCol,
  1038 		       void>::type
  1039     colBounds(T &t, Value lower, Value upper,dummy<1> = 1) { 
  1040       for(typename T::iterator i=t.begin();i!=t.end();++i) {
  1041 	colBounds(i->second, lower, upper);
  1042       }
  1043     }
  1044     template<class T>
  1045     typename enable_if<typename T::MapIt::Value::LpSolverCol,
  1046 		       void>::type
  1047     colBounds(T &t, Value lower, Value upper,dummy<2> = 2) { 
  1048       for(typename T::MapIt i(t); i!=INVALID; ++i){
  1049 	colBounds(*i, lower, upper);
  1050       }
  1051     }
  1052 #endif
  1053     
  1054 //     /// Set the lower bound of a row (i.e a constraint)
  1055 
  1056 //     /// The lower bound of a linear expression (row) has to be given by an 
  1057 //     /// extended number of type Value, i.e. a finite number of type 
  1058 //     /// Value or -\ref INF.
  1059 //     void rowLowerBound(Row r, Value value) {
  1060 //       _setRowLowerBound(rows.floatingId(r.id),value);
  1061 //     };
  1062 //     /// Set the upper bound of a row (i.e a constraint)
  1063 
  1064 //     /// The upper bound of a linear expression (row) has to be given by an 
  1065 //     /// extended number of type Value, i.e. a finite number of type 
  1066 //     /// Value or \ref INF.
  1067 //     void rowUpperBound(Row r, Value value) {
  1068 //       _setRowUpperBound(rows.floatingId(r.id),value);
  1069 //     };
  1070 
  1071     /// Set the lower and the upper bounds of a row (i.e a constraint)
  1072 
  1073     /// The lower and the upper bounds of
  1074     /// a constraint (row) have to be given by an 
  1075     /// extended number of type Value, i.e. a finite number of type 
  1076     /// Value, -\ref INF or \ref INF.
  1077     void rowBounds(Row c, Value lower, Value upper) {
  1078       _setRowBounds(rows.floatingId(c.id),lower, upper);
  1079       // _setRowUpperBound(rows.floatingId(c.id),upper);
  1080     }
  1081     
  1082     ///Set an element of the objective function
  1083     void objCoeff(Col c, Value v) {_setObjCoeff(cols.floatingId(c.id),v); };
  1084     ///Set the objective function
  1085     
  1086     ///\param e is a linear expression of type \ref Expr.
  1087     ///\bug Is should be called obj()
  1088     void setObj(Expr e) {
  1089       _clearObj();
  1090       for (Expr::iterator i=e.begin(); i!=e.end(); ++i)
  1091 	objCoeff((*i).first,(*i).second);
  1092       obj_const_comp=e.constComp();
  1093     }
  1094 
  1095     ///Maximize
  1096     void max() { _setMax(); }
  1097     ///Minimize
  1098     void min() { _setMin(); }
  1099 
  1100     
  1101     ///@}
  1102 
  1103 
  1104     ///\name Solve the LP
  1105 
  1106     ///@{
  1107 
  1108     ///\e Solve the LP problem at hand
  1109     ///
  1110     ///\return The result of the optimization procedure. Possible 
  1111     ///values and their meanings can be found in the documentation of 
  1112     ///\ref SolveExitStatus.
  1113     ///
  1114     ///\todo Which method is used to solve the problem
  1115     SolveExitStatus solve() { return _solve(); }
  1116     
  1117     ///@}
  1118     
  1119     ///\name Obtain the solution
  1120 
  1121     ///@{
  1122 
  1123     /// The status of the primal problem (the original LP problem)
  1124     SolutionStatus primalStatus() {
  1125       return _getPrimalStatus();
  1126     }
  1127 
  1128     /// The status of the dual (of the original LP) problem 
  1129     SolutionStatus dualStatus() {
  1130       return _getDualStatus();
  1131     }
  1132 
  1133     ///The type of the original LP problem
  1134     ProblemTypes problemType() {
  1135       return _getProblemType();
  1136     }
  1137 
  1138     ///\e
  1139     Value primal(Col c) { return _getPrimal(cols.floatingId(c.id)); }
  1140 
  1141     ///\e
  1142     Value dual(Row r) { return _getDual(rows.floatingId(r.id)); }
  1143 
  1144     ///\e
  1145     bool isBasicCol(Col c) { return _isBasicCol(cols.floatingId(c.id)); }
  1146 
  1147     ///\e
  1148 
  1149     ///\return
  1150     ///- \ref INF or -\ref INF means either infeasibility or unboundedness
  1151     /// of the primal problem, depending on whether we minimize or maximize.
