class Game(object):
"""
A Game is a general abstraction for local search based games
Known children are MinConflictsGame and SimulatedAnnealingGame
These children should be also abstract
"""
def __init__(self):
self.state_manager = StateManager()
self.level = None
self.max_steps = 0
self.number_steps = 0
self.max_time = 120
self.date_begin = 0
self.verbose = False
def generate(self, level):
"""
Generates a new game. Level determines how hard the game will be
"""
levels = {
"easy": GameLevel.EASY,
"medium": GameLevel.MEDIUM,
"hard": GameLevel.HARD
}
assert(level in levels)
self.level = levels[level]
self.max_steps = 10000
self.state_manager.build_new_game(self.level)
self.state_manager.indexing_constraints()
self.state_manager.generate_game()
def run(self):
"""
Runs the current game.
It's just an execution of algorithms in the case of MinConflictsGame
and SimulatedAnnealingGame classes
"""
pass
EVAL_MAX = 1000.0
def evaluate(self, state):
"""
Evaluates a state (objective function)
Min value is 1 and max value is EVAL_MAX
"""
# The idea is to subtract the number of violated constraints (NVC) to a
# fix value (EVAL_MAX). But since the NVC can be greater than EVAL_MAX
# we must keep 0 <= NVC <= EVAL_MAX with a rule of three.
# There is still a problem: the number of total constraints is not
# perfect because a constraint can modify the count with a coeff (see
# Magic Square game with SumEqualsConstraint).
# So here we increase this number by a coeff of 1000. It works fine
# for this exercise but this solution is far from perfect.
total_constraints = len(self.state_manager.constraints) * 1000
nb_constraints = self.state_manager.count_constraint_violated(state)
_max = Game.EVAL_MAX
_eval = _max - (nb_constraints * _max) / total_constraints
return max(1, _eval)
def is_terminated(self):
"""
Returns True if the current state is an optimal one, False else
"""
return self.state_manager.is_optimal() or \
self.outofsteps() or self.outoftime()
def outofsteps(self):
return self.number_steps >= self.max_steps
def outoftime(self):
return self.date_begin > 0 and \
time() > (self.date_begin + self.max_time)
def show_state_manager(self):
if self.verbose:
print self.state_manager
