javax.constraints.impl.search.goal.SolverWithGoals Maven / Gradle / Ivy
Go to download
Show more of this group Show more artifacts with this name
Show all versions of jsr331 Show documentation
Show all versions of jsr331 Show documentation
JCP Standard JSR331 “Java Constraint Programming API”. It is used for Modeling and Solving Constraint Satisfaction and Optimization Problems using Java and off-the-shelf Constraint/Linear Solvers
The newest version!
package javax.constraints.impl.search.goal;
import java.util.Vector;
import javax.constraints.Constraint;
import javax.constraints.Objective;
import javax.constraints.Problem;
import javax.constraints.ProblemState;
import javax.constraints.SearchStrategy;
import javax.constraints.SearchStrategy.SearchStrategyType;
import javax.constraints.Solution;
import javax.constraints.Var;
import javax.constraints.VarSet;
import javax.constraints.extra.PropagationEvent;
import javax.constraints.impl.BasicVarSet;
import javax.constraints.impl.search.AbstractSolver;
import javax.constraints.impl.search.goal.Dichotomize;
abstract public class SolverWithGoals extends AbstractSolver {
public SolverWithGoals(Problem problem) {
super(problem);
// if (this.getProblem().getVars() != null ) {
// setSearchStrategy(getSearchStrategy(SearchStrategyType.DEFAULT));
//setSearchStrategy(newSearchStrategy());
// }
// else {
// problem.log("There are no decision variables defined in the problem");
// }
}
/**
* This methods returns a new default search strategy
* @return a new default search strategy
*/
public SearchStrategy newSearchStrategy() {
return new GoalAssignValues(this);
// if (getTimeLimit() <= 0)
// return new GoalAssignValues(this);
// else
// return new GoalAssignValuesTimeLimit(this);
}
// /**
// * Returns a search strategy of the given type.
// * Each Solver' implementation supports their own implementations of search strategies
// * for the standard types.
// * @param strategyType
// */
// public SearchStrategy getSearchStrategy(SearchStrategyType strategyType) {
// switch (strategyType) {
// case DEFAULT:
// return new GoalAssignValues(this);
//
// default:
// return new GoalAssignValues(this);
// }
// }
// public Goal getGoal() {
// SearchStrategyGoal searchStrategy = (SearchStrategyGoal)getSearchStrategy();
// return searchStrategy.getGoal();
// }
// public void setSearchStrategy(Goal goal) {
// SearchStrategy strategy = new SearchStrategyGoal(goal);
// setSearchStrategy(strategy);
// }
// /**
// * This methods adds this strategy
// * to the end of the strategy execution lists.
// * @param strategy
// * @see {@link AbstractSolver#findSolution()} {@link AbstractSolver#findOptimalSolution(Var)}
// */
// public void addSearchStrategy(SearchStrategy strategy) {
// searchStrategies.add(strategy);
// Goal prevGoal = (Goal)getSearchStrategy();
// Goal goal = (Goal) strategy;
// setSearchStrategy(prevGoal.and(goal));
// }
/**
* To create a new goal on the javax.constraints level without necessity to go down to
* the underlying CP solver, one needs to use a goal that will serve as an
* implementation object. "goalThis" serves exactly these purposes and
* actually executes the method execute() from the Goal passed as a
* parameter.
*
* @param goal
*/
abstract public Object goalThis(Goal goal);
abstract public boolean execute(Goal goal, ProblemState restoreOrNot);
public Goal goalVarEqValue(Var var,int value) {
Constraint assign = getProblem().linear(var, "=", value);
Goal goalAssignValue = new GoalConstraint(assign);
goalAssignValue.setName("Assign " + value + " to " + var);
return goalAssignValue;
}
public Goal goalVarNeqValue(Var var,int value) {
Constraint remove = getProblem().linear(var, "!=", value);
Goal goalRemoveValue = new GoalConstraint(remove);
goalRemoveValue.setName("Remove " + value + " from " + var);
return goalRemoveValue;
}
public Goal goalVarLeValue(Var var,int value) {
Constraint c = getProblem().linear(var, "<=", value);
c.setName("" + var + " <= "+ value);
Goal goalLeValue = new GoalConstraint(c);
return goalLeValue;
}
public Goal goalVarGeValue(Var var,int value) {
Constraint c = getProblem().linear(var, ">=", value);
c.setName("" + var + " >= "+ value);
Goal goalGeValue = new GoalConstraint(c);
return goalGeValue;
}
/**
* Returns an "AND" Goal. The Goal "AND" succeeds if both of the Goals
* "g1" and "g2" succeed. The Goal "AND" fails if at least one of the Goals
* "g1" or "g2" fail. It is implementation specific.
*
* @param g1
* the first Goal which is part of the new "AND" Goal.
