def test_get_alpha_num(self):
"""get_alpha_num()"""
# Simple regression test for small automaton.
nfaData = nfa_data()
nfaData.states[0] = b_State(0,set())
nfaData.states[1] = b_State(1,set())
nfaData.states[2] = b_State(2,set())
nfaData.states[3] = b_State(3,set([0]))
nfaData.alphabet[0] = b_Sym_char("a", "a", 0)
nfaData.alphabet[1] = b_Sym_char("b", "b", 1)
nfaData.alphabet[2] = b_Sym_char("c", "c", 2)
nfaData.start = 0
nfaData.transitions.add( (0,0,1) )
nfaData.transitions.add( (1,1,2) )
nfaData.transitions.add( (2,2,3) )
nfaData.final.add(3)
aut = PHF_DFA()
a = bdz()
a.set_limit(128)
aut.set_PHF_class(a)
aut.create_from_nfa_data(nfaData)
aut.compute()
self.assertEqual(aut.get_alpha_num(), 3)
# Manually remove symbol from _automaton1
del aut._automaton1.alphabet[2]
self.assertEqual(aut.get_alpha_num(), 2)
def test_get_table_parameters(self):
"""get_table_parameters"""
# Manually set state_bits and symbol_bits, then check if
# if get_table_parameters returns same values
aut = PHF_DFA()
aut.state_bits = 24
aut.symbol_bits = 20
tmp = aut.get_table_parameters()
self.assertEqual(tmp[0], 24)
self.assertEqual(tmp[1], 20)
def test___init__(self):
"""__init__()"""
# Create PHF_DFA and check the values of internal variables.
aut = PHF_DFA()
self.assertFalse(aut.get_compute())
self.assertEqual(aut._automaton, aut._automaton1)
self.assertEqual(aut.state_bits, 10)
self.assertEqual(aut.symbol_bits, 12)
self.assertEqual(aut.hash_function, None)
self.assertEqual(aut.trans_table, None)
self.assertEqual(aut.ran, 0)
self.assertFalse(aut.fallback)
self.assertEqual(aut.fallback_state, -1)
self.assertFalse(aut.faulty)
self.assertEqual(aut.compress_hash, None)
self.assertEqual(aut.compress_bits, 0)
self.assertEqual(aut.bad_transitions, 0)
self.assertEqual(aut.collisions, dict())
def test_disable_faulty_transitions(self):
"""disable_faulty_transitions()"""
# Check if variables compress_bits, compress_hash, faulty and _compute
# were set to their default values after calling method
# disable_faulty_transitions
aut = PHF_DFA()
aut.enable_faulty_transitions(13)
aut._compute = True
aut.disable_faulty_transitions()
self.assertEqual(aut.compress_bits, 13)
self.assertNotEqual(aut.compress_hash, None)
self.assertFalse(aut.get_compute())
self.assertFalse(aut.faulty)
def get_phf(ruleset):
"""
Generate number of states, transitions and consumed memory for \
Perfect hashing DFA.
"""
# Create parser - use default parser
po = parser.parser()
# Parse input file
po.load_file(ruleset)
# create phf_dfa automaton
aut = PHF_DFA()
# Make automaton from RE which was in input file
aut.create_by_parser(po)
# redefine default PHF class
a = bdz()
a.set_ratio(2.0)
aut.set_PHF_class(a)
# compute dfa and PHF table
aut.compute()
# Return experimental results
return ["Perfect Hashing DFA", aut.get_state_num(), aut.get_trans_num(), aut.report_memory_real(), aut.report_memory_real()]
def test_decode_symbol(self):
"""decode_symbol()"""
# Test if different types of symbols are decoded correctly and
# the symbol was removed from the beginning of input string.
aut = PHF_DFA()
aut._automaton.alphabet[0] = b_Sym_char_class("ch0", set(['a', 'b']), 0)
aut._automaton.alphabet[1] = b_Sym_char_class("ch1", set(['c', 'd']), 1)
aut._automaton.alphabet[2] = b_Sym_char_class("ch2", set(['e', 'f']), 2)
aut._automaton.alphabet[3] = b_Sym_char("ch3", "g", 3)
aut._automaton.alphabet[4] = b_Sym_kchar("ch4", (frozenset(['1', '2']), frozenset(['1', '2'])), 4)
self.assertEqual(aut.decode_symbol("abeg112"), ("beg112", 0))
self.assertEqual(aut.decode_symbol("beg112"), ("eg112", 0))
self.assertEqual(aut.decode_symbol("eg112"), ("g112", 2))
self.assertEqual(aut.decode_symbol("g112"), ("112", 3))
self.assertEqual(aut.decode_symbol("112"), ("2", 4))
# Nonexistent symbol is removed from the string and -1 is returned
self.assertEqual(aut.decode_symbol("2"), ("", -1))
def test_set_table_parameters(self):
"""set_table_parameters()"""
# Set parameters using method set_table_parameters, then check if
# internal variables have the same values
aut = PHF_DFA()
aut._compute = True
tmp = (24, 20)
aut.set_table_parameters(tmp)
self.assertEqual(aut.state_bits, 24)
self.assertEqual(aut.symbol_bits, 20)
self.assertFalse(aut.get_compute())
def test_set_PHF_class(self):
"""set_PHF_class()"""