  1152     ///- \ref NaN if no primal solution is found.
  1153     ///- The (finite) objective value if an optimal solution is found.
  1154     Value primalValue() { return _getPrimalValue()+obj_const_comp;}
  1155     ///@}
  1156     
  1157   };  
  1158 
  1159   ///\e
  1160   
  1161   ///\relates LpSolverBase::Expr
  1162   ///
  1163   inline LpSolverBase::Expr operator+(const LpSolverBase::Expr &a,
  1164 				      const LpSolverBase::Expr &b) 
  1165   {
  1166     LpSolverBase::Expr tmp(a);
  1167     tmp+=b;
  1168     return tmp;
  1169   }
  1170   ///\e
  1171   
  1172   ///\relates LpSolverBase::Expr
  1173   ///
  1174   inline LpSolverBase::Expr operator-(const LpSolverBase::Expr &a,
  1175 				      const LpSolverBase::Expr &b) 
  1176   {
  1177     LpSolverBase::Expr tmp(a);
  1178     tmp-=b;
  1179     return tmp;
  1180   }
  1181   ///\e
  1182   
  1183   ///\relates LpSolverBase::Expr
  1184   ///
  1185   inline LpSolverBase::Expr operator*(const LpSolverBase::Expr &a,
  1186 				      const LpSolverBase::Value &b) 
  1187   {
  1188     LpSolverBase::Expr tmp(a);
  1189     tmp*=b;
  1190     return tmp;
  1191   }
  1192   
  1193   ///\e
  1194   
  1195   ///\relates LpSolverBase::Expr
  1196   ///
  1197   inline LpSolverBase::Expr operator*(const LpSolverBase::Value &a,
  1198 				      const LpSolverBase::Expr &b) 
  1199   {
  1200     LpSolverBase::Expr tmp(b);
  1201     tmp*=a;
  1202     return tmp;
  1203   }
  1204   ///\e
  1205   
  1206   ///\relates LpSolverBase::Expr
  1207   ///
  1208   inline LpSolverBase::Expr operator/(const LpSolverBase::Expr &a,
  1209 				      const LpSolverBase::Value &b) 
  1210   {
  1211     LpSolverBase::Expr tmp(a);
  1212     tmp/=b;
  1213     return tmp;
  1214   }
  1215   
  1216   ///\e
  1217   
  1218   ///\relates LpSolverBase::Constr
  1219   ///
  1220   inline LpSolverBase::Constr operator<=(const LpSolverBase::Expr &e,
  1221 					 const LpSolverBase::Expr &f) 
  1222   {
  1223     return LpSolverBase::Constr(-LpSolverBase::INF,e-f,0);
  1224   }
  1225 
  1226   ///\e
  1227   
  1228   ///\relates LpSolverBase::Constr
  1229   ///
  1230   inline LpSolverBase::Constr operator<=(const LpSolverBase::Value &e,
  1231 					 const LpSolverBase::Expr &f) 
  1232   {
  1233     return LpSolverBase::Constr(e,f);
  1234   }
  1235 
  1236   ///\e
  1237   
  1238   ///\relates LpSolverBase::Constr
  1239   ///
  1240   inline LpSolverBase::Constr operator<=(const LpSolverBase::Expr &e,
  1241 					 const LpSolverBase::Value &f) 
  1242   {
  1243     return LpSolverBase::Constr(e,f);
  1244   }
  1245 
  1246   ///\e
  1247   
  1248   ///\relates LpSolverBase::Constr
  1249   ///
  1250   inline LpSolverBase::Constr operator>=(const LpSolverBase::Expr &e,
  1251 					 const LpSolverBase::Expr &f) 
  1252   {
  1253     return LpSolverBase::Constr(-LpSolverBase::INF,f-e,0);
  1254   }
  1255 
  1256 
  1257   ///\e
  1258   
  1259   ///\relates LpSolverBase::Constr
  1260   ///
  1261   inline LpSolverBase::Constr operator>=(const LpSolverBase::Value &e,
  1262 					 const LpSolverBase::Expr &f) 
  1263   {
  1264     return LpSolverBase::Constr(f,e);
  1265   }
  1266 
  1267 
  1268   ///\e
  1269   
  1270   ///\relates LpSolverBase::Constr
  1271   ///
  1272   inline LpSolverBase::Constr operator>=(const LpSolverBase::Expr &e,
  1273 					 const LpSolverBase::Value &f) 
  1274   {
  1275     return LpSolverBase::Constr(f,e);
  1276   }
  1277 
  1278   ///\e
  1279   
  1280   ///\relates LpSolverBase::Constr
  1281   ///
  1282   inline LpSolverBase::Constr operator==(const LpSolverBase::Expr &e,
  1283 					 const LpSolverBase::Expr &f) 
  1284   {
  1285     return LpSolverBase::Constr(0,e-f,0);
  1286   }
  1287 
  1288   ///\e
  1289   
  1290   ///\relates LpSolverBase::Constr
  1291   ///
  1292   inline LpSolverBase::Constr operator<=(const LpSolverBase::Value &n,
  1293 					 const LpSolverBase::Constr&c) 
  1294   {
  1295     LpSolverBase::Constr tmp(c);
  1296     ///\todo Create an own exception type.
