def solve(self):
start_time = time.time()
total_lp_solve_time = 0
# Create branching options for root
branching_options = []
for person in self.people:
for day in range(1, self.num_days + 1):
branching_options.append(BranchingOption(DecisionType.PERSON_DAY, person.uid, day))
for set_constraint in set_constraints:
if set_constraint.constraintType.value == SetConstraintType.DEPARTMENT.value:
lower_bound = set_constraint.low_bound
upper_bound = set_constraint.up_bound
if upper_bound < 0:
upper_bound = len(set_constraint.personList)
for day in range(1, self.num_days + 1):
branching_options.append(BranchingOption(DecisionType.DEPT_DAY, set_constraint.sid,
day, lower_bound, upper_bound))
elif set_constraint.constraintType.value == SetConstraintType.SYNERGY.value:
for day in range(1, self.num_days + 1):
branching_options.append(BranchingOption(DecisionType.SYNERGY_DAY, set_constraint.sid, day))
else:
pass # Should not reach here unless new set constraint types are added
root = BnbNode(None, branching_options)
root.lp = pulp_utils.build_scheduling_lp(num_days, people, set_constraints)
stack = deque()
stack.append(root)
count_explored_nodes = 0
while stack: # implicit condition: stack is not empty
elapsed_time = time.time() - start_time
if elapsed_time > time_limit:
break
node = stack.pop()
count_explored_nodes += 1
# print(node)
children, lp_solve_time = node.branch()
total_lp_solve_time += lp_solve_time
# print('Node at depth {0}:\n\tLP value: {1:.2f}'.format(node.depth, node.lp_value))
if node.feasible_value > self.best_value:
self.best_value = node.feasible_value
self.best_solution = node.feasible_solution
# Pruning: Don't need to add children if LP relaxation has
# opt value no better than best feasible integer solution seen so far
if node.lp_value <= self.best_value or len(children) == 0:
continue
for child in children:
stack.append(child) # Push children onto stack for DFS
# Print summary stats
print('Explored {0:d} nodes.'.format(count_explored_nodes))
print('Elapsed time: {0:.3f} s'.format(elapsed_time))
print('Total LP solve time: {0:.3f} s'.format(total_lp_solve_time))
print('Best value: {0:d}'.format(int(self.best_value)))
# Update solver status
if elapsed_time > time_limit and time_limit > 0:
if self.best_value <= 0:
self.status = SolverStatus.OUT_OF_TIME
else:
self.status = SolverStatus.FEASIBLE
else:
if self.best_value <= 0:
self.status = SolverStatus.INFEASIBLE
else:
self.status = SolverStatus.OPTIMAL
# Build and return best Schedule
best_schedule = Schedule(people=self.people)
best_schedule.buildFromSolutionVariables(self.best_solution)
return best_schedule