# Create phf class and assing it to PHF_DFA using method set_PHF_class.
# Check if the variable hash_function was set and _compute is false.
aut = PHF_DFA()
aut._compute = True
a = bdz()
a.set_limit(1024)
a.set_iteration_limit(8)
aut.set_PHF_class(a)
self.assertEqual(aut.hash_function, a)
self.assertFalse(aut.get_compute())
def test_remove_fallback_transitions(self):
"""remove_fallback_transitions()"""
# 1. /abc/, state -1 (automatically chosen 0) - 4 transitions removed
nfaData = nfa_data()
nfaData.states[0] = b_State(0,set())
nfaData.states[1] = b_State(1,set())
nfaData.states[2] = b_State(2,set())
nfaData.states[3] = b_State(3,set([0]))
nfaData.alphabet[0] = b_Sym_char("a", "a", 0)
nfaData.alphabet[1] = b_Sym_char("b", "b", 1)
nfaData.alphabet[2] = b_Sym_char("c", "c", 2)
nfaData.start = 0
nfaData.transitions.add( (0,0,1) )
nfaData.transitions.add( (0,1,0) )
nfaData.transitions.add( (0,2,0) )
nfaData.transitions.add( (1,1,2) )
nfaData.transitions.add( (1,0,1) )
nfaData.transitions.add( (1,2,0) )
nfaData.transitions.add( (2,2,3) )
nfaData.transitions.add( (2,0,1) )
nfaData.transitions.add( (2,1,0) )
nfaData.transitions.add( (3,0,3) )
nfaData.transitions.add( (3,1,3) )
nfaData.transitions.add( (3,2,3) )
nfaData.final.add(3)
result = copy.deepcopy(nfaData)
aut = PHF_DFA()
a = bdz()
a.set_limit(128)
aut.set_PHF_class(a)
aut.create_from_nfa_data(nfaData)
aut.enable_fallback_state(warning=False)
aut.compute()
cp = aut._automaton1
self.assertEqual(len(cp.states), len(result.states))
self.assertEqual(len(cp.alphabet), len(result.alphabet))
self.assertEqual(len(cp.transitions), 8) # 4 removed transitions
for i in cp.transitions: # no transitions to fallback_state
self.assertNotEqual(i[2], aut.fallback_state)
self.assertEqual(len(cp.final), len(result.final))
# 2. /abc/, state 1 - 3 transitions removed
aut._automaton1 = aut._automaton
aut.enable_fallback_state(1, False)
aut.compute()
cp = aut._automaton1
self.assertEqual(len(cp.states), len(result.states))
self.assertEqual(len(cp.alphabet), len(result.alphabet))
self.assertEqual(len(cp.transitions), 9) # 3 removed transitions
for i in cp.transitions: # no transitions to fallback_state
self.assertNotEqual(i[2], aut.fallback_state)
self.assertEqual(len(cp.final), len(result.final))
# 3. /^abc/, state 0 - automaton does not change
nfaData = nfa_data()
nfaData.states[0] = b_State(0,set())
nfaData.states[1] = b_State(1,set())
nfaData.states[2] = b_State(2,set())
nfaData.states[3] = b_State(3,set([0]))
nfaData.alphabet[0] = b_Sym_char("a", "a", 0)
nfaData.alphabet[1] = b_Sym_char("b", "b", 1)
nfaData.alphabet[2] = b_Sym_char("c", "c", 2)
nfaData.start = 0
nfaData.transitions.add( (0,0,1) )
nfaData.transitions.add( (1,1,2) )
nfaData.transitions.add( (2,2,3) )
nfaData.final.add(3)
result = copy.deepcopy(nfaData)
aut = PHF_DFA()
a = bdz()
a.set_limit(128)
aut.set_PHF_class(a)
aut.create_from_nfa_data(nfaData)
aut.enable_fallback_state(0, warning=False)
aut.compute()
cp = aut._automaton1
self.assertEqual(len(cp.states), len(result.states))
self.assertEqual(len(cp.alphabet), len(result.alphabet))
self.assertEqual(len(cp.transitions), len(result.transitions))
for i in cp.transitions: # no transitions to fallback_state
self.assertNotEqual(i[2], aut.fallback_state)
self.assertEqual(len(cp.final), len(result.final))
def test_disable_fallback_state(self):
"""disable_fallback_state()"""
# Test if the variables _compute, fallback and fallback_state were set
# to the default values.
nfaData = nfa_data()
nfaData.states[0] = b_State(0,set())
nfaData.states[1] = b_State(1,set())
nfaData.states[2] = b_State(2,set())
nfaData.states[3] = b_State(3,set([0]))
nfaData.alphabet[0] = b_Sym_char("a", "a", 0)
nfaData.alphabet[1] = b_Sym_char("b", "b", 1)
nfaData.alphabet[2] = b_Sym_char("c", "c", 2)
nfaData.start = 0
nfaData.transitions.add( (0,0,1) )
nfaData.transitions.add( (1,1,2) )
nfaData.transitions.add( (2,2,3) )
nfaData.final.add(3)
aut = PHF_DFA()
a = bdz()
a.set_limit(128)
aut.set_PHF_class(a)
aut.create_from_nfa_data(nfaData)
aut.enable_fallback_state(warning=False)
aut.compute()
aut.disable_fallback_state()
self.assertFalse(aut.get_compute())
self.assertFalse(aut.fallback)
self.assertEqual(aut.fallback_state, -1)
def test_enable_fallback_state(self):
"""enable_fallback_state()"""
