You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
216 lines
8.5 KiB
216 lines
8.5 KiB
from . import core
|
|
from .core import *
|
|
from . import storage
|
|
from .storage import *
|
|
from ._config import *
|
|
from .logic import *
|
|
from .exceptions import *
|
|
|
|
from pycarl import Variable # needed for building parametric models
|
|
|
|
__version__ = "unknown"
|
|
try:
|
|
from _version import __version__
|
|
except ImportError:
|
|
# We're running in a tree that doesn't have a _version.py, so we don't know what our version is.
|
|
pass
|
|
|
|
core._set_up("")
|
|
|
|
|
|
def build_model(symbolic_description, properties=None):
|
|
"""
|
|
Build a model from a symbolic description.
|
|
|
|
:param symbolic_description: Symbolic model description to translate into a model.
|
|
:param List[Property] properties: List of properties that should be preserved during the translation. If None, then all properties are preserved.
|
|
:return: Model in sparse representation.
|
|
:rtype: SparseDtmc or SparseMdp
|
|
"""
|
|
if not symbolic_description.undefined_constants_are_graph_preserving:
|
|
raise StormError("Program still contains undefined constants")
|
|
|
|
if properties:
|
|
formulae = [prop.raw_formula for prop in properties]
|
|
intermediate = core._build_sparse_model_from_prism_program(symbolic_description, formulae)
|
|
else:
|
|
intermediate = core._build_sparse_model_from_prism_program(symbolic_description)
|
|
assert not intermediate.supports_parameters
|
|
if intermediate.model_type == ModelType.DTMC:
|
|
return intermediate._as_dtmc()
|
|
elif intermediate.model_type == ModelType.MDP:
|
|
return intermediate._as_mdp()
|
|
elif intermediate.model_type == ModelType.CTMC:
|
|
return intermediate._as_ctmc()
|
|
elif intermediate.model_type == ModelType.MA:
|
|
return intermediate._as_ma()
|
|
else:
|
|
raise StormError("Not supported non-parametric model constructed")
|
|
|
|
|
|
def build_parametric_model(symbolic_description, properties=None):
|
|
"""
|
|
Build a parametric model from a symbolic description.
|
|
|
|
:param symbolic_description: Symbolic model description to translate into a model.
|
|
:param List[Property] properties: List of properties that should be preserved during the translation. If None, then all properties are preserved.
|
|
:return: Parametric model in sparse representation.
|
|
:rtype: SparseParametricDtmc or SparseParametricMdp
|
|
"""
|
|
if not symbolic_description.undefined_constants_are_graph_preserving:
|
|
raise StormError("Program still contains undefined constants")
|
|
|
|
if properties:
|
|
formulae = [prop.raw_formula for prop in properties]
|
|
else:
|
|
formulae = []
|
|
intermediate = core._build_sparse_parametric_model_from_prism_program(symbolic_description, formulae)
|
|
assert intermediate.supports_parameters
|
|
if intermediate.model_type == ModelType.DTMC:
|
|
return intermediate._as_pdtmc()
|
|
elif intermediate.model_type == ModelType.MDP:
|
|
return intermediate._as_pmdp()
|
|
elif intermediate.model_type == ModelType.CTMC:
|
|
return intermediate._as_pctmc()
|
|
elif intermediate.model_type == ModelType.MA:
|
|
return intermediate._as_pma()
|
|
else:
|
|
raise StormError("Not supported parametric model constructed")
|
|
|
|
|
|
def build_model_from_drn(file):
|
|
"""
|
|
Build a model from the explicit DRN representation.
|
|
|
|
:param String file: DRN file containing the model.
|
|
:return: Model in sparse representation.
|
|
:rtype: SparseDtmc or SparseMdp or SparseCTMC or SparseMA
|
|
"""
|
|
intermediate = core._build_sparse_model_from_drn(file)
|
|
assert not intermediate.supports_parameters
|
|
if intermediate.model_type == ModelType.DTMC:
|
|
return intermediate._as_dtmc()
|
|
elif intermediate.model_type == ModelType.MDP:
|
|
return intermediate._as_mdp()
|
|
elif intermediate.model_type == ModelType.CTMC:
|
|
return intermediate._as_ctmc()
|
|
elif intermediate.model_type == ModelType.MA:
|
|
return intermediate._as_ma()
|
|
else:
|
|
raise StormError("Not supported non-parametric model constructed")
|
|
|
|
|
|
def build_parametric_model_from_drn(file):
|
|
"""
|
|
Build a parametric model from the explicit DRN representation.
|
|
|
|
:param String file: DRN file containing the model.
|
|
:return: Parametric model in sparse representation.
