The BCP_lp_user class is the base class from which the user can derive a problem specific class to be used in the LP process. More...

#include <BCP_lp_user.hpp>

Inheritance diagram for BCP_lp_user:

[legend]

Collaboration diagram for BCP_lp_user:

[legend]

Public Member Functions
void	setOsiBabSolver (OsiBabSolver *ptr)

OsiBabSolver *	getOsiBabSolver ()

void	print (const bool ifprint, const char *format,...) const
	A method to print a message with the process id. More...

int	process_id () const
	What is the process id of the current process. More...

int	parent () const
	the process id of the parent More...

void	send_message (const int target, const BCP_buffer &buf, BCP_message_tag tag=BCP_Msg_User)
	Send a message to a particular process. More...

void	receive_message (const int sender, BCP_buffer &buf, BCP_message_tag tag=BCP_Msg_User)
	Wait for a message and receive it. More...

void	broadcast_message (const BCP_process_t proc_type, const BCP_buffer &buf)
	Broadcast the message to all processes of the given type. More...

virtual void	process_message (BCP_buffer &buf)
	Process a message that has been sent by another process' user part to this process' user part. More...

virtual OsiSolverInterface *	initialize_solver_interface ()
	Create LP solver environment. More...

virtual void	initialize_int_and_sos_list (std::vector< OsiObject * > &intAndSosObjects)
	Create the list of objects that can be used for branching (simple integer vars and SOS sets). More...

virtual void	initialize_new_search_tree_node (const BCP_vec< BCP_var * > &vars, const BCP_vec< BCP_cut * > &cuts, const BCP_vec< BCP_obj_status > &var_status, const BCP_vec< BCP_obj_status > &cut_status, BCP_vec< int > &var_changed_pos, BCP_vec< double > &var_new_bd, BCP_vec< int > &cut_changed_pos, BCP_vec< double > &cut_new_bd)
	Initializing a new search tree node. More...

virtual void	load_problem (OsiSolverInterface &osi, BCP_problem_core *core, BCP_var_set &vars, BCP_cut_set &cuts)
	Load the problem specified by core, vars, and cuts into the solver interface. More...

virtual void	modify_lp_parameters (OsiSolverInterface *lp, const int changeType, bool in_strong_branching)
	Modify parameters of the LP solver before optimization. More...

virtual void	process_lp_result (const BCP_lp_result &lpres, const BCP_vec< BCP_var * > &vars, const BCP_vec< BCP_cut * > &cuts, const double old_lower_bound, double &true_lower_bound, BCP_solution &sol, BCP_vec< BCP_cut > &new_cuts, BCP_vec< BCP_row * > &new_rows, BCP_vec< BCP_var * > &new_vars, BCP_vec< BCP_col * > &new_cols)
	Process the result of an iteration. More...

virtual double	compute_lower_bound (const double old_lower_bound, const BCP_lp_result &lpres, const BCP_vec< BCP_var * > &vars, const BCP_vec< BCP_cut * > &cuts)
	Compute a true lower bound for the subproblem. More...

virtual BCP_solution *	generate_heuristic_solution (const BCP_lp_result &lpres, const BCP_vec< BCP_var * > &vars, const BCP_vec< BCP_cut * > &cuts)
	Try to generate a heuristic solution (or return one generated during cut/variable generation. More...

virtual void	restore_feasibility (const BCP_lp_result &lpres, const std::vector< double * > dual_rays, const BCP_vec< BCP_var * > &vars, const BCP_vec< BCP_cut * > &cuts, BCP_vec< BCP_var * > &vars_to_add, BCP_vec< BCP_col * > &cols_to_add)
	Restoring feasibility. More...

virtual void	select_vars_to_delete (const BCP_lp_result &lpres, const BCP_vec< BCP_var * > &vars, const BCP_vec< BCP_cut * > &cuts, const bool before_fathom, BCP_vec< int > &deletable)

virtual void	select_cuts_to_delete (const BCP_lp_result &lpres, const BCP_vec< BCP_var * > &vars, const BCP_vec< BCP_cut * > &cuts, const bool before_fathom, BCP_vec< int > &deletable)

void	reduced_cost_fixing (const double dj, const double x, const double gap, BCP_vec< BCP_var * > &vars, int &newly_changed)
	Reduced cost fixing. More...

virtual BCP_branching_object_relation	compare_branching_candidates (BCP_presolved_lp_brobj new_solved, BCP_presolved_lp_brobj old_solved)
	Decide which branching object is preferred for branching. More...

virtual void	set_actions_for_children (BCP_presolved_lp_brobj *best)
	Decide what to do with the children of the selected branching object. More...

virtual void	set_user_data_for_children (BCP_presolved_lp_brobj *best, const int selected)
	For each child create a user data object and put it into the appropriate entry in `best->user_data()`. More...

virtual void	set_user_data_for_children (BCP_presolved_lp_brobj *best)
	Deprecated version of the previos method (it does not pass the index of the selected branching candidate). More...

Methods to set and get the pointer to the BCP_lp_prob
object. It is unlikely that the users would want to muck around with these (especially with the set method!) but they are here to provide total control.
void	setLpProblemPointer (BCP_lp_prob *ptr)
	Set the pointer. More...

BCP_lp_prob *	getLpProblemPointer ()
	Get the pointer. More...

Informational methods for the user.
double	upper_bound () const
	Return what is the best known upper bound (might be BCP_DBL_MAX) More...

bool	over_ub (double lb) const
	Return true / false depending on whether the lb argument is over the current upper bound or not. More...

int	current_phase () const
	Return the phase the algorithm is in. More...

int	current_level () const
	Return the level of the search tree node being processed. More...

int	current_index () const
	Return the internal index of the search tree node being processed. More...

int	current_iteration () const
	Return the iteration count within the search tree node being processed. More...

double	start_time () const
	Return when the LP process started. More...

BCP_user_data *	get_user_data ()
	Return a pointer to the BCP_user_data structure the user (may have) stored in this node. More...