# Test if fallback and fallback_state is set accordingly, _compute is
# set to False and warning is/is not printed on stdout depending on
# value of parameter warning.
nfaData = nfa_data()
nfaData.states[0] = b_State(0,set())
nfaData.states[1] = b_State(1,set())
nfaData.states[2] = b_State(2,set())
nfaData.states[3] = b_State(3,set([0]))
nfaData.alphabet[0] = b_Sym_char("a", "a", 0)
nfaData.alphabet[1] = b_Sym_char("b", "b", 1)
nfaData.alphabet[2] = b_Sym_char("c", "c", 2)
nfaData.start = 0
nfaData.transitions.add( (0,0,1) )
nfaData.transitions.add( (1,1,2) )
nfaData.transitions.add( (2,2,3) )
nfaData.final.add(3)
aut = PHF_DFA()
a = bdz()
a.set_limit(128)
aut.set_PHF_class(a)
aut.create_from_nfa_data(nfaData)
aut.compute()
# redirect stdout to file
tmp = sys.stdout
f = open("stdout.output", 'w')
sys.stdout = f
aut.enable_fallback_state(2, warning=False)
f.close()
e = open("stdout.output", 'r')
line = e.readline()
# warning was set to False, stdout should be empty
self.assertFalse(line)
# check if the fallback_state was set
self.assertEqual(aut.fallback_state, 2)
self.assertFalse(aut.get_compute())
self.assertTrue(aut.fallback)
f = open("stdout.output", 'w')
sys.stdout = f
aut.enable_fallback_state()
f.close()
e = open("stdout.output", 'r')
line = e.readline()
# warning should be printed by default
self.assertTrue(line)
# check if the fallback_state was chosen correctly
self.assertEqual(aut.fallback_state, 1)
self.assertFalse(aut.get_compute())
self.assertTrue(aut.fallback)
# restore sys.stdout
sys.stdout = tmp
os.remove("stdout.output")
def get_phf(ruleset):
"""
Generate number of states, transitions and consumed memory for \
Perfect hashing DFA.
"""
# Create parser - use default parser
po = parser.parser()
# Parse input file
po.load_file(ruleset)
# create phf_dfa automaton
aut = PHF_DFA()
# Make automaton from RE which was in input file
aut.create_by_parser(po)
# redefine default PHF class
a = bdz()
a.set_ratio(2.0)
aut.set_PHF_class(a)
# compute dfa and PHF table
aut.compute()
# Return experimental results
return [
"Perfect Hashing DFA",
aut.get_state_num(),
aut.get_trans_num(),
aut.report_memory_real(),
aut.report_memory_real()
]
def test_report_memory_real(self):
"""report_memory_real()"""
# Few simple regression tests for different sizes of PHF table, state
# and symbol representations and faulty transitions.
nfaData = nfa_data()
nfaData.states[0] = b_State(0,set())
nfaData.states[1] = b_State(1,set())
nfaData.states[2] = b_State(2,set())
nfaData.states[3] = b_State(3,set([0]))
nfaData.alphabet[0] = b_Sym_char("a", "a", 0)
nfaData.alphabet[1] = b_Sym_char("b", "b", 1)
nfaData.alphabet[2] = b_Sym_char("c", "c", 2)
nfaData.start = 0
nfaData.transitions.add( (0,0,1) )
nfaData.transitions.add( (1,1,2) )
nfaData.transitions.add( (2,2,3) )
nfaData.final.add(3)
aut = PHF_DFA()
a = bdz()
a.set_limit(8)
aut.set_PHF_class(a)
aut.create_from_nfa_data(nfaData)
aut.compute()
self.assertEqual(aut.report_memory_real(), 120)
aut.set_table_parameters((4,6))
self.assertEqual(aut.report_memory_real(), 48)
aut.set_table_parameters((4,7))
self.assertEqual(aut.report_memory_real(), 72)
a.set_limit(5)
aut.set_PHF_class(a)
aut.compute()
self.assertEqual(aut.report_memory_real(), 45)
aut.enable_faulty_transitions(10)
self.assertEqual(aut.report_memory_real(), 30)
aut.enable_faulty_transitions(19)
self.assertEqual(aut.report_memory_real(), 60)
def test_compute(self):
"""compute()"""
# 1. /^abc/ - automaton does not change, PHF table is created
nfaData = nfa_data()
nfaData.states[0] = b_State(0,set())
nfaData.states[1] = b_State(1,set())
nfaData.states[2] = b_State(2,set())
nfaData.states[3] = b_State(3,set([0]))
nfaData.alphabet[0] = b_Sym_char("a", "a", 0)
nfaData.alphabet[1] = b_Sym_char("b", "b", 1)
nfaData.alphabet[2] = b_Sym_char("c", "c", 2)
nfaData.start = 0
nfaData.transitions.add( (0,0,1) )
nfaData.transitions.add( (1,1,2) )
nfaData.transitions.add( (2,2,3) )
nfaData.final.add(3)
result = copy.deepcopy(nfaData)
aut = PHF_DFA()
a = bdz()
a.set_limit(128)
aut.set_PHF_class(a)
aut.create_from_nfa_data(nfaData)
aut.compute()
cp = aut._automaton1
self.assertEqual(len(cp.states), len(result.states))
self.assertEqual(len(cp.alphabet), len(result.alphabet))
self.assertEqual(len(cp.transitions), len(result.transitions))
self.assertEqual(len(cp.final), len(result.final))
self.assertNotEqual(aut.trans_table, None)
self.assertTrue(aut.get_compute())
# 2. determinization of /^ab|ac/, PHF table is created
nfaData = nfa_data()
nfaData.states[0] = b_State(0,set())
nfaData.states[1] = b_State(1,set())
nfaData.states[2] = b_State(2,set([0]))
nfaData.states[3] = b_State(3,set())
nfaData.states[4] = b_State(4,set([0]))
nfaData.alphabet[0] = b_Sym_char("a", "a", 0)
nfaData.alphabet[1] = b_Sym_char("b", "b", 1)
nfaData.alphabet[2] = b_Sym_char("c", "c", 2)
nfaData.start = 0
nfaData.transitions.add( (0,0,1) )
nfaData.transitions.add( (1,1,2) )
nfaData.transitions.add( (0,0,3) )
nfaData.transitions.add( (3,2,4) )
nfaData.final.add(2)
nfaData.final.add(4)
aut = PHF_DFA()
a = bdz()
a.set_limit(128)
aut.set_PHF_class(a)
aut.create_from_nfa_data(nfaData)
aut.compute()
cp = aut._automaton1
self.assertEqual(len(cp.states), 3)
self.assertEqual(len(cp.alphabet), 3)
self.assertEqual(len(cp.transitions), 3)
self.assertEqual(len(cp.final), 1)
