def explainEX(fsm, state, a):
"""
Explain why `state` of `fsm` satisfies `EX phi`, where `a` is
the set of states of `fsm` satisfying `phi`, represented by a BDD.
:param fsm: the system
:type fsm: :class:`BddFsm <pynusmv.fsm.BddFsm>`
:param state: a state of `fsm`
:type state: :class:`State <pynusmv.dd.State>`
:param a: the set of states of `fsm` satisfying `phi`
:type a: :class:`BDD <pynusmv.dd.BDD>`
Return `(s, i, s')` tuple where `s` (:class:`State <pynusmv.dd.State>`)
is the given state, `s'` (:class:`State <pynusmv.dd.State>`) is a successor
of `s` belonging to `a` and `i` (:class:`Inputs <pynusmv.dd.Inputs>`)
is the inputs to go from `s` to `s'` in `fsm`.
"""
enc = fsm.bddEnc
manager = enc.DDmanager
path = nsnode.cons(nsnode.bdd2node(nsdd.bdd_dup(state._ptr)), None)
bddlist = BDDList(nsmc.ex_explain(fsm._ptr, enc._ptr, path, a._ptr),
manager)
# bddlist is reversed!
statep = State.from_bdd(bddlist[0], fsm)
inputs = Inputs.from_bdd(bddlist[1], fsm)
state = State.from_bdd(bddlist[2], fsm)
return (state, inputs, statep)
def from_tuple(bddtuple):
"""
Create a node-based list from the Python tuple `bddtuple`.
:param bddtuple: a Python tuple of BDDs
Return a :class:`BDDList` representing the given tuple,
using NuSMV nodes.
All BDDs are assumed from the same DD manager;
the created list contains the DD manager of the first non-`None` BDD.
If all elements of `bddtuple` are `None`,
the manager of the created :class:`BDDList` is `None`.
"""
# Reverse tuple before, because we build the list reversely.
bddtuple = bddtuple[::-1]
nodes = None
manager = None
for elem in bddtuple:
if elem:
enode = nsnode.bdd2node(nsdd.bdd_dup(elem._ptr))
if manager is None:
manager = elem._manager
else:
enode = elem
nodes = nsnode.cons(enode, nodes)
return BDDList(nodes, manager, freeit=True)
def __getitem__(self, val):
"""
Return the BDD stored at val.
:param val: the index requested OR a slice.
.. note:: cannot access elements with negative indices.
"""
if isinstance(val, int):
if val < 0:
raise IndexError("BDDList index out of range")
ptr = self._ptr
while val > 0:
if ptr is None:
raise IndexError("BDDList index out of range")
val -= 1
ptr = nsnode.cdr(ptr)
if ptr is None:
raise IndexError("BDDList index out of range")
bdd_ptr = nsnode.node2bdd(nsnode.car(ptr))
if bdd_ptr is not None:
return BDD(nsdd.bdd_dup(bdd_ptr), self._manager, freeit=True)
else:
return None
elif isinstance(val, slice):
# TODO Implement slicing
raise NotImplementedError("BDDList slice not implemented")
else:
raise IndexError("BDDList index wrong type")
def explain(fsm, state, spec, context=None):
"""
Explain why `state` of `fsm` satisfies `spec` in `context`.
:param fsm: the system
:type fsm: :class:`BddFsm <pynusmv.fsm.BddFsm>`
:param state: a state of `fsm`
:type state: :class:`State <pynusmv.dd.State>`
:param spec: a specification about `fsm`
:type spec: :class:`Spec <pynusmv.prop.Spec>`
:param context: the context in which evaluate `spec`
:type context: :class:`Spec <pynusmv.prop.Spec>`
Return a tuple `t` composed of states (:class:`State <pynusmv.dd.State>`)
and inputs (:class:`Inputs <pynusmv.dd.Inputs>`),
such that `t[0]` is `state` and `t` represents a path in `fsm` explaining
why `state` satisfies `spec` in `context`.
The returned path is looping if the last state of path is equal to a
previous state along the path.
"""
if context is not None:
context_ptr = context._ptr
else:
context_ptr = None
enc = fsm.bddEnc
manager = enc.DDmanager
path = nsnode.cons(nsnode.bdd2node(nsdd.bdd_dup(state._ptr)), None)
expl = nsmc.explain(fsm._ptr, enc._ptr, path, spec._ptr, context_ptr)
if expl is None:
expl = nsnode.cons(nsnode.bdd2node(nsdd.bdd_dup(state._ptr)), None)
bddlist = BDDList(expl, manager)
bddlist = bddlist.to_tuple()
path = []
path.insert(0, State.from_bdd(bddlist[0], fsm))
for i in range(1, len(bddlist), 2):
inputs = Inputs.from_bdd(bddlist[i], fsm)
state = State.from_bdd(bddlist[i + 1], fsm)
path.insert(0, inputs)
path.insert(0, state)
return tuple(path)
def dup(self):
"""
Return a copy of this BDD.