Methods to get/set BCP parameters on the fly
char	get_param (const BCP_lp_par::chr_params key) const

int	get_param (const BCP_lp_par::int_params key) const

double	get_param (const BCP_lp_par::dbl_params key) const

const BCP_string &	get_param (const BCP_lp_par::str_params key) const

void	set_param (const BCP_lp_par::chr_params key, const bool val)

void	set_param (const BCP_lp_par::chr_params key, const char val)

void	set_param (const BCP_lp_par::int_params key, const int val)

void	set_param (const BCP_lp_par::dbl_params key, const double val)

void	set_param (const BCP_lp_par::str_params key, const char *val)

A methods to send a solution to the Tree Manager. The user can
invoke this method at any time to send off a solution.
void	send_feasible_solution (const BCP_solution *sol)

Constructor, Destructor
	BCP_lp_user ()
	Being virtual, the destructor invokes the destructor for the real type of the object being deleted. More...

virtual	~BCP_lp_user ()
	Being virtual, the destructor invokes the destructor for the real type of the object being deleted. More...

Helper functions for selecting subset of entries from a double
vector. The indices (their position with respect to `first`) of the variables satisfying the criteria are returned in the last argument.
void	select_nonzeros (const double first, const double last, const double etol, BCP_vec< int > &nonzeros) const
	Select all nonzero entries. More...

void	select_zeros (const double first, const double last, const double etol, BCP_vec< int > &zeros) const
	Select all zero entries. More...

void	select_positives (const double first, const double last, const double etol, BCP_vec< int > &positives) const
	Select all positive entries. More...

void	select_fractions (const double first, const double last, const double etol, BCP_vec< int > &fractions) const
	Select all fractional entries. More...

Packing and unpacking methods
virtual void	unpack_module_data (BCP_buffer &buf)
	Unpack the initial information sent to the LP process by the Tree Manager. More...

MIP feasibility testing of LP solutions and heuristics
virtual BCP_solution *	test_feasibility (const BCP_lp_result &lp_result, const BCP_vec< BCP_var * > &vars, const BCP_vec< BCP_cut * > &cuts)
	Evaluate and return MIP feasibility of the current solution. More...


Helper functions for `test_feasibility`. If the solution is feasible a pointer to a BCP_solution_generic object is returned. Note that the solutions generated by these helper functions DO NOT OWN the pointers in the `_vars` member of the solution. Also note that all of these functions assume that the specified integer tolerance in larger than the LP primal tolerance extracted from `lpres` and that the solution in `lpres` do not violate the bounds by more than the LP tolerance.
BCP_solution_generic *	test_binary (const BCP_lp_result &lpres, const BCP_vec< BCP_var * > &vars, const double etol) const
	Test whether all variables are 0/1. More...

BCP_solution_generic *	test_integral (const BCP_lp_result &lpres, const BCP_vec< BCP_var * > &vars, const double etol) const
	Test whether all variables are integer. More...

BCP_solution_generic *	test_full (const BCP_lp_result &lpres, const BCP_vec< BCP_var * > &vars, const double etol) const
	Test whether the variables specified as integers are really integer. More...

Packing of solutions
virtual void	pack_feasible_solution (BCP_buffer &buf, const BCP_solution *sol)
	Pack a MIP feasible solution into a buffer. More...

virtual void	pack_primal_solution (BCP_buffer &buf, const BCP_lp_result &lp_result, const BCP_vec< BCP_var * > &vars, const BCP_vec< BCP_cut * > &cuts)
	Pack the information necessary for cut generation into the buffer. More...

virtual void	pack_dual_solution (BCP_buffer &buf, const BCP_lp_result &lp_result, const BCP_vec< BCP_var * > &vars, const BCP_vec< BCP_cut * > &cuts)
	Pack the information necessary for variable generation into the buffer. More...

Displaying of LP solutions
virtual void	display_lp_solution (const BCP_lp_result &lp_result, const BCP_vec< BCP_var * > &vars, const BCP_vec< BCP_cut * > &cuts, const bool final_lp_solution)
	Display the result of most recent LP optimization. More...

Converting cuts and variables into rows and columns
virtual void	cuts_to_rows (const BCP_vec< BCP_var * > &vars, BCP_vec< BCP_cut * > &cuts, BCP_vec< BCP_row * > &rows, const BCP_lp_result &lpres, BCP_object_origin origin, bool allow_multiple)
	Convert (and possibly lift) a set of cuts into corresponding rows for the current LP relaxation. More...

virtual void	vars_to_cols (const BCP_vec< BCP_cut * > &cuts, BCP_vec< BCP_var * > &vars, BCP_vec< BCP_col * > &cols, const BCP_lp_result &lpres, BCP_object_origin origin, bool allow_multiple)
	Convert a set of variables into corresponding columns for the current LP relaxation. More...

Generating cuts and variables
virtual void	generate_cuts_in_lp (const BCP_lp_result &lpres, const BCP_vec< BCP_var * > &vars, const BCP_vec< BCP_cut * > &cuts, BCP_vec< BCP_cut * > &new_cuts, BCP_vec< BCP_row * > &new_rows)
	Generate cuts within the LP process. More...

virtual void	generate_vars_in_lp (const BCP_lp_result &lpres, const BCP_vec< BCP_var * > &vars, const BCP_vec< BCP_cut * > &cuts, const bool before_fathom, BCP_vec< BCP_var * > &new_vars, BCP_vec< BCP_col * > &new_cols)
	Generate variables within the LP process. More...

virtual BCP_object_compare_result	compare_cuts (const BCP_cut c0, const BCP_cut c1)
	Compare two generated cuts. More...

virtual BCP_object_compare_result	compare_vars (const BCP_var v0, const BCP_var v1)
	Compare two generated variables. More...