* @param g2
* the other Goal which is part of the new "AND" Goal.
* @return a Goal "AND" between the Goals "g1" and "g2".
*/
abstract public Goal and(Goal g1, Goal g2);
/**
* Returns an "OR" Goal. The goal "OR" succeeds if at least one of the
* Goals "g1" or "g2" succeeds. The Goal "OR" fails if both Goals "g1" and
* "g2" fail. It is implementation specific.
*
* @param g1
* the first Goal which is part of the new "OR" Goal.
* @param g2
* the second Goal which is part of the new "OR" Goal.
* @return a Goal "OR" between the Goals "g1" and "g2".
*/
abstract public Goal or(Goal g1, Goal g2);
/**
* This methods is equivalent to execute(goal,ProblemState.DO_NOT_RESTORE);
* @param goal
* @return true if the goal execution succeeds or false if it fails.
*/
public boolean execute(Goal goal) {
return execute(goal, ProblemState.DO_NOT_RESTORE);
}
/**
* This method attempts to find a solution of the problem, for which the solver was defined.
* It uses the search strategy defined by the method setSearchStrategy().
* It returns the found solution (if any) or null. It also saves the solution and makes it
* available through the method getSolution().
* If a solution is not found, the problem state is restored
* to that of before the invocation of this method.
* If a solution is found, the problem state will be restored only if the parameter
* "restore" is true. Otherwise all problem variables will be instantiated with the solution
* values.
*
* The search can be limited by time, number of failed attempts, etc.
*
* @param restoreOrNot defines if the problem state should be restored after a solution is found
* @return a Solution if the search is successful or null
*/
public Solution findSolution(ProblemState restoreOrNot) {
clearSolutions();
Solution solution = null;
// Goal searchStrategy = (Goal)getSearchStrategies().elementAt(0);
// for (int i = 1; i < getSearchStrategies().size(); i++) {
// Goal nextGoal = (Goal)getSearchStrategies().elementAt(i);
// searchStrategy = searchStrategy.and(nextGoal);
// }
Goal searchGoal = combineSearchStrategies();
Goal goalSearchAndSave = searchGoal.and(new GoalAddSolution(this));
if (execute(goalSearchAndSave,restoreOrNot))
solution = getSolution();
if(solution != null && !solution.isBound()) {
if(searchGoal instanceof GoalAssignValuesTimeLimit) {
setTimeLimitExceeded(true);
String msg = "Solver exceeded Time Limit " + getTimeLimit() + " milliseconds";
addExplanation(msg);
getProblem().log(msg);
solution = null;
}
}
return solution;
}
/**
* This method attempts to find all solutions for the Problem.
* It uses the default search strategy or the strategy defined by the latest
* method setSearchStrategy(). It returns an array of found solutions
* or null if there are no solutions. A user has to be careful not to
* overload the available memory because the number of found solutions could be huge.
* The process of finding all solutions can be also controlled by:
*
* - MaxNumberOfSolutions that is the total number of considered solutions
* that may be limited by the method setMaxNumberOfSolutions();
*
- TotalTimeLimit that is the number of seconds allocated for the entire optimization process.
*
* The common implementation is based on the SolutionIterator.
* @return Solution[]
*/
public Solution[] findAllSolutions() {
clearSolutions();
Goal searchGoal = combineSearchStrategies();
Goal addGoal = new GoalAddSolution(this,-1);
//Goal logGoal = new GoalLogSolution(this);
Goal maxGoal = new GoalCheckMaxNumberOfSolutions(this);
Goal backtrackGoal = new GoalBacktrack(this);
//Goal goal1 = searchStrategy.and(addGoal).and(logGoal);
Goal goal1 = searchGoal.and(addGoal);
Goal goal2 = maxGoal.or(backtrackGoal);
Goal goal = goal1.and(goal2);
//execute(goal,ProblemState.DO_NOT_RESTORE);
execute(goal,ProblemState.RESTORE);
return getSolutions();
}
public Goal makeGoal(SearchStrategy strategy) {
Goal goal;
if (strategy.getType().equals(SearchStrategyType.CUSTOM))
goal = new StrategyAsGoal(this, strategy);
else
goal = (Goal) strategy;
return goal;
}
public Goal combineSearchStrategies() {
Vector searchStrategies = getSearchStrategies();
Goal combinedGoal = makeGoal(searchStrategies.elementAt(0));
for (int i = 1; i < searchStrategies.size(); i++) {
Goal nextGoal = makeGoal(searchStrategies.elementAt(i));
combinedGoal = combinedGoal.and(nextGoal);
}
return combinedGoal;
}
/**
* The actual minimization algorithm executes a dichotomized search. During
* the search it modifies an interval [objectiveMin; objectiveMax]. First it
* is trying to find a solution in the [objectiveMin; objectiveMid]. If it
* fails, it is looking at [objectiveMid+1; objectiveMax]. During this
* process it switches the search target: one time in looks at in the upper
* half of the selected interval, another time - to the lower half.