  1297     if(!LpSolverBase::isNaN(tmp.lowerBound())) throw LogicError();
  1298     else tmp.lowerBound()=n;
  1299     return tmp;
  1300   }
  1301   ///\e
  1302   
  1303   ///\relates LpSolverBase::Constr
  1304   ///
  1305   inline LpSolverBase::Constr operator<=(const LpSolverBase::Constr& c,
  1306 					 const LpSolverBase::Value &n)
  1307   {
  1308     LpSolverBase::Constr tmp(c);
  1309     ///\todo Create an own exception type.
  1310     if(!LpSolverBase::isNaN(tmp.upperBound())) throw LogicError();
  1311     else tmp.upperBound()=n;
  1312     return tmp;
  1313   }
  1314 
  1315   ///\e
  1316   
  1317   ///\relates LpSolverBase::Constr
  1318   ///
  1319   inline LpSolverBase::Constr operator>=(const LpSolverBase::Value &n,
  1320 					 const LpSolverBase::Constr&c) 
  1321   {
  1322     LpSolverBase::Constr tmp(c);
  1323     ///\todo Create an own exception type.
  1324     if(!LpSolverBase::isNaN(tmp.upperBound())) throw LogicError();
  1325     else tmp.upperBound()=n;
  1326     return tmp;
  1327   }
  1328   ///\e
  1329   
  1330   ///\relates LpSolverBase::Constr
  1331   ///
  1332   inline LpSolverBase::Constr operator>=(const LpSolverBase::Constr& c,
  1333 					 const LpSolverBase::Value &n)
  1334   {
  1335     LpSolverBase::Constr tmp(c);
  1336     ///\todo Create an own exception type.
  1337     if(!LpSolverBase::isNaN(tmp.lowerBound())) throw LogicError();
  1338     else tmp.lowerBound()=n;
  1339     return tmp;
  1340   }
  1341 
  1342   ///\e
  1343   
  1344   ///\relates LpSolverBase::DualExpr
  1345   ///
  1346   inline LpSolverBase::DualExpr operator+(const LpSolverBase::DualExpr &a,
  1347 				      const LpSolverBase::DualExpr &b) 
  1348   {
  1349     LpSolverBase::DualExpr tmp(a);
  1350     tmp+=b;
  1351     return tmp;
  1352   }
  1353   ///\e
  1354   
  1355   ///\relates LpSolverBase::DualExpr
  1356   ///
  1357   inline LpSolverBase::DualExpr operator-(const LpSolverBase::DualExpr &a,
  1358 				      const LpSolverBase::DualExpr &b) 
  1359   {
  1360     LpSolverBase::DualExpr tmp(a);
  1361     tmp-=b;
  1362     return tmp;
  1363   }
  1364   ///\e
  1365   
  1366   ///\relates LpSolverBase::DualExpr
  1367   ///
  1368   inline LpSolverBase::DualExpr operator*(const LpSolverBase::DualExpr &a,
  1369 				      const LpSolverBase::Value &b) 
  1370   {
  1371     LpSolverBase::DualExpr tmp(a);
  1372     tmp*=b;
  1373     return tmp;
  1374   }
  1375   
  1376   ///\e
  1377   
  1378   ///\relates LpSolverBase::DualExpr
  1379   ///
  1380   inline LpSolverBase::DualExpr operator*(const LpSolverBase::Value &a,
  1381 				      const LpSolverBase::DualExpr &b) 
  1382   {
  1383     LpSolverBase::DualExpr tmp(b);
  1384     tmp*=a;
  1385     return tmp;
  1386   }
  1387   ///\e
  1388   
  1389   ///\relates LpSolverBase::DualExpr
  1390   ///
  1391   inline LpSolverBase::DualExpr operator/(const LpSolverBase::DualExpr &a,
  1392 				      const LpSolverBase::Value &b) 
  1393   {
  1394     LpSolverBase::DualExpr tmp(a);
  1395     tmp/=b;
  1396     return tmp;
  1397   }
  1398   
  1399 
  1400 } //namespace lemon
  1401 
  1402 #endif //LEMON_LP_BASE_H