Logical fixing
virtual void	logical_fixing (const BCP_lp_result &lpres, const BCP_vec< BCP_var * > &vars, const BCP_vec< BCP_cut * > &cuts, const BCP_vec< BCP_obj_status > &var_status, const BCP_vec< BCP_obj_status > &cut_status, const int var_bound_changes_since_logical_fixing, BCP_vec< int > &changed_pos, BCP_vec< double > &new_bd)
	This method provides an opportunity for the user to tighten the bounds of variables. More...

Branching related methods
virtual BCP_branching_decision	select_branching_candidates (const BCP_lp_result &lpres, const BCP_vec< BCP_var * > &vars, const BCP_vec< BCP_cut * > &cuts, const BCP_lp_var_pool &local_var_pool, const BCP_lp_cut_pool &local_cut_pool, BCP_vec< BCP_lp_branching_object * > &cands, bool force_branch=false)
	Decide whether to branch or not and select a set of branching candidates if branching is decided upon. More...

Helper functions for select_branching_candidates()
virtual int	try_to_branch (OsiBranchingInformation &branchInfo, OsiSolverInterface solver, OsiChooseVariable choose, OsiBranchingObject *&branchObject, bool allowVarFix)
	Select the "close-to-half" variables for strong branching. More...

void	branch_close_to_half (const BCP_lp_result &lpres, const BCP_vec< BCP_var * > &vars, const int to_be_selected, const double etol, BCP_vec< BCP_lp_branching_object * > &candidates)
	Select the "close-to-half" variables for strong branching. More...

void	branch_close_to_one (const BCP_lp_result &lpres, const BCP_vec< BCP_var * > &vars, const int to_be_selected, const double etol, BCP_vec< BCP_lp_branching_object * > &candidates)
	Select the "close-to-one" variables for strong branching. More...

void	append_branching_vars (const double x, const BCP_vec< BCP_var > &vars, const BCP_vec< int > &select_pos, BCP_vec< BCP_lp_branching_object * > &candidates)
	This helper method creates branching variable candidates and appends them to `cans`. More...

Purging the slack pool
virtual void	purge_slack_pool (const BCP_vec< BCP_cut * > &slack_pool, BCP_vec< int > &to_be_purged)
	Selectively purge the list of slack cuts. More...

Public Member Functions inherited from BCP_user_class
virtual	~BCP_user_class ()

Private Member Functions
	BCP_lp_user (const BCP_lp_user &)

BCP_lp_user &	operator= (const BCP_lp_user &)

Private Attributes
bool	using_deprecated_set_user_data_for_children

BCP_lp_prob *	p

OsiBabSolver *	babSolver_

Detailed Description

The BCP_lp_user class is the base class from which the user can derive a problem specific class to be used in the LP process.

In that derived class the user can store data to be used in the methods she overrides. Also that is the object the user must return in the USER_initialize::lp_init() method.

There are two kind of methods in the class. The non-virtual methods are helper functions for the built-in defaults, but the user can use them as well. The virtual methods execute steps in the BCP algorithm where the user might want to override the default behavior.

The default implementations fall into three major categories.

Empty; doesn't do anything and immediately returns (e.g., unpack_module_data()).
There is no reasonable default, so throw an exception. This happens if the parameter settings drive the flow of in a way that BCP can't perform the necessary function. This behavior is correct since such methods are invoked only if the parameter settings drive the flow of the algorithm that way, in which case the user better implement those methods
A default is given. Frequently there are multiple defaults and parameters govern which one is selected (e.g., test_feasibility()).

Definition at line 75 of file BCP_lp_user.hpp.

Constructor & Destructor Documentation

BCP_lp_user::BCP_lp_user ( const BCP_lp_user & )

private

BCP_lp_user::BCP_lp_user ( )

inline

Being virtual, the destructor invokes the destructor for the real type of the object being deleted.

Definition at line 154 of file BCP_lp_user.hpp.

virtual BCP_lp_user::~BCP_lp_user ( )

inlinevirtual

Being virtual, the destructor invokes the destructor for the real type of the object being deleted.

Definition at line 157 of file BCP_lp_user.hpp.

Member Function Documentation

BCP_lp_user& BCP_lp_user::operator= ( const BCP_lp_user & )

private

void BCP_lp_user::setLpProblemPointer ( BCP_lp_prob * ptr )

inline

Set the pointer.

Definition at line 93 of file BCP_lp_user.hpp.

BCP_lp_prob* BCP_lp_user::getLpProblemPointer ( )

inline

Get the pointer.

Definition at line 95 of file BCP_lp_user.hpp.

References p.

void BCP_lp_user::setOsiBabSolver ( OsiBabSolver * ptr )

inline

Definition at line 98 of file BCP_lp_user.hpp.

References babSolver_.

OsiBabSolver* BCP_lp_user::getOsiBabSolver ( )

inline

Definition at line 99 of file BCP_lp_user.hpp.

References babSolver_.

double BCP_lp_user::upper_bound ( ) const

Return what is the best known upper bound (might be BCP_DBL_MAX)

bool BCP_lp_user::over_ub ( double lb ) const

Return true / false depending on whether the lb argument is over the current upper bound or not.

int BCP_lp_user::current_phase ( ) const

Return the phase the algorithm is in.

int BCP_lp_user::current_level ( ) const

Return the level of the search tree node being processed.

int BCP_lp_user::current_index ( ) const

Return the internal index of the search tree node being processed.

int BCP_lp_user::current_iteration ( ) const

Return the iteration count within the search tree node being processed.

double BCP_lp_user::start_time ( ) const

Return when the LP process started.