self.assertNotEqual(aut.trans_table, None)
self.assertTrue(aut.get_compute())
# 3. resolve alphabet - /^[a-c][b-d]/, PHF table is created
nfaData = nfa_data()
nfaData.states[0] = b_State(0,set())
nfaData.states[1] = b_State(1,set())
nfaData.states[2] = b_State(2,set([0]))
nfaData.alphabet[0] = b_Sym_char_class("ch0", set(['a', 'b', 'c']), 0)
nfaData.alphabet[1] = b_Sym_char_class("ch1", set(['b', 'c', 'd']), 1)
nfaData.start = 0
nfaData.transitions.add( (0,0,1) )
nfaData.transitions.add( (1,1,2) )
nfaData.final.add(2)
aut = PHF_DFA()
a = bdz()
a.set_limit(128)
aut.set_PHF_class(a)
aut.create_from_nfa_data(nfaData)
aut.compute()
cp = aut._automaton1
self.assertEqual(len(cp.states), 3)
self.assertEqual(len(cp.alphabet), 3)
self.assertEqual(len(cp.transitions), 4)
self.assertEqual(len(cp.final), 1)
self.assertNotEqual(aut.trans_table, None)
self.assertTrue(aut.get_compute())
# 4. /abc/ and enable_fallback_state - some transitions are removed
nfaData = nfa_data()
nfaData.states[0] = b_State(0,set())
nfaData.states[1] = b_State(1,set())
nfaData.states[2] = b_State(2,set())
nfaData.states[3] = b_State(3,set([0]))
nfaData.alphabet[0] = b_Sym_char("a", "a", 0)
nfaData.alphabet[1] = b_Sym_char("b", "b", 1)
nfaData.alphabet[2] = b_Sym_char("c", "c", 2)
nfaData.start = 0
nfaData.transitions.add( (0,0,1) )
nfaData.transitions.add( (0,1,0) )
nfaData.transitions.add( (0,2,0) )
nfaData.transitions.add( (1,1,2) )
nfaData.transitions.add( (1,0,1) )
nfaData.transitions.add( (1,2,0) )
nfaData.transitions.add( (2,2,3) )
nfaData.transitions.add( (2,0,1) )
nfaData.transitions.add( (2,1,0) )
nfaData.transitions.add( (3,0,3) )
nfaData.transitions.add( (3,1,3) )
nfaData.transitions.add( (3,2,3) )
nfaData.final.add(3)
result = copy.deepcopy(nfaData)
aut = PHF_DFA()
a = bdz()
a.set_limit(128)
aut.set_PHF_class(a)
aut.create_from_nfa_data(nfaData)
aut.enable_fallback_state(warning=False)
aut.compute()
cp = aut._automaton1
self.assertEqual(len(cp.states), len(result.states))
self.assertEqual(len(cp.alphabet), len(result.alphabet))
self.assertTrue(len(cp.transitions) < len(result.transitions))
self.assertEqual(len(cp.final), len(result.final))
self.assertNotEqual(aut.trans_table, None)
self.assertTrue(aut.get_compute())
def test_enable_faulty_transitions(self):
"""enable_faulty_transitions()"""
# Check if variables compress_bits, compress_hash, faulty and _compute
# were set appropriately after calling enable_faulty_transitions.
aut = PHF_DFA()
aut._compute = True
aut.enable_faulty_transitions(13)
self.assertEqual(aut.compress_bits, 13)
self.assertNotEqual(aut.compress_hash, None)
self.assertFalse(aut.get_compute())
self.assertTrue(aut.faulty)
# Check if user created compress_hash was used
aut = PHF_DFA()
a = jenkins_compress(4)
a.generate_seed()
aut.enable_faulty_transitions(4, compress_hash = a)
self.assertEqual(aut.compress_bits, 4)
self.assertEqual(aut.compress_hash, a)
self.assertFalse(aut.get_compute())
self.assertTrue(aut.faulty)
# Check if the hash output size is really 4 bits for few inputs
for i in range(0, 255):
val = aut.compress_hash.hash(([chr(i)],[chr(0)]))
self.assertTrue(val < 16 and val >= 0)
def test_validate_transition(self):
"""validate_transition()"""
# Test correct transition validation for both faulty and non-faulty
# transition table.
nfaData = nfa_data()
nfaData.states[0] = b_State(0,set())
nfaData.states[1] = b_State(1,set())
nfaData.states[2] = b_State(2,set())
nfaData.states[3] = b_State(3,set([0]))
nfaData.alphabet[0] = b_Sym_char("a", "a", 0)
nfaData.alphabet[1] = b_Sym_char("b", "b", 1)
nfaData.alphabet[2] = b_Sym_char("c", "c", 2)
nfaData.start = 0
nfaData.transitions.add( (0,0,1) )
nfaData.transitions.add( (1,1,2) )
nfaData.transitions.add( (2,2,3) )
nfaData.final.add(3)
aut = PHF_DFA()
a = bdz()
a.set_limit(128)
aut.set_PHF_class(a)
aut.create_from_nfa_data(nfaData)
aut.compute()
for t in aut._automaton1.transitions: # all transitions must be valid
self.assertTrue(aut.validate_transition(aut._transition_rep(t)))
# some nonexistent transitions -> invalid
t = (0,2,0)
self.assertFalse(aut.validate_transition(aut._transition_rep(t)))
t = (1,0,2)
self.assertFalse(aut.validate_transition(aut._transition_rep(t)))
t = (len(aut._automaton1.states), len(aut._automaton1.alphabet), 0)
self.assertFalse(aut.validate_transition(aut._transition_rep(t)))
t = (0, len(aut._automaton1.alphabet), 0)
self.assertFalse(aut.validate_transition(aut._transition_rep(t)))
t = (len(aut._automaton1.states), 0, 0)
self.assertFalse(aut.validate_transition(aut._transition_rep(t)))
# faulty transitions
aut.enable_faulty_transitions(32)
aut.compute()
for t in aut._automaton1.transitions: # all transitions must be valid
self.assertTrue(aut.validate_transition(aut._transition_rep(t)))
# some nonexistent transitions -> invalid, collisions are improbable
t = (0,2,0)
self.assertFalse(aut.validate_transition(aut._transition_rep(t)))
t = (1,0,2)
self.assertFalse(aut.validate_transition(aut._transition_rep(t)))
t = (10,10,1)
self.assertFalse(aut.validate_transition(aut._transition_rep(t)))
t = (11,11,1)
self.assertFalse(aut.validate_transition(aut._transition_rep(t)))
t = (12,12,1)
self.assertFalse(aut.validate_transition(aut._transition_rep(t)))
def test_generate_PHF_table(self):
"""generate_PHF_table()"""
