def get_transition_function_and_states_and_symbols_non_reduced(functions):
states = set()
symbols = set()
start_state = State("Start")
states.add(start_state)
final_states = set()
counter = 0
transition_function = NondeterministicTransitionFunction()
linear_paths = get_all_linear_paths(functions)
linear_paths = sorted(linear_paths, key=len)
for linear_path in linear_paths:
current_state = start_state
for i, atom in enumerate(linear_path):
symbol = Symbol(atom)
symbols.add(symbol)
next_state = State(str(counter))
states.add(next_state)
counter += 1
transition_function.add_transition(current_state, symbol,
next_state)
current_state = next_state
final_states.add(current_state)
for final_state in final_states:
transition_function.add_transition(final_state,
finite_automaton.Epsilon(),
start_state)
return transition_function, states, symbols, final_states
def test_to_enfa0(self):
""" Tests the transformation to a regex """
symb_a = finite_automaton.Symbol("a")
symb_b = finite_automaton.Symbol("b")
symb_c = finite_automaton.Symbol("c")
epsilon = finite_automaton.Epsilon()
regex = Regex("a|b")
enfa = regex.to_epsilon_nfa()
self.assertTrue(enfa.accepts([symb_a]))
self.assertTrue(enfa.accepts([symb_b]))
self.assertFalse(enfa.accepts([symb_c]))
self.assertFalse(enfa.accepts([epsilon]))
self.assertFalse(enfa.accepts([symb_a, symb_b]))
regex = Regex("a b")
enfa = regex.to_epsilon_nfa()
self.assertFalse(enfa.accepts([symb_a]))
self.assertFalse(enfa.accepts([symb_b]))
self.assertTrue(enfa.accepts([symb_a, symb_b]))
regex = Regex("a b c")
enfa = regex.to_epsilon_nfa()
self.assertFalse(enfa.accepts([symb_a, symb_b]))
self.assertTrue(enfa.accepts([symb_a, symb_b, symb_c]))
self.assertFalse(enfa.accepts([symb_a, symb_b, symb_a]))
regex = Regex("(a b)|c")
enfa = regex.to_epsilon_nfa()
self.assertTrue(enfa.accepts([symb_a, symb_b]))
self.assertFalse(enfa.accepts([symb_a, symb_c]))
self.assertFalse(enfa.accepts([symb_b, symb_c]))
self.assertTrue(enfa.accepts([symb_c]))
regex = Regex("")
enfa = regex.to_epsilon_nfa()
self.assertFalse(enfa.accepts([symb_a]))
self.assertFalse(enfa.accepts([symb_b]))
self.assertFalse(enfa.accepts([symb_c]))
self.assertFalse(enfa.accepts([]))
regex = Regex("a*")
enfa = regex.to_epsilon_nfa()
self.assertTrue(enfa.accepts([symb_a]))
self.assertTrue(enfa.accepts([]))
self.assertTrue(enfa.accepts([symb_a, symb_a]))
self.assertTrue(enfa.accepts([symb_a, symb_a, symb_a]))
def get_folded_automaton(enfa):
in_edges = dict()
out_edges = dict()
processed = set()
to_process = []
final_states = enfa.get_final_states()
start_states = enfa.get_start_states()
alphabet = enfa.get_symbols()
states = enfa.get_states()
star = Symbol("STAR")
for state in states:
in_edges[state] = dict()
out_edges[state] = dict()
for a in alphabet:
in_edges[state][a] = []
out_edges[state][a] = []
# Construction
for from_state in states:
for symbol in alphabet:
for next_state in enfa(from_state, symbol):
if from_state not in in_edges[next_state][symbol]:
in_edges[next_state][symbol].append(from_state)
if next_state not in out_edges[from_state][symbol]:
out_edges[from_state][symbol].append(next_state)
# Initialization
# From final state
for final_state in final_states:
if not in_edges[final_state][star]:
continue
first = in_edges[final_state][star][0]
for relation in in_edges[first]:
if relation == star:
continue
inverse_relation = Symbol(get_inverse_relation(relation.value))
for previous_state in in_edges[first][relation]:
for start_state in start_states:
for next_state in out_edges[start_state].get(
inverse_relation, []):
# previous_state -- relation --> first -- STAR -->
# final_state -- epsilon --> start_state -- inverse -->
# next_state
to_process.append((previous_state, next_state))
processed.add((previous_state, next_state))
# From other states
for state in states:
for relation in out_edges[state]:
if relation == star:
continue
inverse = Symbol(get_inverse_relation(relation.value))
for middle_state in out_edges[state][relation]:
for middle_state2 in out_edges[middle_state][star]:
if inverse not in out_edges[middle_state2]:
continue
for next_state in out_edges[middle_state2][inverse]:
to_process.append((state, next_state))
processed.add((state, next_state))
while to_process:
first, second = to_process.pop()
# first -- epsilon --> second
if first == second:
