In order to allow a flexible combination of constraint and variable generation, every subproblem has its own set of active constraints and variables, which are represented by the generic class ABA_ACTIVE . By default, the variables and the constraints of the last solved linear program of the father of the subproblem are inherited. Therefore, the local constraint and variable sets speed up the optimization. The disadvantage of these local copies is that more memory is allocated per subproblem. However, this local storage of the active constraints and variables will simplify a future parallelization of the framework.
Together with the active constraints and variables we also store in every subproblem the LP-statuses of the variables and slack variables, the upper and lower bounds of the variables, and if a variable is fixed or set.