|
|
:rtype: SparseParametricDtmc or SparseParametricMdp or SparseParametricCTMC or SparseParametricMA
|
|
"""
|
|
intermediate = core._build_sparse_parametric_model_from_drn(file)
|
|
assert intermediate.supports_parameters
|
|
if intermediate.model_type == ModelType.DTMC:
|
|
return intermediate._as_pdtmc()
|
|
elif intermediate.model_type == ModelType.MDP:
|
|
return intermediate._as_pmdp()
|
|
elif intermediate.model_type == ModelType.CTMC:
|
|
return intermediate._as_pctmc()
|
|
elif intermediate.model_type == ModelType.MA:
|
|
return intermediate._as_pma()
|
|
else:
|
|
raise StormError("Not supported parametric model constructed")
|
|
|
|
|
|
def perform_bisimulation(model, properties, bisimulation_type):
|
|
"""
|
|
Perform bisimulation on model.
|
|
:param model: Model.
|
|
:param properties: Properties to preserve during bisimulation.
|
|
:param bisimulation_type: Type of bisimulation (weak or strong).
|
|
:return: Model after bisimulation.
|
|
"""
|
|
formulae = [prop.raw_formula for prop in properties]
|
|
if model.supports_parameters:
|
|
return core._perform_parametric_bisimulation(model, formulae, bisimulation_type)
|
|
else:
|
|
return core._perform_bisimulation(model, formulae, bisimulation_type)
|
|
|
|
|
|
def model_checking(model, property, only_initial_states=False, extract_scheduler=False):
|
|
"""
|
|
Perform model checking on model for property.
|
|
:param model: Model.
|
|
:param property: Property to check for.
|
|
:param only_initial_states: If True, only results for initial states are computed.
|
|
If False, results for all states are computed.
|
|
:param extract_scheduler: If True, try to extract a scheduler
|
|
:return: Model checking result.
|
|
:rtype: CheckResult
|
|
"""
|
|
if isinstance(property, Property):
|
|
formula = property.raw_formula
|
|
else:
|
|
formula = property
|
|
|
|
if model.supports_parameters:
|
|
task = core.ParametricCheckTask(formula, only_initial_states)
|
|
task.set_produce_schedulers(extract_scheduler)
|
|
return core._parametric_model_checking_sparse_engine(model, task)
|
|
else:
|
|
task = core.CheckTask(formula, only_initial_states)
|
|
task.set_produce_schedulers(extract_scheduler)
|
|
return core._model_checking_sparse_engine(model, task)
|
|
|
|
|
|
def prob01min_states(model, eventually_formula):
|
|
assert type(eventually_formula) == logic.EventuallyFormula
|
|
labelform = eventually_formula.subformula
|
|
labelprop = core.Property("label-prop", labelform)
|
|
phiStates = BitVector(model.nr_states, True)
|
|
psiStates = model_checking(model, labelprop).get_truth_values()
|
|
return compute_prob01min_states(model, phiStates, psiStates)
|
|
|
|
|
|
def prob01max_states(model, eventually_formula):
|
|
assert type(eventually_formula) == logic.EventuallyFormula
|
|
labelform = eventually_formula.subformula
|
|
labelprop = core.Property("label-prop", labelform)
|
|
phiStates = BitVector(model.nr_states, True)
|
|
psiStates = model_checking(model, labelprop).get_truth_values()
|
|
return compute_prob01min_states(model, phiStates, psiStates)
|
|
|
|
|
|
def compute_prob01_states(model, phi_states, psi_states):
|
|
"""
|
|
Compute prob01 states for properties of the form phi_states until psi_states
|
|
|
|
:param SparseDTMC model:
|
|
:param BitVector phi_states:
|
|
:param BitVector psi_states: Target states
|
|
"""
|
|
if model.model_type != ModelType.DTMC:
|
|
raise StormError("Prob 01 is only defined for DTMCs -- model must be a DTMC")
|
|
|
|
if model.supports_parameters:
|
|
return core._compute_prob01states_rationalfunc(model, phi_states, psi_states)
|
|
else:
|
|
return core._compute_prob01states_double(model, phi_states, psi_states)
|
|
|
|
|
|
def compute_prob01min_states(model, phi_states, psi_states):
|
|
if model.model_type == ModelType.DTMC:
|
|
return compute_prob01_states(model, phi_states, psi_states)
|
|
if model.supports_parameters:
|
|
return core._compute_prob01states_min_rationalfunc(model, phi_states, psi_states)
|
|
else:
|
|
return core._compute_prob01states_min_double(model, phi_states, psi_states)
|
|
|
|
|
|
def compute_prob01max_states(model, phi_states, psi_states):
|
|
if model.model_type == ModelType.DTMC:
|
|
return compute_prob01_states(model, phi_states, psi_states)
|
|
if model.supports_parameters:
|
|
return core._compute_prob01states_max_rationalfunc(model, phi_states, psi_states)
|
|
else:
|
|
return core._compute_prob01states_max_double(model, phi_states, psi_states)
|