# Test of PHF table generation - the right size of tabel, every
# transition is exactly once in the table and on the right index.
nfaData = nfa_data()
nfaData.states[0] = b_State(0,set())
nfaData.states[1] = b_State(1,set())
nfaData.states[2] = b_State(2,set())
nfaData.states[3] = b_State(3,set([0]))
nfaData.alphabet[0] = b_Sym_char("a", "a", 0)
nfaData.alphabet[1] = b_Sym_char("b", "b", 1)
nfaData.alphabet[2] = b_Sym_char("c", "c", 2)
nfaData.start = 0
nfaData.transitions.add( (0,0,1) )
nfaData.transitions.add( (0,1,0) )
nfaData.transitions.add( (0,2,0) )
nfaData.transitions.add( (1,1,2) )
nfaData.transitions.add( (1,0,1) )
nfaData.transitions.add( (1,2,0) )
nfaData.transitions.add( (2,2,3) )
nfaData.transitions.add( (2,0,1) )
nfaData.transitions.add( (2,1,0) )
nfaData.transitions.add( (3,0,3) )
nfaData.transitions.add( (3,1,3) )
nfaData.transitions.add( (3,2,3) )
nfaData.final.add(3)
aut = PHF_DFA()
a = bdz()
a.set_limit(128)
aut.set_PHF_class(a)
aut._automaton1 = nfaData
aut.generate_PHF_table()
# transition table size
self.assertEqual(aut.ran, len(aut.trans_table))
self.assertEqual(aut.ran, 384)
# count number of unique lines in transition table
tranCount = dict()
for l in aut.trans_table:
tranCount.setdefault(l[1], 0)
tranCount[l[1]] += 1
# test if every automaton transition is just once in the table
for t in aut._automaton1.transitions:
self.assertEqual(tranCount[aut._transition_rep(t)], 1)
t = ([2 ** aut.state_bits - 1, 2 ** aut.symbol_bits - 1, 0])
# rest of trans are the nonexistent transitions
self.assertEqual(tranCount[aut._transition_rep(t)], aut.ran - len(aut._automaton1.transitions))
# check if each transition is on its index returned by hash function
for t in aut._automaton1.transitions:
rep = aut._transition_rep(t)
self.assertEqual(rep, aut.trans_table[aut.hash_function.hash(rep)][1])
# test the representation in faulty table
aut.enable_faulty_transitions(8)
aut.generate_PHF_table()
for t in aut._automaton1.transitions:
rep = aut._transition_rep(t)
self.assertEqual(aut.compress_hash.hash(rep), aut.trans_table[aut.hash_function.hash(rep)][3])
# change the size of PHF table and repeat tests
aut = PHF_DFA()
a = bdz()
a.set_ratio(6.0)
a.set_iteration_limit(10)
aut.set_PHF_class(a)
aut._automaton1 = nfaData
aut.generate_PHF_table()
# transition table size
self.assertEqual(aut.ran, len(aut.trans_table))
self.assertEqual(aut.ran, 72)
# count number of unique lines in transition table
tranCount = dict()
for l in aut.trans_table:
tranCount.setdefault(l[1], 0)
tranCount[l[1]] += 1
# test if every automaton transition is just once in the table
for t in aut._automaton1.transitions:
self.assertEqual(tranCount[aut._transition_rep(t)], 1)
t = ([2 ** aut.state_bits - 1, 2 ** aut.symbol_bits - 1, 0])
# rest of trans are the nonexistent transitions
self.assertEqual(tranCount[aut._transition_rep(t)], aut.ran - len(aut._automaton1.transitions))
# check if each transition is on its index returned by hash function
for t in aut._automaton1.transitions:
rep = aut._transition_rep(t)
self.assertEqual(rep, aut.trans_table[aut.hash_function.hash(rep)][1])
# test the representation in faulty table
aut.enable_faulty_transitions(8)
aut.generate_PHF_table()
for t in aut._automaton1.transitions:
rep = aut._transition_rep(t)
self.assertEqual(aut.compress_hash.hash(rep), aut.trans_table[aut.hash_function.hash(rep)][3])
# RE /#include.*>/ and enable fallback_state
par = pcre_parser()
par.set_text("/#include.*>/s")
aut = PHF_DFA()
a = bdz()
a.set_ratio(2.5)
a.set_iteration_limit(10)
aut.set_PHF_class(a)
aut.create_by_parser(par)
aut.enable_fallback_state(warning=False)
aut.compute()
# transition table size
self.assertEqual(aut.ran, len(aut.trans_table))
self.assertEqual(aut.ran, 90)
# count number of unique lines in transition table
tranCount = dict()
for l in aut.trans_table:
tranCount.setdefault(l[1], 0)
tranCount[l[1]] += 1
# test if every automaton transition is just once in the table
for t in aut._automaton1.transitions:
self.assertEqual(tranCount[aut._transition_rep(t)], 1)
t = ([2 ** aut.state_bits - 1, 2 ** aut.symbol_bits - 1, 0])
# rest of trans are the nonexistent transitions
self.assertEqual(tranCount[aut._transition_rep(t)], aut.ran - len(aut._automaton1.transitions))
# check if each transition is on its index returned by hash function
for t in aut._automaton1.transitions:
rep = aut._transition_rep(t)
self.assertEqual(rep, aut.trans_table[aut.hash_function.hash(rep)][1])
# test the representation in faulty table
aut.enable_faulty_transitions(8)
aut.generate_PHF_table()
for t in aut._automaton1.transitions:
rep = aut._transition_rep(t)
self.assertEqual(aut.compress_hash.hash(rep), aut.trans_table[aut.hash_function.hash(rep)][3])
# disable fallback_state
aut.disable_fallback_state()
aut.compute()
self.assertEqual(aut.ran, len(aut.trans_table))
self.assertEqual(aut.ran, 252)
# count number of unique lines in transition table
tranCount = dict()
for l in aut.trans_table:
tranCount.setdefault(l[1], 0)
tranCount[l[1]] += 1
# test if every automaton transition is just once in the table
for t in aut._automaton1.transitions:
self.assertEqual(tranCount[aut._transition_rep(t)], 1)
t = ([2 ** aut.state_bits - 1, 2 ** aut.symbol_bits - 1, 0])
# rest of trans are the nonexistent transitions
self.assertEqual(tranCount[aut._transition_rep(t)], aut.ran - len(aut._automaton1.transitions))
# check if each transition is on its index returned by hash function
for t in aut._automaton1.transitions:
rep = aut._transition_rep(t)
self.assertEqual(rep, aut.trans_table[aut.hash_function.hash(rep)][1])
# test the representation in faulty table
aut.enable_faulty_transitions(8)
aut.generate_PHF_table()
for t in aut._automaton1.transitions:
rep = aut._transition_rep(t)
self.assertEqual(aut.compress_hash.hash(rep), aut.trans_table[aut.hash_function.hash(rep)][3])
def test_report_memory_naive(self):
"""report_memory_naive()"""