* Successful search stops when (objectiveMax - objectiveMin) is less or equal to tolerance.
*/
public Solution findOptimalSolutionDichotomize(Objective objective, Var objectiveVar) {
long startTime = System.currentTimeMillis();
if (objectiveVar.getName().isEmpty())
objectiveVar.setName("Objective");
if (getProblem().getVar(objectiveVar.getName()) == null) {
getProblem().add(objectiveVar);
}
Var obj = objectiveVar;
if (objective.equals(Objective.MAXIMIZE)) {
obj = objectiveVar.multiply(-1);
obj.setName("-"+objectiveVar.getName());
getProblem().add(obj);
}
addObjective(obj);
Dichotomize dichotomize = new Dichotomize(this, obj);
Solution solution = dichotomize.execute();
if (solution != null)
log("Optimal solution is found. Objective: "
+solution.getValue(obj.getName()));
return solution;
}
// /**
// * Define a goal that will be used by find solutions methods
// * @param goal
// * @see {@link AbstractSolver#findSolution()} {@link AbstractSolver#findOptimalSolution(Var)}
// * {@link AbstractSolver#findAllSolutions()}
// */
// public void setGoal(Goal goal) {
// this.goal = goal;
// }
//
// /**
// * If the goal is not defined by default the solver will use
// * GoalAssignValues(this,problem.getVars());
// * @return a goal to be used with find solutions methods
// */
// public Goal getGoal() {
// if (goal == null)
// return new GoalAssignValues(problem.getVars());
// return goal;
// }
// /**
// * Returns an "AND" Goal. The Goal "AND" succeeds if all three of the
// * Goals "g1", "g2" and "g3" succeed. The Goal "AND" fails if at least one
// * of the Goals "g1", "g2" or "g3" fail. It is implementation specific.
// *
// * @param g1
// * the first Goal which is part of the new "AND" Goal.
// * @param g2
// * the second Goal which is part of the new "AND" Goal.
// * @param g3
// * the third Goal which is part of the new "AND" Goal.
// * @return a Goal "AND" between the Goals "g1", "g2" and "g3".
// */
// public Goal and(Goal g1, Goal g2, Goal g3) {
// Goal g1g2 = and(g1,g2);
// return and(g1g2,g3);
// }
// /**
// * Returns an "AND" Goal. The Goal "AND" succeeds if both of the Goals "g1" and
// * "g2" succeed. The Goal "AND" fails if at least one of
// * the Goals "g1" or "g2" fail. It is implementation specific.
// * @param g1 the first Goal which is part of the new "AND" Goal.
// * @param g2 the other Goal which is part of the new "AND" Goal.
// * @return a Goal "AND" between the Goals "g1" and "g2".
// */
// abstract public Goal and(Goal g1, Goal g2);
// /**
// * Returns an "OR" Goal. The goal "OR" succeeds if at least one of the Goals "g1"
// * or "g2" succeeds. The Goal "OR" fails if both Goals "g1"
// * and "g2" fail. It is implementation specific.
// * @param g1 the first Goal which is part of the new "OR" Goal.
// * @param g2 the second Goal which is part of the new "OR" Goal.
// * @return a Goal "OR" between the Goals "g1" and "g2".
// */
// abstract public Goal or(Goal g1, Goal g2);
/**
* This method attempts to find the solution that minimizes/maximizes the objective variable.
* It uses the search strategy defined by the method setSearchStrategy(strategy).
* The optimization process can be also controlled by:
*
* - OptimizationTolarance that is a difference between optimal solutions - see setOptimizationTolarance()
*
- MaxNumberOfSolutions that is the total number of considered solutions - may be limited by the method
* setMaxNumberOfSolutions()
*
- TotalTimeLimit that is the number of seconds allocated for the entire optimization process.
*
*
At the same time the time for one iteration inside
* optimization loop (a search of one solution) can be also limited by the use of the
* special type of search strategy.
*
The problem state after the execution of this method is always restored. All variables
* that were added to the problems (plus the objectiveVar) will have their assigned values
* saved inside the optimal solution.
*
* @param objective Objective.MINIMIZE or Objective.MAXIMIZE
* @param objectiveVar the variable that is being minimized/maximized
* @return Solution if a solution is found,
* null if there are no solutions.