"""
# Call to bdd_ptr bdd_dup (bdd_ptr);
return BDD(nsdd.bdd_dup(self._ptr), self._manager, freeit=True)
def __iter__(self):
ptr = self._ptr
while ptr:
# Yield BDD copy
bdd_ptr = nsnode.node2bdd(nsnode.car(ptr))
if bdd_ptr is not None:
yield BDD(nsdd.bdd_dup(bdd_ptr), self._manager, freeit=True)
else:
yield None
ptr = nsnode.cdr(ptr)
def from_bdd(bdd, fsm):
"""
Return a new State of fsm from bdd.
:param bdd: a BDD representing a single state
:type bdd: :class:`BDD`
:param fsm: the FSM from which the BDD comes from
:type fsm: :class:`BddFsm <pynusmv.fsm.BddFsm>`
"""
return State(nsdd.bdd_dup(bdd._ptr), fsm)
def from_bdd(bdd, fsm):
"""
Return a new Inputs of fsm from bdd.
:param bdd: a BDD representing a single inputs variables
valuation
:type bdd: :class:`BDD`
:param fsm: the FSM from which the BDD comes from
:type fsm: :class:`BddFsm <pynusmv.fsm.BddFsm>`
"""
return Inputs(nsdd.bdd_dup(bdd._ptr), fsm)
def explainEG(fsm, state, a):
"""
Explain why `state` of `fsm` satisfies `EG phi`, where `a` the set of
states of `fsm` satisfying `phi`, represented by a BDD.
:param fsm: the system
:type fsm: :class:`BddFsm <pynusmv.fsm.BddFsm>`
:param state: a state of `fsm`
:type state: :class:`State <pynusmv.dd.State>`
:param a: the set of states of `fsm` satisfying `phi`
:type a: :class:`BDD <pynusmv.dd.BDD>`
Return a tuple `(t, (i, loop))`
where `t` is a tuple composed of states (:class:`State <pynusmv.dd.State>`)
and inputs (:class:`Inputs <pynusmv.dd.Inputs>`),
such that `t[0]` is state and every other state of `t`
belongs to `a`. The states of `t` are separated by inputs.
Furthermore, `t` represents a path in `fsm`.
`loop` represents the start of the loop contained in `t`,
i.e. `t[-1]` can lead to `loop` through `i`, and `loop` is a state of `t`.
"""
enc = fsm.bddEnc
manager = enc.DDmanager
path = nsnode.cons(nsnode.bdd2node(nsdd.bdd_dup(state._ptr)), None)
bddlist = BDDList(nsmc.eg_explain(fsm._ptr, enc._ptr, path, a._ptr),
manager)
bddlist = bddlist.to_tuple()
path = []
# Discard last state and input, store them as loop indicators
loopstate = State.from_bdd(bddlist[0], fsm)
loopinputs = Inputs.from_bdd(bddlist[1], fsm)
# Consume first state
curstate = State.from_bdd(bddlist[2], fsm)
if curstate._ptr == loopstate._ptr:
loopstate = curstate
path.insert(0, curstate)
for i in range(3, len(bddlist), 2):
inputs = Inputs.from_bdd(bddlist[i], fsm)
curstate = State.from_bdd(bddlist[i + 1], fsm)
if curstate._ptr == loopstate._ptr:
loopstate = curstate
path.insert(0, inputs)
path.insert(0, curstate)
return (tuple(path), (loopinputs, loopstate))
def explainEU(fsm, state, a, b):
"""
Explain why `state` of `fsm` satisfies `E[phi U psi]`,
where `a is the set of states of `fsm` satisfying `phi`
and `b` is the set of states of `fsm` satisfying `psi`, both represented
by BDDs.
:param fsm: the system
:type fsm: :class:`BddFsm <pynusmv.fsm.BddFsm>`
:param state: a state of `fsm`
:type state: :class:`State <pynusmv.dd.State>`
:param a: the set of states of `fsm` satisfying `phi`
:type a: :class:`BDD <pynusmv.dd.BDD>`
:param b: the set of states of `fsm` satisfying `psi`
:type b: :class:`BDD <pynusmv.dd.BDD>`
Return a tuple `t` composed of states (:class:`State <pynusmv.dd.State>`)
and inputs (:class:`Inputs <pynusmv.dd.Inputs>`),
such that `t[0]` is `state`, `t[-1]` belongs to `b`, and every other state
of `t` belongs to `a`. The states of `t` are separated by inputs.
Furthermore, `t` represents a path in `fsm`.
"""
enc = fsm.bddEnc
manager = enc.DDmanager
path = nsnode.cons(nsnode.bdd2node(nsdd.bdd_dup(state._ptr)), None)
bddlist = BDDList(nsmc.eu_explain(fsm._ptr, enc._ptr,
path, a._ptr, b._ptr), manager)
bddlist = bddlist.to_tuple()
path = []
path.insert(0, State.from_bdd(bddlist[0], fsm))
for i in range(1, len(bddlist), 2):
inputs = Inputs.from_bdd(bddlist[i], fsm)
state = State.from_bdd(bddlist[i + 1], fsm)
path.insert(0, inputs)
path.insert(0, state)
return tuple(path)
def check_explain_ltl_spec(spec):
"""
Return whether the loaded SMV model satisfies or not the LTL given `spec`,
that is, whether all initial states of le model satisfies `spec` or not.