continue
# Special cases!
if first in start_states:
for final_state in final_states:
if (final_state, second) not in processed:
processed.add((final_state, second))
to_process.append((final_state, second))
if second in final_states:
for start_state in start_states:
if (first, start_state) not in processed:
processed.add((first, start_state))
to_process.append((first, start_state))
# Reduction only if between two stars
if len(out_edges[second][star]) == 0 or \
len(in_edges[first][star]) == 0:
continue
# Only one possibility normally
first = in_edges[first][star][0]
second = out_edges[second][star][0]
# Apply L -> a- L a
for a in out_edges[second]:
if a == star:
continue
for state in out_edges[second][a]:
# second -- a ---> state
opposite = Symbol(get_inverse_relation(a.value))
if opposite in alphabet:
for begin in in_edges[first][opposite]:
# begin -- a- ---> first
if (begin, state) not in processed:
to_process.append((begin, state))
processed.add((begin, state))
# Apply L -> L * L
for next_state in out_edges[second][star]:
for state in states:
if (next_state, state) in processed:
if (first, state) not in processed:
processed.add((first, state))
to_process.append((first, state))
for previous_state in in_edges[first][star]:
for state in states:
if (state, previous_state) in processed:
if (state, second) not in processed:
processed.add((state, second))
to_process.append((state, second))
new_enfa = enfa.copy()
epsilon = finite_automaton.Epsilon()
for first, second in processed:
new_enfa.add_transition(first, epsilon, second)
# For conveniant reasons, we also transform L -> * L, when not final...
if second not in final_states:
for previous_star in in_edges[first][star]:
new_enfa.add_transition(previous_star, epsilon, second)
return new_enfa
def _add_epsilon_transition_in_enfa_between(self, state0, state1):
self._enfa.add_transition(state0, finite_automaton.Epsilon(), state1)
def intersection(self, other: Any) -> "PDA":
""" Gets the intersection of the current PDA with something else
Equivalent to:
>> pda and regex
Parameters
----------
other : any
The other part of the intersection
Returns
----------
new_pda : :class:`~pyformlang.pda.PDA`
The pda resulting of the intersection
Raises
----------
NotImplementedError
When intersecting with something else than a regex or a finite
automaton
"""
if isinstance(other, Regex):
other = other.to_epsilon_nfa().to_deterministic()
elif isinstance(other, FiniteAutomaton):
if not other.is_deterministic():
other = other.to_deterministic()
else:
raise NotImplementedError
start_state_other = other.start_states
if len(start_state_other) == 0:
return PDA()
pda_state_converter = _PDAStateConverter(self._states, other.states)
start_state_other = list(start_state_other)[0]
final_state_other = other.final_states
start = pda_state_converter.to_pda_combined_state(
self._start_state, start_state_other)
pda = PDA(start_state=start,
start_stack_symbol=self._start_stack_symbol)
symbols = self._input_symbols.copy()
symbols.add(Epsilon())
to_process = [(self._start_state, start_state_other)]
processed = {(self._start_state, start_state_other)}
while to_process:
state_in, state_dfa = to_process.pop()
if (state_in in self._final_states
and state_dfa in final_state_other):
pda.add_final_state(
pda_state_converter.to_pda_combined_state(
state_in, state_dfa))
for symbol in symbols:
if symbol == Epsilon():