*/
// public Solution findOptimalSolution(Objective objective, Var objectiveVar) {
// // Add objective to problem if it was not yet added
// if (objectiveVar.getName() == null || objectiveVar.getName().isEmpty())
// objectiveVar.setName("_OBJECTIVE_");
// if (getProblem().getVar(objectiveVar.getName()) == null) {
// getProblem().add(objectiveVar);
// }
// Var objVar = objectiveVar;
// if (objective == Objective.MAXIMIZE) {
// objVar = objectiveVar.mul(-1);
// objVar.setName("negation_of_" + objectiveVar.getName());
// getProblem().add(objVar);
// }
//
// SolverDichotomize dichomize = new SolverDichotomize(this, objVar);
// Solution solution = dichomize.execute();
// return solution;
// }
// /**
// * This method attempts to find all solutions for the Problem using the solving
// * algorithm as defined by the method setGoal(goal).
// * It returns an array of found solutions of null if there are no solutions.
// * One have to be careful not to overload an available memory
// * because a number of found solutions could be huge.
// *
// * The process of finding all solutions can be also controlled by:
// *
// * - Adding a special goal that exclude some found solutions if they are
// * out of the interests. Such a goal can be added to the search goal using the method "and",
// * e.g. solver.setGoal(searchStrategy.and(filterGoal)).
// *
- MaxNumberOfSolutions that is the total number of considered solutions - may be limited by the method
// * setMaxNumberOfSolutions()
// *
- TotalTimeLimit that is the number of seconds allocated for the entire optimization process.
// *
// *
At the same time the time for one iteration inside
// * optimization loop (a search of one solution) can be also limited by the use of the
// * special type of search goals such as GoalAssignValuesTimeLimit.
// *
The problem state after the execution of this method is always restored.
// *
// * @return an array of solutions found or null
// */
// public Solution[] findAllSolutions() {
// clearSolutions();
// Goal goalCheckMaxNumberOfSolutions = new GoalCheckMaxNumberOfSolutions(this);
// Goal goalCheckTotalTimeLimit = new GoalCheckTotalTimeLimit(this);
// Goal goalCheckLimits = goalCheckMaxNumberOfSolutions.and(goalCheckTotalTimeLimit);
// Goal goal = and(getGoal(), new GoalAddSolution(this).and(goalCheckLimits));
// execute(goal, ProblemState.RESTORE);
// return getSolutions();
// }
/**
* This method tries to resolve the problem by applying the Solution
* "solution" to the problem, and returns true if "solution" satisfies all
* of the posted constraints, false otherwise.
*
* @param solution
* the Solution object being tested on the problem.
* @return true if the Solution satisfies all posted constraints, false
* otherwise.
*/
public boolean applySolution(Solution solution) {
return execute(new GoalApplySolution(solution));
}
/**
* This method forces a solver to "backtrack".
* It is used to emulate a failure, e.g. to produce all solutions.
*
* @throws Exception
*/
abstract public void backtrack() throws Exception;
/**
* This method forces to trace (to log) all modification events for
* the variable "var". All modification events are described at the
* interface {@link Problem}
* @param var
*/
public void trace(Var var) {
trace(var, PropagationEvent.ANY);
}
/**
* This method forces to trace (to log) the modification "event" for
* the variable "var". All modification events are described at the
* interface {@link Problem}
* @param var
*/
public void trace(Var var, PropagationEvent event) {
if (event == PropagationEvent.ANY) {
trace(var, PropagationEvent.MIN);
trace(var, PropagationEvent.MAX);
trace(var, PropagationEvent.REMOVE);
trace(var, PropagationEvent.VALUE);
}
else
if (event == PropagationEvent.RANGE) {
trace(var, PropagationEvent.MIN);
trace(var, PropagationEvent.MAX);
}
else
(new javax.constraints.extra.ConstraintTraceVar(var, event)).post();
}
/**
* This method forces to trace (to log) all modification events for
* all variables "vars"
* @param vars
*/
public void trace(Var[] vars) {
trace(vars, PropagationEvent.ANY);
}
/**
* This method forces solver to trace (to log) the modification "event"
* for all variables "var". All modification events are described at the
* interface {@link Problem}
* @param vars
* @param event
*/
public void trace(Var[] vars, PropagationEvent event) {
for (int i = 0; i < vars.length; i++) {
trace(vars[i],event);
}
}
/**
* This method forces to trace (to log) all modification events for
* the variable "setVar".
* @param setVar
*/
public void trace(VarSet setVar) {
trace(setVar, PropagationEvent.ANY);
}
/**
* This method forces to trace (to log) the modification "event" for
* the variable "setVar". All modification events are described at the
* interface {@link Problem}
* @param setVar
* @param event
*/
public void trace(VarSet setVar, PropagationEvent event) {
javax.constraints.Var indexVar = setVar.getCardinality();
trace(indexVar,event);
BasicVarSet basicSetVar = (BasicVarSet)setVar;
javax.constraints.Var[] requiredVars = basicSetVar.getRequiredVars();
trace(requiredVars,event);
}
}