Return also an explanation for why the model does not satisfy `spec`,
if it is the case, or `None` otherwise.
The result is a tuple where the first element is a boolean telling whether
`spec` is satisfied, and the second element is either `None` if the first
element is `True`, or a path of the SMV model violating `spec` otherwise.
The explanation is a tuple of alternating states and inputs, starting and
ennding with a state. The path is looping if the last state is somewhere
else in the sequence. States and inputs are represented by dictionaries
where keys are state and inputs variable of the loaded SMV model, and
values are their value.
:param spec: a specification
:type spec: :class:`Spec <pynusmv.prop.Spec>`
:rtype: tuple
"""
# Check that a model has been compiled
assert(glob.prop_database().master is not None)
# Create a property from the given spec
prop = nsprop.Prop_create_partial(spec._ptr, propTypes["LTL"])
# Save settings
o = nsopt.OptsHandler_get_enum_option_value(
nsopt.OptsHandler_get_instance(),
"oreg_justice_emptiness_bdd_algorithm")
if (nscompile.
FlatHierarchy_get_compassion(glob.
global_compile_flathierarchy())
is None and
o == nsfsmbdd.BDD_OREG_JUSTICE_EMPTINESS_BDD_ALGORITHM_EL_FWD):
saved_options = nsfsmbdd.Bdd_elfwd_check_set_and_save_options(
nsfsmbdd.BDD_ELFWD_OPT_ALL)
else:
saved_options = None
# Create, build and check the structure for LTL model checking
ltl_struct = nsltl.Ltl_StructCheckLtlSpec_create(prop)
nsltl.Ltl_StructCheckLtlSpec_build(ltl_struct)
nsltl.Ltl_StructCheckLtlSpec_check(ltl_struct)
# Get the result
result = nsprop.Prop_get_status(prop) == nsprop.Prop_True
# explain
if not result:
# Extract explanation
bdd_fsm = BddFsm(ltl_struct.fsm)
full_fairness = (not nsfsmbdd.FairnessList_is_empty(
nsfsmbdd.compassionList2fairnessList(
nsfsmbdd.BddFsm_get_compassion(bdd_fsm._ptr))))
# Counterexample construction for forward Emerson-Lei not yet
# implemented
assert(full_fairness or not o)
tmp = BDD(nsencbdd.BddEnc_pick_one_state(ltl_struct.bdd_enc,
ltl_struct.s0),
dd_manager=DDManager(nsencbdd.BddEnc_get_dd_manager(
ltl_struct.bdd_enc)))
if full_fairness:
exp = BDDList(nsltl.witness(bdd_fsm._ptr,
ltl_struct.bdd_enc,
tmp._ptr),
ddmanager=tmp._manager)
else:
path = nsnode.cons(nsnode.bdd2node(nsdd.bdd_dup(tmp._ptr)), None)
exp = BDDList(nsnode.reverse(
nsmc.explain(bdd_fsm._ptr,
ltl_struct.bdd_enc,
path,
ltl_struct.spec_formula,
None)),
ddmanager=tmp._manager)
if exp is None: # The counterexample consists of one initial state
exp = BDDList(nsnode.cons(nsnode.bdd2node(nsdd.bdd_dup(tmp._ptr)),
None),
ddmanager=tmp._manager)
exp = exp.to_tuple()
# removes all the tableau variables from the result before
# building the resulting trace. This will make simulation
# working, but may show unexistent loops in the shown trace
def clean(bdd):
tableau_cube = Cube(nsencbdd.BddEnc_get_layer_vars_cube(
ltl_struct.bdd_enc,
ltl_struct.tableau_layer,
nssymb_table.VFT_ALL),
dd_manager=bdd._manager)
return bdd.forsome(tableau_cube)
# exp has at least one element
explanation = [bdd_fsm.pick_one_state(clean(exp[0])).get_str_values()]
for inputs_bdd, state_bdd in zip(exp[1::2], exp[2::2]):
explanation.append(bdd_fsm.pick_one_inputs(
clean(inputs_bdd))
.get_str_values())
explanation.append(bdd_fsm.pick_one_state(
clean(state_bdd))
.get_str_values())
else:
explanation = None
# Destroy the intermediate structure
nsltl.Ltl_StructCheckLtlSpec_destroy(ltl_struct)
# Restore settings
if saved_options is not None:
nsfsmbdd.Bdd_elfwd_restore_options(nsfsmbdd.BDD_ELFWD_OPT_ALL,
saved_options)
# Destroy prop
nsprop.Prop_destroy(prop)
return (result, explanation if explanation is None else tuple(explanation))