BCP_user_data* BCP_lp_user::get_user_data ( )

Return a pointer to the BCP_user_data structure the user (may have) stored in this node.

void BCP_lp_user::print	(	const bool	ifprint,
		const char *	format,
			...
	)		const

A method to print a message with the process id.

char BCP_lp_user::get_param ( const BCP_lp_par::chr_params key ) const

int BCP_lp_user::get_param ( const BCP_lp_par::int_params key ) const

double BCP_lp_user::get_param ( const BCP_lp_par::dbl_params key ) const

const BCP_string& BCP_lp_user::get_param ( const BCP_lp_par::str_params key ) const

void BCP_lp_user::set_param	(	const BCP_lp_par::chr_params	key,
		const bool	val
	)

void BCP_lp_user::set_param	(	const BCP_lp_par::chr_params	key,
		const char	val
	)

void BCP_lp_user::set_param	(	const BCP_lp_par::int_params	key,
		const int	val
	)

void BCP_lp_user::set_param	(	const BCP_lp_par::dbl_params	key,
		const double	val
	)

void BCP_lp_user::set_param	(	const BCP_lp_par::str_params	key,
		const char *	val
	)

void BCP_lp_user::send_feasible_solution ( const BCP_solution * sol )

void BCP_lp_user::select_nonzeros	(	const double *	first,
		const double *	last,
		const double	etol,
		BCP_vec< int > &	nonzeros
	)		const

Select all nonzero entries.

Those are considered nonzero that have absolute value greater than etol.

void BCP_lp_user::select_zeros	(	const double *	first,
		const double *	last,
		const double	etol,
		BCP_vec< int > &	zeros
	)		const

Select all zero entries.

Those are considered zero that have absolute value less than etol.

void BCP_lp_user::select_positives	(	const double *	first,
		const double *	last,
		const double	etol,
		BCP_vec< int > &	positives
	)		const

Select all positive entries.

Those are considered positive that have value greater than etol.

void BCP_lp_user::select_fractions	(	const double *	first,
		const double *	last,
		const double	etol,
		BCP_vec< int > &	fractions
	)		const

Select all fractional entries.

Those are considered fractional that are further than etol away from any integer value.

virtual void BCP_lp_user::unpack_module_data ( BCP_buffer & buf )

virtual

Unpack the initial information sent to the LP process by the Tree Manager.

This information was packed by the method BCP_tm_user::pack_module_data() invoked with BCP_ProcessType_LP as the third (target process type) argument.

Default: empty method.

Reimplemented in CSP_lp, BB_lp, MC_lp, MC_lp, MKC_lp, MCF3_lp, MCF1_lp, and MCF2_lp.

int BCP_lp_user::process_id ( ) const

What is the process id of the current process.

int BCP_lp_user::parent ( ) const

the process id of the parent

void BCP_lp_user::send_message	(	const int	target,
		const BCP_buffer &	buf,
		BCP_message_tag	tag = `BCP_Msg_User`
	)

Send a message to a particular process.

void BCP_lp_user::receive_message	(	const int	sender,
		BCP_buffer &	buf,
		BCP_message_tag	tag = `BCP_Msg_User`
	)

Wait for a message and receive it.

void BCP_lp_user::broadcast_message	(	const BCP_process_t	proc_type,
		const BCP_buffer &	buf
	)

Broadcast the message to all processes of the given type.

virtual void BCP_lp_user::process_message ( BCP_buffer & buf )

virtual

Process a message that has been sent by another process' user part to this process' user part.

virtual OsiSolverInterface* BCP_lp_user::initialize_solver_interface ( )

virtual

Create LP solver environment.

Create the LP solver class that will be used for solving the LP relaxations. The default implementation picks up which COIN_USE_XXX is defined and initializes an lp solver of that type. This is probably OK for most users. The only reason to override this method is to be able to choose at runtime which lp solver to instantiate (maybe even different solvers on different processors). In this case she should probably also override the pack_warmstart() and unpack_warmstart() methods in this class and in the BCP_tm_user class.

Reimplemented in CSP_lp, MC_lp, BB_lp, MC_lp, MKC_lp, MCF3_lp, MCF1_lp, and MCF2_lp.

virtual void BCP_lp_user::initialize_int_and_sos_list ( std::vector< OsiObject * > & intAndSosObjects )

virtual

Create the list of objects that can be used for branching (simple integer vars and SOS sets).

If nothing is done here then for each search tree node (just before starting to process the node) BCP will scan the variables and the matrix for candidates.

virtual void BCP_lp_user::initialize_new_search_tree_node	(	const BCP_vec< BCP_var * > &	vars,
		const BCP_vec< BCP_cut * > &	cuts,
		const BCP_vec< BCP_obj_status > &	var_status,
		const BCP_vec< BCP_obj_status > &	cut_status,
		BCP_vec< int > &	var_changed_pos,
		BCP_vec< double > &	var_new_bd,
		BCP_vec< int > &	cut_changed_pos,
		BCP_vec< double > &	cut_new_bd
	)

virtual

Initializing a new search tree node.

This method serves as hook for the user to do some preprocessing on a search tree node before the node is processed. Also, logical fixing results can be returned in the last four parameters. This might be very useful if the branching implies significant tightening.
Default: empty method.

Parameters

vars	(IN) The variables in the current formulation
cuts	(IN) The cuts in the current formulation
var_status	(IN) The stati of the variables
cut_status	(IN) The stati of the cuts
var_changed_pos	(OUT) The positions of the variables whose bounds should be tightened
var_new_bd	(OUT) The new lb/ub of those variables
cut_changed_pos	(OUT) The positions of the cuts whose bounds should be tightened
cut_new_bd	(OUT) The new lb/ub of those cuts

Reimplemented in CSP_lp, MKC_lp, BB_lp, MCF3_lp, MCF1_lp, and MCF2_lp.

virtual void BCP_lp_user::load_problem	(	OsiSolverInterface &	osi,
		BCP_problem_core *	core,
		BCP_var_set &	vars,
		BCP_cut_set &	cuts
	)

virtual

Load the problem specified by core, vars, and cuts into the solver interface.

If the solver is an LP solver then the default is fine. If it's an NLP then the user has to do this herself.

virtual void BCP_lp_user::modify_lp_parameters	(	OsiSolverInterface *	lp,
		const int	changeType,
		bool	in_strong_branching
	)

virtual

Modify parameters of the LP solver before optimization.