# Simple regression test for small automaton.
nfaData = nfa_data()
nfaData.states[0] = b_State(0,set())
nfaData.states[1] = b_State(1,set())
nfaData.states[2] = b_State(2,set())
nfaData.states[3] = b_State(3,set([0]))
nfaData.alphabet[0] = b_Sym_char("a", "a", 0)
nfaData.alphabet[1] = b_Sym_char("b", "b", 1)
nfaData.alphabet[2] = b_Sym_char("c", "c", 2)
nfaData.start = 0
nfaData.transitions.add( (0,0,1) )
nfaData.transitions.add( (1,1,2) )
nfaData.transitions.add( (2,2,3) )
nfaData.final.add(3)
aut = PHF_DFA()
a = bdz()
a.set_limit(128)
aut.set_PHF_class(a)
aut.create_from_nfa_data(nfaData)
aut.compute()
self.assertEqual(aut.report_memory_naive(), 12)
# Test after removing fallback transitions. report_memory_naive depends
# on number of states and symbols, not transitions, so nothing changes
aut.enable_fallback_state(1, warning=False)
aut.remove_fallback_transitions()
self.assertEqual(aut.report_memory_naive(), 12)
# Manually remove symbol and state from _automaton1
del aut._automaton1.states[2]
del aut._automaton1.alphabet[2]
self.assertEqual(aut.report_memory_naive(), 6)
def test_search(self):
"""search()"""
# 1. RE /^abc/
nfaData = nfa_data()
nfaData.states[0] = b_State(0,set())
nfaData.states[1] = b_State(1,set())
nfaData.states[2] = b_State(2,set())
nfaData.states[3] = b_State(3,set([0]))
nfaData.alphabet[0] = b_Sym_char("a", "a", 0)
nfaData.alphabet[1] = b_Sym_char("b", "b", 1)
nfaData.alphabet[2] = b_Sym_char("c", "c", 2)
nfaData.start = 0
nfaData.transitions.add( (0,0,1) )
nfaData.transitions.add( (1,1,2) )
nfaData.transitions.add( (2,2,3) )
nfaData.final.add(3)
aut = PHF_DFA()
a = bdz()
a.set_limit(128)
aut.set_PHF_class(a)
aut.create_from_nfa_data(nfaData)
aut.compute()
self.assertEqual(aut.search("abc"), [1])
self.assertEqual(aut.search("aaaaaaaaaaaaaabc"), [0])
self.assertEqual(aut.search("ccccbbbabc"), [0])
self.assertEqual(aut.search("ababc"), [0])
self.assertEqual(aut.search("d"), [0])
self.assertEqual(aut.search("cbabbacba"), [0])
# 2. RE /abc/
nfaData = nfa_data()
nfaData.states[0] = b_State(0,set())
nfaData.states[1] = b_State(1,set())
nfaData.states[2] = b_State(2,set())
nfaData.states[3] = b_State(3,set([0]))
nfaData.alphabet[0] = b_Sym_char("a", "a", 0)
nfaData.alphabet[1] = b_Sym_char("b", "b", 1)
nfaData.alphabet[2] = b_Sym_char("c", "c", 2)
nfaData.start = 0
nfaData.transitions.add( (0,0,1) )
nfaData.transitions.add( (0,1,0) )
nfaData.transitions.add( (0,2,0) )
nfaData.transitions.add( (1,1,2) )
nfaData.transitions.add( (1,0,1) )
nfaData.transitions.add( (1,2,0) )
nfaData.transitions.add( (2,2,3) )
nfaData.transitions.add( (2,0,1) )
nfaData.transitions.add( (2,1,0) )
nfaData.transitions.add( (3,0,3) )
nfaData.transitions.add( (3,1,3) )
nfaData.transitions.add( (3,2,3) )
nfaData.final.add(3)
aut = PHF_DFA()
a = bdz()
a.set_limit(128)
aut.set_PHF_class(a)
aut.create_from_nfa_data(nfaData)
aut.compute()
self.assertEqual(aut.search("abc"), [1])
self.assertEqual(aut.search("aaaaaaaaaaaaaabc"), [1])
self.assertEqual(aut.search("ccccbbbabc"), [1])
self.assertEqual(aut.search("ababc"), [1])
self.assertEqual(aut.search("d"), [0])
self.assertEqual(aut.search("cbabbacba"), [0])
# 2a. same test with faulty transitions
aut.enable_faulty_transitions(32)
aut.compute()
self.assertEqual(aut.search("abc"), [1])
self.assertEqual(aut.search("aaaaaaaaaaaaaabc"), [1])
self.assertEqual(aut.search("ccccbbbabc"), [1])
self.assertEqual(aut.search("ababc"), [1])
self.assertEqual(aut.search("d"), [0])
self.assertEqual(aut.search("cbabbacba"), [0])
# 3. RE /#include.*>/ with enable_fallback_state
par = pcre_parser()
par.set_text("/#include.*>/")
aut = PHF_DFA()
a = bdz()
a.set_ratio(2.5)
a.set_iteration_limit(10)
aut.set_PHF_class(a)
aut.create_by_parser(par)
aut.enable_fallback_state(warning=False)
aut.compute()
self.assertEqual(aut.search("#include <stdio.h>"), [1])
self.assertEqual(aut.search("#include <stdlib.h>"), [1])
self.assertEqual(aut.search("#include <stdio.h>bba"), [1])
self.assertEqual(aut.search('#include "pcre.h"'), [0])
self.assertEqual(aut.search('asdf#include <stdio.h>'), [1])
def test_get_trans_num(self):
"""get_trans_num()"""
