Class IloSearchPhase

The class IloSearchPhase is an object that is used to define instantiation strategies to help the embedded CP Optimizer search. A search phase is composed of:

• an array of decision variables (IloIntVarArray, IloIntervalVarArray or IloIntervalSequenceVarArray) to instantiate,
• a variable chooser (IloIntVarChooser) that defines how the next variable to instantiate is chosen,
• a value chooser (IloIntValueChooser) that defines how values are chosen when instantiating variables.

A search phase is given to an IloCP instance by invoking the method:

  void IloCP::setSearchPhases(const IloSearchPhase phase) const;
 

The embedded search strategy - that is determined by the value of the parameter IloCP::SearchType - will use the search phases to instantiate the variables for which a search phase is specified.

Several search phases can be given to an IloCP instance by way of an IloSearchPhaseArray. The member function that needs to be invoked for that purpose is:

   void IloCP::setSearchPhases(const IloSearchPhaseArray phaseArray) const;
 

The order of the search phases in the array is important. In the CP Optimizer search strategy, the variables will be instantiated phase by phase starting by the first phase of the array. It is not necessary that the variables in the search phases cover all the variables of the problem. It can be assumed that a search phase containing all the problem variables is implicitly added to the end of the given array of search phases.

For instance, assume we have a model over three arrays of variables x, y and z. In the following search phase array:

   IloSearchPhaseArray phaseArray(env);
   phaseArray.add(IloSearchPhase(env, x, ..., ...));
   phaseArray.add(IloSearchPhase(env, y, ..., ...));
 

the variables x will be instantiated before the variables y and once x and y are instantiated, variables in z will be instantiated. There are some models which have particularly well-understood structure where giving such an order can have a dramatic impact on the solving time.

It is important to note that for constructing a search phase, not all the three parameters above are required. A search phase can be created with an array of variables only. The embedded search will then choose an instantiation strategy automatically. For instance, assuming that x and y are arrays of integer variables, the following code:

   IloCP cp(env);
   IloSearchPhaseArray phaseArray(env);
   phaseArray.add(IloSearchPhase(env, x));
   phaseArray.add(IloSearchPhase(env, y));
   cp.setSearchPhases(phaseArray);
   cp.solve();
 

indicates to CP Optimizer search that variables from the array x must be instantiated before those from the array y. The way to instantiate them will be chosen by the CP Optimizer search.

Similarly, it is not necessary to specify an array of variables to a search phase. A search phase defined this way will be applied to every decision variable of the appropriate type extracted from the model.

This constructor creates a search phase with an array of variables and value choosers. The phase will use varChooser and valueChooser to determine how to instantiate variables in vars.

This constructor creates a search phase with an array of variables only. The variable and value choosers for these variables will be chosen by CP Optimizer search automatically.

This constructor creates a search phase with a variable and a value chooser. The variables instantiated by this phase are those extracted to the IloCP instance.

This constructor creates a search phase with an array of interval variables. During this phase CP Optimizer fixes the value of interval variables specified in the array intervalVars: each interval variable in the array will be assigned a presence status and for each present interval, a start and an end value.

For more information on interval variables, see the concept Interval variables in CP Optimizer.

This constructor creates a search phase with an array of interval sequence variables. During this phase CP Optimizer fixes the value of interval sequence variables specified in the array sequenceVars: each sequence variable in the array will be assigned a totally ordered sequence of present interval variables. Note that this search phase also fixes the presence statuses of the intervals involved in the sequence variables. This phase does not fix the start and end values of interval variables.

For more information on interval sequence variables see the concept Interval variable sequencing in CP Optimizer.

This function releases all the memory allocated internally to create an instance.