This method provides an opportunity for the user to change parameters of the LP solver before optimization in the LP solver starts. The second argument indicates what has changed in the LP before this method is called. 0: no change; 1: changes that affect primal feasibility (change in column/row bounds, added cuts); 2: changes that affect dual feasibility (added columns); 3: both. The last argument indicates whether the optimization is a "regular" optimization or it will take place in strong branching.

Default: If 1 or 2 then the appropriate simplex method will be hinted to the solver.

Reimplemented in BB_lp, and MC_lp.

virtual void BCP_lp_user::process_lp_result	(	const BCP_lp_result &	lpres,
		const BCP_vec< BCP_var * > &	vars,
		const BCP_vec< BCP_cut * > &	cuts,
		const double	old_lower_bound,
		double &	true_lower_bound,
		BCP_solution *&	sol,
		BCP_vec< BCP_cut * > &	new_cuts,
		BCP_vec< BCP_row * > &	new_rows,
		BCP_vec< BCP_var * > &	new_vars,
		BCP_vec< BCP_col * > &	new_cols
	)

virtual

Process the result of an iteration.

This includes:

computing a true lower bound on the subproblem.
In case column generation is done the lower bound for the subproblem might not be the same as the objective value of the current LP relaxation. Here the user has an option to return a true lower bound.
test feasibility of the solution (or generate a heuristic solution)
generating cuts and/or variables

The reason for the existence of this method is that (especially when column generation is done) these tasks are so intertwined that it is much easier to execute them in one method instead of in several separate methods.

The default behavior is to do nothing and invoke the individual methods one-by-one.

Parameters

lp_result	the result of the most recent LP optimization (IN)
vars	variables currently in the formulation (IN)
cuts	variables currently in the formulation (IN)
old_lower_bound	the previously known best lower bound (IN)
new_cuts	the vector of generated cuts (OUT)
new_rows	the correspontding rows(OUT)
new_vars	the vector of generated variables (OUT)
new_cols	the correspontding columns(OUT)

virtual double BCP_lp_user::compute_lower_bound	(	const double	old_lower_bound,
		const BCP_lp_result &	lpres,
		const BCP_vec< BCP_var * > &	vars,
		const BCP_vec< BCP_cut * > &	cuts
	)

virtual

Compute a true lower bound for the subproblem.

In case column generation is done the lower bound for the subproblem might not be the same as the objective value of the current LP relaxation. Here the user has an option to return a true lower bound.
The default implementation returns the objective value of the current LP relaxation if no column generation is done, otherwise returns the current (somehow previously computed) true lower bound.

Reimplemented in CSP_lp, MKC_lp, MCF3_lp, MCF1_lp, and MCF2_lp.

virtual BCP_solution* BCP_lp_user::test_feasibility	(	const BCP_lp_result &	lp_result,
		const BCP_vec< BCP_var * > &	vars,
		const BCP_vec< BCP_cut * > &	cuts
	)

virtual

Evaluate and return MIP feasibility of the current solution.

If the solution is MIP feasible, return a solution object otherwise return a NULL pointer. The useris also welcome to heuristically generate a solution and return a pointer to that solution (although the user will have another chance (after cuts and variables are generated) to return/create heuristically generated solutions. (After all, it's quite possible that solutions are generated during cut/variable generation.)

Default: test feasibility based on the FeeasibilityTest parameter in BCP_lp_par which defults to BCP_FullTest_Feasible.

Parameters

lp_result	the result of the most recent LP optimization
vars	variables currently in the formulation
cuts	variables currently in the formulation

Reimplemented in MKC_lp, BB_lp, MC_lp, MC_lp, MCF3_lp, MCF1_lp, and MCF2_lp.

BCP_solution_generic* BCP_lp_user::test_binary	(	const BCP_lp_result &	lpres,
		const BCP_vec< BCP_var * > &	vars,
		const double	etol
	)		const

Test whether all variables are 0/1.

Note that this method assumes that all variables are binary, i.e., their original lower/upper bounds are 0/1.

BCP_solution_generic* BCP_lp_user::test_integral	(	const BCP_lp_result &	lpres,
		const BCP_vec< BCP_var * > &	vars,
		const double	etol
	)		const

Test whether all variables are integer.

Note that this method assumes that all variables are integer.

BCP_solution_generic* BCP_lp_user::test_full	(	const BCP_lp_result &	lpres,
		const BCP_vec< BCP_var * > &	vars,
		const double	etol
	)		const

Test whether the variables specified as integers are really integer.

virtual BCP_solution* BCP_lp_user::generate_heuristic_solution	(	const BCP_lp_result &	lpres,
		const BCP_vec< BCP_var * > &	vars,
		const BCP_vec< BCP_cut * > &	cuts
	)

virtual

Try to generate a heuristic solution (or return one generated during cut/variable generation.

Return a pointer to the generated solution or return a NULL pointer.

Default: Return a NULL pointer

Reimplemented in CSP_lp, BB_lp, MC_lp, and MC_lp.

virtual void BCP_lp_user::pack_feasible_solution	(	BCP_buffer &	buf,
		const BCP_solution *	sol
	)

virtual

Pack a MIP feasible solution into a buffer.

The solution will be unpacked in the Tree Manager by the BCP_tm_user::unpack_feasible_solution() method.

Default: The default implementation assumes that sol is a BCP_solution_generic object (containing variables at nonzero level) and packs it.

Parameters

buf	(OUT) the buffer to pack into
sol	(IN) the solution to be packed

Reimplemented in MKC_lp, MC_lp, and MC_lp.

virtual void BCP_lp_user::pack_primal_solution	(	BCP_buffer &	buf,
		const BCP_lp_result &	lp_result,
		const BCP_vec< BCP_var * > &	vars,
		const BCP_vec< BCP_cut * > &	cuts
	)

virtual

Pack the information necessary for cut generation into the buffer.

Note that the name of the method is pack_primal_solution because most likely that (or some part of that) will be needed for cut generation. However, if the user overrides the method she is free to pack anything (of course she'll have to unpack it in CG).