# Simple regression test for small automaton.
nfaData = nfa_data()
nfaData.states[0] = b_State(0,set())
nfaData.states[1] = b_State(1,set())
nfaData.states[2] = b_State(2,set())
nfaData.states[3] = b_State(3,set([0]))
nfaData.alphabet[0] = b_Sym_char("a", "a", 0)
nfaData.alphabet[1] = b_Sym_char("b", "b", 1)
nfaData.alphabet[2] = b_Sym_char("c", "c", 2)
nfaData.start = 0
nfaData.transitions.add( (0,0,1) )
nfaData.transitions.add( (1,1,2) )
nfaData.transitions.add( (2,2,3) )
nfaData.final.add(3)
aut = PHF_DFA()
a = bdz()
a.set_limit(128)
aut.set_PHF_class(a)
aut.create_from_nfa_data(nfaData)
aut.compute()
self.assertEqual(aut.get_trans_num(), 3)
# Test after removing fallback transitions
aut.enable_fallback_state(1, warning=False)
aut.remove_fallback_transitions()
self.assertEqual(aut.get_trans_num(), 2)
print("-------------------------------------------------------------------")
print(" Example of use: PHF DFA ")
print("-------------------------------------------------------------------")
print(" Ruleset: /#include.*>/ ")
print(" Faulty Table: No ")
print(" State bits: 10 ")
print(" Symbol bits: 12 ")
print(" Fallback State: No ")
print("-------------------------------------------------------------------")
# create parser and load RE
parser = pcre_parser()
parser.set_text("/#include.*>/")
# create phf_dfa automaton
aut = PHF_DFA()
aut.create_by_parser(parser)
# redefine default PHF class so table generation won't fail in this script
# it's not important right now, more about that later
a = bdz()
a.set_ratio(2.0)
aut.set_PHF_class(a)
# compute dfa and PHF table
aut.compute()
# memory used by PHF table
print "Memory used (Real):", aut.report_memory_real(), "B"
# Print number of symbols, states and transitions
def test():
"""
Run searching using pcregrep and PHF_DFA. Prints out the results.
"""
# parse options
usage = "usage: %prog rules.pcre pcap_dir/ [options]"
optparser = OptionParser(usage=usage)
optparser.add_option("-O", "--outputfile", dest="resultfile", help="output file for results, default is stdout")
optparser.add_option("-P", "--PerPacket", dest="PerPacket", action="store_true", default=False,
help="compare nonfaulty matching for flows and packets, faulty algorithm is used only with flows")
optparser.add_option("-s", "--showprogress", dest="progress", action="store_true", default=False,
help="show progress of computation")
optparser.add_option("-C", "--count", dest="maxiter", type="int", default="1", help="number of test iterations")
optparser.add_option("-F", "--faulty", dest="FAULTY", type="int", default="0", help="number of bits for compress hash, default is 0 (no faulty transitions)")
optparser.add_option("-D", "--debuglevel", dest="DEBUG", type="int", default="0", help="debug output level (0-2)")
optparser.add_option("-S", "--savefile", dest="savefile", default="", metavar="FILE", help="save nfa_data in FILE")
optparser.add_option("-L", "--loadfile", dest="autfile", default="", metavar="FILE", help="load nfa_data from FILE")
optparser.add_option("-N", "--nonfaulty", dest="NonFaulty", action="store_true", default=False,
help="try to generate PHF table without collisions, therefore ensure nonfaulty matching. Experimental code. "
"May take a long time with small compress hash output.")