symbol_dfa = finite_automaton.Epsilon()
else:
symbol_dfa = finite_automaton.Symbol(symbol.value)
if symbol == Epsilon():
next_states_dfa = [state_dfa]
else:
next_states_dfa = other(state_dfa, symbol_dfa)
if len(next_states_dfa) == 0:
continue
for stack_symbol in self._stack_alphabet:
next_states_self = self._transition_function(
state_in, symbol, stack_symbol)
for next_state, next_stack in next_states_self:
for next_state_dfa in next_states_dfa:
pda.add_transition(
pda_state_converter.to_pda_combined_state(
state_in, state_dfa), symbol, stack_symbol,
pda_state_converter.to_pda_combined_state(
next_state, next_state_dfa), next_stack)
if (next_state, next_state_dfa) not in processed:
to_process.append((next_state, next_state_dfa))
processed.add((next_state, next_state_dfa))
return pda
def intersection(self, other: Any) -> "PDA":
""" Gets the intersection of the language L generated by the \
current PDA when accepting by final state with something else
Currently, it only works for regular languages (represented as \
regular expressions or finite automata) as the intersection \
between two PDAs is not context-free (it cannot be represented \
with a PDA).
Equivalent to:
>> pda and regex
Parameters
----------
other : any
The other part of the intersection
Returns
----------
new_pda : :class:`~pyformlang.pda.PDA`
The pda resulting of the intersection
Raises
----------
NotImplementedError
When intersecting with something else than a regex or a finite
automaton
"""
if isinstance(other, regular_expression.Regex):
enfa = other.to_epsilon_nfa()
other = enfa.to_deterministic()
elif isinstance(other, FiniteAutomaton):
is_deterministic = other.is_deterministic()
if not is_deterministic:
other = other.to_deterministic()
else:
raise NotImplementedError
start_state_other = other.start_states
if len(start_state_other) == 0:
return PDA()
pda_state_converter = _PDAStateConverter(self._states, other.states)
start_state_other = list(start_state_other)[0]
final_state_other = other.final_states
start = pda_state_converter.to_pda_combined_state(self._start_state,
start_state_other)
pda = PDA(start_state=start,
start_stack_symbol=self._start_stack_symbol)
symbols = self._input_symbols.copy()
symbols.add(Epsilon())
to_process = [(self._start_state, start_state_other)]
processed = {(self._start_state, start_state_other)}
while to_process:
state_in, state_dfa = to_process.pop()
if (state_in in self._final_states and state_dfa in
final_state_other):
pda.add_final_state(
pda_state_converter.to_pda_combined_state(state_in,
state_dfa))
for symbol in symbols:
if symbol == Epsilon():
symbol_dfa = finite_automaton.Epsilon()
else:
symbol_dfa = finite_automaton.Symbol(symbol.value)
if symbol == Epsilon():
next_states_dfa = [state_dfa]
else:
next_states_dfa = other(state_dfa, symbol_dfa)
if len(next_states_dfa) == 0:
continue
for stack_symbol in self._stack_alphabet:
next_states_self = self._transition_function(state_in,
symbol,
stack_symbol)
for next_state, next_stack in next_states_self:
for next_state_dfa in next_states_dfa:
pda.add_transition(
pda_state_converter.to_pda_combined_state(
state_in,
state_dfa),
symbol,
stack_symbol,
pda_state_converter.to_pda_combined_state(
next_state,
next_state_dfa),
next_stack)
if (next_state, next_state_dfa) not in processed:
to_process.append((next_state, next_state_dfa))
processed.add((next_state, next_state_dfa))
return pda