This information will be sent to the Cut Generator (and possibly to the Cut Pool) where the user has to unpack it. If the user uses the built-in method here, then the built-in method will be used in the Cut Generator as well.

Default: The content of the message depends on the value of the PrimalSolForCG parameter in BCP_lp_par. By default the variables at nonzero level are packed.

Parameters

buf	(OUT) the buffer to pack into
lp_result	(IN) the result of the most recent LP optimization
vars	(IN) variables currently in the formulation
cuts	(IN) cuts currently in the formulation

virtual void BCP_lp_user::pack_dual_solution	(	BCP_buffer &	buf,
		const BCP_lp_result &	lp_result,
		const BCP_vec< BCP_var * > &	vars,
		const BCP_vec< BCP_cut * > &	cuts
	)

virtual

Pack the information necessary for variable generation into the buffer.

Note that the name of the method is pack_dual_solution because most likely that (or some part of that) will be needed for variable generation. However, if the user overrides the method she is free to pack anything (of course she'll have to unpack it in CG).

This information will be sent to the Variable Generator (and possibly to the Variable Pool) where the user has to unpack it. If the user uses the built-in method here, then the built-in method will be used in the Variable Generator as well.

Default: The content of the message depends on the value of the DualSolForVG parameter in BCP_lp_par. By default the full dual solution is packed.

Parameters

buf	(OUT) the buffer to pack into
lp_result	(IN) the result of the most recent LP optimization
vars	(IN) variables currently in the formulation
cuts	(IN) cuts currently in the formulation

virtual void BCP_lp_user::display_lp_solution	(	const BCP_lp_result &	lp_result,
		const BCP_vec< BCP_var * > &	vars,
		const BCP_vec< BCP_cut * > &	cuts,
		const bool	final_lp_solution
	)

virtual

Display the result of most recent LP optimization.

This method is invoked every time an LP relaxation is optimized and the user can display (or not display) it.

Note that this method is invoked only if final_lp_solution is true (i.e., no cuts/variables were found) and the LpVerb_FinalRelaxedSolution parameter of BCP_lp_par is set to true (or alternatively, final_lp_solution is false and LpVerb_RelaxedSolution is true).

Default: display the solution if the appropriate verbosity code entry is set.

Parameters

lp_result	(IN) the result of the most recent LP optimization
vars	(IN) variables currently in the formulation
final_lp_solution	(IN) whether the lp solution is final or not.

virtual void BCP_lp_user::restore_feasibility	(	const BCP_lp_result &	lpres,
		const std::vector< double * >	dual_rays,
		const BCP_vec< BCP_var * > &	vars,
		const BCP_vec< BCP_cut * > &	cuts,
		BCP_vec< BCP_var * > &	vars_to_add,
		BCP_vec< BCP_col * > &	cols_to_add
	)

virtual

Restoring feasibility.

This method is invoked before fathoming a search tree node that has been found infeasible and the variable pricing did not generate any new variables.

Reimplemented in CSP_lp.

virtual void BCP_lp_user::cuts_to_rows	(	const BCP_vec< BCP_var * > &	vars,
		BCP_vec< BCP_cut * > &	cuts,
		BCP_vec< BCP_row * > &	rows,
		const BCP_lp_result &	lpres,
		BCP_object_origin	origin,
		bool	allow_multiple
	)

virtual

Convert (and possibly lift) a set of cuts into corresponding rows for the current LP relaxation.

Converting means computing for each cut the coefficients corresponding to each variable and creating BCP_row objects that can be added to the formulation.

This method has different purposes depending on the value of the last argument. If multiple expansion is not allowed then the user must generate a unique row for each cut. This unique row must always be the same for any given cut. This kind of operation is needed so that an LP relaxation can be exactly recreated.

On the other hand if multiple expansion is allowed then the user has (almost) free reign over what she returns. She can delete some of the cuts or append new ones (e.g., lifted ones) to the end. The result of the LP relaxation and the origin of the cuts are there to help her to make a decision about what to do. For example, she might want to lift cuts coming from the Cut Generator, but not those coming from the Cut Pool. The only requirement is that when this method returns the number of cuts and rows must be the same and the i-th row must be the unique row corresponding to the i-th cut.

Parameters

vars	the variables currently in the relaxation (IN)
cuts	the cuts to be converted (IN/OUT)
rows	the rows into which the cuts are converted (OUT)
lpres	solution to the current LP relaxation (IN)
origin	where the cuts come from (IN)
allow_multiple	whether multiple expansion, i.e., lifting, is allowed (IN)

Default: throw an exception (if this method is invoked then the user must have generated cuts and BCP has no way to know how to convert them).

Reimplemented in BB_lp, MC_lp, and MC_lp.

virtual void BCP_lp_user::vars_to_cols	(	const BCP_vec< BCP_cut * > &	cuts,
		BCP_vec< BCP_var * > &	vars,
		BCP_vec< BCP_col * > &	cols,
		const BCP_lp_result &	lpres,
		BCP_object_origin	origin,
		bool	allow_multiple
	)

virtual

Convert a set of variables into corresponding columns for the current LP relaxation.

Converting means to compute for each variable the coefficients corresponding to each cut and create BCP_col objects that can be added to the formulation.

See the documentation of cuts_to_rows() above for the use of this method (just reverse the role of cuts and variables.)

Parameters

cuts	the cuts currently in the relaxation (IN)
vars	the variables to be converted (IN/OUT)
cols	the colums the variables convert into (OUT)
lpres	solution to the current LP relaxation (IN)
origin	where the do the cuts come from (IN)
allow_multiple	whether multiple expansion, i.e., lifting, is allowed (IN)

Default: throw an exception (if this method is invoked then the user must have generated variables and BCP has no way to know how to convert them).

Reimplemented in CSP_lp, MKC_lp, MCF3_lp, MCF1_lp, and MCF2_lp.

virtual void BCP_lp_user::generate_cuts_in_lp	(	const BCP_lp_result &	lpres,
		const BCP_vec< BCP_var * > &	vars,
		const BCP_vec< BCP_cut * > &	cuts,
		BCP_vec< BCP_cut * > &	new_cuts,
		BCP_vec< BCP_row * > &	new_rows
	)

virtual

Generate cuts within the LP process.