(options, args) = optparser.parse_args()
global FAULTY, DEBUG
if len(args) != 2:
print "You must specify rules.pcre and pcap_dir/"
optparser.print_usage()
exit(1)
rulesfile, inputdir = args
PerPacket, resultfile, maxiter, autfile, savefile, FAULTY, DEBUG = options.PerPacket, options.resultfile, options.maxiter, options.autfile, options.savefile, options.FAULTY, options.DEBUG
progress = options.progress
NonFaulty = options.NonFaulty
if inputdir[-1] == "/":
inputdir = inputdir[:-1] # remove '/' from the end
rules = open(rulesfile, 'rb')
if PerPacket:
packetdir = inputdir + "/packets"
inputdir = inputdir + "/flows"
if resultfile:
sys.stdout = open(resultfile, 'a')
totalhits, totalfp, totalfn = (0, 0, 0)
iter = 0
while iter != maxiter:
if progress:
print >>sys.stderr, "\r", ' '*80, '\r',"pcregrep",
if not iter:
# prepare pcregrep
p = subprocess.Popen("cd pcre-8.20/ && make pcregrep", shell=True, stdout=subprocess.PIPE)
p.wait()
results = dict()
file_list = list()
rule_count = len(open(rulesfile).readlines())
for root, dirs, files in os.walk(inputdir):
for i in files:
i = os.path.join(root, i)
file_list.append(i)
if not iter:
results[i] = [rule_count*[0],rule_count*[0],rule_count*[0]]
else:
results[i][0] = rule_count*[0]
#results = init_results
rule_num = 0
grep_reg_exp = "grep_reg_exp." + str(os.getpid())
for rule in rules:
if not iter:
if DEBUG:
print rule,
(grep_rule, grep_params) = parse_rule(rule)
f = open(grep_reg_exp, 'w')
f.write(grep_rule)
f.close()
p = subprocess.Popen("pcre-8.20/pcregrep --buffer-size 50000 --color=auto -N ANYCRLF" + grep_params + " -r -l -f " + grep_reg_exp + " " + inputdir, shell=True, stdout=subprocess.PIPE)
p.wait()
for out in p.stdout:
item = out.split()[0]
results[item][1][rule_num] = 1
if PerPacket:
p = subprocess.Popen("pcre-8.20/pcregrep --buffer-size 50000 --color=auto -N ANYCRLF" + grep_params + " -r -l -f " + grep_reg_exp + " " + packetdir, shell=True, stdout=subprocess.PIPE)
p.wait()
for out in p.stdout:
item = inputdir + "/" + out.split()[0].split("-")[1].replace("_", "/")
results[item][2][rule_num] = 1
rule_num += 1
try:
os.remove(grep_reg_exp)
except:
pass
if progress:
print >>sys.stderr, "\r", ' '*80, '\r', "create automaton",
#aut = b_Automaton()
aut = PHF_DFA()
if autfile:
aut.create_from_nfa_data(nfa_data().load_from_file(autfile))
else:
par = parser("pcre_parser")
#par.set_text(rule)
par.load_file(rulesfile)
aut.create_by_parser(par)
if DEBUG:
aut.show("NFA.dot")
#aut.remove_epsilons()
if progress:
print >>sys.stderr, "\r", ' '*80, '\r', "resolve alphabet",
aut.resolve_alphabet()
if progress:
print >>sys.stderr, "\r", ' '*80, '\r', "determinise",
aut.determinise()
if progress:
print >>sys.stderr, "\r", ' '*80, '\r', "minimise",
aut.minimise()
if DEBUG:
aut.show("DFA.dot")
if savefile:
aut._automaton.save_to_file(savefile)
aut._automaton1 = aut._automaton
aut.set_table_parameters((20,10))
if DEBUG > 1:
print "Without fallback state:"
print "Symbols:", len(aut._automaton.alphabet)
print "States:", len(aut._automaton.states)
print "Transitions:", aut.get_trans_num(), float(aut.get_trans_num()) / (aut.get_state_num() * aut.get_alpha_num()) * 100, "%"
if isinstance(aut, PHF_DFA):
if progress:
print >>sys.stderr, "\r", ' '*80, '\r', "generate PHF",
if aut.get_trans_num() == (aut.get_state_num() * aut.get_alpha_num()):
aut.enable_fallback_state(warning=False)
if FAULTY:
aut.enable_faulty_transitions(FAULTY)
if NonFaulty:
aut.enable_faulty_check()
aut.compute()
if DEBUG:
print "Fallback state:", aut.fallback_state
print "Symbols:", len(aut._automaton.alphabet)
print "States:", len(aut._automaton.states)
print "Transitions:", aut.get_trans_num(), float(aut.get_trans_num()) / (aut.get_state_num() * aut.get_alpha_num()) * 100, "%"
count = 1
all = len(file_list)
if progress:
print >> sys.stderr, '\r' + 80*' ' + '\r',
for f in file_list:
# progress
if progress:
print >> sys.stderr, '\r',
print >> sys.stderr, str(iter+1)+'/'+str(maxiter)+ ":", count, '/', all,
# sys.stderr.flush()
count += 1
data = open(f, 'rb').read()
results[f][0] = aut.search(data)
if progress:
print >>sys.stderr, "\r", ' '*80, '\r', "compare results",
if isinstance(aut, PHF_DFA) and DEBUG:
if DEBUG > 1:
print "List of collisions:"
print aut.collisions
for tran, i in aut.collisions.iteritems():
#print tran, i
print BitArray(bytes=tran[0], length=aut.symbol_bits).uint, BitArray(bytes=tran[1], length=aut.state_bits).uint, i
print "SYM:", aut._automaton.alphabet[BitArray(bytes=tran[0], length=aut.symbol_bits).uint]
print "Bad transitions:", aut.bad_transitions
print "Collisions:", len(aut.collisions)
print "Compress bits:", aut.compress_bits
stats = compare_results(results)
stats = list(stats)
if stats[0] == 0:
print "Zero hits, cannot compute F-measure!"
stats[0] = 1
if DEBUG:
print "Total number of searched packets/flows:", stats[3]
print "Hits:", stats[0]
totalhits += stats[0]
totalfp += stats[1]
totalfn += stats[2]
precis = float(stats[0])/(stats[1]+stats[0])
recall = float(stats[0])/(stats[0]+stats[2])
fmeas = 2* precis * recall / (precis + recall)
print "False positives:", stats[1], precis*100, "%"
print "False negatives:", stats[2], recall*100, "%"
print "F-measure:", fmeas*100, "%"
if PerPacket:
print "Per packet errors:", stats[4], stats[5]
print '-'*80
iter += 1
print "Total stats:"
precis = float(totalhits)/(totalfp + totalhits)
recall = float(totalhits)/(totalfn + totalhits)
fmeas = 2* precis * recall / (precis + recall)
print "Hits:", totalhits
print "False positives:", totalfp, precis*100, "%"
print "False negatives:", totalfn, recall*100, "%"
print "F-measure:", fmeas*100, "%"
print "_"*80