def setUp(cls):
cls.r1 = 'running(c1,1)'
cls.r2 = 'running(c2,1)'
cls.r3 = 'running(c3,1)'
cls.lp1 = Literal.positive(cls.r1)
cls.ln1 = Literal.negative(cls.r1)
cls.lp2 = Literal.positive(cls.r2)
cls.ln2 = Literal.negative(cls.r2)
cls.lp3 = Literal.positive(cls.r3)
cls.ln3 = Literal.negative(cls.r3)
def p_literal(p):
'''literal : LPAREN NOT_KEY predicate RPAREN
| predicate'''
if len(p) == 2:
p[0] = Literal.positive(p[1])
elif len(p) == 5:
p[0] = Literal.negative(p[3])
def __init__(self, atoms):
self._atoms = atoms
self._terms = []
self._weights = {}
self.__index = {}
for atom in self._atoms:
self.__index[Literal.positive(atom)] = set()
self.__index[Literal.negative(atom)] = set()
def test_combine(self):
parser = Parser()
transition_model = '''
1.000::running(c1, 1) :- running(c1, 0).
'''
atoms1, rules1 = parser.parse(transition_model)
wdnf1 = WeightedDNF(set(atoms1.keys()))
for index, rule in enumerate(rules1):
pos_term = tuple(
sorted([Literal.positive(rule.head)] + list(rule.body)))
wdnf1.add_term(pos_term, rule.probability)
neg_term = tuple(
sorted([Literal.negative(rule.head)] + list(rule.body)))
wdnf1.add_term(neg_term, 1 - rule.probability)
value_model = '''
V(0) :- not(running(c1, 1)), not(running(c2, 1)), not(running(c3, 1)).
V(1) :- running(c1, 1), not(running(c2, 1)), not(running(c3, 1)).
V(1) :- not(running(c1, 1)), running(c2, 1), not(running(c3, 1)).
V(1) :- not(running(c1, 1)), not(running(c2, 1)), running(c3, 1).
V(2) :- running(c1, 1), running(c2, 1), not(running(c3, 1)).
V(2) :- running(c1, 1), not(running(c2, 1)), running(c3, 1).
V(2) :- not(running(c1, 1)), running(c2, 1), running(c3, 1).
V(3) :- running(c1, 1), running(c2, 1), running(c3, 1).
'''
atoms2, rules2 = parser.parse(value_model)
wdnf2 = WeightedDNF(set(atoms2.keys()))
for index, rule in enumerate(rules2):
wdnf2.add_term(sorted(rule.body), float(rule.head[2:-1]))
prod = combine(wdnf1, wdnf2, float.__mul__)
self.assertEqual(len(prod), len(wdnf2))
pos = Literal.positive('running(c1,1)')
neg = Literal.negative('running(c1,1)')
r1 = Literal.positive('running(c1,0)')
for weight, term in prod:
self.assertEqual(len(term), 4)
if neg in term:
self.assertEqual(weight, 0.0)
if pos in term:
count = sum([
literal.is_positive() for literal in term if literal != r1
])
self.assertEqual(weight, float(count))
def p_literal(p):
'''literal : NEGATION LPAREN atom RPAREN
| NEGATION atom
| atom'''
if len(p) == 5:
p[0] = Literal.negative(p[3])
elif len(p) == 3:
p[0] = Literal.negative(p[2])
else:
p[0] = Literal.positive(p[1])
def __update_index(self, term, index):
term = set(term)
for atom in self._atoms:
positive_literal = Literal.positive(atom)
negative_literal = Literal.negative(atom)
if positive_literal in term:
self.__index[positive_literal].add(index)
continue
if negative_literal in term:
self.__index[negative_literal].add(index)
continue
self.__index[positive_literal].add(index)
self.__index[negative_literal].add(index)
def test_add_term(self):
# value function
wdnf1 = wdnf.WeightedDNF(self.atoms1)
self.assertEqual(len(wdnf1), 0)
for index, rule in enumerate(self.rules1):
wdnf1.add_term(rule.body, int(rule.head[2:-1]))
self.assertEqual(len(wdnf1), index + 1)
fluents = ['running(c1,1)', 'running(c2,1)', 'running(c3,1)']
for fluent in fluents:
pos = Literal.positive(fluent)
neg = Literal.negative(fluent)
t1 = set(wdnf1.terms_by_literals([pos]))
t2 = set(wdnf1.terms_by_literals([neg]))
self.assertEqual(len(t1), len(wdnf1) / 2)
self.assertEqual(len(t2), len(wdnf1) / 2)
self.assertEqual(t1 & t2, set())
self.assertEqual(t1 | t2, set(wdnf1.terms))
values = [0, 1, 2, 3]
terms_sets = [set(wdnf1.terms_by_weight(v)) for v in values]
self.assertEqual(sum([len(s) for s in terms_sets]), len(wdnf1))
for i in range(len(terms_sets[-1])):
for j in range(i + 1, len(terms_sets)):
self.assertEqual(terms_sets[i] & terms_sets[j], set())
# transition function
wdnf2 = wdnf.WeightedDNF(self.atoms2)
self.assertEqual(len(wdnf2), 0)
for index, rule in enumerate(self.rules2):
pos_term = tuple([Literal.positive(rule.head)] + list(rule.body))
wdnf2.add_term(pos_term, rule.probability)
neg_term = tuple([Literal.negative(rule.head)] + list(rule.body))
wdnf2.add_term(neg_term, 1 - rule.probability)
self.assertEqual(len(wdnf2), 2 * len(self.rules2))
next_state_fluents = [
'running(c1,1)', 'running(c2,1)', 'running(c3,1)'
]
for fluent in next_state_fluents:
pos = Literal.positive(fluent)
neg = Literal.negative(fluent)
t1 = set(wdnf2.terms_by_literals([pos]))
t2 = set(wdnf2.terms_by_literals([neg]))
self.assertEqual(len(t1), len(t2))
inter = set(term for _, term in t1 & t2)
self.assertTrue(pos not in inter)
self.assertTrue(neg not in inter)
actions = ['reboot(c1)', 'reboot(c2)', 'reboot(c3)']
for action in actions:
pos = Literal.positive(action)
neg = Literal.negative(action)
t1 = set(wdnf2.terms_by_literals([pos]))
t2 = set(wdnf2.terms_by_literals([neg]))
inter = set(term for _, term in t1 & t2)
self.assertTrue(pos not in inter)
self.assertTrue(neg not in inter)
if action == actions[0]:
self.assertEqual(len(t1), len(wdnf2) - 6)
self.assertEqual(len(t2), len(wdnf2) - 2)
if action == actions[2]:
self.assertEqual(len(t1), len(wdnf2) - 6)
self.assertEqual(len(t2), len(wdnf2) - 2)