Sometimes too much information would need to be transmitted for cut generation (e.g., the full tableau for Gomory cuts) or the cut generation is so fast that transmitting the info would take longer than generating the cuts. In such cases it might better to generate the cuts locally. This routine provides the opportunity.
Default: empty for now. To be interfaced to Cgl.

Parameters

lpres	solution to the current LP relaxation (IN)
vars	the variabless currently in the relaxation (IN)
cuts	the cuts currently in the relaxation (IN)
new_cuts	the vector of generated cuts (OUT)
new_rows	the correspontding rows(OUT)

Reimplemented in CSP_lp, MC_lp, BB_lp, and MC_lp.

Referenced by CSP_lp::generate_cuts_in_lp().

virtual void BCP_lp_user::generate_vars_in_lp	(	const BCP_lp_result &	lpres,
		const BCP_vec< BCP_var * > &	vars,
		const BCP_vec< BCP_cut * > &	cuts,
		const bool	before_fathom,
		BCP_vec< BCP_var * > &	new_vars,
		BCP_vec< BCP_col * > &	new_cols
	)

virtual

Generate variables within the LP process.

Sometimes too much information would need to be transmitted for variable generation or the variable generation is so fast that transmitting the info would take longer than generating the variables. In such cases it might be better to generate the variables locally. This routine provides the opportunity.

Default: empty method.

Parameters

lpres	solution to the current LP relaxation (IN)
vars	the variabless currently in the relaxation (IN)
cuts	the cuts currently in the relaxation (IN)
before_fathom	if true then BCP is about to fathom the node, so spend some extra effort generating variables if you want to avoid that...
new_vars	the vector of generated variables (OUT)
new_cols	the correspontding columns(OUT)

Reimplemented in CSP_lp, MKC_lp, MCF3_lp, MCF1_lp, and MCF2_lp.

virtual BCP_object_compare_result BCP_lp_user::compare_cuts	(	const BCP_cut *	c0,
		const BCP_cut *	c1
	)

virtual

Compare two generated cuts.

Cuts are generated in different iterations, they come from the Cut Pool, etc. There is a very real possibility that the LP process receives several cuts that are either identical or one of them is better then another (cuts off everything the other cuts off). This routine is used to decide which one to keep if not both.
Default: Return BCP_DifferentObjs.

Reimplemented in CSP_lp, MC_lp, and MC_lp.

Referenced by CSP_lp::compare_cuts().

virtual BCP_object_compare_result BCP_lp_user::compare_vars	(	const BCP_var *	v0,
		const BCP_var *	v1
	)

virtual

Compare two generated variables.

Variables are generated in different iterations, they come from the Variable Pool, etc. There is a very real possibility that the LP process receives several variables that are either identical or one of them is better then another (e.g., almost identical but has much lower reduced cost). This routine is used to decide which one to keep if not both.
Default: Return BCP_DifferentObjs.

Reimplemented in CSP_lp.

Referenced by CSP_lp::compare_vars().

virtual void BCP_lp_user::select_vars_to_delete	(	const BCP_lp_result &	lpres,
		const BCP_vec< BCP_var * > &	vars,
		const BCP_vec< BCP_cut * > &	cuts,
		const bool	before_fathom,
		BCP_vec< int > &	deletable
	)

virtual

virtual void BCP_lp_user::select_cuts_to_delete	(	const BCP_lp_result &	lpres,
		const BCP_vec< BCP_var * > &	vars,
		const BCP_vec< BCP_cut * > &	cuts,
		const bool	before_fathom,
		BCP_vec< int > &	deletable
	)

virtual

virtual void BCP_lp_user::logical_fixing	(	const BCP_lp_result &	lpres,
		const BCP_vec< BCP_var * > &	vars,
		const BCP_vec< BCP_cut * > &	cuts,
		const BCP_vec< BCP_obj_status > &	var_status,
		const BCP_vec< BCP_obj_status > &	cut_status,
		const int	var_bound_changes_since_logical_fixing,
		BCP_vec< int > &	changed_pos,
		BCP_vec< double > &	new_bd
	)

virtual

This method provides an opportunity for the user to tighten the bounds of variables.

The method is invoked after reduced cost fixing. The results are returned in the last two parameters.
Default: empty method.

Parameters

lpres	the result of the most recent LP optimization,
vars	the variables in the current formulation,
status	the stati of the variables as known to the system,
var_bound_changes_since_logical_fixing	the number of variables whose bounds have changed (by reduced cost fixing) since the most recent invocation of this method that has actually forced changes returned something in the last two arguments,
changed_pos	the positions of the variables whose bounds should be changed
new_bd	the new bounds (lb/ub pairs) of these variables.

Reimplemented in CSP_lp, MC_lp, MKC_lp, MC_lp, and BB_lp.

Referenced by CSP_lp::logical_fixing().

void BCP_lp_user::reduced_cost_fixing	(	const double *	dj,
		const double *	x,
		const double	gap,
		BCP_vec< BCP_var * > &	vars,
		int &	newly_changed
	)

Reduced cost fixing.

This is not exactly a helper function, but the user might want to invoke it...

virtual BCP_branching_decision BCP_lp_user::select_branching_candidates	(	const BCP_lp_result &	lpres,
		const BCP_vec< BCP_var * > &	vars,
		const BCP_vec< BCP_cut * > &	cuts,
		const BCP_lp_var_pool &	local_var_pool,
		const BCP_lp_cut_pool &	local_cut_pool,
		BCP_vec< BCP_lp_branching_object * > &	cands,
		bool	force_branch = `false`
	)

virtual

Decide whether to branch or not and select a set of branching candidates if branching is decided upon.

The return value indicates what should be done: branching, continuing with the same node or abandoning the node completely.

Default: Branch if both local pools are empty. If branching is done then several (based on the StrongBranch_CloseToHalfNum and StrongBranch_CloseToOneNum parameters in BCP_lp_par) variables are selected for strong branching.

"Close-to-half" variables are those that should be integer and are at a fractional level. The measure of their fractionality is their distance from the closest integer. The most fractional variables will be selected, i.e., those that are close to half. If there are too many such variables then those with higher objective value have priority.

"Close-to-on" is interpreted in a more literal sense. It should be used only if the integer variables are binary as it select those fractional variables which are away from 1 but are still close. If there are too many such variables then those with lower objective value have priority.

Parameters

lpres	the result of the most recent LP optimization.
vars	the variables in the current formulation.
cuts	the cuts in the current formulation.
local_var_pool	the local pool that holds variables with negative reduced cost. In case of continuing with the node the best so many variables will be added to the formulation (those with the most negative reduced cost).
local_cut_pool	the local pool that holds violated cuts. In case of continuing with the node the best so many cuts will be added to the formulation (the most violated ones).
cands	the generated branching candidates.
force_branch	indicate whether to force branching regardless of the size of the local cut/var pools

Reimplemented in MC_lp, BB_lp, MCF3_lp, MCF1_lp, and MCF2_lp.

virtual int BCP_lp_user::try_to_branch	(	OsiBranchingInformation &	branchInfo,
		OsiSolverInterface *	solver,
		OsiChooseVariable *	choose,
		OsiBranchingObject *&	branchObject,
		bool	allowVarFix
	)

virtual

Select the "close-to-half" variables for strong branching.

Variables that are at least etol away from integrality are considered and to_be_selected of them will be picked up.

void BCP_lp_user::branch_close_to_half	(	const BCP_lp_result &	lpres,
		const BCP_vec< BCP_var * > &	vars,
		const int	to_be_selected,
		const double	etol,
		BCP_vec< BCP_lp_branching_object * > &	candidates
	)

Select the "close-to-half" variables for strong branching.

Variables that are at least etol away from integrality are considered and to_be_selected of them will be picked up.

void BCP_lp_user::branch_close_to_one	(	const BCP_lp_result &	lpres,
		const BCP_vec< BCP_var * > &	vars,
		const int	to_be_selected,
		const double	etol,
		BCP_vec< BCP_lp_branching_object * > &	candidates
	)

Select the "close-to-one" variables for strong branching.

Variables that are at least etol away from integrality are considered and to_be_selected of them will be picked up.

void BCP_lp_user::append_branching_vars	(	const double *	x,
		const BCP_vec< BCP_var * > &	vars,
		const BCP_vec< int > &	select_pos,
		BCP_vec< BCP_lp_branching_object * > &	candidates
	)

This helper method creates branching variable candidates and appends them to cans.

The indices (in the current formulation) of the variables from which candidates should be created are listed in select_pos.

virtual BCP_branching_object_relation BCP_lp_user::compare_branching_candidates	(	BCP_presolved_lp_brobj *	new_solved,
		BCP_presolved_lp_brobj *	old_solved
	)

virtual

Decide which branching object is preferred for branching.

Based on the member fields of the two presolved candidate branching objects decide which one should be preferred for really branching on it. Possible return values are: BCP_OldPresolvedIsBetter, BCP_NewPresolvedIsBetter and BCP_NewPresolvedIsBetter_BranchOnIt. This last value (besides specifying which candidate is preferred) also indicates that no further candidates should be examined, branching should be done on this candidate.

Default: The behavior of this method is governed by the BranchingObjectComparison parameter in BCP_lp_par.

Reimplemented in CSP_lp, MC_lp, and MC_lp.

Referenced by CSP_lp::compare_branching_candidates().

virtual void BCP_lp_user::set_actions_for_children ( BCP_presolved_lp_brobj * best )

virtual

Decide what to do with the children of the selected branching object.

Fill out the _child_action field in best. This will specify for every child what to do with it. Possible values for each individual child are BCP_FathomChild, BCP_ReturnChild and BCP_KeepChild. There can be at most child with this last action specified. It means that in case of diving this child will be processed by this LP process as the next search tree node.

Default: Every action is BCP_ReturnChild. However, if BCP dives then one child will be mark with BCP_KeepChild. The decision which child to keep is based on the ChildPreference parameter in BCP_lp_par. Also, if a child has a presolved lower bound that is higher than the current upper bound then that child is mark as BCP_FathomChild.

THINK*: Should those children be sent back for processing in the next phase?

Reimplemented in CSP_lp, MC_lp, and MC_lp.

virtual void BCP_lp_user::set_user_data_for_children	(	BCP_presolved_lp_brobj *	best,
		const int	selected
	)

virtual

For each child create a user data object and put it into the appropriate entry in best->user_data().

When this function is called the best->user_data() vector is already the right size and is filled will 0 pointers. The second argument is usefule if strong branching was done. It is the index of the branching candidate that was selected for branching (the one that's the source of best.

Reimplemented in BB_lp, and MCF3_lp.

virtual void BCP_lp_user::set_user_data_for_children ( BCP_presolved_lp_brobj * best )

virtual

Deprecated version of the previos method (it does not pass the index of the selected branching candidate).

virtual void BCP_lp_user::purge_slack_pool	(	const BCP_vec< BCP_cut * > &	slack_pool,
		BCP_vec< int > &	to_be_purged
	)

virtual

Selectively purge the list of slack cuts.

When a cut becomes ineffective and is eventually purged from the LP formulation it is moved into slack_pool. The user might consider cuts might later for branching. This function enables the user to purge any cut from the slack pool (those she wouldn't consider anyway). Of course, the user is not restricted to these cuts when branching, this is only there to help to collect slack cuts. The user should put the indices of the cuts to be purged into the provided vector.

Default: Purges the slack cut pool according to the SlackCutDiscardingStrategy rule in BCP_lp_par (purge everything before every iteration or before a new search tree node).

Parameters